diff options
Diffstat (limited to 'drivers/gpu/drm/amd')
321 files changed, 27231 insertions, 7934 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/Makefile b/drivers/gpu/drm/amd/amdgpu/Makefile index 2ca2b5154d522..bfd332c95b61e 100644 --- a/drivers/gpu/drm/amd/amdgpu/Makefile +++ b/drivers/gpu/drm/amd/amdgpu/Makefile @@ -56,13 +56,18 @@ amdgpu-y += amdgpu_device.o amdgpu_kms.o \ # add asic specific block amdgpu-$(CONFIG_DRM_AMDGPU_CIK)+= cik.o cik_ih.o kv_smc.o kv_dpm.o \ - ci_smc.o ci_dpm.o dce_v8_0.o gfx_v7_0.o cik_sdma.o uvd_v4_2.o vce_v2_0.o \ - amdgpu_amdkfd_gfx_v7.o + ci_smc.o ci_dpm.o dce_v8_0.o gfx_v7_0.o cik_sdma.o uvd_v4_2.o vce_v2_0.o amdgpu-$(CONFIG_DRM_AMDGPU_SI)+= si.o gmc_v6_0.o gfx_v6_0.o si_ih.o si_dma.o dce_v6_0.o si_dpm.o si_smc.o amdgpu-y += \ - vi.o mxgpu_vi.o nbio_v6_1.o soc15.o emu_soc.o mxgpu_ai.o nbio_v7_0.o vega10_reg_init.o + vi.o mxgpu_vi.o nbio_v6_1.o soc15.o emu_soc.o mxgpu_ai.o nbio_v7_0.o vega10_reg_init.o \ + vega20_reg_init.o + +# add DF block +amdgpu-y += \ + df_v1_7.o \ + df_v3_6.o # add GMC block amdgpu-y += \ @@ -126,11 +131,20 @@ amdgpu-y += \ vcn_v1_0.o # add amdkfd interfaces +amdgpu-y += amdgpu_amdkfd.o + +ifneq ($(CONFIG_HSA_AMD),) amdgpu-y += \ - amdgpu_amdkfd.o \ amdgpu_amdkfd_fence.o \ amdgpu_amdkfd_gpuvm.o \ - amdgpu_amdkfd_gfx_v8.o + amdgpu_amdkfd_gfx_v8.o \ + amdgpu_amdkfd_gfx_v9.o + +ifneq ($(CONFIG_DRM_AMDGPU_CIK),) +amdgpu-y += amdgpu_amdkfd_gfx_v7.o +endif + +endif # add cgs amdgpu-y += amdgpu_cgs.o diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu.h b/drivers/gpu/drm/amd/amdgpu/amdgpu.h index c8b605f3dc05d..a59c07590ceec 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu.h @@ -129,6 +129,7 @@ extern int amdgpu_lbpw; extern int amdgpu_compute_multipipe; extern int amdgpu_gpu_recovery; extern int amdgpu_emu_mode; +extern uint amdgpu_smu_memory_pool_size; #ifdef CONFIG_DRM_AMDGPU_SI extern int amdgpu_si_support; @@ -137,6 +138,7 @@ extern int amdgpu_si_support; extern int amdgpu_cik_support; #endif +#define AMDGPU_SG_THRESHOLD (256*1024*1024) #define AMDGPU_DEFAULT_GTT_SIZE_MB 3072ULL /* 3GB by default */ #define AMDGPU_WAIT_IDLE_TIMEOUT_IN_MS 3000 #define AMDGPU_MAX_USEC_TIMEOUT 100000 /* 100 ms */ @@ -222,10 +224,10 @@ enum amdgpu_kiq_irq { AMDGPU_CP_KIQ_IRQ_LAST }; -int amdgpu_device_ip_set_clockgating_state(struct amdgpu_device *adev, +int amdgpu_device_ip_set_clockgating_state(void *dev, enum amd_ip_block_type block_type, enum amd_clockgating_state state); -int amdgpu_device_ip_set_powergating_state(struct amdgpu_device *adev, +int amdgpu_device_ip_set_powergating_state(void *dev, enum amd_ip_block_type block_type, enum amd_powergating_state state); void amdgpu_device_ip_get_clockgating_state(struct amdgpu_device *adev, @@ -681,6 +683,8 @@ int amdgpu_ctx_ioctl(struct drm_device *dev, void *data, int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, unsigned ring_id); void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr); +void amdgpu_ctx_mgr_entity_cleanup(struct amdgpu_ctx_mgr *mgr); +void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr); void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr); @@ -771,9 +775,18 @@ struct amdgpu_rlc { u32 starting_offsets_start; u32 reg_list_format_size_bytes; u32 reg_list_size_bytes; + u32 reg_list_format_direct_reg_list_length; + u32 save_restore_list_cntl_size_bytes; + u32 save_restore_list_gpm_size_bytes; + u32 save_restore_list_srm_size_bytes; u32 *register_list_format; u32 *register_restore; + u8 *save_restore_list_cntl; + u8 *save_restore_list_gpm; + u8 *save_restore_list_srm; + + bool is_rlc_v2_1; }; #define AMDGPU_MAX_COMPUTE_QUEUES KGD_MAX_QUEUES @@ -867,6 +880,8 @@ struct amdgpu_gfx_config { /* gfx configure feature */ uint32_t double_offchip_lds_buf; + /* cached value of DB_DEBUG2 */ + uint32_t db_debug2; }; struct amdgpu_cu_info { @@ -938,6 +953,12 @@ struct amdgpu_gfx { uint32_t ce_feature_version; uint32_t pfp_feature_version; uint32_t rlc_feature_version; + uint32_t rlc_srlc_fw_version; + uint32_t rlc_srlc_feature_version; + uint32_t rlc_srlg_fw_version; + uint32_t rlc_srlg_feature_version; + uint32_t rlc_srls_fw_version; + uint32_t rlc_srls_feature_version; uint32_t mec_feature_version; uint32_t mec2_feature_version; struct amdgpu_ring gfx_ring[AMDGPU_MAX_GFX_RINGS]; @@ -1204,6 +1225,8 @@ struct amdgpu_asic_funcs { /* invalidate hdp read cache */ void (*invalidate_hdp)(struct amdgpu_device *adev, struct amdgpu_ring *ring); + /* check if the asic needs a full reset of if soft reset will work */ + bool (*need_full_reset)(struct amdgpu_device *adev); }; /* @@ -1368,7 +1391,19 @@ struct amdgpu_nbio_funcs { void (*detect_hw_virt)(struct amdgpu_device *adev); }; - +struct amdgpu_df_funcs { + void (*init)(struct amdgpu_device *adev); + void (*enable_broadcast_mode)(struct amdgpu_device *adev, + bool enable); + u32 (*get_fb_channel_number)(struct amdgpu_device *adev); + u32 (*get_hbm_channel_number)(struct amdgpu_device *adev); + void (*update_medium_grain_clock_gating)(struct amdgpu_device *adev, + bool enable); + void (*get_clockgating_state)(struct amdgpu_device *adev, + u32 *flags); + void (*enable_ecc_force_par_wr_rmw)(struct amdgpu_device *adev, + bool enable); +}; /* Define the HW IP blocks will be used in driver , add more if necessary */ enum amd_hw_ip_block_type { GC_HWIP = 1, @@ -1398,6 +1433,7 @@ enum amd_hw_ip_block_type { struct amd_powerplay { void *pp_handle; const struct amd_pm_funcs *pp_funcs; + uint32_t pp_feature; }; #define AMDGPU_RESET_MAGIC_NUM 64 @@ -1590,6 +1626,7 @@ struct amdgpu_device { uint32_t *reg_offset[MAX_HWIP][HWIP_MAX_INSTANCE]; const struct amdgpu_nbio_funcs *nbio_funcs; + const struct amdgpu_df_funcs *df_funcs; /* delayed work_func for deferring clockgating during resume */ struct delayed_work late_init_work; @@ -1764,6 +1801,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_asic_get_config_memsize(adev) (adev)->asic_funcs->get_config_memsize((adev)) #define amdgpu_asic_flush_hdp(adev, r) (adev)->asic_funcs->flush_hdp((adev), (r)) #define amdgpu_asic_invalidate_hdp(adev, r) (adev)->asic_funcs->invalidate_hdp((adev), (r)) +#define amdgpu_asic_need_full_reset(adev) (adev)->asic_funcs->need_full_reset((adev)) #define amdgpu_gmc_flush_gpu_tlb(adev, vmid) (adev)->gmc.gmc_funcs->flush_gpu_tlb((adev), (vmid)) #define amdgpu_gmc_emit_flush_gpu_tlb(r, vmid, addr) (r)->adev->gmc.gmc_funcs->emit_flush_gpu_tlb((r), (vmid), (addr)) #define amdgpu_gmc_emit_pasid_mapping(r, vmid, pasid) (r)->adev->gmc.gmc_funcs->emit_pasid_mapping((r), (vmid), (pasid)) @@ -1790,6 +1828,7 @@ amdgpu_get_sdma_instance(struct amdgpu_ring *ring) #define amdgpu_ring_emit_rreg(r, d) (r)->funcs->emit_rreg((r), (d)) #define amdgpu_ring_emit_wreg(r, d, v) (r)->funcs->emit_wreg((r), (d), (v)) #define amdgpu_ring_emit_reg_wait(r, d, v, m) (r)->funcs->emit_reg_wait((r), (d), (v), (m)) +#define amdgpu_ring_emit_reg_write_reg_wait(r, d0, d1, v, m) (r)->funcs->emit_reg_write_reg_wait((r), (d0), (d1), (v), (m)) #define amdgpu_ring_emit_tmz(r, b) (r)->funcs->emit_tmz((r), (b)) #define amdgpu_ring_pad_ib(r, ib) ((r)->funcs->pad_ib((r), (ib))) #define amdgpu_ring_init_cond_exec(r) (r)->funcs->init_cond_exec((r)) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c index a29362f9ef415..428e5eb3444f0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_acp.c @@ -290,12 +290,11 @@ static int acp_hw_init(void *handle) else if (r) return r; - r = cgs_get_pci_resource(adev->acp.cgs_device, CGS_RESOURCE_TYPE_MMIO, - 0x5289, 0, &acp_base); - if (r == -ENODEV) - return 0; - else if (r) - return r; + if (adev->rmmio_size == 0 || adev->rmmio_size < 0x5289) + return -EINVAL; + + acp_base = adev->rmmio_base; + if (adev->asic_type != CHIP_STONEY) { adev->acp.acp_genpd = kzalloc(sizeof(struct acp_pm_domain), GFP_KERNEL); if (adev->acp.acp_genpd == NULL) @@ -513,7 +512,7 @@ static int acp_hw_fini(void *handle) if (adev->acp.acp_genpd) { for (i = 0; i < ACP_DEVS ; i++) { dev = get_mfd_cell_dev(adev->acp.acp_cell[i].name, i); - ret = pm_genpd_remove_device(&adev->acp.acp_genpd->gpd, dev); + ret = pm_genpd_remove_device(dev); /* If removal fails, dont giveup and try rest */ if (ret) dev_err(dev, "remove dev from genpd failed\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c index 4d36203ffb115..8f6f45567bfa1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.c @@ -50,15 +50,21 @@ int amdgpu_amdkfd_init(void) kgd2kfd = NULL; } + #elif defined(CONFIG_HSA_AMD) + ret = kgd2kfd_init(KFD_INTERFACE_VERSION, &kgd2kfd); if (ret) kgd2kfd = NULL; #else + kgd2kfd = NULL; ret = -ENOENT; #endif + +#if defined(CONFIG_HSA_AMD_MODULE) || defined(CONFIG_HSA_AMD) amdgpu_amdkfd_gpuvm_init_mem_limits(); +#endif return ret; } @@ -92,8 +98,12 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev) case CHIP_POLARIS11: kfd2kgd = amdgpu_amdkfd_gfx_8_0_get_functions(); break; + case CHIP_VEGA10: + case CHIP_RAVEN: + kfd2kgd = amdgpu_amdkfd_gfx_9_0_get_functions(); + break; default: - dev_dbg(adev->dev, "kfd not supported on this ASIC\n"); + dev_info(adev->dev, "kfd not supported on this ASIC\n"); return; } @@ -175,6 +185,28 @@ void amdgpu_amdkfd_device_init(struct amdgpu_device *adev) &gpu_resources.doorbell_physical_address, &gpu_resources.doorbell_aperture_size, &gpu_resources.doorbell_start_offset); + if (adev->asic_type >= CHIP_VEGA10) { + /* On SOC15 the BIF is involved in routing + * doorbells using the low 12 bits of the + * address. Communicate the assignments to + * KFD. KFD uses two doorbell pages per + * process in case of 64-bit doorbells so we + * can use each doorbell assignment twice. + */ + gpu_resources.sdma_doorbell[0][0] = + AMDGPU_DOORBELL64_sDMA_ENGINE0; + gpu_resources.sdma_doorbell[0][1] = + AMDGPU_DOORBELL64_sDMA_ENGINE0 + 0x200; + gpu_resources.sdma_doorbell[1][0] = + AMDGPU_DOORBELL64_sDMA_ENGINE1; + gpu_resources.sdma_doorbell[1][1] = + AMDGPU_DOORBELL64_sDMA_ENGINE1 + 0x200; + /* Doorbells 0x0f0-0ff and 0x2f0-2ff are reserved for + * SDMA, IH and VCN. So don't use them for the CP. + */ + gpu_resources.reserved_doorbell_mask = 0x1f0; + gpu_resources.reserved_doorbell_val = 0x0f0; + } kgd2kfd->device_init(adev->kfd, &gpu_resources); } @@ -217,13 +249,19 @@ int alloc_gtt_mem(struct kgd_dev *kgd, size_t size, { struct amdgpu_device *adev = (struct amdgpu_device *)kgd; struct amdgpu_bo *bo = NULL; + struct amdgpu_bo_param bp; int r; uint64_t gpu_addr_tmp = 0; void *cpu_ptr_tmp = NULL; - r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC, ttm_bo_type_kernel, - NULL, &bo); + memset(&bp, 0, sizeof(bp)); + bp.size = size; + bp.byte_align = PAGE_SIZE; + bp.domain = AMDGPU_GEM_DOMAIN_GTT; + bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC; + bp.type = ttm_bo_type_kernel; + bp.resv = NULL; + r = amdgpu_bo_create(adev, &bp, &bo); if (r) { dev_err(adev->dev, "failed to allocate BO for amdkfd (%d)\n", r); @@ -432,3 +470,44 @@ bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid) return false; } + +#if !defined(CONFIG_HSA_AMD_MODULE) && !defined(CONFIG_HSA_AMD) +bool amdkfd_fence_check_mm(struct dma_fence *f, struct mm_struct *mm) +{ + return false; +} + +void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo) +{ +} + +void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, + struct amdgpu_vm *vm) +{ +} + +struct amdgpu_amdkfd_fence *to_amdgpu_amdkfd_fence(struct dma_fence *f) +{ + return NULL; +} + +int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm) +{ + return 0; +} + +struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void) +{ + return NULL; +} + +struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void) +{ + return NULL; +} + +struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void) +{ + return NULL; +} +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h index c2c2bea731e04..a8418a3f4e9d3 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd.h @@ -28,6 +28,7 @@ #include <linux/types.h> #include <linux/mm.h> #include <linux/mmu_context.h> +#include <linux/workqueue.h> #include <kgd_kfd_interface.h> #include <drm/ttm/ttm_execbuf_util.h> #include "amdgpu_sync.h" @@ -59,7 +60,9 @@ struct kgd_mem { uint32_t mapping_flags; + atomic_t invalid; struct amdkfd_process_info *process_info; + struct page **user_pages; struct amdgpu_sync sync; @@ -84,6 +87,9 @@ struct amdkfd_process_info { struct list_head vm_list_head; /* List head for all KFD BOs that belong to a KFD process. */ struct list_head kfd_bo_list; + /* List of userptr BOs that are valid or invalid */ + struct list_head userptr_valid_list; + struct list_head userptr_inval_list; /* Lock to protect kfd_bo_list */ struct mutex lock; @@ -91,6 +97,11 @@ struct amdkfd_process_info { unsigned int n_vms; /* Eviction Fence */ struct amdgpu_amdkfd_fence *eviction_fence; + + /* MMU-notifier related fields */ + atomic_t evicted_bos; + struct delayed_work restore_userptr_work; + struct pid *pid; }; int amdgpu_amdkfd_init(void); @@ -104,12 +115,14 @@ void amdgpu_amdkfd_device_probe(struct amdgpu_device *adev); void amdgpu_amdkfd_device_init(struct amdgpu_device *adev); void amdgpu_amdkfd_device_fini(struct amdgpu_device *adev); +int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, struct mm_struct *mm); int amdgpu_amdkfd_submit_ib(struct kgd_dev *kgd, enum kgd_engine_type engine, uint32_t vmid, uint64_t gpu_addr, uint32_t *ib_cmd, uint32_t ib_len); struct kfd2kgd_calls *amdgpu_amdkfd_gfx_7_get_functions(void); struct kfd2kgd_calls *amdgpu_amdkfd_gfx_8_0_get_functions(void); +struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void); bool amdgpu_amdkfd_is_kfd_vmid(struct amdgpu_device *adev, u32 vmid); @@ -143,14 +156,14 @@ uint64_t amdgpu_amdkfd_get_vram_usage(struct kgd_dev *kgd); /* GPUVM API */ int amdgpu_amdkfd_gpuvm_create_process_vm(struct kgd_dev *kgd, void **vm, - void **process_info, - struct dma_fence **ef); + void **process_info, + struct dma_fence **ef); int amdgpu_amdkfd_gpuvm_acquire_process_vm(struct kgd_dev *kgd, - struct file *filp, - void **vm, void **process_info, - struct dma_fence **ef); + struct file *filp, + void **vm, void **process_info, + struct dma_fence **ef); void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, - struct amdgpu_vm *vm); + struct amdgpu_vm *vm); void amdgpu_amdkfd_gpuvm_destroy_process_vm(struct kgd_dev *kgd, void *vm); uint32_t amdgpu_amdkfd_gpuvm_get_process_page_dir(void *vm); int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c index ea54e53172b92..0ff36d45a5970 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v7.c @@ -98,8 +98,6 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid, unsigned int vmid); -static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id, - uint32_t hpd_size, uint64_t hpd_gpu_addr); static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id); static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t __user *wptr, @@ -183,7 +181,6 @@ static const struct kfd2kgd_calls kfd2kgd = { .free_pasid = amdgpu_pasid_free, .program_sh_mem_settings = kgd_program_sh_mem_settings, .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, - .init_pipeline = kgd_init_pipeline, .init_interrupts = kgd_init_interrupts, .hqd_load = kgd_hqd_load, .hqd_sdma_load = kgd_hqd_sdma_load, @@ -309,13 +306,6 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid, return 0; } -static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id, - uint32_t hpd_size, uint64_t hpd_gpu_addr) -{ - /* amdgpu owns the per-pipe state */ - return 0; -} - static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) { struct amdgpu_device *adev = get_amdgpu_device(kgd); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c index 89264c9a5e9f9..6ef9762b4b007 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v8.c @@ -57,8 +57,6 @@ static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, uint32_t sh_mem_bases); static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid, unsigned int vmid); -static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id, - uint32_t hpd_size, uint64_t hpd_gpu_addr); static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id); static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, uint32_t queue_id, uint32_t __user *wptr, @@ -141,7 +139,6 @@ static const struct kfd2kgd_calls kfd2kgd = { .free_pasid = amdgpu_pasid_free, .program_sh_mem_settings = kgd_program_sh_mem_settings, .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, - .init_pipeline = kgd_init_pipeline, .init_interrupts = kgd_init_interrupts, .hqd_load = kgd_hqd_load, .hqd_sdma_load = kgd_hqd_sdma_load, @@ -270,13 +267,6 @@ static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid, return 0; } -static int kgd_init_pipeline(struct kgd_dev *kgd, uint32_t pipe_id, - uint32_t hpd_size, uint64_t hpd_gpu_addr) -{ - /* amdgpu owns the per-pipe state */ - return 0; -} - static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) { struct amdgpu_device *adev = get_amdgpu_device(kgd); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c new file mode 100644 index 0000000000000..f0c0d3953f698 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gfx_v9.c @@ -0,0 +1,1043 @@ +/* + * Copyright 2014-2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#define pr_fmt(fmt) "kfd2kgd: " fmt + +#include <linux/module.h> +#include <linux/fdtable.h> +#include <linux/uaccess.h> +#include <linux/firmware.h> +#include <drm/drmP.h> +#include "amdgpu.h" +#include "amdgpu_amdkfd.h" +#include "amdgpu_ucode.h" +#include "soc15_hw_ip.h" +#include "gc/gc_9_0_offset.h" +#include "gc/gc_9_0_sh_mask.h" +#include "vega10_enum.h" +#include "sdma0/sdma0_4_0_offset.h" +#include "sdma0/sdma0_4_0_sh_mask.h" +#include "sdma1/sdma1_4_0_offset.h" +#include "sdma1/sdma1_4_0_sh_mask.h" +#include "athub/athub_1_0_offset.h" +#include "athub/athub_1_0_sh_mask.h" +#include "oss/osssys_4_0_offset.h" +#include "oss/osssys_4_0_sh_mask.h" +#include "soc15_common.h" +#include "v9_structs.h" +#include "soc15.h" +#include "soc15d.h" + +/* HACK: MMHUB and GC both have VM-related register with the same + * names but different offsets. Define the MMHUB register we need here + * with a prefix. A proper solution would be to move the functions + * programming these registers into gfx_v9_0.c and mmhub_v1_0.c + * respectively. + */ +#define mmMMHUB_VM_INVALIDATE_ENG16_REQ 0x06f3 +#define mmMMHUB_VM_INVALIDATE_ENG16_REQ_BASE_IDX 0 + +#define mmMMHUB_VM_INVALIDATE_ENG16_ACK 0x0705 +#define mmMMHUB_VM_INVALIDATE_ENG16_ACK_BASE_IDX 0 + +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32 0x072b +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32_BASE_IDX 0 +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32 0x072c +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32_BASE_IDX 0 + +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32 0x074b +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32_BASE_IDX 0 +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32 0x074c +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32_BASE_IDX 0 + +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32 0x076b +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32_BASE_IDX 0 +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32 0x076c +#define mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32_BASE_IDX 0 + +#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_LO32 0x0727 +#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_LO32_BASE_IDX 0 +#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_HI32 0x0728 +#define mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_HI32_BASE_IDX 0 + +#define V9_PIPE_PER_MEC (4) +#define V9_QUEUES_PER_PIPE_MEC (8) + +enum hqd_dequeue_request_type { + NO_ACTION = 0, + DRAIN_PIPE, + RESET_WAVES +}; + +/* + * Register access functions + */ + +static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, + uint32_t sh_mem_config, + uint32_t sh_mem_ape1_base, uint32_t sh_mem_ape1_limit, + uint32_t sh_mem_bases); +static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid, + unsigned int vmid); +static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id); +static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, + uint32_t queue_id, uint32_t __user *wptr, + uint32_t wptr_shift, uint32_t wptr_mask, + struct mm_struct *mm); +static int kgd_hqd_dump(struct kgd_dev *kgd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs); +static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, + uint32_t __user *wptr, struct mm_struct *mm); +static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, + uint32_t engine_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs); +static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, + uint32_t pipe_id, uint32_t queue_id); +static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd); +static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, + enum kfd_preempt_type reset_type, + unsigned int utimeout, uint32_t pipe_id, + uint32_t queue_id); +static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, + unsigned int utimeout); +static int kgd_address_watch_disable(struct kgd_dev *kgd); +static int kgd_address_watch_execute(struct kgd_dev *kgd, + unsigned int watch_point_id, + uint32_t cntl_val, + uint32_t addr_hi, + uint32_t addr_lo); +static int kgd_wave_control_execute(struct kgd_dev *kgd, + uint32_t gfx_index_val, + uint32_t sq_cmd); +static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd, + unsigned int watch_point_id, + unsigned int reg_offset); + +static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd, + uint8_t vmid); +static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd, + uint8_t vmid); +static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, + uint32_t page_table_base); +static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type); +static void set_scratch_backing_va(struct kgd_dev *kgd, + uint64_t va, uint32_t vmid); +static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid); +static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid); + +/* Because of REG_GET_FIELD() being used, we put this function in the + * asic specific file. + */ +static int amdgpu_amdkfd_get_tile_config(struct kgd_dev *kgd, + struct tile_config *config) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)kgd; + + config->gb_addr_config = adev->gfx.config.gb_addr_config; + + config->tile_config_ptr = adev->gfx.config.tile_mode_array; + config->num_tile_configs = + ARRAY_SIZE(adev->gfx.config.tile_mode_array); + config->macro_tile_config_ptr = + adev->gfx.config.macrotile_mode_array; + config->num_macro_tile_configs = + ARRAY_SIZE(adev->gfx.config.macrotile_mode_array); + + return 0; +} + +static const struct kfd2kgd_calls kfd2kgd = { + .init_gtt_mem_allocation = alloc_gtt_mem, + .free_gtt_mem = free_gtt_mem, + .get_local_mem_info = get_local_mem_info, + .get_gpu_clock_counter = get_gpu_clock_counter, + .get_max_engine_clock_in_mhz = get_max_engine_clock_in_mhz, + .alloc_pasid = amdgpu_pasid_alloc, + .free_pasid = amdgpu_pasid_free, + .program_sh_mem_settings = kgd_program_sh_mem_settings, + .set_pasid_vmid_mapping = kgd_set_pasid_vmid_mapping, + .init_interrupts = kgd_init_interrupts, + .hqd_load = kgd_hqd_load, + .hqd_sdma_load = kgd_hqd_sdma_load, + .hqd_dump = kgd_hqd_dump, + .hqd_sdma_dump = kgd_hqd_sdma_dump, + .hqd_is_occupied = kgd_hqd_is_occupied, + .hqd_sdma_is_occupied = kgd_hqd_sdma_is_occupied, + .hqd_destroy = kgd_hqd_destroy, + .hqd_sdma_destroy = kgd_hqd_sdma_destroy, + .address_watch_disable = kgd_address_watch_disable, + .address_watch_execute = kgd_address_watch_execute, + .wave_control_execute = kgd_wave_control_execute, + .address_watch_get_offset = kgd_address_watch_get_offset, + .get_atc_vmid_pasid_mapping_pasid = + get_atc_vmid_pasid_mapping_pasid, + .get_atc_vmid_pasid_mapping_valid = + get_atc_vmid_pasid_mapping_valid, + .get_fw_version = get_fw_version, + .set_scratch_backing_va = set_scratch_backing_va, + .get_tile_config = amdgpu_amdkfd_get_tile_config, + .get_cu_info = get_cu_info, + .get_vram_usage = amdgpu_amdkfd_get_vram_usage, + .create_process_vm = amdgpu_amdkfd_gpuvm_create_process_vm, + .acquire_process_vm = amdgpu_amdkfd_gpuvm_acquire_process_vm, + .destroy_process_vm = amdgpu_amdkfd_gpuvm_destroy_process_vm, + .get_process_page_dir = amdgpu_amdkfd_gpuvm_get_process_page_dir, + .set_vm_context_page_table_base = set_vm_context_page_table_base, + .alloc_memory_of_gpu = amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu, + .free_memory_of_gpu = amdgpu_amdkfd_gpuvm_free_memory_of_gpu, + .map_memory_to_gpu = amdgpu_amdkfd_gpuvm_map_memory_to_gpu, + .unmap_memory_to_gpu = amdgpu_amdkfd_gpuvm_unmap_memory_from_gpu, + .sync_memory = amdgpu_amdkfd_gpuvm_sync_memory, + .map_gtt_bo_to_kernel = amdgpu_amdkfd_gpuvm_map_gtt_bo_to_kernel, + .restore_process_bos = amdgpu_amdkfd_gpuvm_restore_process_bos, + .invalidate_tlbs = invalidate_tlbs, + .invalidate_tlbs_vmid = invalidate_tlbs_vmid, + .submit_ib = amdgpu_amdkfd_submit_ib, +}; + +struct kfd2kgd_calls *amdgpu_amdkfd_gfx_9_0_get_functions(void) +{ + return (struct kfd2kgd_calls *)&kfd2kgd; +} + +static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) +{ + return (struct amdgpu_device *)kgd; +} + +static void lock_srbm(struct kgd_dev *kgd, uint32_t mec, uint32_t pipe, + uint32_t queue, uint32_t vmid) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + mutex_lock(&adev->srbm_mutex); + soc15_grbm_select(adev, mec, pipe, queue, vmid); +} + +static void unlock_srbm(struct kgd_dev *kgd) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + soc15_grbm_select(adev, 0, 0, 0, 0); + mutex_unlock(&adev->srbm_mutex); +} + +static void acquire_queue(struct kgd_dev *kgd, uint32_t pipe_id, + uint32_t queue_id) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + uint32_t mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + uint32_t pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); + + lock_srbm(kgd, mec, pipe, queue_id, 0); +} + +static uint32_t get_queue_mask(struct amdgpu_device *adev, + uint32_t pipe_id, uint32_t queue_id) +{ + unsigned int bit = (pipe_id * adev->gfx.mec.num_queue_per_pipe + + queue_id) & 31; + + return ((uint32_t)1) << bit; +} + +static void release_queue(struct kgd_dev *kgd) +{ + unlock_srbm(kgd); +} + +static void kgd_program_sh_mem_settings(struct kgd_dev *kgd, uint32_t vmid, + uint32_t sh_mem_config, + uint32_t sh_mem_ape1_base, + uint32_t sh_mem_ape1_limit, + uint32_t sh_mem_bases) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + lock_srbm(kgd, 0, 0, 0, vmid); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_CONFIG), sh_mem_config); + WREG32(SOC15_REG_OFFSET(GC, 0, mmSH_MEM_BASES), sh_mem_bases); + /* APE1 no longer exists on GFX9 */ + + unlock_srbm(kgd); +} + +static int kgd_set_pasid_vmid_mapping(struct kgd_dev *kgd, unsigned int pasid, + unsigned int vmid) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + + /* + * We have to assume that there is no outstanding mapping. + * The ATC_VMID_PASID_MAPPING_UPDATE_STATUS bit could be 0 because + * a mapping is in progress or because a mapping finished + * and the SW cleared it. + * So the protocol is to always wait & clear. + */ + uint32_t pasid_mapping = (pasid == 0) ? 0 : (uint32_t)pasid | + ATC_VMID0_PASID_MAPPING__VALID_MASK; + + /* + * need to do this twice, once for gfx and once for mmhub + * for ATC add 16 to VMID for mmhub, for IH different registers. + * ATC_VMID0..15 registers are separate from ATC_VMID16..31. + */ + + WREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATC_VMID0_PASID_MAPPING) + vmid, + pasid_mapping); + + while (!(RREG32(SOC15_REG_OFFSET( + ATHUB, 0, + mmATC_VMID_PASID_MAPPING_UPDATE_STATUS)) & + (1U << vmid))) + cpu_relax(); + + WREG32(SOC15_REG_OFFSET(ATHUB, 0, + mmATC_VMID_PASID_MAPPING_UPDATE_STATUS), + 1U << vmid); + + /* Mapping vmid to pasid also for IH block */ + WREG32(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT) + vmid, + pasid_mapping); + + WREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATC_VMID16_PASID_MAPPING) + vmid, + pasid_mapping); + + while (!(RREG32(SOC15_REG_OFFSET( + ATHUB, 0, + mmATC_VMID_PASID_MAPPING_UPDATE_STATUS)) & + (1U << (vmid + 16)))) + cpu_relax(); + + WREG32(SOC15_REG_OFFSET(ATHUB, 0, + mmATC_VMID_PASID_MAPPING_UPDATE_STATUS), + 1U << (vmid + 16)); + + /* Mapping vmid to pasid also for IH block */ + WREG32(SOC15_REG_OFFSET(OSSSYS, 0, mmIH_VMID_0_LUT_MM) + vmid, + pasid_mapping); + return 0; +} + +/* TODO - RING0 form of field is obsolete, seems to date back to SI + * but still works + */ + +static int kgd_init_interrupts(struct kgd_dev *kgd, uint32_t pipe_id) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t mec; + uint32_t pipe; + + mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); + + lock_srbm(kgd, mec, pipe, 0, 0); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmCPC_INT_CNTL), + CP_INT_CNTL_RING0__TIME_STAMP_INT_ENABLE_MASK | + CP_INT_CNTL_RING0__OPCODE_ERROR_INT_ENABLE_MASK); + + unlock_srbm(kgd); + + return 0; +} + +static uint32_t get_sdma_base_addr(struct amdgpu_device *adev, + unsigned int engine_id, + unsigned int queue_id) +{ + uint32_t base[2] = { + SOC15_REG_OFFSET(SDMA0, 0, + mmSDMA0_RLC0_RB_CNTL) - mmSDMA0_RLC0_RB_CNTL, + SOC15_REG_OFFSET(SDMA1, 0, + mmSDMA1_RLC0_RB_CNTL) - mmSDMA1_RLC0_RB_CNTL + }; + uint32_t retval; + + retval = base[engine_id] + queue_id * (mmSDMA0_RLC1_RB_CNTL - + mmSDMA0_RLC0_RB_CNTL); + + pr_debug("sdma base address: 0x%x\n", retval); + + return retval; +} + +static inline struct v9_mqd *get_mqd(void *mqd) +{ + return (struct v9_mqd *)mqd; +} + +static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd) +{ + return (struct v9_sdma_mqd *)mqd; +} + +static int kgd_hqd_load(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, + uint32_t queue_id, uint32_t __user *wptr, + uint32_t wptr_shift, uint32_t wptr_mask, + struct mm_struct *mm) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + struct v9_mqd *m; + uint32_t *mqd_hqd; + uint32_t reg, hqd_base, data; + + m = get_mqd(mqd); + + acquire_queue(kgd, pipe_id, queue_id); + + /* HIQ is set during driver init period with vmid set to 0*/ + if (m->cp_hqd_vmid == 0) { + uint32_t value, mec, pipe; + + mec = (pipe_id / adev->gfx.mec.num_pipe_per_mec) + 1; + pipe = (pipe_id % adev->gfx.mec.num_pipe_per_mec); + + pr_debug("kfd: set HIQ, mec:%d, pipe:%d, queue:%d.\n", + mec, pipe, queue_id); + value = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CP_SCHEDULERS)); + value = REG_SET_FIELD(value, RLC_CP_SCHEDULERS, scheduler1, + ((mec << 5) | (pipe << 3) | queue_id | 0x80)); + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_CP_SCHEDULERS), value); + } + + /* HQD registers extend from CP_MQD_BASE_ADDR to CP_HQD_EOP_WPTR_MEM. */ + mqd_hqd = &m->cp_mqd_base_addr_lo; + hqd_base = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR); + + for (reg = hqd_base; + reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++) + WREG32(reg, mqd_hqd[reg - hqd_base]); + + + /* Activate doorbell logic before triggering WPTR poll. */ + data = REG_SET_FIELD(m->cp_hqd_pq_doorbell_control, + CP_HQD_PQ_DOORBELL_CONTROL, DOORBELL_EN, 1); + WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_DOORBELL_CONTROL), data); + + if (wptr) { + /* Don't read wptr with get_user because the user + * context may not be accessible (if this function + * runs in a work queue). Instead trigger a one-shot + * polling read from memory in the CP. This assumes + * that wptr is GPU-accessible in the queue's VMID via + * ATC or SVM. WPTR==RPTR before starting the poll so + * the CP starts fetching new commands from the right + * place. + * + * Guessing a 64-bit WPTR from a 32-bit RPTR is a bit + * tricky. Assume that the queue didn't overflow. The + * number of valid bits in the 32-bit RPTR depends on + * the queue size. The remaining bits are taken from + * the saved 64-bit WPTR. If the WPTR wrapped, add the + * queue size. + */ + uint32_t queue_size = + 2 << REG_GET_FIELD(m->cp_hqd_pq_control, + CP_HQD_PQ_CONTROL, QUEUE_SIZE); + uint64_t guessed_wptr = m->cp_hqd_pq_rptr & (queue_size - 1); + + if ((m->cp_hqd_pq_wptr_lo & (queue_size - 1)) < guessed_wptr) + guessed_wptr += queue_size; + guessed_wptr += m->cp_hqd_pq_wptr_lo & ~(queue_size - 1); + guessed_wptr += (uint64_t)m->cp_hqd_pq_wptr_hi << 32; + + WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_LO), + lower_32_bits(guessed_wptr)); + WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI), + upper_32_bits(guessed_wptr)); + WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR), + lower_32_bits((uintptr_t)wptr)); + WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_POLL_ADDR_HI), + upper_32_bits((uintptr_t)wptr)); + WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_PQ_WPTR_POLL_CNTL1), + get_queue_mask(adev, pipe_id, queue_id)); + } + + /* Start the EOP fetcher */ + WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_EOP_RPTR), + REG_SET_FIELD(m->cp_hqd_eop_rptr, + CP_HQD_EOP_RPTR, INIT_FETCHER, 1)); + + data = REG_SET_FIELD(m->cp_hqd_active, CP_HQD_ACTIVE, ACTIVE, 1); + WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE), data); + + release_queue(kgd); + + return 0; +} + +static int kgd_hqd_dump(struct kgd_dev *kgd, + uint32_t pipe_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t i = 0, reg; +#define HQD_N_REGS 56 +#define DUMP_REG(addr) do { \ + if (WARN_ON_ONCE(i >= HQD_N_REGS)) \ + break; \ + (*dump)[i][0] = (addr) << 2; \ + (*dump)[i++][1] = RREG32(addr); \ + } while (0) + + *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL); + if (*dump == NULL) + return -ENOMEM; + + acquire_queue(kgd, pipe_id, queue_id); + + for (reg = SOC15_REG_OFFSET(GC, 0, mmCP_MQD_BASE_ADDR); + reg <= SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_WPTR_HI); reg++) + DUMP_REG(reg); + + release_queue(kgd); + + WARN_ON_ONCE(i != HQD_N_REGS); + *n_regs = i; + + return 0; +} + +static int kgd_hqd_sdma_load(struct kgd_dev *kgd, void *mqd, + uint32_t __user *wptr, struct mm_struct *mm) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + struct v9_sdma_mqd *m; + uint32_t sdma_base_addr, sdmax_gfx_context_cntl; + unsigned long end_jiffies; + uint32_t data; + uint64_t data64; + uint64_t __user *wptr64 = (uint64_t __user *)wptr; + + m = get_sdma_mqd(mqd); + sdma_base_addr = get_sdma_base_addr(adev, m->sdma_engine_id, + m->sdma_queue_id); + sdmax_gfx_context_cntl = m->sdma_engine_id ? + SOC15_REG_OFFSET(SDMA1, 0, mmSDMA1_GFX_CONTEXT_CNTL) : + SOC15_REG_OFFSET(SDMA0, 0, mmSDMA0_GFX_CONTEXT_CNTL); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, + m->sdmax_rlcx_rb_cntl & (~SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK)); + + end_jiffies = msecs_to_jiffies(2000) + jiffies; + while (true) { + data = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS); + if (data & SDMA0_RLC0_CONTEXT_STATUS__IDLE_MASK) + break; + if (time_after(jiffies, end_jiffies)) + return -ETIME; + usleep_range(500, 1000); + } + data = RREG32(sdmax_gfx_context_cntl); + data = REG_SET_FIELD(data, SDMA0_GFX_CONTEXT_CNTL, + RESUME_CTX, 0); + WREG32(sdmax_gfx_context_cntl, data); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL_OFFSET, + m->sdmax_rlcx_doorbell_offset); + + data = REG_SET_FIELD(m->sdmax_rlcx_doorbell, SDMA0_RLC0_DOORBELL, + ENABLE, 1); + WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, data); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR, m->sdmax_rlcx_rb_rptr); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_HI, + m->sdmax_rlcx_rb_rptr_hi); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_MINOR_PTR_UPDATE, 1); + if (read_user_wptr(mm, wptr64, data64)) { + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, + lower_32_bits(data64)); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR_HI, + upper_32_bits(data64)); + } else { + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR, + m->sdmax_rlcx_rb_rptr); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_WPTR_HI, + m->sdmax_rlcx_rb_rptr_hi); + } + WREG32(sdma_base_addr + mmSDMA0_RLC0_MINOR_PTR_UPDATE, 0); + + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE, m->sdmax_rlcx_rb_base); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_BASE_HI, + m->sdmax_rlcx_rb_base_hi); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_ADDR_LO, + m->sdmax_rlcx_rb_rptr_addr_lo); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_ADDR_HI, + m->sdmax_rlcx_rb_rptr_addr_hi); + + data = REG_SET_FIELD(m->sdmax_rlcx_rb_cntl, SDMA0_RLC0_RB_CNTL, + RB_ENABLE, 1); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, data); + + return 0; +} + +static int kgd_hqd_sdma_dump(struct kgd_dev *kgd, + uint32_t engine_id, uint32_t queue_id, + uint32_t (**dump)[2], uint32_t *n_regs) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t sdma_base_addr = get_sdma_base_addr(adev, engine_id, queue_id); + uint32_t i = 0, reg; +#undef HQD_N_REGS +#define HQD_N_REGS (19+6+7+10) + + *dump = kmalloc(HQD_N_REGS*2*sizeof(uint32_t), GFP_KERNEL); + if (*dump == NULL) + return -ENOMEM; + + for (reg = mmSDMA0_RLC0_RB_CNTL; reg <= mmSDMA0_RLC0_DOORBELL; reg++) + DUMP_REG(sdma_base_addr + reg); + for (reg = mmSDMA0_RLC0_STATUS; reg <= mmSDMA0_RLC0_CSA_ADDR_HI; reg++) + DUMP_REG(sdma_base_addr + reg); + for (reg = mmSDMA0_RLC0_IB_SUB_REMAIN; + reg <= mmSDMA0_RLC0_MINOR_PTR_UPDATE; reg++) + DUMP_REG(sdma_base_addr + reg); + for (reg = mmSDMA0_RLC0_MIDCMD_DATA0; + reg <= mmSDMA0_RLC0_MIDCMD_CNTL; reg++) + DUMP_REG(sdma_base_addr + reg); + + WARN_ON_ONCE(i != HQD_N_REGS); + *n_regs = i; + + return 0; +} + +static bool kgd_hqd_is_occupied(struct kgd_dev *kgd, uint64_t queue_address, + uint32_t pipe_id, uint32_t queue_id) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t act; + bool retval = false; + uint32_t low, high; + + acquire_queue(kgd, pipe_id, queue_id); + act = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE)); + if (act) { + low = lower_32_bits(queue_address >> 8); + high = upper_32_bits(queue_address >> 8); + + if (low == RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_BASE)) && + high == RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_PQ_BASE_HI))) + retval = true; + } + release_queue(kgd); + return retval; +} + +static bool kgd_hqd_sdma_is_occupied(struct kgd_dev *kgd, void *mqd) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + struct v9_sdma_mqd *m; + uint32_t sdma_base_addr; + uint32_t sdma_rlc_rb_cntl; + + m = get_sdma_mqd(mqd); + sdma_base_addr = get_sdma_base_addr(adev, m->sdma_engine_id, + m->sdma_queue_id); + + sdma_rlc_rb_cntl = RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL); + + if (sdma_rlc_rb_cntl & SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK) + return true; + + return false; +} + +static int kgd_hqd_destroy(struct kgd_dev *kgd, void *mqd, + enum kfd_preempt_type reset_type, + unsigned int utimeout, uint32_t pipe_id, + uint32_t queue_id) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + enum hqd_dequeue_request_type type; + unsigned long end_jiffies; + uint32_t temp; + struct v9_mqd *m = get_mqd(mqd); + + acquire_queue(kgd, pipe_id, queue_id); + + if (m->cp_hqd_vmid == 0) + WREG32_FIELD15(GC, 0, RLC_CP_SCHEDULERS, scheduler1, 0); + + switch (reset_type) { + case KFD_PREEMPT_TYPE_WAVEFRONT_DRAIN: + type = DRAIN_PIPE; + break; + case KFD_PREEMPT_TYPE_WAVEFRONT_RESET: + type = RESET_WAVES; + break; + default: + type = DRAIN_PIPE; + break; + } + + WREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_DEQUEUE_REQUEST), type); + + end_jiffies = (utimeout * HZ / 1000) + jiffies; + while (true) { + temp = RREG32(SOC15_REG_OFFSET(GC, 0, mmCP_HQD_ACTIVE)); + if (!(temp & CP_HQD_ACTIVE__ACTIVE_MASK)) + break; + if (time_after(jiffies, end_jiffies)) { + pr_err("cp queue preemption time out.\n"); + release_queue(kgd); + return -ETIME; + } + usleep_range(500, 1000); + } + + release_queue(kgd); + return 0; +} + +static int kgd_hqd_sdma_destroy(struct kgd_dev *kgd, void *mqd, + unsigned int utimeout) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + struct v9_sdma_mqd *m; + uint32_t sdma_base_addr; + uint32_t temp; + unsigned long end_jiffies = (utimeout * HZ / 1000) + jiffies; + + m = get_sdma_mqd(mqd); + sdma_base_addr = get_sdma_base_addr(adev, m->sdma_engine_id, + m->sdma_queue_id); + + temp = RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL); + temp = temp & ~SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK; + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, temp); + + while (true) { + temp = RREG32(sdma_base_addr + mmSDMA0_RLC0_CONTEXT_STATUS); + if (temp & SDMA0_RLC0_CONTEXT_STATUS__IDLE_MASK) + break; + if (time_after(jiffies, end_jiffies)) + return -ETIME; + usleep_range(500, 1000); + } + + WREG32(sdma_base_addr + mmSDMA0_RLC0_DOORBELL, 0); + WREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL, + RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_CNTL) | + SDMA0_RLC0_RB_CNTL__RB_ENABLE_MASK); + + m->sdmax_rlcx_rb_rptr = RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR); + m->sdmax_rlcx_rb_rptr_hi = + RREG32(sdma_base_addr + mmSDMA0_RLC0_RB_RPTR_HI); + + return 0; +} + +static bool get_atc_vmid_pasid_mapping_valid(struct kgd_dev *kgd, + uint8_t vmid) +{ + uint32_t reg; + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + + reg = RREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATC_VMID0_PASID_MAPPING) + + vmid); + return reg & ATC_VMID0_PASID_MAPPING__VALID_MASK; +} + +static uint16_t get_atc_vmid_pasid_mapping_pasid(struct kgd_dev *kgd, + uint8_t vmid) +{ + uint32_t reg; + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + + reg = RREG32(SOC15_REG_OFFSET(ATHUB, 0, mmATC_VMID0_PASID_MAPPING) + + vmid); + return reg & ATC_VMID0_PASID_MAPPING__PASID_MASK; +} + +static void write_vmid_invalidate_request(struct kgd_dev *kgd, uint8_t vmid) +{ + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + uint32_t req = (1 << vmid) | + (0 << VM_INVALIDATE_ENG16_REQ__FLUSH_TYPE__SHIFT) | /* legacy */ + VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PTES_MASK | + VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PDE0_MASK | + VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PDE1_MASK | + VM_INVALIDATE_ENG16_REQ__INVALIDATE_L2_PDE2_MASK | + VM_INVALIDATE_ENG16_REQ__INVALIDATE_L1_PTES_MASK; + + mutex_lock(&adev->srbm_mutex); + + /* Use legacy mode tlb invalidation. + * + * Currently on Raven the code below is broken for anything but + * legacy mode due to a MMHUB power gating problem. A workaround + * is for MMHUB to wait until the condition PER_VMID_INVALIDATE_REQ + * == PER_VMID_INVALIDATE_ACK instead of simply waiting for the ack + * bit. + * + * TODO 1: agree on the right set of invalidation registers for + * KFD use. Use the last one for now. Invalidate both GC and + * MMHUB. + * + * TODO 2: support range-based invalidation, requires kfg2kgd + * interface change + */ + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_ADDR_RANGE_LO32), + 0xffffffff); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_ADDR_RANGE_HI32), + 0x0000001f); + + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_LO32), + 0xffffffff); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmMMHUB_VM_INVALIDATE_ENG16_ADDR_RANGE_HI32), + 0x0000001f); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_REQ), req); + + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_INVALIDATE_ENG16_REQ), + req); + + while (!(RREG32(SOC15_REG_OFFSET(GC, 0, mmVM_INVALIDATE_ENG16_ACK)) & + (1 << vmid))) + cpu_relax(); + + while (!(RREG32(SOC15_REG_OFFSET(MMHUB, 0, + mmMMHUB_VM_INVALIDATE_ENG16_ACK)) & + (1 << vmid))) + cpu_relax(); + + mutex_unlock(&adev->srbm_mutex); + +} + +static int invalidate_tlbs_with_kiq(struct amdgpu_device *adev, uint16_t pasid) +{ + signed long r; + uint32_t seq; + struct amdgpu_ring *ring = &adev->gfx.kiq.ring; + + spin_lock(&adev->gfx.kiq.ring_lock); + amdgpu_ring_alloc(ring, 12); /* fence + invalidate_tlbs package*/ + amdgpu_ring_write(ring, PACKET3(PACKET3_INVALIDATE_TLBS, 0)); + amdgpu_ring_write(ring, + PACKET3_INVALIDATE_TLBS_DST_SEL(1) | + PACKET3_INVALIDATE_TLBS_ALL_HUB(1) | + PACKET3_INVALIDATE_TLBS_PASID(pasid) | + PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(0)); /* legacy */ + amdgpu_fence_emit_polling(ring, &seq); + amdgpu_ring_commit(ring); + spin_unlock(&adev->gfx.kiq.ring_lock); + + r = amdgpu_fence_wait_polling(ring, seq, adev->usec_timeout); + if (r < 1) { + DRM_ERROR("wait for kiq fence error: %ld.\n", r); + return -ETIME; + } + + return 0; +} + +static int invalidate_tlbs(struct kgd_dev *kgd, uint16_t pasid) +{ + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + int vmid; + struct amdgpu_ring *ring = &adev->gfx.kiq.ring; + + if (ring->ready) + return invalidate_tlbs_with_kiq(adev, pasid); + + for (vmid = 0; vmid < 16; vmid++) { + if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) + continue; + if (get_atc_vmid_pasid_mapping_valid(kgd, vmid)) { + if (get_atc_vmid_pasid_mapping_pasid(kgd, vmid) + == pasid) { + write_vmid_invalidate_request(kgd, vmid); + break; + } + } + } + + return 0; +} + +static int invalidate_tlbs_vmid(struct kgd_dev *kgd, uint16_t vmid) +{ + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + + if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) { + pr_err("non kfd vmid %d\n", vmid); + return 0; + } + + write_vmid_invalidate_request(kgd, vmid); + return 0; +} + +static int kgd_address_watch_disable(struct kgd_dev *kgd) +{ + return 0; +} + +static int kgd_address_watch_execute(struct kgd_dev *kgd, + unsigned int watch_point_id, + uint32_t cntl_val, + uint32_t addr_hi, + uint32_t addr_lo) +{ + return 0; +} + +static int kgd_wave_control_execute(struct kgd_dev *kgd, + uint32_t gfx_index_val, + uint32_t sq_cmd) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint32_t data = 0; + + mutex_lock(&adev->grbm_idx_mutex); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_INDEX), gfx_index_val); + WREG32(SOC15_REG_OFFSET(GC, 0, mmSQ_CMD), sq_cmd); + + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, + INSTANCE_BROADCAST_WRITES, 1); + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, + SH_BROADCAST_WRITES, 1); + data = REG_SET_FIELD(data, GRBM_GFX_INDEX, + SE_BROADCAST_WRITES, 1); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmGRBM_GFX_INDEX), data); + mutex_unlock(&adev->grbm_idx_mutex); + + return 0; +} + +static uint32_t kgd_address_watch_get_offset(struct kgd_dev *kgd, + unsigned int watch_point_id, + unsigned int reg_offset) +{ + return 0; +} + +static void set_scratch_backing_va(struct kgd_dev *kgd, + uint64_t va, uint32_t vmid) +{ + /* No longer needed on GFXv9. The scratch base address is + * passed to the shader by the CP. It's the user mode driver's + * responsibility. + */ +} + +/* FIXME: Does this need to be ASIC-specific code? */ +static uint16_t get_fw_version(struct kgd_dev *kgd, enum kgd_engine_type type) +{ + struct amdgpu_device *adev = (struct amdgpu_device *) kgd; + const union amdgpu_firmware_header *hdr; + + switch (type) { + case KGD_ENGINE_PFP: + hdr = (const union amdgpu_firmware_header *)adev->gfx.pfp_fw->data; + break; + + case KGD_ENGINE_ME: + hdr = (const union amdgpu_firmware_header *)adev->gfx.me_fw->data; + break; + + case KGD_ENGINE_CE: + hdr = (const union amdgpu_firmware_header *)adev->gfx.ce_fw->data; + break; + + case KGD_ENGINE_MEC1: + hdr = (const union amdgpu_firmware_header *)adev->gfx.mec_fw->data; + break; + + case KGD_ENGINE_MEC2: + hdr = (const union amdgpu_firmware_header *)adev->gfx.mec2_fw->data; + break; + + case KGD_ENGINE_RLC: + hdr = (const union amdgpu_firmware_header *)adev->gfx.rlc_fw->data; + break; + + case KGD_ENGINE_SDMA1: + hdr = (const union amdgpu_firmware_header *)adev->sdma.instance[0].fw->data; + break; + + case KGD_ENGINE_SDMA2: + hdr = (const union amdgpu_firmware_header *)adev->sdma.instance[1].fw->data; + break; + + default: + return 0; + } + + if (hdr == NULL) + return 0; + + /* Only 12 bit in use*/ + return hdr->common.ucode_version; +} + +static void set_vm_context_page_table_base(struct kgd_dev *kgd, uint32_t vmid, + uint32_t page_table_base) +{ + struct amdgpu_device *adev = get_amdgpu_device(kgd); + uint64_t base = (uint64_t)page_table_base << PAGE_SHIFT | + AMDGPU_PTE_VALID; + + if (!amdgpu_amdkfd_is_kfd_vmid(adev, vmid)) { + pr_err("trying to set page table base for wrong VMID %u\n", + vmid); + return; + } + + /* TODO: take advantage of per-process address space size. For + * now, all processes share the same address space size, like + * on GFX8 and older. + */ + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32) + (vmid*2), 0); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32) + (vmid*2), 0); + + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32) + (vmid*2), + lower_32_bits(adev->vm_manager.max_pfn - 1)); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32) + (vmid*2), + upper_32_bits(adev->vm_manager.max_pfn - 1)); + + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32) + (vmid*2), lower_32_bits(base)); + WREG32(SOC15_REG_OFFSET(MMHUB, 0, mmMMHUB_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32) + (vmid*2), upper_32_bits(base)); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_LO32) + (vmid*2), 0); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_START_ADDR_HI32) + (vmid*2), 0); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_LO32) + (vmid*2), + lower_32_bits(adev->vm_manager.max_pfn - 1)); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_END_ADDR_HI32) + (vmid*2), + upper_32_bits(adev->vm_manager.max_pfn - 1)); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LO32) + (vmid*2), lower_32_bits(base)); + WREG32(SOC15_REG_OFFSET(GC, 0, mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR_HI32) + (vmid*2), upper_32_bits(base)); +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c index 1d6e1479da385..ff8fd75f7ca51 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_amdkfd_gpuvm.c @@ -23,6 +23,8 @@ #define pr_fmt(fmt) "kfd2kgd: " fmt #include <linux/list.h> +#include <linux/pagemap.h> +#include <linux/sched/mm.h> #include <drm/drmP.h> #include "amdgpu_object.h" #include "amdgpu_vm.h" @@ -33,10 +35,20 @@ */ #define VI_BO_SIZE_ALIGN (0x8000) +/* BO flag to indicate a KFD userptr BO */ +#define AMDGPU_AMDKFD_USERPTR_BO (1ULL << 63) + +/* Userptr restore delay, just long enough to allow consecutive VM + * changes to accumulate + */ +#define AMDGPU_USERPTR_RESTORE_DELAY_MS 1 + /* Impose limit on how much memory KFD can use */ static struct { uint64_t max_system_mem_limit; + uint64_t max_userptr_mem_limit; int64_t system_mem_used; + int64_t userptr_mem_used; spinlock_t mem_limit_lock; } kfd_mem_limit; @@ -57,6 +69,7 @@ static const char * const domain_bit_to_string[] = { #define domain_string(domain) domain_bit_to_string[ffs(domain)-1] +static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work); static inline struct amdgpu_device *get_amdgpu_device(struct kgd_dev *kgd) @@ -78,6 +91,7 @@ static bool check_if_add_bo_to_vm(struct amdgpu_vm *avm, /* Set memory usage limits. Current, limits are * System (kernel) memory - 3/8th System RAM + * Userptr memory - 3/4th System RAM */ void amdgpu_amdkfd_gpuvm_init_mem_limits(void) { @@ -90,8 +104,10 @@ void amdgpu_amdkfd_gpuvm_init_mem_limits(void) spin_lock_init(&kfd_mem_limit.mem_limit_lock); kfd_mem_limit.max_system_mem_limit = (mem >> 1) - (mem >> 3); - pr_debug("Kernel memory limit %lluM\n", - (kfd_mem_limit.max_system_mem_limit >> 20)); + kfd_mem_limit.max_userptr_mem_limit = mem - (mem >> 2); + pr_debug("Kernel memory limit %lluM, userptr limit %lluM\n", + (kfd_mem_limit.max_system_mem_limit >> 20), + (kfd_mem_limit.max_userptr_mem_limit >> 20)); } static int amdgpu_amdkfd_reserve_system_mem_limit(struct amdgpu_device *adev, @@ -111,6 +127,16 @@ static int amdgpu_amdkfd_reserve_system_mem_limit(struct amdgpu_device *adev, goto err_no_mem; } kfd_mem_limit.system_mem_used += (acc_size + size); + } else if (domain == AMDGPU_GEM_DOMAIN_CPU) { + if ((kfd_mem_limit.system_mem_used + acc_size > + kfd_mem_limit.max_system_mem_limit) || + (kfd_mem_limit.userptr_mem_used + (size + acc_size) > + kfd_mem_limit.max_userptr_mem_limit)) { + ret = -ENOMEM; + goto err_no_mem; + } + kfd_mem_limit.system_mem_used += acc_size; + kfd_mem_limit.userptr_mem_used += size; } err_no_mem: spin_unlock(&kfd_mem_limit.mem_limit_lock); @@ -126,10 +152,16 @@ static void unreserve_system_mem_limit(struct amdgpu_device *adev, sizeof(struct amdgpu_bo)); spin_lock(&kfd_mem_limit.mem_limit_lock); - if (domain == AMDGPU_GEM_DOMAIN_GTT) + if (domain == AMDGPU_GEM_DOMAIN_GTT) { kfd_mem_limit.system_mem_used -= (acc_size + size); + } else if (domain == AMDGPU_GEM_DOMAIN_CPU) { + kfd_mem_limit.system_mem_used -= acc_size; + kfd_mem_limit.userptr_mem_used -= size; + } WARN_ONCE(kfd_mem_limit.system_mem_used < 0, "kfd system memory accounting unbalanced"); + WARN_ONCE(kfd_mem_limit.userptr_mem_used < 0, + "kfd userptr memory accounting unbalanced"); spin_unlock(&kfd_mem_limit.mem_limit_lock); } @@ -138,12 +170,17 @@ void amdgpu_amdkfd_unreserve_system_memory_limit(struct amdgpu_bo *bo) { spin_lock(&kfd_mem_limit.mem_limit_lock); - if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GTT) { + if (bo->flags & AMDGPU_AMDKFD_USERPTR_BO) { + kfd_mem_limit.system_mem_used -= bo->tbo.acc_size; + kfd_mem_limit.userptr_mem_used -= amdgpu_bo_size(bo); + } else if (bo->preferred_domains == AMDGPU_GEM_DOMAIN_GTT) { kfd_mem_limit.system_mem_used -= (bo->tbo.acc_size + amdgpu_bo_size(bo)); } WARN_ONCE(kfd_mem_limit.system_mem_used < 0, "kfd system memory accounting unbalanced"); + WARN_ONCE(kfd_mem_limit.userptr_mem_used < 0, + "kfd userptr memory accounting unbalanced"); spin_unlock(&kfd_mem_limit.mem_limit_lock); } @@ -506,7 +543,8 @@ static void remove_bo_from_vm(struct amdgpu_device *adev, } static void add_kgd_mem_to_kfd_bo_list(struct kgd_mem *mem, - struct amdkfd_process_info *process_info) + struct amdkfd_process_info *process_info, + bool userptr) { struct ttm_validate_buffer *entry = &mem->validate_list; struct amdgpu_bo *bo = mem->bo; @@ -515,10 +553,95 @@ static void add_kgd_mem_to_kfd_bo_list(struct kgd_mem *mem, entry->shared = true; entry->bo = &bo->tbo; mutex_lock(&process_info->lock); - list_add_tail(&entry->head, &process_info->kfd_bo_list); + if (userptr) + list_add_tail(&entry->head, &process_info->userptr_valid_list); + else + list_add_tail(&entry->head, &process_info->kfd_bo_list); mutex_unlock(&process_info->lock); } +/* Initializes user pages. It registers the MMU notifier and validates + * the userptr BO in the GTT domain. + * + * The BO must already be on the userptr_valid_list. Otherwise an + * eviction and restore may happen that leaves the new BO unmapped + * with the user mode queues running. + * + * Takes the process_info->lock to protect against concurrent restore + * workers. + * + * Returns 0 for success, negative errno for errors. + */ +static int init_user_pages(struct kgd_mem *mem, struct mm_struct *mm, + uint64_t user_addr) +{ + struct amdkfd_process_info *process_info = mem->process_info; + struct amdgpu_bo *bo = mem->bo; + struct ttm_operation_ctx ctx = { true, false }; + int ret = 0; + + mutex_lock(&process_info->lock); + + ret = amdgpu_ttm_tt_set_userptr(bo->tbo.ttm, user_addr, 0); + if (ret) { + pr_err("%s: Failed to set userptr: %d\n", __func__, ret); + goto out; + } + + ret = amdgpu_mn_register(bo, user_addr); + if (ret) { + pr_err("%s: Failed to register MMU notifier: %d\n", + __func__, ret); + goto out; + } + + /* If no restore worker is running concurrently, user_pages + * should not be allocated + */ + WARN(mem->user_pages, "Leaking user_pages array"); + + mem->user_pages = kvmalloc_array(bo->tbo.ttm->num_pages, + sizeof(struct page *), + GFP_KERNEL | __GFP_ZERO); + if (!mem->user_pages) { + pr_err("%s: Failed to allocate pages array\n", __func__); + ret = -ENOMEM; + goto unregister_out; + } + + ret = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm, mem->user_pages); + if (ret) { + pr_err("%s: Failed to get user pages: %d\n", __func__, ret); + goto free_out; + } + + amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, mem->user_pages); + + ret = amdgpu_bo_reserve(bo, true); + if (ret) { + pr_err("%s: Failed to reserve BO\n", __func__); + goto release_out; + } + amdgpu_ttm_placement_from_domain(bo, mem->domain); + ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + if (ret) + pr_err("%s: failed to validate BO\n", __func__); + amdgpu_bo_unreserve(bo); + +release_out: + if (ret) + release_pages(mem->user_pages, bo->tbo.ttm->num_pages); +free_out: + kvfree(mem->user_pages); + mem->user_pages = NULL; +unregister_out: + if (ret) + amdgpu_mn_unregister(bo); +out: + mutex_unlock(&process_info->lock); + return ret; +} + /* Reserving a BO and its page table BOs must happen atomically to * avoid deadlocks. Some operations update multiple VMs at once. Track * all the reservation info in a context structure. Optionally a sync @@ -748,7 +871,8 @@ static int update_gpuvm_pte(struct amdgpu_device *adev, } static int map_bo_to_gpuvm(struct amdgpu_device *adev, - struct kfd_bo_va_list *entry, struct amdgpu_sync *sync) + struct kfd_bo_va_list *entry, struct amdgpu_sync *sync, + bool no_update_pte) { int ret; @@ -762,6 +886,9 @@ static int map_bo_to_gpuvm(struct amdgpu_device *adev, return ret; } + if (no_update_pte) + return 0; + ret = update_gpuvm_pte(adev, entry, sync); if (ret) { pr_err("update_gpuvm_pte() failed\n"); @@ -820,6 +947,8 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, mutex_init(&info->lock); INIT_LIST_HEAD(&info->vm_list_head); INIT_LIST_HEAD(&info->kfd_bo_list); + INIT_LIST_HEAD(&info->userptr_valid_list); + INIT_LIST_HEAD(&info->userptr_inval_list); info->eviction_fence = amdgpu_amdkfd_fence_create(dma_fence_context_alloc(1), @@ -830,6 +959,11 @@ static int init_kfd_vm(struct amdgpu_vm *vm, void **process_info, goto create_evict_fence_fail; } + info->pid = get_task_pid(current->group_leader, PIDTYPE_PID); + atomic_set(&info->evicted_bos, 0); + INIT_DELAYED_WORK(&info->restore_userptr_work, + amdgpu_amdkfd_restore_userptr_worker); + *process_info = info; *ef = dma_fence_get(&info->eviction_fence->base); } @@ -872,6 +1006,7 @@ reserve_pd_fail: dma_fence_put(*ef); *ef = NULL; *process_info = NULL; + put_pid(info->pid); create_evict_fence_fail: mutex_destroy(&info->lock); kfree(info); @@ -967,8 +1102,12 @@ void amdgpu_amdkfd_gpuvm_destroy_cb(struct amdgpu_device *adev, /* Release per-process resources when last compute VM is destroyed */ if (!process_info->n_vms) { WARN_ON(!list_empty(&process_info->kfd_bo_list)); + WARN_ON(!list_empty(&process_info->userptr_valid_list)); + WARN_ON(!list_empty(&process_info->userptr_inval_list)); dma_fence_put(&process_info->eviction_fence->base); + cancel_delayed_work_sync(&process_info->restore_userptr_work); + put_pid(process_info->pid); mutex_destroy(&process_info->lock); kfree(process_info); } @@ -1003,9 +1142,11 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( { struct amdgpu_device *adev = get_amdgpu_device(kgd); struct amdgpu_vm *avm = (struct amdgpu_vm *)vm; + uint64_t user_addr = 0; struct amdgpu_bo *bo; + struct amdgpu_bo_param bp; int byte_align; - u32 alloc_domain; + u32 domain, alloc_domain; u64 alloc_flags; uint32_t mapping_flags; int ret; @@ -1014,14 +1155,21 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( * Check on which domain to allocate BO */ if (flags & ALLOC_MEM_FLAGS_VRAM) { - alloc_domain = AMDGPU_GEM_DOMAIN_VRAM; + domain = alloc_domain = AMDGPU_GEM_DOMAIN_VRAM; alloc_flags = AMDGPU_GEM_CREATE_VRAM_CLEARED; alloc_flags |= (flags & ALLOC_MEM_FLAGS_PUBLIC) ? AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED : AMDGPU_GEM_CREATE_NO_CPU_ACCESS; } else if (flags & ALLOC_MEM_FLAGS_GTT) { - alloc_domain = AMDGPU_GEM_DOMAIN_GTT; + domain = alloc_domain = AMDGPU_GEM_DOMAIN_GTT; + alloc_flags = 0; + } else if (flags & ALLOC_MEM_FLAGS_USERPTR) { + domain = AMDGPU_GEM_DOMAIN_GTT; + alloc_domain = AMDGPU_GEM_DOMAIN_CPU; alloc_flags = 0; + if (!offset || !*offset) + return -EINVAL; + user_addr = *offset; } else { return -EINVAL; } @@ -1069,8 +1217,14 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( pr_debug("\tcreate BO VA 0x%llx size 0x%llx domain %s\n", va, size, domain_string(alloc_domain)); - ret = amdgpu_bo_create(adev, size, byte_align, - alloc_domain, alloc_flags, ttm_bo_type_device, NULL, &bo); + memset(&bp, 0, sizeof(bp)); + bp.size = size; + bp.byte_align = byte_align; + bp.domain = alloc_domain; + bp.flags = alloc_flags; + bp.type = ttm_bo_type_device; + bp.resv = NULL; + ret = amdgpu_bo_create(adev, &bp, &bo); if (ret) { pr_debug("Failed to create BO on domain %s. ret %d\n", domain_string(alloc_domain), ret); @@ -1078,18 +1232,34 @@ int amdgpu_amdkfd_gpuvm_alloc_memory_of_gpu( } bo->kfd_bo = *mem; (*mem)->bo = bo; + if (user_addr) + bo->flags |= AMDGPU_AMDKFD_USERPTR_BO; (*mem)->va = va; - (*mem)->domain = alloc_domain; + (*mem)->domain = domain; (*mem)->mapped_to_gpu_memory = 0; (*mem)->process_info = avm->process_info; - add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info); + add_kgd_mem_to_kfd_bo_list(*mem, avm->process_info, user_addr); + + if (user_addr) { + ret = init_user_pages(*mem, current->mm, user_addr); + if (ret) { + mutex_lock(&avm->process_info->lock); + list_del(&(*mem)->validate_list.head); + mutex_unlock(&avm->process_info->lock); + goto allocate_init_user_pages_failed; + } + } if (offset) *offset = amdgpu_bo_mmap_offset(bo); return 0; +allocate_init_user_pages_failed: + amdgpu_bo_unref(&bo); + /* Don't unreserve system mem limit twice */ + goto err_reserve_system_mem; err_bo_create: unreserve_system_mem_limit(adev, size, alloc_domain); err_reserve_system_mem: @@ -1122,12 +1292,24 @@ int amdgpu_amdkfd_gpuvm_free_memory_of_gpu( * be freed anyway */ + /* No more MMU notifiers */ + amdgpu_mn_unregister(mem->bo); + /* Make sure restore workers don't access the BO any more */ bo_list_entry = &mem->validate_list; mutex_lock(&process_info->lock); list_del(&bo_list_entry->head); mutex_unlock(&process_info->lock); + /* Free user pages if necessary */ + if (mem->user_pages) { + pr_debug("%s: Freeing user_pages array\n", __func__); + if (mem->user_pages[0]) + release_pages(mem->user_pages, + mem->bo->tbo.ttm->num_pages); + kvfree(mem->user_pages); + } + ret = reserve_bo_and_cond_vms(mem, NULL, BO_VM_ALL, &ctx); if (unlikely(ret)) return ret; @@ -1173,21 +1355,32 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( struct kfd_bo_va_list *bo_va_entry = NULL; struct kfd_bo_va_list *bo_va_entry_aql = NULL; unsigned long bo_size; - - /* Make sure restore is not running concurrently. - */ - mutex_lock(&mem->process_info->lock); - - mutex_lock(&mem->lock); + bool is_invalid_userptr = false; bo = mem->bo; - if (!bo) { pr_err("Invalid BO when mapping memory to GPU\n"); - ret = -EINVAL; - goto out; + return -EINVAL; } + /* Make sure restore is not running concurrently. Since we + * don't map invalid userptr BOs, we rely on the next restore + * worker to do the mapping + */ + mutex_lock(&mem->process_info->lock); + + /* Lock mmap-sem. If we find an invalid userptr BO, we can be + * sure that the MMU notifier is no longer running + * concurrently and the queues are actually stopped + */ + if (amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) { + down_write(¤t->mm->mmap_sem); + is_invalid_userptr = atomic_read(&mem->invalid); + up_write(¤t->mm->mmap_sem); + } + + mutex_lock(&mem->lock); + domain = mem->domain; bo_size = bo->tbo.mem.size; @@ -1200,6 +1393,14 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( if (unlikely(ret)) goto out; + /* Userptr can be marked as "not invalid", but not actually be + * validated yet (still in the system domain). In that case + * the queues are still stopped and we can leave mapping for + * the next restore worker + */ + if (bo->tbo.mem.mem_type == TTM_PL_SYSTEM) + is_invalid_userptr = true; + if (check_if_add_bo_to_vm(avm, mem)) { ret = add_bo_to_vm(adev, mem, avm, false, &bo_va_entry); @@ -1217,7 +1418,8 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( goto add_bo_to_vm_failed; } - if (mem->mapped_to_gpu_memory == 0) { + if (mem->mapped_to_gpu_memory == 0 && + !amdgpu_ttm_tt_get_usermm(bo->tbo.ttm)) { /* Validate BO only once. The eviction fence gets added to BO * the first time it is mapped. Validate will wait for all * background evictions to complete. @@ -1235,7 +1437,8 @@ int amdgpu_amdkfd_gpuvm_map_memory_to_gpu( entry->va, entry->va + bo_size, entry); - ret = map_bo_to_gpuvm(adev, entry, ctx.sync); + ret = map_bo_to_gpuvm(adev, entry, ctx.sync, + is_invalid_userptr); if (ret) { pr_err("Failed to map radeon bo to gpuvm\n"); goto map_bo_to_gpuvm_failed; @@ -1418,6 +1621,337 @@ bo_reserve_failed: return ret; } +/* Evict a userptr BO by stopping the queues if necessary + * + * Runs in MMU notifier, may be in RECLAIM_FS context. This means it + * cannot do any memory allocations, and cannot take any locks that + * are held elsewhere while allocating memory. Therefore this is as + * simple as possible, using atomic counters. + * + * It doesn't do anything to the BO itself. The real work happens in + * restore, where we get updated page addresses. This function only + * ensures that GPU access to the BO is stopped. + */ +int amdgpu_amdkfd_evict_userptr(struct kgd_mem *mem, + struct mm_struct *mm) +{ + struct amdkfd_process_info *process_info = mem->process_info; + int invalid, evicted_bos; + int r = 0; + + invalid = atomic_inc_return(&mem->invalid); + evicted_bos = atomic_inc_return(&process_info->evicted_bos); + if (evicted_bos == 1) { + /* First eviction, stop the queues */ + r = kgd2kfd->quiesce_mm(mm); + if (r) + pr_err("Failed to quiesce KFD\n"); + schedule_delayed_work(&process_info->restore_userptr_work, + msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS)); + } + + return r; +} + +/* Update invalid userptr BOs + * + * Moves invalidated (evicted) userptr BOs from userptr_valid_list to + * userptr_inval_list and updates user pages for all BOs that have + * been invalidated since their last update. + */ +static int update_invalid_user_pages(struct amdkfd_process_info *process_info, + struct mm_struct *mm) +{ + struct kgd_mem *mem, *tmp_mem; + struct amdgpu_bo *bo; + struct ttm_operation_ctx ctx = { false, false }; + int invalid, ret; + + /* Move all invalidated BOs to the userptr_inval_list and + * release their user pages by migration to the CPU domain + */ + list_for_each_entry_safe(mem, tmp_mem, + &process_info->userptr_valid_list, + validate_list.head) { + if (!atomic_read(&mem->invalid)) + continue; /* BO is still valid */ + + bo = mem->bo; + + if (amdgpu_bo_reserve(bo, true)) + return -EAGAIN; + amdgpu_ttm_placement_from_domain(bo, AMDGPU_GEM_DOMAIN_CPU); + ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + amdgpu_bo_unreserve(bo); + if (ret) { + pr_err("%s: Failed to invalidate userptr BO\n", + __func__); + return -EAGAIN; + } + + list_move_tail(&mem->validate_list.head, + &process_info->userptr_inval_list); + } + + if (list_empty(&process_info->userptr_inval_list)) + return 0; /* All evicted userptr BOs were freed */ + + /* Go through userptr_inval_list and update any invalid user_pages */ + list_for_each_entry(mem, &process_info->userptr_inval_list, + validate_list.head) { + invalid = atomic_read(&mem->invalid); + if (!invalid) + /* BO hasn't been invalidated since the last + * revalidation attempt. Keep its BO list. + */ + continue; + + bo = mem->bo; + + if (!mem->user_pages) { + mem->user_pages = + kvmalloc_array(bo->tbo.ttm->num_pages, + sizeof(struct page *), + GFP_KERNEL | __GFP_ZERO); + if (!mem->user_pages) { + pr_err("%s: Failed to allocate pages array\n", + __func__); + return -ENOMEM; + } + } else if (mem->user_pages[0]) { + release_pages(mem->user_pages, bo->tbo.ttm->num_pages); + } + + /* Get updated user pages */ + ret = amdgpu_ttm_tt_get_user_pages(bo->tbo.ttm, + mem->user_pages); + if (ret) { + mem->user_pages[0] = NULL; + pr_info("%s: Failed to get user pages: %d\n", + __func__, ret); + /* Pretend it succeeded. It will fail later + * with a VM fault if the GPU tries to access + * it. Better than hanging indefinitely with + * stalled user mode queues. + */ + } + + /* Mark the BO as valid unless it was invalidated + * again concurrently + */ + if (atomic_cmpxchg(&mem->invalid, invalid, 0) != invalid) + return -EAGAIN; + } + + return 0; +} + +/* Validate invalid userptr BOs + * + * Validates BOs on the userptr_inval_list, and moves them back to the + * userptr_valid_list. Also updates GPUVM page tables with new page + * addresses and waits for the page table updates to complete. + */ +static int validate_invalid_user_pages(struct amdkfd_process_info *process_info) +{ + struct amdgpu_bo_list_entry *pd_bo_list_entries; + struct list_head resv_list, duplicates; + struct ww_acquire_ctx ticket; + struct amdgpu_sync sync; + + struct amdgpu_vm *peer_vm; + struct kgd_mem *mem, *tmp_mem; + struct amdgpu_bo *bo; + struct ttm_operation_ctx ctx = { false, false }; + int i, ret; + + pd_bo_list_entries = kcalloc(process_info->n_vms, + sizeof(struct amdgpu_bo_list_entry), + GFP_KERNEL); + if (!pd_bo_list_entries) { + pr_err("%s: Failed to allocate PD BO list entries\n", __func__); + return -ENOMEM; + } + + INIT_LIST_HEAD(&resv_list); + INIT_LIST_HEAD(&duplicates); + + /* Get all the page directory BOs that need to be reserved */ + i = 0; + list_for_each_entry(peer_vm, &process_info->vm_list_head, + vm_list_node) + amdgpu_vm_get_pd_bo(peer_vm, &resv_list, + &pd_bo_list_entries[i++]); + /* Add the userptr_inval_list entries to resv_list */ + list_for_each_entry(mem, &process_info->userptr_inval_list, + validate_list.head) { + list_add_tail(&mem->resv_list.head, &resv_list); + mem->resv_list.bo = mem->validate_list.bo; + mem->resv_list.shared = mem->validate_list.shared; + } + + /* Reserve all BOs and page tables for validation */ + ret = ttm_eu_reserve_buffers(&ticket, &resv_list, false, &duplicates); + WARN(!list_empty(&duplicates), "Duplicates should be empty"); + if (ret) + goto out; + + amdgpu_sync_create(&sync); + + /* Avoid triggering eviction fences when unmapping invalid + * userptr BOs (waits for all fences, doesn't use + * FENCE_OWNER_VM) + */ + list_for_each_entry(peer_vm, &process_info->vm_list_head, + vm_list_node) + amdgpu_amdkfd_remove_eviction_fence(peer_vm->root.base.bo, + process_info->eviction_fence, + NULL, NULL); + + ret = process_validate_vms(process_info); + if (ret) + goto unreserve_out; + + /* Validate BOs and update GPUVM page tables */ + list_for_each_entry_safe(mem, tmp_mem, + &process_info->userptr_inval_list, + validate_list.head) { + struct kfd_bo_va_list *bo_va_entry; + + bo = mem->bo; + + /* Copy pages array and validate the BO if we got user pages */ + if (mem->user_pages[0]) { + amdgpu_ttm_tt_set_user_pages(bo->tbo.ttm, + mem->user_pages); + amdgpu_ttm_placement_from_domain(bo, mem->domain); + ret = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); + if (ret) { + pr_err("%s: failed to validate BO\n", __func__); + goto unreserve_out; + } + } + + /* Validate succeeded, now the BO owns the pages, free + * our copy of the pointer array. Put this BO back on + * the userptr_valid_list. If we need to revalidate + * it, we need to start from scratch. + */ + kvfree(mem->user_pages); + mem->user_pages = NULL; + list_move_tail(&mem->validate_list.head, + &process_info->userptr_valid_list); + + /* Update mapping. If the BO was not validated + * (because we couldn't get user pages), this will + * clear the page table entries, which will result in + * VM faults if the GPU tries to access the invalid + * memory. + */ + list_for_each_entry(bo_va_entry, &mem->bo_va_list, bo_list) { + if (!bo_va_entry->is_mapped) + continue; + + ret = update_gpuvm_pte((struct amdgpu_device *) + bo_va_entry->kgd_dev, + bo_va_entry, &sync); + if (ret) { + pr_err("%s: update PTE failed\n", __func__); + /* make sure this gets validated again */ + atomic_inc(&mem->invalid); + goto unreserve_out; + } + } + } + + /* Update page directories */ + ret = process_update_pds(process_info, &sync); + +unreserve_out: + list_for_each_entry(peer_vm, &process_info->vm_list_head, + vm_list_node) + amdgpu_bo_fence(peer_vm->root.base.bo, + &process_info->eviction_fence->base, true); + ttm_eu_backoff_reservation(&ticket, &resv_list); + amdgpu_sync_wait(&sync, false); + amdgpu_sync_free(&sync); +out: + kfree(pd_bo_list_entries); + + return ret; +} + +/* Worker callback to restore evicted userptr BOs + * + * Tries to update and validate all userptr BOs. If successful and no + * concurrent evictions happened, the queues are restarted. Otherwise, + * reschedule for another attempt later. + */ +static void amdgpu_amdkfd_restore_userptr_worker(struct work_struct *work) +{ + struct delayed_work *dwork = to_delayed_work(work); + struct amdkfd_process_info *process_info = + container_of(dwork, struct amdkfd_process_info, + restore_userptr_work); + struct task_struct *usertask; + struct mm_struct *mm; + int evicted_bos; + + evicted_bos = atomic_read(&process_info->evicted_bos); + if (!evicted_bos) + return; + + /* Reference task and mm in case of concurrent process termination */ + usertask = get_pid_task(process_info->pid, PIDTYPE_PID); + if (!usertask) + return; + mm = get_task_mm(usertask); + if (!mm) { + put_task_struct(usertask); + return; + } + + mutex_lock(&process_info->lock); + + if (update_invalid_user_pages(process_info, mm)) + goto unlock_out; + /* userptr_inval_list can be empty if all evicted userptr BOs + * have been freed. In that case there is nothing to validate + * and we can just restart the queues. + */ + if (!list_empty(&process_info->userptr_inval_list)) { + if (atomic_read(&process_info->evicted_bos) != evicted_bos) + goto unlock_out; /* Concurrent eviction, try again */ + + if (validate_invalid_user_pages(process_info)) + goto unlock_out; + } + /* Final check for concurrent evicton and atomic update. If + * another eviction happens after successful update, it will + * be a first eviction that calls quiesce_mm. The eviction + * reference counting inside KFD will handle this case. + */ + if (atomic_cmpxchg(&process_info->evicted_bos, evicted_bos, 0) != + evicted_bos) + goto unlock_out; + evicted_bos = 0; + if (kgd2kfd->resume_mm(mm)) { + pr_err("%s: Failed to resume KFD\n", __func__); + /* No recovery from this failure. Probably the CP is + * hanging. No point trying again. + */ + } +unlock_out: + mutex_unlock(&process_info->lock); + mmput(mm); + put_task_struct(usertask); + + /* If validation failed, reschedule another attempt */ + if (evicted_bos) + schedule_delayed_work(&process_info->restore_userptr_work, + msecs_to_jiffies(AMDGPU_USERPTR_RESTORE_DELAY_MS)); +} + /** amdgpu_amdkfd_gpuvm_restore_process_bos - Restore all BOs for the given * KFD process identified by process_info * diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c index a0f48cb9b8f06..236915849cfe0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c @@ -322,3 +322,47 @@ int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev) return ret; } + +union gfx_info { + struct atom_gfx_info_v2_4 v24; +}; + +int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev) +{ + struct amdgpu_mode_info *mode_info = &adev->mode_info; + int index; + uint8_t frev, crev; + uint16_t data_offset; + + index = get_index_into_master_table(atom_master_list_of_data_tables_v2_1, + gfx_info); + if (amdgpu_atom_parse_data_header(mode_info->atom_context, index, NULL, + &frev, &crev, &data_offset)) { + union gfx_info *gfx_info = (union gfx_info *) + (mode_info->atom_context->bios + data_offset); + switch (crev) { + case 4: + adev->gfx.config.max_shader_engines = gfx_info->v24.gc_num_se; + adev->gfx.config.max_cu_per_sh = gfx_info->v24.gc_num_cu_per_sh; + adev->gfx.config.max_sh_per_se = gfx_info->v24.gc_num_sh_per_se; + adev->gfx.config.max_backends_per_se = gfx_info->v24.gc_num_rb_per_se; + adev->gfx.config.max_texture_channel_caches = gfx_info->v24.gc_num_tccs; + adev->gfx.config.max_gprs = le16_to_cpu(gfx_info->v24.gc_num_gprs); + adev->gfx.config.max_gs_threads = gfx_info->v24.gc_num_max_gs_thds; + adev->gfx.config.gs_vgt_table_depth = gfx_info->v24.gc_gs_table_depth; + adev->gfx.config.gs_prim_buffer_depth = + le16_to_cpu(gfx_info->v24.gc_gsprim_buff_depth); + adev->gfx.config.double_offchip_lds_buf = + gfx_info->v24.gc_double_offchip_lds_buffer; + adev->gfx.cu_info.wave_front_size = le16_to_cpu(gfx_info->v24.gc_wave_size); + adev->gfx.cu_info.max_waves_per_simd = le16_to_cpu(gfx_info->v24.gc_max_waves_per_simd); + adev->gfx.cu_info.max_scratch_slots_per_cu = gfx_info->v24.gc_max_scratch_slots_per_cu; + adev->gfx.cu_info.lds_size = le16_to_cpu(gfx_info->v24.gc_lds_size); + return 0; + default: + return -EINVAL; + } + + } + return -EINVAL; +} diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h index 7689c961c4ef6..20f158fd3b763 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.h @@ -30,5 +30,6 @@ int amdgpu_atomfirmware_allocate_fb_scratch(struct amdgpu_device *adev); int amdgpu_atomfirmware_get_vram_width(struct amdgpu_device *adev); int amdgpu_atomfirmware_get_vram_type(struct amdgpu_device *adev); int amdgpu_atomfirmware_get_clock_info(struct amdgpu_device *adev); +int amdgpu_atomfirmware_get_gfx_info(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c index 1ae5ae8c45a45..1bcb2b2473357 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atpx_handler.c @@ -550,7 +550,7 @@ static int amdgpu_atpx_init(void) * look up whether we are the integrated or discrete GPU (all asics). * Returns the client id. */ -static int amdgpu_atpx_get_client_id(struct pci_dev *pdev) +static enum vga_switcheroo_client_id amdgpu_atpx_get_client_id(struct pci_dev *pdev) { if (amdgpu_atpx_priv.dhandle == ACPI_HANDLE(&pdev->dev)) return VGA_SWITCHEROO_IGD; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c index 02b849be083bd..19cfff31f2e16 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_benchmark.c @@ -75,13 +75,20 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, { struct amdgpu_bo *dobj = NULL; struct amdgpu_bo *sobj = NULL; + struct amdgpu_bo_param bp; uint64_t saddr, daddr; int r, n; int time; + memset(&bp, 0, sizeof(bp)); + bp.size = size; + bp.byte_align = PAGE_SIZE; + bp.domain = sdomain; + bp.flags = 0; + bp.type = ttm_bo_type_kernel; + bp.resv = NULL; n = AMDGPU_BENCHMARK_ITERATIONS; - r = amdgpu_bo_create(adev, size, PAGE_SIZE,sdomain, 0, - ttm_bo_type_kernel, NULL, &sobj); + r = amdgpu_bo_create(adev, &bp, &sobj); if (r) { goto out_cleanup; } @@ -93,8 +100,8 @@ static void amdgpu_benchmark_move(struct amdgpu_device *adev, unsigned size, if (r) { goto out_cleanup; } - r = amdgpu_bo_create(adev, size, PAGE_SIZE, ddomain, 0, - ttm_bo_type_kernel, NULL, &dobj); + bp.domain = ddomain; + r = amdgpu_bo_create(adev, &bp, &dobj); if (r) { goto out_cleanup; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c index 71a57b2f7f04b..e950730f1933b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cgs.c @@ -23,7 +23,6 @@ */ #include <linux/list.h> #include <linux/slab.h> -#include <linux/pci.h> #include <drm/drmP.h> #include <linux/firmware.h> #include <drm/amdgpu_drm.h> @@ -109,121 +108,6 @@ static void amdgpu_cgs_write_ind_register(struct cgs_device *cgs_device, WARN(1, "Invalid indirect register space"); } -static int amdgpu_cgs_get_pci_resource(struct cgs_device *cgs_device, - enum cgs_resource_type resource_type, - uint64_t size, - uint64_t offset, - uint64_t *resource_base) -{ - CGS_FUNC_ADEV; - - if (resource_base == NULL) - return -EINVAL; - - switch (resource_type) { - case CGS_RESOURCE_TYPE_MMIO: - if (adev->rmmio_size == 0) - return -ENOENT; - if ((offset + size) > adev->rmmio_size) - return -EINVAL; - *resource_base = adev->rmmio_base; - return 0; - case CGS_RESOURCE_TYPE_DOORBELL: - if (adev->doorbell.size == 0) - return -ENOENT; - if ((offset + size) > adev->doorbell.size) - return -EINVAL; - *resource_base = adev->doorbell.base; - return 0; - case CGS_RESOURCE_TYPE_FB: - case CGS_RESOURCE_TYPE_IO: - case CGS_RESOURCE_TYPE_ROM: - default: - return -EINVAL; - } -} - -static const void *amdgpu_cgs_atom_get_data_table(struct cgs_device *cgs_device, - unsigned table, uint16_t *size, - uint8_t *frev, uint8_t *crev) -{ - CGS_FUNC_ADEV; - uint16_t data_start; - - if (amdgpu_atom_parse_data_header( - adev->mode_info.atom_context, table, size, - frev, crev, &data_start)) - return (uint8_t*)adev->mode_info.atom_context->bios + - data_start; - - return NULL; -} - -static int amdgpu_cgs_atom_get_cmd_table_revs(struct cgs_device *cgs_device, unsigned table, - uint8_t *frev, uint8_t *crev) -{ - CGS_FUNC_ADEV; - - if (amdgpu_atom_parse_cmd_header( - adev->mode_info.atom_context, table, - frev, crev)) - return 0; - - return -EINVAL; -} - -static int amdgpu_cgs_atom_exec_cmd_table(struct cgs_device *cgs_device, unsigned table, - void *args) -{ - CGS_FUNC_ADEV; - - return amdgpu_atom_execute_table( - adev->mode_info.atom_context, table, args); -} - -static int amdgpu_cgs_set_clockgating_state(struct cgs_device *cgs_device, - enum amd_ip_block_type block_type, - enum amd_clockgating_state state) -{ - CGS_FUNC_ADEV; - int i, r = -1; - - for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_blocks[i].status.valid) - continue; - - if (adev->ip_blocks[i].version->type == block_type) { - r = adev->ip_blocks[i].version->funcs->set_clockgating_state( - (void *)adev, - state); - break; - } - } - return r; -} - -static int amdgpu_cgs_set_powergating_state(struct cgs_device *cgs_device, - enum amd_ip_block_type block_type, - enum amd_powergating_state state) -{ - CGS_FUNC_ADEV; - int i, r = -1; - - for (i = 0; i < adev->num_ip_blocks; i++) { - if (!adev->ip_blocks[i].status.valid) - continue; - - if (adev->ip_blocks[i].version->type == block_type) { - r = adev->ip_blocks[i].version->funcs->set_powergating_state( - (void *)adev, - state); - break; - } - } - return r; -} - - static uint32_t fw_type_convert(struct cgs_device *cgs_device, uint32_t fw_type) { CGS_FUNC_ADEV; @@ -271,18 +155,6 @@ static uint32_t fw_type_convert(struct cgs_device *cgs_device, uint32_t fw_type) return result; } -static int amdgpu_cgs_rel_firmware(struct cgs_device *cgs_device, enum cgs_ucode_id type) -{ - CGS_FUNC_ADEV; - if ((CGS_UCODE_ID_SMU == type) || (CGS_UCODE_ID_SMU_SK == type)) { - release_firmware(adev->pm.fw); - adev->pm.fw = NULL; - return 0; - } - /* cannot release other firmware because they are not created by cgs */ - return -EINVAL; -} - static uint16_t amdgpu_get_firmware_version(struct cgs_device *cgs_device, enum cgs_ucode_id type) { @@ -326,34 +198,6 @@ static uint16_t amdgpu_get_firmware_version(struct cgs_device *cgs_device, return fw_version; } -static int amdgpu_cgs_enter_safe_mode(struct cgs_device *cgs_device, - bool en) -{ - CGS_FUNC_ADEV; - - if (adev->gfx.rlc.funcs->enter_safe_mode == NULL || - adev->gfx.rlc.funcs->exit_safe_mode == NULL) - return 0; - - if (en) - adev->gfx.rlc.funcs->enter_safe_mode(adev); - else - adev->gfx.rlc.funcs->exit_safe_mode(adev); - - return 0; -} - -static void amdgpu_cgs_lock_grbm_idx(struct cgs_device *cgs_device, - bool lock) -{ - CGS_FUNC_ADEV; - - if (lock) - mutex_lock(&adev->grbm_idx_mutex); - else - mutex_unlock(&adev->grbm_idx_mutex); -} - static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, enum cgs_ucode_id type, struct cgs_firmware_info *info) @@ -541,6 +385,9 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, case CHIP_POLARIS12: strcpy(fw_name, "amdgpu/polaris12_smc.bin"); break; + case CHIP_VEGAM: + strcpy(fw_name, "amdgpu/vegam_smc.bin"); + break; case CHIP_VEGA10: if ((adev->pdev->device == 0x687f) && ((adev->pdev->revision == 0xc0) || @@ -553,6 +400,9 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, case CHIP_VEGA12: strcpy(fw_name, "amdgpu/vega12_smc.bin"); break; + case CHIP_VEGA20: + strcpy(fw_name, "amdgpu/vega20_smc.bin"); + break; default: DRM_ERROR("SMC firmware not supported\n"); return -EINVAL; @@ -598,97 +448,12 @@ static int amdgpu_cgs_get_firmware_info(struct cgs_device *cgs_device, return 0; } -static int amdgpu_cgs_is_virtualization_enabled(void *cgs_device) -{ - CGS_FUNC_ADEV; - return amdgpu_sriov_vf(adev); -} - -static int amdgpu_cgs_get_active_displays_info(struct cgs_device *cgs_device, - struct cgs_display_info *info) -{ - CGS_FUNC_ADEV; - struct cgs_mode_info *mode_info; - - if (info == NULL) - return -EINVAL; - - mode_info = info->mode_info; - if (mode_info) - /* if the displays are off, vblank time is max */ - mode_info->vblank_time_us = 0xffffffff; - - if (!amdgpu_device_has_dc_support(adev)) { - struct amdgpu_crtc *amdgpu_crtc; - struct drm_device *ddev = adev->ddev; - struct drm_crtc *crtc; - uint32_t line_time_us, vblank_lines; - - if (adev->mode_info.num_crtc && adev->mode_info.mode_config_initialized) { - list_for_each_entry(crtc, - &ddev->mode_config.crtc_list, head) { - amdgpu_crtc = to_amdgpu_crtc(crtc); - if (crtc->enabled) { - info->active_display_mask |= (1 << amdgpu_crtc->crtc_id); - info->display_count++; - } - if (mode_info != NULL && - crtc->enabled && amdgpu_crtc->enabled && - amdgpu_crtc->hw_mode.clock) { - line_time_us = (amdgpu_crtc->hw_mode.crtc_htotal * 1000) / - amdgpu_crtc->hw_mode.clock; - vblank_lines = amdgpu_crtc->hw_mode.crtc_vblank_end - - amdgpu_crtc->hw_mode.crtc_vdisplay + - (amdgpu_crtc->v_border * 2); - mode_info->vblank_time_us = vblank_lines * line_time_us; - mode_info->refresh_rate = drm_mode_vrefresh(&amdgpu_crtc->hw_mode); - /* we have issues with mclk switching with refresh rates - * over 120 hz on the non-DC code. - */ - if (mode_info->refresh_rate > 120) - mode_info->vblank_time_us = 0; - mode_info = NULL; - } - } - } - } else { - info->display_count = adev->pm.pm_display_cfg.num_display; - if (mode_info != NULL) { - mode_info->vblank_time_us = adev->pm.pm_display_cfg.min_vblank_time; - mode_info->refresh_rate = adev->pm.pm_display_cfg.vrefresh; - } - } - return 0; -} - - -static int amdgpu_cgs_notify_dpm_enabled(struct cgs_device *cgs_device, bool enabled) -{ - CGS_FUNC_ADEV; - - adev->pm.dpm_enabled = enabled; - - return 0; -} - static const struct cgs_ops amdgpu_cgs_ops = { .read_register = amdgpu_cgs_read_register, .write_register = amdgpu_cgs_write_register, .read_ind_register = amdgpu_cgs_read_ind_register, .write_ind_register = amdgpu_cgs_write_ind_register, - .get_pci_resource = amdgpu_cgs_get_pci_resource, - .atom_get_data_table = amdgpu_cgs_atom_get_data_table, - .atom_get_cmd_table_revs = amdgpu_cgs_atom_get_cmd_table_revs, - .atom_exec_cmd_table = amdgpu_cgs_atom_exec_cmd_table, .get_firmware_info = amdgpu_cgs_get_firmware_info, - .rel_firmware = amdgpu_cgs_rel_firmware, - .set_powergating_state = amdgpu_cgs_set_powergating_state, - .set_clockgating_state = amdgpu_cgs_set_clockgating_state, - .get_active_displays_info = amdgpu_cgs_get_active_displays_info, - .notify_dpm_enabled = amdgpu_cgs_notify_dpm_enabled, - .is_virtualization_enabled = amdgpu_cgs_is_virtualization_enabled, - .enter_safe_mode = amdgpu_cgs_enter_safe_mode, - .lock_grbm_idx = amdgpu_cgs_lock_grbm_idx, }; struct cgs_device *amdgpu_cgs_create_device(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c index 96501ff0e55b4..8e66851eb427b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_connectors.c @@ -691,7 +691,7 @@ static int amdgpu_connector_lvds_get_modes(struct drm_connector *connector) return ret; } -static int amdgpu_connector_lvds_mode_valid(struct drm_connector *connector, +static enum drm_mode_status amdgpu_connector_lvds_mode_valid(struct drm_connector *connector, struct drm_display_mode *mode) { struct drm_encoder *encoder = amdgpu_connector_best_single_encoder(connector); @@ -843,7 +843,7 @@ static int amdgpu_connector_vga_get_modes(struct drm_connector *connector) return ret; } -static int amdgpu_connector_vga_mode_valid(struct drm_connector *connector, +static enum drm_mode_status amdgpu_connector_vga_mode_valid(struct drm_connector *connector, struct drm_display_mode *mode) { struct drm_device *dev = connector->dev; @@ -1172,7 +1172,7 @@ static void amdgpu_connector_dvi_force(struct drm_connector *connector) amdgpu_connector->use_digital = true; } -static int amdgpu_connector_dvi_mode_valid(struct drm_connector *connector, +static enum drm_mode_status amdgpu_connector_dvi_mode_valid(struct drm_connector *connector, struct drm_display_mode *mode) { struct drm_device *dev = connector->dev; @@ -1448,7 +1448,7 @@ out: return ret; } -static int amdgpu_connector_dp_mode_valid(struct drm_connector *connector, +static enum drm_mode_status amdgpu_connector_dp_mode_valid(struct drm_connector *connector, struct drm_display_mode *mode) { struct amdgpu_connector *amdgpu_connector = to_amdgpu_connector(connector); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c index dc34b50e6b29b..9c1d491d742e0 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_cs.c @@ -382,8 +382,7 @@ retry: p->bytes_moved += ctx.bytes_moved; if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size && - bo->tbo.mem.mem_type == TTM_PL_VRAM && - bo->tbo.mem.start < adev->gmc.visible_vram_size >> PAGE_SHIFT) + amdgpu_bo_in_cpu_visible_vram(bo)) p->bytes_moved_vis += ctx.bytes_moved; if (unlikely(r == -ENOMEM) && domain != bo->allowed_domains) { @@ -411,7 +410,6 @@ static bool amdgpu_cs_try_evict(struct amdgpu_cs_parser *p, struct amdgpu_bo_list_entry *candidate = p->evictable; struct amdgpu_bo *bo = candidate->robj; struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); - u64 initial_bytes_moved, bytes_moved; bool update_bytes_moved_vis; uint32_t other; @@ -435,18 +433,14 @@ static bool amdgpu_cs_try_evict(struct amdgpu_cs_parser *p, continue; /* Good we can try to move this BO somewhere else */ - amdgpu_ttm_placement_from_domain(bo, other); update_bytes_moved_vis = adev->gmc.visible_vram_size < adev->gmc.real_vram_size && - bo->tbo.mem.mem_type == TTM_PL_VRAM && - bo->tbo.mem.start < adev->gmc.visible_vram_size >> PAGE_SHIFT; - initial_bytes_moved = atomic64_read(&adev->num_bytes_moved); + amdgpu_bo_in_cpu_visible_vram(bo); + amdgpu_ttm_placement_from_domain(bo, other); r = ttm_bo_validate(&bo->tbo, &bo->placement, &ctx); - bytes_moved = atomic64_read(&adev->num_bytes_moved) - - initial_bytes_moved; - p->bytes_moved += bytes_moved; + p->bytes_moved += ctx.bytes_moved; if (update_bytes_moved_vis) - p->bytes_moved_vis += bytes_moved; + p->bytes_moved_vis += ctx.bytes_moved; if (unlikely(r)) break; @@ -536,7 +530,7 @@ static int amdgpu_cs_parser_bos(struct amdgpu_cs_parser *p, if (p->bo_list) { amdgpu_bo_list_get_list(p->bo_list, &p->validated); if (p->bo_list->first_userptr != p->bo_list->num_entries) - p->mn = amdgpu_mn_get(p->adev); + p->mn = amdgpu_mn_get(p->adev, AMDGPU_MN_TYPE_GFX); } INIT_LIST_HEAD(&duplicates); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c index 09d35051fdd68..c5bb36275e937 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ctx.c @@ -91,7 +91,7 @@ static int amdgpu_ctx_init(struct amdgpu_device *adev, continue; r = drm_sched_entity_init(&ring->sched, &ctx->rings[i].entity, - rq, amdgpu_sched_jobs, &ctx->guilty); + rq, &ctx->guilty); if (r) goto failed; } @@ -111,8 +111,9 @@ failed: return r; } -static void amdgpu_ctx_fini(struct amdgpu_ctx *ctx) +static void amdgpu_ctx_fini(struct kref *ref) { + struct amdgpu_ctx *ctx = container_of(ref, struct amdgpu_ctx, refcount); struct amdgpu_device *adev = ctx->adev; unsigned i, j; @@ -125,13 +126,11 @@ static void amdgpu_ctx_fini(struct amdgpu_ctx *ctx) kfree(ctx->fences); ctx->fences = NULL; - for (i = 0; i < adev->num_rings; i++) - drm_sched_entity_fini(&adev->rings[i]->sched, - &ctx->rings[i].entity); - amdgpu_queue_mgr_fini(adev, &ctx->queue_mgr); mutex_destroy(&ctx->lock); + + kfree(ctx); } static int amdgpu_ctx_alloc(struct amdgpu_device *adev, @@ -170,12 +169,20 @@ static int amdgpu_ctx_alloc(struct amdgpu_device *adev, static void amdgpu_ctx_do_release(struct kref *ref) { struct amdgpu_ctx *ctx; + u32 i; ctx = container_of(ref, struct amdgpu_ctx, refcount); - amdgpu_ctx_fini(ctx); + for (i = 0; i < ctx->adev->num_rings; i++) { - kfree(ctx); + if (ctx->adev->rings[i] == &ctx->adev->gfx.kiq.ring) + continue; + + drm_sched_entity_fini(&ctx->adev->rings[i]->sched, + &ctx->rings[i].entity); + } + + amdgpu_ctx_fini(ref); } static int amdgpu_ctx_free(struct amdgpu_fpriv *fpriv, uint32_t id) @@ -419,9 +426,11 @@ int amdgpu_ctx_wait_prev_fence(struct amdgpu_ctx *ctx, unsigned ring_id) if (other) { signed long r; - r = dma_fence_wait_timeout(other, false, MAX_SCHEDULE_TIMEOUT); + r = dma_fence_wait(other, true); if (r < 0) { - DRM_ERROR("Error (%ld) waiting for fence!\n", r); + if (r != -ERESTARTSYS) + DRM_ERROR("Error (%ld) waiting for fence!\n", r); + return r; } } @@ -435,16 +444,72 @@ void amdgpu_ctx_mgr_init(struct amdgpu_ctx_mgr *mgr) idr_init(&mgr->ctx_handles); } +void amdgpu_ctx_mgr_entity_fini(struct amdgpu_ctx_mgr *mgr) +{ + struct amdgpu_ctx *ctx; + struct idr *idp; + uint32_t id, i; + + idp = &mgr->ctx_handles; + + idr_for_each_entry(idp, ctx, id) { + + if (!ctx->adev) + return; + + for (i = 0; i < ctx->adev->num_rings; i++) { + + if (ctx->adev->rings[i] == &ctx->adev->gfx.kiq.ring) + continue; + + if (kref_read(&ctx->refcount) == 1) + drm_sched_entity_do_release(&ctx->adev->rings[i]->sched, + &ctx->rings[i].entity); + else + DRM_ERROR("ctx %p is still alive\n", ctx); + } + } +} + +void amdgpu_ctx_mgr_entity_cleanup(struct amdgpu_ctx_mgr *mgr) +{ + struct amdgpu_ctx *ctx; + struct idr *idp; + uint32_t id, i; + + idp = &mgr->ctx_handles; + + idr_for_each_entry(idp, ctx, id) { + + if (!ctx->adev) + return; + + for (i = 0; i < ctx->adev->num_rings; i++) { + + if (ctx->adev->rings[i] == &ctx->adev->gfx.kiq.ring) + continue; + + if (kref_read(&ctx->refcount) == 1) + drm_sched_entity_cleanup(&ctx->adev->rings[i]->sched, + &ctx->rings[i].entity); + else + DRM_ERROR("ctx %p is still alive\n", ctx); + } + } +} + void amdgpu_ctx_mgr_fini(struct amdgpu_ctx_mgr *mgr) { struct amdgpu_ctx *ctx; struct idr *idp; uint32_t id; + amdgpu_ctx_mgr_entity_cleanup(mgr); + idp = &mgr->ctx_handles; idr_for_each_entry(idp, ctx, id) { - if (kref_put(&ctx->refcount, amdgpu_ctx_do_release) != 1) + if (kref_put(&ctx->refcount, amdgpu_ctx_fini) != 1) DRM_ERROR("ctx %p is still alive\n", ctx); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c index 448d69fe37561..f5fb93795a69a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_debugfs.c @@ -28,8 +28,13 @@ #include <linux/debugfs.h> #include "amdgpu.h" -/* - * Debugfs +/** + * amdgpu_debugfs_add_files - Add simple debugfs entries + * + * @adev: Device to attach debugfs entries to + * @files: Array of function callbacks that respond to reads + * @nfiles: Number of callbacks to register + * */ int amdgpu_debugfs_add_files(struct amdgpu_device *adev, const struct drm_info_list *files, @@ -64,7 +69,33 @@ int amdgpu_debugfs_add_files(struct amdgpu_device *adev, #if defined(CONFIG_DEBUG_FS) - +/** + * amdgpu_debugfs_process_reg_op - Handle MMIO register reads/writes + * + * @read: True if reading + * @f: open file handle + * @buf: User buffer to write/read to + * @size: Number of bytes to write/read + * @pos: Offset to seek to + * + * This debugfs entry has special meaning on the offset being sought. + * Various bits have different meanings: + * + * Bit 62: Indicates a GRBM bank switch is needed + * Bit 61: Indicates a SRBM bank switch is needed (implies bit 62 is + * zero) + * Bits 24..33: The SE or ME selector if needed + * Bits 34..43: The SH (or SA) or PIPE selector if needed + * Bits 44..53: The INSTANCE (or CU/WGP) or QUEUE selector if needed + * + * Bit 23: Indicates that the PM power gating lock should be held + * This is necessary to read registers that might be + * unreliable during a power gating transistion. + * + * The lower bits are the BYTE offset of the register to read. This + * allows reading multiple registers in a single call and having + * the returned size reflect that. + */ static int amdgpu_debugfs_process_reg_op(bool read, struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -164,19 +195,37 @@ end: return result; } - +/** + * amdgpu_debugfs_regs_read - Callback for reading MMIO registers + */ static ssize_t amdgpu_debugfs_regs_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { return amdgpu_debugfs_process_reg_op(true, f, buf, size, pos); } +/** + * amdgpu_debugfs_regs_write - Callback for writing MMIO registers + */ static ssize_t amdgpu_debugfs_regs_write(struct file *f, const char __user *buf, size_t size, loff_t *pos) { return amdgpu_debugfs_process_reg_op(false, f, (char __user *)buf, size, pos); } + +/** + * amdgpu_debugfs_regs_pcie_read - Read from a PCIE register + * + * @f: open file handle + * @buf: User buffer to store read data in + * @size: Number of bytes to read + * @pos: Offset to seek to + * + * The lower bits are the BYTE offset of the register to read. This + * allows reading multiple registers in a single call and having + * the returned size reflect that. + */ static ssize_t amdgpu_debugfs_regs_pcie_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -204,6 +253,18 @@ static ssize_t amdgpu_debugfs_regs_pcie_read(struct file *f, char __user *buf, return result; } +/** + * amdgpu_debugfs_regs_pcie_write - Write to a PCIE register + * + * @f: open file handle + * @buf: User buffer to write data from + * @size: Number of bytes to write + * @pos: Offset to seek to + * + * The lower bits are the BYTE offset of the register to write. This + * allows writing multiple registers in a single call and having + * the returned size reflect that. + */ static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user *buf, size_t size, loff_t *pos) { @@ -232,6 +293,18 @@ static ssize_t amdgpu_debugfs_regs_pcie_write(struct file *f, const char __user return result; } +/** + * amdgpu_debugfs_regs_didt_read - Read from a DIDT register + * + * @f: open file handle + * @buf: User buffer to store read data in + * @size: Number of bytes to read + * @pos: Offset to seek to + * + * The lower bits are the BYTE offset of the register to read. This + * allows reading multiple registers in a single call and having + * the returned size reflect that. + */ static ssize_t amdgpu_debugfs_regs_didt_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -259,6 +332,18 @@ static ssize_t amdgpu_debugfs_regs_didt_read(struct file *f, char __user *buf, return result; } +/** + * amdgpu_debugfs_regs_didt_write - Write to a DIDT register + * + * @f: open file handle + * @buf: User buffer to write data from + * @size: Number of bytes to write + * @pos: Offset to seek to + * + * The lower bits are the BYTE offset of the register to write. This + * allows writing multiple registers in a single call and having + * the returned size reflect that. + */ static ssize_t amdgpu_debugfs_regs_didt_write(struct file *f, const char __user *buf, size_t size, loff_t *pos) { @@ -287,6 +372,18 @@ static ssize_t amdgpu_debugfs_regs_didt_write(struct file *f, const char __user return result; } +/** + * amdgpu_debugfs_regs_smc_read - Read from a SMC register + * + * @f: open file handle + * @buf: User buffer to store read data in + * @size: Number of bytes to read + * @pos: Offset to seek to + * + * The lower bits are the BYTE offset of the register to read. This + * allows reading multiple registers in a single call and having + * the returned size reflect that. + */ static ssize_t amdgpu_debugfs_regs_smc_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -314,6 +411,18 @@ static ssize_t amdgpu_debugfs_regs_smc_read(struct file *f, char __user *buf, return result; } +/** + * amdgpu_debugfs_regs_smc_write - Write to a SMC register + * + * @f: open file handle + * @buf: User buffer to write data from + * @size: Number of bytes to write + * @pos: Offset to seek to + * + * The lower bits are the BYTE offset of the register to write. This + * allows writing multiple registers in a single call and having + * the returned size reflect that. + */ static ssize_t amdgpu_debugfs_regs_smc_write(struct file *f, const char __user *buf, size_t size, loff_t *pos) { @@ -342,6 +451,20 @@ static ssize_t amdgpu_debugfs_regs_smc_write(struct file *f, const char __user * return result; } +/** + * amdgpu_debugfs_gca_config_read - Read from gfx config data + * + * @f: open file handle + * @buf: User buffer to store read data in + * @size: Number of bytes to read + * @pos: Offset to seek to + * + * This file is used to access configuration data in a somewhat + * stable fashion. The format is a series of DWORDs with the first + * indicating which revision it is. New content is appended to the + * end so that older software can still read the data. + */ + static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -418,6 +541,19 @@ static ssize_t amdgpu_debugfs_gca_config_read(struct file *f, char __user *buf, return result; } +/** + * amdgpu_debugfs_sensor_read - Read from the powerplay sensors + * + * @f: open file handle + * @buf: User buffer to store read data in + * @size: Number of bytes to read + * @pos: Offset to seek to + * + * The offset is treated as the BYTE address of one of the sensors + * enumerated in amd/include/kgd_pp_interface.h under the + * 'amd_pp_sensors' enumeration. For instance to read the UVD VCLK + * you would use the offset 3 * 4 = 12. + */ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -428,7 +564,7 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf, if (size & 3 || *pos & 0x3) return -EINVAL; - if (amdgpu_dpm == 0) + if (!adev->pm.dpm_enabled) return -EINVAL; /* convert offset to sensor number */ @@ -457,6 +593,27 @@ static ssize_t amdgpu_debugfs_sensor_read(struct file *f, char __user *buf, return !r ? outsize : r; } +/** amdgpu_debugfs_wave_read - Read WAVE STATUS data + * + * @f: open file handle + * @buf: User buffer to store read data in + * @size: Number of bytes to read + * @pos: Offset to seek to + * + * The offset being sought changes which wave that the status data + * will be returned for. The bits are used as follows: + * + * Bits 0..6: Byte offset into data + * Bits 7..14: SE selector + * Bits 15..22: SH/SA selector + * Bits 23..30: CU/{WGP+SIMD} selector + * Bits 31..36: WAVE ID selector + * Bits 37..44: SIMD ID selector + * + * The returned data begins with one DWORD of version information + * Followed by WAVE STATUS registers relevant to the GFX IP version + * being used. See gfx_v8_0_read_wave_data() for an example output. + */ static ssize_t amdgpu_debugfs_wave_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -507,6 +664,28 @@ static ssize_t amdgpu_debugfs_wave_read(struct file *f, char __user *buf, return result; } +/** amdgpu_debugfs_gpr_read - Read wave gprs + * + * @f: open file handle + * @buf: User buffer to store read data in + * @size: Number of bytes to read + * @pos: Offset to seek to + * + * The offset being sought changes which wave that the status data + * will be returned for. The bits are used as follows: + * + * Bits 0..11: Byte offset into data + * Bits 12..19: SE selector + * Bits 20..27: SH/SA selector + * Bits 28..35: CU/{WGP+SIMD} selector + * Bits 36..43: WAVE ID selector + * Bits 37..44: SIMD ID selector + * Bits 52..59: Thread selector + * Bits 60..61: Bank selector (VGPR=0,SGPR=1) + * + * The return data comes from the SGPR or VGPR register bank for + * the selected operational unit. + */ static ssize_t amdgpu_debugfs_gpr_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -637,6 +816,12 @@ static const char *debugfs_regs_names[] = { "amdgpu_gpr", }; +/** + * amdgpu_debugfs_regs_init - Initialize debugfs entries that provide + * register access. + * + * @adev: The device to attach the debugfs entries to + */ int amdgpu_debugfs_regs_init(struct amdgpu_device *adev) { struct drm_minor *minor = adev->ddev->primary; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c index 34af664b9f93c..290e279abf0dc 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_device.c @@ -83,8 +83,10 @@ static const char *amdgpu_asic_name[] = { "POLARIS10", "POLARIS11", "POLARIS12", + "VEGAM", "VEGA10", "VEGA12", + "VEGA20", "RAVEN", "LAST", }; @@ -690,6 +692,8 @@ void amdgpu_device_gart_location(struct amdgpu_device *adev, { u64 size_af, size_bf; + mc->gart_size += adev->pm.smu_prv_buffer_size; + size_af = adev->gmc.mc_mask - mc->vram_end; size_bf = mc->vram_start; if (size_bf > size_af) { @@ -907,6 +911,46 @@ static void amdgpu_device_check_vm_size(struct amdgpu_device *adev) } } +static void amdgpu_device_check_smu_prv_buffer_size(struct amdgpu_device *adev) +{ + struct sysinfo si; + bool is_os_64 = (sizeof(void *) == 8) ? true : false; + uint64_t total_memory; + uint64_t dram_size_seven_GB = 0x1B8000000; + uint64_t dram_size_three_GB = 0xB8000000; + + if (amdgpu_smu_memory_pool_size == 0) + return; + + if (!is_os_64) { + DRM_WARN("Not 64-bit OS, feature not supported\n"); + goto def_value; + } + si_meminfo(&si); + total_memory = (uint64_t)si.totalram * si.mem_unit; + + if ((amdgpu_smu_memory_pool_size == 1) || + (amdgpu_smu_memory_pool_size == 2)) { + if (total_memory < dram_size_three_GB) + goto def_value1; + } else if ((amdgpu_smu_memory_pool_size == 4) || + (amdgpu_smu_memory_pool_size == 8)) { + if (total_memory < dram_size_seven_GB) + goto def_value1; + } else { + DRM_WARN("Smu memory pool size not supported\n"); + goto def_value; + } + adev->pm.smu_prv_buffer_size = amdgpu_smu_memory_pool_size << 28; + + return; + +def_value1: + DRM_WARN("No enough system memory\n"); +def_value: + adev->pm.smu_prv_buffer_size = 0; +} + /** * amdgpu_device_check_arguments - validate module params * @@ -948,6 +992,8 @@ static void amdgpu_device_check_arguments(struct amdgpu_device *adev) amdgpu_vm_fragment_size = -1; } + amdgpu_device_check_smu_prv_buffer_size(adev); + amdgpu_device_check_vm_size(adev); amdgpu_device_check_block_size(adev); @@ -1039,10 +1085,11 @@ static const struct vga_switcheroo_client_ops amdgpu_switcheroo_ops = { * the hardware IP specified. * Returns the error code from the last instance. */ -int amdgpu_device_ip_set_clockgating_state(struct amdgpu_device *adev, +int amdgpu_device_ip_set_clockgating_state(void *dev, enum amd_ip_block_type block_type, enum amd_clockgating_state state) { + struct amdgpu_device *adev = dev; int i, r = 0; for (i = 0; i < adev->num_ip_blocks; i++) { @@ -1072,10 +1119,11 @@ int amdgpu_device_ip_set_clockgating_state(struct amdgpu_device *adev, * the hardware IP specified. * Returns the error code from the last instance. */ -int amdgpu_device_ip_set_powergating_state(struct amdgpu_device *adev, +int amdgpu_device_ip_set_powergating_state(void *dev, enum amd_ip_block_type block_type, enum amd_powergating_state state) { + struct amdgpu_device *adev = dev; int i, r = 0; for (i = 0; i < adev->num_ip_blocks; i++) { @@ -1320,9 +1368,10 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev) case CHIP_TOPAZ: case CHIP_TONGA: case CHIP_FIJI: - case CHIP_POLARIS11: case CHIP_POLARIS10: + case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: case CHIP_CARRIZO: case CHIP_STONEY: #ifdef CONFIG_DRM_AMDGPU_SI @@ -1339,6 +1388,7 @@ static int amdgpu_device_parse_gpu_info_fw(struct amdgpu_device *adev) case CHIP_KABINI: case CHIP_MULLINS: #endif + case CHIP_VEGA20: default: return 0; case CHIP_VEGA10: @@ -1428,9 +1478,10 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) case CHIP_TOPAZ: case CHIP_TONGA: case CHIP_FIJI: - case CHIP_POLARIS11: case CHIP_POLARIS10: + case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: case CHIP_CARRIZO: case CHIP_STONEY: if (adev->asic_type == CHIP_CARRIZO || adev->asic_type == CHIP_STONEY) @@ -1472,6 +1523,7 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) #endif case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: case CHIP_RAVEN: if (adev->asic_type == CHIP_RAVEN) adev->family = AMDGPU_FAMILY_RV; @@ -1499,6 +1551,8 @@ static int amdgpu_device_ip_early_init(struct amdgpu_device *adev) return -EAGAIN; } + adev->powerplay.pp_feature = amdgpu_pp_feature_mask; + for (i = 0; i < adev->num_ip_blocks; i++) { if ((amdgpu_ip_block_mask & (1 << i)) == 0) { DRM_ERROR("disabled ip block: %d <%s>\n", @@ -1654,12 +1708,17 @@ static int amdgpu_device_ip_late_set_cg_state(struct amdgpu_device *adev) if (amdgpu_emu_mode == 1) return 0; + r = amdgpu_ib_ring_tests(adev); + if (r) + DRM_ERROR("ib ring test failed (%d).\n", r); + for (i = 0; i < adev->num_ip_blocks; i++) { if (!adev->ip_blocks[i].status.valid) continue; /* skip CG for VCE/UVD, it's handled specially */ if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD && adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCE && + adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCN && adev->ip_blocks[i].version->funcs->set_clockgating_state) { /* enable clockgating to save power */ r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev, @@ -1704,8 +1763,8 @@ static int amdgpu_device_ip_late_init(struct amdgpu_device *adev) } } - mod_delayed_work(system_wq, &adev->late_init_work, - msecs_to_jiffies(AMDGPU_RESUME_MS)); + queue_delayed_work(system_wq, &adev->late_init_work, + msecs_to_jiffies(AMDGPU_RESUME_MS)); amdgpu_device_fill_reset_magic(adev); @@ -1759,6 +1818,7 @@ static int amdgpu_device_ip_fini(struct amdgpu_device *adev) if (adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_UVD && adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCE && + adev->ip_blocks[i].version->type != AMD_IP_BLOCK_TYPE_VCN && adev->ip_blocks[i].version->funcs->set_clockgating_state) { /* ungate blocks before hw fini so that we can shutdown the blocks safely */ r = adev->ip_blocks[i].version->funcs->set_clockgating_state((void *)adev, @@ -1850,6 +1910,12 @@ int amdgpu_device_ip_suspend(struct amdgpu_device *adev) if (amdgpu_sriov_vf(adev)) amdgpu_virt_request_full_gpu(adev, false); + /* ungate SMC block powergating */ + if (adev->powerplay.pp_feature & PP_GFXOFF_MASK) + amdgpu_device_ip_set_powergating_state(adev, + AMD_IP_BLOCK_TYPE_SMC, + AMD_CG_STATE_UNGATE); + /* ungate SMC block first */ r = amdgpu_device_ip_set_clockgating_state(adev, AMD_IP_BLOCK_TYPE_SMC, AMD_CG_STATE_UNGATE); @@ -2086,16 +2152,15 @@ bool amdgpu_device_asic_has_dc_support(enum amd_asic_type asic_type) case CHIP_MULLINS: case CHIP_CARRIZO: case CHIP_STONEY: - case CHIP_POLARIS11: case CHIP_POLARIS10: + case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: case CHIP_TONGA: case CHIP_FIJI: -#if defined(CONFIG_DRM_AMD_DC_PRE_VEGA) - return amdgpu_dc != 0; -#endif case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: #if defined(CONFIG_DRM_AMD_DC_DCN1_0) case CHIP_RAVEN: #endif @@ -2375,10 +2440,6 @@ fence_driver_init: goto failed; } - r = amdgpu_ib_ring_tests(adev); - if (r) - DRM_ERROR("ib ring test failed (%d).\n", r); - if (amdgpu_sriov_vf(adev)) amdgpu_virt_init_data_exchange(adev); @@ -2539,7 +2600,7 @@ int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon) /* unpin the front buffers and cursors */ list_for_each_entry(crtc, &dev->mode_config.crtc_list, head) { struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - struct amdgpu_framebuffer *rfb = to_amdgpu_framebuffer(crtc->primary->fb); + struct drm_framebuffer *fb = crtc->primary->fb; struct amdgpu_bo *robj; if (amdgpu_crtc->cursor_bo) { @@ -2551,10 +2612,10 @@ int amdgpu_device_suspend(struct drm_device *dev, bool suspend, bool fbcon) } } - if (rfb == NULL || rfb->obj == NULL) { + if (fb == NULL || fb->obj[0] == NULL) { continue; } - robj = gem_to_amdgpu_bo(rfb->obj); + robj = gem_to_amdgpu_bo(fb->obj[0]); /* don't unpin kernel fb objects */ if (!amdgpu_fbdev_robj_is_fb(adev, robj)) { r = amdgpu_bo_reserve(robj, true); @@ -2640,11 +2701,6 @@ int amdgpu_device_resume(struct drm_device *dev, bool resume, bool fbcon) } amdgpu_fence_driver_resume(adev); - if (resume) { - r = amdgpu_ib_ring_tests(adev); - if (r) - DRM_ERROR("ib ring test failed (%d).\n", r); - } r = amdgpu_device_ip_late_init(adev); if (r) @@ -2736,6 +2792,9 @@ static bool amdgpu_device_ip_check_soft_reset(struct amdgpu_device *adev) if (amdgpu_sriov_vf(adev)) return true; + if (amdgpu_asic_need_full_reset(adev)) + return true; + for (i = 0; i < adev->num_ip_blocks; i++) { if (!adev->ip_blocks[i].status.valid) continue; @@ -2792,6 +2851,9 @@ static bool amdgpu_device_ip_need_full_reset(struct amdgpu_device *adev) { int i; + if (amdgpu_asic_need_full_reset(adev)) + return true; + for (i = 0; i < adev->num_ip_blocks; i++) { if (!adev->ip_blocks[i].status.valid) continue; @@ -3087,20 +3149,19 @@ static int amdgpu_device_reset_sriov(struct amdgpu_device *adev, /* now we are okay to resume SMC/CP/SDMA */ r = amdgpu_device_ip_reinit_late_sriov(adev); - amdgpu_virt_release_full_gpu(adev, true); if (r) goto error; amdgpu_irq_gpu_reset_resume_helper(adev); r = amdgpu_ib_ring_tests(adev); +error: + amdgpu_virt_release_full_gpu(adev, true); if (!r && adev->virt.gim_feature & AMDGIM_FEATURE_GIM_FLR_VRAMLOST) { atomic_inc(&adev->vram_lost_counter); r = amdgpu_device_handle_vram_lost(adev); } -error: - return r; } @@ -3117,7 +3178,6 @@ error: int amdgpu_device_gpu_recover(struct amdgpu_device *adev, struct amdgpu_job *job, bool force) { - struct drm_atomic_state *state = NULL; int i, r, resched; if (!force && !amdgpu_device_ip_check_soft_reset(adev)) { @@ -3140,10 +3200,6 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, /* block TTM */ resched = ttm_bo_lock_delayed_workqueue(&adev->mman.bdev); - /* store modesetting */ - if (amdgpu_device_has_dc_support(adev)) - state = drm_atomic_helper_suspend(adev->ddev); - /* block all schedulers and reset given job's ring */ for (i = 0; i < AMDGPU_MAX_RINGS; ++i) { struct amdgpu_ring *ring = adev->rings[i]; @@ -3183,10 +3239,7 @@ int amdgpu_device_gpu_recover(struct amdgpu_device *adev, kthread_unpark(ring->sched.thread); } - if (amdgpu_device_has_dc_support(adev)) { - if (drm_atomic_helper_resume(adev->ddev, state)) - dev_info(adev->dev, "drm resume failed:%d\n", r); - } else { + if (!amdgpu_device_has_dc_support(adev)) { drm_helper_resume_force_mode(adev->ddev); } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c index 93f700ab1bfbf..76ee8e04ff117 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.c @@ -35,6 +35,7 @@ #include <linux/pm_runtime.h> #include <drm/drm_crtc_helper.h> #include <drm/drm_edid.h> +#include <drm/drm_gem_framebuffer_helper.h> #include <drm/drm_fb_helper.h> static void amdgpu_display_flip_callback(struct dma_fence *f, @@ -151,8 +152,6 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc, struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); - struct amdgpu_framebuffer *old_amdgpu_fb; - struct amdgpu_framebuffer *new_amdgpu_fb; struct drm_gem_object *obj; struct amdgpu_flip_work *work; struct amdgpu_bo *new_abo; @@ -174,15 +173,13 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc, work->async = (page_flip_flags & DRM_MODE_PAGE_FLIP_ASYNC) != 0; /* schedule unpin of the old buffer */ - old_amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); - obj = old_amdgpu_fb->obj; + obj = crtc->primary->fb->obj[0]; /* take a reference to the old object */ work->old_abo = gem_to_amdgpu_bo(obj); amdgpu_bo_ref(work->old_abo); - new_amdgpu_fb = to_amdgpu_framebuffer(fb); - obj = new_amdgpu_fb->obj; + obj = fb->obj[0]; new_abo = gem_to_amdgpu_bo(obj); /* pin the new buffer */ @@ -192,7 +189,7 @@ int amdgpu_display_crtc_page_flip_target(struct drm_crtc *crtc, goto cleanup; } - r = amdgpu_bo_pin(new_abo, amdgpu_display_framebuffer_domains(adev), &base); + r = amdgpu_bo_pin(new_abo, amdgpu_display_supported_domains(adev), &base); if (unlikely(r != 0)) { DRM_ERROR("failed to pin new abo buffer before flip\n"); goto unreserve; @@ -482,31 +479,12 @@ bool amdgpu_display_ddc_probe(struct amdgpu_connector *amdgpu_connector, return true; } -static void amdgpu_display_user_framebuffer_destroy(struct drm_framebuffer *fb) -{ - struct amdgpu_framebuffer *amdgpu_fb = to_amdgpu_framebuffer(fb); - - drm_gem_object_put_unlocked(amdgpu_fb->obj); - drm_framebuffer_cleanup(fb); - kfree(amdgpu_fb); -} - -static int amdgpu_display_user_framebuffer_create_handle( - struct drm_framebuffer *fb, - struct drm_file *file_priv, - unsigned int *handle) -{ - struct amdgpu_framebuffer *amdgpu_fb = to_amdgpu_framebuffer(fb); - - return drm_gem_handle_create(file_priv, amdgpu_fb->obj, handle); -} - static const struct drm_framebuffer_funcs amdgpu_fb_funcs = { - .destroy = amdgpu_display_user_framebuffer_destroy, - .create_handle = amdgpu_display_user_framebuffer_create_handle, + .destroy = drm_gem_fb_destroy, + .create_handle = drm_gem_fb_create_handle, }; -uint32_t amdgpu_display_framebuffer_domains(struct amdgpu_device *adev) +uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev) { uint32_t domain = AMDGPU_GEM_DOMAIN_VRAM; @@ -526,11 +504,11 @@ int amdgpu_display_framebuffer_init(struct drm_device *dev, struct drm_gem_object *obj) { int ret; - rfb->obj = obj; + rfb->base.obj[0] = obj; drm_helper_mode_fill_fb_struct(dev, &rfb->base, mode_cmd); ret = drm_framebuffer_init(dev, &rfb->base, &amdgpu_fb_funcs); if (ret) { - rfb->obj = NULL; + rfb->base.obj[0] = NULL; return ret; } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h index 2b11d808f2977..f66e3e3fef0a5 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_display.h @@ -23,7 +23,7 @@ #ifndef __AMDGPU_DISPLAY_H__ #define __AMDGPU_DISPLAY_H__ -uint32_t amdgpu_display_framebuffer_domains(struct amdgpu_device *adev); +uint32_t amdgpu_display_supported_domains(struct amdgpu_device *adev); struct drm_framebuffer * amdgpu_display_user_framebuffer_create(struct drm_device *dev, struct drm_file *file_priv, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c index e997ebbe43ea0..def1010ac05e4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.c @@ -115,6 +115,26 @@ void amdgpu_dpm_print_ps_status(struct amdgpu_device *adev, pr_cont("\n"); } +void amdgpu_dpm_get_active_displays(struct amdgpu_device *adev) +{ + struct drm_device *ddev = adev->ddev; + struct drm_crtc *crtc; + struct amdgpu_crtc *amdgpu_crtc; + + adev->pm.dpm.new_active_crtcs = 0; + adev->pm.dpm.new_active_crtc_count = 0; + if (adev->mode_info.num_crtc && adev->mode_info.mode_config_initialized) { + list_for_each_entry(crtc, + &ddev->mode_config.crtc_list, head) { + amdgpu_crtc = to_amdgpu_crtc(crtc); + if (amdgpu_crtc->enabled) { + adev->pm.dpm.new_active_crtcs |= (1 << amdgpu_crtc->crtc_id); + adev->pm.dpm.new_active_crtc_count++; + } + } + } +} + u32 amdgpu_dpm_get_vblank_time(struct amdgpu_device *adev) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h index 643d008410c65..dd6203a0a6b77 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_dpm.h @@ -52,8 +52,6 @@ enum amdgpu_dpm_event_src { AMDGPU_DPM_EVENT_SRC_DIGIAL_OR_EXTERNAL = 4 }; -#define SCLK_DEEP_SLEEP_MASK 0x8 - struct amdgpu_ps { u32 caps; /* vbios flags */ u32 class; /* vbios flags */ @@ -349,12 +347,6 @@ enum amdgpu_pcie_gen { ((adev)->powerplay.pp_funcs->set_clockgating_by_smu(\ (adev)->powerplay.pp_handle, msg_id)) -#define amdgpu_dpm_notify_smu_memory_info(adev, virtual_addr_low, \ - virtual_addr_hi, mc_addr_low, mc_addr_hi, size) \ - ((adev)->powerplay.pp_funcs->notify_smu_memory_info)( \ - (adev)->powerplay.pp_handle, virtual_addr_low, \ - virtual_addr_hi, mc_addr_low, mc_addr_hi, size) - #define amdgpu_dpm_get_power_profile_mode(adev, buf) \ ((adev)->powerplay.pp_funcs->get_power_profile_mode(\ (adev)->powerplay.pp_handle, buf)) @@ -445,6 +437,8 @@ struct amdgpu_pm { uint32_t pcie_gen_mask; uint32_t pcie_mlw_mask; struct amd_pp_display_configuration pm_display_cfg;/* set by dc */ + uint32_t smu_prv_buffer_size; + struct amdgpu_bo *smu_prv_buffer; }; #define R600_SSTU_DFLT 0 @@ -482,6 +476,7 @@ void amdgpu_dpm_print_ps_status(struct amdgpu_device *adev, struct amdgpu_ps *rps); u32 amdgpu_dpm_get_vblank_time(struct amdgpu_device *adev); u32 amdgpu_dpm_get_vrefresh(struct amdgpu_device *adev); +void amdgpu_dpm_get_active_displays(struct amdgpu_device *adev); bool amdgpu_is_uvd_state(u32 class, u32 class2); void amdgpu_calculate_u_and_p(u32 i, u32 r_c, u32 p_b, u32 *p, u32 *u); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c index 0b19482b36b8a..b0bf2f24da48f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c @@ -75,9 +75,10 @@ * - 3.23.0 - Add query for VRAM lost counter * - 3.24.0 - Add high priority compute support for gfx9 * - 3.25.0 - Add support for sensor query info (stable pstate sclk/mclk). + * - 3.26.0 - GFX9: Process AMDGPU_IB_FLAG_TC_WB_NOT_INVALIDATE. */ #define KMS_DRIVER_MAJOR 3 -#define KMS_DRIVER_MINOR 25 +#define KMS_DRIVER_MINOR 26 #define KMS_DRIVER_PATCHLEVEL 0 int amdgpu_vram_limit = 0; @@ -121,7 +122,7 @@ uint amdgpu_pg_mask = 0xffffffff; uint amdgpu_sdma_phase_quantum = 32; char *amdgpu_disable_cu = NULL; char *amdgpu_virtual_display = NULL; -uint amdgpu_pp_feature_mask = 0xffffbfff; +uint amdgpu_pp_feature_mask = 0xffff3fff; /* gfxoff (bit 15) disabled by default */ int amdgpu_ngg = 0; int amdgpu_prim_buf_per_se = 0; int amdgpu_pos_buf_per_se = 0; @@ -132,6 +133,7 @@ int amdgpu_lbpw = -1; int amdgpu_compute_multipipe = -1; int amdgpu_gpu_recovery = -1; /* auto */ int amdgpu_emu_mode = 0; +uint amdgpu_smu_memory_pool_size = 0; MODULE_PARM_DESC(vramlimit, "Restrict VRAM for testing, in megabytes"); module_param_named(vramlimit, amdgpu_vram_limit, int, 0600); @@ -316,6 +318,11 @@ MODULE_PARM_DESC(cik_support, "CIK support (1 = enabled (default), 0 = disabled) module_param_named(cik_support, amdgpu_cik_support, int, 0444); #endif +MODULE_PARM_DESC(smu_memory_pool_size, + "reserve gtt for smu debug usage, 0 = disable," + "0x1 = 256Mbyte, 0x2 = 512Mbyte, 0x4 = 1 Gbyte, 0x8 = 2GByte"); +module_param_named(smu_memory_pool_size, amdgpu_smu_memory_pool_size, uint, 0444); + static const struct pci_device_id pciidlist[] = { #ifdef CONFIG_DRM_AMDGPU_SI {0x1002, 0x6780, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_TAHITI}, @@ -534,6 +541,9 @@ static const struct pci_device_id pciidlist[] = { {0x1002, 0x6995, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12}, {0x1002, 0x6997, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12}, {0x1002, 0x699F, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_POLARIS12}, + /* VEGAM */ + {0x1002, 0x694C, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGAM}, + {0x1002, 0x694E, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGAM}, /* Vega 10 */ {0x1002, 0x6860, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, {0x1002, 0x6861, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA10}, @@ -550,6 +560,13 @@ static const struct pci_device_id pciidlist[] = { {0x1002, 0x69A2, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA12}, {0x1002, 0x69A3, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA12}, {0x1002, 0x69AF, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA12}, + /* Vega 20 */ + {0x1002, 0x66A0, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, + {0x1002, 0x66A1, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, + {0x1002, 0x66A2, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, + {0x1002, 0x66A3, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, + {0x1002, 0x66A7, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, + {0x1002, 0x66AF, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_VEGA20|AMD_EXP_HW_SUPPORT}, /* Raven */ {0x1002, 0x15dd, PCI_ANY_ID, PCI_ANY_ID, 0, 0, CHIP_RAVEN|AMD_IS_APU}, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c index 12063019751bf..bc5fd8ebab5dd 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fb.c @@ -137,7 +137,7 @@ static int amdgpufb_create_pinned_object(struct amdgpu_fbdev *rfbdev, /* need to align pitch with crtc limits */ mode_cmd->pitches[0] = amdgpu_align_pitch(adev, mode_cmd->width, cpp, fb_tiled); - domain = amdgpu_display_framebuffer_domains(adev); + domain = amdgpu_display_supported_domains(adev); height = ALIGN(mode_cmd->height, 8); size = mode_cmd->pitches[0] * height; @@ -292,9 +292,9 @@ static int amdgpu_fbdev_destroy(struct drm_device *dev, struct amdgpu_fbdev *rfb drm_fb_helper_unregister_fbi(&rfbdev->helper); - if (rfb->obj) { - amdgpufb_destroy_pinned_object(rfb->obj); - rfb->obj = NULL; + if (rfb->base.obj[0]) { + amdgpufb_destroy_pinned_object(rfb->base.obj[0]); + rfb->base.obj[0] = NULL; drm_framebuffer_unregister_private(&rfb->base); drm_framebuffer_cleanup(&rfb->base); } @@ -377,7 +377,7 @@ int amdgpu_fbdev_total_size(struct amdgpu_device *adev) if (!adev->mode_info.rfbdev) return 0; - robj = gem_to_amdgpu_bo(adev->mode_info.rfbdev->rfb.obj); + robj = gem_to_amdgpu_bo(adev->mode_info.rfbdev->rfb.base.obj[0]); size += amdgpu_bo_size(robj); return size; } @@ -386,7 +386,7 @@ bool amdgpu_fbdev_robj_is_fb(struct amdgpu_device *adev, struct amdgpu_bo *robj) { if (!adev->mode_info.rfbdev) return false; - if (robj == gem_to_amdgpu_bo(adev->mode_info.rfbdev->rfb.obj)) + if (robj == gem_to_amdgpu_bo(adev->mode_info.rfbdev->rfb.base.obj[0])) return true; return false; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c index 97449e06a2424..39ec6b8890a1b 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_fence.c @@ -131,7 +131,8 @@ static u32 amdgpu_fence_read(struct amdgpu_ring *ring) * Emits a fence command on the requested ring (all asics). * Returns 0 on success, -ENOMEM on failure. */ -int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f) +int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f, + unsigned flags) { struct amdgpu_device *adev = ring->adev; struct amdgpu_fence *fence; @@ -149,7 +150,7 @@ int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **f) adev->fence_context + ring->idx, seq); amdgpu_ring_emit_fence(ring, ring->fence_drv.gpu_addr, - seq, AMDGPU_FENCE_FLAG_INT); + seq, flags | AMDGPU_FENCE_FLAG_INT); ptr = &ring->fence_drv.fences[seq & ring->fence_drv.num_fences_mask]; /* This function can't be called concurrently anyway, otherwise @@ -375,14 +376,14 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, struct amdgpu_device *adev = ring->adev; uint64_t index; - if (ring != &adev->uvd.ring) { + if (ring != &adev->uvd.inst[ring->me].ring) { ring->fence_drv.cpu_addr = &adev->wb.wb[ring->fence_offs]; ring->fence_drv.gpu_addr = adev->wb.gpu_addr + (ring->fence_offs * 4); } else { /* put fence directly behind firmware */ index = ALIGN(adev->uvd.fw->size, 8); - ring->fence_drv.cpu_addr = adev->uvd.cpu_addr + index; - ring->fence_drv.gpu_addr = adev->uvd.gpu_addr + index; + ring->fence_drv.cpu_addr = adev->uvd.inst[ring->me].cpu_addr + index; + ring->fence_drv.gpu_addr = adev->uvd.inst[ring->me].gpu_addr + index; } amdgpu_fence_write(ring, atomic_read(&ring->fence_drv.last_seq)); amdgpu_irq_get(adev, irq_src, irq_type); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c index cf0f186c60925..17d6b9fb6d776 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gart.c @@ -113,12 +113,17 @@ int amdgpu_gart_table_vram_alloc(struct amdgpu_device *adev) int r; if (adev->gart.robj == NULL) { - r = amdgpu_bo_create(adev, adev->gart.table_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - ttm_bo_type_kernel, NULL, - &adev->gart.robj); + struct amdgpu_bo_param bp; + + memset(&bp, 0, sizeof(bp)); + bp.size = adev->gart.table_size; + bp.byte_align = PAGE_SIZE; + bp.domain = AMDGPU_GEM_DOMAIN_VRAM; + bp.flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | + AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; + bp.type = ttm_bo_type_kernel; + bp.resv = NULL; + r = amdgpu_bo_create(adev, &bp, &adev->gart.robj); if (r) { return r; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c index 46b9ea4e61031..2c8e27370284d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_gem.c @@ -48,17 +48,25 @@ int amdgpu_gem_object_create(struct amdgpu_device *adev, unsigned long size, struct drm_gem_object **obj) { struct amdgpu_bo *bo; + struct amdgpu_bo_param bp; int r; + memset(&bp, 0, sizeof(bp)); *obj = NULL; /* At least align on page size */ if (alignment < PAGE_SIZE) { alignment = PAGE_SIZE; } + bp.size = size; + bp.byte_align = alignment; + bp.type = type; + bp.resv = resv; + bp.preferred_domain = initial_domain; retry: - r = amdgpu_bo_create(adev, size, alignment, initial_domain, - flags, type, resv, &bo); + bp.flags = flags; + bp.domain = initial_domain; + r = amdgpu_bo_create(adev, &bp, &bo); if (r) { if (r != -ERESTARTSYS) { if (flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) { @@ -221,12 +229,7 @@ int amdgpu_gem_create_ioctl(struct drm_device *dev, void *data, return -EINVAL; /* reject invalid gem domains */ - if (args->in.domains & ~(AMDGPU_GEM_DOMAIN_CPU | - AMDGPU_GEM_DOMAIN_GTT | - AMDGPU_GEM_DOMAIN_VRAM | - AMDGPU_GEM_DOMAIN_GDS | - AMDGPU_GEM_DOMAIN_GWS | - AMDGPU_GEM_DOMAIN_OA)) + if (args->in.domains & ~AMDGPU_GEM_DOMAIN_MASK) return -EINVAL; /* create a gem object to contain this object in */ @@ -771,16 +774,23 @@ int amdgpu_mode_dumb_create(struct drm_file *file_priv, } #if defined(CONFIG_DEBUG_FS) + +#define amdgpu_debugfs_gem_bo_print_flag(m, bo, flag) \ + if (bo->flags & (AMDGPU_GEM_CREATE_ ## flag)) { \ + seq_printf((m), " " #flag); \ + } + static int amdgpu_debugfs_gem_bo_info(int id, void *ptr, void *data) { struct drm_gem_object *gobj = ptr; struct amdgpu_bo *bo = gem_to_amdgpu_bo(gobj); struct seq_file *m = data; + struct dma_buf_attachment *attachment; + struct dma_buf *dma_buf; unsigned domain; const char *placement; unsigned pin_count; - uint64_t offset; domain = amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type); switch (domain) { @@ -798,13 +808,27 @@ static int amdgpu_debugfs_gem_bo_info(int id, void *ptr, void *data) seq_printf(m, "\t0x%08x: %12ld byte %s", id, amdgpu_bo_size(bo), placement); - offset = READ_ONCE(bo->tbo.mem.start); - if (offset != AMDGPU_BO_INVALID_OFFSET) - seq_printf(m, " @ 0x%010Lx", offset); - pin_count = READ_ONCE(bo->pin_count); if (pin_count) seq_printf(m, " pin count %d", pin_count); + + dma_buf = READ_ONCE(bo->gem_base.dma_buf); + attachment = READ_ONCE(bo->gem_base.import_attach); + + if (attachment) + seq_printf(m, " imported from %p", dma_buf); + else if (dma_buf) + seq_printf(m, " exported as %p", dma_buf); + + amdgpu_debugfs_gem_bo_print_flag(m, bo, CPU_ACCESS_REQUIRED); + amdgpu_debugfs_gem_bo_print_flag(m, bo, NO_CPU_ACCESS); + amdgpu_debugfs_gem_bo_print_flag(m, bo, CPU_GTT_USWC); + amdgpu_debugfs_gem_bo_print_flag(m, bo, VRAM_CLEARED); + amdgpu_debugfs_gem_bo_print_flag(m, bo, SHADOW); + amdgpu_debugfs_gem_bo_print_flag(m, bo, VRAM_CONTIGUOUS); + amdgpu_debugfs_gem_bo_print_flag(m, bo, VM_ALWAYS_VALID); + amdgpu_debugfs_gem_bo_print_flag(m, bo, EXPLICIT_SYNC); + seq_printf(m, "\n"); return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c index 311589e02d175..f70eeed9ed76f 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ib.c @@ -127,6 +127,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, struct amdgpu_vm *vm; uint64_t fence_ctx; uint32_t status = 0, alloc_size; + unsigned fence_flags = 0; unsigned i; int r = 0; @@ -227,7 +228,10 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, #endif amdgpu_asic_invalidate_hdp(adev, ring); - r = amdgpu_fence_emit(ring, f); + if (ib->flags & AMDGPU_IB_FLAG_TC_WB_NOT_INVALIDATE) + fence_flags |= AMDGPU_FENCE_FLAG_TC_WB_ONLY; + + r = amdgpu_fence_emit(ring, f, fence_flags); if (r) { dev_err(adev->dev, "failed to emit fence (%d)\n", r); if (job && job->vmid) @@ -242,7 +246,7 @@ int amdgpu_ib_schedule(struct amdgpu_ring *ring, unsigned num_ibs, /* wrap the last IB with fence */ if (job && job->uf_addr) { amdgpu_ring_emit_fence(ring, job->uf_addr, job->uf_sequence, - AMDGPU_FENCE_FLAG_64BIT); + fence_flags | AMDGPU_FENCE_FLAG_64BIT); } if (patch_offset != ~0 && ring->funcs->patch_cond_exec) diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c index 4b7824d30e730..91517b166a3b8 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c @@ -31,6 +31,7 @@ #include "amdgpu_sched.h" #include "amdgpu_uvd.h" #include "amdgpu_vce.h" +#include "atom.h" #include <linux/vga_switcheroo.h> #include <linux/slab.h> @@ -214,6 +215,18 @@ static int amdgpu_firmware_info(struct drm_amdgpu_info_firmware *fw_info, fw_info->ver = adev->gfx.rlc_fw_version; fw_info->feature = adev->gfx.rlc_feature_version; break; + case AMDGPU_INFO_FW_GFX_RLC_RESTORE_LIST_CNTL: + fw_info->ver = adev->gfx.rlc_srlc_fw_version; + fw_info->feature = adev->gfx.rlc_srlc_feature_version; + break; + case AMDGPU_INFO_FW_GFX_RLC_RESTORE_LIST_GPM_MEM: + fw_info->ver = adev->gfx.rlc_srlg_fw_version; + fw_info->feature = adev->gfx.rlc_srlg_feature_version; + break; + case AMDGPU_INFO_FW_GFX_RLC_RESTORE_LIST_SRM_MEM: + fw_info->ver = adev->gfx.rlc_srls_fw_version; + fw_info->feature = adev->gfx.rlc_srls_feature_version; + break; case AMDGPU_INFO_FW_GFX_MEC: if (query_fw->index == 0) { fw_info->ver = adev->gfx.mec_fw_version; @@ -273,12 +286,15 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file struct drm_crtc *crtc; uint32_t ui32 = 0; uint64_t ui64 = 0; - int i, found; + int i, j, found; int ui32_size = sizeof(ui32); if (!info->return_size || !info->return_pointer) return -EINVAL; + /* Ensure IB tests are run on ring */ + flush_delayed_work(&adev->late_init_work); + switch (info->query) { case AMDGPU_INFO_ACCEL_WORKING: ui32 = adev->accel_working; @@ -332,7 +348,8 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file break; case AMDGPU_HW_IP_UVD: type = AMD_IP_BLOCK_TYPE_UVD; - ring_mask = adev->uvd.ring.ready ? 1 : 0; + for (i = 0; i < adev->uvd.num_uvd_inst; i++) + ring_mask |= ((adev->uvd.inst[i].ring.ready ? 1 : 0) << i); ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; ib_size_alignment = 16; break; @@ -345,8 +362,11 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file break; case AMDGPU_HW_IP_UVD_ENC: type = AMD_IP_BLOCK_TYPE_UVD; - for (i = 0; i < adev->uvd.num_enc_rings; i++) - ring_mask |= ((adev->uvd.ring_enc[i].ready ? 1 : 0) << i); + for (i = 0; i < adev->uvd.num_uvd_inst; i++) + for (j = 0; j < adev->uvd.num_enc_rings; j++) + ring_mask |= + ((adev->uvd.inst[i].ring_enc[j].ready ? 1 : 0) << + (j + i * adev->uvd.num_enc_rings)); ib_start_alignment = AMDGPU_GPU_PAGE_SIZE; ib_size_alignment = 1; break; @@ -701,10 +721,7 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file } } case AMDGPU_INFO_SENSOR: { - struct pp_gpu_power query = {0}; - int query_size = sizeof(query); - - if (amdgpu_dpm == 0) + if (!adev->pm.dpm_enabled) return -ENOENT; switch (info->sensor_info.type) { @@ -746,10 +763,10 @@ static int amdgpu_info_ioctl(struct drm_device *dev, void *data, struct drm_file /* get average GPU power */ if (amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GPU_POWER, - (void *)&query, &query_size)) { + (void *)&ui32, &ui32_size)) { return -EINVAL; } - ui32 = query.average_gpu_power >> 8; + ui32 >>= 8; break; case AMDGPU_INFO_SENSOR_VDDNB: /* get VDDNB in millivolts */ @@ -913,8 +930,7 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, return; pm_runtime_get_sync(dev->dev); - - amdgpu_ctx_mgr_fini(&fpriv->ctx_mgr); + amdgpu_ctx_mgr_entity_fini(&fpriv->ctx_mgr); if (adev->asic_type != CHIP_RAVEN) { amdgpu_uvd_free_handles(adev, file_priv); @@ -935,6 +951,8 @@ void amdgpu_driver_postclose_kms(struct drm_device *dev, pd = amdgpu_bo_ref(fpriv->vm.root.base.bo); amdgpu_vm_fini(adev, &fpriv->vm); + amdgpu_ctx_mgr_fini(&fpriv->ctx_mgr); + if (pasid) amdgpu_pasid_free_delayed(pd->tbo.resv, pasid); amdgpu_bo_unref(&pd); @@ -1088,6 +1106,7 @@ static int amdgpu_debugfs_firmware_info(struct seq_file *m, void *data) struct amdgpu_device *adev = dev->dev_private; struct drm_amdgpu_info_firmware fw_info; struct drm_amdgpu_query_fw query_fw; + struct atom_context *ctx = adev->mode_info.atom_context; int ret, i; /* VCE */ @@ -1146,6 +1165,30 @@ static int amdgpu_debugfs_firmware_info(struct seq_file *m, void *data) seq_printf(m, "RLC feature version: %u, firmware version: 0x%08x\n", fw_info.feature, fw_info.ver); + /* RLC SAVE RESTORE LIST CNTL */ + query_fw.fw_type = AMDGPU_INFO_FW_GFX_RLC_RESTORE_LIST_CNTL; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "RLC SRLC feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* RLC SAVE RESTORE LIST GPM MEM */ + query_fw.fw_type = AMDGPU_INFO_FW_GFX_RLC_RESTORE_LIST_GPM_MEM; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "RLC SRLG feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + + /* RLC SAVE RESTORE LIST SRM MEM */ + query_fw.fw_type = AMDGPU_INFO_FW_GFX_RLC_RESTORE_LIST_SRM_MEM; + ret = amdgpu_firmware_info(&fw_info, &query_fw, adev); + if (ret) + return ret; + seq_printf(m, "RLC SRLS feature version: %u, firmware version: 0x%08x\n", + fw_info.feature, fw_info.ver); + /* MEC */ query_fw.fw_type = AMDGPU_INFO_FW_GFX_MEC; query_fw.index = 0; @@ -1210,6 +1253,9 @@ static int amdgpu_debugfs_firmware_info(struct seq_file *m, void *data) seq_printf(m, "VCN feature version: %u, firmware version: 0x%08x\n", fw_info.feature, fw_info.ver); + + seq_printf(m, "VBIOS version: %s\n", ctx->vbios_version); + return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c index bd67f4cb8e6ce..83e344fbb50a2 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.c @@ -36,12 +36,14 @@ #include <drm/drm.h> #include "amdgpu.h" +#include "amdgpu_amdkfd.h" struct amdgpu_mn { /* constant after initialisation */ struct amdgpu_device *adev; struct mm_struct *mm; struct mmu_notifier mn; + enum amdgpu_mn_type type; /* only used on destruction */ struct work_struct work; @@ -185,7 +187,7 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node, } /** - * amdgpu_mn_invalidate_range_start - callback to notify about mm change + * amdgpu_mn_invalidate_range_start_gfx - callback to notify about mm change * * @mn: our notifier * @mn: the mm this callback is about @@ -195,10 +197,10 @@ static void amdgpu_mn_invalidate_node(struct amdgpu_mn_node *node, * We block for all BOs between start and end to be idle and * unmap them by move them into system domain again. */ -static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn, - struct mm_struct *mm, - unsigned long start, - unsigned long end) +static void amdgpu_mn_invalidate_range_start_gfx(struct mmu_notifier *mn, + struct mm_struct *mm, + unsigned long start, + unsigned long end) { struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); struct interval_tree_node *it; @@ -220,6 +222,49 @@ static void amdgpu_mn_invalidate_range_start(struct mmu_notifier *mn, } /** + * amdgpu_mn_invalidate_range_start_hsa - callback to notify about mm change + * + * @mn: our notifier + * @mn: the mm this callback is about + * @start: start of updated range + * @end: end of updated range + * + * We temporarily evict all BOs between start and end. This + * necessitates evicting all user-mode queues of the process. The BOs + * are restorted in amdgpu_mn_invalidate_range_end_hsa. + */ +static void amdgpu_mn_invalidate_range_start_hsa(struct mmu_notifier *mn, + struct mm_struct *mm, + unsigned long start, + unsigned long end) +{ + struct amdgpu_mn *rmn = container_of(mn, struct amdgpu_mn, mn); + struct interval_tree_node *it; + + /* notification is exclusive, but interval is inclusive */ + end -= 1; + + amdgpu_mn_read_lock(rmn); + + it = interval_tree_iter_first(&rmn->objects, start, end); + while (it) { + struct amdgpu_mn_node *node; + struct amdgpu_bo *bo; + + node = container_of(it, struct amdgpu_mn_node, it); + it = interval_tree_iter_next(it, start, end); + + list_for_each_entry(bo, &node->bos, mn_list) { + struct kgd_mem *mem = bo->kfd_bo; + + if (amdgpu_ttm_tt_affect_userptr(bo->tbo.ttm, + start, end)) + amdgpu_amdkfd_evict_userptr(mem, mm); + } + } +} + +/** * amdgpu_mn_invalidate_range_end - callback to notify about mm change * * @mn: our notifier @@ -239,23 +284,39 @@ static void amdgpu_mn_invalidate_range_end(struct mmu_notifier *mn, amdgpu_mn_read_unlock(rmn); } -static const struct mmu_notifier_ops amdgpu_mn_ops = { - .release = amdgpu_mn_release, - .invalidate_range_start = amdgpu_mn_invalidate_range_start, - .invalidate_range_end = amdgpu_mn_invalidate_range_end, +static const struct mmu_notifier_ops amdgpu_mn_ops[] = { + [AMDGPU_MN_TYPE_GFX] = { + .release = amdgpu_mn_release, + .invalidate_range_start = amdgpu_mn_invalidate_range_start_gfx, + .invalidate_range_end = amdgpu_mn_invalidate_range_end, + }, + [AMDGPU_MN_TYPE_HSA] = { + .release = amdgpu_mn_release, + .invalidate_range_start = amdgpu_mn_invalidate_range_start_hsa, + .invalidate_range_end = amdgpu_mn_invalidate_range_end, + }, }; +/* Low bits of any reasonable mm pointer will be unused due to struct + * alignment. Use these bits to make a unique key from the mm pointer + * and notifier type. + */ +#define AMDGPU_MN_KEY(mm, type) ((unsigned long)(mm) + (type)) + /** * amdgpu_mn_get - create notifier context * * @adev: amdgpu device pointer + * @type: type of MMU notifier context * * Creates a notifier context for current->mm. */ -struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) +struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev, + enum amdgpu_mn_type type) { struct mm_struct *mm = current->mm; struct amdgpu_mn *rmn; + unsigned long key = AMDGPU_MN_KEY(mm, type); int r; mutex_lock(&adev->mn_lock); @@ -264,8 +325,8 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) return ERR_PTR(-EINTR); } - hash_for_each_possible(adev->mn_hash, rmn, node, (unsigned long)mm) - if (rmn->mm == mm) + hash_for_each_possible(adev->mn_hash, rmn, node, key) + if (AMDGPU_MN_KEY(rmn->mm, rmn->type) == key) goto release_locks; rmn = kzalloc(sizeof(*rmn), GFP_KERNEL); @@ -276,8 +337,9 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) rmn->adev = adev; rmn->mm = mm; - rmn->mn.ops = &amdgpu_mn_ops; init_rwsem(&rmn->lock); + rmn->type = type; + rmn->mn.ops = &amdgpu_mn_ops[type]; rmn->objects = RB_ROOT_CACHED; mutex_init(&rmn->read_lock); atomic_set(&rmn->recursion, 0); @@ -286,7 +348,7 @@ struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) if (r) goto free_rmn; - hash_add(adev->mn_hash, &rmn->node, (unsigned long)mm); + hash_add(adev->mn_hash, &rmn->node, AMDGPU_MN_KEY(mm, type)); release_locks: up_write(&mm->mmap_sem); @@ -315,15 +377,21 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) { unsigned long end = addr + amdgpu_bo_size(bo) - 1; struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + enum amdgpu_mn_type type = + bo->kfd_bo ? AMDGPU_MN_TYPE_HSA : AMDGPU_MN_TYPE_GFX; struct amdgpu_mn *rmn; - struct amdgpu_mn_node *node = NULL; + struct amdgpu_mn_node *node = NULL, *new_node; struct list_head bos; struct interval_tree_node *it; - rmn = amdgpu_mn_get(adev); + rmn = amdgpu_mn_get(adev, type); if (IS_ERR(rmn)) return PTR_ERR(rmn); + new_node = kmalloc(sizeof(*new_node), GFP_KERNEL); + if (!new_node) + return -ENOMEM; + INIT_LIST_HEAD(&bos); down_write(&rmn->lock); @@ -337,13 +405,10 @@ int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr) list_splice(&node->bos, &bos); } - if (!node) { - node = kmalloc(sizeof(struct amdgpu_mn_node), GFP_KERNEL); - if (!node) { - up_write(&rmn->lock); - return -ENOMEM; - } - } + if (!node) + node = new_node; + else + kfree(new_node); bo->mn = rmn; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h index d0095a3793b8e..eb0f432f78fee 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mn.h @@ -29,16 +29,23 @@ */ struct amdgpu_mn; +enum amdgpu_mn_type { + AMDGPU_MN_TYPE_GFX, + AMDGPU_MN_TYPE_HSA, +}; + #if defined(CONFIG_MMU_NOTIFIER) void amdgpu_mn_lock(struct amdgpu_mn *mn); void amdgpu_mn_unlock(struct amdgpu_mn *mn); -struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev); +struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev, + enum amdgpu_mn_type type); int amdgpu_mn_register(struct amdgpu_bo *bo, unsigned long addr); void amdgpu_mn_unregister(struct amdgpu_bo *bo); #else static inline void amdgpu_mn_lock(struct amdgpu_mn *mn) {} static inline void amdgpu_mn_unlock(struct amdgpu_mn *mn) {} -static inline struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev) +static inline struct amdgpu_mn *amdgpu_mn_get(struct amdgpu_device *adev, + enum amdgpu_mn_type type) { return NULL; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h index d6416ee52e321..b9e9e8b02fb75 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_mode.h @@ -308,7 +308,6 @@ struct amdgpu_display_funcs { struct amdgpu_framebuffer { struct drm_framebuffer base; - struct drm_gem_object *obj; /* caching for later use */ uint64_t address; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c index 6d08cde8443c6..6a9e46ae7f0a4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c @@ -191,14 +191,21 @@ int amdgpu_bo_create_reserved(struct amdgpu_device *adev, u32 domain, struct amdgpu_bo **bo_ptr, u64 *gpu_addr, void **cpu_addr) { + struct amdgpu_bo_param bp; bool free = false; int r; + memset(&bp, 0, sizeof(bp)); + bp.size = size; + bp.byte_align = align; + bp.domain = domain; + bp.flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | + AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; + bp.type = ttm_bo_type_kernel; + bp.resv = NULL; + if (!*bo_ptr) { - r = amdgpu_bo_create(adev, size, align, domain, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - ttm_bo_type_kernel, NULL, bo_ptr); + r = amdgpu_bo_create(adev, &bp, bo_ptr); if (r) { dev_err(adev->dev, "(%d) failed to allocate kernel bo\n", r); @@ -341,27 +348,25 @@ fail: return false; } -static int amdgpu_bo_do_create(struct amdgpu_device *adev, unsigned long size, - int byte_align, u32 domain, - u64 flags, enum ttm_bo_type type, - struct reservation_object *resv, +static int amdgpu_bo_do_create(struct amdgpu_device *adev, + struct amdgpu_bo_param *bp, struct amdgpu_bo **bo_ptr) { struct ttm_operation_ctx ctx = { - .interruptible = (type != ttm_bo_type_kernel), + .interruptible = (bp->type != ttm_bo_type_kernel), .no_wait_gpu = false, - .resv = resv, + .resv = bp->resv, .flags = TTM_OPT_FLAG_ALLOW_RES_EVICT }; struct amdgpu_bo *bo; - unsigned long page_align; + unsigned long page_align, size = bp->size; size_t acc_size; int r; - page_align = roundup(byte_align, PAGE_SIZE) >> PAGE_SHIFT; + page_align = roundup(bp->byte_align, PAGE_SIZE) >> PAGE_SHIFT; size = ALIGN(size, PAGE_SIZE); - if (!amdgpu_bo_validate_size(adev, size, domain)) + if (!amdgpu_bo_validate_size(adev, size, bp->domain)) return -ENOMEM; *bo_ptr = NULL; @@ -375,18 +380,14 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, unsigned long size, drm_gem_private_object_init(adev->ddev, &bo->gem_base, size); INIT_LIST_HEAD(&bo->shadow_list); INIT_LIST_HEAD(&bo->va); - bo->preferred_domains = domain & (AMDGPU_GEM_DOMAIN_VRAM | - AMDGPU_GEM_DOMAIN_GTT | - AMDGPU_GEM_DOMAIN_CPU | - AMDGPU_GEM_DOMAIN_GDS | - AMDGPU_GEM_DOMAIN_GWS | - AMDGPU_GEM_DOMAIN_OA); + bo->preferred_domains = bp->preferred_domain ? bp->preferred_domain : + bp->domain; bo->allowed_domains = bo->preferred_domains; - if (type != ttm_bo_type_kernel && + if (bp->type != ttm_bo_type_kernel && bo->allowed_domains == AMDGPU_GEM_DOMAIN_VRAM) bo->allowed_domains |= AMDGPU_GEM_DOMAIN_GTT; - bo->flags = flags; + bo->flags = bp->flags; #ifdef CONFIG_X86_32 /* XXX: Write-combined CPU mappings of GTT seem broken on 32-bit @@ -417,11 +418,13 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, unsigned long size, #endif bo->tbo.bdev = &adev->mman.bdev; - amdgpu_ttm_placement_from_domain(bo, domain); + amdgpu_ttm_placement_from_domain(bo, bp->domain); + if (bp->type == ttm_bo_type_kernel) + bo->tbo.priority = 1; - r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, type, + r = ttm_bo_init_reserved(&adev->mman.bdev, &bo->tbo, size, bp->type, &bo->placement, page_align, &ctx, acc_size, - NULL, resv, &amdgpu_ttm_bo_destroy); + NULL, bp->resv, &amdgpu_ttm_bo_destroy); if (unlikely(r != 0)) return r; @@ -433,10 +436,7 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, unsigned long size, else amdgpu_cs_report_moved_bytes(adev, ctx.bytes_moved, 0); - if (type == ttm_bo_type_kernel) - bo->tbo.priority = 1; - - if (flags & AMDGPU_GEM_CREATE_VRAM_CLEARED && + if (bp->flags & AMDGPU_GEM_CREATE_VRAM_CLEARED && bo->tbo.mem.placement & TTM_PL_FLAG_VRAM) { struct dma_fence *fence; @@ -449,20 +449,20 @@ static int amdgpu_bo_do_create(struct amdgpu_device *adev, unsigned long size, bo->tbo.moving = dma_fence_get(fence); dma_fence_put(fence); } - if (!resv) + if (!bp->resv) amdgpu_bo_unreserve(bo); *bo_ptr = bo; trace_amdgpu_bo_create(bo); /* Treat CPU_ACCESS_REQUIRED only as a hint if given by UMD */ - if (type == ttm_bo_type_device) + if (bp->type == ttm_bo_type_device) bo->flags &= ~AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED; return 0; fail_unreserve: - if (!resv) + if (!bp->resv) ww_mutex_unlock(&bo->tbo.resv->lock); amdgpu_bo_unref(&bo); return r; @@ -472,16 +472,22 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, unsigned long size, int byte_align, struct amdgpu_bo *bo) { + struct amdgpu_bo_param bp; int r; if (bo->shadow) return 0; - r = amdgpu_bo_do_create(adev, size, byte_align, AMDGPU_GEM_DOMAIN_GTT, - AMDGPU_GEM_CREATE_CPU_GTT_USWC | - AMDGPU_GEM_CREATE_SHADOW, - ttm_bo_type_kernel, - bo->tbo.resv, &bo->shadow); + memset(&bp, 0, sizeof(bp)); + bp.size = size; + bp.byte_align = byte_align; + bp.domain = AMDGPU_GEM_DOMAIN_GTT; + bp.flags = AMDGPU_GEM_CREATE_CPU_GTT_USWC | + AMDGPU_GEM_CREATE_SHADOW; + bp.type = ttm_bo_type_kernel; + bp.resv = bo->tbo.resv; + + r = amdgpu_bo_do_create(adev, &bp, &bo->shadow); if (!r) { bo->shadow->parent = amdgpu_bo_ref(bo); mutex_lock(&adev->shadow_list_lock); @@ -492,28 +498,26 @@ static int amdgpu_bo_create_shadow(struct amdgpu_device *adev, return r; } -int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size, - int byte_align, u32 domain, - u64 flags, enum ttm_bo_type type, - struct reservation_object *resv, +int amdgpu_bo_create(struct amdgpu_device *adev, + struct amdgpu_bo_param *bp, struct amdgpu_bo **bo_ptr) { - uint64_t parent_flags = flags & ~AMDGPU_GEM_CREATE_SHADOW; + u64 flags = bp->flags; int r; - r = amdgpu_bo_do_create(adev, size, byte_align, domain, - parent_flags, type, resv, bo_ptr); + bp->flags = bp->flags & ~AMDGPU_GEM_CREATE_SHADOW; + r = amdgpu_bo_do_create(adev, bp, bo_ptr); if (r) return r; if ((flags & AMDGPU_GEM_CREATE_SHADOW) && amdgpu_need_backup(adev)) { - if (!resv) + if (!bp->resv) WARN_ON(reservation_object_lock((*bo_ptr)->tbo.resv, NULL)); - r = amdgpu_bo_create_shadow(adev, size, byte_align, (*bo_ptr)); + r = amdgpu_bo_create_shadow(adev, bp->size, bp->byte_align, (*bo_ptr)); - if (!resv) + if (!bp->resv) reservation_object_unlock((*bo_ptr)->tbo.resv); if (r) @@ -689,8 +693,21 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, return -EINVAL; /* A shared bo cannot be migrated to VRAM */ - if (bo->prime_shared_count && (domain == AMDGPU_GEM_DOMAIN_VRAM)) - return -EINVAL; + if (bo->prime_shared_count) { + if (domain & AMDGPU_GEM_DOMAIN_GTT) + domain = AMDGPU_GEM_DOMAIN_GTT; + else + return -EINVAL; + } + + /* This assumes only APU display buffers are pinned with (VRAM|GTT). + * See function amdgpu_display_supported_domains() + */ + if (domain == (AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT)) { + domain = AMDGPU_GEM_DOMAIN_VRAM; + if (adev->gmc.real_vram_size <= AMDGPU_SG_THRESHOLD) + domain = AMDGPU_GEM_DOMAIN_GTT; + } if (bo->pin_count) { uint32_t mem_type = bo->tbo.mem.mem_type; @@ -838,6 +855,13 @@ int amdgpu_bo_init(struct amdgpu_device *adev) return amdgpu_ttm_init(adev); } +int amdgpu_bo_late_init(struct amdgpu_device *adev) +{ + amdgpu_ttm_late_init(adev); + + return 0; +} + void amdgpu_bo_fini(struct amdgpu_device *adev) { amdgpu_ttm_fini(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h index 546f77cb78829..540e03fa159f4 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.h @@ -33,6 +33,16 @@ #define AMDGPU_BO_INVALID_OFFSET LONG_MAX +struct amdgpu_bo_param { + unsigned long size; + int byte_align; + u32 domain; + u32 preferred_domain; + u64 flags; + enum ttm_bo_type type; + struct reservation_object *resv; +}; + /* bo virtual addresses in a vm */ struct amdgpu_bo_va_mapping { struct amdgpu_bo_va *bo_va; @@ -196,6 +206,27 @@ static inline bool amdgpu_bo_gpu_accessible(struct amdgpu_bo *bo) } /** + * amdgpu_bo_in_cpu_visible_vram - check if BO is (partly) in visible VRAM + */ +static inline bool amdgpu_bo_in_cpu_visible_vram(struct amdgpu_bo *bo) +{ + struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); + unsigned fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT; + struct drm_mm_node *node = bo->tbo.mem.mm_node; + unsigned long pages_left; + + if (bo->tbo.mem.mem_type != TTM_PL_VRAM) + return false; + + for (pages_left = bo->tbo.mem.num_pages; pages_left; + pages_left -= node->size, node++) + if (node->start < fpfn) + return true; + + return false; +} + +/** * amdgpu_bo_explicit_sync - return whether the bo is explicitly synced */ static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo) @@ -203,10 +234,8 @@ static inline bool amdgpu_bo_explicit_sync(struct amdgpu_bo *bo) return bo->flags & AMDGPU_GEM_CREATE_EXPLICIT_SYNC; } -int amdgpu_bo_create(struct amdgpu_device *adev, unsigned long size, - int byte_align, u32 domain, - u64 flags, enum ttm_bo_type type, - struct reservation_object *resv, +int amdgpu_bo_create(struct amdgpu_device *adev, + struct amdgpu_bo_param *bp, struct amdgpu_bo **bo_ptr); int amdgpu_bo_create_reserved(struct amdgpu_device *adev, unsigned long size, int align, @@ -230,6 +259,7 @@ int amdgpu_bo_pin_restricted(struct amdgpu_bo *bo, u32 domain, int amdgpu_bo_unpin(struct amdgpu_bo *bo); int amdgpu_bo_evict_vram(struct amdgpu_device *adev); int amdgpu_bo_init(struct amdgpu_device *adev); +int amdgpu_bo_late_init(struct amdgpu_device *adev); void amdgpu_bo_fini(struct amdgpu_device *adev); int amdgpu_bo_fbdev_mmap(struct amdgpu_bo *bo, struct vm_area_struct *vma); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c index 361975cf45a9f..b455da4877829 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_pm.c @@ -77,6 +77,37 @@ void amdgpu_pm_acpi_event_handler(struct amdgpu_device *adev) } } +/** + * DOC: power_dpm_state + * + * This is a legacy interface and is only provided for backwards compatibility. + * The amdgpu driver provides a sysfs API for adjusting certain power + * related parameters. The file power_dpm_state is used for this. + * It accepts the following arguments: + * - battery + * - balanced + * - performance + * + * battery + * + * On older GPUs, the vbios provided a special power state for battery + * operation. Selecting battery switched to this state. This is no + * longer provided on newer GPUs so the option does nothing in that case. + * + * balanced + * + * On older GPUs, the vbios provided a special power state for balanced + * operation. Selecting balanced switched to this state. This is no + * longer provided on newer GPUs so the option does nothing in that case. + * + * performance + * + * On older GPUs, the vbios provided a special power state for performance + * operation. Selecting performance switched to this state. This is no + * longer provided on newer GPUs so the option does nothing in that case. + * + */ + static ssize_t amdgpu_get_dpm_state(struct device *dev, struct device_attribute *attr, char *buf) @@ -131,6 +162,59 @@ fail: return count; } + +/** + * DOC: power_dpm_force_performance_level + * + * The amdgpu driver provides a sysfs API for adjusting certain power + * related parameters. The file power_dpm_force_performance_level is + * used for this. It accepts the following arguments: + * - auto + * - low + * - high + * - manual + * - GPU fan + * - profile_standard + * - profile_min_sclk + * - profile_min_mclk + * - profile_peak + * + * auto + * + * When auto is selected, the driver will attempt to dynamically select + * the optimal power profile for current conditions in the driver. + * + * low + * + * When low is selected, the clocks are forced to the lowest power state. + * + * high + * + * When high is selected, the clocks are forced to the highest power state. + * + * manual + * + * When manual is selected, the user can manually adjust which power states + * are enabled for each clock domain via the sysfs pp_dpm_mclk, pp_dpm_sclk, + * and pp_dpm_pcie files and adjust the power state transition heuristics + * via the pp_power_profile_mode sysfs file. + * + * profile_standard + * profile_min_sclk + * profile_min_mclk + * profile_peak + * + * When the profiling modes are selected, clock and power gating are + * disabled and the clocks are set for different profiling cases. This + * mode is recommended for profiling specific work loads where you do + * not want clock or power gating for clock fluctuation to interfere + * with your results. profile_standard sets the clocks to a fixed clock + * level which varies from asic to asic. profile_min_sclk forces the sclk + * to the lowest level. profile_min_mclk forces the mclk to the lowest level. + * profile_peak sets all clocks (mclk, sclk, pcie) to the highest levels. + * + */ + static ssize_t amdgpu_get_dpm_forced_performance_level(struct device *dev, struct device_attribute *attr, char *buf) @@ -324,6 +408,17 @@ fail: return count; } +/** + * DOC: pp_table + * + * The amdgpu driver provides a sysfs API for uploading new powerplay + * tables. The file pp_table is used for this. Reading the file + * will dump the current power play table. Writing to the file + * will attempt to upload a new powerplay table and re-initialize + * powerplay using that new table. + * + */ + static ssize_t amdgpu_get_pp_table(struct device *dev, struct device_attribute *attr, char *buf) @@ -360,6 +455,29 @@ static ssize_t amdgpu_set_pp_table(struct device *dev, return count; } +/** + * DOC: pp_od_clk_voltage + * + * The amdgpu driver provides a sysfs API for adjusting the clocks and voltages + * in each power level within a power state. The pp_od_clk_voltage is used for + * this. + * + * Reading the file will display: + * - a list of engine clock levels and voltages labeled OD_SCLK + * - a list of memory clock levels and voltages labeled OD_MCLK + * - a list of valid ranges for sclk, mclk, and voltage labeled OD_RANGE + * + * To manually adjust these settings, first select manual using + * power_dpm_force_performance_level. Enter a new value for each + * level by writing a string that contains "s/m level clock voltage" to + * the file. E.g., "s 1 500 820" will update sclk level 1 to be 500 MHz + * at 820 mV; "m 0 350 810" will update mclk level 0 to be 350 MHz at + * 810 mV. When you have edited all of the states as needed, write + * "c" (commit) to the file to commit your changes. If you want to reset to the + * default power levels, write "r" (reset) to the file to reset them. + * + */ + static ssize_t amdgpu_set_pp_od_clk_voltage(struct device *dev, struct device_attribute *attr, const char *buf, @@ -437,6 +555,7 @@ static ssize_t amdgpu_get_pp_od_clk_voltage(struct device *dev, if (adev->powerplay.pp_funcs->print_clock_levels) { size = amdgpu_dpm_print_clock_levels(adev, OD_SCLK, buf); size += amdgpu_dpm_print_clock_levels(adev, OD_MCLK, buf+size); + size += amdgpu_dpm_print_clock_levels(adev, OD_RANGE, buf+size); return size; } else { return snprintf(buf, PAGE_SIZE, "\n"); @@ -444,6 +563,23 @@ static ssize_t amdgpu_get_pp_od_clk_voltage(struct device *dev, } +/** + * DOC: pp_dpm_sclk pp_dpm_mclk pp_dpm_pcie + * + * The amdgpu driver provides a sysfs API for adjusting what power levels + * are enabled for a given power state. The files pp_dpm_sclk, pp_dpm_mclk, + * and pp_dpm_pcie are used for this. + * + * Reading back the files will show you the available power levels within + * the power state and the clock information for those levels. + * + * To manually adjust these states, first select manual using + * power_dpm_force_performance_level. + * Secondly,Enter a new value for each level by inputing a string that + * contains " echo xx xx xx > pp_dpm_sclk/mclk/pcie" + * E.g., echo 4 5 6 to > pp_dpm_sclk will enable sclk levels 4, 5, and 6. + */ + static ssize_t amdgpu_get_pp_dpm_sclk(struct device *dev, struct device_attribute *attr, char *buf) @@ -466,23 +602,27 @@ static ssize_t amdgpu_set_pp_dpm_sclk(struct device *dev, struct amdgpu_device *adev = ddev->dev_private; int ret; long level; - uint32_t i, mask = 0; - char sub_str[2]; + uint32_t mask = 0; + char *sub_str = NULL; + char *tmp; + char buf_cpy[count]; + const char delimiter[3] = {' ', '\n', '\0'}; - for (i = 0; i < strlen(buf); i++) { - if (*(buf + i) == '\n') - continue; - sub_str[0] = *(buf + i); - sub_str[1] = '\0'; - ret = kstrtol(sub_str, 0, &level); + memcpy(buf_cpy, buf, count+1); + tmp = buf_cpy; + while (tmp[0]) { + sub_str = strsep(&tmp, delimiter); + if (strlen(sub_str)) { + ret = kstrtol(sub_str, 0, &level); - if (ret) { - count = -EINVAL; - goto fail; - } - mask |= 1 << level; + if (ret) { + count = -EINVAL; + goto fail; + } + mask |= 1 << level; + } else + break; } - if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_SCLK, mask); @@ -512,21 +652,26 @@ static ssize_t amdgpu_set_pp_dpm_mclk(struct device *dev, struct amdgpu_device *adev = ddev->dev_private; int ret; long level; - uint32_t i, mask = 0; - char sub_str[2]; + uint32_t mask = 0; + char *sub_str = NULL; + char *tmp; + char buf_cpy[count]; + const char delimiter[3] = {' ', '\n', '\0'}; - for (i = 0; i < strlen(buf); i++) { - if (*(buf + i) == '\n') - continue; - sub_str[0] = *(buf + i); - sub_str[1] = '\0'; - ret = kstrtol(sub_str, 0, &level); + memcpy(buf_cpy, buf, count+1); + tmp = buf_cpy; + while (tmp[0]) { + sub_str = strsep(&tmp, delimiter); + if (strlen(sub_str)) { + ret = kstrtol(sub_str, 0, &level); - if (ret) { - count = -EINVAL; - goto fail; - } - mask |= 1 << level; + if (ret) { + count = -EINVAL; + goto fail; + } + mask |= 1 << level; + } else + break; } if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_MCLK, mask); @@ -557,21 +702,27 @@ static ssize_t amdgpu_set_pp_dpm_pcie(struct device *dev, struct amdgpu_device *adev = ddev->dev_private; int ret; long level; - uint32_t i, mask = 0; - char sub_str[2]; + uint32_t mask = 0; + char *sub_str = NULL; + char *tmp; + char buf_cpy[count]; + const char delimiter[3] = {' ', '\n', '\0'}; - for (i = 0; i < strlen(buf); i++) { - if (*(buf + i) == '\n') - continue; - sub_str[0] = *(buf + i); - sub_str[1] = '\0'; - ret = kstrtol(sub_str, 0, &level); + memcpy(buf_cpy, buf, count+1); + tmp = buf_cpy; - if (ret) { - count = -EINVAL; - goto fail; - } - mask |= 1 << level; + while (tmp[0]) { + sub_str = strsep(&tmp, delimiter); + if (strlen(sub_str)) { + ret = kstrtol(sub_str, 0, &level); + + if (ret) { + count = -EINVAL; + goto fail; + } + mask |= 1 << level; + } else + break; } if (adev->powerplay.pp_funcs->force_clock_level) amdgpu_dpm_force_clock_level(adev, PP_PCIE, mask); @@ -668,6 +819,26 @@ fail: return count; } +/** + * DOC: pp_power_profile_mode + * + * The amdgpu driver provides a sysfs API for adjusting the heuristics + * related to switching between power levels in a power state. The file + * pp_power_profile_mode is used for this. + * + * Reading this file outputs a list of all of the predefined power profiles + * and the relevant heuristics settings for that profile. + * + * To select a profile or create a custom profile, first select manual using + * power_dpm_force_performance_level. Writing the number of a predefined + * profile to pp_power_profile_mode will enable those heuristics. To + * create a custom set of heuristics, write a string of numbers to the file + * starting with the number of the custom profile along with a setting + * for each heuristic parameter. Due to differences across asic families + * the heuristic parameters vary from family to family. + * + */ + static ssize_t amdgpu_get_pp_power_profile_mode(struct device *dev, struct device_attribute *attr, char *buf) @@ -1020,8 +1191,8 @@ static ssize_t amdgpu_hwmon_show_power_avg(struct device *dev, { struct amdgpu_device *adev = dev_get_drvdata(dev); struct drm_device *ddev = adev->ddev; - struct pp_gpu_power query = {0}; - int r, size = sizeof(query); + u32 query = 0; + int r, size = sizeof(u32); unsigned uw; /* Can't get power when the card is off */ @@ -1041,7 +1212,7 @@ static ssize_t amdgpu_hwmon_show_power_avg(struct device *dev, return r; /* convert to microwatts */ - uw = (query.average_gpu_power >> 8) * 1000000; + uw = (query >> 8) * 1000000 + (query & 0xff) * 1000; return snprintf(buf, PAGE_SIZE, "%u\n", uw); } @@ -1109,6 +1280,46 @@ static ssize_t amdgpu_hwmon_set_power_cap(struct device *dev, return count; } + +/** + * DOC: hwmon + * + * The amdgpu driver exposes the following sensor interfaces: + * - GPU temperature (via the on-die sensor) + * - GPU voltage + * - Northbridge voltage (APUs only) + * - GPU power + * - GPU fan + * + * hwmon interfaces for GPU temperature: + * - temp1_input: the on die GPU temperature in millidegrees Celsius + * - temp1_crit: temperature critical max value in millidegrees Celsius + * - temp1_crit_hyst: temperature hysteresis for critical limit in millidegrees Celsius + * + * hwmon interfaces for GPU voltage: + * - in0_input: the voltage on the GPU in millivolts + * - in1_input: the voltage on the Northbridge in millivolts + * + * hwmon interfaces for GPU power: + * - power1_average: average power used by the GPU in microWatts + * - power1_cap_min: minimum cap supported in microWatts + * - power1_cap_max: maximum cap supported in microWatts + * - power1_cap: selected power cap in microWatts + * + * hwmon interfaces for GPU fan: + * - pwm1: pulse width modulation fan level (0-255) + * - pwm1_enable: pulse width modulation fan control method + * 0: no fan speed control + * 1: manual fan speed control using pwm interface + * 2: automatic fan speed control + * - pwm1_min: pulse width modulation fan control minimum level (0) + * - pwm1_max: pulse width modulation fan control maximum level (255) + * - fan1_input: fan speed in RPM + * + * You can use hwmon tools like sensors to view this information on your system. + * + */ + static SENSOR_DEVICE_ATTR(temp1_input, S_IRUGO, amdgpu_hwmon_show_temp, NULL, 0); static SENSOR_DEVICE_ATTR(temp1_crit, S_IRUGO, amdgpu_hwmon_show_temp_thresh, NULL, 0); static SENSOR_DEVICE_ATTR(temp1_crit_hyst, S_IRUGO, amdgpu_hwmon_show_temp_thresh, NULL, 1); @@ -1153,19 +1364,14 @@ static umode_t hwmon_attributes_visible(struct kobject *kobj, struct amdgpu_device *adev = dev_get_drvdata(dev); umode_t effective_mode = attr->mode; - /* handle non-powerplay limitations */ - if (!adev->powerplay.pp_handle) { - /* Skip fan attributes if fan is not present */ - if (adev->pm.no_fan && - (attr == &sensor_dev_attr_pwm1.dev_attr.attr || - attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr || - attr == &sensor_dev_attr_pwm1_max.dev_attr.attr || - attr == &sensor_dev_attr_pwm1_min.dev_attr.attr)) - return 0; - /* requires powerplay */ - if (attr == &sensor_dev_attr_fan1_input.dev_attr.attr) - return 0; - } + + /* Skip fan attributes if fan is not present */ + if (adev->pm.no_fan && (attr == &sensor_dev_attr_pwm1.dev_attr.attr || + attr == &sensor_dev_attr_pwm1_enable.dev_attr.attr || + attr == &sensor_dev_attr_pwm1_max.dev_attr.attr || + attr == &sensor_dev_attr_pwm1_min.dev_attr.attr || + attr == &sensor_dev_attr_fan1_input.dev_attr.attr)) + return 0; /* Skip limit attributes if DPM is not enabled */ if (!adev->pm.dpm_enabled && @@ -1658,9 +1864,6 @@ void amdgpu_pm_sysfs_fini(struct amdgpu_device *adev) void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) { - struct drm_device *ddev = adev->ddev; - struct drm_crtc *crtc; - struct amdgpu_crtc *amdgpu_crtc; int i = 0; if (!adev->pm.dpm_enabled) @@ -1676,21 +1879,25 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) } if (adev->powerplay.pp_funcs->dispatch_tasks) { + if (!amdgpu_device_has_dc_support(adev)) { + mutex_lock(&adev->pm.mutex); + amdgpu_dpm_get_active_displays(adev); + adev->pm.pm_display_cfg.num_display = adev->pm.dpm.new_active_crtcs; + adev->pm.pm_display_cfg.vrefresh = amdgpu_dpm_get_vrefresh(adev); + adev->pm.pm_display_cfg.min_vblank_time = amdgpu_dpm_get_vblank_time(adev); + /* we have issues with mclk switching with refresh rates over 120 hz on the non-DC code. */ + if (adev->pm.pm_display_cfg.vrefresh > 120) + adev->pm.pm_display_cfg.min_vblank_time = 0; + if (adev->powerplay.pp_funcs->display_configuration_change) + adev->powerplay.pp_funcs->display_configuration_change( + adev->powerplay.pp_handle, + &adev->pm.pm_display_cfg); + mutex_unlock(&adev->pm.mutex); + } amdgpu_dpm_dispatch_task(adev, AMD_PP_TASK_DISPLAY_CONFIG_CHANGE, NULL); } else { mutex_lock(&adev->pm.mutex); - adev->pm.dpm.new_active_crtcs = 0; - adev->pm.dpm.new_active_crtc_count = 0; - if (adev->mode_info.num_crtc && adev->mode_info.mode_config_initialized) { - list_for_each_entry(crtc, - &ddev->mode_config.crtc_list, head) { - amdgpu_crtc = to_amdgpu_crtc(crtc); - if (amdgpu_crtc->enabled) { - adev->pm.dpm.new_active_crtcs |= (1 << amdgpu_crtc->crtc_id); - adev->pm.dpm.new_active_crtc_count++; - } - } - } + amdgpu_dpm_get_active_displays(adev); /* update battery/ac status */ if (power_supply_is_system_supplied() > 0) adev->pm.dpm.ac_power = true; @@ -1711,7 +1918,7 @@ void amdgpu_pm_compute_clocks(struct amdgpu_device *adev) static int amdgpu_debugfs_pm_info_pp(struct seq_file *m, struct amdgpu_device *adev) { uint32_t value; - struct pp_gpu_power query = {0}; + uint32_t query = 0; int size; /* sanity check PP is enabled */ @@ -1734,17 +1941,9 @@ static int amdgpu_debugfs_pm_info_pp(struct seq_file *m, struct amdgpu_device *a seq_printf(m, "\t%u mV (VDDGFX)\n", value); if (!amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_VDDNB, (void *)&value, &size)) seq_printf(m, "\t%u mV (VDDNB)\n", value); - size = sizeof(query); - if (!amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GPU_POWER, (void *)&query, &size)) { - seq_printf(m, "\t%u.%u W (VDDC)\n", query.vddc_power >> 8, - query.vddc_power & 0xff); - seq_printf(m, "\t%u.%u W (VDDCI)\n", query.vddci_power >> 8, - query.vddci_power & 0xff); - seq_printf(m, "\t%u.%u W (max GPU)\n", query.max_gpu_power >> 8, - query.max_gpu_power & 0xff); - seq_printf(m, "\t%u.%u W (average GPU)\n", query.average_gpu_power >> 8, - query.average_gpu_power & 0xff); - } + size = sizeof(uint32_t); + if (!amdgpu_dpm_read_sensor(adev, AMDGPU_PP_SENSOR_GPU_POWER, (void *)&query, &size)) + seq_printf(m, "\t%u.%u W (average GPU)\n", query >> 8, query & 0xff); size = sizeof(value); seq_printf(m, "\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c index 4b584cb75bf45..4683626b065ff 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_prime.c @@ -102,12 +102,18 @@ amdgpu_gem_prime_import_sg_table(struct drm_device *dev, struct reservation_object *resv = attach->dmabuf->resv; struct amdgpu_device *adev = dev->dev_private; struct amdgpu_bo *bo; + struct amdgpu_bo_param bp; int ret; + memset(&bp, 0, sizeof(bp)); + bp.size = attach->dmabuf->size; + bp.byte_align = PAGE_SIZE; + bp.domain = AMDGPU_GEM_DOMAIN_CPU; + bp.flags = 0; + bp.type = ttm_bo_type_sg; + bp.resv = resv; ww_mutex_lock(&resv->lock, NULL); - ret = amdgpu_bo_create(adev, attach->dmabuf->size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_CPU, 0, ttm_bo_type_sg, - resv, &bo); + ret = amdgpu_bo_create(adev, &bp, &bo); if (ret) goto error; @@ -209,7 +215,7 @@ static int amdgpu_gem_begin_cpu_access(struct dma_buf *dma_buf, struct amdgpu_bo *bo = gem_to_amdgpu_bo(dma_buf->priv); struct amdgpu_device *adev = amdgpu_ttm_adev(bo->tbo.bdev); struct ttm_operation_ctx ctx = { true, false }; - u32 domain = amdgpu_display_framebuffer_domains(adev); + u32 domain = amdgpu_display_supported_domains(adev); int ret; bool reads = (direction == DMA_BIDIRECTIONAL || direction == DMA_FROM_DEVICE); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c index c7d43e064fc76..9f1a5bd39ae8a 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_psp.c @@ -52,6 +52,7 @@ static int psp_sw_init(void *handle) switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: psp_v3_1_set_psp_funcs(psp); break; case CHIP_RAVEN: diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c index 262c1267249eb..8af16e81c7d44 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_queue_mgr.c @@ -66,6 +66,8 @@ static int amdgpu_identity_map(struct amdgpu_device *adev, u32 ring, struct amdgpu_ring **out_ring) { + u32 instance; + switch (mapper->hw_ip) { case AMDGPU_HW_IP_GFX: *out_ring = &adev->gfx.gfx_ring[ring]; @@ -77,13 +79,16 @@ static int amdgpu_identity_map(struct amdgpu_device *adev, *out_ring = &adev->sdma.instance[ring].ring; break; case AMDGPU_HW_IP_UVD: - *out_ring = &adev->uvd.ring; + instance = ring; + *out_ring = &adev->uvd.inst[instance].ring; break; case AMDGPU_HW_IP_VCE: *out_ring = &adev->vce.ring[ring]; break; case AMDGPU_HW_IP_UVD_ENC: - *out_ring = &adev->uvd.ring_enc[ring]; + instance = ring / adev->uvd.num_enc_rings; + *out_ring = + &adev->uvd.inst[instance].ring_enc[ring%adev->uvd.num_enc_rings]; break; case AMDGPU_HW_IP_VCN_DEC: *out_ring = &adev->vcn.ring_dec; @@ -240,13 +245,14 @@ int amdgpu_queue_mgr_map(struct amdgpu_device *adev, ip_num_rings = adev->sdma.num_instances; break; case AMDGPU_HW_IP_UVD: - ip_num_rings = 1; + ip_num_rings = adev->uvd.num_uvd_inst; break; case AMDGPU_HW_IP_VCE: ip_num_rings = adev->vce.num_rings; break; case AMDGPU_HW_IP_UVD_ENC: - ip_num_rings = adev->uvd.num_enc_rings; + ip_num_rings = + adev->uvd.num_enc_rings * adev->uvd.num_uvd_inst; break; case AMDGPU_HW_IP_VCN_DEC: ip_num_rings = 1; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c index d5f526f38e508..c6850b629d0e6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.c @@ -362,6 +362,7 @@ void amdgpu_ring_fini(struct amdgpu_ring *ring) dma_fence_put(ring->vmid_wait); ring->vmid_wait = NULL; + ring->me = 0; ring->adev->rings[ring->idx] = NULL; } @@ -459,6 +460,26 @@ void amdgpu_ring_lru_touch(struct amdgpu_device *adev, struct amdgpu_ring *ring) spin_unlock(&adev->ring_lru_list_lock); } +/** + * amdgpu_ring_emit_reg_write_reg_wait_helper - ring helper + * + * @adev: amdgpu_device pointer + * @reg0: register to write + * @reg1: register to wait on + * @ref: reference value to write/wait on + * @mask: mask to wait on + * + * Helper for rings that don't support write and wait in a + * single oneshot packet. + */ +void amdgpu_ring_emit_reg_write_reg_wait_helper(struct amdgpu_ring *ring, + uint32_t reg0, uint32_t reg1, + uint32_t ref, uint32_t mask) +{ + amdgpu_ring_emit_wreg(ring, reg0, ref); + amdgpu_ring_emit_reg_wait(ring, reg1, mask, mask); +} + /* * Debugfs info */ diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h index 1a5911882657b..1513124c5659d 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ring.h @@ -29,7 +29,7 @@ #include <drm/drm_print.h> /* max number of rings */ -#define AMDGPU_MAX_RINGS 18 +#define AMDGPU_MAX_RINGS 21 #define AMDGPU_MAX_GFX_RINGS 1 #define AMDGPU_MAX_COMPUTE_RINGS 8 #define AMDGPU_MAX_VCE_RINGS 3 @@ -42,6 +42,7 @@ #define AMDGPU_FENCE_FLAG_64BIT (1 << 0) #define AMDGPU_FENCE_FLAG_INT (1 << 1) +#define AMDGPU_FENCE_FLAG_TC_WB_ONLY (1 << 2) enum amdgpu_ring_type { AMDGPU_RING_TYPE_GFX, @@ -90,7 +91,8 @@ int amdgpu_fence_driver_start_ring(struct amdgpu_ring *ring, unsigned irq_type); void amdgpu_fence_driver_suspend(struct amdgpu_device *adev); void amdgpu_fence_driver_resume(struct amdgpu_device *adev); -int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **fence); +int amdgpu_fence_emit(struct amdgpu_ring *ring, struct dma_fence **fence, + unsigned flags); int amdgpu_fence_emit_polling(struct amdgpu_ring *ring, uint32_t *s); void amdgpu_fence_process(struct amdgpu_ring *ring); int amdgpu_fence_wait_empty(struct amdgpu_ring *ring); @@ -154,6 +156,9 @@ struct amdgpu_ring_funcs { void (*emit_wreg)(struct amdgpu_ring *ring, uint32_t reg, uint32_t val); void (*emit_reg_wait)(struct amdgpu_ring *ring, uint32_t reg, uint32_t val, uint32_t mask); + void (*emit_reg_write_reg_wait)(struct amdgpu_ring *ring, + uint32_t reg0, uint32_t reg1, + uint32_t ref, uint32_t mask); void (*emit_tmz)(struct amdgpu_ring *ring, bool start); /* priority functions */ void (*set_priority) (struct amdgpu_ring *ring, @@ -228,6 +233,10 @@ int amdgpu_ring_lru_get(struct amdgpu_device *adev, int type, int *blacklist, int num_blacklist, bool lru_pipe_order, struct amdgpu_ring **ring); void amdgpu_ring_lru_touch(struct amdgpu_device *adev, struct amdgpu_ring *ring); +void amdgpu_ring_emit_reg_write_reg_wait_helper(struct amdgpu_ring *ring, + uint32_t reg0, uint32_t val0, + uint32_t reg1, uint32_t val1); + static inline void amdgpu_ring_clear_ring(struct amdgpu_ring *ring) { int i = 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c index 2dbe87591f813..d167e8ab76d30 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_test.c @@ -33,6 +33,7 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) struct amdgpu_ring *ring = adev->mman.buffer_funcs_ring; struct amdgpu_bo *vram_obj = NULL; struct amdgpu_bo **gtt_obj = NULL; + struct amdgpu_bo_param bp; uint64_t gart_addr, vram_addr; unsigned n, size; int i, r; @@ -58,9 +59,15 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) r = 1; goto out_cleanup; } - - r = amdgpu_bo_create(adev, size, PAGE_SIZE, AMDGPU_GEM_DOMAIN_VRAM, 0, - ttm_bo_type_kernel, NULL, &vram_obj); + memset(&bp, 0, sizeof(bp)); + bp.size = size; + bp.byte_align = PAGE_SIZE; + bp.domain = AMDGPU_GEM_DOMAIN_VRAM; + bp.flags = 0; + bp.type = ttm_bo_type_kernel; + bp.resv = NULL; + + r = amdgpu_bo_create(adev, &bp, &vram_obj); if (r) { DRM_ERROR("Failed to create VRAM object\n"); goto out_cleanup; @@ -79,9 +86,8 @@ static void amdgpu_do_test_moves(struct amdgpu_device *adev) void **vram_start, **vram_end; struct dma_fence *fence = NULL; - r = amdgpu_bo_create(adev, size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_GTT, 0, - ttm_bo_type_kernel, NULL, gtt_obj + i); + bp.domain = AMDGPU_GEM_DOMAIN_GTT; + r = amdgpu_bo_create(adev, &bp, gtt_obj + i); if (r) { DRM_ERROR("Failed to create GTT object %d\n", i); goto out_lclean; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h index 532263ab6e167..e96e26d3f3b08 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_trace.h @@ -275,7 +275,7 @@ TRACE_EVENT(amdgpu_vm_bo_unmap, ), TP_fast_assign( - __entry->bo = bo_va->base.bo; + __entry->bo = bo_va ? bo_va->base.bo : NULL; __entry->start = mapping->start; __entry->last = mapping->last; __entry->offset = mapping->offset; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c index 205da3ff9cd04..e93a0a237dc3e 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.c @@ -63,16 +63,44 @@ static void amdgpu_ttm_debugfs_fini(struct amdgpu_device *adev); /* * Global memory. */ + +/** + * amdgpu_ttm_mem_global_init - Initialize and acquire reference to + * memory object + * + * @ref: Object for initialization. + * + * This is called by drm_global_item_ref() when an object is being + * initialized. + */ static int amdgpu_ttm_mem_global_init(struct drm_global_reference *ref) { return ttm_mem_global_init(ref->object); } +/** + * amdgpu_ttm_mem_global_release - Drop reference to a memory object + * + * @ref: Object being removed + * + * This is called by drm_global_item_unref() when an object is being + * released. + */ static void amdgpu_ttm_mem_global_release(struct drm_global_reference *ref) { ttm_mem_global_release(ref->object); } +/** + * amdgpu_ttm_global_init - Initialize global TTM memory reference + * structures. + * + * @adev: AMDGPU device for which the global structures need to be + * registered. + * + * This is called as part of the AMDGPU ttm init from amdgpu_ttm_init() + * during bring up. + */ static int amdgpu_ttm_global_init(struct amdgpu_device *adev) { struct drm_global_reference *global_ref; @@ -80,7 +108,9 @@ static int amdgpu_ttm_global_init(struct amdgpu_device *adev) struct drm_sched_rq *rq; int r; + /* ensure reference is false in case init fails */ adev->mman.mem_global_referenced = false; + global_ref = &adev->mman.mem_global_ref; global_ref->global_type = DRM_GLOBAL_TTM_MEM; global_ref->size = sizeof(struct ttm_mem_global); @@ -111,7 +141,7 @@ static int amdgpu_ttm_global_init(struct amdgpu_device *adev) ring = adev->mman.buffer_funcs_ring; rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL]; r = drm_sched_entity_init(&ring->sched, &adev->mman.entity, - rq, amdgpu_sched_jobs, NULL); + rq, NULL); if (r) { DRM_ERROR("Failed setting up TTM BO move run queue.\n"); goto error_entity; @@ -146,6 +176,18 @@ static int amdgpu_invalidate_caches(struct ttm_bo_device *bdev, uint32_t flags) return 0; } +/** + * amdgpu_init_mem_type - Initialize a memory manager for a specific + * type of memory request. + * + * @bdev: The TTM BO device object (contains a reference to + * amdgpu_device) + * @type: The type of memory requested + * @man: + * + * This is called by ttm_bo_init_mm() when a buffer object is being + * initialized. + */ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type, struct ttm_mem_type_manager *man) { @@ -161,6 +203,7 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type, man->default_caching = TTM_PL_FLAG_CACHED; break; case TTM_PL_TT: + /* GTT memory */ man->func = &amdgpu_gtt_mgr_func; man->gpu_offset = adev->gmc.gart_start; man->available_caching = TTM_PL_MASK_CACHING; @@ -193,6 +236,14 @@ static int amdgpu_init_mem_type(struct ttm_bo_device *bdev, uint32_t type, return 0; } +/** + * amdgpu_evict_flags - Compute placement flags + * + * @bo: The buffer object to evict + * @placement: Possible destination(s) for evicted BO + * + * Fill in placement data when ttm_bo_evict() is called + */ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, struct ttm_placement *placement) { @@ -204,12 +255,14 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, .flags = TTM_PL_MASK_CACHING | TTM_PL_FLAG_SYSTEM }; + /* Don't handle scatter gather BOs */ if (bo->type == ttm_bo_type_sg) { placement->num_placement = 0; placement->num_busy_placement = 0; return; } + /* Object isn't an AMDGPU object so ignore */ if (!amdgpu_ttm_bo_is_amdgpu_bo(bo)) { placement->placement = &placements; placement->busy_placement = &placements; @@ -217,26 +270,16 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, placement->num_busy_placement = 1; return; } + abo = ttm_to_amdgpu_bo(bo); switch (bo->mem.mem_type) { case TTM_PL_VRAM: if (!adev->mman.buffer_funcs_enabled) { + /* Move to system memory */ amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_CPU); } else if (adev->gmc.visible_vram_size < adev->gmc.real_vram_size && - !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED)) { - unsigned fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT; - struct drm_mm_node *node = bo->mem.mm_node; - unsigned long pages_left; - - for (pages_left = bo->mem.num_pages; - pages_left; - pages_left -= node->size, node++) { - if (node->start < fpfn) - break; - } - - if (!pages_left) - goto gtt; + !(abo->flags & AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED) && + amdgpu_bo_in_cpu_visible_vram(abo)) { /* Try evicting to the CPU inaccessible part of VRAM * first, but only set GTT as busy placement, so this @@ -245,12 +288,12 @@ static void amdgpu_evict_flags(struct ttm_buffer_object *bo, */ amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_VRAM | AMDGPU_GEM_DOMAIN_GTT); - abo->placements[0].fpfn = fpfn; + abo->placements[0].fpfn = adev->gmc.visible_vram_size >> PAGE_SHIFT; abo->placements[0].lpfn = 0; abo->placement.busy_placement = &abo->placements[1]; abo->placement.num_busy_placement = 1; } else { -gtt: + /* Move to GTT memory */ amdgpu_ttm_placement_from_domain(abo, AMDGPU_GEM_DOMAIN_GTT); } break; @@ -261,6 +304,15 @@ gtt: *placement = abo->placement; } +/** + * amdgpu_verify_access - Verify access for a mmap call + * + * @bo: The buffer object to map + * @filp: The file pointer from the process performing the mmap + * + * This is called by ttm_bo_mmap() to verify whether a process + * has the right to mmap a BO to their process space. + */ static int amdgpu_verify_access(struct ttm_buffer_object *bo, struct file *filp) { struct amdgpu_bo *abo = ttm_to_amdgpu_bo(bo); @@ -278,6 +330,15 @@ static int amdgpu_verify_access(struct ttm_buffer_object *bo, struct file *filp) filp->private_data); } +/** + * amdgpu_move_null - Register memory for a buffer object + * + * @bo: The bo to assign the memory to + * @new_mem: The memory to be assigned. + * + * Assign the memory from new_mem to the memory of the buffer object + * bo. + */ static void amdgpu_move_null(struct ttm_buffer_object *bo, struct ttm_mem_reg *new_mem) { @@ -288,6 +349,10 @@ static void amdgpu_move_null(struct ttm_buffer_object *bo, new_mem->mm_node = NULL; } +/** + * amdgpu_mm_node_addr - Compute the GPU relative offset of a GTT + * buffer. + */ static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo, struct drm_mm_node *mm_node, struct ttm_mem_reg *mem) @@ -302,9 +367,10 @@ static uint64_t amdgpu_mm_node_addr(struct ttm_buffer_object *bo, } /** - * amdgpu_find_mm_node - Helper function finds the drm_mm_node - * corresponding to @offset. It also modifies the offset to be - * within the drm_mm_node returned + * amdgpu_find_mm_node - Helper function finds the drm_mm_node + * corresponding to @offset. It also modifies + * the offset to be within the drm_mm_node + * returned */ static struct drm_mm_node *amdgpu_find_mm_node(struct ttm_mem_reg *mem, unsigned long *offset) @@ -443,7 +509,12 @@ error: return r; } - +/** + * amdgpu_move_blit - Copy an entire buffer to another buffer + * + * This is a helper called by amdgpu_bo_move() and + * amdgpu_move_vram_ram() to help move buffers to and from VRAM. + */ static int amdgpu_move_blit(struct ttm_buffer_object *bo, bool evict, bool no_wait_gpu, struct ttm_mem_reg *new_mem, @@ -478,6 +549,11 @@ error: return r; } +/** + * amdgpu_move_vram_ram - Copy VRAM buffer to RAM buffer + * + * Called by amdgpu_bo_move(). + */ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, bool evict, struct ttm_operation_ctx *ctx, struct ttm_mem_reg *new_mem) @@ -490,6 +566,8 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, bool evict, int r; adev = amdgpu_ttm_adev(bo->bdev); + + /* create space/pages for new_mem in GTT space */ tmp_mem = *new_mem; tmp_mem.mm_node = NULL; placement.num_placement = 1; @@ -504,25 +582,36 @@ static int amdgpu_move_vram_ram(struct ttm_buffer_object *bo, bool evict, return r; } + /* set caching flags */ r = ttm_tt_set_placement_caching(bo->ttm, tmp_mem.placement); if (unlikely(r)) { goto out_cleanup; } + /* Bind the memory to the GTT space */ r = ttm_tt_bind(bo->ttm, &tmp_mem, ctx); if (unlikely(r)) { goto out_cleanup; } + + /* blit VRAM to GTT */ r = amdgpu_move_blit(bo, true, ctx->no_wait_gpu, &tmp_mem, old_mem); if (unlikely(r)) { goto out_cleanup; } + + /* move BO (in tmp_mem) to new_mem */ r = ttm_bo_move_ttm(bo, ctx, new_mem); out_cleanup: ttm_bo_mem_put(bo, &tmp_mem); return r; } +/** + * amdgpu_move_ram_vram - Copy buffer from RAM to VRAM + * + * Called by amdgpu_bo_move(). + */ static int amdgpu_move_ram_vram(struct ttm_buffer_object *bo, bool evict, struct ttm_operation_ctx *ctx, struct ttm_mem_reg *new_mem) @@ -535,6 +624,8 @@ static int amdgpu_move_ram_vram(struct ttm_buffer_object *bo, bool evict, int r; adev = amdgpu_ttm_adev(bo->bdev); + + /* make space in GTT for old_mem buffer */ tmp_mem = *new_mem; tmp_mem.mm_node = NULL; placement.num_placement = 1; @@ -548,10 +639,14 @@ static int amdgpu_move_ram_vram(struct ttm_buffer_object *bo, bool evict, if (unlikely(r)) { return r; } + + /* move/bind old memory to GTT space */ r = ttm_bo_move_ttm(bo, ctx, &tmp_mem); if (unlikely(r)) { goto out_cleanup; } + + /* copy to VRAM */ r = amdgpu_move_blit(bo, true, ctx->no_wait_gpu, new_mem, old_mem); if (unlikely(r)) { goto out_cleanup; @@ -561,6 +656,11 @@ out_cleanup: return r; } +/** + * amdgpu_bo_move - Move a buffer object to a new memory location + * + * Called by ttm_bo_handle_move_mem() + */ static int amdgpu_bo_move(struct ttm_buffer_object *bo, bool evict, struct ttm_operation_ctx *ctx, struct ttm_mem_reg *new_mem) @@ -626,6 +726,11 @@ memcpy: return 0; } +/** + * amdgpu_ttm_io_mem_reserve - Reserve a block of memory during a fault + * + * Called by ttm_mem_io_reserve() ultimately via ttm_bo_vm_fault() + */ static int amdgpu_ttm_io_mem_reserve(struct ttm_bo_device *bdev, struct ttm_mem_reg *mem) { struct ttm_mem_type_manager *man = &bdev->man[mem->mem_type]; @@ -695,7 +800,7 @@ struct amdgpu_ttm_tt { struct ttm_dma_tt ttm; u64 offset; uint64_t userptr; - struct mm_struct *usermm; + struct task_struct *usertask; uint32_t userflags; spinlock_t guptasklock; struct list_head guptasks; @@ -703,17 +808,29 @@ struct amdgpu_ttm_tt { uint32_t last_set_pages; }; +/** + * amdgpu_ttm_tt_get_user_pages - Pin pages of memory pointed to + * by a USERPTR pointer to memory + * + * Called by amdgpu_gem_userptr_ioctl() and amdgpu_cs_parser_bos(). + * This provides a wrapper around the get_user_pages() call to provide + * device accessible pages that back user memory. + */ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) { struct amdgpu_ttm_tt *gtt = (void *)ttm; + struct mm_struct *mm = gtt->usertask->mm; unsigned int flags = 0; unsigned pinned = 0; int r; + if (!mm) /* Happens during process shutdown */ + return -ESRCH; + if (!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY)) flags |= FOLL_WRITE; - down_read(¤t->mm->mmap_sem); + down_read(&mm->mmap_sem); if (gtt->userflags & AMDGPU_GEM_USERPTR_ANONONLY) { /* check that we only use anonymous memory @@ -721,13 +838,14 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) unsigned long end = gtt->userptr + ttm->num_pages * PAGE_SIZE; struct vm_area_struct *vma; - vma = find_vma(gtt->usermm, gtt->userptr); + vma = find_vma(mm, gtt->userptr); if (!vma || vma->vm_file || vma->vm_end < end) { - up_read(¤t->mm->mmap_sem); + up_read(&mm->mmap_sem); return -EPERM; } } + /* loop enough times using contiguous pages of memory */ do { unsigned num_pages = ttm->num_pages - pinned; uint64_t userptr = gtt->userptr + pinned * PAGE_SIZE; @@ -739,7 +857,12 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) list_add(&guptask.list, >t->guptasks); spin_unlock(>t->guptasklock); - r = get_user_pages(userptr, num_pages, flags, p, NULL); + if (mm == current->mm) + r = get_user_pages(userptr, num_pages, flags, p, NULL); + else + r = get_user_pages_remote(gtt->usertask, + mm, userptr, num_pages, + flags, p, NULL, NULL); spin_lock(>t->guptasklock); list_del(&guptask.list); @@ -752,15 +875,23 @@ int amdgpu_ttm_tt_get_user_pages(struct ttm_tt *ttm, struct page **pages) } while (pinned < ttm->num_pages); - up_read(¤t->mm->mmap_sem); + up_read(&mm->mmap_sem); return 0; release_pages: release_pages(pages, pinned); - up_read(¤t->mm->mmap_sem); + up_read(&mm->mmap_sem); return r; } +/** + * amdgpu_ttm_tt_set_user_pages - Copy pages in, putting old pages + * as necessary. + * + * Called by amdgpu_cs_list_validate(). This creates the page list + * that backs user memory and will ultimately be mapped into the device + * address space. + */ void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages) { struct amdgpu_ttm_tt *gtt = (void *)ttm; @@ -775,6 +906,11 @@ void amdgpu_ttm_tt_set_user_pages(struct ttm_tt *ttm, struct page **pages) } } +/** + * amdgpu_ttm_tt_mark_user_page - Mark pages as dirty + * + * Called while unpinning userptr pages + */ void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm) { struct amdgpu_ttm_tt *gtt = (void *)ttm; @@ -793,7 +929,12 @@ void amdgpu_ttm_tt_mark_user_pages(struct ttm_tt *ttm) } } -/* prepare the sg table with the user pages */ +/** + * amdgpu_ttm_tt_pin_userptr - prepare the sg table with the + * user pages + * + * Called by amdgpu_ttm_backend_bind() + **/ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); @@ -805,17 +946,20 @@ static int amdgpu_ttm_tt_pin_userptr(struct ttm_tt *ttm) enum dma_data_direction direction = write ? DMA_BIDIRECTIONAL : DMA_TO_DEVICE; + /* Allocate an SG array and squash pages into it */ r = sg_alloc_table_from_pages(ttm->sg, ttm->pages, ttm->num_pages, 0, ttm->num_pages << PAGE_SHIFT, GFP_KERNEL); if (r) goto release_sg; + /* Map SG to device */ r = -ENOMEM; nents = dma_map_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction); if (nents != ttm->sg->nents) goto release_sg; + /* convert SG to linear array of pages and dma addresses */ drm_prime_sg_to_page_addr_arrays(ttm->sg, ttm->pages, gtt->ttm.dma_address, ttm->num_pages); @@ -826,6 +970,9 @@ release_sg: return r; } +/** + * amdgpu_ttm_tt_unpin_userptr - Unpin and unmap userptr pages + */ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); @@ -839,14 +986,60 @@ static void amdgpu_ttm_tt_unpin_userptr(struct ttm_tt *ttm) if (!ttm->sg->sgl) return; - /* free the sg table and pages again */ + /* unmap the pages mapped to the device */ dma_unmap_sg(adev->dev, ttm->sg->sgl, ttm->sg->nents, direction); + /* mark the pages as dirty */ amdgpu_ttm_tt_mark_user_pages(ttm); sg_free_table(ttm->sg); } +int amdgpu_ttm_gart_bind(struct amdgpu_device *adev, + struct ttm_buffer_object *tbo, + uint64_t flags) +{ + struct amdgpu_bo *abo = ttm_to_amdgpu_bo(tbo); + struct ttm_tt *ttm = tbo->ttm; + struct amdgpu_ttm_tt *gtt = (void *)ttm; + int r; + + if (abo->flags & AMDGPU_GEM_CREATE_MQD_GFX9) { + uint64_t page_idx = 1; + + r = amdgpu_gart_bind(adev, gtt->offset, page_idx, + ttm->pages, gtt->ttm.dma_address, flags); + if (r) + goto gart_bind_fail; + + /* Patch mtype of the second part BO */ + flags &= ~AMDGPU_PTE_MTYPE_MASK; + flags |= AMDGPU_PTE_MTYPE(AMDGPU_MTYPE_NC); + + r = amdgpu_gart_bind(adev, + gtt->offset + (page_idx << PAGE_SHIFT), + ttm->num_pages - page_idx, + &ttm->pages[page_idx], + &(gtt->ttm.dma_address[page_idx]), flags); + } else { + r = amdgpu_gart_bind(adev, gtt->offset, ttm->num_pages, + ttm->pages, gtt->ttm.dma_address, flags); + } + +gart_bind_fail: + if (r) + DRM_ERROR("failed to bind %lu pages at 0x%08llX\n", + ttm->num_pages, gtt->offset); + + return r; +} + +/** + * amdgpu_ttm_backend_bind - Bind GTT memory + * + * Called by ttm_tt_bind() on behalf of ttm_bo_handle_move_mem(). + * This handles binding GTT memory to the device address space. + */ static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm, struct ttm_mem_reg *bo_mem) { @@ -877,7 +1070,10 @@ static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm, return 0; } + /* compute PTE flags relevant to this BO memory */ flags = amdgpu_ttm_tt_pte_flags(adev, ttm, bo_mem); + + /* bind pages into GART page tables */ gtt->offset = (u64)bo_mem->start << PAGE_SHIFT; r = amdgpu_gart_bind(adev, gtt->offset, ttm->num_pages, ttm->pages, gtt->ttm.dma_address, flags); @@ -888,6 +1084,9 @@ static int amdgpu_ttm_backend_bind(struct ttm_tt *ttm, return r; } +/** + * amdgpu_ttm_alloc_gart - Allocate GART memory for buffer object + */ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo) { struct amdgpu_device *adev = amdgpu_ttm_adev(bo->bdev); @@ -903,6 +1102,7 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo) amdgpu_gtt_mgr_has_gart_addr(&bo->mem)) return 0; + /* allocate GTT space */ tmp = bo->mem; tmp.mm_node = NULL; placement.num_placement = 1; @@ -918,10 +1118,12 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo) if (unlikely(r)) return r; + /* compute PTE flags for this buffer object */ flags = amdgpu_ttm_tt_pte_flags(adev, bo->ttm, &tmp); + + /* Bind pages */ gtt->offset = (u64)tmp.start << PAGE_SHIFT; - r = amdgpu_gart_bind(adev, gtt->offset, bo->ttm->num_pages, - bo->ttm->pages, gtt->ttm.dma_address, flags); + r = amdgpu_ttm_gart_bind(adev, bo, flags); if (unlikely(r)) { ttm_bo_mem_put(bo, &tmp); return r; @@ -935,31 +1137,40 @@ int amdgpu_ttm_alloc_gart(struct ttm_buffer_object *bo) return 0; } +/** + * amdgpu_ttm_recover_gart - Rebind GTT pages + * + * Called by amdgpu_gtt_mgr_recover() from amdgpu_device_reset() to + * rebind GTT pages during a GPU reset. + */ int amdgpu_ttm_recover_gart(struct ttm_buffer_object *tbo) { struct amdgpu_device *adev = amdgpu_ttm_adev(tbo->bdev); - struct amdgpu_ttm_tt *gtt = (void *)tbo->ttm; uint64_t flags; int r; - if (!gtt) + if (!tbo->ttm) return 0; - flags = amdgpu_ttm_tt_pte_flags(adev, >t->ttm.ttm, &tbo->mem); - r = amdgpu_gart_bind(adev, gtt->offset, gtt->ttm.ttm.num_pages, - gtt->ttm.ttm.pages, gtt->ttm.dma_address, flags); - if (r) - DRM_ERROR("failed to bind %lu pages at 0x%08llX\n", - gtt->ttm.ttm.num_pages, gtt->offset); + flags = amdgpu_ttm_tt_pte_flags(adev, tbo->ttm, &tbo->mem); + r = amdgpu_ttm_gart_bind(adev, tbo, flags); + return r; } +/** + * amdgpu_ttm_backend_unbind - Unbind GTT mapped pages + * + * Called by ttm_tt_unbind() on behalf of ttm_bo_move_ttm() and + * ttm_tt_destroy(). + */ static int amdgpu_ttm_backend_unbind(struct ttm_tt *ttm) { struct amdgpu_device *adev = amdgpu_ttm_adev(ttm->bdev); struct amdgpu_ttm_tt *gtt = (void *)ttm; int r; + /* if the pages have userptr pinning then clear that first */ if (gtt->userptr) amdgpu_ttm_tt_unpin_userptr(ttm); @@ -978,6 +1189,9 @@ static void amdgpu_ttm_backend_destroy(struct ttm_tt *ttm) { struct amdgpu_ttm_tt *gtt = (void *)ttm; + if (gtt->usertask) + put_task_struct(gtt->usertask); + ttm_dma_tt_fini(>t->ttm); kfree(gtt); } @@ -988,6 +1202,13 @@ static struct ttm_backend_func amdgpu_backend_func = { .destroy = &amdgpu_ttm_backend_destroy, }; +/** + * amdgpu_ttm_tt_create - Create a ttm_tt object for a given BO + * + * @bo: The buffer object to create a GTT ttm_tt object around + * + * Called by ttm_tt_create(). + */ static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_buffer_object *bo, uint32_t page_flags) { @@ -1001,6 +1222,8 @@ static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_buffer_object *bo, return NULL; } gtt->ttm.ttm.func = &amdgpu_backend_func; + + /* allocate space for the uninitialized page entries */ if (ttm_sg_tt_init(>t->ttm, bo, page_flags)) { kfree(gtt); return NULL; @@ -1008,6 +1231,12 @@ static struct ttm_tt *amdgpu_ttm_tt_create(struct ttm_buffer_object *bo, return >t->ttm.ttm; } +/** + * amdgpu_ttm_tt_populate - Map GTT pages visible to the device + * + * Map the pages of a ttm_tt object to an address space visible + * to the underlying device. + */ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm, struct ttm_operation_ctx *ctx) { @@ -1015,6 +1244,7 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm, struct amdgpu_ttm_tt *gtt = (void *)ttm; bool slave = !!(ttm->page_flags & TTM_PAGE_FLAG_SG); + /* user pages are bound by amdgpu_ttm_tt_pin_userptr() */ if (gtt && gtt->userptr) { ttm->sg = kzalloc(sizeof(struct sg_table), GFP_KERNEL); if (!ttm->sg) @@ -1039,9 +1269,17 @@ static int amdgpu_ttm_tt_populate(struct ttm_tt *ttm, } #endif + /* fall back to generic helper to populate the page array + * and map them to the device */ return ttm_populate_and_map_pages(adev->dev, >t->ttm, ctx); } +/** + * amdgpu_ttm_tt_unpopulate - unmap GTT pages and unpopulate page arrays + * + * Unmaps pages of a ttm_tt object from the device address space and + * unpopulates the page array backing it. + */ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) { struct amdgpu_device *adev; @@ -1067,9 +1305,21 @@ static void amdgpu_ttm_tt_unpopulate(struct ttm_tt *ttm) } #endif + /* fall back to generic helper to unmap and unpopulate array */ ttm_unmap_and_unpopulate_pages(adev->dev, >t->ttm); } +/** + * amdgpu_ttm_tt_set_userptr - Initialize userptr GTT ttm_tt + * for the current task + * + * @ttm: The ttm_tt object to bind this userptr object to + * @addr: The address in the current tasks VM space to use + * @flags: Requirements of userptr object. + * + * Called by amdgpu_gem_userptr_ioctl() to bind userptr pages + * to current task + */ int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, uint32_t flags) { @@ -1079,8 +1329,13 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, return -EINVAL; gtt->userptr = addr; - gtt->usermm = current->mm; gtt->userflags = flags; + + if (gtt->usertask) + put_task_struct(gtt->usertask); + gtt->usertask = current->group_leader; + get_task_struct(gtt->usertask); + spin_lock_init(>t->guptasklock); INIT_LIST_HEAD(>t->guptasks); atomic_set(>t->mmu_invalidations, 0); @@ -1089,6 +1344,9 @@ int amdgpu_ttm_tt_set_userptr(struct ttm_tt *ttm, uint64_t addr, return 0; } +/** + * amdgpu_ttm_tt_get_usermm - Return memory manager for ttm_tt object + */ struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm) { struct amdgpu_ttm_tt *gtt = (void *)ttm; @@ -1096,9 +1354,18 @@ struct mm_struct *amdgpu_ttm_tt_get_usermm(struct ttm_tt *ttm) if (gtt == NULL) return NULL; - return gtt->usermm; + if (gtt->usertask == NULL) + return NULL; + + return gtt->usertask->mm; } +/** + * amdgpu_ttm_tt_affect_userptr - Determine if a ttm_tt object lays + * inside an address range for the + * current task. + * + */ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, unsigned long end) { @@ -1109,10 +1376,16 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, if (gtt == NULL || !gtt->userptr) return false; + /* Return false if no part of the ttm_tt object lies within + * the range + */ size = (unsigned long)gtt->ttm.ttm.num_pages * PAGE_SIZE; if (gtt->userptr > end || gtt->userptr + size <= start) return false; + /* Search the lists of tasks that hold this mapping and see + * if current is one of them. If it is return false. + */ spin_lock(>t->guptasklock); list_for_each_entry(entry, >t->guptasks, list) { if (entry->task == current) { @@ -1127,6 +1400,10 @@ bool amdgpu_ttm_tt_affect_userptr(struct ttm_tt *ttm, unsigned long start, return true; } +/** + * amdgpu_ttm_tt_userptr_invalidated - Has the ttm_tt object been + * invalidated? + */ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, int *last_invalidated) { @@ -1137,6 +1414,12 @@ bool amdgpu_ttm_tt_userptr_invalidated(struct ttm_tt *ttm, return prev_invalidated != *last_invalidated; } +/** + * amdgpu_ttm_tt_userptr_needs_pages - Have the pages backing this + * ttm_tt object been invalidated + * since the last time they've + * been set? + */ bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm) { struct amdgpu_ttm_tt *gtt = (void *)ttm; @@ -1147,6 +1430,9 @@ bool amdgpu_ttm_tt_userptr_needs_pages(struct ttm_tt *ttm) return atomic_read(>t->mmu_invalidations) != gtt->last_set_pages; } +/** + * amdgpu_ttm_tt_is_readonly - Is the ttm_tt object read only? + */ bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm) { struct amdgpu_ttm_tt *gtt = (void *)ttm; @@ -1157,6 +1443,12 @@ bool amdgpu_ttm_tt_is_readonly(struct ttm_tt *ttm) return !!(gtt->userflags & AMDGPU_GEM_USERPTR_READONLY); } +/** + * amdgpu_ttm_tt_pte_flags - Compute PTE flags for ttm_tt object + * + * @ttm: The ttm_tt object to compute the flags for + * @mem: The memory registry backing this ttm_tt object + */ uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, struct ttm_mem_reg *mem) { @@ -1181,6 +1473,16 @@ uint64_t amdgpu_ttm_tt_pte_flags(struct amdgpu_device *adev, struct ttm_tt *ttm, return flags; } +/** + * amdgpu_ttm_bo_eviction_valuable - Check to see if we can evict + * a buffer object. + * + * Return true if eviction is sensible. Called by + * ttm_mem_evict_first() on behalf of ttm_bo_mem_force_space() + * which tries to evict buffer objects until it can find space + * for a new object and by ttm_bo_force_list_clean() which is + * used to clean out a memory space. + */ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, const struct ttm_place *place) { @@ -1227,6 +1529,19 @@ static bool amdgpu_ttm_bo_eviction_valuable(struct ttm_buffer_object *bo, return ttm_bo_eviction_valuable(bo, place); } +/** + * amdgpu_ttm_access_memory - Read or Write memory that backs a + * buffer object. + * + * @bo: The buffer object to read/write + * @offset: Offset into buffer object + * @buf: Secondary buffer to write/read from + * @len: Length in bytes of access + * @write: true if writing + * + * This is used to access VRAM that backs a buffer object via MMIO + * access for debugging purposes. + */ static int amdgpu_ttm_access_memory(struct ttm_buffer_object *bo, unsigned long offset, void *buf, int len, int write) @@ -1329,6 +1644,7 @@ static void amdgpu_ttm_fw_reserve_vram_fini(struct amdgpu_device *adev) static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev) { struct ttm_operation_ctx ctx = { false, false }; + struct amdgpu_bo_param bp; int r = 0; int i; u64 vram_size = adev->gmc.visible_vram_size; @@ -1336,17 +1652,21 @@ static int amdgpu_ttm_fw_reserve_vram_init(struct amdgpu_device *adev) u64 size = adev->fw_vram_usage.size; struct amdgpu_bo *bo; + memset(&bp, 0, sizeof(bp)); + bp.size = adev->fw_vram_usage.size; + bp.byte_align = PAGE_SIZE; + bp.domain = AMDGPU_GEM_DOMAIN_VRAM; + bp.flags = AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | + AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS; + bp.type = ttm_bo_type_kernel; + bp.resv = NULL; adev->fw_vram_usage.va = NULL; adev->fw_vram_usage.reserved_bo = NULL; if (adev->fw_vram_usage.size > 0 && adev->fw_vram_usage.size <= vram_size) { - r = amdgpu_bo_create(adev, adev->fw_vram_usage.size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - AMDGPU_GEM_CREATE_CPU_ACCESS_REQUIRED | - AMDGPU_GEM_CREATE_VRAM_CONTIGUOUS, - ttm_bo_type_kernel, NULL, + r = amdgpu_bo_create(adev, &bp, &adev->fw_vram_usage.reserved_bo); if (r) goto error_create; @@ -1398,13 +1718,22 @@ error_create: adev->fw_vram_usage.reserved_bo = NULL; return r; } - +/** + * amdgpu_ttm_init - Init the memory management (ttm) as well as + * various gtt/vram related fields. + * + * This initializes all of the memory space pools that the TTM layer + * will need such as the GTT space (system memory mapped to the device), + * VRAM (on-board memory), and on-chip memories (GDS, GWS, OA) which + * can be mapped per VMID. + */ int amdgpu_ttm_init(struct amdgpu_device *adev) { uint64_t gtt_size; int r; u64 vis_vram_limit; + /* initialize global references for vram/gtt */ r = amdgpu_ttm_global_init(adev); if (r) { return r; @@ -1425,6 +1754,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) /* We opt to avoid OOM on system pages allocations */ adev->mman.bdev.no_retry = true; + /* Initialize VRAM pool with all of VRAM divided into pages */ r = ttm_bo_init_mm(&adev->mman.bdev, TTM_PL_VRAM, adev->gmc.real_vram_size >> PAGE_SHIFT); if (r) { @@ -1454,15 +1784,23 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) return r; } - r = amdgpu_bo_create_kernel(adev, adev->gmc.stolen_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, - &adev->stolen_vga_memory, - NULL, NULL); - if (r) - return r; + /* allocate memory as required for VGA + * This is used for VGA emulation and pre-OS scanout buffers to + * avoid display artifacts while transitioning between pre-OS + * and driver. */ + if (adev->gmc.stolen_size) { + r = amdgpu_bo_create_kernel(adev, adev->gmc.stolen_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, + &adev->stolen_vga_memory, + NULL, NULL); + if (r) + return r; + } DRM_INFO("amdgpu: %uM of VRAM memory ready\n", (unsigned) (adev->gmc.real_vram_size / (1024 * 1024))); + /* Compute GTT size, either bsaed on 3/4th the size of RAM size + * or whatever the user passed on module init */ if (amdgpu_gtt_size == -1) { struct sysinfo si; @@ -1473,6 +1811,8 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) } else gtt_size = (uint64_t)amdgpu_gtt_size << 20; + + /* Initialize GTT memory pool */ r = ttm_bo_init_mm(&adev->mman.bdev, TTM_PL_TT, gtt_size >> PAGE_SHIFT); if (r) { DRM_ERROR("Failed initializing GTT heap.\n"); @@ -1481,6 +1821,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) DRM_INFO("amdgpu: %uM of GTT memory ready.\n", (unsigned)(gtt_size / (1024 * 1024))); + /* Initialize various on-chip memory pools */ adev->gds.mem.total_size = adev->gds.mem.total_size << AMDGPU_GDS_SHIFT; adev->gds.mem.gfx_partition_size = adev->gds.mem.gfx_partition_size << AMDGPU_GDS_SHIFT; adev->gds.mem.cs_partition_size = adev->gds.mem.cs_partition_size << AMDGPU_GDS_SHIFT; @@ -1520,6 +1861,7 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) } } + /* Register debugfs entries for amdgpu_ttm */ r = amdgpu_ttm_debugfs_init(adev); if (r) { DRM_ERROR("Failed to init debugfs\n"); @@ -1528,13 +1870,25 @@ int amdgpu_ttm_init(struct amdgpu_device *adev) return 0; } +/** + * amdgpu_ttm_late_init - Handle any late initialization for + * amdgpu_ttm + */ +void amdgpu_ttm_late_init(struct amdgpu_device *adev) +{ + /* return the VGA stolen memory (if any) back to VRAM */ + amdgpu_bo_free_kernel(&adev->stolen_vga_memory, NULL, NULL); +} + +/** + * amdgpu_ttm_fini - De-initialize the TTM memory pools + */ void amdgpu_ttm_fini(struct amdgpu_device *adev) { if (!adev->mman.initialized) return; amdgpu_ttm_debugfs_fini(adev); - amdgpu_bo_free_kernel(&adev->stolen_vga_memory, NULL, NULL); amdgpu_ttm_fw_reserve_vram_fini(adev); if (adev->mman.aper_base_kaddr) iounmap(adev->mman.aper_base_kaddr); @@ -1856,6 +2210,11 @@ static const struct drm_info_list amdgpu_ttm_debugfs_list[] = { #endif }; +/** + * amdgpu_ttm_vram_read - Linear read access to VRAM + * + * Accesses VRAM via MMIO for debugging purposes. + */ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -1895,6 +2254,11 @@ static ssize_t amdgpu_ttm_vram_read(struct file *f, char __user *buf, return result; } +/** + * amdgpu_ttm_vram_write - Linear write access to VRAM + * + * Accesses VRAM via MMIO for debugging purposes. + */ static ssize_t amdgpu_ttm_vram_write(struct file *f, const char __user *buf, size_t size, loff_t *pos) { @@ -1943,6 +2307,9 @@ static const struct file_operations amdgpu_ttm_vram_fops = { #ifdef CONFIG_DRM_AMDGPU_GART_DEBUGFS +/** + * amdgpu_ttm_gtt_read - Linear read access to GTT memory + */ static ssize_t amdgpu_ttm_gtt_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -1990,6 +2357,13 @@ static const struct file_operations amdgpu_ttm_gtt_fops = { #endif +/** + * amdgpu_iomem_read - Virtual read access to GPU mapped memory + * + * This function is used to read memory that has been mapped to the + * GPU and the known addresses are not physical addresses but instead + * bus addresses (e.g., what you'd put in an IB or ring buffer). + */ static ssize_t amdgpu_iomem_read(struct file *f, char __user *buf, size_t size, loff_t *pos) { @@ -1998,6 +2372,7 @@ static ssize_t amdgpu_iomem_read(struct file *f, char __user *buf, ssize_t result = 0; int r; + /* retrieve the IOMMU domain if any for this device */ dom = iommu_get_domain_for_dev(adev->dev); while (size) { @@ -2010,6 +2385,10 @@ static ssize_t amdgpu_iomem_read(struct file *f, char __user *buf, bytes = bytes < size ? bytes : size; + /* Translate the bus address to a physical address. If + * the domain is NULL it means there is no IOMMU active + * and the address translation is the identity + */ addr = dom ? iommu_iova_to_phys(dom, addr) : addr; pfn = addr >> PAGE_SHIFT; @@ -2034,6 +2413,13 @@ static ssize_t amdgpu_iomem_read(struct file *f, char __user *buf, return result; } +/** + * amdgpu_iomem_write - Virtual write access to GPU mapped memory + * + * This function is used to write memory that has been mapped to the + * GPU and the known addresses are not physical addresses but instead + * bus addresses (e.g., what you'd put in an IB or ring buffer). + */ static ssize_t amdgpu_iomem_write(struct file *f, const char __user *buf, size_t size, loff_t *pos) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h index 6ea7de863041b..e969c879d87e6 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ttm.h @@ -77,6 +77,7 @@ uint64_t amdgpu_vram_mgr_usage(struct ttm_mem_type_manager *man); uint64_t amdgpu_vram_mgr_vis_usage(struct ttm_mem_type_manager *man); int amdgpu_ttm_init(struct amdgpu_device *adev); +void amdgpu_ttm_late_init(struct amdgpu_device *adev); void amdgpu_ttm_fini(struct amdgpu_device *adev); void amdgpu_ttm_set_buffer_funcs_status(struct amdgpu_device *adev, bool enable); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c index 5916cc25e28be..f55f72a37ca83 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.c @@ -161,8 +161,38 @@ void amdgpu_ucode_print_rlc_hdr(const struct common_firmware_header *hdr) le32_to_cpu(rlc_hdr->reg_list_format_separate_array_offset_bytes)); DRM_DEBUG("reg_list_separate_size_bytes: %u\n", le32_to_cpu(rlc_hdr->reg_list_separate_size_bytes)); - DRM_DEBUG("reg_list_separate_size_bytes: %u\n", - le32_to_cpu(rlc_hdr->reg_list_separate_size_bytes)); + DRM_DEBUG("reg_list_separate_array_offset_bytes: %u\n", + le32_to_cpu(rlc_hdr->reg_list_separate_array_offset_bytes)); + if (version_minor == 1) { + const struct rlc_firmware_header_v2_1 *v2_1 = + container_of(rlc_hdr, struct rlc_firmware_header_v2_1, v2_0); + DRM_DEBUG("reg_list_format_direct_reg_list_length: %u\n", + le32_to_cpu(v2_1->reg_list_format_direct_reg_list_length)); + DRM_DEBUG("save_restore_list_cntl_ucode_ver: %u\n", + le32_to_cpu(v2_1->save_restore_list_cntl_ucode_ver)); + DRM_DEBUG("save_restore_list_cntl_feature_ver: %u\n", + le32_to_cpu(v2_1->save_restore_list_cntl_feature_ver)); + DRM_DEBUG("save_restore_list_cntl_size_bytes %u\n", + le32_to_cpu(v2_1->save_restore_list_cntl_size_bytes)); + DRM_DEBUG("save_restore_list_cntl_offset_bytes: %u\n", + le32_to_cpu(v2_1->save_restore_list_cntl_offset_bytes)); + DRM_DEBUG("save_restore_list_gpm_ucode_ver: %u\n", + le32_to_cpu(v2_1->save_restore_list_gpm_ucode_ver)); + DRM_DEBUG("save_restore_list_gpm_feature_ver: %u\n", + le32_to_cpu(v2_1->save_restore_list_gpm_feature_ver)); + DRM_DEBUG("save_restore_list_gpm_size_bytes %u\n", + le32_to_cpu(v2_1->save_restore_list_gpm_size_bytes)); + DRM_DEBUG("save_restore_list_gpm_offset_bytes: %u\n", + le32_to_cpu(v2_1->save_restore_list_gpm_offset_bytes)); + DRM_DEBUG("save_restore_list_srm_ucode_ver: %u\n", + le32_to_cpu(v2_1->save_restore_list_srm_ucode_ver)); + DRM_DEBUG("save_restore_list_srm_feature_ver: %u\n", + le32_to_cpu(v2_1->save_restore_list_srm_feature_ver)); + DRM_DEBUG("save_restore_list_srm_size_bytes %u\n", + le32_to_cpu(v2_1->save_restore_list_srm_size_bytes)); + DRM_DEBUG("save_restore_list_srm_offset_bytes: %u\n", + le32_to_cpu(v2_1->save_restore_list_srm_offset_bytes)); + } } else { DRM_ERROR("Unknown RLC ucode version: %u.%u\n", version_major, version_minor); } @@ -265,6 +295,7 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type) case CHIP_POLARIS10: case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: if (!load_type) return AMDGPU_FW_LOAD_DIRECT; else @@ -276,6 +307,8 @@ amdgpu_ucode_get_load_type(struct amdgpu_device *adev, int load_type) return AMDGPU_FW_LOAD_DIRECT; else return AMDGPU_FW_LOAD_PSP; + case CHIP_VEGA20: + return AMDGPU_FW_LOAD_DIRECT; default: DRM_ERROR("Unknown firmware load type\n"); } @@ -307,7 +340,10 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, (ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC1 && ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC2 && ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC1_JT && - ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC2_JT)) { + ucode->ucode_id != AMDGPU_UCODE_ID_CP_MEC2_JT && + ucode->ucode_id != AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL && + ucode->ucode_id != AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM && + ucode->ucode_id != AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM)) { ucode->ucode_size = le32_to_cpu(header->ucode_size_bytes); memcpy(ucode->kaddr, (void *)((uint8_t *)ucode->fw->data + @@ -329,6 +365,18 @@ static int amdgpu_ucode_init_single_fw(struct amdgpu_device *adev, le32_to_cpu(header->ucode_array_offset_bytes) + le32_to_cpu(cp_hdr->jt_offset) * 4), ucode->ucode_size); + } else if (ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL) { + ucode->ucode_size = adev->gfx.rlc.save_restore_list_cntl_size_bytes; + memcpy(ucode->kaddr, adev->gfx.rlc.save_restore_list_cntl, + ucode->ucode_size); + } else if (ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM) { + ucode->ucode_size = adev->gfx.rlc.save_restore_list_gpm_size_bytes; + memcpy(ucode->kaddr, adev->gfx.rlc.save_restore_list_gpm, + ucode->ucode_size); + } else if (ucode->ucode_id == AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM) { + ucode->ucode_size = adev->gfx.rlc.save_restore_list_srm_size_bytes; + memcpy(ucode->kaddr, adev->gfx.rlc.save_restore_list_srm, + ucode->ucode_size); } return 0; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h index 30b5500dc1521..08e38579af246 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_ucode.h @@ -98,6 +98,24 @@ struct rlc_firmware_header_v2_0 { uint32_t reg_list_separate_array_offset_bytes; /* payload offset from the start of the header */ }; +/* version_major=2, version_minor=1 */ +struct rlc_firmware_header_v2_1 { + struct rlc_firmware_header_v2_0 v2_0; + uint32_t reg_list_format_direct_reg_list_length; /* length of direct reg list format array */ + uint32_t save_restore_list_cntl_ucode_ver; + uint32_t save_restore_list_cntl_feature_ver; + uint32_t save_restore_list_cntl_size_bytes; + uint32_t save_restore_list_cntl_offset_bytes; + uint32_t save_restore_list_gpm_ucode_ver; + uint32_t save_restore_list_gpm_feature_ver; + uint32_t save_restore_list_gpm_size_bytes; + uint32_t save_restore_list_gpm_offset_bytes; + uint32_t save_restore_list_srm_ucode_ver; + uint32_t save_restore_list_srm_feature_ver; + uint32_t save_restore_list_srm_size_bytes; + uint32_t save_restore_list_srm_offset_bytes; +}; + /* version_major=1, version_minor=0 */ struct sdma_firmware_header_v1_0 { struct common_firmware_header header; @@ -148,6 +166,7 @@ union amdgpu_firmware_header { struct gfx_firmware_header_v1_0 gfx; struct rlc_firmware_header_v1_0 rlc; struct rlc_firmware_header_v2_0 rlc_v2_0; + struct rlc_firmware_header_v2_1 rlc_v2_1; struct sdma_firmware_header_v1_0 sdma; struct sdma_firmware_header_v1_1 sdma_v1_1; struct gpu_info_firmware_header_v1_0 gpu_info; @@ -168,6 +187,9 @@ enum AMDGPU_UCODE_ID { AMDGPU_UCODE_ID_CP_MEC2, AMDGPU_UCODE_ID_CP_MEC2_JT, AMDGPU_UCODE_ID_RLC_G, + AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL, + AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM, + AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM, AMDGPU_UCODE_ID_STORAGE, AMDGPU_UCODE_ID_SMC, AMDGPU_UCODE_ID_UVD, diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c index 627542b22ae4c..bcf68f80bbf05 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.c @@ -66,15 +66,18 @@ #define FIRMWARE_POLARIS10 "amdgpu/polaris10_uvd.bin" #define FIRMWARE_POLARIS11 "amdgpu/polaris11_uvd.bin" #define FIRMWARE_POLARIS12 "amdgpu/polaris12_uvd.bin" +#define FIRMWARE_VEGAM "amdgpu/vegam_uvd.bin" #define FIRMWARE_VEGA10 "amdgpu/vega10_uvd.bin" #define FIRMWARE_VEGA12 "amdgpu/vega12_uvd.bin" +#define FIRMWARE_VEGA20 "amdgpu/vega20_uvd.bin" -#define mmUVD_GPCOM_VCPU_DATA0_VEGA10 (0x03c4 + 0x7e00) -#define mmUVD_GPCOM_VCPU_DATA1_VEGA10 (0x03c5 + 0x7e00) -#define mmUVD_GPCOM_VCPU_CMD_VEGA10 (0x03c3 + 0x7e00) -#define mmUVD_NO_OP_VEGA10 (0x03ff + 0x7e00) -#define mmUVD_ENGINE_CNTL_VEGA10 (0x03c6 + 0x7e00) +/* These are common relative offsets for all asics, from uvd_7_0_offset.h, */ +#define UVD_GPCOM_VCPU_CMD 0x03c3 +#define UVD_GPCOM_VCPU_DATA0 0x03c4 +#define UVD_GPCOM_VCPU_DATA1 0x03c5 +#define UVD_NO_OP 0x03ff +#define UVD_BASE_SI 0x3800 /** * amdgpu_uvd_cs_ctx - Command submission parser context @@ -109,9 +112,11 @@ MODULE_FIRMWARE(FIRMWARE_STONEY); MODULE_FIRMWARE(FIRMWARE_POLARIS10); MODULE_FIRMWARE(FIRMWARE_POLARIS11); MODULE_FIRMWARE(FIRMWARE_POLARIS12); +MODULE_FIRMWARE(FIRMWARE_VEGAM); MODULE_FIRMWARE(FIRMWARE_VEGA10); MODULE_FIRMWARE(FIRMWARE_VEGA12); +MODULE_FIRMWARE(FIRMWARE_VEGA20); static void amdgpu_uvd_idle_work_handler(struct work_struct *work); @@ -123,9 +128,9 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) const char *fw_name; const struct common_firmware_header *hdr; unsigned version_major, version_minor, family_id; - int i, r; + int i, j, r; - INIT_DELAYED_WORK(&adev->uvd.idle_work, amdgpu_uvd_idle_work_handler); + INIT_DELAYED_WORK(&adev->uvd.inst->idle_work, amdgpu_uvd_idle_work_handler); switch (adev->asic_type) { #ifdef CONFIG_DRM_AMDGPU_CIK @@ -172,6 +177,12 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) case CHIP_VEGA12: fw_name = FIRMWARE_VEGA12; break; + case CHIP_VEGAM: + fw_name = FIRMWARE_VEGAM; + break; + case CHIP_VEGA20: + fw_name = FIRMWARE_VEGA20; + break; default: return -EINVAL; } @@ -226,28 +237,30 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) bo_size += AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(hdr->ucode_size_bytes) + 8); - r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, &adev->uvd.vcpu_bo, - &adev->uvd.gpu_addr, &adev->uvd.cpu_addr); - if (r) { - dev_err(adev->dev, "(%d) failed to allocate UVD bo\n", r); - return r; - } + for (j = 0; j < adev->uvd.num_uvd_inst; j++) { - ring = &adev->uvd.ring; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; - r = drm_sched_entity_init(&ring->sched, &adev->uvd.entity, - rq, amdgpu_sched_jobs, NULL); - if (r != 0) { - DRM_ERROR("Failed setting up UVD run queue.\n"); - return r; - } + r = amdgpu_bo_create_kernel(adev, bo_size, PAGE_SIZE, + AMDGPU_GEM_DOMAIN_VRAM, &adev->uvd.inst[j].vcpu_bo, + &adev->uvd.inst[j].gpu_addr, &adev->uvd.inst[j].cpu_addr); + if (r) { + dev_err(adev->dev, "(%d) failed to allocate UVD bo\n", r); + return r; + } - for (i = 0; i < adev->uvd.max_handles; ++i) { - atomic_set(&adev->uvd.handles[i], 0); - adev->uvd.filp[i] = NULL; - } + ring = &adev->uvd.inst[j].ring; + rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; + r = drm_sched_entity_init(&ring->sched, &adev->uvd.inst[j].entity, + rq, NULL); + if (r != 0) { + DRM_ERROR("Failed setting up UVD(%d) run queue.\n", j); + return r; + } + for (i = 0; i < adev->uvd.max_handles; ++i) { + atomic_set(&adev->uvd.inst[j].handles[i], 0); + adev->uvd.inst[j].filp[i] = NULL; + } + } /* from uvd v5.0 HW addressing capacity increased to 64 bits */ if (!amdgpu_device_ip_block_version_cmp(adev, AMD_IP_BLOCK_TYPE_UVD, 5, 0)) adev->uvd.address_64_bit = true; @@ -274,20 +287,22 @@ int amdgpu_uvd_sw_init(struct amdgpu_device *adev) int amdgpu_uvd_sw_fini(struct amdgpu_device *adev) { - int i; - kfree(adev->uvd.saved_bo); + int i, j; - drm_sched_entity_fini(&adev->uvd.ring.sched, &adev->uvd.entity); + for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + kfree(adev->uvd.inst[j].saved_bo); - amdgpu_bo_free_kernel(&adev->uvd.vcpu_bo, - &adev->uvd.gpu_addr, - (void **)&adev->uvd.cpu_addr); + drm_sched_entity_fini(&adev->uvd.inst[j].ring.sched, &adev->uvd.inst[j].entity); - amdgpu_ring_fini(&adev->uvd.ring); + amdgpu_bo_free_kernel(&adev->uvd.inst[j].vcpu_bo, + &adev->uvd.inst[j].gpu_addr, + (void **)&adev->uvd.inst[j].cpu_addr); - for (i = 0; i < AMDGPU_MAX_UVD_ENC_RINGS; ++i) - amdgpu_ring_fini(&adev->uvd.ring_enc[i]); + amdgpu_ring_fini(&adev->uvd.inst[j].ring); + for (i = 0; i < AMDGPU_MAX_UVD_ENC_RINGS; ++i) + amdgpu_ring_fini(&adev->uvd.inst[j].ring_enc[i]); + } release_firmware(adev->uvd.fw); return 0; @@ -297,32 +312,33 @@ int amdgpu_uvd_suspend(struct amdgpu_device *adev) { unsigned size; void *ptr; - int i; - - if (adev->uvd.vcpu_bo == NULL) - return 0; + int i, j; - cancel_delayed_work_sync(&adev->uvd.idle_work); + for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + if (adev->uvd.inst[j].vcpu_bo == NULL) + continue; - /* only valid for physical mode */ - if (adev->asic_type < CHIP_POLARIS10) { - for (i = 0; i < adev->uvd.max_handles; ++i) - if (atomic_read(&adev->uvd.handles[i])) - break; + cancel_delayed_work_sync(&adev->uvd.inst[j].idle_work); - if (i == adev->uvd.max_handles) - return 0; - } + /* only valid for physical mode */ + if (adev->asic_type < CHIP_POLARIS10) { + for (i = 0; i < adev->uvd.max_handles; ++i) + if (atomic_read(&adev->uvd.inst[j].handles[i])) + break; - size = amdgpu_bo_size(adev->uvd.vcpu_bo); - ptr = adev->uvd.cpu_addr; + if (i == adev->uvd.max_handles) + continue; + } - adev->uvd.saved_bo = kmalloc(size, GFP_KERNEL); - if (!adev->uvd.saved_bo) - return -ENOMEM; + size = amdgpu_bo_size(adev->uvd.inst[j].vcpu_bo); + ptr = adev->uvd.inst[j].cpu_addr; - memcpy_fromio(adev->uvd.saved_bo, ptr, size); + adev->uvd.inst[j].saved_bo = kmalloc(size, GFP_KERNEL); + if (!adev->uvd.inst[j].saved_bo) + return -ENOMEM; + memcpy_fromio(adev->uvd.inst[j].saved_bo, ptr, size); + } return 0; } @@ -330,59 +346,65 @@ int amdgpu_uvd_resume(struct amdgpu_device *adev) { unsigned size; void *ptr; + int i; - if (adev->uvd.vcpu_bo == NULL) - return -EINVAL; + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + if (adev->uvd.inst[i].vcpu_bo == NULL) + return -EINVAL; - size = amdgpu_bo_size(adev->uvd.vcpu_bo); - ptr = adev->uvd.cpu_addr; + size = amdgpu_bo_size(adev->uvd.inst[i].vcpu_bo); + ptr = adev->uvd.inst[i].cpu_addr; - if (adev->uvd.saved_bo != NULL) { - memcpy_toio(ptr, adev->uvd.saved_bo, size); - kfree(adev->uvd.saved_bo); - adev->uvd.saved_bo = NULL; - } else { - const struct common_firmware_header *hdr; - unsigned offset; - - hdr = (const struct common_firmware_header *)adev->uvd.fw->data; - if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { - offset = le32_to_cpu(hdr->ucode_array_offset_bytes); - memcpy_toio(adev->uvd.cpu_addr, adev->uvd.fw->data + offset, - le32_to_cpu(hdr->ucode_size_bytes)); - size -= le32_to_cpu(hdr->ucode_size_bytes); - ptr += le32_to_cpu(hdr->ucode_size_bytes); + if (adev->uvd.inst[i].saved_bo != NULL) { + memcpy_toio(ptr, adev->uvd.inst[i].saved_bo, size); + kfree(adev->uvd.inst[i].saved_bo); + adev->uvd.inst[i].saved_bo = NULL; + } else { + const struct common_firmware_header *hdr; + unsigned offset; + + hdr = (const struct common_firmware_header *)adev->uvd.fw->data; + if (adev->firmware.load_type != AMDGPU_FW_LOAD_PSP) { + offset = le32_to_cpu(hdr->ucode_array_offset_bytes); + memcpy_toio(adev->uvd.inst[i].cpu_addr, adev->uvd.fw->data + offset, + le32_to_cpu(hdr->ucode_size_bytes)); + size -= le32_to_cpu(hdr->ucode_size_bytes); + ptr += le32_to_cpu(hdr->ucode_size_bytes); + } + memset_io(ptr, 0, size); + /* to restore uvd fence seq */ + amdgpu_fence_driver_force_completion(&adev->uvd.inst[i].ring); } - memset_io(ptr, 0, size); - /* to restore uvd fence seq */ - amdgpu_fence_driver_force_completion(&adev->uvd.ring); } - return 0; } void amdgpu_uvd_free_handles(struct amdgpu_device *adev, struct drm_file *filp) { - struct amdgpu_ring *ring = &adev->uvd.ring; - int i, r; + struct amdgpu_ring *ring; + int i, j, r; - for (i = 0; i < adev->uvd.max_handles; ++i) { - uint32_t handle = atomic_read(&adev->uvd.handles[i]); - if (handle != 0 && adev->uvd.filp[i] == filp) { - struct dma_fence *fence; - - r = amdgpu_uvd_get_destroy_msg(ring, handle, - false, &fence); - if (r) { - DRM_ERROR("Error destroying UVD (%d)!\n", r); - continue; - } + for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + ring = &adev->uvd.inst[j].ring; + + for (i = 0; i < adev->uvd.max_handles; ++i) { + uint32_t handle = atomic_read(&adev->uvd.inst[j].handles[i]); + if (handle != 0 && adev->uvd.inst[j].filp[i] == filp) { + struct dma_fence *fence; + + r = amdgpu_uvd_get_destroy_msg(ring, handle, + false, &fence); + if (r) { + DRM_ERROR("Error destroying UVD(%d) %d!\n", j, r); + continue; + } - dma_fence_wait(fence, false); - dma_fence_put(fence); + dma_fence_wait(fence, false); + dma_fence_put(fence); - adev->uvd.filp[i] = NULL; - atomic_set(&adev->uvd.handles[i], 0); + adev->uvd.inst[j].filp[i] = NULL; + atomic_set(&adev->uvd.inst[j].handles[i], 0); + } } } } @@ -657,15 +679,16 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, void *ptr; long r; int i; + uint32_t ip_instance = ctx->parser->job->ring->me; if (offset & 0x3F) { - DRM_ERROR("UVD messages must be 64 byte aligned!\n"); + DRM_ERROR("UVD(%d) messages must be 64 byte aligned!\n", ip_instance); return -EINVAL; } r = amdgpu_bo_kmap(bo, &ptr); if (r) { - DRM_ERROR("Failed mapping the UVD message (%ld)!\n", r); + DRM_ERROR("Failed mapping the UVD(%d) message (%ld)!\n", ip_instance, r); return r; } @@ -675,7 +698,7 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, handle = msg[2]; if (handle == 0) { - DRM_ERROR("Invalid UVD handle!\n"); + DRM_ERROR("Invalid UVD(%d) handle!\n", ip_instance); return -EINVAL; } @@ -686,18 +709,18 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, /* try to alloc a new handle */ for (i = 0; i < adev->uvd.max_handles; ++i) { - if (atomic_read(&adev->uvd.handles[i]) == handle) { - DRM_ERROR("Handle 0x%x already in use!\n", handle); + if (atomic_read(&adev->uvd.inst[ip_instance].handles[i]) == handle) { + DRM_ERROR("(%d)Handle 0x%x already in use!\n", ip_instance, handle); return -EINVAL; } - if (!atomic_cmpxchg(&adev->uvd.handles[i], 0, handle)) { - adev->uvd.filp[i] = ctx->parser->filp; + if (!atomic_cmpxchg(&adev->uvd.inst[ip_instance].handles[i], 0, handle)) { + adev->uvd.inst[ip_instance].filp[i] = ctx->parser->filp; return 0; } } - DRM_ERROR("No more free UVD handles!\n"); + DRM_ERROR("No more free UVD(%d) handles!\n", ip_instance); return -ENOSPC; case 1: @@ -709,27 +732,27 @@ static int amdgpu_uvd_cs_msg(struct amdgpu_uvd_cs_ctx *ctx, /* validate the handle */ for (i = 0; i < adev->uvd.max_handles; ++i) { - if (atomic_read(&adev->uvd.handles[i]) == handle) { - if (adev->uvd.filp[i] != ctx->parser->filp) { - DRM_ERROR("UVD handle collision detected!\n"); + if (atomic_read(&adev->uvd.inst[ip_instance].handles[i]) == handle) { + if (adev->uvd.inst[ip_instance].filp[i] != ctx->parser->filp) { + DRM_ERROR("UVD(%d) handle collision detected!\n", ip_instance); return -EINVAL; } return 0; } } - DRM_ERROR("Invalid UVD handle 0x%x!\n", handle); + DRM_ERROR("Invalid UVD(%d) handle 0x%x!\n", ip_instance, handle); return -ENOENT; case 2: /* it's a destroy msg, free the handle */ for (i = 0; i < adev->uvd.max_handles; ++i) - atomic_cmpxchg(&adev->uvd.handles[i], handle, 0); + atomic_cmpxchg(&adev->uvd.inst[ip_instance].handles[i], handle, 0); amdgpu_bo_kunmap(bo); return 0; default: - DRM_ERROR("Illegal UVD message type (%d)!\n", msg_type); + DRM_ERROR("Illegal UVD(%d) message type (%d)!\n", ip_instance, msg_type); return -EINVAL; } BUG(); @@ -800,7 +823,7 @@ static int amdgpu_uvd_cs_pass2(struct amdgpu_uvd_cs_ctx *ctx) } if ((cmd == 0 || cmd == 0x3) && - (start >> 28) != (ctx->parser->adev->uvd.gpu_addr >> 28)) { + (start >> 28) != (ctx->parser->adev->uvd.inst->gpu_addr >> 28)) { DRM_ERROR("msg/fb buffer %LX-%LX out of 256MB segment!\n", start, end); return -EINVAL; @@ -968,6 +991,8 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, uint64_t addr; long r; int i; + unsigned offset_idx = 0; + unsigned offset[3] = { UVD_BASE_SI, 0, 0 }; amdgpu_bo_kunmap(bo); amdgpu_bo_unpin(bo); @@ -987,17 +1012,16 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, goto err; if (adev->asic_type >= CHIP_VEGA10) { - data[0] = PACKET0(mmUVD_GPCOM_VCPU_DATA0_VEGA10, 0); - data[1] = PACKET0(mmUVD_GPCOM_VCPU_DATA1_VEGA10, 0); - data[2] = PACKET0(mmUVD_GPCOM_VCPU_CMD_VEGA10, 0); - data[3] = PACKET0(mmUVD_NO_OP_VEGA10, 0); - } else { - data[0] = PACKET0(mmUVD_GPCOM_VCPU_DATA0, 0); - data[1] = PACKET0(mmUVD_GPCOM_VCPU_DATA1, 0); - data[2] = PACKET0(mmUVD_GPCOM_VCPU_CMD, 0); - data[3] = PACKET0(mmUVD_NO_OP, 0); + offset_idx = 1 + ring->me; + offset[1] = adev->reg_offset[UVD_HWIP][0][1]; + offset[2] = adev->reg_offset[UVD_HWIP][1][1]; } + data[0] = PACKET0(offset[offset_idx] + UVD_GPCOM_VCPU_DATA0, 0); + data[1] = PACKET0(offset[offset_idx] + UVD_GPCOM_VCPU_DATA1, 0); + data[2] = PACKET0(offset[offset_idx] + UVD_GPCOM_VCPU_CMD, 0); + data[3] = PACKET0(offset[offset_idx] + UVD_NO_OP, 0); + ib = &job->ibs[0]; addr = amdgpu_bo_gpu_offset(bo); ib->ptr[0] = data[0]; @@ -1033,7 +1057,7 @@ static int amdgpu_uvd_send_msg(struct amdgpu_ring *ring, struct amdgpu_bo *bo, if (r) goto err_free; - r = amdgpu_job_submit(job, ring, &adev->uvd.entity, + r = amdgpu_job_submit(job, ring, &adev->uvd.inst[ring->me].entity, AMDGPU_FENCE_OWNER_UNDEFINED, &f); if (r) goto err_free; @@ -1121,8 +1145,15 @@ int amdgpu_uvd_get_destroy_msg(struct amdgpu_ring *ring, uint32_t handle, static void amdgpu_uvd_idle_work_handler(struct work_struct *work) { struct amdgpu_device *adev = - container_of(work, struct amdgpu_device, uvd.idle_work.work); - unsigned fences = amdgpu_fence_count_emitted(&adev->uvd.ring); + container_of(work, struct amdgpu_device, uvd.inst->idle_work.work); + unsigned fences = 0, i, j; + + for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring); + for (j = 0; j < adev->uvd.num_enc_rings; ++j) { + fences += amdgpu_fence_count_emitted(&adev->uvd.inst[i].ring_enc[j]); + } + } if (fences == 0) { if (adev->pm.dpm_enabled) { @@ -1136,7 +1167,7 @@ static void amdgpu_uvd_idle_work_handler(struct work_struct *work) AMD_CG_STATE_GATE); } } else { - schedule_delayed_work(&adev->uvd.idle_work, UVD_IDLE_TIMEOUT); + schedule_delayed_work(&adev->uvd.inst->idle_work, UVD_IDLE_TIMEOUT); } } @@ -1148,7 +1179,7 @@ void amdgpu_uvd_ring_begin_use(struct amdgpu_ring *ring) if (amdgpu_sriov_vf(adev)) return; - set_clocks = !cancel_delayed_work_sync(&adev->uvd.idle_work); + set_clocks = !cancel_delayed_work_sync(&adev->uvd.inst->idle_work); if (set_clocks) { if (adev->pm.dpm_enabled) { amdgpu_dpm_enable_uvd(adev, true); @@ -1165,7 +1196,7 @@ void amdgpu_uvd_ring_begin_use(struct amdgpu_ring *ring) void amdgpu_uvd_ring_end_use(struct amdgpu_ring *ring) { if (!amdgpu_sriov_vf(ring->adev)) - schedule_delayed_work(&ring->adev->uvd.idle_work, UVD_IDLE_TIMEOUT); + schedule_delayed_work(&ring->adev->uvd.inst->idle_work, UVD_IDLE_TIMEOUT); } /** @@ -1179,27 +1210,28 @@ int amdgpu_uvd_ring_test_ib(struct amdgpu_ring *ring, long timeout) { struct dma_fence *fence; long r; + uint32_t ip_instance = ring->me; r = amdgpu_uvd_get_create_msg(ring, 1, NULL); if (r) { - DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r); + DRM_ERROR("amdgpu: (%d)failed to get create msg (%ld).\n", ip_instance, r); goto error; } r = amdgpu_uvd_get_destroy_msg(ring, 1, true, &fence); if (r) { - DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); + DRM_ERROR("amdgpu: (%d)failed to get destroy ib (%ld).\n", ip_instance, r); goto error; } r = dma_fence_wait_timeout(fence, false, timeout); if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out.\n"); + DRM_ERROR("amdgpu: (%d)IB test timed out.\n", ip_instance); r = -ETIMEDOUT; } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); + DRM_ERROR("amdgpu: (%d)fence wait failed (%ld).\n", ip_instance, r); } else { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + DRM_DEBUG("ib test on (%d)ring %d succeeded\n", ip_instance, ring->idx); r = 0; } @@ -1227,7 +1259,7 @@ uint32_t amdgpu_uvd_used_handles(struct amdgpu_device *adev) * necessarily linear. So we need to count * all non-zero handles. */ - if (atomic_read(&adev->uvd.handles[i])) + if (atomic_read(&adev->uvd.inst->handles[i])) used_handles++; } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h index 32ea20b99e53a..b1579fba134c1 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_uvd.h @@ -31,30 +31,37 @@ #define AMDGPU_UVD_SESSION_SIZE (50*1024) #define AMDGPU_UVD_FIRMWARE_OFFSET 256 +#define AMDGPU_MAX_UVD_INSTANCES 2 + #define AMDGPU_UVD_FIRMWARE_SIZE(adev) \ (AMDGPU_GPU_PAGE_ALIGN(le32_to_cpu(((const struct common_firmware_header *)(adev)->uvd.fw->data)->ucode_size_bytes) + \ 8) - AMDGPU_UVD_FIRMWARE_OFFSET) -struct amdgpu_uvd { +struct amdgpu_uvd_inst { struct amdgpu_bo *vcpu_bo; void *cpu_addr; uint64_t gpu_addr; - unsigned fw_version; void *saved_bo; - unsigned max_handles; atomic_t handles[AMDGPU_MAX_UVD_HANDLES]; struct drm_file *filp[AMDGPU_MAX_UVD_HANDLES]; struct delayed_work idle_work; - const struct firmware *fw; /* UVD firmware */ struct amdgpu_ring ring; struct amdgpu_ring ring_enc[AMDGPU_MAX_UVD_ENC_RINGS]; struct amdgpu_irq_src irq; - bool address_64_bit; - bool use_ctx_buf; struct drm_sched_entity entity; struct drm_sched_entity entity_enc; uint32_t srbm_soft_reset; +}; + +struct amdgpu_uvd { + const struct firmware *fw; /* UVD firmware */ + unsigned fw_version; + unsigned max_handles; unsigned num_enc_rings; + uint8_t num_uvd_inst; + bool address_64_bit; + bool use_ctx_buf; + struct amdgpu_uvd_inst inst[AMDGPU_MAX_UVD_INSTANCES]; }; int amdgpu_uvd_sw_init(struct amdgpu_device *adev); diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c index a33804bd33149..23d960ec1cf27 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vce.c @@ -51,11 +51,13 @@ #define FIRMWARE_FIJI "amdgpu/fiji_vce.bin" #define FIRMWARE_STONEY "amdgpu/stoney_vce.bin" #define FIRMWARE_POLARIS10 "amdgpu/polaris10_vce.bin" -#define FIRMWARE_POLARIS11 "amdgpu/polaris11_vce.bin" -#define FIRMWARE_POLARIS12 "amdgpu/polaris12_vce.bin" +#define FIRMWARE_POLARIS11 "amdgpu/polaris11_vce.bin" +#define FIRMWARE_POLARIS12 "amdgpu/polaris12_vce.bin" +#define FIRMWARE_VEGAM "amdgpu/vegam_vce.bin" #define FIRMWARE_VEGA10 "amdgpu/vega10_vce.bin" #define FIRMWARE_VEGA12 "amdgpu/vega12_vce.bin" +#define FIRMWARE_VEGA20 "amdgpu/vega20_vce.bin" #ifdef CONFIG_DRM_AMDGPU_CIK MODULE_FIRMWARE(FIRMWARE_BONAIRE); @@ -71,9 +73,11 @@ MODULE_FIRMWARE(FIRMWARE_STONEY); MODULE_FIRMWARE(FIRMWARE_POLARIS10); MODULE_FIRMWARE(FIRMWARE_POLARIS11); MODULE_FIRMWARE(FIRMWARE_POLARIS12); +MODULE_FIRMWARE(FIRMWARE_VEGAM); MODULE_FIRMWARE(FIRMWARE_VEGA10); MODULE_FIRMWARE(FIRMWARE_VEGA12); +MODULE_FIRMWARE(FIRMWARE_VEGA20); static void amdgpu_vce_idle_work_handler(struct work_struct *work); @@ -132,12 +136,18 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size) case CHIP_POLARIS12: fw_name = FIRMWARE_POLARIS12; break; + case CHIP_VEGAM: + fw_name = FIRMWARE_VEGAM; + break; case CHIP_VEGA10: fw_name = FIRMWARE_VEGA10; break; case CHIP_VEGA12: fw_name = FIRMWARE_VEGA12; break; + case CHIP_VEGA20: + fw_name = FIRMWARE_VEGA20; + break; default: return -EINVAL; @@ -181,7 +191,7 @@ int amdgpu_vce_sw_init(struct amdgpu_device *adev, unsigned long size) ring = &adev->vce.ring[0]; rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; r = drm_sched_entity_init(&ring->sched, &adev->vce.entity, - rq, amdgpu_sched_jobs, NULL); + rq, NULL); if (r != 0) { DRM_ERROR("Failed setting up VCE run queue.\n"); return r; @@ -755,6 +765,18 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) if (r) goto out; break; + + case 0x0500000d: /* MV buffer */ + r = amdgpu_vce_validate_bo(p, ib_idx, idx + 3, + idx + 2, 0, 0); + if (r) + goto out; + + r = amdgpu_vce_validate_bo(p, ib_idx, idx + 8, + idx + 7, 0, 0); + if (r) + goto out; + break; } idx += len / 4; @@ -860,6 +882,18 @@ int amdgpu_vce_ring_parse_cs(struct amdgpu_cs_parser *p, uint32_t ib_idx) goto out; break; + case 0x0500000d: /* MV buffer */ + r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 3, + idx + 2, *size, 0); + if (r) + goto out; + + r = amdgpu_vce_cs_reloc(p, ib_idx, idx + 8, + idx + 7, *size / 12, 0); + if (r) + goto out; + break; + default: DRM_ERROR("invalid VCE command (0x%x)!\n", cmd); r = -EINVAL; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c index 58e495330b38b..8851bcdfc2609 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.c @@ -105,7 +105,7 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev) ring = &adev->vcn.ring_dec; rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; r = drm_sched_entity_init(&ring->sched, &adev->vcn.entity_dec, - rq, amdgpu_sched_jobs, NULL); + rq, NULL); if (r != 0) { DRM_ERROR("Failed setting up VCN dec run queue.\n"); return r; @@ -114,7 +114,7 @@ int amdgpu_vcn_sw_init(struct amdgpu_device *adev) ring = &adev->vcn.ring_enc[0]; rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; r = drm_sched_entity_init(&ring->sched, &adev->vcn.entity_enc, - rq, amdgpu_sched_jobs, NULL); + rq, NULL); if (r != 0) { DRM_ERROR("Failed setting up VCN enc run queue.\n"); return r; @@ -205,13 +205,18 @@ static void amdgpu_vcn_idle_work_handler(struct work_struct *work) struct amdgpu_device *adev = container_of(work, struct amdgpu_device, vcn.idle_work.work); unsigned fences = amdgpu_fence_count_emitted(&adev->vcn.ring_dec); + unsigned i; + + for (i = 0; i < adev->vcn.num_enc_rings; ++i) { + fences += amdgpu_fence_count_emitted(&adev->vcn.ring_enc[i]); + } if (fences == 0) { - if (adev->pm.dpm_enabled) { - /* might be used when with pg/cg + if (adev->pm.dpm_enabled) amdgpu_dpm_enable_uvd(adev, false); - */ - } + else + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, + AMD_PG_STATE_GATE); } else { schedule_delayed_work(&adev->vcn.idle_work, VCN_IDLE_TIMEOUT); } @@ -223,9 +228,11 @@ void amdgpu_vcn_ring_begin_use(struct amdgpu_ring *ring) bool set_clocks = !cancel_delayed_work_sync(&adev->vcn.idle_work); if (set_clocks && adev->pm.dpm_enabled) { - /* might be used when with pg/cg - amdgpu_dpm_enable_uvd(adev, true); - */ + if (adev->pm.dpm_enabled) + amdgpu_dpm_enable_uvd(adev, true); + else + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_VCN, + AMD_PG_STATE_UNGATE); } } diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h index 2fd7db891689f..181e6afa98472 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vcn.h @@ -45,6 +45,17 @@ #define VCN_ENC_CMD_REG_WRITE 0x0000000b #define VCN_ENC_CMD_REG_WAIT 0x0000000c +enum engine_status_constants { + UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON = 0x2AAAA0, + UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON = 0x00000002, + UVD_STATUS__UVD_BUSY = 0x00000004, + GB_ADDR_CONFIG_DEFAULT = 0x26010011, + UVD_STATUS__IDLE = 0x2, + UVD_STATUS__BUSY = 0x5, + UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF = 0x1, + UVD_STATUS__RBC_BUSY = 0x1, +}; + struct amdgpu_vcn { struct amdgpu_bo *vcpu_bo; void *cpu_addr; diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c index da55a78d7380b..ccba88cc8c542 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c @@ -94,6 +94,34 @@ struct amdgpu_prt_cb { struct dma_fence_cb cb; }; +static void amdgpu_vm_bo_base_init(struct amdgpu_vm_bo_base *base, + struct amdgpu_vm *vm, + struct amdgpu_bo *bo) +{ + base->vm = vm; + base->bo = bo; + INIT_LIST_HEAD(&base->bo_list); + INIT_LIST_HEAD(&base->vm_status); + + if (!bo) + return; + list_add_tail(&base->bo_list, &bo->va); + + if (bo->tbo.resv != vm->root.base.bo->tbo.resv) + return; + + if (bo->preferred_domains & + amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type)) + return; + + /* + * we checked all the prerequisites, but it looks like this per vm bo + * is currently evicted. add the bo to the evicted list to make sure it + * is validated on next vm use to avoid fault. + * */ + list_move_tail(&base->vm_status, &vm->evicted); +} + /** * amdgpu_vm_level_shift - return the addr shift for each level * @@ -196,24 +224,16 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, void *param) { struct ttm_bo_global *glob = adev->mman.bdev.glob; - int r; + struct amdgpu_vm_bo_base *bo_base, *tmp; + int r = 0; - spin_lock(&vm->status_lock); - while (!list_empty(&vm->evicted)) { - struct amdgpu_vm_bo_base *bo_base; - struct amdgpu_bo *bo; + list_for_each_entry_safe(bo_base, tmp, &vm->evicted, vm_status) { + struct amdgpu_bo *bo = bo_base->bo; - bo_base = list_first_entry(&vm->evicted, - struct amdgpu_vm_bo_base, - vm_status); - spin_unlock(&vm->status_lock); - - bo = bo_base->bo; - BUG_ON(!bo); if (bo->parent) { r = validate(param, bo); if (r) - return r; + break; spin_lock(&glob->lru_lock); ttm_bo_move_to_lru_tail(&bo->tbo); @@ -222,22 +242,29 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, spin_unlock(&glob->lru_lock); } - if (bo->tbo.type == ttm_bo_type_kernel && - vm->use_cpu_for_update) { - r = amdgpu_bo_kmap(bo, NULL); - if (r) - return r; - } - - spin_lock(&vm->status_lock); - if (bo->tbo.type != ttm_bo_type_kernel) + if (bo->tbo.type != ttm_bo_type_kernel) { + spin_lock(&vm->moved_lock); list_move(&bo_base->vm_status, &vm->moved); - else + spin_unlock(&vm->moved_lock); + } else { list_move(&bo_base->vm_status, &vm->relocated); + } } - spin_unlock(&vm->status_lock); - return 0; + spin_lock(&glob->lru_lock); + list_for_each_entry(bo_base, &vm->idle, vm_status) { + struct amdgpu_bo *bo = bo_base->bo; + + if (!bo->parent) + continue; + + ttm_bo_move_to_lru_tail(&bo->tbo); + if (bo->shadow) + ttm_bo_move_to_lru_tail(&bo->shadow->tbo); + } + spin_unlock(&glob->lru_lock); + + return r; } /** @@ -249,13 +276,7 @@ int amdgpu_vm_validate_pt_bos(struct amdgpu_device *adev, struct amdgpu_vm *vm, */ bool amdgpu_vm_ready(struct amdgpu_vm *vm) { - bool ready; - - spin_lock(&vm->status_lock); - ready = list_empty(&vm->evicted); - spin_unlock(&vm->status_lock); - - return ready; + return list_empty(&vm->evicted); } /** @@ -412,11 +433,16 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, struct amdgpu_bo *pt; if (!entry->base.bo) { - r = amdgpu_bo_create(adev, - amdgpu_vm_bo_size(adev, level), - AMDGPU_GPU_PAGE_SIZE, - AMDGPU_GEM_DOMAIN_VRAM, flags, - ttm_bo_type_kernel, resv, &pt); + struct amdgpu_bo_param bp; + + memset(&bp, 0, sizeof(bp)); + bp.size = amdgpu_vm_bo_size(adev, level); + bp.byte_align = AMDGPU_GPU_PAGE_SIZE; + bp.domain = AMDGPU_GEM_DOMAIN_VRAM; + bp.flags = flags; + bp.type = ttm_bo_type_kernel; + bp.resv = resv; + r = amdgpu_bo_create(adev, &bp, &pt); if (r) return r; @@ -441,12 +467,8 @@ static int amdgpu_vm_alloc_levels(struct amdgpu_device *adev, */ pt->parent = amdgpu_bo_ref(parent->base.bo); - entry->base.vm = vm; - entry->base.bo = pt; - list_add_tail(&entry->base.bo_list, &pt->va); - spin_lock(&vm->status_lock); - list_add(&entry->base.vm_status, &vm->relocated); - spin_unlock(&vm->status_lock); + amdgpu_vm_bo_base_init(&entry->base, vm, pt); + list_move(&entry->base.vm_status, &vm->relocated); } if (level < AMDGPU_VM_PTB) { @@ -628,7 +650,7 @@ int amdgpu_vm_flush(struct amdgpu_ring *ring, struct amdgpu_job *job, bool need_ amdgpu_gmc_emit_pasid_mapping(ring, job->vmid, job->pasid); if (vm_flush_needed || pasid_mapping_needed) { - r = amdgpu_fence_emit(ring, &fence); + r = amdgpu_fence_emit(ring, &fence, 0); if (r) return r; } @@ -893,10 +915,8 @@ static void amdgpu_vm_invalidate_level(struct amdgpu_device *adev, if (!entry->base.bo) continue; - spin_lock(&vm->status_lock); - if (list_empty(&entry->base.vm_status)) - list_add(&entry->base.vm_status, &vm->relocated); - spin_unlock(&vm->status_lock); + if (!entry->base.moved) + list_move(&entry->base.vm_status, &vm->relocated); amdgpu_vm_invalidate_level(adev, vm, entry, level + 1); } } @@ -926,6 +946,14 @@ restart: params.adev = adev; if (vm->use_cpu_for_update) { + struct amdgpu_vm_bo_base *bo_base; + + list_for_each_entry(bo_base, &vm->relocated, vm_status) { + r = amdgpu_bo_kmap(bo_base->bo, NULL); + if (unlikely(r)) + return r; + } + r = amdgpu_vm_wait_pd(adev, vm, AMDGPU_FENCE_OWNER_VM); if (unlikely(r)) return r; @@ -941,7 +969,6 @@ restart: params.func = amdgpu_vm_do_set_ptes; } - spin_lock(&vm->status_lock); while (!list_empty(&vm->relocated)) { struct amdgpu_vm_bo_base *bo_base, *parent; struct amdgpu_vm_pt *pt, *entry; @@ -950,14 +977,12 @@ restart: bo_base = list_first_entry(&vm->relocated, struct amdgpu_vm_bo_base, vm_status); - list_del_init(&bo_base->vm_status); - spin_unlock(&vm->status_lock); + bo_base->moved = false; + list_move(&bo_base->vm_status, &vm->idle); bo = bo_base->bo->parent; - if (!bo) { - spin_lock(&vm->status_lock); + if (!bo) continue; - } parent = list_first_entry(&bo->va, struct amdgpu_vm_bo_base, bo_list); @@ -966,12 +991,10 @@ restart: amdgpu_vm_update_pde(¶ms, vm, pt, entry); - spin_lock(&vm->status_lock); if (!vm->use_cpu_for_update && (ndw - params.ib->length_dw) < 32) break; } - spin_unlock(&vm->status_lock); if (vm->use_cpu_for_update) { /* Flush HDP */ @@ -1074,9 +1097,7 @@ static void amdgpu_vm_handle_huge_pages(struct amdgpu_pte_update_params *p, if (entry->huge) { /* Add the entry to the relocated list to update it. */ entry->huge = false; - spin_lock(&p->vm->status_lock); list_move(&entry->base.vm_status, &p->vm->relocated); - spin_unlock(&p->vm->status_lock); } return; } @@ -1555,9 +1576,22 @@ int amdgpu_vm_bo_update(struct amdgpu_device *adev, amdgpu_asic_flush_hdp(adev, NULL); } - spin_lock(&vm->status_lock); + spin_lock(&vm->moved_lock); list_del_init(&bo_va->base.vm_status); - spin_unlock(&vm->status_lock); + spin_unlock(&vm->moved_lock); + + /* If the BO is not in its preferred location add it back to + * the evicted list so that it gets validated again on the + * next command submission. + */ + if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv) { + uint32_t mem_type = bo->tbo.mem.mem_type; + + if (!(bo->preferred_domains & amdgpu_mem_type_to_domain(mem_type))) + list_add_tail(&bo_va->base.vm_status, &vm->evicted); + else + list_add(&bo_va->base.vm_status, &vm->idle); + } list_splice_init(&bo_va->invalids, &bo_va->valids); bo_va->cleared = clear; @@ -1766,19 +1800,18 @@ int amdgpu_vm_clear_freed(struct amdgpu_device *adev, int amdgpu_vm_handle_moved(struct amdgpu_device *adev, struct amdgpu_vm *vm) { + struct amdgpu_bo_va *bo_va, *tmp; + struct list_head moved; bool clear; - int r = 0; - - spin_lock(&vm->status_lock); - while (!list_empty(&vm->moved)) { - struct amdgpu_bo_va *bo_va; - struct reservation_object *resv; + int r; - bo_va = list_first_entry(&vm->moved, - struct amdgpu_bo_va, base.vm_status); - spin_unlock(&vm->status_lock); + INIT_LIST_HEAD(&moved); + spin_lock(&vm->moved_lock); + list_splice_init(&vm->moved, &moved); + spin_unlock(&vm->moved_lock); - resv = bo_va->base.bo->tbo.resv; + list_for_each_entry_safe(bo_va, tmp, &moved, base.vm_status) { + struct reservation_object *resv = bo_va->base.bo->tbo.resv; /* Per VM BOs never need to bo cleared in the page tables */ if (resv == vm->root.base.bo->tbo.resv) @@ -1791,17 +1824,19 @@ int amdgpu_vm_handle_moved(struct amdgpu_device *adev, clear = true; r = amdgpu_vm_bo_update(adev, bo_va, clear); - if (r) + if (r) { + spin_lock(&vm->moved_lock); + list_splice(&moved, &vm->moved); + spin_unlock(&vm->moved_lock); return r; + } if (!clear && resv != vm->root.base.bo->tbo.resv) reservation_object_unlock(resv); - spin_lock(&vm->status_lock); } - spin_unlock(&vm->status_lock); - return r; + return 0; } /** @@ -1827,36 +1862,12 @@ struct amdgpu_bo_va *amdgpu_vm_bo_add(struct amdgpu_device *adev, if (bo_va == NULL) { return NULL; } - bo_va->base.vm = vm; - bo_va->base.bo = bo; - INIT_LIST_HEAD(&bo_va->base.bo_list); - INIT_LIST_HEAD(&bo_va->base.vm_status); + amdgpu_vm_bo_base_init(&bo_va->base, vm, bo); bo_va->ref_count = 1; INIT_LIST_HEAD(&bo_va->valids); INIT_LIST_HEAD(&bo_va->invalids); - if (!bo) - return bo_va; - - list_add_tail(&bo_va->base.bo_list, &bo->va); - - if (bo->tbo.resv != vm->root.base.bo->tbo.resv) - return bo_va; - - if (bo->preferred_domains & - amdgpu_mem_type_to_domain(bo->tbo.mem.mem_type)) - return bo_va; - - /* - * We checked all the prerequisites, but it looks like this per VM BO - * is currently evicted. add the BO to the evicted list to make sure it - * is validated on next VM use to avoid fault. - * */ - spin_lock(&vm->status_lock); - list_move_tail(&bo_va->base.vm_status, &vm->evicted); - spin_unlock(&vm->status_lock); - return bo_va; } @@ -1884,11 +1895,11 @@ static void amdgpu_vm_bo_insert_map(struct amdgpu_device *adev, if (mapping->flags & AMDGPU_PTE_PRT) amdgpu_vm_prt_get(adev); - if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv) { - spin_lock(&vm->status_lock); - if (list_empty(&bo_va->base.vm_status)) - list_add(&bo_va->base.vm_status, &vm->moved); - spin_unlock(&vm->status_lock); + if (bo && bo->tbo.resv == vm->root.base.bo->tbo.resv && + !bo_va->base.moved) { + spin_lock(&vm->moved_lock); + list_move(&bo_va->base.vm_status, &vm->moved); + spin_unlock(&vm->moved_lock); } trace_amdgpu_vm_bo_map(bo_va, mapping); } @@ -2198,9 +2209,9 @@ void amdgpu_vm_bo_rmv(struct amdgpu_device *adev, list_del(&bo_va->base.bo_list); - spin_lock(&vm->status_lock); + spin_lock(&vm->moved_lock); list_del(&bo_va->base.vm_status); - spin_unlock(&vm->status_lock); + spin_unlock(&vm->moved_lock); list_for_each_entry_safe(mapping, next, &bo_va->valids, list) { list_del(&mapping->list); @@ -2234,33 +2245,34 @@ void amdgpu_vm_bo_invalidate(struct amdgpu_device *adev, { struct amdgpu_vm_bo_base *bo_base; + /* shadow bo doesn't have bo base, its validation needs its parent */ + if (bo->parent && bo->parent->shadow == bo) + bo = bo->parent; + list_for_each_entry(bo_base, &bo->va, bo_list) { struct amdgpu_vm *vm = bo_base->vm; + bool was_moved = bo_base->moved; bo_base->moved = true; if (evicted && bo->tbo.resv == vm->root.base.bo->tbo.resv) { - spin_lock(&bo_base->vm->status_lock); if (bo->tbo.type == ttm_bo_type_kernel) list_move(&bo_base->vm_status, &vm->evicted); else list_move_tail(&bo_base->vm_status, &vm->evicted); - spin_unlock(&bo_base->vm->status_lock); continue; } - if (bo->tbo.type == ttm_bo_type_kernel) { - spin_lock(&bo_base->vm->status_lock); - if (list_empty(&bo_base->vm_status)) - list_add(&bo_base->vm_status, &vm->relocated); - spin_unlock(&bo_base->vm->status_lock); + if (was_moved) continue; - } - spin_lock(&bo_base->vm->status_lock); - if (list_empty(&bo_base->vm_status)) - list_add(&bo_base->vm_status, &vm->moved); - spin_unlock(&bo_base->vm->status_lock); + if (bo->tbo.type == ttm_bo_type_kernel) { + list_move(&bo_base->vm_status, &vm->relocated); + } else { + spin_lock(&bo_base->vm->moved_lock); + list_move(&bo_base->vm_status, &vm->moved); + spin_unlock(&bo_base->vm->moved_lock); + } } } @@ -2355,6 +2367,8 @@ void amdgpu_vm_adjust_size(struct amdgpu_device *adev, uint32_t vm_size, int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, int vm_context, unsigned int pasid) { + struct amdgpu_bo_param bp; + struct amdgpu_bo *root; const unsigned align = min(AMDGPU_VM_PTB_ALIGN_SIZE, AMDGPU_VM_PTE_COUNT(adev) * 8); unsigned ring_instance; @@ -2367,10 +2381,11 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, vm->va = RB_ROOT_CACHED; for (i = 0; i < AMDGPU_MAX_VMHUBS; i++) vm->reserved_vmid[i] = NULL; - spin_lock_init(&vm->status_lock); INIT_LIST_HEAD(&vm->evicted); INIT_LIST_HEAD(&vm->relocated); + spin_lock_init(&vm->moved_lock); INIT_LIST_HEAD(&vm->moved); + INIT_LIST_HEAD(&vm->idle); INIT_LIST_HEAD(&vm->freed); /* create scheduler entity for page table updates */ @@ -2380,7 +2395,7 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, ring = adev->vm_manager.vm_pte_rings[ring_instance]; rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_KERNEL]; r = drm_sched_entity_init(&ring->sched, &vm->entity, - rq, amdgpu_sched_jobs, NULL); + rq, NULL); if (r) return r; @@ -2409,24 +2424,28 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, flags |= AMDGPU_GEM_CREATE_SHADOW; size = amdgpu_vm_bo_size(adev, adev->vm_manager.root_level); - r = amdgpu_bo_create(adev, size, align, AMDGPU_GEM_DOMAIN_VRAM, flags, - ttm_bo_type_kernel, NULL, &vm->root.base.bo); + memset(&bp, 0, sizeof(bp)); + bp.size = size; + bp.byte_align = align; + bp.domain = AMDGPU_GEM_DOMAIN_VRAM; + bp.flags = flags; + bp.type = ttm_bo_type_kernel; + bp.resv = NULL; + r = amdgpu_bo_create(adev, &bp, &root); if (r) goto error_free_sched_entity; - r = amdgpu_bo_reserve(vm->root.base.bo, true); + r = amdgpu_bo_reserve(root, true); if (r) goto error_free_root; - r = amdgpu_vm_clear_bo(adev, vm, vm->root.base.bo, + r = amdgpu_vm_clear_bo(adev, vm, root, adev->vm_manager.root_level, vm->pte_support_ats); if (r) goto error_unreserve; - vm->root.base.vm = vm; - list_add_tail(&vm->root.base.bo_list, &vm->root.base.bo->va); - list_add_tail(&vm->root.base.vm_status, &vm->evicted); + amdgpu_vm_bo_base_init(&vm->root.base, vm, root); amdgpu_bo_unreserve(vm->root.base.bo); if (pasid) { diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h index 30f080364c97b..061b99a18cb81 100644 --- a/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h @@ -75,11 +75,12 @@ struct amdgpu_bo_list_entry; /* PDE Block Fragment Size for VEGA10 */ #define AMDGPU_PDE_BFS(a) ((uint64_t)a << 59) -/* VEGA10 only */ + +/* For GFX9 */ #define AMDGPU_PTE_MTYPE(a) ((uint64_t)a << 57) #define AMDGPU_PTE_MTYPE_MASK AMDGPU_PTE_MTYPE(3ULL) -/* For Raven */ +#define AMDGPU_MTYPE_NC 0 #define AMDGPU_MTYPE_CC 2 #define AMDGPU_PTE_DEFAULT_ATC (AMDGPU_PTE_SYSTEM \ @@ -167,9 +168,6 @@ struct amdgpu_vm { /* tree of virtual addresses mapped */ struct rb_root_cached va; - /* protecting invalidated */ - spinlock_t status_lock; - /* BOs who needs a validation */ struct list_head evicted; @@ -178,6 +176,10 @@ struct amdgpu_vm { /* BOs moved, but not yet updated in the PT */ struct list_head moved; + spinlock_t moved_lock; + + /* All BOs of this VM not currently in the state machine */ + struct list_head idle; /* BO mappings freed, but not yet updated in the PT */ struct list_head freed; @@ -186,9 +188,6 @@ struct amdgpu_vm { struct amdgpu_vm_pt root; struct dma_fence *last_update; - /* protecting freed */ - spinlock_t freed_lock; - /* Scheduler entity for page table updates */ struct drm_sched_entity entity; diff --git a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c index 47ef3e6e71786..a266dcf5daed2 100644 --- a/drivers/gpu/drm/amd/amdgpu/ci_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/ci_dpm.c @@ -5903,7 +5903,7 @@ static int ci_dpm_init(struct amdgpu_device *adev) pi->pcie_dpm_key_disabled = 0; pi->thermal_sclk_dpm_enabled = 0; - if (amdgpu_pp_feature_mask & SCLK_DEEP_SLEEP_MASK) + if (adev->powerplay.pp_feature & PP_SCLK_DEEP_SLEEP_MASK) pi->caps_sclk_ds = true; else pi->caps_sclk_ds = false; @@ -6255,7 +6255,7 @@ static int ci_dpm_late_init(void *handle) int ret; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (!amdgpu_dpm) + if (!adev->pm.dpm_enabled) return 0; /* init the sysfs and debugfs files late */ diff --git a/drivers/gpu/drm/amd/amdgpu/cik.c b/drivers/gpu/drm/amd/amdgpu/cik.c index 0df22030e7137..8ff4c60d1b595 100644 --- a/drivers/gpu/drm/amd/amdgpu/cik.c +++ b/drivers/gpu/drm/amd/amdgpu/cik.c @@ -1735,6 +1735,12 @@ static void cik_invalidate_hdp(struct amdgpu_device *adev, } } +static bool cik_need_full_reset(struct amdgpu_device *adev) +{ + /* change this when we support soft reset */ + return true; +} + static const struct amdgpu_asic_funcs cik_asic_funcs = { .read_disabled_bios = &cik_read_disabled_bios, @@ -1748,6 +1754,7 @@ static const struct amdgpu_asic_funcs cik_asic_funcs = .get_config_memsize = &cik_get_config_memsize, .flush_hdp = &cik_flush_hdp, .invalidate_hdp = &cik_invalidate_hdp, + .need_full_reset = &cik_need_full_reset, }; static int cik_common_early_init(void *handle) diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c index 452f88ea46a2e..ada241bfeee96 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v10_0.c @@ -1823,7 +1823,6 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc, struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; - struct amdgpu_framebuffer *amdgpu_fb; struct drm_framebuffer *target_fb; struct drm_gem_object *obj; struct amdgpu_bo *abo; @@ -1842,18 +1841,15 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc, return 0; } - if (atomic) { - amdgpu_fb = to_amdgpu_framebuffer(fb); + if (atomic) target_fb = fb; - } else { - amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); + else target_fb = crtc->primary->fb; - } /* If atomic, assume fb object is pinned & idle & fenced and * just update base pointers */ - obj = amdgpu_fb->obj; + obj = target_fb->obj[0]; abo = gem_to_amdgpu_bo(obj); r = amdgpu_bo_reserve(abo, false); if (unlikely(r != 0)) @@ -2043,8 +2039,7 @@ static int dce_v10_0_crtc_do_set_base(struct drm_crtc *crtc, WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { - amdgpu_fb = to_amdgpu_framebuffer(fb); - abo = gem_to_amdgpu_bo(amdgpu_fb->obj); + abo = gem_to_amdgpu_bo(fb->obj[0]); r = amdgpu_bo_reserve(abo, true); if (unlikely(r != 0)) return r; @@ -2526,11 +2521,9 @@ static void dce_v10_0_crtc_disable(struct drm_crtc *crtc) dce_v10_0_crtc_dpms(crtc, DRM_MODE_DPMS_OFF); if (crtc->primary->fb) { int r; - struct amdgpu_framebuffer *amdgpu_fb; struct amdgpu_bo *abo; - amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); - abo = gem_to_amdgpu_bo(amdgpu_fb->obj); + abo = gem_to_amdgpu_bo(crtc->primary->fb->obj[0]); r = amdgpu_bo_reserve(abo, true); if (unlikely(r)) DRM_ERROR("failed to reserve abo before unpin\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c index a7c1c584a1913..a5b96eac30332 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v11_0.c @@ -173,6 +173,7 @@ static void dce_v11_0_init_golden_registers(struct amdgpu_device *adev) ARRAY_SIZE(polaris11_golden_settings_a11)); break; case CHIP_POLARIS10: + case CHIP_VEGAM: amdgpu_device_program_register_sequence(adev, polaris10_golden_settings_a11, ARRAY_SIZE(polaris10_golden_settings_a11)); @@ -473,6 +474,7 @@ static int dce_v11_0_get_num_crtc (struct amdgpu_device *adev) num_crtc = 2; break; case CHIP_POLARIS10: + case CHIP_VEGAM: num_crtc = 6; break; case CHIP_POLARIS11: @@ -1445,6 +1447,7 @@ static int dce_v11_0_audio_init(struct amdgpu_device *adev) adev->mode_info.audio.num_pins = 7; break; case CHIP_POLARIS10: + case CHIP_VEGAM: adev->mode_info.audio.num_pins = 8; break; case CHIP_POLARIS11: @@ -1862,7 +1865,6 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc, struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; - struct amdgpu_framebuffer *amdgpu_fb; struct drm_framebuffer *target_fb; struct drm_gem_object *obj; struct amdgpu_bo *abo; @@ -1881,18 +1883,15 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc, return 0; } - if (atomic) { - amdgpu_fb = to_amdgpu_framebuffer(fb); + if (atomic) target_fb = fb; - } else { - amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); + else target_fb = crtc->primary->fb; - } /* If atomic, assume fb object is pinned & idle & fenced and * just update base pointers */ - obj = amdgpu_fb->obj; + obj = target_fb->obj[0]; abo = gem_to_amdgpu_bo(obj); r = amdgpu_bo_reserve(abo, false); if (unlikely(r != 0)) @@ -2082,8 +2081,7 @@ static int dce_v11_0_crtc_do_set_base(struct drm_crtc *crtc, WREG32(mmCRTC_MASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { - amdgpu_fb = to_amdgpu_framebuffer(fb); - abo = gem_to_amdgpu_bo(amdgpu_fb->obj); + abo = gem_to_amdgpu_bo(fb->obj[0]); r = amdgpu_bo_reserve(abo, true); if (unlikely(r != 0)) return r; @@ -2253,7 +2251,8 @@ static u32 dce_v11_0_pick_pll(struct drm_crtc *crtc) if ((adev->asic_type == CHIP_POLARIS10) || (adev->asic_type == CHIP_POLARIS11) || - (adev->asic_type == CHIP_POLARIS12)) { + (adev->asic_type == CHIP_POLARIS12) || + (adev->asic_type == CHIP_VEGAM)) { struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(amdgpu_crtc->encoder); struct amdgpu_encoder_atom_dig *dig = amdgpu_encoder->enc_priv; @@ -2601,11 +2600,9 @@ static void dce_v11_0_crtc_disable(struct drm_crtc *crtc) dce_v11_0_crtc_dpms(crtc, DRM_MODE_DPMS_OFF); if (crtc->primary->fb) { int r; - struct amdgpu_framebuffer *amdgpu_fb; struct amdgpu_bo *abo; - amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); - abo = gem_to_amdgpu_bo(amdgpu_fb->obj); + abo = gem_to_amdgpu_bo(crtc->primary->fb->obj[0]); r = amdgpu_bo_reserve(abo, true); if (unlikely(r)) DRM_ERROR("failed to reserve abo before unpin\n"); @@ -2673,7 +2670,8 @@ static int dce_v11_0_crtc_mode_set(struct drm_crtc *crtc, if ((adev->asic_type == CHIP_POLARIS10) || (adev->asic_type == CHIP_POLARIS11) || - (adev->asic_type == CHIP_POLARIS12)) { + (adev->asic_type == CHIP_POLARIS12) || + (adev->asic_type == CHIP_VEGAM)) { struct amdgpu_encoder *amdgpu_encoder = to_amdgpu_encoder(amdgpu_crtc->encoder); int encoder_mode = @@ -2830,6 +2828,7 @@ static int dce_v11_0_early_init(void *handle) adev->mode_info.num_dig = 9; break; case CHIP_POLARIS10: + case CHIP_VEGAM: adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 6; break; @@ -2949,7 +2948,8 @@ static int dce_v11_0_hw_init(void *handle) amdgpu_atombios_encoder_init_dig(adev); if ((adev->asic_type == CHIP_POLARIS10) || (adev->asic_type == CHIP_POLARIS11) || - (adev->asic_type == CHIP_POLARIS12)) { + (adev->asic_type == CHIP_POLARIS12) || + (adev->asic_type == CHIP_VEGAM)) { amdgpu_atombios_crtc_set_dce_clock(adev, adev->clock.default_dispclk, DCE_CLOCK_TYPE_DISPCLK, ATOM_GCK_DFS); amdgpu_atombios_crtc_set_dce_clock(adev, 0, diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c index 9f67b7fd34879..394cc1e8fe20e 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v6_0.c @@ -1780,7 +1780,6 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc, struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; - struct amdgpu_framebuffer *amdgpu_fb; struct drm_framebuffer *target_fb; struct drm_gem_object *obj; struct amdgpu_bo *abo; @@ -1798,18 +1797,15 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc, return 0; } - if (atomic) { - amdgpu_fb = to_amdgpu_framebuffer(fb); + if (atomic) target_fb = fb; - } else { - amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); + else target_fb = crtc->primary->fb; - } /* If atomic, assume fb object is pinned & idle & fenced and * just update base pointers */ - obj = amdgpu_fb->obj; + obj = target_fb->obj[0]; abo = gem_to_amdgpu_bo(obj); r = amdgpu_bo_reserve(abo, false); if (unlikely(r != 0)) @@ -1978,8 +1974,7 @@ static int dce_v6_0_crtc_do_set_base(struct drm_crtc *crtc, WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { - amdgpu_fb = to_amdgpu_framebuffer(fb); - abo = gem_to_amdgpu_bo(amdgpu_fb->obj); + abo = gem_to_amdgpu_bo(fb->obj[0]); r = amdgpu_bo_reserve(abo, true); if (unlikely(r != 0)) return r; @@ -2414,11 +2409,9 @@ static void dce_v6_0_crtc_disable(struct drm_crtc *crtc) dce_v6_0_crtc_dpms(crtc, DRM_MODE_DPMS_OFF); if (crtc->primary->fb) { int r; - struct amdgpu_framebuffer *amdgpu_fb; struct amdgpu_bo *abo; - amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); - abo = gem_to_amdgpu_bo(amdgpu_fb->obj); + abo = gem_to_amdgpu_bo(crtc->primary->fb->obj[0]); r = amdgpu_bo_reserve(abo, true); if (unlikely(r)) DRM_ERROR("failed to reserve abo before unpin\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c index f55422cbd77ab..c9b9ab8f1b059 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_v8_0.c @@ -1754,7 +1754,6 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); struct drm_device *dev = crtc->dev; struct amdgpu_device *adev = dev->dev_private; - struct amdgpu_framebuffer *amdgpu_fb; struct drm_framebuffer *target_fb; struct drm_gem_object *obj; struct amdgpu_bo *abo; @@ -1773,18 +1772,15 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, return 0; } - if (atomic) { - amdgpu_fb = to_amdgpu_framebuffer(fb); + if (atomic) target_fb = fb; - } else { - amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); + else target_fb = crtc->primary->fb; - } /* If atomic, assume fb object is pinned & idle & fenced and * just update base pointers */ - obj = amdgpu_fb->obj; + obj = target_fb->obj[0]; abo = gem_to_amdgpu_bo(obj); r = amdgpu_bo_reserve(abo, false); if (unlikely(r != 0)) @@ -1955,8 +1951,7 @@ static int dce_v8_0_crtc_do_set_base(struct drm_crtc *crtc, WREG32(mmMASTER_UPDATE_MODE + amdgpu_crtc->crtc_offset, 0); if (!atomic && fb && fb != crtc->primary->fb) { - amdgpu_fb = to_amdgpu_framebuffer(fb); - abo = gem_to_amdgpu_bo(amdgpu_fb->obj); + abo = gem_to_amdgpu_bo(fb->obj[0]); r = amdgpu_bo_reserve(abo, true); if (unlikely(r != 0)) return r; @@ -2430,11 +2425,9 @@ static void dce_v8_0_crtc_disable(struct drm_crtc *crtc) dce_v8_0_crtc_dpms(crtc, DRM_MODE_DPMS_OFF); if (crtc->primary->fb) { int r; - struct amdgpu_framebuffer *amdgpu_fb; struct amdgpu_bo *abo; - amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); - abo = gem_to_amdgpu_bo(amdgpu_fb->obj); + abo = gem_to_amdgpu_bo(crtc->primary->fb->obj[0]); r = amdgpu_bo_reserve(abo, true); if (unlikely(r)) DRM_ERROR("failed to reserve abo before unpin\n"); diff --git a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c index b51f05dc9582c..dbf2ccd0c7447 100644 --- a/drivers/gpu/drm/amd/amdgpu/dce_virtual.c +++ b/drivers/gpu/drm/amd/amdgpu/dce_virtual.c @@ -168,11 +168,9 @@ static void dce_virtual_crtc_disable(struct drm_crtc *crtc) dce_virtual_crtc_dpms(crtc, DRM_MODE_DPMS_OFF); if (crtc->primary->fb) { int r; - struct amdgpu_framebuffer *amdgpu_fb; struct amdgpu_bo *abo; - amdgpu_fb = to_amdgpu_framebuffer(crtc->primary->fb); - abo = gem_to_amdgpu_bo(amdgpu_fb->obj); + abo = gem_to_amdgpu_bo(crtc->primary->fb->obj[0]); r = amdgpu_bo_reserve(abo, true); if (unlikely(r)) DRM_ERROR("failed to reserve abo before unpin\n"); @@ -329,7 +327,7 @@ static int dce_virtual_get_modes(struct drm_connector *connector) return 0; } -static int dce_virtual_mode_valid(struct drm_connector *connector, +static enum drm_mode_status dce_virtual_mode_valid(struct drm_connector *connector, struct drm_display_mode *mode) { return MODE_OK; @@ -462,8 +460,9 @@ static int dce_virtual_hw_init(void *handle) break; case CHIP_CARRIZO: case CHIP_STONEY: - case CHIP_POLARIS11: case CHIP_POLARIS10: + case CHIP_POLARIS11: + case CHIP_VEGAM: dce_v11_0_disable_dce(adev); break; case CHIP_TOPAZ: @@ -474,6 +473,7 @@ static int dce_virtual_hw_init(void *handle) break; case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: break; default: DRM_ERROR("Virtual display unsupported ASIC type: 0x%X\n", adev->asic_type); diff --git a/drivers/gpu/drm/amd/amdgpu/df_v1_7.c b/drivers/gpu/drm/amd/amdgpu/df_v1_7.c new file mode 100644 index 0000000000000..9935371db7ceb --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/df_v1_7.c @@ -0,0 +1,120 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "df_v1_7.h" + +#include "df/df_1_7_default.h" +#include "df/df_1_7_offset.h" +#include "df/df_1_7_sh_mask.h" + +static u32 df_v1_7_channel_number[] = {1, 2, 0, 4, 0, 8, 0, 16, 2}; + +static void df_v1_7_init (struct amdgpu_device *adev) +{ +} + +static void df_v1_7_enable_broadcast_mode(struct amdgpu_device *adev, + bool enable) +{ + u32 tmp; + + if (enable) { + tmp = RREG32_SOC15(DF, 0, mmFabricConfigAccessControl); + tmp &= ~FabricConfigAccessControl__CfgRegInstAccEn_MASK; + WREG32_SOC15(DF, 0, mmFabricConfigAccessControl, tmp); + } else + WREG32_SOC15(DF, 0, mmFabricConfigAccessControl, + mmFabricConfigAccessControl_DEFAULT); +} + +static u32 df_v1_7_get_fb_channel_number(struct amdgpu_device *adev) +{ + u32 tmp; + + tmp = RREG32_SOC15(DF, 0, mmDF_CS_AON0_DramBaseAddress0); + tmp &= DF_CS_AON0_DramBaseAddress0__IntLvNumChan_MASK; + tmp >>= DF_CS_AON0_DramBaseAddress0__IntLvNumChan__SHIFT; + + return tmp; +} + +static u32 df_v1_7_get_hbm_channel_number(struct amdgpu_device *adev) +{ + int fb_channel_number; + + fb_channel_number = adev->df_funcs->get_fb_channel_number(adev); + + return df_v1_7_channel_number[fb_channel_number]; +} + +static void df_v1_7_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + u32 tmp; + + /* Put DF on broadcast mode */ + adev->df_funcs->enable_broadcast_mode(adev, true); + + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_DF_MGCG)) { + tmp = RREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater); + tmp &= ~DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK; + tmp |= DF_V1_7_MGCG_ENABLE_15_CYCLE_DELAY; + WREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater, tmp); + } else { + tmp = RREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater); + tmp &= ~DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK; + tmp |= DF_V1_7_MGCG_DISABLE; + WREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater, tmp); + } + + /* Exit boradcast mode */ + adev->df_funcs->enable_broadcast_mode(adev, false); +} + +static void df_v1_7_get_clockgating_state(struct amdgpu_device *adev, + u32 *flags) +{ + u32 tmp; + + /* AMD_CG_SUPPORT_DF_MGCG */ + tmp = RREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater); + if (tmp & DF_V1_7_MGCG_ENABLE_15_CYCLE_DELAY) + *flags |= AMD_CG_SUPPORT_DF_MGCG; +} + +static void df_v1_7_enable_ecc_force_par_wr_rmw(struct amdgpu_device *adev, + bool enable) +{ + WREG32_FIELD15(DF, 0, DF_CS_AON0_CoherentSlaveModeCtrlA0, + ForceParWrRMW, enable); +} + +const struct amdgpu_df_funcs df_v1_7_funcs = { + .init = df_v1_7_init, + .enable_broadcast_mode = df_v1_7_enable_broadcast_mode, + .get_fb_channel_number = df_v1_7_get_fb_channel_number, + .get_hbm_channel_number = df_v1_7_get_hbm_channel_number, + .update_medium_grain_clock_gating = df_v1_7_update_medium_grain_clock_gating, + .get_clockgating_state = df_v1_7_get_clockgating_state, + .enable_ecc_force_par_wr_rmw = df_v1_7_enable_ecc_force_par_wr_rmw, +}; diff --git a/drivers/gpu/drm/amd/powerplay/inc/pp_soc15.h b/drivers/gpu/drm/amd/amdgpu/df_v1_7.h index 214f370c5efde..74621104c4871 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/pp_soc15.h +++ b/drivers/gpu/drm/amd/amdgpu/df_v1_7.h @@ -1,5 +1,5 @@ /* - * Copyright 2016 Advanced Micro Devices, Inc. + * Copyright 2018 Advanced Micro Devices, Inc. * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), @@ -20,33 +20,21 @@ * OTHER DEALINGS IN THE SOFTWARE. * */ -#ifndef PP_SOC15_H -#define PP_SOC15_H -#include "soc15_hw_ip.h" -#include "vega10_ip_offset.h" +#ifndef __DF_V1_7_H__ +#define __DF_V1_7_H__ -inline static uint32_t soc15_get_register_offset( - uint32_t hw_id, - uint32_t inst, - uint32_t segment, - uint32_t offset) +#include "soc15_common.h" +enum DF_V1_7_MGCG { - uint32_t reg = 0; + DF_V1_7_MGCG_DISABLE = 0, + DF_V1_7_MGCG_ENABLE_00_CYCLE_DELAY =1, + DF_V1_7_MGCG_ENABLE_01_CYCLE_DELAY =2, + DF_V1_7_MGCG_ENABLE_15_CYCLE_DELAY =13, + DF_V1_7_MGCG_ENABLE_31_CYCLE_DELAY =14, + DF_V1_7_MGCG_ENABLE_63_CYCLE_DELAY =15 +}; - if (hw_id == THM_HWID) - reg = THM_BASE.instance[inst].segment[segment] + offset; - else if (hw_id == NBIF_HWID) - reg = NBIF_BASE.instance[inst].segment[segment] + offset; - else if (hw_id == MP1_HWID) - reg = MP1_BASE.instance[inst].segment[segment] + offset; - else if (hw_id == DF_HWID) - reg = DF_BASE.instance[inst].segment[segment] + offset; - else if (hw_id == GC_HWID) - reg = GC_BASE.instance[inst].segment[segment] + offset; - else if (hw_id == SMUIO_HWID) - reg = SMUIO_BASE.instance[inst].segment[segment] + offset; - return reg; -} +extern const struct amdgpu_df_funcs df_v1_7_funcs; #endif diff --git a/drivers/gpu/drm/amd/amdgpu/df_v3_6.c b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c new file mode 100644 index 0000000000000..60608b3df8812 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/df_v3_6.c @@ -0,0 +1,116 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "df_v3_6.h" + +#include "df/df_3_6_default.h" +#include "df/df_3_6_offset.h" +#include "df/df_3_6_sh_mask.h" + +static u32 df_v3_6_channel_number[] = {1, 2, 0, 4, 0, 8, 0, + 16, 32, 0, 0, 0, 2, 4, 8}; + +static void df_v3_6_init(struct amdgpu_device *adev) +{ +} + +static void df_v3_6_enable_broadcast_mode(struct amdgpu_device *adev, + bool enable) +{ + u32 tmp; + + if (enable) { + tmp = RREG32_SOC15(DF, 0, mmFabricConfigAccessControl); + tmp &= ~FabricConfigAccessControl__CfgRegInstAccEn_MASK; + WREG32_SOC15(DF, 0, mmFabricConfigAccessControl, tmp); + } else + WREG32_SOC15(DF, 0, mmFabricConfigAccessControl, + mmFabricConfigAccessControl_DEFAULT); +} + +static u32 df_v3_6_get_fb_channel_number(struct amdgpu_device *adev) +{ + u32 tmp; + + tmp = RREG32_SOC15(DF, 0, mmDF_CS_UMC_AON0_DramBaseAddress0); + tmp &= DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan_MASK; + tmp >>= DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan__SHIFT; + + return tmp; +} + +static u32 df_v3_6_get_hbm_channel_number(struct amdgpu_device *adev) +{ + int fb_channel_number; + + fb_channel_number = adev->df_funcs->get_fb_channel_number(adev); + if (fb_channel_number > ARRAY_SIZE(df_v3_6_channel_number)) + fb_channel_number = 0; + + return df_v3_6_channel_number[fb_channel_number]; +} + +static void df_v3_6_update_medium_grain_clock_gating(struct amdgpu_device *adev, + bool enable) +{ + u32 tmp; + + /* Put DF on broadcast mode */ + adev->df_funcs->enable_broadcast_mode(adev, true); + + if (enable && (adev->cg_flags & AMD_CG_SUPPORT_DF_MGCG)) { + tmp = RREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater); + tmp &= ~DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK; + tmp |= DF_V3_6_MGCG_ENABLE_15_CYCLE_DELAY; + WREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater, tmp); + } else { + tmp = RREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater); + tmp &= ~DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK; + tmp |= DF_V3_6_MGCG_DISABLE; + WREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater, tmp); + } + + /* Exit broadcast mode */ + adev->df_funcs->enable_broadcast_mode(adev, false); +} + +static void df_v3_6_get_clockgating_state(struct amdgpu_device *adev, + u32 *flags) +{ + u32 tmp; + + /* AMD_CG_SUPPORT_DF_MGCG */ + tmp = RREG32_SOC15(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater); + if (tmp & DF_V3_6_MGCG_ENABLE_15_CYCLE_DELAY) + *flags |= AMD_CG_SUPPORT_DF_MGCG; +} + +const struct amdgpu_df_funcs df_v3_6_funcs = { + .init = df_v3_6_init, + .enable_broadcast_mode = df_v3_6_enable_broadcast_mode, + .get_fb_channel_number = df_v3_6_get_fb_channel_number, + .get_hbm_channel_number = df_v3_6_get_hbm_channel_number, + .update_medium_grain_clock_gating = + df_v3_6_update_medium_grain_clock_gating, + .get_clockgating_state = df_v3_6_get_clockgating_state, +}; diff --git a/drivers/gpu/drm/amd/amdgpu/df_v3_6.h b/drivers/gpu/drm/amd/amdgpu/df_v3_6.h new file mode 100644 index 0000000000000..e79c58e5efcb7 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/df_v3_6.h @@ -0,0 +1,40 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef __DF_V3_6_H__ +#define __DF_V3_6_H__ + +#include "soc15_common.h" + +enum DF_V3_6_MGCG { + DF_V3_6_MGCG_DISABLE = 0, + DF_V3_6_MGCG_ENABLE_00_CYCLE_DELAY = 1, + DF_V3_6_MGCG_ENABLE_01_CYCLE_DELAY = 2, + DF_V3_6_MGCG_ENABLE_15_CYCLE_DELAY = 13, + DF_V3_6_MGCG_ENABLE_31_CYCLE_DELAY = 14, + DF_V3_6_MGCG_ENABLE_63_CYCLE_DELAY = 15 +}; + +extern const struct amdgpu_df_funcs df_v3_6_funcs; + +#endif diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index e14263fca1c91..818874b13c99d 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -125,18 +125,6 @@ MODULE_FIRMWARE("amdgpu/fiji_mec.bin"); MODULE_FIRMWARE("amdgpu/fiji_mec2.bin"); MODULE_FIRMWARE("amdgpu/fiji_rlc.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_ce.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_ce_2.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_pfp.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_pfp_2.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_me.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_me_2.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_mec.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_mec_2.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_mec2.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_mec2_2.bin"); -MODULE_FIRMWARE("amdgpu/polaris11_rlc.bin"); - MODULE_FIRMWARE("amdgpu/polaris10_ce.bin"); MODULE_FIRMWARE("amdgpu/polaris10_ce_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_pfp.bin"); @@ -149,6 +137,18 @@ MODULE_FIRMWARE("amdgpu/polaris10_mec2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_mec2_2.bin"); MODULE_FIRMWARE("amdgpu/polaris10_rlc.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_ce.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_ce_2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_pfp.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_pfp_2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_me.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_me_2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec_2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_mec2_2.bin"); +MODULE_FIRMWARE("amdgpu/polaris11_rlc.bin"); + MODULE_FIRMWARE("amdgpu/polaris12_ce.bin"); MODULE_FIRMWARE("amdgpu/polaris12_ce_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_pfp.bin"); @@ -161,6 +161,13 @@ MODULE_FIRMWARE("amdgpu/polaris12_mec2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_mec2_2.bin"); MODULE_FIRMWARE("amdgpu/polaris12_rlc.bin"); +MODULE_FIRMWARE("amdgpu/vegam_ce.bin"); +MODULE_FIRMWARE("amdgpu/vegam_pfp.bin"); +MODULE_FIRMWARE("amdgpu/vegam_me.bin"); +MODULE_FIRMWARE("amdgpu/vegam_mec.bin"); +MODULE_FIRMWARE("amdgpu/vegam_mec2.bin"); +MODULE_FIRMWARE("amdgpu/vegam_rlc.bin"); + static const struct amdgpu_gds_reg_offset amdgpu_gds_reg_offset[] = { {mmGDS_VMID0_BASE, mmGDS_VMID0_SIZE, mmGDS_GWS_VMID0, mmGDS_OA_VMID0}, @@ -292,6 +299,37 @@ static const u32 tonga_mgcg_cgcg_init[] = mmCP_MEM_SLP_CNTL, 0x00000001, 0x00000001, }; +static const u32 golden_settings_vegam_a11[] = +{ + mmCB_HW_CONTROL, 0x0001f3cf, 0x00007208, + mmCB_HW_CONTROL_2, 0x0f000000, 0x0d000000, + mmCB_HW_CONTROL_3, 0x000001ff, 0x00000040, + mmDB_DEBUG2, 0xf00fffff, 0x00000400, + mmPA_SC_ENHANCE, 0xffffffff, 0x20000001, + mmPA_SC_LINE_STIPPLE_STATE, 0x0000ff0f, 0x00000000, + mmPA_SC_RASTER_CONFIG, 0x3f3fffff, 0x3a00161a, + mmPA_SC_RASTER_CONFIG_1, 0x0000003f, 0x0000002e, + mmRLC_CGCG_CGLS_CTRL, 0x00000003, 0x0001003c, + mmRLC_CGCG_CGLS_CTRL_3D, 0xffffffff, 0x0001003c, + mmSQ_CONFIG, 0x07f80000, 0x01180000, + mmTA_CNTL_AUX, 0x000f000f, 0x000b0000, + mmTCC_CTRL, 0x00100000, 0xf31fff7f, + mmTCP_ADDR_CONFIG, 0x000003ff, 0x000000f7, + mmTCP_CHAN_STEER_HI, 0xffffffff, 0x00000000, + mmTCP_CHAN_STEER_LO, 0xffffffff, 0x32761054, + mmVGT_RESET_DEBUG, 0x00000004, 0x00000004, +}; + +static const u32 vegam_golden_common_all[] = +{ + mmGRBM_GFX_INDEX, 0xffffffff, 0xe0000000, + mmGB_ADDR_CONFIG, 0xffffffff, 0x22011003, + mmSPI_RESOURCE_RESERVE_CU_0, 0xffffffff, 0x00000800, + mmSPI_RESOURCE_RESERVE_CU_1, 0xffffffff, 0x00000800, + mmSPI_RESOURCE_RESERVE_EN_CU_0, 0xffffffff, 0x00FF7FBF, + mmSPI_RESOURCE_RESERVE_EN_CU_1, 0xffffffff, 0x00FF7FAF, +}; + static const u32 golden_settings_polaris11_a11[] = { mmCB_HW_CONTROL, 0x0000f3cf, 0x00007208, @@ -712,6 +750,14 @@ static void gfx_v8_0_init_golden_registers(struct amdgpu_device *adev) tonga_golden_common_all, ARRAY_SIZE(tonga_golden_common_all)); break; + case CHIP_VEGAM: + amdgpu_device_program_register_sequence(adev, + golden_settings_vegam_a11, + ARRAY_SIZE(golden_settings_vegam_a11)); + amdgpu_device_program_register_sequence(adev, + vegam_golden_common_all, + ARRAY_SIZE(vegam_golden_common_all)); + break; case CHIP_POLARIS11: case CHIP_POLARIS12: amdgpu_device_program_register_sequence(adev, @@ -918,17 +964,20 @@ static int gfx_v8_0_init_microcode(struct amdgpu_device *adev) case CHIP_FIJI: chip_name = "fiji"; break; - case CHIP_POLARIS11: - chip_name = "polaris11"; + case CHIP_STONEY: + chip_name = "stoney"; break; case CHIP_POLARIS10: chip_name = "polaris10"; break; + case CHIP_POLARIS11: + chip_name = "polaris11"; + break; case CHIP_POLARIS12: chip_name = "polaris12"; break; - case CHIP_STONEY: - chip_name = "stoney"; + case CHIP_VEGAM: + chip_name = "vegam"; break; default: BUG(); @@ -1770,6 +1819,7 @@ static int gfx_v8_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config = POLARIS11_GB_ADDR_CONFIG_GOLDEN; break; case CHIP_POLARIS10: + case CHIP_VEGAM: ret = amdgpu_atombios_get_gfx_info(adev); if (ret) return ret; @@ -1957,12 +2007,13 @@ static int gfx_v8_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; switch (adev->asic_type) { - case CHIP_FIJI: case CHIP_TONGA: + case CHIP_CARRIZO: + case CHIP_FIJI: + case CHIP_POLARIS10: case CHIP_POLARIS11: case CHIP_POLARIS12: - case CHIP_POLARIS10: - case CHIP_CARRIZO: + case CHIP_VEGAM: adev->gfx.mec.num_mec = 2; break; case CHIP_TOPAZ: @@ -2323,6 +2374,7 @@ static void gfx_v8_0_tiling_mode_table_init(struct amdgpu_device *adev) break; case CHIP_FIJI: + case CHIP_VEGAM: modearray[0] = (ARRAY_MODE(ARRAY_2D_TILED_THIN1) | PIPE_CONFIG(ADDR_SURF_P16_32x32_16x16) | TILE_SPLIT(ADDR_SURF_TILE_SPLIT_64B) | @@ -3504,6 +3556,7 @@ gfx_v8_0_raster_config(struct amdgpu_device *adev, u32 *rconf, u32 *rconf1) { switch (adev->asic_type) { case CHIP_FIJI: + case CHIP_VEGAM: *rconf |= RB_MAP_PKR0(2) | RB_MAP_PKR1(2) | RB_XSEL2(1) | PKR_MAP(2) | PKR_XSEL(1) | PKR_YSEL(1) | @@ -4071,7 +4124,8 @@ static void gfx_v8_0_init_pg(struct amdgpu_device *adev) gfx_v8_0_init_power_gating(adev); WREG32(mmRLC_PG_ALWAYS_ON_CU_MASK, adev->gfx.cu_info.ao_cu_mask); } else if ((adev->asic_type == CHIP_POLARIS11) || - (adev->asic_type == CHIP_POLARIS12)) { + (adev->asic_type == CHIP_POLARIS12) || + (adev->asic_type == CHIP_VEGAM)) { gfx_v8_0_init_csb(adev); gfx_v8_0_init_save_restore_list(adev); gfx_v8_0_enable_save_restore_machine(adev); @@ -4146,7 +4200,8 @@ static int gfx_v8_0_rlc_resume(struct amdgpu_device *adev) WREG32(mmRLC_CGCG_CGLS_CTRL, tmp); if (adev->asic_type == CHIP_POLARIS11 || adev->asic_type == CHIP_POLARIS10 || - adev->asic_type == CHIP_POLARIS12) { + adev->asic_type == CHIP_POLARIS12 || + adev->asic_type == CHIP_VEGAM) { tmp = RREG32(mmRLC_CGCG_CGLS_CTRL_3D); tmp &= ~0x3; WREG32(mmRLC_CGCG_CGLS_CTRL_3D, tmp); @@ -5498,7 +5553,8 @@ static void gfx_v8_0_enable_gfx_static_mg_power_gating(struct amdgpu_device *ade bool enable) { if ((adev->asic_type == CHIP_POLARIS11) || - (adev->asic_type == CHIP_POLARIS12)) + (adev->asic_type == CHIP_POLARIS12) || + (adev->asic_type == CHIP_VEGAM)) /* Send msg to SMU via Powerplay */ amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_SMC, @@ -5588,6 +5644,7 @@ static int gfx_v8_0_set_powergating_state(void *handle, break; case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_SMG) && enable) gfx_v8_0_enable_gfx_static_mg_power_gating(adev, true); else @@ -6154,6 +6211,7 @@ static int gfx_v8_0_set_clockgating_state(void *handle, case CHIP_POLARIS10: case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: gfx_v8_0_polaris_update_gfx_clock_gating(adev, state); break; default: diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c index 9d39fd5b1822d..d7530fdfaad51 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v9_0.c @@ -27,6 +27,7 @@ #include "amdgpu_gfx.h" #include "soc15.h" #include "soc15d.h" +#include "amdgpu_atomfirmware.h" #include "gc/gc_9_0_offset.h" #include "gc/gc_9_0_sh_mask.h" @@ -41,7 +42,6 @@ #define GFX9_MEC_HPD_SIZE 2048 #define RLCG_UCODE_LOADING_START_ADDRESS 0x00002000L #define RLC_SAVE_RESTORE_ADDR_STARTING_OFFSET 0x00000000L -#define GFX9_RLC_FORMAT_DIRECT_REG_LIST_LENGTH 34 #define mmPWR_MISC_CNTL_STATUS 0x0183 #define mmPWR_MISC_CNTL_STATUS_BASE_IDX 0 @@ -64,6 +64,13 @@ MODULE_FIRMWARE("amdgpu/vega12_mec.bin"); MODULE_FIRMWARE("amdgpu/vega12_mec2.bin"); MODULE_FIRMWARE("amdgpu/vega12_rlc.bin"); +MODULE_FIRMWARE("amdgpu/vega20_ce.bin"); +MODULE_FIRMWARE("amdgpu/vega20_pfp.bin"); +MODULE_FIRMWARE("amdgpu/vega20_me.bin"); +MODULE_FIRMWARE("amdgpu/vega20_mec.bin"); +MODULE_FIRMWARE("amdgpu/vega20_mec2.bin"); +MODULE_FIRMWARE("amdgpu/vega20_rlc.bin"); + MODULE_FIRMWARE("amdgpu/raven_ce.bin"); MODULE_FIRMWARE("amdgpu/raven_pfp.bin"); MODULE_FIRMWARE("amdgpu/raven_me.bin"); @@ -73,29 +80,22 @@ MODULE_FIRMWARE("amdgpu/raven_rlc.bin"); static const struct soc15_reg_golden golden_settings_gc_9_0[] = { - SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPC_UTCL1_CNTL, 0x08000000, 0x08000080), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPF_UTCL1_CNTL, 0x08000000, 0x08000080), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmCPG_UTCL1_CNTL, 0x08000000, 0x08000080), SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG2, 0xf00fffff, 0x00000420), SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_GPU_ID, 0x0000000f, 0x00000000), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmIA_UTCL1_CNTL, 0x08000000, 0x08000080), SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_BINNER_EVENT_CNTL_3, 0x00000003, 0x82400024), SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE, 0x3fffffff, 0x00000001), SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_LINE_STIPPLE_STATE, 0x0000ff0f, 0x00000000), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_GPM_UTCL1_CNTL_0, 0x08000000, 0x08000080), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_GPM_UTCL1_CNTL_1, 0x08000000, 0x08000080), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_GPM_UTCL1_CNTL_2, 0x08000000, 0x08000080), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_PREWALKER_UTCL1_CNTL, 0x08000000, 0x08000080), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmRLC_SPM_UTCL1_CNTL, 0x08000000, 0x08000080), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSH_MEM_CONFIG, 0x00001000, 0x00001000), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_CONFIG_CNTL_1, 0x0000000f, 0x01000107), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_RESOURCE_RESERVE_CU_0, 0x0007ffff, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_RESOURCE_RESERVE_CU_1, 0x0007ffff, 0x00000800), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_RESOURCE_RESERVE_EN_CU_0, 0x01ffffff, 0x0000ff87), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_RESOURCE_RESERVE_EN_CU_1, 0x01ffffff, 0x0000ff8f), SOC15_REG_GOLDEN_VALUE(GC, 0, mmSQC_CONFIG, 0x03000000, 0x020a2000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX, 0xfffffeef, 0x010b0000), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_HI, 0xffffffff, 0x4a2c0e68), SOC15_REG_GOLDEN_VALUE(GC, 0, mmTCP_CHAN_STEER_LO, 0xffffffff, 0xb5d3f197), SOC15_REG_GOLDEN_VALUE(GC, 0, mmVGT_CACHE_INVALIDATION, 0x3fff3af3, 0x19200000), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmVGT_GS_MAX_WAVE_ID, 0x00000fff, 0x000003ff), - SOC15_REG_GOLDEN_VALUE(GC, 0, mmWD_UTCL1_CNTL, 0x08000000, 0x08000080) + SOC15_REG_GOLDEN_VALUE(GC, 0, mmVGT_GS_MAX_WAVE_ID, 0x00000fff, 0x000003ff) }; static const struct soc15_reg_golden golden_settings_gc_9_0_vg10[] = @@ -109,6 +109,20 @@ static const struct soc15_reg_golden golden_settings_gc_9_0_vg10[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x00001800, 0x00000800) }; +static const struct soc15_reg_golden golden_settings_gc_9_0_vg20[] = +{ + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_2, 0x0f000000, 0x0a000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL_3, 0x30000000, 0x10000000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_ADDR_CONFIG, 0xf3e777ff, 0x22014042), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmGB_ADDR_CONFIG_READ, 0xf3e777ff, 0x22014042), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmDB_DEBUG2, 0x00003e00, 0x00000400), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmPA_SC_ENHANCE_1, 0xff840000, 0x04040000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmRMI_UTCL1_CNTL2, 0x00030000, 0x00030000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmSPI_CONFIG_CNTL_1, 0xffff010f, 0x01000107), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTA_CNTL_AUX, 0x000b0000, 0x000b0000), + SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x01000000, 0x01000000) +}; + static const struct soc15_reg_golden golden_settings_gc_9_1[] = { SOC15_REG_GOLDEN_VALUE(GC, 0, mmCB_HW_CONTROL, 0xfffdf3cf, 0x00014104), @@ -185,6 +199,30 @@ static const struct soc15_reg_golden golden_settings_gc_9_2_1_vg12[] = SOC15_REG_GOLDEN_VALUE(GC, 0, mmTD_CNTL, 0x01bd9f33, 0x01000000) }; +static const u32 GFX_RLC_SRM_INDEX_CNTL_ADDR_OFFSETS[] = +{ + mmRLC_SRM_INDEX_CNTL_ADDR_0 - mmRLC_SRM_INDEX_CNTL_ADDR_0, + mmRLC_SRM_INDEX_CNTL_ADDR_1 - mmRLC_SRM_INDEX_CNTL_ADDR_0, + mmRLC_SRM_INDEX_CNTL_ADDR_2 - mmRLC_SRM_INDEX_CNTL_ADDR_0, + mmRLC_SRM_INDEX_CNTL_ADDR_3 - mmRLC_SRM_INDEX_CNTL_ADDR_0, + mmRLC_SRM_INDEX_CNTL_ADDR_4 - mmRLC_SRM_INDEX_CNTL_ADDR_0, + mmRLC_SRM_INDEX_CNTL_ADDR_5 - mmRLC_SRM_INDEX_CNTL_ADDR_0, + mmRLC_SRM_INDEX_CNTL_ADDR_6 - mmRLC_SRM_INDEX_CNTL_ADDR_0, + mmRLC_SRM_INDEX_CNTL_ADDR_7 - mmRLC_SRM_INDEX_CNTL_ADDR_0, +}; + +static const u32 GFX_RLC_SRM_INDEX_CNTL_DATA_OFFSETS[] = +{ + mmRLC_SRM_INDEX_CNTL_DATA_0 - mmRLC_SRM_INDEX_CNTL_DATA_0, + mmRLC_SRM_INDEX_CNTL_DATA_1 - mmRLC_SRM_INDEX_CNTL_DATA_0, + mmRLC_SRM_INDEX_CNTL_DATA_2 - mmRLC_SRM_INDEX_CNTL_DATA_0, + mmRLC_SRM_INDEX_CNTL_DATA_3 - mmRLC_SRM_INDEX_CNTL_DATA_0, + mmRLC_SRM_INDEX_CNTL_DATA_4 - mmRLC_SRM_INDEX_CNTL_DATA_0, + mmRLC_SRM_INDEX_CNTL_DATA_5 - mmRLC_SRM_INDEX_CNTL_DATA_0, + mmRLC_SRM_INDEX_CNTL_DATA_6 - mmRLC_SRM_INDEX_CNTL_DATA_0, + mmRLC_SRM_INDEX_CNTL_DATA_7 - mmRLC_SRM_INDEX_CNTL_DATA_0, +}; + #define VEGA10_GB_ADDR_CONFIG_GOLDEN 0x2a114042 #define VEGA12_GB_ADDR_CONFIG_GOLDEN 0x24104041 #define RAVEN_GB_ADDR_CONFIG_GOLDEN 0x24000042 @@ -218,6 +256,14 @@ static void gfx_v9_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_gc_9_2_1_vg12, ARRAY_SIZE(golden_settings_gc_9_2_1_vg12)); break; + case CHIP_VEGA20: + soc15_program_register_sequence(adev, + golden_settings_gc_9_0, + ARRAY_SIZE(golden_settings_gc_9_0)); + soc15_program_register_sequence(adev, + golden_settings_gc_9_0_vg20, + ARRAY_SIZE(golden_settings_gc_9_0_vg20)); + break; case CHIP_RAVEN: soc15_program_register_sequence(adev, golden_settings_gc_9_1, @@ -401,6 +447,27 @@ static void gfx_v9_0_free_microcode(struct amdgpu_device *adev) kfree(adev->gfx.rlc.register_list_format); } +static void gfx_v9_0_init_rlc_ext_microcode(struct amdgpu_device *adev) +{ + const struct rlc_firmware_header_v2_1 *rlc_hdr; + + rlc_hdr = (const struct rlc_firmware_header_v2_1 *)adev->gfx.rlc_fw->data; + adev->gfx.rlc_srlc_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_ucode_ver); + adev->gfx.rlc_srlc_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_cntl_feature_ver); + adev->gfx.rlc.save_restore_list_cntl_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_cntl_size_bytes); + adev->gfx.rlc.save_restore_list_cntl = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_cntl_offset_bytes); + adev->gfx.rlc_srlg_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_ucode_ver); + adev->gfx.rlc_srlg_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_gpm_feature_ver); + adev->gfx.rlc.save_restore_list_gpm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_gpm_size_bytes); + adev->gfx.rlc.save_restore_list_gpm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_gpm_offset_bytes); + adev->gfx.rlc_srls_fw_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_ucode_ver); + adev->gfx.rlc_srls_feature_version = le32_to_cpu(rlc_hdr->save_restore_list_srm_feature_ver); + adev->gfx.rlc.save_restore_list_srm_size_bytes = le32_to_cpu(rlc_hdr->save_restore_list_srm_size_bytes); + adev->gfx.rlc.save_restore_list_srm = (u8 *)rlc_hdr + le32_to_cpu(rlc_hdr->save_restore_list_srm_offset_bytes); + adev->gfx.rlc.reg_list_format_direct_reg_list_length = + le32_to_cpu(rlc_hdr->reg_list_format_direct_reg_list_length); +} + static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) { const char *chip_name; @@ -412,6 +479,8 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) const struct rlc_firmware_header_v2_0 *rlc_hdr; unsigned int *tmp = NULL; unsigned int i = 0; + uint16_t version_major; + uint16_t version_minor; DRM_DEBUG("\n"); @@ -422,6 +491,9 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) case CHIP_VEGA12: chip_name = "vega12"; break; + case CHIP_VEGA20: + chip_name = "vega20"; + break; case CHIP_RAVEN: chip_name = "raven"; break; @@ -468,6 +540,12 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) goto out; err = amdgpu_ucode_validate(adev->gfx.rlc_fw); rlc_hdr = (const struct rlc_firmware_header_v2_0 *)adev->gfx.rlc_fw->data; + + version_major = le16_to_cpu(rlc_hdr->header.header_version_major); + version_minor = le16_to_cpu(rlc_hdr->header.header_version_minor); + if (version_major == 2 && version_minor == 1) + adev->gfx.rlc.is_rlc_v2_1 = true; + adev->gfx.rlc_fw_version = le32_to_cpu(rlc_hdr->header.ucode_version); adev->gfx.rlc_feature_version = le32_to_cpu(rlc_hdr->ucode_feature_version); adev->gfx.rlc.save_and_restore_offset = @@ -508,6 +586,9 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) for (i = 0 ; i < (rlc_hdr->reg_list_size_bytes >> 2); i++) adev->gfx.rlc.register_restore[i] = le32_to_cpu(tmp[i]); + if (adev->gfx.rlc.is_rlc_v2_1) + gfx_v9_0_init_rlc_ext_microcode(adev); + snprintf(fw_name, sizeof(fw_name), "amdgpu/%s_mec.bin", chip_name); err = request_firmware(&adev->gfx.mec_fw, fw_name, adev->dev); if (err) @@ -566,6 +647,26 @@ static int gfx_v9_0_init_microcode(struct amdgpu_device *adev) adev->firmware.fw_size += ALIGN(le32_to_cpu(header->ucode_size_bytes), PAGE_SIZE); + if (adev->gfx.rlc.is_rlc_v2_1) { + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.save_restore_list_cntl_size_bytes, PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.save_restore_list_gpm_size_bytes, PAGE_SIZE); + + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM]; + info->ucode_id = AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM; + info->fw = adev->gfx.rlc_fw; + adev->firmware.fw_size += + ALIGN(adev->gfx.rlc.save_restore_list_srm_size_bytes, PAGE_SIZE); + } + info = &adev->firmware.ucode[AMDGPU_UCODE_ID_CP_MEC1]; info->ucode_id = AMDGPU_UCODE_ID_CP_MEC1; info->fw = adev->gfx.mec_fw; @@ -1013,9 +1114,10 @@ static const struct amdgpu_gfx_funcs gfx_v9_0_gfx_funcs = { .select_me_pipe_q = &gfx_v9_0_select_me_pipe_q }; -static void gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) +static int gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) { u32 gb_addr_config; + int err; adev->gfx.funcs = &gfx_v9_0_gfx_funcs; @@ -1037,6 +1139,20 @@ static void gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) gb_addr_config = VEGA12_GB_ADDR_CONFIG_GOLDEN; DRM_INFO("fix gfx.config for vega12\n"); break; + case CHIP_VEGA20: + adev->gfx.config.max_hw_contexts = 8; + adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; + adev->gfx.config.sc_prim_fifo_size_backend = 0x100; + adev->gfx.config.sc_hiz_tile_fifo_size = 0x30; + adev->gfx.config.sc_earlyz_tile_fifo_size = 0x4C0; + gb_addr_config = RREG32_SOC15(GC, 0, mmGB_ADDR_CONFIG); + gb_addr_config &= ~0xf3e777ff; + gb_addr_config |= 0x22014042; + /* check vbios table if gpu info is not available */ + err = amdgpu_atomfirmware_get_gfx_info(adev); + if (err) + return err; + break; case CHIP_RAVEN: adev->gfx.config.max_hw_contexts = 8; adev->gfx.config.sc_prim_fifo_size_frontend = 0x20; @@ -1086,6 +1202,8 @@ static void gfx_v9_0_gpu_early_init(struct amdgpu_device *adev) adev->gfx.config.gb_addr_config, GB_ADDR_CONFIG, PIPE_INTERLEAVE_SIZE)); + + return 0; } static int gfx_v9_0_ngg_create_buf(struct amdgpu_device *adev, @@ -1319,6 +1437,7 @@ static int gfx_v9_0_sw_init(void *handle) switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: case CHIP_RAVEN: adev->gfx.mec.num_mec = 2; break; @@ -1446,7 +1565,9 @@ static int gfx_v9_0_sw_init(void *handle) adev->gfx.ce_ram_size = 0x8000; - gfx_v9_0_gpu_early_init(adev); + r = gfx_v9_0_gpu_early_init(adev); + if (r) + return r; r = gfx_v9_0_ngg_init(adev); if (r) @@ -1600,6 +1721,7 @@ static void gfx_v9_0_gpu_init(struct amdgpu_device *adev) gfx_v9_0_setup_rb(adev); gfx_v9_0_get_cu_info(adev, &adev->gfx.cu_info); + adev->gfx.config.db_debug2 = RREG32_SOC15(GC, 0, mmDB_DEBUG2); /* XXX SH_MEM regs */ /* where to put LDS, scratch, GPUVM in FSA64 space */ @@ -1616,7 +1738,10 @@ static void gfx_v9_0_gpu_init(struct amdgpu_device *adev) tmp = REG_SET_FIELD(0, SH_MEM_CONFIG, ALIGNMENT_MODE, SH_MEM_ALIGNMENT_MODE_UNALIGNED); WREG32_SOC15(GC, 0, mmSH_MEM_CONFIG, tmp); - tmp = adev->gmc.shared_aperture_start >> 48; + tmp = REG_SET_FIELD(0, SH_MEM_BASES, PRIVATE_BASE, + (adev->gmc.private_aperture_start >> 48)); + tmp = REG_SET_FIELD(tmp, SH_MEM_BASES, SHARED_BASE, + (adev->gmc.shared_aperture_start >> 48)); WREG32_SOC15(GC, 0, mmSH_MEM_BASES, tmp); } } @@ -1708,55 +1833,42 @@ static void gfx_v9_0_init_csb(struct amdgpu_device *adev) adev->gfx.rlc.clear_state_size); } -static void gfx_v9_0_parse_ind_reg_list(int *register_list_format, +static void gfx_v9_1_parse_ind_reg_list(int *register_list_format, int indirect_offset, int list_size, int *unique_indirect_regs, int *unique_indirect_reg_count, - int max_indirect_reg_count, int *indirect_start_offsets, - int *indirect_start_offsets_count, - int max_indirect_start_offsets_count) + int *indirect_start_offsets_count) { int idx; - bool new_entry = true; for (; indirect_offset < list_size; indirect_offset++) { + indirect_start_offsets[*indirect_start_offsets_count] = indirect_offset; + *indirect_start_offsets_count = *indirect_start_offsets_count + 1; - if (new_entry) { - new_entry = false; - indirect_start_offsets[*indirect_start_offsets_count] = indirect_offset; - *indirect_start_offsets_count = *indirect_start_offsets_count + 1; - BUG_ON(*indirect_start_offsets_count >= max_indirect_start_offsets_count); - } + while (register_list_format[indirect_offset] != 0xFFFFFFFF) { + indirect_offset += 2; - if (register_list_format[indirect_offset] == 0xFFFFFFFF) { - new_entry = true; - continue; - } + /* look for the matching indice */ + for (idx = 0; idx < *unique_indirect_reg_count; idx++) { + if (unique_indirect_regs[idx] == + register_list_format[indirect_offset] || + !unique_indirect_regs[idx]) + break; + } - indirect_offset += 2; + BUG_ON(idx >= *unique_indirect_reg_count); - /* look for the matching indice */ - for (idx = 0; idx < *unique_indirect_reg_count; idx++) { - if (unique_indirect_regs[idx] == - register_list_format[indirect_offset]) - break; - } + if (!unique_indirect_regs[idx]) + unique_indirect_regs[idx] = register_list_format[indirect_offset]; - if (idx >= *unique_indirect_reg_count) { - unique_indirect_regs[*unique_indirect_reg_count] = - register_list_format[indirect_offset]; - idx = *unique_indirect_reg_count; - *unique_indirect_reg_count = *unique_indirect_reg_count + 1; - BUG_ON(*unique_indirect_reg_count >= max_indirect_reg_count); + indirect_offset++; } - - register_list_format[indirect_offset] = idx; } } -static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) +static int gfx_v9_1_init_rlc_save_restore_list(struct amdgpu_device *adev) { int unique_indirect_regs[] = {0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0, 0x0}; int unique_indirect_reg_count = 0; @@ -1765,7 +1877,7 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) int indirect_start_offsets_count = 0; int list_size = 0; - int i = 0; + int i = 0, j = 0; u32 tmp = 0; u32 *register_list_format = @@ -1776,15 +1888,14 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) adev->gfx.rlc.reg_list_format_size_bytes); /* setup unique_indirect_regs array and indirect_start_offsets array */ - gfx_v9_0_parse_ind_reg_list(register_list_format, - GFX9_RLC_FORMAT_DIRECT_REG_LIST_LENGTH, - adev->gfx.rlc.reg_list_format_size_bytes >> 2, - unique_indirect_regs, - &unique_indirect_reg_count, - ARRAY_SIZE(unique_indirect_regs), - indirect_start_offsets, - &indirect_start_offsets_count, - ARRAY_SIZE(indirect_start_offsets)); + unique_indirect_reg_count = ARRAY_SIZE(unique_indirect_regs); + gfx_v9_1_parse_ind_reg_list(register_list_format, + adev->gfx.rlc.reg_list_format_direct_reg_list_length, + adev->gfx.rlc.reg_list_format_size_bytes >> 2, + unique_indirect_regs, + &unique_indirect_reg_count, + indirect_start_offsets, + &indirect_start_offsets_count); /* enable auto inc in case it is disabled */ tmp = RREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_CNTL)); @@ -1798,19 +1909,37 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_ARAM_DATA), adev->gfx.rlc.register_restore[i]); - /* load direct register */ - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_ARAM_ADDR), 0); - for (i = 0; i < adev->gfx.rlc.reg_list_size_bytes >> 2; i++) - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_ARAM_DATA), - adev->gfx.rlc.register_restore[i]); - /* load indirect register */ WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_ADDR), adev->gfx.rlc.reg_list_format_start); - for (i = 0; i < adev->gfx.rlc.reg_list_format_size_bytes >> 2; i++) + + /* direct register portion */ + for (i = 0; i < adev->gfx.rlc.reg_list_format_direct_reg_list_length; i++) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_DATA), register_list_format[i]); + /* indirect register portion */ + while (i < (adev->gfx.rlc.reg_list_format_size_bytes >> 2)) { + if (register_list_format[i] == 0xFFFFFFFF) { + WREG32_SOC15(GC, 0, mmRLC_GPM_SCRATCH_DATA, register_list_format[i++]); + continue; + } + + WREG32_SOC15(GC, 0, mmRLC_GPM_SCRATCH_DATA, register_list_format[i++]); + WREG32_SOC15(GC, 0, mmRLC_GPM_SCRATCH_DATA, register_list_format[i++]); + + for (j = 0; j < unique_indirect_reg_count; j++) { + if (register_list_format[i] == unique_indirect_regs[j]) { + WREG32_SOC15(GC, 0, mmRLC_GPM_SCRATCH_DATA, j); + break; + } + } + + BUG_ON(j >= unique_indirect_reg_count); + + i++; + } + /* set save/restore list size */ list_size = adev->gfx.rlc.reg_list_size_bytes >> 2; list_size = list_size >> 1; @@ -1823,14 +1952,19 @@ static int gfx_v9_0_init_rlc_save_restore_list(struct amdgpu_device *adev) adev->gfx.rlc.starting_offsets_start); for (i = 0; i < ARRAY_SIZE(indirect_start_offsets); i++) WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_GPM_SCRATCH_DATA), - indirect_start_offsets[i]); + indirect_start_offsets[i]); /* load unique indirect regs*/ for (i = 0; i < ARRAY_SIZE(unique_indirect_regs); i++) { - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_ADDR_0) + i, - unique_indirect_regs[i] & 0x3FFFF); - WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_DATA_0) + i, - unique_indirect_regs[i] >> 20); + if (unique_indirect_regs[i] != 0) { + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_ADDR_0) + + GFX_RLC_SRM_INDEX_CNTL_ADDR_OFFSETS[i], + unique_indirect_regs[i] & 0x3FFFF); + + WREG32(SOC15_REG_OFFSET(GC, 0, mmRLC_SRM_INDEX_CNTL_DATA_0) + + GFX_RLC_SRM_INDEX_CNTL_DATA_OFFSETS[i], + unique_indirect_regs[i] >> 20); + } } kfree(register_list_format); @@ -2010,6 +2144,9 @@ static void gfx_v9_0_enable_gfx_dynamic_mg_power_gating(struct amdgpu_device *ad static void gfx_v9_0_init_pg(struct amdgpu_device *adev) { + if (!adev->gfx.rlc.is_rlc_v2_1) + return; + if (adev->pg_flags & (AMD_PG_SUPPORT_GFX_PG | AMD_PG_SUPPORT_GFX_SMG | AMD_PG_SUPPORT_GFX_DMG | @@ -2017,27 +2154,12 @@ static void gfx_v9_0_init_pg(struct amdgpu_device *adev) AMD_PG_SUPPORT_GDS | AMD_PG_SUPPORT_RLC_SMU_HS)) { gfx_v9_0_init_csb(adev); - gfx_v9_0_init_rlc_save_restore_list(adev); + gfx_v9_1_init_rlc_save_restore_list(adev); gfx_v9_0_enable_save_restore_machine(adev); - if (adev->asic_type == CHIP_RAVEN) { - WREG32(mmRLC_JUMP_TABLE_RESTORE, - adev->gfx.rlc.cp_table_gpu_addr >> 8); - gfx_v9_0_init_gfx_power_gating(adev); - - if (adev->pg_flags & AMD_PG_SUPPORT_RLC_SMU_HS) { - gfx_v9_0_enable_sck_slow_down_on_power_up(adev, true); - gfx_v9_0_enable_sck_slow_down_on_power_down(adev, true); - } else { - gfx_v9_0_enable_sck_slow_down_on_power_up(adev, false); - gfx_v9_0_enable_sck_slow_down_on_power_down(adev, false); - } - - if (adev->pg_flags & AMD_PG_SUPPORT_CP) - gfx_v9_0_enable_cp_power_gating(adev, true); - else - gfx_v9_0_enable_cp_power_gating(adev, false); - } + WREG32(mmRLC_JUMP_TABLE_RESTORE, + adev->gfx.rlc.cp_table_gpu_addr >> 8); + gfx_v9_0_init_gfx_power_gating(adev); } } @@ -3061,6 +3183,9 @@ static int gfx_v9_0_hw_fini(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; int i; + amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_GFX, + AMD_PG_STATE_UNGATE); + amdgpu_irq_put(adev, &adev->gfx.priv_reg_irq, 0); amdgpu_irq_put(adev, &adev->gfx.priv_inst_irq, 0); @@ -3279,6 +3404,11 @@ static int gfx_v9_0_late_init(void *handle) if (r) return r; + r = amdgpu_device_ip_set_powergating_state(adev, AMD_IP_BLOCK_TYPE_GFX, + AMD_PG_STATE_GATE); + if (r) + return r; + return 0; } @@ -3339,8 +3469,7 @@ static void gfx_v9_0_exit_rlc_safe_mode(struct amdgpu_device *adev) static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev, bool enable) { - /* TODO: double check if we need to perform under safe mdoe */ - /* gfx_v9_0_enter_rlc_safe_mode(adev); */ + gfx_v9_0_enter_rlc_safe_mode(adev); if ((adev->pg_flags & AMD_PG_SUPPORT_GFX_PG) && enable) { gfx_v9_0_enable_gfx_cg_power_gating(adev, true); @@ -3351,7 +3480,7 @@ static void gfx_v9_0_update_gfx_cg_power_gating(struct amdgpu_device *adev, gfx_v9_0_enable_gfx_pipeline_powergating(adev, false); } - /* gfx_v9_0_exit_rlc_safe_mode(adev); */ + gfx_v9_0_exit_rlc_safe_mode(adev); } static void gfx_v9_0_update_gfx_mg_power_gating(struct amdgpu_device *adev, @@ -3605,6 +3734,7 @@ static int gfx_v9_0_set_clockgating_state(void *handle, switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: case CHIP_RAVEN: gfx_v9_0_update_gfx_clock_gating(adev, state == AMD_CG_STATE_GATE ? true : false); @@ -3742,7 +3872,7 @@ static void gfx_v9_0_ring_emit_ib_gfx(struct amdgpu_ring *ring, } amdgpu_ring_write(ring, header); -BUG_ON(ib->gpu_addr & 0x3); /* Dword align */ + BUG_ON(ib->gpu_addr & 0x3); /* Dword align */ amdgpu_ring_write(ring, #ifdef __BIG_ENDIAN (2 << 0) | @@ -3774,13 +3904,16 @@ static void gfx_v9_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, { bool write64bit = flags & AMDGPU_FENCE_FLAG_64BIT; bool int_sel = flags & AMDGPU_FENCE_FLAG_INT; + bool writeback = flags & AMDGPU_FENCE_FLAG_TC_WB_ONLY; /* RELEASE_MEM - flush caches, send int */ amdgpu_ring_write(ring, PACKET3(PACKET3_RELEASE_MEM, 6)); - amdgpu_ring_write(ring, (EOP_TCL1_ACTION_EN | - EOP_TC_ACTION_EN | - EOP_TC_WB_ACTION_EN | - EOP_TC_MD_ACTION_EN | + amdgpu_ring_write(ring, ((writeback ? (EOP_TC_WB_ACTION_EN | + EOP_TC_NC_ACTION_EN) : + (EOP_TCL1_ACTION_EN | + EOP_TC_ACTION_EN | + EOP_TC_WB_ACTION_EN | + EOP_TC_MD_ACTION_EN)) | EVENT_TYPE(CACHE_FLUSH_AND_INV_TS_EVENT) | EVENT_INDEX(5))); amdgpu_ring_write(ring, DATA_SEL(write64bit ? 2 : 1) | INT_SEL(int_sel ? 2 : 0)); @@ -4137,6 +4270,20 @@ static void gfx_v9_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, gfx_v9_0_wait_reg_mem(ring, 0, 0, 0, reg, 0, val, mask, 0x20); } +static void gfx_v9_0_ring_emit_reg_write_reg_wait(struct amdgpu_ring *ring, + uint32_t reg0, uint32_t reg1, + uint32_t ref, uint32_t mask) +{ + int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX); + + if (amdgpu_sriov_vf(ring->adev)) + gfx_v9_0_wait_reg_mem(ring, usepfp, 0, 1, reg0, reg1, + ref, mask, 0x20); + else + amdgpu_ring_emit_reg_write_reg_wait_helper(ring, reg0, reg1, + ref, mask); +} + static void gfx_v9_0_set_gfx_eop_interrupt_state(struct amdgpu_device *adev, enum amdgpu_interrupt_state state) { @@ -4458,6 +4605,7 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_gfx = { .emit_tmz = gfx_v9_0_ring_emit_tmz, .emit_wreg = gfx_v9_0_ring_emit_wreg, .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, }; static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { @@ -4492,6 +4640,7 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_compute = { .set_priority = gfx_v9_0_ring_set_priority_compute, .emit_wreg = gfx_v9_0_ring_emit_wreg, .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, }; static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = { @@ -4522,6 +4671,7 @@ static const struct amdgpu_ring_funcs gfx_v9_0_ring_funcs_kiq = { .emit_rreg = gfx_v9_0_ring_emit_rreg, .emit_wreg = gfx_v9_0_ring_emit_wreg, .emit_reg_wait = gfx_v9_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = gfx_v9_0_ring_emit_reg_write_reg_wait, }; static void gfx_v9_0_set_ring_funcs(struct amdgpu_device *adev) @@ -4577,6 +4727,7 @@ static void gfx_v9_0_set_rlc_funcs(struct amdgpu_device *adev) switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: case CHIP_RAVEN: adev->gfx.rlc.funcs = &gfx_v9_0_rlc_funcs; break; @@ -4686,6 +4837,7 @@ static int gfx_v9_0_get_cu_info(struct amdgpu_device *adev, cu_info->number = active_cu_number; cu_info->ao_cu_mask = ao_cu_mask; + cu_info->simd_per_cu = NUM_SIMD_PER_CU; return 0; } diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c index 5617cf62c5668..79f9ac29019bd 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v6_0.c @@ -819,12 +819,33 @@ static int gmc_v6_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + amdgpu_bo_late_init(adev); + if (amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS) return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0); else return 0; } +static unsigned gmc_v6_0_get_vbios_fb_size(struct amdgpu_device *adev) +{ + u32 d1vga_control = RREG32(mmD1VGA_CONTROL); + unsigned size; + + if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) { + size = 9 * 1024 * 1024; /* reserve 8MB for vga emulator and 1 MB for FB */ + } else { + u32 viewport = RREG32(mmVIEWPORT_SIZE); + size = (REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_HEIGHT) * + REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_WIDTH) * + 4); + } + /* return 0 if the pre-OS buffer uses up most of vram */ + if ((adev->gmc.real_vram_size - size) < (8 * 1024 * 1024)) + return 0; + return size; +} + static int gmc_v6_0_sw_init(void *handle) { int r; @@ -851,8 +872,6 @@ static int gmc_v6_0_sw_init(void *handle) adev->gmc.mc_mask = 0xffffffffffULL; - adev->gmc.stolen_size = 256 * 1024; - adev->need_dma32 = false; dma_bits = adev->need_dma32 ? 32 : 40; r = pci_set_dma_mask(adev->pdev, DMA_BIT_MASK(dma_bits)); @@ -878,6 +897,8 @@ static int gmc_v6_0_sw_init(void *handle) if (r) return r; + adev->gmc.stolen_size = gmc_v6_0_get_vbios_fb_size(adev); + r = amdgpu_bo_init(adev); if (r) return r; diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c index 80054f36e4876..7147bfe25a234 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v7_0.c @@ -958,12 +958,33 @@ static int gmc_v7_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + amdgpu_bo_late_init(adev); + if (amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS) return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0); else return 0; } +static unsigned gmc_v7_0_get_vbios_fb_size(struct amdgpu_device *adev) +{ + u32 d1vga_control = RREG32(mmD1VGA_CONTROL); + unsigned size; + + if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) { + size = 9 * 1024 * 1024; /* reserve 8MB for vga emulator and 1 MB for FB */ + } else { + u32 viewport = RREG32(mmVIEWPORT_SIZE); + size = (REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_HEIGHT) * + REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_WIDTH) * + 4); + } + /* return 0 if the pre-OS buffer uses up most of vram */ + if ((adev->gmc.real_vram_size - size) < (8 * 1024 * 1024)) + return 0; + return size; +} + static int gmc_v7_0_sw_init(void *handle) { int r; @@ -998,8 +1019,6 @@ static int gmc_v7_0_sw_init(void *handle) */ adev->gmc.mc_mask = 0xffffffffffULL; /* 40 bit MC */ - adev->gmc.stolen_size = 256 * 1024; - /* set DMA mask + need_dma32 flags. * PCIE - can handle 40-bits. * IGP - can handle 40-bits @@ -1030,6 +1049,8 @@ static int gmc_v7_0_sw_init(void *handle) if (r) return r; + adev->gmc.stolen_size = gmc_v7_0_get_vbios_fb_size(adev); + /* Memory manager */ r = amdgpu_bo_init(adev); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c index d71d4cb68f9c0..1edbe6b477b56 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c @@ -138,6 +138,7 @@ static void gmc_v8_0_init_golden_registers(struct amdgpu_device *adev) break; case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: amdgpu_device_program_register_sequence(adev, golden_settings_polaris11_a11, ARRAY_SIZE(golden_settings_polaris11_a11)); @@ -231,6 +232,7 @@ static int gmc_v8_0_init_microcode(struct amdgpu_device *adev) case CHIP_FIJI: case CHIP_CARRIZO: case CHIP_STONEY: + case CHIP_VEGAM: return 0; default: BUG(); } @@ -567,9 +569,10 @@ static int gmc_v8_0_mc_init(struct amdgpu_device *adev) /* set the gart size */ if (amdgpu_gart_size == -1) { switch (adev->asic_type) { - case CHIP_POLARIS11: /* all engines support GPUVM */ case CHIP_POLARIS10: /* all engines support GPUVM */ + case CHIP_POLARIS11: /* all engines support GPUVM */ case CHIP_POLARIS12: /* all engines support GPUVM */ + case CHIP_VEGAM: /* all engines support GPUVM */ default: adev->gmc.gart_size = 256ULL << 20; break; @@ -1049,12 +1052,33 @@ static int gmc_v8_0_late_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + amdgpu_bo_late_init(adev); + if (amdgpu_vm_fault_stop != AMDGPU_VM_FAULT_STOP_ALWAYS) return amdgpu_irq_get(adev, &adev->gmc.vm_fault, 0); else return 0; } +static unsigned gmc_v8_0_get_vbios_fb_size(struct amdgpu_device *adev) +{ + u32 d1vga_control = RREG32(mmD1VGA_CONTROL); + unsigned size; + + if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) { + size = 9 * 1024 * 1024; /* reserve 8MB for vga emulator and 1 MB for FB */ + } else { + u32 viewport = RREG32(mmVIEWPORT_SIZE); + size = (REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_HEIGHT) * + REG_GET_FIELD(viewport, VIEWPORT_SIZE, VIEWPORT_WIDTH) * + 4); + } + /* return 0 if the pre-OS buffer uses up most of vram */ + if ((adev->gmc.real_vram_size - size) < (8 * 1024 * 1024)) + return 0; + return size; +} + #define mmMC_SEQ_MISC0_FIJI 0xA71 static int gmc_v8_0_sw_init(void *handle) @@ -1068,7 +1092,8 @@ static int gmc_v8_0_sw_init(void *handle) } else { u32 tmp; - if (adev->asic_type == CHIP_FIJI) + if ((adev->asic_type == CHIP_FIJI) || + (adev->asic_type == CHIP_VEGAM)) tmp = RREG32(mmMC_SEQ_MISC0_FIJI); else tmp = RREG32(mmMC_SEQ_MISC0); @@ -1096,8 +1121,6 @@ static int gmc_v8_0_sw_init(void *handle) */ adev->gmc.mc_mask = 0xffffffffffULL; /* 40 bit MC */ - adev->gmc.stolen_size = 256 * 1024; - /* set DMA mask + need_dma32 flags. * PCIE - can handle 40-bits. * IGP - can handle 40-bits @@ -1128,6 +1151,8 @@ static int gmc_v8_0_sw_init(void *handle) if (r) return r; + adev->gmc.stolen_size = gmc_v8_0_get_vbios_fb_size(adev); + /* Memory manager */ r = amdgpu_bo_init(adev); if (r) diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c index e687363900bb1..3c0a85d4e4ab9 100644 --- a/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c @@ -43,19 +43,13 @@ #include "gfxhub_v1_0.h" #include "mmhub_v1_0.h" -#define mmDF_CS_AON0_DramBaseAddress0 0x0044 -#define mmDF_CS_AON0_DramBaseAddress0_BASE_IDX 0 -//DF_CS_AON0_DramBaseAddress0 -#define DF_CS_AON0_DramBaseAddress0__AddrRngVal__SHIFT 0x0 -#define DF_CS_AON0_DramBaseAddress0__LgcyMmioHoleEn__SHIFT 0x1 -#define DF_CS_AON0_DramBaseAddress0__IntLvNumChan__SHIFT 0x4 -#define DF_CS_AON0_DramBaseAddress0__IntLvAddrSel__SHIFT 0x8 -#define DF_CS_AON0_DramBaseAddress0__DramBaseAddr__SHIFT 0xc -#define DF_CS_AON0_DramBaseAddress0__AddrRngVal_MASK 0x00000001L -#define DF_CS_AON0_DramBaseAddress0__LgcyMmioHoleEn_MASK 0x00000002L -#define DF_CS_AON0_DramBaseAddress0__IntLvNumChan_MASK 0x000000F0L -#define DF_CS_AON0_DramBaseAddress0__IntLvAddrSel_MASK 0x00000700L -#define DF_CS_AON0_DramBaseAddress0__DramBaseAddr_MASK 0xFFFFF000L +/* add these here since we already include dce12 headers and these are for DCN */ +#define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION 0x055d +#define mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION_BASE_IDX 2 +#define HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION__PRI_VIEWPORT_WIDTH__SHIFT 0x0 +#define HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION__PRI_VIEWPORT_HEIGHT__SHIFT 0x10 +#define HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION__PRI_VIEWPORT_WIDTH_MASK 0x00003FFFL +#define HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION__PRI_VIEWPORT_HEIGHT_MASK 0x3FFF0000L /* XXX Move this macro to VEGA10 header file, which is like vid.h for VI.*/ #define AMDGPU_NUM_OF_VMIDS 8 @@ -385,11 +379,9 @@ static uint64_t gmc_v9_0_emit_flush_gpu_tlb(struct amdgpu_ring *ring, amdgpu_ring_emit_wreg(ring, hub->ctx0_ptb_addr_hi32 + (2 * vmid), upper_32_bits(pd_addr)); - amdgpu_ring_emit_wreg(ring, hub->vm_inv_eng0_req + eng, req); - - /* wait for the invalidate to complete */ - amdgpu_ring_emit_reg_wait(ring, hub->vm_inv_eng0_ack + eng, - 1 << vmid, 1 << vmid); + amdgpu_ring_emit_reg_write_reg_wait(ring, hub->vm_inv_eng0_req + eng, + hub->vm_inv_eng0_ack + eng, + req, 1 << vmid); return pd_addr; } @@ -556,8 +548,7 @@ static int gmc_v9_0_early_init(void *handle) adev->gmc.shared_aperture_start = 0x2000000000000000ULL; adev->gmc.shared_aperture_end = adev->gmc.shared_aperture_start + (4ULL << 30) - 1; - adev->gmc.private_aperture_start = - adev->gmc.shared_aperture_end + 1; + adev->gmc.private_aperture_start = 0x1000000000000000ULL; adev->gmc.private_aperture_end = adev->gmc.private_aperture_start + (4ULL << 30) - 1; @@ -659,6 +650,11 @@ static int gmc_v9_0_late_init(void *handle) unsigned i; int r; + /* + * TODO - Uncomment once GART corruption issue is fixed. + */ + /* amdgpu_bo_late_init(adev); */ + for(i = 0; i < adev->num_rings; ++i) { struct amdgpu_ring *ring = adev->rings[i]; unsigned vmhub = ring->funcs->vmhub; @@ -679,6 +675,7 @@ static int gmc_v9_0_late_init(void *handle) DRM_INFO("ECC is active.\n"); } else if (r == 0) { DRM_INFO("ECC is not present.\n"); + adev->df_funcs->enable_ecc_force_par_wr_rmw(adev, false); } else { DRM_ERROR("gmc_v9_0_ecc_available() failed. r: %d\n", r); return r; @@ -697,10 +694,7 @@ static void gmc_v9_0_vram_gtt_location(struct amdgpu_device *adev, amdgpu_device_vram_location(adev, &adev->gmc, base); amdgpu_device_gart_location(adev, mc); /* base offset of vram pages */ - if (adev->flags & AMD_IS_APU) - adev->vm_manager.vram_base_offset = gfxhub_v1_0_get_mc_fb_offset(adev); - else - adev->vm_manager.vram_base_offset = 0; + adev->vm_manager.vram_base_offset = gfxhub_v1_0_get_mc_fb_offset(adev); } /** @@ -714,7 +708,6 @@ static void gmc_v9_0_vram_gtt_location(struct amdgpu_device *adev, */ static int gmc_v9_0_mc_init(struct amdgpu_device *adev) { - u32 tmp; int chansize, numchan; int r; @@ -727,39 +720,7 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev) else chansize = 128; - tmp = RREG32_SOC15(DF, 0, mmDF_CS_AON0_DramBaseAddress0); - tmp &= DF_CS_AON0_DramBaseAddress0__IntLvNumChan_MASK; - tmp >>= DF_CS_AON0_DramBaseAddress0__IntLvNumChan__SHIFT; - switch (tmp) { - case 0: - default: - numchan = 1; - break; - case 1: - numchan = 2; - break; - case 2: - numchan = 0; - break; - case 3: - numchan = 4; - break; - case 4: - numchan = 0; - break; - case 5: - numchan = 8; - break; - case 6: - numchan = 0; - break; - case 7: - numchan = 16; - break; - case 8: - numchan = 2; - break; - } + numchan = adev->df_funcs->get_hbm_channel_number(adev); adev->gmc.vram_width = numchan * chansize; } @@ -792,6 +753,7 @@ static int gmc_v9_0_mc_init(struct amdgpu_device *adev) switch (adev->asic_type) { case CHIP_VEGA10: /* all engines support GPUVM */ case CHIP_VEGA12: /* all engines support GPUVM */ + case CHIP_VEGA20: default: adev->gmc.gart_size = 512ULL << 20; break; @@ -826,6 +788,52 @@ static int gmc_v9_0_gart_init(struct amdgpu_device *adev) return amdgpu_gart_table_vram_alloc(adev); } +static unsigned gmc_v9_0_get_vbios_fb_size(struct amdgpu_device *adev) +{ +#if 0 + u32 d1vga_control = RREG32_SOC15(DCE, 0, mmD1VGA_CONTROL); +#endif + unsigned size; + + /* + * TODO Remove once GART corruption is resolved + * Check related code in gmc_v9_0_sw_fini + * */ + size = 9 * 1024 * 1024; + +#if 0 + if (REG_GET_FIELD(d1vga_control, D1VGA_CONTROL, D1VGA_MODE_ENABLE)) { + size = 9 * 1024 * 1024; /* reserve 8MB for vga emulator and 1 MB for FB */ + } else { + u32 viewport; + + switch (adev->asic_type) { + case CHIP_RAVEN: + viewport = RREG32_SOC15(DCE, 0, mmHUBP0_DCSURF_PRI_VIEWPORT_DIMENSION); + size = (REG_GET_FIELD(viewport, + HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION, PRI_VIEWPORT_HEIGHT) * + REG_GET_FIELD(viewport, + HUBP0_DCSURF_PRI_VIEWPORT_DIMENSION, PRI_VIEWPORT_WIDTH) * + 4); + break; + case CHIP_VEGA10: + case CHIP_VEGA12: + default: + viewport = RREG32_SOC15(DCE, 0, mmSCL0_VIEWPORT_SIZE); + size = (REG_GET_FIELD(viewport, SCL0_VIEWPORT_SIZE, VIEWPORT_HEIGHT) * + REG_GET_FIELD(viewport, SCL0_VIEWPORT_SIZE, VIEWPORT_WIDTH) * + 4); + break; + } + } + /* return 0 if the pre-OS buffer uses up most of vram */ + if ((adev->gmc.real_vram_size - size) < (8 * 1024 * 1024)) + return 0; + +#endif + return size; +} + static int gmc_v9_0_sw_init(void *handle) { int r; @@ -851,6 +859,7 @@ static int gmc_v9_0_sw_init(void *handle) break; case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: /* * To fulfill 4-level page support, * vm size is 256TB (48bit), maximum size of Vega10, @@ -877,12 +886,6 @@ static int gmc_v9_0_sw_init(void *handle) */ adev->gmc.mc_mask = 0xffffffffffffULL; /* 48 bit MC */ - /* - * It needs to reserve 8M stolen memory for vega10 - * TODO: Figure out how to avoid that... - */ - adev->gmc.stolen_size = 8 * 1024 * 1024; - /* set DMA mask + need_dma32 flags. * PCIE - can handle 44-bits. * IGP - can handle 44-bits @@ -907,6 +910,8 @@ static int gmc_v9_0_sw_init(void *handle) if (r) return r; + adev->gmc.stolen_size = gmc_v9_0_get_vbios_fb_size(adev); + /* Memory manager */ r = amdgpu_bo_init(adev); if (r) @@ -950,6 +955,18 @@ static int gmc_v9_0_sw_fini(void *handle) amdgpu_gem_force_release(adev); amdgpu_vm_manager_fini(adev); gmc_v9_0_gart_fini(adev); + + /* + * TODO: + * Currently there is a bug where some memory client outside + * of the driver writes to first 8M of VRAM on S3 resume, + * this overrides GART which by default gets placed in first 8M and + * causes VM_FAULTS once GTT is accessed. + * Keep the stolen memory reservation until the while this is not solved. + * Also check code in gmc_v9_0_get_vbios_fb_size and gmc_v9_0_late_init + */ + amdgpu_bo_free_kernel(&adev->stolen_vga_memory, NULL, NULL); + amdgpu_bo_fini(adev); return 0; @@ -960,6 +977,7 @@ static void gmc_v9_0_init_golden_registers(struct amdgpu_device *adev) switch (adev->asic_type) { case CHIP_VEGA10: + case CHIP_VEGA20: soc15_program_register_sequence(adev, golden_settings_mmhub_1_0_0, ARRAY_SIZE(golden_settings_mmhub_1_0_0)); diff --git a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c index 26ba984ab2b7c..17f7f074cedcc 100644 --- a/drivers/gpu/drm/amd/amdgpu/kv_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/kv_dpm.c @@ -2817,7 +2817,7 @@ static int kv_dpm_init(struct amdgpu_device *adev) pi->caps_tcp_ramping = true; } - if (amdgpu_pp_feature_mask & SCLK_DEEP_SLEEP_MASK) + if (adev->powerplay.pp_feature & PP_SCLK_DEEP_SLEEP_MASK) pi->caps_sclk_ds = true; else pi->caps_sclk_ds = false; @@ -2974,7 +2974,7 @@ static int kv_dpm_late_init(void *handle) /* powerdown unused blocks for now */ struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (!amdgpu_dpm) + if (!adev->pm.dpm_enabled) return 0; kv_dpm_powergate_acp(adev, true); diff --git a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c index 43f925773b574..3d53c4413f138 100644 --- a/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/mmhub_v1_0.c @@ -734,6 +734,7 @@ int mmhub_v1_0_set_clockgating(struct amdgpu_device *adev, switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: case CHIP_RAVEN: mmhub_v1_0_update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE ? true : false); diff --git a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c index 4933486724754..078f70faedcbb 100644 --- a/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c +++ b/drivers/gpu/drm/amd/amdgpu/mxgpu_ai.c @@ -260,8 +260,10 @@ static void xgpu_ai_mailbox_flr_work(struct work_struct *work) } while (timeout > 1); flr_done: - if (locked) + if (locked) { + adev->in_gpu_reset = 0; mutex_unlock(&adev->lock_reset); + } /* Trigger recovery for world switch failure if no TDR */ if (amdgpu_lockup_timeout == 0) diff --git a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c index df34dc79d444f..365517c0121e0 100644 --- a/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/nbio_v7_0.c @@ -34,10 +34,19 @@ #define smnCPM_CONTROL 0x11180460 #define smnPCIE_CNTL2 0x11180070 +/* vega20 */ +#define mmRCC_DEV0_EPF0_STRAP0_VG20 0x0011 +#define mmRCC_DEV0_EPF0_STRAP0_VG20_BASE_IDX 2 + static u32 nbio_v7_0_get_rev_id(struct amdgpu_device *adev) { u32 tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0); + if (adev->asic_type == CHIP_VEGA20) + tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0_VG20); + else + tmp = RREG32_SOC15(NBIO, 0, mmRCC_DEV0_EPF0_STRAP0); + tmp &= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0_MASK; tmp >>= RCC_DEV0_EPF0_STRAP0__STRAP_ATI_REV_ID_DEV0_F0__SHIFT; @@ -75,10 +84,14 @@ static void nbio_v7_0_sdma_doorbell_range(struct amdgpu_device *adev, int instan SOC15_REG_OFFSET(NBIO, 0, mmBIF_SDMA1_DOORBELL_RANGE); u32 doorbell_range = RREG32(reg); + u32 range = 2; + + if (adev->asic_type == CHIP_VEGA20) + range = 8; if (use_doorbell) { doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, OFFSET, doorbell_index); - doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 2); + doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, range); } else doorbell_range = REG_SET_FIELD(doorbell_range, BIF_SDMA0_DOORBELL_RANGE, SIZE, 0); @@ -133,6 +146,9 @@ static void nbio_v7_0_update_medium_grain_clock_gating(struct amdgpu_device *ade { uint32_t def, data; + if (adev->asic_type == CHIP_VEGA20) + return; + /* NBIF_MGCG_CTRL_LCLK */ def = data = RREG32_PCIE(smnNBIF_MGCG_CTRL_LCLK); diff --git a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h index 8da6da90b1c99..0cf48d26c6768 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h +++ b/drivers/gpu/drm/amd/amdgpu/psp_gfx_if.h @@ -40,11 +40,20 @@ enum psp_gfx_crtl_cmd_id GFX_CTRL_CMD_ID_INIT_GPCOM_RING = 0x00020000, /* initialize GPCOM ring */ GFX_CTRL_CMD_ID_DESTROY_RINGS = 0x00030000, /* destroy rings */ GFX_CTRL_CMD_ID_CAN_INIT_RINGS = 0x00040000, /* is it allowed to initialized the rings */ + GFX_CTRL_CMD_ID_ENABLE_INT = 0x00050000, /* enable PSP-to-Gfx interrupt */ + GFX_CTRL_CMD_ID_DISABLE_INT = 0x00060000, /* disable PSP-to-Gfx interrupt */ + GFX_CTRL_CMD_ID_MODE1_RST = 0x00070000, /* trigger the Mode 1 reset */ GFX_CTRL_CMD_ID_MAX = 0x000F0000, /* max command ID */ }; +/*----------------------------------------------------------------------------- + NOTE: All physical addresses used in this interface are actually + GPU Virtual Addresses. +*/ + + /* Control registers of the TEE Gfx interface. These are located in * SRBM-to-PSP mailbox registers (total 8 registers). */ @@ -55,8 +64,8 @@ struct psp_gfx_ctrl volatile uint32_t rbi_rptr; /* +8 Read pointer (index) of RBI ring */ volatile uint32_t gpcom_wptr; /* +12 Write pointer (index) of GPCOM ring */ volatile uint32_t gpcom_rptr; /* +16 Read pointer (index) of GPCOM ring */ - volatile uint32_t ring_addr_lo; /* +20 bits [31:0] of physical address of ring buffer */ - volatile uint32_t ring_addr_hi; /* +24 bits [63:32] of physical address of ring buffer */ + volatile uint32_t ring_addr_lo; /* +20 bits [31:0] of GPU Virtual of ring buffer (VMID=0)*/ + volatile uint32_t ring_addr_hi; /* +24 bits [63:32] of GPU Virtual of ring buffer (VMID=0) */ volatile uint32_t ring_buf_size; /* +28 Ring buffer size (in bytes) */ }; @@ -78,6 +87,8 @@ enum psp_gfx_cmd_id GFX_CMD_ID_LOAD_ASD = 0x00000004, /* load ASD Driver */ GFX_CMD_ID_SETUP_TMR = 0x00000005, /* setup TMR region */ GFX_CMD_ID_LOAD_IP_FW = 0x00000006, /* load HW IP FW */ + GFX_CMD_ID_DESTROY_TMR = 0x00000007, /* destroy TMR region */ + GFX_CMD_ID_SAVE_RESTORE = 0x00000008, /* save/restore HW IP FW */ }; @@ -85,11 +96,11 @@ enum psp_gfx_cmd_id /* Command to load Trusted Application binary into PSP OS. */ struct psp_gfx_cmd_load_ta { - uint32_t app_phy_addr_lo; /* bits [31:0] of the physical address of the TA binary (must be 4 KB aligned) */ - uint32_t app_phy_addr_hi; /* bits [63:32] of the physical address of the TA binary */ + uint32_t app_phy_addr_lo; /* bits [31:0] of the GPU Virtual address of the TA binary (must be 4 KB aligned) */ + uint32_t app_phy_addr_hi; /* bits [63:32] of the GPU Virtual address of the TA binary */ uint32_t app_len; /* length of the TA binary in bytes */ - uint32_t cmd_buf_phy_addr_lo; /* bits [31:0] of the physical address of CMD buffer (must be 4 KB aligned) */ - uint32_t cmd_buf_phy_addr_hi; /* bits [63:32] of the physical address of CMD buffer */ + uint32_t cmd_buf_phy_addr_lo; /* bits [31:0] of the GPU Virtual address of CMD buffer (must be 4 KB aligned) */ + uint32_t cmd_buf_phy_addr_hi; /* bits [63:32] of the GPU Virtual address of CMD buffer */ uint32_t cmd_buf_len; /* length of the CMD buffer in bytes; must be multiple of 4 KB */ /* Note: CmdBufLen can be set to 0. In this case no persistent CMD buffer is provided @@ -111,8 +122,8 @@ struct psp_gfx_cmd_unload_ta */ struct psp_gfx_buf_desc { - uint32_t buf_phy_addr_lo; /* bits [31:0] of physical address of the buffer (must be 4 KB aligned) */ - uint32_t buf_phy_addr_hi; /* bits [63:32] of physical address of the buffer */ + uint32_t buf_phy_addr_lo; /* bits [31:0] of GPU Virtual address of the buffer (must be 4 KB aligned) */ + uint32_t buf_phy_addr_hi; /* bits [63:32] of GPU Virtual address of the buffer */ uint32_t buf_size; /* buffer size in bytes (must be multiple of 4 KB and no bigger than 64 MB) */ }; @@ -145,8 +156,8 @@ struct psp_gfx_cmd_invoke_cmd /* Command to setup TMR region. */ struct psp_gfx_cmd_setup_tmr { - uint32_t buf_phy_addr_lo; /* bits [31:0] of physical address of TMR buffer (must be 4 KB aligned) */ - uint32_t buf_phy_addr_hi; /* bits [63:32] of physical address of TMR buffer */ + uint32_t buf_phy_addr_lo; /* bits [31:0] of GPU Virtual address of TMR buffer (must be 4 KB aligned) */ + uint32_t buf_phy_addr_hi; /* bits [63:32] of GPU Virtual address of TMR buffer */ uint32_t buf_size; /* buffer size in bytes (must be multiple of 4 KB) */ }; @@ -174,18 +185,32 @@ enum psp_gfx_fw_type GFX_FW_TYPE_ISP = 16, GFX_FW_TYPE_ACP = 17, GFX_FW_TYPE_SMU = 18, + GFX_FW_TYPE_MMSCH = 19, + GFX_FW_TYPE_RLC_RESTORE_LIST_GPM_MEM = 20, + GFX_FW_TYPE_RLC_RESTORE_LIST_SRM_MEM = 21, + GFX_FW_TYPE_RLC_RESTORE_LIST_CNTL = 22, + GFX_FW_TYPE_MAX = 23 }; /* Command to load HW IP FW. */ struct psp_gfx_cmd_load_ip_fw { - uint32_t fw_phy_addr_lo; /* bits [31:0] of physical address of FW location (must be 4 KB aligned) */ - uint32_t fw_phy_addr_hi; /* bits [63:32] of physical address of FW location */ + uint32_t fw_phy_addr_lo; /* bits [31:0] of GPU Virtual address of FW location (must be 4 KB aligned) */ + uint32_t fw_phy_addr_hi; /* bits [63:32] of GPU Virtual address of FW location */ uint32_t fw_size; /* FW buffer size in bytes */ enum psp_gfx_fw_type fw_type; /* FW type */ }; +/* Command to save/restore HW IP FW. */ +struct psp_gfx_cmd_save_restore_ip_fw +{ + uint32_t save_fw; /* if set, command is used for saving fw otherwise for resetoring*/ + uint32_t save_restore_addr_lo; /* bits [31:0] of FB address of GART memory used as save/restore buffer (must be 4 KB aligned) */ + uint32_t save_restore_addr_hi; /* bits [63:32] of FB address of GART memory used as save/restore buffer */ + uint32_t buf_size; /* Size of the save/restore buffer in bytes */ + enum psp_gfx_fw_type fw_type; /* FW type */ +}; /* All GFX ring buffer commands. */ union psp_gfx_commands @@ -195,7 +220,7 @@ union psp_gfx_commands struct psp_gfx_cmd_invoke_cmd cmd_invoke_cmd; struct psp_gfx_cmd_setup_tmr cmd_setup_tmr; struct psp_gfx_cmd_load_ip_fw cmd_load_ip_fw; - + struct psp_gfx_cmd_save_restore_ip_fw cmd_save_restore_ip_fw; }; @@ -226,8 +251,8 @@ struct psp_gfx_cmd_resp /* These fields are used for RBI only. They are all 0 in GPCOM commands */ - uint32_t resp_buf_addr_lo; /* +12 bits [31:0] of physical address of response buffer (must be 4 KB aligned) */ - uint32_t resp_buf_addr_hi; /* +16 bits [63:32] of physical address of response buffer */ + uint32_t resp_buf_addr_lo; /* +12 bits [31:0] of GPU Virtual address of response buffer (must be 4 KB aligned) */ + uint32_t resp_buf_addr_hi; /* +16 bits [63:32] of GPU Virtual address of response buffer */ uint32_t resp_offset; /* +20 offset within response buffer */ uint32_t resp_buf_size; /* +24 total size of the response buffer in bytes */ @@ -251,19 +276,19 @@ struct psp_gfx_cmd_resp /* Structure of the Ring Buffer Frame */ struct psp_gfx_rb_frame { - uint32_t cmd_buf_addr_lo; /* +0 bits [31:0] of physical address of command buffer (must be 4 KB aligned) */ - uint32_t cmd_buf_addr_hi; /* +4 bits [63:32] of physical address of command buffer */ + uint32_t cmd_buf_addr_lo; /* +0 bits [31:0] of GPU Virtual address of command buffer (must be 4 KB aligned) */ + uint32_t cmd_buf_addr_hi; /* +4 bits [63:32] of GPU Virtual address of command buffer */ uint32_t cmd_buf_size; /* +8 command buffer size in bytes */ - uint32_t fence_addr_lo; /* +12 bits [31:0] of physical address of Fence for this frame */ - uint32_t fence_addr_hi; /* +16 bits [63:32] of physical address of Fence for this frame */ + uint32_t fence_addr_lo; /* +12 bits [31:0] of GPU Virtual address of Fence for this frame */ + uint32_t fence_addr_hi; /* +16 bits [63:32] of GPU Virtual address of Fence for this frame */ uint32_t fence_value; /* +20 Fence value */ uint32_t sid_lo; /* +24 bits [31:0] of SID value (used only for RBI frames) */ uint32_t sid_hi; /* +28 bits [63:32] of SID value (used only for RBI frames) */ uint8_t vmid; /* +32 VMID value used for mapping of all addresses for this frame */ uint8_t frame_type; /* +33 1: destory context frame, 0: all other frames; used only for RBI frames */ uint8_t reserved1[2]; /* +34 reserved, must be 0 */ - uint32_t reserved2[7]; /* +40 reserved, must be 0 */ - /* total 64 bytes */ + uint32_t reserved2[7]; /* +36 reserved, must be 0 */ + /* total 64 bytes */ }; #endif /* _PSP_TEE_GFX_IF_H_ */ diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c index 8873d833a7f7d..0ff136d02d9b3 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v10_0.c @@ -70,6 +70,15 @@ psp_v10_0_get_fw_type(struct amdgpu_firmware_info *ucode, enum psp_gfx_fw_type * case AMDGPU_UCODE_ID_RLC_G: *type = GFX_FW_TYPE_RLC_G; break; + case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_CNTL: + *type = GFX_FW_TYPE_RLC_RESTORE_LIST_CNTL; + break; + case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_GPM_MEM: + *type = GFX_FW_TYPE_RLC_RESTORE_LIST_GPM_MEM; + break; + case AMDGPU_UCODE_ID_RLC_RESTORE_LIST_SRM_MEM: + *type = GFX_FW_TYPE_RLC_RESTORE_LIST_SRM_MEM; + break; case AMDGPU_UCODE_ID_SMC: *type = GFX_FW_TYPE_SMU; break; diff --git a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c index 196e75def1f24..0c768e388ace5 100644 --- a/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c +++ b/drivers/gpu/drm/amd/amdgpu/psp_v3_1.c @@ -41,6 +41,9 @@ MODULE_FIRMWARE("amdgpu/vega10_sos.bin"); MODULE_FIRMWARE("amdgpu/vega10_asd.bin"); MODULE_FIRMWARE("amdgpu/vega12_sos.bin"); MODULE_FIRMWARE("amdgpu/vega12_asd.bin"); +MODULE_FIRMWARE("amdgpu/vega20_sos.bin"); +MODULE_FIRMWARE("amdgpu/vega20_asd.bin"); + #define smnMP1_FIRMWARE_FLAGS 0x3010028 diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c index be20a387d9611..aa9ab299fd32a 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v3_0.c @@ -62,6 +62,8 @@ MODULE_FIRMWARE("amdgpu/polaris11_sdma.bin"); MODULE_FIRMWARE("amdgpu/polaris11_sdma1.bin"); MODULE_FIRMWARE("amdgpu/polaris12_sdma.bin"); MODULE_FIRMWARE("amdgpu/polaris12_sdma1.bin"); +MODULE_FIRMWARE("amdgpu/vegam_sdma.bin"); +MODULE_FIRMWARE("amdgpu/vegam_sdma1.bin"); static const u32 sdma_offsets[SDMA_MAX_INSTANCE] = @@ -209,6 +211,7 @@ static void sdma_v3_0_init_golden_registers(struct amdgpu_device *adev) break; case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: amdgpu_device_program_register_sequence(adev, golden_settings_polaris11_a11, ARRAY_SIZE(golden_settings_polaris11_a11)); @@ -275,15 +278,18 @@ static int sdma_v3_0_init_microcode(struct amdgpu_device *adev) case CHIP_FIJI: chip_name = "fiji"; break; - case CHIP_POLARIS11: - chip_name = "polaris11"; - break; case CHIP_POLARIS10: chip_name = "polaris10"; break; + case CHIP_POLARIS11: + chip_name = "polaris11"; + break; case CHIP_POLARIS12: chip_name = "polaris12"; break; + case CHIP_VEGAM: + chip_name = "vegam"; + break; case CHIP_CARRIZO: chip_name = "carrizo"; break; diff --git a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c index 399f876f9cad3..ca53b3fba422d 100644 --- a/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/sdma_v4_0.c @@ -42,6 +42,8 @@ MODULE_FIRMWARE("amdgpu/vega10_sdma.bin"); MODULE_FIRMWARE("amdgpu/vega10_sdma1.bin"); MODULE_FIRMWARE("amdgpu/vega12_sdma.bin"); MODULE_FIRMWARE("amdgpu/vega12_sdma1.bin"); +MODULE_FIRMWARE("amdgpu/vega20_sdma.bin"); +MODULE_FIRMWARE("amdgpu/vega20_sdma1.bin"); MODULE_FIRMWARE("amdgpu/raven_sdma.bin"); #define SDMA0_POWER_CNTL__ON_OFF_CONDITION_HOLD_TIME_MASK 0x000000F8L @@ -107,6 +109,28 @@ static const struct soc15_reg_golden golden_settings_sdma_4_1[] = SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_PAGE, 0x000003ff, 0x000003c0) }; +static const struct soc15_reg_golden golden_settings_sdma_4_2[] = +{ + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_CHICKEN_BITS, 0xfe931f07, 0x02831d07), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_CLK_CTRL, 0xffffffff, 0x3f000100), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG, 0x0000773f, 0x00004002), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG_READ, 0x0000773f, 0x00004002), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GFX_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_PAGE_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC0_RB_WPTR_POLL_CNTL, 0xfffffff0, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_RLC1_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_UTCL1_PAGE, 0x000003ff, 0x000003c0), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CHICKEN_BITS, 0xfe931f07, 0x02831d07), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_CLK_CTRL, 0xffffffff, 0x3f000100), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG, 0x0000773f, 0x00004002), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GB_ADDR_CONFIG_READ, 0x0000773f, 0x00004002), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_GFX_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_PAGE_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC0_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_RLC1_RB_WPTR_POLL_CNTL, 0xfffffff7, 0x00403000), + SOC15_REG_GOLDEN_VALUE(SDMA1, 0, mmSDMA1_UTCL1_PAGE, 0x000003ff, 0x000003c0) +}; + static const struct soc15_reg_golden golden_settings_sdma_rv1[] = { SOC15_REG_GOLDEN_VALUE(SDMA0, 0, mmSDMA0_GB_ADDR_CONFIG, 0x0018773f, 0x00000002), @@ -139,6 +163,11 @@ static void sdma_v4_0_init_golden_registers(struct amdgpu_device *adev) golden_settings_sdma_vg12, ARRAY_SIZE(golden_settings_sdma_vg12)); break; + case CHIP_VEGA20: + soc15_program_register_sequence(adev, + golden_settings_sdma_4_2, + ARRAY_SIZE(golden_settings_sdma_4_2)); + break; case CHIP_RAVEN: soc15_program_register_sequence(adev, golden_settings_sdma_4_1, @@ -182,6 +211,9 @@ static int sdma_v4_0_init_microcode(struct amdgpu_device *adev) case CHIP_VEGA12: chip_name = "vega12"; break; + case CHIP_VEGA20: + chip_name = "vega20"; + break; case CHIP_RAVEN: chip_name = "raven"; break; @@ -360,6 +392,31 @@ static void sdma_v4_0_ring_emit_ib(struct amdgpu_ring *ring, } +static void sdma_v4_0_wait_reg_mem(struct amdgpu_ring *ring, + int mem_space, int hdp, + uint32_t addr0, uint32_t addr1, + uint32_t ref, uint32_t mask, + uint32_t inv) +{ + amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) | + SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(hdp) | + SDMA_PKT_POLL_REGMEM_HEADER_MEM_POLL(mem_space) | + SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* == */ + if (mem_space) { + /* memory */ + amdgpu_ring_write(ring, addr0); + amdgpu_ring_write(ring, addr1); + } else { + /* registers */ + amdgpu_ring_write(ring, addr0 << 2); + amdgpu_ring_write(ring, addr1 << 2); + } + amdgpu_ring_write(ring, ref); /* reference */ + amdgpu_ring_write(ring, mask); /* mask */ + amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | + SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(inv)); /* retry count, poll interval */ +} + /** * sdma_v4_0_ring_emit_hdp_flush - emit an hdp flush on the DMA ring * @@ -378,15 +435,10 @@ static void sdma_v4_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) else ref_and_mask = nbio_hf_reg->ref_and_mask_sdma1; - amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) | - SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(1) | - SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* == */ - amdgpu_ring_write(ring, (adev->nbio_funcs->get_hdp_flush_done_offset(adev)) << 2); - amdgpu_ring_write(ring, (adev->nbio_funcs->get_hdp_flush_req_offset(adev)) << 2); - amdgpu_ring_write(ring, ref_and_mask); /* reference */ - amdgpu_ring_write(ring, ref_and_mask); /* mask */ - amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | - SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); /* retry count, poll interval */ + sdma_v4_0_wait_reg_mem(ring, 0, 1, + adev->nbio_funcs->get_hdp_flush_done_offset(adev), + adev->nbio_funcs->get_hdp_flush_req_offset(adev), + ref_and_mask, ref_and_mask, 10); } /** @@ -1114,16 +1166,10 @@ static void sdma_v4_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) uint64_t addr = ring->fence_drv.gpu_addr; /* wait for idle */ - amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) | - SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) | - SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3) | /* equal */ - SDMA_PKT_POLL_REGMEM_HEADER_MEM_POLL(1)); - amdgpu_ring_write(ring, addr & 0xfffffffc); - amdgpu_ring_write(ring, upper_32_bits(addr) & 0xffffffff); - amdgpu_ring_write(ring, seq); /* reference */ - amdgpu_ring_write(ring, 0xffffffff); /* mask */ - amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | - SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(4)); /* retry count, poll interval */ + sdma_v4_0_wait_reg_mem(ring, 1, 0, + addr & 0xfffffffc, + upper_32_bits(addr) & 0xffffffff, + seq, 0xffffffff, 4); } @@ -1154,15 +1200,7 @@ static void sdma_v4_0_ring_emit_wreg(struct amdgpu_ring *ring, static void sdma_v4_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, uint32_t val, uint32_t mask) { - amdgpu_ring_write(ring, SDMA_PKT_HEADER_OP(SDMA_OP_POLL_REGMEM) | - SDMA_PKT_POLL_REGMEM_HEADER_HDP_FLUSH(0) | - SDMA_PKT_POLL_REGMEM_HEADER_FUNC(3)); /* equal */ - amdgpu_ring_write(ring, reg << 2); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, val); /* reference */ - amdgpu_ring_write(ring, mask); /* mask */ - amdgpu_ring_write(ring, SDMA_PKT_POLL_REGMEM_DW5_RETRY_COUNT(0xfff) | - SDMA_PKT_POLL_REGMEM_DW5_INTERVAL(10)); + sdma_v4_0_wait_reg_mem(ring, 0, 0, reg, 0, val, mask, 10); } static int sdma_v4_0_early_init(void *handle) @@ -1510,6 +1548,7 @@ static int sdma_v4_0_set_clockgating_state(void *handle, switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: case CHIP_RAVEN: sdma_v4_0_update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE ? true : false); @@ -1605,6 +1644,7 @@ static const struct amdgpu_ring_funcs sdma_v4_0_ring_funcs = { .pad_ib = sdma_v4_0_ring_pad_ib, .emit_wreg = sdma_v4_0_ring_emit_wreg, .emit_reg_wait = sdma_v4_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, }; static void sdma_v4_0_set_ring_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/si.c b/drivers/gpu/drm/amd/amdgpu/si.c index a675ec6d2811e..c364ef94cc366 100644 --- a/drivers/gpu/drm/amd/amdgpu/si.c +++ b/drivers/gpu/drm/amd/amdgpu/si.c @@ -1252,6 +1252,12 @@ static void si_invalidate_hdp(struct amdgpu_device *adev, } } +static bool si_need_full_reset(struct amdgpu_device *adev) +{ + /* change this when we support soft reset */ + return true; +} + static int si_get_pcie_lanes(struct amdgpu_device *adev) { u32 link_width_cntl; @@ -1332,6 +1338,7 @@ static const struct amdgpu_asic_funcs si_asic_funcs = .get_config_memsize = &si_get_config_memsize, .flush_hdp = &si_flush_hdp, .invalidate_hdp = &si_invalidate_hdp, + .need_full_reset = &si_need_full_reset, }; static uint32_t si_get_rev_id(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/si_dpm.c b/drivers/gpu/drm/amd/amdgpu/si_dpm.c index 797d505bf9ee1..b12d7c9d42a05 100644 --- a/drivers/gpu/drm/amd/amdgpu/si_dpm.c +++ b/drivers/gpu/drm/amd/amdgpu/si_dpm.c @@ -7580,7 +7580,7 @@ static int si_dpm_late_init(void *handle) int ret; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (!amdgpu_dpm) + if (!adev->pm.dpm_enabled) return 0; ret = si_set_temperature_range(adev); diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.c b/drivers/gpu/drm/amd/amdgpu/soc15.c index 51cf8a30f6c29..68b4a22a88925 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.c +++ b/drivers/gpu/drm/amd/amdgpu/soc15.c @@ -41,8 +41,6 @@ #include "sdma1/sdma1_4_0_offset.h" #include "hdp/hdp_4_0_offset.h" #include "hdp/hdp_4_0_sh_mask.h" -#include "mp/mp_9_0_offset.h" -#include "mp/mp_9_0_sh_mask.h" #include "smuio/smuio_9_0_offset.h" #include "smuio/smuio_9_0_sh_mask.h" @@ -52,6 +50,8 @@ #include "gmc_v9_0.h" #include "gfxhub_v1_0.h" #include "mmhub_v1_0.h" +#include "df_v1_7.h" +#include "df_v3_6.h" #include "vega10_ih.h" #include "sdma_v4_0.h" #include "uvd_v7_0.h" @@ -60,33 +60,6 @@ #include "dce_virtual.h" #include "mxgpu_ai.h" -#define mmFabricConfigAccessControl 0x0410 -#define mmFabricConfigAccessControl_BASE_IDX 0 -#define mmFabricConfigAccessControl_DEFAULT 0x00000000 -//FabricConfigAccessControl -#define FabricConfigAccessControl__CfgRegInstAccEn__SHIFT 0x0 -#define FabricConfigAccessControl__CfgRegInstAccRegLock__SHIFT 0x1 -#define FabricConfigAccessControl__CfgRegInstID__SHIFT 0x10 -#define FabricConfigAccessControl__CfgRegInstAccEn_MASK 0x00000001L -#define FabricConfigAccessControl__CfgRegInstAccRegLock_MASK 0x00000002L -#define FabricConfigAccessControl__CfgRegInstID_MASK 0x00FF0000L - - -#define mmDF_PIE_AON0_DfGlobalClkGater 0x00fc -#define mmDF_PIE_AON0_DfGlobalClkGater_BASE_IDX 0 -//DF_PIE_AON0_DfGlobalClkGater -#define DF_PIE_AON0_DfGlobalClkGater__MGCGMode__SHIFT 0x0 -#define DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK 0x0000000FL - -enum { - DF_MGCG_DISABLE = 0, - DF_MGCG_ENABLE_00_CYCLE_DELAY =1, - DF_MGCG_ENABLE_01_CYCLE_DELAY =2, - DF_MGCG_ENABLE_15_CYCLE_DELAY =13, - DF_MGCG_ENABLE_31_CYCLE_DELAY =14, - DF_MGCG_ENABLE_63_CYCLE_DELAY =15 -}; - #define mmMP0_MISC_CGTT_CTRL0 0x01b9 #define mmMP0_MISC_CGTT_CTRL0_BASE_IDX 0 #define mmMP0_MISC_LIGHT_SLEEP_CTRL 0x01ba @@ -313,6 +286,7 @@ static struct soc15_allowed_register_entry soc15_allowed_read_registers[] = { { SOC15_REG_ENTRY(GC, 0, mmCP_CPC_STALLED_STAT1)}, { SOC15_REG_ENTRY(GC, 0, mmCP_CPC_STATUS)}, { SOC15_REG_ENTRY(GC, 0, mmGB_ADDR_CONFIG)}, + { SOC15_REG_ENTRY(GC, 0, mmDB_DEBUG2)}, }; static uint32_t soc15_read_indexed_register(struct amdgpu_device *adev, u32 se_num, @@ -341,6 +315,8 @@ static uint32_t soc15_get_register_value(struct amdgpu_device *adev, } else { if (reg_offset == SOC15_REG_OFFSET(GC, 0, mmGB_ADDR_CONFIG)) return adev->gfx.config.gb_addr_config; + else if (reg_offset == SOC15_REG_OFFSET(GC, 0, mmDB_DEBUG2)) + return adev->gfx.config.db_debug2; return RREG32(reg_offset); } } @@ -512,15 +488,24 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) case CHIP_RAVEN: vega10_reg_base_init(adev); break; + case CHIP_VEGA20: + vega20_reg_base_init(adev); + break; default: return -EINVAL; } if (adev->flags & AMD_IS_APU) adev->nbio_funcs = &nbio_v7_0_funcs; + else if (adev->asic_type == CHIP_VEGA20) + adev->nbio_funcs = &nbio_v7_0_funcs; else adev->nbio_funcs = &nbio_v6_1_funcs; + if (adev->asic_type == CHIP_VEGA20) + adev->df_funcs = &df_v3_6_funcs; + else + adev->df_funcs = &df_v1_7_funcs; adev->nbio_funcs->detect_hw_virt(adev); if (amdgpu_sriov_vf(adev)) @@ -529,12 +514,15 @@ int soc15_set_ip_blocks(struct amdgpu_device *adev) switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: amdgpu_device_ip_block_add(adev, &vega10_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v9_0_ip_block); amdgpu_device_ip_block_add(adev, &vega10_ih_ip_block); - amdgpu_device_ip_block_add(adev, &psp_v3_1_ip_block); - if (!amdgpu_sriov_vf(adev)) - amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); + if (adev->asic_type != CHIP_VEGA20) { + amdgpu_device_ip_block_add(adev, &psp_v3_1_ip_block); + if (!amdgpu_sriov_vf(adev)) + amdgpu_device_ip_block_add(adev, &pp_smu_ip_block); + } if (adev->enable_virtual_display || amdgpu_sriov_vf(adev)) amdgpu_device_ip_block_add(adev, &dce_virtual_ip_block); #if defined(CONFIG_DRM_AMD_DC) @@ -593,6 +581,12 @@ static void soc15_invalidate_hdp(struct amdgpu_device *adev, HDP, 0, mmHDP_READ_CACHE_INVALIDATE), 1); } +static bool soc15_need_full_reset(struct amdgpu_device *adev) +{ + /* change this when we implement soft reset */ + return true; +} + static const struct amdgpu_asic_funcs soc15_asic_funcs = { .read_disabled_bios = &soc15_read_disabled_bios, @@ -606,6 +600,7 @@ static const struct amdgpu_asic_funcs soc15_asic_funcs = .get_config_memsize = &soc15_get_config_memsize, .flush_hdp = &soc15_flush_hdp, .invalidate_hdp = &soc15_invalidate_hdp, + .need_full_reset = &soc15_need_full_reset, }; static int soc15_common_early_init(void *handle) @@ -675,6 +670,27 @@ static int soc15_common_early_init(void *handle) adev->pg_flags = 0; adev->external_rev_id = adev->rev_id + 0x14; break; + case CHIP_VEGA20: + adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_MGLS | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGLS | + AMD_CG_SUPPORT_GFX_CP_LS | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_SDMA_MGCG | + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_BIF_MGCG | + AMD_CG_SUPPORT_BIF_LS | + AMD_CG_SUPPORT_HDP_MGCG | + AMD_CG_SUPPORT_ROM_MGCG | + AMD_CG_SUPPORT_VCE_MGCG | + AMD_CG_SUPPORT_UVD_MGCG; + adev->pg_flags = 0; + adev->external_rev_id = adev->rev_id + 0x28; + break; case CHIP_RAVEN: adev->cg_flags = AMD_CG_SUPPORT_GFX_MGCG | AMD_CG_SUPPORT_GFX_MGLS | @@ -694,8 +710,15 @@ static int soc15_common_early_init(void *handle) AMD_CG_SUPPORT_MC_MGCG | AMD_CG_SUPPORT_MC_LS | AMD_CG_SUPPORT_SDMA_MGCG | - AMD_CG_SUPPORT_SDMA_LS; - adev->pg_flags = AMD_PG_SUPPORT_SDMA; + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_VCN_MGCG; + + adev->pg_flags = AMD_PG_SUPPORT_SDMA | AMD_PG_SUPPORT_VCN; + + if (adev->powerplay.pp_feature & PP_GFXOFF_MASK) + adev->pg_flags |= AMD_PG_SUPPORT_GFX_PG | + AMD_PG_SUPPORT_CP | + AMD_PG_SUPPORT_RLC_SMU_HS; adev->external_rev_id = 0x1; break; @@ -871,32 +894,6 @@ static void soc15_update_rom_medium_grain_clock_gating(struct amdgpu_device *ade WREG32(SOC15_REG_OFFSET(SMUIO, 0, mmCGTT_ROM_CLK_CTRL0), data); } -static void soc15_update_df_medium_grain_clock_gating(struct amdgpu_device *adev, - bool enable) -{ - uint32_t data; - - /* Put DF on broadcast mode */ - data = RREG32(SOC15_REG_OFFSET(DF, 0, mmFabricConfigAccessControl)); - data &= ~FabricConfigAccessControl__CfgRegInstAccEn_MASK; - WREG32(SOC15_REG_OFFSET(DF, 0, mmFabricConfigAccessControl), data); - - if (enable && (adev->cg_flags & AMD_CG_SUPPORT_DF_MGCG)) { - data = RREG32(SOC15_REG_OFFSET(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater)); - data &= ~DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK; - data |= DF_MGCG_ENABLE_15_CYCLE_DELAY; - WREG32(SOC15_REG_OFFSET(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater), data); - } else { - data = RREG32(SOC15_REG_OFFSET(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater)); - data &= ~DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK; - data |= DF_MGCG_DISABLE; - WREG32(SOC15_REG_OFFSET(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater), data); - } - - WREG32(SOC15_REG_OFFSET(DF, 0, mmFabricConfigAccessControl), - mmFabricConfigAccessControl_DEFAULT); -} - static int soc15_common_set_clockgating_state(void *handle, enum amd_clockgating_state state) { @@ -908,6 +905,7 @@ static int soc15_common_set_clockgating_state(void *handle, switch (adev->asic_type) { case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: adev->nbio_funcs->update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE ? true : false); adev->nbio_funcs->update_medium_grain_light_sleep(adev, @@ -920,7 +918,7 @@ static int soc15_common_set_clockgating_state(void *handle, state == AMD_CG_STATE_GATE ? true : false); soc15_update_rom_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE ? true : false); - soc15_update_df_medium_grain_clock_gating(adev, + adev->df_funcs->update_medium_grain_clock_gating(adev, state == AMD_CG_STATE_GATE ? true : false); break; case CHIP_RAVEN: @@ -973,10 +971,7 @@ static void soc15_common_get_clockgating_state(void *handle, u32 *flags) if (!(data & CGTT_ROM_CLK_CTRL0__SOFT_OVERRIDE0_MASK)) *flags |= AMD_CG_SUPPORT_ROM_MGCG; - /* AMD_CG_SUPPORT_DF_MGCG */ - data = RREG32(SOC15_REG_OFFSET(DF, 0, mmDF_PIE_AON0_DfGlobalClkGater)); - if (data & DF_MGCG_ENABLE_15_CYCLE_DELAY) - *flags |= AMD_CG_SUPPORT_DF_MGCG; + adev->df_funcs->get_clockgating_state(adev, flags); } static int soc15_common_set_powergating_state(void *handle, diff --git a/drivers/gpu/drm/amd/amdgpu/soc15.h b/drivers/gpu/drm/amd/amdgpu/soc15.h index f70da8a29f864..1f714b7af520d 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15.h @@ -55,5 +55,6 @@ void soc15_program_register_sequence(struct amdgpu_device *adev, const u32 array_size); int vega10_reg_base_init(struct amdgpu_device *adev); +int vega20_reg_base_init(struct amdgpu_device *adev); #endif diff --git a/drivers/gpu/drm/amd/amdgpu/soc15_common.h b/drivers/gpu/drm/amd/amdgpu/soc15_common.h index def865067edd2..0942f492d2e19 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15_common.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15_common.h @@ -47,6 +47,21 @@ #define WREG32_SOC15_OFFSET(ip, inst, reg, offset, value) \ WREG32((adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg) + offset, value) +#define SOC15_WAIT_ON_RREG(ip, inst, reg, expected_value, mask, ret) \ + do { \ + uint32_t tmp_ = RREG32(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg); \ + uint32_t loop = adev->usec_timeout; \ + while ((tmp_ & (mask)) != (expected_value)) { \ + udelay(2); \ + tmp_ = RREG32(adev->reg_offset[ip##_HWIP][inst][reg##_BASE_IDX] + reg); \ + loop--; \ + if (!loop) { \ + ret = -ETIMEDOUT; \ + break; \ + } \ + } \ + } while (0) + #endif diff --git a/drivers/gpu/drm/amd/amdgpu/soc15d.h b/drivers/gpu/drm/amd/amdgpu/soc15d.h index 7f408f85fdb63..8dc29107228fd 100644 --- a/drivers/gpu/drm/amd/amdgpu/soc15d.h +++ b/drivers/gpu/drm/amd/amdgpu/soc15d.h @@ -159,6 +159,7 @@ #define EOP_TC_WB_ACTION_EN (1 << 15) /* L2 */ #define EOP_TCL1_ACTION_EN (1 << 16) #define EOP_TC_ACTION_EN (1 << 17) /* L2 */ +#define EOP_TC_NC_ACTION_EN (1 << 19) #define EOP_TC_MD_ACTION_EN (1 << 21) /* L2 metadata */ #define DATA_SEL(x) ((x) << 29) @@ -268,6 +269,11 @@ * x=1: tmz_end */ +#define PACKET3_INVALIDATE_TLBS 0x98 +# define PACKET3_INVALIDATE_TLBS_DST_SEL(x) ((x) << 0) +# define PACKET3_INVALIDATE_TLBS_ALL_HUB(x) ((x) << 4) +# define PACKET3_INVALIDATE_TLBS_PASID(x) ((x) << 5) +# define PACKET3_INVALIDATE_TLBS_FLUSH_TYPE(x) ((x) << 29) #define PACKET3_SET_RESOURCES 0xA0 /* 1. header * 2. CONTROL diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c index 948bb9437757f..6fed3d7797a8d 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v4_2.c @@ -93,6 +93,7 @@ static void uvd_v4_2_ring_set_wptr(struct amdgpu_ring *ring) static int uvd_v4_2_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + adev->uvd.num_uvd_inst = 1; uvd_v4_2_set_ring_funcs(adev); uvd_v4_2_set_irq_funcs(adev); @@ -107,7 +108,7 @@ static int uvd_v4_2_sw_init(void *handle) int r; /* UVD TRAP */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 124, &adev->uvd.irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 124, &adev->uvd.inst->irq); if (r) return r; @@ -119,9 +120,9 @@ static int uvd_v4_2_sw_init(void *handle) if (r) return r; - ring = &adev->uvd.ring; + ring = &adev->uvd.inst->ring; sprintf(ring->name, "uvd"); - r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); + r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst->irq, 0); return r; } @@ -150,7 +151,7 @@ static void uvd_v4_2_enable_mgcg(struct amdgpu_device *adev, static int uvd_v4_2_hw_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring = &adev->uvd.inst->ring; uint32_t tmp; int r; @@ -208,7 +209,7 @@ done: static int uvd_v4_2_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring = &adev->uvd.inst->ring; if (RREG32(mmUVD_STATUS) != 0) uvd_v4_2_stop(adev); @@ -251,7 +252,7 @@ static int uvd_v4_2_resume(void *handle) */ static int uvd_v4_2_start(struct amdgpu_device *adev) { - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring = &adev->uvd.inst->ring; uint32_t rb_bufsz; int i, j, r; u32 tmp; @@ -523,6 +524,18 @@ static void uvd_v4_2_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, ib->length_dw); } +static void uvd_v4_2_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) +{ + int i; + + WARN_ON(ring->wptr % 2 || count % 2); + + for (i = 0; i < count / 2; i++) { + amdgpu_ring_write(ring, PACKET0(mmUVD_NO_OP, 0)); + amdgpu_ring_write(ring, 0); + } +} + /** * uvd_v4_2_mc_resume - memory controller programming * @@ -536,7 +549,7 @@ static void uvd_v4_2_mc_resume(struct amdgpu_device *adev) uint32_t size; /* programm the VCPU memory controller bits 0-27 */ - addr = (adev->uvd.gpu_addr + AMDGPU_UVD_FIRMWARE_OFFSET) >> 3; + addr = (adev->uvd.inst->gpu_addr + AMDGPU_UVD_FIRMWARE_OFFSET) >> 3; size = AMDGPU_UVD_FIRMWARE_SIZE(adev) >> 3; WREG32(mmUVD_VCPU_CACHE_OFFSET0, addr); WREG32(mmUVD_VCPU_CACHE_SIZE0, size); @@ -553,11 +566,11 @@ static void uvd_v4_2_mc_resume(struct amdgpu_device *adev) WREG32(mmUVD_VCPU_CACHE_SIZE2, size); /* bits 28-31 */ - addr = (adev->uvd.gpu_addr >> 28) & 0xF; + addr = (adev->uvd.inst->gpu_addr >> 28) & 0xF; WREG32(mmUVD_LMI_ADDR_EXT, (addr << 12) | (addr << 0)); /* bits 32-39 */ - addr = (adev->uvd.gpu_addr >> 32) & 0xFF; + addr = (adev->uvd.inst->gpu_addr >> 32) & 0xFF; WREG32(mmUVD_LMI_EXT40_ADDR, addr | (0x9 << 16) | (0x1 << 31)); WREG32(mmUVD_UDEC_ADDR_CONFIG, adev->gfx.config.gb_addr_config); @@ -664,7 +677,7 @@ static int uvd_v4_2_process_interrupt(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { DRM_DEBUG("IH: UVD TRAP\n"); - amdgpu_fence_process(&adev->uvd.ring); + amdgpu_fence_process(&adev->uvd.inst->ring); return 0; } @@ -688,7 +701,7 @@ static int uvd_v4_2_set_powergating_state(void *handle, if (state == AMD_PG_STATE_GATE) { uvd_v4_2_stop(adev); - if (adev->pg_flags & AMD_PG_SUPPORT_UVD && amdgpu_dpm == 0) { + if (adev->pg_flags & AMD_PG_SUPPORT_UVD && !adev->pm.dpm_enabled) { if (!(RREG32_SMC(ixCURRENT_PG_STATUS) & CURRENT_PG_STATUS__UVD_PG_STATUS_MASK)) { WREG32(mmUVD_PGFSM_CONFIG, (UVD_PGFSM_CONFIG__UVD_PGFSM_FSM_ADDR_MASK | @@ -699,7 +712,7 @@ static int uvd_v4_2_set_powergating_state(void *handle, } return 0; } else { - if (adev->pg_flags & AMD_PG_SUPPORT_UVD && amdgpu_dpm == 0) { + if (adev->pg_flags & AMD_PG_SUPPORT_UVD && !adev->pm.dpm_enabled) { if (RREG32_SMC(ixCURRENT_PG_STATUS) & CURRENT_PG_STATUS__UVD_PG_STATUS_MASK) { WREG32(mmUVD_PGFSM_CONFIG, (UVD_PGFSM_CONFIG__UVD_PGFSM_FSM_ADDR_MASK | @@ -732,7 +745,6 @@ static const struct amd_ip_funcs uvd_v4_2_ip_funcs = { static const struct amdgpu_ring_funcs uvd_v4_2_ring_funcs = { .type = AMDGPU_RING_TYPE_UVD, .align_mask = 0xf, - .nop = PACKET0(mmUVD_NO_OP, 0), .support_64bit_ptrs = false, .get_rptr = uvd_v4_2_ring_get_rptr, .get_wptr = uvd_v4_2_ring_get_wptr, @@ -745,7 +757,7 @@ static const struct amdgpu_ring_funcs uvd_v4_2_ring_funcs = { .emit_fence = uvd_v4_2_ring_emit_fence, .test_ring = uvd_v4_2_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, - .insert_nop = amdgpu_ring_insert_nop, + .insert_nop = uvd_v4_2_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .begin_use = amdgpu_uvd_ring_begin_use, .end_use = amdgpu_uvd_ring_end_use, @@ -753,7 +765,7 @@ static const struct amdgpu_ring_funcs uvd_v4_2_ring_funcs = { static void uvd_v4_2_set_ring_funcs(struct amdgpu_device *adev) { - adev->uvd.ring.funcs = &uvd_v4_2_ring_funcs; + adev->uvd.inst->ring.funcs = &uvd_v4_2_ring_funcs; } static const struct amdgpu_irq_src_funcs uvd_v4_2_irq_funcs = { @@ -763,8 +775,8 @@ static const struct amdgpu_irq_src_funcs uvd_v4_2_irq_funcs = { static void uvd_v4_2_set_irq_funcs(struct amdgpu_device *adev) { - adev->uvd.irq.num_types = 1; - adev->uvd.irq.funcs = &uvd_v4_2_irq_funcs; + adev->uvd.inst->irq.num_types = 1; + adev->uvd.inst->irq.funcs = &uvd_v4_2_irq_funcs; } const struct amdgpu_ip_block_version uvd_v4_2_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c index 6445d55e7d5a8..341ee6d55ce89 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v5_0.c @@ -89,6 +89,7 @@ static void uvd_v5_0_ring_set_wptr(struct amdgpu_ring *ring) static int uvd_v5_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + adev->uvd.num_uvd_inst = 1; uvd_v5_0_set_ring_funcs(adev); uvd_v5_0_set_irq_funcs(adev); @@ -103,7 +104,7 @@ static int uvd_v5_0_sw_init(void *handle) int r; /* UVD TRAP */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 124, &adev->uvd.irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 124, &adev->uvd.inst->irq); if (r) return r; @@ -115,9 +116,9 @@ static int uvd_v5_0_sw_init(void *handle) if (r) return r; - ring = &adev->uvd.ring; + ring = &adev->uvd.inst->ring; sprintf(ring->name, "uvd"); - r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); + r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst->irq, 0); return r; } @@ -144,7 +145,7 @@ static int uvd_v5_0_sw_fini(void *handle) static int uvd_v5_0_hw_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring = &adev->uvd.inst->ring; uint32_t tmp; int r; @@ -204,7 +205,7 @@ done: static int uvd_v5_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring = &adev->uvd.inst->ring; if (RREG32(mmUVD_STATUS) != 0) uvd_v5_0_stop(adev); @@ -253,9 +254,9 @@ static void uvd_v5_0_mc_resume(struct amdgpu_device *adev) /* programm memory controller bits 0-27 */ WREG32(mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, - lower_32_bits(adev->uvd.gpu_addr)); + lower_32_bits(adev->uvd.inst->gpu_addr)); WREG32(mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, - upper_32_bits(adev->uvd.gpu_addr)); + upper_32_bits(adev->uvd.inst->gpu_addr)); offset = AMDGPU_UVD_FIRMWARE_OFFSET; size = AMDGPU_UVD_FIRMWARE_SIZE(adev); @@ -287,7 +288,7 @@ static void uvd_v5_0_mc_resume(struct amdgpu_device *adev) */ static int uvd_v5_0_start(struct amdgpu_device *adev) { - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring = &adev->uvd.inst->ring; uint32_t rb_bufsz, tmp; uint32_t lmi_swap_cntl; uint32_t mp_swap_cntl; @@ -540,6 +541,18 @@ static void uvd_v5_0_ring_emit_ib(struct amdgpu_ring *ring, amdgpu_ring_write(ring, ib->length_dw); } +static void uvd_v5_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) +{ + int i; + + WARN_ON(ring->wptr % 2 || count % 2); + + for (i = 0; i < count / 2; i++) { + amdgpu_ring_write(ring, PACKET0(mmUVD_NO_OP, 0)); + amdgpu_ring_write(ring, 0); + } +} + static bool uvd_v5_0_is_idle(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; @@ -586,7 +599,7 @@ static int uvd_v5_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_iv_entry *entry) { DRM_DEBUG("IH: UVD TRAP\n"); - amdgpu_fence_process(&adev->uvd.ring); + amdgpu_fence_process(&adev->uvd.inst->ring); return 0; } @@ -840,7 +853,6 @@ static const struct amd_ip_funcs uvd_v5_0_ip_funcs = { static const struct amdgpu_ring_funcs uvd_v5_0_ring_funcs = { .type = AMDGPU_RING_TYPE_UVD, .align_mask = 0xf, - .nop = PACKET0(mmUVD_NO_OP, 0), .support_64bit_ptrs = false, .get_rptr = uvd_v5_0_ring_get_rptr, .get_wptr = uvd_v5_0_ring_get_wptr, @@ -853,7 +865,7 @@ static const struct amdgpu_ring_funcs uvd_v5_0_ring_funcs = { .emit_fence = uvd_v5_0_ring_emit_fence, .test_ring = uvd_v5_0_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, - .insert_nop = amdgpu_ring_insert_nop, + .insert_nop = uvd_v5_0_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .begin_use = amdgpu_uvd_ring_begin_use, .end_use = amdgpu_uvd_ring_end_use, @@ -861,7 +873,7 @@ static const struct amdgpu_ring_funcs uvd_v5_0_ring_funcs = { static void uvd_v5_0_set_ring_funcs(struct amdgpu_device *adev) { - adev->uvd.ring.funcs = &uvd_v5_0_ring_funcs; + adev->uvd.inst->ring.funcs = &uvd_v5_0_ring_funcs; } static const struct amdgpu_irq_src_funcs uvd_v5_0_irq_funcs = { @@ -871,8 +883,8 @@ static const struct amdgpu_irq_src_funcs uvd_v5_0_irq_funcs = { static void uvd_v5_0_set_irq_funcs(struct amdgpu_device *adev) { - adev->uvd.irq.num_types = 1; - adev->uvd.irq.funcs = &uvd_v5_0_irq_funcs; + adev->uvd.inst->irq.num_types = 1; + adev->uvd.inst->irq.funcs = &uvd_v5_0_irq_funcs; } const struct amdgpu_ip_block_version uvd_v5_0_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c index f26f515db2fb5..bfddf97dd13e6 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v6_0.c @@ -62,7 +62,7 @@ static void uvd_v6_0_enable_mgcg(struct amdgpu_device *adev, static inline bool uvd_v6_0_enc_support(struct amdgpu_device *adev) { return ((adev->asic_type >= CHIP_POLARIS10) && - (adev->asic_type <= CHIP_POLARIS12) && + (adev->asic_type <= CHIP_VEGAM) && (!adev->uvd.fw_version || adev->uvd.fw_version >= FW_1_130_16)); } @@ -91,7 +91,7 @@ static uint64_t uvd_v6_0_enc_ring_get_rptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - if (ring == &adev->uvd.ring_enc[0]) + if (ring == &adev->uvd.inst->ring_enc[0]) return RREG32(mmUVD_RB_RPTR); else return RREG32(mmUVD_RB_RPTR2); @@ -121,7 +121,7 @@ static uint64_t uvd_v6_0_enc_ring_get_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - if (ring == &adev->uvd.ring_enc[0]) + if (ring == &adev->uvd.inst->ring_enc[0]) return RREG32(mmUVD_RB_WPTR); else return RREG32(mmUVD_RB_WPTR2); @@ -152,7 +152,7 @@ static void uvd_v6_0_enc_ring_set_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - if (ring == &adev->uvd.ring_enc[0]) + if (ring == &adev->uvd.inst->ring_enc[0]) WREG32(mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); else @@ -375,6 +375,7 @@ error: static int uvd_v6_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + adev->uvd.num_uvd_inst = 1; if (!(adev->flags & AMD_IS_APU) && (RREG32_SMC(ixCC_HARVEST_FUSES) & CC_HARVEST_FUSES__UVD_DISABLE_MASK)) @@ -399,14 +400,14 @@ static int uvd_v6_0_sw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; /* UVD TRAP */ - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 124, &adev->uvd.irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, 124, &adev->uvd.inst->irq); if (r) return r; /* UVD ENC TRAP */ if (uvd_v6_0_enc_support(adev)) { for (i = 0; i < adev->uvd.num_enc_rings; ++i) { - r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 119, &adev->uvd.irq); + r = amdgpu_irq_add_id(adev, AMDGPU_IH_CLIENTID_LEGACY, i + 119, &adev->uvd.inst->irq); if (r) return r; } @@ -418,18 +419,18 @@ static int uvd_v6_0_sw_init(void *handle) if (!uvd_v6_0_enc_support(adev)) { for (i = 0; i < adev->uvd.num_enc_rings; ++i) - adev->uvd.ring_enc[i].funcs = NULL; + adev->uvd.inst->ring_enc[i].funcs = NULL; - adev->uvd.irq.num_types = 1; + adev->uvd.inst->irq.num_types = 1; adev->uvd.num_enc_rings = 0; DRM_INFO("UVD ENC is disabled\n"); } else { struct drm_sched_rq *rq; - ring = &adev->uvd.ring_enc[0]; + ring = &adev->uvd.inst->ring_enc[0]; rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; - r = drm_sched_entity_init(&ring->sched, &adev->uvd.entity_enc, - rq, amdgpu_sched_jobs, NULL); + r = drm_sched_entity_init(&ring->sched, &adev->uvd.inst->entity_enc, + rq, NULL); if (r) { DRM_ERROR("Failed setting up UVD ENC run queue.\n"); return r; @@ -440,17 +441,17 @@ static int uvd_v6_0_sw_init(void *handle) if (r) return r; - ring = &adev->uvd.ring; + ring = &adev->uvd.inst->ring; sprintf(ring->name, "uvd"); - r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); + r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst->irq, 0); if (r) return r; if (uvd_v6_0_enc_support(adev)) { for (i = 0; i < adev->uvd.num_enc_rings; ++i) { - ring = &adev->uvd.ring_enc[i]; + ring = &adev->uvd.inst->ring_enc[i]; sprintf(ring->name, "uvd_enc%d", i); - r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); + r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst->irq, 0); if (r) return r; } @@ -469,10 +470,10 @@ static int uvd_v6_0_sw_fini(void *handle) return r; if (uvd_v6_0_enc_support(adev)) { - drm_sched_entity_fini(&adev->uvd.ring_enc[0].sched, &adev->uvd.entity_enc); + drm_sched_entity_fini(&adev->uvd.inst->ring_enc[0].sched, &adev->uvd.inst->entity_enc); for (i = 0; i < adev->uvd.num_enc_rings; ++i) - amdgpu_ring_fini(&adev->uvd.ring_enc[i]); + amdgpu_ring_fini(&adev->uvd.inst->ring_enc[i]); } return amdgpu_uvd_sw_fini(adev); @@ -488,7 +489,7 @@ static int uvd_v6_0_sw_fini(void *handle) static int uvd_v6_0_hw_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring = &adev->uvd.inst->ring; uint32_t tmp; int i, r; @@ -532,7 +533,7 @@ static int uvd_v6_0_hw_init(void *handle) if (uvd_v6_0_enc_support(adev)) { for (i = 0; i < adev->uvd.num_enc_rings; ++i) { - ring = &adev->uvd.ring_enc[i]; + ring = &adev->uvd.inst->ring_enc[i]; ring->ready = true; r = amdgpu_ring_test_ring(ring); if (r) { @@ -563,7 +564,7 @@ done: static int uvd_v6_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring = &adev->uvd.inst->ring; if (RREG32(mmUVD_STATUS) != 0) uvd_v6_0_stop(adev); @@ -611,9 +612,9 @@ static void uvd_v6_0_mc_resume(struct amdgpu_device *adev) /* programm memory controller bits 0-27 */ WREG32(mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, - lower_32_bits(adev->uvd.gpu_addr)); + lower_32_bits(adev->uvd.inst->gpu_addr)); WREG32(mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, - upper_32_bits(adev->uvd.gpu_addr)); + upper_32_bits(adev->uvd.inst->gpu_addr)); offset = AMDGPU_UVD_FIRMWARE_OFFSET; size = AMDGPU_UVD_FIRMWARE_SIZE(adev); @@ -726,7 +727,7 @@ static void cz_set_uvd_clock_gating_branches(struct amdgpu_device *adev, */ static int uvd_v6_0_start(struct amdgpu_device *adev) { - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring = &adev->uvd.inst->ring; uint32_t rb_bufsz, tmp; uint32_t lmi_swap_cntl; uint32_t mp_swap_cntl; @@ -866,14 +867,14 @@ static int uvd_v6_0_start(struct amdgpu_device *adev) WREG32_FIELD(UVD_RBC_RB_CNTL, RB_NO_FETCH, 0); if (uvd_v6_0_enc_support(adev)) { - ring = &adev->uvd.ring_enc[0]; + ring = &adev->uvd.inst->ring_enc[0]; WREG32(mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); WREG32(mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); WREG32(mmUVD_RB_BASE_LO, ring->gpu_addr); WREG32(mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); WREG32(mmUVD_RB_SIZE, ring->ring_size / 4); - ring = &adev->uvd.ring_enc[1]; + ring = &adev->uvd.inst->ring_enc[1]; WREG32(mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); WREG32(mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); WREG32(mmUVD_RB_BASE_LO2, ring->gpu_addr); @@ -964,6 +965,16 @@ static void uvd_v6_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, } /** + * uvd_v6_0_ring_emit_hdp_flush - skip HDP flushing + * + * @ring: amdgpu_ring pointer + */ +static void uvd_v6_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) +{ + /* The firmware doesn't seem to like touching registers at this point. */ +} + +/** * uvd_v6_0_ring_test_ring - register write test * * @ring: amdgpu_ring pointer @@ -1089,6 +1100,18 @@ static void uvd_v6_0_ring_emit_pipeline_sync(struct amdgpu_ring *ring) amdgpu_ring_write(ring, 0xE); } +static void uvd_v6_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) +{ + int i; + + WARN_ON(ring->wptr % 2 || count % 2); + + for (i = 0; i < count / 2; i++) { + amdgpu_ring_write(ring, PACKET0(mmUVD_NO_OP, 0)); + amdgpu_ring_write(ring, 0); + } +} + static void uvd_v6_0_enc_ring_emit_pipeline_sync(struct amdgpu_ring *ring) { uint32_t seq = ring->fence_drv.sync_seq; @@ -1148,10 +1171,10 @@ static bool uvd_v6_0_check_soft_reset(void *handle) srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_UVD, 1); if (srbm_soft_reset) { - adev->uvd.srbm_soft_reset = srbm_soft_reset; + adev->uvd.inst->srbm_soft_reset = srbm_soft_reset; return true; } else { - adev->uvd.srbm_soft_reset = 0; + adev->uvd.inst->srbm_soft_reset = 0; return false; } } @@ -1160,7 +1183,7 @@ static int uvd_v6_0_pre_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (!adev->uvd.srbm_soft_reset) + if (!adev->uvd.inst->srbm_soft_reset) return 0; uvd_v6_0_stop(adev); @@ -1172,9 +1195,9 @@ static int uvd_v6_0_soft_reset(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; u32 srbm_soft_reset; - if (!adev->uvd.srbm_soft_reset) + if (!adev->uvd.inst->srbm_soft_reset) return 0; - srbm_soft_reset = adev->uvd.srbm_soft_reset; + srbm_soft_reset = adev->uvd.inst->srbm_soft_reset; if (srbm_soft_reset) { u32 tmp; @@ -1202,7 +1225,7 @@ static int uvd_v6_0_post_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (!adev->uvd.srbm_soft_reset) + if (!adev->uvd.inst->srbm_soft_reset) return 0; mdelay(5); @@ -1228,17 +1251,17 @@ static int uvd_v6_0_process_interrupt(struct amdgpu_device *adev, switch (entry->src_id) { case 124: - amdgpu_fence_process(&adev->uvd.ring); + amdgpu_fence_process(&adev->uvd.inst->ring); break; case 119: if (likely(uvd_v6_0_enc_support(adev))) - amdgpu_fence_process(&adev->uvd.ring_enc[0]); + amdgpu_fence_process(&adev->uvd.inst->ring_enc[0]); else int_handled = false; break; case 120: if (likely(uvd_v6_0_enc_support(adev))) - amdgpu_fence_process(&adev->uvd.ring_enc[1]); + amdgpu_fence_process(&adev->uvd.inst->ring_enc[1]); else int_handled = false; break; @@ -1521,22 +1544,22 @@ static const struct amd_ip_funcs uvd_v6_0_ip_funcs = { static const struct amdgpu_ring_funcs uvd_v6_0_ring_phys_funcs = { .type = AMDGPU_RING_TYPE_UVD, .align_mask = 0xf, - .nop = PACKET0(mmUVD_NO_OP, 0), .support_64bit_ptrs = false, .get_rptr = uvd_v6_0_ring_get_rptr, .get_wptr = uvd_v6_0_ring_get_wptr, .set_wptr = uvd_v6_0_ring_set_wptr, .parse_cs = amdgpu_uvd_ring_parse_cs, .emit_frame_size = - 6 + 6 + /* hdp flush / invalidate */ + 6 + /* hdp invalidate */ 10 + /* uvd_v6_0_ring_emit_pipeline_sync */ 14, /* uvd_v6_0_ring_emit_fence x1 no user fence */ .emit_ib_size = 8, /* uvd_v6_0_ring_emit_ib */ .emit_ib = uvd_v6_0_ring_emit_ib, .emit_fence = uvd_v6_0_ring_emit_fence, + .emit_hdp_flush = uvd_v6_0_ring_emit_hdp_flush, .test_ring = uvd_v6_0_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, - .insert_nop = amdgpu_ring_insert_nop, + .insert_nop = uvd_v6_0_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .begin_use = amdgpu_uvd_ring_begin_use, .end_use = amdgpu_uvd_ring_end_use, @@ -1552,7 +1575,7 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_vm_funcs = { .get_wptr = uvd_v6_0_ring_get_wptr, .set_wptr = uvd_v6_0_ring_set_wptr, .emit_frame_size = - 6 + 6 + /* hdp flush / invalidate */ + 6 + /* hdp invalidate */ 10 + /* uvd_v6_0_ring_emit_pipeline_sync */ VI_FLUSH_GPU_TLB_NUM_WREG * 6 + 8 + /* uvd_v6_0_ring_emit_vm_flush */ 14 + 14, /* uvd_v6_0_ring_emit_fence x2 vm fence */ @@ -1561,6 +1584,7 @@ static const struct amdgpu_ring_funcs uvd_v6_0_ring_vm_funcs = { .emit_fence = uvd_v6_0_ring_emit_fence, .emit_vm_flush = uvd_v6_0_ring_emit_vm_flush, .emit_pipeline_sync = uvd_v6_0_ring_emit_pipeline_sync, + .emit_hdp_flush = uvd_v6_0_ring_emit_hdp_flush, .test_ring = uvd_v6_0_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, @@ -1600,10 +1624,10 @@ static const struct amdgpu_ring_funcs uvd_v6_0_enc_ring_vm_funcs = { static void uvd_v6_0_set_ring_funcs(struct amdgpu_device *adev) { if (adev->asic_type >= CHIP_POLARIS10) { - adev->uvd.ring.funcs = &uvd_v6_0_ring_vm_funcs; + adev->uvd.inst->ring.funcs = &uvd_v6_0_ring_vm_funcs; DRM_INFO("UVD is enabled in VM mode\n"); } else { - adev->uvd.ring.funcs = &uvd_v6_0_ring_phys_funcs; + adev->uvd.inst->ring.funcs = &uvd_v6_0_ring_phys_funcs; DRM_INFO("UVD is enabled in physical mode\n"); } } @@ -1613,7 +1637,7 @@ static void uvd_v6_0_set_enc_ring_funcs(struct amdgpu_device *adev) int i; for (i = 0; i < adev->uvd.num_enc_rings; ++i) - adev->uvd.ring_enc[i].funcs = &uvd_v6_0_enc_ring_vm_funcs; + adev->uvd.inst->ring_enc[i].funcs = &uvd_v6_0_enc_ring_vm_funcs; DRM_INFO("UVD ENC is enabled in VM mode\n"); } @@ -1626,11 +1650,11 @@ static const struct amdgpu_irq_src_funcs uvd_v6_0_irq_funcs = { static void uvd_v6_0_set_irq_funcs(struct amdgpu_device *adev) { if (uvd_v6_0_enc_support(adev)) - adev->uvd.irq.num_types = adev->uvd.num_enc_rings + 1; + adev->uvd.inst->irq.num_types = adev->uvd.num_enc_rings + 1; else - adev->uvd.irq.num_types = 1; + adev->uvd.inst->irq.num_types = 1; - adev->uvd.irq.funcs = &uvd_v6_0_irq_funcs; + adev->uvd.inst->irq.funcs = &uvd_v6_0_irq_funcs; } const struct amdgpu_ip_block_version uvd_v6_0_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c index eddc57f3b72a4..57d32f21b3a61 100644 --- a/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/uvd_v7_0.c @@ -40,6 +40,8 @@ #include "mmhub/mmhub_1_0_offset.h" #include "mmhub/mmhub_1_0_sh_mask.h" +#define UVD7_MAX_HW_INSTANCES_VEGA20 2 + static void uvd_v7_0_set_ring_funcs(struct amdgpu_device *adev); static void uvd_v7_0_set_enc_ring_funcs(struct amdgpu_device *adev); static void uvd_v7_0_set_irq_funcs(struct amdgpu_device *adev); @@ -47,6 +49,11 @@ static int uvd_v7_0_start(struct amdgpu_device *adev); static void uvd_v7_0_stop(struct amdgpu_device *adev); static int uvd_v7_0_sriov_start(struct amdgpu_device *adev); +static int amdgpu_ih_clientid_uvds[] = { + SOC15_IH_CLIENTID_UVD, + SOC15_IH_CLIENTID_UVD1 +}; + /** * uvd_v7_0_ring_get_rptr - get read pointer * @@ -58,7 +65,7 @@ static uint64_t uvd_v7_0_ring_get_rptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR); + return RREG32_SOC15(UVD, ring->me, mmUVD_RBC_RB_RPTR); } /** @@ -72,10 +79,10 @@ static uint64_t uvd_v7_0_enc_ring_get_rptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - if (ring == &adev->uvd.ring_enc[0]) - return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR); + if (ring == &adev->uvd.inst[ring->me].ring_enc[0]) + return RREG32_SOC15(UVD, ring->me, mmUVD_RB_RPTR); else - return RREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2); + return RREG32_SOC15(UVD, ring->me, mmUVD_RB_RPTR2); } /** @@ -89,7 +96,7 @@ static uint64_t uvd_v7_0_ring_get_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - return RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR); + return RREG32_SOC15(UVD, ring->me, mmUVD_RBC_RB_WPTR); } /** @@ -106,10 +113,10 @@ static uint64_t uvd_v7_0_enc_ring_get_wptr(struct amdgpu_ring *ring) if (ring->use_doorbell) return adev->wb.wb[ring->wptr_offs]; - if (ring == &adev->uvd.ring_enc[0]) - return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR); + if (ring == &adev->uvd.inst[ring->me].ring_enc[0]) + return RREG32_SOC15(UVD, ring->me, mmUVD_RB_WPTR); else - return RREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2); + return RREG32_SOC15(UVD, ring->me, mmUVD_RB_WPTR2); } /** @@ -123,7 +130,7 @@ static void uvd_v7_0_ring_set_wptr(struct amdgpu_ring *ring) { struct amdgpu_device *adev = ring->adev; - WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr)); + WREG32_SOC15(UVD, ring->me, mmUVD_RBC_RB_WPTR, lower_32_bits(ring->wptr)); } /** @@ -144,11 +151,11 @@ static void uvd_v7_0_enc_ring_set_wptr(struct amdgpu_ring *ring) return; } - if (ring == &adev->uvd.ring_enc[0]) - WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, + if (ring == &adev->uvd.inst[ring->me].ring_enc[0]) + WREG32_SOC15(UVD, ring->me, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); else - WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, + WREG32_SOC15(UVD, ring->me, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); } @@ -170,8 +177,8 @@ static int uvd_v7_0_enc_ring_test_ring(struct amdgpu_ring *ring) r = amdgpu_ring_alloc(ring, 16); if (r) { - DRM_ERROR("amdgpu: uvd enc failed to lock ring %d (%d).\n", - ring->idx, r); + DRM_ERROR("amdgpu: uvd enc failed to lock (%d)ring %d (%d).\n", + ring->me, ring->idx, r); return r; } amdgpu_ring_write(ring, HEVC_ENC_CMD_END); @@ -184,11 +191,11 @@ static int uvd_v7_0_enc_ring_test_ring(struct amdgpu_ring *ring) } if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); + DRM_DEBUG("(%d)ring test on %d succeeded in %d usecs\n", + ring->me, ring->idx, i); } else { - DRM_ERROR("amdgpu: ring %d test failed\n", - ring->idx); + DRM_ERROR("amdgpu: (%d)ring %d test failed\n", + ring->me, ring->idx); r = -ETIMEDOUT; } @@ -342,24 +349,24 @@ static int uvd_v7_0_enc_ring_test_ib(struct amdgpu_ring *ring, long timeout) r = uvd_v7_0_enc_get_create_msg(ring, 1, NULL); if (r) { - DRM_ERROR("amdgpu: failed to get create msg (%ld).\n", r); + DRM_ERROR("amdgpu: (%d)failed to get create msg (%ld).\n", ring->me, r); goto error; } r = uvd_v7_0_enc_get_destroy_msg(ring, 1, true, &fence); if (r) { - DRM_ERROR("amdgpu: failed to get destroy ib (%ld).\n", r); + DRM_ERROR("amdgpu: (%d)failed to get destroy ib (%ld).\n", ring->me, r); goto error; } r = dma_fence_wait_timeout(fence, false, timeout); if (r == 0) { - DRM_ERROR("amdgpu: IB test timed out.\n"); + DRM_ERROR("amdgpu: (%d)IB test timed out.\n", ring->me); r = -ETIMEDOUT; } else if (r < 0) { - DRM_ERROR("amdgpu: fence wait failed (%ld).\n", r); + DRM_ERROR("amdgpu: (%d)fence wait failed (%ld).\n", ring->me, r); } else { - DRM_DEBUG("ib test on ring %d succeeded\n", ring->idx); + DRM_DEBUG("ib test on (%d)ring %d succeeded\n", ring->me, ring->idx); r = 0; } error: @@ -370,6 +377,10 @@ error: static int uvd_v7_0_early_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; + if (adev->asic_type == CHIP_VEGA20) + adev->uvd.num_uvd_inst = UVD7_MAX_HW_INSTANCES_VEGA20; + else + adev->uvd.num_uvd_inst = 1; if (amdgpu_sriov_vf(adev)) adev->uvd.num_enc_rings = 1; @@ -386,19 +397,21 @@ static int uvd_v7_0_sw_init(void *handle) { struct amdgpu_ring *ring; struct drm_sched_rq *rq; - int i, r; + int i, j, r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; - /* UVD TRAP */ - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_UVD, 124, &adev->uvd.irq); - if (r) - return r; - - /* UVD ENC TRAP */ - for (i = 0; i < adev->uvd.num_enc_rings; ++i) { - r = amdgpu_irq_add_id(adev, SOC15_IH_CLIENTID_UVD, i + 119, &adev->uvd.irq); + for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + /* UVD TRAP */ + r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_uvds[j], 124, &adev->uvd.inst[j].irq); if (r) return r; + + /* UVD ENC TRAP */ + for (i = 0; i < adev->uvd.num_enc_rings; ++i) { + r = amdgpu_irq_add_id(adev, amdgpu_ih_clientid_uvds[j], i + 119, &adev->uvd.inst[j].irq); + if (r) + return r; + } } r = amdgpu_uvd_sw_init(adev); @@ -415,43 +428,48 @@ static int uvd_v7_0_sw_init(void *handle) DRM_INFO("PSP loading UVD firmware\n"); } - ring = &adev->uvd.ring_enc[0]; - rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; - r = drm_sched_entity_init(&ring->sched, &adev->uvd.entity_enc, - rq, amdgpu_sched_jobs, NULL); - if (r) { - DRM_ERROR("Failed setting up UVD ENC run queue.\n"); - return r; + for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + ring = &adev->uvd.inst[j].ring_enc[0]; + rq = &ring->sched.sched_rq[DRM_SCHED_PRIORITY_NORMAL]; + r = drm_sched_entity_init(&ring->sched, &adev->uvd.inst[j].entity_enc, + rq, NULL); + if (r) { + DRM_ERROR("(%d)Failed setting up UVD ENC run queue.\n", j); + return r; + } } r = amdgpu_uvd_resume(adev); if (r) return r; - if (!amdgpu_sriov_vf(adev)) { - ring = &adev->uvd.ring; - sprintf(ring->name, "uvd"); - r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); - if (r) - return r; - } - for (i = 0; i < adev->uvd.num_enc_rings; ++i) { - ring = &adev->uvd.ring_enc[i]; - sprintf(ring->name, "uvd_enc%d", i); - if (amdgpu_sriov_vf(adev)) { - ring->use_doorbell = true; - - /* currently only use the first enconding ring for - * sriov, so set unused location for other unused rings. - */ - if (i == 0) - ring->doorbell_index = AMDGPU_DOORBELL64_UVD_RING0_1 * 2; - else - ring->doorbell_index = AMDGPU_DOORBELL64_UVD_RING2_3 * 2 + 1; + for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + if (!amdgpu_sriov_vf(adev)) { + ring = &adev->uvd.inst[j].ring; + sprintf(ring->name, "uvd<%d>", j); + r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst[j].irq, 0); + if (r) + return r; + } + + for (i = 0; i < adev->uvd.num_enc_rings; ++i) { + ring = &adev->uvd.inst[j].ring_enc[i]; + sprintf(ring->name, "uvd_enc%d<%d>", i, j); + if (amdgpu_sriov_vf(adev)) { + ring->use_doorbell = true; + + /* currently only use the first enconding ring for + * sriov, so set unused location for other unused rings. + */ + if (i == 0) + ring->doorbell_index = AMDGPU_DOORBELL64_UVD_RING0_1 * 2; + else + ring->doorbell_index = AMDGPU_DOORBELL64_UVD_RING2_3 * 2 + 1; + } + r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.inst[j].irq, 0); + if (r) + return r; } - r = amdgpu_ring_init(adev, ring, 512, &adev->uvd.irq, 0); - if (r) - return r; } r = amdgpu_virt_alloc_mm_table(adev); @@ -463,7 +481,7 @@ static int uvd_v7_0_sw_init(void *handle) static int uvd_v7_0_sw_fini(void *handle) { - int i, r; + int i, j, r; struct amdgpu_device *adev = (struct amdgpu_device *)handle; amdgpu_virt_free_mm_table(adev); @@ -472,11 +490,12 @@ static int uvd_v7_0_sw_fini(void *handle) if (r) return r; - drm_sched_entity_fini(&adev->uvd.ring_enc[0].sched, &adev->uvd.entity_enc); - - for (i = 0; i < adev->uvd.num_enc_rings; ++i) - amdgpu_ring_fini(&adev->uvd.ring_enc[i]); + for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + drm_sched_entity_fini(&adev->uvd.inst[j].ring_enc[0].sched, &adev->uvd.inst[j].entity_enc); + for (i = 0; i < adev->uvd.num_enc_rings; ++i) + amdgpu_ring_fini(&adev->uvd.inst[j].ring_enc[i]); + } return amdgpu_uvd_sw_fini(adev); } @@ -490,9 +509,9 @@ static int uvd_v7_0_sw_fini(void *handle) static int uvd_v7_0_hw_init(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring; uint32_t tmp; - int i, r; + int i, j, r; if (amdgpu_sriov_vf(adev)) r = uvd_v7_0_sriov_start(adev); @@ -501,57 +520,60 @@ static int uvd_v7_0_hw_init(void *handle) if (r) goto done; - if (!amdgpu_sriov_vf(adev)) { - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; - goto done; + for (j = 0; j < adev->uvd.num_uvd_inst; ++j) { + ring = &adev->uvd.inst[j].ring; + + if (!amdgpu_sriov_vf(adev)) { + ring->ready = true; + r = amdgpu_ring_test_ring(ring); + if (r) { + ring->ready = false; + goto done; + } + + r = amdgpu_ring_alloc(ring, 10); + if (r) { + DRM_ERROR("amdgpu: (%d)ring failed to lock UVD ring (%d).\n", j, r); + goto done; + } + + tmp = PACKET0(SOC15_REG_OFFSET(UVD, j, + mmUVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL), 0); + amdgpu_ring_write(ring, tmp); + amdgpu_ring_write(ring, 0xFFFFF); + + tmp = PACKET0(SOC15_REG_OFFSET(UVD, j, + mmUVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL), 0); + amdgpu_ring_write(ring, tmp); + amdgpu_ring_write(ring, 0xFFFFF); + + tmp = PACKET0(SOC15_REG_OFFSET(UVD, j, + mmUVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL), 0); + amdgpu_ring_write(ring, tmp); + amdgpu_ring_write(ring, 0xFFFFF); + + /* Clear timeout status bits */ + amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, j, + mmUVD_SEMA_TIMEOUT_STATUS), 0)); + amdgpu_ring_write(ring, 0x8); + + amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, j, + mmUVD_SEMA_CNTL), 0)); + amdgpu_ring_write(ring, 3); + + amdgpu_ring_commit(ring); } - r = amdgpu_ring_alloc(ring, 10); - if (r) { - DRM_ERROR("amdgpu: ring failed to lock UVD ring (%d).\n", r); - goto done; + for (i = 0; i < adev->uvd.num_enc_rings; ++i) { + ring = &adev->uvd.inst[j].ring_enc[i]; + ring->ready = true; + r = amdgpu_ring_test_ring(ring); + if (r) { + ring->ready = false; + goto done; + } } - - tmp = PACKET0(SOC15_REG_OFFSET(UVD, 0, - mmUVD_SEMA_WAIT_FAULT_TIMEOUT_CNTL), 0); - amdgpu_ring_write(ring, tmp); - amdgpu_ring_write(ring, 0xFFFFF); - - tmp = PACKET0(SOC15_REG_OFFSET(UVD, 0, - mmUVD_SEMA_WAIT_INCOMPLETE_TIMEOUT_CNTL), 0); - amdgpu_ring_write(ring, tmp); - amdgpu_ring_write(ring, 0xFFFFF); - - tmp = PACKET0(SOC15_REG_OFFSET(UVD, 0, - mmUVD_SEMA_SIGNAL_INCOMPLETE_TIMEOUT_CNTL), 0); - amdgpu_ring_write(ring, tmp); - amdgpu_ring_write(ring, 0xFFFFF); - - /* Clear timeout status bits */ - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, - mmUVD_SEMA_TIMEOUT_STATUS), 0)); - amdgpu_ring_write(ring, 0x8); - - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, - mmUVD_SEMA_CNTL), 0)); - amdgpu_ring_write(ring, 3); - - amdgpu_ring_commit(ring); } - - for (i = 0; i < adev->uvd.num_enc_rings; ++i) { - ring = &adev->uvd.ring_enc[i]; - ring->ready = true; - r = amdgpu_ring_test_ring(ring); - if (r) { - ring->ready = false; - goto done; - } - } - done: if (!r) DRM_INFO("UVD and UVD ENC initialized successfully.\n"); @@ -569,7 +591,7 @@ done: static int uvd_v7_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - struct amdgpu_ring *ring = &adev->uvd.ring; + int i; if (!amdgpu_sriov_vf(adev)) uvd_v7_0_stop(adev); @@ -578,7 +600,8 @@ static int uvd_v7_0_hw_fini(void *handle) DRM_DEBUG("For SRIOV client, shouldn't do anything.\n"); } - ring->ready = false; + for (i = 0; i < adev->uvd.num_uvd_inst; ++i) + adev->uvd.inst[i].ring.ready = false; return 0; } @@ -618,48 +641,51 @@ static void uvd_v7_0_mc_resume(struct amdgpu_device *adev) { uint32_t size = AMDGPU_UVD_FIRMWARE_SIZE(adev); uint32_t offset; + int i; - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, - lower_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); - WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, - upper_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); - offset = 0; - } else { - WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, - lower_32_bits(adev->uvd.gpu_addr)); - WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, - upper_32_bits(adev->uvd.gpu_addr)); - offset = size; - } + for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, + lower_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); + WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, + upper_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); + offset = 0; + } else { + WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW, + lower_32_bits(adev->uvd.inst[i].gpu_addr)); + WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH, + upper_32_bits(adev->uvd.inst[i].gpu_addr)); + offset = size; + } - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0, - AMDGPU_UVD_FIRMWARE_OFFSET >> 3); - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE0, size); - - WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW, - lower_32_bits(adev->uvd.gpu_addr + offset)); - WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH, - upper_32_bits(adev->uvd.gpu_addr + offset)); - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1, (1 << 21)); - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_UVD_HEAP_SIZE); - - WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW, - lower_32_bits(adev->uvd.gpu_addr + offset + AMDGPU_UVD_HEAP_SIZE)); - WREG32_SOC15(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH, - upper_32_bits(adev->uvd.gpu_addr + offset + AMDGPU_UVD_HEAP_SIZE)); - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2, (2 << 21)); - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CACHE_SIZE2, - AMDGPU_UVD_STACK_SIZE + (AMDGPU_UVD_SESSION_SIZE * 40)); - - WREG32_SOC15(UVD, 0, mmUVD_UDEC_ADDR_CONFIG, - adev->gfx.config.gb_addr_config); - WREG32_SOC15(UVD, 0, mmUVD_UDEC_DB_ADDR_CONFIG, - adev->gfx.config.gb_addr_config); - WREG32_SOC15(UVD, 0, mmUVD_UDEC_DBW_ADDR_CONFIG, - adev->gfx.config.gb_addr_config); - - WREG32_SOC15(UVD, 0, mmUVD_GP_SCRATCH4, adev->uvd.max_handles); + WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_OFFSET0, + AMDGPU_UVD_FIRMWARE_OFFSET >> 3); + WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_SIZE0, size); + + WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW, + lower_32_bits(adev->uvd.inst[i].gpu_addr + offset)); + WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH, + upper_32_bits(adev->uvd.inst[i].gpu_addr + offset)); + WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_OFFSET1, (1 << 21)); + WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_SIZE1, AMDGPU_UVD_HEAP_SIZE); + + WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW, + lower_32_bits(adev->uvd.inst[i].gpu_addr + offset + AMDGPU_UVD_HEAP_SIZE)); + WREG32_SOC15(UVD, i, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH, + upper_32_bits(adev->uvd.inst[i].gpu_addr + offset + AMDGPU_UVD_HEAP_SIZE)); + WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_OFFSET2, (2 << 21)); + WREG32_SOC15(UVD, i, mmUVD_VCPU_CACHE_SIZE2, + AMDGPU_UVD_STACK_SIZE + (AMDGPU_UVD_SESSION_SIZE * 40)); + + WREG32_SOC15(UVD, i, mmUVD_UDEC_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, i, mmUVD_UDEC_DB_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + WREG32_SOC15(UVD, i, mmUVD_UDEC_DBW_ADDR_CONFIG, + adev->gfx.config.gb_addr_config); + + WREG32_SOC15(UVD, i, mmUVD_GP_SCRATCH4, adev->uvd.max_handles); + } } static int uvd_v7_0_mmsch_start(struct amdgpu_device *adev, @@ -669,6 +695,7 @@ static int uvd_v7_0_mmsch_start(struct amdgpu_device *adev, uint64_t addr = table->gpu_addr; struct mmsch_v1_0_init_header *header = (struct mmsch_v1_0_init_header *)table->cpu_addr; uint32_t size; + int i; size = header->header_size + header->vce_table_size + header->uvd_table_size; @@ -688,11 +715,12 @@ static int uvd_v7_0_mmsch_start(struct amdgpu_device *adev, /* 4, set resp to zero */ WREG32_SOC15(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_RESP, 0); - WDOORBELL32(adev->uvd.ring_enc[0].doorbell_index, 0); - adev->wb.wb[adev->uvd.ring_enc[0].wptr_offs] = 0; - adev->uvd.ring_enc[0].wptr = 0; - adev->uvd.ring_enc[0].wptr_old = 0; - + for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + WDOORBELL32(adev->uvd.inst[i].ring_enc[0].doorbell_index, 0); + adev->wb.wb[adev->uvd.inst[i].ring_enc[0].wptr_offs] = 0; + adev->uvd.inst[i].ring_enc[0].wptr = 0; + adev->uvd.inst[i].ring_enc[0].wptr_old = 0; + } /* 5, kick off the initialization and wait until VCE_MMSCH_VF_MAILBOX_RESP becomes non-zero */ WREG32_SOC15(VCE, 0, mmVCE_MMSCH_VF_MAILBOX_HOST, 0x10000001); @@ -725,6 +753,7 @@ static int uvd_v7_0_sriov_start(struct amdgpu_device *adev) struct mmsch_v1_0_cmd_end end = { {0} }; uint32_t *init_table = adev->virt.mm_table.cpu_addr; struct mmsch_v1_0_init_header *header = (struct mmsch_v1_0_init_header *)init_table; + uint8_t i = 0; direct_wt.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_WRITE; direct_rd_mod_wt.cmd_header.command_type = MMSCH_COMMAND__DIRECT_REG_READ_MODIFY_WRITE; @@ -742,120 +771,121 @@ static int uvd_v7_0_sriov_start(struct amdgpu_device *adev) init_table += header->uvd_table_offset; - ring = &adev->uvd.ring; - ring->wptr = 0; - size = AMDGPU_GPU_PAGE_ALIGN(adev->uvd.fw->size + 4); - - MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS), - 0xFFFFFFFF, 0x00000004); - /* mc resume*/ - if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), - lower_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), - upper_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); - offset = 0; - } else { - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), - lower_32_bits(adev->uvd.gpu_addr)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), - upper_32_bits(adev->uvd.gpu_addr)); - offset = size; + for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + ring = &adev->uvd.inst[i].ring; + ring->wptr = 0; + size = AMDGPU_GPU_PAGE_ALIGN(adev->uvd.fw->size + 4); + + MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_STATUS), + 0xFFFFFFFF, 0x00000004); + /* mc resume*/ + if (adev->firmware.load_type == AMDGPU_FW_LOAD_PSP) { + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), + lower_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), + upper_32_bits(adev->firmware.ucode[AMDGPU_UCODE_ID_UVD].mc_addr)); + offset = 0; + } else { + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_LOW), + lower_32_bits(adev->uvd.inst[i].gpu_addr)); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_VCPU_CACHE_64BIT_BAR_HIGH), + upper_32_bits(adev->uvd.inst[i].gpu_addr)); + offset = size; + } + + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_OFFSET0), + AMDGPU_UVD_FIRMWARE_OFFSET >> 3); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_SIZE0), size); + + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), + lower_32_bits(adev->uvd.inst[i].gpu_addr + offset)); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), + upper_32_bits(adev->uvd.inst[i].gpu_addr + offset)); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_OFFSET1), (1 << 21)); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_SIZE1), AMDGPU_UVD_HEAP_SIZE); + + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW), + lower_32_bits(adev->uvd.inst[i].gpu_addr + offset + AMDGPU_UVD_HEAP_SIZE)); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH), + upper_32_bits(adev->uvd.inst[i].gpu_addr + offset + AMDGPU_UVD_HEAP_SIZE)); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_OFFSET2), (2 << 21)); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CACHE_SIZE2), + AMDGPU_UVD_STACK_SIZE + (AMDGPU_UVD_SESSION_SIZE * 40)); + + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_GP_SCRATCH4), adev->uvd.max_handles); + /* mc resume end*/ + + /* disable clock gating */ + MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_CGC_CTRL), + ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK, 0); + + /* disable interupt */ + MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_MASTINT_EN), + ~UVD_MASTINT_EN__VCPU_EN_MASK, 0); + + /* stall UMC and register bus before resetting VCPU */ + MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_CTRL2), + ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, + UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); + + /* put LMI, VCPU, RBC etc... into reset */ + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_SOFT_RESET), + (uint32_t)(UVD_SOFT_RESET__LMI_SOFT_RESET_MASK | + UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK | + UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK | + UVD_SOFT_RESET__RBC_SOFT_RESET_MASK | + UVD_SOFT_RESET__CSM_SOFT_RESET_MASK | + UVD_SOFT_RESET__CXW_SOFT_RESET_MASK | + UVD_SOFT_RESET__TAP_SOFT_RESET_MASK | + UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK)); + + /* initialize UVD memory controller */ + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_CTRL), + (uint32_t)((0x40 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | + UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | + UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__REQ_MODE_MASK | + 0x00100000L)); + + /* take all subblocks out of reset, except VCPU */ + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_SOFT_RESET), + UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + + /* enable VCPU clock */ + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_VCPU_CNTL), + UVD_VCPU_CNTL__CLK_EN_MASK); + + /* enable master interrupt */ + MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_MASTINT_EN), + ~(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK), + (UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK)); + + /* clear the bit 4 of UVD_STATUS */ + MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_STATUS), + ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT), 0); + + /* force RBC into idle state */ + size = order_base_2(ring->ring_size); + tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, size); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_RBC_RB_CNTL), tmp); + + ring = &adev->uvd.inst[i].ring_enc[0]; + ring->wptr = 0; + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_RB_BASE_LO), ring->gpu_addr); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_RB_BASE_HI), upper_32_bits(ring->gpu_addr)); + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_RB_SIZE), ring->ring_size / 4); + + /* boot up the VCPU */ + MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_SOFT_RESET), 0); + + /* enable UMC */ + MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_CTRL2), + ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, 0); + + MMSCH_V1_0_INSERT_DIRECT_POLL(SOC15_REG_OFFSET(UVD, i, mmUVD_STATUS), 0x02, 0x02); } - - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_OFFSET0), - AMDGPU_UVD_FIRMWARE_OFFSET >> 3); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_SIZE0), size); - - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_LOW), - lower_32_bits(adev->uvd.gpu_addr + offset)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE1_64BIT_BAR_HIGH), - upper_32_bits(adev->uvd.gpu_addr + offset)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_OFFSET1), (1 << 21)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_SIZE1), AMDGPU_UVD_HEAP_SIZE); - - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_LOW), - lower_32_bits(adev->uvd.gpu_addr + offset + AMDGPU_UVD_HEAP_SIZE)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_VCPU_CACHE2_64BIT_BAR_HIGH), - upper_32_bits(adev->uvd.gpu_addr + offset + AMDGPU_UVD_HEAP_SIZE)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_OFFSET2), (2 << 21)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CACHE_SIZE2), - AMDGPU_UVD_STACK_SIZE + (AMDGPU_UVD_SESSION_SIZE * 40)); - - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_GP_SCRATCH4), adev->uvd.max_handles); - /* mc resume end*/ - - /* disable clock gating */ - MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_CGC_CTRL), - ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK, 0); - - /* disable interupt */ - MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), - ~UVD_MASTINT_EN__VCPU_EN_MASK, 0); - - /* stall UMC and register bus before resetting VCPU */ - MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), - ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, - UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); - - /* put LMI, VCPU, RBC etc... into reset */ - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), - (uint32_t)(UVD_SOFT_RESET__LMI_SOFT_RESET_MASK | - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK | - UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK | - UVD_SOFT_RESET__RBC_SOFT_RESET_MASK | - UVD_SOFT_RESET__CSM_SOFT_RESET_MASK | - UVD_SOFT_RESET__CXW_SOFT_RESET_MASK | - UVD_SOFT_RESET__TAP_SOFT_RESET_MASK | - UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK)); - - /* initialize UVD memory controller */ - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL), - (uint32_t)((0x40 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | - UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | - UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | - UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | - UVD_LMI_CTRL__REQ_MODE_MASK | - 0x00100000L)); - - /* take all subblocks out of reset, except VCPU */ - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); - - /* enable VCPU clock */ - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_VCPU_CNTL), - UVD_VCPU_CNTL__CLK_EN_MASK); - - /* enable master interrupt */ - MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), - ~(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK), - (UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK)); - - /* clear the bit 4 of UVD_STATUS */ - MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS), - ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT), 0); - - /* force RBC into idle state */ - size = order_base_2(ring->ring_size); - tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, size); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), tmp); - - ring = &adev->uvd.ring_enc[0]; - ring->wptr = 0; - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_BASE_LO), ring->gpu_addr); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_BASE_HI), upper_32_bits(ring->gpu_addr)); - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_RB_SIZE), ring->ring_size / 4); - - /* boot up the VCPU */ - MMSCH_V1_0_INSERT_DIRECT_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0); - - /* enable UMC */ - MMSCH_V1_0_INSERT_DIRECT_RD_MOD_WT(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), - ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, 0); - - MMSCH_V1_0_INSERT_DIRECT_POLL(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS), 0x02, 0x02); - /* add end packet */ memcpy((void *)init_table, &end, sizeof(struct mmsch_v1_0_cmd_end)); table_size += sizeof(struct mmsch_v1_0_cmd_end) / 4; @@ -874,15 +904,17 @@ static int uvd_v7_0_sriov_start(struct amdgpu_device *adev) */ static int uvd_v7_0_start(struct amdgpu_device *adev) { - struct amdgpu_ring *ring = &adev->uvd.ring; + struct amdgpu_ring *ring; uint32_t rb_bufsz, tmp; uint32_t lmi_swap_cntl; uint32_t mp_swap_cntl; - int i, j, r; + int i, j, k, r; - /* disable DPG */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_POWER_STATUS), 0, - ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + for (k = 0; k < adev->uvd.num_uvd_inst; ++k) { + /* disable DPG */ + WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_POWER_STATUS), 0, + ~UVD_POWER_STATUS__UVD_PG_MODE_MASK); + } /* disable byte swapping */ lmi_swap_cntl = 0; @@ -890,157 +922,159 @@ static int uvd_v7_0_start(struct amdgpu_device *adev) uvd_v7_0_mc_resume(adev); - /* disable clock gating */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_CGC_CTRL), 0, - ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK); - - /* disable interupt */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), 0, - ~UVD_MASTINT_EN__VCPU_EN_MASK); - - /* stall UMC and register bus before resetting VCPU */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), - UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, - ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); - mdelay(1); - - /* put LMI, VCPU, RBC etc... into reset */ - WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, - UVD_SOFT_RESET__LMI_SOFT_RESET_MASK | - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK | - UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK | - UVD_SOFT_RESET__RBC_SOFT_RESET_MASK | - UVD_SOFT_RESET__CSM_SOFT_RESET_MASK | - UVD_SOFT_RESET__CXW_SOFT_RESET_MASK | - UVD_SOFT_RESET__TAP_SOFT_RESET_MASK | - UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK); - mdelay(5); + for (k = 0; k < adev->uvd.num_uvd_inst; ++k) { + ring = &adev->uvd.inst[k].ring; + /* disable clock gating */ + WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_CGC_CTRL), 0, + ~UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK); - /* initialize UVD memory controller */ - WREG32_SOC15(UVD, 0, mmUVD_LMI_CTRL, - (0x40 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | - UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | - UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | - UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | - UVD_LMI_CTRL__REQ_MODE_MASK | - 0x00100000L); + /* disable interupt */ + WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_MASTINT_EN), 0, + ~UVD_MASTINT_EN__VCPU_EN_MASK); + + /* stall UMC and register bus before resetting VCPU */ + WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_LMI_CTRL2), + UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, + ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); + mdelay(1); + + /* put LMI, VCPU, RBC etc... into reset */ + WREG32_SOC15(UVD, k, mmUVD_SOFT_RESET, + UVD_SOFT_RESET__LMI_SOFT_RESET_MASK | + UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK | + UVD_SOFT_RESET__LBSI_SOFT_RESET_MASK | + UVD_SOFT_RESET__RBC_SOFT_RESET_MASK | + UVD_SOFT_RESET__CSM_SOFT_RESET_MASK | + UVD_SOFT_RESET__CXW_SOFT_RESET_MASK | + UVD_SOFT_RESET__TAP_SOFT_RESET_MASK | + UVD_SOFT_RESET__LMI_UMC_SOFT_RESET_MASK); + mdelay(5); + + /* initialize UVD memory controller */ + WREG32_SOC15(UVD, k, mmUVD_LMI_CTRL, + (0x40 << UVD_LMI_CTRL__WRITE_CLEAN_TIMER__SHIFT) | + UVD_LMI_CTRL__WRITE_CLEAN_TIMER_EN_MASK | + UVD_LMI_CTRL__DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__VCPU_DATA_COHERENCY_EN_MASK | + UVD_LMI_CTRL__REQ_MODE_MASK | + 0x00100000L); #ifdef __BIG_ENDIAN - /* swap (8 in 32) RB and IB */ - lmi_swap_cntl = 0xa; - mp_swap_cntl = 0; + /* swap (8 in 32) RB and IB */ + lmi_swap_cntl = 0xa; + mp_swap_cntl = 0; #endif - WREG32_SOC15(UVD, 0, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl); - WREG32_SOC15(UVD, 0, mmUVD_MP_SWAP_CNTL, mp_swap_cntl); - - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXA0, 0x40c2040); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXA1, 0x0); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXB0, 0x40c2040); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUXB1, 0x0); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_ALU, 0); - WREG32_SOC15(UVD, 0, mmUVD_MPC_SET_MUX, 0x88); - - /* take all subblocks out of reset, except VCPU */ - WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); - mdelay(5); + WREG32_SOC15(UVD, k, mmUVD_LMI_SWAP_CNTL, lmi_swap_cntl); + WREG32_SOC15(UVD, k, mmUVD_MP_SWAP_CNTL, mp_swap_cntl); - /* enable VCPU clock */ - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CNTL, - UVD_VCPU_CNTL__CLK_EN_MASK); + WREG32_SOC15(UVD, k, mmUVD_MPC_SET_MUXA0, 0x40c2040); + WREG32_SOC15(UVD, k, mmUVD_MPC_SET_MUXA1, 0x0); + WREG32_SOC15(UVD, k, mmUVD_MPC_SET_MUXB0, 0x40c2040); + WREG32_SOC15(UVD, k, mmUVD_MPC_SET_MUXB1, 0x0); + WREG32_SOC15(UVD, k, mmUVD_MPC_SET_ALU, 0); + WREG32_SOC15(UVD, k, mmUVD_MPC_SET_MUX, 0x88); - /* enable UMC */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), 0, - ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); + /* take all subblocks out of reset, except VCPU */ + WREG32_SOC15(UVD, k, mmUVD_SOFT_RESET, + UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + mdelay(5); - /* boot up the VCPU */ - WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, 0); - mdelay(10); + /* enable VCPU clock */ + WREG32_SOC15(UVD, k, mmUVD_VCPU_CNTL, + UVD_VCPU_CNTL__CLK_EN_MASK); - for (i = 0; i < 10; ++i) { - uint32_t status; + /* enable UMC */ + WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_LMI_CTRL2), 0, + ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); - for (j = 0; j < 100; ++j) { - status = RREG32_SOC15(UVD, 0, mmUVD_STATUS); + /* boot up the VCPU */ + WREG32_SOC15(UVD, k, mmUVD_SOFT_RESET, 0); + mdelay(10); + + for (i = 0; i < 10; ++i) { + uint32_t status; + + for (j = 0; j < 100; ++j) { + status = RREG32_SOC15(UVD, k, mmUVD_STATUS); + if (status & 2) + break; + mdelay(10); + } + r = 0; if (status & 2) break; + + DRM_ERROR("UVD(%d) not responding, trying to reset the VCPU!!!\n", k); + WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_SOFT_RESET), + UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK, + ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); mdelay(10); + WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_SOFT_RESET), 0, + ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + mdelay(10); + r = -1; } - r = 0; - if (status & 2) - break; - - DRM_ERROR("UVD not responding, trying to reset the VCPU!!!\n"); - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK, - ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); - mdelay(10); - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_SOFT_RESET), 0, - ~UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); - mdelay(10); - r = -1; - } - - if (r) { - DRM_ERROR("UVD not responding, giving up!!!\n"); - return r; - } - /* enable master interrupt */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), - (UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK), - ~(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK)); - - /* clear the bit 4 of UVD_STATUS */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_STATUS), 0, - ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT)); - - /* force RBC into idle state */ - rb_bufsz = order_base_2(ring->ring_size); - tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_WPTR_POLL_EN, 0); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1); - tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1); - WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, tmp); - - /* set the write pointer delay */ - WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR_CNTL, 0); - - /* set the wb address */ - WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR_ADDR, - (upper_32_bits(ring->gpu_addr) >> 2)); - - /* programm the RB_BASE for ring buffer */ - WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW, - lower_32_bits(ring->gpu_addr)); - WREG32_SOC15(UVD, 0, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH, - upper_32_bits(ring->gpu_addr)); - - /* Initialize the ring buffer's read and write pointers */ - WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR, 0); - - ring->wptr = RREG32_SOC15(UVD, 0, mmUVD_RBC_RB_RPTR); - WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_WPTR, - lower_32_bits(ring->wptr)); - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_RB_CNTL), 0, - ~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK); - - ring = &adev->uvd.ring_enc[0]; - WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); - WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); - WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO, ring->gpu_addr); - WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); - WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE, ring->ring_size / 4); + if (r) { + DRM_ERROR("UVD(%d) not responding, giving up!!!\n", k); + return r; + } + /* enable master interrupt */ + WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_MASTINT_EN), + (UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK), + ~(UVD_MASTINT_EN__VCPU_EN_MASK|UVD_MASTINT_EN__SYS_EN_MASK)); - ring = &adev->uvd.ring_enc[1]; - WREG32_SOC15(UVD, 0, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); - WREG32_SOC15(UVD, 0, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); - WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_LO2, ring->gpu_addr); - WREG32_SOC15(UVD, 0, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); - WREG32_SOC15(UVD, 0, mmUVD_RB_SIZE2, ring->ring_size / 4); + /* clear the bit 4 of UVD_STATUS */ + WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_STATUS), 0, + ~(2 << UVD_STATUS__VCPU_REPORT__SHIFT)); + /* force RBC into idle state */ + rb_bufsz = order_base_2(ring->ring_size); + tmp = REG_SET_FIELD(0, UVD_RBC_RB_CNTL, RB_BUFSZ, rb_bufsz); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_BLKSZ, 1); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_FETCH, 1); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_WPTR_POLL_EN, 0); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_NO_UPDATE, 1); + tmp = REG_SET_FIELD(tmp, UVD_RBC_RB_CNTL, RB_RPTR_WR_EN, 1); + WREG32_SOC15(UVD, k, mmUVD_RBC_RB_CNTL, tmp); + + /* set the write pointer delay */ + WREG32_SOC15(UVD, k, mmUVD_RBC_RB_WPTR_CNTL, 0); + + /* set the wb address */ + WREG32_SOC15(UVD, k, mmUVD_RBC_RB_RPTR_ADDR, + (upper_32_bits(ring->gpu_addr) >> 2)); + + /* programm the RB_BASE for ring buffer */ + WREG32_SOC15(UVD, k, mmUVD_LMI_RBC_RB_64BIT_BAR_LOW, + lower_32_bits(ring->gpu_addr)); + WREG32_SOC15(UVD, k, mmUVD_LMI_RBC_RB_64BIT_BAR_HIGH, + upper_32_bits(ring->gpu_addr)); + + /* Initialize the ring buffer's read and write pointers */ + WREG32_SOC15(UVD, k, mmUVD_RBC_RB_RPTR, 0); + + ring->wptr = RREG32_SOC15(UVD, k, mmUVD_RBC_RB_RPTR); + WREG32_SOC15(UVD, k, mmUVD_RBC_RB_WPTR, + lower_32_bits(ring->wptr)); + + WREG32_P(SOC15_REG_OFFSET(UVD, k, mmUVD_RBC_RB_CNTL), 0, + ~UVD_RBC_RB_CNTL__RB_NO_FETCH_MASK); + + ring = &adev->uvd.inst[k].ring_enc[0]; + WREG32_SOC15(UVD, k, mmUVD_RB_RPTR, lower_32_bits(ring->wptr)); + WREG32_SOC15(UVD, k, mmUVD_RB_WPTR, lower_32_bits(ring->wptr)); + WREG32_SOC15(UVD, k, mmUVD_RB_BASE_LO, ring->gpu_addr); + WREG32_SOC15(UVD, k, mmUVD_RB_BASE_HI, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(UVD, k, mmUVD_RB_SIZE, ring->ring_size / 4); + + ring = &adev->uvd.inst[k].ring_enc[1]; + WREG32_SOC15(UVD, k, mmUVD_RB_RPTR2, lower_32_bits(ring->wptr)); + WREG32_SOC15(UVD, k, mmUVD_RB_WPTR2, lower_32_bits(ring->wptr)); + WREG32_SOC15(UVD, k, mmUVD_RB_BASE_LO2, ring->gpu_addr); + WREG32_SOC15(UVD, k, mmUVD_RB_BASE_HI2, upper_32_bits(ring->gpu_addr)); + WREG32_SOC15(UVD, k, mmUVD_RB_SIZE2, ring->ring_size / 4); + } return 0; } @@ -1053,26 +1087,30 @@ static int uvd_v7_0_start(struct amdgpu_device *adev) */ static void uvd_v7_0_stop(struct amdgpu_device *adev) { - /* force RBC into idle state */ - WREG32_SOC15(UVD, 0, mmUVD_RBC_RB_CNTL, 0x11010101); - - /* Stall UMC and register bus before resetting VCPU */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), - UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, - ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); - mdelay(1); - - /* put VCPU into reset */ - WREG32_SOC15(UVD, 0, mmUVD_SOFT_RESET, - UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); - mdelay(5); + uint8_t i = 0; - /* disable VCPU clock */ - WREG32_SOC15(UVD, 0, mmUVD_VCPU_CNTL, 0x0); + for (i = 0; i < adev->uvd.num_uvd_inst; ++i) { + /* force RBC into idle state */ + WREG32_SOC15(UVD, i, mmUVD_RBC_RB_CNTL, 0x11010101); - /* Unstall UMC and register bus */ - WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), 0, - ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); + /* Stall UMC and register bus before resetting VCPU */ + WREG32_P(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_CTRL2), + UVD_LMI_CTRL2__STALL_ARB_UMC_MASK, + ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); + mdelay(1); + + /* put VCPU into reset */ + WREG32_SOC15(UVD, i, mmUVD_SOFT_RESET, + UVD_SOFT_RESET__VCPU_SOFT_RESET_MASK); + mdelay(5); + + /* disable VCPU clock */ + WREG32_SOC15(UVD, i, mmUVD_VCPU_CNTL, 0x0); + + /* Unstall UMC and register bus */ + WREG32_P(SOC15_REG_OFFSET(UVD, i, mmUVD_LMI_CTRL2), 0, + ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); + } } /** @@ -1091,26 +1129,26 @@ static void uvd_v7_0_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, u64 seq WARN_ON(flags & AMDGPU_FENCE_FLAG_64BIT); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_CONTEXT_ID), 0)); amdgpu_ring_write(ring, seq); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_DATA0), 0)); amdgpu_ring_write(ring, addr & 0xffffffff); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_DATA1), 0)); amdgpu_ring_write(ring, upper_32_bits(addr) & 0xff); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_CMD), 0)); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_DATA0), 0)); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_DATA1), 0)); amdgpu_ring_write(ring, 0); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_CMD), 0)); amdgpu_ring_write(ring, 2); } @@ -1136,6 +1174,16 @@ static void uvd_v7_0_enc_ring_emit_fence(struct amdgpu_ring *ring, u64 addr, } /** + * uvd_v7_0_ring_emit_hdp_flush - skip HDP flushing + * + * @ring: amdgpu_ring pointer + */ +static void uvd_v7_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) +{ + /* The firmware doesn't seem to like touching registers at this point. */ +} + +/** * uvd_v7_0_ring_test_ring - register write test * * @ring: amdgpu_ring pointer @@ -1149,30 +1197,30 @@ static int uvd_v7_0_ring_test_ring(struct amdgpu_ring *ring) unsigned i; int r; - WREG32_SOC15(UVD, 0, mmUVD_CONTEXT_ID, 0xCAFEDEAD); + WREG32_SOC15(UVD, ring->me, mmUVD_CONTEXT_ID, 0xCAFEDEAD); r = amdgpu_ring_alloc(ring, 3); if (r) { - DRM_ERROR("amdgpu: cp failed to lock ring %d (%d).\n", - ring->idx, r); + DRM_ERROR("amdgpu: (%d)cp failed to lock ring %d (%d).\n", + ring->me, ring->idx, r); return r; } amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_CONTEXT_ID), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_CONTEXT_ID), 0)); amdgpu_ring_write(ring, 0xDEADBEEF); amdgpu_ring_commit(ring); for (i = 0; i < adev->usec_timeout; i++) { - tmp = RREG32_SOC15(UVD, 0, mmUVD_CONTEXT_ID); + tmp = RREG32_SOC15(UVD, ring->me, mmUVD_CONTEXT_ID); if (tmp == 0xDEADBEEF) break; DRM_UDELAY(1); } if (i < adev->usec_timeout) { - DRM_DEBUG("ring test on %d succeeded in %d usecs\n", - ring->idx, i); + DRM_DEBUG("(%d)ring test on %d succeeded in %d usecs\n", + ring->me, ring->idx, i); } else { - DRM_ERROR("amdgpu: ring %d test failed (0x%08X)\n", - ring->idx, tmp); + DRM_ERROR("(%d)amdgpu: ring %d test failed (0x%08X)\n", + ring->me, ring->idx, tmp); r = -EINVAL; } return r; @@ -1193,17 +1241,17 @@ static void uvd_v7_0_ring_emit_ib(struct amdgpu_ring *ring, struct amdgpu_device *adev = ring->adev; amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_VMID), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_LMI_RBC_IB_VMID), 0)); amdgpu_ring_write(ring, vmid); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_LOW), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_LMI_RBC_IB_64BIT_BAR_LOW), 0)); amdgpu_ring_write(ring, lower_32_bits(ib->gpu_addr)); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_LMI_RBC_IB_64BIT_BAR_HIGH), 0)); amdgpu_ring_write(ring, upper_32_bits(ib->gpu_addr)); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_RBC_IB_SIZE), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_RBC_IB_SIZE), 0)); amdgpu_ring_write(ring, ib->length_dw); } @@ -1231,13 +1279,13 @@ static void uvd_v7_0_ring_emit_wreg(struct amdgpu_ring *ring, struct amdgpu_device *adev = ring->adev; amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_DATA0), 0)); amdgpu_ring_write(ring, reg << 2); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_DATA1), 0)); amdgpu_ring_write(ring, val); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_CMD), 0)); amdgpu_ring_write(ring, 8); } @@ -1247,16 +1295,16 @@ static void uvd_v7_0_ring_emit_reg_wait(struct amdgpu_ring *ring, uint32_t reg, struct amdgpu_device *adev = ring->adev; amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA0), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_DATA0), 0)); amdgpu_ring_write(ring, reg << 2); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_DATA1), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_DATA1), 0)); amdgpu_ring_write(ring, val); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GP_SCRATCH8), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GP_SCRATCH8), 0)); amdgpu_ring_write(ring, mask); amdgpu_ring_write(ring, - PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_GPCOM_VCPU_CMD), 0)); + PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_GPCOM_VCPU_CMD), 0)); amdgpu_ring_write(ring, 12); } @@ -1277,12 +1325,15 @@ static void uvd_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring, static void uvd_v7_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) { - int i; struct amdgpu_device *adev = ring->adev; + int i; - for (i = 0; i < count; i++) - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP), 0)); + WARN_ON(ring->wptr % 2 || count % 2); + for (i = 0; i < count / 2; i++) { + amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, ring->me, mmUVD_NO_OP), 0)); + amdgpu_ring_write(ring, 0); + } } static void uvd_v7_0_enc_ring_insert_end(struct amdgpu_ring *ring) @@ -1349,16 +1400,16 @@ static bool uvd_v7_0_check_soft_reset(void *handle) if (REG_GET_FIELD(tmp, SRBM_STATUS, UVD_RQ_PENDING) || REG_GET_FIELD(tmp, SRBM_STATUS, UVD_BUSY) || - (RREG32_SOC15(UVD, 0, mmUVD_STATUS) & + (RREG32_SOC15(UVD, ring->me, mmUVD_STATUS) & AMDGPU_UVD_STATUS_BUSY_MASK)) srbm_soft_reset = REG_SET_FIELD(srbm_soft_reset, SRBM_SOFT_RESET, SOFT_RESET_UVD, 1); if (srbm_soft_reset) { - adev->uvd.srbm_soft_reset = srbm_soft_reset; + adev->uvd.inst[ring->me].srbm_soft_reset = srbm_soft_reset; return true; } else { - adev->uvd.srbm_soft_reset = 0; + adev->uvd.inst[ring->me].srbm_soft_reset = 0; return false; } } @@ -1367,7 +1418,7 @@ static int uvd_v7_0_pre_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (!adev->uvd.srbm_soft_reset) + if (!adev->uvd.inst[ring->me].srbm_soft_reset) return 0; uvd_v7_0_stop(adev); @@ -1379,9 +1430,9 @@ static int uvd_v7_0_soft_reset(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; u32 srbm_soft_reset; - if (!adev->uvd.srbm_soft_reset) + if (!adev->uvd.inst[ring->me].srbm_soft_reset) return 0; - srbm_soft_reset = adev->uvd.srbm_soft_reset; + srbm_soft_reset = adev->uvd.inst[ring->me].srbm_soft_reset; if (srbm_soft_reset) { u32 tmp; @@ -1409,7 +1460,7 @@ static int uvd_v7_0_post_soft_reset(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; - if (!adev->uvd.srbm_soft_reset) + if (!adev->uvd.inst[ring->me].srbm_soft_reset) return 0; mdelay(5); @@ -1431,17 +1482,32 @@ static int uvd_v7_0_process_interrupt(struct amdgpu_device *adev, struct amdgpu_irq_src *source, struct amdgpu_iv_entry *entry) { + uint32_t ip_instance; + + switch (entry->client_id) { + case SOC15_IH_CLIENTID_UVD: + ip_instance = 0; + break; + case SOC15_IH_CLIENTID_UVD1: + ip_instance = 1; + break; + default: + DRM_ERROR("Unhandled client id: %d\n", entry->client_id); + return 0; + } + DRM_DEBUG("IH: UVD TRAP\n"); + switch (entry->src_id) { case 124: - amdgpu_fence_process(&adev->uvd.ring); + amdgpu_fence_process(&adev->uvd.inst[ip_instance].ring); break; case 119: - amdgpu_fence_process(&adev->uvd.ring_enc[0]); + amdgpu_fence_process(&adev->uvd.inst[ip_instance].ring_enc[0]); break; case 120: if (!amdgpu_sriov_vf(adev)) - amdgpu_fence_process(&adev->uvd.ring_enc[1]); + amdgpu_fence_process(&adev->uvd.inst[ip_instance].ring_enc[1]); break; default: DRM_ERROR("Unhandled interrupt: %d %d\n", @@ -1457,9 +1523,9 @@ static void uvd_v7_0_set_sw_clock_gating(struct amdgpu_device *adev) { uint32_t data, data1, data2, suvd_flags; - data = RREG32_SOC15(UVD, 0, mmUVD_CGC_CTRL); - data1 = RREG32_SOC15(UVD, 0, mmUVD_SUVD_CGC_GATE); - data2 = RREG32_SOC15(UVD, 0, mmUVD_SUVD_CGC_CTRL); + data = RREG32_SOC15(UVD, ring->me, mmUVD_CGC_CTRL); + data1 = RREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_GATE); + data2 = RREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_CTRL); data &= ~(UVD_CGC_CTRL__CLK_OFF_DELAY_MASK | UVD_CGC_CTRL__CLK_GATE_DLY_TIMER_MASK); @@ -1503,18 +1569,18 @@ static void uvd_v7_0_set_sw_clock_gating(struct amdgpu_device *adev) UVD_SUVD_CGC_CTRL__SDB_MODE_MASK); data1 |= suvd_flags; - WREG32_SOC15(UVD, 0, mmUVD_CGC_CTRL, data); - WREG32_SOC15(UVD, 0, mmUVD_CGC_GATE, 0); - WREG32_SOC15(UVD, 0, mmUVD_SUVD_CGC_GATE, data1); - WREG32_SOC15(UVD, 0, mmUVD_SUVD_CGC_CTRL, data2); + WREG32_SOC15(UVD, ring->me, mmUVD_CGC_CTRL, data); + WREG32_SOC15(UVD, ring->me, mmUVD_CGC_GATE, 0); + WREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_GATE, data1); + WREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_CTRL, data2); } static void uvd_v7_0_set_hw_clock_gating(struct amdgpu_device *adev) { uint32_t data, data1, cgc_flags, suvd_flags; - data = RREG32_SOC15(UVD, 0, mmUVD_CGC_GATE); - data1 = RREG32_SOC15(UVD, 0, mmUVD_SUVD_CGC_GATE); + data = RREG32_SOC15(UVD, ring->me, mmUVD_CGC_GATE); + data1 = RREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_GATE); cgc_flags = UVD_CGC_GATE__SYS_MASK | UVD_CGC_GATE__UDEC_MASK | @@ -1546,8 +1612,8 @@ static void uvd_v7_0_set_hw_clock_gating(struct amdgpu_device *adev) data |= cgc_flags; data1 |= suvd_flags; - WREG32_SOC15(UVD, 0, mmUVD_CGC_GATE, data); - WREG32_SOC15(UVD, 0, mmUVD_SUVD_CGC_GATE, data1); + WREG32_SOC15(UVD, ring->me, mmUVD_CGC_GATE, data); + WREG32_SOC15(UVD, ring->me, mmUVD_SUVD_CGC_GATE, data1); } static void uvd_v7_0_set_bypass_mode(struct amdgpu_device *adev, bool enable) @@ -1606,7 +1672,7 @@ static int uvd_v7_0_set_powergating_state(void *handle, if (!(adev->pg_flags & AMD_PG_SUPPORT_UVD)) return 0; - WREG32_SOC15(UVD, 0, mmUVD_POWER_STATUS, UVD_POWER_STATUS__UVD_PG_EN_MASK); + WREG32_SOC15(UVD, ring->me, mmUVD_POWER_STATUS, UVD_POWER_STATUS__UVD_PG_EN_MASK); if (state == AMD_PG_STATE_GATE) { uvd_v7_0_stop(adev); @@ -1647,14 +1713,13 @@ const struct amd_ip_funcs uvd_v7_0_ip_funcs = { static const struct amdgpu_ring_funcs uvd_v7_0_ring_vm_funcs = { .type = AMDGPU_RING_TYPE_UVD, .align_mask = 0xf, - .nop = PACKET0(0x81ff, 0), .support_64bit_ptrs = false, .vmhub = AMDGPU_MMHUB, .get_rptr = uvd_v7_0_ring_get_rptr, .get_wptr = uvd_v7_0_ring_get_wptr, .set_wptr = uvd_v7_0_ring_set_wptr, .emit_frame_size = - 6 + 6 + /* hdp flush / invalidate */ + 6 + /* hdp invalidate */ SOC15_FLUSH_GPU_TLB_NUM_WREG * 6 + SOC15_FLUSH_GPU_TLB_NUM_REG_WAIT * 8 + 8 + /* uvd_v7_0_ring_emit_vm_flush */ @@ -1663,6 +1728,7 @@ static const struct amdgpu_ring_funcs uvd_v7_0_ring_vm_funcs = { .emit_ib = uvd_v7_0_ring_emit_ib, .emit_fence = uvd_v7_0_ring_emit_fence, .emit_vm_flush = uvd_v7_0_ring_emit_vm_flush, + .emit_hdp_flush = uvd_v7_0_ring_emit_hdp_flush, .test_ring = uvd_v7_0_ring_test_ring, .test_ib = amdgpu_uvd_ring_test_ib, .insert_nop = uvd_v7_0_ring_insert_nop, @@ -1671,6 +1737,7 @@ static const struct amdgpu_ring_funcs uvd_v7_0_ring_vm_funcs = { .end_use = amdgpu_uvd_ring_end_use, .emit_wreg = uvd_v7_0_ring_emit_wreg, .emit_reg_wait = uvd_v7_0_ring_emit_reg_wait, + .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, }; static const struct amdgpu_ring_funcs uvd_v7_0_enc_ring_vm_funcs = { @@ -1702,22 +1769,32 @@ static const struct amdgpu_ring_funcs uvd_v7_0_enc_ring_vm_funcs = { .end_use = amdgpu_uvd_ring_end_use, .emit_wreg = uvd_v7_0_enc_ring_emit_wreg, .emit_reg_wait = uvd_v7_0_enc_ring_emit_reg_wait, + .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, }; static void uvd_v7_0_set_ring_funcs(struct amdgpu_device *adev) { - adev->uvd.ring.funcs = &uvd_v7_0_ring_vm_funcs; - DRM_INFO("UVD is enabled in VM mode\n"); + int i; + + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + adev->uvd.inst[i].ring.funcs = &uvd_v7_0_ring_vm_funcs; + adev->uvd.inst[i].ring.me = i; + DRM_INFO("UVD(%d) is enabled in VM mode\n", i); + } } static void uvd_v7_0_set_enc_ring_funcs(struct amdgpu_device *adev) { - int i; + int i, j; - for (i = 0; i < adev->uvd.num_enc_rings; ++i) - adev->uvd.ring_enc[i].funcs = &uvd_v7_0_enc_ring_vm_funcs; + for (j = 0; j < adev->uvd.num_uvd_inst; j++) { + for (i = 0; i < adev->uvd.num_enc_rings; ++i) { + adev->uvd.inst[j].ring_enc[i].funcs = &uvd_v7_0_enc_ring_vm_funcs; + adev->uvd.inst[j].ring_enc[i].me = j; + } - DRM_INFO("UVD ENC is enabled in VM mode\n"); + DRM_INFO("UVD(%d) ENC is enabled in VM mode\n", j); + } } static const struct amdgpu_irq_src_funcs uvd_v7_0_irq_funcs = { @@ -1727,8 +1804,12 @@ static const struct amdgpu_irq_src_funcs uvd_v7_0_irq_funcs = { static void uvd_v7_0_set_irq_funcs(struct amdgpu_device *adev) { - adev->uvd.irq.num_types = adev->uvd.num_enc_rings + 1; - adev->uvd.irq.funcs = &uvd_v7_0_irq_funcs; + int i; + + for (i = 0; i < adev->uvd.num_uvd_inst; i++) { + adev->uvd.inst[i].irq.num_types = adev->uvd.num_enc_rings + 1; + adev->uvd.inst[i].irq.funcs = &uvd_v7_0_irq_funcs; + } } const struct amdgpu_ip_block_version uvd_v7_0_ip_block = diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c index 428d1928e44e2..0999c843f623c 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v3_0.c @@ -388,7 +388,8 @@ static unsigned vce_v3_0_get_harvest_config(struct amdgpu_device *adev) default: if ((adev->asic_type == CHIP_POLARIS10) || (adev->asic_type == CHIP_POLARIS11) || - (adev->asic_type == CHIP_POLARIS12)) + (adev->asic_type == CHIP_POLARIS12) || + (adev->asic_type == CHIP_VEGAM)) return AMDGPU_VCE_HARVEST_VCE1; return 0; @@ -467,8 +468,8 @@ static int vce_v3_0_hw_init(void *handle) struct amdgpu_device *adev = (struct amdgpu_device *)handle; vce_v3_0_override_vce_clock_gating(adev, true); - if (!(adev->flags & AMD_IS_APU)) - amdgpu_asic_set_vce_clocks(adev, 10000, 10000); + + amdgpu_asic_set_vce_clocks(adev, 10000, 10000); for (i = 0; i < adev->vce.num_rings; i++) adev->vce.ring[i].ready = false; diff --git a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c index 73fd48d6c7566..8fd1b742985ac 100644 --- a/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vce_v4_0.c @@ -1081,6 +1081,7 @@ static const struct amdgpu_ring_funcs vce_v4_0_ring_vm_funcs = { .end_use = amdgpu_vce_ring_end_use, .emit_wreg = vce_v4_0_emit_wreg, .emit_reg_wait = vce_v4_0_emit_reg_wait, + .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, }; static void vce_v4_0_set_ring_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c index 8c132673bc794..110b294ebed3e 100644 --- a/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c +++ b/drivers/gpu/drm/amd/amdgpu/vcn_v1_0.c @@ -35,7 +35,6 @@ #include "mmhub/mmhub_9_1_offset.h" #include "mmhub/mmhub_9_1_sh_mask.h" -static int vcn_v1_0_start(struct amdgpu_device *adev); static int vcn_v1_0_stop(struct amdgpu_device *adev); static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev); static void vcn_v1_0_set_enc_ring_funcs(struct amdgpu_device *adev); @@ -146,10 +145,6 @@ static int vcn_v1_0_hw_init(void *handle) struct amdgpu_ring *ring = &adev->vcn.ring_dec; int i, r; - r = vcn_v1_0_start(adev); - if (r) - goto done; - ring->ready = true; r = amdgpu_ring_test_ring(ring); if (r) { @@ -185,11 +180,9 @@ static int vcn_v1_0_hw_fini(void *handle) { struct amdgpu_device *adev = (struct amdgpu_device *)handle; struct amdgpu_ring *ring = &adev->vcn.ring_dec; - int r; - r = vcn_v1_0_stop(adev); - if (r) - return r; + if (RREG32_SOC15(VCN, 0, mmUVD_STATUS)) + vcn_v1_0_stop(adev); ring->ready = false; @@ -288,14 +281,14 @@ static void vcn_v1_0_mc_resume(struct amdgpu_device *adev) * * Disable clock gating for VCN block */ -static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev, bool sw) +static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev) { uint32_t data; /* JPEG disable CGC */ data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL); - if (sw) + if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; else data &= ~JPEG_CGC_CTRL__DYN_CLOCK_MODE_MASK; @@ -310,7 +303,7 @@ static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev, bool sw) /* UVD disable CGC */ data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL); - if (sw) + if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; else data &= ~ UVD_CGC_CTRL__DYN_CLOCK_MODE_MASK; @@ -415,13 +408,13 @@ static void vcn_v1_0_disable_clock_gating(struct amdgpu_device *adev, bool sw) * * Enable clock gating for VCN block */ -static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev, bool sw) +static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev) { uint32_t data = 0; /* enable JPEG CGC */ data = RREG32_SOC15(VCN, 0, mmJPEG_CGC_CTRL); - if (sw) + if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) data |= 1 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; else data |= 0 << JPEG_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; @@ -435,7 +428,7 @@ static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev, bool sw) /* enable UVD CGC */ data = RREG32_SOC15(VCN, 0, mmUVD_CGC_CTRL); - if (sw) + if (adev->cg_flags & AMD_CG_SUPPORT_VCN_MGCG) data |= 1 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; else data |= 0 << UVD_CGC_CTRL__DYN_CLOCK_MODE__SHIFT; @@ -480,6 +473,94 @@ static void vcn_v1_0_enable_clock_gating(struct amdgpu_device *adev, bool sw) WREG32_SOC15(VCN, 0, mmUVD_SUVD_CGC_CTRL, data); } +static void vcn_1_0_disable_static_power_gating(struct amdgpu_device *adev) +{ + uint32_t data = 0; + int ret; + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN) { + data = (1 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT); + + WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data); + SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, UVD_PGFSM_STATUS__UVDM_UVDU_PWR_ON, 0xFFFFFF, ret); + } else { + data = (1 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT + | 1 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT); + WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data); + SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, 0, 0xFFFFFFFF, ret); + } + + /* polling UVD_PGFSM_STATUS to confirm UVDM_PWR_STATUS , UVDU_PWR_STATUS are 0 (power on) */ + + data = RREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS); + data &= ~0x103; + if (adev->pg_flags & AMD_PG_SUPPORT_VCN) + data |= UVD_PGFSM_CONFIG__UVDM_UVDU_PWR_ON | UVD_POWER_STATUS__UVD_PG_EN_MASK; + + WREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS, data); +} + +static void vcn_1_0_enable_static_power_gating(struct amdgpu_device *adev) +{ + uint32_t data = 0; + int ret; + + if (adev->pg_flags & AMD_PG_SUPPORT_VCN) { + /* Before power off, this indicator has to be turned on */ + data = RREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS); + data &= ~UVD_POWER_STATUS__UVD_POWER_STATUS_MASK; + data |= UVD_POWER_STATUS__UVD_POWER_STATUS_TILES_OFF; + WREG32_SOC15(VCN, 0, mmUVD_POWER_STATUS, data); + + + data = (2 << UVD_PGFSM_CONFIG__UVDM_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDU_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDF_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDC_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDB_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDIL_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDIR_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTD_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDTE_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDE_PWR_CONFIG__SHIFT + | 2 << UVD_PGFSM_CONFIG__UVDW_PWR_CONFIG__SHIFT); + + WREG32_SOC15(VCN, 0, mmUVD_PGFSM_CONFIG, data); + + data = (2 << UVD_PGFSM_STATUS__UVDM_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDU_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDF_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDC_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDB_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDIL_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDIR_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDTD_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDTE_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDE_PWR_STATUS__SHIFT + | 2 << UVD_PGFSM_STATUS__UVDW_PWR_STATUS__SHIFT); + SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_PGFSM_STATUS, data, 0xFFFFFFFF, ret); + } +} + /** * vcn_v1_0_start - start VCN block * @@ -499,8 +580,9 @@ static int vcn_v1_0_start(struct amdgpu_device *adev) vcn_v1_0_mc_resume(adev); + vcn_1_0_disable_static_power_gating(adev); /* disable clock gating */ - vcn_v1_0_disable_clock_gating(adev, true); + vcn_v1_0_disable_clock_gating(adev); /* disable interupt */ WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_MASTINT_EN), 0, @@ -680,16 +762,45 @@ static int vcn_v1_0_stop(struct amdgpu_device *adev) WREG32_P(SOC15_REG_OFFSET(UVD, 0, mmUVD_LMI_CTRL2), 0, ~UVD_LMI_CTRL2__STALL_ARB_UMC_MASK); - /* enable clock gating */ - vcn_v1_0_enable_clock_gating(adev, true); + WREG32_SOC15(VCN, 0, mmUVD_STATUS, 0); + vcn_v1_0_enable_clock_gating(adev); + vcn_1_0_enable_static_power_gating(adev); return 0; } +bool vcn_v1_0_is_idle(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + return (RREG32_SOC15(VCN, 0, mmUVD_STATUS) == 0x2); +} + +int vcn_v1_0_wait_for_idle(void *handle) +{ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + int ret = 0; + + SOC15_WAIT_ON_RREG(VCN, 0, mmUVD_STATUS, 0x2, 0x2, ret); + + return ret; +} + static int vcn_v1_0_set_clockgating_state(void *handle, enum amd_clockgating_state state) { - /* needed for driver unload*/ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + bool enable = (state == AMD_CG_STATE_GATE) ? true : false; + + if (enable) { + /* wait for STATUS to clear */ + if (vcn_v1_0_is_idle(handle)) + return -EBUSY; + vcn_v1_0_enable_clock_gating(adev); + } else { + /* disable HW gating and enable Sw gating */ + vcn_v1_0_disable_clock_gating(adev); + } return 0; } @@ -1048,16 +1159,36 @@ static int vcn_v1_0_process_interrupt(struct amdgpu_device *adev, return 0; } -static void vcn_v1_0_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) +static void vcn_v1_0_dec_ring_insert_nop(struct amdgpu_ring *ring, uint32_t count) { - int i; struct amdgpu_device *adev = ring->adev; + int i; - for (i = 0; i < count; i++) - amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP), 0)); + WARN_ON(ring->wptr % 2 || count % 2); + for (i = 0; i < count / 2; i++) { + amdgpu_ring_write(ring, PACKET0(SOC15_REG_OFFSET(UVD, 0, mmUVD_NO_OP), 0)); + amdgpu_ring_write(ring, 0); + } } +static int vcn_v1_0_set_powergating_state(void *handle, + enum amd_powergating_state state) +{ + /* This doesn't actually powergate the VCN block. + * That's done in the dpm code via the SMC. This + * just re-inits the block as necessary. The actual + * gating still happens in the dpm code. We should + * revisit this when there is a cleaner line between + * the smc and the hw blocks + */ + struct amdgpu_device *adev = (struct amdgpu_device *)handle; + + if (state == AMD_PG_STATE_GATE) + return vcn_v1_0_stop(adev); + else + return vcn_v1_0_start(adev); +} static const struct amd_ip_funcs vcn_v1_0_ip_funcs = { .name = "vcn_v1_0", @@ -1069,20 +1200,19 @@ static const struct amd_ip_funcs vcn_v1_0_ip_funcs = { .hw_fini = vcn_v1_0_hw_fini, .suspend = vcn_v1_0_suspend, .resume = vcn_v1_0_resume, - .is_idle = NULL /* vcn_v1_0_is_idle */, - .wait_for_idle = NULL /* vcn_v1_0_wait_for_idle */, + .is_idle = vcn_v1_0_is_idle, + .wait_for_idle = vcn_v1_0_wait_for_idle, .check_soft_reset = NULL /* vcn_v1_0_check_soft_reset */, .pre_soft_reset = NULL /* vcn_v1_0_pre_soft_reset */, .soft_reset = NULL /* vcn_v1_0_soft_reset */, .post_soft_reset = NULL /* vcn_v1_0_post_soft_reset */, .set_clockgating_state = vcn_v1_0_set_clockgating_state, - .set_powergating_state = NULL /* vcn_v1_0_set_powergating_state */, + .set_powergating_state = vcn_v1_0_set_powergating_state, }; static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = { .type = AMDGPU_RING_TYPE_VCN_DEC, .align_mask = 0xf, - .nop = PACKET0(0x81ff, 0), .support_64bit_ptrs = false, .vmhub = AMDGPU_MMHUB, .get_rptr = vcn_v1_0_dec_ring_get_rptr, @@ -1101,7 +1231,7 @@ static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = { .emit_vm_flush = vcn_v1_0_dec_ring_emit_vm_flush, .test_ring = amdgpu_vcn_dec_ring_test_ring, .test_ib = amdgpu_vcn_dec_ring_test_ib, - .insert_nop = vcn_v1_0_ring_insert_nop, + .insert_nop = vcn_v1_0_dec_ring_insert_nop, .insert_start = vcn_v1_0_dec_ring_insert_start, .insert_end = vcn_v1_0_dec_ring_insert_end, .pad_ib = amdgpu_ring_generic_pad_ib, @@ -1109,6 +1239,7 @@ static const struct amdgpu_ring_funcs vcn_v1_0_dec_ring_vm_funcs = { .end_use = amdgpu_vcn_ring_end_use, .emit_wreg = vcn_v1_0_dec_ring_emit_wreg, .emit_reg_wait = vcn_v1_0_dec_ring_emit_reg_wait, + .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, }; static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = { @@ -1139,6 +1270,7 @@ static const struct amdgpu_ring_funcs vcn_v1_0_enc_ring_vm_funcs = { .end_use = amdgpu_vcn_ring_end_use, .emit_wreg = vcn_v1_0_enc_ring_emit_wreg, .emit_reg_wait = vcn_v1_0_enc_ring_emit_reg_wait, + .emit_reg_write_reg_wait = amdgpu_ring_emit_reg_write_reg_wait_helper, }; static void vcn_v1_0_set_dec_ring_funcs(struct amdgpu_device *adev) diff --git a/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c b/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c new file mode 100644 index 0000000000000..52778de93ab08 --- /dev/null +++ b/drivers/gpu/drm/amd/amdgpu/vega20_reg_init.c @@ -0,0 +1,53 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "amdgpu.h" +#include "soc15.h" + +#include "soc15_common.h" +#include "soc15_hw_ip.h" +#include "vega20_ip_offset.h" + +int vega20_reg_base_init(struct amdgpu_device *adev) +{ + /* HW has more IP blocks, only initialized the blocke beend by our driver */ + uint32_t i; + for (i = 0 ; i < MAX_INSTANCE ; ++i) { + adev->reg_offset[GC_HWIP][i] = (uint32_t *)(&(GC_BASE.instance[i])); + adev->reg_offset[HDP_HWIP][i] = (uint32_t *)(&(HDP_BASE.instance[i])); + adev->reg_offset[MMHUB_HWIP][i] = (uint32_t *)(&(MMHUB_BASE.instance[i])); + adev->reg_offset[ATHUB_HWIP][i] = (uint32_t *)(&(ATHUB_BASE.instance[i])); + adev->reg_offset[NBIO_HWIP][i] = (uint32_t *)(&(NBIO_BASE.instance[i])); + adev->reg_offset[MP0_HWIP][i] = (uint32_t *)(&(MP0_BASE.instance[i])); + adev->reg_offset[UVD_HWIP][i] = (uint32_t *)(&(UVD_BASE.instance[i])); + adev->reg_offset[VCE_HWIP][i] = (uint32_t *)(&(VCE_BASE.instance[i])); + adev->reg_offset[DF_HWIP][i] = (uint32_t *)(&(DF_BASE.instance[i])); + adev->reg_offset[DCE_HWIP][i] = (uint32_t *)(&(DCE_BASE.instance[i])); + adev->reg_offset[OSSSYS_HWIP][i] = (uint32_t *)(&(OSSSYS_BASE.instance[i])); + adev->reg_offset[SDMA0_HWIP][i] = (uint32_t *)(&(SDMA0_BASE.instance[i])); + adev->reg_offset[SDMA1_HWIP][i] = (uint32_t *)(&(SDMA1_BASE.instance[i])); + adev->reg_offset[SMUIO_HWIP][i] = (uint32_t *)(&(SMUIO_BASE.instance[i])); + } + return 0; +} + + diff --git a/drivers/gpu/drm/amd/amdgpu/vi.c b/drivers/gpu/drm/amd/amdgpu/vi.c index 126f1276d3470..4ac1288ab7dff 100644 --- a/drivers/gpu/drm/amd/amdgpu/vi.c +++ b/drivers/gpu/drm/amd/amdgpu/vi.c @@ -305,9 +305,10 @@ static void vi_init_golden_registers(struct amdgpu_device *adev) stoney_mgcg_cgcg_init, ARRAY_SIZE(stoney_mgcg_cgcg_init)); break; - case CHIP_POLARIS11: case CHIP_POLARIS10: + case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: default: break; } @@ -728,33 +729,59 @@ static int vi_set_uvd_clock(struct amdgpu_device *adev, u32 clock, return r; tmp = RREG32_SMC(cntl_reg); - tmp &= ~(CG_DCLK_CNTL__DCLK_DIR_CNTL_EN_MASK | - CG_DCLK_CNTL__DCLK_DIVIDER_MASK); + + if (adev->flags & AMD_IS_APU) + tmp &= ~CG_DCLK_CNTL__DCLK_DIVIDER_MASK; + else + tmp &= ~(CG_DCLK_CNTL__DCLK_DIR_CNTL_EN_MASK | + CG_DCLK_CNTL__DCLK_DIVIDER_MASK); tmp |= dividers.post_divider; WREG32_SMC(cntl_reg, tmp); for (i = 0; i < 100; i++) { - if (RREG32_SMC(status_reg) & CG_DCLK_STATUS__DCLK_STATUS_MASK) - break; + tmp = RREG32_SMC(status_reg); + if (adev->flags & AMD_IS_APU) { + if (tmp & 0x10000) + break; + } else { + if (tmp & CG_DCLK_STATUS__DCLK_STATUS_MASK) + break; + } mdelay(10); } if (i == 100) return -ETIMEDOUT; - return 0; } +#define ixGNB_CLK1_DFS_CNTL 0xD82200F0 +#define ixGNB_CLK1_STATUS 0xD822010C +#define ixGNB_CLK2_DFS_CNTL 0xD8220110 +#define ixGNB_CLK2_STATUS 0xD822012C +#define ixGNB_CLK3_DFS_CNTL 0xD8220130 +#define ixGNB_CLK3_STATUS 0xD822014C + static int vi_set_uvd_clocks(struct amdgpu_device *adev, u32 vclk, u32 dclk) { int r; - r = vi_set_uvd_clock(adev, vclk, ixCG_VCLK_CNTL, ixCG_VCLK_STATUS); - if (r) - return r; + if (adev->flags & AMD_IS_APU) { + r = vi_set_uvd_clock(adev, vclk, ixGNB_CLK2_DFS_CNTL, ixGNB_CLK2_STATUS); + if (r) + return r; - r = vi_set_uvd_clock(adev, dclk, ixCG_DCLK_CNTL, ixCG_DCLK_STATUS); - if (r) - return r; + r = vi_set_uvd_clock(adev, dclk, ixGNB_CLK1_DFS_CNTL, ixGNB_CLK1_STATUS); + if (r) + return r; + } else { + r = vi_set_uvd_clock(adev, vclk, ixCG_VCLK_CNTL, ixCG_VCLK_STATUS); + if (r) + return r; + + r = vi_set_uvd_clock(adev, dclk, ixCG_DCLK_CNTL, ixCG_DCLK_STATUS); + if (r) + return r; + } return 0; } @@ -764,6 +791,22 @@ static int vi_set_vce_clocks(struct amdgpu_device *adev, u32 evclk, u32 ecclk) int r, i; struct atom_clock_dividers dividers; u32 tmp; + u32 reg_ctrl; + u32 reg_status; + u32 status_mask; + u32 reg_mask; + + if (adev->flags & AMD_IS_APU) { + reg_ctrl = ixGNB_CLK3_DFS_CNTL; + reg_status = ixGNB_CLK3_STATUS; + status_mask = 0x00010000; + reg_mask = CG_ECLK_CNTL__ECLK_DIVIDER_MASK; + } else { + reg_ctrl = ixCG_ECLK_CNTL; + reg_status = ixCG_ECLK_STATUS; + status_mask = CG_ECLK_STATUS__ECLK_STATUS_MASK; + reg_mask = CG_ECLK_CNTL__ECLK_DIR_CNTL_EN_MASK | CG_ECLK_CNTL__ECLK_DIVIDER_MASK; + } r = amdgpu_atombios_get_clock_dividers(adev, COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, @@ -772,24 +815,25 @@ static int vi_set_vce_clocks(struct amdgpu_device *adev, u32 evclk, u32 ecclk) return r; for (i = 0; i < 100; i++) { - if (RREG32_SMC(ixCG_ECLK_STATUS) & CG_ECLK_STATUS__ECLK_STATUS_MASK) + if (RREG32_SMC(reg_status) & status_mask) break; mdelay(10); } + if (i == 100) return -ETIMEDOUT; - tmp = RREG32_SMC(ixCG_ECLK_CNTL); - tmp &= ~(CG_ECLK_CNTL__ECLK_DIR_CNTL_EN_MASK | - CG_ECLK_CNTL__ECLK_DIVIDER_MASK); + tmp = RREG32_SMC(reg_ctrl); + tmp &= ~reg_mask; tmp |= dividers.post_divider; - WREG32_SMC(ixCG_ECLK_CNTL, tmp); + WREG32_SMC(reg_ctrl, tmp); for (i = 0; i < 100; i++) { - if (RREG32_SMC(ixCG_ECLK_STATUS) & CG_ECLK_STATUS__ECLK_STATUS_MASK) + if (RREG32_SMC(reg_status) & status_mask) break; mdelay(10); } + if (i == 100) return -ETIMEDOUT; @@ -876,6 +920,27 @@ static void vi_invalidate_hdp(struct amdgpu_device *adev, } } +static bool vi_need_full_reset(struct amdgpu_device *adev) +{ + switch (adev->asic_type) { + case CHIP_CARRIZO: + case CHIP_STONEY: + /* CZ has hang issues with full reset at the moment */ + return false; + case CHIP_FIJI: + case CHIP_TONGA: + /* XXX: soft reset should work on fiji and tonga */ + return true; + case CHIP_POLARIS10: + case CHIP_POLARIS11: + case CHIP_POLARIS12: + case CHIP_TOPAZ: + default: + /* change this when we support soft reset */ + return true; + } +} + static const struct amdgpu_asic_funcs vi_asic_funcs = { .read_disabled_bios = &vi_read_disabled_bios, @@ -889,6 +954,7 @@ static const struct amdgpu_asic_funcs vi_asic_funcs = .get_config_memsize = &vi_get_config_memsize, .flush_hdp = &vi_flush_hdp, .invalidate_hdp = &vi_invalidate_hdp, + .need_full_reset = &vi_need_full_reset, }; #define CZ_REV_BRISTOL(rev) \ @@ -1031,6 +1097,30 @@ static int vi_common_early_init(void *handle) adev->pg_flags = 0; adev->external_rev_id = adev->rev_id + 0x64; break; + case CHIP_VEGAM: + adev->cg_flags = 0; + /*AMD_CG_SUPPORT_GFX_MGCG | + AMD_CG_SUPPORT_GFX_RLC_LS | + AMD_CG_SUPPORT_GFX_CP_LS | + AMD_CG_SUPPORT_GFX_CGCG | + AMD_CG_SUPPORT_GFX_CGLS | + AMD_CG_SUPPORT_GFX_3D_CGCG | + AMD_CG_SUPPORT_GFX_3D_CGLS | + AMD_CG_SUPPORT_SDMA_MGCG | + AMD_CG_SUPPORT_SDMA_LS | + AMD_CG_SUPPORT_BIF_MGCG | + AMD_CG_SUPPORT_BIF_LS | + AMD_CG_SUPPORT_HDP_MGCG | + AMD_CG_SUPPORT_HDP_LS | + AMD_CG_SUPPORT_ROM_MGCG | + AMD_CG_SUPPORT_MC_MGCG | + AMD_CG_SUPPORT_MC_LS | + AMD_CG_SUPPORT_DRM_LS | + AMD_CG_SUPPORT_UVD_MGCG | + AMD_CG_SUPPORT_VCE_MGCG;*/ + adev->pg_flags = 0; + adev->external_rev_id = adev->rev_id + 0x6E; + break; case CHIP_CARRIZO: adev->cg_flags = AMD_CG_SUPPORT_UVD_MGCG | AMD_CG_SUPPORT_GFX_MGCG | @@ -1422,6 +1512,7 @@ static int vi_common_set_clockgating_state(void *handle, case CHIP_POLARIS10: case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: vi_common_set_clockgating_state_by_smu(adev, state); default: break; @@ -1551,9 +1642,10 @@ int vi_set_ip_blocks(struct amdgpu_device *adev) amdgpu_device_ip_block_add(adev, &vce_v3_0_ip_block); } break; - case CHIP_POLARIS11: case CHIP_POLARIS10: + case CHIP_POLARIS11: case CHIP_POLARIS12: + case CHIP_VEGAM: amdgpu_device_ip_block_add(adev, &vi_common_ip_block); amdgpu_device_ip_block_add(adev, &gmc_v8_1_ip_block); amdgpu_device_ip_block_add(adev, &tonga_ih_ip_block); diff --git a/drivers/gpu/drm/amd/amdkfd/Makefile b/drivers/gpu/drm/amd/amdkfd/Makefile index 0d0242240c478..ffd096fffc1c2 100644 --- a/drivers/gpu/drm/amd/amdkfd/Makefile +++ b/drivers/gpu/drm/amd/amdkfd/Makefile @@ -30,12 +30,14 @@ amdkfd-y := kfd_module.o kfd_device.o kfd_chardev.o kfd_topology.o \ kfd_pasid.o kfd_doorbell.o kfd_flat_memory.o \ kfd_process.o kfd_queue.o kfd_mqd_manager.o \ kfd_mqd_manager_cik.o kfd_mqd_manager_vi.o \ + kfd_mqd_manager_v9.o \ kfd_kernel_queue.o kfd_kernel_queue_cik.o \ - kfd_kernel_queue_vi.o kfd_packet_manager.o \ - kfd_process_queue_manager.o kfd_device_queue_manager.o \ - kfd_device_queue_manager_cik.o kfd_device_queue_manager_vi.o \ + kfd_kernel_queue_vi.o kfd_kernel_queue_v9.o \ + kfd_packet_manager.o kfd_process_queue_manager.o \ + kfd_device_queue_manager.o kfd_device_queue_manager_cik.o \ + kfd_device_queue_manager_vi.o kfd_device_queue_manager_v9.o \ kfd_interrupt.o kfd_events.o cik_event_interrupt.o \ - kfd_dbgdev.o kfd_dbgmgr.o kfd_crat.o + kfd_int_process_v9.o kfd_dbgdev.o kfd_dbgmgr.o kfd_crat.o ifneq ($(CONFIG_AMD_IOMMU_V2),) amdkfd-y += kfd_iommu.o diff --git a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c index 3d5ccb3755d4f..49df6c791cfcc 100644 --- a/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c +++ b/drivers/gpu/drm/amd/amdkfd/cik_event_interrupt.c @@ -27,18 +27,28 @@ static bool cik_event_interrupt_isr(struct kfd_dev *dev, const uint32_t *ih_ring_entry) { - unsigned int pasid; const struct cik_ih_ring_entry *ihre = (const struct cik_ih_ring_entry *)ih_ring_entry; + unsigned int vmid, pasid; + + /* Only handle interrupts from KFD VMIDs */ + vmid = (ihre->ring_id & 0x0000ff00) >> 8; + if (vmid < dev->vm_info.first_vmid_kfd || + vmid > dev->vm_info.last_vmid_kfd) + return 0; + /* If there is no valid PASID, it's likely a firmware bug */ pasid = (ihre->ring_id & 0xffff0000) >> 16; + if (WARN_ONCE(pasid == 0, "FW bug: No PASID in KFD interrupt")) + return 0; - /* Do not process in ISR, just request it to be forwarded to WQ. */ - return (pasid != 0) && - (ihre->source_id == CIK_INTSRC_CP_END_OF_PIPE || + /* Interrupt types we care about: various signals and faults. + * They will be forwarded to a work queue (see below). + */ + return ihre->source_id == CIK_INTSRC_CP_END_OF_PIPE || ihre->source_id == CIK_INTSRC_SDMA_TRAP || ihre->source_id == CIK_INTSRC_SQ_INTERRUPT_MSG || - ihre->source_id == CIK_INTSRC_CP_BAD_OPCODE); + ihre->source_id == CIK_INTSRC_CP_BAD_OPCODE; } static void cik_event_interrupt_wq(struct kfd_dev *dev, diff --git a/drivers/gpu/drm/amd/amdkfd/cik_regs.h b/drivers/gpu/drm/amd/amdkfd/cik_regs.h index 48769d12dd7be..37ce6dd653917 100644 --- a/drivers/gpu/drm/amd/amdkfd/cik_regs.h +++ b/drivers/gpu/drm/amd/amdkfd/cik_regs.h @@ -33,7 +33,8 @@ #define APE1_MTYPE(x) ((x) << 7) /* valid for both DEFAULT_MTYPE and APE1_MTYPE */ -#define MTYPE_CACHED 0 +#define MTYPE_CACHED_NV 0 +#define MTYPE_CACHED 1 #define MTYPE_NONCACHED 3 #define DEFAULT_CP_HQD_PERSISTENT_STATE (0x33U << 8) diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h new file mode 100644 index 0000000000000..f68aef02fc1fc --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler.h @@ -0,0 +1,560 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +static const uint32_t cwsr_trap_gfx8_hex[] = { + 0xbf820001, 0xbf820125, + 0xb8f4f802, 0x89748674, + 0xb8f5f803, 0x8675ff75, + 0x00000400, 0xbf850011, + 0xc00a1e37, 0x00000000, + 0xbf8c007f, 0x87777978, + 0xbf840002, 0xb974f802, + 0xbe801d78, 0xb8f5f803, + 0x8675ff75, 0x000001ff, + 0xbf850002, 0x80708470, + 0x82718071, 0x8671ff71, + 0x0000ffff, 0xb974f802, + 0xbe801f70, 0xb8f5f803, + 0x8675ff75, 0x00000100, + 0xbf840006, 0xbefa0080, + 0xb97a0203, 0x8671ff71, + 0x0000ffff, 0x80f08870, + 0x82f18071, 0xbefa0080, + 0xb97a0283, 0xbef60068, + 0xbef70069, 0xb8fa1c07, + 0x8e7a9c7a, 0x87717a71, + 0xb8fa03c7, 0x8e7a9b7a, + 0x87717a71, 0xb8faf807, + 0x867aff7a, 0x00007fff, + 0xb97af807, 0xbef2007e, + 0xbef3007f, 0xbefe0180, + 0xbf900004, 0x877a8474, + 0xb97af802, 0xbf8e0002, + 0xbf88fffe, 0xbef8007e, + 0x8679ff7f, 0x0000ffff, + 0x8779ff79, 0x00040000, + 0xbefa0080, 0xbefb00ff, + 0x00807fac, 0x867aff7f, + 0x08000000, 0x8f7a837a, + 0x877b7a7b, 0x867aff7f, + 0x70000000, 0x8f7a817a, + 0x877b7a7b, 0xbeef007c, + 0xbeee0080, 0xb8ee2a05, + 0x806e816e, 0x8e6e8a6e, + 0xb8fa1605, 0x807a817a, + 0x8e7a867a, 0x806e7a6e, + 0xbefa0084, 0xbefa00ff, + 0x01000000, 0xbefe007c, + 0xbefc006e, 0xc0611bfc, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611c3c, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611c7c, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611cbc, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611cfc, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611d3c, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xb8f5f803, + 0xbefe007c, 0xbefc006e, + 0xc0611d7c, 0x0000007c, + 0x806e846e, 0xbefc007e, + 0xbefe007c, 0xbefc006e, + 0xc0611dbc, 0x0000007c, + 0x806e846e, 0xbefc007e, + 0xbefe007c, 0xbefc006e, + 0xc0611dfc, 0x0000007c, + 0x806e846e, 0xbefc007e, + 0xb8eff801, 0xbefe007c, + 0xbefc006e, 0xc0611bfc, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611b3c, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0xbefe007c, + 0xbefc006e, 0xc0611b7c, + 0x0000007c, 0x806e846e, + 0xbefc007e, 0x867aff7f, + 0x04000000, 0xbef30080, + 0x8773737a, 0xb8ee2a05, + 0x806e816e, 0x8e6e8a6e, + 0xb8f51605, 0x80758175, + 0x8e758475, 0x8e7a8275, + 0xbefa00ff, 0x01000000, + 0xbef60178, 0x80786e78, + 0x82798079, 0xbefc0080, + 0xbe802b00, 0xbe822b02, + 0xbe842b04, 0xbe862b06, + 0xbe882b08, 0xbe8a2b0a, + 0xbe8c2b0c, 0xbe8e2b0e, + 0xc06b003c, 0x00000000, + 0xc06b013c, 0x00000010, + 0xc06b023c, 0x00000020, + 0xc06b033c, 0x00000030, + 0x8078c078, 0x82798079, + 0x807c907c, 0xbf0a757c, + 0xbf85ffeb, 0xbef80176, + 0xbeee0080, 0xbefe00c1, + 0xbeff00c1, 0xbefa00ff, + 0x01000000, 0xe0724000, + 0x6e1e0000, 0xe0724100, + 0x6e1e0100, 0xe0724200, + 0x6e1e0200, 0xe0724300, + 0x6e1e0300, 0xbefe00c1, + 0xbeff00c1, 0xb8f54306, + 0x8675c175, 0xbf84002c, + 0xbf8a0000, 0x867aff73, + 0x04000000, 0xbf840028, + 0x8e758675, 0x8e758275, + 0xbefa0075, 0xb8ee2a05, + 0x806e816e, 0x8e6e8a6e, + 0xb8fa1605, 0x807a817a, + 0x8e7a867a, 0x806e7a6e, + 0x806eff6e, 0x00000080, + 0xbefa00ff, 0x01000000, + 0xbefc0080, 0xd28c0002, + 0x000100c1, 0xd28d0003, + 0x000204c1, 0xd1060002, + 0x00011103, 0x7e0602ff, + 0x00000200, 0xbefc00ff, + 0x00010000, 0xbe80007b, + 0x867bff7b, 0xff7fffff, + 0x877bff7b, 0x00058000, + 0xd8ec0000, 0x00000002, + 0xbf8c007f, 0xe0765000, + 0x6e1e0002, 0x32040702, + 0xd0c9006a, 0x0000eb02, + 0xbf87fff7, 0xbefb0000, + 0xbeee00ff, 0x00000400, + 0xbefe00c1, 0xbeff00c1, + 0xb8f52a05, 0x80758175, + 0x8e758275, 0x8e7a8875, + 0xbefa00ff, 0x01000000, + 0xbefc0084, 0xbf0a757c, + 0xbf840015, 0xbf11017c, + 0x8075ff75, 0x00001000, + 0x7e000300, 0x7e020301, + 0x7e040302, 0x7e060303, + 0xe0724000, 0x6e1e0000, + 0xe0724100, 0x6e1e0100, + 0xe0724200, 0x6e1e0200, + 0xe0724300, 0x6e1e0300, + 0x807c847c, 0x806eff6e, + 0x00000400, 0xbf0a757c, + 0xbf85ffef, 0xbf9c0000, + 0xbf8200ca, 0xbef8007e, + 0x8679ff7f, 0x0000ffff, + 0x8779ff79, 0x00040000, + 0xbefa0080, 0xbefb00ff, + 0x00807fac, 0x8676ff7f, + 0x08000000, 0x8f768376, + 0x877b767b, 0x8676ff7f, + 0x70000000, 0x8f768176, + 0x877b767b, 0x8676ff7f, + 0x04000000, 0xbf84001e, + 0xbefe00c1, 0xbeff00c1, + 0xb8f34306, 0x8673c173, + 0xbf840019, 0x8e738673, + 0x8e738273, 0xbefa0073, + 0xb8f22a05, 0x80728172, + 0x8e728a72, 0xb8f61605, + 0x80768176, 0x8e768676, + 0x80727672, 0x8072ff72, + 0x00000080, 0xbefa00ff, + 0x01000000, 0xbefc0080, + 0xe0510000, 0x721e0000, + 0xe0510100, 0x721e0000, + 0x807cff7c, 0x00000200, + 0x8072ff72, 0x00000200, + 0xbf0a737c, 0xbf85fff6, + 0xbef20080, 0xbefe00c1, + 0xbeff00c1, 0xb8f32a05, + 0x80738173, 0x8e738273, + 0x8e7a8873, 0xbefa00ff, + 0x01000000, 0xbef60072, + 0x8072ff72, 0x00000400, + 0xbefc0084, 0xbf11087c, + 0x8073ff73, 0x00008000, + 0xe0524000, 0x721e0000, + 0xe0524100, 0x721e0100, + 0xe0524200, 0x721e0200, + 0xe0524300, 0x721e0300, + 0xbf8c0f70, 0x7e000300, + 0x7e020301, 0x7e040302, + 0x7e060303, 0x807c847c, + 0x8072ff72, 0x00000400, + 0xbf0a737c, 0xbf85ffee, + 0xbf9c0000, 0xe0524000, + 0x761e0000, 0xe0524100, + 0x761e0100, 0xe0524200, + 0x761e0200, 0xe0524300, + 0x761e0300, 0xb8f22a05, + 0x80728172, 0x8e728a72, + 0xb8f61605, 0x80768176, + 0x8e768676, 0x80727672, + 0x80f2c072, 0xb8f31605, + 0x80738173, 0x8e738473, + 0x8e7a8273, 0xbefa00ff, + 0x01000000, 0xbefc0073, + 0xc031003c, 0x00000072, + 0x80f2c072, 0xbf8c007f, + 0x80fc907c, 0xbe802d00, + 0xbe822d02, 0xbe842d04, + 0xbe862d06, 0xbe882d08, + 0xbe8a2d0a, 0xbe8c2d0c, + 0xbe8e2d0e, 0xbf06807c, + 0xbf84fff1, 0xb8f22a05, + 0x80728172, 0x8e728a72, + 0xb8f61605, 0x80768176, + 0x8e768676, 0x80727672, + 0xbefa0084, 0xbefa00ff, + 0x01000000, 0xc0211cfc, + 0x00000072, 0x80728472, + 0xc0211c3c, 0x00000072, + 0x80728472, 0xc0211c7c, + 0x00000072, 0x80728472, + 0xc0211bbc, 0x00000072, + 0x80728472, 0xc0211bfc, + 0x00000072, 0x80728472, + 0xc0211d3c, 0x00000072, + 0x80728472, 0xc0211d7c, + 0x00000072, 0x80728472, + 0xc0211a3c, 0x00000072, + 0x80728472, 0xc0211a7c, + 0x00000072, 0x80728472, + 0xc0211dfc, 0x00000072, + 0x80728472, 0xc0211b3c, + 0x00000072, 0x80728472, + 0xc0211b7c, 0x00000072, + 0x80728472, 0xbf8c007f, + 0xbefc0073, 0xbefe006e, + 0xbeff006f, 0x867375ff, + 0x000003ff, 0xb9734803, + 0x867375ff, 0xfffff800, + 0x8f738b73, 0xb973a2c3, + 0xb977f801, 0x8673ff71, + 0xf0000000, 0x8f739c73, + 0x8e739073, 0xbef60080, + 0x87767376, 0x8673ff71, + 0x08000000, 0x8f739b73, + 0x8e738f73, 0x87767376, + 0x8673ff74, 0x00800000, + 0x8f739773, 0xb976f807, + 0x8671ff71, 0x0000ffff, + 0x86fe7e7e, 0x86ea6a6a, + 0xb974f802, 0xbf8a0000, + 0x95807370, 0xbf810000, +}; + + +static const uint32_t cwsr_trap_gfx9_hex[] = { + 0xbf820001, 0xbf82015a, + 0xb8f8f802, 0x89788678, + 0xb8f1f803, 0x866eff71, + 0x00000400, 0xbf850034, + 0x866eff71, 0x00000800, + 0xbf850003, 0x866eff71, + 0x00000100, 0xbf840008, + 0x866eff78, 0x00002000, + 0xbf840001, 0xbf810000, + 0x8778ff78, 0x00002000, + 0x80ec886c, 0x82ed806d, + 0xb8eef807, 0x866fff6e, + 0x001f8000, 0x8e6f8b6f, + 0x8977ff77, 0xfc000000, + 0x87776f77, 0x896eff6e, + 0x001f8000, 0xb96ef807, + 0xb8f0f812, 0xb8f1f813, + 0x8ef08870, 0xc0071bb8, + 0x00000000, 0xbf8cc07f, + 0xc0071c38, 0x00000008, + 0xbf8cc07f, 0x86ee6e6e, + 0xbf840001, 0xbe801d6e, + 0xb8f1f803, 0x8671ff71, + 0x000001ff, 0xbf850002, + 0x806c846c, 0x826d806d, + 0x866dff6d, 0x0000ffff, + 0x8f6e8b77, 0x866eff6e, + 0x001f8000, 0xb96ef807, + 0x86fe7e7e, 0x86ea6a6a, + 0xb978f802, 0xbe801f6c, + 0x866dff6d, 0x0000ffff, + 0xbef00080, 0xb9700283, + 0xb8f02407, 0x8e709c70, + 0x876d706d, 0xb8f003c7, + 0x8e709b70, 0x876d706d, + 0xb8f0f807, 0x8670ff70, + 0x00007fff, 0xb970f807, + 0xbeee007e, 0xbeef007f, + 0xbefe0180, 0xbf900004, + 0x87708478, 0xb970f802, + 0xbf8e0002, 0xbf88fffe, + 0xb8f02a05, 0x80708170, + 0x8e708a70, 0xb8f11605, + 0x80718171, 0x8e718671, + 0x80707170, 0x80707e70, + 0x8271807f, 0x8671ff71, + 0x0000ffff, 0xc0471cb8, + 0x00000040, 0xbf8cc07f, + 0xc04b1d38, 0x00000048, + 0xbf8cc07f, 0xc0431e78, + 0x00000058, 0xbf8cc07f, + 0xc0471eb8, 0x0000005c, + 0xbf8cc07f, 0xbef4007e, + 0x8675ff7f, 0x0000ffff, + 0x8775ff75, 0x00040000, + 0xbef60080, 0xbef700ff, + 0x00807fac, 0x8670ff7f, + 0x08000000, 0x8f708370, + 0x87777077, 0x8670ff7f, + 0x70000000, 0x8f708170, + 0x87777077, 0xbefb007c, + 0xbefa0080, 0xb8fa2a05, + 0x807a817a, 0x8e7a8a7a, + 0xb8f01605, 0x80708170, + 0x8e708670, 0x807a707a, + 0xbef60084, 0xbef600ff, + 0x01000000, 0xbefe007c, + 0xbefc007a, 0xc0611efa, + 0x0000007c, 0xbf8cc07f, + 0x807a847a, 0xbefc007e, + 0xbefe007c, 0xbefc007a, + 0xc0611b3a, 0x0000007c, + 0xbf8cc07f, 0x807a847a, + 0xbefc007e, 0xbefe007c, + 0xbefc007a, 0xc0611b7a, + 0x0000007c, 0xbf8cc07f, + 0x807a847a, 0xbefc007e, + 0xbefe007c, 0xbefc007a, + 0xc0611bba, 0x0000007c, + 0xbf8cc07f, 0x807a847a, + 0xbefc007e, 0xbefe007c, + 0xbefc007a, 0xc0611bfa, + 0x0000007c, 0xbf8cc07f, + 0x807a847a, 0xbefc007e, + 0xbefe007c, 0xbefc007a, + 0xc0611e3a, 0x0000007c, + 0xbf8cc07f, 0x807a847a, + 0xbefc007e, 0xb8f1f803, + 0xbefe007c, 0xbefc007a, + 0xc0611c7a, 0x0000007c, + 0xbf8cc07f, 0x807a847a, + 0xbefc007e, 0xbefe007c, + 0xbefc007a, 0xc0611a3a, + 0x0000007c, 0xbf8cc07f, + 0x807a847a, 0xbefc007e, + 0xbefe007c, 0xbefc007a, + 0xc0611a7a, 0x0000007c, + 0xbf8cc07f, 0x807a847a, + 0xbefc007e, 0xb8fbf801, + 0xbefe007c, 0xbefc007a, + 0xc0611efa, 0x0000007c, + 0xbf8cc07f, 0x807a847a, + 0xbefc007e, 0x8670ff7f, + 0x04000000, 0xbeef0080, + 0x876f6f70, 0xb8fa2a05, + 0x807a817a, 0x8e7a8a7a, + 0xb8f11605, 0x80718171, + 0x8e718471, 0x8e768271, + 0xbef600ff, 0x01000000, + 0xbef20174, 0x80747a74, + 0x82758075, 0xbefc0080, + 0xbf800000, 0xbe802b00, + 0xbe822b02, 0xbe842b04, + 0xbe862b06, 0xbe882b08, + 0xbe8a2b0a, 0xbe8c2b0c, + 0xbe8e2b0e, 0xc06b003a, + 0x00000000, 0xbf8cc07f, + 0xc06b013a, 0x00000010, + 0xbf8cc07f, 0xc06b023a, + 0x00000020, 0xbf8cc07f, + 0xc06b033a, 0x00000030, + 0xbf8cc07f, 0x8074c074, + 0x82758075, 0x807c907c, + 0xbf0a717c, 0xbf85ffe7, + 0xbef40172, 0xbefa0080, + 0xbefe00c1, 0xbeff00c1, + 0xbee80080, 0xbee90080, + 0xbef600ff, 0x01000000, + 0xe0724000, 0x7a1d0000, + 0xe0724100, 0x7a1d0100, + 0xe0724200, 0x7a1d0200, + 0xe0724300, 0x7a1d0300, + 0xbefe00c1, 0xbeff00c1, + 0xb8f14306, 0x8671c171, + 0xbf84002c, 0xbf8a0000, + 0x8670ff6f, 0x04000000, + 0xbf840028, 0x8e718671, + 0x8e718271, 0xbef60071, + 0xb8fa2a05, 0x807a817a, + 0x8e7a8a7a, 0xb8f01605, + 0x80708170, 0x8e708670, + 0x807a707a, 0x807aff7a, + 0x00000080, 0xbef600ff, + 0x01000000, 0xbefc0080, + 0xd28c0002, 0x000100c1, + 0xd28d0003, 0x000204c1, + 0xd1060002, 0x00011103, + 0x7e0602ff, 0x00000200, + 0xbefc00ff, 0x00010000, + 0xbe800077, 0x8677ff77, + 0xff7fffff, 0x8777ff77, + 0x00058000, 0xd8ec0000, + 0x00000002, 0xbf8cc07f, + 0xe0765000, 0x7a1d0002, + 0x68040702, 0xd0c9006a, + 0x0000e302, 0xbf87fff7, + 0xbef70000, 0xbefa00ff, + 0x00000400, 0xbefe00c1, + 0xbeff00c1, 0xb8f12a05, + 0x80718171, 0x8e718271, + 0x8e768871, 0xbef600ff, + 0x01000000, 0xbefc0084, + 0xbf0a717c, 0xbf840015, + 0xbf11017c, 0x8071ff71, + 0x00001000, 0x7e000300, + 0x7e020301, 0x7e040302, + 0x7e060303, 0xe0724000, + 0x7a1d0000, 0xe0724100, + 0x7a1d0100, 0xe0724200, + 0x7a1d0200, 0xe0724300, + 0x7a1d0300, 0x807c847c, + 0x807aff7a, 0x00000400, + 0xbf0a717c, 0xbf85ffef, + 0xbf9c0000, 0xbf8200d9, + 0xbef4007e, 0x8675ff7f, + 0x0000ffff, 0x8775ff75, + 0x00040000, 0xbef60080, + 0xbef700ff, 0x00807fac, + 0x866eff7f, 0x08000000, + 0x8f6e836e, 0x87776e77, + 0x866eff7f, 0x70000000, + 0x8f6e816e, 0x87776e77, + 0x866eff7f, 0x04000000, + 0xbf84001e, 0xbefe00c1, + 0xbeff00c1, 0xb8ef4306, + 0x866fc16f, 0xbf840019, + 0x8e6f866f, 0x8e6f826f, + 0xbef6006f, 0xb8f82a05, + 0x80788178, 0x8e788a78, + 0xb8ee1605, 0x806e816e, + 0x8e6e866e, 0x80786e78, + 0x8078ff78, 0x00000080, + 0xbef600ff, 0x01000000, + 0xbefc0080, 0xe0510000, + 0x781d0000, 0xe0510100, + 0x781d0000, 0x807cff7c, + 0x00000200, 0x8078ff78, + 0x00000200, 0xbf0a6f7c, + 0xbf85fff6, 0xbef80080, + 0xbefe00c1, 0xbeff00c1, + 0xb8ef2a05, 0x806f816f, + 0x8e6f826f, 0x8e76886f, + 0xbef600ff, 0x01000000, + 0xbeee0078, 0x8078ff78, + 0x00000400, 0xbefc0084, + 0xbf11087c, 0x806fff6f, + 0x00008000, 0xe0524000, + 0x781d0000, 0xe0524100, + 0x781d0100, 0xe0524200, + 0x781d0200, 0xe0524300, + 0x781d0300, 0xbf8c0f70, + 0x7e000300, 0x7e020301, + 0x7e040302, 0x7e060303, + 0x807c847c, 0x8078ff78, + 0x00000400, 0xbf0a6f7c, + 0xbf85ffee, 0xbf9c0000, + 0xe0524000, 0x6e1d0000, + 0xe0524100, 0x6e1d0100, + 0xe0524200, 0x6e1d0200, + 0xe0524300, 0x6e1d0300, + 0xb8f82a05, 0x80788178, + 0x8e788a78, 0xb8ee1605, + 0x806e816e, 0x8e6e866e, + 0x80786e78, 0x80f8c078, + 0xb8ef1605, 0x806f816f, + 0x8e6f846f, 0x8e76826f, + 0xbef600ff, 0x01000000, + 0xbefc006f, 0xc031003a, + 0x00000078, 0x80f8c078, + 0xbf8cc07f, 0x80fc907c, + 0xbf800000, 0xbe802d00, + 0xbe822d02, 0xbe842d04, + 0xbe862d06, 0xbe882d08, + 0xbe8a2d0a, 0xbe8c2d0c, + 0xbe8e2d0e, 0xbf06807c, + 0xbf84fff0, 0xb8f82a05, + 0x80788178, 0x8e788a78, + 0xb8ee1605, 0x806e816e, + 0x8e6e866e, 0x80786e78, + 0xbef60084, 0xbef600ff, + 0x01000000, 0xc0211bfa, + 0x00000078, 0x80788478, + 0xc0211b3a, 0x00000078, + 0x80788478, 0xc0211b7a, + 0x00000078, 0x80788478, + 0xc0211eba, 0x00000078, + 0x80788478, 0xc0211efa, + 0x00000078, 0x80788478, + 0xc0211c3a, 0x00000078, + 0x80788478, 0xc0211c7a, + 0x00000078, 0x80788478, + 0xc0211a3a, 0x00000078, + 0x80788478, 0xc0211a7a, + 0x00000078, 0x80788478, + 0xc0211cfa, 0x00000078, + 0x80788478, 0xbf8cc07f, + 0xbefc006f, 0xbefe007a, + 0xbeff007b, 0x866f71ff, + 0x000003ff, 0xb96f4803, + 0x866f71ff, 0xfffff800, + 0x8f6f8b6f, 0xb96fa2c3, + 0xb973f801, 0xb8ee2a05, + 0x806e816e, 0x8e6e8a6e, + 0xb8ef1605, 0x806f816f, + 0x8e6f866f, 0x806e6f6e, + 0x806e746e, 0x826f8075, + 0x866fff6f, 0x0000ffff, + 0xc0071cb7, 0x00000040, + 0xc00b1d37, 0x00000048, + 0xc0031e77, 0x00000058, + 0xc0071eb7, 0x0000005c, + 0xbf8cc07f, 0x866fff6d, + 0xf0000000, 0x8f6f9c6f, + 0x8e6f906f, 0xbeee0080, + 0x876e6f6e, 0x866fff6d, + 0x08000000, 0x8f6f9b6f, + 0x8e6f8f6f, 0x876e6f6e, + 0x866fff70, 0x00800000, + 0x8f6f976f, 0xb96ef807, + 0x866dff6d, 0x0000ffff, + 0x86fe7e7e, 0x86ea6a6a, + 0xb970f802, 0xbf8a0000, + 0x95806f6c, 0xbf810000, +}; diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx8.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx8.asm index 997a383dcb8b7..a2a04bb64096f 100644 --- a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx8.asm +++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx8.asm @@ -20,9 +20,12 @@ * OTHER DEALINGS IN THE SOFTWARE. */ -#if 0 -HW (VI) source code for CWSR trap handler -#Version 18 + multiple trap handler +/* To compile this assembly code: + * PROJECT=vi ./sp3 cwsr_trap_handler_gfx8.asm -hex tmp.hex + */ + +/* HW (VI) source code for CWSR trap handler */ +/* Version 18 + multiple trap handler */ // this performance-optimal version was originally from Seven Xu at SRDC @@ -98,6 +101,7 @@ var SWIZZLE_EN = 0 //whether we use swi /**************************************************************************/ var SQ_WAVE_STATUS_INST_ATC_SHIFT = 23 var SQ_WAVE_STATUS_INST_ATC_MASK = 0x00800000 +var SQ_WAVE_STATUS_SPI_PRIO_SHIFT = 1 var SQ_WAVE_STATUS_SPI_PRIO_MASK = 0x00000006 var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT = 12 @@ -149,7 +153,7 @@ var s_save_spi_init_lo = exec_lo var s_save_spi_init_hi = exec_hi //tba_lo and tba_hi need to be saved/restored -var s_save_pc_lo = ttmp0 //{TTMP1, TTMP0} = {3??h0,pc_rewind[3:0], HT[0],trapID[7:0], PC[47:0]} +var s_save_pc_lo = ttmp0 //{TTMP1, TTMP0} = {3'h0,pc_rewind[3:0], HT[0],trapID[7:0], PC[47:0]} var s_save_pc_hi = ttmp1 var s_save_exec_lo = ttmp2 var s_save_exec_hi = ttmp3 @@ -319,6 +323,10 @@ end s_sendmsg sendmsg(MSG_SAVEWAVE) //send SPI a message and wait for SPI's write to EXEC end + // Set SPI_PRIO=2 to avoid starving instruction fetch in the waves we're waiting for. + s_or_b32 s_save_tmp, s_save_status, (2 << SQ_WAVE_STATUS_SPI_PRIO_SHIFT) + s_setreg_b32 hwreg(HW_REG_STATUS), s_save_tmp + L_SLEEP: s_sleep 0x2 // sleep 1 (64clk) is not enough for 8 waves per SIMD, which will cause SQ hang, since the 7,8th wave could not get arbit to exec inst, while other waves are stuck into the sleep-loop and waiting for wrexec!=0 @@ -1007,8 +1015,6 @@ end s_waitcnt lgkmcnt(0) //from now on, it is safe to restore STATUS and IB_STS - s_and_b32 s_restore_pc_hi, s_restore_pc_hi, 0x0000ffff //pc[47:32] //Do it here in order not to affect STATUS - //for normal save & restore, the saved PC points to the next inst to execute, no adjustment needs to be made, otherwise: if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_RESTORE_NORMAL)) s_add_u32 s_restore_pc_lo, s_restore_pc_lo, 8 //pc[31:0]+8 //two back-to-back s_trap are used (first for save and second for restore) @@ -1044,6 +1050,7 @@ end s_lshr_b32 s_restore_m0, s_restore_m0, SQ_WAVE_STATUS_INST_ATC_SHIFT s_setreg_b32 hwreg(HW_REG_IB_STS), s_restore_tmp + s_and_b32 s_restore_pc_hi, s_restore_pc_hi, 0x0000ffff //pc[47:32] //Do it here in order not to affect STATUS s_and_b64 exec, exec, exec // Restore STATUS.EXECZ, not writable by s_setreg_b32 s_and_b64 vcc, vcc, vcc // Restore STATUS.VCCZ, not writable by s_setreg_b32 s_setreg_b32 hwreg(HW_REG_STATUS), s_restore_status // SCC is included, which is changed by previous salu @@ -1127,258 +1134,3 @@ end function get_hwreg_size_bytes return 128 //HWREG size 128 bytes end - - -#endif - -static const uint32_t cwsr_trap_gfx8_hex[] = { - 0xbf820001, 0xbf820123, - 0xb8f4f802, 0x89748674, - 0xb8f5f803, 0x8675ff75, - 0x00000400, 0xbf850011, - 0xc00a1e37, 0x00000000, - 0xbf8c007f, 0x87777978, - 0xbf840002, 0xb974f802, - 0xbe801d78, 0xb8f5f803, - 0x8675ff75, 0x000001ff, - 0xbf850002, 0x80708470, - 0x82718071, 0x8671ff71, - 0x0000ffff, 0xb974f802, - 0xbe801f70, 0xb8f5f803, - 0x8675ff75, 0x00000100, - 0xbf840006, 0xbefa0080, - 0xb97a0203, 0x8671ff71, - 0x0000ffff, 0x80f08870, - 0x82f18071, 0xbefa0080, - 0xb97a0283, 0xbef60068, - 0xbef70069, 0xb8fa1c07, - 0x8e7a9c7a, 0x87717a71, - 0xb8fa03c7, 0x8e7a9b7a, - 0x87717a71, 0xb8faf807, - 0x867aff7a, 0x00007fff, - 0xb97af807, 0xbef2007e, - 0xbef3007f, 0xbefe0180, - 0xbf900004, 0xbf8e0002, - 0xbf88fffe, 0xbef8007e, - 0x8679ff7f, 0x0000ffff, - 0x8779ff79, 0x00040000, - 0xbefa0080, 0xbefb00ff, - 0x00807fac, 0x867aff7f, - 0x08000000, 0x8f7a837a, - 0x877b7a7b, 0x867aff7f, - 0x70000000, 0x8f7a817a, - 0x877b7a7b, 0xbeef007c, - 0xbeee0080, 0xb8ee2a05, - 0x806e816e, 0x8e6e8a6e, - 0xb8fa1605, 0x807a817a, - 0x8e7a867a, 0x806e7a6e, - 0xbefa0084, 0xbefa00ff, - 0x01000000, 0xbefe007c, - 0xbefc006e, 0xc0611bfc, - 0x0000007c, 0x806e846e, - 0xbefc007e, 0xbefe007c, - 0xbefc006e, 0xc0611c3c, - 0x0000007c, 0x806e846e, - 0xbefc007e, 0xbefe007c, - 0xbefc006e, 0xc0611c7c, - 0x0000007c, 0x806e846e, - 0xbefc007e, 0xbefe007c, - 0xbefc006e, 0xc0611cbc, - 0x0000007c, 0x806e846e, - 0xbefc007e, 0xbefe007c, - 0xbefc006e, 0xc0611cfc, - 0x0000007c, 0x806e846e, - 0xbefc007e, 0xbefe007c, - 0xbefc006e, 0xc0611d3c, - 0x0000007c, 0x806e846e, - 0xbefc007e, 0xb8f5f803, - 0xbefe007c, 0xbefc006e, - 0xc0611d7c, 0x0000007c, - 0x806e846e, 0xbefc007e, - 0xbefe007c, 0xbefc006e, - 0xc0611dbc, 0x0000007c, - 0x806e846e, 0xbefc007e, - 0xbefe007c, 0xbefc006e, - 0xc0611dfc, 0x0000007c, - 0x806e846e, 0xbefc007e, - 0xb8eff801, 0xbefe007c, - 0xbefc006e, 0xc0611bfc, - 0x0000007c, 0x806e846e, - 0xbefc007e, 0xbefe007c, - 0xbefc006e, 0xc0611b3c, - 0x0000007c, 0x806e846e, - 0xbefc007e, 0xbefe007c, - 0xbefc006e, 0xc0611b7c, - 0x0000007c, 0x806e846e, - 0xbefc007e, 0x867aff7f, - 0x04000000, 0xbef30080, - 0x8773737a, 0xb8ee2a05, - 0x806e816e, 0x8e6e8a6e, - 0xb8f51605, 0x80758175, - 0x8e758475, 0x8e7a8275, - 0xbefa00ff, 0x01000000, - 0xbef60178, 0x80786e78, - 0x82798079, 0xbefc0080, - 0xbe802b00, 0xbe822b02, - 0xbe842b04, 0xbe862b06, - 0xbe882b08, 0xbe8a2b0a, - 0xbe8c2b0c, 0xbe8e2b0e, - 0xc06b003c, 0x00000000, - 0xc06b013c, 0x00000010, - 0xc06b023c, 0x00000020, - 0xc06b033c, 0x00000030, - 0x8078c078, 0x82798079, - 0x807c907c, 0xbf0a757c, - 0xbf85ffeb, 0xbef80176, - 0xbeee0080, 0xbefe00c1, - 0xbeff00c1, 0xbefa00ff, - 0x01000000, 0xe0724000, - 0x6e1e0000, 0xe0724100, - 0x6e1e0100, 0xe0724200, - 0x6e1e0200, 0xe0724300, - 0x6e1e0300, 0xbefe00c1, - 0xbeff00c1, 0xb8f54306, - 0x8675c175, 0xbf84002c, - 0xbf8a0000, 0x867aff73, - 0x04000000, 0xbf840028, - 0x8e758675, 0x8e758275, - 0xbefa0075, 0xb8ee2a05, - 0x806e816e, 0x8e6e8a6e, - 0xb8fa1605, 0x807a817a, - 0x8e7a867a, 0x806e7a6e, - 0x806eff6e, 0x00000080, - 0xbefa00ff, 0x01000000, - 0xbefc0080, 0xd28c0002, - 0x000100c1, 0xd28d0003, - 0x000204c1, 0xd1060002, - 0x00011103, 0x7e0602ff, - 0x00000200, 0xbefc00ff, - 0x00010000, 0xbe80007b, - 0x867bff7b, 0xff7fffff, - 0x877bff7b, 0x00058000, - 0xd8ec0000, 0x00000002, - 0xbf8c007f, 0xe0765000, - 0x6e1e0002, 0x32040702, - 0xd0c9006a, 0x0000eb02, - 0xbf87fff7, 0xbefb0000, - 0xbeee00ff, 0x00000400, - 0xbefe00c1, 0xbeff00c1, - 0xb8f52a05, 0x80758175, - 0x8e758275, 0x8e7a8875, - 0xbefa00ff, 0x01000000, - 0xbefc0084, 0xbf0a757c, - 0xbf840015, 0xbf11017c, - 0x8075ff75, 0x00001000, - 0x7e000300, 0x7e020301, - 0x7e040302, 0x7e060303, - 0xe0724000, 0x6e1e0000, - 0xe0724100, 0x6e1e0100, - 0xe0724200, 0x6e1e0200, - 0xe0724300, 0x6e1e0300, - 0x807c847c, 0x806eff6e, - 0x00000400, 0xbf0a757c, - 0xbf85ffef, 0xbf9c0000, - 0xbf8200ca, 0xbef8007e, - 0x8679ff7f, 0x0000ffff, - 0x8779ff79, 0x00040000, - 0xbefa0080, 0xbefb00ff, - 0x00807fac, 0x8676ff7f, - 0x08000000, 0x8f768376, - 0x877b767b, 0x8676ff7f, - 0x70000000, 0x8f768176, - 0x877b767b, 0x8676ff7f, - 0x04000000, 0xbf84001e, - 0xbefe00c1, 0xbeff00c1, - 0xb8f34306, 0x8673c173, - 0xbf840019, 0x8e738673, - 0x8e738273, 0xbefa0073, - 0xb8f22a05, 0x80728172, - 0x8e728a72, 0xb8f61605, - 0x80768176, 0x8e768676, - 0x80727672, 0x8072ff72, - 0x00000080, 0xbefa00ff, - 0x01000000, 0xbefc0080, - 0xe0510000, 0x721e0000, - 0xe0510100, 0x721e0000, - 0x807cff7c, 0x00000200, - 0x8072ff72, 0x00000200, - 0xbf0a737c, 0xbf85fff6, - 0xbef20080, 0xbefe00c1, - 0xbeff00c1, 0xb8f32a05, - 0x80738173, 0x8e738273, - 0x8e7a8873, 0xbefa00ff, - 0x01000000, 0xbef60072, - 0x8072ff72, 0x00000400, - 0xbefc0084, 0xbf11087c, - 0x8073ff73, 0x00008000, - 0xe0524000, 0x721e0000, - 0xe0524100, 0x721e0100, - 0xe0524200, 0x721e0200, - 0xe0524300, 0x721e0300, - 0xbf8c0f70, 0x7e000300, - 0x7e020301, 0x7e040302, - 0x7e060303, 0x807c847c, - 0x8072ff72, 0x00000400, - 0xbf0a737c, 0xbf85ffee, - 0xbf9c0000, 0xe0524000, - 0x761e0000, 0xe0524100, - 0x761e0100, 0xe0524200, - 0x761e0200, 0xe0524300, - 0x761e0300, 0xb8f22a05, - 0x80728172, 0x8e728a72, - 0xb8f61605, 0x80768176, - 0x8e768676, 0x80727672, - 0x80f2c072, 0xb8f31605, - 0x80738173, 0x8e738473, - 0x8e7a8273, 0xbefa00ff, - 0x01000000, 0xbefc0073, - 0xc031003c, 0x00000072, - 0x80f2c072, 0xbf8c007f, - 0x80fc907c, 0xbe802d00, - 0xbe822d02, 0xbe842d04, - 0xbe862d06, 0xbe882d08, - 0xbe8a2d0a, 0xbe8c2d0c, - 0xbe8e2d0e, 0xbf06807c, - 0xbf84fff1, 0xb8f22a05, - 0x80728172, 0x8e728a72, - 0xb8f61605, 0x80768176, - 0x8e768676, 0x80727672, - 0xbefa0084, 0xbefa00ff, - 0x01000000, 0xc0211cfc, - 0x00000072, 0x80728472, - 0xc0211c3c, 0x00000072, - 0x80728472, 0xc0211c7c, - 0x00000072, 0x80728472, - 0xc0211bbc, 0x00000072, - 0x80728472, 0xc0211bfc, - 0x00000072, 0x80728472, - 0xc0211d3c, 0x00000072, - 0x80728472, 0xc0211d7c, - 0x00000072, 0x80728472, - 0xc0211a3c, 0x00000072, - 0x80728472, 0xc0211a7c, - 0x00000072, 0x80728472, - 0xc0211dfc, 0x00000072, - 0x80728472, 0xc0211b3c, - 0x00000072, 0x80728472, - 0xc0211b7c, 0x00000072, - 0x80728472, 0xbf8c007f, - 0x8671ff71, 0x0000ffff, - 0xbefc0073, 0xbefe006e, - 0xbeff006f, 0x867375ff, - 0x000003ff, 0xb9734803, - 0x867375ff, 0xfffff800, - 0x8f738b73, 0xb973a2c3, - 0xb977f801, 0x8673ff71, - 0xf0000000, 0x8f739c73, - 0x8e739073, 0xbef60080, - 0x87767376, 0x8673ff71, - 0x08000000, 0x8f739b73, - 0x8e738f73, 0x87767376, - 0x8673ff74, 0x00800000, - 0x8f739773, 0xb976f807, - 0x86fe7e7e, 0x86ea6a6a, - 0xb974f802, 0xbf8a0000, - 0x95807370, 0xbf810000, -}; - diff --git a/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm new file mode 100644 index 0000000000000..998be96be7361 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/cwsr_trap_handler_gfx9.asm @@ -0,0 +1,1214 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +/* To compile this assembly code: + * PROJECT=greenland ./sp3 cwsr_trap_handler_gfx9.asm -hex tmp.hex + */ + +/* HW (GFX9) source code for CWSR trap handler */ +/* Version 18 + multiple trap handler */ + +// this performance-optimal version was originally from Seven Xu at SRDC + +// Revison #18 --... +/* Rev History +** #1. Branch from gc dv. //gfxip/gfx9/main/src/test/suites/block/cs/sr/cs_trap_handler.sp3#1,#50, #51, #52-53(Skip, Already Fixed by PV), #54-56(merged),#57-58(mergerd, skiped-already fixed by PV) +** #4. SR Memory Layout: +** 1. VGPR-SGPR-HWREG-{LDS} +** 2. tba_hi.bits.26 - reconfigured as the first wave in tg bits, for defer Save LDS for a threadgroup.. performance concern.. +** #5. Update: 1. Accurate g8sr_ts_save_d timestamp +** #6. Update: 1. Fix s_barrier usage; 2. VGPR s/r using swizzle buffer?(NoNeed, already matched the swizzle pattern, more investigation) +** #7. Update: 1. don't barrier if noLDS +** #8. Branch: 1. Branch to ver#0, which is very similar to gc dv version +** 2. Fix SQ issue by s_sleep 2 +** #9. Update: 1. Fix scc restore failed issue, restore wave_status at last +** 2. optimize s_buffer save by burst 16sgprs... +** #10. Update 1. Optimize restore sgpr by busrt 16 sgprs. +** #11. Update 1. Add 2 more timestamp for debug version +** #12. Update 1. Add VGPR SR using DWx4, some case improve and some case drop performance +** #13. Integ 1. Always use MUBUF for PV trap shader... +** #14. Update 1. s_buffer_store soft clause... +** #15. Update 1. PERF - sclar write with glc:0/mtype0 to allow L2 combine. perf improvement a lot. +** #16. Update 1. PRRF - UNROLL LDS_DMA got 2500cycle save in IP tree +** #17. Update 1. FUNC - LDS_DMA has issues while ATC, replace with ds_read/buffer_store for save part[TODO restore part] +** 2. PERF - Save LDS before save VGPR to cover LDS save long latency... +** #18. Update 1. FUNC - Implicitly estore STATUS.VCCZ, which is not writable by s_setreg_b32 +** 2. FUNC - Handle non-CWSR traps +*/ + +var G8SR_WDMEM_HWREG_OFFSET = 0 +var G8SR_WDMEM_SGPR_OFFSET = 128 // in bytes + +// Keep definition same as the app shader, These 2 time stamps are part of the app shader... Should before any Save and after restore. + +var G8SR_DEBUG_TIMESTAMP = 0 +var G8SR_DEBUG_TS_SAVE_D_OFFSET = 40*4 // ts_save_d timestamp offset relative to SGPR_SR_memory_offset +var s_g8sr_ts_save_s = s[34:35] // save start +var s_g8sr_ts_sq_save_msg = s[36:37] // The save shader send SAVEWAVE msg to spi +var s_g8sr_ts_spi_wrexec = s[38:39] // the SPI write the sr address to SQ +var s_g8sr_ts_save_d = s[40:41] // save end +var s_g8sr_ts_restore_s = s[42:43] // restore start +var s_g8sr_ts_restore_d = s[44:45] // restore end + +var G8SR_VGPR_SR_IN_DWX4 = 0 +var G8SR_SAVE_BUF_RSRC_WORD1_STRIDE_DWx4 = 0x00100000 // DWx4 stride is 4*4Bytes +var G8SR_RESTORE_BUF_RSRC_WORD1_STRIDE_DWx4 = G8SR_SAVE_BUF_RSRC_WORD1_STRIDE_DWx4 + + +/*************************************************************************/ +/* control on how to run the shader */ +/*************************************************************************/ +//any hack that needs to be made to run this code in EMU (either because various EMU code are not ready or no compute save & restore in EMU run) +var EMU_RUN_HACK = 0 +var EMU_RUN_HACK_RESTORE_NORMAL = 0 +var EMU_RUN_HACK_SAVE_NORMAL_EXIT = 0 +var EMU_RUN_HACK_SAVE_SINGLE_WAVE = 0 +var EMU_RUN_HACK_SAVE_FIRST_TIME = 0 //for interrupted restore in which the first save is through EMU_RUN_HACK +var SAVE_LDS = 1 +var WG_BASE_ADDR_LO = 0x9000a000 +var WG_BASE_ADDR_HI = 0x0 +var WAVE_SPACE = 0x5000 //memory size that each wave occupies in workgroup state mem +var CTX_SAVE_CONTROL = 0x0 +var CTX_RESTORE_CONTROL = CTX_SAVE_CONTROL +var SIM_RUN_HACK = 0 //any hack that needs to be made to run this code in SIM (either because various RTL code are not ready or no compute save & restore in RTL run) +var SGPR_SAVE_USE_SQC = 1 //use SQC D$ to do the write +var USE_MTBUF_INSTEAD_OF_MUBUF = 0 //because TC EMU currently asserts on 0 of // overload DFMT field to carry 4 more bits of stride for MUBUF opcodes +var SWIZZLE_EN = 0 //whether we use swizzled buffer addressing +var ACK_SQC_STORE = 1 //workaround for suspected SQC store bug causing incorrect stores under concurrency + +/**************************************************************************/ +/* variables */ +/**************************************************************************/ +var SQ_WAVE_STATUS_INST_ATC_SHIFT = 23 +var SQ_WAVE_STATUS_INST_ATC_MASK = 0x00800000 +var SQ_WAVE_STATUS_SPI_PRIO_SHIFT = 1 +var SQ_WAVE_STATUS_SPI_PRIO_MASK = 0x00000006 +var SQ_WAVE_STATUS_HALT_MASK = 0x2000 + +var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT = 12 +var SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE = 9 +var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT = 8 +var SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE = 6 +var SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SHIFT = 24 +var SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SIZE = 3 //FIXME sq.blk still has 4 bits at this time while SQ programming guide has 3 bits + +var SQ_WAVE_TRAPSTS_SAVECTX_MASK = 0x400 +var SQ_WAVE_TRAPSTS_EXCE_MASK = 0x1FF // Exception mask +var SQ_WAVE_TRAPSTS_SAVECTX_SHIFT = 10 +var SQ_WAVE_TRAPSTS_MEM_VIOL_MASK = 0x100 +var SQ_WAVE_TRAPSTS_MEM_VIOL_SHIFT = 8 +var SQ_WAVE_TRAPSTS_PRE_SAVECTX_MASK = 0x3FF +var SQ_WAVE_TRAPSTS_PRE_SAVECTX_SHIFT = 0x0 +var SQ_WAVE_TRAPSTS_PRE_SAVECTX_SIZE = 10 +var SQ_WAVE_TRAPSTS_POST_SAVECTX_MASK = 0xFFFFF800 +var SQ_WAVE_TRAPSTS_POST_SAVECTX_SHIFT = 11 +var SQ_WAVE_TRAPSTS_POST_SAVECTX_SIZE = 21 +var SQ_WAVE_TRAPSTS_ILLEGAL_INST_MASK = 0x800 + +var SQ_WAVE_IB_STS_RCNT_SHIFT = 16 //FIXME +var SQ_WAVE_IB_STS_FIRST_REPLAY_SHIFT = 15 //FIXME +var SQ_WAVE_IB_STS_RCNT_FIRST_REPLAY_MASK = 0x1F8000 +var SQ_WAVE_IB_STS_RCNT_FIRST_REPLAY_MASK_NEG = 0x00007FFF //FIXME + +var SQ_BUF_RSRC_WORD1_ATC_SHIFT = 24 +var SQ_BUF_RSRC_WORD3_MTYPE_SHIFT = 27 + +var TTMP11_SAVE_RCNT_FIRST_REPLAY_SHIFT = 26 // bits [31:26] unused by SPI debug data +var TTMP11_SAVE_RCNT_FIRST_REPLAY_MASK = 0xFC000000 + +/* Save */ +var S_SAVE_BUF_RSRC_WORD1_STRIDE = 0x00040000 //stride is 4 bytes +var S_SAVE_BUF_RSRC_WORD3_MISC = 0x00807FAC //SQ_SEL_X/Y/Z/W, BUF_NUM_FORMAT_FLOAT, (0 for MUBUF stride[17:14] when ADD_TID_ENABLE and BUF_DATA_FORMAT_32 for MTBUF), ADD_TID_ENABLE + +var S_SAVE_SPI_INIT_ATC_MASK = 0x08000000 //bit[27]: ATC bit +var S_SAVE_SPI_INIT_ATC_SHIFT = 27 +var S_SAVE_SPI_INIT_MTYPE_MASK = 0x70000000 //bit[30:28]: Mtype +var S_SAVE_SPI_INIT_MTYPE_SHIFT = 28 +var S_SAVE_SPI_INIT_FIRST_WAVE_MASK = 0x04000000 //bit[26]: FirstWaveInTG +var S_SAVE_SPI_INIT_FIRST_WAVE_SHIFT = 26 + +var S_SAVE_PC_HI_RCNT_SHIFT = 28 //FIXME check with Brian to ensure all fields other than PC[47:0] can be used +var S_SAVE_PC_HI_RCNT_MASK = 0xF0000000 //FIXME +var S_SAVE_PC_HI_FIRST_REPLAY_SHIFT = 27 //FIXME +var S_SAVE_PC_HI_FIRST_REPLAY_MASK = 0x08000000 //FIXME + +var s_save_spi_init_lo = exec_lo +var s_save_spi_init_hi = exec_hi + +var s_save_pc_lo = ttmp0 //{TTMP1, TTMP0} = {3'h0,pc_rewind[3:0], HT[0],trapID[7:0], PC[47:0]} +var s_save_pc_hi = ttmp1 +var s_save_exec_lo = ttmp2 +var s_save_exec_hi = ttmp3 +var s_save_tmp = ttmp4 +var s_save_trapsts = ttmp5 //not really used until the end of the SAVE routine +var s_save_xnack_mask_lo = ttmp6 +var s_save_xnack_mask_hi = ttmp7 +var s_save_buf_rsrc0 = ttmp8 +var s_save_buf_rsrc1 = ttmp9 +var s_save_buf_rsrc2 = ttmp10 +var s_save_buf_rsrc3 = ttmp11 +var s_save_status = ttmp12 +var s_save_mem_offset = ttmp14 +var s_save_alloc_size = s_save_trapsts //conflict +var s_save_m0 = ttmp15 +var s_save_ttmps_lo = s_save_tmp //no conflict +var s_save_ttmps_hi = s_save_trapsts //no conflict + +/* Restore */ +var S_RESTORE_BUF_RSRC_WORD1_STRIDE = S_SAVE_BUF_RSRC_WORD1_STRIDE +var S_RESTORE_BUF_RSRC_WORD3_MISC = S_SAVE_BUF_RSRC_WORD3_MISC + +var S_RESTORE_SPI_INIT_ATC_MASK = 0x08000000 //bit[27]: ATC bit +var S_RESTORE_SPI_INIT_ATC_SHIFT = 27 +var S_RESTORE_SPI_INIT_MTYPE_MASK = 0x70000000 //bit[30:28]: Mtype +var S_RESTORE_SPI_INIT_MTYPE_SHIFT = 28 +var S_RESTORE_SPI_INIT_FIRST_WAVE_MASK = 0x04000000 //bit[26]: FirstWaveInTG +var S_RESTORE_SPI_INIT_FIRST_WAVE_SHIFT = 26 + +var S_RESTORE_PC_HI_RCNT_SHIFT = S_SAVE_PC_HI_RCNT_SHIFT +var S_RESTORE_PC_HI_RCNT_MASK = S_SAVE_PC_HI_RCNT_MASK +var S_RESTORE_PC_HI_FIRST_REPLAY_SHIFT = S_SAVE_PC_HI_FIRST_REPLAY_SHIFT +var S_RESTORE_PC_HI_FIRST_REPLAY_MASK = S_SAVE_PC_HI_FIRST_REPLAY_MASK + +var s_restore_spi_init_lo = exec_lo +var s_restore_spi_init_hi = exec_hi + +var s_restore_mem_offset = ttmp12 +var s_restore_alloc_size = ttmp3 +var s_restore_tmp = ttmp2 +var s_restore_mem_offset_save = s_restore_tmp //no conflict + +var s_restore_m0 = s_restore_alloc_size //no conflict + +var s_restore_mode = ttmp7 + +var s_restore_pc_lo = ttmp0 +var s_restore_pc_hi = ttmp1 +var s_restore_exec_lo = ttmp14 +var s_restore_exec_hi = ttmp15 +var s_restore_status = ttmp4 +var s_restore_trapsts = ttmp5 +var s_restore_xnack_mask_lo = xnack_mask_lo +var s_restore_xnack_mask_hi = xnack_mask_hi +var s_restore_buf_rsrc0 = ttmp8 +var s_restore_buf_rsrc1 = ttmp9 +var s_restore_buf_rsrc2 = ttmp10 +var s_restore_buf_rsrc3 = ttmp11 +var s_restore_ttmps_lo = s_restore_tmp //no conflict +var s_restore_ttmps_hi = s_restore_alloc_size //no conflict + +/**************************************************************************/ +/* trap handler entry points */ +/**************************************************************************/ +/* Shader Main*/ + +shader main + asic(GFX9) + type(CS) + + + if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_RESTORE_NORMAL)) //hack to use trap_id for determining save/restore + //FIXME VCCZ un-init assertion s_getreg_b32 s_save_status, hwreg(HW_REG_STATUS) //save STATUS since we will change SCC + s_and_b32 s_save_tmp, s_save_pc_hi, 0xffff0000 //change SCC + s_cmp_eq_u32 s_save_tmp, 0x007e0000 //Save: trap_id = 0x7e. Restore: trap_id = 0x7f. + s_cbranch_scc0 L_JUMP_TO_RESTORE //do not need to recover STATUS here since we are going to RESTORE + //FIXME s_setreg_b32 hwreg(HW_REG_STATUS), s_save_status //need to recover STATUS since we are going to SAVE + s_branch L_SKIP_RESTORE //NOT restore, SAVE actually + else + s_branch L_SKIP_RESTORE //NOT restore. might be a regular trap or save + end + +L_JUMP_TO_RESTORE: + s_branch L_RESTORE //restore + +L_SKIP_RESTORE: + + s_getreg_b32 s_save_status, hwreg(HW_REG_STATUS) //save STATUS since we will change SCC + s_andn2_b32 s_save_status, s_save_status, SQ_WAVE_STATUS_SPI_PRIO_MASK //check whether this is for save + s_getreg_b32 s_save_trapsts, hwreg(HW_REG_TRAPSTS) + s_and_b32 ttmp2, s_save_trapsts, SQ_WAVE_TRAPSTS_SAVECTX_MASK //check whether this is for save + s_cbranch_scc1 L_SAVE //this is the operation for save + + // ********* Handle non-CWSR traps ******************* +if (!EMU_RUN_HACK) + // Illegal instruction is a non-maskable exception which blocks context save. + // Halt the wavefront and return from the trap. + s_and_b32 ttmp2, s_save_trapsts, SQ_WAVE_TRAPSTS_ILLEGAL_INST_MASK + s_cbranch_scc1 L_HALT_WAVE + + // If STATUS.MEM_VIOL is asserted then we cannot fetch from the TMA. + // Instead, halt the wavefront and return from the trap. + s_and_b32 ttmp2, s_save_trapsts, SQ_WAVE_TRAPSTS_MEM_VIOL_MASK + s_cbranch_scc0 L_FETCH_2ND_TRAP + +L_HALT_WAVE: + // If STATUS.HALT is set then this fault must come from SQC instruction fetch. + // We cannot prevent further faults so just terminate the wavefront. + s_and_b32 ttmp2, s_save_status, SQ_WAVE_STATUS_HALT_MASK + s_cbranch_scc0 L_NOT_ALREADY_HALTED + s_endpgm +L_NOT_ALREADY_HALTED: + s_or_b32 s_save_status, s_save_status, SQ_WAVE_STATUS_HALT_MASK + + // If the PC points to S_ENDPGM then context save will fail if STATUS.HALT is set. + // Rewind the PC to prevent this from occurring. The debugger compensates for this. + s_sub_u32 ttmp0, ttmp0, 0x8 + s_subb_u32 ttmp1, ttmp1, 0x0 + +L_FETCH_2ND_TRAP: + // Preserve and clear scalar XNACK state before issuing scalar reads. + // Save IB_STS.FIRST_REPLAY[15] and IB_STS.RCNT[20:16] into unused space ttmp11[31:26]. + s_getreg_b32 ttmp2, hwreg(HW_REG_IB_STS) + s_and_b32 ttmp3, ttmp2, SQ_WAVE_IB_STS_RCNT_FIRST_REPLAY_MASK + s_lshl_b32 ttmp3, ttmp3, (TTMP11_SAVE_RCNT_FIRST_REPLAY_SHIFT - SQ_WAVE_IB_STS_FIRST_REPLAY_SHIFT) + s_andn2_b32 ttmp11, ttmp11, TTMP11_SAVE_RCNT_FIRST_REPLAY_MASK + s_or_b32 ttmp11, ttmp11, ttmp3 + + s_andn2_b32 ttmp2, ttmp2, SQ_WAVE_IB_STS_RCNT_FIRST_REPLAY_MASK + s_setreg_b32 hwreg(HW_REG_IB_STS), ttmp2 + + // Read second-level TBA/TMA from first-level TMA and jump if available. + // ttmp[2:5] and ttmp12 can be used (others hold SPI-initialized debug data) + // ttmp12 holds SQ_WAVE_STATUS + s_getreg_b32 ttmp4, hwreg(HW_REG_SQ_SHADER_TMA_LO) + s_getreg_b32 ttmp5, hwreg(HW_REG_SQ_SHADER_TMA_HI) + s_lshl_b64 [ttmp4, ttmp5], [ttmp4, ttmp5], 0x8 + s_load_dwordx2 [ttmp2, ttmp3], [ttmp4, ttmp5], 0x0 glc:1 // second-level TBA + s_waitcnt lgkmcnt(0) + s_load_dwordx2 [ttmp4, ttmp5], [ttmp4, ttmp5], 0x8 glc:1 // second-level TMA + s_waitcnt lgkmcnt(0) + s_and_b64 [ttmp2, ttmp3], [ttmp2, ttmp3], [ttmp2, ttmp3] + s_cbranch_scc0 L_NO_NEXT_TRAP // second-level trap handler not been set + s_setpc_b64 [ttmp2, ttmp3] // jump to second-level trap handler + +L_NO_NEXT_TRAP: + s_getreg_b32 s_save_trapsts, hwreg(HW_REG_TRAPSTS) + s_and_b32 s_save_trapsts, s_save_trapsts, SQ_WAVE_TRAPSTS_EXCE_MASK // Check whether it is an exception + s_cbranch_scc1 L_EXCP_CASE // Exception, jump back to the shader program directly. + s_add_u32 ttmp0, ttmp0, 4 // S_TRAP case, add 4 to ttmp0 + s_addc_u32 ttmp1, ttmp1, 0 +L_EXCP_CASE: + s_and_b32 ttmp1, ttmp1, 0xFFFF + + // Restore SQ_WAVE_IB_STS. + s_lshr_b32 ttmp2, ttmp11, (TTMP11_SAVE_RCNT_FIRST_REPLAY_SHIFT - SQ_WAVE_IB_STS_FIRST_REPLAY_SHIFT) + s_and_b32 ttmp2, ttmp2, SQ_WAVE_IB_STS_RCNT_FIRST_REPLAY_MASK + s_setreg_b32 hwreg(HW_REG_IB_STS), ttmp2 + + // Restore SQ_WAVE_STATUS. + s_and_b64 exec, exec, exec // Restore STATUS.EXECZ, not writable by s_setreg_b32 + s_and_b64 vcc, vcc, vcc // Restore STATUS.VCCZ, not writable by s_setreg_b32 + s_setreg_b32 hwreg(HW_REG_STATUS), s_save_status + + s_rfe_b64 [ttmp0, ttmp1] +end + // ********* End handling of non-CWSR traps ******************* + +/**************************************************************************/ +/* save routine */ +/**************************************************************************/ + +L_SAVE: + +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_save_s + s_waitcnt lgkmcnt(0) //FIXME, will cause xnack?? +end + + s_and_b32 s_save_pc_hi, s_save_pc_hi, 0x0000ffff //pc[47:32] + + s_mov_b32 s_save_tmp, 0 //clear saveCtx bit + s_setreg_b32 hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_SAVECTX_SHIFT, 1), s_save_tmp //clear saveCtx bit + + s_getreg_b32 s_save_tmp, hwreg(HW_REG_IB_STS, SQ_WAVE_IB_STS_RCNT_SHIFT, SQ_WAVE_IB_STS_RCNT_SIZE) //save RCNT + s_lshl_b32 s_save_tmp, s_save_tmp, S_SAVE_PC_HI_RCNT_SHIFT + s_or_b32 s_save_pc_hi, s_save_pc_hi, s_save_tmp + s_getreg_b32 s_save_tmp, hwreg(HW_REG_IB_STS, SQ_WAVE_IB_STS_FIRST_REPLAY_SHIFT, SQ_WAVE_IB_STS_FIRST_REPLAY_SIZE) //save FIRST_REPLAY + s_lshl_b32 s_save_tmp, s_save_tmp, S_SAVE_PC_HI_FIRST_REPLAY_SHIFT + s_or_b32 s_save_pc_hi, s_save_pc_hi, s_save_tmp + s_getreg_b32 s_save_tmp, hwreg(HW_REG_IB_STS) //clear RCNT and FIRST_REPLAY in IB_STS + s_and_b32 s_save_tmp, s_save_tmp, SQ_WAVE_IB_STS_RCNT_FIRST_REPLAY_MASK_NEG + + s_setreg_b32 hwreg(HW_REG_IB_STS), s_save_tmp + + /* inform SPI the readiness and wait for SPI's go signal */ + s_mov_b32 s_save_exec_lo, exec_lo //save EXEC and use EXEC for the go signal from SPI + s_mov_b32 s_save_exec_hi, exec_hi + s_mov_b64 exec, 0x0 //clear EXEC to get ready to receive + +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_sq_save_msg + s_waitcnt lgkmcnt(0) +end + + if (EMU_RUN_HACK) + + else + s_sendmsg sendmsg(MSG_SAVEWAVE) //send SPI a message and wait for SPI's write to EXEC + end + + // Set SPI_PRIO=2 to avoid starving instruction fetch in the waves we're waiting for. + s_or_b32 s_save_tmp, s_save_status, (2 << SQ_WAVE_STATUS_SPI_PRIO_SHIFT) + s_setreg_b32 hwreg(HW_REG_STATUS), s_save_tmp + + L_SLEEP: + s_sleep 0x2 // sleep 1 (64clk) is not enough for 8 waves per SIMD, which will cause SQ hang, since the 7,8th wave could not get arbit to exec inst, while other waves are stuck into the sleep-loop and waiting for wrexec!=0 + + if (EMU_RUN_HACK) + + else + s_cbranch_execz L_SLEEP + end + +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_spi_wrexec + s_waitcnt lgkmcnt(0) +end + + if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_SAVE_SINGLE_WAVE)) + //calculate wd_addr using absolute thread id + v_readlane_b32 s_save_tmp, v9, 0 + s_lshr_b32 s_save_tmp, s_save_tmp, 6 + s_mul_i32 s_save_tmp, s_save_tmp, WAVE_SPACE + s_add_i32 s_save_spi_init_lo, s_save_tmp, WG_BASE_ADDR_LO + s_mov_b32 s_save_spi_init_hi, WG_BASE_ADDR_HI + s_and_b32 s_save_spi_init_hi, s_save_spi_init_hi, CTX_SAVE_CONTROL + else + end + if ((EMU_RUN_HACK) && (EMU_RUN_HACK_SAVE_SINGLE_WAVE)) + s_add_i32 s_save_spi_init_lo, s_save_tmp, WG_BASE_ADDR_LO + s_mov_b32 s_save_spi_init_hi, WG_BASE_ADDR_HI + s_and_b32 s_save_spi_init_hi, s_save_spi_init_hi, CTX_SAVE_CONTROL + else + end + + // Save trap temporaries 6-11, 13-15 initialized by SPI debug dispatch logic + // ttmp SR memory offset : size(VGPR)+size(SGPR)+0x40 + get_vgpr_size_bytes(s_save_ttmps_lo) + get_sgpr_size_bytes(s_save_ttmps_hi) + s_add_u32 s_save_ttmps_lo, s_save_ttmps_lo, s_save_ttmps_hi + s_add_u32 s_save_ttmps_lo, s_save_ttmps_lo, s_save_spi_init_lo + s_addc_u32 s_save_ttmps_hi, s_save_spi_init_hi, 0x0 + s_and_b32 s_save_ttmps_hi, s_save_ttmps_hi, 0xFFFF + s_store_dwordx2 [ttmp6, ttmp7], [s_save_ttmps_lo, s_save_ttmps_hi], 0x40 glc:1 + ack_sqc_store_workaround() + s_store_dwordx4 [ttmp8, ttmp9, ttmp10, ttmp11], [s_save_ttmps_lo, s_save_ttmps_hi], 0x48 glc:1 + ack_sqc_store_workaround() + s_store_dword ttmp13, [s_save_ttmps_lo, s_save_ttmps_hi], 0x58 glc:1 + ack_sqc_store_workaround() + s_store_dwordx2 [ttmp14, ttmp15], [s_save_ttmps_lo, s_save_ttmps_hi], 0x5C glc:1 + ack_sqc_store_workaround() + + /* setup Resource Contants */ + s_mov_b32 s_save_buf_rsrc0, s_save_spi_init_lo //base_addr_lo + s_and_b32 s_save_buf_rsrc1, s_save_spi_init_hi, 0x0000FFFF //base_addr_hi + s_or_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, S_SAVE_BUF_RSRC_WORD1_STRIDE + s_mov_b32 s_save_buf_rsrc2, 0 //NUM_RECORDS initial value = 0 (in bytes) although not neccessarily inited + s_mov_b32 s_save_buf_rsrc3, S_SAVE_BUF_RSRC_WORD3_MISC + s_and_b32 s_save_tmp, s_save_spi_init_hi, S_SAVE_SPI_INIT_ATC_MASK + s_lshr_b32 s_save_tmp, s_save_tmp, (S_SAVE_SPI_INIT_ATC_SHIFT-SQ_BUF_RSRC_WORD1_ATC_SHIFT) //get ATC bit into position + s_or_b32 s_save_buf_rsrc3, s_save_buf_rsrc3, s_save_tmp //or ATC + s_and_b32 s_save_tmp, s_save_spi_init_hi, S_SAVE_SPI_INIT_MTYPE_MASK + s_lshr_b32 s_save_tmp, s_save_tmp, (S_SAVE_SPI_INIT_MTYPE_SHIFT-SQ_BUF_RSRC_WORD3_MTYPE_SHIFT) //get MTYPE bits into position + s_or_b32 s_save_buf_rsrc3, s_save_buf_rsrc3, s_save_tmp //or MTYPE + + //FIXME right now s_save_m0/s_save_mem_offset use tma_lo/tma_hi (might need to save them before using them?) + s_mov_b32 s_save_m0, m0 //save M0 + + /* global mem offset */ + s_mov_b32 s_save_mem_offset, 0x0 //mem offset initial value = 0 + + + + + /* save HW registers */ + ////////////////////////////// + + L_SAVE_HWREG: + // HWREG SR memory offset : size(VGPR)+size(SGPR) + get_vgpr_size_bytes(s_save_mem_offset) + get_sgpr_size_bytes(s_save_tmp) + s_add_u32 s_save_mem_offset, s_save_mem_offset, s_save_tmp + + + s_mov_b32 s_save_buf_rsrc2, 0x4 //NUM_RECORDS in bytes + if (SWIZZLE_EN) + s_add_u32 s_save_buf_rsrc2, s_save_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + + write_hwreg_to_mem(s_save_m0, s_save_buf_rsrc0, s_save_mem_offset) //M0 + + if ((EMU_RUN_HACK) && (EMU_RUN_HACK_SAVE_FIRST_TIME)) + s_add_u32 s_save_pc_lo, s_save_pc_lo, 4 //pc[31:0]+4 + s_addc_u32 s_save_pc_hi, s_save_pc_hi, 0x0 //carry bit over + end + + write_hwreg_to_mem(s_save_pc_lo, s_save_buf_rsrc0, s_save_mem_offset) //PC + write_hwreg_to_mem(s_save_pc_hi, s_save_buf_rsrc0, s_save_mem_offset) + write_hwreg_to_mem(s_save_exec_lo, s_save_buf_rsrc0, s_save_mem_offset) //EXEC + write_hwreg_to_mem(s_save_exec_hi, s_save_buf_rsrc0, s_save_mem_offset) + write_hwreg_to_mem(s_save_status, s_save_buf_rsrc0, s_save_mem_offset) //STATUS + + //s_save_trapsts conflicts with s_save_alloc_size + s_getreg_b32 s_save_trapsts, hwreg(HW_REG_TRAPSTS) + write_hwreg_to_mem(s_save_trapsts, s_save_buf_rsrc0, s_save_mem_offset) //TRAPSTS + + write_hwreg_to_mem(xnack_mask_lo, s_save_buf_rsrc0, s_save_mem_offset) //XNACK_MASK_LO + write_hwreg_to_mem(xnack_mask_hi, s_save_buf_rsrc0, s_save_mem_offset) //XNACK_MASK_HI + + //use s_save_tmp would introduce conflict here between s_save_tmp and s_save_buf_rsrc2 + s_getreg_b32 s_save_m0, hwreg(HW_REG_MODE) //MODE + write_hwreg_to_mem(s_save_m0, s_save_buf_rsrc0, s_save_mem_offset) + + + + /* the first wave in the threadgroup */ + s_and_b32 s_save_tmp, s_save_spi_init_hi, S_SAVE_SPI_INIT_FIRST_WAVE_MASK // extract fisrt wave bit + s_mov_b32 s_save_exec_hi, 0x0 + s_or_b32 s_save_exec_hi, s_save_tmp, s_save_exec_hi // save first wave bit to s_save_exec_hi.bits[26] + + + /* save SGPRs */ + // Save SGPR before LDS save, then the s0 to s4 can be used during LDS save... + ////////////////////////////// + + // SGPR SR memory offset : size(VGPR) + get_vgpr_size_bytes(s_save_mem_offset) + // TODO, change RSRC word to rearrange memory layout for SGPRS + + s_getreg_b32 s_save_alloc_size, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SIZE) //spgr_size + s_add_u32 s_save_alloc_size, s_save_alloc_size, 1 + s_lshl_b32 s_save_alloc_size, s_save_alloc_size, 4 //Number of SGPRs = (sgpr_size + 1) * 16 (non-zero value) + + if (SGPR_SAVE_USE_SQC) + s_lshl_b32 s_save_buf_rsrc2, s_save_alloc_size, 2 //NUM_RECORDS in bytes + else + s_lshl_b32 s_save_buf_rsrc2, s_save_alloc_size, 8 //NUM_RECORDS in bytes (64 threads) + end + + if (SWIZZLE_EN) + s_add_u32 s_save_buf_rsrc2, s_save_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + + // backup s_save_buf_rsrc0,1 to s_save_pc_lo/hi, since write_16sgpr_to_mem function will change the rsrc0 + //s_mov_b64 s_save_pc_lo, s_save_buf_rsrc0 + s_mov_b64 s_save_xnack_mask_lo, s_save_buf_rsrc0 + s_add_u32 s_save_buf_rsrc0, s_save_buf_rsrc0, s_save_mem_offset + s_addc_u32 s_save_buf_rsrc1, s_save_buf_rsrc1, 0 + + s_mov_b32 m0, 0x0 //SGPR initial index value =0 + s_nop 0x0 //Manually inserted wait states + L_SAVE_SGPR_LOOP: + // SGPR is allocated in 16 SGPR granularity + s_movrels_b64 s0, s0 //s0 = s[0+m0], s1 = s[1+m0] + s_movrels_b64 s2, s2 //s2 = s[2+m0], s3 = s[3+m0] + s_movrels_b64 s4, s4 //s4 = s[4+m0], s5 = s[5+m0] + s_movrels_b64 s6, s6 //s6 = s[6+m0], s7 = s[7+m0] + s_movrels_b64 s8, s8 //s8 = s[8+m0], s9 = s[9+m0] + s_movrels_b64 s10, s10 //s10 = s[10+m0], s11 = s[11+m0] + s_movrels_b64 s12, s12 //s12 = s[12+m0], s13 = s[13+m0] + s_movrels_b64 s14, s14 //s14 = s[14+m0], s15 = s[15+m0] + + write_16sgpr_to_mem(s0, s_save_buf_rsrc0, s_save_mem_offset) //PV: the best performance should be using s_buffer_store_dwordx4 + s_add_u32 m0, m0, 16 //next sgpr index + s_cmp_lt_u32 m0, s_save_alloc_size //scc = (m0 < s_save_alloc_size) ? 1 : 0 + s_cbranch_scc1 L_SAVE_SGPR_LOOP //SGPR save is complete? + // restore s_save_buf_rsrc0,1 + //s_mov_b64 s_save_buf_rsrc0, s_save_pc_lo + s_mov_b64 s_save_buf_rsrc0, s_save_xnack_mask_lo + + + + + /* save first 4 VGPR, then LDS save could use */ + // each wave will alloc 4 vgprs at least... + ///////////////////////////////////////////////////////////////////////////////////// + + s_mov_b32 s_save_mem_offset, 0 + s_mov_b32 exec_lo, 0xFFFFFFFF //need every thread from now on + s_mov_b32 exec_hi, 0xFFFFFFFF + s_mov_b32 xnack_mask_lo, 0x0 + s_mov_b32 xnack_mask_hi, 0x0 + + if (SWIZZLE_EN) + s_add_u32 s_save_buf_rsrc2, s_save_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + + // VGPR Allocated in 4-GPR granularity + +if G8SR_VGPR_SR_IN_DWX4 + // the const stride for DWx4 is 4*4 bytes + s_and_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, G8SR_SAVE_BUF_RSRC_WORD1_STRIDE_DWx4 // const stride to 4*4 bytes + + buffer_store_dwordx4 v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 + + s_and_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, S_SAVE_BUF_RSRC_WORD1_STRIDE // reset const stride to 4 bytes +else + buffer_store_dword v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 + buffer_store_dword v1, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256 + buffer_store_dword v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*2 + buffer_store_dword v3, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*3 +end + + + + /* save LDS */ + ////////////////////////////// + + L_SAVE_LDS: + + // Change EXEC to all threads... + s_mov_b32 exec_lo, 0xFFFFFFFF //need every thread from now on + s_mov_b32 exec_hi, 0xFFFFFFFF + + s_getreg_b32 s_save_alloc_size, hwreg(HW_REG_LDS_ALLOC,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE) //lds_size + s_and_b32 s_save_alloc_size, s_save_alloc_size, 0xFFFFFFFF //lds_size is zero? + s_cbranch_scc0 L_SAVE_LDS_DONE //no lds used? jump to L_SAVE_DONE + + s_barrier //LDS is used? wait for other waves in the same TG + s_and_b32 s_save_tmp, s_save_exec_hi, S_SAVE_SPI_INIT_FIRST_WAVE_MASK //exec is still used here + s_cbranch_scc0 L_SAVE_LDS_DONE + + // first wave do LDS save; + + s_lshl_b32 s_save_alloc_size, s_save_alloc_size, 6 //LDS size in dwords = lds_size * 64dw + s_lshl_b32 s_save_alloc_size, s_save_alloc_size, 2 //LDS size in bytes + s_mov_b32 s_save_buf_rsrc2, s_save_alloc_size //NUM_RECORDS in bytes + + // LDS at offset: size(VGPR)+SIZE(SGPR)+SIZE(HWREG) + // + get_vgpr_size_bytes(s_save_mem_offset) + get_sgpr_size_bytes(s_save_tmp) + s_add_u32 s_save_mem_offset, s_save_mem_offset, s_save_tmp + s_add_u32 s_save_mem_offset, s_save_mem_offset, get_hwreg_size_bytes() + + + if (SWIZZLE_EN) + s_add_u32 s_save_buf_rsrc2, s_save_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + s_mov_b32 m0, 0x0 //lds_offset initial value = 0 + + +var LDS_DMA_ENABLE = 0 +var UNROLL = 0 +if UNROLL==0 && LDS_DMA_ENABLE==1 + s_mov_b32 s3, 256*2 + s_nop 0 + s_nop 0 + s_nop 0 + L_SAVE_LDS_LOOP: + //TODO: looks the 2 buffer_store/load clause for s/r will hurt performance.??? + if (SAVE_LDS) //SPI always alloc LDS space in 128DW granularity + buffer_store_lds_dword s_save_buf_rsrc0, s_save_mem_offset lds:1 // first 64DW + buffer_store_lds_dword s_save_buf_rsrc0, s_save_mem_offset lds:1 offset:256 // second 64DW + end + + s_add_u32 m0, m0, s3 //every buffer_store_lds does 256 bytes + s_add_u32 s_save_mem_offset, s_save_mem_offset, s3 //mem offset increased by 256 bytes + s_cmp_lt_u32 m0, s_save_alloc_size //scc=(m0 < s_save_alloc_size) ? 1 : 0 + s_cbranch_scc1 L_SAVE_LDS_LOOP //LDS save is complete? + +elsif LDS_DMA_ENABLE==1 && UNROLL==1 // UNROOL , has ichace miss + // store from higest LDS address to lowest + s_mov_b32 s3, 256*2 + s_sub_u32 m0, s_save_alloc_size, s3 + s_add_u32 s_save_mem_offset, s_save_mem_offset, m0 + s_lshr_b32 s_save_alloc_size, s_save_alloc_size, 9 // how many 128 trunks... + s_sub_u32 s_save_alloc_size, 128, s_save_alloc_size // store from higheset addr to lowest + s_mul_i32 s_save_alloc_size, s_save_alloc_size, 6*4 // PC offset increment, each LDS save block cost 6*4 Bytes instruction + s_add_u32 s_save_alloc_size, s_save_alloc_size, 3*4 //2is the below 2 inst...//s_addc and s_setpc + s_nop 0 + s_nop 0 + s_nop 0 //pad 3 dw to let LDS_DMA align with 64Bytes + s_getpc_b64 s[0:1] // reuse s[0:1], since s[0:1] already saved + s_add_u32 s0, s0,s_save_alloc_size + s_addc_u32 s1, s1, 0 + s_setpc_b64 s[0:1] + + + for var i =0; i< 128; i++ + // be careful to make here a 64Byte aligned address, which could improve performance... + buffer_store_lds_dword s_save_buf_rsrc0, s_save_mem_offset lds:1 offset:0 // first 64DW + buffer_store_lds_dword s_save_buf_rsrc0, s_save_mem_offset lds:1 offset:256 // second 64DW + + if i!=127 + s_sub_u32 m0, m0, s3 // use a sgpr to shrink 2DW-inst to 1DW inst to improve performance , i.e. pack more LDS_DMA inst to one Cacheline + s_sub_u32 s_save_mem_offset, s_save_mem_offset, s3 + end + end + +else // BUFFER_STORE + v_mbcnt_lo_u32_b32 v2, 0xffffffff, 0x0 + v_mbcnt_hi_u32_b32 v3, 0xffffffff, v2 // tid + v_mul_i32_i24 v2, v3, 8 // tid*8 + v_mov_b32 v3, 256*2 + s_mov_b32 m0, 0x10000 + s_mov_b32 s0, s_save_buf_rsrc3 + s_and_b32 s_save_buf_rsrc3, s_save_buf_rsrc3, 0xFF7FFFFF // disable add_tid + s_or_b32 s_save_buf_rsrc3, s_save_buf_rsrc3, 0x58000 //DFMT + +L_SAVE_LDS_LOOP_VECTOR: + ds_read_b64 v[0:1], v2 //x =LDS[a], byte address + s_waitcnt lgkmcnt(0) + buffer_store_dwordx2 v[0:1], v2, s_save_buf_rsrc0, s_save_mem_offset offen:1 glc:1 slc:1 +// s_waitcnt vmcnt(0) +// v_add_u32 v2, vcc[0:1], v2, v3 + v_add_u32 v2, v2, v3 + v_cmp_lt_u32 vcc[0:1], v2, s_save_alloc_size + s_cbranch_vccnz L_SAVE_LDS_LOOP_VECTOR + + // restore rsrc3 + s_mov_b32 s_save_buf_rsrc3, s0 + +end + +L_SAVE_LDS_DONE: + + + /* save VGPRs - set the Rest VGPRs */ + ////////////////////////////////////////////////////////////////////////////////////// + L_SAVE_VGPR: + // VGPR SR memory offset: 0 + // TODO rearrange the RSRC words to use swizzle for VGPR save... + + s_mov_b32 s_save_mem_offset, (0+256*4) // for the rest VGPRs + s_mov_b32 exec_lo, 0xFFFFFFFF //need every thread from now on + s_mov_b32 exec_hi, 0xFFFFFFFF + + s_getreg_b32 s_save_alloc_size, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE) //vpgr_size + s_add_u32 s_save_alloc_size, s_save_alloc_size, 1 + s_lshl_b32 s_save_alloc_size, s_save_alloc_size, 2 //Number of VGPRs = (vgpr_size + 1) * 4 (non-zero value) //FIXME for GFX, zero is possible + s_lshl_b32 s_save_buf_rsrc2, s_save_alloc_size, 8 //NUM_RECORDS in bytes (64 threads*4) + if (SWIZZLE_EN) + s_add_u32 s_save_buf_rsrc2, s_save_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + + // VGPR Allocated in 4-GPR granularity + +if G8SR_VGPR_SR_IN_DWX4 + // the const stride for DWx4 is 4*4 bytes + s_and_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, G8SR_SAVE_BUF_RSRC_WORD1_STRIDE_DWx4 // const stride to 4*4 bytes + + s_mov_b32 m0, 4 // skip first 4 VGPRs + s_cmp_lt_u32 m0, s_save_alloc_size + s_cbranch_scc0 L_SAVE_VGPR_LOOP_END // no more vgprs + + s_set_gpr_idx_on m0, 0x1 // This will change M0 + s_add_u32 s_save_alloc_size, s_save_alloc_size, 0x1000 // because above inst change m0 +L_SAVE_VGPR_LOOP: + v_mov_b32 v0, v0 // v0 = v[0+m0] + v_mov_b32 v1, v1 + v_mov_b32 v2, v2 + v_mov_b32 v3, v3 + + + buffer_store_dwordx4 v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 + s_add_u32 m0, m0, 4 + s_add_u32 s_save_mem_offset, s_save_mem_offset, 256*4 + s_cmp_lt_u32 m0, s_save_alloc_size + s_cbranch_scc1 L_SAVE_VGPR_LOOP //VGPR save is complete? + s_set_gpr_idx_off +L_SAVE_VGPR_LOOP_END: + + s_and_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_save_buf_rsrc1, s_save_buf_rsrc1, S_SAVE_BUF_RSRC_WORD1_STRIDE // reset const stride to 4 bytes +else + // VGPR store using dw burst + s_mov_b32 m0, 0x4 //VGPR initial index value =0 + s_cmp_lt_u32 m0, s_save_alloc_size + s_cbranch_scc0 L_SAVE_VGPR_END + + + s_set_gpr_idx_on m0, 0x1 //M0[7:0] = M0[7:0] and M0[15:12] = 0x1 + s_add_u32 s_save_alloc_size, s_save_alloc_size, 0x1000 //add 0x1000 since we compare m0 against it later + + L_SAVE_VGPR_LOOP: + v_mov_b32 v0, v0 //v0 = v[0+m0] + v_mov_b32 v1, v1 //v0 = v[0+m0] + v_mov_b32 v2, v2 //v0 = v[0+m0] + v_mov_b32 v3, v3 //v0 = v[0+m0] + + if(USE_MTBUF_INSTEAD_OF_MUBUF) + tbuffer_store_format_x v0, v0, s_save_buf_rsrc0, s_save_mem_offset format:BUF_NUM_FORMAT_FLOAT format: BUF_DATA_FORMAT_32 slc:1 glc:1 + else + buffer_store_dword v0, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 + buffer_store_dword v1, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256 + buffer_store_dword v2, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*2 + buffer_store_dword v3, v0, s_save_buf_rsrc0, s_save_mem_offset slc:1 glc:1 offset:256*3 + end + + s_add_u32 m0, m0, 4 //next vgpr index + s_add_u32 s_save_mem_offset, s_save_mem_offset, 256*4 //every buffer_store_dword does 256 bytes + s_cmp_lt_u32 m0, s_save_alloc_size //scc = (m0 < s_save_alloc_size) ? 1 : 0 + s_cbranch_scc1 L_SAVE_VGPR_LOOP //VGPR save is complete? + s_set_gpr_idx_off +end + +L_SAVE_VGPR_END: + + + + + + + /* S_PGM_END_SAVED */ //FIXME graphics ONLY + if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_SAVE_NORMAL_EXIT)) + s_and_b32 s_save_pc_hi, s_save_pc_hi, 0x0000ffff //pc[47:32] + s_add_u32 s_save_pc_lo, s_save_pc_lo, 4 //pc[31:0]+4 + s_addc_u32 s_save_pc_hi, s_save_pc_hi, 0x0 //carry bit over + s_rfe_b64 s_save_pc_lo //Return to the main shader program + else + end + +// Save Done timestamp +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_save_d + // SGPR SR memory offset : size(VGPR) + get_vgpr_size_bytes(s_save_mem_offset) + s_add_u32 s_save_mem_offset, s_save_mem_offset, G8SR_DEBUG_TS_SAVE_D_OFFSET + s_waitcnt lgkmcnt(0) //FIXME, will cause xnack?? + // Need reset rsrc2?? + s_mov_b32 m0, s_save_mem_offset + s_mov_b32 s_save_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + s_buffer_store_dwordx2 s_g8sr_ts_save_d, s_save_buf_rsrc0, m0 glc:1 +end + + + s_branch L_END_PGM + + + +/**************************************************************************/ +/* restore routine */ +/**************************************************************************/ + +L_RESTORE: + /* Setup Resource Contants */ + if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_RESTORE_NORMAL)) + //calculate wd_addr using absolute thread id + v_readlane_b32 s_restore_tmp, v9, 0 + s_lshr_b32 s_restore_tmp, s_restore_tmp, 6 + s_mul_i32 s_restore_tmp, s_restore_tmp, WAVE_SPACE + s_add_i32 s_restore_spi_init_lo, s_restore_tmp, WG_BASE_ADDR_LO + s_mov_b32 s_restore_spi_init_hi, WG_BASE_ADDR_HI + s_and_b32 s_restore_spi_init_hi, s_restore_spi_init_hi, CTX_RESTORE_CONTROL + else + end + +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_restore_s + s_waitcnt lgkmcnt(0) //FIXME, will cause xnack?? + // tma_lo/hi are sgpr 110, 111, which will not used for 112 SGPR allocated case... + s_mov_b32 s_restore_pc_lo, s_g8sr_ts_restore_s[0] + s_mov_b32 s_restore_pc_hi, s_g8sr_ts_restore_s[1] //backup ts to ttmp0/1, sicne exec will be finally restored.. +end + + + + s_mov_b32 s_restore_buf_rsrc0, s_restore_spi_init_lo //base_addr_lo + s_and_b32 s_restore_buf_rsrc1, s_restore_spi_init_hi, 0x0000FFFF //base_addr_hi + s_or_b32 s_restore_buf_rsrc1, s_restore_buf_rsrc1, S_RESTORE_BUF_RSRC_WORD1_STRIDE + s_mov_b32 s_restore_buf_rsrc2, 0 //NUM_RECORDS initial value = 0 (in bytes) + s_mov_b32 s_restore_buf_rsrc3, S_RESTORE_BUF_RSRC_WORD3_MISC + s_and_b32 s_restore_tmp, s_restore_spi_init_hi, S_RESTORE_SPI_INIT_ATC_MASK + s_lshr_b32 s_restore_tmp, s_restore_tmp, (S_RESTORE_SPI_INIT_ATC_SHIFT-SQ_BUF_RSRC_WORD1_ATC_SHIFT) //get ATC bit into position + s_or_b32 s_restore_buf_rsrc3, s_restore_buf_rsrc3, s_restore_tmp //or ATC + s_and_b32 s_restore_tmp, s_restore_spi_init_hi, S_RESTORE_SPI_INIT_MTYPE_MASK + s_lshr_b32 s_restore_tmp, s_restore_tmp, (S_RESTORE_SPI_INIT_MTYPE_SHIFT-SQ_BUF_RSRC_WORD3_MTYPE_SHIFT) //get MTYPE bits into position + s_or_b32 s_restore_buf_rsrc3, s_restore_buf_rsrc3, s_restore_tmp //or MTYPE + + /* global mem offset */ +// s_mov_b32 s_restore_mem_offset, 0x0 //mem offset initial value = 0 + + /* the first wave in the threadgroup */ + s_and_b32 s_restore_tmp, s_restore_spi_init_hi, S_RESTORE_SPI_INIT_FIRST_WAVE_MASK + s_cbranch_scc0 L_RESTORE_VGPR + + /* restore LDS */ + ////////////////////////////// + L_RESTORE_LDS: + + s_mov_b32 exec_lo, 0xFFFFFFFF //need every thread from now on //be consistent with SAVE although can be moved ahead + s_mov_b32 exec_hi, 0xFFFFFFFF + + s_getreg_b32 s_restore_alloc_size, hwreg(HW_REG_LDS_ALLOC,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT,SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE) //lds_size + s_and_b32 s_restore_alloc_size, s_restore_alloc_size, 0xFFFFFFFF //lds_size is zero? + s_cbranch_scc0 L_RESTORE_VGPR //no lds used? jump to L_RESTORE_VGPR + s_lshl_b32 s_restore_alloc_size, s_restore_alloc_size, 6 //LDS size in dwords = lds_size * 64dw + s_lshl_b32 s_restore_alloc_size, s_restore_alloc_size, 2 //LDS size in bytes + s_mov_b32 s_restore_buf_rsrc2, s_restore_alloc_size //NUM_RECORDS in bytes + + // LDS at offset: size(VGPR)+SIZE(SGPR)+SIZE(HWREG) + // + get_vgpr_size_bytes(s_restore_mem_offset) + get_sgpr_size_bytes(s_restore_tmp) + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, s_restore_tmp + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, get_hwreg_size_bytes() //FIXME, Check if offset overflow??? + + + if (SWIZZLE_EN) + s_add_u32 s_restore_buf_rsrc2, s_restore_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_restore_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + s_mov_b32 m0, 0x0 //lds_offset initial value = 0 + + L_RESTORE_LDS_LOOP: + if (SAVE_LDS) + buffer_load_dword v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1 // first 64DW + buffer_load_dword v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset lds:1 offset:256 // second 64DW + end + s_add_u32 m0, m0, 256*2 // 128 DW + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, 256*2 //mem offset increased by 128DW + s_cmp_lt_u32 m0, s_restore_alloc_size //scc=(m0 < s_restore_alloc_size) ? 1 : 0 + s_cbranch_scc1 L_RESTORE_LDS_LOOP //LDS restore is complete? + + + /* restore VGPRs */ + ////////////////////////////// + L_RESTORE_VGPR: + // VGPR SR memory offset : 0 + s_mov_b32 s_restore_mem_offset, 0x0 + s_mov_b32 exec_lo, 0xFFFFFFFF //need every thread from now on //be consistent with SAVE although can be moved ahead + s_mov_b32 exec_hi, 0xFFFFFFFF + + s_getreg_b32 s_restore_alloc_size, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE) //vpgr_size + s_add_u32 s_restore_alloc_size, s_restore_alloc_size, 1 + s_lshl_b32 s_restore_alloc_size, s_restore_alloc_size, 2 //Number of VGPRs = (vgpr_size + 1) * 4 (non-zero value) + s_lshl_b32 s_restore_buf_rsrc2, s_restore_alloc_size, 8 //NUM_RECORDS in bytes (64 threads*4) + if (SWIZZLE_EN) + s_add_u32 s_restore_buf_rsrc2, s_restore_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_restore_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + +if G8SR_VGPR_SR_IN_DWX4 + get_vgpr_size_bytes(s_restore_mem_offset) + s_sub_u32 s_restore_mem_offset, s_restore_mem_offset, 256*4 + + // the const stride for DWx4 is 4*4 bytes + s_and_b32 s_restore_buf_rsrc1, s_restore_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_restore_buf_rsrc1, s_restore_buf_rsrc1, G8SR_RESTORE_BUF_RSRC_WORD1_STRIDE_DWx4 // const stride to 4*4 bytes + + s_mov_b32 m0, s_restore_alloc_size + s_set_gpr_idx_on m0, 0x8 // Note.. This will change m0 + +L_RESTORE_VGPR_LOOP: + buffer_load_dwordx4 v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 + s_waitcnt vmcnt(0) + s_sub_u32 m0, m0, 4 + v_mov_b32 v0, v0 // v[0+m0] = v0 + v_mov_b32 v1, v1 + v_mov_b32 v2, v2 + v_mov_b32 v3, v3 + s_sub_u32 s_restore_mem_offset, s_restore_mem_offset, 256*4 + s_cmp_eq_u32 m0, 0x8000 + s_cbranch_scc0 L_RESTORE_VGPR_LOOP + s_set_gpr_idx_off + + s_and_b32 s_restore_buf_rsrc1, s_restore_buf_rsrc1, 0x0000FFFF // reset const stride to 0 + s_or_b32 s_restore_buf_rsrc1, s_restore_buf_rsrc1, S_RESTORE_BUF_RSRC_WORD1_STRIDE // const stride to 4*4 bytes + +else + // VGPR load using dw burst + s_mov_b32 s_restore_mem_offset_save, s_restore_mem_offset // restore start with v1, v0 will be the last + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, 256*4 + s_mov_b32 m0, 4 //VGPR initial index value = 1 + s_set_gpr_idx_on m0, 0x8 //M0[7:0] = M0[7:0] and M0[15:12] = 0x8 + s_add_u32 s_restore_alloc_size, s_restore_alloc_size, 0x8000 //add 0x8000 since we compare m0 against it later + + L_RESTORE_VGPR_LOOP: + if(USE_MTBUF_INSTEAD_OF_MUBUF) + tbuffer_load_format_x v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset format:BUF_NUM_FORMAT_FLOAT format: BUF_DATA_FORMAT_32 slc:1 glc:1 + else + buffer_load_dword v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 + buffer_load_dword v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:256 + buffer_load_dword v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:256*2 + buffer_load_dword v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset slc:1 glc:1 offset:256*3 + end + s_waitcnt vmcnt(0) //ensure data ready + v_mov_b32 v0, v0 //v[0+m0] = v0 + v_mov_b32 v1, v1 + v_mov_b32 v2, v2 + v_mov_b32 v3, v3 + s_add_u32 m0, m0, 4 //next vgpr index + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, 256*4 //every buffer_load_dword does 256 bytes + s_cmp_lt_u32 m0, s_restore_alloc_size //scc = (m0 < s_restore_alloc_size) ? 1 : 0 + s_cbranch_scc1 L_RESTORE_VGPR_LOOP //VGPR restore (except v0) is complete? + s_set_gpr_idx_off + /* VGPR restore on v0 */ + if(USE_MTBUF_INSTEAD_OF_MUBUF) + tbuffer_load_format_x v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save format:BUF_NUM_FORMAT_FLOAT format: BUF_DATA_FORMAT_32 slc:1 glc:1 + else + buffer_load_dword v0, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 + buffer_load_dword v1, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:256 + buffer_load_dword v2, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:256*2 + buffer_load_dword v3, v0, s_restore_buf_rsrc0, s_restore_mem_offset_save slc:1 glc:1 offset:256*3 + end + +end + + /* restore SGPRs */ + ////////////////////////////// + + // SGPR SR memory offset : size(VGPR) + get_vgpr_size_bytes(s_restore_mem_offset) + get_sgpr_size_bytes(s_restore_tmp) + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, s_restore_tmp + s_sub_u32 s_restore_mem_offset, s_restore_mem_offset, 16*4 // restore SGPR from S[n] to S[0], by 16 sgprs group + // TODO, change RSRC word to rearrange memory layout for SGPRS + + s_getreg_b32 s_restore_alloc_size, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SIZE) //spgr_size + s_add_u32 s_restore_alloc_size, s_restore_alloc_size, 1 + s_lshl_b32 s_restore_alloc_size, s_restore_alloc_size, 4 //Number of SGPRs = (sgpr_size + 1) * 16 (non-zero value) + + if (SGPR_SAVE_USE_SQC) + s_lshl_b32 s_restore_buf_rsrc2, s_restore_alloc_size, 2 //NUM_RECORDS in bytes + else + s_lshl_b32 s_restore_buf_rsrc2, s_restore_alloc_size, 8 //NUM_RECORDS in bytes (64 threads) + end + if (SWIZZLE_EN) + s_add_u32 s_restore_buf_rsrc2, s_restore_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_restore_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + s_mov_b32 m0, s_restore_alloc_size + + L_RESTORE_SGPR_LOOP: + read_16sgpr_from_mem(s0, s_restore_buf_rsrc0, s_restore_mem_offset) //PV: further performance improvement can be made + s_waitcnt lgkmcnt(0) //ensure data ready + + s_sub_u32 m0, m0, 16 // Restore from S[n] to S[0] + s_nop 0 // hazard SALU M0=> S_MOVREL + + s_movreld_b64 s0, s0 //s[0+m0] = s0 + s_movreld_b64 s2, s2 + s_movreld_b64 s4, s4 + s_movreld_b64 s6, s6 + s_movreld_b64 s8, s8 + s_movreld_b64 s10, s10 + s_movreld_b64 s12, s12 + s_movreld_b64 s14, s14 + + s_cmp_eq_u32 m0, 0 //scc = (m0 < s_restore_alloc_size) ? 1 : 0 + s_cbranch_scc0 L_RESTORE_SGPR_LOOP //SGPR restore (except s0) is complete? + + /* restore HW registers */ + ////////////////////////////// + L_RESTORE_HWREG: + + +if G8SR_DEBUG_TIMESTAMP + s_mov_b32 s_g8sr_ts_restore_s[0], s_restore_pc_lo + s_mov_b32 s_g8sr_ts_restore_s[1], s_restore_pc_hi +end + + // HWREG SR memory offset : size(VGPR)+size(SGPR) + get_vgpr_size_bytes(s_restore_mem_offset) + get_sgpr_size_bytes(s_restore_tmp) + s_add_u32 s_restore_mem_offset, s_restore_mem_offset, s_restore_tmp + + + s_mov_b32 s_restore_buf_rsrc2, 0x4 //NUM_RECORDS in bytes + if (SWIZZLE_EN) + s_add_u32 s_restore_buf_rsrc2, s_restore_buf_rsrc2, 0x0 //FIXME need to use swizzle to enable bounds checking? + else + s_mov_b32 s_restore_buf_rsrc2, 0x1000000 //NUM_RECORDS in bytes + end + + read_hwreg_from_mem(s_restore_m0, s_restore_buf_rsrc0, s_restore_mem_offset) //M0 + read_hwreg_from_mem(s_restore_pc_lo, s_restore_buf_rsrc0, s_restore_mem_offset) //PC + read_hwreg_from_mem(s_restore_pc_hi, s_restore_buf_rsrc0, s_restore_mem_offset) + read_hwreg_from_mem(s_restore_exec_lo, s_restore_buf_rsrc0, s_restore_mem_offset) //EXEC + read_hwreg_from_mem(s_restore_exec_hi, s_restore_buf_rsrc0, s_restore_mem_offset) + read_hwreg_from_mem(s_restore_status, s_restore_buf_rsrc0, s_restore_mem_offset) //STATUS + read_hwreg_from_mem(s_restore_trapsts, s_restore_buf_rsrc0, s_restore_mem_offset) //TRAPSTS + read_hwreg_from_mem(xnack_mask_lo, s_restore_buf_rsrc0, s_restore_mem_offset) //XNACK_MASK_LO + read_hwreg_from_mem(xnack_mask_hi, s_restore_buf_rsrc0, s_restore_mem_offset) //XNACK_MASK_HI + read_hwreg_from_mem(s_restore_mode, s_restore_buf_rsrc0, s_restore_mem_offset) //MODE + + s_waitcnt lgkmcnt(0) //from now on, it is safe to restore STATUS and IB_STS + + //for normal save & restore, the saved PC points to the next inst to execute, no adjustment needs to be made, otherwise: + if ((EMU_RUN_HACK) && (!EMU_RUN_HACK_RESTORE_NORMAL)) + s_add_u32 s_restore_pc_lo, s_restore_pc_lo, 8 //pc[31:0]+8 //two back-to-back s_trap are used (first for save and second for restore) + s_addc_u32 s_restore_pc_hi, s_restore_pc_hi, 0x0 //carry bit over + end + if ((EMU_RUN_HACK) && (EMU_RUN_HACK_RESTORE_NORMAL)) + s_add_u32 s_restore_pc_lo, s_restore_pc_lo, 4 //pc[31:0]+4 // save is hack through s_trap but restore is normal + s_addc_u32 s_restore_pc_hi, s_restore_pc_hi, 0x0 //carry bit over + end + + s_mov_b32 m0, s_restore_m0 + s_mov_b32 exec_lo, s_restore_exec_lo + s_mov_b32 exec_hi, s_restore_exec_hi + + s_and_b32 s_restore_m0, SQ_WAVE_TRAPSTS_PRE_SAVECTX_MASK, s_restore_trapsts + s_setreg_b32 hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_PRE_SAVECTX_SHIFT, SQ_WAVE_TRAPSTS_PRE_SAVECTX_SIZE), s_restore_m0 + s_and_b32 s_restore_m0, SQ_WAVE_TRAPSTS_POST_SAVECTX_MASK, s_restore_trapsts + s_lshr_b32 s_restore_m0, s_restore_m0, SQ_WAVE_TRAPSTS_POST_SAVECTX_SHIFT + s_setreg_b32 hwreg(HW_REG_TRAPSTS, SQ_WAVE_TRAPSTS_POST_SAVECTX_SHIFT, SQ_WAVE_TRAPSTS_POST_SAVECTX_SIZE), s_restore_m0 + //s_setreg_b32 hwreg(HW_REG_TRAPSTS), s_restore_trapsts //don't overwrite SAVECTX bit as it may be set through external SAVECTX during restore + s_setreg_b32 hwreg(HW_REG_MODE), s_restore_mode + + // Restore trap temporaries 6-11, 13-15 initialized by SPI debug dispatch logic + // ttmp SR memory offset : size(VGPR)+size(SGPR)+0x40 + get_vgpr_size_bytes(s_restore_ttmps_lo) + get_sgpr_size_bytes(s_restore_ttmps_hi) + s_add_u32 s_restore_ttmps_lo, s_restore_ttmps_lo, s_restore_ttmps_hi + s_add_u32 s_restore_ttmps_lo, s_restore_ttmps_lo, s_restore_buf_rsrc0 + s_addc_u32 s_restore_ttmps_hi, s_restore_buf_rsrc1, 0x0 + s_and_b32 s_restore_ttmps_hi, s_restore_ttmps_hi, 0xFFFF + s_load_dwordx2 [ttmp6, ttmp7], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x40 glc:1 + s_load_dwordx4 [ttmp8, ttmp9, ttmp10, ttmp11], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x48 glc:1 + s_load_dword ttmp13, [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x58 glc:1 + s_load_dwordx2 [ttmp14, ttmp15], [s_restore_ttmps_lo, s_restore_ttmps_hi], 0x5C glc:1 + s_waitcnt lgkmcnt(0) + + //reuse s_restore_m0 as a temp register + s_and_b32 s_restore_m0, s_restore_pc_hi, S_SAVE_PC_HI_RCNT_MASK + s_lshr_b32 s_restore_m0, s_restore_m0, S_SAVE_PC_HI_RCNT_SHIFT + s_lshl_b32 s_restore_m0, s_restore_m0, SQ_WAVE_IB_STS_RCNT_SHIFT + s_mov_b32 s_restore_tmp, 0x0 //IB_STS is zero + s_or_b32 s_restore_tmp, s_restore_tmp, s_restore_m0 + s_and_b32 s_restore_m0, s_restore_pc_hi, S_SAVE_PC_HI_FIRST_REPLAY_MASK + s_lshr_b32 s_restore_m0, s_restore_m0, S_SAVE_PC_HI_FIRST_REPLAY_SHIFT + s_lshl_b32 s_restore_m0, s_restore_m0, SQ_WAVE_IB_STS_FIRST_REPLAY_SHIFT + s_or_b32 s_restore_tmp, s_restore_tmp, s_restore_m0 + s_and_b32 s_restore_m0, s_restore_status, SQ_WAVE_STATUS_INST_ATC_MASK + s_lshr_b32 s_restore_m0, s_restore_m0, SQ_WAVE_STATUS_INST_ATC_SHIFT + s_setreg_b32 hwreg(HW_REG_IB_STS), s_restore_tmp + + s_and_b32 s_restore_pc_hi, s_restore_pc_hi, 0x0000ffff //pc[47:32] //Do it here in order not to affect STATUS + s_and_b64 exec, exec, exec // Restore STATUS.EXECZ, not writable by s_setreg_b32 + s_and_b64 vcc, vcc, vcc // Restore STATUS.VCCZ, not writable by s_setreg_b32 + s_setreg_b32 hwreg(HW_REG_STATUS), s_restore_status // SCC is included, which is changed by previous salu + + s_barrier //barrier to ensure the readiness of LDS before access attempts from any other wave in the same TG //FIXME not performance-optimal at this time + +if G8SR_DEBUG_TIMESTAMP + s_memrealtime s_g8sr_ts_restore_d + s_waitcnt lgkmcnt(0) +end + +// s_rfe_b64 s_restore_pc_lo //Return to the main shader program and resume execution + s_rfe_restore_b64 s_restore_pc_lo, s_restore_m0 // s_restore_m0[0] is used to set STATUS.inst_atc + + +/**************************************************************************/ +/* the END */ +/**************************************************************************/ +L_END_PGM: + s_endpgm + +end + + +/**************************************************************************/ +/* the helper functions */ +/**************************************************************************/ + +//Only for save hwreg to mem +function write_hwreg_to_mem(s, s_rsrc, s_mem_offset) + s_mov_b32 exec_lo, m0 //assuming exec_lo is not needed anymore from this point on + s_mov_b32 m0, s_mem_offset + s_buffer_store_dword s, s_rsrc, m0 glc:1 + ack_sqc_store_workaround() + s_add_u32 s_mem_offset, s_mem_offset, 4 + s_mov_b32 m0, exec_lo +end + + +// HWREG are saved before SGPRs, so all HWREG could be use. +function write_16sgpr_to_mem(s, s_rsrc, s_mem_offset) + + s_buffer_store_dwordx4 s[0], s_rsrc, 0 glc:1 + ack_sqc_store_workaround() + s_buffer_store_dwordx4 s[4], s_rsrc, 16 glc:1 + ack_sqc_store_workaround() + s_buffer_store_dwordx4 s[8], s_rsrc, 32 glc:1 + ack_sqc_store_workaround() + s_buffer_store_dwordx4 s[12], s_rsrc, 48 glc:1 + ack_sqc_store_workaround() + s_add_u32 s_rsrc[0], s_rsrc[0], 4*16 + s_addc_u32 s_rsrc[1], s_rsrc[1], 0x0 // +scc +end + + +function read_hwreg_from_mem(s, s_rsrc, s_mem_offset) + s_buffer_load_dword s, s_rsrc, s_mem_offset glc:1 + s_add_u32 s_mem_offset, s_mem_offset, 4 +end + +function read_16sgpr_from_mem(s, s_rsrc, s_mem_offset) + s_buffer_load_dwordx16 s, s_rsrc, s_mem_offset glc:1 + s_sub_u32 s_mem_offset, s_mem_offset, 4*16 +end + + + +function get_lds_size_bytes(s_lds_size_byte) + // SQ LDS granularity is 64DW, while PGM_RSRC2.lds_size is in granularity 128DW + s_getreg_b32 s_lds_size_byte, hwreg(HW_REG_LDS_ALLOC, SQ_WAVE_LDS_ALLOC_LDS_SIZE_SHIFT, SQ_WAVE_LDS_ALLOC_LDS_SIZE_SIZE) // lds_size + s_lshl_b32 s_lds_size_byte, s_lds_size_byte, 8 //LDS size in dwords = lds_size * 64 *4Bytes // granularity 64DW +end + +function get_vgpr_size_bytes(s_vgpr_size_byte) + s_getreg_b32 s_vgpr_size_byte, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_VGPR_SIZE_SIZE) //vpgr_size + s_add_u32 s_vgpr_size_byte, s_vgpr_size_byte, 1 + s_lshl_b32 s_vgpr_size_byte, s_vgpr_size_byte, (2+8) //Number of VGPRs = (vgpr_size + 1) * 4 * 64 * 4 (non-zero value) //FIXME for GFX, zero is possible +end + +function get_sgpr_size_bytes(s_sgpr_size_byte) + s_getreg_b32 s_sgpr_size_byte, hwreg(HW_REG_GPR_ALLOC,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SHIFT,SQ_WAVE_GPR_ALLOC_SGPR_SIZE_SIZE) //spgr_size + s_add_u32 s_sgpr_size_byte, s_sgpr_size_byte, 1 + s_lshl_b32 s_sgpr_size_byte, s_sgpr_size_byte, 6 //Number of SGPRs = (sgpr_size + 1) * 16 *4 (non-zero value) +end + +function get_hwreg_size_bytes + return 128 //HWREG size 128 bytes +end + +function ack_sqc_store_workaround + if ACK_SQC_STORE + s_waitcnt lgkmcnt(0) + end +end diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c index 59808a39ecf4a..f64c5551cdba0 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_chardev.c @@ -233,7 +233,7 @@ static int set_queue_properties_from_user(struct queue_properties *q_properties, pr_debug("Queue Size: 0x%llX, %u\n", q_properties->queue_size, args->ring_size); - pr_debug("Queue r/w Pointers: %p, %p\n", + pr_debug("Queue r/w Pointers: %px, %px\n", q_properties->read_ptr, q_properties->write_ptr); @@ -292,8 +292,16 @@ static int kfd_ioctl_create_queue(struct file *filep, struct kfd_process *p, /* Return gpu_id as doorbell offset for mmap usage */ - args->doorbell_offset = (KFD_MMAP_DOORBELL_MASK | args->gpu_id); + args->doorbell_offset = KFD_MMAP_TYPE_DOORBELL; + args->doorbell_offset |= KFD_MMAP_GPU_ID(args->gpu_id); args->doorbell_offset <<= PAGE_SHIFT; + if (KFD_IS_SOC15(dev->device_info->asic_family)) + /* On SOC15 ASICs, doorbell allocation must be + * per-device, and independent from the per-process + * queue_id. Return the doorbell offset within the + * doorbell aperture to user mode. + */ + args->doorbell_offset |= q_properties.doorbell_off; mutex_unlock(&p->mutex); @@ -1296,8 +1304,8 @@ static int kfd_ioctl_map_memory_to_gpu(struct file *filep, return -EINVAL; } - devices_arr = kmalloc(args->n_devices * sizeof(*devices_arr), - GFP_KERNEL); + devices_arr = kmalloc_array(args->n_devices, sizeof(*devices_arr), + GFP_KERNEL); if (!devices_arr) return -ENOMEM; @@ -1405,8 +1413,8 @@ static int kfd_ioctl_unmap_memory_from_gpu(struct file *filep, return -EINVAL; } - devices_arr = kmalloc(args->n_devices * sizeof(*devices_arr), - GFP_KERNEL); + devices_arr = kmalloc_array(args->n_devices, sizeof(*devices_arr), + GFP_KERNEL); if (!devices_arr) return -ENOMEM; @@ -1645,23 +1653,33 @@ err_i1: static int kfd_mmap(struct file *filp, struct vm_area_struct *vma) { struct kfd_process *process; + struct kfd_dev *dev = NULL; + unsigned long vm_pgoff; + unsigned int gpu_id; process = kfd_get_process(current); if (IS_ERR(process)) return PTR_ERR(process); - if ((vma->vm_pgoff & KFD_MMAP_DOORBELL_MASK) == - KFD_MMAP_DOORBELL_MASK) { - vma->vm_pgoff = vma->vm_pgoff ^ KFD_MMAP_DOORBELL_MASK; - return kfd_doorbell_mmap(process, vma); - } else if ((vma->vm_pgoff & KFD_MMAP_EVENTS_MASK) == - KFD_MMAP_EVENTS_MASK) { - vma->vm_pgoff = vma->vm_pgoff ^ KFD_MMAP_EVENTS_MASK; + vm_pgoff = vma->vm_pgoff; + vma->vm_pgoff = KFD_MMAP_OFFSET_VALUE_GET(vm_pgoff); + gpu_id = KFD_MMAP_GPU_ID_GET(vm_pgoff); + if (gpu_id) + dev = kfd_device_by_id(gpu_id); + + switch (vm_pgoff & KFD_MMAP_TYPE_MASK) { + case KFD_MMAP_TYPE_DOORBELL: + if (!dev) + return -ENODEV; + return kfd_doorbell_mmap(dev, process, vma); + + case KFD_MMAP_TYPE_EVENTS: return kfd_event_mmap(process, vma); - } else if ((vma->vm_pgoff & KFD_MMAP_RESERVED_MEM_MASK) == - KFD_MMAP_RESERVED_MEM_MASK) { - vma->vm_pgoff = vma->vm_pgoff ^ KFD_MMAP_RESERVED_MEM_MASK; - return kfd_reserved_mem_mmap(process, vma); + + case KFD_MMAP_TYPE_RESERVED_MEM: + if (!dev) + return -ENODEV; + return kfd_reserved_mem_mmap(dev, process, vma); } return -EFAULT; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c index 4f126ef6139b7..296b3f230280b 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_crat.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_crat.c @@ -132,6 +132,9 @@ static struct kfd_gpu_cache_info carrizo_cache_info[] = { #define fiji_cache_info carrizo_cache_info #define polaris10_cache_info carrizo_cache_info #define polaris11_cache_info carrizo_cache_info +/* TODO - check & update Vega10 cache details */ +#define vega10_cache_info carrizo_cache_info +#define raven_cache_info carrizo_cache_info static void kfd_populated_cu_info_cpu(struct kfd_topology_device *dev, struct crat_subtype_computeunit *cu) @@ -603,6 +606,14 @@ static int kfd_fill_gpu_cache_info(struct kfd_dev *kdev, pcache_info = polaris11_cache_info; num_of_cache_types = ARRAY_SIZE(polaris11_cache_info); break; + case CHIP_VEGA10: + pcache_info = vega10_cache_info; + num_of_cache_types = ARRAY_SIZE(vega10_cache_info); + break; + case CHIP_RAVEN: + pcache_info = raven_cache_info; + num_of_cache_types = ARRAY_SIZE(raven_cache_info); + break; default: return -EINVAL; } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device.c b/drivers/gpu/drm/amd/amdkfd/kfd_device.c index 3346699960ddb..7ee6cec2c0602 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device.c @@ -20,16 +20,13 @@ * OTHER DEALINGS IN THE SOFTWARE. */ -#if defined(CONFIG_AMD_IOMMU_V2_MODULE) || defined(CONFIG_AMD_IOMMU_V2) -#include <linux/amd-iommu.h> -#endif #include <linux/bsearch.h> #include <linux/pci.h> #include <linux/slab.h> #include "kfd_priv.h" #include "kfd_device_queue_manager.h" #include "kfd_pm4_headers_vi.h" -#include "cwsr_trap_handler_gfx8.asm" +#include "cwsr_trap_handler.h" #include "kfd_iommu.h" #define MQD_SIZE_ALIGNED 768 @@ -41,6 +38,7 @@ static const struct kfd_device_info kaveri_device_info = { .max_pasid_bits = 16, /* max num of queues for KV.TODO should be a dynamic value */ .max_no_of_hqd = 24, + .doorbell_size = 4, .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, @@ -55,6 +53,7 @@ static const struct kfd_device_info carrizo_device_info = { .max_pasid_bits = 16, /* max num of queues for CZ.TODO should be a dynamic value */ .max_no_of_hqd = 24, + .doorbell_size = 4, .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, @@ -70,6 +69,7 @@ static const struct kfd_device_info hawaii_device_info = { .max_pasid_bits = 16, /* max num of queues for KV.TODO should be a dynamic value */ .max_no_of_hqd = 24, + .doorbell_size = 4, .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, @@ -83,6 +83,7 @@ static const struct kfd_device_info tonga_device_info = { .asic_family = CHIP_TONGA, .max_pasid_bits = 16, .max_no_of_hqd = 24, + .doorbell_size = 4, .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, @@ -96,6 +97,7 @@ static const struct kfd_device_info tonga_vf_device_info = { .asic_family = CHIP_TONGA, .max_pasid_bits = 16, .max_no_of_hqd = 24, + .doorbell_size = 4, .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, @@ -109,6 +111,7 @@ static const struct kfd_device_info fiji_device_info = { .asic_family = CHIP_FIJI, .max_pasid_bits = 16, .max_no_of_hqd = 24, + .doorbell_size = 4, .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, @@ -122,6 +125,7 @@ static const struct kfd_device_info fiji_vf_device_info = { .asic_family = CHIP_FIJI, .max_pasid_bits = 16, .max_no_of_hqd = 24, + .doorbell_size = 4, .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, @@ -136,6 +140,7 @@ static const struct kfd_device_info polaris10_device_info = { .asic_family = CHIP_POLARIS10, .max_pasid_bits = 16, .max_no_of_hqd = 24, + .doorbell_size = 4, .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, @@ -149,6 +154,7 @@ static const struct kfd_device_info polaris10_vf_device_info = { .asic_family = CHIP_POLARIS10, .max_pasid_bits = 16, .max_no_of_hqd = 24, + .doorbell_size = 4, .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, @@ -162,6 +168,7 @@ static const struct kfd_device_info polaris11_device_info = { .asic_family = CHIP_POLARIS11, .max_pasid_bits = 16, .max_no_of_hqd = 24, + .doorbell_size = 4, .ih_ring_entry_size = 4 * sizeof(uint32_t), .event_interrupt_class = &event_interrupt_class_cik, .num_of_watch_points = 4, @@ -171,6 +178,34 @@ static const struct kfd_device_info polaris11_device_info = { .needs_pci_atomics = true, }; +static const struct kfd_device_info vega10_device_info = { + .asic_family = CHIP_VEGA10, + .max_pasid_bits = 16, + .max_no_of_hqd = 24, + .doorbell_size = 8, + .ih_ring_entry_size = 8 * sizeof(uint32_t), + .event_interrupt_class = &event_interrupt_class_v9, + .num_of_watch_points = 4, + .mqd_size_aligned = MQD_SIZE_ALIGNED, + .supports_cwsr = true, + .needs_iommu_device = false, + .needs_pci_atomics = false, +}; + +static const struct kfd_device_info vega10_vf_device_info = { + .asic_family = CHIP_VEGA10, + .max_pasid_bits = 16, + .max_no_of_hqd = 24, + .doorbell_size = 8, + .ih_ring_entry_size = 8 * sizeof(uint32_t), + .event_interrupt_class = &event_interrupt_class_v9, + .num_of_watch_points = 4, + .mqd_size_aligned = MQD_SIZE_ALIGNED, + .supports_cwsr = true, + .needs_iommu_device = false, + .needs_pci_atomics = false, +}; + struct kfd_deviceid { unsigned short did; @@ -250,6 +285,15 @@ static const struct kfd_deviceid supported_devices[] = { { 0x67EB, &polaris11_device_info }, /* Polaris11 */ { 0x67EF, &polaris11_device_info }, /* Polaris11 */ { 0x67FF, &polaris11_device_info }, /* Polaris11 */ + { 0x6860, &vega10_device_info }, /* Vega10 */ + { 0x6861, &vega10_device_info }, /* Vega10 */ + { 0x6862, &vega10_device_info }, /* Vega10 */ + { 0x6863, &vega10_device_info }, /* Vega10 */ + { 0x6864, &vega10_device_info }, /* Vega10 */ + { 0x6867, &vega10_device_info }, /* Vega10 */ + { 0x6868, &vega10_device_info }, /* Vega10 */ + { 0x686C, &vega10_vf_device_info }, /* Vega10 vf*/ + { 0x687F, &vega10_device_info }, /* Vega10 */ }; static int kfd_gtt_sa_init(struct kfd_dev *kfd, unsigned int buf_size, @@ -279,7 +323,7 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, struct pci_dev *pdev, const struct kfd2kgd_calls *f2g) { struct kfd_dev *kfd; - + int ret; const struct kfd_device_info *device_info = lookup_device_info(pdev->device); @@ -288,19 +332,18 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, return NULL; } - if (device_info->needs_pci_atomics) { - /* Allow BIF to recode atomics to PCIe 3.0 - * AtomicOps. 32 and 64-bit requests are possible and - * must be supported. - */ - if (pci_enable_atomic_ops_to_root(pdev, - PCI_EXP_DEVCAP2_ATOMIC_COMP32 | - PCI_EXP_DEVCAP2_ATOMIC_COMP64) < 0) { - dev_info(kfd_device, - "skipped device %x:%x, PCI rejects atomics", - pdev->vendor, pdev->device); - return NULL; - } + /* Allow BIF to recode atomics to PCIe 3.0 AtomicOps. + * 32 and 64-bit requests are possible and must be + * supported. + */ + ret = pci_enable_atomic_ops_to_root(pdev, + PCI_EXP_DEVCAP2_ATOMIC_COMP32 | + PCI_EXP_DEVCAP2_ATOMIC_COMP64); + if (device_info->needs_pci_atomics && ret < 0) { + dev_info(kfd_device, + "skipped device %x:%x, PCI rejects atomics\n", + pdev->vendor, pdev->device); + return NULL; } kfd = kzalloc(sizeof(*kfd), GFP_KERNEL); @@ -323,10 +366,16 @@ struct kfd_dev *kgd2kfd_probe(struct kgd_dev *kgd, static void kfd_cwsr_init(struct kfd_dev *kfd) { if (cwsr_enable && kfd->device_info->supports_cwsr) { - BUILD_BUG_ON(sizeof(cwsr_trap_gfx8_hex) > PAGE_SIZE); + if (kfd->device_info->asic_family < CHIP_VEGA10) { + BUILD_BUG_ON(sizeof(cwsr_trap_gfx8_hex) > PAGE_SIZE); + kfd->cwsr_isa = cwsr_trap_gfx8_hex; + kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx8_hex); + } else { + BUILD_BUG_ON(sizeof(cwsr_trap_gfx9_hex) > PAGE_SIZE); + kfd->cwsr_isa = cwsr_trap_gfx9_hex; + kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx9_hex); + } - kfd->cwsr_isa = cwsr_trap_gfx8_hex; - kfd->cwsr_isa_size = sizeof(cwsr_trap_gfx8_hex); kfd->cwsr_enabled = true; } } @@ -541,6 +590,44 @@ void kgd2kfd_interrupt(struct kfd_dev *kfd, const void *ih_ring_entry) spin_unlock(&kfd->interrupt_lock); } +int kgd2kfd_quiesce_mm(struct mm_struct *mm) +{ + struct kfd_process *p; + int r; + + /* Because we are called from arbitrary context (workqueue) as opposed + * to process context, kfd_process could attempt to exit while we are + * running so the lookup function increments the process ref count. + */ + p = kfd_lookup_process_by_mm(mm); + if (!p) + return -ESRCH; + + r = kfd_process_evict_queues(p); + + kfd_unref_process(p); + return r; +} + +int kgd2kfd_resume_mm(struct mm_struct *mm) +{ + struct kfd_process *p; + int r; + + /* Because we are called from arbitrary context (workqueue) as opposed + * to process context, kfd_process could attempt to exit while we are + * running so the lookup function increments the process ref count. + */ + p = kfd_lookup_process_by_mm(mm); + if (!p) + return -ESRCH; + + r = kfd_process_restore_queues(p); + + kfd_unref_process(p); + return r; +} + /** kgd2kfd_schedule_evict_and_restore_process - Schedules work queue that will * prepare for safe eviction of KFD BOs that belong to the specified * process. @@ -652,7 +739,7 @@ int kfd_gtt_sa_allocate(struct kfd_dev *kfd, unsigned int size, if (size > kfd->gtt_sa_num_of_chunks * kfd->gtt_sa_chunk_size) return -ENOMEM; - *mem_obj = kmalloc(sizeof(struct kfd_mem_obj), GFP_KERNEL); + *mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_NOIO); if ((*mem_obj) == NULL) return -ENOMEM; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c index d55d29d31da4e..668ad07ebe1fd 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.c @@ -110,6 +110,57 @@ void program_sh_mem_settings(struct device_queue_manager *dqm, qpd->sh_mem_bases); } +static int allocate_doorbell(struct qcm_process_device *qpd, struct queue *q) +{ + struct kfd_dev *dev = qpd->dqm->dev; + + if (!KFD_IS_SOC15(dev->device_info->asic_family)) { + /* On pre-SOC15 chips we need to use the queue ID to + * preserve the user mode ABI. + */ + q->doorbell_id = q->properties.queue_id; + } else if (q->properties.type == KFD_QUEUE_TYPE_SDMA) { + /* For SDMA queues on SOC15, use static doorbell + * assignments based on the engine and queue. + */ + q->doorbell_id = dev->shared_resources.sdma_doorbell + [q->properties.sdma_engine_id] + [q->properties.sdma_queue_id]; + } else { + /* For CP queues on SOC15 reserve a free doorbell ID */ + unsigned int found; + + found = find_first_zero_bit(qpd->doorbell_bitmap, + KFD_MAX_NUM_OF_QUEUES_PER_PROCESS); + if (found >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) { + pr_debug("No doorbells available"); + return -EBUSY; + } + set_bit(found, qpd->doorbell_bitmap); + q->doorbell_id = found; + } + + q->properties.doorbell_off = + kfd_doorbell_id_to_offset(dev, q->process, + q->doorbell_id); + + return 0; +} + +static void deallocate_doorbell(struct qcm_process_device *qpd, + struct queue *q) +{ + unsigned int old; + struct kfd_dev *dev = qpd->dqm->dev; + + if (!KFD_IS_SOC15(dev->device_info->asic_family) || + q->properties.type == KFD_QUEUE_TYPE_SDMA) + return; + + old = test_and_clear_bit(q->doorbell_id, qpd->doorbell_bitmap); + WARN_ON(!old); +} + static int allocate_vmid(struct device_queue_manager *dqm, struct qcm_process_device *qpd, struct queue *q) @@ -145,15 +196,19 @@ static int allocate_vmid(struct device_queue_manager *dqm, static int flush_texture_cache_nocpsch(struct kfd_dev *kdev, struct qcm_process_device *qpd) { - uint32_t len; + const struct packet_manager_funcs *pmf = qpd->dqm->packets.pmf; + int ret; if (!qpd->ib_kaddr) return -ENOMEM; - len = pm_create_release_mem(qpd->ib_base, (uint32_t *)qpd->ib_kaddr); + ret = pmf->release_mem(qpd->ib_base, (uint32_t *)qpd->ib_kaddr); + if (ret) + return ret; return kdev->kfd2kgd->submit_ib(kdev->kgd, KGD_ENGINE_MEC1, qpd->vmid, - qpd->ib_base, (uint32_t *)qpd->ib_kaddr, len); + qpd->ib_base, (uint32_t *)qpd->ib_kaddr, + pmf->release_mem_size / sizeof(uint32_t)); } static void deallocate_vmid(struct device_queue_manager *dqm, @@ -301,10 +356,14 @@ static int create_compute_queue_nocpsch(struct device_queue_manager *dqm, if (retval) return retval; + retval = allocate_doorbell(qpd, q); + if (retval) + goto out_deallocate_hqd; + retval = mqd->init_mqd(mqd, &q->mqd, &q->mqd_mem_obj, &q->gart_mqd_addr, &q->properties); if (retval) - goto out_deallocate_hqd; + goto out_deallocate_doorbell; pr_debug("Loading mqd to hqd on pipe %d, queue %d\n", q->pipe, q->queue); @@ -324,6 +383,8 @@ static int create_compute_queue_nocpsch(struct device_queue_manager *dqm, out_uninit_mqd: mqd->uninit_mqd(mqd, q->mqd, q->mqd_mem_obj); +out_deallocate_doorbell: + deallocate_doorbell(qpd, q); out_deallocate_hqd: deallocate_hqd(dqm, q); @@ -357,6 +418,8 @@ static int destroy_queue_nocpsch_locked(struct device_queue_manager *dqm, } dqm->total_queue_count--; + deallocate_doorbell(qpd, q); + retval = mqd->destroy_mqd(mqd, q->mqd, KFD_PREEMPT_TYPE_WAVEFRONT_RESET, KFD_UNMAP_LATENCY_MS, @@ -861,6 +924,10 @@ static int create_sdma_queue_nocpsch(struct device_queue_manager *dqm, q->properties.sdma_queue_id = q->sdma_id / CIK_SDMA_QUEUES_PER_ENGINE; q->properties.sdma_engine_id = q->sdma_id % CIK_SDMA_QUEUES_PER_ENGINE; + retval = allocate_doorbell(qpd, q); + if (retval) + goto out_deallocate_sdma_queue; + pr_debug("SDMA id is: %d\n", q->sdma_id); pr_debug("SDMA queue id: %d\n", q->properties.sdma_queue_id); pr_debug("SDMA engine id: %d\n", q->properties.sdma_engine_id); @@ -869,7 +936,7 @@ static int create_sdma_queue_nocpsch(struct device_queue_manager *dqm, retval = mqd->init_mqd(mqd, &q->mqd, &q->mqd_mem_obj, &q->gart_mqd_addr, &q->properties); if (retval) - goto out_deallocate_sdma_queue; + goto out_deallocate_doorbell; retval = mqd->load_mqd(mqd, q->mqd, 0, 0, &q->properties, NULL); if (retval) @@ -879,6 +946,8 @@ static int create_sdma_queue_nocpsch(struct device_queue_manager *dqm, out_uninit_mqd: mqd->uninit_mqd(mqd, q->mqd, q->mqd_mem_obj); +out_deallocate_doorbell: + deallocate_doorbell(qpd, q); out_deallocate_sdma_queue: deallocate_sdma_queue(dqm, q->sdma_id); @@ -1070,12 +1139,17 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q, q->properties.sdma_engine_id = q->sdma_id % CIK_SDMA_QUEUES_PER_ENGINE; } + + retval = allocate_doorbell(qpd, q); + if (retval) + goto out_deallocate_sdma_queue; + mqd = dqm->ops.get_mqd_manager(dqm, get_mqd_type_from_queue_type(q->properties.type)); if (!mqd) { retval = -ENOMEM; - goto out_deallocate_sdma_queue; + goto out_deallocate_doorbell; } /* * Eviction state logic: we only mark active queues as evicted @@ -1093,7 +1167,7 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q, retval = mqd->init_mqd(mqd, &q->mqd, &q->mqd_mem_obj, &q->gart_mqd_addr, &q->properties); if (retval) - goto out_deallocate_sdma_queue; + goto out_deallocate_doorbell; list_add(&q->list, &qpd->queues_list); qpd->queue_count++; @@ -1117,6 +1191,8 @@ static int create_queue_cpsch(struct device_queue_manager *dqm, struct queue *q, mutex_unlock(&dqm->lock); return retval; +out_deallocate_doorbell: + deallocate_doorbell(qpd, q); out_deallocate_sdma_queue: if (q->properties.type == KFD_QUEUE_TYPE_SDMA) deallocate_sdma_queue(dqm, q->sdma_id); @@ -1257,6 +1333,8 @@ static int destroy_queue_cpsch(struct device_queue_manager *dqm, goto failed; } + deallocate_doorbell(qpd, q); + if (q->properties.type == KFD_QUEUE_TYPE_SDMA) { dqm->sdma_queue_count--; deallocate_sdma_queue(dqm, q->sdma_id); @@ -1308,7 +1386,10 @@ static bool set_cache_memory_policy(struct device_queue_manager *dqm, void __user *alternate_aperture_base, uint64_t alternate_aperture_size) { - bool retval; + bool retval = true; + + if (!dqm->asic_ops.set_cache_memory_policy) + return retval; mutex_lock(&dqm->lock); @@ -1577,6 +1658,11 @@ struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev) case CHIP_POLARIS11: device_queue_manager_init_vi_tonga(&dqm->asic_ops); break; + + case CHIP_VEGA10: + case CHIP_RAVEN: + device_queue_manager_init_v9(&dqm->asic_ops); + break; default: WARN(1, "Unexpected ASIC family %u", dev->device_info->asic_family); @@ -1627,6 +1713,18 @@ int dqm_debugfs_hqds(struct seq_file *m, void *data) int pipe, queue; int r = 0; + r = dqm->dev->kfd2kgd->hqd_dump(dqm->dev->kgd, + KFD_CIK_HIQ_PIPE, KFD_CIK_HIQ_QUEUE, &dump, &n_regs); + if (!r) { + seq_printf(m, " HIQ on MEC %d Pipe %d Queue %d\n", + KFD_CIK_HIQ_PIPE/get_pipes_per_mec(dqm)+1, + KFD_CIK_HIQ_PIPE%get_pipes_per_mec(dqm), + KFD_CIK_HIQ_QUEUE); + seq_reg_dump(m, dump, n_regs); + + kfree(dump); + } + for (pipe = 0; pipe < get_pipes_per_mec(dqm); pipe++) { int pipe_offset = pipe * get_queues_per_pipe(dqm); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h index 412beff3281da..59a6b1956932f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager.h @@ -200,6 +200,8 @@ void device_queue_manager_init_vi( struct device_queue_manager_asic_ops *asic_ops); void device_queue_manager_init_vi_tonga( struct device_queue_manager_asic_ops *asic_ops); +void device_queue_manager_init_v9( + struct device_queue_manager_asic_ops *asic_ops); void program_sh_mem_settings(struct device_queue_manager *dqm, struct qcm_process_device *qpd); unsigned int get_queues_num(struct device_queue_manager *dqm); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c new file mode 100644 index 0000000000000..79e5bcf6367cc --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_device_queue_manager_v9.c @@ -0,0 +1,84 @@ +/* + * Copyright 2016-2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "kfd_device_queue_manager.h" +#include "vega10_enum.h" +#include "gc/gc_9_0_offset.h" +#include "gc/gc_9_0_sh_mask.h" +#include "sdma0/sdma0_4_0_sh_mask.h" + +static int update_qpd_v9(struct device_queue_manager *dqm, + struct qcm_process_device *qpd); +static void init_sdma_vm_v9(struct device_queue_manager *dqm, struct queue *q, + struct qcm_process_device *qpd); + +void device_queue_manager_init_v9( + struct device_queue_manager_asic_ops *asic_ops) +{ + asic_ops->update_qpd = update_qpd_v9; + asic_ops->init_sdma_vm = init_sdma_vm_v9; +} + +static uint32_t compute_sh_mem_bases_64bit(struct kfd_process_device *pdd) +{ + uint32_t shared_base = pdd->lds_base >> 48; + uint32_t private_base = pdd->scratch_base >> 48; + + return (shared_base << SH_MEM_BASES__SHARED_BASE__SHIFT) | + private_base; +} + +static int update_qpd_v9(struct device_queue_manager *dqm, + struct qcm_process_device *qpd) +{ + struct kfd_process_device *pdd; + + pdd = qpd_to_pdd(qpd); + + /* check if sh_mem_config register already configured */ + if (qpd->sh_mem_config == 0) { + qpd->sh_mem_config = + SH_MEM_ALIGNMENT_MODE_UNALIGNED << + SH_MEM_CONFIG__ALIGNMENT_MODE__SHIFT; + if (vega10_noretry && + !dqm->dev->device_info->needs_iommu_device) + qpd->sh_mem_config |= + 1 << SH_MEM_CONFIG__RETRY_DISABLE__SHIFT; + + qpd->sh_mem_ape1_limit = 0; + qpd->sh_mem_ape1_base = 0; + } + + qpd->sh_mem_bases = compute_sh_mem_bases_64bit(pdd); + + pr_debug("sh_mem_bases 0x%X\n", qpd->sh_mem_bases); + + return 0; +} + +static void init_sdma_vm_v9(struct device_queue_manager *dqm, struct queue *q, + struct qcm_process_device *qpd) +{ + /* Not needed on SDMAv4 any more */ + q->properties.sdma_vm_addr = 0; +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c index ebb4da14e3df7..c3744d89352c1 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_doorbell.c @@ -33,7 +33,6 @@ static DEFINE_IDA(doorbell_ida); static unsigned int max_doorbell_slices; -#define KFD_SIZE_OF_DOORBELL_IN_BYTES 4 /* * Each device exposes a doorbell aperture, a PCI MMIO aperture that @@ -50,9 +49,9 @@ static unsigned int max_doorbell_slices; */ /* # of doorbell bytes allocated for each process. */ -static inline size_t doorbell_process_allocation(void) +size_t kfd_doorbell_process_slice(struct kfd_dev *kfd) { - return roundup(KFD_SIZE_OF_DOORBELL_IN_BYTES * + return roundup(kfd->device_info->doorbell_size * KFD_MAX_NUM_OF_QUEUES_PER_PROCESS, PAGE_SIZE); } @@ -72,16 +71,16 @@ int kfd_doorbell_init(struct kfd_dev *kfd) doorbell_start_offset = roundup(kfd->shared_resources.doorbell_start_offset, - doorbell_process_allocation()); + kfd_doorbell_process_slice(kfd)); doorbell_aperture_size = rounddown(kfd->shared_resources.doorbell_aperture_size, - doorbell_process_allocation()); + kfd_doorbell_process_slice(kfd)); if (doorbell_aperture_size > doorbell_start_offset) doorbell_process_limit = (doorbell_aperture_size - doorbell_start_offset) / - doorbell_process_allocation(); + kfd_doorbell_process_slice(kfd); else return -ENOSPC; @@ -95,7 +94,7 @@ int kfd_doorbell_init(struct kfd_dev *kfd) kfd->doorbell_id_offset = doorbell_start_offset / sizeof(u32); kfd->doorbell_kernel_ptr = ioremap(kfd->doorbell_base, - doorbell_process_allocation()); + kfd_doorbell_process_slice(kfd)); if (!kfd->doorbell_kernel_ptr) return -ENOMEM; @@ -127,21 +126,16 @@ void kfd_doorbell_fini(struct kfd_dev *kfd) iounmap(kfd->doorbell_kernel_ptr); } -int kfd_doorbell_mmap(struct kfd_process *process, struct vm_area_struct *vma) +int kfd_doorbell_mmap(struct kfd_dev *dev, struct kfd_process *process, + struct vm_area_struct *vma) { phys_addr_t address; - struct kfd_dev *dev; /* * For simplicitly we only allow mapping of the entire doorbell * allocation of a single device & process. */ - if (vma->vm_end - vma->vm_start != doorbell_process_allocation()) - return -EINVAL; - - /* Find kfd device according to gpu id */ - dev = kfd_device_by_id(vma->vm_pgoff); - if (!dev) + if (vma->vm_end - vma->vm_start != kfd_doorbell_process_slice(dev)) return -EINVAL; /* Calculate physical address of doorbell */ @@ -158,19 +152,19 @@ int kfd_doorbell_mmap(struct kfd_process *process, struct vm_area_struct *vma) " vm_flags == 0x%04lX\n" " size == 0x%04lX\n", (unsigned long long) vma->vm_start, address, vma->vm_flags, - doorbell_process_allocation()); + kfd_doorbell_process_slice(dev)); return io_remap_pfn_range(vma, vma->vm_start, address >> PAGE_SHIFT, - doorbell_process_allocation(), + kfd_doorbell_process_slice(dev), vma->vm_page_prot); } /* get kernel iomem pointer for a doorbell */ -u32 __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd, +void __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd, unsigned int *doorbell_off) { u32 inx; @@ -185,6 +179,8 @@ u32 __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd, if (inx >= KFD_MAX_NUM_OF_QUEUES_PER_PROCESS) return NULL; + inx *= kfd->device_info->doorbell_size / sizeof(u32); + /* * Calculating the kernel doorbell offset using the first * doorbell page. @@ -210,7 +206,7 @@ void kfd_release_kernel_doorbell(struct kfd_dev *kfd, u32 __iomem *db_addr) mutex_unlock(&kfd->doorbell_mutex); } -inline void write_kernel_doorbell(u32 __iomem *db, u32 value) +void write_kernel_doorbell(void __iomem *db, u32 value) { if (db) { writel(value, db); @@ -218,30 +214,37 @@ inline void write_kernel_doorbell(u32 __iomem *db, u32 value) } } -/* - * queue_ids are in the range [0,MAX_PROCESS_QUEUES) and are mapped 1:1 - * to doorbells with the process's doorbell page - */ -unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd, +void write_kernel_doorbell64(void __iomem *db, u64 value) +{ + if (db) { + WARN(((unsigned long)db & 7) != 0, + "Unaligned 64-bit doorbell"); + writeq(value, (u64 __iomem *)db); + pr_debug("writing %llu to doorbell address %p\n", value, db); + } +} + +unsigned int kfd_doorbell_id_to_offset(struct kfd_dev *kfd, struct kfd_process *process, - unsigned int queue_id) + unsigned int doorbell_id) { /* * doorbell_id_offset accounts for doorbells taken by KGD. - * index * doorbell_process_allocation/sizeof(u32) adjusts to - * the process's doorbells. + * index * kfd_doorbell_process_slice/sizeof(u32) adjusts to + * the process's doorbells. The offset returned is in dword + * units regardless of the ASIC-dependent doorbell size. */ return kfd->doorbell_id_offset + process->doorbell_index - * doorbell_process_allocation() / sizeof(u32) + - queue_id; + * kfd_doorbell_process_slice(kfd) / sizeof(u32) + + doorbell_id * kfd->device_info->doorbell_size / sizeof(u32); } uint64_t kfd_get_number_elems(struct kfd_dev *kfd) { uint64_t num_of_elems = (kfd->shared_resources.doorbell_aperture_size - kfd->shared_resources.doorbell_start_offset) / - doorbell_process_allocation() + 1; + kfd_doorbell_process_slice(kfd) + 1; return num_of_elems; @@ -251,7 +254,7 @@ phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev, struct kfd_process *process) { return dev->doorbell_base + - process->doorbell_index * doorbell_process_allocation(); + process->doorbell_index * kfd_doorbell_process_slice(dev); } int kfd_alloc_process_doorbells(struct kfd_process *process) diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_events.c b/drivers/gpu/drm/amd/amdkfd/kfd_events.c index 4890a90f1e444..5562e94e786ae 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_events.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_events.c @@ -345,7 +345,7 @@ int kfd_event_create(struct file *devkfd, struct kfd_process *p, case KFD_EVENT_TYPE_DEBUG: ret = create_signal_event(devkfd, p, ev); if (!ret) { - *event_page_offset = KFD_MMAP_EVENTS_MASK; + *event_page_offset = KFD_MMAP_TYPE_EVENTS; *event_page_offset <<= PAGE_SHIFT; *event_slot_index = ev->event_id; } @@ -496,7 +496,7 @@ void kfd_signal_event_interrupt(unsigned int pasid, uint32_t partial_id, pr_debug_ratelimited("Partial ID invalid: %u (%u valid bits)\n", partial_id, valid_id_bits); - if (p->signal_event_count < KFD_SIGNAL_EVENT_LIMIT/2) { + if (p->signal_event_count < KFD_SIGNAL_EVENT_LIMIT / 64) { /* With relatively few events, it's faster to * iterate over the event IDR */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c index 66852de410c82..97d5423c56732 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_flat_memory.c @@ -275,23 +275,35 @@ * for FLAT_* / S_LOAD operations. */ -#define MAKE_GPUVM_APP_BASE(gpu_num) \ +#define MAKE_GPUVM_APP_BASE_VI(gpu_num) \ (((uint64_t)(gpu_num) << 61) + 0x1000000000000L) #define MAKE_GPUVM_APP_LIMIT(base, size) \ (((uint64_t)(base) & 0xFFFFFF0000000000UL) + (size) - 1) -#define MAKE_SCRATCH_APP_BASE() \ +#define MAKE_SCRATCH_APP_BASE_VI() \ (((uint64_t)(0x1UL) << 61) + 0x100000000L) #define MAKE_SCRATCH_APP_LIMIT(base) \ (((uint64_t)base & 0xFFFFFFFF00000000UL) | 0xFFFFFFFF) -#define MAKE_LDS_APP_BASE() \ +#define MAKE_LDS_APP_BASE_VI() \ (((uint64_t)(0x1UL) << 61) + 0x0) #define MAKE_LDS_APP_LIMIT(base) \ (((uint64_t)(base) & 0xFFFFFFFF00000000UL) | 0xFFFFFFFF) +/* On GFXv9 the LDS and scratch apertures are programmed independently + * using the high 16 bits of the 64-bit virtual address. They must be + * in the hole, which will be the case as long as the high 16 bits are + * not 0. + * + * The aperture sizes are still 4GB implicitly. + * + * A GPUVM aperture is not applicable on GFXv9. + */ +#define MAKE_LDS_APP_BASE_V9() ((uint64_t)(0x1UL) << 48) +#define MAKE_SCRATCH_APP_BASE_V9() ((uint64_t)(0x2UL) << 48) + /* User mode manages most of the SVM aperture address space. The low * 16MB are reserved for kernel use (CWSR trap handler and kernel IB * for now). @@ -300,6 +312,55 @@ #define SVM_CWSR_BASE (SVM_USER_BASE - KFD_CWSR_TBA_TMA_SIZE) #define SVM_IB_BASE (SVM_CWSR_BASE - PAGE_SIZE) +static void kfd_init_apertures_vi(struct kfd_process_device *pdd, uint8_t id) +{ + /* + * node id couldn't be 0 - the three MSB bits of + * aperture shoudn't be 0 + */ + pdd->lds_base = MAKE_LDS_APP_BASE_VI(); + pdd->lds_limit = MAKE_LDS_APP_LIMIT(pdd->lds_base); + + if (!pdd->dev->device_info->needs_iommu_device) { + /* dGPUs: SVM aperture starting at 0 + * with small reserved space for kernel. + * Set them to CANONICAL addresses. + */ + pdd->gpuvm_base = SVM_USER_BASE; + pdd->gpuvm_limit = + pdd->dev->shared_resources.gpuvm_size - 1; + } else { + /* set them to non CANONICAL addresses, and no SVM is + * allocated. + */ + pdd->gpuvm_base = MAKE_GPUVM_APP_BASE_VI(id + 1); + pdd->gpuvm_limit = MAKE_GPUVM_APP_LIMIT(pdd->gpuvm_base, + pdd->dev->shared_resources.gpuvm_size); + } + + pdd->scratch_base = MAKE_SCRATCH_APP_BASE_VI(); + pdd->scratch_limit = MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base); +} + +static void kfd_init_apertures_v9(struct kfd_process_device *pdd, uint8_t id) +{ + pdd->lds_base = MAKE_LDS_APP_BASE_V9(); + pdd->lds_limit = MAKE_LDS_APP_LIMIT(pdd->lds_base); + + /* Raven needs SVM to support graphic handle, etc. Leave the small + * reserved space before SVM on Raven as well, even though we don't + * have to. + * Set gpuvm_base and gpuvm_limit to CANONICAL addresses so that they + * are used in Thunk to reserve SVM. + */ + pdd->gpuvm_base = SVM_USER_BASE; + pdd->gpuvm_limit = + pdd->dev->shared_resources.gpuvm_size - 1; + + pdd->scratch_base = MAKE_SCRATCH_APP_BASE_V9(); + pdd->scratch_limit = MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base); +} + int kfd_init_apertures(struct kfd_process *process) { uint8_t id = 0; @@ -307,9 +368,7 @@ int kfd_init_apertures(struct kfd_process *process) struct kfd_process_device *pdd; /*Iterating over all devices*/ - while (kfd_topology_enum_kfd_devices(id, &dev) == 0 && - id < NUM_OF_SUPPORTED_GPUS) { - + while (kfd_topology_enum_kfd_devices(id, &dev) == 0) { if (!dev) { id++; /* Skip non GPU devices */ continue; @@ -318,7 +377,7 @@ int kfd_init_apertures(struct kfd_process *process) pdd = kfd_create_process_device_data(dev, process); if (!pdd) { pr_err("Failed to create process device data\n"); - return -1; + return -ENOMEM; } /* * For 64 bit process apertures will be statically reserved in @@ -330,32 +389,30 @@ int kfd_init_apertures(struct kfd_process *process) pdd->gpuvm_base = pdd->gpuvm_limit = 0; pdd->scratch_base = pdd->scratch_limit = 0; } else { - /* Same LDS and scratch apertures can be used - * on all GPUs. This allows using more dGPUs - * than placement options for apertures. - */ - pdd->lds_base = MAKE_LDS_APP_BASE(); - pdd->lds_limit = MAKE_LDS_APP_LIMIT(pdd->lds_base); - - pdd->scratch_base = MAKE_SCRATCH_APP_BASE(); - pdd->scratch_limit = - MAKE_SCRATCH_APP_LIMIT(pdd->scratch_base); + switch (dev->device_info->asic_family) { + case CHIP_KAVERI: + case CHIP_HAWAII: + case CHIP_CARRIZO: + case CHIP_TONGA: + case CHIP_FIJI: + case CHIP_POLARIS10: + case CHIP_POLARIS11: + kfd_init_apertures_vi(pdd, id); + break; + case CHIP_VEGA10: + case CHIP_RAVEN: + kfd_init_apertures_v9(pdd, id); + break; + default: + WARN(1, "Unexpected ASIC family %u", + dev->device_info->asic_family); + return -EINVAL; + } - if (dev->device_info->needs_iommu_device) { - /* APUs: GPUVM aperture in - * non-canonical address space - */ - pdd->gpuvm_base = MAKE_GPUVM_APP_BASE(id + 1); - pdd->gpuvm_limit = MAKE_GPUVM_APP_LIMIT( - pdd->gpuvm_base, - dev->shared_resources.gpuvm_size); - } else { - /* dGPUs: SVM aperture starting at 0 - * with small reserved space for kernel + if (!dev->device_info->needs_iommu_device) { + /* dGPUs: the reserved space for kernel + * before SVM */ - pdd->gpuvm_base = SVM_USER_BASE; - pdd->gpuvm_limit = - dev->shared_resources.gpuvm_size - 1; pdd->qpd.cwsr_base = SVM_CWSR_BASE; pdd->qpd.ib_base = SVM_IB_BASE; } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c new file mode 100644 index 0000000000000..37029baa33460 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_int_process_v9.c @@ -0,0 +1,92 @@ +/* + * Copyright 2016-2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#include "kfd_priv.h" +#include "kfd_events.h" +#include "soc15_int.h" + + +static bool event_interrupt_isr_v9(struct kfd_dev *dev, + const uint32_t *ih_ring_entry) +{ + uint16_t source_id, client_id, pasid, vmid; + const uint32_t *data = ih_ring_entry; + + /* Only handle interrupts from KFD VMIDs */ + vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry); + if (vmid < dev->vm_info.first_vmid_kfd || + vmid > dev->vm_info.last_vmid_kfd) + return 0; + + /* If there is no valid PASID, it's likely a firmware bug */ + pasid = SOC15_PASID_FROM_IH_ENTRY(ih_ring_entry); + if (WARN_ONCE(pasid == 0, "FW bug: No PASID in KFD interrupt")) + return 0; + + source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry); + client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry); + + pr_debug("client id 0x%x, source id %d, pasid 0x%x. raw data:\n", + client_id, source_id, pasid); + pr_debug("%8X, %8X, %8X, %8X, %8X, %8X, %8X, %8X.\n", + data[0], data[1], data[2], data[3], + data[4], data[5], data[6], data[7]); + + /* Interrupt types we care about: various signals and faults. + * They will be forwarded to a work queue (see below). + */ + return source_id == SOC15_INTSRC_CP_END_OF_PIPE || + source_id == SOC15_INTSRC_SDMA_TRAP || + source_id == SOC15_INTSRC_SQ_INTERRUPT_MSG || + source_id == SOC15_INTSRC_CP_BAD_OPCODE; +} + +static void event_interrupt_wq_v9(struct kfd_dev *dev, + const uint32_t *ih_ring_entry) +{ + uint16_t source_id, client_id, pasid, vmid; + uint32_t context_id; + + source_id = SOC15_SOURCE_ID_FROM_IH_ENTRY(ih_ring_entry); + client_id = SOC15_CLIENT_ID_FROM_IH_ENTRY(ih_ring_entry); + pasid = SOC15_PASID_FROM_IH_ENTRY(ih_ring_entry); + vmid = SOC15_VMID_FROM_IH_ENTRY(ih_ring_entry); + context_id = SOC15_CONTEXT_ID0_FROM_IH_ENTRY(ih_ring_entry); + + if (source_id == SOC15_INTSRC_CP_END_OF_PIPE) + kfd_signal_event_interrupt(pasid, context_id, 32); + else if (source_id == SOC15_INTSRC_SDMA_TRAP) + kfd_signal_event_interrupt(pasid, context_id & 0xfffffff, 28); + else if (source_id == SOC15_INTSRC_SQ_INTERRUPT_MSG) + kfd_signal_event_interrupt(pasid, context_id & 0xffffff, 24); + else if (source_id == SOC15_INTSRC_CP_BAD_OPCODE) + kfd_signal_hw_exception_event(pasid); + else if (client_id == SOC15_IH_CLIENTID_VMC || + client_id == SOC15_IH_CLIENTID_UTCL2) { + /* TODO */ + } +} + +const struct kfd_event_interrupt_class event_interrupt_class_v9 = { + .interrupt_isr = event_interrupt_isr_v9, + .interrupt_wq = event_interrupt_wq_v9, +}; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c index 035c351f47c5a..db6d9336b80d2 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_interrupt.c @@ -139,10 +139,12 @@ static void interrupt_wq(struct work_struct *work) { struct kfd_dev *dev = container_of(work, struct kfd_dev, interrupt_work); + uint32_t ih_ring_entry[KFD_MAX_RING_ENTRY_SIZE]; - uint32_t ih_ring_entry[DIV_ROUND_UP( - dev->device_info->ih_ring_entry_size, - sizeof(uint32_t))]; + if (dev->device_info->ih_ring_entry_size > sizeof(ih_ring_entry)) { + dev_err_once(kfd_chardev(), "Ring entry too small\n"); + return; + } while (dequeue_ih_ring_entry(dev, ih_ring_entry)) dev->device_info->event_interrupt_class->interrupt_wq(dev, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c index 69f4964853310..476951d8c91cc 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.c @@ -99,7 +99,7 @@ static bool initialize(struct kernel_queue *kq, struct kfd_dev *dev, kq->rptr_kernel = kq->rptr_mem->cpu_ptr; kq->rptr_gpu_addr = kq->rptr_mem->gpu_addr; - retval = kfd_gtt_sa_allocate(dev, sizeof(*kq->wptr_kernel), + retval = kfd_gtt_sa_allocate(dev, dev->device_info->doorbell_size, &kq->wptr_mem); if (retval != 0) @@ -208,6 +208,7 @@ static int acquire_packet_buffer(struct kernel_queue *kq, size_t available_size; size_t queue_size_dwords; uint32_t wptr, rptr; + uint64_t wptr64; unsigned int *queue_address; /* When rptr == wptr, the buffer is empty. @@ -216,7 +217,8 @@ static int acquire_packet_buffer(struct kernel_queue *kq, * the opposite. So we can only use up to queue_size_dwords - 1 dwords. */ rptr = *kq->rptr_kernel; - wptr = *kq->wptr_kernel; + wptr = kq->pending_wptr; + wptr64 = kq->pending_wptr64; queue_address = (unsigned int *)kq->pq_kernel_addr; queue_size_dwords = kq->queue->properties.queue_size / 4; @@ -232,29 +234,33 @@ static int acquire_packet_buffer(struct kernel_queue *kq, * make sure calling functions know * acquire_packet_buffer() failed */ - *buffer_ptr = NULL; - return -ENOMEM; + goto err_no_space; } if (wptr + packet_size_in_dwords >= queue_size_dwords) { /* make sure after rolling back to position 0, there is * still enough space. */ - if (packet_size_in_dwords >= rptr) { - *buffer_ptr = NULL; - return -ENOMEM; - } + if (packet_size_in_dwords >= rptr) + goto err_no_space; + /* fill nops, roll back and start at position 0 */ while (wptr > 0) { queue_address[wptr] = kq->nop_packet; wptr = (wptr + 1) % queue_size_dwords; + wptr64++; } } *buffer_ptr = &queue_address[wptr]; kq->pending_wptr = wptr + packet_size_in_dwords; + kq->pending_wptr64 = wptr64 + packet_size_in_dwords; return 0; + +err_no_space: + *buffer_ptr = NULL; + return -ENOMEM; } static void submit_packet(struct kernel_queue *kq) @@ -270,14 +276,18 @@ static void submit_packet(struct kernel_queue *kq) pr_debug("\n"); #endif - *kq->wptr_kernel = kq->pending_wptr; - write_kernel_doorbell(kq->queue->properties.doorbell_ptr, - kq->pending_wptr); + kq->ops_asic_specific.submit_packet(kq); } static void rollback_packet(struct kernel_queue *kq) { - kq->pending_wptr = *kq->queue->properties.write_ptr; + if (kq->dev->device_info->doorbell_size == 8) { + kq->pending_wptr64 = *kq->wptr64_kernel; + kq->pending_wptr = *kq->wptr_kernel % + (kq->queue->properties.queue_size / 4); + } else { + kq->pending_wptr = *kq->wptr_kernel; + } } struct kernel_queue *kernel_queue_init(struct kfd_dev *dev, @@ -308,6 +318,11 @@ struct kernel_queue *kernel_queue_init(struct kfd_dev *dev, case CHIP_HAWAII: kernel_queue_init_cik(&kq->ops_asic_specific); break; + + case CHIP_VEGA10: + case CHIP_RAVEN: + kernel_queue_init_v9(&kq->ops_asic_specific); + break; default: WARN(1, "Unexpected ASIC family %u", dev->device_info->asic_family); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h index 594053136ee4d..97aff2041a5d1 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue.h @@ -72,6 +72,7 @@ struct kernel_queue { struct kfd_dev *dev; struct mqd_manager *mqd; struct queue *queue; + uint64_t pending_wptr64; uint32_t pending_wptr; unsigned int nop_packet; @@ -79,7 +80,10 @@ struct kernel_queue { uint32_t *rptr_kernel; uint64_t rptr_gpu_addr; struct kfd_mem_obj *wptr_mem; - uint32_t *wptr_kernel; + union { + uint64_t *wptr64_kernel; + uint32_t *wptr_kernel; + }; uint64_t wptr_gpu_addr; struct kfd_mem_obj *pq; uint64_t pq_gpu_addr; @@ -97,5 +101,6 @@ struct kernel_queue { void kernel_queue_init_cik(struct kernel_queue_ops *ops); void kernel_queue_init_vi(struct kernel_queue_ops *ops); +void kernel_queue_init_v9(struct kernel_queue_ops *ops); #endif /* KFD_KERNEL_QUEUE_H_ */ diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_cik.c index a90eb440b1fbb..19e54acb4125c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_cik.c @@ -26,11 +26,13 @@ static bool initialize_cik(struct kernel_queue *kq, struct kfd_dev *dev, enum kfd_queue_type type, unsigned int queue_size); static void uninitialize_cik(struct kernel_queue *kq); +static void submit_packet_cik(struct kernel_queue *kq); void kernel_queue_init_cik(struct kernel_queue_ops *ops) { ops->initialize = initialize_cik; ops->uninitialize = uninitialize_cik; + ops->submit_packet = submit_packet_cik; } static bool initialize_cik(struct kernel_queue *kq, struct kfd_dev *dev, @@ -42,3 +44,10 @@ static bool initialize_cik(struct kernel_queue *kq, struct kfd_dev *dev, static void uninitialize_cik(struct kernel_queue *kq) { } + +static void submit_packet_cik(struct kernel_queue *kq) +{ + *kq->wptr_kernel = kq->pending_wptr; + write_kernel_doorbell(kq->queue->properties.doorbell_ptr, + kq->pending_wptr); +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c new file mode 100644 index 0000000000000..684a3bf07efde --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_v9.c @@ -0,0 +1,340 @@ +/* + * Copyright 2016-2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include "kfd_kernel_queue.h" +#include "kfd_device_queue_manager.h" +#include "kfd_pm4_headers_ai.h" +#include "kfd_pm4_opcodes.h" + +static bool initialize_v9(struct kernel_queue *kq, struct kfd_dev *dev, + enum kfd_queue_type type, unsigned int queue_size); +static void uninitialize_v9(struct kernel_queue *kq); +static void submit_packet_v9(struct kernel_queue *kq); + +void kernel_queue_init_v9(struct kernel_queue_ops *ops) +{ + ops->initialize = initialize_v9; + ops->uninitialize = uninitialize_v9; + ops->submit_packet = submit_packet_v9; +} + +static bool initialize_v9(struct kernel_queue *kq, struct kfd_dev *dev, + enum kfd_queue_type type, unsigned int queue_size) +{ + int retval; + + retval = kfd_gtt_sa_allocate(dev, PAGE_SIZE, &kq->eop_mem); + if (retval) + return false; + + kq->eop_gpu_addr = kq->eop_mem->gpu_addr; + kq->eop_kernel_addr = kq->eop_mem->cpu_ptr; + + memset(kq->eop_kernel_addr, 0, PAGE_SIZE); + + return true; +} + +static void uninitialize_v9(struct kernel_queue *kq) +{ + kfd_gtt_sa_free(kq->dev, kq->eop_mem); +} + +static void submit_packet_v9(struct kernel_queue *kq) +{ + *kq->wptr64_kernel = kq->pending_wptr64; + write_kernel_doorbell64(kq->queue->properties.doorbell_ptr, + kq->pending_wptr64); +} + +static int pm_map_process_v9(struct packet_manager *pm, + uint32_t *buffer, struct qcm_process_device *qpd) +{ + struct pm4_mes_map_process *packet; + uint64_t vm_page_table_base_addr = + (uint64_t)(qpd->page_table_base) << 12; + + packet = (struct pm4_mes_map_process *)buffer; + memset(buffer, 0, sizeof(struct pm4_mes_map_process)); + + packet->header.u32All = pm_build_pm4_header(IT_MAP_PROCESS, + sizeof(struct pm4_mes_map_process)); + packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0; + packet->bitfields2.process_quantum = 1; + packet->bitfields2.pasid = qpd->pqm->process->pasid; + packet->bitfields14.gds_size = qpd->gds_size; + packet->bitfields14.num_gws = qpd->num_gws; + packet->bitfields14.num_oac = qpd->num_oac; + packet->bitfields14.sdma_enable = 1; + packet->bitfields14.num_queues = (qpd->is_debug) ? 0 : qpd->queue_count; + + packet->sh_mem_config = qpd->sh_mem_config; + packet->sh_mem_bases = qpd->sh_mem_bases; + packet->sq_shader_tba_lo = lower_32_bits(qpd->tba_addr >> 8); + packet->sq_shader_tba_hi = upper_32_bits(qpd->tba_addr >> 8); + packet->sq_shader_tma_lo = lower_32_bits(qpd->tma_addr >> 8); + packet->sq_shader_tma_hi = upper_32_bits(qpd->tma_addr >> 8); + + packet->gds_addr_lo = lower_32_bits(qpd->gds_context_area); + packet->gds_addr_hi = upper_32_bits(qpd->gds_context_area); + + packet->vm_context_page_table_base_addr_lo32 = + lower_32_bits(vm_page_table_base_addr); + packet->vm_context_page_table_base_addr_hi32 = + upper_32_bits(vm_page_table_base_addr); + + return 0; +} + +static int pm_runlist_v9(struct packet_manager *pm, uint32_t *buffer, + uint64_t ib, size_t ib_size_in_dwords, bool chain) +{ + struct pm4_mes_runlist *packet; + + int concurrent_proc_cnt = 0; + struct kfd_dev *kfd = pm->dqm->dev; + + /* Determine the number of processes to map together to HW: + * it can not exceed the number of VMIDs available to the + * scheduler, and it is determined by the smaller of the number + * of processes in the runlist and kfd module parameter + * hws_max_conc_proc. + * Note: the arbitration between the number of VMIDs and + * hws_max_conc_proc has been done in + * kgd2kfd_device_init(). + */ + concurrent_proc_cnt = min(pm->dqm->processes_count, + kfd->max_proc_per_quantum); + + packet = (struct pm4_mes_runlist *)buffer; + + memset(buffer, 0, sizeof(struct pm4_mes_runlist)); + packet->header.u32All = pm_build_pm4_header(IT_RUN_LIST, + sizeof(struct pm4_mes_runlist)); + + packet->bitfields4.ib_size = ib_size_in_dwords; + packet->bitfields4.chain = chain ? 1 : 0; + packet->bitfields4.offload_polling = 0; + packet->bitfields4.valid = 1; + packet->bitfields4.process_cnt = concurrent_proc_cnt; + packet->ordinal2 = lower_32_bits(ib); + packet->ib_base_hi = upper_32_bits(ib); + + return 0; +} + +static int pm_map_queues_v9(struct packet_manager *pm, uint32_t *buffer, + struct queue *q, bool is_static) +{ + struct pm4_mes_map_queues *packet; + bool use_static = is_static; + + packet = (struct pm4_mes_map_queues *)buffer; + memset(buffer, 0, sizeof(struct pm4_mes_map_queues)); + + packet->header.u32All = pm_build_pm4_header(IT_MAP_QUEUES, + sizeof(struct pm4_mes_map_queues)); + packet->bitfields2.alloc_format = + alloc_format__mes_map_queues__one_per_pipe_vi; + packet->bitfields2.num_queues = 1; + packet->bitfields2.queue_sel = + queue_sel__mes_map_queues__map_to_hws_determined_queue_slots_vi; + + packet->bitfields2.engine_sel = + engine_sel__mes_map_queues__compute_vi; + packet->bitfields2.queue_type = + queue_type__mes_map_queues__normal_compute_vi; + + switch (q->properties.type) { + case KFD_QUEUE_TYPE_COMPUTE: + if (use_static) + packet->bitfields2.queue_type = + queue_type__mes_map_queues__normal_latency_static_queue_vi; + break; + case KFD_QUEUE_TYPE_DIQ: + packet->bitfields2.queue_type = + queue_type__mes_map_queues__debug_interface_queue_vi; + break; + case KFD_QUEUE_TYPE_SDMA: + packet->bitfields2.engine_sel = q->properties.sdma_engine_id + + engine_sel__mes_map_queues__sdma0_vi; + use_static = false; /* no static queues under SDMA */ + break; + default: + WARN(1, "queue type %d", q->properties.type); + return -EINVAL; + } + packet->bitfields3.doorbell_offset = + q->properties.doorbell_off; + + packet->mqd_addr_lo = + lower_32_bits(q->gart_mqd_addr); + + packet->mqd_addr_hi = + upper_32_bits(q->gart_mqd_addr); + + packet->wptr_addr_lo = + lower_32_bits((uint64_t)q->properties.write_ptr); + + packet->wptr_addr_hi = + upper_32_bits((uint64_t)q->properties.write_ptr); + + return 0; +} + +static int pm_unmap_queues_v9(struct packet_manager *pm, uint32_t *buffer, + enum kfd_queue_type type, + enum kfd_unmap_queues_filter filter, + uint32_t filter_param, bool reset, + unsigned int sdma_engine) +{ + struct pm4_mes_unmap_queues *packet; + + packet = (struct pm4_mes_unmap_queues *)buffer; + memset(buffer, 0, sizeof(struct pm4_mes_unmap_queues)); + + packet->header.u32All = pm_build_pm4_header(IT_UNMAP_QUEUES, + sizeof(struct pm4_mes_unmap_queues)); + switch (type) { + case KFD_QUEUE_TYPE_COMPUTE: + case KFD_QUEUE_TYPE_DIQ: + packet->bitfields2.engine_sel = + engine_sel__mes_unmap_queues__compute; + break; + case KFD_QUEUE_TYPE_SDMA: + packet->bitfields2.engine_sel = + engine_sel__mes_unmap_queues__sdma0 + sdma_engine; + break; + default: + WARN(1, "queue type %d", type); + return -EINVAL; + } + + if (reset) + packet->bitfields2.action = + action__mes_unmap_queues__reset_queues; + else + packet->bitfields2.action = + action__mes_unmap_queues__preempt_queues; + + switch (filter) { + case KFD_UNMAP_QUEUES_FILTER_SINGLE_QUEUE: + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__perform_request_on_specified_queues; + packet->bitfields2.num_queues = 1; + packet->bitfields3b.doorbell_offset0 = filter_param; + break; + case KFD_UNMAP_QUEUES_FILTER_BY_PASID: + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__perform_request_on_pasid_queues; + packet->bitfields3a.pasid = filter_param; + break; + case KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES: + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__unmap_all_queues; + break; + case KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES: + /* in this case, we do not preempt static queues */ + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__unmap_all_non_static_queues; + break; + default: + WARN(1, "filter %d", filter); + return -EINVAL; + } + + return 0; + +} + +static int pm_query_status_v9(struct packet_manager *pm, uint32_t *buffer, + uint64_t fence_address, uint32_t fence_value) +{ + struct pm4_mes_query_status *packet; + + packet = (struct pm4_mes_query_status *)buffer; + memset(buffer, 0, sizeof(struct pm4_mes_query_status)); + + + packet->header.u32All = pm_build_pm4_header(IT_QUERY_STATUS, + sizeof(struct pm4_mes_query_status)); + + packet->bitfields2.context_id = 0; + packet->bitfields2.interrupt_sel = + interrupt_sel__mes_query_status__completion_status; + packet->bitfields2.command = + command__mes_query_status__fence_only_after_write_ack; + + packet->addr_hi = upper_32_bits((uint64_t)fence_address); + packet->addr_lo = lower_32_bits((uint64_t)fence_address); + packet->data_hi = upper_32_bits((uint64_t)fence_value); + packet->data_lo = lower_32_bits((uint64_t)fence_value); + + return 0; +} + + +static int pm_release_mem_v9(uint64_t gpu_addr, uint32_t *buffer) +{ + struct pm4_mec_release_mem *packet; + + packet = (struct pm4_mec_release_mem *)buffer; + memset(buffer, 0, sizeof(struct pm4_mec_release_mem)); + + packet->header.u32All = pm_build_pm4_header(IT_RELEASE_MEM, + sizeof(struct pm4_mec_release_mem)); + + packet->bitfields2.event_type = CACHE_FLUSH_AND_INV_TS_EVENT; + packet->bitfields2.event_index = event_index__mec_release_mem__end_of_pipe; + packet->bitfields2.tcl1_action_ena = 1; + packet->bitfields2.tc_action_ena = 1; + packet->bitfields2.cache_policy = cache_policy__mec_release_mem__lru; + + packet->bitfields3.data_sel = data_sel__mec_release_mem__send_32_bit_low; + packet->bitfields3.int_sel = + int_sel__mec_release_mem__send_interrupt_after_write_confirm; + + packet->bitfields4.address_lo_32b = (gpu_addr & 0xffffffff) >> 2; + packet->address_hi = upper_32_bits(gpu_addr); + + packet->data_lo = 0; + + return 0; +} + +const struct packet_manager_funcs kfd_v9_pm_funcs = { + .map_process = pm_map_process_v9, + .runlist = pm_runlist_v9, + .set_resources = pm_set_resources_vi, + .map_queues = pm_map_queues_v9, + .unmap_queues = pm_unmap_queues_v9, + .query_status = pm_query_status_v9, + .release_mem = pm_release_mem_v9, + .map_process_size = sizeof(struct pm4_mes_map_process), + .runlist_size = sizeof(struct pm4_mes_runlist), + .set_resources_size = sizeof(struct pm4_mes_set_resources), + .map_queues_size = sizeof(struct pm4_mes_map_queues), + .unmap_queues_size = sizeof(struct pm4_mes_unmap_queues), + .query_status_size = sizeof(struct pm4_mes_query_status), + .release_mem_size = sizeof(struct pm4_mec_release_mem) +}; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_vi.c index f1d48281e322a..bf20c6d32ef33 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_kernel_queue_vi.c @@ -22,15 +22,20 @@ */ #include "kfd_kernel_queue.h" +#include "kfd_device_queue_manager.h" +#include "kfd_pm4_headers_vi.h" +#include "kfd_pm4_opcodes.h" static bool initialize_vi(struct kernel_queue *kq, struct kfd_dev *dev, enum kfd_queue_type type, unsigned int queue_size); static void uninitialize_vi(struct kernel_queue *kq); +static void submit_packet_vi(struct kernel_queue *kq); void kernel_queue_init_vi(struct kernel_queue_ops *ops) { ops->initialize = initialize_vi; ops->uninitialize = uninitialize_vi; + ops->submit_packet = submit_packet_vi; } static bool initialize_vi(struct kernel_queue *kq, struct kfd_dev *dev, @@ -54,3 +59,317 @@ static void uninitialize_vi(struct kernel_queue *kq) { kfd_gtt_sa_free(kq->dev, kq->eop_mem); } + +static void submit_packet_vi(struct kernel_queue *kq) +{ + *kq->wptr_kernel = kq->pending_wptr; + write_kernel_doorbell(kq->queue->properties.doorbell_ptr, + kq->pending_wptr); +} + +unsigned int pm_build_pm4_header(unsigned int opcode, size_t packet_size) +{ + union PM4_MES_TYPE_3_HEADER header; + + header.u32All = 0; + header.opcode = opcode; + header.count = packet_size / 4 - 2; + header.type = PM4_TYPE_3; + + return header.u32All; +} + +static int pm_map_process_vi(struct packet_manager *pm, uint32_t *buffer, + struct qcm_process_device *qpd) +{ + struct pm4_mes_map_process *packet; + + packet = (struct pm4_mes_map_process *)buffer; + + memset(buffer, 0, sizeof(struct pm4_mes_map_process)); + + packet->header.u32All = pm_build_pm4_header(IT_MAP_PROCESS, + sizeof(struct pm4_mes_map_process)); + packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0; + packet->bitfields2.process_quantum = 1; + packet->bitfields2.pasid = qpd->pqm->process->pasid; + packet->bitfields3.page_table_base = qpd->page_table_base; + packet->bitfields10.gds_size = qpd->gds_size; + packet->bitfields10.num_gws = qpd->num_gws; + packet->bitfields10.num_oac = qpd->num_oac; + packet->bitfields10.num_queues = (qpd->is_debug) ? 0 : qpd->queue_count; + + packet->sh_mem_config = qpd->sh_mem_config; + packet->sh_mem_bases = qpd->sh_mem_bases; + packet->sh_mem_ape1_base = qpd->sh_mem_ape1_base; + packet->sh_mem_ape1_limit = qpd->sh_mem_ape1_limit; + + packet->sh_hidden_private_base_vmid = qpd->sh_hidden_private_base; + + packet->gds_addr_lo = lower_32_bits(qpd->gds_context_area); + packet->gds_addr_hi = upper_32_bits(qpd->gds_context_area); + + return 0; +} + +static int pm_runlist_vi(struct packet_manager *pm, uint32_t *buffer, + uint64_t ib, size_t ib_size_in_dwords, bool chain) +{ + struct pm4_mes_runlist *packet; + int concurrent_proc_cnt = 0; + struct kfd_dev *kfd = pm->dqm->dev; + + if (WARN_ON(!ib)) + return -EFAULT; + + /* Determine the number of processes to map together to HW: + * it can not exceed the number of VMIDs available to the + * scheduler, and it is determined by the smaller of the number + * of processes in the runlist and kfd module parameter + * hws_max_conc_proc. + * Note: the arbitration between the number of VMIDs and + * hws_max_conc_proc has been done in + * kgd2kfd_device_init(). + */ + concurrent_proc_cnt = min(pm->dqm->processes_count, + kfd->max_proc_per_quantum); + + packet = (struct pm4_mes_runlist *)buffer; + + memset(buffer, 0, sizeof(struct pm4_mes_runlist)); + packet->header.u32All = pm_build_pm4_header(IT_RUN_LIST, + sizeof(struct pm4_mes_runlist)); + + packet->bitfields4.ib_size = ib_size_in_dwords; + packet->bitfields4.chain = chain ? 1 : 0; + packet->bitfields4.offload_polling = 0; + packet->bitfields4.valid = 1; + packet->bitfields4.process_cnt = concurrent_proc_cnt; + packet->ordinal2 = lower_32_bits(ib); + packet->bitfields3.ib_base_hi = upper_32_bits(ib); + + return 0; +} + +int pm_set_resources_vi(struct packet_manager *pm, uint32_t *buffer, + struct scheduling_resources *res) +{ + struct pm4_mes_set_resources *packet; + + packet = (struct pm4_mes_set_resources *)buffer; + memset(buffer, 0, sizeof(struct pm4_mes_set_resources)); + + packet->header.u32All = pm_build_pm4_header(IT_SET_RESOURCES, + sizeof(struct pm4_mes_set_resources)); + + packet->bitfields2.queue_type = + queue_type__mes_set_resources__hsa_interface_queue_hiq; + packet->bitfields2.vmid_mask = res->vmid_mask; + packet->bitfields2.unmap_latency = KFD_UNMAP_LATENCY_MS / 100; + packet->bitfields7.oac_mask = res->oac_mask; + packet->bitfields8.gds_heap_base = res->gds_heap_base; + packet->bitfields8.gds_heap_size = res->gds_heap_size; + + packet->gws_mask_lo = lower_32_bits(res->gws_mask); + packet->gws_mask_hi = upper_32_bits(res->gws_mask); + + packet->queue_mask_lo = lower_32_bits(res->queue_mask); + packet->queue_mask_hi = upper_32_bits(res->queue_mask); + + return 0; +} + +static int pm_map_queues_vi(struct packet_manager *pm, uint32_t *buffer, + struct queue *q, bool is_static) +{ + struct pm4_mes_map_queues *packet; + bool use_static = is_static; + + packet = (struct pm4_mes_map_queues *)buffer; + memset(buffer, 0, sizeof(struct pm4_mes_map_queues)); + + packet->header.u32All = pm_build_pm4_header(IT_MAP_QUEUES, + sizeof(struct pm4_mes_map_queues)); + packet->bitfields2.alloc_format = + alloc_format__mes_map_queues__one_per_pipe_vi; + packet->bitfields2.num_queues = 1; + packet->bitfields2.queue_sel = + queue_sel__mes_map_queues__map_to_hws_determined_queue_slots_vi; + + packet->bitfields2.engine_sel = + engine_sel__mes_map_queues__compute_vi; + packet->bitfields2.queue_type = + queue_type__mes_map_queues__normal_compute_vi; + + switch (q->properties.type) { + case KFD_QUEUE_TYPE_COMPUTE: + if (use_static) + packet->bitfields2.queue_type = + queue_type__mes_map_queues__normal_latency_static_queue_vi; + break; + case KFD_QUEUE_TYPE_DIQ: + packet->bitfields2.queue_type = + queue_type__mes_map_queues__debug_interface_queue_vi; + break; + case KFD_QUEUE_TYPE_SDMA: + packet->bitfields2.engine_sel = q->properties.sdma_engine_id + + engine_sel__mes_map_queues__sdma0_vi; + use_static = false; /* no static queues under SDMA */ + break; + default: + WARN(1, "queue type %d", q->properties.type); + return -EINVAL; + } + packet->bitfields3.doorbell_offset = + q->properties.doorbell_off; + + packet->mqd_addr_lo = + lower_32_bits(q->gart_mqd_addr); + + packet->mqd_addr_hi = + upper_32_bits(q->gart_mqd_addr); + + packet->wptr_addr_lo = + lower_32_bits((uint64_t)q->properties.write_ptr); + + packet->wptr_addr_hi = + upper_32_bits((uint64_t)q->properties.write_ptr); + + return 0; +} + +static int pm_unmap_queues_vi(struct packet_manager *pm, uint32_t *buffer, + enum kfd_queue_type type, + enum kfd_unmap_queues_filter filter, + uint32_t filter_param, bool reset, + unsigned int sdma_engine) +{ + struct pm4_mes_unmap_queues *packet; + + packet = (struct pm4_mes_unmap_queues *)buffer; + memset(buffer, 0, sizeof(struct pm4_mes_unmap_queues)); + + packet->header.u32All = pm_build_pm4_header(IT_UNMAP_QUEUES, + sizeof(struct pm4_mes_unmap_queues)); + switch (type) { + case KFD_QUEUE_TYPE_COMPUTE: + case KFD_QUEUE_TYPE_DIQ: + packet->bitfields2.engine_sel = + engine_sel__mes_unmap_queues__compute; + break; + case KFD_QUEUE_TYPE_SDMA: + packet->bitfields2.engine_sel = + engine_sel__mes_unmap_queues__sdma0 + sdma_engine; + break; + default: + WARN(1, "queue type %d", type); + return -EINVAL; + } + + if (reset) + packet->bitfields2.action = + action__mes_unmap_queues__reset_queues; + else + packet->bitfields2.action = + action__mes_unmap_queues__preempt_queues; + + switch (filter) { + case KFD_UNMAP_QUEUES_FILTER_SINGLE_QUEUE: + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__perform_request_on_specified_queues; + packet->bitfields2.num_queues = 1; + packet->bitfields3b.doorbell_offset0 = filter_param; + break; + case KFD_UNMAP_QUEUES_FILTER_BY_PASID: + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__perform_request_on_pasid_queues; + packet->bitfields3a.pasid = filter_param; + break; + case KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES: + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__unmap_all_queues; + break; + case KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES: + /* in this case, we do not preempt static queues */ + packet->bitfields2.queue_sel = + queue_sel__mes_unmap_queues__unmap_all_non_static_queues; + break; + default: + WARN(1, "filter %d", filter); + return -EINVAL; + } + + return 0; + +} + +static int pm_query_status_vi(struct packet_manager *pm, uint32_t *buffer, + uint64_t fence_address, uint32_t fence_value) +{ + struct pm4_mes_query_status *packet; + + packet = (struct pm4_mes_query_status *)buffer; + memset(buffer, 0, sizeof(struct pm4_mes_query_status)); + + packet->header.u32All = pm_build_pm4_header(IT_QUERY_STATUS, + sizeof(struct pm4_mes_query_status)); + + packet->bitfields2.context_id = 0; + packet->bitfields2.interrupt_sel = + interrupt_sel__mes_query_status__completion_status; + packet->bitfields2.command = + command__mes_query_status__fence_only_after_write_ack; + + packet->addr_hi = upper_32_bits((uint64_t)fence_address); + packet->addr_lo = lower_32_bits((uint64_t)fence_address); + packet->data_hi = upper_32_bits((uint64_t)fence_value); + packet->data_lo = lower_32_bits((uint64_t)fence_value); + + return 0; +} + +static int pm_release_mem_vi(uint64_t gpu_addr, uint32_t *buffer) +{ + struct pm4_mec_release_mem *packet; + + packet = (struct pm4_mec_release_mem *)buffer; + memset(buffer, 0, sizeof(*packet)); + + packet->header.u32All = pm_build_pm4_header(IT_RELEASE_MEM, + sizeof(*packet)); + + packet->bitfields2.event_type = CACHE_FLUSH_AND_INV_TS_EVENT; + packet->bitfields2.event_index = event_index___release_mem__end_of_pipe; + packet->bitfields2.tcl1_action_ena = 1; + packet->bitfields2.tc_action_ena = 1; + packet->bitfields2.cache_policy = cache_policy___release_mem__lru; + packet->bitfields2.atc = 0; + + packet->bitfields3.data_sel = data_sel___release_mem__send_32_bit_low; + packet->bitfields3.int_sel = + int_sel___release_mem__send_interrupt_after_write_confirm; + + packet->bitfields4.address_lo_32b = (gpu_addr & 0xffffffff) >> 2; + packet->address_hi = upper_32_bits(gpu_addr); + + packet->data_lo = 0; + + return 0; +} + +const struct packet_manager_funcs kfd_vi_pm_funcs = { + .map_process = pm_map_process_vi, + .runlist = pm_runlist_vi, + .set_resources = pm_set_resources_vi, + .map_queues = pm_map_queues_vi, + .unmap_queues = pm_unmap_queues_vi, + .query_status = pm_query_status_vi, + .release_mem = pm_release_mem_vi, + .map_process_size = sizeof(struct pm4_mes_map_process), + .runlist_size = sizeof(struct pm4_mes_runlist), + .set_resources_size = sizeof(struct pm4_mes_set_resources), + .map_queues_size = sizeof(struct pm4_mes_map_queues), + .unmap_queues_size = sizeof(struct pm4_mes_unmap_queues), + .query_status_size = sizeof(struct pm4_mes_query_status), + .release_mem_size = sizeof(struct pm4_mec_release_mem) +}; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_module.c b/drivers/gpu/drm/amd/amdkfd/kfd_module.c index e0c07d24d2517..76bf2dc8aec43 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_module.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_module.c @@ -43,6 +43,8 @@ static const struct kgd2kfd_calls kgd2kfd = { .interrupt = kgd2kfd_interrupt, .suspend = kgd2kfd_suspend, .resume = kgd2kfd_resume, + .quiesce_mm = kgd2kfd_quiesce_mm, + .resume_mm = kgd2kfd_resume_mm, .schedule_evict_and_restore_process = kgd2kfd_schedule_evict_and_restore_process, }; @@ -81,6 +83,11 @@ module_param(ignore_crat, int, 0444); MODULE_PARM_DESC(ignore_crat, "Ignore CRAT table during KFD initialization (0 = use CRAT (default), 1 = ignore CRAT)"); +int vega10_noretry; +module_param_named(noretry, vega10_noretry, int, 0644); +MODULE_PARM_DESC(noretry, + "Set sh_mem_config.retry_disable on Vega10 (0 = retry enabled (default), 1 = retry disabled)"); + static int amdkfd_init_completed; int kgd2kfd_init(unsigned int interface_version, diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c index ee7061e1c466f..4b8eb506642b8 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager.c @@ -38,6 +38,9 @@ struct mqd_manager *mqd_manager_init(enum KFD_MQD_TYPE type, case CHIP_POLARIS10: case CHIP_POLARIS11: return mqd_manager_init_vi_tonga(type, dev); + case CHIP_VEGA10: + case CHIP_RAVEN: + return mqd_manager_init_v9(type, dev); default: WARN(1, "Unexpected ASIC family %u", dev->device_info->asic_family); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c index c00c325ed3c97..06eaa218eba65 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_cik.c @@ -79,10 +79,6 @@ static int init_mqd(struct mqd_manager *mm, void **mqd, m->cp_mqd_base_addr_lo = lower_32_bits(addr); m->cp_mqd_base_addr_hi = upper_32_bits(addr); - m->cp_hqd_ib_control = DEFAULT_MIN_IB_AVAIL_SIZE | IB_ATC_EN; - /* Although WinKFD writes this, I suspect it should not be necessary */ - m->cp_hqd_ib_control = IB_ATC_EN | DEFAULT_MIN_IB_AVAIL_SIZE; - m->cp_hqd_quantum = QUANTUM_EN | QUANTUM_SCALE_1MS | QUANTUM_DURATION(10); @@ -412,7 +408,7 @@ struct mqd_manager *mqd_manager_init_cik(enum KFD_MQD_TYPE type, if (WARN_ON(type >= KFD_MQD_TYPE_MAX)) return NULL; - mqd = kzalloc(sizeof(*mqd), GFP_KERNEL); + mqd = kzalloc(sizeof(*mqd), GFP_NOIO); if (!mqd) return NULL; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c new file mode 100644 index 0000000000000..684054ff02cdc --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_v9.c @@ -0,0 +1,443 @@ +/* + * Copyright 2016-2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#include <linux/printk.h> +#include <linux/slab.h> +#include <linux/uaccess.h> +#include "kfd_priv.h" +#include "kfd_mqd_manager.h" +#include "v9_structs.h" +#include "gc/gc_9_0_offset.h" +#include "gc/gc_9_0_sh_mask.h" +#include "sdma0/sdma0_4_0_sh_mask.h" + +static inline struct v9_mqd *get_mqd(void *mqd) +{ + return (struct v9_mqd *)mqd; +} + +static inline struct v9_sdma_mqd *get_sdma_mqd(void *mqd) +{ + return (struct v9_sdma_mqd *)mqd; +} + +static int init_mqd(struct mqd_manager *mm, void **mqd, + struct kfd_mem_obj **mqd_mem_obj, uint64_t *gart_addr, + struct queue_properties *q) +{ + int retval; + uint64_t addr; + struct v9_mqd *m; + struct kfd_dev *kfd = mm->dev; + + /* From V9, for CWSR, the control stack is located on the next page + * boundary after the mqd, we will use the gtt allocation function + * instead of sub-allocation function. + */ + if (kfd->cwsr_enabled && (q->type == KFD_QUEUE_TYPE_COMPUTE)) { + *mqd_mem_obj = kzalloc(sizeof(struct kfd_mem_obj), GFP_NOIO); + if (!*mqd_mem_obj) + return -ENOMEM; + retval = kfd->kfd2kgd->init_gtt_mem_allocation(kfd->kgd, + ALIGN(q->ctl_stack_size, PAGE_SIZE) + + ALIGN(sizeof(struct v9_mqd), PAGE_SIZE), + &((*mqd_mem_obj)->gtt_mem), + &((*mqd_mem_obj)->gpu_addr), + (void *)&((*mqd_mem_obj)->cpu_ptr)); + } else + retval = kfd_gtt_sa_allocate(mm->dev, sizeof(struct v9_mqd), + mqd_mem_obj); + if (retval != 0) + return -ENOMEM; + + m = (struct v9_mqd *) (*mqd_mem_obj)->cpu_ptr; + addr = (*mqd_mem_obj)->gpu_addr; + + memset(m, 0, sizeof(struct v9_mqd)); + + m->header = 0xC0310800; + m->compute_pipelinestat_enable = 1; + m->compute_static_thread_mgmt_se0 = 0xFFFFFFFF; + m->compute_static_thread_mgmt_se1 = 0xFFFFFFFF; + m->compute_static_thread_mgmt_se2 = 0xFFFFFFFF; + m->compute_static_thread_mgmt_se3 = 0xFFFFFFFF; + + m->cp_hqd_persistent_state = CP_HQD_PERSISTENT_STATE__PRELOAD_REQ_MASK | + 0x53 << CP_HQD_PERSISTENT_STATE__PRELOAD_SIZE__SHIFT; + + m->cp_mqd_control = 1 << CP_MQD_CONTROL__PRIV_STATE__SHIFT; + + m->cp_mqd_base_addr_lo = lower_32_bits(addr); + m->cp_mqd_base_addr_hi = upper_32_bits(addr); + + m->cp_hqd_quantum = 1 << CP_HQD_QUANTUM__QUANTUM_EN__SHIFT | + 1 << CP_HQD_QUANTUM__QUANTUM_SCALE__SHIFT | + 10 << CP_HQD_QUANTUM__QUANTUM_DURATION__SHIFT; + + m->cp_hqd_pipe_priority = 1; + m->cp_hqd_queue_priority = 15; + + if (q->format == KFD_QUEUE_FORMAT_AQL) { + m->cp_hqd_aql_control = + 1 << CP_HQD_AQL_CONTROL__CONTROL0__SHIFT; + } + + if (q->tba_addr) { + m->compute_pgm_rsrc2 |= + (1 << COMPUTE_PGM_RSRC2__TRAP_PRESENT__SHIFT); + } + + if (mm->dev->cwsr_enabled && q->ctx_save_restore_area_address) { + m->cp_hqd_persistent_state |= + (1 << CP_HQD_PERSISTENT_STATE__QSWITCH_MODE__SHIFT); + m->cp_hqd_ctx_save_base_addr_lo = + lower_32_bits(q->ctx_save_restore_area_address); + m->cp_hqd_ctx_save_base_addr_hi = + upper_32_bits(q->ctx_save_restore_area_address); + m->cp_hqd_ctx_save_size = q->ctx_save_restore_area_size; + m->cp_hqd_cntl_stack_size = q->ctl_stack_size; + m->cp_hqd_cntl_stack_offset = q->ctl_stack_size; + m->cp_hqd_wg_state_offset = q->ctl_stack_size; + } + + *mqd = m; + if (gart_addr) + *gart_addr = addr; + retval = mm->update_mqd(mm, m, q); + + return retval; +} + +static int load_mqd(struct mqd_manager *mm, void *mqd, + uint32_t pipe_id, uint32_t queue_id, + struct queue_properties *p, struct mm_struct *mms) +{ + /* AQL write pointer counts in 64B packets, PM4/CP counts in dwords. */ + uint32_t wptr_shift = (p->format == KFD_QUEUE_FORMAT_AQL ? 4 : 0); + + return mm->dev->kfd2kgd->hqd_load(mm->dev->kgd, mqd, pipe_id, queue_id, + (uint32_t __user *)p->write_ptr, + wptr_shift, 0, mms); +} + +static int update_mqd(struct mqd_manager *mm, void *mqd, + struct queue_properties *q) +{ + struct v9_mqd *m; + + m = get_mqd(mqd); + + m->cp_hqd_pq_control = 5 << CP_HQD_PQ_CONTROL__RPTR_BLOCK_SIZE__SHIFT; + m->cp_hqd_pq_control |= order_base_2(q->queue_size / 4) - 1; + pr_debug("cp_hqd_pq_control 0x%x\n", m->cp_hqd_pq_control); + + m->cp_hqd_pq_base_lo = lower_32_bits((uint64_t)q->queue_address >> 8); + m->cp_hqd_pq_base_hi = upper_32_bits((uint64_t)q->queue_address >> 8); + + m->cp_hqd_pq_rptr_report_addr_lo = lower_32_bits((uint64_t)q->read_ptr); + m->cp_hqd_pq_rptr_report_addr_hi = upper_32_bits((uint64_t)q->read_ptr); + m->cp_hqd_pq_wptr_poll_addr_lo = lower_32_bits((uint64_t)q->write_ptr); + m->cp_hqd_pq_wptr_poll_addr_hi = upper_32_bits((uint64_t)q->write_ptr); + + m->cp_hqd_pq_doorbell_control = + q->doorbell_off << + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_OFFSET__SHIFT; + pr_debug("cp_hqd_pq_doorbell_control 0x%x\n", + m->cp_hqd_pq_doorbell_control); + + m->cp_hqd_ib_control = + 3 << CP_HQD_IB_CONTROL__MIN_IB_AVAIL_SIZE__SHIFT | + 1 << CP_HQD_IB_CONTROL__IB_EXE_DISABLE__SHIFT; + + /* + * HW does not clamp this field correctly. Maximum EOP queue size + * is constrained by per-SE EOP done signal count, which is 8-bit. + * Limit is 0xFF EOP entries (= 0x7F8 dwords). CP will not submit + * more than (EOP entry count - 1) so a queue size of 0x800 dwords + * is safe, giving a maximum field value of 0xA. + */ + m->cp_hqd_eop_control = min(0xA, + order_base_2(q->eop_ring_buffer_size / 4) - 1); + m->cp_hqd_eop_base_addr_lo = + lower_32_bits(q->eop_ring_buffer_address >> 8); + m->cp_hqd_eop_base_addr_hi = + upper_32_bits(q->eop_ring_buffer_address >> 8); + + m->cp_hqd_iq_timer = 0; + + m->cp_hqd_vmid = q->vmid; + + if (q->format == KFD_QUEUE_FORMAT_AQL) { + m->cp_hqd_pq_control |= CP_HQD_PQ_CONTROL__NO_UPDATE_RPTR_MASK | + 2 << CP_HQD_PQ_CONTROL__SLOT_BASED_WPTR__SHIFT | + 1 << CP_HQD_PQ_CONTROL__QUEUE_FULL_EN__SHIFT | + 1 << CP_HQD_PQ_CONTROL__WPP_CLAMP_EN__SHIFT; + m->cp_hqd_pq_doorbell_control |= 1 << + CP_HQD_PQ_DOORBELL_CONTROL__DOORBELL_BIF_DROP__SHIFT; + } + if (mm->dev->cwsr_enabled && q->ctx_save_restore_area_address) + m->cp_hqd_ctx_save_control = 0; + + q->is_active = (q->queue_size > 0 && + q->queue_address != 0 && + q->queue_percent > 0 && + !q->is_evicted); + + return 0; +} + + +static int destroy_mqd(struct mqd_manager *mm, void *mqd, + enum kfd_preempt_type type, + unsigned int timeout, uint32_t pipe_id, + uint32_t queue_id) +{ + return mm->dev->kfd2kgd->hqd_destroy + (mm->dev->kgd, mqd, type, timeout, + pipe_id, queue_id); +} + +static void uninit_mqd(struct mqd_manager *mm, void *mqd, + struct kfd_mem_obj *mqd_mem_obj) +{ + struct kfd_dev *kfd = mm->dev; + + if (mqd_mem_obj->gtt_mem) { + kfd->kfd2kgd->free_gtt_mem(kfd->kgd, mqd_mem_obj->gtt_mem); + kfree(mqd_mem_obj); + } else { + kfd_gtt_sa_free(mm->dev, mqd_mem_obj); + } +} + +static bool is_occupied(struct mqd_manager *mm, void *mqd, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id) +{ + return mm->dev->kfd2kgd->hqd_is_occupied( + mm->dev->kgd, queue_address, + pipe_id, queue_id); +} + +static int init_mqd_hiq(struct mqd_manager *mm, void **mqd, + struct kfd_mem_obj **mqd_mem_obj, uint64_t *gart_addr, + struct queue_properties *q) +{ + struct v9_mqd *m; + int retval = init_mqd(mm, mqd, mqd_mem_obj, gart_addr, q); + + if (retval != 0) + return retval; + + m = get_mqd(*mqd); + + m->cp_hqd_pq_control |= 1 << CP_HQD_PQ_CONTROL__PRIV_STATE__SHIFT | + 1 << CP_HQD_PQ_CONTROL__KMD_QUEUE__SHIFT; + + return retval; +} + +static int update_mqd_hiq(struct mqd_manager *mm, void *mqd, + struct queue_properties *q) +{ + struct v9_mqd *m; + int retval = update_mqd(mm, mqd, q); + + if (retval != 0) + return retval; + + /* TODO: what's the point? update_mqd already does this. */ + m = get_mqd(mqd); + m->cp_hqd_vmid = q->vmid; + return retval; +} + +static int init_mqd_sdma(struct mqd_manager *mm, void **mqd, + struct kfd_mem_obj **mqd_mem_obj, uint64_t *gart_addr, + struct queue_properties *q) +{ + int retval; + struct v9_sdma_mqd *m; + + + retval = kfd_gtt_sa_allocate(mm->dev, + sizeof(struct v9_sdma_mqd), + mqd_mem_obj); + + if (retval != 0) + return -ENOMEM; + + m = (struct v9_sdma_mqd *) (*mqd_mem_obj)->cpu_ptr; + + memset(m, 0, sizeof(struct v9_sdma_mqd)); + + *mqd = m; + if (gart_addr) + *gart_addr = (*mqd_mem_obj)->gpu_addr; + + retval = mm->update_mqd(mm, m, q); + + return retval; +} + +static void uninit_mqd_sdma(struct mqd_manager *mm, void *mqd, + struct kfd_mem_obj *mqd_mem_obj) +{ + kfd_gtt_sa_free(mm->dev, mqd_mem_obj); +} + +static int load_mqd_sdma(struct mqd_manager *mm, void *mqd, + uint32_t pipe_id, uint32_t queue_id, + struct queue_properties *p, struct mm_struct *mms) +{ + return mm->dev->kfd2kgd->hqd_sdma_load(mm->dev->kgd, mqd, + (uint32_t __user *)p->write_ptr, + mms); +} + +#define SDMA_RLC_DUMMY_DEFAULT 0xf + +static int update_mqd_sdma(struct mqd_manager *mm, void *mqd, + struct queue_properties *q) +{ + struct v9_sdma_mqd *m; + + m = get_sdma_mqd(mqd); + m->sdmax_rlcx_rb_cntl = order_base_2(q->queue_size / 4) + << SDMA0_RLC0_RB_CNTL__RB_SIZE__SHIFT | + q->vmid << SDMA0_RLC0_RB_CNTL__RB_VMID__SHIFT | + 1 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_ENABLE__SHIFT | + 6 << SDMA0_RLC0_RB_CNTL__RPTR_WRITEBACK_TIMER__SHIFT; + + m->sdmax_rlcx_rb_base = lower_32_bits(q->queue_address >> 8); + m->sdmax_rlcx_rb_base_hi = upper_32_bits(q->queue_address >> 8); + m->sdmax_rlcx_rb_rptr_addr_lo = lower_32_bits((uint64_t)q->read_ptr); + m->sdmax_rlcx_rb_rptr_addr_hi = upper_32_bits((uint64_t)q->read_ptr); + m->sdmax_rlcx_doorbell_offset = + q->doorbell_off << SDMA0_RLC0_DOORBELL_OFFSET__OFFSET__SHIFT; + + m->sdma_engine_id = q->sdma_engine_id; + m->sdma_queue_id = q->sdma_queue_id; + m->sdmax_rlcx_dummy_reg = SDMA_RLC_DUMMY_DEFAULT; + + q->is_active = (q->queue_size > 0 && + q->queue_address != 0 && + q->queue_percent > 0 && + !q->is_evicted); + + return 0; +} + +/* + * * preempt type here is ignored because there is only one way + * * to preempt sdma queue + */ +static int destroy_mqd_sdma(struct mqd_manager *mm, void *mqd, + enum kfd_preempt_type type, + unsigned int timeout, uint32_t pipe_id, + uint32_t queue_id) +{ + return mm->dev->kfd2kgd->hqd_sdma_destroy(mm->dev->kgd, mqd, timeout); +} + +static bool is_occupied_sdma(struct mqd_manager *mm, void *mqd, + uint64_t queue_address, uint32_t pipe_id, + uint32_t queue_id) +{ + return mm->dev->kfd2kgd->hqd_sdma_is_occupied(mm->dev->kgd, mqd); +} + +#if defined(CONFIG_DEBUG_FS) + +static int debugfs_show_mqd(struct seq_file *m, void *data) +{ + seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4, + data, sizeof(struct v9_mqd), false); + return 0; +} + +static int debugfs_show_mqd_sdma(struct seq_file *m, void *data) +{ + seq_hex_dump(m, " ", DUMP_PREFIX_OFFSET, 32, 4, + data, sizeof(struct v9_sdma_mqd), false); + return 0; +} + +#endif + +struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type, + struct kfd_dev *dev) +{ + struct mqd_manager *mqd; + + if (WARN_ON(type >= KFD_MQD_TYPE_MAX)) + return NULL; + + mqd = kzalloc(sizeof(*mqd), GFP_NOIO); + if (!mqd) + return NULL; + + mqd->dev = dev; + + switch (type) { + case KFD_MQD_TYPE_CP: + case KFD_MQD_TYPE_COMPUTE: + mqd->init_mqd = init_mqd; + mqd->uninit_mqd = uninit_mqd; + mqd->load_mqd = load_mqd; + mqd->update_mqd = update_mqd; + mqd->destroy_mqd = destroy_mqd; + mqd->is_occupied = is_occupied; +#if defined(CONFIG_DEBUG_FS) + mqd->debugfs_show_mqd = debugfs_show_mqd; +#endif + break; + case KFD_MQD_TYPE_HIQ: + mqd->init_mqd = init_mqd_hiq; + mqd->uninit_mqd = uninit_mqd; + mqd->load_mqd = load_mqd; + mqd->update_mqd = update_mqd_hiq; + mqd->destroy_mqd = destroy_mqd; + mqd->is_occupied = is_occupied; +#if defined(CONFIG_DEBUG_FS) + mqd->debugfs_show_mqd = debugfs_show_mqd; +#endif + break; + case KFD_MQD_TYPE_SDMA: + mqd->init_mqd = init_mqd_sdma; + mqd->uninit_mqd = uninit_mqd_sdma; + mqd->load_mqd = load_mqd_sdma; + mqd->update_mqd = update_mqd_sdma; + mqd->destroy_mqd = destroy_mqd_sdma; + mqd->is_occupied = is_occupied_sdma; +#if defined(CONFIG_DEBUG_FS) + mqd->debugfs_show_mqd = debugfs_show_mqd_sdma; +#endif + break; + default: + kfree(mqd); + return NULL; + } + + return mqd; +} diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c index 89e4242e43e71..481307b8b4dbd 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_mqd_manager_vi.c @@ -394,7 +394,7 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, if (WARN_ON(type >= KFD_MQD_TYPE_MAX)) return NULL; - mqd = kzalloc(sizeof(*mqd), GFP_KERNEL); + mqd = kzalloc(sizeof(*mqd), GFP_NOIO); if (!mqd) return NULL; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c index 89ba4c670ec58..c317feb43f69d 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_packet_manager.c @@ -26,8 +26,6 @@ #include "kfd_device_queue_manager.h" #include "kfd_kernel_queue.h" #include "kfd_priv.h" -#include "kfd_pm4_headers_vi.h" -#include "kfd_pm4_opcodes.h" static inline void inc_wptr(unsigned int *wptr, unsigned int increment_bytes, unsigned int buffer_size_bytes) @@ -39,18 +37,6 @@ static inline void inc_wptr(unsigned int *wptr, unsigned int increment_bytes, *wptr = temp; } -static unsigned int build_pm4_header(unsigned int opcode, size_t packet_size) -{ - union PM4_MES_TYPE_3_HEADER header; - - header.u32All = 0; - header.opcode = opcode; - header.count = packet_size / 4 - 2; - header.type = PM4_TYPE_3; - - return header.u32All; -} - static void pm_calc_rlib_size(struct packet_manager *pm, unsigned int *rlib_size, bool *over_subscription) @@ -80,9 +66,9 @@ static void pm_calc_rlib_size(struct packet_manager *pm, pr_debug("Over subscribed runlist\n"); } - map_queue_size = sizeof(struct pm4_mes_map_queues); + map_queue_size = pm->pmf->map_queues_size; /* calculate run list ib allocation size */ - *rlib_size = process_count * sizeof(struct pm4_mes_map_process) + + *rlib_size = process_count * pm->pmf->map_process_size + queue_count * map_queue_size; /* @@ -90,7 +76,7 @@ static void pm_calc_rlib_size(struct packet_manager *pm, * when over subscription */ if (*over_subscription) - *rlib_size += sizeof(struct pm4_mes_runlist); + *rlib_size += pm->pmf->runlist_size; pr_debug("runlist ib size %d\n", *rlib_size); } @@ -108,12 +94,14 @@ static int pm_allocate_runlist_ib(struct packet_manager *pm, pm_calc_rlib_size(pm, rl_buffer_size, is_over_subscription); + mutex_lock(&pm->lock); + retval = kfd_gtt_sa_allocate(pm->dqm->dev, *rl_buffer_size, &pm->ib_buffer_obj); if (retval) { pr_err("Failed to allocate runlist IB\n"); - return retval; + goto out; } *(void **)rl_buffer = pm->ib_buffer_obj->cpu_ptr; @@ -121,138 +109,10 @@ static int pm_allocate_runlist_ib(struct packet_manager *pm, memset(*rl_buffer, 0, *rl_buffer_size); pm->allocated = true; - return retval; -} - -static int pm_create_runlist(struct packet_manager *pm, uint32_t *buffer, - uint64_t ib, size_t ib_size_in_dwords, bool chain) -{ - struct pm4_mes_runlist *packet; - int concurrent_proc_cnt = 0; - struct kfd_dev *kfd = pm->dqm->dev; - - if (WARN_ON(!ib)) - return -EFAULT; - - /* Determine the number of processes to map together to HW: - * it can not exceed the number of VMIDs available to the - * scheduler, and it is determined by the smaller of the number - * of processes in the runlist and kfd module parameter - * hws_max_conc_proc. - * Note: the arbitration between the number of VMIDs and - * hws_max_conc_proc has been done in - * kgd2kfd_device_init(). - */ - concurrent_proc_cnt = min(pm->dqm->processes_count, - kfd->max_proc_per_quantum); - - packet = (struct pm4_mes_runlist *)buffer; - - memset(buffer, 0, sizeof(struct pm4_mes_runlist)); - packet->header.u32All = build_pm4_header(IT_RUN_LIST, - sizeof(struct pm4_mes_runlist)); - - packet->bitfields4.ib_size = ib_size_in_dwords; - packet->bitfields4.chain = chain ? 1 : 0; - packet->bitfields4.offload_polling = 0; - packet->bitfields4.valid = 1; - packet->bitfields4.process_cnt = concurrent_proc_cnt; - packet->ordinal2 = lower_32_bits(ib); - packet->bitfields3.ib_base_hi = upper_32_bits(ib); - - return 0; -} - -static int pm_create_map_process(struct packet_manager *pm, uint32_t *buffer, - struct qcm_process_device *qpd) -{ - struct pm4_mes_map_process *packet; - - packet = (struct pm4_mes_map_process *)buffer; - memset(buffer, 0, sizeof(struct pm4_mes_map_process)); - - packet->header.u32All = build_pm4_header(IT_MAP_PROCESS, - sizeof(struct pm4_mes_map_process)); - packet->bitfields2.diq_enable = (qpd->is_debug) ? 1 : 0; - packet->bitfields2.process_quantum = 1; - packet->bitfields2.pasid = qpd->pqm->process->pasid; - packet->bitfields3.page_table_base = qpd->page_table_base; - packet->bitfields10.gds_size = qpd->gds_size; - packet->bitfields10.num_gws = qpd->num_gws; - packet->bitfields10.num_oac = qpd->num_oac; - packet->bitfields10.num_queues = (qpd->is_debug) ? 0 : qpd->queue_count; - - packet->sh_mem_config = qpd->sh_mem_config; - packet->sh_mem_bases = qpd->sh_mem_bases; - packet->sh_mem_ape1_base = qpd->sh_mem_ape1_base; - packet->sh_mem_ape1_limit = qpd->sh_mem_ape1_limit; - - packet->sh_hidden_private_base_vmid = qpd->sh_hidden_private_base; - - packet->gds_addr_lo = lower_32_bits(qpd->gds_context_area); - packet->gds_addr_hi = upper_32_bits(qpd->gds_context_area); - - return 0; -} - -static int pm_create_map_queue(struct packet_manager *pm, uint32_t *buffer, - struct queue *q, bool is_static) -{ - struct pm4_mes_map_queues *packet; - bool use_static = is_static; - - packet = (struct pm4_mes_map_queues *)buffer; - memset(buffer, 0, sizeof(struct pm4_mes_map_queues)); - - packet->header.u32All = build_pm4_header(IT_MAP_QUEUES, - sizeof(struct pm4_mes_map_queues)); - packet->bitfields2.alloc_format = - alloc_format__mes_map_queues__one_per_pipe_vi; - packet->bitfields2.num_queues = 1; - packet->bitfields2.queue_sel = - queue_sel__mes_map_queues__map_to_hws_determined_queue_slots_vi; - - packet->bitfields2.engine_sel = - engine_sel__mes_map_queues__compute_vi; - packet->bitfields2.queue_type = - queue_type__mes_map_queues__normal_compute_vi; - - switch (q->properties.type) { - case KFD_QUEUE_TYPE_COMPUTE: - if (use_static) - packet->bitfields2.queue_type = - queue_type__mes_map_queues__normal_latency_static_queue_vi; - break; - case KFD_QUEUE_TYPE_DIQ: - packet->bitfields2.queue_type = - queue_type__mes_map_queues__debug_interface_queue_vi; - break; - case KFD_QUEUE_TYPE_SDMA: - packet->bitfields2.engine_sel = q->properties.sdma_engine_id + - engine_sel__mes_map_queues__sdma0_vi; - use_static = false; /* no static queues under SDMA */ - break; - default: - WARN(1, "queue type %d", q->properties.type); - return -EINVAL; - } - packet->bitfields3.doorbell_offset = - q->properties.doorbell_off; - - packet->mqd_addr_lo = - lower_32_bits(q->gart_mqd_addr); - - packet->mqd_addr_hi = - upper_32_bits(q->gart_mqd_addr); - - packet->wptr_addr_lo = - lower_32_bits((uint64_t)q->properties.write_ptr); - - packet->wptr_addr_hi = - upper_32_bits((uint64_t)q->properties.write_ptr); - - return 0; +out: + mutex_unlock(&pm->lock); + return retval; } static int pm_create_runlist_ib(struct packet_manager *pm, @@ -292,12 +152,12 @@ static int pm_create_runlist_ib(struct packet_manager *pm, return -ENOMEM; } - retval = pm_create_map_process(pm, &rl_buffer[rl_wptr], qpd); + retval = pm->pmf->map_process(pm, &rl_buffer[rl_wptr], qpd); if (retval) return retval; proccesses_mapped++; - inc_wptr(&rl_wptr, sizeof(struct pm4_mes_map_process), + inc_wptr(&rl_wptr, pm->pmf->map_process_size, alloc_size_bytes); list_for_each_entry(kq, &qpd->priv_queue_list, list) { @@ -307,7 +167,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm, pr_debug("static_queue, mapping kernel q %d, is debug status %d\n", kq->queue->queue, qpd->is_debug); - retval = pm_create_map_queue(pm, + retval = pm->pmf->map_queues(pm, &rl_buffer[rl_wptr], kq->queue, qpd->is_debug); @@ -315,7 +175,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm, return retval; inc_wptr(&rl_wptr, - sizeof(struct pm4_mes_map_queues), + pm->pmf->map_queues_size, alloc_size_bytes); } @@ -326,7 +186,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm, pr_debug("static_queue, mapping user queue %d, is debug status %d\n", q->queue, qpd->is_debug); - retval = pm_create_map_queue(pm, + retval = pm->pmf->map_queues(pm, &rl_buffer[rl_wptr], q, qpd->is_debug); @@ -335,7 +195,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm, return retval; inc_wptr(&rl_wptr, - sizeof(struct pm4_mes_map_queues), + pm->pmf->map_queues_size, alloc_size_bytes); } } @@ -343,7 +203,7 @@ static int pm_create_runlist_ib(struct packet_manager *pm, pr_debug("Finished map process and queues to runlist\n"); if (is_over_subscription) - retval = pm_create_runlist(pm, &rl_buffer[rl_wptr], + retval = pm->pmf->runlist(pm, &rl_buffer[rl_wptr], *rl_gpu_addr, alloc_size_bytes / sizeof(uint32_t), true); @@ -355,45 +215,29 @@ static int pm_create_runlist_ib(struct packet_manager *pm, return retval; } -/* pm_create_release_mem - Create a RELEASE_MEM packet and return the size - * of this packet - * @gpu_addr - GPU address of the packet. It's a virtual address. - * @buffer - buffer to fill up with the packet. It's a CPU kernel pointer - * Return - length of the packet - */ -uint32_t pm_create_release_mem(uint64_t gpu_addr, uint32_t *buffer) -{ - struct pm4_mec_release_mem *packet; - - WARN_ON(!buffer); - - packet = (struct pm4_mec_release_mem *)buffer; - memset(buffer, 0, sizeof(*packet)); - - packet->header.u32All = build_pm4_header(IT_RELEASE_MEM, - sizeof(*packet)); - - packet->bitfields2.event_type = CACHE_FLUSH_AND_INV_TS_EVENT; - packet->bitfields2.event_index = event_index___release_mem__end_of_pipe; - packet->bitfields2.tcl1_action_ena = 1; - packet->bitfields2.tc_action_ena = 1; - packet->bitfields2.cache_policy = cache_policy___release_mem__lru; - packet->bitfields2.atc = 0; - - packet->bitfields3.data_sel = data_sel___release_mem__send_32_bit_low; - packet->bitfields3.int_sel = - int_sel___release_mem__send_interrupt_after_write_confirm; - - packet->bitfields4.address_lo_32b = (gpu_addr & 0xffffffff) >> 2; - packet->address_hi = upper_32_bits(gpu_addr); - - packet->data_lo = 0; - - return sizeof(*packet) / sizeof(unsigned int); -} - int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm) { + switch (dqm->dev->device_info->asic_family) { + case CHIP_KAVERI: + case CHIP_HAWAII: + /* PM4 packet structures on CIK are the same as on VI */ + case CHIP_CARRIZO: + case CHIP_TONGA: + case CHIP_FIJI: + case CHIP_POLARIS10: + case CHIP_POLARIS11: + pm->pmf = &kfd_vi_pm_funcs; + break; + case CHIP_VEGA10: + case CHIP_RAVEN: + pm->pmf = &kfd_v9_pm_funcs; + break; + default: + WARN(1, "Unexpected ASIC family %u", + dqm->dev->device_info->asic_family); + return -EINVAL; + } + pm->dqm = dqm; mutex_init(&pm->lock); pm->priv_queue = kernel_queue_init(dqm->dev, KFD_QUEUE_TYPE_HIQ); @@ -415,38 +259,25 @@ void pm_uninit(struct packet_manager *pm) int pm_send_set_resources(struct packet_manager *pm, struct scheduling_resources *res) { - struct pm4_mes_set_resources *packet; + uint32_t *buffer, size; int retval = 0; + size = pm->pmf->set_resources_size; mutex_lock(&pm->lock); pm->priv_queue->ops.acquire_packet_buffer(pm->priv_queue, - sizeof(*packet) / sizeof(uint32_t), - (unsigned int **)&packet); - if (!packet) { + size / sizeof(uint32_t), + (unsigned int **)&buffer); + if (!buffer) { pr_err("Failed to allocate buffer on kernel queue\n"); retval = -ENOMEM; goto out; } - memset(packet, 0, sizeof(struct pm4_mes_set_resources)); - packet->header.u32All = build_pm4_header(IT_SET_RESOURCES, - sizeof(struct pm4_mes_set_resources)); - - packet->bitfields2.queue_type = - queue_type__mes_set_resources__hsa_interface_queue_hiq; - packet->bitfields2.vmid_mask = res->vmid_mask; - packet->bitfields2.unmap_latency = KFD_UNMAP_LATENCY_MS / 100; - packet->bitfields7.oac_mask = res->oac_mask; - packet->bitfields8.gds_heap_base = res->gds_heap_base; - packet->bitfields8.gds_heap_size = res->gds_heap_size; - - packet->gws_mask_lo = lower_32_bits(res->gws_mask); - packet->gws_mask_hi = upper_32_bits(res->gws_mask); - - packet->queue_mask_lo = lower_32_bits(res->queue_mask); - packet->queue_mask_hi = upper_32_bits(res->queue_mask); - - pm->priv_queue->ops.submit_packet(pm->priv_queue); + retval = pm->pmf->set_resources(pm, buffer, res); + if (!retval) + pm->priv_queue->ops.submit_packet(pm->priv_queue); + else + pm->priv_queue->ops.rollback_packet(pm->priv_queue); out: mutex_unlock(&pm->lock); @@ -468,7 +299,7 @@ int pm_send_runlist(struct packet_manager *pm, struct list_head *dqm_queues) pr_debug("runlist IB address: 0x%llX\n", rl_gpu_ib_addr); - packet_size_dwords = sizeof(struct pm4_mes_runlist) / sizeof(uint32_t); + packet_size_dwords = pm->pmf->runlist_size / sizeof(uint32_t); mutex_lock(&pm->lock); retval = pm->priv_queue->ops.acquire_packet_buffer(pm->priv_queue, @@ -476,7 +307,7 @@ int pm_send_runlist(struct packet_manager *pm, struct list_head *dqm_queues) if (retval) goto fail_acquire_packet_buffer; - retval = pm_create_runlist(pm, rl_buffer, rl_gpu_ib_addr, + retval = pm->pmf->runlist(pm, rl_buffer, rl_gpu_ib_addr, rl_ib_size / sizeof(uint32_t), false); if (retval) goto fail_create_runlist; @@ -499,37 +330,29 @@ fail_create_runlist_ib: int pm_send_query_status(struct packet_manager *pm, uint64_t fence_address, uint32_t fence_value) { - int retval; - struct pm4_mes_query_status *packet; + uint32_t *buffer, size; + int retval = 0; if (WARN_ON(!fence_address)) return -EFAULT; + size = pm->pmf->query_status_size; mutex_lock(&pm->lock); - retval = pm->priv_queue->ops.acquire_packet_buffer( - pm->priv_queue, - sizeof(struct pm4_mes_query_status) / sizeof(uint32_t), - (unsigned int **)&packet); - if (retval) - goto fail_acquire_packet_buffer; - - packet->header.u32All = build_pm4_header(IT_QUERY_STATUS, - sizeof(struct pm4_mes_query_status)); - - packet->bitfields2.context_id = 0; - packet->bitfields2.interrupt_sel = - interrupt_sel__mes_query_status__completion_status; - packet->bitfields2.command = - command__mes_query_status__fence_only_after_write_ack; - - packet->addr_hi = upper_32_bits((uint64_t)fence_address); - packet->addr_lo = lower_32_bits((uint64_t)fence_address); - packet->data_hi = upper_32_bits((uint64_t)fence_value); - packet->data_lo = lower_32_bits((uint64_t)fence_value); + pm->priv_queue->ops.acquire_packet_buffer(pm->priv_queue, + size / sizeof(uint32_t), (unsigned int **)&buffer); + if (!buffer) { + pr_err("Failed to allocate buffer on kernel queue\n"); + retval = -ENOMEM; + goto out; + } - pm->priv_queue->ops.submit_packet(pm->priv_queue); + retval = pm->pmf->query_status(pm, buffer, fence_address, fence_value); + if (!retval) + pm->priv_queue->ops.submit_packet(pm->priv_queue); + else + pm->priv_queue->ops.rollback_packet(pm->priv_queue); -fail_acquire_packet_buffer: +out: mutex_unlock(&pm->lock); return retval; } @@ -539,82 +362,27 @@ int pm_send_unmap_queue(struct packet_manager *pm, enum kfd_queue_type type, uint32_t filter_param, bool reset, unsigned int sdma_engine) { - int retval; - uint32_t *buffer; - struct pm4_mes_unmap_queues *packet; + uint32_t *buffer, size; + int retval = 0; + size = pm->pmf->unmap_queues_size; mutex_lock(&pm->lock); - retval = pm->priv_queue->ops.acquire_packet_buffer( - pm->priv_queue, - sizeof(struct pm4_mes_unmap_queues) / sizeof(uint32_t), - &buffer); - if (retval) - goto err_acquire_packet_buffer; - - packet = (struct pm4_mes_unmap_queues *)buffer; - memset(buffer, 0, sizeof(struct pm4_mes_unmap_queues)); - pr_debug("static_queue: unmapping queues: filter is %d , reset is %d , type is %d\n", - filter, reset, type); - packet->header.u32All = build_pm4_header(IT_UNMAP_QUEUES, - sizeof(struct pm4_mes_unmap_queues)); - switch (type) { - case KFD_QUEUE_TYPE_COMPUTE: - case KFD_QUEUE_TYPE_DIQ: - packet->bitfields2.engine_sel = - engine_sel__mes_unmap_queues__compute; - break; - case KFD_QUEUE_TYPE_SDMA: - packet->bitfields2.engine_sel = - engine_sel__mes_unmap_queues__sdma0 + sdma_engine; - break; - default: - WARN(1, "queue type %d", type); - retval = -EINVAL; - goto err_invalid; + pm->priv_queue->ops.acquire_packet_buffer(pm->priv_queue, + size / sizeof(uint32_t), (unsigned int **)&buffer); + if (!buffer) { + pr_err("Failed to allocate buffer on kernel queue\n"); + retval = -ENOMEM; + goto out; } - if (reset) - packet->bitfields2.action = - action__mes_unmap_queues__reset_queues; + retval = pm->pmf->unmap_queues(pm, buffer, type, filter, filter_param, + reset, sdma_engine); + if (!retval) + pm->priv_queue->ops.submit_packet(pm->priv_queue); else - packet->bitfields2.action = - action__mes_unmap_queues__preempt_queues; - - switch (filter) { - case KFD_UNMAP_QUEUES_FILTER_SINGLE_QUEUE: - packet->bitfields2.queue_sel = - queue_sel__mes_unmap_queues__perform_request_on_specified_queues; - packet->bitfields2.num_queues = 1; - packet->bitfields3b.doorbell_offset0 = filter_param; - break; - case KFD_UNMAP_QUEUES_FILTER_BY_PASID: - packet->bitfields2.queue_sel = - queue_sel__mes_unmap_queues__perform_request_on_pasid_queues; - packet->bitfields3a.pasid = filter_param; - break; - case KFD_UNMAP_QUEUES_FILTER_ALL_QUEUES: - packet->bitfields2.queue_sel = - queue_sel__mes_unmap_queues__unmap_all_queues; - break; - case KFD_UNMAP_QUEUES_FILTER_DYNAMIC_QUEUES: - /* in this case, we do not preempt static queues */ - packet->bitfields2.queue_sel = - queue_sel__mes_unmap_queues__unmap_all_non_static_queues; - break; - default: - WARN(1, "filter %d", filter); - retval = -EINVAL; - goto err_invalid; - } + pm->priv_queue->ops.rollback_packet(pm->priv_queue); - pm->priv_queue->ops.submit_packet(pm->priv_queue); - - mutex_unlock(&pm->lock); - return 0; - -err_invalid: - pm->priv_queue->ops.rollback_packet(pm->priv_queue); -err_acquire_packet_buffer: +out: mutex_unlock(&pm->lock); return retval; } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h new file mode 100644 index 0000000000000..f2bcf5c092ea2 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/kfd_pm4_headers_ai.h @@ -0,0 +1,583 @@ +/* + * Copyright 2016 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef F32_MES_PM4_PACKETS_H +#define F32_MES_PM4_PACKETS_H + +#ifndef PM4_MES_HEADER_DEFINED +#define PM4_MES_HEADER_DEFINED +union PM4_MES_TYPE_3_HEADER { + struct { + uint32_t reserved1 : 8; /* < reserved */ + uint32_t opcode : 8; /* < IT opcode */ + uint32_t count : 14;/* < number of DWORDs - 1 in the + * information body. + */ + uint32_t type : 2; /* < packet identifier. + * It should be 3 for type 3 packets + */ + }; + uint32_t u32All; +}; +#endif /* PM4_MES_HEADER_DEFINED */ + +/*--------------------MES_SET_RESOURCES--------------------*/ + +#ifndef PM4_MES_SET_RESOURCES_DEFINED +#define PM4_MES_SET_RESOURCES_DEFINED +enum mes_set_resources_queue_type_enum { + queue_type__mes_set_resources__kernel_interface_queue_kiq = 0, + queue_type__mes_set_resources__hsa_interface_queue_hiq = 1, + queue_type__mes_set_resources__hsa_debug_interface_queue = 4 +}; + + +struct pm4_mes_set_resources { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t vmid_mask:16; + uint32_t unmap_latency:8; + uint32_t reserved1:5; + enum mes_set_resources_queue_type_enum queue_type:3; + } bitfields2; + uint32_t ordinal2; + }; + + uint32_t queue_mask_lo; + uint32_t queue_mask_hi; + uint32_t gws_mask_lo; + uint32_t gws_mask_hi; + + union { + struct { + uint32_t oac_mask:16; + uint32_t reserved2:16; + } bitfields7; + uint32_t ordinal7; + }; + + union { + struct { + uint32_t gds_heap_base:6; + uint32_t reserved3:5; + uint32_t gds_heap_size:6; + uint32_t reserved4:15; + } bitfields8; + uint32_t ordinal8; + }; + +}; +#endif + +/*--------------------MES_RUN_LIST--------------------*/ + +#ifndef PM4_MES_RUN_LIST_DEFINED +#define PM4_MES_RUN_LIST_DEFINED + +struct pm4_mes_runlist { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t reserved1:2; + uint32_t ib_base_lo:30; + } bitfields2; + uint32_t ordinal2; + }; + + uint32_t ib_base_hi; + + union { + struct { + uint32_t ib_size:20; + uint32_t chain:1; + uint32_t offload_polling:1; + uint32_t reserved2:1; + uint32_t valid:1; + uint32_t process_cnt:4; + uint32_t reserved3:4; + } bitfields4; + uint32_t ordinal4; + }; + +}; +#endif + +/*--------------------MES_MAP_PROCESS--------------------*/ + +#ifndef PM4_MES_MAP_PROCESS_DEFINED +#define PM4_MES_MAP_PROCESS_DEFINED + +struct pm4_mes_map_process { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t pasid:16; + uint32_t reserved1:8; + uint32_t diq_enable:1; + uint32_t process_quantum:7; + } bitfields2; + uint32_t ordinal2; + }; + + uint32_t vm_context_page_table_base_addr_lo32; + + uint32_t vm_context_page_table_base_addr_hi32; + + uint32_t sh_mem_bases; + + uint32_t sh_mem_config; + + uint32_t sq_shader_tba_lo; + + uint32_t sq_shader_tba_hi; + + uint32_t sq_shader_tma_lo; + + uint32_t sq_shader_tma_hi; + + uint32_t reserved6; + + uint32_t gds_addr_lo; + + uint32_t gds_addr_hi; + + union { + struct { + uint32_t num_gws:6; + uint32_t reserved7:1; + uint32_t sdma_enable:1; + uint32_t num_oac:4; + uint32_t reserved8:4; + uint32_t gds_size:6; + uint32_t num_queues:10; + } bitfields14; + uint32_t ordinal14; + }; + + uint32_t completion_signal_lo; + + uint32_t completion_signal_hi; + +}; + +#endif + +/*--------------------MES_MAP_PROCESS_VM--------------------*/ + +#ifndef PM4_MES_MAP_PROCESS_VM_DEFINED +#define PM4_MES_MAP_PROCESS_VM_DEFINED + +struct PM4_MES_MAP_PROCESS_VM { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + uint32_t reserved1; + + uint32_t vm_context_cntl; + + uint32_t reserved2; + + uint32_t vm_context_page_table_end_addr_lo32; + + uint32_t vm_context_page_table_end_addr_hi32; + + uint32_t vm_context_page_table_start_addr_lo32; + + uint32_t vm_context_page_table_start_addr_hi32; + + uint32_t reserved3; + + uint32_t reserved4; + + uint32_t reserved5; + + uint32_t reserved6; + + uint32_t reserved7; + + uint32_t reserved8; + + uint32_t completion_signal_lo32; + + uint32_t completion_signal_hi32; + +}; +#endif + +/*--------------------MES_MAP_QUEUES--------------------*/ + +#ifndef PM4_MES_MAP_QUEUES_VI_DEFINED +#define PM4_MES_MAP_QUEUES_VI_DEFINED +enum mes_map_queues_queue_sel_enum { + queue_sel__mes_map_queues__map_to_specified_queue_slots_vi = 0, +queue_sel__mes_map_queues__map_to_hws_determined_queue_slots_vi = 1 +}; + +enum mes_map_queues_queue_type_enum { + queue_type__mes_map_queues__normal_compute_vi = 0, + queue_type__mes_map_queues__debug_interface_queue_vi = 1, + queue_type__mes_map_queues__normal_latency_static_queue_vi = 2, +queue_type__mes_map_queues__low_latency_static_queue_vi = 3 +}; + +enum mes_map_queues_alloc_format_enum { + alloc_format__mes_map_queues__one_per_pipe_vi = 0, +alloc_format__mes_map_queues__all_on_one_pipe_vi = 1 +}; + +enum mes_map_queues_engine_sel_enum { + engine_sel__mes_map_queues__compute_vi = 0, + engine_sel__mes_map_queues__sdma0_vi = 2, + engine_sel__mes_map_queues__sdma1_vi = 3 +}; + + +struct pm4_mes_map_queues { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t reserved1:4; + enum mes_map_queues_queue_sel_enum queue_sel:2; + uint32_t reserved2:15; + enum mes_map_queues_queue_type_enum queue_type:3; + enum mes_map_queues_alloc_format_enum alloc_format:2; + enum mes_map_queues_engine_sel_enum engine_sel:3; + uint32_t num_queues:3; + } bitfields2; + uint32_t ordinal2; + }; + + union { + struct { + uint32_t reserved3:1; + uint32_t check_disable:1; + uint32_t doorbell_offset:26; + uint32_t reserved4:4; + } bitfields3; + uint32_t ordinal3; + }; + + uint32_t mqd_addr_lo; + uint32_t mqd_addr_hi; + uint32_t wptr_addr_lo; + uint32_t wptr_addr_hi; +}; +#endif + +/*--------------------MES_QUERY_STATUS--------------------*/ + +#ifndef PM4_MES_QUERY_STATUS_DEFINED +#define PM4_MES_QUERY_STATUS_DEFINED +enum mes_query_status_interrupt_sel_enum { + interrupt_sel__mes_query_status__completion_status = 0, + interrupt_sel__mes_query_status__process_status = 1, + interrupt_sel__mes_query_status__queue_status = 2 +}; + +enum mes_query_status_command_enum { + command__mes_query_status__interrupt_only = 0, + command__mes_query_status__fence_only_immediate = 1, + command__mes_query_status__fence_only_after_write_ack = 2, + command__mes_query_status__fence_wait_for_write_ack_send_interrupt = 3 +}; + +enum mes_query_status_engine_sel_enum { + engine_sel__mes_query_status__compute = 0, + engine_sel__mes_query_status__sdma0_queue = 2, + engine_sel__mes_query_status__sdma1_queue = 3 +}; + +struct pm4_mes_query_status { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + uint32_t context_id:28; + enum mes_query_status_interrupt_sel_enum interrupt_sel:2; + enum mes_query_status_command_enum command:2; + } bitfields2; + uint32_t ordinal2; + }; + + union { + struct { + uint32_t pasid:16; + uint32_t reserved1:16; + } bitfields3a; + struct { + uint32_t reserved2:2; + uint32_t doorbell_offset:26; + enum mes_query_status_engine_sel_enum engine_sel:3; + uint32_t reserved3:1; + } bitfields3b; + uint32_t ordinal3; + }; + + uint32_t addr_lo; + uint32_t addr_hi; + uint32_t data_lo; + uint32_t data_hi; +}; +#endif + +/*--------------------MES_UNMAP_QUEUES--------------------*/ + +#ifndef PM4_MES_UNMAP_QUEUES_DEFINED +#define PM4_MES_UNMAP_QUEUES_DEFINED +enum mes_unmap_queues_action_enum { + action__mes_unmap_queues__preempt_queues = 0, + action__mes_unmap_queues__reset_queues = 1, + action__mes_unmap_queues__disable_process_queues = 2, + action__mes_unmap_queues__reserved = 3 +}; + +enum mes_unmap_queues_queue_sel_enum { + queue_sel__mes_unmap_queues__perform_request_on_specified_queues = 0, + queue_sel__mes_unmap_queues__perform_request_on_pasid_queues = 1, + queue_sel__mes_unmap_queues__unmap_all_queues = 2, + queue_sel__mes_unmap_queues__unmap_all_non_static_queues = 3 +}; + +enum mes_unmap_queues_engine_sel_enum { + engine_sel__mes_unmap_queues__compute = 0, + engine_sel__mes_unmap_queues__sdma0 = 2, + engine_sel__mes_unmap_queues__sdmal = 3 +}; + +struct pm4_mes_unmap_queues { + union { + union PM4_MES_TYPE_3_HEADER header; /* header */ + uint32_t ordinal1; + }; + + union { + struct { + enum mes_unmap_queues_action_enum action:2; + uint32_t reserved1:2; + enum mes_unmap_queues_queue_sel_enum queue_sel:2; + uint32_t reserved2:20; + enum mes_unmap_queues_engine_sel_enum engine_sel:3; + uint32_t num_queues:3; + } bitfields2; + uint32_t ordinal2; + }; + + union { + struct { + uint32_t pasid:16; + uint32_t reserved3:16; + } bitfields3a; + struct { + uint32_t reserved4:2; + uint32_t doorbell_offset0:26; + int32_t reserved5:4; + } bitfields3b; + uint32_t ordinal3; + }; + + union { + struct { + uint32_t reserved6:2; + uint32_t doorbell_offset1:26; + uint32_t reserved7:4; + } bitfields4; + uint32_t ordinal4; + }; + + union { + struct { + uint32_t reserved8:2; + uint32_t doorbell_offset2:26; + uint32_t reserved9:4; + } bitfields5; + uint32_t ordinal5; + }; + + union { + struct { + uint32_t reserved10:2; + uint32_t doorbell_offset3:26; + uint32_t reserved11:4; + } bitfields6; + uint32_t ordinal6; + }; +}; +#endif + +#ifndef PM4_MEC_RELEASE_MEM_DEFINED +#define PM4_MEC_RELEASE_MEM_DEFINED + +enum mec_release_mem_event_index_enum { + event_index__mec_release_mem__end_of_pipe = 5, + event_index__mec_release_mem__shader_done = 6 +}; + +enum mec_release_mem_cache_policy_enum { + cache_policy__mec_release_mem__lru = 0, + cache_policy__mec_release_mem__stream = 1 +}; + +enum mec_release_mem_pq_exe_status_enum { + pq_exe_status__mec_release_mem__default = 0, + pq_exe_status__mec_release_mem__phase_update = 1 +}; + +enum mec_release_mem_dst_sel_enum { + dst_sel__mec_release_mem__memory_controller = 0, + dst_sel__mec_release_mem__tc_l2 = 1, + dst_sel__mec_release_mem__queue_write_pointer_register = 2, + dst_sel__mec_release_mem__queue_write_pointer_poll_mask_bit = 3 +}; + +enum mec_release_mem_int_sel_enum { + int_sel__mec_release_mem__none = 0, + int_sel__mec_release_mem__send_interrupt_only = 1, + int_sel__mec_release_mem__send_interrupt_after_write_confirm = 2, + int_sel__mec_release_mem__send_data_after_write_confirm = 3, + int_sel__mec_release_mem__unconditionally_send_int_ctxid = 4, + int_sel__mec_release_mem__conditionally_send_int_ctxid_based_on_32_bit_compare = 5, + int_sel__mec_release_mem__conditionally_send_int_ctxid_based_on_64_bit_compare = 6 +}; + +enum mec_release_mem_data_sel_enum { + data_sel__mec_release_mem__none = 0, + data_sel__mec_release_mem__send_32_bit_low = 1, + data_sel__mec_release_mem__send_64_bit_data = 2, + data_sel__mec_release_mem__send_gpu_clock_counter = 3, + data_sel__mec_release_mem__send_cp_perfcounter_hi_lo = 4, + data_sel__mec_release_mem__store_gds_data_to_memory = 5 +}; + +struct pm4_mec_release_mem { + union { + union PM4_MES_TYPE_3_HEADER header; /*header */ + unsigned int ordinal1; + }; + + union { + struct { + unsigned int event_type:6; + unsigned int reserved1:2; + enum mec_release_mem_event_index_enum event_index:4; + unsigned int tcl1_vol_action_ena:1; + unsigned int tc_vol_action_ena:1; + unsigned int reserved2:1; + unsigned int tc_wb_action_ena:1; + unsigned int tcl1_action_ena:1; + unsigned int tc_action_ena:1; + uint32_t reserved3:1; + uint32_t tc_nc_action_ena:1; + uint32_t tc_wc_action_ena:1; + uint32_t tc_md_action_ena:1; + uint32_t reserved4:3; + enum mec_release_mem_cache_policy_enum cache_policy:2; + uint32_t reserved5:2; + enum mec_release_mem_pq_exe_status_enum pq_exe_status:1; + uint32_t reserved6:2; + } bitfields2; + unsigned int ordinal2; + }; + + union { + struct { + uint32_t reserved7:16; + enum mec_release_mem_dst_sel_enum dst_sel:2; + uint32_t reserved8:6; + enum mec_release_mem_int_sel_enum int_sel:3; + uint32_t reserved9:2; + enum mec_release_mem_data_sel_enum data_sel:3; + } bitfields3; + unsigned int ordinal3; + }; + + union { + struct { + uint32_t reserved10:2; + unsigned int address_lo_32b:30; + } bitfields4; + struct { + uint32_t reserved11:3; + uint32_t address_lo_64b:29; + } bitfields4b; + uint32_t reserved12; + unsigned int ordinal4; + }; + + union { + uint32_t address_hi; + uint32_t reserved13; + uint32_t ordinal5; + }; + + union { + uint32_t data_lo; + uint32_t cmp_data_lo; + struct { + uint32_t dw_offset:16; + uint32_t num_dwords:16; + } bitfields6c; + uint32_t reserved14; + uint32_t ordinal6; + }; + + union { + uint32_t data_hi; + uint32_t cmp_data_hi; + uint32_t reserved15; + uint32_t reserved16; + uint32_t ordinal7; + }; + + uint32_t int_ctxid; + +}; + +#endif + +enum { + CACHE_FLUSH_AND_INV_TS_EVENT = 0x00000014 +}; +#endif + diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h index 96a9cc0f02c97..5e3990bb4c4be 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_priv.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_priv.h @@ -39,11 +39,37 @@ #include "amd_shared.h" +#define KFD_MAX_RING_ENTRY_SIZE 8 + #define KFD_SYSFS_FILE_MODE 0444 -#define KFD_MMAP_DOORBELL_MASK 0x8000000000000ull -#define KFD_MMAP_EVENTS_MASK 0x4000000000000ull -#define KFD_MMAP_RESERVED_MEM_MASK 0x2000000000000ull +/* GPU ID hash width in bits */ +#define KFD_GPU_ID_HASH_WIDTH 16 + +/* Use upper bits of mmap offset to store KFD driver specific information. + * BITS[63:62] - Encode MMAP type + * BITS[61:46] - Encode gpu_id. To identify to which GPU the offset belongs to + * BITS[45:0] - MMAP offset value + * + * NOTE: struct vm_area_struct.vm_pgoff uses offset in pages. Hence, these + * defines are w.r.t to PAGE_SIZE + */ +#define KFD_MMAP_TYPE_SHIFT (62 - PAGE_SHIFT) +#define KFD_MMAP_TYPE_MASK (0x3ULL << KFD_MMAP_TYPE_SHIFT) +#define KFD_MMAP_TYPE_DOORBELL (0x3ULL << KFD_MMAP_TYPE_SHIFT) +#define KFD_MMAP_TYPE_EVENTS (0x2ULL << KFD_MMAP_TYPE_SHIFT) +#define KFD_MMAP_TYPE_RESERVED_MEM (0x1ULL << KFD_MMAP_TYPE_SHIFT) + +#define KFD_MMAP_GPU_ID_SHIFT (46 - PAGE_SHIFT) +#define KFD_MMAP_GPU_ID_MASK (((1ULL << KFD_GPU_ID_HASH_WIDTH) - 1) \ + << KFD_MMAP_GPU_ID_SHIFT) +#define KFD_MMAP_GPU_ID(gpu_id) ((((uint64_t)gpu_id) << KFD_MMAP_GPU_ID_SHIFT)\ + & KFD_MMAP_GPU_ID_MASK) +#define KFD_MMAP_GPU_ID_GET(offset) ((offset & KFD_MMAP_GPU_ID_MASK) \ + >> KFD_MMAP_GPU_ID_SHIFT) + +#define KFD_MMAP_OFFSET_VALUE_MASK (0x3FFFFFFFFFFFULL >> PAGE_SHIFT) +#define KFD_MMAP_OFFSET_VALUE_GET(offset) (offset & KFD_MMAP_OFFSET_VALUE_MASK) /* * When working with cp scheduler we should assign the HIQ manually or via @@ -55,9 +81,6 @@ #define KFD_CIK_HIQ_PIPE 4 #define KFD_CIK_HIQ_QUEUE 0 -/* GPU ID hash width in bits */ -#define KFD_GPU_ID_HASH_WIDTH 16 - /* Macro for allocating structures */ #define kfd_alloc_struct(ptr_to_struct) \ ((typeof(ptr_to_struct)) kzalloc(sizeof(*ptr_to_struct), GFP_KERNEL)) @@ -116,6 +139,11 @@ extern int debug_largebar; */ extern int ignore_crat; +/* + * Set sh_mem_config.retry_disable on Vega10 + */ +extern int vega10_noretry; + /** * enum kfd_sched_policy * @@ -148,6 +176,8 @@ enum cache_policy { cache_policy_noncoherent }; +#define KFD_IS_SOC15(chip) ((chip) >= CHIP_VEGA10) + struct kfd_event_interrupt_class { bool (*interrupt_isr)(struct kfd_dev *dev, const uint32_t *ih_ring_entry); @@ -160,6 +190,7 @@ struct kfd_device_info { const struct kfd_event_interrupt_class *event_interrupt_class; unsigned int max_pasid_bits; unsigned int max_no_of_hqd; + unsigned int doorbell_size; size_t ih_ring_entry_size; uint8_t num_of_watch_points; uint16_t mqd_size_aligned; @@ -173,6 +204,7 @@ struct kfd_mem_obj { uint32_t range_end; uint64_t gpu_addr; uint32_t *cpu_ptr; + void *gtt_mem; }; struct kfd_vmid_info { @@ -364,7 +396,7 @@ struct queue_properties { uint32_t queue_percent; uint32_t *read_ptr; uint32_t *write_ptr; - uint32_t __iomem *doorbell_ptr; + void __iomem *doorbell_ptr; uint32_t doorbell_off; bool is_interop; bool is_evicted; @@ -427,6 +459,7 @@ struct queue { uint32_t queue; unsigned int sdma_id; + unsigned int doorbell_id; struct kfd_process *process; struct kfd_dev *device; @@ -501,6 +534,9 @@ struct qcm_process_device { /* IB memory */ uint64_t ib_base; void *ib_kaddr; + + /* doorbell resources per process per device */ + unsigned long *doorbell_bitmap; }; /* KFD Memory Eviction */ @@ -512,6 +548,8 @@ struct qcm_process_device { /* Approx. time before evicting the process again */ #define PROCESS_ACTIVE_TIME_MS 10 +int kgd2kfd_quiesce_mm(struct mm_struct *mm); +int kgd2kfd_resume_mm(struct mm_struct *mm); int kgd2kfd_schedule_evict_and_restore_process(struct mm_struct *mm, struct dma_fence *fence); @@ -681,6 +719,8 @@ struct kfd_process *kfd_get_process(const struct task_struct *); struct kfd_process *kfd_lookup_process_by_pasid(unsigned int pasid); struct kfd_process *kfd_lookup_process_by_mm(const struct mm_struct *mm); void kfd_unref_process(struct kfd_process *p); +int kfd_process_evict_queues(struct kfd_process *p); +int kfd_process_restore_queues(struct kfd_process *p); void kfd_suspend_all_processes(void); int kfd_resume_all_processes(void); @@ -693,7 +733,7 @@ struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev, struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev, struct kfd_process *p); -int kfd_reserved_mem_mmap(struct kfd_process *process, +int kfd_reserved_mem_mmap(struct kfd_dev *dev, struct kfd_process *process, struct vm_area_struct *vma); /* KFD process API for creating and translating handles */ @@ -721,17 +761,20 @@ unsigned int kfd_pasid_alloc(void); void kfd_pasid_free(unsigned int pasid); /* Doorbells */ +size_t kfd_doorbell_process_slice(struct kfd_dev *kfd); int kfd_doorbell_init(struct kfd_dev *kfd); void kfd_doorbell_fini(struct kfd_dev *kfd); -int kfd_doorbell_mmap(struct kfd_process *process, struct vm_area_struct *vma); -u32 __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd, +int kfd_doorbell_mmap(struct kfd_dev *dev, struct kfd_process *process, + struct vm_area_struct *vma); +void __iomem *kfd_get_kernel_doorbell(struct kfd_dev *kfd, unsigned int *doorbell_off); void kfd_release_kernel_doorbell(struct kfd_dev *kfd, u32 __iomem *db_addr); u32 read_kernel_doorbell(u32 __iomem *db); -void write_kernel_doorbell(u32 __iomem *db, u32 value); -unsigned int kfd_queue_id_to_doorbell(struct kfd_dev *kfd, +void write_kernel_doorbell(void __iomem *db, u32 value); +void write_kernel_doorbell64(void __iomem *db, u64 value); +unsigned int kfd_doorbell_id_to_offset(struct kfd_dev *kfd, struct kfd_process *process, - unsigned int queue_id); + unsigned int doorbell_id); phys_addr_t kfd_get_process_doorbells(struct kfd_dev *dev, struct kfd_process *process); int kfd_alloc_process_doorbells(struct kfd_process *process); @@ -788,6 +831,8 @@ struct mqd_manager *mqd_manager_init_vi(enum KFD_MQD_TYPE type, struct kfd_dev *dev); struct mqd_manager *mqd_manager_init_vi_tonga(enum KFD_MQD_TYPE type, struct kfd_dev *dev); +struct mqd_manager *mqd_manager_init_v9(enum KFD_MQD_TYPE type, + struct kfd_dev *dev); struct device_queue_manager *device_queue_manager_init(struct kfd_dev *dev); void device_queue_manager_uninit(struct device_queue_manager *dqm); struct kernel_queue *kernel_queue_init(struct kfd_dev *dev, @@ -832,8 +877,42 @@ struct packet_manager { bool allocated; struct kfd_mem_obj *ib_buffer_obj; unsigned int ib_size_bytes; + + const struct packet_manager_funcs *pmf; +}; + +struct packet_manager_funcs { + /* Support ASIC-specific packet formats for PM4 packets */ + int (*map_process)(struct packet_manager *pm, uint32_t *buffer, + struct qcm_process_device *qpd); + int (*runlist)(struct packet_manager *pm, uint32_t *buffer, + uint64_t ib, size_t ib_size_in_dwords, bool chain); + int (*set_resources)(struct packet_manager *pm, uint32_t *buffer, + struct scheduling_resources *res); + int (*map_queues)(struct packet_manager *pm, uint32_t *buffer, + struct queue *q, bool is_static); + int (*unmap_queues)(struct packet_manager *pm, uint32_t *buffer, + enum kfd_queue_type type, + enum kfd_unmap_queues_filter mode, + uint32_t filter_param, bool reset, + unsigned int sdma_engine); + int (*query_status)(struct packet_manager *pm, uint32_t *buffer, + uint64_t fence_address, uint32_t fence_value); + int (*release_mem)(uint64_t gpu_addr, uint32_t *buffer); + + /* Packet sizes */ + int map_process_size; + int runlist_size; + int set_resources_size; + int map_queues_size; + int unmap_queues_size; + int query_status_size; + int release_mem_size; }; +extern const struct packet_manager_funcs kfd_vi_pm_funcs; +extern const struct packet_manager_funcs kfd_v9_pm_funcs; + int pm_init(struct packet_manager *pm, struct device_queue_manager *dqm); void pm_uninit(struct packet_manager *pm); int pm_send_set_resources(struct packet_manager *pm, @@ -849,12 +928,17 @@ int pm_send_unmap_queue(struct packet_manager *pm, enum kfd_queue_type type, void pm_release_ib(struct packet_manager *pm); -uint32_t pm_create_release_mem(uint64_t gpu_addr, uint32_t *buffer); +/* Following PM funcs can be shared among VI and AI */ +unsigned int pm_build_pm4_header(unsigned int opcode, size_t packet_size); +int pm_set_resources_vi(struct packet_manager *pm, uint32_t *buffer, + struct scheduling_resources *res); uint64_t kfd_get_number_elems(struct kfd_dev *kfd); /* Events */ extern const struct kfd_event_interrupt_class event_interrupt_class_cik; +extern const struct kfd_event_interrupt_class event_interrupt_class_v9; + extern const struct kfd_device_global_init_class device_global_init_class_cik; void kfd_event_init_process(struct kfd_process *p); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process.c b/drivers/gpu/drm/amd/amdkfd/kfd_process.c index 1711ad0642f74..1d80b4f7c681c 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process.c @@ -332,6 +332,7 @@ static void kfd_process_destroy_pdds(struct kfd_process *p) free_pages((unsigned long)pdd->qpd.cwsr_kaddr, get_order(KFD_CWSR_TBA_TMA_SIZE)); + kfree(pdd->qpd.doorbell_bitmap); idr_destroy(&pdd->alloc_idr); kfree(pdd); @@ -451,7 +452,8 @@ static int kfd_process_init_cwsr_apu(struct kfd_process *p, struct file *filep) if (!dev->cwsr_enabled || qpd->cwsr_kaddr || qpd->cwsr_base) continue; - offset = (dev->id | KFD_MMAP_RESERVED_MEM_MASK) << PAGE_SHIFT; + offset = (KFD_MMAP_TYPE_RESERVED_MEM | KFD_MMAP_GPU_ID(dev->id)) + << PAGE_SHIFT; qpd->tba_addr = (int64_t)vm_mmap(filep, 0, KFD_CWSR_TBA_TMA_SIZE, PROT_READ | PROT_EXEC, MAP_SHARED, offset); @@ -585,6 +587,31 @@ err_alloc_process: return ERR_PTR(err); } +static int init_doorbell_bitmap(struct qcm_process_device *qpd, + struct kfd_dev *dev) +{ + unsigned int i; + + if (!KFD_IS_SOC15(dev->device_info->asic_family)) + return 0; + + qpd->doorbell_bitmap = + kzalloc(DIV_ROUND_UP(KFD_MAX_NUM_OF_QUEUES_PER_PROCESS, + BITS_PER_BYTE), GFP_KERNEL); + if (!qpd->doorbell_bitmap) + return -ENOMEM; + + /* Mask out any reserved doorbells */ + for (i = 0; i < KFD_MAX_NUM_OF_QUEUES_PER_PROCESS; i++) + if ((dev->shared_resources.reserved_doorbell_mask & i) == + dev->shared_resources.reserved_doorbell_val) { + set_bit(i, qpd->doorbell_bitmap); + pr_debug("reserved doorbell 0x%03x\n", i); + } + + return 0; +} + struct kfd_process_device *kfd_get_process_device_data(struct kfd_dev *dev, struct kfd_process *p) { @@ -606,6 +633,12 @@ struct kfd_process_device *kfd_create_process_device_data(struct kfd_dev *dev, if (!pdd) return NULL; + if (init_doorbell_bitmap(&pdd->qpd, dev)) { + pr_err("Failed to init doorbell for process\n"); + kfree(pdd); + return NULL; + } + pdd->dev = dev; INIT_LIST_HEAD(&pdd->qpd.queues_list); INIT_LIST_HEAD(&pdd->qpd.priv_queue_list); @@ -808,7 +841,7 @@ struct kfd_process *kfd_lookup_process_by_mm(const struct mm_struct *mm) * Eviction is reference-counted per process-device. This means multiple * evictions from different sources can be nested safely. */ -static int process_evict_queues(struct kfd_process *p) +int kfd_process_evict_queues(struct kfd_process *p) { struct kfd_process_device *pdd; int r = 0; @@ -844,7 +877,7 @@ fail: } /* process_restore_queues - Restore all user queues of a process */ -static int process_restore_queues(struct kfd_process *p) +int kfd_process_restore_queues(struct kfd_process *p) { struct kfd_process_device *pdd; int r, ret = 0; @@ -886,7 +919,7 @@ static void evict_process_worker(struct work_struct *work) flush_delayed_work(&p->restore_work); pr_debug("Started evicting pasid %d\n", p->pasid); - ret = process_evict_queues(p); + ret = kfd_process_evict_queues(p); if (!ret) { dma_fence_signal(p->ef); dma_fence_put(p->ef); @@ -946,7 +979,7 @@ static void restore_process_worker(struct work_struct *work) return; } - ret = process_restore_queues(p); + ret = kfd_process_restore_queues(p); if (!ret) pr_debug("Finished restoring pasid %d\n", p->pasid); else @@ -963,7 +996,7 @@ void kfd_suspend_all_processes(void) cancel_delayed_work_sync(&p->eviction_work); cancel_delayed_work_sync(&p->restore_work); - if (process_evict_queues(p)) + if (kfd_process_evict_queues(p)) pr_err("Failed to suspend process %d\n", p->pasid); dma_fence_signal(p->ef); dma_fence_put(p->ef); @@ -989,15 +1022,12 @@ int kfd_resume_all_processes(void) return ret; } -int kfd_reserved_mem_mmap(struct kfd_process *process, +int kfd_reserved_mem_mmap(struct kfd_dev *dev, struct kfd_process *process, struct vm_area_struct *vma) { - struct kfd_dev *dev = kfd_device_by_id(vma->vm_pgoff); struct kfd_process_device *pdd; struct qcm_process_device *qpd; - if (!dev) - return -EINVAL; if ((vma->vm_end - vma->vm_start) != KFD_CWSR_TBA_TMA_SIZE) { pr_err("Incorrect CWSR mapping size.\n"); return -EINVAL; diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c index 7817e327ea6dd..d65ce0436b31b 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_process_queue_manager.c @@ -119,9 +119,6 @@ static int create_cp_queue(struct process_queue_manager *pqm, /* Doorbell initialized in user space*/ q_properties->doorbell_ptr = NULL; - q_properties->doorbell_off = - kfd_queue_id_to_doorbell(dev, pqm->process, qid); - /* let DQM handle it*/ q_properties->vmid = 0; q_properties->queue_id = qid; @@ -244,10 +241,20 @@ int pqm_create_queue(struct process_queue_manager *pqm, } if (retval != 0) { - pr_err("DQM create queue failed\n"); + pr_err("Pasid %d DQM create queue %d failed. ret %d\n", + pqm->process->pasid, type, retval); goto err_create_queue; } + if (q) + /* Return the doorbell offset within the doorbell page + * to the caller so it can be passed up to user mode + * (in bytes). + */ + properties->doorbell_off = + (q->properties.doorbell_off * sizeof(uint32_t)) & + (kfd_doorbell_process_slice(dev) - 1); + pr_debug("PQM After DQM create queue\n"); list_add(&pqn->process_queue_list, &pqm->queues); @@ -313,8 +320,11 @@ int pqm_destroy_queue(struct process_queue_manager *pqm, unsigned int qid) dqm = pqn->q->device->dqm; retval = dqm->ops.destroy_queue(dqm, &pdd->qpd, pqn->q); if (retval) { - pr_debug("Destroy queue failed, returned %d\n", retval); - goto err_destroy_queue; + pr_err("Pasid %d destroy queue %d failed, ret %d\n", + pqm->process->pasid, + pqn->q->properties.queue_id, retval); + if (retval != -ETIME) + goto err_destroy_queue; } uninit_queue(pqn->q); } diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_queue.c b/drivers/gpu/drm/amd/amdkfd/kfd_queue.c index a5315d4f1c954..6dcd621e5b71f 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_queue.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_queue.c @@ -36,8 +36,8 @@ void print_queue_properties(struct queue_properties *q) pr_debug("Queue Address: 0x%llX\n", q->queue_address); pr_debug("Queue Id: %u\n", q->queue_id); pr_debug("Queue Process Vmid: %u\n", q->vmid); - pr_debug("Queue Read Pointer: 0x%p\n", q->read_ptr); - pr_debug("Queue Write Pointer: 0x%p\n", q->write_ptr); + pr_debug("Queue Read Pointer: 0x%px\n", q->read_ptr); + pr_debug("Queue Write Pointer: 0x%px\n", q->write_ptr); pr_debug("Queue Doorbell Pointer: 0x%p\n", q->doorbell_ptr); pr_debug("Queue Doorbell Offset: %u\n", q->doorbell_off); } @@ -53,8 +53,8 @@ void print_queue(struct queue *q) pr_debug("Queue Address: 0x%llX\n", q->properties.queue_address); pr_debug("Queue Id: %u\n", q->properties.queue_id); pr_debug("Queue Process Vmid: %u\n", q->properties.vmid); - pr_debug("Queue Read Pointer: 0x%p\n", q->properties.read_ptr); - pr_debug("Queue Write Pointer: 0x%p\n", q->properties.write_ptr); + pr_debug("Queue Read Pointer: 0x%px\n", q->properties.read_ptr); + pr_debug("Queue Write Pointer: 0x%px\n", q->properties.write_ptr); pr_debug("Queue Doorbell Pointer: 0x%p\n", q->properties.doorbell_ptr); pr_debug("Queue Doorbell Offset: %u\n", q->properties.doorbell_off); pr_debug("Queue MQD Address: 0x%p\n", q->mqd); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c index ac28abc94e57b..bc95d4dfee2e9 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.c +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.c @@ -1239,6 +1239,12 @@ int kfd_topology_add_device(struct kfd_dev *gpu) HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT) & HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK); break; + case CHIP_VEGA10: + case CHIP_RAVEN: + dev->node_props.capability |= ((HSA_CAP_DOORBELL_TYPE_2_0 << + HSA_CAP_DOORBELL_TYPE_TOTALBITS_SHIFT) & + HSA_CAP_DOORBELL_TYPE_TOTALBITS_MASK); + break; default: WARN(1, "Unexpected ASIC family %u", dev->gpu->device_info->asic_family); diff --git a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h index eb54cfcaf0395..7d9c3f948dff2 100644 --- a/drivers/gpu/drm/amd/amdkfd/kfd_topology.h +++ b/drivers/gpu/drm/amd/amdkfd/kfd_topology.h @@ -45,6 +45,7 @@ #define HSA_CAP_DOORBELL_TYPE_PRE_1_0 0x0 #define HSA_CAP_DOORBELL_TYPE_1_0 0x1 +#define HSA_CAP_DOORBELL_TYPE_2_0 0x2 #define HSA_CAP_AQL_QUEUE_DOUBLE_MAP 0x00004000 struct kfd_node_properties { diff --git a/drivers/gpu/drm/amd/amdkfd/soc15_int.h b/drivers/gpu/drm/amd/amdkfd/soc15_int.h new file mode 100644 index 0000000000000..0bc0b25cb4105 --- /dev/null +++ b/drivers/gpu/drm/amd/amdkfd/soc15_int.h @@ -0,0 +1,47 @@ +/* + * Copyright 2016-2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + */ + +#ifndef HSA_SOC15_INT_H_INCLUDED +#define HSA_SOC15_INT_H_INCLUDED + +#include "soc15_ih_clientid.h" + +#define SOC15_INTSRC_CP_END_OF_PIPE 181 +#define SOC15_INTSRC_CP_BAD_OPCODE 183 +#define SOC15_INTSRC_SQ_INTERRUPT_MSG 239 +#define SOC15_INTSRC_VMC_FAULT 0 +#define SOC15_INTSRC_SDMA_TRAP 224 + + +#define SOC15_CLIENT_ID_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[0]) & 0xff) +#define SOC15_SOURCE_ID_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[0]) >> 8 & 0xff) +#define SOC15_RING_ID_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[0]) >> 16 & 0xff) +#define SOC15_VMID_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[0]) >> 24 & 0xf) +#define SOC15_VMID_TYPE_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[0]) >> 31 & 0x1) +#define SOC15_PASID_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[3]) & 0xffff) +#define SOC15_CONTEXT_ID0_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[4])) +#define SOC15_CONTEXT_ID1_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[5])) +#define SOC15_CONTEXT_ID2_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[6])) +#define SOC15_CONTEXT_ID3_FROM_IH_ENTRY(entry) (le32_to_cpu(entry[7])) + +#endif + diff --git a/drivers/gpu/drm/amd/display/Kconfig b/drivers/gpu/drm/amd/display/Kconfig index 5b124a67404cc..d5d4586e61765 100644 --- a/drivers/gpu/drm/amd/display/Kconfig +++ b/drivers/gpu/drm/amd/display/Kconfig @@ -9,14 +9,6 @@ config DRM_AMD_DC support for AMDGPU. This adds required support for Vega and Raven ASICs. -config DRM_AMD_DC_PRE_VEGA - bool "DC support for Polaris and older ASICs" - default y - help - Choose this option to enable the new DC support for older asics - by default. This includes Polaris, Carrizo, Tonga, Bonaire, - and Hawaii. - config DRM_AMD_DC_FBC bool "AMD FBC - Enable Frame Buffer Compression" depends on DRM_AMD_DC diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c index 1dd1142246c25..f9b9ab90558c9 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.c @@ -433,11 +433,6 @@ static int amdgpu_dm_init(struct amdgpu_device *adev) init_data.dce_environment = DCE_ENV_PRODUCTION_DRV; - if (amdgpu_dc_log) - init_data.log_mask = DC_DEFAULT_LOG_MASK; - else - init_data.log_mask = DC_MIN_LOG_MASK; - /* * TODO debug why this doesn't work on Raven */ @@ -649,18 +644,6 @@ amdgpu_dm_find_first_crtc_matching_connector(struct drm_atomic_state *state, static int dm_resume(void *handle) { struct amdgpu_device *adev = handle; - struct amdgpu_display_manager *dm = &adev->dm; - int ret = 0; - - /* power on hardware */ - dc_set_power_state(dm->dc, DC_ACPI_CM_POWER_STATE_D0); - - ret = amdgpu_dm_display_resume(adev); - return ret; -} - -int amdgpu_dm_display_resume(struct amdgpu_device *adev) -{ struct drm_device *ddev = adev->ddev; struct amdgpu_display_manager *dm = &adev->dm; struct amdgpu_dm_connector *aconnector; @@ -671,10 +654,12 @@ int amdgpu_dm_display_resume(struct amdgpu_device *adev) struct drm_plane *plane; struct drm_plane_state *new_plane_state; struct dm_plane_state *dm_new_plane_state; - - int ret = 0; + int ret; int i; + /* power on hardware */ + dc_set_power_state(dm->dc, DC_ACPI_CM_POWER_STATE_D0); + /* program HPD filter */ dc_resume(dm->dc); @@ -688,8 +673,7 @@ int amdgpu_dm_display_resume(struct amdgpu_device *adev) amdgpu_dm_irq_resume_early(adev); /* Do detection*/ - list_for_each_entry(connector, - &ddev->mode_config.connector_list, head) { + list_for_each_entry(connector, &ddev->mode_config.connector_list, head) { aconnector = to_amdgpu_dm_connector(connector); /* @@ -711,7 +695,7 @@ int amdgpu_dm_display_resume(struct amdgpu_device *adev) } /* Force mode set in atomic comit */ - for_each_new_crtc_in_state(adev->dm.cached_state, crtc, new_crtc_state, i) + for_each_new_crtc_in_state(dm->cached_state, crtc, new_crtc_state, i) new_crtc_state->active_changed = true; /* @@ -719,7 +703,7 @@ int amdgpu_dm_display_resume(struct amdgpu_device *adev) * them here, since they were duplicated as part of the suspend * procedure. */ - for_each_new_crtc_in_state(adev->dm.cached_state, crtc, new_crtc_state, i) { + for_each_new_crtc_in_state(dm->cached_state, crtc, new_crtc_state, i) { dm_new_crtc_state = to_dm_crtc_state(new_crtc_state); if (dm_new_crtc_state->stream) { WARN_ON(kref_read(&dm_new_crtc_state->stream->refcount) > 1); @@ -728,7 +712,7 @@ int amdgpu_dm_display_resume(struct amdgpu_device *adev) } } - for_each_new_plane_in_state(adev->dm.cached_state, plane, new_plane_state, i) { + for_each_new_plane_in_state(dm->cached_state, plane, new_plane_state, i) { dm_new_plane_state = to_dm_plane_state(new_plane_state); if (dm_new_plane_state->dc_state) { WARN_ON(kref_read(&dm_new_plane_state->dc_state->refcount) > 1); @@ -737,9 +721,9 @@ int amdgpu_dm_display_resume(struct amdgpu_device *adev) } } - ret = drm_atomic_helper_resume(ddev, adev->dm.cached_state); + ret = drm_atomic_helper_resume(ddev, dm->cached_state); - adev->dm.cached_state = NULL; + dm->cached_state = NULL; amdgpu_dm_irq_resume_late(adev); @@ -927,6 +911,7 @@ amdgpu_dm_update_connector_after_detect(struct amdgpu_dm_connector *aconnector) drm_mode_connector_update_edid_property(connector, NULL); aconnector->num_modes = 0; aconnector->dc_sink = NULL; + aconnector->edid = NULL; } mutex_unlock(&dev->mode_config.mutex); @@ -1131,6 +1116,7 @@ static int dce110_register_irq_handlers(struct amdgpu_device *adev) if (adev->asic_type == CHIP_VEGA10 || adev->asic_type == CHIP_VEGA12 || + adev->asic_type == CHIP_VEGA20 || adev->asic_type == CHIP_RAVEN) client_id = SOC15_IH_CLIENTID_DCE; @@ -1529,8 +1515,10 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev) case CHIP_POLARIS11: case CHIP_POLARIS10: case CHIP_POLARIS12: + case CHIP_VEGAM: case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: if (dce110_register_irq_handlers(dm->adev)) { DRM_ERROR("DM: Failed to initialize IRQ\n"); goto fail; @@ -1549,7 +1537,7 @@ static int amdgpu_dm_initialize_drm_device(struct amdgpu_device *adev) break; #endif default: - DRM_ERROR("Usupported ASIC type: 0x%X\n", adev->asic_type); + DRM_ERROR("Unsupported ASIC type: 0x%X\n", adev->asic_type); goto fail; } @@ -1657,7 +1645,6 @@ static ssize_t s3_debug_store(struct device *device, if (ret == 0) { if (s3_state) { dm_resume(adev); - amdgpu_dm_display_resume(adev); drm_kms_helper_hotplug_event(adev->ddev); } else dm_suspend(adev); @@ -1722,6 +1709,7 @@ static int dm_early_init(void *handle) adev->mode_info.plane_type = dm_plane_type_default; break; case CHIP_POLARIS10: + case CHIP_VEGAM: adev->mode_info.num_crtc = 6; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 6; @@ -1729,6 +1717,7 @@ static int dm_early_init(void *handle) break; case CHIP_VEGA10: case CHIP_VEGA12: + case CHIP_VEGA20: adev->mode_info.num_crtc = 6; adev->mode_info.num_hpd = 6; adev->mode_info.num_dig = 6; @@ -1743,7 +1732,7 @@ static int dm_early_init(void *handle) break; #endif default: - DRM_ERROR("Usupported ASIC type: 0x%X\n", adev->asic_type); + DRM_ERROR("Unsupported ASIC type: 0x%X\n", adev->asic_type); return -EINVAL; } @@ -1848,7 +1837,7 @@ static bool fill_rects_from_plane_state(const struct drm_plane_state *state, static int get_fb_info(const struct amdgpu_framebuffer *amdgpu_fb, uint64_t *tiling_flags) { - struct amdgpu_bo *rbo = gem_to_amdgpu_bo(amdgpu_fb->obj); + struct amdgpu_bo *rbo = gem_to_amdgpu_bo(amdgpu_fb->base.obj[0]); int r = amdgpu_bo_reserve(rbo, false); if (unlikely(r)) { @@ -1977,6 +1966,7 @@ static int fill_plane_attributes_from_fb(struct amdgpu_device *adev, if (adev->asic_type == CHIP_VEGA10 || adev->asic_type == CHIP_VEGA12 || + adev->asic_type == CHIP_VEGA20 || adev->asic_type == CHIP_RAVEN) { /* Fill GFX9 params */ plane_state->tiling_info.gfx9.num_pipes = @@ -2017,7 +2007,6 @@ static int fill_plane_attributes(struct amdgpu_device *adev, const struct amdgpu_framebuffer *amdgpu_fb = to_amdgpu_framebuffer(plane_state->fb); const struct drm_crtc *crtc = plane_state->crtc; - struct dc_transfer_func *input_tf; int ret = 0; if (!fill_rects_from_plane_state(plane_state, dc_plane_state)) @@ -2031,13 +2020,6 @@ static int fill_plane_attributes(struct amdgpu_device *adev, if (ret) return ret; - input_tf = dc_create_transfer_func(); - - if (input_tf == NULL) - return -ENOMEM; - - dc_plane_state->in_transfer_func = input_tf; - /* * Always set input transfer function, since plane state is refreshed * every time. @@ -2206,7 +2188,6 @@ fill_stream_properties_from_drm_display_mode(struct dc_stream_state *stream, const struct drm_connector *connector) { struct dc_crtc_timing *timing_out = &stream->timing; - struct dc_transfer_func *tf = dc_create_transfer_func(); memset(timing_out, 0, sizeof(struct dc_crtc_timing)); @@ -2250,9 +2231,8 @@ fill_stream_properties_from_drm_display_mode(struct dc_stream_state *stream, stream->output_color_space = get_output_color_space(timing_out); - tf->type = TF_TYPE_PREDEFINED; - tf->tf = TRANSFER_FUNCTION_SRGB; - stream->out_transfer_func = tf; + stream->out_transfer_func->type = TF_TYPE_PREDEFINED; + stream->out_transfer_func->tf = TRANSFER_FUNCTION_SRGB; } static void fill_audio_info(struct audio_info *audio_info, @@ -2488,6 +2468,9 @@ create_stream_for_sink(struct amdgpu_dm_connector *aconnector, update_stream_signal(stream); + if (dm_state && dm_state->freesync_capable) + stream->ignore_msa_timing_param = true; + return stream; } @@ -2710,18 +2693,15 @@ static void amdgpu_dm_connector_destroy(struct drm_connector *connector) const struct dc_link *link = aconnector->dc_link; struct amdgpu_device *adev = connector->dev->dev_private; struct amdgpu_display_manager *dm = &adev->dm; + #if defined(CONFIG_BACKLIGHT_CLASS_DEVICE) ||\ defined(CONFIG_BACKLIGHT_CLASS_DEVICE_MODULE) if ((link->connector_signal & (SIGNAL_TYPE_EDP | SIGNAL_TYPE_LVDS)) && - link->type != dc_connection_none) { - amdgpu_dm_register_backlight_device(dm); - - if (dm->backlight_dev) { - backlight_device_unregister(dm->backlight_dev); - dm->backlight_dev = NULL; - } - + link->type != dc_connection_none && + dm->backlight_dev) { + backlight_device_unregister(dm->backlight_dev); + dm->backlight_dev = NULL; } #endif drm_connector_unregister(connector); @@ -2855,7 +2835,7 @@ static void handle_edid_mgmt(struct amdgpu_dm_connector *aconnector) create_eml_sink(aconnector); } -int amdgpu_dm_connector_mode_valid(struct drm_connector *connector, +enum drm_mode_status amdgpu_dm_connector_mode_valid(struct drm_connector *connector, struct drm_display_mode *mode) { int result = MODE_ERROR; @@ -3058,8 +3038,7 @@ static int dm_plane_helper_prepare_fb(struct drm_plane *plane, } afb = to_amdgpu_framebuffer(new_state->fb); - - obj = afb->obj; + obj = new_state->fb->obj[0]; rbo = gem_to_amdgpu_bo(obj); adev = amdgpu_ttm_adev(rbo->tbo.bdev); r = amdgpu_bo_reserve(rbo, false); @@ -3067,12 +3046,11 @@ static int dm_plane_helper_prepare_fb(struct drm_plane *plane, return r; if (plane->type != DRM_PLANE_TYPE_CURSOR) - domain = amdgpu_display_framebuffer_domains(adev); + domain = amdgpu_display_supported_domains(adev); else domain = AMDGPU_GEM_DOMAIN_VRAM; r = amdgpu_bo_pin(rbo, domain, &afb->address); - amdgpu_bo_unreserve(rbo); if (unlikely(r != 0)) { @@ -3123,14 +3101,12 @@ static void dm_plane_helper_cleanup_fb(struct drm_plane *plane, struct drm_plane_state *old_state) { struct amdgpu_bo *rbo; - struct amdgpu_framebuffer *afb; int r; if (!old_state->fb) return; - afb = to_amdgpu_framebuffer(old_state->fb); - rbo = gem_to_amdgpu_bo(afb->obj); + rbo = gem_to_amdgpu_bo(old_state->fb->obj[0]); r = amdgpu_bo_reserve(rbo, false); if (unlikely(r)) { DRM_ERROR("failed to reserve rbo before unpin\n"); @@ -3773,7 +3749,7 @@ static void remove_stream(struct amdgpu_device *adev, static int get_cursor_position(struct drm_plane *plane, struct drm_crtc *crtc, struct dc_cursor_position *position) { - struct amdgpu_crtc *amdgpu_crtc = amdgpu_crtc = to_amdgpu_crtc(crtc); + struct amdgpu_crtc *amdgpu_crtc = to_amdgpu_crtc(crtc); int x, y; int xorigin = 0, yorigin = 0; @@ -3905,7 +3881,7 @@ static void amdgpu_dm_do_flip(struct drm_crtc *crtc, int r, vpos, hpos; struct amdgpu_crtc *acrtc = to_amdgpu_crtc(crtc); struct amdgpu_framebuffer *afb = to_amdgpu_framebuffer(fb); - struct amdgpu_bo *abo = gem_to_amdgpu_bo(afb->obj); + struct amdgpu_bo *abo = gem_to_amdgpu_bo(fb->obj[0]); struct amdgpu_device *adev = crtc->dev->dev_private; bool async_flip = (crtc->state->pageflip_flags & DRM_MODE_PAGE_FLIP_ASYNC) != 0; struct dc_flip_addrs addr = { {0} }; @@ -3986,6 +3962,96 @@ static void amdgpu_dm_do_flip(struct drm_crtc *crtc, spin_unlock_irqrestore(&crtc->dev->event_lock, flags); } +/* + * TODO this whole function needs to go + * + * dc_surface_update is needlessly complex. See if we can just replace this + * with a dc_plane_state and follow the atomic model a bit more closely here. + */ +static bool commit_planes_to_stream( + struct dc *dc, + struct dc_plane_state **plane_states, + uint8_t new_plane_count, + struct dm_crtc_state *dm_new_crtc_state, + struct dm_crtc_state *dm_old_crtc_state, + struct dc_state *state) +{ + /* no need to dynamically allocate this. it's pretty small */ + struct dc_surface_update updates[MAX_SURFACES]; + struct dc_flip_addrs *flip_addr; + struct dc_plane_info *plane_info; + struct dc_scaling_info *scaling_info; + int i; + struct dc_stream_state *dc_stream = dm_new_crtc_state->stream; + struct dc_stream_update *stream_update = + kzalloc(sizeof(struct dc_stream_update), GFP_KERNEL); + + if (!stream_update) { + BREAK_TO_DEBUGGER(); + return false; + } + + flip_addr = kcalloc(MAX_SURFACES, sizeof(struct dc_flip_addrs), + GFP_KERNEL); + plane_info = kcalloc(MAX_SURFACES, sizeof(struct dc_plane_info), + GFP_KERNEL); + scaling_info = kcalloc(MAX_SURFACES, sizeof(struct dc_scaling_info), + GFP_KERNEL); + + if (!flip_addr || !plane_info || !scaling_info) { + kfree(flip_addr); + kfree(plane_info); + kfree(scaling_info); + kfree(stream_update); + return false; + } + + memset(updates, 0, sizeof(updates)); + + stream_update->src = dc_stream->src; + stream_update->dst = dc_stream->dst; + stream_update->out_transfer_func = dc_stream->out_transfer_func; + + for (i = 0; i < new_plane_count; i++) { + updates[i].surface = plane_states[i]; + updates[i].gamma = + (struct dc_gamma *)plane_states[i]->gamma_correction; + updates[i].in_transfer_func = plane_states[i]->in_transfer_func; + flip_addr[i].address = plane_states[i]->address; + flip_addr[i].flip_immediate = plane_states[i]->flip_immediate; + plane_info[i].color_space = plane_states[i]->color_space; + plane_info[i].format = plane_states[i]->format; + plane_info[i].plane_size = plane_states[i]->plane_size; + plane_info[i].rotation = plane_states[i]->rotation; + plane_info[i].horizontal_mirror = plane_states[i]->horizontal_mirror; + plane_info[i].stereo_format = plane_states[i]->stereo_format; + plane_info[i].tiling_info = plane_states[i]->tiling_info; + plane_info[i].visible = plane_states[i]->visible; + plane_info[i].per_pixel_alpha = plane_states[i]->per_pixel_alpha; + plane_info[i].dcc = plane_states[i]->dcc; + scaling_info[i].scaling_quality = plane_states[i]->scaling_quality; + scaling_info[i].src_rect = plane_states[i]->src_rect; + scaling_info[i].dst_rect = plane_states[i]->dst_rect; + scaling_info[i].clip_rect = plane_states[i]->clip_rect; + + updates[i].flip_addr = &flip_addr[i]; + updates[i].plane_info = &plane_info[i]; + updates[i].scaling_info = &scaling_info[i]; + } + + dc_commit_updates_for_stream( + dc, + updates, + new_plane_count, + dc_stream, stream_update, plane_states, state); + + kfree(flip_addr); + kfree(plane_info); + kfree(scaling_info); + kfree(stream_update); + return true; +} + static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, struct drm_device *dev, struct amdgpu_display_manager *dm, @@ -4001,6 +4067,8 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, struct drm_crtc_state *new_pcrtc_state = drm_atomic_get_new_crtc_state(state, pcrtc); struct dm_crtc_state *acrtc_state = to_dm_crtc_state(new_pcrtc_state); + struct dm_crtc_state *dm_old_crtc_state = + to_dm_crtc_state(drm_atomic_get_old_crtc_state(state, pcrtc)); struct dm_atomic_state *dm_state = to_dm_atomic_state(state); int planes_count = 0; unsigned long flags; @@ -4037,7 +4105,7 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, } spin_unlock_irqrestore(&crtc->dev->event_lock, flags); - if (!pflip_needed) { + if (!pflip_needed || plane->type == DRM_PLANE_TYPE_OVERLAY) { WARN_ON(!dm_new_plane_state->dc_state); plane_states_constructed[planes_count] = dm_new_plane_state->dc_state; @@ -4079,10 +4147,12 @@ static void amdgpu_dm_commit_planes(struct drm_atomic_state *state, spin_unlock_irqrestore(&pcrtc->dev->event_lock, flags); } - if (false == dc_commit_planes_to_stream(dm->dc, + + if (false == commit_planes_to_stream(dm->dc, plane_states_constructed, planes_count, - dc_stream_attach, + acrtc_state, + dm_old_crtc_state, dm_state->context)) dm_error("%s: Failed to attach plane!\n", __func__); } else { @@ -4307,8 +4377,10 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) struct amdgpu_crtc *acrtc = to_amdgpu_crtc(dm_new_con_state->base.crtc); struct dc_stream_status *status = NULL; - if (acrtc) + if (acrtc) { new_crtc_state = drm_atomic_get_new_crtc_state(state, &acrtc->base); + old_crtc_state = drm_atomic_get_old_crtc_state(state, &acrtc->base); + } /* Skip any modesets/resets */ if (!acrtc || drm_atomic_crtc_needs_modeset(new_crtc_state)) @@ -4331,11 +4403,12 @@ static void amdgpu_dm_atomic_commit_tail(struct drm_atomic_state *state) WARN_ON(!status->plane_count); /*TODO How it works with MPO ?*/ - if (!dc_commit_planes_to_stream( + if (!commit_planes_to_stream( dm->dc, status->plane_states, status->plane_count, - dm_new_crtc_state->stream, + dm_new_crtc_state, + to_dm_crtc_state(old_crtc_state), dm_state->context)) dm_error("%s: Failed to update stream scaling!\n", __func__); } @@ -4555,8 +4628,8 @@ static int dm_update_crtcs_state(struct dc *dc, for_each_oldnew_crtc_in_state(state, crtc, old_crtc_state, new_crtc_state, i) { struct amdgpu_crtc *acrtc = NULL; struct amdgpu_dm_connector *aconnector = NULL; - struct drm_connector_state *new_con_state = NULL; - struct dm_connector_state *dm_conn_state = NULL; + struct drm_connector_state *drm_new_conn_state = NULL, *drm_old_conn_state = NULL; + struct dm_connector_state *dm_new_conn_state = NULL, *dm_old_conn_state = NULL; struct drm_plane_state *new_plane_state = NULL; new_stream = NULL; @@ -4577,19 +4650,22 @@ static int dm_update_crtcs_state(struct dc *dc, /* TODO This hack should go away */ if (aconnector && enable) { // Make sure fake sink is created in plug-in scenario - new_con_state = drm_atomic_get_connector_state(state, + drm_new_conn_state = drm_atomic_get_new_connector_state(state, &aconnector->base); + drm_old_conn_state = drm_atomic_get_old_connector_state(state, + &aconnector->base); - if (IS_ERR(new_con_state)) { - ret = PTR_ERR_OR_ZERO(new_con_state); + if (IS_ERR(drm_new_conn_state)) { + ret = PTR_ERR_OR_ZERO(drm_new_conn_state); break; } - dm_conn_state = to_dm_connector_state(new_con_state); + dm_new_conn_state = to_dm_connector_state(drm_new_conn_state); + dm_old_conn_state = to_dm_connector_state(drm_old_conn_state); new_stream = create_stream_for_sink(aconnector, &new_crtc_state->mode, - dm_conn_state); + dm_new_conn_state); /* * we can have no stream on ACTION_SET if a display @@ -4695,20 +4771,30 @@ next_crtc: * We want to do dc stream updates that do not require a * full modeset below. */ - if (!enable || !aconnector || modereset_required(new_crtc_state)) + if (!(enable && aconnector && new_crtc_state->enable && + new_crtc_state->active)) continue; /* * Given above conditions, the dc state cannot be NULL because: - * 1. We're attempting to enable a CRTC. Which has a... - * 2. Valid connector attached, and - * 3. User does not want to reset it (disable or mark inactive, - * which can happen on a CRTC that's already disabled). - * => It currently exists. + * 1. We're in the process of enabling CRTCs (just been added + * to the dc context, or already is on the context) + * 2. Has a valid connector attached, and + * 3. Is currently active and enabled. + * => The dc stream state currently exists. */ BUG_ON(dm_new_crtc_state->stream == NULL); - /* Color managment settings */ - if (dm_new_crtc_state->base.color_mgmt_changed) { + /* Scaling or underscan settings */ + if (is_scaling_state_different(dm_old_conn_state, dm_new_conn_state)) + update_stream_scaling_settings( + &new_crtc_state->mode, dm_new_conn_state, dm_new_crtc_state->stream); + + /* + * Color management settings. We also update color properties + * when a modeset is needed, to ensure it gets reprogrammed. + */ + if (dm_new_crtc_state->base.color_mgmt_changed || + drm_atomic_crtc_needs_modeset(new_crtc_state)) { ret = amdgpu_dm_set_regamma_lut(dm_new_crtc_state); if (ret) goto fail; @@ -4755,7 +4841,8 @@ static int dm_update_planes_state(struct dc *dc, /* Remove any changed/removed planes */ if (!enable) { - if (pflip_needed) + if (pflip_needed && + plane->type != DRM_PLANE_TYPE_OVERLAY) continue; if (!old_plane_crtc) @@ -4802,7 +4889,8 @@ static int dm_update_planes_state(struct dc *dc, if (!dm_new_crtc_state->stream) continue; - if (pflip_needed) + if (pflip_needed && + plane->type != DRM_PLANE_TYPE_OVERLAY) continue; WARN_ON(dm_new_plane_state->dc_state); @@ -5009,17 +5097,24 @@ void amdgpu_dm_add_sink_to_freesync_module(struct drm_connector *connector, struct edid *edid) { int i; - uint64_t val_capable; bool edid_check_required; struct detailed_timing *timing; struct detailed_non_pixel *data; struct detailed_data_monitor_range *range; struct amdgpu_dm_connector *amdgpu_dm_connector = to_amdgpu_dm_connector(connector); + struct dm_connector_state *dm_con_state; struct drm_device *dev = connector->dev; struct amdgpu_device *adev = dev->dev_private; + if (!connector->state) { + DRM_ERROR("%s - Connector has no state", __func__); + return; + } + + dm_con_state = to_dm_connector_state(connector->state); + edid_check_required = false; if (!amdgpu_dm_connector->dc_sink) { DRM_ERROR("dc_sink NULL, could not add free_sync module.\n"); @@ -5038,7 +5133,7 @@ void amdgpu_dm_add_sink_to_freesync_module(struct drm_connector *connector, amdgpu_dm_connector); } } - val_capable = 0; + dm_con_state->freesync_capable = false; if (edid_check_required == true && (edid->version > 1 || (edid->version == 1 && edid->revision > 1))) { for (i = 0; i < 4; i++) { @@ -5074,7 +5169,7 @@ void amdgpu_dm_add_sink_to_freesync_module(struct drm_connector *connector, amdgpu_dm_connector->min_vfreq * 1000000; amdgpu_dm_connector->caps.max_refresh_in_micro_hz = amdgpu_dm_connector->max_vfreq * 1000000; - val_capable = 1; + dm_con_state->freesync_capable = true; } } diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h index b68400c1154b4..d5aa89ad5571d 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm.h @@ -28,7 +28,6 @@ #include <drm/drmP.h> #include <drm/drm_atomic.h> -#include "dc.h" /* * This file contains the definition for amdgpu_display_manager @@ -53,6 +52,7 @@ struct amdgpu_device; struct drm_device; struct amdgpu_dm_irq_handler_data; +struct dc; struct amdgpu_dm_prev_state { struct drm_framebuffer *fb; @@ -220,6 +220,7 @@ struct dm_connector_state { uint8_t underscan_hborder; bool underscan_enable; struct mod_freesync_user_enable user_enable; + bool freesync_capable; }; #define to_dm_connector_state(x)\ @@ -246,7 +247,7 @@ void amdgpu_dm_connector_init_helper(struct amdgpu_display_manager *dm, struct dc_link *link, int link_index); -int amdgpu_dm_connector_mode_valid(struct drm_connector *connector, +enum drm_mode_status amdgpu_dm_connector_mode_valid(struct drm_connector *connector, struct drm_display_mode *mode); void dm_restore_drm_connector_state(struct drm_device *dev, diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c index 25f064c010384..b329393307e59 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_color.c @@ -25,6 +25,7 @@ #include "amdgpu_mode.h" #include "amdgpu_dm.h" +#include "dc.h" #include "modules/color/color_gamma.h" #define MAX_DRM_LUT_VALUE 0xFFFF @@ -87,9 +88,9 @@ static void __drm_lut_to_dc_gamma(struct drm_color_lut *lut, g = drm_color_lut_extract(lut[i].green, 16); b = drm_color_lut_extract(lut[i].blue, 16); - gamma->entries.red[i] = dal_fixed31_32_from_int(r); - gamma->entries.green[i] = dal_fixed31_32_from_int(g); - gamma->entries.blue[i] = dal_fixed31_32_from_int(b); + gamma->entries.red[i] = dc_fixpt_from_int(r); + gamma->entries.green[i] = dc_fixpt_from_int(g); + gamma->entries.blue[i] = dc_fixpt_from_int(b); } return; } @@ -100,9 +101,9 @@ static void __drm_lut_to_dc_gamma(struct drm_color_lut *lut, g = drm_color_lut_extract(lut[i].green, 16); b = drm_color_lut_extract(lut[i].blue, 16); - gamma->entries.red[i] = dal_fixed31_32_from_fraction(r, MAX_DRM_LUT_VALUE); - gamma->entries.green[i] = dal_fixed31_32_from_fraction(g, MAX_DRM_LUT_VALUE); - gamma->entries.blue[i] = dal_fixed31_32_from_fraction(b, MAX_DRM_LUT_VALUE); + gamma->entries.red[i] = dc_fixpt_from_fraction(r, MAX_DRM_LUT_VALUE); + gamma->entries.green[i] = dc_fixpt_from_fraction(g, MAX_DRM_LUT_VALUE); + gamma->entries.blue[i] = dc_fixpt_from_fraction(b, MAX_DRM_LUT_VALUE); } } @@ -207,7 +208,7 @@ void amdgpu_dm_set_ctm(struct dm_crtc_state *crtc) for (i = 0; i < 12; i++) { /* Skip 4th element */ if (i % 4 == 3) { - stream->gamut_remap_matrix.matrix[i] = dal_fixed31_32_zero; + stream->gamut_remap_matrix.matrix[i] = dc_fixpt_zero; continue; } diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c index ca0b08bfa2cf6..bd449351803fb 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_helpers.c @@ -330,11 +330,6 @@ bool dm_helpers_dp_mst_send_payload_allocation( return true; } -bool dm_helpers_dc_conn_log(struct dc_context *ctx, struct log_entry *entry, enum dc_log_type event) -{ - return true; -} - void dm_dtn_log_begin(struct dc_context *ctx) {} diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c index ace9ad578ca08..4304d9e408b88 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_mst_types.c @@ -83,21 +83,22 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux, enum i2c_mot_mode mot = (msg->request & DP_AUX_I2C_MOT) ? I2C_MOT_TRUE : I2C_MOT_FALSE; enum ddc_result res; - ssize_t read_bytes; + uint32_t read_bytes = msg->size; if (WARN_ON(msg->size > 16)) return -E2BIG; switch (msg->request & ~DP_AUX_I2C_MOT) { case DP_AUX_NATIVE_READ: - read_bytes = dal_ddc_service_read_dpcd_data( + res = dal_ddc_service_read_dpcd_data( TO_DM_AUX(aux)->ddc_service, false, I2C_MOT_UNDEF, msg->address, msg->buffer, - msg->size); - return read_bytes; + msg->size, + &read_bytes); + break; case DP_AUX_NATIVE_WRITE: res = dal_ddc_service_write_dpcd_data( TO_DM_AUX(aux)->ddc_service, @@ -108,14 +109,15 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux, msg->size); break; case DP_AUX_I2C_READ: - read_bytes = dal_ddc_service_read_dpcd_data( + res = dal_ddc_service_read_dpcd_data( TO_DM_AUX(aux)->ddc_service, true, mot, msg->address, msg->buffer, - msg->size); - return read_bytes; + msg->size, + &read_bytes); + break; case DP_AUX_I2C_WRITE: res = dal_ddc_service_write_dpcd_data( TO_DM_AUX(aux)->ddc_service, @@ -137,7 +139,9 @@ static ssize_t dm_dp_aux_transfer(struct drm_dp_aux *aux, r == DDC_RESULT_SUCESSFULL); #endif - return msg->size; + if (res != DDC_RESULT_SUCESSFULL) + return -EIO; + return read_bytes; } static enum drm_connector_status diff --git a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c index 89342b48be6b4..0229c7edb8ad7 100644 --- a/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c +++ b/drivers/gpu/drm/amd/display/amdgpu_dm/amdgpu_dm_services.c @@ -37,8 +37,17 @@ unsigned long long dm_get_timestamp(struct dc_context *ctx) { - /* TODO: return actual timestamp */ - return 0; + struct timespec64 time; + + getrawmonotonic64(&time); + return timespec64_to_ns(&time); +} + +unsigned long long dm_get_elapse_time_in_ns(struct dc_context *ctx, + unsigned long long current_time_stamp, + unsigned long long last_time_stamp) +{ + return current_time_stamp - last_time_stamp; } void dm_perf_trace_timestamp(const char *func_name, unsigned int line) diff --git a/drivers/gpu/drm/amd/display/dc/basics/Makefile b/drivers/gpu/drm/amd/display/dc/basics/Makefile index bca33bd9a0d25..b49ea96b5daee 100644 --- a/drivers/gpu/drm/amd/display/dc/basics/Makefile +++ b/drivers/gpu/drm/amd/display/dc/basics/Makefile @@ -24,7 +24,7 @@ # It provides the general basic services required by other DAL # subcomponents. -BASICS = conversion.o fixpt31_32.o fixpt32_32.o \ +BASICS = conversion.o fixpt31_32.o \ logger.o log_helpers.o vector.o AMD_DAL_BASICS = $(addprefix $(AMDDALPATH)/dc/basics/,$(BASICS)) diff --git a/drivers/gpu/drm/amd/display/dc/basics/conversion.c b/drivers/gpu/drm/amd/display/dc/basics/conversion.c index 310964915a838..50b47f11875c9 100644 --- a/drivers/gpu/drm/amd/display/dc/basics/conversion.c +++ b/drivers/gpu/drm/amd/display/dc/basics/conversion.c @@ -41,22 +41,22 @@ uint16_t fixed_point_to_int_frac( uint16_t result; - uint16_t d = (uint16_t)dal_fixed31_32_floor( - dal_fixed31_32_abs( + uint16_t d = (uint16_t)dc_fixpt_floor( + dc_fixpt_abs( arg)); if (d <= (uint16_t)(1 << integer_bits) - (1 / (uint16_t)divisor)) - numerator = (uint16_t)dal_fixed31_32_round( - dal_fixed31_32_mul_int( + numerator = (uint16_t)dc_fixpt_round( + dc_fixpt_mul_int( arg, divisor)); else { - numerator = dal_fixed31_32_floor( - dal_fixed31_32_sub( - dal_fixed31_32_from_int( + numerator = dc_fixpt_floor( + dc_fixpt_sub( + dc_fixpt_from_int( 1LL << integer_bits), - dal_fixed31_32_recip( - dal_fixed31_32_from_int( + dc_fixpt_recip( + dc_fixpt_from_int( divisor)))); } @@ -66,8 +66,8 @@ uint16_t fixed_point_to_int_frac( result = (uint16_t)( (1 << (integer_bits + fractional_bits + 1)) + numerator); - if ((result != 0) && dal_fixed31_32_lt( - arg, dal_fixed31_32_zero)) + if ((result != 0) && dc_fixpt_lt( + arg, dc_fixpt_zero)) result |= 1 << (integer_bits + fractional_bits); return result; @@ -84,15 +84,15 @@ void convert_float_matrix( uint32_t buffer_size) { const struct fixed31_32 min_2_13 = - dal_fixed31_32_from_fraction(S2D13_MIN, DIVIDER); + dc_fixpt_from_fraction(S2D13_MIN, DIVIDER); const struct fixed31_32 max_2_13 = - dal_fixed31_32_from_fraction(S2D13_MAX, DIVIDER); + dc_fixpt_from_fraction(S2D13_MAX, DIVIDER); uint32_t i; for (i = 0; i < buffer_size; ++i) { uint32_t reg_value = fixed_point_to_int_frac( - dal_fixed31_32_clamp( + dc_fixpt_clamp( flt[i], min_2_13, max_2_13), diff --git a/drivers/gpu/drm/amd/display/dc/basics/fixpt31_32.c b/drivers/gpu/drm/amd/display/dc/basics/fixpt31_32.c index 8a9bba879207d..e61dd97d0928c 100644 --- a/drivers/gpu/drm/amd/display/dc/basics/fixpt31_32.c +++ b/drivers/gpu/drm/amd/display/dc/basics/fixpt31_32.c @@ -26,13 +26,13 @@ #include "dm_services.h" #include "include/fixed31_32.h" -static inline uint64_t abs_i64( - int64_t arg) +static inline unsigned long long abs_i64( + long long arg) { if (arg > 0) - return (uint64_t)arg; + return (unsigned long long)arg; else - return (uint64_t)(-arg); + return (unsigned long long)(-arg); } /* @@ -40,12 +40,12 @@ static inline uint64_t abs_i64( * result = dividend / divisor * *remainder = dividend % divisor */ -static inline uint64_t complete_integer_division_u64( - uint64_t dividend, - uint64_t divisor, - uint64_t *remainder) +static inline unsigned long long complete_integer_division_u64( + unsigned long long dividend, + unsigned long long divisor, + unsigned long long *remainder) { - uint64_t result; + unsigned long long result; ASSERT(divisor); @@ -64,30 +64,28 @@ static inline uint64_t complete_integer_division_u64( #define GET_FRACTIONAL_PART(x) \ (FRACTIONAL_PART_MASK & (x)) -struct fixed31_32 dal_fixed31_32_from_fraction( - int64_t numerator, - int64_t denominator) +struct fixed31_32 dc_fixpt_from_fraction(long long numerator, long long denominator) { struct fixed31_32 res; bool arg1_negative = numerator < 0; bool arg2_negative = denominator < 0; - uint64_t arg1_value = arg1_negative ? -numerator : numerator; - uint64_t arg2_value = arg2_negative ? -denominator : denominator; + unsigned long long arg1_value = arg1_negative ? -numerator : numerator; + unsigned long long arg2_value = arg2_negative ? -denominator : denominator; - uint64_t remainder; + unsigned long long remainder; /* determine integer part */ - uint64_t res_value = complete_integer_division_u64( + unsigned long long res_value = complete_integer_division_u64( arg1_value, arg2_value, &remainder); ASSERT(res_value <= LONG_MAX); /* determine fractional part */ { - uint32_t i = FIXED31_32_BITS_PER_FRACTIONAL_PART; + unsigned int i = FIXED31_32_BITS_PER_FRACTIONAL_PART; do { remainder <<= 1; @@ -103,14 +101,14 @@ struct fixed31_32 dal_fixed31_32_from_fraction( /* round up LSB */ { - uint64_t summand = (remainder << 1) >= arg2_value; + unsigned long long summand = (remainder << 1) >= arg2_value; ASSERT(res_value <= LLONG_MAX - summand); res_value += summand; } - res.value = (int64_t)res_value; + res.value = (long long)res_value; if (arg1_negative ^ arg2_negative) res.value = -res.value; @@ -118,79 +116,23 @@ struct fixed31_32 dal_fixed31_32_from_fraction( return res; } -struct fixed31_32 dal_fixed31_32_from_int_nonconst( - int64_t arg) -{ - struct fixed31_32 res; - - ASSERT((LONG_MIN <= arg) && (arg <= LONG_MAX)); - - res.value = arg << FIXED31_32_BITS_PER_FRACTIONAL_PART; - - return res; -} - -struct fixed31_32 dal_fixed31_32_shl( - struct fixed31_32 arg, - uint8_t shift) -{ - struct fixed31_32 res; - - ASSERT(((arg.value >= 0) && (arg.value <= LLONG_MAX >> shift)) || - ((arg.value < 0) && (arg.value >= LLONG_MIN >> shift))); - - res.value = arg.value << shift; - - return res; -} - -struct fixed31_32 dal_fixed31_32_add( - struct fixed31_32 arg1, - struct fixed31_32 arg2) -{ - struct fixed31_32 res; - - ASSERT(((arg1.value >= 0) && (LLONG_MAX - arg1.value >= arg2.value)) || - ((arg1.value < 0) && (LLONG_MIN - arg1.value <= arg2.value))); - - res.value = arg1.value + arg2.value; - - return res; -} - -struct fixed31_32 dal_fixed31_32_sub( - struct fixed31_32 arg1, - struct fixed31_32 arg2) -{ - struct fixed31_32 res; - - ASSERT(((arg2.value >= 0) && (LLONG_MIN + arg2.value <= arg1.value)) || - ((arg2.value < 0) && (LLONG_MAX + arg2.value >= arg1.value))); - - res.value = arg1.value - arg2.value; - - return res; -} - -struct fixed31_32 dal_fixed31_32_mul( - struct fixed31_32 arg1, - struct fixed31_32 arg2) +struct fixed31_32 dc_fixpt_mul(struct fixed31_32 arg1, struct fixed31_32 arg2) { struct fixed31_32 res; bool arg1_negative = arg1.value < 0; bool arg2_negative = arg2.value < 0; - uint64_t arg1_value = arg1_negative ? -arg1.value : arg1.value; - uint64_t arg2_value = arg2_negative ? -arg2.value : arg2.value; + unsigned long long arg1_value = arg1_negative ? -arg1.value : arg1.value; + unsigned long long arg2_value = arg2_negative ? -arg2.value : arg2.value; - uint64_t arg1_int = GET_INTEGER_PART(arg1_value); - uint64_t arg2_int = GET_INTEGER_PART(arg2_value); + unsigned long long arg1_int = GET_INTEGER_PART(arg1_value); + unsigned long long arg2_int = GET_INTEGER_PART(arg2_value); - uint64_t arg1_fra = GET_FRACTIONAL_PART(arg1_value); - uint64_t arg2_fra = GET_FRACTIONAL_PART(arg2_value); + unsigned long long arg1_fra = GET_FRACTIONAL_PART(arg1_value); + unsigned long long arg2_fra = GET_FRACTIONAL_PART(arg2_value); - uint64_t tmp; + unsigned long long tmp; res.value = arg1_int * arg2_int; @@ -200,22 +142,22 @@ struct fixed31_32 dal_fixed31_32_mul( tmp = arg1_int * arg2_fra; - ASSERT(tmp <= (uint64_t)(LLONG_MAX - res.value)); + ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value)); res.value += tmp; tmp = arg2_int * arg1_fra; - ASSERT(tmp <= (uint64_t)(LLONG_MAX - res.value)); + ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value)); res.value += tmp; tmp = arg1_fra * arg2_fra; tmp = (tmp >> FIXED31_32_BITS_PER_FRACTIONAL_PART) + - (tmp >= (uint64_t)dal_fixed31_32_half.value); + (tmp >= (unsigned long long)dc_fixpt_half.value); - ASSERT(tmp <= (uint64_t)(LLONG_MAX - res.value)); + ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value)); res.value += tmp; @@ -225,18 +167,17 @@ struct fixed31_32 dal_fixed31_32_mul( return res; } -struct fixed31_32 dal_fixed31_32_sqr( - struct fixed31_32 arg) +struct fixed31_32 dc_fixpt_sqr(struct fixed31_32 arg) { struct fixed31_32 res; - uint64_t arg_value = abs_i64(arg.value); + unsigned long long arg_value = abs_i64(arg.value); - uint64_t arg_int = GET_INTEGER_PART(arg_value); + unsigned long long arg_int = GET_INTEGER_PART(arg_value); - uint64_t arg_fra = GET_FRACTIONAL_PART(arg_value); + unsigned long long arg_fra = GET_FRACTIONAL_PART(arg_value); - uint64_t tmp; + unsigned long long tmp; res.value = arg_int * arg_int; @@ -246,28 +187,27 @@ struct fixed31_32 dal_fixed31_32_sqr( tmp = arg_int * arg_fra; - ASSERT(tmp <= (uint64_t)(LLONG_MAX - res.value)); + ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value)); res.value += tmp; - ASSERT(tmp <= (uint64_t)(LLONG_MAX - res.value)); + ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value)); res.value += tmp; tmp = arg_fra * arg_fra; tmp = (tmp >> FIXED31_32_BITS_PER_FRACTIONAL_PART) + - (tmp >= (uint64_t)dal_fixed31_32_half.value); + (tmp >= (unsigned long long)dc_fixpt_half.value); - ASSERT(tmp <= (uint64_t)(LLONG_MAX - res.value)); + ASSERT(tmp <= (unsigned long long)(LLONG_MAX - res.value)); res.value += tmp; return res; } -struct fixed31_32 dal_fixed31_32_recip( - struct fixed31_32 arg) +struct fixed31_32 dc_fixpt_recip(struct fixed31_32 arg) { /* * @note @@ -276,41 +216,40 @@ struct fixed31_32 dal_fixed31_32_recip( ASSERT(arg.value); - return dal_fixed31_32_from_fraction( - dal_fixed31_32_one.value, + return dc_fixpt_from_fraction( + dc_fixpt_one.value, arg.value); } -struct fixed31_32 dal_fixed31_32_sinc( - struct fixed31_32 arg) +struct fixed31_32 dc_fixpt_sinc(struct fixed31_32 arg) { struct fixed31_32 square; - struct fixed31_32 res = dal_fixed31_32_one; + struct fixed31_32 res = dc_fixpt_one; - int32_t n = 27; + int n = 27; struct fixed31_32 arg_norm = arg; - if (dal_fixed31_32_le( - dal_fixed31_32_two_pi, - dal_fixed31_32_abs(arg))) { - arg_norm = dal_fixed31_32_sub( + if (dc_fixpt_le( + dc_fixpt_two_pi, + dc_fixpt_abs(arg))) { + arg_norm = dc_fixpt_sub( arg_norm, - dal_fixed31_32_mul_int( - dal_fixed31_32_two_pi, - (int32_t)div64_s64( + dc_fixpt_mul_int( + dc_fixpt_two_pi, + (int)div64_s64( arg_norm.value, - dal_fixed31_32_two_pi.value))); + dc_fixpt_two_pi.value))); } - square = dal_fixed31_32_sqr(arg_norm); + square = dc_fixpt_sqr(arg_norm); do { - res = dal_fixed31_32_sub( - dal_fixed31_32_one, - dal_fixed31_32_div_int( - dal_fixed31_32_mul( + res = dc_fixpt_sub( + dc_fixpt_one, + dc_fixpt_div_int( + dc_fixpt_mul( square, res), n * (n - 1))); @@ -319,37 +258,35 @@ struct fixed31_32 dal_fixed31_32_sinc( } while (n > 2); if (arg.value != arg_norm.value) - res = dal_fixed31_32_div( - dal_fixed31_32_mul(res, arg_norm), + res = dc_fixpt_div( + dc_fixpt_mul(res, arg_norm), arg); return res; } -struct fixed31_32 dal_fixed31_32_sin( - struct fixed31_32 arg) +struct fixed31_32 dc_fixpt_sin(struct fixed31_32 arg) { - return dal_fixed31_32_mul( + return dc_fixpt_mul( arg, - dal_fixed31_32_sinc(arg)); + dc_fixpt_sinc(arg)); } -struct fixed31_32 dal_fixed31_32_cos( - struct fixed31_32 arg) +struct fixed31_32 dc_fixpt_cos(struct fixed31_32 arg) { /* TODO implement argument normalization */ - const struct fixed31_32 square = dal_fixed31_32_sqr(arg); + const struct fixed31_32 square = dc_fixpt_sqr(arg); - struct fixed31_32 res = dal_fixed31_32_one; + struct fixed31_32 res = dc_fixpt_one; - int32_t n = 26; + int n = 26; do { - res = dal_fixed31_32_sub( - dal_fixed31_32_one, - dal_fixed31_32_div_int( - dal_fixed31_32_mul( + res = dc_fixpt_sub( + dc_fixpt_one, + dc_fixpt_div_int( + dc_fixpt_mul( square, res), n * (n - 1))); @@ -367,37 +304,35 @@ struct fixed31_32 dal_fixed31_32_cos( * * Calculated as Taylor series. */ -static struct fixed31_32 fixed31_32_exp_from_taylor_series( - struct fixed31_32 arg) +static struct fixed31_32 fixed31_32_exp_from_taylor_series(struct fixed31_32 arg) { - uint32_t n = 9; + unsigned int n = 9; - struct fixed31_32 res = dal_fixed31_32_from_fraction( + struct fixed31_32 res = dc_fixpt_from_fraction( n + 2, n + 1); /* TODO find correct res */ - ASSERT(dal_fixed31_32_lt(arg, dal_fixed31_32_one)); + ASSERT(dc_fixpt_lt(arg, dc_fixpt_one)); do - res = dal_fixed31_32_add( - dal_fixed31_32_one, - dal_fixed31_32_div_int( - dal_fixed31_32_mul( + res = dc_fixpt_add( + dc_fixpt_one, + dc_fixpt_div_int( + dc_fixpt_mul( arg, res), n)); while (--n != 1); - return dal_fixed31_32_add( - dal_fixed31_32_one, - dal_fixed31_32_mul( + return dc_fixpt_add( + dc_fixpt_one, + dc_fixpt_mul( arg, res)); } -struct fixed31_32 dal_fixed31_32_exp( - struct fixed31_32 arg) +struct fixed31_32 dc_fixpt_exp(struct fixed31_32 arg) { /* * @brief @@ -406,44 +341,43 @@ struct fixed31_32 dal_fixed31_32_exp( * where m = round(x / ln(2)), r = x - m * ln(2) */ - if (dal_fixed31_32_le( - dal_fixed31_32_ln2_div_2, - dal_fixed31_32_abs(arg))) { - int32_t m = dal_fixed31_32_round( - dal_fixed31_32_div( + if (dc_fixpt_le( + dc_fixpt_ln2_div_2, + dc_fixpt_abs(arg))) { + int m = dc_fixpt_round( + dc_fixpt_div( arg, - dal_fixed31_32_ln2)); + dc_fixpt_ln2)); - struct fixed31_32 r = dal_fixed31_32_sub( + struct fixed31_32 r = dc_fixpt_sub( arg, - dal_fixed31_32_mul_int( - dal_fixed31_32_ln2, + dc_fixpt_mul_int( + dc_fixpt_ln2, m)); ASSERT(m != 0); - ASSERT(dal_fixed31_32_lt( - dal_fixed31_32_abs(r), - dal_fixed31_32_one)); + ASSERT(dc_fixpt_lt( + dc_fixpt_abs(r), + dc_fixpt_one)); if (m > 0) - return dal_fixed31_32_shl( + return dc_fixpt_shl( fixed31_32_exp_from_taylor_series(r), - (uint8_t)m); + (unsigned char)m); else - return dal_fixed31_32_div_int( + return dc_fixpt_div_int( fixed31_32_exp_from_taylor_series(r), 1LL << -m); } else if (arg.value != 0) return fixed31_32_exp_from_taylor_series(arg); else - return dal_fixed31_32_one; + return dc_fixpt_one; } -struct fixed31_32 dal_fixed31_32_log( - struct fixed31_32 arg) +struct fixed31_32 dc_fixpt_log(struct fixed31_32 arg) { - struct fixed31_32 res = dal_fixed31_32_neg(dal_fixed31_32_one); + struct fixed31_32 res = dc_fixpt_neg(dc_fixpt_one); /* TODO improve 1st estimation */ struct fixed31_32 error; @@ -453,15 +387,15 @@ struct fixed31_32 dal_fixed31_32_log( /* TODO if arg is zero, return -INF */ do { - struct fixed31_32 res1 = dal_fixed31_32_add( - dal_fixed31_32_sub( + struct fixed31_32 res1 = dc_fixpt_add( + dc_fixpt_sub( res, - dal_fixed31_32_one), - dal_fixed31_32_div( + dc_fixpt_one), + dc_fixpt_div( arg, - dal_fixed31_32_exp(res))); + dc_fixpt_exp(res))); - error = dal_fixed31_32_sub( + error = dc_fixpt_sub( res, res1); @@ -472,78 +406,23 @@ struct fixed31_32 dal_fixed31_32_log( return res; } -struct fixed31_32 dal_fixed31_32_pow( - struct fixed31_32 arg1, - struct fixed31_32 arg2) -{ - return dal_fixed31_32_exp( - dal_fixed31_32_mul( - dal_fixed31_32_log(arg1), - arg2)); -} - -int32_t dal_fixed31_32_floor( - struct fixed31_32 arg) -{ - uint64_t arg_value = abs_i64(arg.value); - - if (arg.value >= 0) - return (int32_t)GET_INTEGER_PART(arg_value); - else - return -(int32_t)GET_INTEGER_PART(arg_value); -} - -int32_t dal_fixed31_32_round( - struct fixed31_32 arg) -{ - uint64_t arg_value = abs_i64(arg.value); - - const int64_t summand = dal_fixed31_32_half.value; - - ASSERT(LLONG_MAX - (int64_t)arg_value >= summand); - - arg_value += summand; - - if (arg.value >= 0) - return (int32_t)GET_INTEGER_PART(arg_value); - else - return -(int32_t)GET_INTEGER_PART(arg_value); -} - -int32_t dal_fixed31_32_ceil( - struct fixed31_32 arg) -{ - uint64_t arg_value = abs_i64(arg.value); - - const int64_t summand = dal_fixed31_32_one.value - - dal_fixed31_32_epsilon.value; - - ASSERT(LLONG_MAX - (int64_t)arg_value >= summand); - - arg_value += summand; - - if (arg.value >= 0) - return (int32_t)GET_INTEGER_PART(arg_value); - else - return -(int32_t)GET_INTEGER_PART(arg_value); -} /* this function is a generic helper to translate fixed point value to * specified integer format that will consist of integer_bits integer part and * fractional_bits fractional part. For example it is used in - * dal_fixed31_32_u2d19 to receive 2 bits integer part and 19 bits fractional + * dc_fixpt_u2d19 to receive 2 bits integer part and 19 bits fractional * part in 32 bits. It is used in hw programming (scaler) */ -static inline uint32_t ux_dy( - int64_t value, - uint32_t integer_bits, - uint32_t fractional_bits) +static inline unsigned int ux_dy( + long long value, + unsigned int integer_bits, + unsigned int fractional_bits) { /* 1. create mask of integer part */ - uint32_t result = (1 << integer_bits) - 1; + unsigned int result = (1 << integer_bits) - 1; /* 2. mask out fractional part */ - uint32_t fractional_part = FRACTIONAL_PART_MASK & value; + unsigned int fractional_part = FRACTIONAL_PART_MASK & value; /* 3. shrink fixed point integer part to be of integer_bits width*/ result &= GET_INTEGER_PART(value); /* 4. make space for fractional part to be filled in after integer */ @@ -554,13 +433,13 @@ static inline uint32_t ux_dy( return result | fractional_part; } -static inline uint32_t clamp_ux_dy( - int64_t value, - uint32_t integer_bits, - uint32_t fractional_bits, - uint32_t min_clamp) +static inline unsigned int clamp_ux_dy( + long long value, + unsigned int integer_bits, + unsigned int fractional_bits, + unsigned int min_clamp) { - uint32_t truncated_val = ux_dy(value, integer_bits, fractional_bits); + unsigned int truncated_val = ux_dy(value, integer_bits, fractional_bits); if (value >= (1LL << (integer_bits + FIXED31_32_BITS_PER_FRACTIONAL_PART))) return (1 << (integer_bits + fractional_bits)) - 1; @@ -570,35 +449,30 @@ static inline uint32_t clamp_ux_dy( return min_clamp; } -uint32_t dal_fixed31_32_u2d19( - struct fixed31_32 arg) +unsigned int dc_fixpt_u2d19(struct fixed31_32 arg) { return ux_dy(arg.value, 2, 19); } -uint32_t dal_fixed31_32_u0d19( - struct fixed31_32 arg) +unsigned int dc_fixpt_u0d19(struct fixed31_32 arg) { return ux_dy(arg.value, 0, 19); } -uint32_t dal_fixed31_32_clamp_u0d14( - struct fixed31_32 arg) +unsigned int dc_fixpt_clamp_u0d14(struct fixed31_32 arg) { return clamp_ux_dy(arg.value, 0, 14, 1); } -uint32_t dal_fixed31_32_clamp_u0d10( - struct fixed31_32 arg) +unsigned int dc_fixpt_clamp_u0d10(struct fixed31_32 arg) { return clamp_ux_dy(arg.value, 0, 10, 1); } -int32_t dal_fixed31_32_s4d19( - struct fixed31_32 arg) +int dc_fixpt_s4d19(struct fixed31_32 arg) { if (arg.value < 0) - return -(int32_t)ux_dy(dal_fixed31_32_abs(arg).value, 4, 19); + return -(int)ux_dy(dc_fixpt_abs(arg).value, 4, 19); else return ux_dy(arg.value, 4, 19); } diff --git a/drivers/gpu/drm/amd/display/dc/basics/fixpt32_32.c b/drivers/gpu/drm/amd/display/dc/basics/fixpt32_32.c deleted file mode 100644 index 4d3aaa82a07b5..0000000000000 --- a/drivers/gpu/drm/amd/display/dc/basics/fixpt32_32.c +++ /dev/null @@ -1,161 +0,0 @@ -/* - * Copyright 2012-15 Advanced Micro Devices, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included in - * all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR - * OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: AMD - * - */ - -#include "dm_services.h" -#include "include/fixed32_32.h" - -static uint64_t u64_div(uint64_t n, uint64_t d) -{ - uint32_t i = 0; - uint64_t r; - uint64_t q = div64_u64_rem(n, d, &r); - - for (i = 0; i < 32; ++i) { - uint64_t sbit = q & (1ULL<<63); - - r <<= 1; - r |= sbit ? 1 : 0; - q <<= 1; - if (r >= d) { - r -= d; - q |= 1; - } - } - - if (2*r >= d) - q += 1; - return q; -} - -struct fixed32_32 dal_fixed32_32_from_fraction(uint32_t n, uint32_t d) -{ - struct fixed32_32 fx; - - fx.value = u64_div((uint64_t)n << 32, (uint64_t)d << 32); - return fx; -} - -struct fixed32_32 dal_fixed32_32_add( - struct fixed32_32 lhs, - struct fixed32_32 rhs) -{ - struct fixed32_32 fx = {lhs.value + rhs.value}; - - ASSERT(fx.value >= rhs.value); - return fx; -} - -struct fixed32_32 dal_fixed32_32_add_int(struct fixed32_32 lhs, uint32_t rhs) -{ - struct fixed32_32 fx = {lhs.value + ((uint64_t)rhs << 32)}; - - ASSERT(fx.value >= (uint64_t)rhs << 32); - return fx; - -} -struct fixed32_32 dal_fixed32_32_sub( - struct fixed32_32 lhs, - struct fixed32_32 rhs) -{ - struct fixed32_32 fx; - - ASSERT(lhs.value >= rhs.value); - fx.value = lhs.value - rhs.value; - return fx; -} - -struct fixed32_32 dal_fixed32_32_sub_int(struct fixed32_32 lhs, uint32_t rhs) -{ - struct fixed32_32 fx; - - ASSERT(lhs.value >= ((uint64_t)rhs<<32)); - fx.value = lhs.value - ((uint64_t)rhs<<32); - return fx; -} - -struct fixed32_32 dal_fixed32_32_mul( - struct fixed32_32 lhs, - struct fixed32_32 rhs) -{ - struct fixed32_32 fx; - uint64_t lhs_int = lhs.value>>32; - uint64_t lhs_frac = (uint32_t)lhs.value; - uint64_t rhs_int = rhs.value>>32; - uint64_t rhs_frac = (uint32_t)rhs.value; - uint64_t ahbh = lhs_int * rhs_int; - uint64_t ahbl = lhs_int * rhs_frac; - uint64_t albh = lhs_frac * rhs_int; - uint64_t albl = lhs_frac * rhs_frac; - - ASSERT((ahbh>>32) == 0); - - fx.value = (ahbh<<32) + ahbl + albh + (albl>>32); - return fx; - -} - -struct fixed32_32 dal_fixed32_32_mul_int(struct fixed32_32 lhs, uint32_t rhs) -{ - struct fixed32_32 fx; - uint64_t lhsi = (lhs.value>>32) * (uint64_t)rhs; - uint64_t lhsf; - - ASSERT((lhsi>>32) == 0); - lhsf = ((uint32_t)lhs.value) * (uint64_t)rhs; - ASSERT((lhsi<<32) + lhsf >= lhsf); - fx.value = (lhsi<<32) + lhsf; - return fx; -} - -struct fixed32_32 dal_fixed32_32_div( - struct fixed32_32 lhs, - struct fixed32_32 rhs) -{ - struct fixed32_32 fx; - - fx.value = u64_div(lhs.value, rhs.value); - return fx; -} - -struct fixed32_32 dal_fixed32_32_div_int(struct fixed32_32 lhs, uint32_t rhs) -{ - struct fixed32_32 fx; - - fx.value = u64_div(lhs.value, (uint64_t)rhs << 32); - return fx; -} - -uint32_t dal_fixed32_32_ceil(struct fixed32_32 v) -{ - ASSERT((uint32_t)v.value ? (v.value >> 32) + 1 >= 1 : true); - return (v.value>>32) + ((uint32_t)v.value ? 1 : 0); -} - -uint32_t dal_fixed32_32_round(struct fixed32_32 v) -{ - ASSERT(v.value + (1ULL<<31) >= (1ULL<<31)); - return (v.value + (1ULL<<31))>>32; -} - diff --git a/drivers/gpu/drm/amd/display/dc/basics/log_helpers.c b/drivers/gpu/drm/amd/display/dc/basics/log_helpers.c index 854678a0c54ba..021451549ff78 100644 --- a/drivers/gpu/drm/amd/display/dc/basics/log_helpers.c +++ b/drivers/gpu/drm/amd/display/dc/basics/log_helpers.c @@ -94,7 +94,6 @@ void dc_conn_log(struct dc_context *ctx, dm_logger_append(&entry, "%2.2X ", hex_data[i]); dm_logger_append(&entry, "^\n"); - dm_helpers_dc_conn_log(ctx, &entry, event); fail: dm_logger_close(&entry); diff --git a/drivers/gpu/drm/amd/display/dc/basics/logger.c b/drivers/gpu/drm/amd/display/dc/basics/logger.c index 31bee054f43a7..738a818d58d1c 100644 --- a/drivers/gpu/drm/amd/display/dc/basics/logger.c +++ b/drivers/gpu/drm/amd/display/dc/basics/logger.c @@ -61,7 +61,7 @@ static const struct dc_log_type_info log_type_info_tbl[] = { {LOG_EVENT_UNDERFLOW, "Underflow"}, {LOG_IF_TRACE, "InterfaceTrace"}, {LOG_DTN, "DTN"}, - {LOG_PROFILING, "Profiling"} + {LOG_DISPLAYSTATS, "DisplayStats"} }; @@ -402,3 +402,4 @@ cleanup: entry->max_buf_bytes = 0; } } + diff --git a/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c b/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c index 985fe8c22875a..b8cef7af3c4af 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c +++ b/drivers/gpu/drm/amd/display/dc/bios/bios_parser2.c @@ -70,6 +70,10 @@ static enum bp_result get_firmware_info_v3_1( struct bios_parser *bp, struct dc_firmware_info *info); +static enum bp_result get_firmware_info_v3_2( + struct bios_parser *bp, + struct dc_firmware_info *info); + static struct atom_hpd_int_record *get_hpd_record(struct bios_parser *bp, struct atom_display_object_path_v2 *object); @@ -1321,9 +1325,14 @@ static enum bp_result bios_parser_get_firmware_info( case 3: switch (revision.minor) { case 1: - case 2: result = get_firmware_info_v3_1(bp, info); break; + case 2: + result = get_firmware_info_v3_2(bp, info); + break; + case 3: + result = get_firmware_info_v3_2(bp, info); + break; default: break; } @@ -1383,6 +1392,84 @@ static enum bp_result get_firmware_info_v3_1( return BP_RESULT_OK; } +static enum bp_result get_firmware_info_v3_2( + struct bios_parser *bp, + struct dc_firmware_info *info) +{ + struct atom_firmware_info_v3_2 *firmware_info; + struct atom_display_controller_info_v4_1 *dce_info = NULL; + struct atom_common_table_header *header; + struct atom_data_revision revision; + struct atom_smu_info_v3_2 *smu_info_v3_2 = NULL; + struct atom_smu_info_v3_3 *smu_info_v3_3 = NULL; + + if (!info) + return BP_RESULT_BADINPUT; + + firmware_info = GET_IMAGE(struct atom_firmware_info_v3_2, + DATA_TABLES(firmwareinfo)); + + dce_info = GET_IMAGE(struct atom_display_controller_info_v4_1, + DATA_TABLES(dce_info)); + + if (!firmware_info || !dce_info) + return BP_RESULT_BADBIOSTABLE; + + memset(info, 0, sizeof(*info)); + + header = GET_IMAGE(struct atom_common_table_header, + DATA_TABLES(smu_info)); + get_atom_data_table_revision(header, &revision); + + if (revision.minor == 2) { + /* Vega12 */ + smu_info_v3_2 = GET_IMAGE(struct atom_smu_info_v3_2, + DATA_TABLES(smu_info)); + + if (!smu_info_v3_2) + return BP_RESULT_BADBIOSTABLE; + + info->default_engine_clk = smu_info_v3_2->bootup_dcefclk_10khz * 10; + } else if (revision.minor == 3) { + /* Vega20 */ + smu_info_v3_3 = GET_IMAGE(struct atom_smu_info_v3_3, + DATA_TABLES(smu_info)); + + if (!smu_info_v3_3) + return BP_RESULT_BADBIOSTABLE; + + info->default_engine_clk = smu_info_v3_3->bootup_dcefclk_10khz * 10; + } + + // We need to convert from 10KHz units into KHz units. + info->default_memory_clk = firmware_info->bootup_mclk_in10khz * 10; + + /* 27MHz for Vega10 & Vega12; 100MHz for Vega20 */ + info->pll_info.crystal_frequency = dce_info->dce_refclk_10khz * 10; + /* Hardcode frequency if BIOS gives no DCE Ref Clk */ + if (info->pll_info.crystal_frequency == 0) { + if (revision.minor == 2) + info->pll_info.crystal_frequency = 27000; + else if (revision.minor == 3) + info->pll_info.crystal_frequency = 100000; + } + /*dp_phy_ref_clk is not correct for atom_display_controller_info_v4_2, but we don't use it*/ + info->dp_phy_ref_clk = dce_info->dpphy_refclk_10khz * 10; + info->i2c_engine_ref_clk = dce_info->i2c_engine_refclk_10khz * 10; + + /* Get GPU PLL VCO Clock */ + if (bp->cmd_tbl.get_smu_clock_info != NULL) { + if (revision.minor == 2) + info->smu_gpu_pll_output_freq = + bp->cmd_tbl.get_smu_clock_info(bp, SMU9_SYSPLL0_ID) * 10; + else if (revision.minor == 3) + info->smu_gpu_pll_output_freq = + bp->cmd_tbl.get_smu_clock_info(bp, SMU11_SYSPLL3_0_ID) * 10; + } + + return BP_RESULT_OK; +} + static enum bp_result bios_parser_get_encoder_cap_info( struct dc_bios *dcb, struct graphics_object_id object_id, diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table.c b/drivers/gpu/drm/amd/display/dc/bios/command_table.c index 4b5fdd577848a..651e1fd4622f9 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/command_table.c +++ b/drivers/gpu/drm/amd/display/dc/bios/command_table.c @@ -24,7 +24,7 @@ */ #include "dm_services.h" - +#include "amdgpu.h" #include "atom.h" #include "include/bios_parser_interface.h" @@ -35,16 +35,16 @@ #include "bios_parser_types_internal.h" #define EXEC_BIOS_CMD_TABLE(command, params)\ - (cgs_atom_exec_cmd_table(bp->base.ctx->cgs_device, \ + (amdgpu_atom_execute_table(((struct amdgpu_device *)bp->base.ctx->driver_context)->mode_info.atom_context, \ GetIndexIntoMasterTable(COMMAND, command), \ - ¶ms) == 0) + (uint32_t *)¶ms) == 0) #define BIOS_CMD_TABLE_REVISION(command, frev, crev)\ - cgs_atom_get_cmd_table_revs(bp->base.ctx->cgs_device, \ + amdgpu_atom_parse_cmd_header(((struct amdgpu_device *)bp->base.ctx->driver_context)->mode_info.atom_context, \ GetIndexIntoMasterTable(COMMAND, command), &frev, &crev) #define BIOS_CMD_TABLE_PARA_REVISION(command)\ - bios_cmd_table_para_revision(bp->base.ctx->cgs_device, \ + bios_cmd_table_para_revision(bp->base.ctx->driver_context, \ GetIndexIntoMasterTable(COMMAND, command)) static void init_dig_encoder_control(struct bios_parser *bp); @@ -82,16 +82,18 @@ void dal_bios_parser_init_cmd_tbl(struct bios_parser *bp) init_set_dce_clock(bp); } -static uint32_t bios_cmd_table_para_revision(void *cgs_device, +static uint32_t bios_cmd_table_para_revision(void *dev, uint32_t index) { + struct amdgpu_device *adev = dev; uint8_t frev, crev; - if (cgs_atom_get_cmd_table_revs(cgs_device, + if (amdgpu_atom_parse_cmd_header(adev->mode_info.atom_context, index, - &frev, &crev) != 0) + &frev, &crev)) + return crev; + else return 0; - return crev; } /******************************************************************************* @@ -368,7 +370,7 @@ static void init_transmitter_control(struct bios_parser *bp) uint8_t crev; if (BIOS_CMD_TABLE_REVISION(UNIPHYTransmitterControl, - frev, crev) != 0) + frev, crev) == false) BREAK_TO_DEBUGGER(); switch (crev) { case 2: diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table2.c b/drivers/gpu/drm/amd/display/dc/bios/command_table2.c index 3f63f712c8a41..752b08a42d3ec 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/command_table2.c +++ b/drivers/gpu/drm/amd/display/dc/bios/command_table2.c @@ -26,14 +26,18 @@ #include "dm_services.h" #include "ObjectID.h" -#include "atomfirmware.h" +#include "atomfirmware.h" +#include "atom.h" #include "include/bios_parser_interface.h" #include "command_table2.h" #include "command_table_helper2.h" #include "bios_parser_helper.h" #include "bios_parser_types_internal2.h" +#include "amdgpu.h" + + #define DC_LOGGER \ bp->base.ctx->logger @@ -43,16 +47,16 @@ ->FieldName)-(char *)0)/sizeof(uint16_t)) #define EXEC_BIOS_CMD_TABLE(fname, params)\ - (cgs_atom_exec_cmd_table(bp->base.ctx->cgs_device, \ + (amdgpu_atom_execute_table(((struct amdgpu_device *)bp->base.ctx->driver_context)->mode_info.atom_context, \ GET_INDEX_INTO_MASTER_TABLE(command, fname), \ - ¶ms) == 0) + (uint32_t *)¶ms) == 0) #define BIOS_CMD_TABLE_REVISION(fname, frev, crev)\ - cgs_atom_get_cmd_table_revs(bp->base.ctx->cgs_device, \ + amdgpu_atom_parse_cmd_header(((struct amdgpu_device *)bp->base.ctx->driver_context)->mode_info.atom_context, \ GET_INDEX_INTO_MASTER_TABLE(command, fname), &frev, &crev) #define BIOS_CMD_TABLE_PARA_REVISION(fname)\ - bios_cmd_table_para_revision(bp->base.ctx->cgs_device, \ + bios_cmd_table_para_revision(bp->base.ctx->driver_context, \ GET_INDEX_INTO_MASTER_TABLE(command, fname)) static void init_dig_encoder_control(struct bios_parser *bp); @@ -86,16 +90,18 @@ void dal_firmware_parser_init_cmd_tbl(struct bios_parser *bp) init_get_smu_clock_info(bp); } -static uint32_t bios_cmd_table_para_revision(void *cgs_device, +static uint32_t bios_cmd_table_para_revision(void *dev, uint32_t index) { + struct amdgpu_device *adev = dev; uint8_t frev, crev; - if (cgs_atom_get_cmd_table_revs(cgs_device, + if (amdgpu_atom_parse_cmd_header(adev->mode_info.atom_context, index, - &frev, &crev) != 0) + &frev, &crev)) + return crev; + else return 0; - return crev; } /****************************************************************************** @@ -201,7 +207,7 @@ static void init_transmitter_control(struct bios_parser *bp) uint8_t frev; uint8_t crev; - if (BIOS_CMD_TABLE_REVISION(dig1transmittercontrol, frev, crev) != 0) + if (BIOS_CMD_TABLE_REVISION(dig1transmittercontrol, frev, crev) == false) BREAK_TO_DEBUGGER(); switch (crev) { case 6: diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.c b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.c index 2979358c6a55c..253bbb1eea609 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.c +++ b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper.c @@ -51,6 +51,7 @@ bool dal_bios_parser_init_cmd_tbl_helper( return true; case DCE_VERSION_11_2: + case DCE_VERSION_11_22: *h = dal_cmd_tbl_helper_dce112_get_table(); return true; diff --git a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c index 9a4d30dd49693..bbbcef566c551 100644 --- a/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c +++ b/drivers/gpu/drm/amd/display/dc/bios/command_table_helper2.c @@ -52,6 +52,7 @@ bool dal_bios_parser_init_cmd_tbl_helper2( return true; case DCE_VERSION_11_2: + case DCE_VERSION_11_22: *h = dal_cmd_tbl_helper_dce112_get_table2(); return true; #if defined(CONFIG_DRM_AMD_DC_DCN1_0) diff --git a/drivers/gpu/drm/amd/display/dc/calcs/calcs_logger.h b/drivers/gpu/drm/amd/display/dc/calcs/calcs_logger.h new file mode 100644 index 0000000000000..fc3f98fb09ead --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/calcs/calcs_logger.h @@ -0,0 +1,579 @@ +/* + * Copyright 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + +#ifndef _CALCS_CALCS_LOGGER_H_ +#define _CALCS_CALCS_LOGGER_H_ +#define DC_LOGGER \ + logger + +static void print_bw_calcs_dceip(struct dal_logger *logger, const struct bw_calcs_dceip *dceip) +{ + + DC_LOG_BANDWIDTH_CALCS("#####################################################################"); + DC_LOG_BANDWIDTH_CALCS("struct bw_calcs_dceip"); + DC_LOG_BANDWIDTH_CALCS("#####################################################################"); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_calcs_version version %d", dceip->version); + DC_LOG_BANDWIDTH_CALCS(" [bool] large_cursor: %d", dceip->large_cursor); + DC_LOG_BANDWIDTH_CALCS(" [bool] dmif_pipe_en_fbc_chunk_tracker: %d", dceip->dmif_pipe_en_fbc_chunk_tracker); + DC_LOG_BANDWIDTH_CALCS(" [bool] display_write_back_supported: %d", dceip->display_write_back_supported); + DC_LOG_BANDWIDTH_CALCS(" [bool] argb_compression_support: %d", dceip->argb_compression_support); + DC_LOG_BANDWIDTH_CALCS(" [bool] pre_downscaler_enabled: %d", dceip->pre_downscaler_enabled); + DC_LOG_BANDWIDTH_CALCS(" [bool] underlay_downscale_prefetch_enabled: %d", + dceip->underlay_downscale_prefetch_enabled); + DC_LOG_BANDWIDTH_CALCS(" [bool] graphics_lb_nodownscaling_multi_line_prefetching: %d", + dceip->graphics_lb_nodownscaling_multi_line_prefetching); + DC_LOG_BANDWIDTH_CALCS(" [bool] limit_excessive_outstanding_dmif_requests: %d", + dceip->limit_excessive_outstanding_dmif_requests); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] cursor_max_outstanding_group_num: %d", + dceip->cursor_max_outstanding_group_num); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] lines_interleaved_into_lb: %d", dceip->lines_interleaved_into_lb); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] low_power_tiling_mode: %d", dceip->low_power_tiling_mode); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] chunk_width: %d", dceip->chunk_width); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] number_of_graphics_pipes: %d", dceip->number_of_graphics_pipes); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] number_of_underlay_pipes: %d", dceip->number_of_underlay_pipes); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] max_dmif_buffer_allocated: %d", dceip->max_dmif_buffer_allocated); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] graphics_dmif_size: %d", dceip->graphics_dmif_size); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] underlay_luma_dmif_size: %d", dceip->underlay_luma_dmif_size); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] underlay_chroma_dmif_size: %d", dceip->underlay_chroma_dmif_size); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] scatter_gather_lines_of_pte_prefetching_in_linear_mode: %d", + dceip->scatter_gather_lines_of_pte_prefetching_in_linear_mode); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] display_write_back420_luma_mcifwr_buffer_size: %d", + dceip->display_write_back420_luma_mcifwr_buffer_size); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] display_write_back420_chroma_mcifwr_buffer_size: %d", + dceip->display_write_back420_chroma_mcifwr_buffer_size); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] scatter_gather_pte_request_rows_in_tiling_mode: %d", + dceip->scatter_gather_pte_request_rows_in_tiling_mode); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] underlay_vscaler_efficiency10_bit_per_component: %d", + bw_fixed_to_int(dceip->underlay_vscaler_efficiency10_bit_per_component)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] underlay_vscaler_efficiency12_bit_per_component: %d", + bw_fixed_to_int(dceip->underlay_vscaler_efficiency12_bit_per_component)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] graphics_vscaler_efficiency6_bit_per_component: %d", + bw_fixed_to_int(dceip->graphics_vscaler_efficiency6_bit_per_component)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] graphics_vscaler_efficiency8_bit_per_component: %d", + bw_fixed_to_int(dceip->graphics_vscaler_efficiency8_bit_per_component)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] graphics_vscaler_efficiency10_bit_per_component: %d", + bw_fixed_to_int(dceip->graphics_vscaler_efficiency10_bit_per_component)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] graphics_vscaler_efficiency12_bit_per_component: %d", + bw_fixed_to_int(dceip->graphics_vscaler_efficiency12_bit_per_component)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] alpha_vscaler_efficiency: %d", + bw_fixed_to_int(dceip->alpha_vscaler_efficiency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] lb_write_pixels_per_dispclk: %d", + bw_fixed_to_int(dceip->lb_write_pixels_per_dispclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] lb_size_per_component444: %d", + bw_fixed_to_int(dceip->lb_size_per_component444)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] stutter_and_dram_clock_state_change_gated_before_cursor: %d", + bw_fixed_to_int(dceip->stutter_and_dram_clock_state_change_gated_before_cursor)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] underlay420_luma_lb_size_per_component: %d", + bw_fixed_to_int(dceip->underlay420_luma_lb_size_per_component)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] underlay420_chroma_lb_size_per_component: %d", + bw_fixed_to_int(dceip->underlay420_chroma_lb_size_per_component)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] underlay422_lb_size_per_component: %d", + bw_fixed_to_int(dceip->underlay422_lb_size_per_component)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] cursor_chunk_width: %d", bw_fixed_to_int(dceip->cursor_chunk_width)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] cursor_dcp_buffer_lines: %d", + bw_fixed_to_int(dceip->cursor_dcp_buffer_lines)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] underlay_maximum_width_efficient_for_tiling: %d", + bw_fixed_to_int(dceip->underlay_maximum_width_efficient_for_tiling)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] underlay_maximum_height_efficient_for_tiling: %d", + bw_fixed_to_int(dceip->underlay_maximum_height_efficient_for_tiling)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] peak_pte_request_to_eviction_ratio_limiting_multiple_displays_or_single_rotated_display: %d", + bw_fixed_to_int(dceip->peak_pte_request_to_eviction_ratio_limiting_multiple_displays_or_single_rotated_display)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] peak_pte_request_to_eviction_ratio_limiting_single_display_no_rotation: %d", + bw_fixed_to_int(dceip->peak_pte_request_to_eviction_ratio_limiting_single_display_no_rotation)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] minimum_outstanding_pte_request_limit: %d", + bw_fixed_to_int(dceip->minimum_outstanding_pte_request_limit)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] maximum_total_outstanding_pte_requests_allowed_by_saw: %d", + bw_fixed_to_int(dceip->maximum_total_outstanding_pte_requests_allowed_by_saw)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] linear_mode_line_request_alternation_slice: %d", + bw_fixed_to_int(dceip->linear_mode_line_request_alternation_slice)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] request_efficiency: %d", bw_fixed_to_int(dceip->request_efficiency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dispclk_per_request: %d", bw_fixed_to_int(dceip->dispclk_per_request)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dispclk_ramping_factor: %d", + bw_fixed_to_int(dceip->dispclk_ramping_factor)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] display_pipe_throughput_factor: %d", + bw_fixed_to_int(dceip->display_pipe_throughput_factor)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mcifwr_all_surfaces_burst_time: %d", + bw_fixed_to_int(dceip->mcifwr_all_surfaces_burst_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmif_request_buffer_size: %d", + bw_fixed_to_int(dceip->dmif_request_buffer_size)); + + +} + +static void print_bw_calcs_vbios(struct dal_logger *logger, const struct bw_calcs_vbios *vbios) +{ + + DC_LOG_BANDWIDTH_CALCS("#####################################################################"); + DC_LOG_BANDWIDTH_CALCS("struct bw_calcs_vbios vbios"); + DC_LOG_BANDWIDTH_CALCS("#####################################################################"); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines memory_type: %d", vbios->memory_type); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines memory_type: %d", vbios->memory_type); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] dram_channel_width_in_bits: %d", vbios->dram_channel_width_in_bits); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] number_of_dram_channels: %d", vbios->number_of_dram_channels); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] number_of_dram_banks: %d", vbios->number_of_dram_banks); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] low_yclk: %d", bw_fixed_to_int(vbios->low_yclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mid_yclk: %d", bw_fixed_to_int(vbios->mid_yclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] high_yclk: %d", bw_fixed_to_int(vbios->high_yclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] low_sclk: %d", bw_fixed_to_int(vbios->low_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mid1_sclk: %d", bw_fixed_to_int(vbios->mid1_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mid2_sclk: %d", bw_fixed_to_int(vbios->mid2_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mid3_sclk: %d", bw_fixed_to_int(vbios->mid3_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mid4_sclk: %d", bw_fixed_to_int(vbios->mid4_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mid5_sclk: %d", bw_fixed_to_int(vbios->mid5_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mid6_sclk: %d", bw_fixed_to_int(vbios->mid6_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] high_sclk: %d", bw_fixed_to_int(vbios->high_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] low_voltage_max_dispclk: %d", + bw_fixed_to_int(vbios->low_voltage_max_dispclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mid_voltage_max_dispclk;: %d", + bw_fixed_to_int(vbios->mid_voltage_max_dispclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] high_voltage_max_dispclk;: %d", + bw_fixed_to_int(vbios->high_voltage_max_dispclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] low_voltage_max_phyclk: %d", + bw_fixed_to_int(vbios->low_voltage_max_phyclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mid_voltage_max_phyclk: %d", + bw_fixed_to_int(vbios->mid_voltage_max_phyclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] high_voltage_max_phyclk: %d", + bw_fixed_to_int(vbios->high_voltage_max_phyclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] data_return_bus_width: %d", bw_fixed_to_int(vbios->data_return_bus_width)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] trc: %d", bw_fixed_to_int(vbios->trc)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmifmc_urgent_latency: %d", bw_fixed_to_int(vbios->dmifmc_urgent_latency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] stutter_self_refresh_exit_latency: %d", + bw_fixed_to_int(vbios->stutter_self_refresh_exit_latency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] stutter_self_refresh_entry_latency: %d", + bw_fixed_to_int(vbios->stutter_self_refresh_entry_latency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] nbp_state_change_latency: %d", + bw_fixed_to_int(vbios->nbp_state_change_latency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mcifwrmc_urgent_latency: %d", + bw_fixed_to_int(vbios->mcifwrmc_urgent_latency)); + DC_LOG_BANDWIDTH_CALCS(" [bool] scatter_gather_enable: %d", vbios->scatter_gather_enable); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] down_spread_percentage: %d", + bw_fixed_to_int(vbios->down_spread_percentage)); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] cursor_width: %d", vbios->cursor_width); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] average_compression_rate: %d", vbios->average_compression_rate); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] number_of_request_slots_gmc_reserves_for_dmif_per_channel: %d", + vbios->number_of_request_slots_gmc_reserves_for_dmif_per_channel); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] blackout_duration: %d", bw_fixed_to_int(vbios->blackout_duration)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] maximum_blackout_recovery_time: %d", + bw_fixed_to_int(vbios->maximum_blackout_recovery_time)); + + +} + +static void print_bw_calcs_data(struct dal_logger *logger, struct bw_calcs_data *data) +{ + + int i, j, k; + + DC_LOG_BANDWIDTH_CALCS("#####################################################################"); + DC_LOG_BANDWIDTH_CALCS("struct bw_calcs_data data"); + DC_LOG_BANDWIDTH_CALCS("#####################################################################"); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] number_of_displays: %d", data->number_of_displays); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines underlay_surface_type: %d", data->underlay_surface_type); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines panning_and_bezel_adjustment: %d", + data->panning_and_bezel_adjustment); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines graphics_tiling_mode: %d", data->graphics_tiling_mode); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] graphics_lb_bpc: %d", data->graphics_lb_bpc); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] underlay_lb_bpc: %d", data->underlay_lb_bpc); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines underlay_tiling_mode: %d", data->underlay_tiling_mode); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines d0_underlay_mode: %d", data->d0_underlay_mode); + DC_LOG_BANDWIDTH_CALCS(" [bool] d1_display_write_back_dwb_enable: %d", data->d1_display_write_back_dwb_enable); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines d1_underlay_mode: %d", data->d1_underlay_mode); + DC_LOG_BANDWIDTH_CALCS(" [bool] cpup_state_change_enable: %d", data->cpup_state_change_enable); + DC_LOG_BANDWIDTH_CALCS(" [bool] cpuc_state_change_enable: %d", data->cpuc_state_change_enable); + DC_LOG_BANDWIDTH_CALCS(" [bool] nbp_state_change_enable: %d", data->nbp_state_change_enable); + DC_LOG_BANDWIDTH_CALCS(" [bool] stutter_mode_enable: %d", data->stutter_mode_enable); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] y_clk_level: %d", data->y_clk_level); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] sclk_level: %d", data->sclk_level); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] number_of_underlay_surfaces: %d", data->number_of_underlay_surfaces); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] number_of_dram_wrchannels: %d", data->number_of_dram_wrchannels); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] chunk_request_delay: %d", data->chunk_request_delay); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] number_of_dram_channels: %d", data->number_of_dram_channels); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines underlay_micro_tile_mode: %d", data->underlay_micro_tile_mode); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines graphics_micro_tile_mode: %d", data->graphics_micro_tile_mode); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] max_phyclk: %d", bw_fixed_to_int(data->max_phyclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dram_efficiency: %d", bw_fixed_to_int(data->dram_efficiency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] src_width_after_surface_type: %d", + bw_fixed_to_int(data->src_width_after_surface_type)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] src_height_after_surface_type: %d", + bw_fixed_to_int(data->src_height_after_surface_type)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] hsr_after_surface_type: %d", + bw_fixed_to_int(data->hsr_after_surface_type)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] vsr_after_surface_type: %d", bw_fixed_to_int(data->vsr_after_surface_type)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] src_width_after_rotation: %d", + bw_fixed_to_int(data->src_width_after_rotation)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] src_height_after_rotation: %d", + bw_fixed_to_int(data->src_height_after_rotation)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] hsr_after_rotation: %d", bw_fixed_to_int(data->hsr_after_rotation)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] vsr_after_rotation: %d", bw_fixed_to_int(data->vsr_after_rotation)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] source_height_pixels: %d", bw_fixed_to_int(data->source_height_pixels)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] hsr_after_stereo: %d", bw_fixed_to_int(data->hsr_after_stereo)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] vsr_after_stereo: %d", bw_fixed_to_int(data->vsr_after_stereo)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] source_width_in_lb: %d", bw_fixed_to_int(data->source_width_in_lb)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] lb_line_pitch: %d", bw_fixed_to_int(data->lb_line_pitch)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] underlay_maximum_source_efficient_for_tiling: %d", + bw_fixed_to_int(data->underlay_maximum_source_efficient_for_tiling)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] num_lines_at_frame_start: %d", + bw_fixed_to_int(data->num_lines_at_frame_start)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] min_dmif_size_in_time: %d", bw_fixed_to_int(data->min_dmif_size_in_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] min_mcifwr_size_in_time: %d", + bw_fixed_to_int(data->min_mcifwr_size_in_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_requests_for_dmif_size: %d", + bw_fixed_to_int(data->total_requests_for_dmif_size)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] peak_pte_request_to_eviction_ratio_limiting: %d", + bw_fixed_to_int(data->peak_pte_request_to_eviction_ratio_limiting)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] useful_pte_per_pte_request: %d", + bw_fixed_to_int(data->useful_pte_per_pte_request)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scatter_gather_pte_request_rows: %d", + bw_fixed_to_int(data->scatter_gather_pte_request_rows)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scatter_gather_row_height: %d", + bw_fixed_to_int(data->scatter_gather_row_height)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scatter_gather_pte_requests_in_vblank: %d", + bw_fixed_to_int(data->scatter_gather_pte_requests_in_vblank)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] inefficient_linear_pitch_in_bytes: %d", + bw_fixed_to_int(data->inefficient_linear_pitch_in_bytes)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] cursor_total_data: %d", bw_fixed_to_int(data->cursor_total_data)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] cursor_total_request_groups: %d", + bw_fixed_to_int(data->cursor_total_request_groups)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scatter_gather_total_pte_requests: %d", + bw_fixed_to_int(data->scatter_gather_total_pte_requests)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scatter_gather_total_pte_request_groups: %d", + bw_fixed_to_int(data->scatter_gather_total_pte_request_groups)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] tile_width_in_pixels: %d", bw_fixed_to_int(data->tile_width_in_pixels)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmif_total_number_of_data_request_page_close_open: %d", + bw_fixed_to_int(data->dmif_total_number_of_data_request_page_close_open)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mcifwr_total_number_of_data_request_page_close_open: %d", + bw_fixed_to_int(data->mcifwr_total_number_of_data_request_page_close_open)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] bytes_per_page_close_open: %d", + bw_fixed_to_int(data->bytes_per_page_close_open)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mcifwr_total_page_close_open_time: %d", + bw_fixed_to_int(data->mcifwr_total_page_close_open_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_requests_for_adjusted_dmif_size: %d", + bw_fixed_to_int(data->total_requests_for_adjusted_dmif_size)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_dmifmc_urgent_trips: %d", + bw_fixed_to_int(data->total_dmifmc_urgent_trips)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_dmifmc_urgent_latency: %d", + bw_fixed_to_int(data->total_dmifmc_urgent_latency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_display_reads_required_data: %d", + bw_fixed_to_int(data->total_display_reads_required_data)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_display_reads_required_dram_access_data: %d", + bw_fixed_to_int(data->total_display_reads_required_dram_access_data)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_display_writes_required_data: %d", + bw_fixed_to_int(data->total_display_writes_required_data)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_display_writes_required_dram_access_data: %d", + bw_fixed_to_int(data->total_display_writes_required_dram_access_data)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] display_reads_required_data: %d", + bw_fixed_to_int(data->display_reads_required_data)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] display_reads_required_dram_access_data: %d", + bw_fixed_to_int(data->display_reads_required_dram_access_data)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmif_total_page_close_open_time: %d", + bw_fixed_to_int(data->dmif_total_page_close_open_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] min_cursor_memory_interface_buffer_size_in_time: %d", + bw_fixed_to_int(data->min_cursor_memory_interface_buffer_size_in_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] min_read_buffer_size_in_time: %d", + bw_fixed_to_int(data->min_read_buffer_size_in_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] display_reads_time_for_data_transfer: %d", + bw_fixed_to_int(data->display_reads_time_for_data_transfer)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] display_writes_time_for_data_transfer: %d", + bw_fixed_to_int(data->display_writes_time_for_data_transfer)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmif_required_dram_bandwidth: %d", + bw_fixed_to_int(data->dmif_required_dram_bandwidth)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mcifwr_required_dram_bandwidth: %d", + bw_fixed_to_int(data->mcifwr_required_dram_bandwidth)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] required_dmifmc_urgent_latency_for_page_close_open: %d", + bw_fixed_to_int(data->required_dmifmc_urgent_latency_for_page_close_open)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] required_mcifmcwr_urgent_latency: %d", + bw_fixed_to_int(data->required_mcifmcwr_urgent_latency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] required_dram_bandwidth_gbyte_per_second: %d", + bw_fixed_to_int(data->required_dram_bandwidth_gbyte_per_second)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dram_bandwidth: %d", bw_fixed_to_int(data->dram_bandwidth)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmif_required_sclk: %d", bw_fixed_to_int(data->dmif_required_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mcifwr_required_sclk: %d", bw_fixed_to_int(data->mcifwr_required_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] required_sclk: %d", bw_fixed_to_int(data->required_sclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] downspread_factor: %d", bw_fixed_to_int(data->downspread_factor)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] v_scaler_efficiency: %d", bw_fixed_to_int(data->v_scaler_efficiency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scaler_limits_factor: %d", bw_fixed_to_int(data->scaler_limits_factor)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] display_pipe_pixel_throughput: %d", + bw_fixed_to_int(data->display_pipe_pixel_throughput)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_dispclk_required_with_ramping: %d", + bw_fixed_to_int(data->total_dispclk_required_with_ramping)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_dispclk_required_without_ramping: %d", + bw_fixed_to_int(data->total_dispclk_required_without_ramping)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_read_request_bandwidth: %d", + bw_fixed_to_int(data->total_read_request_bandwidth)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_write_request_bandwidth: %d", + bw_fixed_to_int(data->total_write_request_bandwidth)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dispclk_required_for_total_read_request_bandwidth: %d", + bw_fixed_to_int(data->dispclk_required_for_total_read_request_bandwidth)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_dispclk_required_with_ramping_with_request_bandwidth: %d", + bw_fixed_to_int(data->total_dispclk_required_with_ramping_with_request_bandwidth)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_dispclk_required_without_ramping_with_request_bandwidth: %d", + bw_fixed_to_int(data->total_dispclk_required_without_ramping_with_request_bandwidth)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dispclk: %d", bw_fixed_to_int(data->dispclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] blackout_recovery_time: %d", bw_fixed_to_int(data->blackout_recovery_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] min_pixels_per_data_fifo_entry: %d", + bw_fixed_to_int(data->min_pixels_per_data_fifo_entry)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] sclk_deep_sleep: %d", bw_fixed_to_int(data->sclk_deep_sleep)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] chunk_request_time: %d", bw_fixed_to_int(data->chunk_request_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] cursor_request_time: %d", bw_fixed_to_int(data->cursor_request_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] line_source_pixels_transfer_time: %d", + bw_fixed_to_int(data->line_source_pixels_transfer_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmifdram_access_efficiency: %d", + bw_fixed_to_int(data->dmifdram_access_efficiency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mcifwrdram_access_efficiency: %d", + bw_fixed_to_int(data->mcifwrdram_access_efficiency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_average_bandwidth_no_compression: %d", + bw_fixed_to_int(data->total_average_bandwidth_no_compression)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_average_bandwidth: %d", + bw_fixed_to_int(data->total_average_bandwidth)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] total_stutter_cycle_duration: %d", + bw_fixed_to_int(data->total_stutter_cycle_duration)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] stutter_burst_time: %d", bw_fixed_to_int(data->stutter_burst_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] time_in_self_refresh: %d", bw_fixed_to_int(data->time_in_self_refresh)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] stutter_efficiency: %d", bw_fixed_to_int(data->stutter_efficiency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] worst_number_of_trips_to_memory: %d", + bw_fixed_to_int(data->worst_number_of_trips_to_memory)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] immediate_flip_time: %d", bw_fixed_to_int(data->immediate_flip_time)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] latency_for_non_dmif_clients: %d", + bw_fixed_to_int(data->latency_for_non_dmif_clients)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] latency_for_non_mcifwr_clients: %d", + bw_fixed_to_int(data->latency_for_non_mcifwr_clients)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmifmc_urgent_latency_supported_in_high_sclk_and_yclk: %d", + bw_fixed_to_int(data->dmifmc_urgent_latency_supported_in_high_sclk_and_yclk)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] nbp_state_dram_speed_change_margin: %d", + bw_fixed_to_int(data->nbp_state_dram_speed_change_margin)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] display_reads_time_for_data_transfer_and_urgent_latency: %d", + bw_fixed_to_int(data->display_reads_time_for_data_transfer_and_urgent_latency)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dram_speed_change_margin: %d", + bw_fixed_to_int(data->dram_speed_change_margin)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] min_vblank_dram_speed_change_margin: %d", + bw_fixed_to_int(data->min_vblank_dram_speed_change_margin)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] min_stutter_refresh_duration: %d", + bw_fixed_to_int(data->min_stutter_refresh_duration)); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] total_stutter_dmif_buffer_size: %d", data->total_stutter_dmif_buffer_size); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] total_bytes_requested: %d", data->total_bytes_requested); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] min_stutter_dmif_buffer_size: %d", data->min_stutter_dmif_buffer_size); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] num_stutter_bursts: %d", data->num_stutter_bursts); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] v_blank_nbp_state_dram_speed_change_latency_supported: %d", + bw_fixed_to_int(data->v_blank_nbp_state_dram_speed_change_latency_supported)); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] nbp_state_dram_speed_change_latency_supported: %d", + bw_fixed_to_int(data->nbp_state_dram_speed_change_latency_supported)); + + for (i = 0; i < maximum_number_of_surfaces; i++) { + DC_LOG_BANDWIDTH_CALCS(" [bool] fbc_en[%d]:%d\n", i, data->fbc_en[i]); + DC_LOG_BANDWIDTH_CALCS(" [bool] lpt_en[%d]:%d", i, data->lpt_en[i]); + DC_LOG_BANDWIDTH_CALCS(" [bool] displays_match_flag[%d]:%d", i, data->displays_match_flag[i]); + DC_LOG_BANDWIDTH_CALCS(" [bool] use_alpha[%d]:%d", i, data->use_alpha[i]); + DC_LOG_BANDWIDTH_CALCS(" [bool] orthogonal_rotation[%d]:%d", i, data->orthogonal_rotation[i]); + DC_LOG_BANDWIDTH_CALCS(" [bool] enable[%d]:%d", i, data->enable[i]); + DC_LOG_BANDWIDTH_CALCS(" [bool] access_one_channel_only[%d]:%d", i, data->access_one_channel_only[i]); + DC_LOG_BANDWIDTH_CALCS(" [bool] scatter_gather_enable_for_pipe[%d]:%d", + i, data->scatter_gather_enable_for_pipe[i]); + DC_LOG_BANDWIDTH_CALCS(" [bool] interlace_mode[%d]:%d", + i, data->interlace_mode[i]); + DC_LOG_BANDWIDTH_CALCS(" [bool] display_pstate_change_enable[%d]:%d", + i, data->display_pstate_change_enable[i]); + DC_LOG_BANDWIDTH_CALCS(" [bool] line_buffer_prefetch[%d]:%d", i, data->line_buffer_prefetch[i]); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] bytes_per_pixel[%d]:%d", i, data->bytes_per_pixel[i]); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] max_chunks_non_fbc_mode[%d]:%d", + i, data->max_chunks_non_fbc_mode[i]); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] lb_bpc[%d]:%d", i, data->lb_bpc[i]); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] output_bpphdmi[%d]:%d", i, data->output_bpphdmi[i]); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] output_bppdp4_lane_hbr[%d]:%d", i, data->output_bppdp4_lane_hbr[i]); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] output_bppdp4_lane_hbr2[%d]:%d", + i, data->output_bppdp4_lane_hbr2[i]); + DC_LOG_BANDWIDTH_CALCS(" [uint32_t] output_bppdp4_lane_hbr3[%d]:%d", + i, data->output_bppdp4_lane_hbr3[i]); + DC_LOG_BANDWIDTH_CALCS(" [enum] bw_defines stereo_mode[%d]:%d", i, data->stereo_mode[i]); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmif_buffer_transfer_time[%d]:%d", + i, bw_fixed_to_int(data->dmif_buffer_transfer_time[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] displays_with_same_mode[%d]:%d", + i, bw_fixed_to_int(data->displays_with_same_mode[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] stutter_dmif_buffer_size[%d]:%d", + i, bw_fixed_to_int(data->stutter_dmif_buffer_size[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] stutter_refresh_duration[%d]:%d", + i, bw_fixed_to_int(data->stutter_refresh_duration[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] stutter_exit_watermark[%d]:%d", + i, bw_fixed_to_int(data->stutter_exit_watermark[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] stutter_entry_watermark[%d]:%d", + i, bw_fixed_to_int(data->stutter_entry_watermark[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] h_total[%d]:%d", i, bw_fixed_to_int(data->h_total[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] v_total[%d]:%d", i, bw_fixed_to_int(data->v_total[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] pixel_rate[%d]:%d", i, bw_fixed_to_int(data->pixel_rate[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] src_width[%d]:%d", i, bw_fixed_to_int(data->src_width[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] pitch_in_pixels[%d]:%d", + i, bw_fixed_to_int(data->pitch_in_pixels[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] pitch_in_pixels_after_surface_type[%d]:%d", + i, bw_fixed_to_int(data->pitch_in_pixels_after_surface_type[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] src_height[%d]:%d", i, bw_fixed_to_int(data->src_height[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scale_ratio[%d]:%d", i, bw_fixed_to_int(data->scale_ratio[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] h_taps[%d]:%d", i, bw_fixed_to_int(data->h_taps[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] v_taps[%d]:%d", i, bw_fixed_to_int(data->v_taps[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] h_scale_ratio[%d]:%d", i, bw_fixed_to_int(data->h_scale_ratio[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] v_scale_ratio[%d]:%d", i, bw_fixed_to_int(data->v_scale_ratio[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] rotation_angle[%d]:%d", + i, bw_fixed_to_int(data->rotation_angle[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] compression_rate[%d]:%d", + i, bw_fixed_to_int(data->compression_rate[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] hsr[%d]:%d", i, bw_fixed_to_int(data->hsr[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] vsr[%d]:%d", i, bw_fixed_to_int(data->vsr[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] source_width_rounded_up_to_chunks[%d]:%d", + i, bw_fixed_to_int(data->source_width_rounded_up_to_chunks[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] source_width_pixels[%d]:%d", + i, bw_fixed_to_int(data->source_width_pixels[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] source_height_rounded_up_to_chunks[%d]:%d", + i, bw_fixed_to_int(data->source_height_rounded_up_to_chunks[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] display_bandwidth[%d]:%d", + i, bw_fixed_to_int(data->display_bandwidth[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] request_bandwidth[%d]:%d", + i, bw_fixed_to_int(data->request_bandwidth[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] bytes_per_request[%d]:%d", + i, bw_fixed_to_int(data->bytes_per_request[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] useful_bytes_per_request[%d]:%d", + i, bw_fixed_to_int(data->useful_bytes_per_request[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] lines_interleaved_in_mem_access[%d]:%d", + i, bw_fixed_to_int(data->lines_interleaved_in_mem_access[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] latency_hiding_lines[%d]:%d", + i, bw_fixed_to_int(data->latency_hiding_lines[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] lb_partitions[%d]:%d", + i, bw_fixed_to_int(data->lb_partitions[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] lb_partitions_max[%d]:%d", + i, bw_fixed_to_int(data->lb_partitions_max[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dispclk_required_with_ramping[%d]:%d", + i, bw_fixed_to_int(data->dispclk_required_with_ramping[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dispclk_required_without_ramping[%d]:%d", + i, bw_fixed_to_int(data->dispclk_required_without_ramping[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] data_buffer_size[%d]:%d", + i, bw_fixed_to_int(data->data_buffer_size[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] outstanding_chunk_request_limit[%d]:%d", + i, bw_fixed_to_int(data->outstanding_chunk_request_limit[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] urgent_watermark[%d]:%d", + i, bw_fixed_to_int(data->urgent_watermark[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] nbp_state_change_watermark[%d]:%d", + i, bw_fixed_to_int(data->nbp_state_change_watermark[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] v_filter_init[%d]:%d", i, bw_fixed_to_int(data->v_filter_init[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] stutter_cycle_duration[%d]:%d", + i, bw_fixed_to_int(data->stutter_cycle_duration[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] average_bandwidth[%d]:%d", + i, bw_fixed_to_int(data->average_bandwidth[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] average_bandwidth_no_compression[%d]:%d", + i, bw_fixed_to_int(data->average_bandwidth_no_compression[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scatter_gather_pte_request_limit[%d]:%d", + i, bw_fixed_to_int(data->scatter_gather_pte_request_limit[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] lb_size_per_component[%d]:%d", + i, bw_fixed_to_int(data->lb_size_per_component[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] memory_chunk_size_in_bytes[%d]:%d", + i, bw_fixed_to_int(data->memory_chunk_size_in_bytes[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] pipe_chunk_size_in_bytes[%d]:%d", + i, bw_fixed_to_int(data->pipe_chunk_size_in_bytes[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] number_of_trips_to_memory_for_getting_apte_row[%d]:%d", + i, bw_fixed_to_int(data->number_of_trips_to_memory_for_getting_apte_row[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] adjusted_data_buffer_size[%d]:%d", + i, bw_fixed_to_int(data->adjusted_data_buffer_size[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] adjusted_data_buffer_size_in_memory[%d]:%d", + i, bw_fixed_to_int(data->adjusted_data_buffer_size_in_memory[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] pixels_per_data_fifo_entry[%d]:%d", + i, bw_fixed_to_int(data->pixels_per_data_fifo_entry[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scatter_gather_pte_requests_in_row[%d]:%d", + i, bw_fixed_to_int(data->scatter_gather_pte_requests_in_row[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] pte_request_per_chunk[%d]:%d", + i, bw_fixed_to_int(data->pte_request_per_chunk[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scatter_gather_page_width[%d]:%d", + i, bw_fixed_to_int(data->scatter_gather_page_width[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] scatter_gather_page_height[%d]:%d", + i, bw_fixed_to_int(data->scatter_gather_page_height[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] lb_lines_in_per_line_out_in_beginning_of_frame[%d]:%d", + i, bw_fixed_to_int(data->lb_lines_in_per_line_out_in_beginning_of_frame[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] lb_lines_in_per_line_out_in_middle_of_frame[%d]:%d", + i, bw_fixed_to_int(data->lb_lines_in_per_line_out_in_middle_of_frame[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] cursor_width_pixels[%d]:%d", + i, bw_fixed_to_int(data->cursor_width_pixels[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] minimum_latency_hiding[%d]:%d", + i, bw_fixed_to_int(data->minimum_latency_hiding[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] maximum_latency_hiding[%d]:%d", + i, bw_fixed_to_int(data->maximum_latency_hiding[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] minimum_latency_hiding_with_cursor[%d]:%d", + i, bw_fixed_to_int(data->minimum_latency_hiding_with_cursor[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] maximum_latency_hiding_with_cursor[%d]:%d", + i, bw_fixed_to_int(data->maximum_latency_hiding_with_cursor[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] src_pixels_for_first_output_pixel[%d]:%d", + i, bw_fixed_to_int(data->src_pixels_for_first_output_pixel[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] src_pixels_for_last_output_pixel[%d]:%d", + i, bw_fixed_to_int(data->src_pixels_for_last_output_pixel[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] src_data_for_first_output_pixel[%d]:%d", + i, bw_fixed_to_int(data->src_data_for_first_output_pixel[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] src_data_for_last_output_pixel[%d]:%d", + i, bw_fixed_to_int(data->src_data_for_last_output_pixel[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] active_time[%d]:%d", i, bw_fixed_to_int(data->active_time[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] horizontal_blank_and_chunk_granularity_factor[%d]:%d", + i, bw_fixed_to_int(data->horizontal_blank_and_chunk_granularity_factor[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] cursor_latency_hiding[%d]:%d", + i, bw_fixed_to_int(data->cursor_latency_hiding[i])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] v_blank_dram_speed_change_margin[%d]:%d", + i, bw_fixed_to_int(data->v_blank_dram_speed_change_margin[i])); + } + + for (i = 0; i < maximum_number_of_surfaces; i++) { + for (j = 0; j < 3; j++) { + for (k = 0; k < 8; k++) { + + DC_LOG_BANDWIDTH_CALCS("\n [bw_fixed] line_source_transfer_time[%d][%d][%d]:%d", + i, j, k, bw_fixed_to_int(data->line_source_transfer_time[i][j][k])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dram_speed_change_line_source_transfer_time[%d][%d][%d]:%d", + i, j, k, + bw_fixed_to_int(data->dram_speed_change_line_source_transfer_time[i][j][k])); + } + } + } + + for (i = 0; i < 3; i++) { + for (j = 0; j < 8; j++) { + + DC_LOG_BANDWIDTH_CALCS("\n [uint32_t] num_displays_with_margin[%d][%d]:%d", + i, j, data->num_displays_with_margin[i][j]); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmif_burst_time[%d][%d]:%d", + i, j, bw_fixed_to_int(data->dmif_burst_time[i][j])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] mcifwr_burst_time[%d][%d]:%d", + i, j, bw_fixed_to_int(data->mcifwr_burst_time[i][j])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] min_dram_speed_change_margin[%d][%d]:%d", + i, j, bw_fixed_to_int(data->min_dram_speed_change_margin[i][j])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dispclk_required_for_dram_speed_change[%d][%d]:%d", + i, j, bw_fixed_to_int(data->dispclk_required_for_dram_speed_change[i][j])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] blackout_duration_margin[%d][%d]:%d", + i, j, bw_fixed_to_int(data->blackout_duration_margin[i][j])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dispclk_required_for_blackout_duration[%d][%d]:%d", + i, j, bw_fixed_to_int(data->dispclk_required_for_blackout_duration[i][j])); + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dispclk_required_for_blackout_recovery[%d][%d]:%d", + i, j, bw_fixed_to_int(data->dispclk_required_for_blackout_recovery[i][j])); + } + } + + for (i = 0; i < 6; i++) { + DC_LOG_BANDWIDTH_CALCS(" [bw_fixed] dmif_required_sclk_for_urgent_latency[%d]:%d", + i, bw_fixed_to_int(data->dmif_required_sclk_for_urgent_latency[i])); + } +} +; + +#endif /* _CALCS_CALCS_LOGGER_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/calcs/custom_float.c b/drivers/gpu/drm/amd/display/dc/calcs/custom_float.c index 7243c37f569e2..31d167bc548fd 100644 --- a/drivers/gpu/drm/amd/display/dc/calcs/custom_float.c +++ b/drivers/gpu/drm/amd/display/dc/calcs/custom_float.c @@ -36,41 +36,41 @@ static bool build_custom_float( uint32_t exp_offset = (1 << (format->exponenta_bits - 1)) - 1; const struct fixed31_32 mantissa_constant_plus_max_fraction = - dal_fixed31_32_from_fraction( + dc_fixpt_from_fraction( (1LL << (format->mantissa_bits + 1)) - 1, 1LL << format->mantissa_bits); struct fixed31_32 mantiss; - if (dal_fixed31_32_eq( + if (dc_fixpt_eq( value, - dal_fixed31_32_zero)) { + dc_fixpt_zero)) { *negative = false; *mantissa = 0; *exponenta = 0; return true; } - if (dal_fixed31_32_lt( + if (dc_fixpt_lt( value, - dal_fixed31_32_zero)) { + dc_fixpt_zero)) { *negative = format->sign; - value = dal_fixed31_32_neg(value); + value = dc_fixpt_neg(value); } else { *negative = false; } - if (dal_fixed31_32_lt( + if (dc_fixpt_lt( value, - dal_fixed31_32_one)) { + dc_fixpt_one)) { uint32_t i = 1; do { - value = dal_fixed31_32_shl(value, 1); + value = dc_fixpt_shl(value, 1); ++i; - } while (dal_fixed31_32_lt( + } while (dc_fixpt_lt( value, - dal_fixed31_32_one)); + dc_fixpt_one)); --i; @@ -81,15 +81,15 @@ static bool build_custom_float( } *exponenta = exp_offset - i; - } else if (dal_fixed31_32_le( + } else if (dc_fixpt_le( mantissa_constant_plus_max_fraction, value)) { uint32_t i = 1; do { - value = dal_fixed31_32_shr(value, 1); + value = dc_fixpt_shr(value, 1); ++i; - } while (dal_fixed31_32_lt( + } while (dc_fixpt_lt( mantissa_constant_plus_max_fraction, value)); @@ -98,23 +98,23 @@ static bool build_custom_float( *exponenta = exp_offset; } - mantiss = dal_fixed31_32_sub( + mantiss = dc_fixpt_sub( value, - dal_fixed31_32_one); + dc_fixpt_one); - if (dal_fixed31_32_lt( + if (dc_fixpt_lt( mantiss, - dal_fixed31_32_zero) || - dal_fixed31_32_lt( - dal_fixed31_32_one, + dc_fixpt_zero) || + dc_fixpt_lt( + dc_fixpt_one, mantiss)) - mantiss = dal_fixed31_32_zero; + mantiss = dc_fixpt_zero; else - mantiss = dal_fixed31_32_shl( + mantiss = dc_fixpt_shl( mantiss, format->mantissa_bits); - *mantissa = dal_fixed31_32_floor(mantiss); + *mantissa = dc_fixpt_floor(mantiss); return true; } diff --git a/drivers/gpu/drm/amd/display/dc/calcs/dce_calcs.c b/drivers/gpu/drm/amd/display/dc/calcs/dce_calcs.c index 0cbab81ab304a..2c4e8f0cb2dcd 100644 --- a/drivers/gpu/drm/amd/display/dc/calcs/dce_calcs.c +++ b/drivers/gpu/drm/amd/display/dc/calcs/dce_calcs.c @@ -28,6 +28,7 @@ #include "dc.h" #include "core_types.h" #include "dal_asic_id.h" +#include "calcs_logger.h" /* * NOTE: @@ -52,11 +53,14 @@ static enum bw_calcs_version bw_calcs_version_from_asic_id(struct hw_asic_id asi return BW_CALCS_VERSION_CARRIZO; case FAMILY_VI: + if (ASIC_REV_IS_POLARIS12_V(asic_id.hw_internal_rev)) + return BW_CALCS_VERSION_POLARIS12; if (ASIC_REV_IS_POLARIS10_P(asic_id.hw_internal_rev)) return BW_CALCS_VERSION_POLARIS10; - if (ASIC_REV_IS_POLARIS11_M(asic_id.hw_internal_rev) || - ASIC_REV_IS_POLARIS12_V(asic_id.hw_internal_rev)) + if (ASIC_REV_IS_POLARIS11_M(asic_id.hw_internal_rev)) return BW_CALCS_VERSION_POLARIS11; + if (ASIC_REV_IS_VEGAM(asic_id.hw_internal_rev)) + return BW_CALCS_VERSION_VEGAM; return BW_CALCS_VERSION_INVALID; case FAMILY_AI: @@ -2145,6 +2149,9 @@ void bw_calcs_init(struct bw_calcs_dceip *bw_dceip, dceip.mcifwr_all_surfaces_burst_time = bw_int_to_fixed(0); /* todo: this is a bug*/ break; case BW_CALCS_VERSION_POLARIS10: + /* TODO: Treat VEGAM the same as P10 for now + * Need to tune the para for VEGAM if needed */ + case BW_CALCS_VERSION_VEGAM: vbios.memory_type = bw_def_gddr5; vbios.dram_channel_width_in_bits = 32; vbios.number_of_dram_channels = asic_id.vram_width / vbios.dram_channel_width_in_bits; @@ -2373,6 +2380,122 @@ void bw_calcs_init(struct bw_calcs_dceip *bw_dceip, dceip.scatter_gather_pte_request_rows_in_tiling_mode = 2; dceip.mcifwr_all_surfaces_burst_time = bw_int_to_fixed(0); break; + case BW_CALCS_VERSION_POLARIS12: + vbios.memory_type = bw_def_gddr5; + vbios.dram_channel_width_in_bits = 32; + vbios.number_of_dram_channels = asic_id.vram_width / vbios.dram_channel_width_in_bits; + vbios.number_of_dram_banks = 8; + vbios.high_yclk = bw_int_to_fixed(6000); + vbios.mid_yclk = bw_int_to_fixed(3200); + vbios.low_yclk = bw_int_to_fixed(1000); + vbios.low_sclk = bw_int_to_fixed(678); + vbios.mid1_sclk = bw_int_to_fixed(864); + vbios.mid2_sclk = bw_int_to_fixed(900); + vbios.mid3_sclk = bw_int_to_fixed(920); + vbios.mid4_sclk = bw_int_to_fixed(940); + vbios.mid5_sclk = bw_int_to_fixed(960); + vbios.mid6_sclk = bw_int_to_fixed(980); + vbios.high_sclk = bw_int_to_fixed(1049); + vbios.low_voltage_max_dispclk = bw_int_to_fixed(459); + vbios.mid_voltage_max_dispclk = bw_int_to_fixed(654); + vbios.high_voltage_max_dispclk = bw_int_to_fixed(1108); + vbios.low_voltage_max_phyclk = bw_int_to_fixed(540); + vbios.mid_voltage_max_phyclk = bw_int_to_fixed(810); + vbios.high_voltage_max_phyclk = bw_int_to_fixed(810); + vbios.data_return_bus_width = bw_int_to_fixed(32); + vbios.trc = bw_int_to_fixed(48); + if (vbios.number_of_dram_channels == 2) // 64-bit + vbios.dmifmc_urgent_latency = bw_int_to_fixed(4); + else + vbios.dmifmc_urgent_latency = bw_int_to_fixed(3); + vbios.stutter_self_refresh_exit_latency = bw_int_to_fixed(5); + vbios.stutter_self_refresh_entry_latency = bw_int_to_fixed(0); + vbios.nbp_state_change_latency = bw_int_to_fixed(250); + vbios.mcifwrmc_urgent_latency = bw_int_to_fixed(10); + vbios.scatter_gather_enable = false; + vbios.down_spread_percentage = bw_frc_to_fixed(5, 10); + vbios.cursor_width = 32; + vbios.average_compression_rate = 4; + vbios.number_of_request_slots_gmc_reserves_for_dmif_per_channel = 256; + vbios.blackout_duration = bw_int_to_fixed(0); /* us */ + vbios.maximum_blackout_recovery_time = bw_int_to_fixed(0); + + dceip.max_average_percent_of_ideal_port_bw_display_can_use_in_normal_system_operation = 100; + dceip.max_average_percent_of_ideal_drambw_display_can_use_in_normal_system_operation = 100; + dceip.percent_of_ideal_port_bw_received_after_urgent_latency = 100; + dceip.large_cursor = false; + dceip.dmif_request_buffer_size = bw_int_to_fixed(768); + dceip.dmif_pipe_en_fbc_chunk_tracker = false; + dceip.cursor_max_outstanding_group_num = 1; + dceip.lines_interleaved_into_lb = 2; + dceip.chunk_width = 256; + dceip.number_of_graphics_pipes = 5; + dceip.number_of_underlay_pipes = 0; + dceip.low_power_tiling_mode = 0; + dceip.display_write_back_supported = true; + dceip.argb_compression_support = true; + dceip.underlay_vscaler_efficiency6_bit_per_component = + bw_frc_to_fixed(35556, 10000); + dceip.underlay_vscaler_efficiency8_bit_per_component = + bw_frc_to_fixed(34286, 10000); + dceip.underlay_vscaler_efficiency10_bit_per_component = + bw_frc_to_fixed(32, 10); + dceip.underlay_vscaler_efficiency12_bit_per_component = + bw_int_to_fixed(3); + dceip.graphics_vscaler_efficiency6_bit_per_component = + bw_frc_to_fixed(35, 10); + dceip.graphics_vscaler_efficiency8_bit_per_component = + bw_frc_to_fixed(34286, 10000); + dceip.graphics_vscaler_efficiency10_bit_per_component = + bw_frc_to_fixed(32, 10); + dceip.graphics_vscaler_efficiency12_bit_per_component = + bw_int_to_fixed(3); + dceip.alpha_vscaler_efficiency = bw_int_to_fixed(3); + dceip.max_dmif_buffer_allocated = 4; + dceip.graphics_dmif_size = 12288; + dceip.underlay_luma_dmif_size = 19456; + dceip.underlay_chroma_dmif_size = 23552; + dceip.pre_downscaler_enabled = true; + dceip.underlay_downscale_prefetch_enabled = true; + dceip.lb_write_pixels_per_dispclk = bw_int_to_fixed(1); + dceip.lb_size_per_component444 = bw_int_to_fixed(245952); + dceip.graphics_lb_nodownscaling_multi_line_prefetching = true; + dceip.stutter_and_dram_clock_state_change_gated_before_cursor = + bw_int_to_fixed(1); + dceip.underlay420_luma_lb_size_per_component = bw_int_to_fixed( + 82176); + dceip.underlay420_chroma_lb_size_per_component = + bw_int_to_fixed(164352); + dceip.underlay422_lb_size_per_component = bw_int_to_fixed( + 82176); + dceip.cursor_chunk_width = bw_int_to_fixed(64); + dceip.cursor_dcp_buffer_lines = bw_int_to_fixed(4); + dceip.underlay_maximum_width_efficient_for_tiling = + bw_int_to_fixed(1920); + dceip.underlay_maximum_height_efficient_for_tiling = + bw_int_to_fixed(1080); + dceip.peak_pte_request_to_eviction_ratio_limiting_multiple_displays_or_single_rotated_display = + bw_frc_to_fixed(3, 10); + dceip.peak_pte_request_to_eviction_ratio_limiting_single_display_no_rotation = + bw_int_to_fixed(25); + dceip.minimum_outstanding_pte_request_limit = bw_int_to_fixed( + 2); + dceip.maximum_total_outstanding_pte_requests_allowed_by_saw = + bw_int_to_fixed(128); + dceip.limit_excessive_outstanding_dmif_requests = true; + dceip.linear_mode_line_request_alternation_slice = + bw_int_to_fixed(64); + dceip.scatter_gather_lines_of_pte_prefetching_in_linear_mode = + 32; + dceip.display_write_back420_luma_mcifwr_buffer_size = 12288; + dceip.display_write_back420_chroma_mcifwr_buffer_size = 8192; + dceip.request_efficiency = bw_frc_to_fixed(8, 10); + dceip.dispclk_per_request = bw_int_to_fixed(2); + dceip.dispclk_ramping_factor = bw_frc_to_fixed(105, 100); + dceip.display_pipe_throughput_factor = bw_frc_to_fixed(105, 100); + dceip.scatter_gather_pte_request_rows_in_tiling_mode = 2; + dceip.mcifwr_all_surfaces_burst_time = bw_int_to_fixed(0); + break; case BW_CALCS_VERSION_STONEY: vbios.memory_type = bw_def_gddr5; vbios.dram_channel_width_in_bits = 64; @@ -2815,6 +2938,19 @@ static void populate_initial_data( data->bytes_per_pixel[num_displays + 4] = 4; break; } + } else if (pipe[i].stream->dst.width != 0 && + pipe[i].stream->dst.height != 0 && + pipe[i].stream->src.width != 0 && + pipe[i].stream->src.height != 0) { + data->src_width[num_displays + 4] = bw_int_to_fixed(pipe[i].stream->src.width); + data->pitch_in_pixels[num_displays + 4] = data->src_width[num_displays + 4]; + data->src_height[num_displays + 4] = bw_int_to_fixed(pipe[i].stream->src.height); + data->h_taps[num_displays + 4] = pipe[i].stream->src.width == pipe[i].stream->dst.width ? bw_int_to_fixed(1) : bw_int_to_fixed(2); + data->v_taps[num_displays + 4] = pipe[i].stream->src.height == pipe[i].stream->dst.height ? bw_int_to_fixed(1) : bw_int_to_fixed(2); + data->h_scale_ratio[num_displays + 4] = bw_frc_to_fixed(pipe[i].stream->src.width, pipe[i].stream->dst.width); + data->v_scale_ratio[num_displays + 4] = bw_frc_to_fixed(pipe[i].stream->src.height, pipe[i].stream->dst.height); + data->rotation_angle[num_displays + 4] = bw_int_to_fixed(0); + data->bytes_per_pixel[num_displays + 4] = 4; } else { data->src_width[num_displays + 4] = bw_int_to_fixed(pipe[i].stream->timing.h_addressable); data->pitch_in_pixels[num_displays + 4] = data->src_width[num_displays + 4]; @@ -2873,6 +3009,11 @@ bool bw_calcs(struct dc_context *ctx, struct bw_fixed mid_yclk = vbios->mid_yclk; struct bw_fixed low_yclk = vbios->low_yclk; + if (ctx->dc->debug.bandwidth_calcs_trace) { + print_bw_calcs_dceip(ctx->logger, dceip); + print_bw_calcs_vbios(ctx->logger, vbios); + print_bw_calcs_data(ctx->logger, data); + } calculate_bandwidth(dceip, vbios, data); yclk_lvl = data->y_clk_level; @@ -2968,7 +3109,33 @@ bool bw_calcs(struct dc_context *ctx, bw_fixed_to_int(bw_mul(data-> stutter_exit_watermark[9], bw_int_to_fixed(1000))); - + calcs_output->stutter_entry_wm_ns[0].a_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[4], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[1].a_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[5], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[2].a_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[6], bw_int_to_fixed(1000))); + if (ctx->dc->caps.max_slave_planes) { + calcs_output->stutter_entry_wm_ns[3].a_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[0], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[4].a_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[1], bw_int_to_fixed(1000))); + } else { + calcs_output->stutter_entry_wm_ns[3].a_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[7], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[4].a_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[8], bw_int_to_fixed(1000))); + } + calcs_output->stutter_entry_wm_ns[5].a_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[9], bw_int_to_fixed(1000))); calcs_output->urgent_wm_ns[0].a_mark = bw_fixed_to_int(bw_mul(data-> @@ -3063,7 +3230,33 @@ bool bw_calcs(struct dc_context *ctx, bw_fixed_to_int(bw_mul(data-> stutter_exit_watermark[9], bw_int_to_fixed(1000))); - + calcs_output->stutter_entry_wm_ns[0].b_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[4], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[1].b_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[5], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[2].b_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[6], bw_int_to_fixed(1000))); + if (ctx->dc->caps.max_slave_planes) { + calcs_output->stutter_entry_wm_ns[3].b_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[0], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[4].b_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[1], bw_int_to_fixed(1000))); + } else { + calcs_output->stutter_entry_wm_ns[3].b_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[7], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[4].b_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[8], bw_int_to_fixed(1000))); + } + calcs_output->stutter_entry_wm_ns[5].b_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[9], bw_int_to_fixed(1000))); calcs_output->urgent_wm_ns[0].b_mark = bw_fixed_to_int(bw_mul(data-> @@ -3156,6 +3349,34 @@ bool bw_calcs(struct dc_context *ctx, bw_fixed_to_int(bw_mul(data-> stutter_exit_watermark[9], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[0].c_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[4], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[1].c_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[5], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[2].c_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[6], bw_int_to_fixed(1000))); + if (ctx->dc->caps.max_slave_planes) { + calcs_output->stutter_entry_wm_ns[3].c_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[0], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[4].c_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[1], bw_int_to_fixed(1000))); + } else { + calcs_output->stutter_entry_wm_ns[3].c_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[7], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[4].c_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[8], bw_int_to_fixed(1000))); + } + calcs_output->stutter_entry_wm_ns[5].c_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[9], bw_int_to_fixed(1000))); + calcs_output->urgent_wm_ns[0].c_mark = bw_fixed_to_int(bw_mul(data-> urgent_watermark[4], bw_int_to_fixed(1000))); @@ -3260,6 +3481,33 @@ bool bw_calcs(struct dc_context *ctx, bw_fixed_to_int(bw_mul(data-> stutter_exit_watermark[9], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[0].d_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[4], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[1].d_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[5], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[2].d_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[6], bw_int_to_fixed(1000))); + if (ctx->dc->caps.max_slave_planes) { + calcs_output->stutter_entry_wm_ns[3].d_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[0], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[4].d_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[1], bw_int_to_fixed(1000))); + } else { + calcs_output->stutter_entry_wm_ns[3].d_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[7], bw_int_to_fixed(1000))); + calcs_output->stutter_entry_wm_ns[4].d_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[8], bw_int_to_fixed(1000))); + } + calcs_output->stutter_entry_wm_ns[5].d_mark = + bw_fixed_to_int(bw_mul(data-> + stutter_entry_watermark[9], bw_int_to_fixed(1000))); calcs_output->urgent_wm_ns[0].d_mark = bw_fixed_to_int(bw_mul(data-> diff --git a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c index 4bb43a371292a..49a4ea45466d3 100644 --- a/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c +++ b/drivers/gpu/drm/amd/display/dc/calcs/dcn_calcs.c @@ -873,14 +873,14 @@ bool dcn_validate_bandwidth( } if (pipe->plane_state->rotation % 2 == 0) { - ASSERT(pipe->plane_res.scl_data.ratios.horz.value != dal_fixed31_32_one.value + ASSERT(pipe->plane_res.scl_data.ratios.horz.value != dc_fixpt_one.value || v->scaler_rec_out_width[input_idx] == v->viewport_width[input_idx]); - ASSERT(pipe->plane_res.scl_data.ratios.vert.value != dal_fixed31_32_one.value + ASSERT(pipe->plane_res.scl_data.ratios.vert.value != dc_fixpt_one.value || v->scaler_recout_height[input_idx] == v->viewport_height[input_idx]); } else { - ASSERT(pipe->plane_res.scl_data.ratios.horz.value != dal_fixed31_32_one.value + ASSERT(pipe->plane_res.scl_data.ratios.horz.value != dc_fixpt_one.value || v->scaler_recout_height[input_idx] == v->viewport_width[input_idx]); - ASSERT(pipe->plane_res.scl_data.ratios.vert.value != dal_fixed31_32_one.value + ASSERT(pipe->plane_res.scl_data.ratios.vert.value != dc_fixpt_one.value || v->scaler_rec_out_width[input_idx] == v->viewport_height[input_idx]); } v->dcc_enable[input_idx] = pipe->plane_state->dcc.enable ? dcn_bw_yes : dcn_bw_no; @@ -1459,39 +1459,39 @@ void dcn_bw_notify_pplib_of_wm_ranges(struct dc *dc) void dcn_bw_sync_calcs_and_dml(struct dc *dc) { kernel_fpu_begin(); - DC_LOG_BANDWIDTH_CALCS("sr_exit_time: %d ns\n" - "sr_enter_plus_exit_time: %d ns\n" - "urgent_latency: %d ns\n" - "write_back_latency: %d ns\n" - "percent_of_ideal_drambw_received_after_urg_latency: %d %\n" + DC_LOG_BANDWIDTH_CALCS("sr_exit_time: %f ns\n" + "sr_enter_plus_exit_time: %f ns\n" + "urgent_latency: %f ns\n" + "write_back_latency: %f ns\n" + "percent_of_ideal_drambw_received_after_urg_latency: %f %%\n" "max_request_size: %d bytes\n" - "dcfclkv_max0p9: %d kHz\n" - "dcfclkv_nom0p8: %d kHz\n" - "dcfclkv_mid0p72: %d kHz\n" - "dcfclkv_min0p65: %d kHz\n" - "max_dispclk_vmax0p9: %d kHz\n" - "max_dispclk_vnom0p8: %d kHz\n" - "max_dispclk_vmid0p72: %d kHz\n" - "max_dispclk_vmin0p65: %d kHz\n" - "max_dppclk_vmax0p9: %d kHz\n" - "max_dppclk_vnom0p8: %d kHz\n" - "max_dppclk_vmid0p72: %d kHz\n" - "max_dppclk_vmin0p65: %d kHz\n" - "socclk: %d kHz\n" - "fabric_and_dram_bandwidth_vmax0p9: %d MB/s\n" - "fabric_and_dram_bandwidth_vnom0p8: %d MB/s\n" - "fabric_and_dram_bandwidth_vmid0p72: %d MB/s\n" - "fabric_and_dram_bandwidth_vmin0p65: %d MB/s\n" - "phyclkv_max0p9: %d kHz\n" - "phyclkv_nom0p8: %d kHz\n" - "phyclkv_mid0p72: %d kHz\n" - "phyclkv_min0p65: %d kHz\n" - "downspreading: %d %\n" + "dcfclkv_max0p9: %f kHz\n" + "dcfclkv_nom0p8: %f kHz\n" + "dcfclkv_mid0p72: %f kHz\n" + "dcfclkv_min0p65: %f kHz\n" + "max_dispclk_vmax0p9: %f kHz\n" + "max_dispclk_vnom0p8: %f kHz\n" + "max_dispclk_vmid0p72: %f kHz\n" + "max_dispclk_vmin0p65: %f kHz\n" + "max_dppclk_vmax0p9: %f kHz\n" + "max_dppclk_vnom0p8: %f kHz\n" + "max_dppclk_vmid0p72: %f kHz\n" + "max_dppclk_vmin0p65: %f kHz\n" + "socclk: %f kHz\n" + "fabric_and_dram_bandwidth_vmax0p9: %f MB/s\n" + "fabric_and_dram_bandwidth_vnom0p8: %f MB/s\n" + "fabric_and_dram_bandwidth_vmid0p72: %f MB/s\n" + "fabric_and_dram_bandwidth_vmin0p65: %f MB/s\n" + "phyclkv_max0p9: %f kHz\n" + "phyclkv_nom0p8: %f kHz\n" + "phyclkv_mid0p72: %f kHz\n" + "phyclkv_min0p65: %f kHz\n" + "downspreading: %f %%\n" "round_trip_ping_latency_cycles: %d DCFCLK Cycles\n" "urgent_out_of_order_return_per_channel: %d Bytes\n" "number_of_channels: %d\n" "vmm_page_size: %d Bytes\n" - "dram_clock_change_latency: %d ns\n" + "dram_clock_change_latency: %f ns\n" "return_bus_width: %d Bytes\n", dc->dcn_soc->sr_exit_time * 1000, dc->dcn_soc->sr_enter_plus_exit_time * 1000, @@ -1527,11 +1527,11 @@ void dcn_bw_sync_calcs_and_dml(struct dc *dc) dc->dcn_soc->vmm_page_size, dc->dcn_soc->dram_clock_change_latency * 1000, dc->dcn_soc->return_bus_width); - DC_LOG_BANDWIDTH_CALCS("rob_buffer_size_in_kbyte: %d\n" - "det_buffer_size_in_kbyte: %d\n" - "dpp_output_buffer_pixels: %d\n" - "opp_output_buffer_lines: %d\n" - "pixel_chunk_size_in_kbyte: %d\n" + DC_LOG_BANDWIDTH_CALCS("rob_buffer_size_in_kbyte: %f\n" + "det_buffer_size_in_kbyte: %f\n" + "dpp_output_buffer_pixels: %f\n" + "opp_output_buffer_lines: %f\n" + "pixel_chunk_size_in_kbyte: %f\n" "pte_enable: %d\n" "pte_chunk_size: %d kbytes\n" "meta_chunk_size: %d kbytes\n" @@ -1550,13 +1550,13 @@ void dcn_bw_sync_calcs_and_dml(struct dc *dc) "max_pscl_tolb_throughput: %d pixels/dppclk\n" "max_lb_tovscl_throughput: %d pixels/dppclk\n" "max_vscl_tohscl_throughput: %d pixels/dppclk\n" - "max_hscl_ratio: %d\n" - "max_vscl_ratio: %d\n" + "max_hscl_ratio: %f\n" + "max_vscl_ratio: %f\n" "max_hscl_taps: %d\n" "max_vscl_taps: %d\n" "pte_buffer_size_in_requests: %d\n" - "dispclk_ramping_margin: %d %\n" - "under_scan_factor: %d %\n" + "dispclk_ramping_margin: %f %%\n" + "under_scan_factor: %f %%\n" "max_inter_dcn_tile_repeaters: %d\n" "can_vstartup_lines_exceed_vsync_plus_back_porch_lines_minus_one: %d\n" "bug_forcing_luma_and_chroma_request_to_same_size_fixed: %d\n" diff --git a/drivers/gpu/drm/amd/display/dc/core/dc.c b/drivers/gpu/drm/amd/display/dc/core/dc.c index 9cd3566def8d3..644b2187507b9 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc.c @@ -936,95 +936,6 @@ bool dc_post_update_surfaces_to_stream(struct dc *dc) return true; } -/* - * TODO this whole function needs to go - * - * dc_surface_update is needlessly complex. See if we can just replace this - * with a dc_plane_state and follow the atomic model a bit more closely here. - */ -bool dc_commit_planes_to_stream( - struct dc *dc, - struct dc_plane_state **plane_states, - uint8_t new_plane_count, - struct dc_stream_state *dc_stream, - struct dc_state *state) -{ - /* no need to dynamically allocate this. it's pretty small */ - struct dc_surface_update updates[MAX_SURFACES]; - struct dc_flip_addrs *flip_addr; - struct dc_plane_info *plane_info; - struct dc_scaling_info *scaling_info; - int i; - struct dc_stream_update *stream_update = - kzalloc(sizeof(struct dc_stream_update), GFP_KERNEL); - - if (!stream_update) { - BREAK_TO_DEBUGGER(); - return false; - } - - flip_addr = kcalloc(MAX_SURFACES, sizeof(struct dc_flip_addrs), - GFP_KERNEL); - plane_info = kcalloc(MAX_SURFACES, sizeof(struct dc_plane_info), - GFP_KERNEL); - scaling_info = kcalloc(MAX_SURFACES, sizeof(struct dc_scaling_info), - GFP_KERNEL); - - if (!flip_addr || !plane_info || !scaling_info) { - kfree(flip_addr); - kfree(plane_info); - kfree(scaling_info); - kfree(stream_update); - return false; - } - - memset(updates, 0, sizeof(updates)); - - stream_update->src = dc_stream->src; - stream_update->dst = dc_stream->dst; - stream_update->out_transfer_func = dc_stream->out_transfer_func; - - for (i = 0; i < new_plane_count; i++) { - updates[i].surface = plane_states[i]; - updates[i].gamma = - (struct dc_gamma *)plane_states[i]->gamma_correction; - updates[i].in_transfer_func = plane_states[i]->in_transfer_func; - flip_addr[i].address = plane_states[i]->address; - flip_addr[i].flip_immediate = plane_states[i]->flip_immediate; - plane_info[i].color_space = plane_states[i]->color_space; - plane_info[i].input_tf = plane_states[i]->input_tf; - plane_info[i].format = plane_states[i]->format; - plane_info[i].plane_size = plane_states[i]->plane_size; - plane_info[i].rotation = plane_states[i]->rotation; - plane_info[i].horizontal_mirror = plane_states[i]->horizontal_mirror; - plane_info[i].stereo_format = plane_states[i]->stereo_format; - plane_info[i].tiling_info = plane_states[i]->tiling_info; - plane_info[i].visible = plane_states[i]->visible; - plane_info[i].per_pixel_alpha = plane_states[i]->per_pixel_alpha; - plane_info[i].dcc = plane_states[i]->dcc; - scaling_info[i].scaling_quality = plane_states[i]->scaling_quality; - scaling_info[i].src_rect = plane_states[i]->src_rect; - scaling_info[i].dst_rect = plane_states[i]->dst_rect; - scaling_info[i].clip_rect = plane_states[i]->clip_rect; - - updates[i].flip_addr = &flip_addr[i]; - updates[i].plane_info = &plane_info[i]; - updates[i].scaling_info = &scaling_info[i]; - } - - dc_commit_updates_for_stream( - dc, - updates, - new_plane_count, - dc_stream, stream_update, plane_states, state); - - kfree(flip_addr); - kfree(plane_info); - kfree(scaling_info); - kfree(stream_update); - return true; -} - struct dc_state *dc_create_state(void) { struct dc_state *context = kzalloc(sizeof(struct dc_state), @@ -1107,9 +1018,6 @@ static enum surface_update_type get_plane_info_update_type(const struct dc_surfa if (u->plane_info->color_space != u->surface->color_space) update_flags->bits.color_space_change = 1; - if (u->plane_info->input_tf != u->surface->input_tf) - update_flags->bits.input_tf_change = 1; - if (u->plane_info->horizontal_mirror != u->surface->horizontal_mirror) update_flags->bits.horizontal_mirror_change = 1; @@ -1243,12 +1151,20 @@ static enum surface_update_type det_surface_update(const struct dc *dc, if (u->input_csc_color_matrix) update_flags->bits.input_csc_change = 1; - if (update_flags->bits.in_transfer_func_change - || update_flags->bits.input_csc_change) { + if (u->coeff_reduction_factor) + update_flags->bits.coeff_reduction_change = 1; + + if (update_flags->bits.in_transfer_func_change) { type = UPDATE_TYPE_MED; elevate_update_type(&overall_type, type); } + if (update_flags->bits.input_csc_change + || update_flags->bits.coeff_reduction_change) { + type = UPDATE_TYPE_FULL; + elevate_update_type(&overall_type, type); + } + return overall_type; } @@ -1297,7 +1213,7 @@ enum surface_update_type dc_check_update_surfaces_for_stream( type = check_update_surfaces_for_stream(dc, updates, surface_count, stream_update, stream_status); if (type == UPDATE_TYPE_FULL) for (i = 0; i < surface_count; i++) - updates[i].surface->update_flags.bits.full_update = 1; + updates[i].surface->update_flags.raw = 0xFFFFFFFF; return type; } @@ -1375,6 +1291,12 @@ static void commit_planes_for_stream(struct dc *dc, pipe_ctx->stream_res.abm->funcs->set_abm_level( pipe_ctx->stream_res.abm, stream->abm_level); } + + if (stream_update && stream_update->periodic_fn_vsync_delta && + pipe_ctx->stream_res.tg->funcs->program_vline_interrupt) + pipe_ctx->stream_res.tg->funcs->program_vline_interrupt( + pipe_ctx->stream_res.tg, &pipe_ctx->stream->timing, + pipe_ctx->stream->periodic_fn_vsync_delta); } } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_debug.c b/drivers/gpu/drm/amd/display/dc/core/dc_debug.c index 5a552cb3f8a74..267c76766deae 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_debug.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_debug.c @@ -36,8 +36,9 @@ #include "hw_sequencer.h" #include "resource.h" -#define DC_LOGGER \ - logger + +#define DC_LOGGER_INIT(logger) + #define SURFACE_TRACE(...) do {\ if (dc->debug.surface_trace) \ @@ -60,8 +61,7 @@ void pre_surface_trace( int surface_count) { int i; - struct dc *core_dc = dc; - struct dal_logger *logger = core_dc->ctx->logger; + DC_LOGGER_INIT(dc->ctx->logger); for (i = 0; i < surface_count; i++) { const struct dc_plane_state *plane_state = plane_states[i]; @@ -72,8 +72,8 @@ void pre_surface_trace( "plane_state->visible = %d;\n" "plane_state->flip_immediate = %d;\n" "plane_state->address.type = %d;\n" - "plane_state->address.grph.addr.quad_part = 0x%X;\n" - "plane_state->address.grph.meta_addr.quad_part = 0x%X;\n" + "plane_state->address.grph.addr.quad_part = 0x%llX;\n" + "plane_state->address.grph.meta_addr.quad_part = 0x%llX;\n" "plane_state->scaling_quality.h_taps = %d;\n" "plane_state->scaling_quality.v_taps = %d;\n" "plane_state->scaling_quality.h_taps_c = %d;\n" @@ -155,7 +155,6 @@ void pre_surface_trace( "plane_state->tiling_info.gfx8.pipe_config = %d;\n" "plane_state->tiling_info.gfx8.array_mode = %d;\n" "plane_state->color_space = %d;\n" - "plane_state->input_tf = %d;\n" "plane_state->dcc.enable = %d;\n" "plane_state->format = %d;\n" "plane_state->rotation = %d;\n" @@ -163,7 +162,6 @@ void pre_surface_trace( plane_state->tiling_info.gfx8.pipe_config, plane_state->tiling_info.gfx8.array_mode, plane_state->color_space, - plane_state->input_tf, plane_state->dcc.enable, plane_state->format, plane_state->rotation, @@ -183,8 +181,7 @@ void update_surface_trace( int surface_count) { int i; - struct dc *core_dc = dc; - struct dal_logger *logger = core_dc->ctx->logger; + DC_LOGGER_INIT(dc->ctx->logger); for (i = 0; i < surface_count; i++) { const struct dc_surface_update *update = &updates[i]; @@ -192,8 +189,8 @@ void update_surface_trace( SURFACE_TRACE("Update %d\n", i); if (update->flip_addr) { SURFACE_TRACE("flip_addr->address.type = %d;\n" - "flip_addr->address.grph.addr.quad_part = 0x%X;\n" - "flip_addr->address.grph.meta_addr.quad_part = 0x%X;\n" + "flip_addr->address.grph.addr.quad_part = 0x%llX;\n" + "flip_addr->address.grph.meta_addr.quad_part = 0x%llX;\n" "flip_addr->flip_immediate = %d;\n", update->flip_addr->address.type, update->flip_addr->address.grph.addr.quad_part, @@ -204,16 +201,15 @@ void update_surface_trace( if (update->plane_info) { SURFACE_TRACE( "plane_info->color_space = %d;\n" - "plane_info->input_tf = %d;\n" "plane_info->format = %d;\n" "plane_info->plane_size.grph.surface_pitch = %d;\n" "plane_info->plane_size.grph.surface_size.height = %d;\n" "plane_info->plane_size.grph.surface_size.width = %d;\n" "plane_info->plane_size.grph.surface_size.x = %d;\n" "plane_info->plane_size.grph.surface_size.y = %d;\n" - "plane_info->rotation = %d;\n", + "plane_info->rotation = %d;\n" + "plane_info->stereo_format = %d;\n", update->plane_info->color_space, - update->plane_info->input_tf, update->plane_info->format, update->plane_info->plane_size.grph.surface_pitch, update->plane_info->plane_size.grph.surface_size.height, @@ -303,8 +299,7 @@ void update_surface_trace( void post_surface_trace(struct dc *dc) { - struct dc *core_dc = dc; - struct dal_logger *logger = core_dc->ctx->logger; + DC_LOGGER_INIT(dc->ctx->logger); SURFACE_TRACE("post surface process.\n"); @@ -316,10 +311,10 @@ void context_timing_trace( { int i; struct dc *core_dc = dc; - struct dal_logger *logger = core_dc->ctx->logger; int h_pos[MAX_PIPES], v_pos[MAX_PIPES]; struct crtc_position position; unsigned int underlay_idx = core_dc->res_pool->underlay_pipe_index; + DC_LOGGER_INIT(dc->ctx->logger); for (i = 0; i < core_dc->res_pool->pipe_count; i++) { @@ -354,9 +349,7 @@ void context_clock_trace( struct dc_state *context) { #if defined(CONFIG_DRM_AMD_DC_DCN1_0) - struct dc *core_dc = dc; - struct dal_logger *logger = core_dc->ctx->logger; - + DC_LOGGER_INIT(dc->ctx->logger); CLOCK_TRACE("Current: dispclk_khz:%d max_dppclk_khz:%d dcfclk_khz:%d\n" "dcfclk_deep_sleep_khz:%d fclk_khz:%d socclk_khz:%d\n", context->bw.dcn.calc_clk.dispclk_khz, @@ -371,6 +364,7 @@ void context_clock_trace( context->bw.dcn.calc_clk.dppclk_khz, context->bw.dcn.calc_clk.dcfclk_khz, context->bw.dcn.calc_clk.dcfclk_deep_sleep_khz, - context->bw.dcn.calc_clk.fclk_khz); + context->bw.dcn.calc_clk.fclk_khz, + context->bw.dcn.calc_clk.socclk_khz); #endif } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/core/dc_hw_sequencer.c index ebc96b7200835..83d121510ef52 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_hw_sequencer.c @@ -208,6 +208,7 @@ void color_space_to_black_color( case COLOR_SPACE_YCBCR709: case COLOR_SPACE_YCBCR601_LIMITED: case COLOR_SPACE_YCBCR709_LIMITED: + case COLOR_SPACE_2020_YCBCR: *black_color = black_color_format[BLACK_COLOR_FORMAT_YUV_CV]; break; @@ -216,7 +217,25 @@ void color_space_to_black_color( black_color_format[BLACK_COLOR_FORMAT_RGB_LIMITED]; break; - default: + /** + * Remove default and add case for all color space + * so when we forget to add new color space + * compiler will give a warning + */ + case COLOR_SPACE_UNKNOWN: + case COLOR_SPACE_SRGB: + case COLOR_SPACE_XR_RGB: + case COLOR_SPACE_MSREF_SCRGB: + case COLOR_SPACE_XV_YCC_709: + case COLOR_SPACE_XV_YCC_601: + case COLOR_SPACE_2020_RGB_FULLRANGE: + case COLOR_SPACE_2020_RGB_LIMITEDRANGE: + case COLOR_SPACE_ADOBERGB: + case COLOR_SPACE_DCIP3: + case COLOR_SPACE_DISPLAYNATIVE: + case COLOR_SPACE_DOLBYVISION: + case COLOR_SPACE_APPCTRL: + case COLOR_SPACE_CUSTOMPOINTS: /* fefault is sRGB black (full range). */ *black_color = black_color_format[BLACK_COLOR_FORMAT_RGB_FULLRANGE]; @@ -230,6 +249,9 @@ bool hwss_wait_for_blank_complete( { int counter; + /* Not applicable if the pipe is not primary, save 300ms of boot time */ + if (!tg->funcs->is_blanked) + return true; for (counter = 0; counter < 100; counter++) { if (tg->funcs->is_blanked(tg)) break; diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link.c b/drivers/gpu/drm/amd/display/dc/core/dc_link.c index 6d1c4981a1851..2fa521812d23b 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link.c @@ -45,8 +45,9 @@ #include "dce/dce_11_0_d.h" #include "dce/dce_11_0_enum.h" #include "dce/dce_11_0_sh_mask.h" -#define DC_LOGGER \ - dc_ctx->logger + +#define DC_LOGGER_INIT(logger) + #define LINK_INFO(...) \ DC_LOG_HW_HOTPLUG( \ @@ -468,6 +469,13 @@ static void link_disconnect_sink(struct dc_link *link) link->dpcd_sink_count = 0; } +static void link_disconnect_remap(struct dc_sink *prev_sink, struct dc_link *link) +{ + dc_sink_release(link->local_sink); + link->local_sink = prev_sink; +} + + static bool detect_dp( struct dc_link *link, struct display_sink_capability *sink_caps, @@ -550,6 +558,17 @@ static bool detect_dp( return true; } +static bool is_same_edid(struct dc_edid *old_edid, struct dc_edid *new_edid) +{ + if (old_edid->length != new_edid->length) + return false; + + if (new_edid->length == 0) + return false; + + return (memcmp(old_edid->raw_edid, new_edid->raw_edid, new_edid->length) == 0); +} + bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) { struct dc_sink_init_data sink_init_data = { 0 }; @@ -557,11 +576,15 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) uint8_t i; bool converter_disable_audio = false; struct audio_support *aud_support = &link->dc->res_pool->audio_support; + bool same_edid = false; enum dc_edid_status edid_status; struct dc_context *dc_ctx = link->ctx; struct dc_sink *sink = NULL; + struct dc_sink *prev_sink = NULL; + struct dpcd_caps prev_dpcd_caps; + bool same_dpcd = true; enum dc_connection_type new_connection_type = dc_connection_none; - + DC_LOGGER_INIT(link->ctx->logger); if (link->connector_signal == SIGNAL_TYPE_VIRTUAL) return false; @@ -574,6 +597,11 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) link->local_sink) return true; + prev_sink = link->local_sink; + if (prev_sink != NULL) { + dc_sink_retain(prev_sink); + memcpy(&prev_dpcd_caps, &link->dpcd_caps, sizeof(struct dpcd_caps)); + } link_disconnect_sink(link); if (new_connection_type != dc_connection_none) { @@ -615,14 +643,25 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) link, &sink_caps, &converter_disable_audio, - aud_support, reason)) + aud_support, reason)) { + if (prev_sink != NULL) + dc_sink_release(prev_sink); return false; + } + // Check if dpcp block is the same + if (prev_sink != NULL) { + if (memcmp(&link->dpcd_caps, &prev_dpcd_caps, sizeof(struct dpcd_caps))) + same_dpcd = false; + } /* Active dongle downstream unplug */ if (link->type == dc_connection_active_dongle && link->dpcd_caps.sink_count. - bits.SINK_COUNT == 0) + bits.SINK_COUNT == 0) { + if (prev_sink != NULL) + dc_sink_release(prev_sink); return true; + } if (link->type == dc_connection_mst_branch) { LINK_INFO("link=%d, mst branch is now Connected\n", @@ -630,9 +669,11 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) /* Need to setup mst link_cap struct here * otherwise dc_link_detect() will leave mst link_cap * empty which leads to allocate_mst_payload() has "0" - * pbn_per_slot value leading to exception on dal_fixed31_32_div() + * pbn_per_slot value leading to exception on dc_fixpt_div() */ link->verified_link_cap = link->reported_link_cap; + if (prev_sink != NULL) + dc_sink_release(prev_sink); return false; } @@ -642,6 +683,8 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) default: DC_ERROR("Invalid connector type! signal:%d\n", link->connector_signal); + if (prev_sink != NULL) + dc_sink_release(prev_sink); return false; } /* switch() */ @@ -664,6 +707,8 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) sink = dc_sink_create(&sink_init_data); if (!sink) { DC_ERROR("Failed to create sink!\n"); + if (prev_sink != NULL) + dc_sink_release(prev_sink); return false; } @@ -687,22 +732,33 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) break; } - if (link->connector_signal == SIGNAL_TYPE_DISPLAY_PORT && - sink_caps.transaction_type == - DDC_TRANSACTION_TYPE_I2C_OVER_AUX) { - /* - * TODO debug why Dell 2413 doesn't like - * two link trainings - */ + // Check if edid is the same + if ((prev_sink != NULL) && ((edid_status == EDID_THE_SAME) || (edid_status == EDID_OK))) + same_edid = is_same_edid(&prev_sink->dc_edid, &sink->dc_edid); - /* deal with non-mst cases */ - dp_hbr_verify_link_cap(link, &link->reported_link_cap); - } + // If both edid and dpcd are the same, then discard new sink and revert back to original sink + if ((same_edid) && (same_dpcd)) { + link_disconnect_remap(prev_sink, link); + sink = prev_sink; + prev_sink = NULL; + } else { + if (link->connector_signal == SIGNAL_TYPE_DISPLAY_PORT && + sink_caps.transaction_type == + DDC_TRANSACTION_TYPE_I2C_OVER_AUX) { + /* + * TODO debug why Dell 2413 doesn't like + * two link trainings + */ - /* HDMI-DVI Dongle */ - if (sink->sink_signal == SIGNAL_TYPE_HDMI_TYPE_A && - !sink->edid_caps.edid_hdmi) - sink->sink_signal = SIGNAL_TYPE_DVI_SINGLE_LINK; + /* deal with non-mst cases */ + dp_hbr_verify_link_cap(link, &link->reported_link_cap); + } + + /* HDMI-DVI Dongle */ + if (sink->sink_signal == SIGNAL_TYPE_HDMI_TYPE_A && + !sink->edid_caps.edid_hdmi) + sink->sink_signal = SIGNAL_TYPE_DVI_SINGLE_LINK; + } /* Connectivity log: detection */ for (i = 0; i < sink->dc_edid.length / EDID_BLOCK_SIZE; i++) { @@ -761,10 +817,14 @@ bool dc_link_detect(struct dc_link *link, enum dc_detect_reason reason) sink_caps.signal = SIGNAL_TYPE_NONE; } - LINK_INFO("link=%d, dc_sink_in=%p is now %s\n", + LINK_INFO("link=%d, dc_sink_in=%p is now %s prev_sink=%p dpcd same=%d edid same=%d\n", link->link_index, sink, (sink_caps.signal == SIGNAL_TYPE_NONE ? - "Disconnected":"Connected")); + "Disconnected":"Connected"), prev_sink, + same_dpcd, same_edid); + + if (prev_sink != NULL) + dc_sink_release(prev_sink); return true; } @@ -927,6 +987,7 @@ static bool construct( struct integrated_info info = {{{ 0 }}}; struct dc_bios *bios = init_params->dc->ctx->dc_bios; const struct dc_vbios_funcs *bp_funcs = bios->funcs; + DC_LOGGER_INIT(dc_ctx->logger); link->irq_source_hpd = DC_IRQ_SOURCE_INVALID; link->irq_source_hpd_rx = DC_IRQ_SOURCE_INVALID; @@ -1135,7 +1196,8 @@ static void dpcd_configure_panel_mode( { union dpcd_edp_config edp_config_set; bool panel_mode_edp = false; - struct dc_context *dc_ctx = link->ctx; + DC_LOGGER_INIT(link->ctx->logger); + memset(&edp_config_set, '\0', sizeof(union dpcd_edp_config)); if (DP_PANEL_MODE_DEFAULT != panel_mode) { @@ -1183,16 +1245,21 @@ static void enable_stream_features(struct pipe_ctx *pipe_ctx) { struct dc_stream_state *stream = pipe_ctx->stream; struct dc_link *link = stream->sink->link; - union down_spread_ctrl downspread; + union down_spread_ctrl old_downspread; + union down_spread_ctrl new_downspread; core_link_read_dpcd(link, DP_DOWNSPREAD_CTRL, - &downspread.raw, sizeof(downspread)); + &old_downspread.raw, sizeof(old_downspread)); - downspread.bits.IGNORE_MSA_TIMING_PARAM = + new_downspread.raw = old_downspread.raw; + + new_downspread.bits.IGNORE_MSA_TIMING_PARAM = (stream->ignore_msa_timing_param) ? 1 : 0; - core_link_write_dpcd(link, DP_DOWNSPREAD_CTRL, - &downspread.raw, sizeof(downspread)); + if (new_downspread.raw != old_downspread.raw) { + core_link_write_dpcd(link, DP_DOWNSPREAD_CTRL, + &new_downspread.raw, sizeof(new_downspread)); + } } static enum dc_status enable_link_dp( @@ -1843,9 +1910,22 @@ static void disable_link(struct dc_link *link, enum signal_type signal) static bool dp_active_dongle_validate_timing( const struct dc_crtc_timing *timing, - const struct dc_dongle_caps *dongle_caps) + const struct dpcd_caps *dpcd_caps) { unsigned int required_pix_clk = timing->pix_clk_khz; + const struct dc_dongle_caps *dongle_caps = &dpcd_caps->dongle_caps; + + switch (dpcd_caps->dongle_type) { + case DISPLAY_DONGLE_DP_VGA_CONVERTER: + case DISPLAY_DONGLE_DP_DVI_CONVERTER: + case DISPLAY_DONGLE_DP_DVI_DONGLE: + if (timing->pixel_encoding == PIXEL_ENCODING_RGB) + return true; + else + return false; + default: + break; + } if (dongle_caps->dongle_type != DISPLAY_DONGLE_DP_HDMI_CONVERTER || dongle_caps->extendedCapValid == false) @@ -1911,7 +1991,7 @@ enum dc_status dc_link_validate_mode_timing( const struct dc_crtc_timing *timing) { uint32_t max_pix_clk = stream->sink->dongle_max_pix_clk; - struct dc_dongle_caps *dongle_caps = &link->dpcd_caps.dongle_caps; + struct dpcd_caps *dpcd_caps = &link->dpcd_caps; /* A hack to avoid failing any modes for EDID override feature on * topology change such as lower quality cable for DP or different dongle @@ -1924,7 +2004,7 @@ enum dc_status dc_link_validate_mode_timing( return DC_EXCEED_DONGLE_CAP; /* Active Dongle*/ - if (!dp_active_dongle_validate_timing(timing, dongle_caps)) + if (!dp_active_dongle_validate_timing(timing, dpcd_caps)) return DC_EXCEED_DONGLE_CAP; switch (stream->signal) { @@ -1950,10 +2030,10 @@ bool dc_link_set_backlight_level(const struct dc_link *link, uint32_t level, struct dc *core_dc = link->ctx->dc; struct abm *abm = core_dc->res_pool->abm; struct dmcu *dmcu = core_dc->res_pool->dmcu; - struct dc_context *dc_ctx = link->ctx; unsigned int controller_id = 0; bool use_smooth_brightness = true; int i; + DC_LOGGER_INIT(link->ctx->logger); if ((dmcu == NULL) || (abm == NULL) || @@ -1961,7 +2041,7 @@ bool dc_link_set_backlight_level(const struct dc_link *link, uint32_t level, return false; if (stream) { - if (stream->bl_pwm_level == 0) + if (stream->bl_pwm_level == EDP_BACKLIGHT_RAMP_DISABLE_LEVEL) frame_ramp = 0; ((struct dc_stream_state *)stream)->bl_pwm_level = level; @@ -2038,10 +2118,10 @@ static struct fixed31_32 get_pbn_per_slot(struct dc_stream_state *stream) &stream->sink->link->cur_link_settings; uint32_t link_rate_in_mbps = link_settings->link_rate * LINK_RATE_REF_FREQ_IN_MHZ; - struct fixed31_32 mbps = dal_fixed31_32_from_int( + struct fixed31_32 mbps = dc_fixpt_from_int( link_rate_in_mbps * link_settings->lane_count); - return dal_fixed31_32_div_int(mbps, 54); + return dc_fixpt_div_int(mbps, 54); } static int get_color_depth(enum dc_color_depth color_depth) @@ -2082,7 +2162,7 @@ static struct fixed31_32 get_pbn_from_timing(struct pipe_ctx *pipe_ctx) numerator = 64 * PEAK_FACTOR_X1000; denominator = 54 * 8 * 1000 * 1000; kbps *= numerator; - peak_kbps = dal_fixed31_32_from_fraction(kbps, denominator); + peak_kbps = dc_fixpt_from_fraction(kbps, denominator); return peak_kbps; } @@ -2149,8 +2229,8 @@ static enum dc_status allocate_mst_payload(struct pipe_ctx *pipe_ctx) struct fixed31_32 avg_time_slots_per_mtp; struct fixed31_32 pbn; struct fixed31_32 pbn_per_slot; - struct dc_context *dc_ctx = link->ctx; uint8_t i; + DC_LOGGER_INIT(link->ctx->logger); /* enable_link_dp_mst already check link->enabled_stream_count * and stream is in link->stream[]. This is called during set mode, @@ -2178,11 +2258,11 @@ static enum dc_status allocate_mst_payload(struct pipe_ctx *pipe_ctx) link->mst_stream_alloc_table.stream_count); for (i = 0; i < MAX_CONTROLLER_NUM; i++) { - DC_LOG_MST("stream_enc[%d]: 0x%x " + DC_LOG_MST("stream_enc[%d]: %p " "stream[%d].vcp_id: %d " "stream[%d].slot_count: %d\n", i, - link->mst_stream_alloc_table.stream_allocations[i].stream_enc, + (void *) link->mst_stream_alloc_table.stream_allocations[i].stream_enc, i, link->mst_stream_alloc_table.stream_allocations[i].vcp_id, i, @@ -2209,7 +2289,7 @@ static enum dc_status allocate_mst_payload(struct pipe_ctx *pipe_ctx) /* slot X.Y for only current stream */ pbn_per_slot = get_pbn_per_slot(stream); pbn = get_pbn_from_timing(pipe_ctx); - avg_time_slots_per_mtp = dal_fixed31_32_div(pbn, pbn_per_slot); + avg_time_slots_per_mtp = dc_fixpt_div(pbn, pbn_per_slot); stream_encoder->funcs->set_mst_bandwidth( stream_encoder, @@ -2226,10 +2306,10 @@ static enum dc_status deallocate_mst_payload(struct pipe_ctx *pipe_ctx) struct link_encoder *link_encoder = link->link_enc; struct stream_encoder *stream_encoder = pipe_ctx->stream_res.stream_enc; struct dp_mst_stream_allocation_table proposed_table = {0}; - struct fixed31_32 avg_time_slots_per_mtp = dal_fixed31_32_from_int(0); + struct fixed31_32 avg_time_slots_per_mtp = dc_fixpt_from_int(0); uint8_t i; bool mst_mode = (link->type == dc_connection_mst_branch); - struct dc_context *dc_ctx = link->ctx; + DC_LOGGER_INIT(link->ctx->logger); /* deallocate_mst_payload is called before disable link. When mode or * disable/enable monitor, new stream is created which is not in link @@ -2268,11 +2348,11 @@ static enum dc_status deallocate_mst_payload(struct pipe_ctx *pipe_ctx) link->mst_stream_alloc_table.stream_count); for (i = 0; i < MAX_CONTROLLER_NUM; i++) { - DC_LOG_MST("stream_enc[%d]: 0x%x " + DC_LOG_MST("stream_enc[%d]: %p " "stream[%d].vcp_id: %d " "stream[%d].slot_count: %d\n", i, - link->mst_stream_alloc_table.stream_allocations[i].stream_enc, + (void *) link->mst_stream_alloc_table.stream_allocations[i].stream_enc, i, link->mst_stream_alloc_table.stream_allocations[i].vcp_id, i, @@ -2302,8 +2382,8 @@ void core_link_enable_stream( struct pipe_ctx *pipe_ctx) { struct dc *core_dc = pipe_ctx->stream->ctx->dc; - struct dc_context *dc_ctx = pipe_ctx->stream->ctx; enum dc_status status; + DC_LOGGER_INIT(pipe_ctx->stream->ctx->logger); /* eDP lit up by bios already, no need to enable again. */ if (pipe_ctx->stream->signal == SIGNAL_TYPE_EDP && diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c index 49c2face1e7a8..ae48d603ebd6c 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_ddc.c @@ -629,13 +629,14 @@ bool dal_ddc_service_query_ddc_data( return ret; } -ssize_t dal_ddc_service_read_dpcd_data( +enum ddc_result dal_ddc_service_read_dpcd_data( struct ddc_service *ddc, bool i2c, enum i2c_mot_mode mot, uint32_t address, uint8_t *data, - uint32_t len) + uint32_t len, + uint32_t *read) { struct aux_payload read_payload = { .i2c_over_aux = i2c, @@ -652,6 +653,8 @@ ssize_t dal_ddc_service_read_dpcd_data( .mot = mot }; + *read = 0; + if (len > DEFAULT_AUX_MAX_DATA_SIZE) { BREAK_TO_DEBUGGER(); return DDC_RESULT_FAILED_INVALID_OPERATION; @@ -661,7 +664,8 @@ ssize_t dal_ddc_service_read_dpcd_data( ddc->ctx->i2caux, ddc->ddc_pin, &command)) { - return (ssize_t)command.payloads->length; + *read = command.payloads->length; + return DDC_RESULT_SUCESSFULL; } return DDC_RESULT_FAILED_OPERATION; diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c index 3b5053570229c..7d609c71394bb 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_dp.c @@ -1378,8 +1378,8 @@ static uint32_t bandwidth_in_kbps_from_timing( { uint32_t bits_per_channel = 0; uint32_t kbps; - switch (timing->display_color_depth) { + switch (timing->display_color_depth) { case COLOR_DEPTH_666: bits_per_channel = 6; break; @@ -1401,14 +1401,20 @@ static uint32_t bandwidth_in_kbps_from_timing( default: break; } + ASSERT(bits_per_channel != 0); kbps = timing->pix_clk_khz; kbps *= bits_per_channel; - if (timing->flags.Y_ONLY != 1) + if (timing->flags.Y_ONLY != 1) { /*Only YOnly make reduce bandwidth by 1/3 compares to RGB*/ kbps *= 3; + if (timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) + kbps /= 2; + else if (timing->pixel_encoding == PIXEL_ENCODING_YCBCR422) + kbps = kbps * 2 / 3; + } return kbps; @@ -2278,6 +2284,8 @@ static bool retrieve_link_cap(struct dc_link *link) union edp_configuration_cap edp_config_cap; union dp_downstream_port_present ds_port = { 0 }; enum dc_status status = DC_ERROR_UNEXPECTED; + uint32_t read_dpcd_retry_cnt = 3; + int i; memset(dpcd_data, '\0', sizeof(dpcd_data)); memset(&down_strm_port_count, @@ -2285,11 +2293,15 @@ static bool retrieve_link_cap(struct dc_link *link) memset(&edp_config_cap, '\0', sizeof(union edp_configuration_cap)); - status = core_link_read_dpcd( - link, - DP_DPCD_REV, - dpcd_data, - sizeof(dpcd_data)); + for (i = 0; i < read_dpcd_retry_cnt; i++) { + status = core_link_read_dpcd( + link, + DP_DPCD_REV, + dpcd_data, + sizeof(dpcd_data)); + if (status == DC_OK) + break; + } if (status != DC_OK) { dm_error("%s: Read dpcd data failed.\n", __func__); @@ -2376,6 +2388,10 @@ bool detect_dp_sink_caps(struct dc_link *link) void detect_edp_sink_caps(struct dc_link *link) { retrieve_link_cap(link); + + if (link->reported_link_cap.link_rate == LINK_RATE_UNKNOWN) + link->reported_link_cap.link_rate = LINK_RATE_HIGH2; + link->verified_link_cap = link->reported_link_cap; } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c b/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c index 7c866a7d5e775..82cd1d6e6e59d 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_link_hwss.c @@ -11,8 +11,6 @@ #include "dc_link_dp.h" #include "dc_link_ddc.h" #include "dm_helpers.h" -#include "dce/dce_link_encoder.h" -#include "dce/dce_stream_encoder.h" #include "dpcd_defs.h" enum dc_status core_link_read_dpcd( diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c index ba3487e97361c..751f3ac9d9214 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_resource.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_resource.c @@ -45,8 +45,9 @@ #include "dcn10/dcn10_resource.h" #endif #include "dce120/dce120_resource.h" -#define DC_LOGGER \ - ctx->logger + +#define DC_LOGGER_INIT(logger) + enum dce_version resource_parse_asic_id(struct hw_asic_id asic_id) { enum dce_version dc_version = DCE_VERSION_UNKNOWN; @@ -78,6 +79,8 @@ enum dce_version resource_parse_asic_id(struct hw_asic_id asic_id) ASIC_REV_IS_POLARIS12_V(asic_id.hw_internal_rev)) { dc_version = DCE_VERSION_11_2; } + if (ASIC_REV_IS_VEGAM(asic_id.hw_internal_rev)) + dc_version = DCE_VERSION_11_22; break; case FAMILY_AI: dc_version = DCE_VERSION_12_0; @@ -124,6 +127,7 @@ struct resource_pool *dc_create_resource_pool( num_virtual_links, dc, asic_id); break; case DCE_VERSION_11_2: + case DCE_VERSION_11_22: res_pool = dce112_create_resource_pool( num_virtual_links, dc); break; @@ -492,9 +496,9 @@ static void calculate_viewport(struct pipe_ctx *pipe_ctx) data->viewport_c.x = data->viewport.x / vpc_div; data->viewport_c.y = data->viewport.y / vpc_div; data->inits.h_c = (data->viewport.x % vpc_div) != 0 ? - dal_fixed31_32_half : dal_fixed31_32_zero; + dc_fixpt_half : dc_fixpt_zero; data->inits.v_c = (data->viewport.y % vpc_div) != 0 ? - dal_fixed31_32_half : dal_fixed31_32_zero; + dc_fixpt_half : dc_fixpt_zero; /* Round up, assume original video size always even dimensions */ data->viewport_c.width = (data->viewport.width + vpc_div - 1) / vpc_div; data->viewport_c.height = (data->viewport.height + vpc_div - 1) / vpc_div; @@ -623,10 +627,10 @@ static void calculate_scaling_ratios(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_state->rotation == ROTATION_ANGLE_270) rect_swap_helper(&surf_src); - pipe_ctx->plane_res.scl_data.ratios.horz = dal_fixed31_32_from_fraction( + pipe_ctx->plane_res.scl_data.ratios.horz = dc_fixpt_from_fraction( surf_src.width, plane_state->dst_rect.width); - pipe_ctx->plane_res.scl_data.ratios.vert = dal_fixed31_32_from_fraction( + pipe_ctx->plane_res.scl_data.ratios.vert = dc_fixpt_from_fraction( surf_src.height, plane_state->dst_rect.height); @@ -648,6 +652,14 @@ static void calculate_scaling_ratios(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.scl_data.ratios.horz_c.value /= 2; pipe_ctx->plane_res.scl_data.ratios.vert_c.value /= 2; } + pipe_ctx->plane_res.scl_data.ratios.horz = dc_fixpt_truncate( + pipe_ctx->plane_res.scl_data.ratios.horz, 19); + pipe_ctx->plane_res.scl_data.ratios.vert = dc_fixpt_truncate( + pipe_ctx->plane_res.scl_data.ratios.vert, 19); + pipe_ctx->plane_res.scl_data.ratios.horz_c = dc_fixpt_truncate( + pipe_ctx->plane_res.scl_data.ratios.horz_c, 19); + pipe_ctx->plane_res.scl_data.ratios.vert_c = dc_fixpt_truncate( + pipe_ctx->plane_res.scl_data.ratios.vert_c, 19); } static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx, struct view *recout_skip) @@ -684,32 +696,33 @@ static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx, struct view *r * init_bot = init + scaling_ratio * init_c = init + truncated_vp_c_offset(from calculate viewport) */ - data->inits.h = dal_fixed31_32_div_int( - dal_fixed31_32_add_int(data->ratios.horz, data->taps.h_taps + 1), 2); + data->inits.h = dc_fixpt_truncate(dc_fixpt_div_int( + dc_fixpt_add_int(data->ratios.horz, data->taps.h_taps + 1), 2), 19); - data->inits.h_c = dal_fixed31_32_add(data->inits.h_c, dal_fixed31_32_div_int( - dal_fixed31_32_add_int(data->ratios.horz_c, data->taps.h_taps_c + 1), 2)); + data->inits.h_c = dc_fixpt_truncate(dc_fixpt_add(data->inits.h_c, dc_fixpt_div_int( + dc_fixpt_add_int(data->ratios.horz_c, data->taps.h_taps_c + 1), 2)), 19); - data->inits.v = dal_fixed31_32_div_int( - dal_fixed31_32_add_int(data->ratios.vert, data->taps.v_taps + 1), 2); + data->inits.v = dc_fixpt_truncate(dc_fixpt_div_int( + dc_fixpt_add_int(data->ratios.vert, data->taps.v_taps + 1), 2), 19); + + data->inits.v_c = dc_fixpt_truncate(dc_fixpt_add(data->inits.v_c, dc_fixpt_div_int( + dc_fixpt_add_int(data->ratios.vert_c, data->taps.v_taps_c + 1), 2)), 19); - data->inits.v_c = dal_fixed31_32_add(data->inits.v_c, dal_fixed31_32_div_int( - dal_fixed31_32_add_int(data->ratios.vert_c, data->taps.v_taps_c + 1), 2)); /* Adjust for viewport end clip-off */ if ((data->viewport.x + data->viewport.width) < (src.x + src.width) && !flip_horz_scan_dir) { int vp_clip = src.x + src.width - data->viewport.width - data->viewport.x; - int int_part = dal_fixed31_32_floor( - dal_fixed31_32_sub(data->inits.h, data->ratios.horz)); + int int_part = dc_fixpt_floor( + dc_fixpt_sub(data->inits.h, data->ratios.horz)); int_part = int_part > 0 ? int_part : 0; data->viewport.width += int_part < vp_clip ? int_part : vp_clip; } if ((data->viewport.y + data->viewport.height) < (src.y + src.height) && !flip_vert_scan_dir) { int vp_clip = src.y + src.height - data->viewport.height - data->viewport.y; - int int_part = dal_fixed31_32_floor( - dal_fixed31_32_sub(data->inits.v, data->ratios.vert)); + int int_part = dc_fixpt_floor( + dc_fixpt_sub(data->inits.v, data->ratios.vert)); int_part = int_part > 0 ? int_part : 0; data->viewport.height += int_part < vp_clip ? int_part : vp_clip; @@ -717,8 +730,8 @@ static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx, struct view *r if ((data->viewport_c.x + data->viewport_c.width) < (src.x + src.width) / vpc_div && !flip_horz_scan_dir) { int vp_clip = (src.x + src.width) / vpc_div - data->viewport_c.width - data->viewport_c.x; - int int_part = dal_fixed31_32_floor( - dal_fixed31_32_sub(data->inits.h_c, data->ratios.horz_c)); + int int_part = dc_fixpt_floor( + dc_fixpt_sub(data->inits.h_c, data->ratios.horz_c)); int_part = int_part > 0 ? int_part : 0; data->viewport_c.width += int_part < vp_clip ? int_part : vp_clip; @@ -726,8 +739,8 @@ static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx, struct view *r if ((data->viewport_c.y + data->viewport_c.height) < (src.y + src.height) / vpc_div && !flip_vert_scan_dir) { int vp_clip = (src.y + src.height) / vpc_div - data->viewport_c.height - data->viewport_c.y; - int int_part = dal_fixed31_32_floor( - dal_fixed31_32_sub(data->inits.v_c, data->ratios.vert_c)); + int int_part = dc_fixpt_floor( + dc_fixpt_sub(data->inits.v_c, data->ratios.vert_c)); int_part = int_part > 0 ? int_part : 0; data->viewport_c.height += int_part < vp_clip ? int_part : vp_clip; @@ -737,9 +750,9 @@ static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx, struct view *r if (data->viewport.x && !flip_horz_scan_dir) { int int_part; - data->inits.h = dal_fixed31_32_add(data->inits.h, dal_fixed31_32_mul_int( + data->inits.h = dc_fixpt_add(data->inits.h, dc_fixpt_mul_int( data->ratios.horz, recout_skip->width)); - int_part = dal_fixed31_32_floor(data->inits.h) - data->viewport.x; + int_part = dc_fixpt_floor(data->inits.h) - data->viewport.x; if (int_part < data->taps.h_taps) { int int_adj = data->viewport.x >= (data->taps.h_taps - int_part) ? (data->taps.h_taps - int_part) : data->viewport.x; @@ -752,15 +765,15 @@ static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx, struct view *r int_part = data->taps.h_taps; } data->inits.h.value &= 0xffffffff; - data->inits.h = dal_fixed31_32_add_int(data->inits.h, int_part); + data->inits.h = dc_fixpt_add_int(data->inits.h, int_part); } if (data->viewport_c.x && !flip_horz_scan_dir) { int int_part; - data->inits.h_c = dal_fixed31_32_add(data->inits.h_c, dal_fixed31_32_mul_int( + data->inits.h_c = dc_fixpt_add(data->inits.h_c, dc_fixpt_mul_int( data->ratios.horz_c, recout_skip->width)); - int_part = dal_fixed31_32_floor(data->inits.h_c) - data->viewport_c.x; + int_part = dc_fixpt_floor(data->inits.h_c) - data->viewport_c.x; if (int_part < data->taps.h_taps_c) { int int_adj = data->viewport_c.x >= (data->taps.h_taps_c - int_part) ? (data->taps.h_taps_c - int_part) : data->viewport_c.x; @@ -773,15 +786,15 @@ static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx, struct view *r int_part = data->taps.h_taps_c; } data->inits.h_c.value &= 0xffffffff; - data->inits.h_c = dal_fixed31_32_add_int(data->inits.h_c, int_part); + data->inits.h_c = dc_fixpt_add_int(data->inits.h_c, int_part); } if (data->viewport.y && !flip_vert_scan_dir) { int int_part; - data->inits.v = dal_fixed31_32_add(data->inits.v, dal_fixed31_32_mul_int( + data->inits.v = dc_fixpt_add(data->inits.v, dc_fixpt_mul_int( data->ratios.vert, recout_skip->height)); - int_part = dal_fixed31_32_floor(data->inits.v) - data->viewport.y; + int_part = dc_fixpt_floor(data->inits.v) - data->viewport.y; if (int_part < data->taps.v_taps) { int int_adj = data->viewport.y >= (data->taps.v_taps - int_part) ? (data->taps.v_taps - int_part) : data->viewport.y; @@ -794,15 +807,15 @@ static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx, struct view *r int_part = data->taps.v_taps; } data->inits.v.value &= 0xffffffff; - data->inits.v = dal_fixed31_32_add_int(data->inits.v, int_part); + data->inits.v = dc_fixpt_add_int(data->inits.v, int_part); } if (data->viewport_c.y && !flip_vert_scan_dir) { int int_part; - data->inits.v_c = dal_fixed31_32_add(data->inits.v_c, dal_fixed31_32_mul_int( + data->inits.v_c = dc_fixpt_add(data->inits.v_c, dc_fixpt_mul_int( data->ratios.vert_c, recout_skip->height)); - int_part = dal_fixed31_32_floor(data->inits.v_c) - data->viewport_c.y; + int_part = dc_fixpt_floor(data->inits.v_c) - data->viewport_c.y; if (int_part < data->taps.v_taps_c) { int int_adj = data->viewport_c.y >= (data->taps.v_taps_c - int_part) ? (data->taps.v_taps_c - int_part) : data->viewport_c.y; @@ -815,12 +828,12 @@ static void calculate_inits_and_adj_vp(struct pipe_ctx *pipe_ctx, struct view *r int_part = data->taps.v_taps_c; } data->inits.v_c.value &= 0xffffffff; - data->inits.v_c = dal_fixed31_32_add_int(data->inits.v_c, int_part); + data->inits.v_c = dc_fixpt_add_int(data->inits.v_c, int_part); } /* Interlaced inits based on final vert inits */ - data->inits.v_bot = dal_fixed31_32_add(data->inits.v, data->ratios.vert); - data->inits.v_c_bot = dal_fixed31_32_add(data->inits.v_c, data->ratios.vert_c); + data->inits.v_bot = dc_fixpt_add(data->inits.v, data->ratios.vert); + data->inits.v_c_bot = dc_fixpt_add(data->inits.v_c, data->ratios.vert_c); if (pipe_ctx->plane_state->rotation == ROTATION_ANGLE_90 || pipe_ctx->plane_state->rotation == ROTATION_ANGLE_270) { @@ -835,7 +848,7 @@ bool resource_build_scaling_params(struct pipe_ctx *pipe_ctx) struct dc_crtc_timing *timing = &pipe_ctx->stream->timing; struct view recout_skip = { 0 }; bool res = false; - struct dc_context *ctx = pipe_ctx->stream->ctx; + DC_LOGGER_INIT(pipe_ctx->stream->ctx->logger); /* Important: scaling ratio calculation requires pixel format, * lb depth calculation requires recout and taps require scaling ratios. * Inits require viewport, taps, ratios and recout of split pipe @@ -843,6 +856,9 @@ bool resource_build_scaling_params(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.scl_data.format = convert_pixel_format_to_dalsurface( pipe_ctx->plane_state->format); + if (pipe_ctx->stream->timing.flags.INTERLACE) + pipe_ctx->stream->dst.height *= 2; + calculate_scaling_ratios(pipe_ctx); calculate_viewport(pipe_ctx); @@ -863,6 +879,8 @@ bool resource_build_scaling_params(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.scl_data.h_active = timing->h_addressable + timing->h_border_left + timing->h_border_right; pipe_ctx->plane_res.scl_data.v_active = timing->v_addressable + timing->v_border_top + timing->v_border_bottom; + if (pipe_ctx->stream->timing.flags.INTERLACE) + pipe_ctx->plane_res.scl_data.v_active *= 2; /* Taps calculations */ @@ -908,6 +926,9 @@ bool resource_build_scaling_params(struct pipe_ctx *pipe_ctx) plane_state->dst_rect.x, plane_state->dst_rect.y); + if (pipe_ctx->stream->timing.flags.INTERLACE) + pipe_ctx->stream->dst.height /= 2; + return res; } @@ -1294,6 +1315,19 @@ bool dc_add_all_planes_for_stream( } +static bool is_hdr_static_meta_changed(struct dc_stream_state *cur_stream, + struct dc_stream_state *new_stream) +{ + if (cur_stream == NULL) + return true; + + if (memcmp(&cur_stream->hdr_static_metadata, + &new_stream->hdr_static_metadata, + sizeof(struct dc_info_packet)) != 0) + return true; + + return false; +} static bool is_timing_changed(struct dc_stream_state *cur_stream, struct dc_stream_state *new_stream) @@ -1329,6 +1363,9 @@ static bool are_stream_backends_same( if (is_timing_changed(stream_a, stream_b)) return false; + if (is_hdr_static_meta_changed(stream_a, stream_b)) + return false; + return true; } @@ -1599,18 +1636,6 @@ enum dc_status dc_remove_stream_from_ctx( return DC_OK; } -static void copy_pipe_ctx( - const struct pipe_ctx *from_pipe_ctx, struct pipe_ctx *to_pipe_ctx) -{ - struct dc_plane_state *plane_state = to_pipe_ctx->plane_state; - struct dc_stream_state *stream = to_pipe_ctx->stream; - - *to_pipe_ctx = *from_pipe_ctx; - to_pipe_ctx->stream = stream; - if (plane_state != NULL) - to_pipe_ctx->plane_state = plane_state; -} - static struct dc_stream_state *find_pll_sharable_stream( struct dc_stream_state *stream_needs_pll, struct dc_state *context) @@ -1703,7 +1728,7 @@ enum dc_status resource_map_pool_resources( pipe_idx = acquire_first_split_pipe(&context->res_ctx, pool, stream); #endif - if (pipe_idx < 0) + if (pipe_idx < 0 || context->res_ctx.pipe_ctx[pipe_idx].stream_res.tg == NULL) return DC_NO_CONTROLLER_RESOURCE; pipe_ctx = &context->res_ctx.pipe_ctx[pipe_idx]; @@ -1752,26 +1777,6 @@ enum dc_status resource_map_pool_resources( return DC_ERROR_UNEXPECTED; } -/* first stream in the context is used to populate the rest */ -void validate_guaranteed_copy_streams( - struct dc_state *context, - int max_streams) -{ - int i; - - for (i = 1; i < max_streams; i++) { - context->streams[i] = context->streams[0]; - - copy_pipe_ctx(&context->res_ctx.pipe_ctx[0], - &context->res_ctx.pipe_ctx[i]); - context->res_ctx.pipe_ctx[i].stream = - context->res_ctx.pipe_ctx[0].stream; - - dc_stream_retain(context->streams[i]); - context->stream_count++; - } -} - void dc_resource_state_copy_construct_current( const struct dc *dc, struct dc_state *dst_ctx) @@ -1798,9 +1803,9 @@ enum dc_status dc_validate_global_state( return DC_ERROR_UNEXPECTED; if (dc->res_pool->funcs->validate_global) { - result = dc->res_pool->funcs->validate_global(dc, new_ctx); - if (result != DC_OK) - return result; + result = dc->res_pool->funcs->validate_global(dc, new_ctx); + if (result != DC_OK) + return result; } for (i = 0; i < new_ctx->stream_count; i++) { @@ -1843,7 +1848,7 @@ enum dc_status dc_validate_global_state( } static void patch_gamut_packet_checksum( - struct encoder_info_packet *gamut_packet) + struct dc_info_packet *gamut_packet) { /* For gamut we recalc checksum */ if (gamut_packet->valid) { @@ -1862,12 +1867,11 @@ static void patch_gamut_packet_checksum( } static void set_avi_info_frame( - struct encoder_info_packet *info_packet, + struct dc_info_packet *info_packet, struct pipe_ctx *pipe_ctx) { struct dc_stream_state *stream = pipe_ctx->stream; enum dc_color_space color_space = COLOR_SPACE_UNKNOWN; - struct info_frame info_frame = { {0} }; uint32_t pixel_encoding = 0; enum scanning_type scan_type = SCANNING_TYPE_NODATA; enum dc_aspect_ratio aspect = ASPECT_RATIO_NO_DATA; @@ -1877,22 +1881,24 @@ static void set_avi_info_frame( unsigned int cn0_cn1_value = 0; uint8_t *check_sum = NULL; uint8_t byte_index = 0; - union hdmi_info_packet *hdmi_info = &info_frame.avi_info_packet.info_packet_hdmi; + union hdmi_info_packet hdmi_info; union display_content_support support = {0}; unsigned int vic = pipe_ctx->stream->timing.vic; enum dc_timing_3d_format format; + memset(&hdmi_info, 0, sizeof(union hdmi_info_packet)); + color_space = pipe_ctx->stream->output_color_space; if (color_space == COLOR_SPACE_UNKNOWN) color_space = (stream->timing.pixel_encoding == PIXEL_ENCODING_RGB) ? COLOR_SPACE_SRGB:COLOR_SPACE_YCBCR709; /* Initialize header */ - hdmi_info->bits.header.info_frame_type = HDMI_INFOFRAME_TYPE_AVI; + hdmi_info.bits.header.info_frame_type = HDMI_INFOFRAME_TYPE_AVI; /* InfoFrameVersion_3 is defined by CEA861F (Section 6.4), but shall * not be used in HDMI 2.0 (Section 10.1) */ - hdmi_info->bits.header.version = 2; - hdmi_info->bits.header.length = HDMI_AVI_INFOFRAME_SIZE; + hdmi_info.bits.header.version = 2; + hdmi_info.bits.header.length = HDMI_AVI_INFOFRAME_SIZE; /* * IDO-defined (Y2,Y1,Y0 = 1,1,1) shall not be used by devices built @@ -1918,39 +1924,39 @@ static void set_avi_info_frame( /* Y0_Y1_Y2 : The pixel encoding */ /* H14b AVI InfoFrame has extension on Y-field from 2 bits to 3 bits */ - hdmi_info->bits.Y0_Y1_Y2 = pixel_encoding; + hdmi_info.bits.Y0_Y1_Y2 = pixel_encoding; /* A0 = 1 Active Format Information valid */ - hdmi_info->bits.A0 = ACTIVE_FORMAT_VALID; + hdmi_info.bits.A0 = ACTIVE_FORMAT_VALID; /* B0, B1 = 3; Bar info data is valid */ - hdmi_info->bits.B0_B1 = BAR_INFO_BOTH_VALID; + hdmi_info.bits.B0_B1 = BAR_INFO_BOTH_VALID; - hdmi_info->bits.SC0_SC1 = PICTURE_SCALING_UNIFORM; + hdmi_info.bits.SC0_SC1 = PICTURE_SCALING_UNIFORM; /* S0, S1 : Underscan / Overscan */ /* TODO: un-hardcode scan type */ scan_type = SCANNING_TYPE_UNDERSCAN; - hdmi_info->bits.S0_S1 = scan_type; + hdmi_info.bits.S0_S1 = scan_type; /* C0, C1 : Colorimetry */ if (color_space == COLOR_SPACE_YCBCR709 || color_space == COLOR_SPACE_YCBCR709_LIMITED) - hdmi_info->bits.C0_C1 = COLORIMETRY_ITU709; + hdmi_info.bits.C0_C1 = COLORIMETRY_ITU709; else if (color_space == COLOR_SPACE_YCBCR601 || color_space == COLOR_SPACE_YCBCR601_LIMITED) - hdmi_info->bits.C0_C1 = COLORIMETRY_ITU601; + hdmi_info.bits.C0_C1 = COLORIMETRY_ITU601; else { - hdmi_info->bits.C0_C1 = COLORIMETRY_NO_DATA; + hdmi_info.bits.C0_C1 = COLORIMETRY_NO_DATA; } if (color_space == COLOR_SPACE_2020_RGB_FULLRANGE || color_space == COLOR_SPACE_2020_RGB_LIMITEDRANGE || color_space == COLOR_SPACE_2020_YCBCR) { - hdmi_info->bits.EC0_EC2 = COLORIMETRYEX_BT2020RGBYCBCR; - hdmi_info->bits.C0_C1 = COLORIMETRY_EXTENDED; + hdmi_info.bits.EC0_EC2 = COLORIMETRYEX_BT2020RGBYCBCR; + hdmi_info.bits.C0_C1 = COLORIMETRY_EXTENDED; } else if (color_space == COLOR_SPACE_ADOBERGB) { - hdmi_info->bits.EC0_EC2 = COLORIMETRYEX_ADOBERGB; - hdmi_info->bits.C0_C1 = COLORIMETRY_EXTENDED; + hdmi_info.bits.EC0_EC2 = COLORIMETRYEX_ADOBERGB; + hdmi_info.bits.C0_C1 = COLORIMETRY_EXTENDED; } /* TODO: un-hardcode aspect ratio */ @@ -1959,18 +1965,18 @@ static void set_avi_info_frame( switch (aspect) { case ASPECT_RATIO_4_3: case ASPECT_RATIO_16_9: - hdmi_info->bits.M0_M1 = aspect; + hdmi_info.bits.M0_M1 = aspect; break; case ASPECT_RATIO_NO_DATA: case ASPECT_RATIO_64_27: case ASPECT_RATIO_256_135: default: - hdmi_info->bits.M0_M1 = 0; + hdmi_info.bits.M0_M1 = 0; } /* Active Format Aspect ratio - same as Picture Aspect Ratio. */ - hdmi_info->bits.R0_R3 = ACTIVE_FORMAT_ASPECT_RATIO_SAME_AS_PICTURE; + hdmi_info.bits.R0_R3 = ACTIVE_FORMAT_ASPECT_RATIO_SAME_AS_PICTURE; /* TODO: un-hardcode cn0_cn1 and itc */ @@ -2013,8 +2019,8 @@ static void set_avi_info_frame( } } } - hdmi_info->bits.CN0_CN1 = cn0_cn1_value; - hdmi_info->bits.ITC = itc_value; + hdmi_info.bits.CN0_CN1 = cn0_cn1_value; + hdmi_info.bits.ITC = itc_value; } /* TODO : We should handle YCC quantization */ @@ -2023,19 +2029,19 @@ static void set_avi_info_frame( stream->sink->edid_caps.qy_bit == 1) { if (color_space == COLOR_SPACE_SRGB || color_space == COLOR_SPACE_2020_RGB_FULLRANGE) { - hdmi_info->bits.Q0_Q1 = RGB_QUANTIZATION_FULL_RANGE; - hdmi_info->bits.YQ0_YQ1 = YYC_QUANTIZATION_FULL_RANGE; + hdmi_info.bits.Q0_Q1 = RGB_QUANTIZATION_FULL_RANGE; + hdmi_info.bits.YQ0_YQ1 = YYC_QUANTIZATION_FULL_RANGE; } else if (color_space == COLOR_SPACE_SRGB_LIMITED || color_space == COLOR_SPACE_2020_RGB_LIMITEDRANGE) { - hdmi_info->bits.Q0_Q1 = RGB_QUANTIZATION_LIMITED_RANGE; - hdmi_info->bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; + hdmi_info.bits.Q0_Q1 = RGB_QUANTIZATION_LIMITED_RANGE; + hdmi_info.bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; } else { - hdmi_info->bits.Q0_Q1 = RGB_QUANTIZATION_DEFAULT_RANGE; - hdmi_info->bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; + hdmi_info.bits.Q0_Q1 = RGB_QUANTIZATION_DEFAULT_RANGE; + hdmi_info.bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; } } else { - hdmi_info->bits.Q0_Q1 = RGB_QUANTIZATION_DEFAULT_RANGE; - hdmi_info->bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; + hdmi_info.bits.Q0_Q1 = RGB_QUANTIZATION_DEFAULT_RANGE; + hdmi_info.bits.YQ0_YQ1 = YYC_QUANTIZATION_LIMITED_RANGE; } ///VIC @@ -2060,51 +2066,49 @@ static void set_avi_info_frame( break; } } - hdmi_info->bits.VIC0_VIC7 = vic; + hdmi_info.bits.VIC0_VIC7 = vic; /* pixel repetition * PR0 - PR3 start from 0 whereas pHwPathMode->mode.timing.flags.pixel * repetition start from 1 */ - hdmi_info->bits.PR0_PR3 = 0; + hdmi_info.bits.PR0_PR3 = 0; /* Bar Info * barTop: Line Number of End of Top Bar. * barBottom: Line Number of Start of Bottom Bar. * barLeft: Pixel Number of End of Left Bar. * barRight: Pixel Number of Start of Right Bar. */ - hdmi_info->bits.bar_top = stream->timing.v_border_top; - hdmi_info->bits.bar_bottom = (stream->timing.v_total + hdmi_info.bits.bar_top = stream->timing.v_border_top; + hdmi_info.bits.bar_bottom = (stream->timing.v_total - stream->timing.v_border_bottom + 1); - hdmi_info->bits.bar_left = stream->timing.h_border_left; - hdmi_info->bits.bar_right = (stream->timing.h_total + hdmi_info.bits.bar_left = stream->timing.h_border_left; + hdmi_info.bits.bar_right = (stream->timing.h_total - stream->timing.h_border_right + 1); /* check_sum - Calculate AFMT_AVI_INFO0 ~ AFMT_AVI_INFO3 */ - check_sum = &info_frame.avi_info_packet.info_packet_hdmi.packet_raw_data.sb[0]; + check_sum = &hdmi_info.packet_raw_data.sb[0]; *check_sum = HDMI_INFOFRAME_TYPE_AVI + HDMI_AVI_INFOFRAME_SIZE + 2; for (byte_index = 1; byte_index <= HDMI_AVI_INFOFRAME_SIZE; byte_index++) - *check_sum += hdmi_info->packet_raw_data.sb[byte_index]; + *check_sum += hdmi_info.packet_raw_data.sb[byte_index]; /* one byte complement */ *check_sum = (uint8_t) (0x100 - *check_sum); /* Store in hw_path_mode */ - info_packet->hb0 = hdmi_info->packet_raw_data.hb0; - info_packet->hb1 = hdmi_info->packet_raw_data.hb1; - info_packet->hb2 = hdmi_info->packet_raw_data.hb2; + info_packet->hb0 = hdmi_info.packet_raw_data.hb0; + info_packet->hb1 = hdmi_info.packet_raw_data.hb1; + info_packet->hb2 = hdmi_info.packet_raw_data.hb2; - for (byte_index = 0; byte_index < sizeof(info_frame.avi_info_packet. - info_packet_hdmi.packet_raw_data.sb); byte_index++) - info_packet->sb[byte_index] = info_frame.avi_info_packet. - info_packet_hdmi.packet_raw_data.sb[byte_index]; + for (byte_index = 0; byte_index < sizeof(hdmi_info.packet_raw_data.sb); byte_index++) + info_packet->sb[byte_index] = hdmi_info.packet_raw_data.sb[byte_index]; info_packet->valid = true; } static void set_vendor_info_packet( - struct encoder_info_packet *info_packet, + struct dc_info_packet *info_packet, struct dc_stream_state *stream) { uint32_t length = 0; @@ -2217,7 +2221,7 @@ static void set_vendor_info_packet( } static void set_spd_info_packet( - struct encoder_info_packet *info_packet, + struct dc_info_packet *info_packet, struct dc_stream_state *stream) { /* SPD info packet for FreeSync */ @@ -2338,104 +2342,19 @@ static void set_spd_info_packet( } static void set_hdr_static_info_packet( - struct encoder_info_packet *info_packet, + struct dc_info_packet *info_packet, struct dc_stream_state *stream) { - uint16_t i = 0; - enum signal_type signal = stream->signal; - uint32_t data; + /* HDR Static Metadata info packet for HDR10 */ - if (!stream->hdr_static_metadata.hdr_supported) + if (!stream->hdr_static_metadata.valid) return; - if (dc_is_hdmi_signal(signal)) { - info_packet->valid = true; - - info_packet->hb0 = 0x87; - info_packet->hb1 = 0x01; - info_packet->hb2 = 0x1A; - i = 1; - } else if (dc_is_dp_signal(signal)) { - info_packet->valid = true; - - info_packet->hb0 = 0x00; - info_packet->hb1 = 0x87; - info_packet->hb2 = 0x1D; - info_packet->hb3 = (0x13 << 2); - i = 2; - } - - data = stream->hdr_static_metadata.is_hdr; - info_packet->sb[i++] = data ? 0x02 : 0x00; - info_packet->sb[i++] = 0x00; - - data = stream->hdr_static_metadata.chromaticity_green_x / 2; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.chromaticity_green_y / 2; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.chromaticity_blue_x / 2; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.chromaticity_blue_y / 2; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.chromaticity_red_x / 2; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.chromaticity_red_y / 2; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.chromaticity_white_point_x / 2; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.chromaticity_white_point_y / 2; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.max_luminance; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.min_luminance; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.maximum_content_light_level; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - data = stream->hdr_static_metadata.maximum_frame_average_light_level; - info_packet->sb[i++] = data & 0xFF; - info_packet->sb[i++] = (data & 0xFF00) >> 8; - - if (dc_is_hdmi_signal(signal)) { - uint32_t checksum = 0; - - checksum += info_packet->hb0; - checksum += info_packet->hb1; - checksum += info_packet->hb2; - - for (i = 1; i <= info_packet->hb2; i++) - checksum += info_packet->sb[i]; - - info_packet->sb[0] = 0x100 - checksum; - } else if (dc_is_dp_signal(signal)) { - info_packet->sb[0] = 0x01; - info_packet->sb[1] = 0x1A; - } + *info_packet = stream->hdr_static_metadata; } static void set_vsc_info_packet( - struct encoder_info_packet *info_packet, + struct dc_info_packet *info_packet, struct dc_stream_state *stream) { unsigned int vscPacketRevision = 0; @@ -2650,6 +2569,8 @@ bool pipe_need_reprogram( if (is_timing_changed(pipe_ctx_old->stream, pipe_ctx->stream)) return true; + if (is_hdr_static_meta_changed(pipe_ctx_old->stream, pipe_ctx->stream)) + return true; return false; } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_stream.c b/drivers/gpu/drm/amd/display/dc/core/dc_stream.c index ce0747ed0f008..3732a1de9d6c2 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_stream.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_stream.c @@ -101,14 +101,16 @@ static void construct(struct dc_stream_state *stream, stream->status.link = stream->sink->link; update_stream_signal(stream); + + stream->out_transfer_func = dc_create_transfer_func(); + stream->out_transfer_func->type = TF_TYPE_BYPASS; } static void destruct(struct dc_stream_state *stream) { dc_sink_release(stream->sink); if (stream->out_transfer_func != NULL) { - dc_transfer_func_release( - stream->out_transfer_func); + dc_transfer_func_release(stream->out_transfer_func); stream->out_transfer_func = NULL; } } @@ -176,6 +178,7 @@ bool dc_stream_set_cursor_attributes( int i; struct dc *core_dc; struct resource_context *res_ctx; + struct pipe_ctx *pipe_to_program = NULL; if (NULL == stream) { dm_error("DC: dc_stream is NULL!\n"); @@ -203,9 +206,17 @@ bool dc_stream_set_cursor_attributes( if (pipe_ctx->top_pipe && pipe_ctx->plane_state != pipe_ctx->top_pipe->plane_state) continue; + if (!pipe_to_program) { + pipe_to_program = pipe_ctx; + core_dc->hwss.pipe_control_lock(core_dc, pipe_to_program, true); + } core_dc->hwss.set_cursor_attribute(pipe_ctx); } + + if (pipe_to_program) + core_dc->hwss.pipe_control_lock(core_dc, pipe_to_program, false); + return true; } @@ -216,6 +227,7 @@ bool dc_stream_set_cursor_position( int i; struct dc *core_dc; struct resource_context *res_ctx; + struct pipe_ctx *pipe_to_program = NULL; if (NULL == stream) { dm_error("DC: dc_stream is NULL!\n"); @@ -241,9 +253,17 @@ bool dc_stream_set_cursor_position( !pipe_ctx->plane_res.ipp) continue; + if (!pipe_to_program) { + pipe_to_program = pipe_ctx; + core_dc->hwss.pipe_control_lock(core_dc, pipe_to_program, true); + } + core_dc->hwss.set_cursor_position(pipe_ctx); } + if (pipe_to_program) + core_dc->hwss.pipe_control_lock(core_dc, pipe_to_program, false); + return true; } diff --git a/drivers/gpu/drm/amd/display/dc/core/dc_surface.c b/drivers/gpu/drm/amd/display/dc/core/dc_surface.c index ade5b8ee9c3c0..68a71adeb12e4 100644 --- a/drivers/gpu/drm/amd/display/dc/core/dc_surface.c +++ b/drivers/gpu/drm/amd/display/dc/core/dc_surface.c @@ -38,6 +38,12 @@ static void construct(struct dc_context *ctx, struct dc_plane_state *plane_state) { plane_state->ctx = ctx; + + plane_state->gamma_correction = dc_create_gamma(); + plane_state->gamma_correction->is_identity = true; + + plane_state->in_transfer_func = dc_create_transfer_func(); + plane_state->in_transfer_func->type = TF_TYPE_BYPASS; } static void destruct(struct dc_plane_state *plane_state) @@ -66,8 +72,8 @@ struct dc_plane_state *dc_create_plane_state(struct dc *dc) { struct dc *core_dc = dc; - struct dc_plane_state *plane_state = kzalloc(sizeof(*plane_state), - GFP_KERNEL); + struct dc_plane_state *plane_state = kvzalloc(sizeof(*plane_state), + GFP_KERNEL); if (NULL == plane_state) return NULL; @@ -120,7 +126,7 @@ static void dc_plane_state_free(struct kref *kref) { struct dc_plane_state *plane_state = container_of(kref, struct dc_plane_state, refcount); destruct(plane_state); - kfree(plane_state); + kvfree(plane_state); } void dc_plane_state_release(struct dc_plane_state *plane_state) @@ -136,7 +142,7 @@ void dc_gamma_retain(struct dc_gamma *gamma) static void dc_gamma_free(struct kref *kref) { struct dc_gamma *gamma = container_of(kref, struct dc_gamma, refcount); - kfree(gamma); + kvfree(gamma); } void dc_gamma_release(struct dc_gamma **gamma) @@ -147,7 +153,7 @@ void dc_gamma_release(struct dc_gamma **gamma) struct dc_gamma *dc_create_gamma(void) { - struct dc_gamma *gamma = kzalloc(sizeof(*gamma), GFP_KERNEL); + struct dc_gamma *gamma = kvzalloc(sizeof(*gamma), GFP_KERNEL); if (gamma == NULL) goto alloc_fail; @@ -167,7 +173,7 @@ void dc_transfer_func_retain(struct dc_transfer_func *tf) static void dc_transfer_func_free(struct kref *kref) { struct dc_transfer_func *tf = container_of(kref, struct dc_transfer_func, refcount); - kfree(tf); + kvfree(tf); } void dc_transfer_func_release(struct dc_transfer_func *tf) @@ -175,9 +181,9 @@ void dc_transfer_func_release(struct dc_transfer_func *tf) kref_put(&tf->refcount, dc_transfer_func_free); } -struct dc_transfer_func *dc_create_transfer_func(void) +struct dc_transfer_func *dc_create_transfer_func() { - struct dc_transfer_func *tf = kzalloc(sizeof(*tf), GFP_KERNEL); + struct dc_transfer_func *tf = kvzalloc(sizeof(*tf), GFP_KERNEL); if (tf == NULL) goto alloc_fail; diff --git a/drivers/gpu/drm/amd/display/dc/dc.h b/drivers/gpu/drm/amd/display/dc/dc.h index fa4b3c8b3bb75..9cfde0ccf4e9d 100644 --- a/drivers/gpu/drm/amd/display/dc/dc.h +++ b/drivers/gpu/drm/amd/display/dc/dc.h @@ -38,7 +38,7 @@ #include "inc/compressor.h" #include "dml/display_mode_lib.h" -#define DC_VER "3.1.38" +#define DC_VER "3.1.44" #define MAX_SURFACES 3 #define MAX_STREAMS 6 @@ -75,6 +75,7 @@ struct dc_caps { bool dynamic_audio; bool is_apu; bool dual_link_dvi; + bool post_blend_color_processing; }; struct dc_dcc_surface_param { @@ -202,6 +203,8 @@ struct dc_debug { bool timing_trace; bool clock_trace; bool validation_trace; + bool bandwidth_calcs_trace; + int max_downscale_src_width; /* stutter efficiency related */ bool disable_stutter; @@ -238,6 +241,8 @@ struct dc_debug { bool az_endpoint_mute_only; bool always_use_regamma; bool p010_mpo_support; + bool recovery_enabled; + }; struct dc_state; struct resource_pool; @@ -332,20 +337,6 @@ enum { TRANSFER_FUNC_POINTS = 1025 }; -// Moved here from color module for linux -enum color_transfer_func { - transfer_func_unknown, - transfer_func_srgb, - transfer_func_bt709, - transfer_func_pq2084, - transfer_func_pq2084_interim, - transfer_func_linear_0_1, - transfer_func_linear_0_125, - transfer_func_dolbyvision, - transfer_func_gamma_22, - transfer_func_gamma_26 -}; - struct dc_hdr_static_metadata { /* display chromaticities and white point in units of 0.00001 */ unsigned int chromaticity_green_x; @@ -361,9 +352,6 @@ struct dc_hdr_static_metadata { uint32_t max_luminance; uint32_t maximum_content_light_level; uint32_t maximum_frame_average_light_level; - - bool hdr_supported; - bool is_hdr; }; enum dc_transfer_func_type { @@ -419,7 +407,6 @@ union surface_update_flags { /* Medium updates */ uint32_t dcc_change:1; uint32_t color_space_change:1; - uint32_t input_tf_change:1; uint32_t horizontal_mirror_change:1; uint32_t per_pixel_alpha_change:1; uint32_t rotation_change:1; @@ -428,6 +415,7 @@ union surface_update_flags { uint32_t position_change:1; uint32_t in_transfer_func_change:1; uint32_t input_csc_change:1; + uint32_t coeff_reduction_change:1; uint32_t output_tf_change:1; uint32_t pixel_format_change:1; @@ -460,7 +448,7 @@ struct dc_plane_state { struct dc_gamma *gamma_correction; struct dc_transfer_func *in_transfer_func; struct dc_bias_and_scale *bias_and_scale; - struct csc_transform input_csc_color_matrix; + struct dc_csc_transform input_csc_color_matrix; struct fixed31_32 coeff_reduction_factor; uint32_t sdr_white_level; @@ -468,7 +456,6 @@ struct dc_plane_state { struct dc_hdr_static_metadata hdr_static_ctx; enum dc_color_space color_space; - enum color_transfer_func input_tf; enum surface_pixel_format format; enum dc_rotation_angle rotation; @@ -498,7 +485,6 @@ struct dc_plane_info { enum dc_rotation_angle rotation; enum plane_stereo_format stereo_format; enum dc_color_space color_space; - enum color_transfer_func input_tf; unsigned int sdr_white_level; bool horizontal_mirror; bool visible; @@ -517,19 +503,18 @@ struct dc_surface_update { struct dc_plane_state *surface; /* isr safe update parameters. null means no updates */ - struct dc_flip_addrs *flip_addr; - struct dc_plane_info *plane_info; - struct dc_scaling_info *scaling_info; + const struct dc_flip_addrs *flip_addr; + const struct dc_plane_info *plane_info; + const struct dc_scaling_info *scaling_info; /* following updates require alloc/sleep/spin that is not isr safe, * null means no updates */ - struct dc_gamma *gamma; - enum color_transfer_func color_input_tf; - struct dc_transfer_func *in_transfer_func; + const struct dc_gamma *gamma; + const struct dc_transfer_func *in_transfer_func; - struct csc_transform *input_csc_color_matrix; - struct fixed31_32 *coeff_reduction_factor; + const struct dc_csc_transform *input_csc_color_matrix; + const struct fixed31_32 *coeff_reduction_factor; }; /* @@ -699,6 +684,7 @@ struct dc_cursor { struct dc_cursor_attributes attributes; }; + /******************************************************************************* * Interrupt interfaces ******************************************************************************/ diff --git a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h index 2726b02e006bc..90bccd5ccaa2b 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_dp_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_dp_types.h @@ -26,6 +26,8 @@ #ifndef DC_DP_TYPES_H #define DC_DP_TYPES_H +#include "os_types.h" + enum dc_lane_count { LANE_COUNT_UNKNOWN = 0, LANE_COUNT_ONE = 1, diff --git a/drivers/gpu/drm/amd/display/dc/dc_helper.c b/drivers/gpu/drm/amd/display/dc/dc_helper.c index 48e1fcf53d434..bd0fda0ceb919 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_helper.c +++ b/drivers/gpu/drm/amd/display/dc/dc_helper.c @@ -117,6 +117,65 @@ uint32_t generic_reg_get5(const struct dc_context *ctx, uint32_t addr, return reg_val; } +uint32_t generic_reg_get6(const struct dc_context *ctx, uint32_t addr, + uint8_t shift1, uint32_t mask1, uint32_t *field_value1, + uint8_t shift2, uint32_t mask2, uint32_t *field_value2, + uint8_t shift3, uint32_t mask3, uint32_t *field_value3, + uint8_t shift4, uint32_t mask4, uint32_t *field_value4, + uint8_t shift5, uint32_t mask5, uint32_t *field_value5, + uint8_t shift6, uint32_t mask6, uint32_t *field_value6) +{ + uint32_t reg_val = dm_read_reg(ctx, addr); + *field_value1 = get_reg_field_value_ex(reg_val, mask1, shift1); + *field_value2 = get_reg_field_value_ex(reg_val, mask2, shift2); + *field_value3 = get_reg_field_value_ex(reg_val, mask3, shift3); + *field_value4 = get_reg_field_value_ex(reg_val, mask4, shift4); + *field_value5 = get_reg_field_value_ex(reg_val, mask5, shift5); + *field_value6 = get_reg_field_value_ex(reg_val, mask6, shift6); + return reg_val; +} + +uint32_t generic_reg_get7(const struct dc_context *ctx, uint32_t addr, + uint8_t shift1, uint32_t mask1, uint32_t *field_value1, + uint8_t shift2, uint32_t mask2, uint32_t *field_value2, + uint8_t shift3, uint32_t mask3, uint32_t *field_value3, + uint8_t shift4, uint32_t mask4, uint32_t *field_value4, + uint8_t shift5, uint32_t mask5, uint32_t *field_value5, + uint8_t shift6, uint32_t mask6, uint32_t *field_value6, + uint8_t shift7, uint32_t mask7, uint32_t *field_value7) +{ + uint32_t reg_val = dm_read_reg(ctx, addr); + *field_value1 = get_reg_field_value_ex(reg_val, mask1, shift1); + *field_value2 = get_reg_field_value_ex(reg_val, mask2, shift2); + *field_value3 = get_reg_field_value_ex(reg_val, mask3, shift3); + *field_value4 = get_reg_field_value_ex(reg_val, mask4, shift4); + *field_value5 = get_reg_field_value_ex(reg_val, mask5, shift5); + *field_value6 = get_reg_field_value_ex(reg_val, mask6, shift6); + *field_value7 = get_reg_field_value_ex(reg_val, mask7, shift7); + return reg_val; +} + +uint32_t generic_reg_get8(const struct dc_context *ctx, uint32_t addr, + uint8_t shift1, uint32_t mask1, uint32_t *field_value1, + uint8_t shift2, uint32_t mask2, uint32_t *field_value2, + uint8_t shift3, uint32_t mask3, uint32_t *field_value3, + uint8_t shift4, uint32_t mask4, uint32_t *field_value4, + uint8_t shift5, uint32_t mask5, uint32_t *field_value5, + uint8_t shift6, uint32_t mask6, uint32_t *field_value6, + uint8_t shift7, uint32_t mask7, uint32_t *field_value7, + uint8_t shift8, uint32_t mask8, uint32_t *field_value8) +{ + uint32_t reg_val = dm_read_reg(ctx, addr); + *field_value1 = get_reg_field_value_ex(reg_val, mask1, shift1); + *field_value2 = get_reg_field_value_ex(reg_val, mask2, shift2); + *field_value3 = get_reg_field_value_ex(reg_val, mask3, shift3); + *field_value4 = get_reg_field_value_ex(reg_val, mask4, shift4); + *field_value5 = get_reg_field_value_ex(reg_val, mask5, shift5); + *field_value6 = get_reg_field_value_ex(reg_val, mask6, shift6); + *field_value7 = get_reg_field_value_ex(reg_val, mask7, shift7); + *field_value8 = get_reg_field_value_ex(reg_val, mask8, shift8); + return reg_val; +} /* note: va version of this is pretty bad idea, since there is a output parameter pass by pointer * compiler won't be able to check for size match and is prone to stack corruption type of bugs diff --git a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h index b83a7dc2f5a97..b1f70579d61b8 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_hw_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_hw_types.h @@ -423,6 +423,11 @@ enum dc_gamma_type { GAMMA_CS_TFM_1D = 3, }; +struct dc_csc_transform { + uint16_t matrix[12]; + bool enable_adjustment; +}; + struct dc_gamma { struct kref refcount; enum dc_gamma_type type; diff --git a/drivers/gpu/drm/amd/display/dc/dc_link.h b/drivers/gpu/drm/amd/display/dc/dc_link.h index dc34515ef01fd..8a716baa1203b 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_link.h +++ b/drivers/gpu/drm/amd/display/dc/dc_link.h @@ -51,6 +51,14 @@ struct link_mst_stream_allocation_table { struct link_mst_stream_allocation stream_allocations[MAX_CONTROLLER_NUM]; }; +struct time_stamp { + uint64_t edp_poweroff; + uint64_t edp_poweron; +}; + +struct link_trace { + struct time_stamp time_stamp; +}; /* * A link contains one or more sinks and their connected status. * The currently active signal type (HDMI, DP-SST, DP-MST) is also reported. @@ -114,6 +122,7 @@ struct dc_link { struct dc_link_status link_status; + struct link_trace link_trace; }; const struct dc_link_status *dc_link_get_status(const struct dc_link *dc_link); diff --git a/drivers/gpu/drm/amd/display/dc/dc_stream.h b/drivers/gpu/drm/amd/display/dc/dc_stream.h index d017df56b2ba1..d7e6d53bb3834 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_stream.h +++ b/drivers/gpu/drm/amd/display/dc/dc_stream.h @@ -58,18 +58,20 @@ struct dc_stream_state { struct freesync_context freesync_ctx; - struct dc_hdr_static_metadata hdr_static_metadata; + struct dc_info_packet hdr_static_metadata; struct dc_transfer_func *out_transfer_func; struct colorspace_transform gamut_remap_matrix; - struct csc_transform csc_color_matrix; + struct dc_csc_transform csc_color_matrix; enum dc_color_space output_color_space; enum dc_dither_option dither_option; enum view_3d_format view_format; - enum color_transfer_func output_tf; bool ignore_msa_timing_param; + + unsigned long long periodic_fn_vsync_delta; + /* TODO: custom INFO packets */ /* TODO: ABM info (DMCU) */ /* PSR info */ @@ -110,9 +112,10 @@ struct dc_stream_update { struct rect src; struct rect dst; struct dc_transfer_func *out_transfer_func; - struct dc_hdr_static_metadata *hdr_static_metadata; - enum color_transfer_func color_output_tf; + struct dc_info_packet *hdr_static_metadata; unsigned int *abm_level; + + unsigned long long *periodic_fn_vsync_delta; }; bool dc_is_stream_unchanged( @@ -131,13 +134,6 @@ bool dc_is_stream_scaling_unchanged( * This does not trigger a flip. No surface address is programmed. */ -bool dc_commit_planes_to_stream( - struct dc *dc, - struct dc_plane_state **plane_states, - uint8_t new_plane_count, - struct dc_stream_state *dc_stream, - struct dc_state *state); - void dc_commit_updates_for_stream(struct dc *dc, struct dc_surface_update *srf_updates, int surface_count, @@ -209,14 +205,6 @@ bool dc_add_all_planes_for_stream( enum dc_status dc_validate_stream(struct dc *dc, struct dc_stream_state *stream); /* - * This function takes a stream and checks if it is guaranteed to be supported. - * Guaranteed means that MAX_COFUNC similar streams are supported. - * - * After this call: - * No hardware is programmed for call. Only validation is done. - */ - -/* * Set up streams and links associated to drive sinks * The streams parameter is an absolute set of all active streams. * diff --git a/drivers/gpu/drm/amd/display/dc/dc_types.h b/drivers/gpu/drm/amd/display/dc/dc_types.h index 9441305d3ab58..76df2534c4a41 100644 --- a/drivers/gpu/drm/amd/display/dc/dc_types.h +++ b/drivers/gpu/drm/amd/display/dc/dc_types.h @@ -25,7 +25,7 @@ #ifndef DC_TYPES_H_ #define DC_TYPES_H_ -#include "fixed32_32.h" +#include "os_types.h" #include "fixed31_32.h" #include "irq_types.h" #include "dc_dp_types.h" @@ -370,12 +370,6 @@ struct dc_csc_adjustments { struct fixed31_32 hue; }; -enum { - MAX_LANES = 2, - MAX_COFUNC_PATH = 6, - LAYER_INDEX_PRIMARY = -1, -}; - enum dpcd_downstream_port_max_bpc { DOWN_STREAM_MAX_8BPC = 0, DOWN_STREAM_MAX_10BPC, @@ -530,6 +524,15 @@ struct vrr_params { uint32_t frame_counter; }; +struct dc_info_packet { + bool valid; + uint8_t hb0; + uint8_t hb1; + uint8_t hb2; + uint8_t hb3; + uint8_t sb[32]; +}; + #define DC_PLANE_UPDATE_TIMES_MAX 10 struct dc_plane_flip_time { diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_abm.c b/drivers/gpu/drm/amd/display/dc/dce/dce_abm.c index fe92a12228031..29294db1a96b7 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_abm.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_abm.c @@ -26,7 +26,7 @@ #include "dce_abm.h" #include "dm_services.h" #include "reg_helper.h" -#include "fixed32_32.h" +#include "fixed31_32.h" #include "dc.h" #include "atom.h" diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_audio.c b/drivers/gpu/drm/amd/display/dc/dce/dce_audio.c index 6d5cdcdc8ec9d..7f6d724686f1a 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_audio.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_audio.c @@ -33,8 +33,9 @@ #define CTX \ aud->base.ctx -#define DC_LOGGER \ - aud->base.ctx->logger + +#define DC_LOGGER_INIT() + #define REG(reg)\ (aud->regs->reg) @@ -348,8 +349,8 @@ static void set_audio_latency( void dce_aud_az_enable(struct audio *audio) { - struct dce_audio *aud = DCE_AUD(audio); uint32_t value = AZ_REG_READ(AZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL); + DC_LOGGER_INIT(); set_reg_field_value(value, 1, AZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL, @@ -371,7 +372,7 @@ void dce_aud_az_enable(struct audio *audio) void dce_aud_az_disable(struct audio *audio) { uint32_t value; - struct dce_audio *aud = DCE_AUD(audio); + DC_LOGGER_INIT(); value = AZ_REG_READ(AZALIA_F0_CODEC_PIN_CONTROL_HOT_PLUG_CONTROL); set_reg_field_value(value, 1, diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_clock_source.c b/drivers/gpu/drm/amd/display/dc/dce/dce_clock_source.c index 0aa2cda608909..599c7ab6befef 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_clock_source.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_clock_source.c @@ -41,8 +41,9 @@ #define CTX \ clk_src->base.ctx -#define DC_LOGGER \ - calc_pll_cs->ctx->logger + +#define DC_LOGGER_INIT() + #undef FN #define FN(reg_name, field_name) \ clk_src->cs_shift->field_name, clk_src->cs_mask->field_name @@ -467,7 +468,7 @@ static uint32_t dce110_get_pix_clk_dividers_helper ( { uint32_t field = 0; uint32_t pll_calc_error = MAX_PLL_CALC_ERROR; - struct calc_pll_clock_source *calc_pll_cs = &clk_src->calc_pll; + DC_LOGGER_INIT(); /* Check if reference clock is external (not pcie/xtalin) * HW Dce80 spec: * 00 - PCIE_REFCLK, 01 - XTALIN, 02 - GENERICA, 03 - GENERICB @@ -557,8 +558,8 @@ static uint32_t dce110_get_pix_clk_dividers( struct pll_settings *pll_settings) { struct dce110_clk_src *clk_src = TO_DCE110_CLK_SRC(cs); - struct calc_pll_clock_source *calc_pll_cs = &clk_src->calc_pll; uint32_t pll_calc_error = MAX_PLL_CALC_ERROR; + DC_LOGGER_INIT(); if (pix_clk_params == NULL || pll_settings == NULL || pix_clk_params->requested_pix_clk == 0) { @@ -589,6 +590,7 @@ static uint32_t dce110_get_pix_clk_dividers( pll_settings, pix_clk_params); break; case DCE_VERSION_11_2: + case DCE_VERSION_11_22: case DCE_VERSION_12_0: #if defined(CONFIG_DRM_AMD_DC_DCN1_0) case DCN_VERSION_1_0: @@ -655,12 +657,12 @@ static uint32_t dce110_get_d_to_pixel_rate_in_hz( return 0; } - pix_rate = dal_fixed31_32_from_int(clk_src->ref_freq_khz); - pix_rate = dal_fixed31_32_mul_int(pix_rate, 1000); - pix_rate = dal_fixed31_32_mul_int(pix_rate, phase); - pix_rate = dal_fixed31_32_div_int(pix_rate, modulo); + pix_rate = dc_fixpt_from_int(clk_src->ref_freq_khz); + pix_rate = dc_fixpt_mul_int(pix_rate, 1000); + pix_rate = dc_fixpt_mul_int(pix_rate, phase); + pix_rate = dc_fixpt_div_int(pix_rate, modulo); - return dal_fixed31_32_round(pix_rate); + return dc_fixpt_round(pix_rate); } else { return dce110_get_dp_pixel_rate_from_combo_phy_pll(cs, pix_clk_params, pll_settings); } @@ -709,12 +711,12 @@ static bool calculate_ss( const struct spread_spectrum_data *ss_data, struct delta_sigma_data *ds_data) { - struct fixed32_32 fb_div; - struct fixed32_32 ss_amount; - struct fixed32_32 ss_nslip_amount; - struct fixed32_32 ss_ds_frac_amount; - struct fixed32_32 ss_step_size; - struct fixed32_32 modulation_time; + struct fixed31_32 fb_div; + struct fixed31_32 ss_amount; + struct fixed31_32 ss_nslip_amount; + struct fixed31_32 ss_ds_frac_amount; + struct fixed31_32 ss_step_size; + struct fixed31_32 modulation_time; if (ds_data == NULL) return false; @@ -729,42 +731,42 @@ static bool calculate_ss( /* compute SS_AMOUNT_FBDIV & SS_AMOUNT_NFRAC_SLIP & SS_AMOUNT_DSFRAC*/ /* 6 decimal point support in fractional feedback divider */ - fb_div = dal_fixed32_32_from_fraction( + fb_div = dc_fixpt_from_fraction( pll_settings->fract_feedback_divider, 1000000); - fb_div = dal_fixed32_32_add_int(fb_div, pll_settings->feedback_divider); + fb_div = dc_fixpt_add_int(fb_div, pll_settings->feedback_divider); ds_data->ds_frac_amount = 0; /*spreadSpectrumPercentage is in the unit of .01%, * so have to divided by 100 * 100*/ - ss_amount = dal_fixed32_32_mul( - fb_div, dal_fixed32_32_from_fraction(ss_data->percentage, + ss_amount = dc_fixpt_mul( + fb_div, dc_fixpt_from_fraction(ss_data->percentage, 100 * ss_data->percentage_divider)); - ds_data->feedback_amount = dal_fixed32_32_floor(ss_amount); + ds_data->feedback_amount = dc_fixpt_floor(ss_amount); - ss_nslip_amount = dal_fixed32_32_sub(ss_amount, - dal_fixed32_32_from_int(ds_data->feedback_amount)); - ss_nslip_amount = dal_fixed32_32_mul_int(ss_nslip_amount, 10); - ds_data->nfrac_amount = dal_fixed32_32_floor(ss_nslip_amount); + ss_nslip_amount = dc_fixpt_sub(ss_amount, + dc_fixpt_from_int(ds_data->feedback_amount)); + ss_nslip_amount = dc_fixpt_mul_int(ss_nslip_amount, 10); + ds_data->nfrac_amount = dc_fixpt_floor(ss_nslip_amount); - ss_ds_frac_amount = dal_fixed32_32_sub(ss_nslip_amount, - dal_fixed32_32_from_int(ds_data->nfrac_amount)); - ss_ds_frac_amount = dal_fixed32_32_mul_int(ss_ds_frac_amount, 65536); - ds_data->ds_frac_amount = dal_fixed32_32_floor(ss_ds_frac_amount); + ss_ds_frac_amount = dc_fixpt_sub(ss_nslip_amount, + dc_fixpt_from_int(ds_data->nfrac_amount)); + ss_ds_frac_amount = dc_fixpt_mul_int(ss_ds_frac_amount, 65536); + ds_data->ds_frac_amount = dc_fixpt_floor(ss_ds_frac_amount); /* compute SS_STEP_SIZE_DSFRAC */ - modulation_time = dal_fixed32_32_from_fraction( + modulation_time = dc_fixpt_from_fraction( pll_settings->reference_freq * 1000, pll_settings->reference_divider * ss_data->modulation_freq_hz); if (ss_data->flags.CENTER_SPREAD) - modulation_time = dal_fixed32_32_div_int(modulation_time, 4); + modulation_time = dc_fixpt_div_int(modulation_time, 4); else - modulation_time = dal_fixed32_32_div_int(modulation_time, 2); + modulation_time = dc_fixpt_div_int(modulation_time, 2); - ss_step_size = dal_fixed32_32_div(ss_amount, modulation_time); + ss_step_size = dc_fixpt_div(ss_amount, modulation_time); /* SS_STEP_SIZE_DSFRAC_DEC = Int(SS_STEP_SIZE * 2 ^ 16 * 10)*/ - ss_step_size = dal_fixed32_32_mul_int(ss_step_size, 65536 * 10); - ds_data->ds_frac_size = dal_fixed32_32_floor(ss_step_size); + ss_step_size = dc_fixpt_mul_int(ss_step_size, 65536 * 10); + ds_data->ds_frac_size = dc_fixpt_floor(ss_step_size); return true; } @@ -978,6 +980,7 @@ static bool dce110_program_pix_clk( break; case DCE_VERSION_11_2: + case DCE_VERSION_11_22: case DCE_VERSION_12_0: #if defined(CONFIG_DRM_AMD_DC_DCN1_0) case DCN_VERSION_1_0: @@ -1054,7 +1057,7 @@ static void get_ss_info_from_atombios( struct spread_spectrum_info *ss_info_cur; struct spread_spectrum_data *ss_data_cur; uint32_t i; - struct calc_pll_clock_source *calc_pll_cs = &clk_src->calc_pll; + DC_LOGGER_INIT(); if (ss_entries_num == NULL) { DC_LOG_SYNC( "Invalid entry !!!\n"); diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_clocks.c b/drivers/gpu/drm/amd/display/dc/dce/dce_clocks.c index 78e6beb6cf268..8a581c67bf2d1 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_clocks.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_clocks.c @@ -26,7 +26,7 @@ #include "dce_clocks.h" #include "dm_services.h" #include "reg_helper.h" -#include "fixed32_32.h" +#include "fixed31_32.h" #include "bios_parser_interface.h" #include "dc.h" #include "dmcu.h" @@ -35,7 +35,7 @@ #endif #include "core_types.h" #include "dc_types.h" - +#include "dal_asic_id.h" #define TO_DCE_CLOCKS(clocks)\ container_of(clocks, struct dce_disp_clk, base) @@ -228,19 +228,19 @@ static int dce_clocks_get_dp_ref_freq(struct display_clock *clk) generated according to average value (case as with previous ASICs) */ if (clk_dce->ss_on_dprefclk && clk_dce->dprefclk_ss_divider != 0) { - struct fixed32_32 ss_percentage = dal_fixed32_32_div_int( - dal_fixed32_32_from_fraction( + struct fixed31_32 ss_percentage = dc_fixpt_div_int( + dc_fixpt_from_fraction( clk_dce->dprefclk_ss_percentage, clk_dce->dprefclk_ss_divider), 200); - struct fixed32_32 adj_dp_ref_clk_khz; + struct fixed31_32 adj_dp_ref_clk_khz; - ss_percentage = dal_fixed32_32_sub(dal_fixed32_32_one, + ss_percentage = dc_fixpt_sub(dc_fixpt_one, ss_percentage); adj_dp_ref_clk_khz = - dal_fixed32_32_mul_int( + dc_fixpt_mul_int( ss_percentage, dp_ref_clk_khz); - dp_ref_clk_khz = dal_fixed32_32_floor(adj_dp_ref_clk_khz); + dp_ref_clk_khz = dc_fixpt_floor(adj_dp_ref_clk_khz); } return dp_ref_clk_khz; @@ -256,19 +256,19 @@ static int dce_clocks_get_dp_ref_freq_wrkaround(struct display_clock *clk) int dp_ref_clk_khz = 600000; if (clk_dce->ss_on_dprefclk && clk_dce->dprefclk_ss_divider != 0) { - struct fixed32_32 ss_percentage = dal_fixed32_32_div_int( - dal_fixed32_32_from_fraction( + struct fixed31_32 ss_percentage = dc_fixpt_div_int( + dc_fixpt_from_fraction( clk_dce->dprefclk_ss_percentage, clk_dce->dprefclk_ss_divider), 200); - struct fixed32_32 adj_dp_ref_clk_khz; + struct fixed31_32 adj_dp_ref_clk_khz; - ss_percentage = dal_fixed32_32_sub(dal_fixed32_32_one, + ss_percentage = dc_fixpt_sub(dc_fixpt_one, ss_percentage); adj_dp_ref_clk_khz = - dal_fixed32_32_mul_int( + dc_fixpt_mul_int( ss_percentage, dp_ref_clk_khz); - dp_ref_clk_khz = dal_fixed32_32_floor(adj_dp_ref_clk_khz); + dp_ref_clk_khz = dc_fixpt_floor(adj_dp_ref_clk_khz); } return dp_ref_clk_khz; @@ -413,9 +413,12 @@ static int dce112_set_clock( /*VBIOS will determine DPREFCLK frequency, so we don't set it*/ dce_clk_params.target_clock_frequency = 0; dce_clk_params.clock_type = DCECLOCK_TYPE_DPREFCLK; - dce_clk_params.flags.USE_GENLOCK_AS_SOURCE_FOR_DPREFCLK = + if (!ASICREV_IS_VEGA20_P(clk->ctx->asic_id.hw_internal_rev)) + dce_clk_params.flags.USE_GENLOCK_AS_SOURCE_FOR_DPREFCLK = (dce_clk_params.pll_id == CLOCK_SOURCE_COMBO_DISPLAY_PLL0); + else + dce_clk_params.flags.USE_GENLOCK_AS_SOURCE_FOR_DPREFCLK = false; bp->funcs->set_dce_clock(bp, &dce_clk_params); diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.c b/drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.c index 2ee3d9bf10622..a576b8bbb3cd4 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_dmcu.c @@ -28,7 +28,7 @@ #include "dce_dmcu.h" #include "dm_services.h" #include "reg_helper.h" -#include "fixed32_32.h" +#include "fixed31_32.h" #include "dc.h" #define TO_DCE_DMCU(dmcu)\ diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_hwseq.c b/drivers/gpu/drm/amd/display/dc/dce/dce_hwseq.c index 487724345d9d8..0275d6d60da42 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_hwseq.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_hwseq.c @@ -53,7 +53,8 @@ void dce_pipe_control_lock(struct dc *dc, struct dce_hwseq *hws = dc->hwseq; /* Not lock pipe when blank */ - if (lock && pipe->stream_res.tg->funcs->is_blanked(pipe->stream_res.tg)) + if (lock && pipe->stream_res.tg->funcs->is_blanked && + pipe->stream_res.tg->funcs->is_blanked(pipe->stream_res.tg)) return; val = REG_GET_4(BLND_V_UPDATE_LOCK[pipe->stream_res.tg->inst], diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_ipp.c b/drivers/gpu/drm/amd/display/dc/dce/dce_ipp.c index d737e911971bc..5d9506b3d46bf 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_ipp.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_ipp.c @@ -195,13 +195,13 @@ static void dce_ipp_program_input_lut( for (i = 0; i < gamma->num_entries; i++) { REG_SET(DC_LUT_SEQ_COLOR, 0, DC_LUT_SEQ_COLOR, - dal_fixed31_32_round( + dc_fixpt_round( gamma->entries.red[i])); REG_SET(DC_LUT_SEQ_COLOR, 0, DC_LUT_SEQ_COLOR, - dal_fixed31_32_round( + dc_fixpt_round( gamma->entries.green[i])); REG_SET(DC_LUT_SEQ_COLOR, 0, DC_LUT_SEQ_COLOR, - dal_fixed31_32_round( + dc_fixpt_round( gamma->entries.blue[i])); } diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c b/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c index 8167cad7bcf74..dbe3b26b6d9eb 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.c @@ -113,6 +113,7 @@ static const struct link_encoder_funcs dce110_lnk_enc_funcs = { .connect_dig_be_to_fe = dce110_link_encoder_connect_dig_be_to_fe, .enable_hpd = dce110_link_encoder_enable_hpd, .disable_hpd = dce110_link_encoder_disable_hpd, + .is_dig_enabled = dce110_is_dig_enabled, .destroy = dce110_link_encoder_destroy }; @@ -535,8 +536,9 @@ void dce110_psr_program_secondary_packet(struct link_encoder *enc, DP_SEC_GSP0_PRIORITY, 1); } -static bool is_dig_enabled(const struct dce110_link_encoder *enc110) +bool dce110_is_dig_enabled(struct link_encoder *enc) { + struct dce110_link_encoder *enc110 = TO_DCE110_LINK_ENC(enc); uint32_t value; REG_GET(DIG_BE_EN_CNTL, DIG_ENABLE, &value); @@ -1031,7 +1033,7 @@ void dce110_link_encoder_disable_output( struct bp_transmitter_control cntl = { 0 }; enum bp_result result; - if (!is_dig_enabled(enc110)) { + if (!dce110_is_dig_enabled(enc)) { /* OF_SKIP_POWER_DOWN_INACTIVE_ENCODER */ return; } diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.h b/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.h index 0ec3433d34b62..347069461a22a 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_link_encoder.h @@ -263,4 +263,6 @@ void dce110_psr_program_dp_dphy_fast_training(struct link_encoder *enc, void dce110_psr_program_secondary_packet(struct link_encoder *enc, unsigned int sdp_transmit_line_num_deadline); +bool dce110_is_dig_enabled(struct link_encoder *enc); + #endif /* __DC_LINK_ENCODER__DCE110_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.c b/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.c index 0790f25c7b3b0..b235a75355b85 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.c @@ -174,6 +174,25 @@ static void program_urgency_watermark( URGENCY_HIGH_WATERMARK, urgency_high_wm); } +static void dce120_program_urgency_watermark( + struct dce_mem_input *dce_mi, + uint32_t wm_select, + uint32_t urgency_low_wm, + uint32_t urgency_high_wm) +{ + REG_UPDATE(DPG_WATERMARK_MASK_CONTROL, + URGENCY_WATERMARK_MASK, wm_select); + + REG_SET_2(DPG_PIPE_URGENCY_CONTROL, 0, + URGENCY_LOW_WATERMARK, urgency_low_wm, + URGENCY_HIGH_WATERMARK, urgency_high_wm); + + REG_SET_2(DPG_PIPE_URGENT_LEVEL_CONTROL, 0, + URGENT_LEVEL_LOW_WATERMARK, urgency_low_wm, + URGENT_LEVEL_HIGH_WATERMARK, urgency_high_wm); + +} + static void program_nbp_watermark( struct dce_mem_input *dce_mi, uint32_t wm_select, @@ -206,6 +225,25 @@ static void program_nbp_watermark( } } +static void dce120_program_stutter_watermark( + struct dce_mem_input *dce_mi, + uint32_t wm_select, + uint32_t stutter_mark, + uint32_t stutter_entry) +{ + REG_UPDATE(DPG_WATERMARK_MASK_CONTROL, + STUTTER_EXIT_SELF_REFRESH_WATERMARK_MASK, wm_select); + + if (REG(DPG_PIPE_STUTTER_CONTROL2)) + REG_UPDATE_2(DPG_PIPE_STUTTER_CONTROL2, + STUTTER_EXIT_SELF_REFRESH_WATERMARK, stutter_mark, + STUTTER_ENTER_SELF_REFRESH_WATERMARK, stutter_entry); + else + REG_UPDATE_2(DPG_PIPE_STUTTER_CONTROL, + STUTTER_EXIT_SELF_REFRESH_WATERMARK, stutter_mark, + STUTTER_ENTER_SELF_REFRESH_WATERMARK, stutter_entry); +} + static void program_stutter_watermark( struct dce_mem_input *dce_mi, uint32_t wm_select, @@ -225,7 +263,8 @@ static void program_stutter_watermark( static void dce_mi_program_display_marks( struct mem_input *mi, struct dce_watermarks nbp, - struct dce_watermarks stutter, + struct dce_watermarks stutter_exit, + struct dce_watermarks stutter_enter, struct dce_watermarks urgent, uint32_t total_dest_line_time_ns) { @@ -243,13 +282,14 @@ static void dce_mi_program_display_marks( program_nbp_watermark(dce_mi, 2, nbp.a_mark); /* set a */ program_nbp_watermark(dce_mi, 1, nbp.d_mark); /* set d */ - program_stutter_watermark(dce_mi, 2, stutter.a_mark); /* set a */ - program_stutter_watermark(dce_mi, 1, stutter.d_mark); /* set d */ + program_stutter_watermark(dce_mi, 2, stutter_exit.a_mark); /* set a */ + program_stutter_watermark(dce_mi, 1, stutter_exit.d_mark); /* set d */ } -static void dce120_mi_program_display_marks(struct mem_input *mi, +static void dce112_mi_program_display_marks(struct mem_input *mi, struct dce_watermarks nbp, - struct dce_watermarks stutter, + struct dce_watermarks stutter_exit, + struct dce_watermarks stutter_entry, struct dce_watermarks urgent, uint32_t total_dest_line_time_ns) { @@ -273,10 +313,43 @@ static void dce120_mi_program_display_marks(struct mem_input *mi, program_nbp_watermark(dce_mi, 2, nbp.c_mark); /* set c */ program_nbp_watermark(dce_mi, 3, nbp.d_mark); /* set d */ - program_stutter_watermark(dce_mi, 0, stutter.a_mark); /* set a */ - program_stutter_watermark(dce_mi, 1, stutter.b_mark); /* set b */ - program_stutter_watermark(dce_mi, 2, stutter.c_mark); /* set c */ - program_stutter_watermark(dce_mi, 3, stutter.d_mark); /* set d */ + program_stutter_watermark(dce_mi, 0, stutter_exit.a_mark); /* set a */ + program_stutter_watermark(dce_mi, 1, stutter_exit.b_mark); /* set b */ + program_stutter_watermark(dce_mi, 2, stutter_exit.c_mark); /* set c */ + program_stutter_watermark(dce_mi, 3, stutter_exit.d_mark); /* set d */ +} + +static void dce120_mi_program_display_marks(struct mem_input *mi, + struct dce_watermarks nbp, + struct dce_watermarks stutter_exit, + struct dce_watermarks stutter_entry, + struct dce_watermarks urgent, + uint32_t total_dest_line_time_ns) +{ + struct dce_mem_input *dce_mi = TO_DCE_MEM_INPUT(mi); + uint32_t stutter_en = mi->ctx->dc->debug.disable_stutter ? 0 : 1; + + dce120_program_urgency_watermark(dce_mi, 0, /* set a */ + urgent.a_mark, total_dest_line_time_ns); + dce120_program_urgency_watermark(dce_mi, 1, /* set b */ + urgent.b_mark, total_dest_line_time_ns); + dce120_program_urgency_watermark(dce_mi, 2, /* set c */ + urgent.c_mark, total_dest_line_time_ns); + dce120_program_urgency_watermark(dce_mi, 3, /* set d */ + urgent.d_mark, total_dest_line_time_ns); + + REG_UPDATE_2(DPG_PIPE_STUTTER_CONTROL, + STUTTER_ENABLE, stutter_en, + STUTTER_IGNORE_FBC, 1); + program_nbp_watermark(dce_mi, 0, nbp.a_mark); /* set a */ + program_nbp_watermark(dce_mi, 1, nbp.b_mark); /* set b */ + program_nbp_watermark(dce_mi, 2, nbp.c_mark); /* set c */ + program_nbp_watermark(dce_mi, 3, nbp.d_mark); /* set d */ + + dce120_program_stutter_watermark(dce_mi, 0, stutter_exit.a_mark, stutter_entry.a_mark); /* set a */ + dce120_program_stutter_watermark(dce_mi, 1, stutter_exit.b_mark, stutter_entry.b_mark); /* set b */ + dce120_program_stutter_watermark(dce_mi, 2, stutter_exit.c_mark, stutter_entry.c_mark); /* set c */ + dce120_program_stutter_watermark(dce_mi, 3, stutter_exit.d_mark, stutter_entry.d_mark); /* set d */ } static void program_tiling( @@ -696,5 +769,17 @@ void dce112_mem_input_construct( const struct dce_mem_input_mask *mi_mask) { dce_mem_input_construct(dce_mi, ctx, inst, regs, mi_shift, mi_mask); + dce_mi->base.funcs->mem_input_program_display_marks = dce112_mi_program_display_marks; +} + +void dce120_mem_input_construct( + struct dce_mem_input *dce_mi, + struct dc_context *ctx, + int inst, + const struct dce_mem_input_registers *regs, + const struct dce_mem_input_shift *mi_shift, + const struct dce_mem_input_mask *mi_mask) +{ + dce_mem_input_construct(dce_mi, ctx, inst, regs, mi_shift, mi_mask); dce_mi->base.funcs->mem_input_program_display_marks = dce120_mi_program_display_marks; } diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.h b/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.h index 05d39c0cbe87c..d15b0d7f47fcd 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.h +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_mem_input.h @@ -106,6 +106,7 @@ struct dce_mem_input_registers { uint32_t DPG_PIPE_ARBITRATION_CONTROL1; uint32_t DPG_WATERMARK_MASK_CONTROL; uint32_t DPG_PIPE_URGENCY_CONTROL; + uint32_t DPG_PIPE_URGENT_LEVEL_CONTROL; uint32_t DPG_PIPE_NB_PSTATE_CHANGE_CONTROL; uint32_t DPG_PIPE_LOW_POWER_CONTROL; uint32_t DPG_PIPE_STUTTER_CONTROL; @@ -213,6 +214,11 @@ struct dce_mem_input_registers { #define MI_DCE12_DMIF_PG_MASK_SH_LIST(mask_sh, blk)\ SFB(blk, DPG_PIPE_STUTTER_CONTROL2, STUTTER_EXIT_SELF_REFRESH_WATERMARK, mask_sh),\ + SFB(blk, DPG_PIPE_STUTTER_CONTROL2, STUTTER_ENTER_SELF_REFRESH_WATERMARK, mask_sh),\ + SFB(blk, DPG_PIPE_URGENT_LEVEL_CONTROL, URGENT_LEVEL_LOW_WATERMARK, mask_sh),\ + SFB(blk, DPG_PIPE_URGENT_LEVEL_CONTROL, URGENT_LEVEL_HIGH_WATERMARK, mask_sh),\ + SFB(blk, DPG_PIPE_URGENCY_CONTROL, URGENCY_LOW_WATERMARK, mask_sh),\ + SFB(blk, DPG_PIPE_URGENCY_CONTROL, URGENCY_HIGH_WATERMARK, mask_sh),\ SFB(blk, DPG_WATERMARK_MASK_CONTROL, PSTATE_CHANGE_WATERMARK_MASK, mask_sh),\ SFB(blk, DPG_PIPE_LOW_POWER_CONTROL, PSTATE_CHANGE_ENABLE, mask_sh),\ SFB(blk, DPG_PIPE_LOW_POWER_CONTROL, PSTATE_CHANGE_URGENT_DURING_REQUEST, mask_sh),\ @@ -286,6 +292,8 @@ struct dce_mem_input_registers { type STUTTER_EXIT_SELF_REFRESH_WATERMARK_MASK; \ type URGENCY_LOW_WATERMARK; \ type URGENCY_HIGH_WATERMARK; \ + type URGENT_LEVEL_LOW_WATERMARK;\ + type URGENT_LEVEL_HIGH_WATERMARK;\ type NB_PSTATE_CHANGE_ENABLE; \ type NB_PSTATE_CHANGE_URGENT_DURING_REQUEST; \ type NB_PSTATE_CHANGE_NOT_SELF_REFRESH_DURING_REQUEST; \ @@ -297,6 +305,7 @@ struct dce_mem_input_registers { type STUTTER_ENABLE; \ type STUTTER_IGNORE_FBC; \ type STUTTER_EXIT_SELF_REFRESH_WATERMARK; \ + type STUTTER_ENTER_SELF_REFRESH_WATERMARK; \ type DMIF_BUFFERS_ALLOCATED; \ type DMIF_BUFFERS_ALLOCATION_COMPLETED; \ type ENABLE; /* MC_HUB_RDREQ_DMIF_LIMIT */\ @@ -344,4 +353,12 @@ void dce112_mem_input_construct( const struct dce_mem_input_shift *mi_shift, const struct dce_mem_input_mask *mi_mask); +void dce120_mem_input_construct( + struct dce_mem_input *dce_mi, + struct dc_context *ctx, + int inst, + const struct dce_mem_input_registers *regs, + const struct dce_mem_input_shift *mi_shift, + const struct dce_mem_input_mask *mi_mask); + #endif /*__DCE_MEM_INPUT_H__*/ diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_scl_filters.c b/drivers/gpu/drm/amd/display/dc/dce/dce_scl_filters.c index 6243450b41b7e..48862bebf29e9 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_scl_filters.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_scl_filters.c @@ -1014,11 +1014,11 @@ static const uint16_t filter_8tap_64p_183[264] = { const uint16_t *get_filter_3tap_16p(struct fixed31_32 ratio) { - if (ratio.value < dal_fixed31_32_one.value) + if (ratio.value < dc_fixpt_one.value) return filter_3tap_16p_upscale; - else if (ratio.value < dal_fixed31_32_from_fraction(4, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(4, 3).value) return filter_3tap_16p_117; - else if (ratio.value < dal_fixed31_32_from_fraction(5, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(5, 3).value) return filter_3tap_16p_150; else return filter_3tap_16p_183; @@ -1026,11 +1026,11 @@ const uint16_t *get_filter_3tap_16p(struct fixed31_32 ratio) const uint16_t *get_filter_3tap_64p(struct fixed31_32 ratio) { - if (ratio.value < dal_fixed31_32_one.value) + if (ratio.value < dc_fixpt_one.value) return filter_3tap_64p_upscale; - else if (ratio.value < dal_fixed31_32_from_fraction(4, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(4, 3).value) return filter_3tap_64p_117; - else if (ratio.value < dal_fixed31_32_from_fraction(5, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(5, 3).value) return filter_3tap_64p_150; else return filter_3tap_64p_183; @@ -1038,11 +1038,11 @@ const uint16_t *get_filter_3tap_64p(struct fixed31_32 ratio) const uint16_t *get_filter_4tap_16p(struct fixed31_32 ratio) { - if (ratio.value < dal_fixed31_32_one.value) + if (ratio.value < dc_fixpt_one.value) return filter_4tap_16p_upscale; - else if (ratio.value < dal_fixed31_32_from_fraction(4, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(4, 3).value) return filter_4tap_16p_117; - else if (ratio.value < dal_fixed31_32_from_fraction(5, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(5, 3).value) return filter_4tap_16p_150; else return filter_4tap_16p_183; @@ -1050,11 +1050,11 @@ const uint16_t *get_filter_4tap_16p(struct fixed31_32 ratio) const uint16_t *get_filter_4tap_64p(struct fixed31_32 ratio) { - if (ratio.value < dal_fixed31_32_one.value) + if (ratio.value < dc_fixpt_one.value) return filter_4tap_64p_upscale; - else if (ratio.value < dal_fixed31_32_from_fraction(4, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(4, 3).value) return filter_4tap_64p_117; - else if (ratio.value < dal_fixed31_32_from_fraction(5, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(5, 3).value) return filter_4tap_64p_150; else return filter_4tap_64p_183; @@ -1062,11 +1062,11 @@ const uint16_t *get_filter_4tap_64p(struct fixed31_32 ratio) const uint16_t *get_filter_5tap_64p(struct fixed31_32 ratio) { - if (ratio.value < dal_fixed31_32_one.value) + if (ratio.value < dc_fixpt_one.value) return filter_5tap_64p_upscale; - else if (ratio.value < dal_fixed31_32_from_fraction(4, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(4, 3).value) return filter_5tap_64p_117; - else if (ratio.value < dal_fixed31_32_from_fraction(5, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(5, 3).value) return filter_5tap_64p_150; else return filter_5tap_64p_183; @@ -1074,11 +1074,11 @@ const uint16_t *get_filter_5tap_64p(struct fixed31_32 ratio) const uint16_t *get_filter_6tap_64p(struct fixed31_32 ratio) { - if (ratio.value < dal_fixed31_32_one.value) + if (ratio.value < dc_fixpt_one.value) return filter_6tap_64p_upscale; - else if (ratio.value < dal_fixed31_32_from_fraction(4, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(4, 3).value) return filter_6tap_64p_117; - else if (ratio.value < dal_fixed31_32_from_fraction(5, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(5, 3).value) return filter_6tap_64p_150; else return filter_6tap_64p_183; @@ -1086,11 +1086,11 @@ const uint16_t *get_filter_6tap_64p(struct fixed31_32 ratio) const uint16_t *get_filter_7tap_64p(struct fixed31_32 ratio) { - if (ratio.value < dal_fixed31_32_one.value) + if (ratio.value < dc_fixpt_one.value) return filter_7tap_64p_upscale; - else if (ratio.value < dal_fixed31_32_from_fraction(4, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(4, 3).value) return filter_7tap_64p_117; - else if (ratio.value < dal_fixed31_32_from_fraction(5, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(5, 3).value) return filter_7tap_64p_150; else return filter_7tap_64p_183; @@ -1098,11 +1098,11 @@ const uint16_t *get_filter_7tap_64p(struct fixed31_32 ratio) const uint16_t *get_filter_8tap_64p(struct fixed31_32 ratio) { - if (ratio.value < dal_fixed31_32_one.value) + if (ratio.value < dc_fixpt_one.value) return filter_8tap_64p_upscale; - else if (ratio.value < dal_fixed31_32_from_fraction(4, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(4, 3).value) return filter_8tap_64p_117; - else if (ratio.value < dal_fixed31_32_from_fraction(5, 3).value) + else if (ratio.value < dc_fixpt_from_fraction(5, 3).value) return filter_8tap_64p_150; else return filter_8tap_64p_183; diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.c b/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.c index 162f6a6c4208f..0a6d483dc046a 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_stream_encoder.c @@ -26,27 +26,10 @@ #include "dc_bios_types.h" #include "dce_stream_encoder.h" #include "reg_helper.h" +#include "hw_shared.h" + #define DC_LOGGER \ enc110->base.ctx->logger -enum DP_PIXEL_ENCODING { -DP_PIXEL_ENCODING_RGB444 = 0x00000000, -DP_PIXEL_ENCODING_YCBCR422 = 0x00000001, -DP_PIXEL_ENCODING_YCBCR444 = 0x00000002, -DP_PIXEL_ENCODING_RGB_WIDE_GAMUT = 0x00000003, -DP_PIXEL_ENCODING_Y_ONLY = 0x00000004, -DP_PIXEL_ENCODING_YCBCR420 = 0x00000005, -DP_PIXEL_ENCODING_RESERVED = 0x00000006, -}; - - -enum DP_COMPONENT_DEPTH { -DP_COMPONENT_DEPTH_6BPC = 0x00000000, -DP_COMPONENT_DEPTH_8BPC = 0x00000001, -DP_COMPONENT_DEPTH_10BPC = 0x00000002, -DP_COMPONENT_DEPTH_12BPC = 0x00000003, -DP_COMPONENT_DEPTH_16BPC = 0x00000004, -DP_COMPONENT_DEPTH_RESERVED = 0x00000005, -}; #define REG(reg)\ @@ -80,7 +63,7 @@ enum { static void dce110_update_generic_info_packet( struct dce110_stream_encoder *enc110, uint32_t packet_index, - const struct encoder_info_packet *info_packet) + const struct dc_info_packet *info_packet) { uint32_t regval; /* TODOFPGA Figure out a proper number for max_retries polling for lock @@ -196,7 +179,7 @@ static void dce110_update_generic_info_packet( static void dce110_update_hdmi_info_packet( struct dce110_stream_encoder *enc110, uint32_t packet_index, - const struct encoder_info_packet *info_packet) + const struct dc_info_packet *info_packet) { uint32_t cont, send, line; @@ -314,11 +297,11 @@ static void dce110_stream_encoder_dp_set_stream_attribute( switch (crtc_timing->pixel_encoding) { case PIXEL_ENCODING_YCBCR422: REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, - DP_PIXEL_ENCODING_YCBCR422); + DP_PIXEL_ENCODING_TYPE_YCBCR422); break; case PIXEL_ENCODING_YCBCR444: REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, - DP_PIXEL_ENCODING_YCBCR444); + DP_PIXEL_ENCODING_TYPE_YCBCR444); if (crtc_timing->flags.Y_ONLY) if (crtc_timing->display_color_depth != COLOR_DEPTH_666) @@ -326,7 +309,7 @@ static void dce110_stream_encoder_dp_set_stream_attribute( * Color depth of Y-only could be * 8, 10, 12, 16 bits */ REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, - DP_PIXEL_ENCODING_Y_ONLY); + DP_PIXEL_ENCODING_TYPE_Y_ONLY); /* Note: DP_MSA_MISC1 bit 7 is the indicator * of Y-only mode. * This bit is set in HW if register @@ -334,7 +317,7 @@ static void dce110_stream_encoder_dp_set_stream_attribute( break; case PIXEL_ENCODING_YCBCR420: REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, - DP_PIXEL_ENCODING_YCBCR420); + DP_PIXEL_ENCODING_TYPE_YCBCR420); if (enc110->se_mask->DP_VID_M_DOUBLE_VALUE_EN) REG_UPDATE(DP_VID_TIMING, DP_VID_M_DOUBLE_VALUE_EN, 1); @@ -345,7 +328,7 @@ static void dce110_stream_encoder_dp_set_stream_attribute( break; default: REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, - DP_PIXEL_ENCODING_RGB444); + DP_PIXEL_ENCODING_TYPE_RGB444); break; } @@ -363,20 +346,20 @@ static void dce110_stream_encoder_dp_set_stream_attribute( break; case COLOR_DEPTH_888: REG_UPDATE(DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, - DP_COMPONENT_DEPTH_8BPC); + DP_COMPONENT_PIXEL_DEPTH_8BPC); break; case COLOR_DEPTH_101010: REG_UPDATE(DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, - DP_COMPONENT_DEPTH_10BPC); + DP_COMPONENT_PIXEL_DEPTH_10BPC); break; case COLOR_DEPTH_121212: REG_UPDATE(DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, - DP_COMPONENT_DEPTH_12BPC); + DP_COMPONENT_PIXEL_DEPTH_12BPC); break; default: REG_UPDATE(DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, - DP_COMPONENT_DEPTH_6BPC); + DP_COMPONENT_PIXEL_DEPTH_6BPC); break; } @@ -700,11 +683,11 @@ static void dce110_stream_encoder_set_mst_bandwidth( struct fixed31_32 avg_time_slots_per_mtp) { struct dce110_stream_encoder *enc110 = DCE110STRENC_FROM_STRENC(enc); - uint32_t x = dal_fixed31_32_floor( + uint32_t x = dc_fixpt_floor( avg_time_slots_per_mtp); - uint32_t y = dal_fixed31_32_ceil( - dal_fixed31_32_shl( - dal_fixed31_32_sub_int( + uint32_t y = dc_fixpt_ceil( + dc_fixpt_shl( + dc_fixpt_sub_int( avg_time_slots_per_mtp, x), 26)); @@ -836,7 +819,7 @@ static void dce110_stream_encoder_update_dp_info_packets( const struct encoder_info_frame *info_frame) { struct dce110_stream_encoder *enc110 = DCE110STRENC_FROM_STRENC(enc); - uint32_t value = REG_READ(DP_SEC_CNTL); + uint32_t value = 0; if (info_frame->vsc.valid) dce110_update_generic_info_packet( @@ -870,6 +853,7 @@ static void dce110_stream_encoder_update_dp_info_packets( * Therefore we need to enable master bit * if at least on of the fields is not 0 */ + value = REG_READ(DP_SEC_CNTL); if (value) REG_UPDATE(DP_SEC_CNTL, DP_SEC_STREAM_ENABLE, 1); } @@ -879,7 +863,7 @@ static void dce110_stream_encoder_stop_dp_info_packets( { /* stop generic packets on DP */ struct dce110_stream_encoder *enc110 = DCE110STRENC_FROM_STRENC(enc); - uint32_t value = REG_READ(DP_SEC_CNTL); + uint32_t value = 0; if (enc110->se_mask->DP_SEC_AVI_ENABLE) { REG_SET_7(DP_SEC_CNTL, 0, @@ -892,25 +876,10 @@ static void dce110_stream_encoder_stop_dp_info_packets( DP_SEC_STREAM_ENABLE, 0); } -#if defined(CONFIG_DRM_AMD_DC_DCN1_0) - if (enc110->se_mask->DP_SEC_GSP7_ENABLE) { - REG_SET_10(DP_SEC_CNTL, 0, - DP_SEC_GSP0_ENABLE, 0, - DP_SEC_GSP1_ENABLE, 0, - DP_SEC_GSP2_ENABLE, 0, - DP_SEC_GSP3_ENABLE, 0, - DP_SEC_GSP4_ENABLE, 0, - DP_SEC_GSP5_ENABLE, 0, - DP_SEC_GSP6_ENABLE, 0, - DP_SEC_GSP7_ENABLE, 0, - DP_SEC_MPG_ENABLE, 0, - DP_SEC_STREAM_ENABLE, 0); - } -#endif /* this register shared with audio info frame. * therefore we need to keep master enabled * if at least one of the fields is not 0 */ - + value = REG_READ(DP_SEC_CNTL); if (value) REG_UPDATE(DP_SEC_CNTL, DP_SEC_STREAM_ENABLE, 1); @@ -1513,7 +1482,7 @@ static void dce110_se_disable_dp_audio( struct stream_encoder *enc) { struct dce110_stream_encoder *enc110 = DCE110STRENC_FROM_STRENC(enc); - uint32_t value = REG_READ(DP_SEC_CNTL); + uint32_t value = 0; /* Disable Audio packets */ REG_UPDATE_5(DP_SEC_CNTL, @@ -1525,6 +1494,7 @@ static void dce110_se_disable_dp_audio( /* This register shared with encoder info frame. Therefore we need to keep master enabled if at least on of the fields is not 0 */ + value = REG_READ(DP_SEC_CNTL); if (value != 0) REG_UPDATE(DP_SEC_CNTL, DP_SEC_STREAM_ENABLE, 1); diff --git a/drivers/gpu/drm/amd/display/dc/dce/dce_transform.c b/drivers/gpu/drm/amd/display/dc/dce/dce_transform.c index 832c5daada354..a02e719d77944 100644 --- a/drivers/gpu/drm/amd/display/dc/dce/dce_transform.c +++ b/drivers/gpu/drm/amd/display/dc/dce/dce_transform.c @@ -41,7 +41,7 @@ #define DC_LOGGER \ xfm_dce->base.ctx->logger -#define IDENTITY_RATIO(ratio) (dal_fixed31_32_u2d19(ratio) == (1 << 19)) +#define IDENTITY_RATIO(ratio) (dc_fixpt_u2d19(ratio) == (1 << 19)) #define GAMUT_MATRIX_SIZE 12 #define SCL_PHASES 16 @@ -256,27 +256,27 @@ static void calculate_inits( struct fixed31_32 v_init; inits->h_int_scale_ratio = - dal_fixed31_32_u2d19(data->ratios.horz) << 5; + dc_fixpt_u2d19(data->ratios.horz) << 5; inits->v_int_scale_ratio = - dal_fixed31_32_u2d19(data->ratios.vert) << 5; + dc_fixpt_u2d19(data->ratios.vert) << 5; h_init = - dal_fixed31_32_div_int( - dal_fixed31_32_add( + dc_fixpt_div_int( + dc_fixpt_add( data->ratios.horz, - dal_fixed31_32_from_int(data->taps.h_taps + 1)), + dc_fixpt_from_int(data->taps.h_taps + 1)), 2); - inits->h_init.integer = dal_fixed31_32_floor(h_init); - inits->h_init.fraction = dal_fixed31_32_u0d19(h_init) << 5; + inits->h_init.integer = dc_fixpt_floor(h_init); + inits->h_init.fraction = dc_fixpt_u0d19(h_init) << 5; v_init = - dal_fixed31_32_div_int( - dal_fixed31_32_add( + dc_fixpt_div_int( + dc_fixpt_add( data->ratios.vert, - dal_fixed31_32_from_int(data->taps.v_taps + 1)), + dc_fixpt_from_int(data->taps.v_taps + 1)), 2); - inits->v_init.integer = dal_fixed31_32_floor(v_init); - inits->v_init.fraction = dal_fixed31_32_u0d19(v_init) << 5; + inits->v_init.integer = dc_fixpt_floor(v_init); + inits->v_init.fraction = dc_fixpt_u0d19(v_init) << 5; } static void program_scl_ratios_inits( diff --git a/drivers/gpu/drm/amd/display/dc/dce100/dce100_resource.c b/drivers/gpu/drm/amd/display/dc/dce100/dce100_resource.c index 3092f76bdb75b..38ec0d609297f 100644 --- a/drivers/gpu/drm/amd/display/dc/dce100/dce100_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dce100/dce100_resource.c @@ -733,38 +733,6 @@ enum dc_status dce100_add_stream_to_ctx( return result; } -enum dc_status dce100_validate_guaranteed( - struct dc *dc, - struct dc_stream_state *dc_stream, - struct dc_state *context) -{ - enum dc_status result = DC_ERROR_UNEXPECTED; - - context->streams[0] = dc_stream; - dc_stream_retain(context->streams[0]); - context->stream_count++; - - result = resource_map_pool_resources(dc, context, dc_stream); - - if (result == DC_OK) - result = resource_map_clock_resources(dc, context, dc_stream); - - if (result == DC_OK) - result = build_mapped_resource(dc, context, dc_stream); - - if (result == DC_OK) { - validate_guaranteed_copy_streams( - context, dc->caps.max_streams); - result = resource_build_scaling_params_for_context(dc, context); - } - - if (result == DC_OK) - if (!dce100_validate_bandwidth(dc, context)) - result = DC_FAIL_BANDWIDTH_VALIDATE; - - return result; -} - static void dce100_destroy_resource_pool(struct resource_pool **pool) { struct dce110_resource_pool *dce110_pool = TO_DCE110_RES_POOL(*pool); @@ -786,7 +754,6 @@ enum dc_status dce100_validate_plane(const struct dc_plane_state *plane_state, s static const struct resource_funcs dce100_res_pool_funcs = { .destroy = dce100_destroy_resource_pool, .link_enc_create = dce100_link_encoder_create, - .validate_guaranteed = dce100_validate_guaranteed, .validate_bandwidth = dce100_validate_bandwidth, .validate_plane = dce100_validate_plane, .add_stream_to_ctx = dce100_add_stream_to_ctx, diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c index d0575999f1728..a92fb0aa2ff3b 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_hw_sequencer.c @@ -70,8 +70,9 @@ #define CTX \ hws->ctx -#define DC_LOGGER \ - ctx->logger + +#define DC_LOGGER_INIT() + #define REG(reg)\ hws->regs->reg @@ -279,7 +280,9 @@ dce110_set_input_transfer_func(struct pipe_ctx *pipe_ctx, build_prescale_params(&prescale_params, plane_state); ipp->funcs->ipp_program_prescale(ipp, &prescale_params); - if (plane_state->gamma_correction && dce_use_lut(plane_state->format)) + if (plane_state->gamma_correction && + !plane_state->gamma_correction->is_identity && + dce_use_lut(plane_state->format)) ipp->funcs->ipp_program_input_lut(ipp, plane_state->gamma_correction); if (tf == NULL) { @@ -506,19 +509,19 @@ dce110_translate_regamma_to_hw_format(const struct dc_transfer_func *output_tf, rgb_resulted[hw_points - 1].green = output_tf->tf_pts.green[start_index]; rgb_resulted[hw_points - 1].blue = output_tf->tf_pts.blue[start_index]; - arr_points[0].x = dal_fixed31_32_pow(dal_fixed31_32_from_int(2), - dal_fixed31_32_from_int(region_start)); - arr_points[1].x = dal_fixed31_32_pow(dal_fixed31_32_from_int(2), - dal_fixed31_32_from_int(region_end)); + arr_points[0].x = dc_fixpt_pow(dc_fixpt_from_int(2), + dc_fixpt_from_int(region_start)); + arr_points[1].x = dc_fixpt_pow(dc_fixpt_from_int(2), + dc_fixpt_from_int(region_end)); y_r = rgb_resulted[0].red; y_g = rgb_resulted[0].green; y_b = rgb_resulted[0].blue; - y1_min = dal_fixed31_32_min(y_r, dal_fixed31_32_min(y_g, y_b)); + y1_min = dc_fixpt_min(y_r, dc_fixpt_min(y_g, y_b)); arr_points[0].y = y1_min; - arr_points[0].slope = dal_fixed31_32_div(arr_points[0].y, + arr_points[0].slope = dc_fixpt_div(arr_points[0].y, arr_points[0].x); y_r = rgb_resulted[hw_points - 1].red; @@ -528,21 +531,21 @@ dce110_translate_regamma_to_hw_format(const struct dc_transfer_func *output_tf, /* see comment above, m_arrPoints[1].y should be the Y value for the * region end (m_numOfHwPoints), not last HW point(m_numOfHwPoints - 1) */ - y3_max = dal_fixed31_32_max(y_r, dal_fixed31_32_max(y_g, y_b)); + y3_max = dc_fixpt_max(y_r, dc_fixpt_max(y_g, y_b)); arr_points[1].y = y3_max; - arr_points[1].slope = dal_fixed31_32_zero; + arr_points[1].slope = dc_fixpt_zero; if (output_tf->tf == TRANSFER_FUNCTION_PQ) { /* for PQ, we want to have a straight line from last HW X point, * and the slope to be such that we hit 1.0 at 10000 nits. */ - const struct fixed31_32 end_value = dal_fixed31_32_from_int(125); + const struct fixed31_32 end_value = dc_fixpt_from_int(125); - arr_points[1].slope = dal_fixed31_32_div( - dal_fixed31_32_sub(dal_fixed31_32_one, arr_points[1].y), - dal_fixed31_32_sub(end_value, arr_points[1].x)); + arr_points[1].slope = dc_fixpt_div( + dc_fixpt_sub(dc_fixpt_one, arr_points[1].y), + dc_fixpt_sub(end_value, arr_points[1].x)); } regamma_params->hw_points_num = hw_points; @@ -566,16 +569,16 @@ dce110_translate_regamma_to_hw_format(const struct dc_transfer_func *output_tf, i = 1; while (i != hw_points + 1) { - if (dal_fixed31_32_lt(rgb_plus_1->red, rgb->red)) + if (dc_fixpt_lt(rgb_plus_1->red, rgb->red)) rgb_plus_1->red = rgb->red; - if (dal_fixed31_32_lt(rgb_plus_1->green, rgb->green)) + if (dc_fixpt_lt(rgb_plus_1->green, rgb->green)) rgb_plus_1->green = rgb->green; - if (dal_fixed31_32_lt(rgb_plus_1->blue, rgb->blue)) + if (dc_fixpt_lt(rgb_plus_1->blue, rgb->blue)) rgb_plus_1->blue = rgb->blue; - rgb->delta_red = dal_fixed31_32_sub(rgb_plus_1->red, rgb->red); - rgb->delta_green = dal_fixed31_32_sub(rgb_plus_1->green, rgb->green); - rgb->delta_blue = dal_fixed31_32_sub(rgb_plus_1->blue, rgb->blue); + rgb->delta_red = dc_fixpt_sub(rgb_plus_1->red, rgb->red); + rgb->delta_green = dc_fixpt_sub(rgb_plus_1->green, rgb->green); + rgb->delta_blue = dc_fixpt_sub(rgb_plus_1->blue, rgb->blue); ++rgb_plus_1; ++rgb; @@ -851,6 +854,28 @@ void hwss_edp_power_control( if (power_up != is_panel_powered_on(hwseq)) { /* Send VBIOS command to prompt eDP panel power */ + if (power_up) { + unsigned long long current_ts = dm_get_timestamp(ctx); + unsigned long long duration_in_ms = + dm_get_elapse_time_in_ns( + ctx, + current_ts, + div64_u64(link->link_trace.time_stamp.edp_poweroff, 1000000)); + unsigned long long wait_time_ms = 0; + + /* max 500ms from LCDVDD off to on */ + if (link->link_trace.time_stamp.edp_poweroff == 0) + wait_time_ms = 500; + else if (duration_in_ms < 500) + wait_time_ms = 500 - duration_in_ms; + + if (wait_time_ms) { + msleep(wait_time_ms); + dm_output_to_console("%s: wait %lld ms to power on eDP.\n", + __func__, wait_time_ms); + } + + } DC_LOG_HW_RESUME_S3( "%s: Panel Power action: %s\n", @@ -864,9 +889,14 @@ void hwss_edp_power_control( cntl.coherent = false; cntl.lanes_number = LANE_COUNT_FOUR; cntl.hpd_sel = link->link_enc->hpd_source; - bp_result = link_transmitter_control(ctx->dc_bios, &cntl); + if (!power_up) + /*save driver power off time stamp*/ + link->link_trace.time_stamp.edp_poweroff = dm_get_timestamp(ctx); + else + link->link_trace.time_stamp.edp_poweron = dm_get_timestamp(ctx); + if (bp_result != BP_RESULT_OK) DC_LOG_ERROR( "%s: Panel Power bp_result: %d\n", @@ -1011,7 +1041,7 @@ void dce110_unblank_stream(struct pipe_ctx *pipe_ctx, if (link->local_sink && link->local_sink->sink_signal == SIGNAL_TYPE_EDP) { link->dc->hwss.edp_backlight_control(link, true); - stream->bl_pwm_level = 0; + stream->bl_pwm_level = EDP_BACKLIGHT_RAMP_DISABLE_LEVEL; } } void dce110_blank_stream(struct pipe_ctx *pipe_ctx) @@ -1203,7 +1233,7 @@ static void program_scaler(const struct dc *dc, &pipe_ctx->plane_res.scl_data); } -static enum dc_status dce110_prog_pixclk_crtc_otg( +static enum dc_status dce110_enable_stream_timing( struct pipe_ctx *pipe_ctx, struct dc_state *context, struct dc *dc) @@ -1269,7 +1299,7 @@ static enum dc_status apply_single_controller_ctx_to_hw( pipe_ctx[pipe_ctx->pipe_idx]; /* */ - dc->hwss.prog_pixclk_crtc_otg(pipe_ctx, context, dc); + dc->hwss.enable_stream_timing(pipe_ctx, context, dc); /* FPGA does not program backend */ if (IS_FPGA_MAXIMUS_DC(dc->ctx->dce_environment)) { @@ -1441,6 +1471,17 @@ static void disable_vga_and_power_gate_all_controllers( } } +static struct dc_link *get_link_for_edp(struct dc *dc) +{ + int i; + + for (i = 0; i < dc->link_count; i++) { + if (dc->links[i]->connector_signal == SIGNAL_TYPE_EDP) + return dc->links[i]; + } + return NULL; +} + static struct dc_link *get_link_for_edp_not_in_use( struct dc *dc, struct dc_state *context) @@ -1475,20 +1516,21 @@ static struct dc_link *get_link_for_edp_not_in_use( */ void dce110_enable_accelerated_mode(struct dc *dc, struct dc_state *context) { - struct dc_bios *dcb = dc->ctx->dc_bios; - - /* vbios already light up eDP, so we can leverage vbios and skip eDP - * programming - */ - bool can_eDP_fast_boot_optimize = - (dcb->funcs->get_vga_enabled_displays(dc->ctx->dc_bios) == ATOM_DISPLAY_LCD1_ACTIVE); - - /* if OS doesn't light up eDP and eDP link is available, we want to disable */ struct dc_link *edp_link_to_turnoff = NULL; + struct dc_link *edp_link = get_link_for_edp(dc); + bool can_eDP_fast_boot_optimize = false; + + if (edp_link) { + can_eDP_fast_boot_optimize = + edp_link->link_enc->funcs->is_dig_enabled(edp_link->link_enc); + } if (can_eDP_fast_boot_optimize) { edp_link_to_turnoff = get_link_for_edp_not_in_use(dc, context); + /* if OS doesn't light up eDP and eDP link is available, we want to disable + * If resume from S4/S5, should optimization. + */ if (!edp_link_to_turnoff) dc->apply_edp_fast_boot_optimization = true; } @@ -1544,6 +1586,7 @@ static void dce110_set_displaymarks( pipe_ctx->plane_res.mi, context->bw.dce.nbp_state_change_wm_ns[num_pipes], context->bw.dce.stutter_exit_wm_ns[num_pipes], + context->bw.dce.stutter_entry_wm_ns[num_pipes], context->bw.dce.urgent_wm_ns[num_pipes], total_dest_line_time_ns); if (i == underlay_idx) { @@ -1569,6 +1612,7 @@ static void set_safe_displaymarks( MAX_WATERMARK, MAX_WATERMARK, MAX_WATERMARK, MAX_WATERMARK }; struct dce_watermarks nbp_marks = { SAFE_NBP_MARK, SAFE_NBP_MARK, SAFE_NBP_MARK, SAFE_NBP_MARK }; + struct dce_watermarks min_marks = { 0, 0, 0, 0}; for (i = 0; i < MAX_PIPES; i++) { if (res_ctx->pipe_ctx[i].stream == NULL || res_ctx->pipe_ctx[i].plane_res.mi == NULL) @@ -1578,6 +1622,7 @@ static void set_safe_displaymarks( res_ctx->pipe_ctx[i].plane_res.mi, nbp_marks, max_marks, + min_marks, max_marks, MAX_WATERMARK); @@ -1803,6 +1848,9 @@ static bool should_enable_fbc(struct dc *dc, } } + /* Pipe context should be found */ + ASSERT(pipe_ctx); + /* Only supports eDP */ if (pipe_ctx->stream->sink->link->connector_signal != SIGNAL_TYPE_EDP) return false; @@ -2221,74 +2269,6 @@ static void program_gamut_remap(struct pipe_ctx *pipe_ctx) pipe_ctx->plane_res.xfm->funcs->transform_set_gamut_remap(pipe_ctx->plane_res.xfm, &adjust); } - -/** - * TODO REMOVE, USE UPDATE INSTEAD - */ -static void set_plane_config( - const struct dc *dc, - struct pipe_ctx *pipe_ctx, - struct resource_context *res_ctx) -{ - struct mem_input *mi = pipe_ctx->plane_res.mi; - struct dc_plane_state *plane_state = pipe_ctx->plane_state; - struct xfm_grph_csc_adjustment adjust; - struct out_csc_color_matrix tbl_entry; - unsigned int i; - - memset(&adjust, 0, sizeof(adjust)); - memset(&tbl_entry, 0, sizeof(tbl_entry)); - adjust.gamut_adjust_type = GRAPHICS_GAMUT_ADJUST_TYPE_BYPASS; - - dce_enable_fe_clock(dc->hwseq, mi->inst, true); - - set_default_colors(pipe_ctx); - if (pipe_ctx->stream->csc_color_matrix.enable_adjustment == true) { - tbl_entry.color_space = - pipe_ctx->stream->output_color_space; - - for (i = 0; i < 12; i++) - tbl_entry.regval[i] = - pipe_ctx->stream->csc_color_matrix.matrix[i]; - - pipe_ctx->plane_res.xfm->funcs->opp_set_csc_adjustment - (pipe_ctx->plane_res.xfm, &tbl_entry); - } - - if (pipe_ctx->stream->gamut_remap_matrix.enable_remap == true) { - adjust.gamut_adjust_type = GRAPHICS_GAMUT_ADJUST_TYPE_SW; - - for (i = 0; i < CSC_TEMPERATURE_MATRIX_SIZE; i++) - adjust.temperature_matrix[i] = - pipe_ctx->stream->gamut_remap_matrix.matrix[i]; - } - - pipe_ctx->plane_res.xfm->funcs->transform_set_gamut_remap(pipe_ctx->plane_res.xfm, &adjust); - - pipe_ctx->plane_res.scl_data.lb_params.alpha_en = pipe_ctx->bottom_pipe != 0; - program_scaler(dc, pipe_ctx); - - program_surface_visibility(dc, pipe_ctx); - - mi->funcs->mem_input_program_surface_config( - mi, - plane_state->format, - &plane_state->tiling_info, - &plane_state->plane_size, - plane_state->rotation, - NULL, - false); - if (mi->funcs->set_blank) - mi->funcs->set_blank(mi, pipe_ctx->plane_state->visible); - - if (dc->config.gpu_vm_support) - mi->funcs->mem_input_program_pte_vm( - pipe_ctx->plane_res.mi, - plane_state->format, - &plane_state->tiling_info, - plane_state->rotation); -} - static void update_plane_addr(const struct dc *dc, struct pipe_ctx *pipe_ctx) { @@ -2699,8 +2679,11 @@ static void dce110_program_front_end_for_pipe( struct dc_plane_state *plane_state = pipe_ctx->plane_state; struct xfm_grph_csc_adjustment adjust; struct out_csc_color_matrix tbl_entry; +#if defined(CONFIG_DRM_AMD_DC_FBC) + unsigned int underlay_idx = dc->res_pool->underlay_pipe_index; +#endif unsigned int i; - struct dc_context *ctx = dc->ctx; + DC_LOGGER_INIT(); memset(&tbl_entry, 0, sizeof(tbl_entry)); if (dc->current_state) @@ -2740,7 +2723,9 @@ static void dce110_program_front_end_for_pipe( program_scaler(dc, pipe_ctx); #if defined(CONFIG_DRM_AMD_DC_FBC) - if (dc->fbc_compressor && old_pipe->stream) { + /* fbc not applicable on Underlay pipe */ + if (dc->fbc_compressor && old_pipe->stream && + pipe_ctx->pipe_idx != underlay_idx) { if (plane_state->tiling_info.gfx8.array_mode == DC_ARRAY_LINEAR_GENERAL) dc->fbc_compressor->funcs->disable_fbc(dc->fbc_compressor); else @@ -2776,13 +2761,13 @@ static void dce110_program_front_end_for_pipe( dc->hwss.set_output_transfer_func(pipe_ctx, pipe_ctx->stream); DC_LOG_SURFACE( - "Pipe:%d 0x%x: addr hi:0x%x, " + "Pipe:%d %p: addr hi:0x%x, " "addr low:0x%x, " "src: %d, %d, %d," " %d; dst: %d, %d, %d, %d;" "clip: %d, %d, %d, %d\n", pipe_ctx->pipe_idx, - pipe_ctx->plane_state, + (void *) pipe_ctx->plane_state, pipe_ctx->plane_state->address.grph.addr.high_part, pipe_ctx->plane_state->address.grph.addr.low_part, pipe_ctx->plane_state->src_rect.x, @@ -2970,7 +2955,6 @@ static const struct hw_sequencer_funcs dce110_funcs = { .init_hw = init_hw, .apply_ctx_to_hw = dce110_apply_ctx_to_hw, .apply_ctx_for_surface = dce110_apply_ctx_for_surface, - .set_plane_config = set_plane_config, .update_plane_addr = update_plane_addr, .update_pending_status = dce110_update_pending_status, .set_input_transfer_func = dce110_set_input_transfer_func, @@ -2993,7 +2977,7 @@ static const struct hw_sequencer_funcs dce110_funcs = { .get_position = get_position, .set_static_screen_control = set_static_screen_control, .reset_hw_ctx_wrap = dce110_reset_hw_ctx_wrap, - .prog_pixclk_crtc_otg = dce110_prog_pixclk_crtc_otg, + .enable_stream_timing = dce110_enable_stream_timing, .setup_stereo = NULL, .set_avmute = dce110_set_avmute, .wait_for_mpcc_disconnect = dce110_wait_for_mpcc_disconnect, diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_mem_input_v.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_mem_input_v.c index 7bab8c6d2a738..0564c8e312529 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_mem_input_v.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_mem_input_v.c @@ -923,6 +923,7 @@ void dce_mem_input_v_program_display_marks( struct mem_input *mem_input, struct dce_watermarks nbp, struct dce_watermarks stutter, + struct dce_watermarks stutter_enter, struct dce_watermarks urgent, uint32_t total_dest_line_time_ns) { diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c index b1f14be20fdff..ee33786bdef63 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_resource.c @@ -930,38 +930,6 @@ static enum dc_status dce110_add_stream_to_ctx( return result; } -static enum dc_status dce110_validate_guaranteed( - struct dc *dc, - struct dc_stream_state *dc_stream, - struct dc_state *context) -{ - enum dc_status result = DC_ERROR_UNEXPECTED; - - context->streams[0] = dc_stream; - dc_stream_retain(context->streams[0]); - context->stream_count++; - - result = resource_map_pool_resources(dc, context, dc_stream); - - if (result == DC_OK) - result = resource_map_clock_resources(dc, context, dc_stream); - - if (result == DC_OK) - result = build_mapped_resource(dc, context, dc_stream); - - if (result == DC_OK) { - validate_guaranteed_copy_streams( - context, dc->caps.max_streams); - result = resource_build_scaling_params_for_context(dc, context); - } - - if (result == DC_OK) - if (!dce110_validate_bandwidth(dc, context)) - result = DC_FAIL_BANDWIDTH_VALIDATE; - - return result; -} - static struct pipe_ctx *dce110_acquire_underlay( struct dc_state *context, const struct resource_pool *pool, @@ -1036,7 +1004,6 @@ static void dce110_destroy_resource_pool(struct resource_pool **pool) static const struct resource_funcs dce110_res_pool_funcs = { .destroy = dce110_destroy_resource_pool, .link_enc_create = dce110_link_encoder_create, - .validate_guaranteed = dce110_validate_guaranteed, .validate_bandwidth = dce110_validate_bandwidth, .validate_plane = dce110_validate_plane, .acquire_idle_pipe_for_layer = dce110_acquire_underlay, diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_timing_generator.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_timing_generator.c index be7153924a708..1b2fe0df347f7 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_timing_generator.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_timing_generator.c @@ -431,14 +431,6 @@ void dce110_timing_generator_set_drr( 0, CRTC_V_TOTAL_CONTROL, CRTC_SET_V_TOTAL_MIN_MASK); - set_reg_field_value(v_total_min, - 0, - CRTC_V_TOTAL_MIN, - CRTC_V_TOTAL_MIN); - set_reg_field_value(v_total_max, - 0, - CRTC_V_TOTAL_MAX, - CRTC_V_TOTAL_MAX); set_reg_field_value(v_total_cntl, 0, CRTC_V_TOTAL_CONTROL, @@ -447,6 +439,14 @@ void dce110_timing_generator_set_drr( 0, CRTC_V_TOTAL_CONTROL, CRTC_V_TOTAL_MAX_SEL); + set_reg_field_value(v_total_min, + 0, + CRTC_V_TOTAL_MIN, + CRTC_V_TOTAL_MIN); + set_reg_field_value(v_total_max, + 0, + CRTC_V_TOTAL_MAX, + CRTC_V_TOTAL_MAX); set_reg_field_value(v_total_cntl, 0, CRTC_V_TOTAL_CONTROL, diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_timing_generator_v.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_timing_generator_v.c index 8ad04816e7d31..a3cef60380edb 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_timing_generator_v.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_timing_generator_v.c @@ -648,12 +648,6 @@ static void dce110_timing_generator_v_disable_vga( return; } -static bool dce110_tg_v_is_blanked(struct timing_generator *tg) -{ - /* Signal comes from the primary pipe, underlay is never blanked. */ - return false; -} - /** ******************************************************************************************** * * DCE11 Timing Generator Constructor / Destructor @@ -670,7 +664,6 @@ static const struct timing_generator_funcs dce110_tg_v_funcs = { .set_early_control = dce110_timing_generator_v_set_early_control, .wait_for_state = dce110_timing_generator_v_wait_for_state, .set_blank = dce110_timing_generator_v_set_blank, - .is_blanked = dce110_tg_v_is_blanked, .set_colors = dce110_timing_generator_v_set_colors, .set_overscan_blank_color = dce110_timing_generator_v_set_overscan_color_black, diff --git a/drivers/gpu/drm/amd/display/dc/dce110/dce110_transform_v.c b/drivers/gpu/drm/amd/display/dc/dce110/dce110_transform_v.c index 8ba3c12fc6084..a7dce060204fc 100644 --- a/drivers/gpu/drm/amd/display/dc/dce110/dce110_transform_v.c +++ b/drivers/gpu/drm/amd/display/dc/dce110/dce110_transform_v.c @@ -373,13 +373,13 @@ static void calculate_inits( struct rect *chroma_viewport) { inits->h_int_scale_ratio_luma = - dal_fixed31_32_u2d19(data->ratios.horz) << 5; + dc_fixpt_u2d19(data->ratios.horz) << 5; inits->v_int_scale_ratio_luma = - dal_fixed31_32_u2d19(data->ratios.vert) << 5; + dc_fixpt_u2d19(data->ratios.vert) << 5; inits->h_int_scale_ratio_chroma = - dal_fixed31_32_u2d19(data->ratios.horz_c) << 5; + dc_fixpt_u2d19(data->ratios.horz_c) << 5; inits->v_int_scale_ratio_chroma = - dal_fixed31_32_u2d19(data->ratios.vert_c) << 5; + dc_fixpt_u2d19(data->ratios.vert_c) << 5; inits->h_init_luma.integer = 1; inits->v_init_luma.integer = 1; diff --git a/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.c b/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.c index cd1e3f72c44e1..00c0a1ef15ebd 100644 --- a/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.c @@ -430,7 +430,7 @@ static struct stream_encoder *dce112_stream_encoder_create( if (!enc110) return NULL; - + dce110_stream_encoder_construct(enc110, ctx, ctx->dc_bios, eng_id, &stream_enc_regs[eng_id], &se_shift, &se_mask); @@ -867,38 +867,6 @@ enum dc_status dce112_add_stream_to_ctx( return result; } -enum dc_status dce112_validate_guaranteed( - struct dc *dc, - struct dc_stream_state *stream, - struct dc_state *context) -{ - enum dc_status result = DC_ERROR_UNEXPECTED; - - context->streams[0] = stream; - dc_stream_retain(context->streams[0]); - context->stream_count++; - - result = resource_map_pool_resources(dc, context, stream); - - if (result == DC_OK) - result = resource_map_phy_clock_resources(dc, context, stream); - - if (result == DC_OK) - result = build_mapped_resource(dc, context, stream); - - if (result == DC_OK) { - validate_guaranteed_copy_streams( - context, dc->caps.max_streams); - result = resource_build_scaling_params_for_context(dc, context); - } - - if (result == DC_OK) - if (!dce112_validate_bandwidth(dc, context)) - result = DC_FAIL_BANDWIDTH_VALIDATE; - - return result; -} - enum dc_status dce112_validate_global( struct dc *dc, struct dc_state *context) @@ -921,7 +889,6 @@ static void dce112_destroy_resource_pool(struct resource_pool **pool) static const struct resource_funcs dce112_res_pool_funcs = { .destroy = dce112_destroy_resource_pool, .link_enc_create = dce112_link_encoder_create, - .validate_guaranteed = dce112_validate_guaranteed, .validate_bandwidth = dce112_validate_bandwidth, .validate_plane = dce100_validate_plane, .add_stream_to_ctx = dce112_add_stream_to_ctx, diff --git a/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.h b/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.h index d5c19d34eb0a4..95a4033962196 100644 --- a/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.h +++ b/drivers/gpu/drm/amd/display/dc/dce112/dce112_resource.h @@ -42,11 +42,6 @@ enum dc_status dce112_validate_with_context( struct dc_state *context, struct dc_state *old_context); -enum dc_status dce112_validate_guaranteed( - struct dc *dc, - struct dc_stream_state *dc_stream, - struct dc_state *context); - bool dce112_validate_bandwidth( struct dc *dc, struct dc_state *context); diff --git a/drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c b/drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c index 4659a4bfabaa2..2d58daccc0056 100644 --- a/drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dce120/dce120_resource.c @@ -652,7 +652,7 @@ static struct mem_input *dce120_mem_input_create( return NULL; } - dce112_mem_input_construct(dce_mi, ctx, inst, &mi_regs[inst], &mi_shifts, &mi_masks); + dce120_mem_input_construct(dce_mi, ctx, inst, &mi_regs[inst], &mi_shifts, &mi_masks); return &dce_mi->base; } @@ -684,7 +684,6 @@ static void dce120_destroy_resource_pool(struct resource_pool **pool) static const struct resource_funcs dce120_res_pool_funcs = { .destroy = dce120_destroy_resource_pool, .link_enc_create = dce120_link_encoder_create, - .validate_guaranteed = dce112_validate_guaranteed, .validate_bandwidth = dce112_validate_bandwidth, .validate_plane = dce100_validate_plane, .add_stream_to_ctx = dce112_add_stream_to_ctx @@ -815,14 +814,25 @@ static void bw_calcs_data_update_from_pplib(struct dc *dc) dm_pp_notify_wm_clock_changes(dc->ctx, &clk_ranges); } +static uint32_t read_pipe_fuses(struct dc_context *ctx) +{ + uint32_t value = dm_read_reg_soc15(ctx, mmCC_DC_PIPE_DIS, 0); + /* VG20 support max 6 pipes */ + value = value & 0x3f; + return value; +} + static bool construct( uint8_t num_virtual_links, struct dc *dc, struct dce110_resource_pool *pool) { unsigned int i; + int j; struct dc_context *ctx = dc->ctx; struct irq_service_init_data irq_init_data; + bool harvest_enabled = ASICREV_IS_VEGA20_P(ctx->asic_id.hw_internal_rev); + uint32_t pipe_fuses; ctx->dc_bios->regs = &bios_regs; @@ -916,28 +926,41 @@ static bool construct( if (!pool->base.irqs) goto irqs_create_fail; + /* retrieve valid pipe fuses */ + if (harvest_enabled) + pipe_fuses = read_pipe_fuses(ctx); + + /* index to valid pipe resource */ + j = 0; for (i = 0; i < pool->base.pipe_count; i++) { - pool->base.timing_generators[i] = + if (harvest_enabled) { + if ((pipe_fuses & (1 << i)) != 0) { + dm_error("DC: skip invalid pipe %d!\n", i); + continue; + } + } + + pool->base.timing_generators[j] = dce120_timing_generator_create( ctx, i, &dce120_tg_offsets[i]); - if (pool->base.timing_generators[i] == NULL) { + if (pool->base.timing_generators[j] == NULL) { BREAK_TO_DEBUGGER(); dm_error("DC: failed to create tg!\n"); goto controller_create_fail; } - pool->base.mis[i] = dce120_mem_input_create(ctx, i); + pool->base.mis[j] = dce120_mem_input_create(ctx, i); - if (pool->base.mis[i] == NULL) { + if (pool->base.mis[j] == NULL) { BREAK_TO_DEBUGGER(); dm_error( "DC: failed to create memory input!\n"); goto controller_create_fail; } - pool->base.ipps[i] = dce120_ipp_create(ctx, i); + pool->base.ipps[j] = dce120_ipp_create(ctx, i); if (pool->base.ipps[i] == NULL) { BREAK_TO_DEBUGGER(); dm_error( @@ -945,7 +968,7 @@ static bool construct( goto controller_create_fail; } - pool->base.transforms[i] = dce120_transform_create(ctx, i); + pool->base.transforms[j] = dce120_transform_create(ctx, i); if (pool->base.transforms[i] == NULL) { BREAK_TO_DEBUGGER(); dm_error( @@ -953,16 +976,23 @@ static bool construct( goto res_create_fail; } - pool->base.opps[i] = dce120_opp_create( + pool->base.opps[j] = dce120_opp_create( ctx, i); - if (pool->base.opps[i] == NULL) { + if (pool->base.opps[j] == NULL) { BREAK_TO_DEBUGGER(); dm_error( "DC: failed to create output pixel processor!\n"); } + + /* check next valid pipe */ + j++; } + /* valid pipe num */ + pool->base.pipe_count = j; + pool->base.timing_generator_count = j; + if (!resource_construct(num_virtual_links, dc, &pool->base, &res_create_funcs)) goto res_create_fail; diff --git a/drivers/gpu/drm/amd/display/dc/dce120/dce120_timing_generator.c b/drivers/gpu/drm/amd/display/dc/dce120/dce120_timing_generator.c index 7bee78172d853..2ea490f8482e0 100644 --- a/drivers/gpu/drm/amd/display/dc/dce120/dce120_timing_generator.c +++ b/drivers/gpu/drm/amd/display/dc/dce120/dce120_timing_generator.c @@ -570,12 +570,6 @@ void dce120_timing_generator_set_drr( 0x180); } else { - CRTC_REG_UPDATE( - CRTC0_CRTC_V_TOTAL_MIN, - CRTC_V_TOTAL_MIN, 0); - CRTC_REG_UPDATE( - CRTC0_CRTC_V_TOTAL_MAX, - CRTC_V_TOTAL_MAX, 0); CRTC_REG_SET_N(CRTC0_CRTC_V_TOTAL_CONTROL, 5, FD(CRTC0_CRTC_V_TOTAL_CONTROL__CRTC_V_TOTAL_MIN_SEL), 0, FD(CRTC0_CRTC_V_TOTAL_CONTROL__CRTC_V_TOTAL_MAX_SEL), 0, @@ -583,6 +577,12 @@ void dce120_timing_generator_set_drr( FD(CRTC0_CRTC_V_TOTAL_CONTROL__CRTC_FORCE_LOCK_TO_MASTER_VSYNC), 0, FD(CRTC0_CRTC_V_TOTAL_CONTROL__CRTC_SET_V_TOTAL_MIN_MASK), 0); CRTC_REG_UPDATE( + CRTC0_CRTC_V_TOTAL_MIN, + CRTC_V_TOTAL_MIN, 0); + CRTC_REG_UPDATE( + CRTC0_CRTC_V_TOTAL_MAX, + CRTC_V_TOTAL_MAX, 0); + CRTC_REG_UPDATE( CRTC0_CRTC_STATIC_SCREEN_CONTROL, CRTC_STATIC_SCREEN_EVENT_MASK, 0); diff --git a/drivers/gpu/drm/amd/display/dc/dce80/dce80_resource.c b/drivers/gpu/drm/amd/display/dc/dce80/dce80_resource.c index 5d854a37a9789..48a0689647225 100644 --- a/drivers/gpu/drm/amd/display/dc/dce80/dce80_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dce80/dce80_resource.c @@ -691,23 +691,6 @@ static void destruct(struct dce110_resource_pool *pool) } } -static enum dc_status build_mapped_resource( - const struct dc *dc, - struct dc_state *context, - struct dc_stream_state *stream) -{ - struct pipe_ctx *pipe_ctx = resource_get_head_pipe_for_stream(&context->res_ctx, stream); - - if (!pipe_ctx) - return DC_ERROR_UNEXPECTED; - - dce110_resource_build_pipe_hw_param(pipe_ctx); - - resource_build_info_frame(pipe_ctx); - - return DC_OK; -} - bool dce80_validate_bandwidth( struct dc *dc, struct dc_state *context) @@ -749,37 +732,6 @@ enum dc_status dce80_validate_global( return DC_OK; } -enum dc_status dce80_validate_guaranteed( - struct dc *dc, - struct dc_stream_state *dc_stream, - struct dc_state *context) -{ - enum dc_status result = DC_ERROR_UNEXPECTED; - - context->streams[0] = dc_stream; - dc_stream_retain(context->streams[0]); - context->stream_count++; - - result = resource_map_pool_resources(dc, context, dc_stream); - - if (result == DC_OK) - result = resource_map_clock_resources(dc, context, dc_stream); - - if (result == DC_OK) - result = build_mapped_resource(dc, context, dc_stream); - - if (result == DC_OK) { - validate_guaranteed_copy_streams( - context, dc->caps.max_streams); - result = resource_build_scaling_params_for_context(dc, context); - } - - if (result == DC_OK) - result = dce80_validate_bandwidth(dc, context); - - return result; -} - static void dce80_destroy_resource_pool(struct resource_pool **pool) { struct dce110_resource_pool *dce110_pool = TO_DCE110_RES_POOL(*pool); @@ -792,7 +744,6 @@ static void dce80_destroy_resource_pool(struct resource_pool **pool) static const struct resource_funcs dce80_res_pool_funcs = { .destroy = dce80_destroy_resource_pool, .link_enc_create = dce80_link_encoder_create, - .validate_guaranteed = dce80_validate_guaranteed, .validate_bandwidth = dce80_validate_bandwidth, .validate_plane = dce100_validate_plane, .add_stream_to_ctx = dce100_add_stream_to_ctx, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/Makefile b/drivers/gpu/drm/amd/display/dc/dcn10/Makefile index 5469bdfe19f35..84f52c63d95c7 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/Makefile +++ b/drivers/gpu/drm/amd/display/dc/dcn10/Makefile @@ -26,7 +26,7 @@ DCN10 = dcn10_resource.o dcn10_ipp.o dcn10_hw_sequencer.o \ dcn10_dpp.o dcn10_opp.o dcn10_optc.o \ dcn10_hubp.o dcn10_mpc.o \ dcn10_dpp_dscl.o dcn10_dpp_cm.o dcn10_cm_common.o \ - dcn10_hubbub.o + dcn10_hubbub.o dcn10_stream_encoder.o dcn10_link_encoder.o AMD_DAL_DCN10 = $(addprefix $(AMDDALPATH)/dc/dcn10/,$(DCN10)) diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c index 881a1bff94d2c..5d95a997fd9f9 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_cm_common.c @@ -169,7 +169,7 @@ bool cm_helper_convert_to_custom_float( } if (fixpoint == true) - arr_points[1].custom_float_y = dal_fixed31_32_clamp_u0d14(arr_points[1].y); + arr_points[1].custom_float_y = dc_fixpt_clamp_u0d14(arr_points[1].y); else if (!convert_to_custom_float_format(arr_points[1].y, &fmt, &arr_points[1].custom_float_y)) { BREAK_TO_DEBUGGER(); @@ -327,19 +327,19 @@ bool cm_helper_translate_curve_to_hw_format( rgb_resulted[hw_points - 1].green = output_tf->tf_pts.green[start_index]; rgb_resulted[hw_points - 1].blue = output_tf->tf_pts.blue[start_index]; - arr_points[0].x = dal_fixed31_32_pow(dal_fixed31_32_from_int(2), - dal_fixed31_32_from_int(region_start)); - arr_points[1].x = dal_fixed31_32_pow(dal_fixed31_32_from_int(2), - dal_fixed31_32_from_int(region_end)); + arr_points[0].x = dc_fixpt_pow(dc_fixpt_from_int(2), + dc_fixpt_from_int(region_start)); + arr_points[1].x = dc_fixpt_pow(dc_fixpt_from_int(2), + dc_fixpt_from_int(region_end)); y_r = rgb_resulted[0].red; y_g = rgb_resulted[0].green; y_b = rgb_resulted[0].blue; - y1_min = dal_fixed31_32_min(y_r, dal_fixed31_32_min(y_g, y_b)); + y1_min = dc_fixpt_min(y_r, dc_fixpt_min(y_g, y_b)); arr_points[0].y = y1_min; - arr_points[0].slope = dal_fixed31_32_div(arr_points[0].y, arr_points[0].x); + arr_points[0].slope = dc_fixpt_div(arr_points[0].y, arr_points[0].x); y_r = rgb_resulted[hw_points - 1].red; y_g = rgb_resulted[hw_points - 1].green; y_b = rgb_resulted[hw_points - 1].blue; @@ -347,35 +347,35 @@ bool cm_helper_translate_curve_to_hw_format( /* see comment above, m_arrPoints[1].y should be the Y value for the * region end (m_numOfHwPoints), not last HW point(m_numOfHwPoints - 1) */ - y3_max = dal_fixed31_32_max(y_r, dal_fixed31_32_max(y_g, y_b)); + y3_max = dc_fixpt_max(y_r, dc_fixpt_max(y_g, y_b)); arr_points[1].y = y3_max; - arr_points[1].slope = dal_fixed31_32_zero; + arr_points[1].slope = dc_fixpt_zero; if (output_tf->tf == TRANSFER_FUNCTION_PQ) { /* for PQ, we want to have a straight line from last HW X point, * and the slope to be such that we hit 1.0 at 10000 nits. */ const struct fixed31_32 end_value = - dal_fixed31_32_from_int(125); + dc_fixpt_from_int(125); - arr_points[1].slope = dal_fixed31_32_div( - dal_fixed31_32_sub(dal_fixed31_32_one, arr_points[1].y), - dal_fixed31_32_sub(end_value, arr_points[1].x)); + arr_points[1].slope = dc_fixpt_div( + dc_fixpt_sub(dc_fixpt_one, arr_points[1].y), + dc_fixpt_sub(end_value, arr_points[1].x)); } lut_params->hw_points_num = hw_points; - i = 1; - for (k = 0; k < MAX_REGIONS_NUMBER && i < MAX_REGIONS_NUMBER; k++) { + k = 0; + for (i = 1; i < MAX_REGIONS_NUMBER; i++) { if (seg_distr[k] != -1) { lut_params->arr_curve_points[k].segments_num = seg_distr[k]; lut_params->arr_curve_points[i].offset = lut_params->arr_curve_points[k].offset + (1 << seg_distr[k]); } - i++; + k++; } if (seg_distr[k] != -1) @@ -386,24 +386,24 @@ bool cm_helper_translate_curve_to_hw_format( i = 1; while (i != hw_points + 1) { - if (dal_fixed31_32_lt(rgb_plus_1->red, rgb->red)) + if (dc_fixpt_lt(rgb_plus_1->red, rgb->red)) rgb_plus_1->red = rgb->red; - if (dal_fixed31_32_lt(rgb_plus_1->green, rgb->green)) + if (dc_fixpt_lt(rgb_plus_1->green, rgb->green)) rgb_plus_1->green = rgb->green; - if (dal_fixed31_32_lt(rgb_plus_1->blue, rgb->blue)) + if (dc_fixpt_lt(rgb_plus_1->blue, rgb->blue)) rgb_plus_1->blue = rgb->blue; - rgb->delta_red = dal_fixed31_32_sub(rgb_plus_1->red, rgb->red); - rgb->delta_green = dal_fixed31_32_sub(rgb_plus_1->green, rgb->green); - rgb->delta_blue = dal_fixed31_32_sub(rgb_plus_1->blue, rgb->blue); + rgb->delta_red = dc_fixpt_sub(rgb_plus_1->red, rgb->red); + rgb->delta_green = dc_fixpt_sub(rgb_plus_1->green, rgb->green); + rgb->delta_blue = dc_fixpt_sub(rgb_plus_1->blue, rgb->blue); if (fixpoint == true) { - rgb->delta_red_reg = dal_fixed31_32_clamp_u0d10(rgb->delta_red); - rgb->delta_green_reg = dal_fixed31_32_clamp_u0d10(rgb->delta_green); - rgb->delta_blue_reg = dal_fixed31_32_clamp_u0d10(rgb->delta_blue); - rgb->red_reg = dal_fixed31_32_clamp_u0d14(rgb->red); - rgb->green_reg = dal_fixed31_32_clamp_u0d14(rgb->green); - rgb->blue_reg = dal_fixed31_32_clamp_u0d14(rgb->blue); + rgb->delta_red_reg = dc_fixpt_clamp_u0d10(rgb->delta_red); + rgb->delta_green_reg = dc_fixpt_clamp_u0d10(rgb->delta_green); + rgb->delta_blue_reg = dc_fixpt_clamp_u0d10(rgb->delta_blue); + rgb->red_reg = dc_fixpt_clamp_u0d14(rgb->red); + rgb->green_reg = dc_fixpt_clamp_u0d14(rgb->green); + rgb->blue_reg = dc_fixpt_clamp_u0d14(rgb->blue); } ++rgb_plus_1; @@ -489,19 +489,19 @@ bool cm_helper_translate_curve_to_degamma_hw_format( rgb_resulted[hw_points - 1].green = output_tf->tf_pts.green[start_index]; rgb_resulted[hw_points - 1].blue = output_tf->tf_pts.blue[start_index]; - arr_points[0].x = dal_fixed31_32_pow(dal_fixed31_32_from_int(2), - dal_fixed31_32_from_int(region_start)); - arr_points[1].x = dal_fixed31_32_pow(dal_fixed31_32_from_int(2), - dal_fixed31_32_from_int(region_end)); + arr_points[0].x = dc_fixpt_pow(dc_fixpt_from_int(2), + dc_fixpt_from_int(region_start)); + arr_points[1].x = dc_fixpt_pow(dc_fixpt_from_int(2), + dc_fixpt_from_int(region_end)); y_r = rgb_resulted[0].red; y_g = rgb_resulted[0].green; y_b = rgb_resulted[0].blue; - y1_min = dal_fixed31_32_min(y_r, dal_fixed31_32_min(y_g, y_b)); + y1_min = dc_fixpt_min(y_r, dc_fixpt_min(y_g, y_b)); arr_points[0].y = y1_min; - arr_points[0].slope = dal_fixed31_32_div(arr_points[0].y, arr_points[0].x); + arr_points[0].slope = dc_fixpt_div(arr_points[0].y, arr_points[0].x); y_r = rgb_resulted[hw_points - 1].red; y_g = rgb_resulted[hw_points - 1].green; y_b = rgb_resulted[hw_points - 1].blue; @@ -509,35 +509,35 @@ bool cm_helper_translate_curve_to_degamma_hw_format( /* see comment above, m_arrPoints[1].y should be the Y value for the * region end (m_numOfHwPoints), not last HW point(m_numOfHwPoints - 1) */ - y3_max = dal_fixed31_32_max(y_r, dal_fixed31_32_max(y_g, y_b)); + y3_max = dc_fixpt_max(y_r, dc_fixpt_max(y_g, y_b)); arr_points[1].y = y3_max; - arr_points[1].slope = dal_fixed31_32_zero; + arr_points[1].slope = dc_fixpt_zero; if (output_tf->tf == TRANSFER_FUNCTION_PQ) { /* for PQ, we want to have a straight line from last HW X point, * and the slope to be such that we hit 1.0 at 10000 nits. */ const struct fixed31_32 end_value = - dal_fixed31_32_from_int(125); + dc_fixpt_from_int(125); - arr_points[1].slope = dal_fixed31_32_div( - dal_fixed31_32_sub(dal_fixed31_32_one, arr_points[1].y), - dal_fixed31_32_sub(end_value, arr_points[1].x)); + arr_points[1].slope = dc_fixpt_div( + dc_fixpt_sub(dc_fixpt_one, arr_points[1].y), + dc_fixpt_sub(end_value, arr_points[1].x)); } lut_params->hw_points_num = hw_points; - i = 1; - for (k = 0; k < MAX_REGIONS_NUMBER && i < MAX_REGIONS_NUMBER; k++) { + k = 0; + for (i = 1; i < MAX_REGIONS_NUMBER; i++) { if (seg_distr[k] != -1) { lut_params->arr_curve_points[k].segments_num = seg_distr[k]; lut_params->arr_curve_points[i].offset = lut_params->arr_curve_points[k].offset + (1 << seg_distr[k]); } - i++; + k++; } if (seg_distr[k] != -1) @@ -548,16 +548,16 @@ bool cm_helper_translate_curve_to_degamma_hw_format( i = 1; while (i != hw_points + 1) { - if (dal_fixed31_32_lt(rgb_plus_1->red, rgb->red)) + if (dc_fixpt_lt(rgb_plus_1->red, rgb->red)) rgb_plus_1->red = rgb->red; - if (dal_fixed31_32_lt(rgb_plus_1->green, rgb->green)) + if (dc_fixpt_lt(rgb_plus_1->green, rgb->green)) rgb_plus_1->green = rgb->green; - if (dal_fixed31_32_lt(rgb_plus_1->blue, rgb->blue)) + if (dc_fixpt_lt(rgb_plus_1->blue, rgb->blue)) rgb_plus_1->blue = rgb->blue; - rgb->delta_red = dal_fixed31_32_sub(rgb_plus_1->red, rgb->red); - rgb->delta_green = dal_fixed31_32_sub(rgb_plus_1->green, rgb->green); - rgb->delta_blue = dal_fixed31_32_sub(rgb_plus_1->blue, rgb->blue); + rgb->delta_red = dc_fixpt_sub(rgb_plus_1->red, rgb->red); + rgb->delta_green = dc_fixpt_sub(rgb_plus_1->green, rgb->green); + rgb->delta_blue = dc_fixpt_sub(rgb_plus_1->blue, rgb->blue); ++rgb_plus_1; ++rgb; diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.c index e305c28c98def..46a35c7f01df6 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.c @@ -98,6 +98,30 @@ enum gamut_remap_select { GAMUT_REMAP_COMB_COEFF }; +void dpp_read_state(struct dpp *dpp_base, + struct dcn_dpp_state *s) +{ + struct dcn10_dpp *dpp = TO_DCN10_DPP(dpp_base); + + REG_GET(CM_IGAM_CONTROL, + CM_IGAM_LUT_MODE, &s->igam_lut_mode); + REG_GET(CM_IGAM_CONTROL, + CM_IGAM_INPUT_FORMAT, &s->igam_input_format); + REG_GET(CM_DGAM_CONTROL, + CM_DGAM_LUT_MODE, &s->dgam_lut_mode); + REG_GET(CM_RGAM_CONTROL, + CM_RGAM_LUT_MODE, &s->rgam_lut_mode); + REG_GET(CM_GAMUT_REMAP_CONTROL, + CM_GAMUT_REMAP_MODE, &s->gamut_remap_mode); + + s->gamut_remap_c11_c12 = REG_READ(CM_GAMUT_REMAP_C11_C12); + s->gamut_remap_c13_c14 = REG_READ(CM_GAMUT_REMAP_C13_C14); + s->gamut_remap_c21_c22 = REG_READ(CM_GAMUT_REMAP_C21_C22); + s->gamut_remap_c23_c24 = REG_READ(CM_GAMUT_REMAP_C23_C24); + s->gamut_remap_c31_c32 = REG_READ(CM_GAMUT_REMAP_C31_C32); + s->gamut_remap_c33_c34 = REG_READ(CM_GAMUT_REMAP_C33_C34); +} + /* Program gamut remap in bypass mode */ void dpp_set_gamut_remap_bypass(struct dcn10_dpp *dpp) { @@ -106,7 +130,7 @@ void dpp_set_gamut_remap_bypass(struct dcn10_dpp *dpp) /* Gamut remap in bypass */ } -#define IDENTITY_RATIO(ratio) (dal_fixed31_32_u2d19(ratio) == (1 << 19)) +#define IDENTITY_RATIO(ratio) (dc_fixpt_u2d19(ratio) == (1 << 19)) bool dpp_get_optimal_number_of_taps( @@ -121,6 +145,18 @@ bool dpp_get_optimal_number_of_taps( else pixel_width = scl_data->viewport.width; + /* Some ASICs does not support FP16 scaling, so we reject modes require this*/ + if (scl_data->viewport.width != scl_data->h_active && + scl_data->viewport.height != scl_data->v_active && + dpp->caps->dscl_data_proc_format == DSCL_DATA_PRCESSING_FIXED_FORMAT && + scl_data->format == PIXEL_FORMAT_FP16) + return false; + + if (scl_data->viewport.width > scl_data->h_active && + dpp->ctx->dc->debug.max_downscale_src_width != 0 && + scl_data->viewport.width > dpp->ctx->dc->debug.max_downscale_src_width) + return false; + /* TODO: add lb check */ /* No support for programming ratio of 4, drop to 3.99999.. */ @@ -257,7 +293,7 @@ void dpp1_cnv_setup ( struct dpp *dpp_base, enum surface_pixel_format format, enum expansion_mode mode, - struct csc_transform input_csc_color_matrix, + struct dc_csc_transform input_csc_color_matrix, enum dc_color_space input_color_space) { uint32_t pixel_format; @@ -416,7 +452,7 @@ void dpp1_set_cursor_position( if (src_x_offset >= (int)param->viewport_width) cur_en = 0; /* not visible beyond right edge*/ - if (src_x_offset + (int)width < 0) + if (src_x_offset + (int)width <= 0) cur_en = 0; /* not visible beyond left edge*/ REG_UPDATE(CURSOR0_CONTROL, @@ -443,6 +479,7 @@ void dpp1_dppclk_control( } static const struct dpp_funcs dcn10_dpp_funcs = { + .dpp_read_state = dpp_read_state, .dpp_reset = dpp_reset, .dpp_set_scaler = dpp1_dscl_set_scaler_manual_scale, .dpp_get_optimal_number_of_taps = dpp_get_optimal_number_of_taps, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.h index 17b062a8f88a4..5944a3ba04090 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp.h @@ -44,6 +44,10 @@ #define TF_REG_LIST_DCN(id) \ SRI(CM_GAMUT_REMAP_CONTROL, CM, id),\ SRI(CM_GAMUT_REMAP_C11_C12, CM, id),\ + SRI(CM_GAMUT_REMAP_C13_C14, CM, id),\ + SRI(CM_GAMUT_REMAP_C21_C22, CM, id),\ + SRI(CM_GAMUT_REMAP_C23_C24, CM, id),\ + SRI(CM_GAMUT_REMAP_C31_C32, CM, id),\ SRI(CM_GAMUT_REMAP_C33_C34, CM, id),\ SRI(DSCL_EXT_OVERSCAN_LEFT_RIGHT, DSCL, id), \ SRI(DSCL_EXT_OVERSCAN_TOP_BOTTOM, DSCL, id), \ @@ -108,6 +112,8 @@ SRI(CM_DGAM_LUT_DATA, CM, id), \ SRI(CM_CONTROL, CM, id), \ SRI(CM_DGAM_CONTROL, CM, id), \ + SRI(CM_TEST_DEBUG_INDEX, CM, id), \ + SRI(CM_TEST_DEBUG_DATA, CM, id), \ SRI(FORMAT_CONTROL, CNVC_CFG, id), \ SRI(CNVC_SURFACE_PIXEL_FORMAT, CNVC_CFG, id), \ SRI(CURSOR0_CONTROL, CNVC_CUR, id), \ @@ -175,6 +181,14 @@ TF_SF(CM0_CM_GAMUT_REMAP_CONTROL, CM_GAMUT_REMAP_MODE, mask_sh),\ TF_SF(CM0_CM_GAMUT_REMAP_C11_C12, CM_GAMUT_REMAP_C11, mask_sh),\ TF_SF(CM0_CM_GAMUT_REMAP_C11_C12, CM_GAMUT_REMAP_C12, mask_sh),\ + TF_SF(CM0_CM_GAMUT_REMAP_C13_C14, CM_GAMUT_REMAP_C13, mask_sh),\ + TF_SF(CM0_CM_GAMUT_REMAP_C13_C14, CM_GAMUT_REMAP_C14, mask_sh),\ + TF_SF(CM0_CM_GAMUT_REMAP_C21_C22, CM_GAMUT_REMAP_C21, mask_sh),\ + TF_SF(CM0_CM_GAMUT_REMAP_C21_C22, CM_GAMUT_REMAP_C22, mask_sh),\ + TF_SF(CM0_CM_GAMUT_REMAP_C23_C24, CM_GAMUT_REMAP_C23, mask_sh),\ + TF_SF(CM0_CM_GAMUT_REMAP_C23_C24, CM_GAMUT_REMAP_C24, mask_sh),\ + TF_SF(CM0_CM_GAMUT_REMAP_C31_C32, CM_GAMUT_REMAP_C31, mask_sh),\ + TF_SF(CM0_CM_GAMUT_REMAP_C31_C32, CM_GAMUT_REMAP_C32, mask_sh),\ TF_SF(CM0_CM_GAMUT_REMAP_C33_C34, CM_GAMUT_REMAP_C33, mask_sh),\ TF_SF(CM0_CM_GAMUT_REMAP_C33_C34, CM_GAMUT_REMAP_C34, mask_sh),\ TF_SF(DSCL0_DSCL_EXT_OVERSCAN_LEFT_RIGHT, EXT_OVERSCAN_LEFT, mask_sh),\ @@ -300,6 +314,7 @@ TF_SF(CM0_CM_DGAM_LUT_INDEX, CM_DGAM_LUT_INDEX, mask_sh), \ TF_SF(CM0_CM_DGAM_LUT_DATA, CM_DGAM_LUT_DATA, mask_sh), \ TF_SF(CM0_CM_DGAM_CONTROL, CM_DGAM_LUT_MODE, mask_sh), \ + TF_SF(CM0_CM_TEST_DEBUG_INDEX, CM_TEST_DEBUG_INDEX, mask_sh), \ TF_SF(CNVC_CFG0_FORMAT_CONTROL, CNVC_BYPASS, mask_sh), \ TF2_SF(CNVC_CFG0, FORMAT_CONTROL__ALPHA_EN, mask_sh), \ TF_SF(CNVC_CFG0_FORMAT_CONTROL, FORMAT_EXPANSION_MODE, mask_sh), \ @@ -417,6 +432,41 @@ TF_SF(CURSOR0_CURSOR_CONTROL, CURSOR_ENABLE, mask_sh), \ TF_SF(DPP_TOP0_DPP_CONTROL, DPPCLK_RATE_CONTROL, mask_sh) +/* + * + DCN1 CM debug status register definition + + register :ID9_CM_STATUS do + implement_ref :cm + map to: :cmdebugind, at: j + width 32 + disclosure NEVER + + field :ID9_VUPDATE_CFG, [0], R + field :ID9_IGAM_LUT_MODE, [2..1], R + field :ID9_BNS_BYPASS, [3], R + field :ID9_ICSC_MODE, [5..4], R + field :ID9_DGAM_LUT_MODE, [8..6], R + field :ID9_HDR_BYPASS, [9], R + field :ID9_GAMUT_REMAP_MODE, [11..10], R + field :ID9_RGAM_LUT_MODE, [14..12], R + #1 free bit + field :ID9_OCSC_MODE, [18..16], R + field :ID9_DENORM_MODE, [21..19], R + field :ID9_ROUND_TRUNC_MODE, [25..22], R + field :ID9_DITHER_EN, [26], R + field :ID9_DITHER_MODE, [28..27], R + end +*/ + +#define TF_DEBUG_REG_LIST_SH_DCN10 \ + .CM_TEST_DEBUG_DATA_ID9_ICSC_MODE = 4, \ + .CM_TEST_DEBUG_DATA_ID9_OCSC_MODE = 16 + +#define TF_DEBUG_REG_LIST_MASK_DCN10 \ + .CM_TEST_DEBUG_DATA_ID9_ICSC_MODE = 0x30, \ + .CM_TEST_DEBUG_DATA_ID9_OCSC_MODE = 0x70000 + #define TF_REG_FIELD_LIST(type) \ type EXT_OVERSCAN_LEFT; \ type EXT_OVERSCAN_RIGHT; \ @@ -486,6 +536,14 @@ type CM_GAMUT_REMAP_MODE; \ type CM_GAMUT_REMAP_C11; \ type CM_GAMUT_REMAP_C12; \ + type CM_GAMUT_REMAP_C13; \ + type CM_GAMUT_REMAP_C14; \ + type CM_GAMUT_REMAP_C21; \ + type CM_GAMUT_REMAP_C22; \ + type CM_GAMUT_REMAP_C23; \ + type CM_GAMUT_REMAP_C24; \ + type CM_GAMUT_REMAP_C31; \ + type CM_GAMUT_REMAP_C32; \ type CM_GAMUT_REMAP_C33; \ type CM_GAMUT_REMAP_C34; \ type CM_COMA_C11; \ @@ -1010,6 +1068,9 @@ type CUR0_EXPANSION_MODE; \ type CUR0_ENABLE; \ type CM_BYPASS; \ + type CM_TEST_DEBUG_INDEX; \ + type CM_TEST_DEBUG_DATA_ID9_ICSC_MODE; \ + type CM_TEST_DEBUG_DATA_ID9_OCSC_MODE;\ type FORMAT_CONTROL__ALPHA_EN; \ type CUR0_COLOR0; \ type CUR0_COLOR1; \ @@ -1054,6 +1115,10 @@ struct dcn_dpp_mask { uint32_t RECOUT_SIZE; \ uint32_t CM_GAMUT_REMAP_CONTROL; \ uint32_t CM_GAMUT_REMAP_C11_C12; \ + uint32_t CM_GAMUT_REMAP_C13_C14; \ + uint32_t CM_GAMUT_REMAP_C21_C22; \ + uint32_t CM_GAMUT_REMAP_C23_C24; \ + uint32_t CM_GAMUT_REMAP_C31_C32; \ uint32_t CM_GAMUT_REMAP_C33_C34; \ uint32_t CM_COMA_C11_C12; \ uint32_t CM_COMA_C33_C34; \ @@ -1255,6 +1320,8 @@ struct dcn_dpp_mask { uint32_t CM_IGAM_LUT_RW_CONTROL; \ uint32_t CM_IGAM_LUT_RW_INDEX; \ uint32_t CM_IGAM_LUT_SEQ_COLOR; \ + uint32_t CM_TEST_DEBUG_INDEX; \ + uint32_t CM_TEST_DEBUG_DATA; \ uint32_t FORMAT_CONTROL; \ uint32_t CNVC_SURFACE_PIXEL_FORMAT; \ uint32_t CURSOR_CONTROL; \ @@ -1289,8 +1356,8 @@ struct dcn10_dpp { enum dcn10_input_csc_select { INPUT_CSC_SELECT_BYPASS = 0, - INPUT_CSC_SELECT_ICSC, - INPUT_CSC_SELECT_COMA + INPUT_CSC_SELECT_ICSC = 1, + INPUT_CSC_SELECT_COMA = 2 }; void dpp1_set_cursor_attributes( @@ -1364,6 +1431,9 @@ bool dpp_get_optimal_number_of_taps( struct scaler_data *scl_data, const struct scaling_taps *in_taps); +void dpp_read_state(struct dpp *dpp_base, + struct dcn_dpp_state *s); + void dpp_reset(struct dpp *dpp_base); void dpp1_cm_program_regamma_lut( @@ -1408,7 +1478,7 @@ void dpp1_cnv_setup ( struct dpp *dpp_base, enum surface_pixel_format format, enum expansion_mode mode, - struct csc_transform input_csc_color_matrix, + struct dc_csc_transform input_csc_color_matrix, enum dc_color_space input_color_space); void dpp1_full_bypass(struct dpp *dpp_base); diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_cm.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_cm.c index fb32975e4b67f..116977eb24e22 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_cm.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_cm.c @@ -216,41 +216,55 @@ static void dpp1_cm_program_color_matrix( struct dcn10_dpp *dpp, const uint16_t *regval) { - uint32_t mode; + uint32_t ocsc_mode; + uint32_t cur_mode; struct color_matrices_reg gam_regs; - REG_GET(CM_OCSC_CONTROL, CM_OCSC_MODE, &mode); - if (regval == NULL) { BREAK_TO_DEBUGGER(); return; } - mode = 4; + + /* determine which CSC matrix (ocsc or comb) we are using + * currently. select the alternate set to double buffer + * the CSC update so CSC is updated on frame boundary + */ + REG_SET(CM_TEST_DEBUG_INDEX, 0, + CM_TEST_DEBUG_INDEX, 9); + + REG_GET(CM_TEST_DEBUG_DATA, + CM_TEST_DEBUG_DATA_ID9_OCSC_MODE, &cur_mode); + + if (cur_mode != 4) + ocsc_mode = 4; + else + ocsc_mode = 5; + + gam_regs.shifts.csc_c11 = dpp->tf_shift->CM_OCSC_C11; gam_regs.masks.csc_c11 = dpp->tf_mask->CM_OCSC_C11; gam_regs.shifts.csc_c12 = dpp->tf_shift->CM_OCSC_C12; gam_regs.masks.csc_c12 = dpp->tf_mask->CM_OCSC_C12; - if (mode == 4) { + if (ocsc_mode == 4) { gam_regs.csc_c11_c12 = REG(CM_OCSC_C11_C12); gam_regs.csc_c33_c34 = REG(CM_OCSC_C33_C34); - cm_helper_program_color_matrices( - dpp->base.ctx, - regval, - &gam_regs); - } else { gam_regs.csc_c11_c12 = REG(CM_COMB_C11_C12); gam_regs.csc_c33_c34 = REG(CM_COMB_C33_C34); - cm_helper_program_color_matrices( - dpp->base.ctx, - regval, - &gam_regs); } + + cm_helper_program_color_matrices( + dpp->base.ctx, + regval, + &gam_regs); + + REG_SET(CM_OCSC_CONTROL, 0, CM_OCSC_MODE, ocsc_mode); + } void dpp1_cm_set_output_csc_default( @@ -260,15 +274,14 @@ void dpp1_cm_set_output_csc_default( struct dcn10_dpp *dpp = TO_DCN10_DPP(dpp_base); const uint16_t *regval = NULL; int arr_size; - uint32_t ocsc_mode = 4; regval = find_color_matrix(colorspace, &arr_size); if (regval == NULL) { BREAK_TO_DEBUGGER(); return; } + dpp1_cm_program_color_matrix(dpp, regval); - REG_SET(CM_OCSC_CONTROL, 0, CM_OCSC_MODE, ocsc_mode); } static void dpp1_cm_get_reg_field( @@ -329,9 +342,8 @@ void dpp1_cm_set_output_csc_adjustment( const uint16_t *regval) { struct dcn10_dpp *dpp = TO_DCN10_DPP(dpp_base); - uint32_t ocsc_mode = 4; + dpp1_cm_program_color_matrix(dpp, regval); - REG_SET(CM_OCSC_CONTROL, 0, CM_OCSC_MODE, ocsc_mode); } void dpp1_cm_power_on_regamma_lut(struct dpp *dpp_base, @@ -437,17 +449,18 @@ void dpp1_cm_program_regamma_lutb_settings( void dpp1_program_input_csc( struct dpp *dpp_base, enum dc_color_space color_space, - enum dcn10_input_csc_select select, + enum dcn10_input_csc_select input_select, const struct out_csc_color_matrix *tbl_entry) { struct dcn10_dpp *dpp = TO_DCN10_DPP(dpp_base); int i; int arr_size = sizeof(dcn10_input_csc_matrix)/sizeof(struct dcn10_input_csc_matrix); const uint16_t *regval = NULL; - uint32_t selection = 1; + uint32_t cur_select = 0; + enum dcn10_input_csc_select select; struct color_matrices_reg gam_regs; - if (select == INPUT_CSC_SELECT_BYPASS) { + if (input_select == INPUT_CSC_SELECT_BYPASS) { REG_SET(CM_ICSC_CONTROL, 0, CM_ICSC_MODE, 0); return; } @@ -467,36 +480,45 @@ void dpp1_program_input_csc( regval = tbl_entry->regval; } - if (select == INPUT_CSC_SELECT_COMA) - selection = 2; - REG_SET(CM_ICSC_CONTROL, 0, - CM_ICSC_MODE, selection); + /* determine which CSC matrix (icsc or coma) we are using + * currently. select the alternate set to double buffer + * the CSC update so CSC is updated on frame boundary + */ + REG_SET(CM_TEST_DEBUG_INDEX, 0, + CM_TEST_DEBUG_INDEX, 9); + + REG_GET(CM_TEST_DEBUG_DATA, + CM_TEST_DEBUG_DATA_ID9_ICSC_MODE, &cur_select); + + if (cur_select != INPUT_CSC_SELECT_ICSC) + select = INPUT_CSC_SELECT_ICSC; + else + select = INPUT_CSC_SELECT_COMA; gam_regs.shifts.csc_c11 = dpp->tf_shift->CM_ICSC_C11; gam_regs.masks.csc_c11 = dpp->tf_mask->CM_ICSC_C11; gam_regs.shifts.csc_c12 = dpp->tf_shift->CM_ICSC_C12; gam_regs.masks.csc_c12 = dpp->tf_mask->CM_ICSC_C12; - if (select == INPUT_CSC_SELECT_ICSC) { gam_regs.csc_c11_c12 = REG(CM_ICSC_C11_C12); gam_regs.csc_c33_c34 = REG(CM_ICSC_C33_C34); - cm_helper_program_color_matrices( - dpp->base.ctx, - regval, - &gam_regs); } else { gam_regs.csc_c11_c12 = REG(CM_COMA_C11_C12); gam_regs.csc_c33_c34 = REG(CM_COMA_C33_C34); - cm_helper_program_color_matrices( - dpp->base.ctx, - regval, - &gam_regs); } + + cm_helper_program_color_matrices( + dpp->base.ctx, + regval, + &gam_regs); + + REG_SET(CM_ICSC_CONTROL, 0, + CM_ICSC_MODE, select); } //keep here for now, decide multi dce support later @@ -789,13 +811,13 @@ void dpp1_program_input_lut( REG_UPDATE(CM_IGAM_LUT_RW_INDEX, CM_IGAM_LUT_RW_INDEX, 0); for (i = 0; i < gamma->num_entries; i++) { REG_SET(CM_IGAM_LUT_SEQ_COLOR, 0, CM_IGAM_LUT_SEQ_COLOR, - dal_fixed31_32_round( + dc_fixpt_round( gamma->entries.red[i])); REG_SET(CM_IGAM_LUT_SEQ_COLOR, 0, CM_IGAM_LUT_SEQ_COLOR, - dal_fixed31_32_round( + dc_fixpt_round( gamma->entries.green[i])); REG_SET(CM_IGAM_LUT_SEQ_COLOR, 0, CM_IGAM_LUT_SEQ_COLOR, - dal_fixed31_32_round( + dc_fixpt_round( gamma->entries.blue[i])); } // Power off LUT memory diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c index 3eb824debf432..4ddd6273d5a5b 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_dpp_dscl.c @@ -169,7 +169,7 @@ static enum dscl_mode_sel dpp1_dscl_get_dscl_mode( const struct scaler_data *data, bool dbg_always_scale) { - const long long one = dal_fixed31_32_one.value; + const long long one = dc_fixpt_one.value; if (dpp_base->caps->dscl_data_proc_format == DSCL_DATA_PRCESSING_FIXED_FORMAT) { /* DSCL is processing data in fixed format */ @@ -464,8 +464,8 @@ static enum lb_memory_config dpp1_dscl_find_lb_memory_config(struct dcn10_dpp *d int num_part_y, num_part_c; int vtaps = scl_data->taps.v_taps; int vtaps_c = scl_data->taps.v_taps_c; - int ceil_vratio = dal_fixed31_32_ceil(scl_data->ratios.vert); - int ceil_vratio_c = dal_fixed31_32_ceil(scl_data->ratios.vert_c); + int ceil_vratio = dc_fixpt_ceil(scl_data->ratios.vert); + int ceil_vratio_c = dc_fixpt_ceil(scl_data->ratios.vert_c); enum lb_memory_config mem_cfg = LB_MEMORY_CONFIG_0; if (dpp->base.ctx->dc->debug.use_max_lb) @@ -565,52 +565,52 @@ static void dpp1_dscl_set_manual_ratio_init( uint32_t init_int = 0; REG_SET(SCL_HORZ_FILTER_SCALE_RATIO, 0, - SCL_H_SCALE_RATIO, dal_fixed31_32_u2d19(data->ratios.horz) << 5); + SCL_H_SCALE_RATIO, dc_fixpt_u2d19(data->ratios.horz) << 5); REG_SET(SCL_VERT_FILTER_SCALE_RATIO, 0, - SCL_V_SCALE_RATIO, dal_fixed31_32_u2d19(data->ratios.vert) << 5); + SCL_V_SCALE_RATIO, dc_fixpt_u2d19(data->ratios.vert) << 5); REG_SET(SCL_HORZ_FILTER_SCALE_RATIO_C, 0, - SCL_H_SCALE_RATIO_C, dal_fixed31_32_u2d19(data->ratios.horz_c) << 5); + SCL_H_SCALE_RATIO_C, dc_fixpt_u2d19(data->ratios.horz_c) << 5); REG_SET(SCL_VERT_FILTER_SCALE_RATIO_C, 0, - SCL_V_SCALE_RATIO_C, dal_fixed31_32_u2d19(data->ratios.vert_c) << 5); + SCL_V_SCALE_RATIO_C, dc_fixpt_u2d19(data->ratios.vert_c) << 5); /* * 0.24 format for fraction, first five bits zeroed */ - init_frac = dal_fixed31_32_u0d19(data->inits.h) << 5; - init_int = dal_fixed31_32_floor(data->inits.h); + init_frac = dc_fixpt_u0d19(data->inits.h) << 5; + init_int = dc_fixpt_floor(data->inits.h); REG_SET_2(SCL_HORZ_FILTER_INIT, 0, SCL_H_INIT_FRAC, init_frac, SCL_H_INIT_INT, init_int); - init_frac = dal_fixed31_32_u0d19(data->inits.h_c) << 5; - init_int = dal_fixed31_32_floor(data->inits.h_c); + init_frac = dc_fixpt_u0d19(data->inits.h_c) << 5; + init_int = dc_fixpt_floor(data->inits.h_c); REG_SET_2(SCL_HORZ_FILTER_INIT_C, 0, SCL_H_INIT_FRAC_C, init_frac, SCL_H_INIT_INT_C, init_int); - init_frac = dal_fixed31_32_u0d19(data->inits.v) << 5; - init_int = dal_fixed31_32_floor(data->inits.v); + init_frac = dc_fixpt_u0d19(data->inits.v) << 5; + init_int = dc_fixpt_floor(data->inits.v); REG_SET_2(SCL_VERT_FILTER_INIT, 0, SCL_V_INIT_FRAC, init_frac, SCL_V_INIT_INT, init_int); - init_frac = dal_fixed31_32_u0d19(data->inits.v_bot) << 5; - init_int = dal_fixed31_32_floor(data->inits.v_bot); + init_frac = dc_fixpt_u0d19(data->inits.v_bot) << 5; + init_int = dc_fixpt_floor(data->inits.v_bot); REG_SET_2(SCL_VERT_FILTER_INIT_BOT, 0, SCL_V_INIT_FRAC_BOT, init_frac, SCL_V_INIT_INT_BOT, init_int); - init_frac = dal_fixed31_32_u0d19(data->inits.v_c) << 5; - init_int = dal_fixed31_32_floor(data->inits.v_c); + init_frac = dc_fixpt_u0d19(data->inits.v_c) << 5; + init_int = dc_fixpt_floor(data->inits.v_c); REG_SET_2(SCL_VERT_FILTER_INIT_C, 0, SCL_V_INIT_FRAC_C, init_frac, SCL_V_INIT_INT_C, init_int); - init_frac = dal_fixed31_32_u0d19(data->inits.v_c_bot) << 5; - init_int = dal_fixed31_32_floor(data->inits.v_c_bot); + init_frac = dc_fixpt_u0d19(data->inits.v_c_bot) << 5; + init_int = dc_fixpt_floor(data->inits.v_c_bot); REG_SET_2(SCL_VERT_FILTER_INIT_BOT_C, 0, SCL_V_INIT_FRAC_BOT_C, init_frac, SCL_V_INIT_INT_BOT_C, init_int); diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.c index 738f67ffd1b40..943143efbb823 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.c @@ -476,8 +476,235 @@ void hubbub1_toggle_watermark_change_req(struct hubbub *hubbub) DCHUBBUB_ARB_WATERMARK_CHANGE_REQUEST, watermark_change_req); } +void hubbub1_soft_reset(struct hubbub *hubbub, bool reset) +{ + uint32_t reset_en = reset ? 1 : 0; + + REG_UPDATE(DCHUBBUB_SOFT_RESET, + DCHUBBUB_GLOBAL_SOFT_RESET, reset_en); +} + +static bool hubbub1_dcc_support_swizzle( + enum swizzle_mode_values swizzle, + unsigned int bytes_per_element, + enum segment_order *segment_order_horz, + enum segment_order *segment_order_vert) +{ + bool standard_swizzle = false; + bool display_swizzle = false; + + switch (swizzle) { + case DC_SW_4KB_S: + case DC_SW_64KB_S: + case DC_SW_VAR_S: + case DC_SW_4KB_S_X: + case DC_SW_64KB_S_X: + case DC_SW_VAR_S_X: + standard_swizzle = true; + break; + case DC_SW_4KB_D: + case DC_SW_64KB_D: + case DC_SW_VAR_D: + case DC_SW_4KB_D_X: + case DC_SW_64KB_D_X: + case DC_SW_VAR_D_X: + display_swizzle = true; + break; + default: + break; + } + + if (bytes_per_element == 1 && standard_swizzle) { + *segment_order_horz = segment_order__contiguous; + *segment_order_vert = segment_order__na; + return true; + } + if (bytes_per_element == 2 && standard_swizzle) { + *segment_order_horz = segment_order__non_contiguous; + *segment_order_vert = segment_order__contiguous; + return true; + } + if (bytes_per_element == 4 && standard_swizzle) { + *segment_order_horz = segment_order__non_contiguous; + *segment_order_vert = segment_order__contiguous; + return true; + } + if (bytes_per_element == 8 && standard_swizzle) { + *segment_order_horz = segment_order__na; + *segment_order_vert = segment_order__contiguous; + return true; + } + if (bytes_per_element == 8 && display_swizzle) { + *segment_order_horz = segment_order__contiguous; + *segment_order_vert = segment_order__non_contiguous; + return true; + } + + return false; +} + +static bool hubbub1_dcc_support_pixel_format( + enum surface_pixel_format format, + unsigned int *bytes_per_element) +{ + /* DML: get_bytes_per_element */ + switch (format) { + case SURFACE_PIXEL_FORMAT_GRPH_ARGB1555: + case SURFACE_PIXEL_FORMAT_GRPH_RGB565: + *bytes_per_element = 2; + return true; + case SURFACE_PIXEL_FORMAT_GRPH_ARGB8888: + case SURFACE_PIXEL_FORMAT_GRPH_ABGR8888: + case SURFACE_PIXEL_FORMAT_GRPH_ARGB2101010: + case SURFACE_PIXEL_FORMAT_GRPH_ABGR2101010: + *bytes_per_element = 4; + return true; + case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616: + case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616F: + case SURFACE_PIXEL_FORMAT_GRPH_ABGR16161616F: + *bytes_per_element = 8; + return true; + default: + return false; + } +} + +static void hubbub1_get_blk256_size(unsigned int *blk256_width, unsigned int *blk256_height, + unsigned int bytes_per_element) +{ + /* copied from DML. might want to refactor DML to leverage from DML */ + /* DML : get_blk256_size */ + if (bytes_per_element == 1) { + *blk256_width = 16; + *blk256_height = 16; + } else if (bytes_per_element == 2) { + *blk256_width = 16; + *blk256_height = 8; + } else if (bytes_per_element == 4) { + *blk256_width = 8; + *blk256_height = 8; + } else if (bytes_per_element == 8) { + *blk256_width = 8; + *blk256_height = 4; + } +} + +static void hubbub1_det_request_size( + unsigned int height, + unsigned int width, + unsigned int bpe, + bool *req128_horz_wc, + bool *req128_vert_wc) +{ + unsigned int detile_buf_size = 164 * 1024; /* 164KB for DCN1.0 */ + + unsigned int blk256_height = 0; + unsigned int blk256_width = 0; + unsigned int swath_bytes_horz_wc, swath_bytes_vert_wc; + + hubbub1_get_blk256_size(&blk256_width, &blk256_height, bpe); + + swath_bytes_horz_wc = height * blk256_height * bpe; + swath_bytes_vert_wc = width * blk256_width * bpe; + + *req128_horz_wc = (2 * swath_bytes_horz_wc <= detile_buf_size) ? + false : /* full 256B request */ + true; /* half 128b request */ + + *req128_vert_wc = (2 * swath_bytes_vert_wc <= detile_buf_size) ? + false : /* full 256B request */ + true; /* half 128b request */ +} + +static bool hubbub1_get_dcc_compression_cap(struct hubbub *hubbub, + const struct dc_dcc_surface_param *input, + struct dc_surface_dcc_cap *output) +{ + struct dc *dc = hubbub->ctx->dc; + /* implement section 1.6.2.1 of DCN1_Programming_Guide.docx */ + enum dcc_control dcc_control; + unsigned int bpe; + enum segment_order segment_order_horz, segment_order_vert; + bool req128_horz_wc, req128_vert_wc; + + memset(output, 0, sizeof(*output)); + + if (dc->debug.disable_dcc == DCC_DISABLE) + return false; + + if (!hubbub->funcs->dcc_support_pixel_format(input->format, &bpe)) + return false; + + if (!hubbub->funcs->dcc_support_swizzle(input->swizzle_mode, bpe, + &segment_order_horz, &segment_order_vert)) + return false; + + hubbub1_det_request_size(input->surface_size.height, input->surface_size.width, + bpe, &req128_horz_wc, &req128_vert_wc); + + if (!req128_horz_wc && !req128_vert_wc) { + dcc_control = dcc_control__256_256_xxx; + } else if (input->scan == SCAN_DIRECTION_HORIZONTAL) { + if (!req128_horz_wc) + dcc_control = dcc_control__256_256_xxx; + else if (segment_order_horz == segment_order__contiguous) + dcc_control = dcc_control__128_128_xxx; + else + dcc_control = dcc_control__256_64_64; + } else if (input->scan == SCAN_DIRECTION_VERTICAL) { + if (!req128_vert_wc) + dcc_control = dcc_control__256_256_xxx; + else if (segment_order_vert == segment_order__contiguous) + dcc_control = dcc_control__128_128_xxx; + else + dcc_control = dcc_control__256_64_64; + } else { + if ((req128_horz_wc && + segment_order_horz == segment_order__non_contiguous) || + (req128_vert_wc && + segment_order_vert == segment_order__non_contiguous)) + /* access_dir not known, must use most constraining */ + dcc_control = dcc_control__256_64_64; + else + /* reg128 is true for either horz and vert + * but segment_order is contiguous + */ + dcc_control = dcc_control__128_128_xxx; + } + + if (dc->debug.disable_dcc == DCC_HALF_REQ_DISALBE && + dcc_control != dcc_control__256_256_xxx) + return false; + + switch (dcc_control) { + case dcc_control__256_256_xxx: + output->grph.rgb.max_uncompressed_blk_size = 256; + output->grph.rgb.max_compressed_blk_size = 256; + output->grph.rgb.independent_64b_blks = false; + break; + case dcc_control__128_128_xxx: + output->grph.rgb.max_uncompressed_blk_size = 128; + output->grph.rgb.max_compressed_blk_size = 128; + output->grph.rgb.independent_64b_blks = false; + break; + case dcc_control__256_64_64: + output->grph.rgb.max_uncompressed_blk_size = 256; + output->grph.rgb.max_compressed_blk_size = 64; + output->grph.rgb.independent_64b_blks = true; + break; + } + + output->capable = true; + output->const_color_support = false; + + return true; +} + static const struct hubbub_funcs hubbub1_funcs = { - .update_dchub = hubbub1_update_dchub + .update_dchub = hubbub1_update_dchub, + .dcc_support_swizzle = hubbub1_dcc_support_swizzle, + .dcc_support_pixel_format = hubbub1_dcc_support_pixel_format, + .get_dcc_compression_cap = hubbub1_get_dcc_compression_cap, }; void hubbub1_construct(struct hubbub *hubbub, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.h index a16e908821a0d..6315a0e6b0d69 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubbub.h @@ -27,6 +27,7 @@ #define __DC_HUBBUB_DCN10_H__ #include "core_types.h" +#include "dchubbub.h" #define HUBHUB_REG_LIST_DCN()\ SR(DCHUBBUB_ARB_DATA_URGENCY_WATERMARK_A),\ @@ -47,7 +48,8 @@ SR(DCHUBBUB_ARB_DF_REQ_OUTSTAND),\ SR(DCHUBBUB_GLOBAL_TIMER_CNTL), \ SR(DCHUBBUB_TEST_DEBUG_INDEX), \ - SR(DCHUBBUB_TEST_DEBUG_DATA) + SR(DCHUBBUB_TEST_DEBUG_DATA),\ + SR(DCHUBBUB_SOFT_RESET) #define HUBBUB_SR_WATERMARK_REG_LIST()\ SR(DCHUBBUB_ARB_ALLOW_SR_ENTER_WATERMARK_A),\ @@ -104,6 +106,7 @@ struct dcn_hubbub_registers { uint32_t DCHUBBUB_SDPIF_AGP_BOT; uint32_t DCHUBBUB_SDPIF_AGP_TOP; uint32_t DCHUBBUB_CRC_CTRL; + uint32_t DCHUBBUB_SOFT_RESET; }; /* set field name */ @@ -113,6 +116,7 @@ struct dcn_hubbub_registers { #define HUBBUB_MASK_SH_LIST_DCN(mask_sh)\ HUBBUB_SF(DCHUBBUB_GLOBAL_TIMER_CNTL, DCHUBBUB_GLOBAL_TIMER_ENABLE, mask_sh), \ + HUBBUB_SF(DCHUBBUB_SOFT_RESET, DCHUBBUB_GLOBAL_SOFT_RESET, mask_sh), \ HUBBUB_SF(DCHUBBUB_ARB_WATERMARK_CHANGE_CNTL, DCHUBBUB_ARB_WATERMARK_CHANGE_REQUEST, mask_sh), \ HUBBUB_SF(DCHUBBUB_ARB_WATERMARK_CHANGE_CNTL, DCHUBBUB_ARB_WATERMARK_CHANGE_DONE_INTERRUPT_DISABLE, mask_sh), \ HUBBUB_SF(DCHUBBUB_ARB_DRAM_STATE_CNTL, DCHUBBUB_ARB_ALLOW_SELF_REFRESH_FORCE_VALUE, mask_sh), \ @@ -142,6 +146,7 @@ struct dcn_hubbub_registers { type DCHUBBUB_ARB_SAT_LEVEL;\ type DCHUBBUB_ARB_MIN_REQ_OUTSTAND;\ type DCHUBBUB_GLOBAL_TIMER_REFDIV;\ + type DCHUBBUB_GLOBAL_SOFT_RESET; \ type SDPIF_FB_TOP;\ type SDPIF_FB_BASE;\ type SDPIF_FB_OFFSET;\ @@ -173,12 +178,6 @@ struct dcn_hubbub_wm { struct dcn_hubbub_wm_set sets[4]; }; -struct hubbub_funcs { - void (*update_dchub)( - struct hubbub *hubbub, - struct dchub_init_data *dh_data); -}; - struct hubbub { const struct hubbub_funcs *funcs; struct dc_context *ctx; @@ -206,6 +205,7 @@ void hubbub1_toggle_watermark_change_req( void hubbub1_wm_read_state(struct hubbub *hubbub, struct dcn_hubbub_wm *wm); +void hubbub1_soft_reset(struct hubbub *hubbub, bool reset); void hubbub1_construct(struct hubbub *hubbub, struct dc_context *ctx, const struct dcn_hubbub_registers *hubbub_regs, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.c index 39b72f696ae9d..d2ab78b35a7ae 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.c @@ -78,6 +78,27 @@ static void hubp1_disconnect(struct hubp *hubp) CURSOR_ENABLE, 0); } +static void hubp1_disable_control(struct hubp *hubp, bool disable_hubp) +{ + struct dcn10_hubp *hubp1 = TO_DCN10_HUBP(hubp); + uint32_t disable = disable_hubp ? 1 : 0; + + REG_UPDATE(DCHUBP_CNTL, + HUBP_DISABLE, disable); +} + +static unsigned int hubp1_get_underflow_status(struct hubp *hubp) +{ + uint32_t hubp_underflow = 0; + struct dcn10_hubp *hubp1 = TO_DCN10_HUBP(hubp); + + REG_GET(DCHUBP_CNTL, + HUBP_UNDERFLOW_STATUS, + &hubp_underflow); + + return hubp_underflow; +} + static void hubp1_set_hubp_blank_en(struct hubp *hubp, bool blank) { struct dcn10_hubp *hubp1 = TO_DCN10_HUBP(hubp); @@ -146,6 +167,9 @@ void hubp1_program_size_and_rotation( * 444 or 420 luma */ if (format >= SURFACE_PIXEL_FORMAT_VIDEO_BEGIN) { + ASSERT(plane_size->video.chroma_pitch != 0); + /* Chroma pitch zero can cause system hang! */ + pitch = plane_size->video.luma_pitch - 1; meta_pitch = dcc->video.meta_pitch_l - 1; pitch_c = plane_size->video.chroma_pitch - 1; @@ -535,11 +559,13 @@ void hubp1_program_deadline( REG_SET(VBLANK_PARAMETERS_3, 0, REFCYC_PER_META_CHUNK_VBLANK_L, dlg_attr->refcyc_per_meta_chunk_vblank_l); - REG_SET(NOM_PARAMETERS_0, 0, - DST_Y_PER_PTE_ROW_NOM_L, dlg_attr->dst_y_per_pte_row_nom_l); + if (REG(NOM_PARAMETERS_0)) + REG_SET(NOM_PARAMETERS_0, 0, + DST_Y_PER_PTE_ROW_NOM_L, dlg_attr->dst_y_per_pte_row_nom_l); - REG_SET(NOM_PARAMETERS_1, 0, - REFCYC_PER_PTE_GROUP_NOM_L, dlg_attr->refcyc_per_pte_group_nom_l); + if (REG(NOM_PARAMETERS_1)) + REG_SET(NOM_PARAMETERS_1, 0, + REFCYC_PER_PTE_GROUP_NOM_L, dlg_attr->refcyc_per_pte_group_nom_l); REG_SET(NOM_PARAMETERS_4, 0, DST_Y_PER_META_ROW_NOM_L, dlg_attr->dst_y_per_meta_row_nom_l); @@ -568,11 +594,13 @@ void hubp1_program_deadline( REG_SET(VBLANK_PARAMETERS_4, 0, REFCYC_PER_META_CHUNK_VBLANK_C, dlg_attr->refcyc_per_meta_chunk_vblank_c); - REG_SET(NOM_PARAMETERS_2, 0, - DST_Y_PER_PTE_ROW_NOM_C, dlg_attr->dst_y_per_pte_row_nom_c); + if (REG(NOM_PARAMETERS_2)) + REG_SET(NOM_PARAMETERS_2, 0, + DST_Y_PER_PTE_ROW_NOM_C, dlg_attr->dst_y_per_pte_row_nom_c); - REG_SET(NOM_PARAMETERS_3, 0, - REFCYC_PER_PTE_GROUP_NOM_C, dlg_attr->refcyc_per_pte_group_nom_c); + if (REG(NOM_PARAMETERS_3)) + REG_SET(NOM_PARAMETERS_3, 0, + REFCYC_PER_PTE_GROUP_NOM_C, dlg_attr->refcyc_per_pte_group_nom_c); REG_SET(NOM_PARAMETERS_6, 0, DST_Y_PER_META_ROW_NOM_C, dlg_attr->dst_y_per_meta_row_nom_c); @@ -609,6 +637,13 @@ void hubp1_program_deadline( REG_SET(DCN_SURF1_TTU_CNTL1, 0, REFCYC_PER_REQ_DELIVERY_PRE, ttu_attr->refcyc_per_req_delivery_pre_c); + + REG_SET_3(DCN_CUR0_TTU_CNTL0, 0, + REFCYC_PER_REQ_DELIVERY, ttu_attr->refcyc_per_req_delivery_cur0, + QoS_LEVEL_FIXED, ttu_attr->qos_level_fixed_cur0, + QoS_RAMP_DISABLE, ttu_attr->qos_ramp_disable_cur0); + REG_SET(DCN_CUR0_TTU_CNTL1, 0, + REFCYC_PER_REQ_DELIVERY_PRE, ttu_attr->refcyc_per_req_delivery_pre_cur0); } static void hubp1_setup( @@ -752,9 +787,159 @@ void min_set_viewport( PRI_VIEWPORT_Y_START_C, viewport_c->y); } -void hubp1_read_state(struct dcn10_hubp *hubp1, - struct dcn_hubp_state *s) +void hubp1_read_state(struct hubp *hubp) { + struct dcn10_hubp *hubp1 = TO_DCN10_HUBP(hubp); + struct dcn_hubp_state *s = &hubp1->state; + struct _vcs_dpi_display_dlg_regs_st *dlg_attr = &s->dlg_attr; + struct _vcs_dpi_display_ttu_regs_st *ttu_attr = &s->ttu_attr; + struct _vcs_dpi_display_rq_regs_st *rq_regs = &s->rq_regs; + + /* Requester */ + REG_GET(HUBPRET_CONTROL, + DET_BUF_PLANE1_BASE_ADDRESS, &rq_regs->plane1_base_address); + REG_GET_4(DCN_EXPANSION_MODE, + DRQ_EXPANSION_MODE, &rq_regs->drq_expansion_mode, + PRQ_EXPANSION_MODE, &rq_regs->prq_expansion_mode, + MRQ_EXPANSION_MODE, &rq_regs->mrq_expansion_mode, + CRQ_EXPANSION_MODE, &rq_regs->crq_expansion_mode); + REG_GET_8(DCHUBP_REQ_SIZE_CONFIG, + CHUNK_SIZE, &rq_regs->rq_regs_l.chunk_size, + MIN_CHUNK_SIZE, &rq_regs->rq_regs_l.min_chunk_size, + META_CHUNK_SIZE, &rq_regs->rq_regs_l.meta_chunk_size, + MIN_META_CHUNK_SIZE, &rq_regs->rq_regs_l.min_meta_chunk_size, + DPTE_GROUP_SIZE, &rq_regs->rq_regs_l.dpte_group_size, + MPTE_GROUP_SIZE, &rq_regs->rq_regs_l.mpte_group_size, + SWATH_HEIGHT, &rq_regs->rq_regs_l.swath_height, + PTE_ROW_HEIGHT_LINEAR, &rq_regs->rq_regs_l.pte_row_height_linear); + REG_GET_8(DCHUBP_REQ_SIZE_CONFIG_C, + CHUNK_SIZE_C, &rq_regs->rq_regs_c.chunk_size, + MIN_CHUNK_SIZE_C, &rq_regs->rq_regs_c.min_chunk_size, + META_CHUNK_SIZE_C, &rq_regs->rq_regs_c.meta_chunk_size, + MIN_META_CHUNK_SIZE_C, &rq_regs->rq_regs_c.min_meta_chunk_size, + DPTE_GROUP_SIZE_C, &rq_regs->rq_regs_c.dpte_group_size, + MPTE_GROUP_SIZE_C, &rq_regs->rq_regs_c.mpte_group_size, + SWATH_HEIGHT_C, &rq_regs->rq_regs_c.swath_height, + PTE_ROW_HEIGHT_LINEAR_C, &rq_regs->rq_regs_c.pte_row_height_linear); + + /* DLG - Per hubp */ + REG_GET_2(BLANK_OFFSET_0, + REFCYC_H_BLANK_END, &dlg_attr->refcyc_h_blank_end, + DLG_V_BLANK_END, &dlg_attr->dlg_vblank_end); + + REG_GET(BLANK_OFFSET_1, + MIN_DST_Y_NEXT_START, &dlg_attr->min_dst_y_next_start); + + REG_GET(DST_DIMENSIONS, + REFCYC_PER_HTOTAL, &dlg_attr->refcyc_per_htotal); + + REG_GET_2(DST_AFTER_SCALER, + REFCYC_X_AFTER_SCALER, &dlg_attr->refcyc_x_after_scaler, + DST_Y_AFTER_SCALER, &dlg_attr->dst_y_after_scaler); + + if (REG(PREFETCH_SETTINS)) + REG_GET_2(PREFETCH_SETTINS, + DST_Y_PREFETCH, &dlg_attr->dst_y_prefetch, + VRATIO_PREFETCH, &dlg_attr->vratio_prefetch); + else + REG_GET_2(PREFETCH_SETTINGS, + DST_Y_PREFETCH, &dlg_attr->dst_y_prefetch, + VRATIO_PREFETCH, &dlg_attr->vratio_prefetch); + + REG_GET_2(VBLANK_PARAMETERS_0, + DST_Y_PER_VM_VBLANK, &dlg_attr->dst_y_per_vm_vblank, + DST_Y_PER_ROW_VBLANK, &dlg_attr->dst_y_per_row_vblank); + + REG_GET(REF_FREQ_TO_PIX_FREQ, + REF_FREQ_TO_PIX_FREQ, &dlg_attr->ref_freq_to_pix_freq); + + /* DLG - Per luma/chroma */ + REG_GET(VBLANK_PARAMETERS_1, + REFCYC_PER_PTE_GROUP_VBLANK_L, &dlg_attr->refcyc_per_pte_group_vblank_l); + + REG_GET(VBLANK_PARAMETERS_3, + REFCYC_PER_META_CHUNK_VBLANK_L, &dlg_attr->refcyc_per_meta_chunk_vblank_l); + + if (REG(NOM_PARAMETERS_0)) + REG_GET(NOM_PARAMETERS_0, + DST_Y_PER_PTE_ROW_NOM_L, &dlg_attr->dst_y_per_pte_row_nom_l); + + if (REG(NOM_PARAMETERS_1)) + REG_GET(NOM_PARAMETERS_1, + REFCYC_PER_PTE_GROUP_NOM_L, &dlg_attr->refcyc_per_pte_group_nom_l); + + REG_GET(NOM_PARAMETERS_4, + DST_Y_PER_META_ROW_NOM_L, &dlg_attr->dst_y_per_meta_row_nom_l); + + REG_GET(NOM_PARAMETERS_5, + REFCYC_PER_META_CHUNK_NOM_L, &dlg_attr->refcyc_per_meta_chunk_nom_l); + + REG_GET_2(PER_LINE_DELIVERY_PRE, + REFCYC_PER_LINE_DELIVERY_PRE_L, &dlg_attr->refcyc_per_line_delivery_pre_l, + REFCYC_PER_LINE_DELIVERY_PRE_C, &dlg_attr->refcyc_per_line_delivery_pre_c); + + REG_GET_2(PER_LINE_DELIVERY, + REFCYC_PER_LINE_DELIVERY_L, &dlg_attr->refcyc_per_line_delivery_l, + REFCYC_PER_LINE_DELIVERY_C, &dlg_attr->refcyc_per_line_delivery_c); + + if (REG(PREFETCH_SETTINS_C)) + REG_GET(PREFETCH_SETTINS_C, + VRATIO_PREFETCH_C, &dlg_attr->vratio_prefetch_c); + else + REG_GET(PREFETCH_SETTINGS_C, + VRATIO_PREFETCH_C, &dlg_attr->vratio_prefetch_c); + + REG_GET(VBLANK_PARAMETERS_2, + REFCYC_PER_PTE_GROUP_VBLANK_C, &dlg_attr->refcyc_per_pte_group_vblank_c); + + REG_GET(VBLANK_PARAMETERS_4, + REFCYC_PER_META_CHUNK_VBLANK_C, &dlg_attr->refcyc_per_meta_chunk_vblank_c); + + if (REG(NOM_PARAMETERS_2)) + REG_GET(NOM_PARAMETERS_2, + DST_Y_PER_PTE_ROW_NOM_C, &dlg_attr->dst_y_per_pte_row_nom_c); + + if (REG(NOM_PARAMETERS_3)) + REG_GET(NOM_PARAMETERS_3, + REFCYC_PER_PTE_GROUP_NOM_C, &dlg_attr->refcyc_per_pte_group_nom_c); + + REG_GET(NOM_PARAMETERS_6, + DST_Y_PER_META_ROW_NOM_C, &dlg_attr->dst_y_per_meta_row_nom_c); + + REG_GET(NOM_PARAMETERS_7, + REFCYC_PER_META_CHUNK_NOM_C, &dlg_attr->refcyc_per_meta_chunk_nom_c); + + /* TTU - per hubp */ + REG_GET_2(DCN_TTU_QOS_WM, + QoS_LEVEL_LOW_WM, &ttu_attr->qos_level_low_wm, + QoS_LEVEL_HIGH_WM, &ttu_attr->qos_level_high_wm); + + REG_GET_2(DCN_GLOBAL_TTU_CNTL, + MIN_TTU_VBLANK, &ttu_attr->min_ttu_vblank, + QoS_LEVEL_FLIP, &ttu_attr->qos_level_flip); + + /* TTU - per luma/chroma */ + /* Assumed surf0 is luma and 1 is chroma */ + + REG_GET_3(DCN_SURF0_TTU_CNTL0, + REFCYC_PER_REQ_DELIVERY, &ttu_attr->refcyc_per_req_delivery_l, + QoS_LEVEL_FIXED, &ttu_attr->qos_level_fixed_l, + QoS_RAMP_DISABLE, &ttu_attr->qos_ramp_disable_l); + + REG_GET(DCN_SURF0_TTU_CNTL1, + REFCYC_PER_REQ_DELIVERY_PRE, + &ttu_attr->refcyc_per_req_delivery_pre_l); + + REG_GET_3(DCN_SURF1_TTU_CNTL0, + REFCYC_PER_REQ_DELIVERY, &ttu_attr->refcyc_per_req_delivery_c, + QoS_LEVEL_FIXED, &ttu_attr->qos_level_fixed_c, + QoS_RAMP_DISABLE, &ttu_attr->qos_ramp_disable_c); + + REG_GET(DCN_SURF1_TTU_CNTL1, + REFCYC_PER_REQ_DELIVERY_PRE, + &ttu_attr->refcyc_per_req_delivery_pre_c); + + /* Rest of hubp */ REG_GET(DCSURF_SURFACE_CONFIG, SURFACE_PIXEL_FORMAT, &s->pixel_format); @@ -890,14 +1075,14 @@ void hubp1_cursor_set_position( ASSERT(param->h_scale_ratio.value); if (param->h_scale_ratio.value) - dst_x_offset = dal_fixed31_32_floor(dal_fixed31_32_div( - dal_fixed31_32_from_int(dst_x_offset), + dst_x_offset = dc_fixpt_floor(dc_fixpt_div( + dc_fixpt_from_int(dst_x_offset), param->h_scale_ratio)); if (src_x_offset >= (int)param->viewport_width) cur_en = 0; /* not visible beyond right edge*/ - if (src_x_offset + (int)hubp->curs_attr.width < 0) + if (src_x_offset + (int)hubp->curs_attr.width <= 0) cur_en = 0; /* not visible beyond left edge*/ if (cur_en && REG_READ(CURSOR_SURFACE_ADDRESS) == 0) @@ -952,6 +1137,10 @@ static struct hubp_funcs dcn10_hubp_funcs = { .hubp_disconnect = hubp1_disconnect, .hubp_clk_cntl = hubp1_clk_cntl, .hubp_vtg_sel = hubp1_vtg_sel, + .hubp_read_state = hubp1_read_state, + .hubp_disable_control = hubp1_disable_control, + .hubp_get_underflow_status = hubp1_get_underflow_status, + }; /*****************************************/ diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.h index 4a3703e12ea1c..af384034398f0 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hubp.h @@ -30,6 +30,7 @@ #define TO_DCN10_HUBP(hubp)\ container_of(hubp, struct dcn10_hubp, base) +/* Register address initialization macro for all ASICs (including those with reduced functionality) */ #define HUBP_REG_LIST_DCN(id)\ SRI(DCHUBP_CNTL, HUBP, id),\ SRI(HUBPREQ_DEBUG_DB, HUBP, id),\ @@ -78,16 +79,12 @@ SRI(REF_FREQ_TO_PIX_FREQ, HUBPREQ, id),\ SRI(VBLANK_PARAMETERS_1, HUBPREQ, id),\ SRI(VBLANK_PARAMETERS_3, HUBPREQ, id),\ - SRI(NOM_PARAMETERS_0, HUBPREQ, id),\ - SRI(NOM_PARAMETERS_1, HUBPREQ, id),\ SRI(NOM_PARAMETERS_4, HUBPREQ, id),\ SRI(NOM_PARAMETERS_5, HUBPREQ, id),\ SRI(PER_LINE_DELIVERY_PRE, HUBPREQ, id),\ SRI(PER_LINE_DELIVERY, HUBPREQ, id),\ SRI(VBLANK_PARAMETERS_2, HUBPREQ, id),\ SRI(VBLANK_PARAMETERS_4, HUBPREQ, id),\ - SRI(NOM_PARAMETERS_2, HUBPREQ, id),\ - SRI(NOM_PARAMETERS_3, HUBPREQ, id),\ SRI(NOM_PARAMETERS_6, HUBPREQ, id),\ SRI(NOM_PARAMETERS_7, HUBPREQ, id),\ SRI(DCN_TTU_QOS_WM, HUBPREQ, id),\ @@ -96,11 +93,21 @@ SRI(DCN_SURF0_TTU_CNTL1, HUBPREQ, id),\ SRI(DCN_SURF1_TTU_CNTL0, HUBPREQ, id),\ SRI(DCN_SURF1_TTU_CNTL1, HUBPREQ, id),\ - SRI(DCN_VM_MX_L1_TLB_CNTL, HUBPREQ, id),\ + SRI(DCN_CUR0_TTU_CNTL0, HUBPREQ, id),\ + SRI(DCN_CUR0_TTU_CNTL1, HUBPREQ, id),\ SRI(HUBP_CLK_CNTL, HUBP, id) +/* Register address initialization macro for ASICs with VM */ +#define HUBP_REG_LIST_DCN_VM(id)\ + SRI(NOM_PARAMETERS_0, HUBPREQ, id),\ + SRI(NOM_PARAMETERS_1, HUBPREQ, id),\ + SRI(NOM_PARAMETERS_2, HUBPREQ, id),\ + SRI(NOM_PARAMETERS_3, HUBPREQ, id),\ + SRI(DCN_VM_MX_L1_TLB_CNTL, HUBPREQ, id) + #define HUBP_REG_LIST_DCN10(id)\ HUBP_REG_LIST_DCN(id),\ + HUBP_REG_LIST_DCN_VM(id),\ SRI(PREFETCH_SETTINS, HUBPREQ, id),\ SRI(PREFETCH_SETTINS_C, HUBPREQ, id),\ SRI(DCN_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_MSB, HUBPREQ, id),\ @@ -198,6 +205,8 @@ uint32_t DCN_SURF0_TTU_CNTL1; \ uint32_t DCN_SURF1_TTU_CNTL0; \ uint32_t DCN_SURF1_TTU_CNTL1; \ + uint32_t DCN_CUR0_TTU_CNTL0; \ + uint32_t DCN_CUR0_TTU_CNTL1; \ uint32_t DCN_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_MSB; \ uint32_t DCN_VM_CONTEXT0_PAGE_TABLE_BASE_ADDR_LSB; \ uint32_t DCN_VM_CONTEXT0_PAGE_TABLE_START_ADDR_MSB; \ @@ -237,12 +246,14 @@ #define HUBP_SF(reg_name, field_name, post_fix)\ .field_name = reg_name ## __ ## field_name ## post_fix +/* Mask/shift struct generation macro for all ASICs (including those with reduced functionality) */ #define HUBP_MASK_SH_LIST_DCN(mask_sh)\ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_BLANK_EN, mask_sh),\ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_TTU_DISABLE, mask_sh),\ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_UNDERFLOW_STATUS, mask_sh),\ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_NO_OUTSTANDING_REQ, mask_sh),\ HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_VTG_SEL, mask_sh),\ + HUBP_SF(HUBP0_DCHUBP_CNTL, HUBP_DISABLE, mask_sh),\ HUBP_SF(HUBP0_DCSURF_ADDR_CONFIG, NUM_PIPES, mask_sh),\ HUBP_SF(HUBP0_DCSURF_ADDR_CONFIG, NUM_BANKS, mask_sh),\ HUBP_SF(HUBP0_DCSURF_ADDR_CONFIG, PIPE_INTERLEAVE, mask_sh),\ @@ -335,8 +346,6 @@ HUBP_SF(HUBPREQ0_REF_FREQ_TO_PIX_FREQ, REF_FREQ_TO_PIX_FREQ, mask_sh),\ HUBP_SF(HUBPREQ0_VBLANK_PARAMETERS_1, REFCYC_PER_PTE_GROUP_VBLANK_L, mask_sh),\ HUBP_SF(HUBPREQ0_VBLANK_PARAMETERS_3, REFCYC_PER_META_CHUNK_VBLANK_L, mask_sh),\ - HUBP_SF(HUBPREQ0_NOM_PARAMETERS_0, DST_Y_PER_PTE_ROW_NOM_L, mask_sh),\ - HUBP_SF(HUBPREQ0_NOM_PARAMETERS_1, REFCYC_PER_PTE_GROUP_NOM_L, mask_sh),\ HUBP_SF(HUBPREQ0_NOM_PARAMETERS_4, DST_Y_PER_META_ROW_NOM_L, mask_sh),\ HUBP_SF(HUBPREQ0_NOM_PARAMETERS_5, REFCYC_PER_META_CHUNK_NOM_L, mask_sh),\ HUBP_SF(HUBPREQ0_PER_LINE_DELIVERY_PRE, REFCYC_PER_LINE_DELIVERY_PRE_L, mask_sh),\ @@ -345,8 +354,6 @@ HUBP_SF(HUBPREQ0_PER_LINE_DELIVERY, REFCYC_PER_LINE_DELIVERY_C, mask_sh),\ HUBP_SF(HUBPREQ0_VBLANK_PARAMETERS_2, REFCYC_PER_PTE_GROUP_VBLANK_C, mask_sh),\ HUBP_SF(HUBPREQ0_VBLANK_PARAMETERS_4, REFCYC_PER_META_CHUNK_VBLANK_C, mask_sh),\ - HUBP_SF(HUBPREQ0_NOM_PARAMETERS_2, DST_Y_PER_PTE_ROW_NOM_C, mask_sh),\ - HUBP_SF(HUBPREQ0_NOM_PARAMETERS_3, REFCYC_PER_PTE_GROUP_NOM_C, mask_sh),\ HUBP_SF(HUBPREQ0_NOM_PARAMETERS_6, DST_Y_PER_META_ROW_NOM_C, mask_sh),\ HUBP_SF(HUBPREQ0_NOM_PARAMETERS_7, REFCYC_PER_META_CHUNK_NOM_C, mask_sh),\ HUBP_SF(HUBPREQ0_DCN_TTU_QOS_WM, QoS_LEVEL_LOW_WM, mask_sh),\ @@ -357,12 +364,24 @@ HUBP_SF(HUBPREQ0_DCN_SURF0_TTU_CNTL0, QoS_LEVEL_FIXED, mask_sh),\ HUBP_SF(HUBPREQ0_DCN_SURF0_TTU_CNTL0, QoS_RAMP_DISABLE, mask_sh),\ HUBP_SF(HUBPREQ0_DCN_SURF0_TTU_CNTL1, REFCYC_PER_REQ_DELIVERY_PRE, mask_sh),\ + HUBP_SF(HUBP0_HUBP_CLK_CNTL, HUBP_CLOCK_ENABLE, mask_sh) + +/* Mask/shift struct generation macro for ASICs with VM */ +#define HUBP_MASK_SH_LIST_DCN_VM(mask_sh)\ + HUBP_SF(HUBPREQ0_NOM_PARAMETERS_0, DST_Y_PER_PTE_ROW_NOM_L, mask_sh),\ + HUBP_SF(HUBPREQ0_NOM_PARAMETERS_1, REFCYC_PER_PTE_GROUP_NOM_L, mask_sh),\ + HUBP_SF(HUBPREQ0_NOM_PARAMETERS_2, DST_Y_PER_PTE_ROW_NOM_C, mask_sh),\ + HUBP_SF(HUBPREQ0_NOM_PARAMETERS_3, REFCYC_PER_PTE_GROUP_NOM_C, mask_sh),\ HUBP_SF(HUBPREQ0_DCN_VM_MX_L1_TLB_CNTL, ENABLE_L1_TLB, mask_sh),\ HUBP_SF(HUBPREQ0_DCN_VM_MX_L1_TLB_CNTL, SYSTEM_ACCESS_MODE, mask_sh),\ - HUBP_SF(HUBP0_HUBP_CLK_CNTL, HUBP_CLOCK_ENABLE, mask_sh) + HUBP_SF(HUBPREQ0_DCN_CUR0_TTU_CNTL0, REFCYC_PER_REQ_DELIVERY, mask_sh),\ + HUBP_SF(HUBPREQ0_DCN_CUR0_TTU_CNTL0, QoS_LEVEL_FIXED, mask_sh),\ + HUBP_SF(HUBPREQ0_DCN_CUR0_TTU_CNTL0, QoS_RAMP_DISABLE, mask_sh),\ + HUBP_SF(HUBPREQ0_DCN_CUR0_TTU_CNTL1, REFCYC_PER_REQ_DELIVERY_PRE, mask_sh) #define HUBP_MASK_SH_LIST_DCN10(mask_sh)\ HUBP_MASK_SH_LIST_DCN(mask_sh),\ + HUBP_MASK_SH_LIST_DCN_VM(mask_sh),\ HUBP_SF(HUBPREQ0_PREFETCH_SETTINS, DST_Y_PREFETCH, mask_sh),\ HUBP_SF(HUBPREQ0_PREFETCH_SETTINS, VRATIO_PREFETCH, mask_sh),\ HUBP_SF(HUBPREQ0_PREFETCH_SETTINS_C, VRATIO_PREFETCH_C, mask_sh),\ @@ -403,6 +422,7 @@ #define DCN_HUBP_REG_FIELD_LIST(type) \ type HUBP_BLANK_EN;\ + type HUBP_DISABLE;\ type HUBP_TTU_DISABLE;\ type HUBP_NO_OUTSTANDING_REQ;\ type HUBP_VTG_SEL;\ @@ -601,8 +621,29 @@ struct dcn_mi_mask { DCN_HUBP_REG_FIELD_LIST(uint32_t); }; +struct dcn_hubp_state { + struct _vcs_dpi_display_dlg_regs_st dlg_attr; + struct _vcs_dpi_display_ttu_regs_st ttu_attr; + struct _vcs_dpi_display_rq_regs_st rq_regs; + uint32_t pixel_format; + uint32_t inuse_addr_hi; + uint32_t viewport_width; + uint32_t viewport_height; + uint32_t rotation_angle; + uint32_t h_mirror_en; + uint32_t sw_mode; + uint32_t dcc_en; + uint32_t blank_en; + uint32_t underflow_status; + uint32_t ttu_disable; + uint32_t min_ttu_vblank; + uint32_t qos_level_low_wm; + uint32_t qos_level_high_wm; +}; + struct dcn10_hubp { struct hubp base; + struct dcn_hubp_state state; const struct dcn_mi_registers *hubp_regs; const struct dcn_mi_shift *hubp_shift; const struct dcn_mi_mask *hubp_mask; @@ -680,26 +721,9 @@ void dcn10_hubp_construct( const struct dcn_mi_shift *hubp_shift, const struct dcn_mi_mask *hubp_mask); - -struct dcn_hubp_state { - uint32_t pixel_format; - uint32_t inuse_addr_hi; - uint32_t viewport_width; - uint32_t viewport_height; - uint32_t rotation_angle; - uint32_t h_mirror_en; - uint32_t sw_mode; - uint32_t dcc_en; - uint32_t blank_en; - uint32_t underflow_status; - uint32_t ttu_disable; - uint32_t min_ttu_vblank; - uint32_t qos_level_low_wm; - uint32_t qos_level_high_wm; -}; -void hubp1_read_state(struct dcn10_hubp *hubp1, - struct dcn_hubp_state *s); +void hubp1_read_state(struct hubp *hubp); enum cursor_pitch hubp1_get_cursor_pitch(unsigned int pitch); + #endif diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c index 8b0f6b8a56272..f8e0576af6e0b 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.c @@ -45,8 +45,8 @@ #include "dcn10_hubbub.h" #include "dcn10_cm_common.h" -#define DC_LOGGER \ - ctx->logger +#define DC_LOGGER_INIT(logger) + #define CTX \ hws->ctx #define REG(reg)\ @@ -56,16 +56,17 @@ #define FN(reg_name, field_name) \ hws->shifts->field_name, hws->masks->field_name +/*print is 17 wide, first two characters are spaces*/ #define DTN_INFO_MICRO_SEC(ref_cycle) \ print_microsec(dc_ctx, ref_cycle) void print_microsec(struct dc_context *dc_ctx, uint32_t ref_cycle) { - static const uint32_t ref_clk_mhz = 48; - static const unsigned int frac = 10; + const uint32_t ref_clk_mhz = dc_ctx->dc->res_pool->ref_clock_inKhz / 1000; + static const unsigned int frac = 1000; uint32_t us_x10 = (ref_cycle * frac) / ref_clk_mhz; - DTN_INFO("%d.%d \t ", + DTN_INFO(" %11d.%03d", us_x10 / frac, us_x10 % frac); } @@ -92,14 +93,14 @@ void dcn10_log_hubbub_state(struct dc *dc) hubbub1_wm_read_state(dc->res_pool->hubbub, &wm); - DTN_INFO("HUBBUB WM: \t data_urgent \t pte_meta_urgent \t " - "sr_enter \t sr_exit \t dram_clk_change \n"); + DTN_INFO("HUBBUB WM: data_urgent pte_meta_urgent" + " sr_enter sr_exit dram_clk_change\n"); for (i = 0; i < 4; i++) { struct dcn_hubbub_wm_set *s; s = &wm.sets[i]; - DTN_INFO("WM_Set[%d]:\t ", s->wm_set); + DTN_INFO("WM_Set[%d]:", s->wm_set); DTN_INFO_MICRO_SEC(s->data_urgent); DTN_INFO_MICRO_SEC(s->pte_meta_urgent); DTN_INFO_MICRO_SEC(s->sr_enter); @@ -111,6 +112,121 @@ void dcn10_log_hubbub_state(struct dc *dc) DTN_INFO("\n"); } +static void dcn10_log_hubp_states(struct dc *dc) +{ + struct dc_context *dc_ctx = dc->ctx; + struct resource_pool *pool = dc->res_pool; + int i; + + DTN_INFO("HUBP: format addr_hi width height" + " rot mir sw_mode dcc_en blank_en ttu_dis underflow" + " min_ttu_vblank qos_low_wm qos_high_wm\n"); + for (i = 0; i < pool->pipe_count; i++) { + struct hubp *hubp = pool->hubps[i]; + struct dcn_hubp_state *s = &(TO_DCN10_HUBP(hubp)->state); + + hubp->funcs->hubp_read_state(hubp); + + if (!s->blank_en) { + DTN_INFO("[%2d]: %5xh %6xh %5d %6d %2xh %2xh %6xh" + " %6d %8d %7d %8xh", + hubp->inst, + s->pixel_format, + s->inuse_addr_hi, + s->viewport_width, + s->viewport_height, + s->rotation_angle, + s->h_mirror_en, + s->sw_mode, + s->dcc_en, + s->blank_en, + s->ttu_disable, + s->underflow_status); + DTN_INFO_MICRO_SEC(s->min_ttu_vblank); + DTN_INFO_MICRO_SEC(s->qos_level_low_wm); + DTN_INFO_MICRO_SEC(s->qos_level_high_wm); + DTN_INFO("\n"); + } + } + + DTN_INFO("\n=========RQ========\n"); + DTN_INFO("HUBP: drq_exp_m prq_exp_m mrq_exp_m crq_exp_m plane1_ba L:chunk_s min_chu_s meta_ch_s" + " min_m_c_s dpte_gr_s mpte_gr_s swath_hei pte_row_h C:chunk_s min_chu_s meta_ch_s" + " min_m_c_s dpte_gr_s mpte_gr_s swath_hei pte_row_h\n"); + for (i = 0; i < pool->pipe_count; i++) { + struct dcn_hubp_state *s = &(TO_DCN10_HUBP(pool->hubps[i])->state); + struct _vcs_dpi_display_rq_regs_st *rq_regs = &s->rq_regs; + + if (!s->blank_en) + DTN_INFO("[%2d]: %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh\n", + pool->hubps[i]->inst, rq_regs->drq_expansion_mode, rq_regs->prq_expansion_mode, rq_regs->mrq_expansion_mode, + rq_regs->crq_expansion_mode, rq_regs->plane1_base_address, rq_regs->rq_regs_l.chunk_size, + rq_regs->rq_regs_l.min_chunk_size, rq_regs->rq_regs_l.meta_chunk_size, + rq_regs->rq_regs_l.min_meta_chunk_size, rq_regs->rq_regs_l.dpte_group_size, + rq_regs->rq_regs_l.mpte_group_size, rq_regs->rq_regs_l.swath_height, + rq_regs->rq_regs_l.pte_row_height_linear, rq_regs->rq_regs_c.chunk_size, rq_regs->rq_regs_c.min_chunk_size, + rq_regs->rq_regs_c.meta_chunk_size, rq_regs->rq_regs_c.min_meta_chunk_size, + rq_regs->rq_regs_c.dpte_group_size, rq_regs->rq_regs_c.mpte_group_size, + rq_regs->rq_regs_c.swath_height, rq_regs->rq_regs_c.pte_row_height_linear); + } + + DTN_INFO("========DLG========\n"); + DTN_INFO("HUBP: rc_hbe dlg_vbe min_d_y_n rc_per_ht rc_x_a_s " + " dst_y_a_s dst_y_pf dst_y_vvb dst_y_rvb dst_y_vfl dst_y_rfl rf_pix_fq" + " vratio_pf vrat_pf_c rc_pg_vbl rc_pg_vbc rc_mc_vbl rc_mc_vbc rc_pg_fll" + " rc_pg_flc rc_mc_fll rc_mc_flc pr_nom_l pr_nom_c rc_pg_nl rc_pg_nc " + " mr_nom_l mr_nom_c rc_mc_nl rc_mc_nc rc_ld_pl rc_ld_pc rc_ld_l " + " rc_ld_c cha_cur0 ofst_cur1 cha_cur1 vr_af_vc0 ddrq_limt x_rt_dlay" + " x_rp_dlay x_rr_sfl\n"); + for (i = 0; i < pool->pipe_count; i++) { + struct dcn_hubp_state *s = &(TO_DCN10_HUBP(pool->hubps[i])->state); + struct _vcs_dpi_display_dlg_regs_st *dlg_regs = &s->dlg_attr; + + if (!s->blank_en) + DTN_INFO("[%2d]: %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh" + "% 8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh" + " %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh\n", + pool->hubps[i]->inst, dlg_regs->refcyc_h_blank_end, dlg_regs->dlg_vblank_end, dlg_regs->min_dst_y_next_start, + dlg_regs->refcyc_per_htotal, dlg_regs->refcyc_x_after_scaler, dlg_regs->dst_y_after_scaler, + dlg_regs->dst_y_prefetch, dlg_regs->dst_y_per_vm_vblank, dlg_regs->dst_y_per_row_vblank, + dlg_regs->dst_y_per_vm_flip, dlg_regs->dst_y_per_row_flip, dlg_regs->ref_freq_to_pix_freq, + dlg_regs->vratio_prefetch, dlg_regs->vratio_prefetch_c, dlg_regs->refcyc_per_pte_group_vblank_l, + dlg_regs->refcyc_per_pte_group_vblank_c, dlg_regs->refcyc_per_meta_chunk_vblank_l, + dlg_regs->refcyc_per_meta_chunk_vblank_c, dlg_regs->refcyc_per_pte_group_flip_l, + dlg_regs->refcyc_per_pte_group_flip_c, dlg_regs->refcyc_per_meta_chunk_flip_l, + dlg_regs->refcyc_per_meta_chunk_flip_c, dlg_regs->dst_y_per_pte_row_nom_l, + dlg_regs->dst_y_per_pte_row_nom_c, dlg_regs->refcyc_per_pte_group_nom_l, + dlg_regs->refcyc_per_pte_group_nom_c, dlg_regs->dst_y_per_meta_row_nom_l, + dlg_regs->dst_y_per_meta_row_nom_c, dlg_regs->refcyc_per_meta_chunk_nom_l, + dlg_regs->refcyc_per_meta_chunk_nom_c, dlg_regs->refcyc_per_line_delivery_pre_l, + dlg_regs->refcyc_per_line_delivery_pre_c, dlg_regs->refcyc_per_line_delivery_l, + dlg_regs->refcyc_per_line_delivery_c, dlg_regs->chunk_hdl_adjust_cur0, dlg_regs->dst_y_offset_cur1, + dlg_regs->chunk_hdl_adjust_cur1, dlg_regs->vready_after_vcount0, dlg_regs->dst_y_delta_drq_limit, + dlg_regs->xfc_reg_transfer_delay, dlg_regs->xfc_reg_precharge_delay, + dlg_regs->xfc_reg_remote_surface_flip_latency); + } + + DTN_INFO("========TTU========\n"); + DTN_INFO("HUBP: qos_ll_wm qos_lh_wm mn_ttu_vb qos_l_flp rc_rd_p_l rc_rd_l rc_rd_p_c" + " rc_rd_c rc_rd_c0 rc_rd_pc0 rc_rd_c1 rc_rd_pc1 qos_lf_l qos_rds_l" + " qos_lf_c qos_rds_c qos_lf_c0 qos_rds_c0 qos_lf_c1 qos_rds_c1\n"); + for (i = 0; i < pool->pipe_count; i++) { + struct dcn_hubp_state *s = &(TO_DCN10_HUBP(pool->hubps[i])->state); + struct _vcs_dpi_display_ttu_regs_st *ttu_regs = &s->ttu_attr; + + if (!s->blank_en) + DTN_INFO("[%2d]: %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh %8xh\n", + pool->hubps[i]->inst, ttu_regs->qos_level_low_wm, ttu_regs->qos_level_high_wm, ttu_regs->min_ttu_vblank, + ttu_regs->qos_level_flip, ttu_regs->refcyc_per_req_delivery_pre_l, ttu_regs->refcyc_per_req_delivery_l, + ttu_regs->refcyc_per_req_delivery_pre_c, ttu_regs->refcyc_per_req_delivery_c, ttu_regs->refcyc_per_req_delivery_cur0, + ttu_regs->refcyc_per_req_delivery_pre_cur0, ttu_regs->refcyc_per_req_delivery_cur1, + ttu_regs->refcyc_per_req_delivery_pre_cur1, ttu_regs->qos_level_fixed_l, ttu_regs->qos_ramp_disable_l, + ttu_regs->qos_level_fixed_c, ttu_regs->qos_ramp_disable_c, ttu_regs->qos_level_fixed_cur0, + ttu_regs->qos_ramp_disable_cur0, ttu_regs->qos_level_fixed_cur1, ttu_regs->qos_ramp_disable_cur1); + } + DTN_INFO("\n"); +} + void dcn10_log_hw_state(struct dc *dc) { struct dc_context *dc_ctx = dc->ctx; @@ -121,41 +237,64 @@ void dcn10_log_hw_state(struct dc *dc) dcn10_log_hubbub_state(dc); - DTN_INFO("HUBP:\t format \t addr_hi \t width \t height \t " - "rotation \t mirror \t sw_mode \t " - "dcc_en \t blank_en \t ttu_dis \t underflow \t " - "min_ttu_vblank \t qos_low_wm \t qos_high_wm \n"); + dcn10_log_hubp_states(dc); + DTN_INFO("DPP: IGAM format IGAM mode DGAM mode RGAM mode" + " GAMUT mode C11 C12 C13 C14 C21 C22 C23 C24 " + "C31 C32 C33 C34\n"); for (i = 0; i < pool->pipe_count; i++) { - struct hubp *hubp = pool->hubps[i]; - struct dcn_hubp_state s; - - hubp1_read_state(TO_DCN10_HUBP(hubp), &s); - - DTN_INFO("[%d]:\t %xh \t %xh \t %d \t %d \t " - "%xh \t %xh \t %xh \t " - "%d \t %d \t %d \t %xh \t", - hubp->inst, - s.pixel_format, - s.inuse_addr_hi, - s.viewport_width, - s.viewport_height, - s.rotation_angle, - s.h_mirror_en, - s.sw_mode, - s.dcc_en, - s.blank_en, - s.ttu_disable, - s.underflow_status); - DTN_INFO_MICRO_SEC(s.min_ttu_vblank); - DTN_INFO_MICRO_SEC(s.qos_level_low_wm); - DTN_INFO_MICRO_SEC(s.qos_level_high_wm); + struct dpp *dpp = pool->dpps[i]; + struct dcn_dpp_state s; + + dpp->funcs->dpp_read_state(dpp, &s); + + DTN_INFO("[%2d]: %11xh %-11s %-11s %-11s" + "%8x %08xh %08xh %08xh %08xh %08xh %08xh", + dpp->inst, + s.igam_input_format, + (s.igam_lut_mode == 0) ? "BypassFixed" : + ((s.igam_lut_mode == 1) ? "BypassFloat" : + ((s.igam_lut_mode == 2) ? "RAM" : + ((s.igam_lut_mode == 3) ? "RAM" : + "Unknown"))), + (s.dgam_lut_mode == 0) ? "Bypass" : + ((s.dgam_lut_mode == 1) ? "sRGB" : + ((s.dgam_lut_mode == 2) ? "Ycc" : + ((s.dgam_lut_mode == 3) ? "RAM" : + ((s.dgam_lut_mode == 4) ? "RAM" : + "Unknown")))), + (s.rgam_lut_mode == 0) ? "Bypass" : + ((s.rgam_lut_mode == 1) ? "sRGB" : + ((s.rgam_lut_mode == 2) ? "Ycc" : + ((s.rgam_lut_mode == 3) ? "RAM" : + ((s.rgam_lut_mode == 4) ? "RAM" : + "Unknown")))), + s.gamut_remap_mode, + s.gamut_remap_c11_c12, + s.gamut_remap_c13_c14, + s.gamut_remap_c21_c22, + s.gamut_remap_c23_c24, + s.gamut_remap_c31_c32, + s.gamut_remap_c33_c34); DTN_INFO("\n"); } DTN_INFO("\n"); - DTN_INFO("OTG:\t v_bs \t v_be \t v_ss \t v_se \t vpol \t vmax \t vmin \t " - "h_bs \t h_be \t h_ss \t h_se \t hpol \t htot \t vtot \t underflow\n"); + DTN_INFO("MPCC: OPP DPP MPCCBOT MODE ALPHA_MODE PREMULT OVERLAP_ONLY IDLE\n"); + for (i = 0; i < pool->pipe_count; i++) { + struct mpcc_state s = {0}; + + pool->mpc->funcs->read_mpcc_state(pool->mpc, i, &s); + if (s.opp_id != 0xf) + DTN_INFO("[%2d]: %2xh %2xh %6xh %4d %10d %7d %12d %4d\n", + i, s.opp_id, s.dpp_id, s.bot_mpcc_id, + s.mode, s.alpha_mode, s.pre_multiplied_alpha, s.overlap_only, + s.idle); + } + DTN_INFO("\n"); + + DTN_INFO("OTG: v_bs v_be v_ss v_se vpol vmax vmin vmax_sel vmin_sel" + " h_bs h_be h_ss h_se hpol htot vtot underflow\n"); for (i = 0; i < pool->timing_generator_count; i++) { struct timing_generator *tg = pool->timing_generators[i]; @@ -167,9 +306,8 @@ void dcn10_log_hw_state(struct dc *dc) if ((s.otg_enabled & 1) == 0) continue; - DTN_INFO("[%d]:\t %d \t %d \t %d \t %d \t " - "%d \t %d \t %d \t %d \t %d \t %d \t " - "%d \t %d \t %d \t %d \t %d \t ", + DTN_INFO("[%d]: %5d %5d %5d %5d %5d %5d %5d %9d %9d %5d %5d %5d" + " %5d %5d %5d %5d %9d\n", tg->inst, s.v_blank_start, s.v_blank_end, @@ -178,6 +316,8 @@ void dcn10_log_hw_state(struct dc *dc) s.v_sync_a_pol, s.v_total_max, s.v_total_min, + s.v_total_max_sel, + s.v_total_min_sel, s.h_blank_start, s.h_blank_end, s.h_sync_a_start, @@ -186,10 +326,25 @@ void dcn10_log_hw_state(struct dc *dc) s.h_total, s.v_total, s.underflow_occurred_status); - DTN_INFO("\n"); + + // Clear underflow for debug purposes + // We want to keep underflow sticky bit on for the longevity tests outside of test environment. + // This function is called only from Windows or Diags test environment, hence it's safe to clear + // it from here without affecting the original intent. + tg->funcs->clear_optc_underflow(tg); } DTN_INFO("\n"); + DTN_INFO("\nCALCULATED Clocks: dcfclk_khz:%d dcfclk_deep_sleep_khz:%d dispclk_khz:%d\n" + "dppclk_khz:%d max_supported_dppclk_khz:%d fclk_khz:%d socclk_khz:%d\n\n", + dc->current_state->bw.dcn.calc_clk.dcfclk_khz, + dc->current_state->bw.dcn.calc_clk.dcfclk_deep_sleep_khz, + dc->current_state->bw.dcn.calc_clk.dispclk_khz, + dc->current_state->bw.dcn.calc_clk.dppclk_khz, + dc->current_state->bw.dcn.calc_clk.max_supported_dppclk_khz, + dc->current_state->bw.dcn.calc_clk.fclk_khz, + dc->current_state->bw.dcn.calc_clk.socclk_khz); + log_mpc_crc(dc); DTN_INFO_END(); @@ -354,7 +509,7 @@ static void power_on_plane( struct dce_hwseq *hws, int plane_id) { - struct dc_context *ctx = hws->ctx; + DC_LOGGER_INIT(hws->ctx->logger); if (REG(DC_IP_REQUEST_CNTL)) { REG_SET(DC_IP_REQUEST_CNTL, 0, IP_REQUEST_EN, 1); @@ -461,7 +616,7 @@ static void false_optc_underflow_wa( tg->funcs->clear_optc_underflow(tg); } -static enum dc_status dcn10_prog_pixclk_crtc_otg( +static enum dc_status dcn10_enable_stream_timing( struct pipe_ctx *pipe_ctx, struct dc_state *context, struct dc *dc) @@ -553,7 +708,7 @@ static void reset_back_end_for_pipe( struct dc_state *context) { int i; - struct dc_context *ctx = dc->ctx; + DC_LOGGER_INIT(dc->ctx->logger); if (pipe_ctx->stream_res.stream_enc == NULL) { pipe_ctx->stream = NULL; return; @@ -603,6 +758,90 @@ static void reset_back_end_for_pipe( pipe_ctx->pipe_idx, pipe_ctx->stream_res.tg->inst); } +static bool dcn10_hw_wa_force_recovery(struct dc *dc) +{ + struct hubp *hubp ; + unsigned int i; + bool need_recover = true; + + if (!dc->debug.recovery_enabled) + return false; + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = + &dc->current_state->res_ctx.pipe_ctx[i]; + if (pipe_ctx != NULL) { + hubp = pipe_ctx->plane_res.hubp; + if (hubp != NULL) { + if (hubp->funcs->hubp_get_underflow_status(hubp) != 0) { + /* one pipe underflow, we will reset all the pipes*/ + need_recover = true; + } + } + } + } + if (!need_recover) + return false; + /* + DCHUBP_CNTL:HUBP_BLANK_EN=1 + DCHUBBUB_SOFT_RESET:DCHUBBUB_GLOBAL_SOFT_RESET=1 + DCHUBP_CNTL:HUBP_DISABLE=1 + DCHUBP_CNTL:HUBP_DISABLE=0 + DCHUBBUB_SOFT_RESET:DCHUBBUB_GLOBAL_SOFT_RESET=0 + DCSURF_PRIMARY_SURFACE_ADDRESS + DCHUBP_CNTL:HUBP_BLANK_EN=0 + */ + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = + &dc->current_state->res_ctx.pipe_ctx[i]; + if (pipe_ctx != NULL) { + hubp = pipe_ctx->plane_res.hubp; + /*DCHUBP_CNTL:HUBP_BLANK_EN=1*/ + if (hubp != NULL) + hubp->funcs->set_hubp_blank_en(hubp, true); + } + } + /*DCHUBBUB_SOFT_RESET:DCHUBBUB_GLOBAL_SOFT_RESET=1*/ + hubbub1_soft_reset(dc->res_pool->hubbub, true); + + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = + &dc->current_state->res_ctx.pipe_ctx[i]; + if (pipe_ctx != NULL) { + hubp = pipe_ctx->plane_res.hubp; + /*DCHUBP_CNTL:HUBP_DISABLE=1*/ + if (hubp != NULL) + hubp->funcs->hubp_disable_control(hubp, true); + } + } + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = + &dc->current_state->res_ctx.pipe_ctx[i]; + if (pipe_ctx != NULL) { + hubp = pipe_ctx->plane_res.hubp; + /*DCHUBP_CNTL:HUBP_DISABLE=0*/ + if (hubp != NULL) + hubp->funcs->hubp_disable_control(hubp, true); + } + } + /*DCHUBBUB_SOFT_RESET:DCHUBBUB_GLOBAL_SOFT_RESET=0*/ + hubbub1_soft_reset(dc->res_pool->hubbub, false); + for (i = 0; i < dc->res_pool->pipe_count; i++) { + struct pipe_ctx *pipe_ctx = + &dc->current_state->res_ctx.pipe_ctx[i]; + if (pipe_ctx != NULL) { + hubp = pipe_ctx->plane_res.hubp; + /*DCHUBP_CNTL:HUBP_BLANK_EN=0*/ + if (hubp != NULL) + hubp->funcs->set_hubp_blank_en(hubp, true); + } + } + return true; + +} + + static void dcn10_verify_allow_pstate_change_high(struct dc *dc) { static bool should_log_hw_state; /* prevent hw state log by default */ @@ -611,13 +850,17 @@ static void dcn10_verify_allow_pstate_change_high(struct dc *dc) if (should_log_hw_state) { dcn10_log_hw_state(dc); } - BREAK_TO_DEBUGGER(); + if (dcn10_hw_wa_force_recovery(dc)) { + /*check again*/ + if (!hubbub1_verify_allow_pstate_change_high(dc->res_pool->hubbub)) + BREAK_TO_DEBUGGER(); + } } } /* trigger HW to start disconnect plane from stream on the next vsync */ -static void plane_atomic_disconnect(struct dc *dc, struct pipe_ctx *pipe_ctx) +void hwss1_plane_atomic_disconnect(struct dc *dc, struct pipe_ctx *pipe_ctx) { struct hubp *hubp = pipe_ctx->plane_res.hubp; int dpp_id = pipe_ctx->plane_res.dpp->inst; @@ -649,7 +892,7 @@ static void plane_atomic_power_down(struct dc *dc, struct pipe_ctx *pipe_ctx) { struct dce_hwseq *hws = dc->hwseq; struct dpp *dpp = pipe_ctx->plane_res.dpp; - struct dc_context *ctx = dc->ctx; + DC_LOGGER_INIT(dc->ctx->logger); if (REG(DC_IP_REQUEST_CNTL)) { REG_SET(DC_IP_REQUEST_CNTL, 0, @@ -699,7 +942,7 @@ static void plane_atomic_disable(struct dc *dc, struct pipe_ctx *pipe_ctx) static void dcn10_disable_plane(struct dc *dc, struct pipe_ctx *pipe_ctx) { - struct dc_context *ctx = dc->ctx; + DC_LOGGER_INIT(dc->ctx->logger); if (!pipe_ctx->plane_res.hubp || pipe_ctx->plane_res.hubp->power_gated) return; @@ -800,7 +1043,7 @@ static void dcn10_init_hw(struct dc *dc) dc->res_pool->opps[i]->mpcc_disconnect_pending[pipe_ctx->plane_res.mpcc_inst] = true; pipe_ctx->stream_res.opp = dc->res_pool->opps[i]; - plane_atomic_disconnect(dc, pipe_ctx); + hwss1_plane_atomic_disconnect(dc, pipe_ctx); } for (i = 0; i < dc->res_pool->pipe_count; i++) { @@ -945,9 +1188,8 @@ static bool dcn10_set_input_transfer_func(struct pipe_ctx *pipe_ctx, tf = plane_state->in_transfer_func; if (plane_state->gamma_correction && - plane_state->gamma_correction->is_identity) - dpp_base->funcs->dpp_set_degamma(dpp_base, IPP_DEGAMMA_MODE_BYPASS); - else if (plane_state->gamma_correction && dce_use_lut(plane_state->format)) + !plane_state->gamma_correction->is_identity + && dce_use_lut(plane_state->format)) dpp_base->funcs->dpp_program_input_lut(dpp_base, plane_state->gamma_correction); if (tf == NULL) @@ -1433,7 +1675,7 @@ static void program_csc_matrix(struct pipe_ctx *pipe_ctx, } } -static void program_output_csc(struct dc *dc, +static void dcn10_program_output_csc(struct dc *dc, struct pipe_ctx *pipe_ctx, enum dc_color_space colorspace, uint16_t *matrix, @@ -1542,22 +1784,22 @@ static uint16_t fixed_point_to_int_frac( uint16_t result; - uint16_t d = (uint16_t)dal_fixed31_32_floor( - dal_fixed31_32_abs( + uint16_t d = (uint16_t)dc_fixpt_floor( + dc_fixpt_abs( arg)); if (d <= (uint16_t)(1 << integer_bits) - (1 / (uint16_t)divisor)) - numerator = (uint16_t)dal_fixed31_32_floor( - dal_fixed31_32_mul_int( + numerator = (uint16_t)dc_fixpt_floor( + dc_fixpt_mul_int( arg, divisor)); else { - numerator = dal_fixed31_32_floor( - dal_fixed31_32_sub( - dal_fixed31_32_from_int( + numerator = dc_fixpt_floor( + dc_fixpt_sub( + dc_fixpt_from_int( 1LL << integer_bits), - dal_fixed31_32_recip( - dal_fixed31_32_from_int( + dc_fixpt_recip( + dc_fixpt_from_int( divisor)))); } @@ -1567,8 +1809,8 @@ static uint16_t fixed_point_to_int_frac( result = (uint16_t)( (1 << (integer_bits + fractional_bits + 1)) + numerator); - if ((result != 0) && dal_fixed31_32_lt( - arg, dal_fixed31_32_zero)) + if ((result != 0) && dc_fixpt_lt( + arg, dc_fixpt_zero)) result |= 1 << (integer_bits + fractional_bits); return result; @@ -1582,8 +1824,8 @@ void build_prescale_params(struct dc_bias_and_scale *bias_and_scale, && plane_state->input_csc_color_matrix.enable_adjustment && plane_state->coeff_reduction_factor.value != 0) { bias_and_scale->scale_blue = fixed_point_to_int_frac( - dal_fixed31_32_mul(plane_state->coeff_reduction_factor, - dal_fixed31_32_from_fraction(256, 255)), + dc_fixpt_mul(plane_state->coeff_reduction_factor, + dc_fixpt_from_fraction(256, 255)), 2, 13); bias_and_scale->scale_red = bias_and_scale->scale_blue; @@ -1623,6 +1865,8 @@ static void update_mpcc(struct dc *dc, struct pipe_ctx *pipe_ctx) struct mpc *mpc = dc->res_pool->mpc; struct mpc_tree *mpc_tree_params = &(pipe_ctx->stream_res.opp->mpc_tree_params); + + /* TODO: proper fix once fpga works */ if (dc->debug.surface_visual_confirm) @@ -1649,6 +1893,7 @@ static void update_mpcc(struct dc *dc, struct pipe_ctx *pipe_ctx) pipe_ctx->stream->output_color_space) && per_pixel_alpha; + /* * TODO: remove hack * Note: currently there is a bug in init_hw such that @@ -1659,6 +1904,12 @@ static void update_mpcc(struct dc *dc, struct pipe_ctx *pipe_ctx) */ mpcc_id = hubp->inst; + /* If there is no full update, don't need to touch MPC tree*/ + if (!pipe_ctx->plane_state->update_flags.bits.full_update) { + mpc->funcs->update_blending(mpc, &blnd_cfg, mpcc_id); + return; + } + /* check if this MPCC is already being used */ new_mpcc = mpc->funcs->get_mpcc_for_dpp(mpc_tree_params, mpcc_id); /* remove MPCC if being used */ @@ -1777,7 +2028,7 @@ static void update_dchubp_dpp( /*gamut remap*/ program_gamut_remap(pipe_ctx); - program_output_csc(dc, + dc->hwss.program_output_csc(dc, pipe_ctx, pipe_ctx->stream->output_color_space, pipe_ctx->stream->csc_color_matrix.matrix, @@ -1810,9 +2061,9 @@ static void update_dchubp_dpp( hubp->funcs->set_blank(hubp, false); } -static void dcn10_otg_blank( +static void dcn10_blank_pixel_data( struct dc *dc, - struct stream_resource stream_res, + struct stream_resource *stream_res, struct dc_stream_state *stream, bool blank) { @@ -1823,27 +2074,27 @@ static void dcn10_otg_blank( color_space = stream->output_color_space; color_space_to_black_color(dc, color_space, &black_color); - if (stream_res.tg->funcs->set_blank_color) - stream_res.tg->funcs->set_blank_color( - stream_res.tg, + if (stream_res->tg->funcs->set_blank_color) + stream_res->tg->funcs->set_blank_color( + stream_res->tg, &black_color); if (!blank) { - if (stream_res.tg->funcs->set_blank) - stream_res.tg->funcs->set_blank(stream_res.tg, blank); - if (stream_res.abm) - stream_res.abm->funcs->set_abm_level(stream_res.abm, stream->abm_level); + if (stream_res->tg->funcs->set_blank) + stream_res->tg->funcs->set_blank(stream_res->tg, blank); + if (stream_res->abm) + stream_res->abm->funcs->set_abm_level(stream_res->abm, stream->abm_level); } else if (blank) { - if (stream_res.abm) - stream_res.abm->funcs->set_abm_immediate_disable(stream_res.abm); - if (stream_res.tg->funcs->set_blank) - stream_res.tg->funcs->set_blank(stream_res.tg, blank); + if (stream_res->abm) + stream_res->abm->funcs->set_abm_immediate_disable(stream_res->abm); + if (stream_res->tg->funcs->set_blank) + stream_res->tg->funcs->set_blank(stream_res->tg, blank); } } static void set_hdr_multiplier(struct pipe_ctx *pipe_ctx) { - struct fixed31_32 multiplier = dal_fixed31_32_from_fraction( + struct fixed31_32 multiplier = dc_fixpt_from_fraction( pipe_ctx->plane_state->sdr_white_level, 80); uint32_t hw_mult = 0x1f000; // 1.0 default multiplier struct custom_float_format fmt; @@ -1876,7 +2127,7 @@ static void program_all_pipe_in_tree( pipe_ctx->stream_res.tg->funcs->program_global_sync( pipe_ctx->stream_res.tg); - dcn10_otg_blank(dc, pipe_ctx->stream_res, + dc->hwss.blank_pixel_data(dc, &pipe_ctx->stream_res, pipe_ctx->stream, blank); } @@ -1983,9 +2234,9 @@ static void dcn10_apply_ctx_for_surface( bool removed_pipe[4] = { false }; unsigned int ref_clk_mhz = dc->res_pool->ref_clock_inKhz/1000; bool program_water_mark = false; - struct dc_context *ctx = dc->ctx; struct pipe_ctx *top_pipe_to_program = find_top_pipe_for_stream(dc, context, stream); + DC_LOGGER_INIT(dc->ctx->logger); if (!top_pipe_to_program) return; @@ -1996,7 +2247,7 @@ static void dcn10_apply_ctx_for_surface( if (num_planes == 0) { /* OTG blank before remove all front end */ - dcn10_otg_blank(dc, top_pipe_to_program->stream_res, top_pipe_to_program->stream, true); + dc->hwss.blank_pixel_data(dc, &top_pipe_to_program->stream_res, top_pipe_to_program->stream, true); } /* Disconnect unused mpcc */ @@ -2027,7 +2278,7 @@ static void dcn10_apply_ctx_for_surface( old_pipe_ctx->plane_state && old_pipe_ctx->stream_res.tg == tg) { - plane_atomic_disconnect(dc, old_pipe_ctx); + hwss1_plane_atomic_disconnect(dc, old_pipe_ctx); removed_pipe[i] = true; DC_LOG_DC( @@ -2335,15 +2586,6 @@ static void set_static_screen_control(struct pipe_ctx **pipe_ctx, set_static_screen_control(pipe_ctx[i]->stream_res.tg, value); } -static void set_plane_config( - const struct dc *dc, - struct pipe_ctx *pipe_ctx, - struct resource_context *res_ctx) -{ - /* TODO */ - program_gamut_remap(pipe_ctx); -} - static void dcn10_config_stereo_parameters( struct dc_stream_state *stream, struct crtc_stereo_flags *flags) { @@ -2521,12 +2763,12 @@ static const struct hw_sequencer_funcs dcn10_funcs = { .init_hw = dcn10_init_hw, .apply_ctx_to_hw = dce110_apply_ctx_to_hw, .apply_ctx_for_surface = dcn10_apply_ctx_for_surface, - .set_plane_config = set_plane_config, .update_plane_addr = dcn10_update_plane_addr, .update_dchub = dcn10_update_dchub, .update_pending_status = dcn10_update_pending_status, .set_input_transfer_func = dcn10_set_input_transfer_func, .set_output_transfer_func = dcn10_set_output_transfer_func, + .program_output_csc = dcn10_program_output_csc, .power_down = dce110_power_down, .enable_accelerated_mode = dce110_enable_accelerated_mode, .enable_timing_synchronization = dcn10_enable_timing_synchronization, @@ -2538,10 +2780,11 @@ static const struct hw_sequencer_funcs dcn10_funcs = { .blank_stream = dce110_blank_stream, .enable_display_power_gating = dcn10_dummy_display_power_gating, .disable_plane = dcn10_disable_plane, + .blank_pixel_data = dcn10_blank_pixel_data, .pipe_control_lock = dcn10_pipe_control_lock, .set_bandwidth = dcn10_set_bandwidth, .reset_hw_ctx_wrap = reset_hw_ctx_wrap, - .prog_pixclk_crtc_otg = dcn10_prog_pixclk_crtc_otg, + .enable_stream_timing = dcn10_enable_stream_timing, .set_drr = set_drr, .get_position = get_position, .set_static_screen_control = set_static_screen_control, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h index 6c526b5095d90..44f734b73f9ea 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_hw_sequencer.h @@ -37,4 +37,6 @@ extern void fill_display_configs( bool is_rgb_cspace(enum dc_color_space output_color_space); +void hwss1_plane_atomic_disconnect(struct dc *dc, struct pipe_ctx *pipe_ctx); + #endif /* __DC_HWSS_DCN10_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.c new file mode 100644 index 0000000000000..21fa40ac0786b --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.c @@ -0,0 +1,1362 @@ +/* + * Copyright 2012-15 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + +#include "reg_helper.h" + +#include "core_types.h" +#include "link_encoder.h" +#include "dcn10_link_encoder.h" +#include "stream_encoder.h" +#include "i2caux_interface.h" +#include "dc_bios_types.h" + +#include "gpio_service_interface.h" + +#define CTX \ + enc10->base.ctx +#define DC_LOGGER \ + enc10->base.ctx->logger + +#define REG(reg)\ + (enc10->link_regs->reg) + +#undef FN +#define FN(reg_name, field_name) \ + enc10->link_shift->field_name, enc10->link_mask->field_name + + +/* + * @brief + * Trigger Source Select + * ASIC-dependent, actual values for register programming + */ +#define DCN10_DIG_FE_SOURCE_SELECT_INVALID 0x0 +#define DCN10_DIG_FE_SOURCE_SELECT_DIGA 0x1 +#define DCN10_DIG_FE_SOURCE_SELECT_DIGB 0x2 +#define DCN10_DIG_FE_SOURCE_SELECT_DIGC 0x4 +#define DCN10_DIG_FE_SOURCE_SELECT_DIGD 0x08 +#define DCN10_DIG_FE_SOURCE_SELECT_DIGE 0x10 +#define DCN10_DIG_FE_SOURCE_SELECT_DIGF 0x20 +#define DCN10_DIG_FE_SOURCE_SELECT_DIGG 0x40 + +enum { + DP_MST_UPDATE_MAX_RETRY = 50 +}; + + + +static void aux_initialize(struct dcn10_link_encoder *enc10); + + +static const struct link_encoder_funcs dcn10_lnk_enc_funcs = { + .validate_output_with_stream = + dcn10_link_encoder_validate_output_with_stream, + .hw_init = dcn10_link_encoder_hw_init, + .setup = dcn10_link_encoder_setup, + .enable_tmds_output = dcn10_link_encoder_enable_tmds_output, + .enable_dp_output = dcn10_link_encoder_enable_dp_output, + .enable_dp_mst_output = dcn10_link_encoder_enable_dp_mst_output, + .disable_output = dcn10_link_encoder_disable_output, + .dp_set_lane_settings = dcn10_link_encoder_dp_set_lane_settings, + .dp_set_phy_pattern = dcn10_link_encoder_dp_set_phy_pattern, + .update_mst_stream_allocation_table = + dcn10_link_encoder_update_mst_stream_allocation_table, + .psr_program_dp_dphy_fast_training = + dcn10_psr_program_dp_dphy_fast_training, + .psr_program_secondary_packet = dcn10_psr_program_secondary_packet, + .connect_dig_be_to_fe = dcn10_link_encoder_connect_dig_be_to_fe, + .enable_hpd = dcn10_link_encoder_enable_hpd, + .disable_hpd = dcn10_link_encoder_disable_hpd, + .is_dig_enabled = dcn10_is_dig_enabled, + .destroy = dcn10_link_encoder_destroy +}; + +static enum bp_result link_transmitter_control( + struct dcn10_link_encoder *enc10, + struct bp_transmitter_control *cntl) +{ + enum bp_result result; + struct dc_bios *bp = enc10->base.ctx->dc_bios; + + result = bp->funcs->transmitter_control(bp, cntl); + + return result; +} + +static void enable_phy_bypass_mode( + struct dcn10_link_encoder *enc10, + bool enable) +{ + /* This register resides in DP back end block; + * transmitter is used for the offset + */ + REG_UPDATE(DP_DPHY_CNTL, DPHY_BYPASS, enable); + +} + +static void disable_prbs_symbols( + struct dcn10_link_encoder *enc10, + bool disable) +{ + /* This register resides in DP back end block; + * transmitter is used for the offset + */ + REG_UPDATE_4(DP_DPHY_CNTL, + DPHY_ATEST_SEL_LANE0, disable, + DPHY_ATEST_SEL_LANE1, disable, + DPHY_ATEST_SEL_LANE2, disable, + DPHY_ATEST_SEL_LANE3, disable); +} + +static void disable_prbs_mode( + struct dcn10_link_encoder *enc10) +{ + REG_UPDATE(DP_DPHY_PRBS_CNTL, DPHY_PRBS_EN, 0); +} + +static void program_pattern_symbols( + struct dcn10_link_encoder *enc10, + uint16_t pattern_symbols[8]) +{ + /* This register resides in DP back end block; + * transmitter is used for the offset + */ + REG_SET_3(DP_DPHY_SYM0, 0, + DPHY_SYM1, pattern_symbols[0], + DPHY_SYM2, pattern_symbols[1], + DPHY_SYM3, pattern_symbols[2]); + + /* This register resides in DP back end block; + * transmitter is used for the offset + */ + REG_SET_3(DP_DPHY_SYM1, 0, + DPHY_SYM4, pattern_symbols[3], + DPHY_SYM5, pattern_symbols[4], + DPHY_SYM6, pattern_symbols[5]); + + /* This register resides in DP back end block; + * transmitter is used for the offset + */ + REG_SET_2(DP_DPHY_SYM2, 0, + DPHY_SYM7, pattern_symbols[6], + DPHY_SYM8, pattern_symbols[7]); +} + +static void set_dp_phy_pattern_d102( + struct dcn10_link_encoder *enc10) +{ + /* Disable PHY Bypass mode to setup the test pattern */ + enable_phy_bypass_mode(enc10, false); + + /* For 10-bit PRBS or debug symbols + * please use the following sequence: + * + * Enable debug symbols on the lanes + */ + disable_prbs_symbols(enc10, true); + + /* Disable PRBS mode */ + disable_prbs_mode(enc10); + + /* Program debug symbols to be output */ + { + uint16_t pattern_symbols[8] = { + 0x2AA, 0x2AA, 0x2AA, 0x2AA, + 0x2AA, 0x2AA, 0x2AA, 0x2AA + }; + + program_pattern_symbols(enc10, pattern_symbols); + } + + /* Enable phy bypass mode to enable the test pattern */ + + enable_phy_bypass_mode(enc10, true); +} + +static void set_link_training_complete( + struct dcn10_link_encoder *enc10, + bool complete) +{ + /* This register resides in DP back end block; + * transmitter is used for the offset + */ + REG_UPDATE(DP_LINK_CNTL, DP_LINK_TRAINING_COMPLETE, complete); + +} + +void dcn10_link_encoder_set_dp_phy_pattern_training_pattern( + struct link_encoder *enc, + uint32_t index) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + /* Write Training Pattern */ + + REG_WRITE(DP_DPHY_TRAINING_PATTERN_SEL, index); + + /* Set HW Register Training Complete to false */ + + set_link_training_complete(enc10, false); + + /* Disable PHY Bypass mode to output Training Pattern */ + + enable_phy_bypass_mode(enc10, false); + + /* Disable PRBS mode */ + disable_prbs_mode(enc10); +} + +static void setup_panel_mode( + struct dcn10_link_encoder *enc10, + enum dp_panel_mode panel_mode) +{ + uint32_t value; + + ASSERT(REG(DP_DPHY_INTERNAL_CTRL)); + value = REG_READ(DP_DPHY_INTERNAL_CTRL); + + switch (panel_mode) { + case DP_PANEL_MODE_EDP: + value = 0x1; + break; + case DP_PANEL_MODE_SPECIAL: + value = 0x11; + break; + default: + value = 0x0; + break; + } + + REG_WRITE(DP_DPHY_INTERNAL_CTRL, value); +} + +static void set_dp_phy_pattern_symbol_error( + struct dcn10_link_encoder *enc10) +{ + /* Disable PHY Bypass mode to setup the test pattern */ + enable_phy_bypass_mode(enc10, false); + + /* program correct panel mode*/ + setup_panel_mode(enc10, DP_PANEL_MODE_DEFAULT); + + /* A PRBS23 pattern is used for most DP electrical measurements. */ + + /* Enable PRBS symbols on the lanes */ + disable_prbs_symbols(enc10, false); + + /* For PRBS23 Set bit DPHY_PRBS_SEL=1 and Set bit DPHY_PRBS_EN=1 */ + REG_UPDATE_2(DP_DPHY_PRBS_CNTL, + DPHY_PRBS_SEL, 1, + DPHY_PRBS_EN, 1); + + /* Enable phy bypass mode to enable the test pattern */ + enable_phy_bypass_mode(enc10, true); +} + +static void set_dp_phy_pattern_prbs7( + struct dcn10_link_encoder *enc10) +{ + /* Disable PHY Bypass mode to setup the test pattern */ + enable_phy_bypass_mode(enc10, false); + + /* A PRBS7 pattern is used for most DP electrical measurements. */ + + /* Enable PRBS symbols on the lanes */ + disable_prbs_symbols(enc10, false); + + /* For PRBS7 Set bit DPHY_PRBS_SEL=0 and Set bit DPHY_PRBS_EN=1 */ + REG_UPDATE_2(DP_DPHY_PRBS_CNTL, + DPHY_PRBS_SEL, 0, + DPHY_PRBS_EN, 1); + + /* Enable phy bypass mode to enable the test pattern */ + enable_phy_bypass_mode(enc10, true); +} + +static void set_dp_phy_pattern_80bit_custom( + struct dcn10_link_encoder *enc10, + const uint8_t *pattern) +{ + /* Disable PHY Bypass mode to setup the test pattern */ + enable_phy_bypass_mode(enc10, false); + + /* Enable debug symbols on the lanes */ + + disable_prbs_symbols(enc10, true); + + /* Enable PHY bypass mode to enable the test pattern */ + /* TODO is it really needed ? */ + + enable_phy_bypass_mode(enc10, true); + + /* Program 80 bit custom pattern */ + { + uint16_t pattern_symbols[8]; + + pattern_symbols[0] = + ((pattern[1] & 0x03) << 8) | pattern[0]; + pattern_symbols[1] = + ((pattern[2] & 0x0f) << 6) | ((pattern[1] >> 2) & 0x3f); + pattern_symbols[2] = + ((pattern[3] & 0x3f) << 4) | ((pattern[2] >> 4) & 0x0f); + pattern_symbols[3] = + (pattern[4] << 2) | ((pattern[3] >> 6) & 0x03); + pattern_symbols[4] = + ((pattern[6] & 0x03) << 8) | pattern[5]; + pattern_symbols[5] = + ((pattern[7] & 0x0f) << 6) | ((pattern[6] >> 2) & 0x3f); + pattern_symbols[6] = + ((pattern[8] & 0x3f) << 4) | ((pattern[7] >> 4) & 0x0f); + pattern_symbols[7] = + (pattern[9] << 2) | ((pattern[8] >> 6) & 0x03); + + program_pattern_symbols(enc10, pattern_symbols); + } + + /* Enable phy bypass mode to enable the test pattern */ + + enable_phy_bypass_mode(enc10, true); +} + +static void set_dp_phy_pattern_hbr2_compliance_cp2520_2( + struct dcn10_link_encoder *enc10, + unsigned int cp2520_pattern) +{ + + /* previously there is a register DP_HBR2_EYE_PATTERN + * that is enabled to get the pattern. + * But it does not work with the latest spec change, + * so we are programming the following registers manually. + * + * The following settings have been confirmed + * by Nick Chorney and Sandra Liu + */ + + /* Disable PHY Bypass mode to setup the test pattern */ + + enable_phy_bypass_mode(enc10, false); + + /* Setup DIG encoder in DP SST mode */ + enc10->base.funcs->setup(&enc10->base, SIGNAL_TYPE_DISPLAY_PORT); + + /* ensure normal panel mode. */ + setup_panel_mode(enc10, DP_PANEL_MODE_DEFAULT); + + /* no vbid after BS (SR) + * DP_LINK_FRAMING_CNTL changed history Sandra Liu + * 11000260 / 11000104 / 110000FC + */ + REG_UPDATE_3(DP_LINK_FRAMING_CNTL, + DP_IDLE_BS_INTERVAL, 0xFC, + DP_VBID_DISABLE, 1, + DP_VID_ENHANCED_FRAME_MODE, 1); + + /* swap every BS with SR */ + REG_UPDATE(DP_DPHY_SCRAM_CNTL, DPHY_SCRAMBLER_BS_COUNT, 0); + + /* select cp2520 patterns */ + if (REG(DP_DPHY_HBR2_PATTERN_CONTROL)) + REG_UPDATE(DP_DPHY_HBR2_PATTERN_CONTROL, + DP_DPHY_HBR2_PATTERN_CONTROL, cp2520_pattern); + else + /* pre-DCE11 can only generate CP2520 pattern 2 */ + ASSERT(cp2520_pattern == 2); + + /* set link training complete */ + set_link_training_complete(enc10, true); + + /* disable video stream */ + REG_UPDATE(DP_VID_STREAM_CNTL, DP_VID_STREAM_ENABLE, 0); + + /* Disable PHY Bypass mode to setup the test pattern */ + enable_phy_bypass_mode(enc10, false); +} + +static void set_dp_phy_pattern_passthrough_mode( + struct dcn10_link_encoder *enc10, + enum dp_panel_mode panel_mode) +{ + /* program correct panel mode */ + setup_panel_mode(enc10, panel_mode); + + /* restore LINK_FRAMING_CNTL and DPHY_SCRAMBLER_BS_COUNT + * in case we were doing HBR2 compliance pattern before + */ + REG_UPDATE_3(DP_LINK_FRAMING_CNTL, + DP_IDLE_BS_INTERVAL, 0x2000, + DP_VBID_DISABLE, 0, + DP_VID_ENHANCED_FRAME_MODE, 1); + + REG_UPDATE(DP_DPHY_SCRAM_CNTL, DPHY_SCRAMBLER_BS_COUNT, 0x1FF); + + /* set link training complete */ + set_link_training_complete(enc10, true); + + /* Disable PHY Bypass mode to setup the test pattern */ + enable_phy_bypass_mode(enc10, false); + + /* Disable PRBS mode */ + disable_prbs_mode(enc10); +} + +/* return value is bit-vector */ +static uint8_t get_frontend_source( + enum engine_id engine) +{ + switch (engine) { + case ENGINE_ID_DIGA: + return DCN10_DIG_FE_SOURCE_SELECT_DIGA; + case ENGINE_ID_DIGB: + return DCN10_DIG_FE_SOURCE_SELECT_DIGB; + case ENGINE_ID_DIGC: + return DCN10_DIG_FE_SOURCE_SELECT_DIGC; + case ENGINE_ID_DIGD: + return DCN10_DIG_FE_SOURCE_SELECT_DIGD; + case ENGINE_ID_DIGE: + return DCN10_DIG_FE_SOURCE_SELECT_DIGE; + case ENGINE_ID_DIGF: + return DCN10_DIG_FE_SOURCE_SELECT_DIGF; + case ENGINE_ID_DIGG: + return DCN10_DIG_FE_SOURCE_SELECT_DIGG; + default: + ASSERT_CRITICAL(false); + return DCN10_DIG_FE_SOURCE_SELECT_INVALID; + } +} + +static void configure_encoder( + struct dcn10_link_encoder *enc10, + const struct dc_link_settings *link_settings) +{ + /* set number of lanes */ + + REG_SET(DP_CONFIG, 0, + DP_UDI_LANES, link_settings->lane_count - LANE_COUNT_ONE); + + /* setup scrambler */ + REG_UPDATE(DP_DPHY_SCRAM_CNTL, DPHY_SCRAMBLER_ADVANCE, 1); +} + +void dcn10_psr_program_dp_dphy_fast_training(struct link_encoder *enc, + bool exit_link_training_required) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + + if (exit_link_training_required) + REG_UPDATE(DP_DPHY_FAST_TRAINING, + DPHY_RX_FAST_TRAINING_CAPABLE, 1); + else { + REG_UPDATE(DP_DPHY_FAST_TRAINING, + DPHY_RX_FAST_TRAINING_CAPABLE, 0); + /*In DCE 11, we are able to pre-program a Force SR register + * to be able to trigger SR symbol after 5 idle patterns + * transmitted. Upon PSR Exit, DMCU can trigger + * DPHY_LOAD_BS_COUNT_START = 1. Upon writing 1 to + * DPHY_LOAD_BS_COUNT_START and the internal counter + * reaches DPHY_LOAD_BS_COUNT, the next BS symbol will be + * replaced by SR symbol once. + */ + + REG_UPDATE(DP_DPHY_BS_SR_SWAP_CNTL, DPHY_LOAD_BS_COUNT, 0x5); + } +} + +void dcn10_psr_program_secondary_packet(struct link_encoder *enc, + unsigned int sdp_transmit_line_num_deadline) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + + REG_UPDATE_2(DP_SEC_CNTL1, + DP_SEC_GSP0_LINE_NUM, sdp_transmit_line_num_deadline, + DP_SEC_GSP0_PRIORITY, 1); +} + +bool dcn10_is_dig_enabled(struct link_encoder *enc) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + uint32_t value; + + REG_GET(DIG_BE_EN_CNTL, DIG_ENABLE, &value); + return value; +} + +static void link_encoder_disable(struct dcn10_link_encoder *enc10) +{ + /* reset training pattern */ + REG_SET(DP_DPHY_TRAINING_PATTERN_SEL, 0, + DPHY_TRAINING_PATTERN_SEL, 0); + + /* reset training complete */ + REG_UPDATE(DP_LINK_CNTL, DP_LINK_TRAINING_COMPLETE, 0); + + /* reset panel mode */ + setup_panel_mode(enc10, DP_PANEL_MODE_DEFAULT); +} + +static void hpd_initialize( + struct dcn10_link_encoder *enc10) +{ + /* Associate HPD with DIG_BE */ + enum hpd_source_id hpd_source = enc10->base.hpd_source; + + REG_UPDATE(DIG_BE_CNTL, DIG_HPD_SELECT, hpd_source); +} + +bool dcn10_link_encoder_validate_dvi_output( + const struct dcn10_link_encoder *enc10, + enum signal_type connector_signal, + enum signal_type signal, + const struct dc_crtc_timing *crtc_timing) +{ + uint32_t max_pixel_clock = TMDS_MAX_PIXEL_CLOCK; + + if (signal == SIGNAL_TYPE_DVI_DUAL_LINK) + max_pixel_clock *= 2; + + /* This handles the case of HDMI downgrade to DVI we don't want to + * we don't want to cap the pixel clock if the DDI is not DVI. + */ + if (connector_signal != SIGNAL_TYPE_DVI_DUAL_LINK && + connector_signal != SIGNAL_TYPE_DVI_SINGLE_LINK) + max_pixel_clock = enc10->base.features.max_hdmi_pixel_clock; + + /* DVI only support RGB pixel encoding */ + if (crtc_timing->pixel_encoding != PIXEL_ENCODING_RGB) + return false; + + /*connect DVI via adpater's HDMI connector*/ + if ((connector_signal == SIGNAL_TYPE_DVI_SINGLE_LINK || + connector_signal == SIGNAL_TYPE_HDMI_TYPE_A) && + signal != SIGNAL_TYPE_HDMI_TYPE_A && + crtc_timing->pix_clk_khz > TMDS_MAX_PIXEL_CLOCK) + return false; + if (crtc_timing->pix_clk_khz < TMDS_MIN_PIXEL_CLOCK) + return false; + + if (crtc_timing->pix_clk_khz > max_pixel_clock) + return false; + + /* DVI supports 6/8bpp single-link and 10/16bpp dual-link */ + switch (crtc_timing->display_color_depth) { + case COLOR_DEPTH_666: + case COLOR_DEPTH_888: + break; + case COLOR_DEPTH_101010: + case COLOR_DEPTH_161616: + if (signal != SIGNAL_TYPE_DVI_DUAL_LINK) + return false; + break; + default: + return false; + } + + return true; +} + +static bool dcn10_link_encoder_validate_hdmi_output( + const struct dcn10_link_encoder *enc10, + const struct dc_crtc_timing *crtc_timing, + int adjusted_pix_clk_khz) +{ + enum dc_color_depth max_deep_color = + enc10->base.features.max_hdmi_deep_color; + + if (max_deep_color < crtc_timing->display_color_depth) + return false; + + if (crtc_timing->display_color_depth < COLOR_DEPTH_888) + return false; + if (adjusted_pix_clk_khz < TMDS_MIN_PIXEL_CLOCK) + return false; + + if ((adjusted_pix_clk_khz == 0) || + (adjusted_pix_clk_khz > enc10->base.features.max_hdmi_pixel_clock)) + return false; + + /* DCE11 HW does not support 420 */ + if (!enc10->base.features.ycbcr420_supported && + crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR420) + return false; + + if (!enc10->base.features.flags.bits.HDMI_6GB_EN && + adjusted_pix_clk_khz >= 300000) + return false; + return true; +} + +bool dcn10_link_encoder_validate_dp_output( + const struct dcn10_link_encoder *enc10, + const struct dc_crtc_timing *crtc_timing) +{ + /* default RGB only */ + if (crtc_timing->pixel_encoding == PIXEL_ENCODING_RGB) + return true; + + if (enc10->base.features.flags.bits.IS_YCBCR_CAPABLE) + return true; + + /* for DCE 8.x or later DP Y-only feature, + * we need ASIC cap + FeatureSupportDPYonly, not support 666 + */ + if (crtc_timing->flags.Y_ONLY && + enc10->base.features.flags.bits.IS_YCBCR_CAPABLE && + crtc_timing->display_color_depth != COLOR_DEPTH_666) + return true; + + return false; +} + +void dcn10_link_encoder_construct( + struct dcn10_link_encoder *enc10, + const struct encoder_init_data *init_data, + const struct encoder_feature_support *enc_features, + const struct dcn10_link_enc_registers *link_regs, + const struct dcn10_link_enc_aux_registers *aux_regs, + const struct dcn10_link_enc_hpd_registers *hpd_regs, + const struct dcn10_link_enc_shift *link_shift, + const struct dcn10_link_enc_mask *link_mask) +{ + struct bp_encoder_cap_info bp_cap_info = {0}; + const struct dc_vbios_funcs *bp_funcs = init_data->ctx->dc_bios->funcs; + enum bp_result result = BP_RESULT_OK; + + enc10->base.funcs = &dcn10_lnk_enc_funcs; + enc10->base.ctx = init_data->ctx; + enc10->base.id = init_data->encoder; + + enc10->base.hpd_source = init_data->hpd_source; + enc10->base.connector = init_data->connector; + + enc10->base.preferred_engine = ENGINE_ID_UNKNOWN; + + enc10->base.features = *enc_features; + + enc10->base.transmitter = init_data->transmitter; + + /* set the flag to indicate whether driver poll the I2C data pin + * while doing the DP sink detect + */ + +/* if (dal_adapter_service_is_feature_supported(as, + FEATURE_DP_SINK_DETECT_POLL_DATA_PIN)) + enc10->base.features.flags.bits. + DP_SINK_DETECT_POLL_DATA_PIN = true;*/ + + enc10->base.output_signals = + SIGNAL_TYPE_DVI_SINGLE_LINK | + SIGNAL_TYPE_DVI_DUAL_LINK | + SIGNAL_TYPE_LVDS | + SIGNAL_TYPE_DISPLAY_PORT | + SIGNAL_TYPE_DISPLAY_PORT_MST | + SIGNAL_TYPE_EDP | + SIGNAL_TYPE_HDMI_TYPE_A; + + /* For DCE 8.0 and 8.1, by design, UNIPHY is hardwired to DIG_BE. + * SW always assign DIG_FE 1:1 mapped to DIG_FE for non-MST UNIPHY. + * SW assign DIG_FE to non-MST UNIPHY first and MST last. So prefer + * DIG is per UNIPHY and used by SST DP, eDP, HDMI, DVI and LVDS. + * Prefer DIG assignment is decided by board design. + * For DCE 8.0, there are only max 6 UNIPHYs, we assume board design + * and VBIOS will filter out 7 UNIPHY for DCE 8.0. + * By this, adding DIGG should not hurt DCE 8.0. + * This will let DCE 8.1 share DCE 8.0 as much as possible + */ + + enc10->link_regs = link_regs; + enc10->aux_regs = aux_regs; + enc10->hpd_regs = hpd_regs; + enc10->link_shift = link_shift; + enc10->link_mask = link_mask; + + switch (enc10->base.transmitter) { + case TRANSMITTER_UNIPHY_A: + enc10->base.preferred_engine = ENGINE_ID_DIGA; + break; + case TRANSMITTER_UNIPHY_B: + enc10->base.preferred_engine = ENGINE_ID_DIGB; + break; + case TRANSMITTER_UNIPHY_C: + enc10->base.preferred_engine = ENGINE_ID_DIGC; + break; + case TRANSMITTER_UNIPHY_D: + enc10->base.preferred_engine = ENGINE_ID_DIGD; + break; + case TRANSMITTER_UNIPHY_E: + enc10->base.preferred_engine = ENGINE_ID_DIGE; + break; + case TRANSMITTER_UNIPHY_F: + enc10->base.preferred_engine = ENGINE_ID_DIGF; + break; + case TRANSMITTER_UNIPHY_G: + enc10->base.preferred_engine = ENGINE_ID_DIGG; + break; + default: + ASSERT_CRITICAL(false); + enc10->base.preferred_engine = ENGINE_ID_UNKNOWN; + } + + /* default to one to mirror Windows behavior */ + enc10->base.features.flags.bits.HDMI_6GB_EN = 1; + + result = bp_funcs->get_encoder_cap_info(enc10->base.ctx->dc_bios, + enc10->base.id, &bp_cap_info); + + /* Override features with DCE-specific values */ + if (result == BP_RESULT_OK) { + enc10->base.features.flags.bits.IS_HBR2_CAPABLE = + bp_cap_info.DP_HBR2_EN; + enc10->base.features.flags.bits.IS_HBR3_CAPABLE = + bp_cap_info.DP_HBR3_EN; + enc10->base.features.flags.bits.HDMI_6GB_EN = bp_cap_info.HDMI_6GB_EN; + } else { + DC_LOG_WARNING("%s: Failed to get encoder_cap_info from VBIOS with error code %d!\n", + __func__, + result); + } +} + +bool dcn10_link_encoder_validate_output_with_stream( + struct link_encoder *enc, + const struct dc_stream_state *stream) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + bool is_valid; + + switch (stream->signal) { + case SIGNAL_TYPE_DVI_SINGLE_LINK: + case SIGNAL_TYPE_DVI_DUAL_LINK: + is_valid = dcn10_link_encoder_validate_dvi_output( + enc10, + stream->sink->link->connector_signal, + stream->signal, + &stream->timing); + break; + case SIGNAL_TYPE_HDMI_TYPE_A: + is_valid = dcn10_link_encoder_validate_hdmi_output( + enc10, + &stream->timing, + stream->phy_pix_clk); + break; + case SIGNAL_TYPE_DISPLAY_PORT: + case SIGNAL_TYPE_DISPLAY_PORT_MST: + is_valid = dcn10_link_encoder_validate_dp_output( + enc10, &stream->timing); + break; + case SIGNAL_TYPE_EDP: + is_valid = (stream->timing.pixel_encoding == PIXEL_ENCODING_RGB) ? true : false; + break; + case SIGNAL_TYPE_VIRTUAL: + is_valid = true; + break; + default: + is_valid = false; + break; + } + + return is_valid; +} + +void dcn10_link_encoder_hw_init( + struct link_encoder *enc) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + struct bp_transmitter_control cntl = { 0 }; + enum bp_result result; + + cntl.action = TRANSMITTER_CONTROL_INIT; + cntl.engine_id = ENGINE_ID_UNKNOWN; + cntl.transmitter = enc10->base.transmitter; + cntl.connector_obj_id = enc10->base.connector; + cntl.lanes_number = LANE_COUNT_FOUR; + cntl.coherent = false; + cntl.hpd_sel = enc10->base.hpd_source; + + if (enc10->base.connector.id == CONNECTOR_ID_EDP) + cntl.signal = SIGNAL_TYPE_EDP; + + result = link_transmitter_control(enc10, &cntl); + + if (result != BP_RESULT_OK) { + DC_LOG_ERROR("%s: Failed to execute VBIOS command table!\n", + __func__); + BREAK_TO_DEBUGGER(); + return; + } + + if (enc10->base.connector.id == CONNECTOR_ID_LVDS) { + cntl.action = TRANSMITTER_CONTROL_BACKLIGHT_BRIGHTNESS; + + result = link_transmitter_control(enc10, &cntl); + + ASSERT(result == BP_RESULT_OK); + + } + aux_initialize(enc10); + + /* reinitialize HPD. + * hpd_initialize() will pass DIG_FE id to HW context. + * All other routine within HW context will use fe_engine_offset + * as DIG_FE id even caller pass DIG_FE id. + * So this routine must be called first. + */ + hpd_initialize(enc10); +} + +void dcn10_link_encoder_destroy(struct link_encoder **enc) +{ + kfree(TO_DCN10_LINK_ENC(*enc)); + *enc = NULL; +} + +void dcn10_link_encoder_setup( + struct link_encoder *enc, + enum signal_type signal) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + + switch (signal) { + case SIGNAL_TYPE_EDP: + case SIGNAL_TYPE_DISPLAY_PORT: + /* DP SST */ + REG_UPDATE(DIG_BE_CNTL, DIG_MODE, 0); + break; + case SIGNAL_TYPE_LVDS: + /* LVDS */ + REG_UPDATE(DIG_BE_CNTL, DIG_MODE, 1); + break; + case SIGNAL_TYPE_DVI_SINGLE_LINK: + case SIGNAL_TYPE_DVI_DUAL_LINK: + /* TMDS-DVI */ + REG_UPDATE(DIG_BE_CNTL, DIG_MODE, 2); + break; + case SIGNAL_TYPE_HDMI_TYPE_A: + /* TMDS-HDMI */ + REG_UPDATE(DIG_BE_CNTL, DIG_MODE, 3); + break; + case SIGNAL_TYPE_DISPLAY_PORT_MST: + /* DP MST */ + REG_UPDATE(DIG_BE_CNTL, DIG_MODE, 5); + break; + default: + ASSERT_CRITICAL(false); + /* invalid mode ! */ + break; + } + +} + +/* TODO: still need depth or just pass in adjusted pixel clock? */ +void dcn10_link_encoder_enable_tmds_output( + struct link_encoder *enc, + enum clock_source_id clock_source, + enum dc_color_depth color_depth, + enum signal_type signal, + uint32_t pixel_clock) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + struct bp_transmitter_control cntl = { 0 }; + enum bp_result result; + + /* Enable the PHY */ + + cntl.action = TRANSMITTER_CONTROL_ENABLE; + cntl.engine_id = enc->preferred_engine; + cntl.transmitter = enc10->base.transmitter; + cntl.pll_id = clock_source; + cntl.signal = signal; + if (cntl.signal == SIGNAL_TYPE_DVI_DUAL_LINK) + cntl.lanes_number = 8; + else + cntl.lanes_number = 4; + + cntl.hpd_sel = enc10->base.hpd_source; + + cntl.pixel_clock = pixel_clock; + cntl.color_depth = color_depth; + + result = link_transmitter_control(enc10, &cntl); + + if (result != BP_RESULT_OK) { + DC_LOG_ERROR("%s: Failed to execute VBIOS command table!\n", + __func__); + BREAK_TO_DEBUGGER(); + } +} + +/* enables DP PHY output */ +void dcn10_link_encoder_enable_dp_output( + struct link_encoder *enc, + const struct dc_link_settings *link_settings, + enum clock_source_id clock_source) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + struct bp_transmitter_control cntl = { 0 }; + enum bp_result result; + + /* Enable the PHY */ + + /* number_of_lanes is used for pixel clock adjust, + * but it's not passed to asic_control. + * We need to set number of lanes manually. + */ + configure_encoder(enc10, link_settings); + + cntl.action = TRANSMITTER_CONTROL_ENABLE; + cntl.engine_id = enc->preferred_engine; + cntl.transmitter = enc10->base.transmitter; + cntl.pll_id = clock_source; + cntl.signal = SIGNAL_TYPE_DISPLAY_PORT; + cntl.lanes_number = link_settings->lane_count; + cntl.hpd_sel = enc10->base.hpd_source; + cntl.pixel_clock = link_settings->link_rate + * LINK_RATE_REF_FREQ_IN_KHZ; + /* TODO: check if undefined works */ + cntl.color_depth = COLOR_DEPTH_UNDEFINED; + + result = link_transmitter_control(enc10, &cntl); + + if (result != BP_RESULT_OK) { + DC_LOG_ERROR("%s: Failed to execute VBIOS command table!\n", + __func__); + BREAK_TO_DEBUGGER(); + } +} + +/* enables DP PHY output in MST mode */ +void dcn10_link_encoder_enable_dp_mst_output( + struct link_encoder *enc, + const struct dc_link_settings *link_settings, + enum clock_source_id clock_source) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + struct bp_transmitter_control cntl = { 0 }; + enum bp_result result; + + /* Enable the PHY */ + + /* number_of_lanes is used for pixel clock adjust, + * but it's not passed to asic_control. + * We need to set number of lanes manually. + */ + configure_encoder(enc10, link_settings); + + cntl.action = TRANSMITTER_CONTROL_ENABLE; + cntl.engine_id = ENGINE_ID_UNKNOWN; + cntl.transmitter = enc10->base.transmitter; + cntl.pll_id = clock_source; + cntl.signal = SIGNAL_TYPE_DISPLAY_PORT_MST; + cntl.lanes_number = link_settings->lane_count; + cntl.hpd_sel = enc10->base.hpd_source; + cntl.pixel_clock = link_settings->link_rate + * LINK_RATE_REF_FREQ_IN_KHZ; + /* TODO: check if undefined works */ + cntl.color_depth = COLOR_DEPTH_UNDEFINED; + + result = link_transmitter_control(enc10, &cntl); + + if (result != BP_RESULT_OK) { + DC_LOG_ERROR("%s: Failed to execute VBIOS command table!\n", + __func__); + BREAK_TO_DEBUGGER(); + } +} +/* + * @brief + * Disable transmitter and its encoder + */ +void dcn10_link_encoder_disable_output( + struct link_encoder *enc, + enum signal_type signal) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + struct bp_transmitter_control cntl = { 0 }; + enum bp_result result; + + if (!dcn10_is_dig_enabled(enc)) { + /* OF_SKIP_POWER_DOWN_INACTIVE_ENCODER */ + return; + } + /* Power-down RX and disable GPU PHY should be paired. + * Disabling PHY without powering down RX may cause + * symbol lock loss, on which we will get DP Sink interrupt. + */ + + /* There is a case for the DP active dongles + * where we want to disable the PHY but keep RX powered, + * for those we need to ignore DP Sink interrupt + * by checking lane count that has been set + * on the last do_enable_output(). + */ + + /* disable transmitter */ + cntl.action = TRANSMITTER_CONTROL_DISABLE; + cntl.transmitter = enc10->base.transmitter; + cntl.hpd_sel = enc10->base.hpd_source; + cntl.signal = signal; + cntl.connector_obj_id = enc10->base.connector; + + result = link_transmitter_control(enc10, &cntl); + + if (result != BP_RESULT_OK) { + DC_LOG_ERROR("%s: Failed to execute VBIOS command table!\n", + __func__); + BREAK_TO_DEBUGGER(); + return; + } + + /* disable encoder */ + if (dc_is_dp_signal(signal)) + link_encoder_disable(enc10); +} + +void dcn10_link_encoder_dp_set_lane_settings( + struct link_encoder *enc, + const struct link_training_settings *link_settings) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + union dpcd_training_lane_set training_lane_set = { { 0 } }; + int32_t lane = 0; + struct bp_transmitter_control cntl = { 0 }; + + if (!link_settings) { + BREAK_TO_DEBUGGER(); + return; + } + + cntl.action = TRANSMITTER_CONTROL_SET_VOLTAGE_AND_PREEMPASIS; + cntl.transmitter = enc10->base.transmitter; + cntl.connector_obj_id = enc10->base.connector; + cntl.lanes_number = link_settings->link_settings.lane_count; + cntl.hpd_sel = enc10->base.hpd_source; + cntl.pixel_clock = link_settings->link_settings.link_rate * + LINK_RATE_REF_FREQ_IN_KHZ; + + for (lane = 0; lane < link_settings->link_settings.lane_count; lane++) { + /* translate lane settings */ + + training_lane_set.bits.VOLTAGE_SWING_SET = + link_settings->lane_settings[lane].VOLTAGE_SWING; + training_lane_set.bits.PRE_EMPHASIS_SET = + link_settings->lane_settings[lane].PRE_EMPHASIS; + + /* post cursor 2 setting only applies to HBR2 link rate */ + if (link_settings->link_settings.link_rate == LINK_RATE_HIGH2) { + /* this is passed to VBIOS + * to program post cursor 2 level + */ + training_lane_set.bits.POST_CURSOR2_SET = + link_settings->lane_settings[lane].POST_CURSOR2; + } + + cntl.lane_select = lane; + cntl.lane_settings = training_lane_set.raw; + + /* call VBIOS table to set voltage swing and pre-emphasis */ + link_transmitter_control(enc10, &cntl); + } +} + +/* set DP PHY test and training patterns */ +void dcn10_link_encoder_dp_set_phy_pattern( + struct link_encoder *enc, + const struct encoder_set_dp_phy_pattern_param *param) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + + switch (param->dp_phy_pattern) { + case DP_TEST_PATTERN_TRAINING_PATTERN1: + dcn10_link_encoder_set_dp_phy_pattern_training_pattern(enc, 0); + break; + case DP_TEST_PATTERN_TRAINING_PATTERN2: + dcn10_link_encoder_set_dp_phy_pattern_training_pattern(enc, 1); + break; + case DP_TEST_PATTERN_TRAINING_PATTERN3: + dcn10_link_encoder_set_dp_phy_pattern_training_pattern(enc, 2); + break; + case DP_TEST_PATTERN_TRAINING_PATTERN4: + dcn10_link_encoder_set_dp_phy_pattern_training_pattern(enc, 3); + break; + case DP_TEST_PATTERN_D102: + set_dp_phy_pattern_d102(enc10); + break; + case DP_TEST_PATTERN_SYMBOL_ERROR: + set_dp_phy_pattern_symbol_error(enc10); + break; + case DP_TEST_PATTERN_PRBS7: + set_dp_phy_pattern_prbs7(enc10); + break; + case DP_TEST_PATTERN_80BIT_CUSTOM: + set_dp_phy_pattern_80bit_custom( + enc10, param->custom_pattern); + break; + case DP_TEST_PATTERN_CP2520_1: + set_dp_phy_pattern_hbr2_compliance_cp2520_2(enc10, 1); + break; + case DP_TEST_PATTERN_CP2520_2: + set_dp_phy_pattern_hbr2_compliance_cp2520_2(enc10, 2); + break; + case DP_TEST_PATTERN_CP2520_3: + set_dp_phy_pattern_hbr2_compliance_cp2520_2(enc10, 3); + break; + case DP_TEST_PATTERN_VIDEO_MODE: { + set_dp_phy_pattern_passthrough_mode( + enc10, param->dp_panel_mode); + break; + } + + default: + /* invalid phy pattern */ + ASSERT_CRITICAL(false); + break; + } +} + +static void fill_stream_allocation_row_info( + const struct link_mst_stream_allocation *stream_allocation, + uint32_t *src, + uint32_t *slots) +{ + const struct stream_encoder *stream_enc = stream_allocation->stream_enc; + + if (stream_enc) { + *src = stream_enc->id; + *slots = stream_allocation->slot_count; + } else { + *src = 0; + *slots = 0; + } +} + +/* programs DP MST VC payload allocation */ +void dcn10_link_encoder_update_mst_stream_allocation_table( + struct link_encoder *enc, + const struct link_mst_stream_allocation_table *table) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + uint32_t value0 = 0; + uint32_t value1 = 0; + uint32_t value2 = 0; + uint32_t slots = 0; + uint32_t src = 0; + uint32_t retries = 0; + + /* For CZ, there are only 3 pipes. So Virtual channel is up 3.*/ + + /* --- Set MSE Stream Attribute - + * Setup VC Payload Table on Tx Side, + * Issue allocation change trigger + * to commit payload on both tx and rx side + */ + + /* we should clean-up table each time */ + + if (table->stream_count >= 1) { + fill_stream_allocation_row_info( + &table->stream_allocations[0], + &src, + &slots); + } else { + src = 0; + slots = 0; + } + + REG_UPDATE_2(DP_MSE_SAT0, + DP_MSE_SAT_SRC0, src, + DP_MSE_SAT_SLOT_COUNT0, slots); + + if (table->stream_count >= 2) { + fill_stream_allocation_row_info( + &table->stream_allocations[1], + &src, + &slots); + } else { + src = 0; + slots = 0; + } + + REG_UPDATE_2(DP_MSE_SAT0, + DP_MSE_SAT_SRC1, src, + DP_MSE_SAT_SLOT_COUNT1, slots); + + if (table->stream_count >= 3) { + fill_stream_allocation_row_info( + &table->stream_allocations[2], + &src, + &slots); + } else { + src = 0; + slots = 0; + } + + REG_UPDATE_2(DP_MSE_SAT1, + DP_MSE_SAT_SRC2, src, + DP_MSE_SAT_SLOT_COUNT2, slots); + + if (table->stream_count >= 4) { + fill_stream_allocation_row_info( + &table->stream_allocations[3], + &src, + &slots); + } else { + src = 0; + slots = 0; + } + + REG_UPDATE_2(DP_MSE_SAT1, + DP_MSE_SAT_SRC3, src, + DP_MSE_SAT_SLOT_COUNT3, slots); + + /* --- wait for transaction finish */ + + /* send allocation change trigger (ACT) ? + * this step first sends the ACT, + * then double buffers the SAT into the hardware + * making the new allocation active on the DP MST mode link + */ + + /* DP_MSE_SAT_UPDATE: + * 0 - No Action + * 1 - Update SAT with trigger + * 2 - Update SAT without trigger + */ + REG_UPDATE(DP_MSE_SAT_UPDATE, + DP_MSE_SAT_UPDATE, 1); + + /* wait for update to complete + * (i.e. DP_MSE_SAT_UPDATE field is reset to 0) + * then wait for the transmission + * of at least 16 MTP headers on immediate local link. + * i.e. DP_MSE_16_MTP_KEEPOUT field (read only) is reset to 0 + * a value of 1 indicates that DP MST mode + * is in the 16 MTP keepout region after a VC has been added. + * MST stream bandwidth (VC rate) can be configured + * after this bit is cleared + */ + do { + udelay(10); + + value0 = REG_READ(DP_MSE_SAT_UPDATE); + + REG_GET(DP_MSE_SAT_UPDATE, + DP_MSE_SAT_UPDATE, &value1); + + REG_GET(DP_MSE_SAT_UPDATE, + DP_MSE_16_MTP_KEEPOUT, &value2); + + /* bit field DP_MSE_SAT_UPDATE is set to 1 already */ + if (!value1 && !value2) + break; + ++retries; + } while (retries < DP_MST_UPDATE_MAX_RETRY); +} + +void dcn10_link_encoder_connect_dig_be_to_fe( + struct link_encoder *enc, + enum engine_id engine, + bool connect) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + uint32_t field; + + if (engine != ENGINE_ID_UNKNOWN) { + + REG_GET(DIG_BE_CNTL, DIG_FE_SOURCE_SELECT, &field); + + if (connect) + field |= get_frontend_source(engine); + else + field &= ~get_frontend_source(engine); + + REG_UPDATE(DIG_BE_CNTL, DIG_FE_SOURCE_SELECT, field); + } +} + + +#define HPD_REG(reg)\ + (enc10->hpd_regs->reg) + +#define HPD_REG_READ(reg_name) \ + dm_read_reg(CTX, HPD_REG(reg_name)) + +#define HPD_REG_UPDATE_N(reg_name, n, ...) \ + generic_reg_update_ex(CTX, \ + HPD_REG(reg_name), \ + HPD_REG_READ(reg_name), \ + n, __VA_ARGS__) + +#define HPD_REG_UPDATE(reg_name, field, val) \ + HPD_REG_UPDATE_N(reg_name, 1, \ + FN(reg_name, field), val) + +void dcn10_link_encoder_enable_hpd(struct link_encoder *enc) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + + HPD_REG_UPDATE(DC_HPD_CONTROL, + DC_HPD_EN, 1); +} + +void dcn10_link_encoder_disable_hpd(struct link_encoder *enc) +{ + struct dcn10_link_encoder *enc10 = TO_DCN10_LINK_ENC(enc); + + HPD_REG_UPDATE(DC_HPD_CONTROL, + DC_HPD_EN, 0); +} + + +#define AUX_REG(reg)\ + (enc10->aux_regs->reg) + +#define AUX_REG_READ(reg_name) \ + dm_read_reg(CTX, AUX_REG(reg_name)) + +#define AUX_REG_UPDATE_N(reg_name, n, ...) \ + generic_reg_update_ex(CTX, \ + AUX_REG(reg_name), \ + AUX_REG_READ(reg_name), \ + n, __VA_ARGS__) + +#define AUX_REG_UPDATE(reg_name, field, val) \ + AUX_REG_UPDATE_N(reg_name, 1, \ + FN(reg_name, field), val) + +#define AUX_REG_UPDATE_2(reg, f1, v1, f2, v2) \ + AUX_REG_UPDATE_N(reg, 2,\ + FN(reg, f1), v1,\ + FN(reg, f2), v2) + +static void aux_initialize( + struct dcn10_link_encoder *enc10) +{ + enum hpd_source_id hpd_source = enc10->base.hpd_source; + + AUX_REG_UPDATE_2(AUX_CONTROL, + AUX_HPD_SEL, hpd_source, + AUX_LS_READ_EN, 0); + + /* 1/4 window (the maximum allowed) */ + AUX_REG_UPDATE(AUX_DPHY_RX_CONTROL0, + AUX_RX_RECEIVE_WINDOW, 1); +} diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.h new file mode 100644 index 0000000000000..2a97cdb2cfbb3 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_link_encoder.h @@ -0,0 +1,330 @@ +/* + * Copyright 2012-15 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + +#ifndef __DC_LINK_ENCODER__DCN10_H__ +#define __DC_LINK_ENCODER__DCN10_H__ + +#include "link_encoder.h" + +#define TO_DCN10_LINK_ENC(link_encoder)\ + container_of(link_encoder, struct dcn10_link_encoder, base) + + +#define AUX_REG_LIST(id)\ + SRI(AUX_CONTROL, DP_AUX, id), \ + SRI(AUX_DPHY_RX_CONTROL0, DP_AUX, id) + +#define HPD_REG_LIST(id)\ + SRI(DC_HPD_CONTROL, HPD, id) + +#define LE_DCN_COMMON_REG_LIST(id) \ + SRI(DIG_BE_CNTL, DIG, id), \ + SRI(DIG_BE_EN_CNTL, DIG, id), \ + SRI(DP_CONFIG, DP, id), \ + SRI(DP_DPHY_CNTL, DP, id), \ + SRI(DP_DPHY_PRBS_CNTL, DP, id), \ + SRI(DP_DPHY_SCRAM_CNTL, DP, id),\ + SRI(DP_DPHY_SYM0, DP, id), \ + SRI(DP_DPHY_SYM1, DP, id), \ + SRI(DP_DPHY_SYM2, DP, id), \ + SRI(DP_DPHY_TRAINING_PATTERN_SEL, DP, id), \ + SRI(DP_LINK_CNTL, DP, id), \ + SRI(DP_LINK_FRAMING_CNTL, DP, id), \ + SRI(DP_MSE_SAT0, DP, id), \ + SRI(DP_MSE_SAT1, DP, id), \ + SRI(DP_MSE_SAT2, DP, id), \ + SRI(DP_MSE_SAT_UPDATE, DP, id), \ + SRI(DP_SEC_CNTL, DP, id), \ + SRI(DP_VID_STREAM_CNTL, DP, id), \ + SRI(DP_DPHY_FAST_TRAINING, DP, id), \ + SRI(DP_SEC_CNTL1, DP, id), \ + SRI(DP_DPHY_BS_SR_SWAP_CNTL, DP, id), \ + SRI(DP_DPHY_INTERNAL_CTRL, DP, id), \ + SRI(DP_DPHY_HBR2_PATTERN_CONTROL, DP, id) + +#define LE_DCN10_REG_LIST(id)\ + LE_DCN_COMMON_REG_LIST(id) + +struct dcn10_link_enc_aux_registers { + uint32_t AUX_CONTROL; + uint32_t AUX_DPHY_RX_CONTROL0; +}; + +struct dcn10_link_enc_hpd_registers { + uint32_t DC_HPD_CONTROL; +}; + +struct dcn10_link_enc_registers { + uint32_t DIG_BE_CNTL; + uint32_t DIG_BE_EN_CNTL; + uint32_t DP_CONFIG; + uint32_t DP_DPHY_CNTL; + uint32_t DP_DPHY_INTERNAL_CTRL; + uint32_t DP_DPHY_PRBS_CNTL; + uint32_t DP_DPHY_SCRAM_CNTL; + uint32_t DP_DPHY_SYM0; + uint32_t DP_DPHY_SYM1; + uint32_t DP_DPHY_SYM2; + uint32_t DP_DPHY_TRAINING_PATTERN_SEL; + uint32_t DP_LINK_CNTL; + uint32_t DP_LINK_FRAMING_CNTL; + uint32_t DP_MSE_SAT0; + uint32_t DP_MSE_SAT1; + uint32_t DP_MSE_SAT2; + uint32_t DP_MSE_SAT_UPDATE; + uint32_t DP_SEC_CNTL; + uint32_t DP_VID_STREAM_CNTL; + uint32_t DP_DPHY_FAST_TRAINING; + uint32_t DP_DPHY_BS_SR_SWAP_CNTL; + uint32_t DP_DPHY_HBR2_PATTERN_CONTROL; + uint32_t DP_SEC_CNTL1; +}; + +#define LE_SF(reg_name, field_name, post_fix)\ + .field_name = reg_name ## __ ## field_name ## post_fix + +#define LINK_ENCODER_MASK_SH_LIST_DCN10(mask_sh)\ + LE_SF(DIG0_DIG_BE_EN_CNTL, DIG_ENABLE, mask_sh),\ + LE_SF(DIG0_DIG_BE_CNTL, DIG_HPD_SELECT, mask_sh),\ + LE_SF(DIG0_DIG_BE_CNTL, DIG_MODE, mask_sh),\ + LE_SF(DIG0_DIG_BE_CNTL, DIG_FE_SOURCE_SELECT, mask_sh),\ + LE_SF(DP0_DP_DPHY_CNTL, DPHY_BYPASS, mask_sh),\ + LE_SF(DP0_DP_DPHY_CNTL, DPHY_ATEST_SEL_LANE0, mask_sh),\ + LE_SF(DP0_DP_DPHY_CNTL, DPHY_ATEST_SEL_LANE1, mask_sh),\ + LE_SF(DP0_DP_DPHY_CNTL, DPHY_ATEST_SEL_LANE2, mask_sh),\ + LE_SF(DP0_DP_DPHY_CNTL, DPHY_ATEST_SEL_LANE3, mask_sh),\ + LE_SF(DP0_DP_DPHY_PRBS_CNTL, DPHY_PRBS_EN, mask_sh),\ + LE_SF(DP0_DP_DPHY_PRBS_CNTL, DPHY_PRBS_SEL, mask_sh),\ + LE_SF(DP0_DP_DPHY_SYM0, DPHY_SYM1, mask_sh),\ + LE_SF(DP0_DP_DPHY_SYM0, DPHY_SYM2, mask_sh),\ + LE_SF(DP0_DP_DPHY_SYM0, DPHY_SYM3, mask_sh),\ + LE_SF(DP0_DP_DPHY_SYM1, DPHY_SYM4, mask_sh),\ + LE_SF(DP0_DP_DPHY_SYM1, DPHY_SYM5, mask_sh),\ + LE_SF(DP0_DP_DPHY_SYM1, DPHY_SYM6, mask_sh),\ + LE_SF(DP0_DP_DPHY_SYM2, DPHY_SYM7, mask_sh),\ + LE_SF(DP0_DP_DPHY_SYM2, DPHY_SYM8, mask_sh),\ + LE_SF(DP0_DP_DPHY_SCRAM_CNTL, DPHY_SCRAMBLER_BS_COUNT, mask_sh),\ + LE_SF(DP0_DP_DPHY_SCRAM_CNTL, DPHY_SCRAMBLER_ADVANCE, mask_sh),\ + LE_SF(DP0_DP_DPHY_FAST_TRAINING, DPHY_RX_FAST_TRAINING_CAPABLE, mask_sh),\ + LE_SF(DP0_DP_DPHY_BS_SR_SWAP_CNTL, DPHY_LOAD_BS_COUNT, mask_sh),\ + LE_SF(DP0_DP_DPHY_TRAINING_PATTERN_SEL, DPHY_TRAINING_PATTERN_SEL, mask_sh),\ + LE_SF(DP0_DP_DPHY_HBR2_PATTERN_CONTROL, DP_DPHY_HBR2_PATTERN_CONTROL, mask_sh),\ + LE_SF(DP0_DP_LINK_CNTL, DP_LINK_TRAINING_COMPLETE, mask_sh),\ + LE_SF(DP0_DP_LINK_FRAMING_CNTL, DP_IDLE_BS_INTERVAL, mask_sh),\ + LE_SF(DP0_DP_LINK_FRAMING_CNTL, DP_VBID_DISABLE, mask_sh),\ + LE_SF(DP0_DP_LINK_FRAMING_CNTL, DP_VID_ENHANCED_FRAME_MODE, mask_sh),\ + LE_SF(DP0_DP_VID_STREAM_CNTL, DP_VID_STREAM_ENABLE, mask_sh),\ + LE_SF(DP0_DP_CONFIG, DP_UDI_LANES, mask_sh),\ + LE_SF(DP0_DP_SEC_CNTL1, DP_SEC_GSP0_LINE_NUM, mask_sh),\ + LE_SF(DP0_DP_SEC_CNTL1, DP_SEC_GSP0_PRIORITY, mask_sh),\ + LE_SF(DP0_DP_MSE_SAT0, DP_MSE_SAT_SRC0, mask_sh),\ + LE_SF(DP0_DP_MSE_SAT0, DP_MSE_SAT_SRC1, mask_sh),\ + LE_SF(DP0_DP_MSE_SAT0, DP_MSE_SAT_SLOT_COUNT0, mask_sh),\ + LE_SF(DP0_DP_MSE_SAT0, DP_MSE_SAT_SLOT_COUNT1, mask_sh),\ + LE_SF(DP0_DP_MSE_SAT1, DP_MSE_SAT_SRC2, mask_sh),\ + LE_SF(DP0_DP_MSE_SAT1, DP_MSE_SAT_SRC3, mask_sh),\ + LE_SF(DP0_DP_MSE_SAT1, DP_MSE_SAT_SLOT_COUNT2, mask_sh),\ + LE_SF(DP0_DP_MSE_SAT1, DP_MSE_SAT_SLOT_COUNT3, mask_sh),\ + LE_SF(DP0_DP_MSE_SAT_UPDATE, DP_MSE_SAT_UPDATE, mask_sh),\ + LE_SF(DP0_DP_MSE_SAT_UPDATE, DP_MSE_16_MTP_KEEPOUT, mask_sh),\ + LE_SF(DP_AUX0_AUX_CONTROL, AUX_HPD_SEL, mask_sh),\ + LE_SF(DP_AUX0_AUX_CONTROL, AUX_LS_READ_EN, mask_sh),\ + LE_SF(DP_AUX0_AUX_DPHY_RX_CONTROL0, AUX_RX_RECEIVE_WINDOW, mask_sh),\ + LE_SF(HPD0_DC_HPD_CONTROL, DC_HPD_EN, mask_sh) + +#define DCN_LINK_ENCODER_REG_FIELD_LIST(type) \ + type DIG_ENABLE;\ + type DIG_HPD_SELECT;\ + type DIG_MODE;\ + type DIG_FE_SOURCE_SELECT;\ + type DPHY_BYPASS;\ + type DPHY_ATEST_SEL_LANE0;\ + type DPHY_ATEST_SEL_LANE1;\ + type DPHY_ATEST_SEL_LANE2;\ + type DPHY_ATEST_SEL_LANE3;\ + type DPHY_PRBS_EN;\ + type DPHY_PRBS_SEL;\ + type DPHY_SYM1;\ + type DPHY_SYM2;\ + type DPHY_SYM3;\ + type DPHY_SYM4;\ + type DPHY_SYM5;\ + type DPHY_SYM6;\ + type DPHY_SYM7;\ + type DPHY_SYM8;\ + type DPHY_SCRAMBLER_BS_COUNT;\ + type DPHY_SCRAMBLER_ADVANCE;\ + type DPHY_RX_FAST_TRAINING_CAPABLE;\ + type DPHY_LOAD_BS_COUNT;\ + type DPHY_TRAINING_PATTERN_SEL;\ + type DP_DPHY_HBR2_PATTERN_CONTROL;\ + type DP_LINK_TRAINING_COMPLETE;\ + type DP_IDLE_BS_INTERVAL;\ + type DP_VBID_DISABLE;\ + type DP_VID_ENHANCED_FRAME_MODE;\ + type DP_VID_STREAM_ENABLE;\ + type DP_UDI_LANES;\ + type DP_SEC_GSP0_LINE_NUM;\ + type DP_SEC_GSP0_PRIORITY;\ + type DP_MSE_SAT_SRC0;\ + type DP_MSE_SAT_SRC1;\ + type DP_MSE_SAT_SRC2;\ + type DP_MSE_SAT_SRC3;\ + type DP_MSE_SAT_SLOT_COUNT0;\ + type DP_MSE_SAT_SLOT_COUNT1;\ + type DP_MSE_SAT_SLOT_COUNT2;\ + type DP_MSE_SAT_SLOT_COUNT3;\ + type DP_MSE_SAT_UPDATE;\ + type DP_MSE_16_MTP_KEEPOUT;\ + type AUX_HPD_SEL;\ + type AUX_LS_READ_EN;\ + type AUX_RX_RECEIVE_WINDOW;\ + type DC_HPD_EN + +struct dcn10_link_enc_shift { + DCN_LINK_ENCODER_REG_FIELD_LIST(uint8_t); +}; + +struct dcn10_link_enc_mask { + DCN_LINK_ENCODER_REG_FIELD_LIST(uint32_t); +}; + +struct dcn10_link_encoder { + struct link_encoder base; + const struct dcn10_link_enc_registers *link_regs; + const struct dcn10_link_enc_aux_registers *aux_regs; + const struct dcn10_link_enc_hpd_registers *hpd_regs; + const struct dcn10_link_enc_shift *link_shift; + const struct dcn10_link_enc_mask *link_mask; +}; + + +void dcn10_link_encoder_construct( + struct dcn10_link_encoder *enc10, + const struct encoder_init_data *init_data, + const struct encoder_feature_support *enc_features, + const struct dcn10_link_enc_registers *link_regs, + const struct dcn10_link_enc_aux_registers *aux_regs, + const struct dcn10_link_enc_hpd_registers *hpd_regs, + const struct dcn10_link_enc_shift *link_shift, + const struct dcn10_link_enc_mask *link_mask); + +bool dcn10_link_encoder_validate_dvi_output( + const struct dcn10_link_encoder *enc10, + enum signal_type connector_signal, + enum signal_type signal, + const struct dc_crtc_timing *crtc_timing); + +bool dcn10_link_encoder_validate_rgb_output( + const struct dcn10_link_encoder *enc10, + const struct dc_crtc_timing *crtc_timing); + +bool dcn10_link_encoder_validate_dp_output( + const struct dcn10_link_encoder *enc10, + const struct dc_crtc_timing *crtc_timing); + +bool dcn10_link_encoder_validate_wireless_output( + const struct dcn10_link_encoder *enc10, + const struct dc_crtc_timing *crtc_timing); + +bool dcn10_link_encoder_validate_output_with_stream( + struct link_encoder *enc, + const struct dc_stream_state *stream); + +/****************** HW programming ************************/ + +/* initialize HW */ /* why do we initialze aux in here? */ +void dcn10_link_encoder_hw_init(struct link_encoder *enc); + +void dcn10_link_encoder_destroy(struct link_encoder **enc); + +/* program DIG_MODE in DIG_BE */ +/* TODO can this be combined with enable_output? */ +void dcn10_link_encoder_setup( + struct link_encoder *enc, + enum signal_type signal); + +/* enables TMDS PHY output */ +/* TODO: still need depth or just pass in adjusted pixel clock? */ +void dcn10_link_encoder_enable_tmds_output( + struct link_encoder *enc, + enum clock_source_id clock_source, + enum dc_color_depth color_depth, + enum signal_type signal, + uint32_t pixel_clock); + +/* enables DP PHY output */ +void dcn10_link_encoder_enable_dp_output( + struct link_encoder *enc, + const struct dc_link_settings *link_settings, + enum clock_source_id clock_source); + +/* enables DP PHY output in MST mode */ +void dcn10_link_encoder_enable_dp_mst_output( + struct link_encoder *enc, + const struct dc_link_settings *link_settings, + enum clock_source_id clock_source); + +/* disable PHY output */ +void dcn10_link_encoder_disable_output( + struct link_encoder *enc, + enum signal_type signal); + +/* set DP lane settings */ +void dcn10_link_encoder_dp_set_lane_settings( + struct link_encoder *enc, + const struct link_training_settings *link_settings); + +void dcn10_link_encoder_dp_set_phy_pattern( + struct link_encoder *enc, + const struct encoder_set_dp_phy_pattern_param *param); + +/* programs DP MST VC payload allocation */ +void dcn10_link_encoder_update_mst_stream_allocation_table( + struct link_encoder *enc, + const struct link_mst_stream_allocation_table *table); + +void dcn10_link_encoder_connect_dig_be_to_fe( + struct link_encoder *enc, + enum engine_id engine, + bool connect); + +void dcn10_link_encoder_set_dp_phy_pattern_training_pattern( + struct link_encoder *enc, + uint32_t index); + +void dcn10_link_encoder_enable_hpd(struct link_encoder *enc); + +void dcn10_link_encoder_disable_hpd(struct link_encoder *enc); + +void dcn10_psr_program_dp_dphy_fast_training(struct link_encoder *enc, + bool exit_link_training_required); + +void dcn10_psr_program_secondary_packet(struct link_encoder *enc, + unsigned int sdp_transmit_line_num_deadline); + +bool dcn10_is_dig_enabled(struct link_encoder *enc); + +#endif /* __DC_LINK_ENCODER__DCN10_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.c index 179890b1a8c4e..9ca51ae46de74 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.c @@ -65,6 +65,7 @@ static void mpc1_update_blending( int mpcc_id) { struct dcn10_mpc *mpc10 = TO_DCN10_MPC(mpc); + struct mpcc *mpcc = mpc1_get_mpcc(mpc, mpcc_id); REG_UPDATE_5(MPCC_CONTROL[mpcc_id], MPCC_ALPHA_BLND_MODE, blnd_cfg->alpha_mode, @@ -74,6 +75,7 @@ static void mpc1_update_blending( MPCC_GLOBAL_GAIN, blnd_cfg->global_gain); mpc1_set_bg_color(mpc, &blnd_cfg->black_color, mpcc_id); + mpcc->blnd_cfg = *blnd_cfg; } void mpc1_update_stereo_mix( @@ -235,8 +237,7 @@ struct mpcc *mpc1_insert_plane( } /* update the blending configuration */ - new_mpcc->blnd_cfg = *blnd_cfg; - mpc->funcs->update_blending(mpc, &new_mpcc->blnd_cfg, mpcc_id); + mpc->funcs->update_blending(mpc, blnd_cfg, mpcc_id); /* update the stereo mix settings, if provided */ if (sm_cfg != NULL) { @@ -409,7 +410,26 @@ void mpc1_init_mpcc_list_from_hw( } } +void mpc1_read_mpcc_state( + struct mpc *mpc, + int mpcc_inst, + struct mpcc_state *s) +{ + struct dcn10_mpc *mpc10 = TO_DCN10_MPC(mpc); + + REG_GET(MPCC_OPP_ID[mpcc_inst], MPCC_OPP_ID, &s->opp_id); + REG_GET(MPCC_TOP_SEL[mpcc_inst], MPCC_TOP_SEL, &s->dpp_id); + REG_GET(MPCC_BOT_SEL[mpcc_inst], MPCC_BOT_SEL, &s->bot_mpcc_id); + REG_GET_4(MPCC_CONTROL[mpcc_inst], MPCC_MODE, &s->mode, + MPCC_ALPHA_BLND_MODE, &s->alpha_mode, + MPCC_ALPHA_MULTIPLIED_MODE, &s->pre_multiplied_alpha, + MPCC_BLND_ACTIVE_OVERLAP_ONLY, &s->overlap_only); + REG_GET_2(MPCC_STATUS[mpcc_inst], MPCC_IDLE, &s->idle, + MPCC_BUSY, &s->busy); +} + const struct mpc_funcs dcn10_mpc_funcs = { + .read_mpcc_state = mpc1_read_mpcc_state, .insert_plane = mpc1_insert_plane, .remove_mpcc = mpc1_remove_mpcc, .mpc_init = mpc1_mpc_init, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.h index 267a2995ef6e7..d3d16c4cbea38 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_mpc.h @@ -183,4 +183,9 @@ struct mpcc *mpc1_get_mpcc_for_dpp( struct mpc_tree *tree, int dpp_id); +void mpc1_read_mpcc_state( + struct mpc *mpc, + int mpcc_inst, + struct mpcc_state *s); + #endif diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.c index 4bf64d1b2c602..f2fbce0e3fc56 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.c @@ -93,6 +93,81 @@ static void optc1_disable_stereo(struct timing_generator *optc) OTG_3D_STRUCTURE_STEREO_SEL_OVR, 0); } +static uint32_t get_start_vline(struct timing_generator *optc, const struct dc_crtc_timing *dc_crtc_timing) +{ + struct dc_crtc_timing patched_crtc_timing; + int vesa_sync_start; + int asic_blank_end; + int interlace_factor; + int vertical_line_start; + + patched_crtc_timing = *dc_crtc_timing; + optc1_apply_front_porch_workaround(optc, &patched_crtc_timing); + + vesa_sync_start = patched_crtc_timing.h_addressable + + patched_crtc_timing.h_border_right + + patched_crtc_timing.h_front_porch; + + asic_blank_end = patched_crtc_timing.h_total - + vesa_sync_start - + patched_crtc_timing.h_border_left; + + interlace_factor = patched_crtc_timing.flags.INTERLACE ? 2 : 1; + + vesa_sync_start = patched_crtc_timing.v_addressable + + patched_crtc_timing.v_border_bottom + + patched_crtc_timing.v_front_porch; + + asic_blank_end = (patched_crtc_timing.v_total - + vesa_sync_start - + patched_crtc_timing.v_border_top) + * interlace_factor; + + vertical_line_start = asic_blank_end - optc->dlg_otg_param.vstartup_start + 1; + if (vertical_line_start < 0) { + ASSERT(0); + vertical_line_start = 0; + } + + return vertical_line_start; +} + +void optc1_program_vline_interrupt( + struct timing_generator *optc, + const struct dc_crtc_timing *dc_crtc_timing, + unsigned long long vsync_delta) +{ + + struct optc *optc1 = DCN10TG_FROM_TG(optc); + + unsigned long long req_delta_tens_of_usec = div64_u64((vsync_delta + 9999), 10000); + unsigned long long pix_clk_hundreds_khz = div64_u64((dc_crtc_timing->pix_clk_khz + 99), 100); + uint32_t req_delta_lines = (uint32_t) div64_u64( + (req_delta_tens_of_usec * pix_clk_hundreds_khz + dc_crtc_timing->h_total - 1), + dc_crtc_timing->h_total); + + uint32_t vsync_line = get_start_vline(optc, dc_crtc_timing); + uint32_t start_line = 0; + uint32_t endLine = 0; + + if (req_delta_lines != 0) + req_delta_lines--; + + if (req_delta_lines > vsync_line) + start_line = dc_crtc_timing->v_total - (req_delta_lines - vsync_line) - 1; + else + start_line = vsync_line - req_delta_lines; + + endLine = start_line + 2; + + if (endLine >= dc_crtc_timing->v_total) + endLine = 2; + + REG_SET_2(OTG_VERTICAL_INTERRUPT0_POSITION, 0, + OTG_VERTICAL_INTERRUPT0_LINE_START, start_line, + OTG_VERTICAL_INTERRUPT0_LINE_END, endLine); +} + /** * program_timing_generator used by mode timing set * Program CRTC Timing Registers - OTG_H_*, OTG_V_*, Pixel repetition. @@ -285,7 +360,7 @@ void optc1_program_timing( } -static void optc1_set_blank_data_double_buffer(struct timing_generator *optc, bool enable) +void optc1_set_blank_data_double_buffer(struct timing_generator *optc, bool enable) { struct optc *optc1 = DCN10TG_FROM_TG(optc); @@ -780,17 +855,17 @@ void optc1_set_drr( OTG_SET_V_TOTAL_MIN_MASK_EN, 0, OTG_SET_V_TOTAL_MIN_MASK, 0); } else { - REG_SET(OTG_V_TOTAL_MIN, 0, - OTG_V_TOTAL_MIN, 0); - - REG_SET(OTG_V_TOTAL_MAX, 0, - OTG_V_TOTAL_MAX, 0); - REG_UPDATE_4(OTG_V_TOTAL_CONTROL, OTG_SET_V_TOTAL_MIN_MASK, 0, OTG_V_TOTAL_MIN_SEL, 0, OTG_V_TOTAL_MAX_SEL, 0, OTG_FORCE_LOCK_ON_EVENT, 0); + + REG_SET(OTG_V_TOTAL_MIN, 0, + OTG_V_TOTAL_MIN, 0); + + REG_SET(OTG_V_TOTAL_MAX, 0, + OTG_V_TOTAL_MAX, 0); } } @@ -1154,6 +1229,12 @@ void optc1_read_otg_state(struct optc *optc1, REG_GET(OTG_V_TOTAL_MIN, OTG_V_TOTAL_MIN, &s->v_total_min); + REG_GET(OTG_V_TOTAL_CONTROL, + OTG_V_TOTAL_MAX_SEL, &s->v_total_max_sel); + + REG_GET(OTG_V_TOTAL_CONTROL, + OTG_V_TOTAL_MIN_SEL, &s->v_total_min_sel); + REG_GET_2(OTG_V_SYNC_A, OTG_V_SYNC_A_START, &s->v_sync_a_start, OTG_V_SYNC_A_END, &s->v_sync_a_end); @@ -1176,20 +1257,20 @@ void optc1_read_otg_state(struct optc *optc1, OPTC_UNDERFLOW_OCCURRED_STATUS, &s->underflow_occurred_status); } -static void optc1_clear_optc_underflow(struct timing_generator *optc) +void optc1_clear_optc_underflow(struct timing_generator *optc) { struct optc *optc1 = DCN10TG_FROM_TG(optc); REG_UPDATE(OPTC_INPUT_GLOBAL_CONTROL, OPTC_UNDERFLOW_CLEAR, 1); } -static void optc1_tg_init(struct timing_generator *optc) +void optc1_tg_init(struct timing_generator *optc) { optc1_set_blank_data_double_buffer(optc, true); optc1_clear_optc_underflow(optc); } -static bool optc1_is_tg_enabled(struct timing_generator *optc) +bool optc1_is_tg_enabled(struct timing_generator *optc) { struct optc *optc1 = DCN10TG_FROM_TG(optc); uint32_t otg_enabled = 0; @@ -1200,7 +1281,7 @@ static bool optc1_is_tg_enabled(struct timing_generator *optc) } -static bool optc1_is_optc_underflow_occurred(struct timing_generator *optc) +bool optc1_is_optc_underflow_occurred(struct timing_generator *optc) { struct optc *optc1 = DCN10TG_FROM_TG(optc); uint32_t underflow_occurred = 0; @@ -1215,6 +1296,7 @@ static bool optc1_is_optc_underflow_occurred(struct timing_generator *optc) static const struct timing_generator_funcs dcn10_tg_funcs = { .validate_timing = optc1_validate_timing, .program_timing = optc1_program_timing, + .program_vline_interrupt = optc1_program_vline_interrupt, .program_global_sync = optc1_program_global_sync, .enable_crtc = optc1_enable_crtc, .disable_crtc = optc1_disable_crtc, diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.h index d25e7bf0d0d77..c62052f46460e 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.h +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_optc.h @@ -65,6 +65,8 @@ SRI(OTG_NOM_VERT_POSITION, OTG, inst),\ SRI(OTG_BLACK_COLOR, OTG, inst),\ SRI(OTG_CLOCK_CONTROL, OTG, inst),\ + SRI(OTG_VERTICAL_INTERRUPT0_CONTROL, OTG, inst),\ + SRI(OTG_VERTICAL_INTERRUPT0_POSITION, OTG, inst),\ SRI(OTG_VERTICAL_INTERRUPT2_CONTROL, OTG, inst),\ SRI(OTG_VERTICAL_INTERRUPT2_POSITION, OTG, inst),\ SRI(OPTC_INPUT_CLOCK_CONTROL, ODM, inst),\ @@ -124,6 +126,8 @@ struct dcn_optc_registers { uint32_t OTG_TEST_PATTERN_CONTROL; uint32_t OTG_TEST_PATTERN_COLOR; uint32_t OTG_CLOCK_CONTROL; + uint32_t OTG_VERTICAL_INTERRUPT0_CONTROL; + uint32_t OTG_VERTICAL_INTERRUPT0_POSITION; uint32_t OTG_VERTICAL_INTERRUPT2_CONTROL; uint32_t OTG_VERTICAL_INTERRUPT2_POSITION; uint32_t OPTC_INPUT_CLOCK_CONTROL; @@ -206,6 +210,9 @@ struct dcn_optc_registers { SF(OTG0_OTG_CLOCK_CONTROL, OTG_CLOCK_EN, mask_sh),\ SF(OTG0_OTG_CLOCK_CONTROL, OTG_CLOCK_ON, mask_sh),\ SF(OTG0_OTG_CLOCK_CONTROL, OTG_CLOCK_GATE_DIS, mask_sh),\ + SF(OTG0_OTG_VERTICAL_INTERRUPT0_CONTROL, OTG_VERTICAL_INTERRUPT0_INT_ENABLE, mask_sh),\ + SF(OTG0_OTG_VERTICAL_INTERRUPT0_POSITION, OTG_VERTICAL_INTERRUPT0_LINE_START, mask_sh),\ + SF(OTG0_OTG_VERTICAL_INTERRUPT0_POSITION, OTG_VERTICAL_INTERRUPT0_LINE_END, mask_sh),\ SF(OTG0_OTG_VERTICAL_INTERRUPT2_CONTROL, OTG_VERTICAL_INTERRUPT2_INT_ENABLE, mask_sh),\ SF(OTG0_OTG_VERTICAL_INTERRUPT2_POSITION, OTG_VERTICAL_INTERRUPT2_LINE_START, mask_sh),\ SF(ODM0_OPTC_INPUT_CLOCK_CONTROL, OPTC_INPUT_CLK_EN, mask_sh),\ @@ -323,6 +330,9 @@ struct dcn_optc_registers { type OTG_CLOCK_EN;\ type OTG_CLOCK_ON;\ type OTG_CLOCK_GATE_DIS;\ + type OTG_VERTICAL_INTERRUPT0_INT_ENABLE;\ + type OTG_VERTICAL_INTERRUPT0_LINE_START;\ + type OTG_VERTICAL_INTERRUPT0_LINE_END;\ type OTG_VERTICAL_INTERRUPT2_INT_ENABLE;\ type OTG_VERTICAL_INTERRUPT2_LINE_START;\ type OPTC_INPUT_CLK_EN;\ @@ -396,6 +406,8 @@ struct dcn_otg_state { uint32_t v_total; uint32_t v_total_max; uint32_t v_total_min; + uint32_t v_total_min_sel; + uint32_t v_total_max_sel; uint32_t v_sync_a_start; uint32_t v_sync_a_end; uint32_t h_blank_start; @@ -420,6 +432,10 @@ void optc1_program_timing( const struct dc_crtc_timing *dc_crtc_timing, bool use_vbios); +void optc1_program_vline_interrupt(struct timing_generator *optc, + const struct dc_crtc_timing *dc_crtc_timing, + unsigned long long vsync_delta); + void optc1_program_global_sync( struct timing_generator *optc); @@ -481,4 +497,14 @@ void optc1_program_stereo(struct timing_generator *optc, bool optc1_is_stereo_left_eye(struct timing_generator *optc); +void optc1_clear_optc_underflow(struct timing_generator *optc); + +void optc1_tg_init(struct timing_generator *optc); + +bool optc1_is_tg_enabled(struct timing_generator *optc); + +bool optc1_is_optc_underflow_occurred(struct timing_generator *optc); + +void optc1_set_blank_data_double_buffer(struct timing_generator *optc, bool enable); + #endif /* __DC_TIMING_GENERATOR_DCN10_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c index 02bd664aed3ec..df5cb2d1d1645 100644 --- a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_resource.c @@ -38,8 +38,8 @@ #include "dcn10/dcn10_hw_sequencer.h" #include "dce110/dce110_hw_sequencer.h" #include "dcn10/dcn10_opp.h" -#include "dce/dce_link_encoder.h" -#include "dce/dce_stream_encoder.h" +#include "dcn10/dcn10_link_encoder.h" +#include "dcn10/dcn10_stream_encoder.h" #include "dce/dce_clocks.h" #include "dce/dce_clock_source.h" #include "dce/dce_audio.h" @@ -166,36 +166,22 @@ static const struct dce_abm_mask abm_mask = { #define stream_enc_regs(id)\ [id] = {\ - SE_DCN_REG_LIST(id),\ - .TMDS_CNTL = 0,\ - .AFMT_AVI_INFO0 = 0,\ - .AFMT_AVI_INFO1 = 0,\ - .AFMT_AVI_INFO2 = 0,\ - .AFMT_AVI_INFO3 = 0,\ + SE_DCN_REG_LIST(id)\ } -static const struct dce110_stream_enc_registers stream_enc_regs[] = { +static const struct dcn10_stream_enc_registers stream_enc_regs[] = { stream_enc_regs(0), stream_enc_regs(1), stream_enc_regs(2), stream_enc_regs(3), }; -static const struct dce_stream_encoder_shift se_shift = { +static const struct dcn10_stream_encoder_shift se_shift = { SE_COMMON_MASK_SH_LIST_DCN10(__SHIFT) }; -static const struct dce_stream_encoder_mask se_mask = { - SE_COMMON_MASK_SH_LIST_DCN10(_MASK), - .AFMT_GENERIC0_UPDATE = 0, - .AFMT_GENERIC2_UPDATE = 0, - .DP_DYN_RANGE = 0, - .DP_YCBCR_RANGE = 0, - .HDMI_AVI_INFO_SEND = 0, - .HDMI_AVI_INFO_CONT = 0, - .HDMI_AVI_INFO_LINE = 0, - .DP_SEC_AVI_ENABLE = 0, - .AFMT_AVI_INFO_VERSION = 0 +static const struct dcn10_stream_encoder_mask se_mask = { + SE_COMMON_MASK_SH_LIST_DCN10(_MASK) }; #define audio_regs(id)\ @@ -228,13 +214,11 @@ static const struct dce_aduio_mask audio_mask = { AUX_REG_LIST(id)\ } -static const struct dce110_link_enc_aux_registers link_enc_aux_regs[] = { +static const struct dcn10_link_enc_aux_registers link_enc_aux_regs[] = { aux_regs(0), aux_regs(1), aux_regs(2), - aux_regs(3), - aux_regs(4), - aux_regs(5) + aux_regs(3) }; #define hpd_regs(id)\ @@ -242,13 +226,11 @@ static const struct dce110_link_enc_aux_registers link_enc_aux_regs[] = { HPD_REG_LIST(id)\ } -static const struct dce110_link_enc_hpd_registers link_enc_hpd_regs[] = { +static const struct dcn10_link_enc_hpd_registers link_enc_hpd_regs[] = { hpd_regs(0), hpd_regs(1), hpd_regs(2), - hpd_regs(3), - hpd_regs(4), - hpd_regs(5) + hpd_regs(3) }; #define link_regs(id)\ @@ -257,14 +239,19 @@ static const struct dce110_link_enc_hpd_registers link_enc_hpd_regs[] = { SRI(DP_DPHY_INTERNAL_CTRL, DP, id) \ } -static const struct dce110_link_enc_registers link_enc_regs[] = { +static const struct dcn10_link_enc_registers link_enc_regs[] = { link_regs(0), link_regs(1), link_regs(2), - link_regs(3), - link_regs(4), - link_regs(5), - link_regs(6), + link_regs(3) +}; + +static const struct dcn10_link_enc_shift le_shift = { + LINK_ENCODER_MASK_SH_LIST_DCN10(__SHIFT) +}; + +static const struct dcn10_link_enc_mask le_mask = { + LINK_ENCODER_MASK_SH_LIST_DCN10(_MASK) }; #define ipp_regs(id)\ @@ -320,11 +307,14 @@ static const struct dcn_dpp_registers tf_regs[] = { }; static const struct dcn_dpp_shift tf_shift = { - TF_REG_LIST_SH_MASK_DCN10(__SHIFT) + TF_REG_LIST_SH_MASK_DCN10(__SHIFT), + TF_DEBUG_REG_LIST_SH_DCN10 + }; static const struct dcn_dpp_mask tf_mask = { TF_REG_LIST_SH_MASK_DCN10(_MASK), + TF_DEBUG_REG_LIST_MASK_DCN10 }; static const struct dcn_mpc_registers mpc_regs = { @@ -457,6 +447,8 @@ static const struct dc_debug debug_defaults_drv = { .vsr_support = true, .performance_trace = false, .az_endpoint_mute_only = true, + .recovery_enabled = false, /*enable this by default after testing.*/ + .max_downscale_src_width = 3840, }; static const struct dc_debug debug_defaults_diags = { @@ -592,20 +584,22 @@ static const struct encoder_feature_support link_enc_feature = { struct link_encoder *dcn10_link_encoder_create( const struct encoder_init_data *enc_init_data) { - struct dce110_link_encoder *enc110 = - kzalloc(sizeof(struct dce110_link_encoder), GFP_KERNEL); + struct dcn10_link_encoder *enc10 = + kzalloc(sizeof(struct dcn10_link_encoder), GFP_KERNEL); - if (!enc110) + if (!enc10) return NULL; - dce110_link_encoder_construct(enc110, + dcn10_link_encoder_construct(enc10, enc_init_data, &link_enc_feature, &link_enc_regs[enc_init_data->transmitter], &link_enc_aux_regs[enc_init_data->channel - 1], - &link_enc_hpd_regs[enc_init_data->hpd_source]); + &link_enc_hpd_regs[enc_init_data->hpd_source], + &le_shift, + &le_mask); - return &enc110->base; + return &enc10->base; } struct clock_source *dcn10_clock_source_create( @@ -650,16 +644,16 @@ static struct stream_encoder *dcn10_stream_encoder_create( enum engine_id eng_id, struct dc_context *ctx) { - struct dce110_stream_encoder *enc110 = - kzalloc(sizeof(struct dce110_stream_encoder), GFP_KERNEL); + struct dcn10_stream_encoder *enc1 = + kzalloc(sizeof(struct dcn10_stream_encoder), GFP_KERNEL); - if (!enc110) + if (!enc1) return NULL; - dce110_stream_encoder_construct(enc110, ctx, ctx->dc_bios, eng_id, + dcn10_stream_encoder_construct(enc1, ctx, ctx->dc_bios, eng_id, &stream_enc_regs[eng_id], &se_shift, &se_mask); - return &enc110->base; + return &enc1->base; } static const struct dce_hwseq_registers hwseq_reg = { @@ -918,36 +912,6 @@ enum dc_status dcn10_add_stream_to_ctx( return result; } -enum dc_status dcn10_validate_guaranteed( - struct dc *dc, - struct dc_stream_state *dc_stream, - struct dc_state *context) -{ - enum dc_status result = DC_ERROR_UNEXPECTED; - - context->streams[0] = dc_stream; - dc_stream_retain(context->streams[0]); - context->stream_count++; - - result = resource_map_pool_resources(dc, context, dc_stream); - - if (result == DC_OK) - result = resource_map_phy_clock_resources(dc, context, dc_stream); - - if (result == DC_OK) - result = build_mapped_resource(dc, context, dc_stream); - - if (result == DC_OK) { - validate_guaranteed_copy_streams( - context, dc->caps.max_streams); - result = resource_build_scaling_params_for_context(dc, context); - } - if (result == DC_OK && !dcn_validate_bandwidth(dc, context)) - return DC_FAIL_BANDWIDTH_VALIDATE; - - return result; -} - static struct pipe_ctx *dcn10_acquire_idle_pipe_for_layer( struct dc_state *context, const struct resource_pool *pool, @@ -978,235 +942,16 @@ static struct pipe_ctx *dcn10_acquire_idle_pipe_for_layer( return idle_pipe; } -enum dcc_control { - dcc_control__256_256_xxx, - dcc_control__128_128_xxx, - dcc_control__256_64_64, -}; - -enum segment_order { - segment_order__na, - segment_order__contiguous, - segment_order__non_contiguous, -}; - -static bool dcc_support_pixel_format( - enum surface_pixel_format format, - unsigned int *bytes_per_element) -{ - /* DML: get_bytes_per_element */ - switch (format) { - case SURFACE_PIXEL_FORMAT_GRPH_ARGB1555: - case SURFACE_PIXEL_FORMAT_GRPH_RGB565: - *bytes_per_element = 2; - return true; - case SURFACE_PIXEL_FORMAT_GRPH_ARGB8888: - case SURFACE_PIXEL_FORMAT_GRPH_ABGR8888: - case SURFACE_PIXEL_FORMAT_GRPH_ARGB2101010: - case SURFACE_PIXEL_FORMAT_GRPH_ABGR2101010: - *bytes_per_element = 4; - return true; - case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616: - case SURFACE_PIXEL_FORMAT_GRPH_ARGB16161616F: - case SURFACE_PIXEL_FORMAT_GRPH_ABGR16161616F: - *bytes_per_element = 8; - return true; - default: - return false; - } -} - -static bool dcc_support_swizzle( - enum swizzle_mode_values swizzle, - unsigned int bytes_per_element, - enum segment_order *segment_order_horz, - enum segment_order *segment_order_vert) -{ - bool standard_swizzle = false; - bool display_swizzle = false; - - switch (swizzle) { - case DC_SW_4KB_S: - case DC_SW_64KB_S: - case DC_SW_VAR_S: - case DC_SW_4KB_S_X: - case DC_SW_64KB_S_X: - case DC_SW_VAR_S_X: - standard_swizzle = true; - break; - case DC_SW_4KB_D: - case DC_SW_64KB_D: - case DC_SW_VAR_D: - case DC_SW_4KB_D_X: - case DC_SW_64KB_D_X: - case DC_SW_VAR_D_X: - display_swizzle = true; - break; - default: - break; - } - - if (bytes_per_element == 1 && standard_swizzle) { - *segment_order_horz = segment_order__contiguous; - *segment_order_vert = segment_order__na; - return true; - } - if (bytes_per_element == 2 && standard_swizzle) { - *segment_order_horz = segment_order__non_contiguous; - *segment_order_vert = segment_order__contiguous; - return true; - } - if (bytes_per_element == 4 && standard_swizzle) { - *segment_order_horz = segment_order__non_contiguous; - *segment_order_vert = segment_order__contiguous; - return true; - } - if (bytes_per_element == 8 && standard_swizzle) { - *segment_order_horz = segment_order__na; - *segment_order_vert = segment_order__contiguous; - return true; - } - if (bytes_per_element == 8 && display_swizzle) { - *segment_order_horz = segment_order__contiguous; - *segment_order_vert = segment_order__non_contiguous; - return true; - } - - return false; -} - -static void get_blk256_size(unsigned int *blk256_width, unsigned int *blk256_height, - unsigned int bytes_per_element) -{ - /* copied from DML. might want to refactor DML to leverage from DML */ - /* DML : get_blk256_size */ - if (bytes_per_element == 1) { - *blk256_width = 16; - *blk256_height = 16; - } else if (bytes_per_element == 2) { - *blk256_width = 16; - *blk256_height = 8; - } else if (bytes_per_element == 4) { - *blk256_width = 8; - *blk256_height = 8; - } else if (bytes_per_element == 8) { - *blk256_width = 8; - *blk256_height = 4; - } -} - -static void det_request_size( - unsigned int height, - unsigned int width, - unsigned int bpe, - bool *req128_horz_wc, - bool *req128_vert_wc) -{ - unsigned int detile_buf_size = 164 * 1024; /* 164KB for DCN1.0 */ - - unsigned int blk256_height = 0; - unsigned int blk256_width = 0; - unsigned int swath_bytes_horz_wc, swath_bytes_vert_wc; - - get_blk256_size(&blk256_width, &blk256_height, bpe); - - swath_bytes_horz_wc = height * blk256_height * bpe; - swath_bytes_vert_wc = width * blk256_width * bpe; - - *req128_horz_wc = (2 * swath_bytes_horz_wc <= detile_buf_size) ? - false : /* full 256B request */ - true; /* half 128b request */ - - *req128_vert_wc = (2 * swath_bytes_vert_wc <= detile_buf_size) ? - false : /* full 256B request */ - true; /* half 128b request */ -} - -static bool get_dcc_compression_cap(const struct dc *dc, +static bool dcn10_get_dcc_compression_cap(const struct dc *dc, const struct dc_dcc_surface_param *input, struct dc_surface_dcc_cap *output) { - /* implement section 1.6.2.1 of DCN1_Programming_Guide.docx */ - enum dcc_control dcc_control; - unsigned int bpe; - enum segment_order segment_order_horz, segment_order_vert; - bool req128_horz_wc, req128_vert_wc; - - memset(output, 0, sizeof(*output)); - - if (dc->debug.disable_dcc == DCC_DISABLE) - return false; - - if (!dcc_support_pixel_format(input->format, - &bpe)) - return false; - - if (!dcc_support_swizzle(input->swizzle_mode, bpe, - &segment_order_horz, &segment_order_vert)) - return false; - - det_request_size(input->surface_size.height, input->surface_size.width, - bpe, &req128_horz_wc, &req128_vert_wc); - - if (!req128_horz_wc && !req128_vert_wc) { - dcc_control = dcc_control__256_256_xxx; - } else if (input->scan == SCAN_DIRECTION_HORIZONTAL) { - if (!req128_horz_wc) - dcc_control = dcc_control__256_256_xxx; - else if (segment_order_horz == segment_order__contiguous) - dcc_control = dcc_control__128_128_xxx; - else - dcc_control = dcc_control__256_64_64; - } else if (input->scan == SCAN_DIRECTION_VERTICAL) { - if (!req128_vert_wc) - dcc_control = dcc_control__256_256_xxx; - else if (segment_order_vert == segment_order__contiguous) - dcc_control = dcc_control__128_128_xxx; - else - dcc_control = dcc_control__256_64_64; - } else { - if ((req128_horz_wc && - segment_order_horz == segment_order__non_contiguous) || - (req128_vert_wc && - segment_order_vert == segment_order__non_contiguous)) - /* access_dir not known, must use most constraining */ - dcc_control = dcc_control__256_64_64; - else - /* reg128 is true for either horz and vert - * but segment_order is contiguous - */ - dcc_control = dcc_control__128_128_xxx; - } - - if (dc->debug.disable_dcc == DCC_HALF_REQ_DISALBE && - dcc_control != dcc_control__256_256_xxx) - return false; - - switch (dcc_control) { - case dcc_control__256_256_xxx: - output->grph.rgb.max_uncompressed_blk_size = 256; - output->grph.rgb.max_compressed_blk_size = 256; - output->grph.rgb.independent_64b_blks = false; - break; - case dcc_control__128_128_xxx: - output->grph.rgb.max_uncompressed_blk_size = 128; - output->grph.rgb.max_compressed_blk_size = 128; - output->grph.rgb.independent_64b_blks = false; - break; - case dcc_control__256_64_64: - output->grph.rgb.max_uncompressed_blk_size = 256; - output->grph.rgb.max_compressed_blk_size = 64; - output->grph.rgb.independent_64b_blks = true; - break; - } - - output->capable = true; - output->const_color_support = false; - - return true; + return dc->res_pool->hubbub->funcs->get_dcc_compression_cap( + dc->res_pool->hubbub, + input, + output); } - static void dcn10_destroy_resource_pool(struct resource_pool **pool) { struct dcn10_resource_pool *dcn10_pool = TO_DCN10_RES_POOL(*pool); @@ -1227,13 +972,12 @@ static enum dc_status dcn10_validate_plane(const struct dc_plane_state *plane_st } static struct dc_cap_funcs cap_funcs = { - .get_dcc_compression_cap = get_dcc_compression_cap + .get_dcc_compression_cap = dcn10_get_dcc_compression_cap }; static struct resource_funcs dcn10_res_pool_funcs = { .destroy = dcn10_destroy_resource_pool, .link_enc_create = dcn10_link_encoder_create, - .validate_guaranteed = dcn10_validate_guaranteed, .validate_bandwidth = dcn_validate_bandwidth, .acquire_idle_pipe_for_layer = dcn10_acquire_idle_pipe_for_layer, .validate_plane = dcn10_validate_plane, @@ -1282,6 +1026,7 @@ static bool construct( dc->caps.max_cursor_size = 256; dc->caps.max_slave_planes = 1; dc->caps.is_apu = true; + dc->caps.post_blend_color_processing = false; if (dc->ctx->dce_environment == DCE_ENV_PRODUCTION_DRV) dc->debug = debug_defaults_drv; diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c new file mode 100644 index 0000000000000..653b7b2efe2e4 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.c @@ -0,0 +1,1490 @@ +/* + * Copyright 2012-15 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + + +#include "dc_bios_types.h" +#include "dcn10_stream_encoder.h" +#include "reg_helper.h" +#include "hw_shared.h" + +#define DC_LOGGER \ + enc1->base.ctx->logger + + +#define REG(reg)\ + (enc1->regs->reg) + +#undef FN +#define FN(reg_name, field_name) \ + enc1->se_shift->field_name, enc1->se_mask->field_name + +#define VBI_LINE_0 0 +#define DP_BLANK_MAX_RETRY 20 +#define HDMI_CLOCK_CHANNEL_RATE_MORE_340M 340000 + + +enum { + DP_MST_UPDATE_MAX_RETRY = 50 +}; + +#define CTX \ + enc1->base.ctx + +void enc1_update_generic_info_packet( + struct dcn10_stream_encoder *enc1, + uint32_t packet_index, + const struct dc_info_packet *info_packet) +{ + uint32_t regval; + /* TODOFPGA Figure out a proper number for max_retries polling for lock + * use 50 for now. + */ + uint32_t max_retries = 50; + + /*we need turn on clock before programming AFMT block*/ + REG_UPDATE(AFMT_CNTL, AFMT_AUDIO_CLOCK_EN, 1); + + if (packet_index >= 8) + ASSERT(0); + + /* poll dig_update_lock is not locked -> asic internal signal + * assume otg master lock will unlock it + */ +/* REG_WAIT(AFMT_VBI_PACKET_CONTROL, AFMT_GENERIC_LOCK_STATUS, + 0, 10, max_retries);*/ + + /* check if HW reading GSP memory */ + REG_WAIT(AFMT_VBI_PACKET_CONTROL, AFMT_GENERIC_CONFLICT, + 0, 10, max_retries); + + /* HW does is not reading GSP memory not reading too long -> + * something wrong. clear GPS memory access and notify? + * hw SW is writing to GSP memory + */ + REG_UPDATE(AFMT_VBI_PACKET_CONTROL, AFMT_GENERIC_CONFLICT_CLR, 1); + + /* choose which generic packet to use */ + regval = REG_READ(AFMT_VBI_PACKET_CONTROL); + REG_UPDATE(AFMT_VBI_PACKET_CONTROL, + AFMT_GENERIC_INDEX, packet_index); + + /* write generic packet header + * (4th byte is for GENERIC0 only) + */ + REG_SET_4(AFMT_GENERIC_HDR, 0, + AFMT_GENERIC_HB0, info_packet->hb0, + AFMT_GENERIC_HB1, info_packet->hb1, + AFMT_GENERIC_HB2, info_packet->hb2, + AFMT_GENERIC_HB3, info_packet->hb3); + + /* write generic packet contents + * (we never use last 4 bytes) + * there are 8 (0-7) mmDIG0_AFMT_GENERIC0_x registers + */ + { + const uint32_t *content = + (const uint32_t *) &info_packet->sb[0]; + + REG_WRITE(AFMT_GENERIC_0, *content++); + REG_WRITE(AFMT_GENERIC_1, *content++); + REG_WRITE(AFMT_GENERIC_2, *content++); + REG_WRITE(AFMT_GENERIC_3, *content++); + REG_WRITE(AFMT_GENERIC_4, *content++); + REG_WRITE(AFMT_GENERIC_5, *content++); + REG_WRITE(AFMT_GENERIC_6, *content++); + REG_WRITE(AFMT_GENERIC_7, *content); + } + + switch (packet_index) { + case 0: + REG_UPDATE(AFMT_VBI_PACKET_CONTROL1, + AFMT_GENERIC0_FRAME_UPDATE, 1); + break; + case 1: + REG_UPDATE(AFMT_VBI_PACKET_CONTROL1, + AFMT_GENERIC1_FRAME_UPDATE, 1); + break; + case 2: + REG_UPDATE(AFMT_VBI_PACKET_CONTROL1, + AFMT_GENERIC2_FRAME_UPDATE, 1); + break; + case 3: + REG_UPDATE(AFMT_VBI_PACKET_CONTROL1, + AFMT_GENERIC3_FRAME_UPDATE, 1); + break; + case 4: + REG_UPDATE(AFMT_VBI_PACKET_CONTROL1, + AFMT_GENERIC4_FRAME_UPDATE, 1); + break; + case 5: + REG_UPDATE(AFMT_VBI_PACKET_CONTROL1, + AFMT_GENERIC5_FRAME_UPDATE, 1); + break; + case 6: + REG_UPDATE(AFMT_VBI_PACKET_CONTROL1, + AFMT_GENERIC6_FRAME_UPDATE, 1); + break; + case 7: + REG_UPDATE(AFMT_VBI_PACKET_CONTROL1, + AFMT_GENERIC7_FRAME_UPDATE, 1); + break; + default: + break; + } +} + +static void enc1_update_hdmi_info_packet( + struct dcn10_stream_encoder *enc1, + uint32_t packet_index, + const struct dc_info_packet *info_packet) +{ + uint32_t cont, send, line; + + if (info_packet->valid) { + enc1_update_generic_info_packet( + enc1, + packet_index, + info_packet); + + /* enable transmission of packet(s) - + * packet transmission begins on the next frame + */ + cont = 1; + /* send packet(s) every frame */ + send = 1; + /* select line number to send packets on */ + line = 2; + } else { + cont = 0; + send = 0; + line = 0; + } + + /* choose which generic packet control to use */ + switch (packet_index) { + case 0: + REG_UPDATE_3(HDMI_GENERIC_PACKET_CONTROL0, + HDMI_GENERIC0_CONT, cont, + HDMI_GENERIC0_SEND, send, + HDMI_GENERIC0_LINE, line); + break; + case 1: + REG_UPDATE_3(HDMI_GENERIC_PACKET_CONTROL0, + HDMI_GENERIC1_CONT, cont, + HDMI_GENERIC1_SEND, send, + HDMI_GENERIC1_LINE, line); + break; + case 2: + REG_UPDATE_3(HDMI_GENERIC_PACKET_CONTROL1, + HDMI_GENERIC0_CONT, cont, + HDMI_GENERIC0_SEND, send, + HDMI_GENERIC0_LINE, line); + break; + case 3: + REG_UPDATE_3(HDMI_GENERIC_PACKET_CONTROL1, + HDMI_GENERIC1_CONT, cont, + HDMI_GENERIC1_SEND, send, + HDMI_GENERIC1_LINE, line); + break; + case 4: + REG_UPDATE_3(HDMI_GENERIC_PACKET_CONTROL2, + HDMI_GENERIC0_CONT, cont, + HDMI_GENERIC0_SEND, send, + HDMI_GENERIC0_LINE, line); + break; + case 5: + REG_UPDATE_3(HDMI_GENERIC_PACKET_CONTROL2, + HDMI_GENERIC1_CONT, cont, + HDMI_GENERIC1_SEND, send, + HDMI_GENERIC1_LINE, line); + break; + case 6: + REG_UPDATE_3(HDMI_GENERIC_PACKET_CONTROL3, + HDMI_GENERIC0_CONT, cont, + HDMI_GENERIC0_SEND, send, + HDMI_GENERIC0_LINE, line); + break; + case 7: + REG_UPDATE_3(HDMI_GENERIC_PACKET_CONTROL3, + HDMI_GENERIC1_CONT, cont, + HDMI_GENERIC1_SEND, send, + HDMI_GENERIC1_LINE, line); + break; + default: + /* invalid HW packet index */ + DC_LOG_WARNING( + "Invalid HW packet index: %s()\n", + __func__); + return; + } +} + +/* setup stream encoder in dp mode */ +void enc1_stream_encoder_dp_set_stream_attribute( + struct stream_encoder *enc, + struct dc_crtc_timing *crtc_timing, + enum dc_color_space output_color_space) +{ + uint32_t h_active_start; + uint32_t v_active_start; + uint32_t misc0 = 0; + uint32_t misc1 = 0; + uint32_t h_blank; + uint32_t h_back_porch; + uint8_t synchronous_clock = 0; /* asynchronous mode */ + uint8_t colorimetry_bpc; + uint8_t dynamic_range_rgb = 0; /*full range*/ + uint8_t dynamic_range_ycbcr = 1; /*bt709*/ + + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + REG_UPDATE(DP_DB_CNTL, DP_DB_DISABLE, 1); + + /* set pixel encoding */ + switch (crtc_timing->pixel_encoding) { + case PIXEL_ENCODING_YCBCR422: + REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, + DP_PIXEL_ENCODING_TYPE_YCBCR422); + break; + case PIXEL_ENCODING_YCBCR444: + REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, + DP_PIXEL_ENCODING_TYPE_YCBCR444); + + if (crtc_timing->flags.Y_ONLY) + if (crtc_timing->display_color_depth != COLOR_DEPTH_666) + /* HW testing only, no use case yet. + * Color depth of Y-only could be + * 8, 10, 12, 16 bits + */ + REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, + DP_PIXEL_ENCODING_TYPE_Y_ONLY); + /* Note: DP_MSA_MISC1 bit 7 is the indicator + * of Y-only mode. + * This bit is set in HW if register + * DP_PIXEL_ENCODING is programmed to 0x4 + */ + break; + case PIXEL_ENCODING_YCBCR420: + REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, + DP_PIXEL_ENCODING_TYPE_YCBCR420); + REG_UPDATE(DP_VID_TIMING, DP_VID_N_MUL, 1); + break; + default: + REG_UPDATE(DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, + DP_PIXEL_ENCODING_TYPE_RGB444); + break; + } + + misc1 = REG_READ(DP_MSA_MISC); + + /* set color depth */ + + switch (crtc_timing->display_color_depth) { + case COLOR_DEPTH_666: + REG_UPDATE(DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, + 0); + break; + case COLOR_DEPTH_888: + REG_UPDATE(DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, + DP_COMPONENT_PIXEL_DEPTH_8BPC); + break; + case COLOR_DEPTH_101010: + REG_UPDATE(DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, + DP_COMPONENT_PIXEL_DEPTH_10BPC); + + break; + case COLOR_DEPTH_121212: + REG_UPDATE(DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, + DP_COMPONENT_PIXEL_DEPTH_12BPC); + break; + default: + REG_UPDATE(DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, + DP_COMPONENT_PIXEL_DEPTH_6BPC); + break; + } + + /* set dynamic range and YCbCr range */ + + switch (crtc_timing->display_color_depth) { + case COLOR_DEPTH_666: + colorimetry_bpc = 0; + break; + case COLOR_DEPTH_888: + colorimetry_bpc = 1; + break; + case COLOR_DEPTH_101010: + colorimetry_bpc = 2; + break; + case COLOR_DEPTH_121212: + colorimetry_bpc = 3; + break; + default: + colorimetry_bpc = 0; + break; + } + + misc0 = misc0 | synchronous_clock; + misc0 = colorimetry_bpc << 5; + + switch (output_color_space) { + case COLOR_SPACE_SRGB: + misc0 = misc0 | 0x0; + misc1 = misc1 & ~0x80; /* bit7 = 0*/ + dynamic_range_rgb = 0; /*full range*/ + break; + case COLOR_SPACE_SRGB_LIMITED: + misc0 = misc0 | 0x8; /* bit3=1 */ + misc1 = misc1 & ~0x80; /* bit7 = 0*/ + dynamic_range_rgb = 1; /*limited range*/ + break; + case COLOR_SPACE_YCBCR601: + case COLOR_SPACE_YCBCR601_LIMITED: + misc0 = misc0 | 0x8; /* bit3=1, bit4=0 */ + misc1 = misc1 & ~0x80; /* bit7 = 0*/ + dynamic_range_ycbcr = 0; /*bt601*/ + if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR422) + misc0 = misc0 | 0x2; /* bit2=0, bit1=1 */ + else if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR444) + misc0 = misc0 | 0x4; /* bit2=1, bit1=0 */ + break; + case COLOR_SPACE_YCBCR709: + case COLOR_SPACE_YCBCR709_LIMITED: + misc0 = misc0 | 0x18; /* bit3=1, bit4=1 */ + misc1 = misc1 & ~0x80; /* bit7 = 0*/ + dynamic_range_ycbcr = 1; /*bt709*/ + if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR422) + misc0 = misc0 | 0x2; /* bit2=0, bit1=1 */ + else if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR444) + misc0 = misc0 | 0x4; /* bit2=1, bit1=0 */ + break; + case COLOR_SPACE_2020_RGB_LIMITEDRANGE: + dynamic_range_rgb = 1; /*limited range*/ + break; + case COLOR_SPACE_2020_RGB_FULLRANGE: + case COLOR_SPACE_2020_YCBCR: + case COLOR_SPACE_XR_RGB: + case COLOR_SPACE_MSREF_SCRGB: + case COLOR_SPACE_ADOBERGB: + case COLOR_SPACE_DCIP3: + case COLOR_SPACE_XV_YCC_709: + case COLOR_SPACE_XV_YCC_601: + case COLOR_SPACE_DISPLAYNATIVE: + case COLOR_SPACE_DOLBYVISION: + case COLOR_SPACE_APPCTRL: + case COLOR_SPACE_CUSTOMPOINTS: + case COLOR_SPACE_UNKNOWN: + /* do nothing */ + break; + } + + REG_SET(DP_MSA_COLORIMETRY, 0, DP_MSA_MISC0, misc0); + REG_WRITE(DP_MSA_MISC, misc1); /* MSA_MISC1 */ + + /* dcn new register + * dc_crtc_timing is vesa dmt struct. data from edid + */ + REG_SET_2(DP_MSA_TIMING_PARAM1, 0, + DP_MSA_HTOTAL, crtc_timing->h_total, + DP_MSA_VTOTAL, crtc_timing->v_total); + + /* calculate from vesa timing parameters + * h_active_start related to leading edge of sync + */ + + h_blank = crtc_timing->h_total - crtc_timing->h_border_left - + crtc_timing->h_addressable - crtc_timing->h_border_right; + + h_back_porch = h_blank - crtc_timing->h_front_porch - + crtc_timing->h_sync_width; + + /* start at beginning of left border */ + h_active_start = crtc_timing->h_sync_width + h_back_porch; + + + v_active_start = crtc_timing->v_total - crtc_timing->v_border_top - + crtc_timing->v_addressable - crtc_timing->v_border_bottom - + crtc_timing->v_front_porch; + + + /* start at beginning of left border */ + REG_SET_2(DP_MSA_TIMING_PARAM2, 0, + DP_MSA_HSTART, h_active_start, + DP_MSA_VSTART, v_active_start); + + REG_SET_4(DP_MSA_TIMING_PARAM3, 0, + DP_MSA_HSYNCWIDTH, + crtc_timing->h_sync_width, + DP_MSA_HSYNCPOLARITY, + !crtc_timing->flags.HSYNC_POSITIVE_POLARITY, + DP_MSA_VSYNCWIDTH, + crtc_timing->v_sync_width, + DP_MSA_VSYNCPOLARITY, + !crtc_timing->flags.VSYNC_POSITIVE_POLARITY); + + /* HWDITH include border or overscan */ + REG_SET_2(DP_MSA_TIMING_PARAM4, 0, + DP_MSA_HWIDTH, crtc_timing->h_border_left + + crtc_timing->h_addressable + crtc_timing->h_border_right, + DP_MSA_VHEIGHT, crtc_timing->v_border_top + + crtc_timing->v_addressable + crtc_timing->v_border_bottom); +} + +static void enc1_stream_encoder_set_stream_attribute_helper( + struct dcn10_stream_encoder *enc1, + struct dc_crtc_timing *crtc_timing) +{ + switch (crtc_timing->pixel_encoding) { + case PIXEL_ENCODING_YCBCR422: + REG_UPDATE(DIG_FE_CNTL, TMDS_PIXEL_ENCODING, 1); + break; + default: + REG_UPDATE(DIG_FE_CNTL, TMDS_PIXEL_ENCODING, 0); + break; + } + REG_UPDATE(DIG_FE_CNTL, TMDS_COLOR_FORMAT, 0); +} + +/* setup stream encoder in hdmi mode */ +void enc1_stream_encoder_hdmi_set_stream_attribute( + struct stream_encoder *enc, + struct dc_crtc_timing *crtc_timing, + int actual_pix_clk_khz, + bool enable_audio) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + struct bp_encoder_control cntl = {0}; + + cntl.action = ENCODER_CONTROL_SETUP; + cntl.engine_id = enc1->base.id; + cntl.signal = SIGNAL_TYPE_HDMI_TYPE_A; + cntl.enable_dp_audio = enable_audio; + cntl.pixel_clock = actual_pix_clk_khz; + cntl.lanes_number = LANE_COUNT_FOUR; + + if (enc1->base.bp->funcs->encoder_control( + enc1->base.bp, &cntl) != BP_RESULT_OK) + return; + + enc1_stream_encoder_set_stream_attribute_helper(enc1, crtc_timing); + + /* setup HDMI engine */ + REG_UPDATE_5(HDMI_CONTROL, + HDMI_PACKET_GEN_VERSION, 1, + HDMI_KEEPOUT_MODE, 1, + HDMI_DEEP_COLOR_ENABLE, 0, + HDMI_DATA_SCRAMBLE_EN, 0, + HDMI_CLOCK_CHANNEL_RATE, 0); + + + switch (crtc_timing->display_color_depth) { + case COLOR_DEPTH_888: + REG_UPDATE(HDMI_CONTROL, HDMI_DEEP_COLOR_DEPTH, 0); + break; + case COLOR_DEPTH_101010: + if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR422) { + REG_UPDATE_2(HDMI_CONTROL, + HDMI_DEEP_COLOR_DEPTH, 1, + HDMI_DEEP_COLOR_ENABLE, 0); + } else { + REG_UPDATE_2(HDMI_CONTROL, + HDMI_DEEP_COLOR_DEPTH, 1, + HDMI_DEEP_COLOR_ENABLE, 1); + } + break; + case COLOR_DEPTH_121212: + if (crtc_timing->pixel_encoding == PIXEL_ENCODING_YCBCR422) { + REG_UPDATE_2(HDMI_CONTROL, + HDMI_DEEP_COLOR_DEPTH, 2, + HDMI_DEEP_COLOR_ENABLE, 0); + } else { + REG_UPDATE_2(HDMI_CONTROL, + HDMI_DEEP_COLOR_DEPTH, 2, + HDMI_DEEP_COLOR_ENABLE, 1); + } + break; + case COLOR_DEPTH_161616: + REG_UPDATE_2(HDMI_CONTROL, + HDMI_DEEP_COLOR_DEPTH, 3, + HDMI_DEEP_COLOR_ENABLE, 1); + break; + default: + break; + } + + if (actual_pix_clk_khz >= HDMI_CLOCK_CHANNEL_RATE_MORE_340M) { + /* enable HDMI data scrambler + * HDMI_CLOCK_CHANNEL_RATE_MORE_340M + * Clock channel frequency is 1/4 of character rate. + */ + REG_UPDATE_2(HDMI_CONTROL, + HDMI_DATA_SCRAMBLE_EN, 1, + HDMI_CLOCK_CHANNEL_RATE, 1); + } else if (crtc_timing->flags.LTE_340MCSC_SCRAMBLE) { + + /* TODO: New feature for DCE11, still need to implement */ + + /* enable HDMI data scrambler + * HDMI_CLOCK_CHANNEL_FREQ_EQUAL_TO_CHAR_RATE + * Clock channel frequency is the same + * as character rate + */ + REG_UPDATE_2(HDMI_CONTROL, + HDMI_DATA_SCRAMBLE_EN, 1, + HDMI_CLOCK_CHANNEL_RATE, 0); + } + + + REG_UPDATE_3(HDMI_VBI_PACKET_CONTROL, + HDMI_GC_CONT, 1, + HDMI_GC_SEND, 1, + HDMI_NULL_SEND, 1); + + /* following belongs to audio */ + REG_UPDATE(HDMI_INFOFRAME_CONTROL0, HDMI_AUDIO_INFO_SEND, 1); + + REG_UPDATE(AFMT_INFOFRAME_CONTROL0, AFMT_AUDIO_INFO_UPDATE, 1); + + REG_UPDATE(HDMI_INFOFRAME_CONTROL1, HDMI_AUDIO_INFO_LINE, + VBI_LINE_0 + 2); + + REG_UPDATE(HDMI_GC, HDMI_GC_AVMUTE, 0); +} + +/* setup stream encoder in dvi mode */ +void enc1_stream_encoder_dvi_set_stream_attribute( + struct stream_encoder *enc, + struct dc_crtc_timing *crtc_timing, + bool is_dual_link) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + struct bp_encoder_control cntl = {0}; + + cntl.action = ENCODER_CONTROL_SETUP; + cntl.engine_id = enc1->base.id; + cntl.signal = is_dual_link ? + SIGNAL_TYPE_DVI_DUAL_LINK : SIGNAL_TYPE_DVI_SINGLE_LINK; + cntl.enable_dp_audio = false; + cntl.pixel_clock = crtc_timing->pix_clk_khz; + cntl.lanes_number = (is_dual_link) ? LANE_COUNT_EIGHT : LANE_COUNT_FOUR; + + if (enc1->base.bp->funcs->encoder_control( + enc1->base.bp, &cntl) != BP_RESULT_OK) + return; + + ASSERT(crtc_timing->pixel_encoding == PIXEL_ENCODING_RGB); + ASSERT(crtc_timing->display_color_depth == COLOR_DEPTH_888); + enc1_stream_encoder_set_stream_attribute_helper(enc1, crtc_timing); +} + +void enc1_stream_encoder_set_mst_bandwidth( + struct stream_encoder *enc, + struct fixed31_32 avg_time_slots_per_mtp) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + uint32_t x = dc_fixpt_floor( + avg_time_slots_per_mtp); + uint32_t y = dc_fixpt_ceil( + dc_fixpt_shl( + dc_fixpt_sub_int( + avg_time_slots_per_mtp, + x), + 26)); + + REG_SET_2(DP_MSE_RATE_CNTL, 0, + DP_MSE_RATE_X, x, + DP_MSE_RATE_Y, y); + + /* wait for update to be completed on the link */ + /* i.e. DP_MSE_RATE_UPDATE_PENDING field (read only) */ + /* is reset to 0 (not pending) */ + REG_WAIT(DP_MSE_RATE_UPDATE, DP_MSE_RATE_UPDATE_PENDING, + 0, + 10, DP_MST_UPDATE_MAX_RETRY); +} + +static void enc1_stream_encoder_update_hdmi_info_packets( + struct stream_encoder *enc, + const struct encoder_info_frame *info_frame) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + /* for bring up, disable dp double TODO */ + REG_UPDATE(HDMI_DB_CONTROL, HDMI_DB_DISABLE, 1); + + enc1_update_hdmi_info_packet(enc1, 0, &info_frame->avi); + enc1_update_hdmi_info_packet(enc1, 1, &info_frame->vendor); + enc1_update_hdmi_info_packet(enc1, 2, &info_frame->gamut); + enc1_update_hdmi_info_packet(enc1, 3, &info_frame->spd); + enc1_update_hdmi_info_packet(enc1, 4, &info_frame->hdrsmd); +} + +static void enc1_stream_encoder_stop_hdmi_info_packets( + struct stream_encoder *enc) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + /* stop generic packets 0 & 1 on HDMI */ + REG_SET_6(HDMI_GENERIC_PACKET_CONTROL0, 0, + HDMI_GENERIC1_CONT, 0, + HDMI_GENERIC1_LINE, 0, + HDMI_GENERIC1_SEND, 0, + HDMI_GENERIC0_CONT, 0, + HDMI_GENERIC0_LINE, 0, + HDMI_GENERIC0_SEND, 0); + + /* stop generic packets 2 & 3 on HDMI */ + REG_SET_6(HDMI_GENERIC_PACKET_CONTROL1, 0, + HDMI_GENERIC0_CONT, 0, + HDMI_GENERIC0_LINE, 0, + HDMI_GENERIC0_SEND, 0, + HDMI_GENERIC1_CONT, 0, + HDMI_GENERIC1_LINE, 0, + HDMI_GENERIC1_SEND, 0); + + /* stop generic packets 2 & 3 on HDMI */ + REG_SET_6(HDMI_GENERIC_PACKET_CONTROL2, 0, + HDMI_GENERIC0_CONT, 0, + HDMI_GENERIC0_LINE, 0, + HDMI_GENERIC0_SEND, 0, + HDMI_GENERIC1_CONT, 0, + HDMI_GENERIC1_LINE, 0, + HDMI_GENERIC1_SEND, 0); + + REG_SET_6(HDMI_GENERIC_PACKET_CONTROL3, 0, + HDMI_GENERIC0_CONT, 0, + HDMI_GENERIC0_LINE, 0, + HDMI_GENERIC0_SEND, 0, + HDMI_GENERIC1_CONT, 0, + HDMI_GENERIC1_LINE, 0, + HDMI_GENERIC1_SEND, 0); +} + +void enc1_stream_encoder_update_dp_info_packets( + struct stream_encoder *enc, + const struct encoder_info_frame *info_frame) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + uint32_t value = 0; + + if (info_frame->vsc.valid) + enc1_update_generic_info_packet( + enc1, + 0, /* packetIndex */ + &info_frame->vsc); + + if (info_frame->spd.valid) + enc1_update_generic_info_packet( + enc1, + 2, /* packetIndex */ + &info_frame->spd); + + if (info_frame->hdrsmd.valid) + enc1_update_generic_info_packet( + enc1, + 3, /* packetIndex */ + &info_frame->hdrsmd); + + /* enable/disable transmission of packet(s). + * If enabled, packet transmission begins on the next frame + */ + REG_UPDATE(DP_SEC_CNTL, DP_SEC_GSP0_ENABLE, info_frame->vsc.valid); + REG_UPDATE(DP_SEC_CNTL, DP_SEC_GSP2_ENABLE, info_frame->spd.valid); + REG_UPDATE(DP_SEC_CNTL, DP_SEC_GSP3_ENABLE, info_frame->hdrsmd.valid); + + + /* This bit is the master enable bit. + * When enabling secondary stream engine, + * this master bit must also be set. + * This register shared with audio info frame. + * Therefore we need to enable master bit + * if at least on of the fields is not 0 + */ + value = REG_READ(DP_SEC_CNTL); + if (value) + REG_UPDATE(DP_SEC_CNTL, DP_SEC_STREAM_ENABLE, 1); +} + +void enc1_stream_encoder_stop_dp_info_packets( + struct stream_encoder *enc) +{ + /* stop generic packets on DP */ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + uint32_t value = 0; + + REG_SET_10(DP_SEC_CNTL, 0, + DP_SEC_GSP0_ENABLE, 0, + DP_SEC_GSP1_ENABLE, 0, + DP_SEC_GSP2_ENABLE, 0, + DP_SEC_GSP3_ENABLE, 0, + DP_SEC_GSP4_ENABLE, 0, + DP_SEC_GSP5_ENABLE, 0, + DP_SEC_GSP6_ENABLE, 0, + DP_SEC_GSP7_ENABLE, 0, + DP_SEC_MPG_ENABLE, 0, + DP_SEC_STREAM_ENABLE, 0); + + /* this register shared with audio info frame. + * therefore we need to keep master enabled + * if at least one of the fields is not 0 */ + value = REG_READ(DP_SEC_CNTL); + if (value) + REG_UPDATE(DP_SEC_CNTL, DP_SEC_STREAM_ENABLE, 1); + +} + +void enc1_stream_encoder_dp_blank( + struct stream_encoder *enc) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + uint32_t retries = 0; + uint32_t reg1 = 0; + uint32_t max_retries = DP_BLANK_MAX_RETRY * 10; + + /* Note: For CZ, we are changing driver default to disable + * stream deferred to next VBLANK. If results are positive, we + * will make the same change to all DCE versions. There are a + * handful of panels that cannot handle disable stream at + * HBLANK and will result in a white line flash across the + * screen on stream disable. + */ + REG_GET(DP_VID_STREAM_CNTL, DP_VID_STREAM_ENABLE, ®1); + if ((reg1 & 0x1) == 0) + /*stream not enabled*/ + return; + /* Specify the video stream disable point + * (2 = start of the next vertical blank) + */ + REG_UPDATE(DP_VID_STREAM_CNTL, DP_VID_STREAM_DIS_DEFER, 2); + /* Larger delay to wait until VBLANK - use max retry of + * 10us*3000=30ms. This covers 16.6ms of typical 60 Hz mode + + * a little more because we may not trust delay accuracy. + */ + max_retries = DP_BLANK_MAX_RETRY * 150; + + /* disable DP stream */ + REG_UPDATE(DP_VID_STREAM_CNTL, DP_VID_STREAM_ENABLE, 0); + + /* the encoder stops sending the video stream + * at the start of the vertical blanking. + * Poll for DP_VID_STREAM_STATUS == 0 + */ + + REG_WAIT(DP_VID_STREAM_CNTL, DP_VID_STREAM_STATUS, + 0, + 10, max_retries); + + ASSERT(retries <= max_retries); + + /* Tell the DP encoder to ignore timing from CRTC, must be done after + * the polling. If we set DP_STEER_FIFO_RESET before DP stream blank is + * complete, stream status will be stuck in video stream enabled state, + * i.e. DP_VID_STREAM_STATUS stuck at 1. + */ + + REG_UPDATE(DP_STEER_FIFO, DP_STEER_FIFO_RESET, true); +} + +/* output video stream to link encoder */ +void enc1_stream_encoder_dp_unblank( + struct stream_encoder *enc, + const struct encoder_unblank_param *param) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + if (param->link_settings.link_rate != LINK_RATE_UNKNOWN) { + uint32_t n_vid = 0x8000; + uint32_t m_vid; + + /* M / N = Fstream / Flink + * m_vid / n_vid = pixel rate / link rate + */ + + uint64_t m_vid_l = n_vid; + + m_vid_l *= param->pixel_clk_khz; + m_vid_l = div_u64(m_vid_l, + param->link_settings.link_rate + * LINK_RATE_REF_FREQ_IN_KHZ); + + m_vid = (uint32_t) m_vid_l; + + /* enable auto measurement */ + + REG_UPDATE(DP_VID_TIMING, DP_VID_M_N_GEN_EN, 0); + + /* auto measurement need 1 full 0x8000 symbol cycle to kick in, + * therefore program initial value for Mvid and Nvid + */ + + REG_UPDATE(DP_VID_N, DP_VID_N, n_vid); + + REG_UPDATE(DP_VID_M, DP_VID_M, m_vid); + + REG_UPDATE(DP_VID_TIMING, DP_VID_M_N_GEN_EN, 1); + } + + /* set DIG_START to 0x1 to resync FIFO */ + + REG_UPDATE(DIG_FE_CNTL, DIG_START, 1); + + /* switch DP encoder to CRTC data */ + + REG_UPDATE(DP_STEER_FIFO, DP_STEER_FIFO_RESET, 0); + + /* wait 100us for DIG/DP logic to prime + * (i.e. a few video lines) + */ + udelay(100); + + /* the hardware would start sending video at the start of the next DP + * frame (i.e. rising edge of the vblank). + * NOTE: We used to program DP_VID_STREAM_DIS_DEFER = 2 here, but this + * register has no effect on enable transition! HW always guarantees + * VID_STREAM enable at start of next frame, and this is not + * programmable + */ + + REG_UPDATE(DP_VID_STREAM_CNTL, DP_VID_STREAM_ENABLE, true); +} + +void enc1_stream_encoder_set_avmute( + struct stream_encoder *enc, + bool enable) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + unsigned int value = enable ? 1 : 0; + + REG_UPDATE(HDMI_GC, HDMI_GC_AVMUTE, value); +} + + +#define DP_SEC_AUD_N__DP_SEC_AUD_N__DEFAULT 0x8000 +#define DP_SEC_TIMESTAMP__DP_SEC_TIMESTAMP_MODE__AUTO_CALC 1 + +#include "include/audio_types.h" + +/** +* speakersToChannels +* +* @brief +* translate speakers to channels +* +* FL - Front Left +* FR - Front Right +* RL - Rear Left +* RR - Rear Right +* RC - Rear Center +* FC - Front Center +* FLC - Front Left Center +* FRC - Front Right Center +* RLC - Rear Left Center +* RRC - Rear Right Center +* LFE - Low Freq Effect +* +* FC +* FLC FRC +* FL FR +* +* LFE +* () +* +* +* RL RR +* RLC RRC +* RC +* +* ch 8 7 6 5 4 3 2 1 +* 0b00000011 - - - - - - FR FL +* 0b00000111 - - - - - LFE FR FL +* 0b00001011 - - - - FC - FR FL +* 0b00001111 - - - - FC LFE FR FL +* 0b00010011 - - - RC - - FR FL +* 0b00010111 - - - RC - LFE FR FL +* 0b00011011 - - - RC FC - FR FL +* 0b00011111 - - - RC FC LFE FR FL +* 0b00110011 - - RR RL - - FR FL +* 0b00110111 - - RR RL - LFE FR FL +* 0b00111011 - - RR RL FC - FR FL +* 0b00111111 - - RR RL FC LFE FR FL +* 0b01110011 - RC RR RL - - FR FL +* 0b01110111 - RC RR RL - LFE FR FL +* 0b01111011 - RC RR RL FC - FR FL +* 0b01111111 - RC RR RL FC LFE FR FL +* 0b11110011 RRC RLC RR RL - - FR FL +* 0b11110111 RRC RLC RR RL - LFE FR FL +* 0b11111011 RRC RLC RR RL FC - FR FL +* 0b11111111 RRC RLC RR RL FC LFE FR FL +* 0b11000011 FRC FLC - - - - FR FL +* 0b11000111 FRC FLC - - - LFE FR FL +* 0b11001011 FRC FLC - - FC - FR FL +* 0b11001111 FRC FLC - - FC LFE FR FL +* 0b11010011 FRC FLC - RC - - FR FL +* 0b11010111 FRC FLC - RC - LFE FR FL +* 0b11011011 FRC FLC - RC FC - FR FL +* 0b11011111 FRC FLC - RC FC LFE FR FL +* 0b11110011 FRC FLC RR RL - - FR FL +* 0b11110111 FRC FLC RR RL - LFE FR FL +* 0b11111011 FRC FLC RR RL FC - FR FL +* 0b11111111 FRC FLC RR RL FC LFE FR FL +* +* @param +* speakers - speaker information as it comes from CEA audio block +*/ +/* translate speakers to channels */ + +union audio_cea_channels { + uint8_t all; + struct audio_cea_channels_bits { + uint32_t FL:1; + uint32_t FR:1; + uint32_t LFE:1; + uint32_t FC:1; + uint32_t RL_RC:1; + uint32_t RR:1; + uint32_t RC_RLC_FLC:1; + uint32_t RRC_FRC:1; + } channels; +}; + +struct audio_clock_info { + /* pixel clock frequency*/ + uint32_t pixel_clock_in_10khz; + /* N - 32KHz audio */ + uint32_t n_32khz; + /* CTS - 32KHz audio*/ + uint32_t cts_32khz; + uint32_t n_44khz; + uint32_t cts_44khz; + uint32_t n_48khz; + uint32_t cts_48khz; +}; + +/* 25.2MHz/1.001*/ +/* 25.2MHz/1.001*/ +/* 25.2MHz*/ +/* 27MHz */ +/* 27MHz*1.001*/ +/* 27MHz*1.001*/ +/* 54MHz*/ +/* 54MHz*1.001*/ +/* 74.25MHz/1.001*/ +/* 74.25MHz*/ +/* 148.5MHz/1.001*/ +/* 148.5MHz*/ + +static const struct audio_clock_info audio_clock_info_table[16] = { + {2517, 4576, 28125, 7007, 31250, 6864, 28125}, + {2518, 4576, 28125, 7007, 31250, 6864, 28125}, + {2520, 4096, 25200, 6272, 28000, 6144, 25200}, + {2700, 4096, 27000, 6272, 30000, 6144, 27000}, + {2702, 4096, 27027, 6272, 30030, 6144, 27027}, + {2703, 4096, 27027, 6272, 30030, 6144, 27027}, + {5400, 4096, 54000, 6272, 60000, 6144, 54000}, + {5405, 4096, 54054, 6272, 60060, 6144, 54054}, + {7417, 11648, 210937, 17836, 234375, 11648, 140625}, + {7425, 4096, 74250, 6272, 82500, 6144, 74250}, + {14835, 11648, 421875, 8918, 234375, 5824, 140625}, + {14850, 4096, 148500, 6272, 165000, 6144, 148500}, + {29670, 5824, 421875, 4459, 234375, 5824, 281250}, + {29700, 3072, 222750, 4704, 247500, 5120, 247500}, + {59340, 5824, 843750, 8918, 937500, 5824, 562500}, + {59400, 3072, 445500, 9408, 990000, 6144, 594000} +}; + +static const struct audio_clock_info audio_clock_info_table_36bpc[14] = { + {2517, 9152, 84375, 7007, 48875, 9152, 56250}, + {2518, 9152, 84375, 7007, 48875, 9152, 56250}, + {2520, 4096, 37800, 6272, 42000, 6144, 37800}, + {2700, 4096, 40500, 6272, 45000, 6144, 40500}, + {2702, 8192, 81081, 6272, 45045, 8192, 54054}, + {2703, 8192, 81081, 6272, 45045, 8192, 54054}, + {5400, 4096, 81000, 6272, 90000, 6144, 81000}, + {5405, 4096, 81081, 6272, 90090, 6144, 81081}, + {7417, 11648, 316406, 17836, 351562, 11648, 210937}, + {7425, 4096, 111375, 6272, 123750, 6144, 111375}, + {14835, 11648, 632812, 17836, 703125, 11648, 421875}, + {14850, 4096, 222750, 6272, 247500, 6144, 222750}, + {29670, 5824, 632812, 8918, 703125, 5824, 421875}, + {29700, 4096, 445500, 4704, 371250, 5120, 371250} +}; + +static const struct audio_clock_info audio_clock_info_table_48bpc[14] = { + {2517, 4576, 56250, 7007, 62500, 6864, 56250}, + {2518, 4576, 56250, 7007, 62500, 6864, 56250}, + {2520, 4096, 50400, 6272, 56000, 6144, 50400}, + {2700, 4096, 54000, 6272, 60000, 6144, 54000}, + {2702, 4096, 54054, 6267, 60060, 8192, 54054}, + {2703, 4096, 54054, 6272, 60060, 8192, 54054}, + {5400, 4096, 108000, 6272, 120000, 6144, 108000}, + {5405, 4096, 108108, 6272, 120120, 6144, 108108}, + {7417, 11648, 421875, 17836, 468750, 11648, 281250}, + {7425, 4096, 148500, 6272, 165000, 6144, 148500}, + {14835, 11648, 843750, 8918, 468750, 11648, 281250}, + {14850, 4096, 297000, 6272, 330000, 6144, 297000}, + {29670, 5824, 843750, 4459, 468750, 5824, 562500}, + {29700, 3072, 445500, 4704, 495000, 5120, 495000} + + +}; + +static union audio_cea_channels speakers_to_channels( + struct audio_speaker_flags speaker_flags) +{ + union audio_cea_channels cea_channels = {0}; + + /* these are one to one */ + cea_channels.channels.FL = speaker_flags.FL_FR; + cea_channels.channels.FR = speaker_flags.FL_FR; + cea_channels.channels.LFE = speaker_flags.LFE; + cea_channels.channels.FC = speaker_flags.FC; + + /* if Rear Left and Right exist move RC speaker to channel 7 + * otherwise to channel 5 + */ + if (speaker_flags.RL_RR) { + cea_channels.channels.RL_RC = speaker_flags.RL_RR; + cea_channels.channels.RR = speaker_flags.RL_RR; + cea_channels.channels.RC_RLC_FLC = speaker_flags.RC; + } else { + cea_channels.channels.RL_RC = speaker_flags.RC; + } + + /* FRONT Left Right Center and REAR Left Right Center are exclusive */ + if (speaker_flags.FLC_FRC) { + cea_channels.channels.RC_RLC_FLC = speaker_flags.FLC_FRC; + cea_channels.channels.RRC_FRC = speaker_flags.FLC_FRC; + } else { + cea_channels.channels.RC_RLC_FLC = speaker_flags.RLC_RRC; + cea_channels.channels.RRC_FRC = speaker_flags.RLC_RRC; + } + + return cea_channels; +} + +static uint32_t calc_max_audio_packets_per_line( + const struct audio_crtc_info *crtc_info) +{ + uint32_t max_packets_per_line; + + max_packets_per_line = + crtc_info->h_total - crtc_info->h_active; + + if (crtc_info->pixel_repetition) + max_packets_per_line *= crtc_info->pixel_repetition; + + /* for other hdmi features */ + max_packets_per_line -= 58; + /* for Control Period */ + max_packets_per_line -= 16; + /* Number of Audio Packets per Line */ + max_packets_per_line /= 32; + + return max_packets_per_line; +} + +static void get_audio_clock_info( + enum dc_color_depth color_depth, + uint32_t crtc_pixel_clock_in_khz, + uint32_t actual_pixel_clock_in_khz, + struct audio_clock_info *audio_clock_info) +{ + const struct audio_clock_info *clock_info; + uint32_t index; + uint32_t crtc_pixel_clock_in_10khz = crtc_pixel_clock_in_khz / 10; + uint32_t audio_array_size; + + switch (color_depth) { + case COLOR_DEPTH_161616: + clock_info = audio_clock_info_table_48bpc; + audio_array_size = ARRAY_SIZE( + audio_clock_info_table_48bpc); + break; + case COLOR_DEPTH_121212: + clock_info = audio_clock_info_table_36bpc; + audio_array_size = ARRAY_SIZE( + audio_clock_info_table_36bpc); + break; + default: + clock_info = audio_clock_info_table; + audio_array_size = ARRAY_SIZE( + audio_clock_info_table); + break; + } + + if (clock_info != NULL) { + /* search for exact pixel clock in table */ + for (index = 0; index < audio_array_size; index++) { + if (clock_info[index].pixel_clock_in_10khz > + crtc_pixel_clock_in_10khz) + break; /* not match */ + else if (clock_info[index].pixel_clock_in_10khz == + crtc_pixel_clock_in_10khz) { + /* match found */ + *audio_clock_info = clock_info[index]; + return; + } + } + } + + /* not found */ + if (actual_pixel_clock_in_khz == 0) + actual_pixel_clock_in_khz = crtc_pixel_clock_in_khz; + + /* See HDMI spec the table entry under + * pixel clock of "Other". */ + audio_clock_info->pixel_clock_in_10khz = + actual_pixel_clock_in_khz / 10; + audio_clock_info->cts_32khz = actual_pixel_clock_in_khz; + audio_clock_info->cts_44khz = actual_pixel_clock_in_khz; + audio_clock_info->cts_48khz = actual_pixel_clock_in_khz; + + audio_clock_info->n_32khz = 4096; + audio_clock_info->n_44khz = 6272; + audio_clock_info->n_48khz = 6144; +} + +static void enc1_se_audio_setup( + struct stream_encoder *enc, + unsigned int az_inst, + struct audio_info *audio_info) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + uint32_t speakers = 0; + uint32_t channels = 0; + + ASSERT(audio_info); + if (audio_info == NULL) + /* This should not happen.it does so we don't get BSOD*/ + return; + + speakers = audio_info->flags.info.ALLSPEAKERS; + channels = speakers_to_channels(audio_info->flags.speaker_flags).all; + + /* setup the audio stream source select (audio -> dig mapping) */ + REG_SET(AFMT_AUDIO_SRC_CONTROL, 0, AFMT_AUDIO_SRC_SELECT, az_inst); + + /* Channel allocation */ + REG_UPDATE(AFMT_AUDIO_PACKET_CONTROL2, AFMT_AUDIO_CHANNEL_ENABLE, channels); +} + +static void enc1_se_setup_hdmi_audio( + struct stream_encoder *enc, + const struct audio_crtc_info *crtc_info) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + struct audio_clock_info audio_clock_info = {0}; + uint32_t max_packets_per_line; + + /* For now still do calculation, although this field is ignored when + * above HDMI_PACKET_GEN_VERSION set to 1 + */ + max_packets_per_line = calc_max_audio_packets_per_line(crtc_info); + + /* HDMI_AUDIO_PACKET_CONTROL */ + REG_UPDATE_2(HDMI_AUDIO_PACKET_CONTROL, + HDMI_AUDIO_PACKETS_PER_LINE, max_packets_per_line, + HDMI_AUDIO_DELAY_EN, 1); + + /* AFMT_AUDIO_PACKET_CONTROL */ + REG_UPDATE(AFMT_AUDIO_PACKET_CONTROL, AFMT_60958_CS_UPDATE, 1); + + /* AFMT_AUDIO_PACKET_CONTROL2 */ + REG_UPDATE_2(AFMT_AUDIO_PACKET_CONTROL2, + AFMT_AUDIO_LAYOUT_OVRD, 0, + AFMT_60958_OSF_OVRD, 0); + + /* HDMI_ACR_PACKET_CONTROL */ + REG_UPDATE_3(HDMI_ACR_PACKET_CONTROL, + HDMI_ACR_AUTO_SEND, 1, + HDMI_ACR_SOURCE, 0, + HDMI_ACR_AUDIO_PRIORITY, 0); + + /* Program audio clock sample/regeneration parameters */ + get_audio_clock_info(crtc_info->color_depth, + crtc_info->requested_pixel_clock, + crtc_info->calculated_pixel_clock, + &audio_clock_info); + DC_LOG_HW_AUDIO( + "\n%s:Input::requested_pixel_clock = %d" \ + "calculated_pixel_clock = %d \n", __func__, \ + crtc_info->requested_pixel_clock, \ + crtc_info->calculated_pixel_clock); + + /* HDMI_ACR_32_0__HDMI_ACR_CTS_32_MASK */ + REG_UPDATE(HDMI_ACR_32_0, HDMI_ACR_CTS_32, audio_clock_info.cts_32khz); + + /* HDMI_ACR_32_1__HDMI_ACR_N_32_MASK */ + REG_UPDATE(HDMI_ACR_32_1, HDMI_ACR_N_32, audio_clock_info.n_32khz); + + /* HDMI_ACR_44_0__HDMI_ACR_CTS_44_MASK */ + REG_UPDATE(HDMI_ACR_44_0, HDMI_ACR_CTS_44, audio_clock_info.cts_44khz); + + /* HDMI_ACR_44_1__HDMI_ACR_N_44_MASK */ + REG_UPDATE(HDMI_ACR_44_1, HDMI_ACR_N_44, audio_clock_info.n_44khz); + + /* HDMI_ACR_48_0__HDMI_ACR_CTS_48_MASK */ + REG_UPDATE(HDMI_ACR_48_0, HDMI_ACR_CTS_48, audio_clock_info.cts_48khz); + + /* HDMI_ACR_48_1__HDMI_ACR_N_48_MASK */ + REG_UPDATE(HDMI_ACR_48_1, HDMI_ACR_N_48, audio_clock_info.n_48khz); + + /* Video driver cannot know in advance which sample rate will + * be used by HD Audio driver + * HDMI_ACR_PACKET_CONTROL__HDMI_ACR_N_MULTIPLE field is + * programmed below in interruppt callback + */ + + /* AFMT_60958_0__AFMT_60958_CS_CHANNEL_NUMBER_L_MASK & + * AFMT_60958_0__AFMT_60958_CS_CLOCK_ACCURACY_MASK + */ + REG_UPDATE_2(AFMT_60958_0, + AFMT_60958_CS_CHANNEL_NUMBER_L, 1, + AFMT_60958_CS_CLOCK_ACCURACY, 0); + + /* AFMT_60958_1 AFMT_60958_CS_CHALNNEL_NUMBER_R */ + REG_UPDATE(AFMT_60958_1, AFMT_60958_CS_CHANNEL_NUMBER_R, 2); + + /* AFMT_60958_2 now keep this settings until + * Programming guide comes out + */ + REG_UPDATE_6(AFMT_60958_2, + AFMT_60958_CS_CHANNEL_NUMBER_2, 3, + AFMT_60958_CS_CHANNEL_NUMBER_3, 4, + AFMT_60958_CS_CHANNEL_NUMBER_4, 5, + AFMT_60958_CS_CHANNEL_NUMBER_5, 6, + AFMT_60958_CS_CHANNEL_NUMBER_6, 7, + AFMT_60958_CS_CHANNEL_NUMBER_7, 8); +} + +static void enc1_se_setup_dp_audio( + struct stream_encoder *enc) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + /* --- DP Audio packet configurations --- */ + + /* ATP Configuration */ + REG_SET(DP_SEC_AUD_N, 0, + DP_SEC_AUD_N, DP_SEC_AUD_N__DP_SEC_AUD_N__DEFAULT); + + /* Async/auto-calc timestamp mode */ + REG_SET(DP_SEC_TIMESTAMP, 0, DP_SEC_TIMESTAMP_MODE, + DP_SEC_TIMESTAMP__DP_SEC_TIMESTAMP_MODE__AUTO_CALC); + + /* --- The following are the registers + * copied from the SetupHDMI --- + */ + + /* AFMT_AUDIO_PACKET_CONTROL */ + REG_UPDATE(AFMT_AUDIO_PACKET_CONTROL, AFMT_60958_CS_UPDATE, 1); + + /* AFMT_AUDIO_PACKET_CONTROL2 */ + /* Program the ATP and AIP next */ + REG_UPDATE_2(AFMT_AUDIO_PACKET_CONTROL2, + AFMT_AUDIO_LAYOUT_OVRD, 0, + AFMT_60958_OSF_OVRD, 0); + + /* AFMT_INFOFRAME_CONTROL0 */ + REG_UPDATE(AFMT_INFOFRAME_CONTROL0, AFMT_AUDIO_INFO_UPDATE, 1); + + /* AFMT_60958_0__AFMT_60958_CS_CLOCK_ACCURACY_MASK */ + REG_UPDATE(AFMT_60958_0, AFMT_60958_CS_CLOCK_ACCURACY, 0); +} + +static void enc1_se_enable_audio_clock( + struct stream_encoder *enc, + bool enable) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + if (REG(AFMT_CNTL) == 0) + return; /* DCE8/10 does not have this register */ + + REG_UPDATE(AFMT_CNTL, AFMT_AUDIO_CLOCK_EN, !!enable); + + /* wait for AFMT clock to turn on, + * expectation: this should complete in 1-2 reads + * + * REG_WAIT(AFMT_CNTL, AFMT_AUDIO_CLOCK_ON, !!enable, 1, 10); + * + * TODO: wait for clock_on does not work well. May need HW + * program sequence. But audio seems work normally even without wait + * for clock_on status change + */ +} + +static void enc1_se_enable_dp_audio( + struct stream_encoder *enc) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + /* Enable Audio packets */ + REG_UPDATE(DP_SEC_CNTL, DP_SEC_ASP_ENABLE, 1); + + /* Program the ATP and AIP next */ + REG_UPDATE_2(DP_SEC_CNTL, + DP_SEC_ATP_ENABLE, 1, + DP_SEC_AIP_ENABLE, 1); + + /* Program STREAM_ENABLE after all the other enables. */ + REG_UPDATE(DP_SEC_CNTL, DP_SEC_STREAM_ENABLE, 1); +} + +static void enc1_se_disable_dp_audio( + struct stream_encoder *enc) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + uint32_t value = 0; + + /* Disable Audio packets */ + REG_UPDATE_5(DP_SEC_CNTL, + DP_SEC_ASP_ENABLE, 0, + DP_SEC_ATP_ENABLE, 0, + DP_SEC_AIP_ENABLE, 0, + DP_SEC_ACM_ENABLE, 0, + DP_SEC_STREAM_ENABLE, 0); + + /* This register shared with encoder info frame. Therefore we need to + * keep master enabled if at least on of the fields is not 0 + */ + value = REG_READ(DP_SEC_CNTL); + if (value != 0) + REG_UPDATE(DP_SEC_CNTL, DP_SEC_STREAM_ENABLE, 1); + +} + +void enc1_se_audio_mute_control( + struct stream_encoder *enc, + bool mute) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + + REG_UPDATE(AFMT_AUDIO_PACKET_CONTROL, AFMT_AUDIO_SAMPLE_SEND, !mute); +} + +void enc1_se_dp_audio_setup( + struct stream_encoder *enc, + unsigned int az_inst, + struct audio_info *info) +{ + enc1_se_audio_setup(enc, az_inst, info); +} + +void enc1_se_dp_audio_enable( + struct stream_encoder *enc) +{ + enc1_se_enable_audio_clock(enc, true); + enc1_se_setup_dp_audio(enc); + enc1_se_enable_dp_audio(enc); +} + +void enc1_se_dp_audio_disable( + struct stream_encoder *enc) +{ + enc1_se_disable_dp_audio(enc); + enc1_se_enable_audio_clock(enc, false); +} + +void enc1_se_hdmi_audio_setup( + struct stream_encoder *enc, + unsigned int az_inst, + struct audio_info *info, + struct audio_crtc_info *audio_crtc_info) +{ + enc1_se_enable_audio_clock(enc, true); + enc1_se_setup_hdmi_audio(enc, audio_crtc_info); + enc1_se_audio_setup(enc, az_inst, info); +} + +void enc1_se_hdmi_audio_disable( + struct stream_encoder *enc) +{ + enc1_se_enable_audio_clock(enc, false); +} + + +void enc1_setup_stereo_sync( + struct stream_encoder *enc, + int tg_inst, bool enable) +{ + struct dcn10_stream_encoder *enc1 = DCN10STRENC_FROM_STRENC(enc); + REG_UPDATE(DIG_FE_CNTL, DIG_STEREOSYNC_SELECT, tg_inst); + REG_UPDATE(DIG_FE_CNTL, DIG_STEREOSYNC_GATE_EN, !enable); +} + + +static const struct stream_encoder_funcs dcn10_str_enc_funcs = { + .dp_set_stream_attribute = + enc1_stream_encoder_dp_set_stream_attribute, + .hdmi_set_stream_attribute = + enc1_stream_encoder_hdmi_set_stream_attribute, + .dvi_set_stream_attribute = + enc1_stream_encoder_dvi_set_stream_attribute, + .set_mst_bandwidth = + enc1_stream_encoder_set_mst_bandwidth, + .update_hdmi_info_packets = + enc1_stream_encoder_update_hdmi_info_packets, + .stop_hdmi_info_packets = + enc1_stream_encoder_stop_hdmi_info_packets, + .update_dp_info_packets = + enc1_stream_encoder_update_dp_info_packets, + .stop_dp_info_packets = + enc1_stream_encoder_stop_dp_info_packets, + .dp_blank = + enc1_stream_encoder_dp_blank, + .dp_unblank = + enc1_stream_encoder_dp_unblank, + .audio_mute_control = enc1_se_audio_mute_control, + + .dp_audio_setup = enc1_se_dp_audio_setup, + .dp_audio_enable = enc1_se_dp_audio_enable, + .dp_audio_disable = enc1_se_dp_audio_disable, + + .hdmi_audio_setup = enc1_se_hdmi_audio_setup, + .hdmi_audio_disable = enc1_se_hdmi_audio_disable, + .setup_stereo_sync = enc1_setup_stereo_sync, + .set_avmute = enc1_stream_encoder_set_avmute, +}; + +void dcn10_stream_encoder_construct( + struct dcn10_stream_encoder *enc1, + struct dc_context *ctx, + struct dc_bios *bp, + enum engine_id eng_id, + const struct dcn10_stream_enc_registers *regs, + const struct dcn10_stream_encoder_shift *se_shift, + const struct dcn10_stream_encoder_mask *se_mask) +{ + enc1->base.funcs = &dcn10_str_enc_funcs; + enc1->base.ctx = ctx; + enc1->base.id = eng_id; + enc1->base.bp = bp; + enc1->regs = regs; + enc1->se_shift = se_shift; + enc1->se_mask = se_mask; +} + diff --git a/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h new file mode 100644 index 0000000000000..6b3e4ded155bc --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/dcn10/dcn10_stream_encoder.h @@ -0,0 +1,524 @@ +/* + * Copyright 2012-15 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + +#ifndef __DC_STREAM_ENCODER_DCN10_H__ +#define __DC_STREAM_ENCODER_DCN10_H__ + +#include "stream_encoder.h" + +#define DCN10STRENC_FROM_STRENC(stream_encoder)\ + container_of(stream_encoder, struct dcn10_stream_encoder, base) + +#define SE_COMMON_DCN_REG_LIST(id) \ + SRI(AFMT_CNTL, DIG, id), \ + SRI(AFMT_GENERIC_0, DIG, id), \ + SRI(AFMT_GENERIC_1, DIG, id), \ + SRI(AFMT_GENERIC_2, DIG, id), \ + SRI(AFMT_GENERIC_3, DIG, id), \ + SRI(AFMT_GENERIC_4, DIG, id), \ + SRI(AFMT_GENERIC_5, DIG, id), \ + SRI(AFMT_GENERIC_6, DIG, id), \ + SRI(AFMT_GENERIC_7, DIG, id), \ + SRI(AFMT_GENERIC_HDR, DIG, id), \ + SRI(AFMT_INFOFRAME_CONTROL0, DIG, id), \ + SRI(AFMT_VBI_PACKET_CONTROL, DIG, id), \ + SRI(AFMT_VBI_PACKET_CONTROL1, DIG, id), \ + SRI(AFMT_AUDIO_PACKET_CONTROL, DIG, id), \ + SRI(AFMT_AUDIO_PACKET_CONTROL2, DIG, id), \ + SRI(AFMT_AUDIO_SRC_CONTROL, DIG, id), \ + SRI(AFMT_60958_0, DIG, id), \ + SRI(AFMT_60958_1, DIG, id), \ + SRI(AFMT_60958_2, DIG, id), \ + SRI(DIG_FE_CNTL, DIG, id), \ + SRI(HDMI_CONTROL, DIG, id), \ + SRI(HDMI_DB_CONTROL, DIG, id), \ + SRI(HDMI_GC, DIG, id), \ + SRI(HDMI_GENERIC_PACKET_CONTROL0, DIG, id), \ + SRI(HDMI_GENERIC_PACKET_CONTROL1, DIG, id), \ + SRI(HDMI_GENERIC_PACKET_CONTROL2, DIG, id), \ + SRI(HDMI_GENERIC_PACKET_CONTROL3, DIG, id), \ + SRI(HDMI_INFOFRAME_CONTROL0, DIG, id), \ + SRI(HDMI_INFOFRAME_CONTROL1, DIG, id), \ + SRI(HDMI_VBI_PACKET_CONTROL, DIG, id), \ + SRI(HDMI_AUDIO_PACKET_CONTROL, DIG, id),\ + SRI(HDMI_ACR_PACKET_CONTROL, DIG, id),\ + SRI(HDMI_ACR_32_0, DIG, id),\ + SRI(HDMI_ACR_32_1, DIG, id),\ + SRI(HDMI_ACR_44_0, DIG, id),\ + SRI(HDMI_ACR_44_1, DIG, id),\ + SRI(HDMI_ACR_48_0, DIG, id),\ + SRI(HDMI_ACR_48_1, DIG, id),\ + SRI(DP_DB_CNTL, DP, id), \ + SRI(DP_MSA_MISC, DP, id), \ + SRI(DP_MSA_COLORIMETRY, DP, id), \ + SRI(DP_MSA_TIMING_PARAM1, DP, id), \ + SRI(DP_MSA_TIMING_PARAM2, DP, id), \ + SRI(DP_MSA_TIMING_PARAM3, DP, id), \ + SRI(DP_MSA_TIMING_PARAM4, DP, id), \ + SRI(DP_MSE_RATE_CNTL, DP, id), \ + SRI(DP_MSE_RATE_UPDATE, DP, id), \ + SRI(DP_PIXEL_FORMAT, DP, id), \ + SRI(DP_SEC_CNTL, DP, id), \ + SRI(DP_STEER_FIFO, DP, id), \ + SRI(DP_VID_M, DP, id), \ + SRI(DP_VID_N, DP, id), \ + SRI(DP_VID_STREAM_CNTL, DP, id), \ + SRI(DP_VID_TIMING, DP, id), \ + SRI(DP_SEC_AUD_N, DP, id), \ + SRI(DP_SEC_TIMESTAMP, DP, id) + +#define SE_DCN_REG_LIST(id)\ + SE_COMMON_DCN_REG_LIST(id) + + +struct dcn10_stream_enc_registers { + uint32_t AFMT_CNTL; + uint32_t AFMT_AVI_INFO0; + uint32_t AFMT_AVI_INFO1; + uint32_t AFMT_AVI_INFO2; + uint32_t AFMT_AVI_INFO3; + uint32_t AFMT_GENERIC_0; + uint32_t AFMT_GENERIC_1; + uint32_t AFMT_GENERIC_2; + uint32_t AFMT_GENERIC_3; + uint32_t AFMT_GENERIC_4; + uint32_t AFMT_GENERIC_5; + uint32_t AFMT_GENERIC_6; + uint32_t AFMT_GENERIC_7; + uint32_t AFMT_GENERIC_HDR; + uint32_t AFMT_INFOFRAME_CONTROL0; + uint32_t AFMT_VBI_PACKET_CONTROL; + uint32_t AFMT_VBI_PACKET_CONTROL1; + uint32_t AFMT_AUDIO_PACKET_CONTROL; + uint32_t AFMT_AUDIO_PACKET_CONTROL2; + uint32_t AFMT_AUDIO_SRC_CONTROL; + uint32_t AFMT_60958_0; + uint32_t AFMT_60958_1; + uint32_t AFMT_60958_2; + uint32_t DIG_FE_CNTL; + uint32_t DP_MSE_RATE_CNTL; + uint32_t DP_MSE_RATE_UPDATE; + uint32_t DP_PIXEL_FORMAT; + uint32_t DP_SEC_CNTL; + uint32_t DP_STEER_FIFO; + uint32_t DP_VID_M; + uint32_t DP_VID_N; + uint32_t DP_VID_STREAM_CNTL; + uint32_t DP_VID_TIMING; + uint32_t DP_SEC_AUD_N; + uint32_t DP_SEC_TIMESTAMP; + uint32_t HDMI_CONTROL; + uint32_t HDMI_GC; + uint32_t HDMI_GENERIC_PACKET_CONTROL0; + uint32_t HDMI_GENERIC_PACKET_CONTROL1; + uint32_t HDMI_GENERIC_PACKET_CONTROL2; + uint32_t HDMI_GENERIC_PACKET_CONTROL3; + uint32_t HDMI_GENERIC_PACKET_CONTROL4; + uint32_t HDMI_GENERIC_PACKET_CONTROL5; + uint32_t HDMI_INFOFRAME_CONTROL0; + uint32_t HDMI_INFOFRAME_CONTROL1; + uint32_t HDMI_VBI_PACKET_CONTROL; + uint32_t HDMI_AUDIO_PACKET_CONTROL; + uint32_t HDMI_ACR_PACKET_CONTROL; + uint32_t HDMI_ACR_32_0; + uint32_t HDMI_ACR_32_1; + uint32_t HDMI_ACR_44_0; + uint32_t HDMI_ACR_44_1; + uint32_t HDMI_ACR_48_0; + uint32_t HDMI_ACR_48_1; + uint32_t DP_DB_CNTL; + uint32_t DP_MSA_MISC; + uint32_t DP_MSA_COLORIMETRY; + uint32_t DP_MSA_TIMING_PARAM1; + uint32_t DP_MSA_TIMING_PARAM2; + uint32_t DP_MSA_TIMING_PARAM3; + uint32_t DP_MSA_TIMING_PARAM4; + uint32_t HDMI_DB_CONTROL; +}; + + +#define SE_SF(reg_name, field_name, post_fix)\ + .field_name = reg_name ## __ ## field_name ## post_fix + +#define SE_COMMON_MASK_SH_LIST_SOC_BASE(mask_sh)\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL, AFMT_GENERIC_INDEX, mask_sh),\ + SE_SF(DIG0_AFMT_GENERIC_HDR, AFMT_GENERIC_HB0, mask_sh),\ + SE_SF(DIG0_AFMT_GENERIC_HDR, AFMT_GENERIC_HB1, mask_sh),\ + SE_SF(DIG0_AFMT_GENERIC_HDR, AFMT_GENERIC_HB2, mask_sh),\ + SE_SF(DIG0_AFMT_GENERIC_HDR, AFMT_GENERIC_HB3, mask_sh),\ + SE_SF(DP0_DP_PIXEL_FORMAT, DP_PIXEL_ENCODING, mask_sh),\ + SE_SF(DP0_DP_PIXEL_FORMAT, DP_COMPONENT_DEPTH, mask_sh),\ + SE_SF(DIG0_HDMI_CONTROL, HDMI_PACKET_GEN_VERSION, mask_sh),\ + SE_SF(DIG0_HDMI_CONTROL, HDMI_KEEPOUT_MODE, mask_sh),\ + SE_SF(DIG0_HDMI_CONTROL, HDMI_DEEP_COLOR_ENABLE, mask_sh),\ + SE_SF(DIG0_HDMI_CONTROL, HDMI_DEEP_COLOR_DEPTH, mask_sh),\ + SE_SF(DIG0_HDMI_CONTROL, HDMI_DATA_SCRAMBLE_EN, mask_sh),\ + SE_SF(DIG0_HDMI_VBI_PACKET_CONTROL, HDMI_GC_CONT, mask_sh),\ + SE_SF(DIG0_HDMI_VBI_PACKET_CONTROL, HDMI_GC_SEND, mask_sh),\ + SE_SF(DIG0_HDMI_VBI_PACKET_CONTROL, HDMI_NULL_SEND, mask_sh),\ + SE_SF(DIG0_HDMI_INFOFRAME_CONTROL0, HDMI_AUDIO_INFO_SEND, mask_sh),\ + SE_SF(DIG0_AFMT_INFOFRAME_CONTROL0, AFMT_AUDIO_INFO_UPDATE, mask_sh),\ + SE_SF(DIG0_HDMI_INFOFRAME_CONTROL1, HDMI_AUDIO_INFO_LINE, mask_sh),\ + SE_SF(DIG0_HDMI_GC, HDMI_GC_AVMUTE, mask_sh),\ + SE_SF(DP0_DP_MSE_RATE_CNTL, DP_MSE_RATE_X, mask_sh),\ + SE_SF(DP0_DP_MSE_RATE_CNTL, DP_MSE_RATE_Y, mask_sh),\ + SE_SF(DP0_DP_MSE_RATE_UPDATE, DP_MSE_RATE_UPDATE_PENDING, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP0_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_STREAM_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP1_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP2_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP3_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_MPG_ENABLE, mask_sh),\ + SE_SF(DP0_DP_VID_STREAM_CNTL, DP_VID_STREAM_DIS_DEFER, mask_sh),\ + SE_SF(DP0_DP_VID_STREAM_CNTL, DP_VID_STREAM_ENABLE, mask_sh),\ + SE_SF(DP0_DP_VID_STREAM_CNTL, DP_VID_STREAM_STATUS, mask_sh),\ + SE_SF(DP0_DP_STEER_FIFO, DP_STEER_FIFO_RESET, mask_sh),\ + SE_SF(DP0_DP_VID_TIMING, DP_VID_M_N_GEN_EN, mask_sh),\ + SE_SF(DP0_DP_VID_N, DP_VID_N, mask_sh),\ + SE_SF(DP0_DP_VID_M, DP_VID_M, mask_sh),\ + SE_SF(DIG0_DIG_FE_CNTL, DIG_START, mask_sh),\ + SE_SF(DIG0_AFMT_AUDIO_SRC_CONTROL, AFMT_AUDIO_SRC_SELECT, mask_sh),\ + SE_SF(DIG0_AFMT_AUDIO_PACKET_CONTROL2, AFMT_AUDIO_CHANNEL_ENABLE, mask_sh),\ + SE_SF(DIG0_HDMI_AUDIO_PACKET_CONTROL, HDMI_AUDIO_PACKETS_PER_LINE, mask_sh),\ + SE_SF(DIG0_HDMI_AUDIO_PACKET_CONTROL, HDMI_AUDIO_DELAY_EN, mask_sh),\ + SE_SF(DIG0_AFMT_AUDIO_PACKET_CONTROL, AFMT_60958_CS_UPDATE, mask_sh),\ + SE_SF(DIG0_AFMT_AUDIO_PACKET_CONTROL2, AFMT_AUDIO_LAYOUT_OVRD, mask_sh),\ + SE_SF(DIG0_AFMT_AUDIO_PACKET_CONTROL2, AFMT_60958_OSF_OVRD, mask_sh),\ + SE_SF(DIG0_HDMI_ACR_PACKET_CONTROL, HDMI_ACR_AUTO_SEND, mask_sh),\ + SE_SF(DIG0_HDMI_ACR_PACKET_CONTROL, HDMI_ACR_SOURCE, mask_sh),\ + SE_SF(DIG0_HDMI_ACR_PACKET_CONTROL, HDMI_ACR_AUDIO_PRIORITY, mask_sh),\ + SE_SF(DIG0_HDMI_ACR_32_0, HDMI_ACR_CTS_32, mask_sh),\ + SE_SF(DIG0_HDMI_ACR_32_1, HDMI_ACR_N_32, mask_sh),\ + SE_SF(DIG0_HDMI_ACR_44_0, HDMI_ACR_CTS_44, mask_sh),\ + SE_SF(DIG0_HDMI_ACR_44_1, HDMI_ACR_N_44, mask_sh),\ + SE_SF(DIG0_HDMI_ACR_48_0, HDMI_ACR_CTS_48, mask_sh),\ + SE_SF(DIG0_HDMI_ACR_48_1, HDMI_ACR_N_48, mask_sh),\ + SE_SF(DIG0_AFMT_60958_0, AFMT_60958_CS_CHANNEL_NUMBER_L, mask_sh),\ + SE_SF(DIG0_AFMT_60958_0, AFMT_60958_CS_CLOCK_ACCURACY, mask_sh),\ + SE_SF(DIG0_AFMT_60958_1, AFMT_60958_CS_CHANNEL_NUMBER_R, mask_sh),\ + SE_SF(DIG0_AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_2, mask_sh),\ + SE_SF(DIG0_AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_3, mask_sh),\ + SE_SF(DIG0_AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_4, mask_sh),\ + SE_SF(DIG0_AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_5, mask_sh),\ + SE_SF(DIG0_AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_6, mask_sh),\ + SE_SF(DIG0_AFMT_60958_2, AFMT_60958_CS_CHANNEL_NUMBER_7, mask_sh),\ + SE_SF(DP0_DP_SEC_AUD_N, DP_SEC_AUD_N, mask_sh),\ + SE_SF(DP0_DP_SEC_TIMESTAMP, DP_SEC_TIMESTAMP_MODE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_ASP_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_ATP_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_AIP_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_ACM_ENABLE, mask_sh),\ + SE_SF(DIG0_AFMT_AUDIO_PACKET_CONTROL, AFMT_AUDIO_SAMPLE_SEND, mask_sh),\ + SE_SF(DIG0_AFMT_CNTL, AFMT_AUDIO_CLOCK_EN, mask_sh),\ + SE_SF(DIG0_HDMI_CONTROL, HDMI_CLOCK_CHANNEL_RATE, mask_sh),\ + SE_SF(DIG0_DIG_FE_CNTL, TMDS_PIXEL_ENCODING, mask_sh),\ + SE_SF(DIG0_DIG_FE_CNTL, TMDS_COLOR_FORMAT, mask_sh),\ + SE_SF(DIG0_DIG_FE_CNTL, DIG_STEREOSYNC_SELECT, mask_sh),\ + SE_SF(DIG0_DIG_FE_CNTL, DIG_STEREOSYNC_GATE_EN, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL, AFMT_GENERIC_LOCK_STATUS, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL, AFMT_GENERIC_CONFLICT, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL, AFMT_GENERIC_CONFLICT_CLR, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC0_FRAME_UPDATE_PENDING, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC1_FRAME_UPDATE_PENDING, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC2_FRAME_UPDATE_PENDING, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC3_FRAME_UPDATE_PENDING, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC4_FRAME_UPDATE_PENDING, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC5_FRAME_UPDATE_PENDING, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC6_FRAME_UPDATE_PENDING, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC7_FRAME_UPDATE_PENDING, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC0_FRAME_UPDATE, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC1_FRAME_UPDATE, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC2_FRAME_UPDATE, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC3_FRAME_UPDATE, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC4_FRAME_UPDATE, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC5_FRAME_UPDATE, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC6_FRAME_UPDATE, mask_sh),\ + SE_SF(DIG0_AFMT_VBI_PACKET_CONTROL1, AFMT_GENERIC7_FRAME_UPDATE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP4_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP5_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP6_ENABLE, mask_sh),\ + SE_SF(DP0_DP_SEC_CNTL, DP_SEC_GSP7_ENABLE, mask_sh),\ + SE_SF(DP0_DP_DB_CNTL, DP_DB_DISABLE, mask_sh),\ + SE_SF(DP0_DP_MSA_COLORIMETRY, DP_MSA_MISC0, mask_sh),\ + SE_SF(DP0_DP_MSA_TIMING_PARAM1, DP_MSA_HTOTAL, mask_sh),\ + SE_SF(DP0_DP_MSA_TIMING_PARAM1, DP_MSA_VTOTAL, mask_sh),\ + SE_SF(DP0_DP_MSA_TIMING_PARAM2, DP_MSA_HSTART, mask_sh),\ + SE_SF(DP0_DP_MSA_TIMING_PARAM2, DP_MSA_VSTART, mask_sh),\ + SE_SF(DP0_DP_MSA_TIMING_PARAM3, DP_MSA_HSYNCWIDTH, mask_sh),\ + SE_SF(DP0_DP_MSA_TIMING_PARAM3, DP_MSA_HSYNCPOLARITY, mask_sh),\ + SE_SF(DP0_DP_MSA_TIMING_PARAM3, DP_MSA_VSYNCWIDTH, mask_sh),\ + SE_SF(DP0_DP_MSA_TIMING_PARAM3, DP_MSA_VSYNCPOLARITY, mask_sh),\ + SE_SF(DP0_DP_MSA_TIMING_PARAM4, DP_MSA_HWIDTH, mask_sh),\ + SE_SF(DP0_DP_MSA_TIMING_PARAM4, DP_MSA_VHEIGHT, mask_sh),\ + SE_SF(DIG0_HDMI_DB_CONTROL, HDMI_DB_DISABLE, mask_sh),\ + SE_SF(DP0_DP_VID_TIMING, DP_VID_N_MUL, mask_sh) + +#define SE_COMMON_MASK_SH_LIST_SOC(mask_sh)\ + SE_COMMON_MASK_SH_LIST_SOC_BASE(mask_sh) + +#define SE_COMMON_MASK_SH_LIST_DCN10(mask_sh)\ + SE_COMMON_MASK_SH_LIST_SOC(mask_sh),\ + SE_SF(DIG0_HDMI_GENERIC_PACKET_CONTROL0, HDMI_GENERIC0_CONT, mask_sh),\ + SE_SF(DIG0_HDMI_GENERIC_PACKET_CONTROL0, HDMI_GENERIC0_SEND, mask_sh),\ + SE_SF(DIG0_HDMI_GENERIC_PACKET_CONTROL0, HDMI_GENERIC0_LINE, mask_sh),\ + SE_SF(DIG0_HDMI_GENERIC_PACKET_CONTROL0, HDMI_GENERIC1_CONT, mask_sh),\ + SE_SF(DIG0_HDMI_GENERIC_PACKET_CONTROL0, HDMI_GENERIC1_SEND, mask_sh),\ + SE_SF(DIG0_HDMI_GENERIC_PACKET_CONTROL0, HDMI_GENERIC1_LINE, mask_sh) + + +#define SE_REG_FIELD_LIST_DCN1_0(type) \ + type AFMT_GENERIC_INDEX;\ + type AFMT_GENERIC_HB0;\ + type AFMT_GENERIC_HB1;\ + type AFMT_GENERIC_HB2;\ + type AFMT_GENERIC_HB3;\ + type AFMT_GENERIC_LOCK_STATUS;\ + type AFMT_GENERIC_CONFLICT;\ + type AFMT_GENERIC_CONFLICT_CLR;\ + type AFMT_GENERIC0_FRAME_UPDATE_PENDING;\ + type AFMT_GENERIC1_FRAME_UPDATE_PENDING;\ + type AFMT_GENERIC2_FRAME_UPDATE_PENDING;\ + type AFMT_GENERIC3_FRAME_UPDATE_PENDING;\ + type AFMT_GENERIC4_FRAME_UPDATE_PENDING;\ + type AFMT_GENERIC5_FRAME_UPDATE_PENDING;\ + type AFMT_GENERIC6_FRAME_UPDATE_PENDING;\ + type AFMT_GENERIC7_FRAME_UPDATE_PENDING;\ + type AFMT_GENERIC0_FRAME_UPDATE;\ + type AFMT_GENERIC1_FRAME_UPDATE;\ + type AFMT_GENERIC2_FRAME_UPDATE;\ + type AFMT_GENERIC3_FRAME_UPDATE;\ + type AFMT_GENERIC4_FRAME_UPDATE;\ + type AFMT_GENERIC5_FRAME_UPDATE;\ + type AFMT_GENERIC6_FRAME_UPDATE;\ + type AFMT_GENERIC7_FRAME_UPDATE;\ + type HDMI_GENERIC0_CONT;\ + type HDMI_GENERIC0_SEND;\ + type HDMI_GENERIC0_LINE;\ + type HDMI_GENERIC1_CONT;\ + type HDMI_GENERIC1_SEND;\ + type HDMI_GENERIC1_LINE;\ + type HDMI_GENERIC2_CONT;\ + type HDMI_GENERIC2_SEND;\ + type HDMI_GENERIC2_LINE;\ + type HDMI_GENERIC3_CONT;\ + type HDMI_GENERIC3_SEND;\ + type HDMI_GENERIC3_LINE;\ + type HDMI_GENERIC4_CONT;\ + type HDMI_GENERIC4_SEND;\ + type HDMI_GENERIC4_LINE;\ + type HDMI_GENERIC5_CONT;\ + type HDMI_GENERIC5_SEND;\ + type HDMI_GENERIC5_LINE;\ + type HDMI_GENERIC6_CONT;\ + type HDMI_GENERIC6_SEND;\ + type HDMI_GENERIC6_LINE;\ + type HDMI_GENERIC7_CONT;\ + type HDMI_GENERIC7_SEND;\ + type HDMI_GENERIC7_LINE;\ + type DP_PIXEL_ENCODING;\ + type DP_COMPONENT_DEPTH;\ + type HDMI_PACKET_GEN_VERSION;\ + type HDMI_KEEPOUT_MODE;\ + type HDMI_DEEP_COLOR_ENABLE;\ + type HDMI_CLOCK_CHANNEL_RATE;\ + type HDMI_DEEP_COLOR_DEPTH;\ + type HDMI_GC_CONT;\ + type HDMI_GC_SEND;\ + type HDMI_NULL_SEND;\ + type HDMI_DATA_SCRAMBLE_EN;\ + type HDMI_AUDIO_INFO_SEND;\ + type AFMT_AUDIO_INFO_UPDATE;\ + type HDMI_AUDIO_INFO_LINE;\ + type HDMI_GC_AVMUTE;\ + type DP_MSE_RATE_X;\ + type DP_MSE_RATE_Y;\ + type DP_MSE_RATE_UPDATE_PENDING;\ + type DP_SEC_GSP0_ENABLE;\ + type DP_SEC_STREAM_ENABLE;\ + type DP_SEC_GSP1_ENABLE;\ + type DP_SEC_GSP2_ENABLE;\ + type DP_SEC_GSP3_ENABLE;\ + type DP_SEC_GSP4_ENABLE;\ + type DP_SEC_GSP5_ENABLE;\ + type DP_SEC_GSP6_ENABLE;\ + type DP_SEC_GSP7_ENABLE;\ + type DP_SEC_MPG_ENABLE;\ + type DP_VID_STREAM_DIS_DEFER;\ + type DP_VID_STREAM_ENABLE;\ + type DP_VID_STREAM_STATUS;\ + type DP_STEER_FIFO_RESET;\ + type DP_VID_M_N_GEN_EN;\ + type DP_VID_N;\ + type DP_VID_M;\ + type DIG_START;\ + type AFMT_AUDIO_SRC_SELECT;\ + type AFMT_AUDIO_CHANNEL_ENABLE;\ + type HDMI_AUDIO_PACKETS_PER_LINE;\ + type HDMI_AUDIO_DELAY_EN;\ + type AFMT_60958_CS_UPDATE;\ + type AFMT_AUDIO_LAYOUT_OVRD;\ + type AFMT_60958_OSF_OVRD;\ + type HDMI_ACR_AUTO_SEND;\ + type HDMI_ACR_SOURCE;\ + type HDMI_ACR_AUDIO_PRIORITY;\ + type HDMI_ACR_CTS_32;\ + type HDMI_ACR_N_32;\ + type HDMI_ACR_CTS_44;\ + type HDMI_ACR_N_44;\ + type HDMI_ACR_CTS_48;\ + type HDMI_ACR_N_48;\ + type AFMT_60958_CS_CHANNEL_NUMBER_L;\ + type AFMT_60958_CS_CLOCK_ACCURACY;\ + type AFMT_60958_CS_CHANNEL_NUMBER_R;\ + type AFMT_60958_CS_CHANNEL_NUMBER_2;\ + type AFMT_60958_CS_CHANNEL_NUMBER_3;\ + type AFMT_60958_CS_CHANNEL_NUMBER_4;\ + type AFMT_60958_CS_CHANNEL_NUMBER_5;\ + type AFMT_60958_CS_CHANNEL_NUMBER_6;\ + type AFMT_60958_CS_CHANNEL_NUMBER_7;\ + type DP_SEC_AUD_N;\ + type DP_SEC_TIMESTAMP_MODE;\ + type DP_SEC_ASP_ENABLE;\ + type DP_SEC_ATP_ENABLE;\ + type DP_SEC_AIP_ENABLE;\ + type DP_SEC_ACM_ENABLE;\ + type AFMT_AUDIO_SAMPLE_SEND;\ + type AFMT_AUDIO_CLOCK_EN;\ + type TMDS_PIXEL_ENCODING;\ + type TMDS_COLOR_FORMAT;\ + type DIG_STEREOSYNC_SELECT;\ + type DIG_STEREOSYNC_GATE_EN;\ + type DP_DB_DISABLE;\ + type DP_MSA_MISC0;\ + type DP_MSA_HTOTAL;\ + type DP_MSA_VTOTAL;\ + type DP_MSA_HSTART;\ + type DP_MSA_VSTART;\ + type DP_MSA_HSYNCWIDTH;\ + type DP_MSA_HSYNCPOLARITY;\ + type DP_MSA_VSYNCWIDTH;\ + type DP_MSA_VSYNCPOLARITY;\ + type DP_MSA_HWIDTH;\ + type DP_MSA_VHEIGHT;\ + type HDMI_DB_DISABLE;\ + type DP_VID_N_MUL;\ + type DP_VID_M_DOUBLE_VALUE_EN + +struct dcn10_stream_encoder_shift { + SE_REG_FIELD_LIST_DCN1_0(uint8_t); +}; + +struct dcn10_stream_encoder_mask { + SE_REG_FIELD_LIST_DCN1_0(uint32_t); +}; + +struct dcn10_stream_encoder { + struct stream_encoder base; + const struct dcn10_stream_enc_registers *regs; + const struct dcn10_stream_encoder_shift *se_shift; + const struct dcn10_stream_encoder_mask *se_mask; +}; + +void dcn10_stream_encoder_construct( + struct dcn10_stream_encoder *enc1, + struct dc_context *ctx, + struct dc_bios *bp, + enum engine_id eng_id, + const struct dcn10_stream_enc_registers *regs, + const struct dcn10_stream_encoder_shift *se_shift, + const struct dcn10_stream_encoder_mask *se_mask); + +void enc1_update_generic_info_packet( + struct dcn10_stream_encoder *enc1, + uint32_t packet_index, + const struct dc_info_packet *info_packet); + +void enc1_stream_encoder_dp_set_stream_attribute( + struct stream_encoder *enc, + struct dc_crtc_timing *crtc_timing, + enum dc_color_space output_color_space); + +void enc1_stream_encoder_hdmi_set_stream_attribute( + struct stream_encoder *enc, + struct dc_crtc_timing *crtc_timing, + int actual_pix_clk_khz, + bool enable_audio); + +void enc1_stream_encoder_dvi_set_stream_attribute( + struct stream_encoder *enc, + struct dc_crtc_timing *crtc_timing, + bool is_dual_link); + +void enc1_stream_encoder_set_mst_bandwidth( + struct stream_encoder *enc, + struct fixed31_32 avg_time_slots_per_mtp); + +void enc1_stream_encoder_update_dp_info_packets( + struct stream_encoder *enc, + const struct encoder_info_frame *info_frame); + +void enc1_stream_encoder_stop_dp_info_packets( + struct stream_encoder *enc); + +void enc1_stream_encoder_dp_blank( + struct stream_encoder *enc); + +void enc1_stream_encoder_dp_unblank( + struct stream_encoder *enc, + const struct encoder_unblank_param *param); + +void enc1_setup_stereo_sync( + struct stream_encoder *enc, + int tg_inst, bool enable); + +void enc1_stream_encoder_set_avmute( + struct stream_encoder *enc, + bool enable); + +void enc1_se_audio_mute_control( + struct stream_encoder *enc, + bool mute); + +void enc1_se_dp_audio_setup( + struct stream_encoder *enc, + unsigned int az_inst, + struct audio_info *info); + +void enc1_se_dp_audio_enable( + struct stream_encoder *enc); + +void enc1_se_dp_audio_disable( + struct stream_encoder *enc); + +void enc1_se_hdmi_audio_setup( + struct stream_encoder *enc, + unsigned int az_inst, + struct audio_info *info, + struct audio_crtc_info *audio_crtc_info); + +void enc1_se_hdmi_audio_disable( + struct stream_encoder *enc); + +#endif /* __DC_STREAM_ENCODER_DCN10_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/dm_services.h b/drivers/gpu/drm/amd/display/dc/dm_services.h index 22e7ee7dcd26d..4ff9b2bba1782 100644 --- a/drivers/gpu/drm/amd/display/dc/dm_services.h +++ b/drivers/gpu/drm/amd/display/dc/dm_services.h @@ -341,6 +341,10 @@ bool dm_dmcu_set_pipe(struct dc_context *ctx, unsigned int controller_id); unsigned long long dm_get_timestamp(struct dc_context *ctx); +unsigned long long dm_get_elapse_time_in_ns(struct dc_context *ctx, + unsigned long long current_time_stamp, + unsigned long long last_time_stamp); + /* * performance tracing */ @@ -351,10 +355,6 @@ void dm_perf_trace_timestamp(const char *func_name, unsigned int line); /* * Debug and verification hooks */ -bool dm_helpers_dc_conn_log( - struct dc_context *ctx, - struct log_entry *entry, - enum dc_log_type event); void dm_dtn_log_begin(struct dc_context *ctx); void dm_dtn_log_append_v(struct dc_context *ctx, const char *msg, ...); diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_enums.h b/drivers/gpu/drm/amd/display/dc/dml/display_mode_enums.h index b1ad3553f900d..47c19f8fe7d1e 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_enums.h +++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_enums.h @@ -108,4 +108,17 @@ enum output_standard { dm_std_uninitialized = 0, dm_std_cvtr2, dm_std_cvt }; +enum mpc_combine_affinity { + dm_mpc_always_when_possible, + dm_mpc_reduce_voltage, + dm_mpc_reduce_voltage_and_clocks +}; + +enum self_refresh_affinity { + dm_try_to_allow_self_refresh_and_mclk_switch, + dm_allow_self_refresh_and_mclk_switch, + dm_allow_self_refresh, + dm_neither_self_refresh_nor_mclk_switch +}; + #endif diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.c b/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.c index c109b2c34c8fc..fd9d97aab0715 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.c +++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_lib.c @@ -26,75 +26,89 @@ #include "display_mode_lib.h" #include "dc_features.h" +static const struct _vcs_dpi_ip_params_st dcn1_0_ip = { + .rob_buffer_size_kbytes = 64, + .det_buffer_size_kbytes = 164, + .dpte_buffer_size_in_pte_reqs = 42, + .dpp_output_buffer_pixels = 2560, + .opp_output_buffer_lines = 1, + .pixel_chunk_size_kbytes = 8, + .pte_enable = 1, + .pte_chunk_size_kbytes = 2, + .meta_chunk_size_kbytes = 2, + .writeback_chunk_size_kbytes = 2, + .line_buffer_size_bits = 589824, + .max_line_buffer_lines = 12, + .IsLineBufferBppFixed = 0, + .LineBufferFixedBpp = -1, + .writeback_luma_buffer_size_kbytes = 12, + .writeback_chroma_buffer_size_kbytes = 8, + .max_num_dpp = 4, + .max_num_wb = 2, + .max_dchub_pscl_bw_pix_per_clk = 4, + .max_pscl_lb_bw_pix_per_clk = 2, + .max_lb_vscl_bw_pix_per_clk = 4, + .max_vscl_hscl_bw_pix_per_clk = 4, + .max_hscl_ratio = 4, + .max_vscl_ratio = 4, + .hscl_mults = 4, + .vscl_mults = 4, + .max_hscl_taps = 8, + .max_vscl_taps = 8, + .dispclk_ramp_margin_percent = 1, + .underscan_factor = 1.10, + .min_vblank_lines = 14, + .dppclk_delay_subtotal = 90, + .dispclk_delay_subtotal = 42, + .dcfclk_cstate_latency = 10, + .max_inter_dcn_tile_repeaters = 8, + .can_vstartup_lines_exceed_vsync_plus_back_porch_lines_minus_one = 0, + .bug_forcing_LC_req_same_size_fixed = 0, +}; + +static const struct _vcs_dpi_soc_bounding_box_st dcn1_0_soc = { + .sr_exit_time_us = 9.0, + .sr_enter_plus_exit_time_us = 11.0, + .urgent_latency_us = 4.0, + .writeback_latency_us = 12.0, + .ideal_dram_bw_after_urgent_percent = 80.0, + .max_request_size_bytes = 256, + .downspread_percent = 0.5, + .dram_page_open_time_ns = 50.0, + .dram_rw_turnaround_time_ns = 17.5, + .dram_return_buffer_per_channel_bytes = 8192, + .round_trip_ping_latency_dcfclk_cycles = 128, + .urgent_out_of_order_return_per_channel_bytes = 256, + .channel_interleave_bytes = 256, + .num_banks = 8, + .num_chans = 2, + .vmm_page_size_bytes = 4096, + .dram_clock_change_latency_us = 17.0, + .writeback_dram_clock_change_latency_us = 23.0, + .return_bus_width_bytes = 64, +}; + static void set_soc_bounding_box(struct _vcs_dpi_soc_bounding_box_st *soc, enum dml_project project) { - if (project == DML_PROJECT_RAVEN1) { - soc->sr_exit_time_us = 9.0; - soc->sr_enter_plus_exit_time_us = 11.0; - soc->urgent_latency_us = 4.0; - soc->writeback_latency_us = 12.0; - soc->ideal_dram_bw_after_urgent_percent = 80.0; - soc->max_request_size_bytes = 256; - soc->downspread_percent = 0.5; - soc->dram_page_open_time_ns = 50.0; - soc->dram_rw_turnaround_time_ns = 17.5; - soc->dram_return_buffer_per_channel_bytes = 8192; - soc->round_trip_ping_latency_dcfclk_cycles = 128; - soc->urgent_out_of_order_return_per_channel_bytes = 256; - soc->channel_interleave_bytes = 256; - soc->num_banks = 8; - soc->num_chans = 2; - soc->vmm_page_size_bytes = 4096; - soc->dram_clock_change_latency_us = 17.0; - soc->writeback_dram_clock_change_latency_us = 23.0; - soc->return_bus_width_bytes = 64; - } else { - BREAK_TO_DEBUGGER(); /* Invalid Project Specified */ + switch (project) { + case DML_PROJECT_RAVEN1: + *soc = dcn1_0_soc; + break; + default: + ASSERT(0); + break; } } static void set_ip_params(struct _vcs_dpi_ip_params_st *ip, enum dml_project project) { - if (project == DML_PROJECT_RAVEN1) { - ip->rob_buffer_size_kbytes = 64; - ip->det_buffer_size_kbytes = 164; - ip->dpte_buffer_size_in_pte_reqs = 42; - ip->dpp_output_buffer_pixels = 2560; - ip->opp_output_buffer_lines = 1; - ip->pixel_chunk_size_kbytes = 8; - ip->pte_enable = 1; - ip->pte_chunk_size_kbytes = 2; - ip->meta_chunk_size_kbytes = 2; - ip->writeback_chunk_size_kbytes = 2; - ip->line_buffer_size_bits = 589824; - ip->max_line_buffer_lines = 12; - ip->IsLineBufferBppFixed = 0; - ip->LineBufferFixedBpp = -1; - ip->writeback_luma_buffer_size_kbytes = 12; - ip->writeback_chroma_buffer_size_kbytes = 8; - ip->max_num_dpp = 4; - ip->max_num_wb = 2; - ip->max_dchub_pscl_bw_pix_per_clk = 4; - ip->max_pscl_lb_bw_pix_per_clk = 2; - ip->max_lb_vscl_bw_pix_per_clk = 4; - ip->max_vscl_hscl_bw_pix_per_clk = 4; - ip->max_hscl_ratio = 4; - ip->max_vscl_ratio = 4; - ip->hscl_mults = 4; - ip->vscl_mults = 4; - ip->max_hscl_taps = 8; - ip->max_vscl_taps = 8; - ip->dispclk_ramp_margin_percent = 1; - ip->underscan_factor = 1.10; - ip->min_vblank_lines = 14; - ip->dppclk_delay_subtotal = 90; - ip->dispclk_delay_subtotal = 42; - ip->dcfclk_cstate_latency = 10; - ip->max_inter_dcn_tile_repeaters = 8; - ip->can_vstartup_lines_exceed_vsync_plus_back_porch_lines_minus_one = 0; - ip->bug_forcing_LC_req_same_size_fixed = 0; - } else { - BREAK_TO_DEBUGGER(); /* Invalid Project Specified */ + switch (project) { + case DML_PROJECT_RAVEN1: + *ip = dcn1_0_ip; + break; + default: + ASSERT(0); + break; } } diff --git a/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h b/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h index 09affa16cc43b..7fa0375939aec 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h +++ b/drivers/gpu/drm/amd/display/dc/dml/display_mode_structs.h @@ -25,39 +25,39 @@ #ifndef __DISPLAY_MODE_STRUCTS_H__ #define __DISPLAY_MODE_STRUCTS_H__ -typedef struct _vcs_dpi_voltage_scaling_st voltage_scaling_st; -typedef struct _vcs_dpi_soc_bounding_box_st soc_bounding_box_st; -typedef struct _vcs_dpi_ip_params_st ip_params_st; -typedef struct _vcs_dpi_display_pipe_source_params_st display_pipe_source_params_st; -typedef struct _vcs_dpi_display_output_params_st display_output_params_st; -typedef struct _vcs_dpi_display_bandwidth_st display_bandwidth_st; -typedef struct _vcs_dpi_scaler_ratio_depth_st scaler_ratio_depth_st; -typedef struct _vcs_dpi_scaler_taps_st scaler_taps_st; -typedef struct _vcs_dpi_display_pipe_dest_params_st display_pipe_dest_params_st; -typedef struct _vcs_dpi_display_pipe_params_st display_pipe_params_st; -typedef struct _vcs_dpi_display_clocks_and_cfg_st display_clocks_and_cfg_st; -typedef struct _vcs_dpi_display_e2e_pipe_params_st display_e2e_pipe_params_st; -typedef struct _vcs_dpi_dchub_buffer_sizing_st dchub_buffer_sizing_st; -typedef struct _vcs_dpi_watermarks_perf_st watermarks_perf_st; -typedef struct _vcs_dpi_cstate_pstate_watermarks_st cstate_pstate_watermarks_st; -typedef struct _vcs_dpi_wm_calc_pipe_params_st wm_calc_pipe_params_st; -typedef struct _vcs_dpi_vratio_pre_st vratio_pre_st; -typedef struct _vcs_dpi_display_data_rq_misc_params_st display_data_rq_misc_params_st; -typedef struct _vcs_dpi_display_data_rq_sizing_params_st display_data_rq_sizing_params_st; -typedef struct _vcs_dpi_display_data_rq_dlg_params_st display_data_rq_dlg_params_st; -typedef struct _vcs_dpi_display_cur_rq_dlg_params_st display_cur_rq_dlg_params_st; -typedef struct _vcs_dpi_display_rq_dlg_params_st display_rq_dlg_params_st; -typedef struct _vcs_dpi_display_rq_sizing_params_st display_rq_sizing_params_st; -typedef struct _vcs_dpi_display_rq_misc_params_st display_rq_misc_params_st; -typedef struct _vcs_dpi_display_rq_params_st display_rq_params_st; -typedef struct _vcs_dpi_display_dlg_regs_st display_dlg_regs_st; -typedef struct _vcs_dpi_display_ttu_regs_st display_ttu_regs_st; -typedef struct _vcs_dpi_display_data_rq_regs_st display_data_rq_regs_st; -typedef struct _vcs_dpi_display_rq_regs_st display_rq_regs_st; -typedef struct _vcs_dpi_display_dlg_sys_params_st display_dlg_sys_params_st; -typedef struct _vcs_dpi_display_dlg_prefetch_param_st display_dlg_prefetch_param_st; -typedef struct _vcs_dpi_display_pipe_clock_st display_pipe_clock_st; -typedef struct _vcs_dpi_display_arb_params_st display_arb_params_st; +typedef struct _vcs_dpi_voltage_scaling_st voltage_scaling_st; +typedef struct _vcs_dpi_soc_bounding_box_st soc_bounding_box_st; +typedef struct _vcs_dpi_ip_params_st ip_params_st; +typedef struct _vcs_dpi_display_pipe_source_params_st display_pipe_source_params_st; +typedef struct _vcs_dpi_display_output_params_st display_output_params_st; +typedef struct _vcs_dpi_display_bandwidth_st display_bandwidth_st; +typedef struct _vcs_dpi_scaler_ratio_depth_st scaler_ratio_depth_st; +typedef struct _vcs_dpi_scaler_taps_st scaler_taps_st; +typedef struct _vcs_dpi_display_pipe_dest_params_st display_pipe_dest_params_st; +typedef struct _vcs_dpi_display_pipe_params_st display_pipe_params_st; +typedef struct _vcs_dpi_display_clocks_and_cfg_st display_clocks_and_cfg_st; +typedef struct _vcs_dpi_display_e2e_pipe_params_st display_e2e_pipe_params_st; +typedef struct _vcs_dpi_dchub_buffer_sizing_st dchub_buffer_sizing_st; +typedef struct _vcs_dpi_watermarks_perf_st watermarks_perf_st; +typedef struct _vcs_dpi_cstate_pstate_watermarks_st cstate_pstate_watermarks_st; +typedef struct _vcs_dpi_wm_calc_pipe_params_st wm_calc_pipe_params_st; +typedef struct _vcs_dpi_vratio_pre_st vratio_pre_st; +typedef struct _vcs_dpi_display_data_rq_misc_params_st display_data_rq_misc_params_st; +typedef struct _vcs_dpi_display_data_rq_sizing_params_st display_data_rq_sizing_params_st; +typedef struct _vcs_dpi_display_data_rq_dlg_params_st display_data_rq_dlg_params_st; +typedef struct _vcs_dpi_display_cur_rq_dlg_params_st display_cur_rq_dlg_params_st; +typedef struct _vcs_dpi_display_rq_dlg_params_st display_rq_dlg_params_st; +typedef struct _vcs_dpi_display_rq_sizing_params_st display_rq_sizing_params_st; +typedef struct _vcs_dpi_display_rq_misc_params_st display_rq_misc_params_st; +typedef struct _vcs_dpi_display_rq_params_st display_rq_params_st; +typedef struct _vcs_dpi_display_dlg_regs_st display_dlg_regs_st; +typedef struct _vcs_dpi_display_ttu_regs_st display_ttu_regs_st; +typedef struct _vcs_dpi_display_data_rq_regs_st display_data_rq_regs_st; +typedef struct _vcs_dpi_display_rq_regs_st display_rq_regs_st; +typedef struct _vcs_dpi_display_dlg_sys_params_st display_dlg_sys_params_st; +typedef struct _vcs_dpi_display_dlg_prefetch_param_st display_dlg_prefetch_param_st; +typedef struct _vcs_dpi_display_pipe_clock_st display_pipe_clock_st; +typedef struct _vcs_dpi_display_arb_params_st display_arb_params_st; struct _vcs_dpi_voltage_scaling_st { int state; @@ -72,89 +72,107 @@ struct _vcs_dpi_voltage_scaling_st { double dppclk_mhz; }; -struct _vcs_dpi_soc_bounding_box_st { - double sr_exit_time_us; - double sr_enter_plus_exit_time_us; - double urgent_latency_us; - double writeback_latency_us; - double ideal_dram_bw_after_urgent_percent; - unsigned int max_request_size_bytes; - double downspread_percent; - double dram_page_open_time_ns; - double dram_rw_turnaround_time_ns; - double dram_return_buffer_per_channel_bytes; - double dram_channel_width_bytes; +struct _vcs_dpi_soc_bounding_box_st { + double sr_exit_time_us; + double sr_enter_plus_exit_time_us; + double urgent_latency_us; + double urgent_latency_pixel_data_only_us; + double urgent_latency_pixel_mixed_with_vm_data_us; + double urgent_latency_vm_data_only_us; + double writeback_latency_us; + double ideal_dram_bw_after_urgent_percent; + double pct_ideal_dram_sdp_bw_after_urgent_pixel_only; // PercentOfIdealDRAMFabricAndSDPPortBWReceivedAfterUrgLatencyPixelDataOnly + double pct_ideal_dram_sdp_bw_after_urgent_pixel_and_vm; + double pct_ideal_dram_sdp_bw_after_urgent_vm_only; + double max_avg_sdp_bw_use_normal_percent; + double max_avg_dram_bw_use_normal_percent; + unsigned int max_request_size_bytes; + double downspread_percent; + double dram_page_open_time_ns; + double dram_rw_turnaround_time_ns; + double dram_return_buffer_per_channel_bytes; + double dram_channel_width_bytes; double fabric_datapath_to_dcn_data_return_bytes; double dcn_downspread_percent; double dispclk_dppclk_vco_speed_mhz; double dfs_vco_period_ps; - unsigned int round_trip_ping_latency_dcfclk_cycles; - unsigned int urgent_out_of_order_return_per_channel_bytes; - unsigned int channel_interleave_bytes; - unsigned int num_banks; - unsigned int num_chans; - unsigned int vmm_page_size_bytes; - double dram_clock_change_latency_us; - double writeback_dram_clock_change_latency_us; - unsigned int return_bus_width_bytes; - unsigned int voltage_override; - double xfc_bus_transport_time_us; - double xfc_xbuf_latency_tolerance_us; + unsigned int urgent_out_of_order_return_per_channel_pixel_only_bytes; + unsigned int urgent_out_of_order_return_per_channel_pixel_and_vm_bytes; + unsigned int urgent_out_of_order_return_per_channel_vm_only_bytes; + unsigned int round_trip_ping_latency_dcfclk_cycles; + unsigned int urgent_out_of_order_return_per_channel_bytes; + unsigned int channel_interleave_bytes; + unsigned int num_banks; + unsigned int num_chans; + unsigned int vmm_page_size_bytes; + double dram_clock_change_latency_us; + double writeback_dram_clock_change_latency_us; + unsigned int return_bus_width_bytes; + unsigned int voltage_override; + double xfc_bus_transport_time_us; + double xfc_xbuf_latency_tolerance_us; + int use_urgent_burst_bw; struct _vcs_dpi_voltage_scaling_st clock_limits[7]; }; -struct _vcs_dpi_ip_params_st { - unsigned int max_inter_dcn_tile_repeaters; - unsigned int num_dsc; - unsigned int odm_capable; - unsigned int rob_buffer_size_kbytes; - unsigned int det_buffer_size_kbytes; - unsigned int dpte_buffer_size_in_pte_reqs; - unsigned int pde_proc_buffer_size_64k_reqs; - unsigned int dpp_output_buffer_pixels; - unsigned int opp_output_buffer_lines; - unsigned int pixel_chunk_size_kbytes; - unsigned char pte_enable; - unsigned int pte_chunk_size_kbytes; - unsigned int meta_chunk_size_kbytes; - unsigned int writeback_chunk_size_kbytes; - unsigned int line_buffer_size_bits; - unsigned int max_line_buffer_lines; - unsigned int writeback_luma_buffer_size_kbytes; - unsigned int writeback_chroma_buffer_size_kbytes; - unsigned int writeback_chroma_line_buffer_width_pixels; - unsigned int max_page_table_levels; - unsigned int max_num_dpp; - unsigned int max_num_otg; - unsigned int cursor_chunk_size; - unsigned int cursor_buffer_size; - unsigned int max_num_wb; - unsigned int max_dchub_pscl_bw_pix_per_clk; - unsigned int max_pscl_lb_bw_pix_per_clk; - unsigned int max_lb_vscl_bw_pix_per_clk; - unsigned int max_vscl_hscl_bw_pix_per_clk; - double max_hscl_ratio; - double max_vscl_ratio; - unsigned int hscl_mults; - unsigned int vscl_mults; - unsigned int max_hscl_taps; - unsigned int max_vscl_taps; - unsigned int xfc_supported; - unsigned int xfc_fill_constant_bytes; - double dispclk_ramp_margin_percent; - double xfc_fill_bw_overhead_percent; - double underscan_factor; - unsigned int min_vblank_lines; - unsigned int dppclk_delay_subtotal; - unsigned int dispclk_delay_subtotal; - unsigned int dcfclk_cstate_latency; - unsigned int dppclk_delay_scl; - unsigned int dppclk_delay_scl_lb_only; - unsigned int dppclk_delay_cnvc_formatter; - unsigned int dppclk_delay_cnvc_cursor; - unsigned int is_line_buffer_bpp_fixed; - unsigned int line_buffer_fixed_bpp; - unsigned int dcc_supported; +struct _vcs_dpi_ip_params_st { + bool gpuvm_enable; + bool hostvm_enable; + unsigned int gpuvm_max_page_table_levels; + unsigned int hostvm_max_page_table_levels; + unsigned int hostvm_cached_page_table_levels; + unsigned int pte_group_size_bytes; + unsigned int max_inter_dcn_tile_repeaters; + unsigned int num_dsc; + unsigned int odm_capable; + unsigned int rob_buffer_size_kbytes; + unsigned int det_buffer_size_kbytes; + unsigned int dpte_buffer_size_in_pte_reqs; + unsigned int pde_proc_buffer_size_64k_reqs; + unsigned int dpp_output_buffer_pixels; + unsigned int opp_output_buffer_lines; + unsigned int pixel_chunk_size_kbytes; + unsigned char pte_enable; + unsigned int pte_chunk_size_kbytes; + unsigned int meta_chunk_size_kbytes; + unsigned int writeback_chunk_size_kbytes; + unsigned int line_buffer_size_bits; + unsigned int max_line_buffer_lines; + unsigned int writeback_luma_buffer_size_kbytes; + unsigned int writeback_chroma_buffer_size_kbytes; + unsigned int writeback_chroma_line_buffer_width_pixels; + unsigned int max_page_table_levels; + unsigned int max_num_dpp; + unsigned int max_num_otg; + unsigned int cursor_chunk_size; + unsigned int cursor_buffer_size; + unsigned int max_num_wb; + unsigned int max_dchub_pscl_bw_pix_per_clk; + unsigned int max_pscl_lb_bw_pix_per_clk; + unsigned int max_lb_vscl_bw_pix_per_clk; + unsigned int max_vscl_hscl_bw_pix_per_clk; + double max_hscl_ratio; + double max_vscl_ratio; + unsigned int hscl_mults; + unsigned int vscl_mults; + unsigned int max_hscl_taps; + unsigned int max_vscl_taps; + unsigned int xfc_supported; + unsigned int xfc_fill_constant_bytes; + double dispclk_ramp_margin_percent; + double xfc_fill_bw_overhead_percent; + double underscan_factor; + unsigned int min_vblank_lines; + unsigned int dppclk_delay_subtotal; + unsigned int dispclk_delay_subtotal; + unsigned int dcfclk_cstate_latency; + unsigned int dppclk_delay_scl; + unsigned int dppclk_delay_scl_lb_only; + unsigned int dppclk_delay_cnvc_formatter; + unsigned int dppclk_delay_cnvc_cursor; + unsigned int is_line_buffer_bpp_fixed; + unsigned int line_buffer_fixed_bpp; + unsigned int dcc_supported; unsigned int IsLineBufferBppFixed; unsigned int LineBufferFixedBpp; @@ -169,41 +187,45 @@ struct _vcs_dpi_display_xfc_params_st { int xfc_slv_chunk_size_bytes; }; -struct _vcs_dpi_display_pipe_source_params_st { - int source_format; - unsigned char dcc; - unsigned int dcc_override; - unsigned int dcc_rate; - unsigned char dcc_use_global; - unsigned char vm; - unsigned char vm_levels_force_en; - unsigned int vm_levels_force; - int source_scan; - int sw_mode; - int macro_tile_size; - unsigned char is_display_sw; - unsigned int viewport_width; - unsigned int viewport_height; - unsigned int viewport_y_y; - unsigned int viewport_y_c; - unsigned int viewport_width_c; - unsigned int viewport_height_c; - unsigned int data_pitch; - unsigned int data_pitch_c; - unsigned int meta_pitch; - unsigned int meta_pitch_c; - unsigned int cur0_src_width; - int cur0_bpp; - unsigned int cur1_src_width; - int cur1_bpp; - int num_cursors; - unsigned char is_hsplit; - unsigned char dynamic_metadata_enable; - unsigned int dynamic_metadata_lines_before_active; - unsigned int dynamic_metadata_xmit_bytes; - unsigned int hsplit_grp; - unsigned char xfc_enable; - unsigned char xfc_slave; +struct _vcs_dpi_display_pipe_source_params_st { + int source_format; + unsigned char dcc; + unsigned int dcc_override; + unsigned int dcc_rate; + unsigned char dcc_use_global; + unsigned char vm; + bool gpuvm; // gpuvm enabled + bool hostvm; // hostvm enabled + bool gpuvm_levels_force_en; + unsigned int gpuvm_levels_force; + bool hostvm_levels_force_en; + unsigned int hostvm_levels_force; + int source_scan; + int sw_mode; + int macro_tile_size; + unsigned char is_display_sw; + unsigned int viewport_width; + unsigned int viewport_height; + unsigned int viewport_y_y; + unsigned int viewport_y_c; + unsigned int viewport_width_c; + unsigned int viewport_height_c; + unsigned int data_pitch; + unsigned int data_pitch_c; + unsigned int meta_pitch; + unsigned int meta_pitch_c; + unsigned int cur0_src_width; + int cur0_bpp; + unsigned int cur1_src_width; + int cur1_bpp; + int num_cursors; + unsigned char is_hsplit; + unsigned char dynamic_metadata_enable; + unsigned int dynamic_metadata_lines_before_active; + unsigned int dynamic_metadata_xmit_bytes; + unsigned int hsplit_grp; + unsigned char xfc_enable; + unsigned char xfc_slave; struct _vcs_dpi_display_xfc_params_st xfc_params; }; struct writeback_st { @@ -215,338 +237,339 @@ struct writeback_st { int wb_vtaps_luma; int wb_htaps_chroma; int wb_vtaps_chroma; - int wb_hratio; - int wb_vratio; -}; - -struct _vcs_dpi_display_output_params_st { - int dp_lanes; - int output_bpp; - int dsc_enable; - int wb_enable; - int opp_input_bpc; - int output_type; - int output_format; - int output_standard; - int dsc_slices; + double wb_hratio; + double wb_vratio; +}; + +struct _vcs_dpi_display_output_params_st { + int dp_lanes; + int output_bpp; + int dsc_enable; + int wb_enable; + int num_active_wb; + int output_bpc; + int output_type; + int output_format; + int output_standard; + int dsc_slices; struct writeback_st wb; }; -struct _vcs_dpi_display_bandwidth_st { - double total_bw_consumed_gbps; - double guaranteed_urgent_return_bw_gbps; -}; - -struct _vcs_dpi_scaler_ratio_depth_st { - double hscl_ratio; - double vscl_ratio; - double hscl_ratio_c; - double vscl_ratio_c; - double vinit; - double vinit_c; - double vinit_bot; - double vinit_bot_c; - int lb_depth; - int scl_enable; -}; - -struct _vcs_dpi_scaler_taps_st { - unsigned int htaps; - unsigned int vtaps; - unsigned int htaps_c; - unsigned int vtaps_c; -}; - -struct _vcs_dpi_display_pipe_dest_params_st { - unsigned int recout_width; - unsigned int recout_height; - unsigned int full_recout_width; - unsigned int full_recout_height; - unsigned int hblank_start; - unsigned int hblank_end; - unsigned int vblank_start; - unsigned int vblank_end; - unsigned int htotal; - unsigned int vtotal; - unsigned int vactive; - unsigned int hactive; - unsigned int vstartup_start; - unsigned int vupdate_offset; - unsigned int vupdate_width; - unsigned int vready_offset; - unsigned char interlaced; - unsigned char underscan; - double pixel_rate_mhz; - unsigned char synchronized_vblank_all_planes; - unsigned char otg_inst; - unsigned char odm_split_cnt; - unsigned char odm_combine; -}; - -struct _vcs_dpi_display_pipe_params_st { - display_pipe_source_params_st src; - display_pipe_dest_params_st dest; - scaler_ratio_depth_st scale_ratio_depth; - scaler_taps_st scale_taps; -}; - -struct _vcs_dpi_display_clocks_and_cfg_st { - int voltage; - double dppclk_mhz; - double refclk_mhz; - double dispclk_mhz; - double dcfclk_mhz; - double socclk_mhz; -}; - -struct _vcs_dpi_display_e2e_pipe_params_st { - display_pipe_params_st pipe; - display_output_params_st dout; - display_clocks_and_cfg_st clks_cfg; -}; - -struct _vcs_dpi_dchub_buffer_sizing_st { - unsigned int swath_width_y; - unsigned int swath_height_y; - unsigned int swath_height_c; - unsigned int detail_buffer_size_y; -}; - -struct _vcs_dpi_watermarks_perf_st { - double stutter_eff_in_active_region_percent; - double urgent_latency_supported_us; - double non_urgent_latency_supported_us; - double dram_clock_change_margin_us; - double dram_access_eff_percent; -}; - -struct _vcs_dpi_cstate_pstate_watermarks_st { - double cstate_exit_us; - double cstate_enter_plus_exit_us; - double pstate_change_us; -}; - -struct _vcs_dpi_wm_calc_pipe_params_st { - unsigned int num_dpp; - int voltage; - int output_type; - double dcfclk_mhz; - double socclk_mhz; - double dppclk_mhz; - double pixclk_mhz; - unsigned char interlace_en; - unsigned char pte_enable; - unsigned char dcc_enable; - double dcc_rate; - double bytes_per_pixel_c; - double bytes_per_pixel_y; - unsigned int swath_width_y; - unsigned int swath_height_y; - unsigned int swath_height_c; - unsigned int det_buffer_size_y; - double h_ratio; - double v_ratio; - unsigned int h_taps; - unsigned int h_total; - unsigned int v_total; - unsigned int v_active; - unsigned int e2e_index; - double display_pipe_line_delivery_time; - double read_bw; - unsigned int lines_in_det_y; - unsigned int lines_in_det_y_rounded_down_to_swath; - double full_det_buffering_time; - double dcfclk_deepsleep_mhz_per_plane; -}; - -struct _vcs_dpi_vratio_pre_st { - double vratio_pre_l; - double vratio_pre_c; -}; - -struct _vcs_dpi_display_data_rq_misc_params_st { - unsigned int full_swath_bytes; - unsigned int stored_swath_bytes; - unsigned int blk256_height; - unsigned int blk256_width; - unsigned int req_height; - unsigned int req_width; -}; - -struct _vcs_dpi_display_data_rq_sizing_params_st { - unsigned int chunk_bytes; - unsigned int min_chunk_bytes; - unsigned int meta_chunk_bytes; - unsigned int min_meta_chunk_bytes; - unsigned int mpte_group_bytes; - unsigned int dpte_group_bytes; -}; - -struct _vcs_dpi_display_data_rq_dlg_params_st { - unsigned int swath_width_ub; - unsigned int swath_height; - unsigned int req_per_swath_ub; - unsigned int meta_pte_bytes_per_frame_ub; - unsigned int dpte_req_per_row_ub; - unsigned int dpte_groups_per_row_ub; - unsigned int dpte_row_height; - unsigned int dpte_bytes_per_row_ub; - unsigned int meta_chunks_per_row_ub; - unsigned int meta_req_per_row_ub; - unsigned int meta_row_height; - unsigned int meta_bytes_per_row_ub; -}; - -struct _vcs_dpi_display_cur_rq_dlg_params_st { - unsigned char enable; - unsigned int swath_height; - unsigned int req_per_line; -}; - -struct _vcs_dpi_display_rq_dlg_params_st { - display_data_rq_dlg_params_st rq_l; - display_data_rq_dlg_params_st rq_c; - display_cur_rq_dlg_params_st rq_cur0; -}; - -struct _vcs_dpi_display_rq_sizing_params_st { - display_data_rq_sizing_params_st rq_l; - display_data_rq_sizing_params_st rq_c; -}; - -struct _vcs_dpi_display_rq_misc_params_st { - display_data_rq_misc_params_st rq_l; - display_data_rq_misc_params_st rq_c; -}; - -struct _vcs_dpi_display_rq_params_st { - unsigned char yuv420; - unsigned char yuv420_10bpc; - display_rq_misc_params_st misc; - display_rq_sizing_params_st sizing; - display_rq_dlg_params_st dlg; -}; - -struct _vcs_dpi_display_dlg_regs_st { - unsigned int refcyc_h_blank_end; - unsigned int dlg_vblank_end; - unsigned int min_dst_y_next_start; - unsigned int refcyc_per_htotal; - unsigned int refcyc_x_after_scaler; - unsigned int dst_y_after_scaler; - unsigned int dst_y_prefetch; - unsigned int dst_y_per_vm_vblank; - unsigned int dst_y_per_row_vblank; - unsigned int dst_y_per_vm_flip; - unsigned int dst_y_per_row_flip; - unsigned int ref_freq_to_pix_freq; - unsigned int vratio_prefetch; - unsigned int vratio_prefetch_c; - unsigned int refcyc_per_pte_group_vblank_l; - unsigned int refcyc_per_pte_group_vblank_c; - unsigned int refcyc_per_meta_chunk_vblank_l; - unsigned int refcyc_per_meta_chunk_vblank_c; - unsigned int refcyc_per_pte_group_flip_l; - unsigned int refcyc_per_pte_group_flip_c; - unsigned int refcyc_per_meta_chunk_flip_l; - unsigned int refcyc_per_meta_chunk_flip_c; - unsigned int dst_y_per_pte_row_nom_l; - unsigned int dst_y_per_pte_row_nom_c; - unsigned int refcyc_per_pte_group_nom_l; - unsigned int refcyc_per_pte_group_nom_c; - unsigned int dst_y_per_meta_row_nom_l; - unsigned int dst_y_per_meta_row_nom_c; - unsigned int refcyc_per_meta_chunk_nom_l; - unsigned int refcyc_per_meta_chunk_nom_c; - unsigned int refcyc_per_line_delivery_pre_l; - unsigned int refcyc_per_line_delivery_pre_c; - unsigned int refcyc_per_line_delivery_l; - unsigned int refcyc_per_line_delivery_c; - unsigned int chunk_hdl_adjust_cur0; - unsigned int chunk_hdl_adjust_cur1; - unsigned int vready_after_vcount0; - unsigned int dst_y_offset_cur0; - unsigned int dst_y_offset_cur1; - unsigned int xfc_reg_transfer_delay; - unsigned int xfc_reg_precharge_delay; - unsigned int xfc_reg_remote_surface_flip_latency; - unsigned int xfc_reg_prefetch_margin; - unsigned int dst_y_delta_drq_limit; -}; - -struct _vcs_dpi_display_ttu_regs_st { - unsigned int qos_level_low_wm; - unsigned int qos_level_high_wm; - unsigned int min_ttu_vblank; - unsigned int qos_level_flip; - unsigned int refcyc_per_req_delivery_l; - unsigned int refcyc_per_req_delivery_c; - unsigned int refcyc_per_req_delivery_cur0; - unsigned int refcyc_per_req_delivery_cur1; - unsigned int refcyc_per_req_delivery_pre_l; - unsigned int refcyc_per_req_delivery_pre_c; - unsigned int refcyc_per_req_delivery_pre_cur0; - unsigned int refcyc_per_req_delivery_pre_cur1; - unsigned int qos_level_fixed_l; - unsigned int qos_level_fixed_c; - unsigned int qos_level_fixed_cur0; - unsigned int qos_level_fixed_cur1; - unsigned int qos_ramp_disable_l; - unsigned int qos_ramp_disable_c; - unsigned int qos_ramp_disable_cur0; - unsigned int qos_ramp_disable_cur1; -}; - -struct _vcs_dpi_display_data_rq_regs_st { - unsigned int chunk_size; - unsigned int min_chunk_size; - unsigned int meta_chunk_size; - unsigned int min_meta_chunk_size; - unsigned int dpte_group_size; - unsigned int mpte_group_size; - unsigned int swath_height; - unsigned int pte_row_height_linear; -}; - -struct _vcs_dpi_display_rq_regs_st { - display_data_rq_regs_st rq_regs_l; - display_data_rq_regs_st rq_regs_c; - unsigned int drq_expansion_mode; - unsigned int prq_expansion_mode; - unsigned int mrq_expansion_mode; - unsigned int crq_expansion_mode; - unsigned int plane1_base_address; -}; - -struct _vcs_dpi_display_dlg_sys_params_st { - double t_mclk_wm_us; - double t_urg_wm_us; - double t_sr_wm_us; - double t_extra_us; - double mem_trip_us; - double t_srx_delay_us; - double deepsleep_dcfclk_mhz; - double total_flip_bw; - unsigned int total_flip_bytes; -}; - -struct _vcs_dpi_display_dlg_prefetch_param_st { - double prefetch_bw; - unsigned int flip_bytes; -}; - -struct _vcs_dpi_display_pipe_clock_st { - double dcfclk_mhz; - double dispclk_mhz; - double socclk_mhz; - double dscclk_mhz[6]; - double dppclk_mhz[6]; -}; - -struct _vcs_dpi_display_arb_params_st { - int max_req_outstanding; - int min_req_outstanding; - int sat_level_us; +struct _vcs_dpi_display_bandwidth_st { + double total_bw_consumed_gbps; + double guaranteed_urgent_return_bw_gbps; +}; + +struct _vcs_dpi_scaler_ratio_depth_st { + double hscl_ratio; + double vscl_ratio; + double hscl_ratio_c; + double vscl_ratio_c; + double vinit; + double vinit_c; + double vinit_bot; + double vinit_bot_c; + int lb_depth; + int scl_enable; +}; + +struct _vcs_dpi_scaler_taps_st { + unsigned int htaps; + unsigned int vtaps; + unsigned int htaps_c; + unsigned int vtaps_c; +}; + +struct _vcs_dpi_display_pipe_dest_params_st { + unsigned int recout_width; + unsigned int recout_height; + unsigned int full_recout_width; + unsigned int full_recout_height; + unsigned int hblank_start; + unsigned int hblank_end; + unsigned int vblank_start; + unsigned int vblank_end; + unsigned int htotal; + unsigned int vtotal; + unsigned int vactive; + unsigned int hactive; + unsigned int vstartup_start; + unsigned int vupdate_offset; + unsigned int vupdate_width; + unsigned int vready_offset; + unsigned char interlaced; + unsigned char underscan; + double pixel_rate_mhz; + unsigned char synchronized_vblank_all_planes; + unsigned char otg_inst; + unsigned char odm_split_cnt; + unsigned char odm_combine; +}; + +struct _vcs_dpi_display_pipe_params_st { + display_pipe_source_params_st src; + display_pipe_dest_params_st dest; + scaler_ratio_depth_st scale_ratio_depth; + scaler_taps_st scale_taps; +}; + +struct _vcs_dpi_display_clocks_and_cfg_st { + int voltage; + double dppclk_mhz; + double refclk_mhz; + double dispclk_mhz; + double dcfclk_mhz; + double socclk_mhz; +}; + +struct _vcs_dpi_display_e2e_pipe_params_st { + display_pipe_params_st pipe; + display_output_params_st dout; + display_clocks_and_cfg_st clks_cfg; +}; + +struct _vcs_dpi_dchub_buffer_sizing_st { + unsigned int swath_width_y; + unsigned int swath_height_y; + unsigned int swath_height_c; + unsigned int detail_buffer_size_y; +}; + +struct _vcs_dpi_watermarks_perf_st { + double stutter_eff_in_active_region_percent; + double urgent_latency_supported_us; + double non_urgent_latency_supported_us; + double dram_clock_change_margin_us; + double dram_access_eff_percent; +}; + +struct _vcs_dpi_cstate_pstate_watermarks_st { + double cstate_exit_us; + double cstate_enter_plus_exit_us; + double pstate_change_us; +}; + +struct _vcs_dpi_wm_calc_pipe_params_st { + unsigned int num_dpp; + int voltage; + int output_type; + double dcfclk_mhz; + double socclk_mhz; + double dppclk_mhz; + double pixclk_mhz; + unsigned char interlace_en; + unsigned char pte_enable; + unsigned char dcc_enable; + double dcc_rate; + double bytes_per_pixel_c; + double bytes_per_pixel_y; + unsigned int swath_width_y; + unsigned int swath_height_y; + unsigned int swath_height_c; + unsigned int det_buffer_size_y; + double h_ratio; + double v_ratio; + unsigned int h_taps; + unsigned int h_total; + unsigned int v_total; + unsigned int v_active; + unsigned int e2e_index; + double display_pipe_line_delivery_time; + double read_bw; + unsigned int lines_in_det_y; + unsigned int lines_in_det_y_rounded_down_to_swath; + double full_det_buffering_time; + double dcfclk_deepsleep_mhz_per_plane; +}; + +struct _vcs_dpi_vratio_pre_st { + double vratio_pre_l; + double vratio_pre_c; +}; + +struct _vcs_dpi_display_data_rq_misc_params_st { + unsigned int full_swath_bytes; + unsigned int stored_swath_bytes; + unsigned int blk256_height; + unsigned int blk256_width; + unsigned int req_height; + unsigned int req_width; +}; + +struct _vcs_dpi_display_data_rq_sizing_params_st { + unsigned int chunk_bytes; + unsigned int min_chunk_bytes; + unsigned int meta_chunk_bytes; + unsigned int min_meta_chunk_bytes; + unsigned int mpte_group_bytes; + unsigned int dpte_group_bytes; +}; + +struct _vcs_dpi_display_data_rq_dlg_params_st { + unsigned int swath_width_ub; + unsigned int swath_height; + unsigned int req_per_swath_ub; + unsigned int meta_pte_bytes_per_frame_ub; + unsigned int dpte_req_per_row_ub; + unsigned int dpte_groups_per_row_ub; + unsigned int dpte_row_height; + unsigned int dpte_bytes_per_row_ub; + unsigned int meta_chunks_per_row_ub; + unsigned int meta_req_per_row_ub; + unsigned int meta_row_height; + unsigned int meta_bytes_per_row_ub; +}; + +struct _vcs_dpi_display_cur_rq_dlg_params_st { + unsigned char enable; + unsigned int swath_height; + unsigned int req_per_line; +}; + +struct _vcs_dpi_display_rq_dlg_params_st { + display_data_rq_dlg_params_st rq_l; + display_data_rq_dlg_params_st rq_c; + display_cur_rq_dlg_params_st rq_cur0; +}; + +struct _vcs_dpi_display_rq_sizing_params_st { + display_data_rq_sizing_params_st rq_l; + display_data_rq_sizing_params_st rq_c; +}; + +struct _vcs_dpi_display_rq_misc_params_st { + display_data_rq_misc_params_st rq_l; + display_data_rq_misc_params_st rq_c; +}; + +struct _vcs_dpi_display_rq_params_st { + unsigned char yuv420; + unsigned char yuv420_10bpc; + display_rq_misc_params_st misc; + display_rq_sizing_params_st sizing; + display_rq_dlg_params_st dlg; +}; + +struct _vcs_dpi_display_dlg_regs_st { + unsigned int refcyc_h_blank_end; + unsigned int dlg_vblank_end; + unsigned int min_dst_y_next_start; + unsigned int refcyc_per_htotal; + unsigned int refcyc_x_after_scaler; + unsigned int dst_y_after_scaler; + unsigned int dst_y_prefetch; + unsigned int dst_y_per_vm_vblank; + unsigned int dst_y_per_row_vblank; + unsigned int dst_y_per_vm_flip; + unsigned int dst_y_per_row_flip; + unsigned int ref_freq_to_pix_freq; + unsigned int vratio_prefetch; + unsigned int vratio_prefetch_c; + unsigned int refcyc_per_pte_group_vblank_l; + unsigned int refcyc_per_pte_group_vblank_c; + unsigned int refcyc_per_meta_chunk_vblank_l; + unsigned int refcyc_per_meta_chunk_vblank_c; + unsigned int refcyc_per_pte_group_flip_l; + unsigned int refcyc_per_pte_group_flip_c; + unsigned int refcyc_per_meta_chunk_flip_l; + unsigned int refcyc_per_meta_chunk_flip_c; + unsigned int dst_y_per_pte_row_nom_l; + unsigned int dst_y_per_pte_row_nom_c; + unsigned int refcyc_per_pte_group_nom_l; + unsigned int refcyc_per_pte_group_nom_c; + unsigned int dst_y_per_meta_row_nom_l; + unsigned int dst_y_per_meta_row_nom_c; + unsigned int refcyc_per_meta_chunk_nom_l; + unsigned int refcyc_per_meta_chunk_nom_c; + unsigned int refcyc_per_line_delivery_pre_l; + unsigned int refcyc_per_line_delivery_pre_c; + unsigned int refcyc_per_line_delivery_l; + unsigned int refcyc_per_line_delivery_c; + unsigned int chunk_hdl_adjust_cur0; + unsigned int chunk_hdl_adjust_cur1; + unsigned int vready_after_vcount0; + unsigned int dst_y_offset_cur0; + unsigned int dst_y_offset_cur1; + unsigned int xfc_reg_transfer_delay; + unsigned int xfc_reg_precharge_delay; + unsigned int xfc_reg_remote_surface_flip_latency; + unsigned int xfc_reg_prefetch_margin; + unsigned int dst_y_delta_drq_limit; +}; + +struct _vcs_dpi_display_ttu_regs_st { + unsigned int qos_level_low_wm; + unsigned int qos_level_high_wm; + unsigned int min_ttu_vblank; + unsigned int qos_level_flip; + unsigned int refcyc_per_req_delivery_l; + unsigned int refcyc_per_req_delivery_c; + unsigned int refcyc_per_req_delivery_cur0; + unsigned int refcyc_per_req_delivery_cur1; + unsigned int refcyc_per_req_delivery_pre_l; + unsigned int refcyc_per_req_delivery_pre_c; + unsigned int refcyc_per_req_delivery_pre_cur0; + unsigned int refcyc_per_req_delivery_pre_cur1; + unsigned int qos_level_fixed_l; + unsigned int qos_level_fixed_c; + unsigned int qos_level_fixed_cur0; + unsigned int qos_level_fixed_cur1; + unsigned int qos_ramp_disable_l; + unsigned int qos_ramp_disable_c; + unsigned int qos_ramp_disable_cur0; + unsigned int qos_ramp_disable_cur1; +}; + +struct _vcs_dpi_display_data_rq_regs_st { + unsigned int chunk_size; + unsigned int min_chunk_size; + unsigned int meta_chunk_size; + unsigned int min_meta_chunk_size; + unsigned int dpte_group_size; + unsigned int mpte_group_size; + unsigned int swath_height; + unsigned int pte_row_height_linear; +}; + +struct _vcs_dpi_display_rq_regs_st { + display_data_rq_regs_st rq_regs_l; + display_data_rq_regs_st rq_regs_c; + unsigned int drq_expansion_mode; + unsigned int prq_expansion_mode; + unsigned int mrq_expansion_mode; + unsigned int crq_expansion_mode; + unsigned int plane1_base_address; +}; + +struct _vcs_dpi_display_dlg_sys_params_st { + double t_mclk_wm_us; + double t_urg_wm_us; + double t_sr_wm_us; + double t_extra_us; + double mem_trip_us; + double t_srx_delay_us; + double deepsleep_dcfclk_mhz; + double total_flip_bw; + unsigned int total_flip_bytes; +}; + +struct _vcs_dpi_display_dlg_prefetch_param_st { + double prefetch_bw; + unsigned int flip_bytes; +}; + +struct _vcs_dpi_display_pipe_clock_st { + double dcfclk_mhz; + double dispclk_mhz; + double socclk_mhz; + double dscclk_mhz[6]; + double dppclk_mhz[6]; +}; + +struct _vcs_dpi_display_arb_params_st { + int max_req_outstanding; + int min_req_outstanding; + int sat_level_us; }; #endif /*__DISPLAY_MODE_STRUCTS_H__*/ diff --git a/drivers/gpu/drm/amd/display/dc/dml/dml_inline_defs.h b/drivers/gpu/drm/amd/display/dc/dml/dml_inline_defs.h index f9cf083579899..e8ce08567cd8e 100644 --- a/drivers/gpu/drm/amd/display/dc/dml/dml_inline_defs.h +++ b/drivers/gpu/drm/amd/display/dc/dml/dml_inline_defs.h @@ -35,6 +35,16 @@ static inline double dml_min(double a, double b) return (double) dcn_bw_min2(a, b); } +static inline double dml_min3(double a, double b, double c) +{ + return dml_min(dml_min(a, b), c); +} + +static inline double dml_min4(double a, double b, double c, double d) +{ + return dml_min(dml_min(a, b), dml_min(c, d)); +} + static inline double dml_max(double a, double b) { return (double) dcn_bw_max2(a, b); diff --git a/drivers/gpu/drm/amd/display/dc/gpio/hw_factory.c b/drivers/gpu/drm/amd/display/dc/gpio/hw_factory.c index 87b580fa4bc90..0caee3523017f 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/hw_factory.c +++ b/drivers/gpu/drm/amd/display/dc/gpio/hw_factory.c @@ -75,6 +75,7 @@ bool dal_hw_factory_init( return true; case DCE_VERSION_11_0: case DCE_VERSION_11_2: + case DCE_VERSION_11_22: dal_hw_factory_dce110_init(factory); return true; case DCE_VERSION_12_0: diff --git a/drivers/gpu/drm/amd/display/dc/gpio/hw_translate.c b/drivers/gpu/drm/amd/display/dc/gpio/hw_translate.c index 0ae8ace257391..55c7074885413 100644 --- a/drivers/gpu/drm/amd/display/dc/gpio/hw_translate.c +++ b/drivers/gpu/drm/amd/display/dc/gpio/hw_translate.c @@ -72,6 +72,7 @@ bool dal_hw_translate_init( case DCE_VERSION_10_0: case DCE_VERSION_11_0: case DCE_VERSION_11_2: + case DCE_VERSION_11_22: dal_hw_translate_dce110_init(translate); return true; case DCE_VERSION_12_0: diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_hw_engine_dce110.c b/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_hw_engine_dce110.c index abd0095ced30a..b7256f595052b 100644 --- a/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_hw_engine_dce110.c +++ b/drivers/gpu/drm/amd/display/dc/i2caux/dce110/i2c_hw_engine_dce110.c @@ -527,7 +527,7 @@ static void construct( REG_GET(MICROSECOND_TIME_BASE_DIV, XTAL_REF_DIV, &xtal_ref_div); if (xtal_ref_div == 0) { - DC_LOG_WARNING("Invalid base timer divider\n", + DC_LOG_WARNING("Invalid base timer divider [%s]\n", __func__); xtal_ref_div = 2; } diff --git a/drivers/gpu/drm/amd/display/dc/i2caux/i2caux.c b/drivers/gpu/drm/amd/display/dc/i2caux/i2caux.c index 5cbf6626b8d40..14dc8c94d862b 100644 --- a/drivers/gpu/drm/amd/display/dc/i2caux/i2caux.c +++ b/drivers/gpu/drm/amd/display/dc/i2caux/i2caux.c @@ -83,6 +83,7 @@ struct i2caux *dal_i2caux_create( case DCE_VERSION_8_3: return dal_i2caux_dce80_create(ctx); case DCE_VERSION_11_2: + case DCE_VERSION_11_22: return dal_i2caux_dce112_create(ctx); case DCE_VERSION_11_0: return dal_i2caux_dce110_create(ctx); diff --git a/drivers/gpu/drm/amd/display/dc/inc/core_types.h b/drivers/gpu/drm/amd/display/dc/inc/core_types.h index 8c51ad70caceb..a94942d4e66b9 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/core_types.h +++ b/drivers/gpu/drm/amd/display/dc/inc/core_types.h @@ -95,11 +95,6 @@ struct resource_funcs { struct link_encoder *(*link_enc_create)( const struct encoder_init_data *init); - enum dc_status (*validate_guaranteed)( - struct dc *dc, - struct dc_stream_state *stream, - struct dc_state *context); - bool (*validate_bandwidth)( struct dc *dc, struct dc_state *context); @@ -250,6 +245,7 @@ struct dce_bw_output { bool all_displays_in_sync; struct dce_watermarks urgent_wm_ns[MAX_PIPES]; struct dce_watermarks stutter_exit_wm_ns[MAX_PIPES]; + struct dce_watermarks stutter_entry_wm_ns[MAX_PIPES]; struct dce_watermarks nbp_state_change_wm_ns[MAX_PIPES]; int sclk_khz; int sclk_deep_sleep_khz; diff --git a/drivers/gpu/drm/amd/display/dc/inc/dc_link_ddc.h b/drivers/gpu/drm/amd/display/dc/inc/dc_link_ddc.h index 090b7a8dd67bd..30b3a08b91be2 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/dc_link_ddc.h +++ b/drivers/gpu/drm/amd/display/dc/inc/dc_link_ddc.h @@ -102,13 +102,14 @@ bool dal_ddc_service_query_ddc_data( uint8_t *read_buf, uint32_t read_size); -ssize_t dal_ddc_service_read_dpcd_data( +enum ddc_result dal_ddc_service_read_dpcd_data( struct ddc_service *ddc, bool i2c, enum i2c_mot_mode mot, uint32_t address, uint8_t *data, - uint32_t len); + uint32_t len, + uint32_t *read); enum ddc_result dal_ddc_service_write_dpcd_data( struct ddc_service *ddc, diff --git a/drivers/gpu/drm/amd/display/dc/inc/dce_calcs.h b/drivers/gpu/drm/amd/display/dc/inc/dce_calcs.h index a9bfe9ff8ce66..eece165206f9e 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/dce_calcs.h +++ b/drivers/gpu/drm/amd/display/dc/inc/dce_calcs.h @@ -42,6 +42,8 @@ enum bw_calcs_version { BW_CALCS_VERSION_CARRIZO, BW_CALCS_VERSION_POLARIS10, BW_CALCS_VERSION_POLARIS11, + BW_CALCS_VERSION_POLARIS12, + BW_CALCS_VERSION_VEGAM, BW_CALCS_VERSION_STONEY, BW_CALCS_VERSION_VEGA10 }; diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/dchubbub.h b/drivers/gpu/drm/amd/display/dc/inc/hw/dchubbub.h new file mode 100644 index 0000000000000..02f757dd70d44 --- /dev/null +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/dchubbub.h @@ -0,0 +1,64 @@ +/* + * Copyright 2012-15 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + * Authors: AMD + * + */ + +#ifndef __DAL_DCHUBBUB_H__ +#define __DAL_DCHUBBUB_H__ + + +enum dcc_control { + dcc_control__256_256_xxx, + dcc_control__128_128_xxx, + dcc_control__256_64_64, +}; + +enum segment_order { + segment_order__na, + segment_order__contiguous, + segment_order__non_contiguous, +}; + + +struct hubbub_funcs { + void (*update_dchub)( + struct hubbub *hubbub, + struct dchub_init_data *dh_data); + + bool (*get_dcc_compression_cap)(struct hubbub *hubbub, + const struct dc_dcc_surface_param *input, + struct dc_surface_dcc_cap *output); + + bool (*dcc_support_swizzle)( + enum swizzle_mode_values swizzle, + unsigned int bytes_per_element, + enum segment_order *segment_order_horz, + enum segment_order *segment_order_vert); + + bool (*dcc_support_pixel_format)( + enum surface_pixel_format format, + unsigned int *bytes_per_element); +}; + + +#endif diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/dpp.h b/drivers/gpu/drm/amd/display/dc/inc/hw/dpp.h index 99995608b6200..582458f028f84 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/dpp.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/dpp.h @@ -44,7 +44,23 @@ struct dpp_grph_csc_adjustment { enum graphics_gamut_adjust_type gamut_adjust_type; }; +struct dcn_dpp_state { + uint32_t igam_lut_mode; + uint32_t igam_input_format; + uint32_t dgam_lut_mode; + uint32_t rgam_lut_mode; + uint32_t gamut_remap_mode; + uint32_t gamut_remap_c11_c12; + uint32_t gamut_remap_c13_c14; + uint32_t gamut_remap_c21_c22; + uint32_t gamut_remap_c23_c24; + uint32_t gamut_remap_c31_c32; + uint32_t gamut_remap_c33_c34; +}; + struct dpp_funcs { + void (*dpp_read_state)(struct dpp *dpp, struct dcn_dpp_state *s); + void (*dpp_reset)(struct dpp *dpp); void (*dpp_set_scaler)(struct dpp *dpp, @@ -117,7 +133,7 @@ struct dpp_funcs { struct dpp *dpp_base, enum surface_pixel_format format, enum expansion_mode mode, - struct csc_transform input_csc_color_matrix, + struct dc_csc_transform input_csc_color_matrix, enum dc_color_space input_color_space); void (*dpp_full_bypass)(struct dpp *dpp_base); diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h b/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h index 9ced254e652c9..97df82cddf829 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/hubp.h @@ -56,7 +56,6 @@ struct hubp { bool power_gated; }; - struct hubp_funcs { void (*hubp_setup)( struct hubp *hubp, @@ -121,6 +120,9 @@ struct hubp_funcs { void (*hubp_clk_cntl)(struct hubp *hubp, bool enable); void (*hubp_vtg_sel)(struct hubp *hubp, uint32_t otg_inst); + void (*hubp_read_state)(struct hubp *hubp); + void (*hubp_disable_control)(struct hubp *hubp, bool disable_hubp); + unsigned int (*hubp_get_underflow_status)(struct hubp *hubp); }; diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h b/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h index b22158190262f..cf7433ebf91a0 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/hw_shared.h @@ -140,11 +140,6 @@ enum opp_regamma { OPP_REGAMMA_USER }; -struct csc_transform { - uint16_t matrix[12]; - bool enable_adjustment; -}; - struct dc_bias_and_scale { uint16_t scale_red; uint16_t bias_red; @@ -191,4 +186,9 @@ enum controller_dp_test_pattern { CONTROLLER_DP_TEST_PATTERN_COLORSQUARES_CEA }; +enum dc_lut_mode { + LUT_BYPASS, + LUT_RAM_A, + LUT_RAM_B +}; #endif /* __DAL_HW_SHARED_H__ */ diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/ipp.h b/drivers/gpu/drm/amd/display/dc/inc/hw/ipp.h index 2109eac20a3d3..b2fa4c4cd9206 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/ipp.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/ipp.h @@ -87,7 +87,7 @@ struct ipp_funcs { struct input_pixel_processor *ipp, enum surface_pixel_format format, enum expansion_mode mode, - struct csc_transform input_csc_color_matrix, + struct dc_csc_transform input_csc_color_matrix, enum dc_color_space input_color_space); /* DCE function to setup IPP. TODO: see if we can consolidate to setup */ diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/link_encoder.h b/drivers/gpu/drm/amd/display/dc/inc/hw/link_encoder.h index 54d8a13861423..cf6df2e7beb2d 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/link_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/link_encoder.h @@ -149,6 +149,7 @@ struct link_encoder_funcs { bool connect); void (*enable_hpd)(struct link_encoder *enc); void (*disable_hpd)(struct link_encoder *enc); + bool (*is_dig_enabled)(struct link_encoder *enc); void (*destroy)(struct link_encoder **enc); }; diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/mem_input.h b/drivers/gpu/drm/amd/display/dc/inc/hw/mem_input.h index 3e1e7e6a8792d..47f1dc5a43b74 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/mem_input.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/mem_input.h @@ -104,6 +104,7 @@ struct mem_input_funcs { struct mem_input *mem_input, struct dce_watermarks nbp, struct dce_watermarks stutter, + struct dce_watermarks stutter_enter, struct dce_watermarks urgent, uint32_t total_dest_line_time_ns); diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h b/drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h index 23a8d5e53a89d..caf74e3c836f5 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/mpc.h @@ -105,7 +105,24 @@ struct mpc { struct mpcc mpcc_array[MAX_MPCC]; }; +struct mpcc_state { + uint32_t opp_id; + uint32_t dpp_id; + uint32_t bot_mpcc_id; + uint32_t mode; + uint32_t alpha_mode; + uint32_t pre_multiplied_alpha; + uint32_t overlap_only; + uint32_t idle; + uint32_t busy; +}; + struct mpc_funcs { + void (*read_mpcc_state)( + struct mpc *mpc, + int mpcc_inst, + struct mpcc_state *s); + /* * Insert DPP into MPC tree based on specified blending position. * Only used for planes that are part of blending chain for OPP output diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h b/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h index b5db1692393c9..cfa7ec9517ae1 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/stream_encoder.h @@ -29,31 +29,40 @@ #define STREAM_ENCODER_H_ #include "audio_types.h" +#include "hw_shared.h" struct dc_bios; struct dc_context; struct dc_crtc_timing; -struct encoder_info_packet { - bool valid; - uint8_t hb0; - uint8_t hb1; - uint8_t hb2; - uint8_t hb3; - uint8_t sb[32]; +enum dp_pixel_encoding_type { + DP_PIXEL_ENCODING_TYPE_RGB444 = 0x00000000, + DP_PIXEL_ENCODING_TYPE_YCBCR422 = 0x00000001, + DP_PIXEL_ENCODING_TYPE_YCBCR444 = 0x00000002, + DP_PIXEL_ENCODING_TYPE_RGB_WIDE_GAMUT = 0x00000003, + DP_PIXEL_ENCODING_TYPE_Y_ONLY = 0x00000004, + DP_PIXEL_ENCODING_TYPE_YCBCR420 = 0x00000005 +}; + +enum dp_component_depth { + DP_COMPONENT_PIXEL_DEPTH_6BPC = 0x00000000, + DP_COMPONENT_PIXEL_DEPTH_8BPC = 0x00000001, + DP_COMPONENT_PIXEL_DEPTH_10BPC = 0x00000002, + DP_COMPONENT_PIXEL_DEPTH_12BPC = 0x00000003, + DP_COMPONENT_PIXEL_DEPTH_16BPC = 0x00000004 }; struct encoder_info_frame { /* auxiliary video information */ - struct encoder_info_packet avi; - struct encoder_info_packet gamut; - struct encoder_info_packet vendor; + struct dc_info_packet avi; + struct dc_info_packet gamut; + struct dc_info_packet vendor; /* source product description */ - struct encoder_info_packet spd; + struct dc_info_packet spd; /* video stream configuration */ - struct encoder_info_packet vsc; + struct dc_info_packet vsc; /* HDR Static MetaData */ - struct encoder_info_packet hdrsmd; + struct dc_info_packet hdrsmd; }; struct encoder_unblank_param { @@ -147,6 +156,7 @@ struct stream_encoder_funcs { void (*set_avmute)( struct stream_encoder *enc, bool enable); + }; #endif /* STREAM_ENCODER_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h b/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h index 3217b5bf6c7a3..69cb0a1053000 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/timing_generator.h @@ -140,6 +140,9 @@ struct timing_generator_funcs { void (*program_timing)(struct timing_generator *tg, const struct dc_crtc_timing *timing, bool use_vbios); + void (*program_vline_interrupt)(struct timing_generator *optc, + const struct dc_crtc_timing *dc_crtc_timing, + unsigned long long vsync_delta); bool (*enable_crtc)(struct timing_generator *tg); bool (*disable_crtc)(struct timing_generator *tg); bool (*is_counter_moving)(struct timing_generator *tg); diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw/transform.h b/drivers/gpu/drm/amd/display/dc/inc/hw/transform.h index c5b3623bcbd95..fecc80c47c267 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw/transform.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw/transform.h @@ -252,7 +252,7 @@ struct transform_funcs { struct transform *xfm_base, enum surface_pixel_format format, enum expansion_mode mode, - struct csc_transform input_csc_color_matrix, + struct dc_csc_transform input_csc_color_matrix, enum dc_color_space input_color_space); void (*ipp_full_bypass)(struct transform *xfm_base); diff --git a/drivers/gpu/drm/amd/display/dc/inc/hw_sequencer.h b/drivers/gpu/drm/amd/display/dc/inc/hw_sequencer.h index e764cbad881b7..63fc6c4997892 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/hw_sequencer.h +++ b/drivers/gpu/drm/amd/display/dc/inc/hw_sequencer.h @@ -32,6 +32,8 @@ #include "inc/hw/link_encoder.h" #include "core_status.h" +#define EDP_BACKLIGHT_RAMP_DISABLE_LEVEL 0xFFFFFFFF + enum pipe_gating_control { PIPE_GATING_CONTROL_DISABLE = 0, PIPE_GATING_CONTROL_ENABLE, @@ -63,6 +65,7 @@ struct dchub_init_data; struct dc_static_screen_events; struct resource_pool; struct resource_context; +struct stream_resource; struct hw_sequencer_funcs { @@ -80,11 +83,6 @@ struct hw_sequencer_funcs { int num_planes, struct dc_state *context); - void (*set_plane_config)( - const struct dc *dc, - struct pipe_ctx *pipe_ctx, - struct resource_context *res_ctx); - void (*program_gamut_remap)( struct pipe_ctx *pipe_ctx); @@ -93,6 +91,12 @@ struct hw_sequencer_funcs { enum dc_color_space colorspace, uint16_t *matrix); + void (*program_output_csc)(struct dc *dc, + struct pipe_ctx *pipe_ctx, + enum dc_color_space colorspace, + uint16_t *matrix, + int opp_id); + void (*update_plane_addr)( const struct dc *dc, struct pipe_ctx *pipe_ctx); @@ -154,6 +158,11 @@ struct hw_sequencer_funcs { struct dc *dc, struct pipe_ctx *pipe, bool lock); + void (*blank_pixel_data)( + struct dc *dc, + struct stream_resource *stream_res, + struct dc_stream_state *stream, + bool blank); void (*set_bandwidth)( struct dc *dc, @@ -169,7 +178,7 @@ struct hw_sequencer_funcs { void (*set_static_screen_control)(struct pipe_ctx **pipe_ctx, int num_pipes, const struct dc_static_screen_events *events); - enum dc_status (*prog_pixclk_crtc_otg)( + enum dc_status (*enable_stream_timing)( struct pipe_ctx *pipe_ctx, struct dc_state *context, struct dc *dc); @@ -201,6 +210,7 @@ struct hw_sequencer_funcs { void (*set_cursor_position)(struct pipe_ctx *pipe); void (*set_cursor_attribute)(struct pipe_ctx *pipe); + }; void color_space_to_black_color( diff --git a/drivers/gpu/drm/amd/display/dc/inc/reg_helper.h b/drivers/gpu/drm/amd/display/dc/inc/reg_helper.h index 77eb72874e901..3306e7b0b3e34 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/reg_helper.h +++ b/drivers/gpu/drm/amd/display/dc/inc/reg_helper.h @@ -183,6 +183,36 @@ FN(reg_name, f4), v4, \ FN(reg_name, f5), v5) +#define REG_GET_6(reg_name, f1, v1, f2, v2, f3, v3, f4, v4, f5, v5, f6, v6) \ + generic_reg_get6(CTX, REG(reg_name), \ + FN(reg_name, f1), v1, \ + FN(reg_name, f2), v2, \ + FN(reg_name, f3), v3, \ + FN(reg_name, f4), v4, \ + FN(reg_name, f5), v5, \ + FN(reg_name, f6), v6) + +#define REG_GET_7(reg_name, f1, v1, f2, v2, f3, v3, f4, v4, f5, v5, f6, v6, f7, v7) \ + generic_reg_get7(CTX, REG(reg_name), \ + FN(reg_name, f1), v1, \ + FN(reg_name, f2), v2, \ + FN(reg_name, f3), v3, \ + FN(reg_name, f4), v4, \ + FN(reg_name, f5), v5, \ + FN(reg_name, f6), v6, \ + FN(reg_name, f7), v7) + +#define REG_GET_8(reg_name, f1, v1, f2, v2, f3, v3, f4, v4, f5, v5, f6, v6, f7, v7, f8, v8) \ + generic_reg_get8(CTX, REG(reg_name), \ + FN(reg_name, f1), v1, \ + FN(reg_name, f2), v2, \ + FN(reg_name, f3), v3, \ + FN(reg_name, f4), v4, \ + FN(reg_name, f5), v5, \ + FN(reg_name, f6), v6, \ + FN(reg_name, f7), v7, \ + FN(reg_name, f8), v8) + /* macro to poll and wait for a register field to read back given value */ #define REG_WAIT(reg_name, field, val, delay_between_poll_us, max_try) \ @@ -389,4 +419,30 @@ uint32_t generic_reg_get5(const struct dc_context *ctx, uint32_t addr, uint8_t shift4, uint32_t mask4, uint32_t *field_value4, uint8_t shift5, uint32_t mask5, uint32_t *field_value5); +uint32_t generic_reg_get6(const struct dc_context *ctx, uint32_t addr, + uint8_t shift1, uint32_t mask1, uint32_t *field_value1, + uint8_t shift2, uint32_t mask2, uint32_t *field_value2, + uint8_t shift3, uint32_t mask3, uint32_t *field_value3, + uint8_t shift4, uint32_t mask4, uint32_t *field_value4, + uint8_t shift5, uint32_t mask5, uint32_t *field_value5, + uint8_t shift6, uint32_t mask6, uint32_t *field_value6); + +uint32_t generic_reg_get7(const struct dc_context *ctx, uint32_t addr, + uint8_t shift1, uint32_t mask1, uint32_t *field_value1, + uint8_t shift2, uint32_t mask2, uint32_t *field_value2, + uint8_t shift3, uint32_t mask3, uint32_t *field_value3, + uint8_t shift4, uint32_t mask4, uint32_t *field_value4, + uint8_t shift5, uint32_t mask5, uint32_t *field_value5, + uint8_t shift6, uint32_t mask6, uint32_t *field_value6, + uint8_t shift7, uint32_t mask7, uint32_t *field_value7); + +uint32_t generic_reg_get8(const struct dc_context *ctx, uint32_t addr, + uint8_t shift1, uint32_t mask1, uint32_t *field_value1, + uint8_t shift2, uint32_t mask2, uint32_t *field_value2, + uint8_t shift3, uint32_t mask3, uint32_t *field_value3, + uint8_t shift4, uint32_t mask4, uint32_t *field_value4, + uint8_t shift5, uint32_t mask5, uint32_t *field_value5, + uint8_t shift6, uint32_t mask6, uint32_t *field_value6, + uint8_t shift7, uint32_t mask7, uint32_t *field_value7, + uint8_t shift8, uint32_t mask8, uint32_t *field_value8); #endif /* DRIVERS_GPU_DRM_AMD_DC_DEV_DC_INC_REG_HELPER_H_ */ diff --git a/drivers/gpu/drm/amd/display/dc/inc/resource.h b/drivers/gpu/drm/amd/display/dc/inc/resource.h index 5467332faf7b0..640a647f4611f 100644 --- a/drivers/gpu/drm/amd/display/dc/inc/resource.h +++ b/drivers/gpu/drm/amd/display/dc/inc/resource.h @@ -139,10 +139,6 @@ bool resource_validate_attach_surfaces( struct dc_state *context, const struct resource_pool *pool); -void validate_guaranteed_copy_streams( - struct dc_state *context, - int max_streams); - void resource_validate_ctx_update_pointer_after_copy( const struct dc_state *src_ctx, struct dc_state *dst_ctx); diff --git a/drivers/gpu/drm/amd/display/dc/irq_types.h b/drivers/gpu/drm/amd/display/dc/irq_types.h index a506c2e939f52..0b5f3a278c227 100644 --- a/drivers/gpu/drm/amd/display/dc/irq_types.h +++ b/drivers/gpu/drm/amd/display/dc/irq_types.h @@ -26,6 +26,8 @@ #ifndef __DAL_IRQ_TYPES_H__ #define __DAL_IRQ_TYPES_H__ +#include "os_types.h" + struct dc_context; typedef void (*interrupt_handler)(void *); @@ -135,6 +137,13 @@ enum dc_irq_source { DC_IRQ_SOURCE_VBLANK5, DC_IRQ_SOURCE_VBLANK6, + DC_IRQ_SOURCE_DC1_VLINE0, + DC_IRQ_SOURCE_DC2_VLINE0, + DC_IRQ_SOURCE_DC3_VLINE0, + DC_IRQ_SOURCE_DC4_VLINE0, + DC_IRQ_SOURCE_DC5_VLINE0, + DC_IRQ_SOURCE_DC6_VLINE0, + DAL_IRQ_SOURCES_NUMBER }; diff --git a/drivers/gpu/drm/amd/display/include/dal_asic_id.h b/drivers/gpu/drm/amd/display/include/dal_asic_id.h index 9831cb5eaa7cf..25029ed42d893 100644 --- a/drivers/gpu/drm/amd/display/include/dal_asic_id.h +++ b/drivers/gpu/drm/amd/display/include/dal_asic_id.h @@ -86,6 +86,7 @@ #define VI_POLARIS10_P_A0 80 #define VI_POLARIS11_M_A0 90 #define VI_POLARIS12_V_A0 100 +#define VI_VEGAM_A0 110 #define VI_UNKNOWN 0xFF @@ -98,7 +99,9 @@ (eChipRev < VI_POLARIS11_M_A0)) #define ASIC_REV_IS_POLARIS11_M(eChipRev) ((eChipRev >= VI_POLARIS11_M_A0) && \ (eChipRev < VI_POLARIS12_V_A0)) -#define ASIC_REV_IS_POLARIS12_V(eChipRev) (eChipRev >= VI_POLARIS12_V_A0) +#define ASIC_REV_IS_POLARIS12_V(eChipRev) ((eChipRev >= VI_POLARIS12_V_A0) && \ + (eChipRev < VI_VEGAM_A0)) +#define ASIC_REV_IS_VEGAM(eChipRev) (eChipRev >= VI_VEGAM_A0) /* DCE11 */ #define CZ_CARRIZO_A0 0x01 @@ -110,12 +113,19 @@ ((rev >= STONEY_A0) && (rev < CZ_UNKNOWN)) /* DCE12 */ +#define AI_UNKNOWN 0xFF #define AI_GREENLAND_P_A0 1 #define AI_GREENLAND_P_A1 2 +#define AI_UNKNOWN 0xFF -#define ASICREV_IS_GREENLAND_M(eChipRev) (eChipRev < AI_UNKNOWN) -#define ASICREV_IS_GREENLAND_P(eChipRev) (eChipRev < AI_UNKNOWN) +#define AI_VEGA12_P_A0 20 +#define AI_VEGA20_P_A0 40 +#define ASICREV_IS_GREENLAND_M(eChipRev) (eChipRev < AI_VEGA12_P_A0) +#define ASICREV_IS_GREENLAND_P(eChipRev) (eChipRev < AI_VEGA12_P_A0) + +#define ASICREV_IS_VEGA12_P(eChipRev) ((eChipRev >= AI_VEGA12_P_A0) && (eChipRev < AI_VEGA20_P_A0)) +#define ASICREV_IS_VEGA20_P(eChipRev) ((eChipRev >= AI_VEGA20_P_A0) && (eChipRev < AI_UNKNOWN)) /* DCN1_0 */ #define INTERNAL_REV_RAVEN_A0 0x00 /* First spin of Raven */ diff --git a/drivers/gpu/drm/amd/display/include/dal_types.h b/drivers/gpu/drm/amd/display/include/dal_types.h index fa543965feb51..840142b65f8b0 100644 --- a/drivers/gpu/drm/amd/display/include/dal_types.h +++ b/drivers/gpu/drm/amd/display/include/dal_types.h @@ -40,6 +40,7 @@ enum dce_version { DCE_VERSION_10_0, DCE_VERSION_11_0, DCE_VERSION_11_2, + DCE_VERSION_11_22, DCE_VERSION_12_0, DCE_VERSION_MAX, DCN_VERSION_1_0, diff --git a/drivers/gpu/drm/amd/display/include/fixed31_32.h b/drivers/gpu/drm/amd/display/include/fixed31_32.h index 0de258622c12d..bb0d4ebba9f08 100644 --- a/drivers/gpu/drm/amd/display/include/fixed31_32.h +++ b/drivers/gpu/drm/amd/display/include/fixed31_32.h @@ -26,9 +26,13 @@ #ifndef __DAL_FIXED31_32_H__ #define __DAL_FIXED31_32_H__ -#include "os_types.h" - #define FIXED31_32_BITS_PER_FRACTIONAL_PART 32 +#ifndef LLONG_MIN +#define LLONG_MIN (1LL<<63) +#endif +#ifndef LLONG_MAX +#define LLONG_MAX (-1LL>>1) +#endif /* * @brief @@ -44,24 +48,25 @@ */ struct fixed31_32 { - int64_t value; + long long value; }; + /* * @brief * Useful constants */ -static const struct fixed31_32 dal_fixed31_32_zero = { 0 }; -static const struct fixed31_32 dal_fixed31_32_epsilon = { 1LL }; -static const struct fixed31_32 dal_fixed31_32_half = { 0x80000000LL }; -static const struct fixed31_32 dal_fixed31_32_one = { 0x100000000LL }; +static const struct fixed31_32 dc_fixpt_zero = { 0 }; +static const struct fixed31_32 dc_fixpt_epsilon = { 1LL }; +static const struct fixed31_32 dc_fixpt_half = { 0x80000000LL }; +static const struct fixed31_32 dc_fixpt_one = { 0x100000000LL }; -static const struct fixed31_32 dal_fixed31_32_pi = { 13493037705LL }; -static const struct fixed31_32 dal_fixed31_32_two_pi = { 26986075409LL }; -static const struct fixed31_32 dal_fixed31_32_e = { 11674931555LL }; -static const struct fixed31_32 dal_fixed31_32_ln2 = { 2977044471LL }; -static const struct fixed31_32 dal_fixed31_32_ln2_div_2 = { 1488522236LL }; +static const struct fixed31_32 dc_fixpt_pi = { 13493037705LL }; +static const struct fixed31_32 dc_fixpt_two_pi = { 26986075409LL }; +static const struct fixed31_32 dc_fixpt_e = { 11674931555LL }; +static const struct fixed31_32 dc_fixpt_ln2 = { 2977044471LL }; +static const struct fixed31_32 dc_fixpt_ln2_div_2 = { 1488522236LL }; /* * @brief @@ -72,24 +77,19 @@ static const struct fixed31_32 dal_fixed31_32_ln2_div_2 = { 1488522236LL }; * @brief * result = numerator / denominator */ -struct fixed31_32 dal_fixed31_32_from_fraction( - int64_t numerator, - int64_t denominator); +struct fixed31_32 dc_fixpt_from_fraction(long long numerator, long long denominator); /* * @brief * result = arg */ -struct fixed31_32 dal_fixed31_32_from_int_nonconst(int64_t arg); -static inline struct fixed31_32 dal_fixed31_32_from_int(int64_t arg) +static inline struct fixed31_32 dc_fixpt_from_int(int arg) { - if (__builtin_constant_p(arg)) { - struct fixed31_32 res; - BUILD_BUG_ON((LONG_MIN > arg) || (arg > LONG_MAX)); - res.value = arg << FIXED31_32_BITS_PER_FRACTIONAL_PART; - return res; - } else - return dal_fixed31_32_from_int_nonconst(arg); + struct fixed31_32 res; + + res.value = (long long) arg << FIXED31_32_BITS_PER_FRACTIONAL_PART; + + return res; } /* @@ -101,7 +101,7 @@ static inline struct fixed31_32 dal_fixed31_32_from_int(int64_t arg) * @brief * result = -arg */ -static inline struct fixed31_32 dal_fixed31_32_neg(struct fixed31_32 arg) +static inline struct fixed31_32 dc_fixpt_neg(struct fixed31_32 arg) { struct fixed31_32 res; @@ -114,10 +114,10 @@ static inline struct fixed31_32 dal_fixed31_32_neg(struct fixed31_32 arg) * @brief * result = abs(arg) := (arg >= 0) ? arg : -arg */ -static inline struct fixed31_32 dal_fixed31_32_abs(struct fixed31_32 arg) +static inline struct fixed31_32 dc_fixpt_abs(struct fixed31_32 arg) { if (arg.value < 0) - return dal_fixed31_32_neg(arg); + return dc_fixpt_neg(arg); else return arg; } @@ -131,8 +131,7 @@ static inline struct fixed31_32 dal_fixed31_32_abs(struct fixed31_32 arg) * @brief * result = arg1 < arg2 */ -static inline bool dal_fixed31_32_lt(struct fixed31_32 arg1, - struct fixed31_32 arg2) +static inline bool dc_fixpt_lt(struct fixed31_32 arg1, struct fixed31_32 arg2) { return arg1.value < arg2.value; } @@ -141,8 +140,7 @@ static inline bool dal_fixed31_32_lt(struct fixed31_32 arg1, * @brief * result = arg1 <= arg2 */ -static inline bool dal_fixed31_32_le(struct fixed31_32 arg1, - struct fixed31_32 arg2) +static inline bool dc_fixpt_le(struct fixed31_32 arg1, struct fixed31_32 arg2) { return arg1.value <= arg2.value; } @@ -151,8 +149,7 @@ static inline bool dal_fixed31_32_le(struct fixed31_32 arg1, * @brief * result = arg1 == arg2 */ -static inline bool dal_fixed31_32_eq(struct fixed31_32 arg1, - struct fixed31_32 arg2) +static inline bool dc_fixpt_eq(struct fixed31_32 arg1, struct fixed31_32 arg2) { return arg1.value == arg2.value; } @@ -161,8 +158,7 @@ static inline bool dal_fixed31_32_eq(struct fixed31_32 arg1, * @brief * result = min(arg1, arg2) := (arg1 <= arg2) ? arg1 : arg2 */ -static inline struct fixed31_32 dal_fixed31_32_min(struct fixed31_32 arg1, - struct fixed31_32 arg2) +static inline struct fixed31_32 dc_fixpt_min(struct fixed31_32 arg1, struct fixed31_32 arg2) { if (arg1.value <= arg2.value) return arg1; @@ -174,8 +170,7 @@ static inline struct fixed31_32 dal_fixed31_32_min(struct fixed31_32 arg1, * @brief * result = max(arg1, arg2) := (arg1 <= arg2) ? arg2 : arg1 */ -static inline struct fixed31_32 dal_fixed31_32_max(struct fixed31_32 arg1, - struct fixed31_32 arg2) +static inline struct fixed31_32 dc_fixpt_max(struct fixed31_32 arg1, struct fixed31_32 arg2) { if (arg1.value <= arg2.value) return arg2; @@ -189,14 +184,14 @@ static inline struct fixed31_32 dal_fixed31_32_max(struct fixed31_32 arg1, * result = | arg, when min_value < arg < max_value * | max_value, when arg >= max_value */ -static inline struct fixed31_32 dal_fixed31_32_clamp( +static inline struct fixed31_32 dc_fixpt_clamp( struct fixed31_32 arg, struct fixed31_32 min_value, struct fixed31_32 max_value) { - if (dal_fixed31_32_le(arg, min_value)) + if (dc_fixpt_le(arg, min_value)) return min_value; - else if (dal_fixed31_32_le(max_value, arg)) + else if (dc_fixpt_le(max_value, arg)) return max_value; else return arg; @@ -211,21 +206,30 @@ static inline struct fixed31_32 dal_fixed31_32_clamp( * @brief * result = arg << shift */ -struct fixed31_32 dal_fixed31_32_shl( - struct fixed31_32 arg, - uint8_t shift); +static inline struct fixed31_32 dc_fixpt_shl(struct fixed31_32 arg, unsigned char shift) +{ + ASSERT(((arg.value >= 0) && (arg.value <= LLONG_MAX >> shift)) || + ((arg.value < 0) && (arg.value >= ~(LLONG_MAX >> shift)))); + + arg.value = arg.value << shift; + + return arg; +} /* * @brief * result = arg >> shift */ -static inline struct fixed31_32 dal_fixed31_32_shr( - struct fixed31_32 arg, - uint8_t shift) +static inline struct fixed31_32 dc_fixpt_shr(struct fixed31_32 arg, unsigned char shift) { - struct fixed31_32 res; - res.value = arg.value >> shift; - return res; + bool negative = arg.value < 0; + + if (negative) + arg.value = -arg.value; + arg.value = arg.value >> shift; + if (negative) + arg.value = -arg.value; + return arg; } /* @@ -237,38 +241,50 @@ static inline struct fixed31_32 dal_fixed31_32_shr( * @brief * result = arg1 + arg2 */ -struct fixed31_32 dal_fixed31_32_add( - struct fixed31_32 arg1, - struct fixed31_32 arg2); +static inline struct fixed31_32 dc_fixpt_add(struct fixed31_32 arg1, struct fixed31_32 arg2) +{ + struct fixed31_32 res; + + ASSERT(((arg1.value >= 0) && (LLONG_MAX - arg1.value >= arg2.value)) || + ((arg1.value < 0) && (LLONG_MIN - arg1.value <= arg2.value))); + + res.value = arg1.value + arg2.value; + + return res; +} /* * @brief * result = arg1 + arg2 */ -static inline struct fixed31_32 dal_fixed31_32_add_int(struct fixed31_32 arg1, - int32_t arg2) +static inline struct fixed31_32 dc_fixpt_add_int(struct fixed31_32 arg1, int arg2) { - return dal_fixed31_32_add(arg1, - dal_fixed31_32_from_int(arg2)); + return dc_fixpt_add(arg1, dc_fixpt_from_int(arg2)); } /* * @brief * result = arg1 - arg2 */ -struct fixed31_32 dal_fixed31_32_sub( - struct fixed31_32 arg1, - struct fixed31_32 arg2); +static inline struct fixed31_32 dc_fixpt_sub(struct fixed31_32 arg1, struct fixed31_32 arg2) +{ + struct fixed31_32 res; + + ASSERT(((arg2.value >= 0) && (LLONG_MIN + arg2.value <= arg1.value)) || + ((arg2.value < 0) && (LLONG_MAX + arg2.value >= arg1.value))); + + res.value = arg1.value - arg2.value; + + return res; +} /* * @brief * result = arg1 - arg2 */ -static inline struct fixed31_32 dal_fixed31_32_sub_int(struct fixed31_32 arg1, - int32_t arg2) +static inline struct fixed31_32 dc_fixpt_sub_int(struct fixed31_32 arg1, int arg2) { - return dal_fixed31_32_sub(arg1, - dal_fixed31_32_from_int(arg2)); + return dc_fixpt_sub(arg1, dc_fixpt_from_int(arg2)); } @@ -281,49 +297,40 @@ static inline struct fixed31_32 dal_fixed31_32_sub_int(struct fixed31_32 arg1, * @brief * result = arg1 * arg2 */ -struct fixed31_32 dal_fixed31_32_mul( - struct fixed31_32 arg1, - struct fixed31_32 arg2); +struct fixed31_32 dc_fixpt_mul(struct fixed31_32 arg1, struct fixed31_32 arg2); /* * @brief * result = arg1 * arg2 */ -static inline struct fixed31_32 dal_fixed31_32_mul_int(struct fixed31_32 arg1, - int32_t arg2) +static inline struct fixed31_32 dc_fixpt_mul_int(struct fixed31_32 arg1, int arg2) { - return dal_fixed31_32_mul(arg1, - dal_fixed31_32_from_int(arg2)); + return dc_fixpt_mul(arg1, dc_fixpt_from_int(arg2)); } /* * @brief * result = square(arg) := arg * arg */ -struct fixed31_32 dal_fixed31_32_sqr( - struct fixed31_32 arg); +struct fixed31_32 dc_fixpt_sqr(struct fixed31_32 arg); /* * @brief * result = arg1 / arg2 */ -static inline struct fixed31_32 dal_fixed31_32_div_int(struct fixed31_32 arg1, - int64_t arg2) +static inline struct fixed31_32 dc_fixpt_div_int(struct fixed31_32 arg1, long long arg2) { - return dal_fixed31_32_from_fraction(arg1.value, - dal_fixed31_32_from_int(arg2).value); + return dc_fixpt_from_fraction(arg1.value, dc_fixpt_from_int(arg2).value); } /* * @brief * result = arg1 / arg2 */ -static inline struct fixed31_32 dal_fixed31_32_div(struct fixed31_32 arg1, - struct fixed31_32 arg2) +static inline struct fixed31_32 dc_fixpt_div(struct fixed31_32 arg1, struct fixed31_32 arg2) { - return dal_fixed31_32_from_fraction(arg1.value, - arg2.value); + return dc_fixpt_from_fraction(arg1.value, arg2.value); } /* @@ -338,8 +345,7 @@ static inline struct fixed31_32 dal_fixed31_32_div(struct fixed31_32 arg1, * @note * No special actions taken in case argument is zero. */ -struct fixed31_32 dal_fixed31_32_recip( - struct fixed31_32 arg); +struct fixed31_32 dc_fixpt_recip(struct fixed31_32 arg); /* * @brief @@ -354,8 +360,7 @@ struct fixed31_32 dal_fixed31_32_recip( * Argument specified in radians, * internally it's normalized to [-2pi...2pi] range. */ -struct fixed31_32 dal_fixed31_32_sinc( - struct fixed31_32 arg); +struct fixed31_32 dc_fixpt_sinc(struct fixed31_32 arg); /* * @brief @@ -365,8 +370,7 @@ struct fixed31_32 dal_fixed31_32_sinc( * Argument specified in radians, * internally it's normalized to [-2pi...2pi] range. */ -struct fixed31_32 dal_fixed31_32_sin( - struct fixed31_32 arg); +struct fixed31_32 dc_fixpt_sin(struct fixed31_32 arg); /* * @brief @@ -378,8 +382,7 @@ struct fixed31_32 dal_fixed31_32_sin( * passing arguments outside that range * will cause incorrect result! */ -struct fixed31_32 dal_fixed31_32_cos( - struct fixed31_32 arg); +struct fixed31_32 dc_fixpt_cos(struct fixed31_32 arg); /* * @brief @@ -393,8 +396,7 @@ struct fixed31_32 dal_fixed31_32_cos( * @note * Currently, function is verified for abs(arg) <= 1. */ -struct fixed31_32 dal_fixed31_32_exp( - struct fixed31_32 arg); +struct fixed31_32 dc_fixpt_exp(struct fixed31_32 arg); /* * @brief @@ -406,8 +408,7 @@ struct fixed31_32 dal_fixed31_32_exp( * Currently, no special actions taken * in case of invalid argument(s). Take care! */ -struct fixed31_32 dal_fixed31_32_log( - struct fixed31_32 arg); +struct fixed31_32 dc_fixpt_log(struct fixed31_32 arg); /* * @brief @@ -421,9 +422,13 @@ struct fixed31_32 dal_fixed31_32_log( * @note * Currently, abs(arg1) should be less than 1. Take care! */ -struct fixed31_32 dal_fixed31_32_pow( - struct fixed31_32 arg1, - struct fixed31_32 arg2); +static inline struct fixed31_32 dc_fixpt_pow(struct fixed31_32 arg1, struct fixed31_32 arg2) +{ + return dc_fixpt_exp( + dc_fixpt_mul( + dc_fixpt_log(arg1), + arg2)); +} /* * @brief @@ -434,22 +439,56 @@ struct fixed31_32 dal_fixed31_32_pow( * @brief * result = floor(arg) := greatest integer lower than or equal to arg */ -int32_t dal_fixed31_32_floor( - struct fixed31_32 arg); +static inline int dc_fixpt_floor(struct fixed31_32 arg) +{ + unsigned long long arg_value = arg.value > 0 ? arg.value : -arg.value; + + if (arg.value >= 0) + return (int)(arg_value >> FIXED31_32_BITS_PER_FRACTIONAL_PART); + else + return -(int)(arg_value >> FIXED31_32_BITS_PER_FRACTIONAL_PART); +} /* * @brief * result = round(arg) := integer nearest to arg */ -int32_t dal_fixed31_32_round( - struct fixed31_32 arg); +static inline int dc_fixpt_round(struct fixed31_32 arg) +{ + unsigned long long arg_value = arg.value > 0 ? arg.value : -arg.value; + + const long long summand = dc_fixpt_half.value; + + ASSERT(LLONG_MAX - (long long)arg_value >= summand); + + arg_value += summand; + + if (arg.value >= 0) + return (int)(arg_value >> FIXED31_32_BITS_PER_FRACTIONAL_PART); + else + return -(int)(arg_value >> FIXED31_32_BITS_PER_FRACTIONAL_PART); +} /* * @brief * result = ceil(arg) := lowest integer greater than or equal to arg */ -int32_t dal_fixed31_32_ceil( - struct fixed31_32 arg); +static inline int dc_fixpt_ceil(struct fixed31_32 arg) +{ + unsigned long long arg_value = arg.value > 0 ? arg.value : -arg.value; + + const long long summand = dc_fixpt_one.value - + dc_fixpt_epsilon.value; + + ASSERT(LLONG_MAX - (long long)arg_value >= summand); + + arg_value += summand; + + if (arg.value >= 0) + return (int)(arg_value >> FIXED31_32_BITS_PER_FRACTIONAL_PART); + else + return -(int)(arg_value >> FIXED31_32_BITS_PER_FRACTIONAL_PART); +} /* the following two function are used in scaler hw programming to convert fixed * point value to format 2 bits from integer part and 19 bits from fractional @@ -457,20 +496,31 @@ int32_t dal_fixed31_32_ceil( * fractional */ -uint32_t dal_fixed31_32_u2d19( - struct fixed31_32 arg); +unsigned int dc_fixpt_u2d19(struct fixed31_32 arg); + +unsigned int dc_fixpt_u0d19(struct fixed31_32 arg); -uint32_t dal_fixed31_32_u0d19( - struct fixed31_32 arg); +unsigned int dc_fixpt_clamp_u0d14(struct fixed31_32 arg); +unsigned int dc_fixpt_clamp_u0d10(struct fixed31_32 arg); -uint32_t dal_fixed31_32_clamp_u0d14( - struct fixed31_32 arg); +int dc_fixpt_s4d19(struct fixed31_32 arg); -uint32_t dal_fixed31_32_clamp_u0d10( - struct fixed31_32 arg); +static inline struct fixed31_32 dc_fixpt_truncate(struct fixed31_32 arg, unsigned int frac_bits) +{ + bool negative = arg.value < 0; -int32_t dal_fixed31_32_s4d19( - struct fixed31_32 arg); + if (frac_bits >= FIXED31_32_BITS_PER_FRACTIONAL_PART) { + ASSERT(frac_bits == FIXED31_32_BITS_PER_FRACTIONAL_PART); + return arg; + } + + if (negative) + arg.value = -arg.value; + arg.value &= (~0LL) << (FIXED31_32_BITS_PER_FRACTIONAL_PART - frac_bits); + if (negative) + arg.value = -arg.value; + return arg; +} #endif diff --git a/drivers/gpu/drm/amd/display/include/fixed32_32.h b/drivers/gpu/drm/amd/display/include/fixed32_32.h deleted file mode 100644 index 9c70341fe0263..0000000000000 --- a/drivers/gpu/drm/amd/display/include/fixed32_32.h +++ /dev/null @@ -1,129 +0,0 @@ -/* - * Copyright 2012-15 Advanced Micro Devices, Inc. - * - * Permission is hereby granted, free of charge, to any person obtaining a - * copy of this software and associated documentation files (the "Software"), - * to deal in the Software without restriction, including without limitation - * the rights to use, copy, modify, merge, publish, distribute, sublicense, - * and/or sell copies of the Software, and to permit persons to whom the - * Software is furnished to do so, subject to the following conditions: - * - * The above copyright notice and this permission notice shall be included in - * all copies or substantial portions of the Software. - * - * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR - * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, - * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL - * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR - * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, - * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR - * OTHER DEALINGS IN THE SOFTWARE. - * - * Authors: AMD - * - */ - - -#ifndef __DAL_FIXED32_32_H__ -#define __DAL_FIXED32_32_H__ - -#include "os_types.h" - -struct fixed32_32 { - uint64_t value; -}; - -static const struct fixed32_32 dal_fixed32_32_zero = { 0 }; -static const struct fixed32_32 dal_fixed32_32_one = { 0x100000000LL }; -static const struct fixed32_32 dal_fixed32_32_half = { 0x80000000LL }; - -struct fixed32_32 dal_fixed32_32_from_fraction(uint32_t n, uint32_t d); -static inline struct fixed32_32 dal_fixed32_32_from_int(uint32_t value) -{ - struct fixed32_32 fx; - - fx.value = (uint64_t)value<<32; - return fx; -} - -struct fixed32_32 dal_fixed32_32_add( - struct fixed32_32 lhs, - struct fixed32_32 rhs); -struct fixed32_32 dal_fixed32_32_add_int( - struct fixed32_32 lhs, - uint32_t rhs); -struct fixed32_32 dal_fixed32_32_sub( - struct fixed32_32 lhs, - struct fixed32_32 rhs); -struct fixed32_32 dal_fixed32_32_sub_int( - struct fixed32_32 lhs, - uint32_t rhs); -struct fixed32_32 dal_fixed32_32_mul( - struct fixed32_32 lhs, - struct fixed32_32 rhs); -struct fixed32_32 dal_fixed32_32_mul_int( - struct fixed32_32 lhs, - uint32_t rhs); -struct fixed32_32 dal_fixed32_32_div( - struct fixed32_32 lhs, - struct fixed32_32 rhs); -struct fixed32_32 dal_fixed32_32_div_int( - struct fixed32_32 lhs, - uint32_t rhs); - -static inline struct fixed32_32 dal_fixed32_32_min(struct fixed32_32 lhs, - struct fixed32_32 rhs) -{ - return (lhs.value < rhs.value) ? lhs : rhs; -} - -static inline struct fixed32_32 dal_fixed32_32_max(struct fixed32_32 lhs, - struct fixed32_32 rhs) -{ - return (lhs.value > rhs.value) ? lhs : rhs; -} - -static inline bool dal_fixed32_32_gt(struct fixed32_32 lhs, struct fixed32_32 rhs) -{ - return lhs.value > rhs.value; -} - -static inline bool dal_fixed32_32_gt_int(struct fixed32_32 lhs, uint32_t rhs) -{ - return lhs.value > ((uint64_t)rhs<<32); -} - -static inline bool dal_fixed32_32_lt(struct fixed32_32 lhs, struct fixed32_32 rhs) -{ - return lhs.value < rhs.value; -} - -static inline bool dal_fixed32_32_lt_int(struct fixed32_32 lhs, uint32_t rhs) -{ - return lhs.value < ((uint64_t)rhs<<32); -} - -static inline bool dal_fixed32_32_le(struct fixed32_32 lhs, struct fixed32_32 rhs) -{ - return lhs.value <= rhs.value; -} - -static inline bool dal_fixed32_32_le_int(struct fixed32_32 lhs, uint32_t rhs) -{ - return lhs.value <= ((uint64_t)rhs<<32); -} - -static inline bool dal_fixed32_32_eq(struct fixed32_32 lhs, struct fixed32_32 rhs) -{ - return lhs.value == rhs.value; -} - -uint32_t dal_fixed32_32_ceil(struct fixed32_32 value); -static inline uint32_t dal_fixed32_32_floor(struct fixed32_32 value) -{ - return value.value>>32; -} - -uint32_t dal_fixed32_32_round(struct fixed32_32 value); - -#endif diff --git a/drivers/gpu/drm/amd/display/include/logger_interface.h b/drivers/gpu/drm/amd/display/include/logger_interface.h index 28dee960d5095..dc98d6d4b2bd0 100644 --- a/drivers/gpu/drm/amd/display/include/logger_interface.h +++ b/drivers/gpu/drm/amd/display/include/logger_interface.h @@ -190,4 +190,13 @@ void context_clock_trace( } \ } while (0) +#define DISPLAY_STATS_BEGIN(entry) \ + dm_logger_open(dc->ctx->logger, &entry, LOG_DISPLAYSTATS) + +#define DISPLAY_STATS(msg, ...) \ + dm_logger_append(&log_entry, msg, ##__VA_ARGS__) + +#define DISPLAY_STATS_END(entry) \ + dm_logger_close(&entry) + #endif /* __DAL_LOGGER_INTERFACE_H__ */ diff --git a/drivers/gpu/drm/amd/display/include/logger_types.h b/drivers/gpu/drm/amd/display/include/logger_types.h index 427796bdc14ae..0a540b9897a67 100644 --- a/drivers/gpu/drm/amd/display/include/logger_types.h +++ b/drivers/gpu/drm/amd/display/include/logger_types.h @@ -29,39 +29,39 @@ #include "os_types.h" #define MAX_NAME_LEN 32 -#define DC_LOG_ERROR(a, ...) dm_logger_write(DC_LOGGER, LOG_ERROR, a, ## __VA_ARGS__) -#define DC_LOG_WARNING(a, ...) dm_logger_write(DC_LOGGER, LOG_WARNING, a, ## __VA_ARGS__) -#define DC_LOG_DEBUG(a, ...) dm_logger_write(DC_LOGGER, LOG_DEBUG, a, ## __VA_ARGS__) -#define DC_LOG_DC(a, ...) dm_logger_write(DC_LOGGER, LOG_DC, a, ## __VA_ARGS__) -#define DC_LOG_DTN(a, ...) dm_logger_write(DC_LOGGER, LOG_DTN, a, ## __VA_ARGS__) -#define DC_LOG_SURFACE(a, ...) dm_logger_write(DC_LOGGER, LOG_SURFACE, a, ## __VA_ARGS__) -#define DC_LOG_HW_HOTPLUG(a, ...) dm_logger_write(DC_LOGGER, LOG_HW_HOTPLUG, a, ## __VA_ARGS__) -#define DC_LOG_HW_LINK_TRAINING(a, ...) dm_logger_write(DC_LOGGER, LOG_HW_LINK_TRAINING, a, ## __VA_ARGS__) -#define DC_LOG_HW_SET_MODE(a, ...) dm_logger_write(DC_LOGGER, LOG_HW_SET_MODE, a, ## __VA_ARGS__) -#define DC_LOG_HW_RESUME_S3(a, ...) dm_logger_write(DC_LOGGER, LOG_HW_RESUME_S3, a, ## __VA_ARGS__) -#define DC_LOG_HW_AUDIO(a, ...) dm_logger_write(DC_LOGGER, LOG_HW_AUDIO, a, ## __VA_ARGS__) -#define DC_LOG_HW_HPD_IRQ(a, ...) dm_logger_write(DC_LOGGER, LOG_HW_HPD_IRQ, a, ## __VA_ARGS__) -#define DC_LOG_MST(a, ...) dm_logger_write(DC_LOGGER, LOG_MST, a, ## __VA_ARGS__) -#define DC_LOG_SCALER(a, ...) dm_logger_write(DC_LOGGER, LOG_SCALER, a, ## __VA_ARGS__) -#define DC_LOG_BIOS(a, ...) dm_logger_write(DC_LOGGER, LOG_BIOS, a, ## __VA_ARGS__) -#define DC_LOG_BANDWIDTH_CALCS(a, ...) dm_logger_write(DC_LOGGER, LOG_BANDWIDTH_CALCS, a, ## __VA_ARGS__) -#define DC_LOG_BANDWIDTH_VALIDATION(a, ...) dm_logger_write(DC_LOGGER, LOG_BANDWIDTH_VALIDATION, a, ## __VA_ARGS__) -#define DC_LOG_I2C_AUX(a, ...) dm_logger_write(DC_LOGGER, LOG_I2C_AUX, a, ## __VA_ARGS__) -#define DC_LOG_SYNC(a, ...) dm_logger_write(DC_LOGGER, LOG_SYNC, a, ## __VA_ARGS__) -#define DC_LOG_BACKLIGHT(a, ...) dm_logger_write(DC_LOGGER, LOG_BACKLIGHT, a, ## __VA_ARGS__) -#define DC_LOG_FEATURE_OVERRIDE(a, ...) dm_logger_write(DC_LOGGER, LOG_FEATURE_OVERRIDE, a, ## __VA_ARGS__) -#define DC_LOG_DETECTION_EDID_PARSER(a, ...) dm_logger_write(DC_LOGGER, LOG_DETECTION_EDID_PARSER, a, ## __VA_ARGS__) -#define DC_LOG_DETECTION_DP_CAPS(a, ...) dm_logger_write(DC_LOGGER, LOG_DETECTION_DP_CAPS, a, ## __VA_ARGS__) -#define DC_LOG_RESOURCE(a, ...) dm_logger_write(DC_LOGGER, LOG_RESOURCE, a, ## __VA_ARGS__) -#define DC_LOG_DML(a, ...) dm_logger_write(DC_LOGGER, LOG_DML, a, ## __VA_ARGS__) -#define DC_LOG_EVENT_MODE_SET(a, ...) dm_logger_write(DC_LOGGER, LOG_EVENT_MODE_SET, a, ## __VA_ARGS__) -#define DC_LOG_EVENT_DETECTION(a, ...) dm_logger_write(DC_LOGGER, LOG_EVENT_DETECTION, a, ## __VA_ARGS__) -#define DC_LOG_EVENT_LINK_TRAINING(a, ...) dm_logger_write(DC_LOGGER, LOG_EVENT_LINK_TRAINING, a, ## __VA_ARGS__) -#define DC_LOG_EVENT_LINK_LOSS(a, ...) dm_logger_write(DC_LOGGER, LOG_EVENT_LINK_LOSS, a, ## __VA_ARGS__) -#define DC_LOG_EVENT_UNDERFLOW(a, ...) dm_logger_write(DC_LOGGER, LOG_EVENT_UNDERFLOW, a, ## __VA_ARGS__) -#define DC_LOG_IF_TRACE(a, ...) dm_logger_write(DC_LOGGER, LOG_IF_TRACE, a, ## __VA_ARGS__) -#define DC_LOG_PERF_TRACE(a, ...) dm_logger_write(DC_LOGGER, LOG_PERF_TRACE, a, ## __VA_ARGS__) +#define DC_LOG_ERROR(...) DRM_ERROR(__VA_ARGS__) +#define DC_LOG_WARNING(...) DRM_WARN(__VA_ARGS__) +#define DC_LOG_DEBUG(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_DC(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_DTN(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_SURFACE(...) pr_debug("[SURFACE]:"__VA_ARGS__) +#define DC_LOG_HW_HOTPLUG(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_HW_LINK_TRAINING(...) pr_debug("[HW_LINK_TRAINING]:"__VA_ARGS__) +#define DC_LOG_HW_SET_MODE(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_HW_RESUME_S3(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_HW_AUDIO(...) pr_debug("[HW_AUDIO]:"__VA_ARGS__) +#define DC_LOG_HW_HPD_IRQ(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_MST(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_SCALER(...) pr_debug("[SCALER]:"__VA_ARGS__) +#define DC_LOG_BIOS(...) pr_debug("[BIOS]:"__VA_ARGS__) +#define DC_LOG_BANDWIDTH_CALCS(...) pr_debug("[BANDWIDTH_CALCS]:"__VA_ARGS__) +#define DC_LOG_BANDWIDTH_VALIDATION(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_I2C_AUX(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_SYNC(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_BACKLIGHT(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_FEATURE_OVERRIDE(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_DETECTION_EDID_PARSER(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_DETECTION_DP_CAPS(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_RESOURCE(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_DML(...) pr_debug("[DML]:"__VA_ARGS__) +#define DC_LOG_EVENT_MODE_SET(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_EVENT_DETECTION(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_EVENT_LINK_TRAINING(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_EVENT_LINK_LOSS(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_EVENT_UNDERFLOW(...) DRM_DEBUG_KMS(__VA_ARGS__) +#define DC_LOG_IF_TRACE(...) pr_debug("[IF_TRACE]:"__VA_ARGS__) +#define DC_LOG_PERF_TRACE(...) DRM_DEBUG_KMS(__VA_ARGS__) struct dal_logger; @@ -98,7 +98,7 @@ enum dc_log_type { LOG_EVENT_UNDERFLOW, LOG_IF_TRACE, LOG_PERF_TRACE, - LOG_PROFILING, + LOG_DISPLAYSTATS, LOG_SECTION_TOTAL_COUNT }; diff --git a/drivers/gpu/drm/amd/display/modules/color/color_gamma.c b/drivers/gpu/drm/amd/display/modules/color/color_gamma.c index e7e374f568644..0cd111d590183 100644 --- a/drivers/gpu/drm/amd/display/modules/color/color_gamma.c +++ b/drivers/gpu/drm/amd/display/modules/color/color_gamma.c @@ -43,7 +43,7 @@ static bool de_pq_initialized; /* = false; */ /* one-time setup of X points */ void setup_x_points_distribution(void) { - struct fixed31_32 region_size = dal_fixed31_32_from_int(128); + struct fixed31_32 region_size = dc_fixpt_from_int(128); int32_t segment; uint32_t seg_offset; uint32_t index; @@ -53,8 +53,8 @@ void setup_x_points_distribution(void) coordinates_x[MAX_HW_POINTS + 1].x = region_size; for (segment = 6; segment > (6 - NUM_REGIONS); segment--) { - region_size = dal_fixed31_32_div_int(region_size, 2); - increment = dal_fixed31_32_div_int(region_size, + region_size = dc_fixpt_div_int(region_size, 2); + increment = dc_fixpt_div_int(region_size, NUM_PTS_IN_REGION); seg_offset = (segment + (NUM_REGIONS - 7)) * NUM_PTS_IN_REGION; coordinates_x[seg_offset].x = region_size; @@ -62,7 +62,7 @@ void setup_x_points_distribution(void) for (index = seg_offset + 1; index < seg_offset + NUM_PTS_IN_REGION; index++) { - coordinates_x[index].x = dal_fixed31_32_add + coordinates_x[index].x = dc_fixpt_add (coordinates_x[index-1].x, increment); } } @@ -72,63 +72,63 @@ static void compute_pq(struct fixed31_32 in_x, struct fixed31_32 *out_y) { /* consts for PQ gamma formula. */ const struct fixed31_32 m1 = - dal_fixed31_32_from_fraction(159301758, 1000000000); + dc_fixpt_from_fraction(159301758, 1000000000); const struct fixed31_32 m2 = - dal_fixed31_32_from_fraction(7884375, 100000); + dc_fixpt_from_fraction(7884375, 100000); const struct fixed31_32 c1 = - dal_fixed31_32_from_fraction(8359375, 10000000); + dc_fixpt_from_fraction(8359375, 10000000); const struct fixed31_32 c2 = - dal_fixed31_32_from_fraction(188515625, 10000000); + dc_fixpt_from_fraction(188515625, 10000000); const struct fixed31_32 c3 = - dal_fixed31_32_from_fraction(186875, 10000); + dc_fixpt_from_fraction(186875, 10000); struct fixed31_32 l_pow_m1; struct fixed31_32 base; - if (dal_fixed31_32_lt(in_x, dal_fixed31_32_zero)) - in_x = dal_fixed31_32_zero; + if (dc_fixpt_lt(in_x, dc_fixpt_zero)) + in_x = dc_fixpt_zero; - l_pow_m1 = dal_fixed31_32_pow(in_x, m1); - base = dal_fixed31_32_div( - dal_fixed31_32_add(c1, - (dal_fixed31_32_mul(c2, l_pow_m1))), - dal_fixed31_32_add(dal_fixed31_32_one, - (dal_fixed31_32_mul(c3, l_pow_m1)))); - *out_y = dal_fixed31_32_pow(base, m2); + l_pow_m1 = dc_fixpt_pow(in_x, m1); + base = dc_fixpt_div( + dc_fixpt_add(c1, + (dc_fixpt_mul(c2, l_pow_m1))), + dc_fixpt_add(dc_fixpt_one, + (dc_fixpt_mul(c3, l_pow_m1)))); + *out_y = dc_fixpt_pow(base, m2); } static void compute_de_pq(struct fixed31_32 in_x, struct fixed31_32 *out_y) { /* consts for dePQ gamma formula. */ const struct fixed31_32 m1 = - dal_fixed31_32_from_fraction(159301758, 1000000000); + dc_fixpt_from_fraction(159301758, 1000000000); const struct fixed31_32 m2 = - dal_fixed31_32_from_fraction(7884375, 100000); + dc_fixpt_from_fraction(7884375, 100000); const struct fixed31_32 c1 = - dal_fixed31_32_from_fraction(8359375, 10000000); + dc_fixpt_from_fraction(8359375, 10000000); const struct fixed31_32 c2 = - dal_fixed31_32_from_fraction(188515625, 10000000); + dc_fixpt_from_fraction(188515625, 10000000); const struct fixed31_32 c3 = - dal_fixed31_32_from_fraction(186875, 10000); + dc_fixpt_from_fraction(186875, 10000); struct fixed31_32 l_pow_m1; struct fixed31_32 base, div; - if (dal_fixed31_32_lt(in_x, dal_fixed31_32_zero)) - in_x = dal_fixed31_32_zero; + if (dc_fixpt_lt(in_x, dc_fixpt_zero)) + in_x = dc_fixpt_zero; - l_pow_m1 = dal_fixed31_32_pow(in_x, - dal_fixed31_32_div(dal_fixed31_32_one, m2)); - base = dal_fixed31_32_sub(l_pow_m1, c1); + l_pow_m1 = dc_fixpt_pow(in_x, + dc_fixpt_div(dc_fixpt_one, m2)); + base = dc_fixpt_sub(l_pow_m1, c1); - if (dal_fixed31_32_lt(base, dal_fixed31_32_zero)) - base = dal_fixed31_32_zero; + if (dc_fixpt_lt(base, dc_fixpt_zero)) + base = dc_fixpt_zero; - div = dal_fixed31_32_sub(c2, dal_fixed31_32_mul(c3, l_pow_m1)); + div = dc_fixpt_sub(c2, dc_fixpt_mul(c3, l_pow_m1)); - *out_y = dal_fixed31_32_pow(dal_fixed31_32_div(base, div), - dal_fixed31_32_div(dal_fixed31_32_one, m1)); + *out_y = dc_fixpt_pow(dc_fixpt_div(base, div), + dc_fixpt_div(dc_fixpt_one, m1)); } /* one-time pre-compute PQ values - only for sdr_white_level 80 */ @@ -138,14 +138,14 @@ void precompute_pq(void) struct fixed31_32 x; const struct hw_x_point *coord_x = coordinates_x + 32; struct fixed31_32 scaling_factor = - dal_fixed31_32_from_fraction(80, 10000); + dc_fixpt_from_fraction(80, 10000); /* pow function has problems with arguments too small */ for (i = 0; i < 32; i++) - pq_table[i] = dal_fixed31_32_zero; + pq_table[i] = dc_fixpt_zero; for (i = 32; i <= MAX_HW_POINTS; i++) { - x = dal_fixed31_32_mul(coord_x->x, scaling_factor); + x = dc_fixpt_mul(coord_x->x, scaling_factor); compute_pq(x, &pq_table[i]); ++coord_x; } @@ -158,7 +158,7 @@ void precompute_de_pq(void) struct fixed31_32 y; uint32_t begin_index, end_index; - struct fixed31_32 scaling_factor = dal_fixed31_32_from_int(125); + struct fixed31_32 scaling_factor = dc_fixpt_from_int(125); /* X points is 2^-25 to 2^7 * De-gamma X is 2^-12 to 2^0 – we are skipping first -12-(-25) = 13 regions @@ -167,11 +167,11 @@ void precompute_de_pq(void) end_index = begin_index + 12 * NUM_PTS_IN_REGION; for (i = 0; i <= begin_index; i++) - de_pq_table[i] = dal_fixed31_32_zero; + de_pq_table[i] = dc_fixpt_zero; for (; i <= end_index; i++) { compute_de_pq(coordinates_x[i].x, &y); - de_pq_table[i] = dal_fixed31_32_mul(y, scaling_factor); + de_pq_table[i] = dc_fixpt_mul(y, scaling_factor); } for (; i <= MAX_HW_POINTS; i++) @@ -185,25 +185,25 @@ struct dividers { static void build_coefficients(struct gamma_coefficients *coefficients, bool is_2_4) { - static const int32_t numerator01[] = { 31308, 180000}; - static const int32_t numerator02[] = { 12920, 4500}; - static const int32_t numerator03[] = { 55, 99}; - static const int32_t numerator04[] = { 55, 99}; - static const int32_t numerator05[] = { 2400, 2200}; + static const int32_t numerator01[] = { 31308, 180000}; + static const int32_t numerator02[] = { 12920, 4500}; + static const int32_t numerator03[] = { 55, 99}; + static const int32_t numerator04[] = { 55, 99}; + static const int32_t numerator05[] = { 2400, 2200}; - uint32_t i = 0; - uint32_t index = is_2_4 == true ? 0:1; + uint32_t i = 0; + uint32_t index = is_2_4 == true ? 0:1; do { - coefficients->a0[i] = dal_fixed31_32_from_fraction( + coefficients->a0[i] = dc_fixpt_from_fraction( numerator01[index], 10000000); - coefficients->a1[i] = dal_fixed31_32_from_fraction( + coefficients->a1[i] = dc_fixpt_from_fraction( numerator02[index], 1000); - coefficients->a2[i] = dal_fixed31_32_from_fraction( + coefficients->a2[i] = dc_fixpt_from_fraction( numerator03[index], 1000); - coefficients->a3[i] = dal_fixed31_32_from_fraction( + coefficients->a3[i] = dc_fixpt_from_fraction( numerator04[index], 1000); - coefficients->user_gamma[i] = dal_fixed31_32_from_fraction( + coefficients->user_gamma[i] = dc_fixpt_from_fraction( numerator05[index], 1000); ++i; @@ -218,33 +218,33 @@ static struct fixed31_32 translate_from_linear_space( struct fixed31_32 a3, struct fixed31_32 gamma) { - const struct fixed31_32 one = dal_fixed31_32_from_int(1); + const struct fixed31_32 one = dc_fixpt_from_int(1); - if (dal_fixed31_32_lt(one, arg)) + if (dc_fixpt_lt(one, arg)) return one; - if (dal_fixed31_32_le(arg, dal_fixed31_32_neg(a0))) - return dal_fixed31_32_sub( + if (dc_fixpt_le(arg, dc_fixpt_neg(a0))) + return dc_fixpt_sub( a2, - dal_fixed31_32_mul( - dal_fixed31_32_add( + dc_fixpt_mul( + dc_fixpt_add( one, a3), - dal_fixed31_32_pow( - dal_fixed31_32_neg(arg), - dal_fixed31_32_recip(gamma)))); - else if (dal_fixed31_32_le(a0, arg)) - return dal_fixed31_32_sub( - dal_fixed31_32_mul( - dal_fixed31_32_add( + dc_fixpt_pow( + dc_fixpt_neg(arg), + dc_fixpt_recip(gamma)))); + else if (dc_fixpt_le(a0, arg)) + return dc_fixpt_sub( + dc_fixpt_mul( + dc_fixpt_add( one, a3), - dal_fixed31_32_pow( + dc_fixpt_pow( arg, - dal_fixed31_32_recip(gamma))), + dc_fixpt_recip(gamma))), a2); else - return dal_fixed31_32_mul( + return dc_fixpt_mul( arg, a1); } @@ -259,25 +259,25 @@ static struct fixed31_32 translate_to_linear_space( { struct fixed31_32 linear; - a0 = dal_fixed31_32_mul(a0, a1); - if (dal_fixed31_32_le(arg, dal_fixed31_32_neg(a0))) + a0 = dc_fixpt_mul(a0, a1); + if (dc_fixpt_le(arg, dc_fixpt_neg(a0))) - linear = dal_fixed31_32_neg( - dal_fixed31_32_pow( - dal_fixed31_32_div( - dal_fixed31_32_sub(a2, arg), - dal_fixed31_32_add( - dal_fixed31_32_one, a3)), gamma)); + linear = dc_fixpt_neg( + dc_fixpt_pow( + dc_fixpt_div( + dc_fixpt_sub(a2, arg), + dc_fixpt_add( + dc_fixpt_one, a3)), gamma)); - else if (dal_fixed31_32_le(dal_fixed31_32_neg(a0), arg) && - dal_fixed31_32_le(arg, a0)) - linear = dal_fixed31_32_div(arg, a1); + else if (dc_fixpt_le(dc_fixpt_neg(a0), arg) && + dc_fixpt_le(arg, a0)) + linear = dc_fixpt_div(arg, a1); else - linear = dal_fixed31_32_pow( - dal_fixed31_32_div( - dal_fixed31_32_add(a2, arg), - dal_fixed31_32_add( - dal_fixed31_32_one, a3)), gamma); + linear = dc_fixpt_pow( + dc_fixpt_div( + dc_fixpt_add(a2, arg), + dc_fixpt_add( + dc_fixpt_one, a3)), gamma); return linear; } @@ -352,8 +352,8 @@ static bool find_software_points( right = axis_x[max_number - 1].b; } - if (dal_fixed31_32_le(left, hw_point) && - dal_fixed31_32_le(hw_point, right)) { + if (dc_fixpt_le(left, hw_point) && + dc_fixpt_le(hw_point, right)) { *index_to_start = i; *index_left = i; @@ -366,7 +366,7 @@ static bool find_software_points( return true; } else if ((i == *index_to_start) && - dal_fixed31_32_le(hw_point, left)) { + dc_fixpt_le(hw_point, left)) { *index_to_start = i; *index_left = i; *index_right = i; @@ -375,7 +375,7 @@ static bool find_software_points( return true; } else if ((i == max_number - 1) && - dal_fixed31_32_le(right, hw_point)) { + dc_fixpt_le(right, hw_point)) { *index_to_start = i; *index_left = i; *index_right = i; @@ -457,17 +457,17 @@ static bool build_custom_gamma_mapping_coefficients_worker( } if (hw_pos == HW_POINT_POSITION_MIDDLE) - point->coeff = dal_fixed31_32_div( - dal_fixed31_32_sub( + point->coeff = dc_fixpt_div( + dc_fixpt_sub( coord_x, left_pos), - dal_fixed31_32_sub( + dc_fixpt_sub( right_pos, left_pos)); else if (hw_pos == HW_POINT_POSITION_LEFT) - point->coeff = dal_fixed31_32_zero; + point->coeff = dc_fixpt_zero; else if (hw_pos == HW_POINT_POSITION_RIGHT) - point->coeff = dal_fixed31_32_from_int(2); + point->coeff = dc_fixpt_from_int(2); else { BREAK_TO_DEBUGGER(); return false; @@ -502,45 +502,45 @@ static struct fixed31_32 calculate_mapped_value( if ((point->left_index < 0) || (point->left_index > max_index)) { BREAK_TO_DEBUGGER(); - return dal_fixed31_32_zero; + return dc_fixpt_zero; } if ((point->right_index < 0) || (point->right_index > max_index)) { BREAK_TO_DEBUGGER(); - return dal_fixed31_32_zero; + return dc_fixpt_zero; } if (point->pos == HW_POINT_POSITION_MIDDLE) if (channel == CHANNEL_NAME_RED) - result = dal_fixed31_32_add( - dal_fixed31_32_mul( + result = dc_fixpt_add( + dc_fixpt_mul( point->coeff, - dal_fixed31_32_sub( + dc_fixpt_sub( rgb[point->right_index].r, rgb[point->left_index].r)), rgb[point->left_index].r); else if (channel == CHANNEL_NAME_GREEN) - result = dal_fixed31_32_add( - dal_fixed31_32_mul( + result = dc_fixpt_add( + dc_fixpt_mul( point->coeff, - dal_fixed31_32_sub( + dc_fixpt_sub( rgb[point->right_index].g, rgb[point->left_index].g)), rgb[point->left_index].g); else - result = dal_fixed31_32_add( - dal_fixed31_32_mul( + result = dc_fixpt_add( + dc_fixpt_mul( point->coeff, - dal_fixed31_32_sub( + dc_fixpt_sub( rgb[point->right_index].b, rgb[point->left_index].b)), rgb[point->left_index].b); else if (point->pos == HW_POINT_POSITION_LEFT) { BREAK_TO_DEBUGGER(); - result = dal_fixed31_32_zero; + result = dc_fixpt_zero; } else { BREAK_TO_DEBUGGER(); - result = dal_fixed31_32_one; + result = dc_fixpt_one; } return result; @@ -558,7 +558,7 @@ static void build_pq(struct pwl_float_data_ex *rgb_regamma, struct fixed31_32 x; struct fixed31_32 output; struct fixed31_32 scaling_factor = - dal_fixed31_32_from_fraction(sdr_white_level, 10000); + dc_fixpt_from_fraction(sdr_white_level, 10000); if (!pq_initialized && sdr_white_level == 80) { precompute_pq(); @@ -579,15 +579,15 @@ static void build_pq(struct pwl_float_data_ex *rgb_regamma, if (sdr_white_level == 80) { output = pq_table[i]; } else { - x = dal_fixed31_32_mul(coord_x->x, scaling_factor); + x = dc_fixpt_mul(coord_x->x, scaling_factor); compute_pq(x, &output); } /* should really not happen? */ - if (dal_fixed31_32_lt(output, dal_fixed31_32_zero)) - output = dal_fixed31_32_zero; - else if (dal_fixed31_32_lt(dal_fixed31_32_one, output)) - output = dal_fixed31_32_one; + if (dc_fixpt_lt(output, dc_fixpt_zero)) + output = dc_fixpt_zero; + else if (dc_fixpt_lt(dc_fixpt_one, output)) + output = dc_fixpt_one; rgb->r = output; rgb->g = output; @@ -605,7 +605,7 @@ static void build_de_pq(struct pwl_float_data_ex *de_pq, uint32_t i; struct fixed31_32 output; - struct fixed31_32 scaling_factor = dal_fixed31_32_from_int(125); + struct fixed31_32 scaling_factor = dc_fixpt_from_int(125); if (!de_pq_initialized) { precompute_de_pq(); @@ -616,9 +616,9 @@ static void build_de_pq(struct pwl_float_data_ex *de_pq, for (i = 0; i <= hw_points_num; i++) { output = de_pq_table[i]; /* should really not happen? */ - if (dal_fixed31_32_lt(output, dal_fixed31_32_zero)) - output = dal_fixed31_32_zero; - else if (dal_fixed31_32_lt(scaling_factor, output)) + if (dc_fixpt_lt(output, dc_fixpt_zero)) + output = dc_fixpt_zero; + else if (dc_fixpt_lt(scaling_factor, output)) output = scaling_factor; de_pq[i].r = output; de_pq[i].g = output; @@ -670,9 +670,9 @@ static void build_degamma(struct pwl_float_data_ex *curve, end_index = begin_index + 12 * NUM_PTS_IN_REGION; while (i != begin_index) { - curve[i].r = dal_fixed31_32_zero; - curve[i].g = dal_fixed31_32_zero; - curve[i].b = dal_fixed31_32_zero; + curve[i].r = dc_fixpt_zero; + curve[i].g = dc_fixpt_zero; + curve[i].b = dc_fixpt_zero; i++; } @@ -684,19 +684,19 @@ static void build_degamma(struct pwl_float_data_ex *curve, i++; } while (i != hw_points_num + 1) { - curve[i].r = dal_fixed31_32_one; - curve[i].g = dal_fixed31_32_one; - curve[i].b = dal_fixed31_32_one; + curve[i].r = dc_fixpt_one; + curve[i].g = dc_fixpt_one; + curve[i].b = dc_fixpt_one; i++; } } -static bool scale_gamma(struct pwl_float_data *pwl_rgb, +static void scale_gamma(struct pwl_float_data *pwl_rgb, const struct dc_gamma *ramp, struct dividers dividers) { - const struct fixed31_32 max_driver = dal_fixed31_32_from_int(0xFFFF); - const struct fixed31_32 max_os = dal_fixed31_32_from_int(0xFF00); + const struct fixed31_32 max_driver = dc_fixpt_from_int(0xFFFF); + const struct fixed31_32 max_os = dc_fixpt_from_int(0xFF00); struct fixed31_32 scaler = max_os; uint32_t i; struct pwl_float_data *rgb = pwl_rgb; @@ -705,9 +705,9 @@ static bool scale_gamma(struct pwl_float_data *pwl_rgb, i = 0; do { - if (dal_fixed31_32_lt(max_os, ramp->entries.red[i]) || - dal_fixed31_32_lt(max_os, ramp->entries.green[i]) || - dal_fixed31_32_lt(max_os, ramp->entries.blue[i])) { + if (dc_fixpt_lt(max_os, ramp->entries.red[i]) || + dc_fixpt_lt(max_os, ramp->entries.green[i]) || + dc_fixpt_lt(max_os, ramp->entries.blue[i])) { scaler = max_driver; break; } @@ -717,109 +717,170 @@ static bool scale_gamma(struct pwl_float_data *pwl_rgb, i = 0; do { - rgb->r = dal_fixed31_32_div( + rgb->r = dc_fixpt_div( ramp->entries.red[i], scaler); - rgb->g = dal_fixed31_32_div( + rgb->g = dc_fixpt_div( ramp->entries.green[i], scaler); - rgb->b = dal_fixed31_32_div( + rgb->b = dc_fixpt_div( ramp->entries.blue[i], scaler); ++rgb; ++i; } while (i != ramp->num_entries); - rgb->r = dal_fixed31_32_mul(rgb_last->r, + rgb->r = dc_fixpt_mul(rgb_last->r, dividers.divider1); - rgb->g = dal_fixed31_32_mul(rgb_last->g, + rgb->g = dc_fixpt_mul(rgb_last->g, dividers.divider1); - rgb->b = dal_fixed31_32_mul(rgb_last->b, + rgb->b = dc_fixpt_mul(rgb_last->b, dividers.divider1); ++rgb; - rgb->r = dal_fixed31_32_mul(rgb_last->r, + rgb->r = dc_fixpt_mul(rgb_last->r, dividers.divider2); - rgb->g = dal_fixed31_32_mul(rgb_last->g, + rgb->g = dc_fixpt_mul(rgb_last->g, dividers.divider2); - rgb->b = dal_fixed31_32_mul(rgb_last->b, + rgb->b = dc_fixpt_mul(rgb_last->b, dividers.divider2); ++rgb; - rgb->r = dal_fixed31_32_mul(rgb_last->r, + rgb->r = dc_fixpt_mul(rgb_last->r, dividers.divider3); - rgb->g = dal_fixed31_32_mul(rgb_last->g, + rgb->g = dc_fixpt_mul(rgb_last->g, dividers.divider3); - rgb->b = dal_fixed31_32_mul(rgb_last->b, + rgb->b = dc_fixpt_mul(rgb_last->b, dividers.divider3); - - return true; } -static bool scale_gamma_dx(struct pwl_float_data *pwl_rgb, +static void scale_gamma_dx(struct pwl_float_data *pwl_rgb, const struct dc_gamma *ramp, struct dividers dividers) { uint32_t i; - struct fixed31_32 min = dal_fixed31_32_zero; - struct fixed31_32 max = dal_fixed31_32_one; + struct fixed31_32 min = dc_fixpt_zero; + struct fixed31_32 max = dc_fixpt_one; - struct fixed31_32 delta = dal_fixed31_32_zero; - struct fixed31_32 offset = dal_fixed31_32_zero; + struct fixed31_32 delta = dc_fixpt_zero; + struct fixed31_32 offset = dc_fixpt_zero; for (i = 0 ; i < ramp->num_entries; i++) { - if (dal_fixed31_32_lt(ramp->entries.red[i], min)) + if (dc_fixpt_lt(ramp->entries.red[i], min)) min = ramp->entries.red[i]; - if (dal_fixed31_32_lt(ramp->entries.green[i], min)) + if (dc_fixpt_lt(ramp->entries.green[i], min)) min = ramp->entries.green[i]; - if (dal_fixed31_32_lt(ramp->entries.blue[i], min)) + if (dc_fixpt_lt(ramp->entries.blue[i], min)) min = ramp->entries.blue[i]; - if (dal_fixed31_32_lt(max, ramp->entries.red[i])) + if (dc_fixpt_lt(max, ramp->entries.red[i])) max = ramp->entries.red[i]; - if (dal_fixed31_32_lt(max, ramp->entries.green[i])) + if (dc_fixpt_lt(max, ramp->entries.green[i])) max = ramp->entries.green[i]; - if (dal_fixed31_32_lt(max, ramp->entries.blue[i])) + if (dc_fixpt_lt(max, ramp->entries.blue[i])) max = ramp->entries.blue[i]; } - if (dal_fixed31_32_lt(min, dal_fixed31_32_zero)) - delta = dal_fixed31_32_neg(min); + if (dc_fixpt_lt(min, dc_fixpt_zero)) + delta = dc_fixpt_neg(min); - offset = dal_fixed31_32_add(min, max); + offset = dc_fixpt_add(min, max); for (i = 0 ; i < ramp->num_entries; i++) { - pwl_rgb[i].r = dal_fixed31_32_div( - dal_fixed31_32_add( + pwl_rgb[i].r = dc_fixpt_div( + dc_fixpt_add( ramp->entries.red[i], delta), offset); - pwl_rgb[i].g = dal_fixed31_32_div( - dal_fixed31_32_add( + pwl_rgb[i].g = dc_fixpt_div( + dc_fixpt_add( ramp->entries.green[i], delta), offset); - pwl_rgb[i].b = dal_fixed31_32_div( - dal_fixed31_32_add( + pwl_rgb[i].b = dc_fixpt_div( + dc_fixpt_add( ramp->entries.blue[i], delta), offset); } - pwl_rgb[i].r = dal_fixed31_32_sub(dal_fixed31_32_mul_int( + pwl_rgb[i].r = dc_fixpt_sub(dc_fixpt_mul_int( pwl_rgb[i-1].r, 2), pwl_rgb[i-2].r); - pwl_rgb[i].g = dal_fixed31_32_sub(dal_fixed31_32_mul_int( + pwl_rgb[i].g = dc_fixpt_sub(dc_fixpt_mul_int( pwl_rgb[i-1].g, 2), pwl_rgb[i-2].g); - pwl_rgb[i].b = dal_fixed31_32_sub(dal_fixed31_32_mul_int( + pwl_rgb[i].b = dc_fixpt_sub(dc_fixpt_mul_int( pwl_rgb[i-1].b, 2), pwl_rgb[i-2].b); ++i; - pwl_rgb[i].r = dal_fixed31_32_sub(dal_fixed31_32_mul_int( + pwl_rgb[i].r = dc_fixpt_sub(dc_fixpt_mul_int( pwl_rgb[i-1].r, 2), pwl_rgb[i-2].r); - pwl_rgb[i].g = dal_fixed31_32_sub(dal_fixed31_32_mul_int( + pwl_rgb[i].g = dc_fixpt_sub(dc_fixpt_mul_int( pwl_rgb[i-1].g, 2), pwl_rgb[i-2].g); - pwl_rgb[i].b = dal_fixed31_32_sub(dal_fixed31_32_mul_int( + pwl_rgb[i].b = dc_fixpt_sub(dc_fixpt_mul_int( pwl_rgb[i-1].b, 2), pwl_rgb[i-2].b); +} - return true; +/* todo: all these scale_gamma functions are inherently the same but + * take different structures as params or different format for ramp + * values. We could probably implement it in a more generic fashion + */ +static void scale_user_regamma_ramp(struct pwl_float_data *pwl_rgb, + const struct regamma_ramp *ramp, + struct dividers dividers) +{ + unsigned short max_driver = 0xFFFF; + unsigned short max_os = 0xFF00; + unsigned short scaler = max_os; + uint32_t i; + struct pwl_float_data *rgb = pwl_rgb; + struct pwl_float_data *rgb_last = rgb + GAMMA_RGB_256_ENTRIES - 1; + + i = 0; + do { + if (ramp->gamma[i] > max_os || + ramp->gamma[i + 256] > max_os || + ramp->gamma[i + 512] > max_os) { + scaler = max_driver; + break; + } + i++; + } while (i != GAMMA_RGB_256_ENTRIES); + + i = 0; + do { + rgb->r = dc_fixpt_from_fraction( + ramp->gamma[i], scaler); + rgb->g = dc_fixpt_from_fraction( + ramp->gamma[i + 256], scaler); + rgb->b = dc_fixpt_from_fraction( + ramp->gamma[i + 512], scaler); + + ++rgb; + ++i; + } while (i != GAMMA_RGB_256_ENTRIES); + + rgb->r = dc_fixpt_mul(rgb_last->r, + dividers.divider1); + rgb->g = dc_fixpt_mul(rgb_last->g, + dividers.divider1); + rgb->b = dc_fixpt_mul(rgb_last->b, + dividers.divider1); + + ++rgb; + + rgb->r = dc_fixpt_mul(rgb_last->r, + dividers.divider2); + rgb->g = dc_fixpt_mul(rgb_last->g, + dividers.divider2); + rgb->b = dc_fixpt_mul(rgb_last->b, + dividers.divider2); + + ++rgb; + + rgb->r = dc_fixpt_mul(rgb_last->r, + dividers.divider3); + rgb->g = dc_fixpt_mul(rgb_last->g, + dividers.divider3); + rgb->b = dc_fixpt_mul(rgb_last->b, + dividers.divider3); } /* @@ -852,7 +913,7 @@ static void apply_lut_1d( struct fixed31_32 lut2; const int max_lut_index = 4095; const struct fixed31_32 max_lut_index_f = - dal_fixed31_32_from_int_nonconst(max_lut_index); + dc_fixpt_from_int(max_lut_index); int32_t index = 0, index_next = 0; struct fixed31_32 index_f; struct fixed31_32 delta_lut; @@ -870,10 +931,10 @@ static void apply_lut_1d( else regamma_y = &tf_pts->blue[i]; - norm_y = dal_fixed31_32_mul(max_lut_index_f, + norm_y = dc_fixpt_mul(max_lut_index_f, *regamma_y); - index = dal_fixed31_32_floor(norm_y); - index_f = dal_fixed31_32_from_int_nonconst(index); + index = dc_fixpt_floor(norm_y); + index_f = dc_fixpt_from_int(index); if (index < 0 || index > max_lut_index) continue; @@ -892,11 +953,11 @@ static void apply_lut_1d( } // we have everything now, so interpolate - delta_lut = dal_fixed31_32_sub(lut2, lut1); - delta_index = dal_fixed31_32_sub(norm_y, index_f); + delta_lut = dc_fixpt_sub(lut2, lut1); + delta_index = dc_fixpt_sub(norm_y, index_f); - *regamma_y = dal_fixed31_32_add(lut1, - dal_fixed31_32_mul(delta_index, delta_lut)); + *regamma_y = dc_fixpt_add(lut1, + dc_fixpt_mul(delta_index, delta_lut)); } } } @@ -912,7 +973,7 @@ static void build_evenly_distributed_points( uint32_t i = 0; do { - struct fixed31_32 value = dal_fixed31_32_from_fraction(i, + struct fixed31_32 value = dc_fixpt_from_fraction(i, numberof_points - 1); p->r = value; @@ -923,21 +984,21 @@ static void build_evenly_distributed_points( ++i; } while (i != numberof_points); - p->r = dal_fixed31_32_div(p_last->r, dividers.divider1); - p->g = dal_fixed31_32_div(p_last->g, dividers.divider1); - p->b = dal_fixed31_32_div(p_last->b, dividers.divider1); + p->r = dc_fixpt_div(p_last->r, dividers.divider1); + p->g = dc_fixpt_div(p_last->g, dividers.divider1); + p->b = dc_fixpt_div(p_last->b, dividers.divider1); ++p; - p->r = dal_fixed31_32_div(p_last->r, dividers.divider2); - p->g = dal_fixed31_32_div(p_last->g, dividers.divider2); - p->b = dal_fixed31_32_div(p_last->b, dividers.divider2); + p->r = dc_fixpt_div(p_last->r, dividers.divider2); + p->g = dc_fixpt_div(p_last->g, dividers.divider2); + p->b = dc_fixpt_div(p_last->b, dividers.divider2); ++p; - p->r = dal_fixed31_32_div(p_last->r, dividers.divider3); - p->g = dal_fixed31_32_div(p_last->g, dividers.divider3); - p->b = dal_fixed31_32_div(p_last->b, dividers.divider3); + p->r = dc_fixpt_div(p_last->r, dividers.divider3); + p->g = dc_fixpt_div(p_last->g, dividers.divider3); + p->b = dc_fixpt_div(p_last->b, dividers.divider3); } static inline void copy_rgb_regamma_to_coordinates_x( @@ -949,7 +1010,7 @@ static inline void copy_rgb_regamma_to_coordinates_x( uint32_t i = 0; const struct pwl_float_data_ex *rgb_regamma = rgb_ex; - while (i <= hw_points_num) { + while (i <= hw_points_num + 1) { coords->regamma_y_red = rgb_regamma->r; coords->regamma_y_green = rgb_regamma->g; coords->regamma_y_blue = rgb_regamma->b; @@ -1002,6 +1063,102 @@ static bool calculate_interpolated_hardware_curve( return true; } +/* The "old" interpolation uses a complicated scheme to build an array of + * coefficients while also using an array of 0-255 normalized to 0-1 + * Then there's another loop using both of the above + new scaled user ramp + * and we concatenate them. It also searches for points of interpolation and + * uses enums for positions. + * + * This function uses a different approach: + * user ramp is always applied on X with 0/255, 1/255, 2/255, ..., 255/255 + * To find index for hwX , we notice the following: + * i/255 <= hwX < (i+1)/255 <=> i <= 255*hwX < i+1 + * See apply_lut_1d which is the same principle, but on 4K entry 1D LUT + * + * Once the index is known, combined Y is simply: + * user_ramp(index) + (hwX-index/255)*(user_ramp(index+1) - user_ramp(index) + * + * We should switch to this method in all cases, it's simpler and faster + * ToDo one day - for now this only applies to ADL regamma to avoid regression + * for regular use cases (sRGB and PQ) + */ +static void interpolate_user_regamma(uint32_t hw_points_num, + struct pwl_float_data *rgb_user, + bool apply_degamma, + struct dc_transfer_func_distributed_points *tf_pts) +{ + uint32_t i; + uint32_t color = 0; + int32_t index; + int32_t index_next; + struct fixed31_32 *tf_point; + struct fixed31_32 hw_x; + struct fixed31_32 norm_factor = + dc_fixpt_from_int(255); + struct fixed31_32 norm_x; + struct fixed31_32 index_f; + struct fixed31_32 lut1; + struct fixed31_32 lut2; + struct fixed31_32 delta_lut; + struct fixed31_32 delta_index; + + i = 0; + /* fixed_pt library has problems handling too small values */ + while (i != 32) { + tf_pts->red[i] = dc_fixpt_zero; + tf_pts->green[i] = dc_fixpt_zero; + tf_pts->blue[i] = dc_fixpt_zero; + ++i; + } + while (i <= hw_points_num + 1) { + for (color = 0; color < 3; color++) { + if (color == 0) + tf_point = &tf_pts->red[i]; + else if (color == 1) + tf_point = &tf_pts->green[i]; + else + tf_point = &tf_pts->blue[i]; + + if (apply_degamma) { + if (color == 0) + hw_x = coordinates_x[i].regamma_y_red; + else if (color == 1) + hw_x = coordinates_x[i].regamma_y_green; + else + hw_x = coordinates_x[i].regamma_y_blue; + } else + hw_x = coordinates_x[i].x; + + norm_x = dc_fixpt_mul(norm_factor, hw_x); + index = dc_fixpt_floor(norm_x); + if (index < 0 || index > 255) + continue; + + index_f = dc_fixpt_from_int(index); + index_next = (index == 255) ? index : index + 1; + + if (color == 0) { + lut1 = rgb_user[index].r; + lut2 = rgb_user[index_next].r; + } else if (color == 1) { + lut1 = rgb_user[index].g; + lut2 = rgb_user[index_next].g; + } else { + lut1 = rgb_user[index].b; + lut2 = rgb_user[index_next].b; + } + + // we have everything now, so interpolate + delta_lut = dc_fixpt_sub(lut2, lut1); + delta_index = dc_fixpt_sub(norm_x, index_f); + + *tf_point = dc_fixpt_add(lut1, + dc_fixpt_mul(delta_index, delta_lut)); + } + ++i; + } +} + static void build_new_custom_resulted_curve( uint32_t hw_points_num, struct dc_transfer_func_distributed_points *tf_pts) @@ -1011,16 +1168,39 @@ static void build_new_custom_resulted_curve( i = 0; while (i != hw_points_num + 1) { - tf_pts->red[i] = dal_fixed31_32_clamp( - tf_pts->red[i], dal_fixed31_32_zero, - dal_fixed31_32_one); - tf_pts->green[i] = dal_fixed31_32_clamp( - tf_pts->green[i], dal_fixed31_32_zero, - dal_fixed31_32_one); - tf_pts->blue[i] = dal_fixed31_32_clamp( - tf_pts->blue[i], dal_fixed31_32_zero, - dal_fixed31_32_one); + tf_pts->red[i] = dc_fixpt_clamp( + tf_pts->red[i], dc_fixpt_zero, + dc_fixpt_one); + tf_pts->green[i] = dc_fixpt_clamp( + tf_pts->green[i], dc_fixpt_zero, + dc_fixpt_one); + tf_pts->blue[i] = dc_fixpt_clamp( + tf_pts->blue[i], dc_fixpt_zero, + dc_fixpt_one); + + ++i; + } +} + +static void apply_degamma_for_user_regamma(struct pwl_float_data_ex *rgb_regamma, + uint32_t hw_points_num) +{ + uint32_t i; + + struct gamma_coefficients coeff; + struct pwl_float_data_ex *rgb = rgb_regamma; + const struct hw_x_point *coord_x = coordinates_x; + build_coefficients(&coeff, true); + + i = 0; + while (i != hw_points_num + 1) { + rgb->r = translate_from_linear_space_ex( + coord_x->x, &coeff, 0); + rgb->g = rgb->r; + rgb->b = rgb->r; + ++coord_x; + ++rgb; ++i; } } @@ -1062,6 +1242,7 @@ static bool map_regamma_hw_to_x_user( } } + /* this should be named differently, all it does is clamp to 0-1 */ build_new_custom_resulted_curve(hw_points_num, tf_pts); return true; @@ -1093,25 +1274,25 @@ bool mod_color_calculate_regamma_params(struct dc_transfer_func *output_tf, output_tf->type = TF_TYPE_DISTRIBUTED_POINTS; - rgb_user = kzalloc(sizeof(*rgb_user) * (ramp->num_entries + _EXTRA_POINTS), - GFP_KERNEL); + rgb_user = kvzalloc(sizeof(*rgb_user) * (ramp->num_entries + _EXTRA_POINTS), + GFP_KERNEL); if (!rgb_user) goto rgb_user_alloc_fail; - rgb_regamma = kzalloc(sizeof(*rgb_regamma) * (MAX_HW_POINTS + _EXTRA_POINTS), - GFP_KERNEL); + rgb_regamma = kvzalloc(sizeof(*rgb_regamma) * (MAX_HW_POINTS + _EXTRA_POINTS), + GFP_KERNEL); if (!rgb_regamma) goto rgb_regamma_alloc_fail; - axix_x = kzalloc(sizeof(*axix_x) * (ramp->num_entries + 3), - GFP_KERNEL); + axix_x = kvzalloc(sizeof(*axix_x) * (ramp->num_entries + 3), + GFP_KERNEL); if (!axix_x) goto axix_x_alloc_fail; - coeff = kzalloc(sizeof(*coeff) * (MAX_HW_POINTS + _EXTRA_POINTS), GFP_KERNEL); + coeff = kvzalloc(sizeof(*coeff) * (MAX_HW_POINTS + _EXTRA_POINTS), GFP_KERNEL); if (!coeff) goto coeff_alloc_fail; - dividers.divider1 = dal_fixed31_32_from_fraction(3, 2); - dividers.divider2 = dal_fixed31_32_from_int(2); - dividers.divider3 = dal_fixed31_32_from_fraction(5, 2); + dividers.divider1 = dc_fixpt_from_fraction(3, 2); + dividers.divider2 = dc_fixpt_from_int(2); + dividers.divider3 = dc_fixpt_from_fraction(5, 2); tf = output_tf->tf; @@ -1157,13 +1338,120 @@ bool mod_color_calculate_regamma_params(struct dc_transfer_func *output_tf, ret = true; - kfree(coeff); + kvfree(coeff); coeff_alloc_fail: - kfree(axix_x); + kvfree(axix_x); axix_x_alloc_fail: + kvfree(rgb_regamma); +rgb_regamma_alloc_fail: + kvfree(rgb_user); +rgb_user_alloc_fail: + return ret; +} + +bool calculate_user_regamma_coeff(struct dc_transfer_func *output_tf, + const struct regamma_lut *regamma) +{ + struct gamma_coefficients coeff; + const struct hw_x_point *coord_x = coordinates_x; + uint32_t i = 0; + + do { + coeff.a0[i] = dc_fixpt_from_fraction( + regamma->coeff.A0[i], 10000000); + coeff.a1[i] = dc_fixpt_from_fraction( + regamma->coeff.A1[i], 1000); + coeff.a2[i] = dc_fixpt_from_fraction( + regamma->coeff.A2[i], 1000); + coeff.a3[i] = dc_fixpt_from_fraction( + regamma->coeff.A3[i], 1000); + coeff.user_gamma[i] = dc_fixpt_from_fraction( + regamma->coeff.gamma[i], 1000); + + ++i; + } while (i != 3); + + i = 0; + /* fixed_pt library has problems handling too small values */ + while (i != 32) { + output_tf->tf_pts.red[i] = dc_fixpt_zero; + output_tf->tf_pts.green[i] = dc_fixpt_zero; + output_tf->tf_pts.blue[i] = dc_fixpt_zero; + ++coord_x; + ++i; + } + while (i != MAX_HW_POINTS + 1) { + output_tf->tf_pts.red[i] = translate_from_linear_space_ex( + coord_x->x, &coeff, 0); + output_tf->tf_pts.green[i] = translate_from_linear_space_ex( + coord_x->x, &coeff, 1); + output_tf->tf_pts.blue[i] = translate_from_linear_space_ex( + coord_x->x, &coeff, 2); + ++coord_x; + ++i; + } + + // this function just clamps output to 0-1 + build_new_custom_resulted_curve(MAX_HW_POINTS, &output_tf->tf_pts); + output_tf->type = TF_TYPE_DISTRIBUTED_POINTS; + + return true; +} + +bool calculate_user_regamma_ramp(struct dc_transfer_func *output_tf, + const struct regamma_lut *regamma) +{ + struct dc_transfer_func_distributed_points *tf_pts = &output_tf->tf_pts; + struct dividers dividers; + + struct pwl_float_data *rgb_user = NULL; + struct pwl_float_data_ex *rgb_regamma = NULL; + bool ret = false; + + if (regamma == NULL) + return false; + + output_tf->type = TF_TYPE_DISTRIBUTED_POINTS; + + rgb_user = kzalloc(sizeof(*rgb_user) * (GAMMA_RGB_256_ENTRIES + _EXTRA_POINTS), + GFP_KERNEL); + if (!rgb_user) + goto rgb_user_alloc_fail; + + rgb_regamma = kzalloc(sizeof(*rgb_regamma) * (MAX_HW_POINTS + _EXTRA_POINTS), + GFP_KERNEL); + if (!rgb_regamma) + goto rgb_regamma_alloc_fail; + + dividers.divider1 = dc_fixpt_from_fraction(3, 2); + dividers.divider2 = dc_fixpt_from_int(2); + dividers.divider3 = dc_fixpt_from_fraction(5, 2); + + scale_user_regamma_ramp(rgb_user, ®amma->ramp, dividers); + + if (regamma->flags.bits.applyDegamma == 1) { + apply_degamma_for_user_regamma(rgb_regamma, MAX_HW_POINTS); + copy_rgb_regamma_to_coordinates_x(coordinates_x, + MAX_HW_POINTS, rgb_regamma); + } + + interpolate_user_regamma(MAX_HW_POINTS, rgb_user, + regamma->flags.bits.applyDegamma, tf_pts); + + // no custom HDR curves! + tf_pts->end_exponent = 0; + tf_pts->x_point_at_y1_red = 1; + tf_pts->x_point_at_y1_green = 1; + tf_pts->x_point_at_y1_blue = 1; + + // this function just clamps output to 0-1 + build_new_custom_resulted_curve(MAX_HW_POINTS, tf_pts); + + ret = true; + kfree(rgb_regamma); rgb_regamma_alloc_fail: - kfree(rgb_user); + kvfree(rgb_user); rgb_user_alloc_fail: return ret; } @@ -1192,25 +1480,25 @@ bool mod_color_calculate_degamma_params(struct dc_transfer_func *input_tf, input_tf->type = TF_TYPE_DISTRIBUTED_POINTS; - rgb_user = kzalloc(sizeof(*rgb_user) * (ramp->num_entries + _EXTRA_POINTS), - GFP_KERNEL); + rgb_user = kvzalloc(sizeof(*rgb_user) * (ramp->num_entries + _EXTRA_POINTS), + GFP_KERNEL); if (!rgb_user) goto rgb_user_alloc_fail; - curve = kzalloc(sizeof(*curve) * (MAX_HW_POINTS + _EXTRA_POINTS), - GFP_KERNEL); + curve = kvzalloc(sizeof(*curve) * (MAX_HW_POINTS + _EXTRA_POINTS), + GFP_KERNEL); if (!curve) goto curve_alloc_fail; - axix_x = kzalloc(sizeof(*axix_x) * (ramp->num_entries + _EXTRA_POINTS), - GFP_KERNEL); + axix_x = kvzalloc(sizeof(*axix_x) * (ramp->num_entries + _EXTRA_POINTS), + GFP_KERNEL); if (!axix_x) goto axix_x_alloc_fail; - coeff = kzalloc(sizeof(*coeff) * (MAX_HW_POINTS + _EXTRA_POINTS), GFP_KERNEL); + coeff = kvzalloc(sizeof(*coeff) * (MAX_HW_POINTS + _EXTRA_POINTS), GFP_KERNEL); if (!coeff) goto coeff_alloc_fail; - dividers.divider1 = dal_fixed31_32_from_fraction(3, 2); - dividers.divider2 = dal_fixed31_32_from_int(2); - dividers.divider3 = dal_fixed31_32_from_fraction(5, 2); + dividers.divider1 = dc_fixpt_from_fraction(3, 2); + dividers.divider2 = dc_fixpt_from_int(2); + dividers.divider3 = dc_fixpt_from_fraction(5, 2); tf = input_tf->tf; @@ -1246,13 +1534,13 @@ bool mod_color_calculate_degamma_params(struct dc_transfer_func *input_tf, ret = true; - kfree(coeff); + kvfree(coeff); coeff_alloc_fail: - kfree(axix_x); + kvfree(axix_x); axix_x_alloc_fail: - kfree(curve); + kvfree(curve); curve_alloc_fail: - kfree(rgb_user); + kvfree(rgb_user); rgb_user_alloc_fail: return ret; @@ -1281,8 +1569,9 @@ bool mod_color_calculate_curve(enum dc_transfer_func_predefined trans, } ret = true; } else if (trans == TRANSFER_FUNCTION_PQ) { - rgb_regamma = kzalloc(sizeof(*rgb_regamma) * (MAX_HW_POINTS + - _EXTRA_POINTS), GFP_KERNEL); + rgb_regamma = kvzalloc(sizeof(*rgb_regamma) * + (MAX_HW_POINTS + _EXTRA_POINTS), + GFP_KERNEL); if (!rgb_regamma) goto rgb_regamma_alloc_fail; points->end_exponent = 7; @@ -1302,11 +1591,12 @@ bool mod_color_calculate_curve(enum dc_transfer_func_predefined trans, } ret = true; - kfree(rgb_regamma); + kvfree(rgb_regamma); } else if (trans == TRANSFER_FUNCTION_SRGB || trans == TRANSFER_FUNCTION_BT709) { - rgb_regamma = kzalloc(sizeof(*rgb_regamma) * (MAX_HW_POINTS + - _EXTRA_POINTS), GFP_KERNEL); + rgb_regamma = kvzalloc(sizeof(*rgb_regamma) * + (MAX_HW_POINTS + _EXTRA_POINTS), + GFP_KERNEL); if (!rgb_regamma) goto rgb_regamma_alloc_fail; points->end_exponent = 0; @@ -1324,7 +1614,7 @@ bool mod_color_calculate_curve(enum dc_transfer_func_predefined trans, } ret = true; - kfree(rgb_regamma); + kvfree(rgb_regamma); } rgb_regamma_alloc_fail: return ret; @@ -1348,8 +1638,9 @@ bool mod_color_calculate_degamma_curve(enum dc_transfer_func_predefined trans, } ret = true; } else if (trans == TRANSFER_FUNCTION_PQ) { - rgb_degamma = kzalloc(sizeof(*rgb_degamma) * (MAX_HW_POINTS + - _EXTRA_POINTS), GFP_KERNEL); + rgb_degamma = kvzalloc(sizeof(*rgb_degamma) * + (MAX_HW_POINTS + _EXTRA_POINTS), + GFP_KERNEL); if (!rgb_degamma) goto rgb_degamma_alloc_fail; @@ -1364,11 +1655,12 @@ bool mod_color_calculate_degamma_curve(enum dc_transfer_func_predefined trans, } ret = true; - kfree(rgb_degamma); + kvfree(rgb_degamma); } else if (trans == TRANSFER_FUNCTION_SRGB || trans == TRANSFER_FUNCTION_BT709) { - rgb_degamma = kzalloc(sizeof(*rgb_degamma) * (MAX_HW_POINTS + - _EXTRA_POINTS), GFP_KERNEL); + rgb_degamma = kvzalloc(sizeof(*rgb_degamma) * + (MAX_HW_POINTS + _EXTRA_POINTS), + GFP_KERNEL); if (!rgb_degamma) goto rgb_degamma_alloc_fail; @@ -1382,7 +1674,7 @@ bool mod_color_calculate_degamma_curve(enum dc_transfer_func_predefined trans, } ret = true; - kfree(rgb_degamma); + kvfree(rgb_degamma); } points->end_exponent = 0; points->x_point_at_y1_red = 1; diff --git a/drivers/gpu/drm/amd/display/modules/color/color_gamma.h b/drivers/gpu/drm/amd/display/modules/color/color_gamma.h index b7f9bc27d1019..b64048991a95d 100644 --- a/drivers/gpu/drm/amd/display/modules/color/color_gamma.h +++ b/drivers/gpu/drm/amd/display/modules/color/color_gamma.h @@ -32,6 +32,47 @@ struct dc_transfer_func_distributed_points; struct dc_rgb_fixed; enum dc_transfer_func_predefined; +/* For SetRegamma ADL interface support + * Must match escape type + */ +union regamma_flags { + unsigned int raw; + struct { + unsigned int gammaRampArray :1; // RegammaRamp is in use + unsigned int gammaFromEdid :1; //gamma from edid is in use + unsigned int gammaFromEdidEx :1; //gamma from edid is in use , but only for Display Id 1.2 + unsigned int gammaFromUser :1; //user custom gamma is used + unsigned int coeffFromUser :1; //coeff. A0-A3 from user is in use + unsigned int coeffFromEdid :1; //coeff. A0-A3 from edid is in use + unsigned int applyDegamma :1; //flag for additional degamma correction in driver + unsigned int gammaPredefinedSRGB :1; //flag for SRGB gamma + unsigned int gammaPredefinedPQ :1; //flag for PQ gamma + unsigned int gammaPredefinedPQ2084Interim :1; //flag for PQ gamma, lower max nits + unsigned int gammaPredefined36 :1; //flag for 3.6 gamma + unsigned int gammaPredefinedReset :1; //flag to return to previous gamma + } bits; +}; + +struct regamma_ramp { + unsigned short gamma[256*3]; // gamma ramp packed in same way as OS windows ,r , g & b +}; + +struct regamma_coeff { + int gamma[3]; + int A0[3]; + int A1[3]; + int A2[3]; + int A3[3]; +}; + +struct regamma_lut { + union regamma_flags flags; + union { + struct regamma_ramp ramp; + struct regamma_coeff coeff; + }; +}; + void setup_x_points_distribution(void); void precompute_pq(void); void precompute_de_pq(void); @@ -45,9 +86,14 @@ bool mod_color_calculate_degamma_params(struct dc_transfer_func *output_tf, bool mod_color_calculate_curve(enum dc_transfer_func_predefined trans, struct dc_transfer_func_distributed_points *points); -bool mod_color_calculate_degamma_curve(enum dc_transfer_func_predefined trans, +bool mod_color_calculate_degamma_curve(enum dc_transfer_func_predefined trans, struct dc_transfer_func_distributed_points *points); +bool calculate_user_regamma_coeff(struct dc_transfer_func *output_tf, + const struct regamma_lut *regamma); + +bool calculate_user_regamma_ramp(struct dc_transfer_func *output_tf, + const struct regamma_lut *regamma); #endif /* COLOR_MOD_COLOR_GAMMA_H_ */ diff --git a/drivers/gpu/drm/amd/display/modules/inc/mod_stats.h b/drivers/gpu/drm/amd/display/modules/inc/mod_stats.h index 3230e2adb8701..3812094b52e8f 100644 --- a/drivers/gpu/drm/amd/display/modules/inc/mod_stats.h +++ b/drivers/gpu/drm/amd/display/modules/inc/mod_stats.h @@ -46,6 +46,10 @@ void mod_stats_dump(struct mod_stats *mod_stats); void mod_stats_reset_data(struct mod_stats *mod_stats); +void mod_stats_update_event(struct mod_stats *mod_stats, + char *event_string, + unsigned int length); + void mod_stats_update_flip(struct mod_stats *mod_stats, unsigned long timestamp_in_ns); diff --git a/drivers/gpu/drm/amd/display/modules/stats/stats.c b/drivers/gpu/drm/amd/display/modules/stats/stats.c index 041f87b73d5fe..3f7d47fdc3679 100644 --- a/drivers/gpu/drm/amd/display/modules/stats/stats.c +++ b/drivers/gpu/drm/amd/display/modules/stats/stats.c @@ -36,9 +36,14 @@ #define DAL_STATS_ENTRIES_REGKEY_DEFAULT 0x00350000 #define DAL_STATS_ENTRIES_REGKEY_MAX 0x01000000 +#define DAL_STATS_EVENT_ENTRIES_DEFAULT 0x00000100 + #define MOD_STATS_NUM_VSYNCS 5 +#define MOD_STATS_EVENT_STRING_MAX 512 struct stats_time_cache { + unsigned int entry_id; + unsigned long flip_timestamp_in_ns; unsigned long vupdate_timestamp_in_ns; @@ -63,15 +68,26 @@ struct stats_time_cache { unsigned int flags; }; +struct stats_event_cache { + unsigned int entry_id; + char event_string[MOD_STATS_EVENT_STRING_MAX]; +}; + struct core_stats { struct mod_stats public; struct dc *dc; + bool enabled; + unsigned int entries; + unsigned int event_entries; + unsigned int entry_id; + struct stats_time_cache *time; unsigned int index; - bool enabled; - unsigned int entries; + struct stats_event_cache *events; + unsigned int event_index; + }; #define MOD_STATS_TO_CORE(mod_stats)\ @@ -99,12 +115,12 @@ struct mod_stats *mod_stats_create(struct dc *dc) unsigned int reg_data; int i = 0; + if (dc == NULL) + goto fail_construct; + core_stats = kzalloc(sizeof(struct core_stats), GFP_KERNEL); if (core_stats == NULL) - goto fail_alloc_context; - - if (dc == NULL) goto fail_construct; core_stats->dc = dc; @@ -115,33 +131,55 @@ struct mod_stats *mod_stats_create(struct dc *dc) ®_data, sizeof(unsigned int), &flag)) core_stats->enabled = reg_data; - core_stats->entries = DAL_STATS_ENTRIES_REGKEY_DEFAULT; - if (dm_read_persistent_data(dc->ctx, NULL, NULL, - DAL_STATS_ENTRIES_REGKEY, - ®_data, sizeof(unsigned int), &flag)) { - if (reg_data > DAL_STATS_ENTRIES_REGKEY_MAX) - core_stats->entries = DAL_STATS_ENTRIES_REGKEY_MAX; - else - core_stats->entries = reg_data; - } + if (core_stats->enabled) { + core_stats->entries = DAL_STATS_ENTRIES_REGKEY_DEFAULT; + if (dm_read_persistent_data(dc->ctx, NULL, NULL, + DAL_STATS_ENTRIES_REGKEY, + ®_data, sizeof(unsigned int), &flag)) { + if (reg_data > DAL_STATS_ENTRIES_REGKEY_MAX) + core_stats->entries = DAL_STATS_ENTRIES_REGKEY_MAX; + else + core_stats->entries = reg_data; + } + core_stats->time = kzalloc( + sizeof(struct stats_time_cache) * + core_stats->entries, + GFP_KERNEL); - core_stats->time = kzalloc(sizeof(struct stats_time_cache) * core_stats->entries, - GFP_KERNEL); + if (core_stats->time == NULL) + goto fail_construct_time; - if (core_stats->time == NULL) - goto fail_construct; + core_stats->event_entries = DAL_STATS_EVENT_ENTRIES_DEFAULT; + core_stats->events = kzalloc( + sizeof(struct stats_event_cache) * + core_stats->event_entries, + GFP_KERNEL); + + if (core_stats->events == NULL) + goto fail_construct_events; + + } else { + core_stats->entries = 0; + } /* Purposely leave index 0 unused so we don't need special logic to * handle calculation cases that depend on previous flip data. */ core_stats->index = 1; + core_stats->event_index = 0; + + // Keeps track of ordering within the different stats structures + core_stats->entry_id = 0; return &core_stats->public; -fail_construct: +fail_construct_events: + kfree(core_stats->time); + +fail_construct_time: kfree(core_stats); -fail_alloc_context: +fail_construct: return NULL; } @@ -153,6 +191,9 @@ void mod_stats_destroy(struct mod_stats *mod_stats) if (core_stats->time != NULL) kfree(core_stats->time); + if (core_stats->events != NULL) + kfree(core_stats->events); + kfree(core_stats); } } @@ -163,7 +204,11 @@ void mod_stats_dump(struct mod_stats *mod_stats) struct dal_logger *logger = NULL; struct core_stats *core_stats = NULL; struct stats_time_cache *time = NULL; + struct stats_event_cache *events = NULL; + unsigned int time_index = 1; + unsigned int event_index = 0; unsigned int index = 0; + struct log_entry log_entry; if (mod_stats == NULL) return; @@ -172,45 +217,62 @@ void mod_stats_dump(struct mod_stats *mod_stats) dc = core_stats->dc; logger = dc->ctx->logger; time = core_stats->time; - - //LogEntry* pLog = GetLog()->Open(LogMajor_ISR, LogMinor_ISR_FreeSyncSW); - - //if (!pLog->IsDummyEntry()) - { - dm_logger_write(logger, LOG_PROFILING, "==Display Caps==\n"); - dm_logger_write(logger, LOG_PROFILING, "\n"); - dm_logger_write(logger, LOG_PROFILING, "\n"); - - dm_logger_write(logger, LOG_PROFILING, "==Stats==\n"); - dm_logger_write(logger, LOG_PROFILING, - "render avgRender minWindow midPoint maxWindow vsyncToFlip flipToVsync #vsyncBetweenFlip #frame insertDuration vTotalMin vTotalMax eventTrigs vSyncTime1 vSyncTime2 vSyncTime3 vSyncTime4 vSyncTime5 flags\n"); - - for (int i = 0; i < core_stats->index && i < core_stats->entries; i++) { - dm_logger_write(logger, LOG_PROFILING, - "%u %u %u %u %u %u %u %u %u %u %u %u %u %u %u %u %u %u %u\n", - time[i].render_time_in_us, - time[i].avg_render_time_in_us_last_ten, - time[i].min_window, - time[i].lfc_mid_point_in_us, - time[i].max_window, - time[i].vsync_to_flip_time_in_us, - time[i].flip_to_vsync_time_in_us, - time[i].num_vsync_between_flips, - time[i].num_frames_inserted, - time[i].inserted_duration_in_us, - time[i].v_total_min, - time[i].v_total_max, - time[i].event_triggers, - time[i].v_sync_time_in_us[0], - time[i].v_sync_time_in_us[1], - time[i].v_sync_time_in_us[2], - time[i].v_sync_time_in_us[3], - time[i].v_sync_time_in_us[4], - time[i].flags); + events = core_stats->events; + + DISPLAY_STATS_BEGIN(log_entry); + + DISPLAY_STATS("==Display Caps==\n"); + + DISPLAY_STATS("==Display Stats==\n"); + + DISPLAY_STATS("%10s %10s %10s %10s %10s" + " %11s %11s %17s %10s %14s" + " %10s %10s %10s %10s %10s" + " %10s %10s %10s %10s\n", + "render", "avgRender", + "minWindow", "midPoint", "maxWindow", + "vsyncToFlip", "flipToVsync", "vsyncsBetweenFlip", + "numFrame", "insertDuration", + "vTotalMin", "vTotalMax", "eventTrigs", + "vSyncTime1", "vSyncTime2", "vSyncTime3", + "vSyncTime4", "vSyncTime5", "flags"); + + for (int i = 0; i < core_stats->entry_id; i++) { + if (event_index < core_stats->event_index && + i == events[event_index].entry_id) { + DISPLAY_STATS("%s\n", events[event_index].event_string); + event_index++; + } else if (time_index < core_stats->index && + i == time[time_index].entry_id) { + DISPLAY_STATS("%10u %10u %10u %10u %10u" + " %11u %11u %17u %10u %14u" + " %10u %10u %10u %10u %10u" + " %10u %10u %10u %10u\n", + time[time_index].render_time_in_us, + time[time_index].avg_render_time_in_us_last_ten, + time[time_index].min_window, + time[time_index].lfc_mid_point_in_us, + time[time_index].max_window, + time[time_index].vsync_to_flip_time_in_us, + time[time_index].flip_to_vsync_time_in_us, + time[time_index].num_vsync_between_flips, + time[time_index].num_frames_inserted, + time[time_index].inserted_duration_in_us, + time[time_index].v_total_min, + time[time_index].v_total_max, + time[time_index].event_triggers, + time[time_index].v_sync_time_in_us[0], + time[time_index].v_sync_time_in_us[1], + time[time_index].v_sync_time_in_us[2], + time[time_index].v_sync_time_in_us[3], + time[time_index].v_sync_time_in_us[4], + time[time_index].flags); + + time_index++; } } - //GetLog()->Close(pLog); - //GetLog()->UnSetLogMask(LogMajor_ISR, LogMinor_ISR_FreeSyncSW); + + DISPLAY_STATS_END(log_entry); } void mod_stats_reset_data(struct mod_stats *mod_stats) @@ -227,7 +289,46 @@ void mod_stats_reset_data(struct mod_stats *mod_stats) memset(core_stats->time, 0, sizeof(struct stats_time_cache) * core_stats->entries); - core_stats->index = 0; + memset(core_stats->events, 0, + sizeof(struct stats_event_cache) * core_stats->event_entries); + + core_stats->index = 1; + core_stats->event_index = 0; + + // Keeps track of ordering within the different stats structures + core_stats->entry_id = 0; +} + +void mod_stats_update_event(struct mod_stats *mod_stats, + char *event_string, + unsigned int length) +{ + struct core_stats *core_stats = NULL; + struct stats_event_cache *events = NULL; + unsigned int index = 0; + unsigned int copy_length = 0; + + if (mod_stats == NULL) + return; + + core_stats = MOD_STATS_TO_CORE(mod_stats); + + if (core_stats->event_index >= core_stats->event_entries) + return; + + events = core_stats->events; + index = core_stats->event_index; + + copy_length = length; + if (length > MOD_STATS_EVENT_STRING_MAX) + copy_length = MOD_STATS_EVENT_STRING_MAX; + + memcpy(&events[index].event_string, event_string, copy_length); + events[index].event_string[copy_length - 1] = '\0'; + + events[index].entry_id = core_stats->entry_id; + core_stats->event_index++; + core_stats->entry_id++; } void mod_stats_update_flip(struct mod_stats *mod_stats, @@ -250,7 +351,7 @@ void mod_stats_update_flip(struct mod_stats *mod_stats, time[index].flip_timestamp_in_ns = timestamp_in_ns; time[index].render_time_in_us = - timestamp_in_ns - time[index - 1].flip_timestamp_in_ns; + (timestamp_in_ns - time[index - 1].flip_timestamp_in_ns) / 1000; if (index >= 10) { for (unsigned int i = 0; i < 10; i++) @@ -261,12 +362,16 @@ void mod_stats_update_flip(struct mod_stats *mod_stats, if (time[index].num_vsync_between_flips > 0) time[index].vsync_to_flip_time_in_us = - timestamp_in_ns - time[index].vupdate_timestamp_in_ns; + (timestamp_in_ns - + time[index].vupdate_timestamp_in_ns) / 1000; else time[index].vsync_to_flip_time_in_us = - timestamp_in_ns - time[index - 1].vupdate_timestamp_in_ns; + (timestamp_in_ns - + time[index - 1].vupdate_timestamp_in_ns) / 1000; + time[index].entry_id = core_stats->entry_id; core_stats->index++; + core_stats->entry_id++; } void mod_stats_update_vupdate(struct mod_stats *mod_stats, @@ -275,6 +380,8 @@ void mod_stats_update_vupdate(struct mod_stats *mod_stats, struct core_stats *core_stats = NULL; struct stats_time_cache *time = NULL; unsigned int index = 0; + unsigned int num_vsyncs = 0; + unsigned int prev_vsync_in_ns = 0; if (mod_stats == NULL) return; @@ -286,14 +393,27 @@ void mod_stats_update_vupdate(struct mod_stats *mod_stats, time = core_stats->time; index = core_stats->index; + num_vsyncs = time[index].num_vsync_between_flips; + + if (num_vsyncs < MOD_STATS_NUM_VSYNCS) { + if (num_vsyncs == 0) { + prev_vsync_in_ns = + time[index - 1].vupdate_timestamp_in_ns; + + time[index].flip_to_vsync_time_in_us = + (timestamp_in_ns - + time[index - 1].flip_timestamp_in_ns) / + 1000; + } else { + prev_vsync_in_ns = + time[index].vupdate_timestamp_in_ns; + } - time[index].vupdate_timestamp_in_ns = timestamp_in_ns; - if (time[index].num_vsync_between_flips < MOD_STATS_NUM_VSYNCS) - time[index].v_sync_time_in_us[time[index].num_vsync_between_flips] = - timestamp_in_ns - time[index - 1].vupdate_timestamp_in_ns; - time[index].flip_to_vsync_time_in_us = - timestamp_in_ns - time[index - 1].flip_timestamp_in_ns; + time[index].v_sync_time_in_us[num_vsyncs] = + (timestamp_in_ns - prev_vsync_in_ns) / 1000; + } + time[index].vupdate_timestamp_in_ns = timestamp_in_ns; time[index].num_vsync_between_flips++; } diff --git a/drivers/gpu/drm/amd/include/amd_shared.h b/drivers/gpu/drm/amd/include/amd_shared.h index 9fa3aaef3f331..b178176b72ac6 100644 --- a/drivers/gpu/drm/amd/include/amd_shared.h +++ b/drivers/gpu/drm/amd/include/amd_shared.h @@ -92,7 +92,7 @@ enum amd_powergating_state { #define AMD_CG_SUPPORT_GFX_3D_CGLS (1 << 21) #define AMD_CG_SUPPORT_DRM_MGCG (1 << 22) #define AMD_CG_SUPPORT_DF_MGCG (1 << 23) - +#define AMD_CG_SUPPORT_VCN_MGCG (1 << 24) /* PG flags */ #define AMD_PG_SUPPORT_GFX_PG (1 << 0) #define AMD_PG_SUPPORT_GFX_SMG (1 << 1) @@ -108,6 +108,27 @@ enum amd_powergating_state { #define AMD_PG_SUPPORT_GFX_QUICK_MG (1 << 11) #define AMD_PG_SUPPORT_GFX_PIPELINE (1 << 12) #define AMD_PG_SUPPORT_MMHUB (1 << 13) +#define AMD_PG_SUPPORT_VCN (1 << 14) + +enum PP_FEATURE_MASK { + PP_SCLK_DPM_MASK = 0x1, + PP_MCLK_DPM_MASK = 0x2, + PP_PCIE_DPM_MASK = 0x4, + PP_SCLK_DEEP_SLEEP_MASK = 0x8, + PP_POWER_CONTAINMENT_MASK = 0x10, + PP_UVD_HANDSHAKE_MASK = 0x20, + PP_SMC_VOLTAGE_CONTROL_MASK = 0x40, + PP_VBI_TIME_SUPPORT_MASK = 0x80, + PP_ULV_MASK = 0x100, + PP_ENABLE_GFX_CG_THRU_SMU = 0x200, + PP_CLOCK_STRETCH_MASK = 0x400, + PP_OD_FUZZY_FAN_CONTROL_MASK = 0x800, + PP_SOCCLK_DPM_MASK = 0x1000, + PP_DCEFCLK_DPM_MASK = 0x2000, + PP_OVERDRIVE_MASK = 0x4000, + PP_GFXOFF_MASK = 0x8000, + PP_ACG_MASK = 0x10000, +}; struct amd_ip_funcs { /* Name of IP block */ diff --git a/drivers/gpu/drm/amd/include/asic_reg/dce/dce_12_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/dce/dce_12_0_offset.h index f730d06290201..b6f74bf4af023 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/dce/dce_12_0_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/dce/dce_12_0_offset.h @@ -2095,6 +2095,18 @@ #define mmDC_GPIO_AUX_CTRL_2_BASE_IDX 2 #define mmDC_GPIO_RXEN 0x212f #define mmDC_GPIO_RXEN_BASE_IDX 2 +#define mmDC_GPIO_AUX_CTRL_3 0x2130 +#define mmDC_GPIO_AUX_CTRL_3_BASE_IDX 2 +#define mmDC_GPIO_AUX_CTRL_4 0x2131 +#define mmDC_GPIO_AUX_CTRL_4_BASE_IDX 2 +#define mmDC_GPIO_AUX_CTRL_5 0x2132 +#define mmDC_GPIO_AUX_CTRL_5_BASE_IDX 2 +#define mmAUXI2C_PAD_ALL_PWR_OK 0x2133 +#define mmAUXI2C_PAD_ALL_PWR_OK_BASE_IDX 2 +#define mmDC_GPIO_PULLUPEN 0x2134 +#define mmDC_GPIO_PULLUPEN_BASE_IDX 2 +#define mmDC_GPIO_AUX_CTRL_6 0x2135 +#define mmDC_GPIO_AUX_CTRL_6_BASE_IDX 2 #define mmBPHYC_DAC_MACRO_CNTL 0x2136 #define mmBPHYC_DAC_MACRO_CNTL_BASE_IDX 2 #define mmDAC_MACRO_CNTL_RESERVED0 0x2136 diff --git a/drivers/gpu/drm/amd/include/asic_reg/dce/dce_12_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/dce/dce_12_0_sh_mask.h index 6d3162c42957e..bcd190a3fcdd6 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/dce/dce_12_0_sh_mask.h +++ b/drivers/gpu/drm/amd/include/asic_reg/dce/dce_12_0_sh_mask.h @@ -10971,6 +10971,158 @@ #define DC_GPIO_RXEN__DC_GPIO_BLON_RXEN_MASK 0x00100000L #define DC_GPIO_RXEN__DC_GPIO_DIGON_RXEN_MASK 0x00200000L #define DC_GPIO_RXEN__DC_GPIO_ENA_BL_RXEN_MASK 0x00400000L +//DC_GPIO_AUX_CTRL_3 +#define DC_GPIO_AUX_CTRL_3__AUX1_NEN_RTERM__SHIFT 0x0 +#define DC_GPIO_AUX_CTRL_3__AUX2_NEN_RTERM__SHIFT 0x1 +#define DC_GPIO_AUX_CTRL_3__AUX3_NEN_RTERM__SHIFT 0x2 +#define DC_GPIO_AUX_CTRL_3__AUX4_NEN_RTERM__SHIFT 0x3 +#define DC_GPIO_AUX_CTRL_3__AUX5_NEN_RTERM__SHIFT 0x4 +#define DC_GPIO_AUX_CTRL_3__AUX6_NEN_RTERM__SHIFT 0x5 +#define DC_GPIO_AUX_CTRL_3__AUX1_DP_DN_SWAP__SHIFT 0x8 +#define DC_GPIO_AUX_CTRL_3__AUX2_DP_DN_SWAP__SHIFT 0x9 +#define DC_GPIO_AUX_CTRL_3__AUX3_DP_DN_SWAP__SHIFT 0xa +#define DC_GPIO_AUX_CTRL_3__AUX4_DP_DN_SWAP__SHIFT 0xb +#define DC_GPIO_AUX_CTRL_3__AUX5_DP_DN_SWAP__SHIFT 0xc +#define DC_GPIO_AUX_CTRL_3__AUX6_DP_DN_SWAP__SHIFT 0xd +#define DC_GPIO_AUX_CTRL_3__AUX1_HYS_TUNE__SHIFT 0x10 +#define DC_GPIO_AUX_CTRL_3__AUX2_HYS_TUNE__SHIFT 0x12 +#define DC_GPIO_AUX_CTRL_3__AUX3_HYS_TUNE__SHIFT 0x14 +#define DC_GPIO_AUX_CTRL_3__AUX4_HYS_TUNE__SHIFT 0x16 +#define DC_GPIO_AUX_CTRL_3__AUX5_HYS_TUNE__SHIFT 0x18 +#define DC_GPIO_AUX_CTRL_3__AUX6_HYS_TUNE__SHIFT 0x1a +#define DC_GPIO_AUX_CTRL_3__AUX1_NEN_RTERM_MASK 0x00000001L +#define DC_GPIO_AUX_CTRL_3__AUX2_NEN_RTERM_MASK 0x00000002L +#define DC_GPIO_AUX_CTRL_3__AUX3_NEN_RTERM_MASK 0x00000004L +#define DC_GPIO_AUX_CTRL_3__AUX4_NEN_RTERM_MASK 0x00000008L +#define DC_GPIO_AUX_CTRL_3__AUX5_NEN_RTERM_MASK 0x00000010L +#define DC_GPIO_AUX_CTRL_3__AUX6_NEN_RTERM_MASK 0x00000020L +#define DC_GPIO_AUX_CTRL_3__AUX1_DP_DN_SWAP_MASK 0x00000100L +#define DC_GPIO_AUX_CTRL_3__AUX2_DP_DN_SWAP_MASK 0x00000200L +#define DC_GPIO_AUX_CTRL_3__AUX3_DP_DN_SWAP_MASK 0x00000400L +#define DC_GPIO_AUX_CTRL_3__AUX4_DP_DN_SWAP_MASK 0x00000800L +#define DC_GPIO_AUX_CTRL_3__AUX5_DP_DN_SWAP_MASK 0x00001000L +#define DC_GPIO_AUX_CTRL_3__AUX6_DP_DN_SWAP_MASK 0x00002000L +#define DC_GPIO_AUX_CTRL_3__AUX1_HYS_TUNE_MASK 0x00030000L +#define DC_GPIO_AUX_CTRL_3__AUX2_HYS_TUNE_MASK 0x000C0000L +#define DC_GPIO_AUX_CTRL_3__AUX3_HYS_TUNE_MASK 0x00300000L +#define DC_GPIO_AUX_CTRL_3__AUX4_HYS_TUNE_MASK 0x00C00000L +#define DC_GPIO_AUX_CTRL_3__AUX5_HYS_TUNE_MASK 0x03000000L +#define DC_GPIO_AUX_CTRL_3__AUX6_HYS_TUNE_MASK 0x0C000000L +//DC_GPIO_AUX_CTRL_4 +#define DC_GPIO_AUX_CTRL_4__AUX1_AUX_CTRL__SHIFT 0x0 +#define DC_GPIO_AUX_CTRL_4__AUX2_AUX_CTRL__SHIFT 0x4 +#define DC_GPIO_AUX_CTRL_4__AUX3_AUX_CTRL__SHIFT 0x8 +#define DC_GPIO_AUX_CTRL_4__AUX4_AUX_CTRL__SHIFT 0xc +#define DC_GPIO_AUX_CTRL_4__AUX5_AUX_CTRL__SHIFT 0x10 +#define DC_GPIO_AUX_CTRL_4__AUX6_AUX_CTRL__SHIFT 0x14 +#define DC_GPIO_AUX_CTRL_4__AUX1_AUX_CTRL_MASK 0x0000000FL +#define DC_GPIO_AUX_CTRL_4__AUX2_AUX_CTRL_MASK 0x000000F0L +#define DC_GPIO_AUX_CTRL_4__AUX3_AUX_CTRL_MASK 0x00000F00L +#define DC_GPIO_AUX_CTRL_4__AUX4_AUX_CTRL_MASK 0x0000F000L +#define DC_GPIO_AUX_CTRL_4__AUX5_AUX_CTRL_MASK 0x000F0000L +#define DC_GPIO_AUX_CTRL_4__AUX6_AUX_CTRL_MASK 0x00F00000L +//DC_GPIO_AUX_CTRL_5 +#define DC_GPIO_AUX_CTRL_5__AUX1_VOD_TUNE__SHIFT 0x0 +#define DC_GPIO_AUX_CTRL_5__AUX2_VOD_TUNE__SHIFT 0x2 +#define DC_GPIO_AUX_CTRL_5__AUX3_VOD_TUNE__SHIFT 0x4 +#define DC_GPIO_AUX_CTRL_5__AUX4_VOD_TUNE__SHIFT 0x6 +#define DC_GPIO_AUX_CTRL_5__AUX5_VOD_TUNE__SHIFT 0x8 +#define DC_GPIO_AUX_CTRL_5__AUX6_VOD_TUNE__SHIFT 0xa +#define DC_GPIO_AUX_CTRL_5__DDC_PAD1_I2CMODE__SHIFT 0xc +#define DC_GPIO_AUX_CTRL_5__DDC_PAD2_I2CMODE__SHIFT 0xd +#define DC_GPIO_AUX_CTRL_5__DDC_PAD3_I2CMODE__SHIFT 0xe +#define DC_GPIO_AUX_CTRL_5__DDC_PAD4_I2CMODE__SHIFT 0xf +#define DC_GPIO_AUX_CTRL_5__DDC_PAD5_I2CMODE__SHIFT 0x10 +#define DC_GPIO_AUX_CTRL_5__DDC_PAD6_I2CMODE__SHIFT 0x11 +#define DC_GPIO_AUX_CTRL_5__DDC1_I2C_VPH_1V2_EN__SHIFT 0x12 +#define DC_GPIO_AUX_CTRL_5__DDC2_I2C_VPH_1V2_EN__SHIFT 0x13 +#define DC_GPIO_AUX_CTRL_5__DDC3_I2C_VPH_1V2_EN__SHIFT 0x14 +#define DC_GPIO_AUX_CTRL_5__DDC4_I2C_VPH_1V2_EN__SHIFT 0x15 +#define DC_GPIO_AUX_CTRL_5__DDC5_I2C_VPH_1V2_EN__SHIFT 0x16 +#define DC_GPIO_AUX_CTRL_5__DDC6_I2C_VPH_1V2_EN__SHIFT 0x17 +#define DC_GPIO_AUX_CTRL_5__DDC1_PAD_I2C_CTRL__SHIFT 0x18 +#define DC_GPIO_AUX_CTRL_5__DDC2_PAD_I2C_CTRL__SHIFT 0x19 +#define DC_GPIO_AUX_CTRL_5__DDC3_PAD_I2C_CTRL__SHIFT 0x1a +#define DC_GPIO_AUX_CTRL_5__DDC4_PAD_I2C_CTRL__SHIFT 0x1b +#define DC_GPIO_AUX_CTRL_5__DDC5_PAD_I2C_CTRL__SHIFT 0x1c +#define DC_GPIO_AUX_CTRL_5__DDC6_PAD_I2C_CTRL__SHIFT 0x1d +#define DC_GPIO_AUX_CTRL_5__AUX1_VOD_TUNE_MASK 0x00000003L +#define DC_GPIO_AUX_CTRL_5__AUX2_VOD_TUNE_MASK 0x0000000CL +#define DC_GPIO_AUX_CTRL_5__AUX3_VOD_TUNE_MASK 0x00000030L +#define DC_GPIO_AUX_CTRL_5__AUX4_VOD_TUNE_MASK 0x000000C0L +#define DC_GPIO_AUX_CTRL_5__AUX5_VOD_TUNE_MASK 0x00000300L +#define DC_GPIO_AUX_CTRL_5__AUX6_VOD_TUNE_MASK 0x00000C00L +#define DC_GPIO_AUX_CTRL_5__DDC_PAD1_I2CMODE_MASK 0x00001000L +#define DC_GPIO_AUX_CTRL_5__DDC_PAD2_I2CMODE_MASK 0x00002000L +#define DC_GPIO_AUX_CTRL_5__DDC_PAD3_I2CMODE_MASK 0x00004000L +#define DC_GPIO_AUX_CTRL_5__DDC_PAD4_I2CMODE_MASK 0x00008000L +#define DC_GPIO_AUX_CTRL_5__DDC_PAD5_I2CMODE_MASK 0x00010000L +#define DC_GPIO_AUX_CTRL_5__DDC_PAD6_I2CMODE_MASK 0x00020000L +#define DC_GPIO_AUX_CTRL_5__DDC1_I2C_VPH_1V2_EN_MASK 0x00040000L +#define DC_GPIO_AUX_CTRL_5__DDC2_I2C_VPH_1V2_EN_MASK 0x00080000L +#define DC_GPIO_AUX_CTRL_5__DDC3_I2C_VPH_1V2_EN_MASK 0x00100000L +#define DC_GPIO_AUX_CTRL_5__DDC4_I2C_VPH_1V2_EN_MASK 0x00200000L +#define DC_GPIO_AUX_CTRL_5__DDC5_I2C_VPH_1V2_EN_MASK 0x00400000L +#define DC_GPIO_AUX_CTRL_5__DDC6_I2C_VPH_1V2_EN_MASK 0x00800000L +#define DC_GPIO_AUX_CTRL_5__DDC1_PAD_I2C_CTRL_MASK 0x01000000L +#define DC_GPIO_AUX_CTRL_5__DDC2_PAD_I2C_CTRL_MASK 0x02000000L +#define DC_GPIO_AUX_CTRL_5__DDC3_PAD_I2C_CTRL_MASK 0x04000000L +#define DC_GPIO_AUX_CTRL_5__DDC4_PAD_I2C_CTRL_MASK 0x08000000L +#define DC_GPIO_AUX_CTRL_5__DDC5_PAD_I2C_CTRL_MASK 0x10000000L +#define DC_GPIO_AUX_CTRL_5__DDC6_PAD_I2C_CTRL_MASK 0x20000000L +//AUXI2C_PAD_ALL_PWR_OK +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY1_ALL_PWR_OK__SHIFT 0x0 +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY2_ALL_PWR_OK__SHIFT 0x1 +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY3_ALL_PWR_OK__SHIFT 0x2 +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY4_ALL_PWR_OK__SHIFT 0x3 +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY5_ALL_PWR_OK__SHIFT 0x4 +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY6_ALL_PWR_OK__SHIFT 0x5 +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY1_ALL_PWR_OK_MASK 0x00000001L +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY2_ALL_PWR_OK_MASK 0x00000002L +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY3_ALL_PWR_OK_MASK 0x00000004L +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY4_ALL_PWR_OK_MASK 0x00000008L +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY5_ALL_PWR_OK_MASK 0x00000010L +#define AUXI2C_PAD_ALL_PWR_OK__AUXI2C_PHY6_ALL_PWR_OK_MASK 0x00000020L +//DC_GPIO_PULLUPEN +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICA_PU_EN__SHIFT 0x0 +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICB_PU_EN__SHIFT 0x1 +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICC_PU_EN__SHIFT 0x2 +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICD_PU_EN__SHIFT 0x3 +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICE_PU_EN__SHIFT 0x4 +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICF_PU_EN__SHIFT 0x5 +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICG_PU_EN__SHIFT 0x6 +#define DC_GPIO_PULLUPEN__DC_GPIO_HSYNCA_PU_EN__SHIFT 0x8 +#define DC_GPIO_PULLUPEN__DC_GPIO_VSYNCA_PU_EN__SHIFT 0x9 +#define DC_GPIO_PULLUPEN__DC_GPIO_HPD1_PU_EN__SHIFT 0xe +#define DC_GPIO_PULLUPEN__DC_GPIO_BLON_PU_EN__SHIFT 0x14 +#define DC_GPIO_PULLUPEN__DC_GPIO_DIGON_PU_EN__SHIFT 0x15 +#define DC_GPIO_PULLUPEN__DC_GPIO_ENA_BL_PU_EN__SHIFT 0x16 +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICA_PU_EN_MASK 0x00000001L +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICB_PU_EN_MASK 0x00000002L +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICC_PU_EN_MASK 0x00000004L +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICD_PU_EN_MASK 0x00000008L +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICE_PU_EN_MASK 0x00000010L +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICF_PU_EN_MASK 0x00000020L +#define DC_GPIO_PULLUPEN__DC_GPIO_GENERICG_PU_EN_MASK 0x00000040L +#define DC_GPIO_PULLUPEN__DC_GPIO_HSYNCA_PU_EN_MASK 0x00000100L +#define DC_GPIO_PULLUPEN__DC_GPIO_VSYNCA_PU_EN_MASK 0x00000200L +#define DC_GPIO_PULLUPEN__DC_GPIO_HPD1_PU_EN_MASK 0x00004000L +#define DC_GPIO_PULLUPEN__DC_GPIO_BLON_PU_EN_MASK 0x00100000L +#define DC_GPIO_PULLUPEN__DC_GPIO_DIGON_PU_EN_MASK 0x00200000L +#define DC_GPIO_PULLUPEN__DC_GPIO_ENA_BL_PU_EN_MASK 0x00400000L +//DC_GPIO_AUX_CTRL_6 +#define DC_GPIO_AUX_CTRL_6__AUX1_PAD_RXSEL__SHIFT 0x0 +#define DC_GPIO_AUX_CTRL_6__AUX2_PAD_RXSEL__SHIFT 0x2 +#define DC_GPIO_AUX_CTRL_6__AUX3_PAD_RXSEL__SHIFT 0x4 +#define DC_GPIO_AUX_CTRL_6__AUX4_PAD_RXSEL__SHIFT 0x6 +#define DC_GPIO_AUX_CTRL_6__AUX5_PAD_RXSEL__SHIFT 0x8 +#define DC_GPIO_AUX_CTRL_6__AUX6_PAD_RXSEL__SHIFT 0xa +#define DC_GPIO_AUX_CTRL_6__AUX1_PAD_RXSEL_MASK 0x00000003L +#define DC_GPIO_AUX_CTRL_6__AUX2_PAD_RXSEL_MASK 0x0000000CL +#define DC_GPIO_AUX_CTRL_6__AUX3_PAD_RXSEL_MASK 0x00000030L +#define DC_GPIO_AUX_CTRL_6__AUX4_PAD_RXSEL_MASK 0x000000C0L +#define DC_GPIO_AUX_CTRL_6__AUX5_PAD_RXSEL_MASK 0x00000300L +#define DC_GPIO_AUX_CTRL_6__AUX6_PAD_RXSEL_MASK 0x00000C00L //BPHYC_DAC_MACRO_CNTL #define BPHYC_DAC_MACRO_CNTL__BPHYC_DAC_WHITE_LEVEL__SHIFT 0x0 #define BPHYC_DAC_MACRO_CNTL__BPHYC_DAC_WHITE_FINE_CONTROL__SHIFT 0x8 diff --git a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_1_0_offset.h b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_1_0_offset.h index 4ccf9681c45d0..721c611710452 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_1_0_offset.h +++ b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_1_0_offset.h @@ -3895,6 +3895,10 @@ #define mmCM0_CM_MEM_PWR_CTRL_BASE_IDX 2 #define mmCM0_CM_MEM_PWR_STATUS 0x0d33 #define mmCM0_CM_MEM_PWR_STATUS_BASE_IDX 2 +#define mmCM0_CM_TEST_DEBUG_INDEX 0x0d35 +#define mmCM0_CM_TEST_DEBUG_INDEX_BASE_IDX 2 +#define mmCM0_CM_TEST_DEBUG_DATA 0x0d36 +#define mmCM0_CM_TEST_DEBUG_DATA_BASE_IDX 2 // addressBlock: dce_dc_dpp0_dispdec_dpp_dcperfmon_dc_perfmon_dispdec @@ -4367,7 +4371,10 @@ #define mmCM1_CM_MEM_PWR_CTRL_BASE_IDX 2 #define mmCM1_CM_MEM_PWR_STATUS 0x0e4e #define mmCM1_CM_MEM_PWR_STATUS_BASE_IDX 2 - +#define mmCM1_CM_TEST_DEBUG_INDEX 0x0e50 +#define mmCM1_CM_TEST_DEBUG_INDEX_BASE_IDX 2 +#define mmCM1_CM_TEST_DEBUG_DATA 0x0e51 +#define mmCM1_CM_TEST_DEBUG_DATA_BASE_IDX 2 // addressBlock: dce_dc_dpp1_dispdec_dpp_dcperfmon_dc_perfmon_dispdec // base address: 0x399c @@ -4839,7 +4846,10 @@ #define mmCM2_CM_MEM_PWR_CTRL_BASE_IDX 2 #define mmCM2_CM_MEM_PWR_STATUS 0x0f69 #define mmCM2_CM_MEM_PWR_STATUS_BASE_IDX 2 - +#define mmCM2_CM_TEST_DEBUG_INDEX 0x0f6b +#define mmCM2_CM_TEST_DEBUG_INDEX_BASE_IDX 2 +#define mmCM2_CM_TEST_DEBUG_DATA 0x0f6c +#define mmCM2_CM_TEST_DEBUG_DATA_BASE_IDX 2 // addressBlock: dce_dc_dpp2_dispdec_dpp_dcperfmon_dc_perfmon_dispdec // base address: 0x3e08 @@ -5311,7 +5321,10 @@ #define mmCM3_CM_MEM_PWR_CTRL_BASE_IDX 2 #define mmCM3_CM_MEM_PWR_STATUS 0x1084 #define mmCM3_CM_MEM_PWR_STATUS_BASE_IDX 2 - +#define mmCM3_CM_TEST_DEBUG_INDEX 0x1086 +#define mmCM3_CM_TEST_DEBUG_INDEX_BASE_IDX 2 +#define mmCM3_CM_TEST_DEBUG_DATA 0x1087 +#define mmCM3_CM_TEST_DEBUG_DATA_BASE_IDX 2 // addressBlock: dce_dc_dpp3_dispdec_dpp_dcperfmon_dc_perfmon_dispdec // base address: 0x4274 diff --git a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_1_0_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_1_0_sh_mask.h index e2a2f114bd8ed..e7c0cad41081c 100644 --- a/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_1_0_sh_mask.h +++ b/drivers/gpu/drm/amd/include/asic_reg/dcn/dcn_1_0_sh_mask.h @@ -14049,6 +14049,14 @@ #define CM0_CM_MEM_PWR_STATUS__RGAM_MEM_PWR_STATE__SHIFT 0x2 #define CM0_CM_MEM_PWR_STATUS__SHARED_MEM_PWR_STATE_MASK 0x00000003L #define CM0_CM_MEM_PWR_STATUS__RGAM_MEM_PWR_STATE_MASK 0x0000000CL +//CM0_CM_TEST_DEBUG_INDEX +#define CM0_CM_TEST_DEBUG_INDEX__CM_TEST_DEBUG_INDEX__SHIFT 0x0 +#define CM0_CM_TEST_DEBUG_INDEX__CM_TEST_DEBUG_WRITE_EN__SHIFT 0x8 +#define CM0_CM_TEST_DEBUG_INDEX__CM_TEST_DEBUG_INDEX_MASK 0x000000FFL +#define CM0_CM_TEST_DEBUG_INDEX__CM_TEST_DEBUG_WRITE_EN_MASK 0x00000100L +//CM0_CM_TEST_DEBUG_DATA +#define CM0_CM_TEST_DEBUG_DATA__CM_TEST_DEBUG_DATA__SHIFT 0x0 +#define CM0_CM_TEST_DEBUG_DATA__CM_TEST_DEBUG_DATA_MASK 0xFFFFFFFFL // addressBlock: dce_dc_dpp0_dispdec_dpp_dcperfmon_dc_perfmon_dispdec diff --git a/drivers/gpu/drm/amd/include/asic_reg/df/df_1_7_default.h b/drivers/gpu/drm/amd/include/asic_reg/df/df_1_7_default.h new file mode 100644 index 0000000000000..9e19e723081b6 --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/df/df_1_7_default.h @@ -0,0 +1,26 @@ +/* + * Copyright (C) 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef _df_1_7_DEFAULT_HEADER +#define _df_1_7_DEFAULT_HEADER + +#define mmFabricConfigAccessControl_DEFAULT 0x00000000 + +#endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/df/df_1_7_offset.h b/drivers/gpu/drm/amd/include/asic_reg/df/df_1_7_offset.h new file mode 100644 index 0000000000000..e6044e27a913e --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/df/df_1_7_offset.h @@ -0,0 +1,37 @@ +/* + * Copyright (C) 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef _df_1_7_OFFSET_HEADER +#define _df_1_7_OFFSET_HEADER + +#define mmFabricConfigAccessControl 0x0410 +#define mmFabricConfigAccessControl_BASE_IDX 0 + +#define mmDF_PIE_AON0_DfGlobalClkGater 0x00fc +#define mmDF_PIE_AON0_DfGlobalClkGater_BASE_IDX 0 + +#define mmDF_CS_AON0_DramBaseAddress0 0x0044 +#define mmDF_CS_AON0_DramBaseAddress0_BASE_IDX 0 + +#define mmDF_CS_AON0_CoherentSlaveModeCtrlA0 0x0214 +#define mmDF_CS_AON0_CoherentSlaveModeCtrlA0_BASE_IDX 0 + + +#endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/df/df_1_7_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/df/df_1_7_sh_mask.h new file mode 100644 index 0000000000000..a78c99480e2dd --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/df/df_1_7_sh_mask.h @@ -0,0 +1,52 @@ +/* + * Copyright (C) 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef _df_1_7_SH_MASK_HEADER +#define _df_1_7_SH_MASK_HEADER + +/* FabricConfigAccessControl */ +#define FabricConfigAccessControl__CfgRegInstAccEn__SHIFT 0x0 +#define FabricConfigAccessControl__CfgRegInstAccRegLock__SHIFT 0x1 +#define FabricConfigAccessControl__CfgRegInstID__SHIFT 0x10 +#define FabricConfigAccessControl__CfgRegInstAccEn_MASK 0x00000001L +#define FabricConfigAccessControl__CfgRegInstAccRegLock_MASK 0x00000002L +#define FabricConfigAccessControl__CfgRegInstID_MASK 0x00FF0000L + +/* DF_PIE_AON0_DfGlobalClkGater */ +#define DF_PIE_AON0_DfGlobalClkGater__MGCGMode__SHIFT 0x0 +#define DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK 0x0000000FL + +/* DF_CS_AON0_DramBaseAddress0 */ +#define DF_CS_AON0_DramBaseAddress0__AddrRngVal__SHIFT 0x0 +#define DF_CS_AON0_DramBaseAddress0__LgcyMmioHoleEn__SHIFT 0x1 +#define DF_CS_AON0_DramBaseAddress0__IntLvNumChan__SHIFT 0x4 +#define DF_CS_AON0_DramBaseAddress0__IntLvAddrSel__SHIFT 0x8 +#define DF_CS_AON0_DramBaseAddress0__DramBaseAddr__SHIFT 0xc +#define DF_CS_AON0_DramBaseAddress0__AddrRngVal_MASK 0x00000001L +#define DF_CS_AON0_DramBaseAddress0__LgcyMmioHoleEn_MASK 0x00000002L +#define DF_CS_AON0_DramBaseAddress0__IntLvNumChan_MASK 0x000000F0L +#define DF_CS_AON0_DramBaseAddress0__IntLvAddrSel_MASK 0x00000700L +#define DF_CS_AON0_DramBaseAddress0__DramBaseAddr_MASK 0xFFFFF000L + +//DF_CS_AON0_CoherentSlaveModeCtrlA0 +#define DF_CS_AON0_CoherentSlaveModeCtrlA0__ForceParWrRMW__SHIFT 0x3 +#define DF_CS_AON0_CoherentSlaveModeCtrlA0__ForceParWrRMW_MASK 0x00000008L + +#endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_default.h b/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_default.h new file mode 100644 index 0000000000000..e58c207ac9809 --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_default.h @@ -0,0 +1,26 @@ +/* + * Copyright (C) 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef _df_3_6_DEFAULT_HEADER +#define _df_3_6_DEFAULT_HEADER + +#define mmFabricConfigAccessControl_DEFAULT 0x00000000 + +#endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_offset.h b/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_offset.h new file mode 100644 index 0000000000000..a9575db8d7aab --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_offset.h @@ -0,0 +1,33 @@ +/* + * Copyright (C) 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef _df_3_6_OFFSET_HEADER +#define _df_3_6_OFFSET_HEADER + +#define mmFabricConfigAccessControl 0x0410 +#define mmFabricConfigAccessControl_BASE_IDX 0 + +#define mmDF_PIE_AON0_DfGlobalClkGater 0x00fc +#define mmDF_PIE_AON0_DfGlobalClkGater_BASE_IDX 0 + +#define mmDF_CS_UMC_AON0_DramBaseAddress0 0x0044 +#define mmDF_CS_UMC_AON0_DramBaseAddress0_BASE_IDX 0 + +#endif diff --git a/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_sh_mask.h b/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_sh_mask.h new file mode 100644 index 0000000000000..88f7c69df6b96 --- /dev/null +++ b/drivers/gpu/drm/amd/include/asic_reg/df/df_3_6_sh_mask.h @@ -0,0 +1,48 @@ +/* + * Copyright (C) 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef _df_3_6_SH_MASK_HEADER +#define _df_3_6_SH_MASK_HEADER + +/* FabricConfigAccessControl */ +#define FabricConfigAccessControl__CfgRegInstAccEn__SHIFT 0x0 +#define FabricConfigAccessControl__CfgRegInstAccRegLock__SHIFT 0x1 +#define FabricConfigAccessControl__CfgRegInstID__SHIFT 0x10 +#define FabricConfigAccessControl__CfgRegInstAccEn_MASK 0x00000001L +#define FabricConfigAccessControl__CfgRegInstAccRegLock_MASK 0x00000002L +#define FabricConfigAccessControl__CfgRegInstID_MASK 0x00FF0000L + +/* DF_PIE_AON0_DfGlobalClkGater */ +#define DF_PIE_AON0_DfGlobalClkGater__MGCGMode__SHIFT 0x0 +#define DF_PIE_AON0_DfGlobalClkGater__MGCGMode_MASK 0x0000000FL + +/* DF_CS_AON0_DramBaseAddress0 */ +#define DF_CS_UMC_AON0_DramBaseAddress0__AddrRngVal__SHIFT 0x0 +#define DF_CS_UMC_AON0_DramBaseAddress0__LgcyMmioHoleEn__SHIFT 0x1 +#define DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan__SHIFT 0x4 +#define DF_CS_UMC_AON0_DramBaseAddress0__IntLvAddrSel__SHIFT 0x8 +#define DF_CS_UMC_AON0_DramBaseAddress0__DramBaseAddr__SHIFT 0xc +#define DF_CS_UMC_AON0_DramBaseAddress0__AddrRngVal_MASK 0x00000001L +#define DF_CS_UMC_AON0_DramBaseAddress0__LgcyMmioHoleEn_MASK 0x00000002L +#define DF_CS_UMC_AON0_DramBaseAddress0__IntLvNumChan_MASK 0x000000F0L +#define DF_CS_UMC_AON0_DramBaseAddress0__IntLvAddrSel_MASK 0x00000700L +#define DF_CS_UMC_AON0_DramBaseAddress0__DramBaseAddr_MASK 0xFFFFF000L + +#endif diff --git a/drivers/gpu/drm/amd/include/atombios.h b/drivers/gpu/drm/amd/include/atombios.h index f696bbb643efb..7931502fa54fa 100644 --- a/drivers/gpu/drm/amd/include/atombios.h +++ b/drivers/gpu/drm/amd/include/atombios.h @@ -632,6 +632,13 @@ typedef struct _COMPUTE_MEMORY_CLOCK_PARAM_PARAMETERS_V2_2 ULONG ulReserved; }COMPUTE_MEMORY_CLOCK_PARAM_PARAMETERS_V2_2; +typedef struct _COMPUTE_MEMORY_CLOCK_PARAM_PARAMETERS_V2_3 +{ + COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V4 ulClock; + USHORT usMclk_fcw_frac; //fractional divider of fcw = usSclk_fcw_frac/65536 + USHORT usMclk_fcw_int; //integer divider of fcwc +}COMPUTE_MEMORY_CLOCK_PARAM_PARAMETERS_V2_3; + //Input parameter of DynamicMemorySettingsTable //when ATOM_COMPUTE_CLOCK_FREQ.ulComputeClockFlag = COMPUTE_MEMORY_PLL_PARAM typedef struct _DYNAMICE_MEMORY_SETTINGS_PARAMETER diff --git a/drivers/gpu/drm/amd/include/atomfirmware.h b/drivers/gpu/drm/amd/include/atomfirmware.h index 0f5ad54d3fd3d..c6c1666ac1201 100644 --- a/drivers/gpu/drm/amd/include/atomfirmware.h +++ b/drivers/gpu/drm/amd/include/atomfirmware.h @@ -501,6 +501,32 @@ enum atom_cooling_solution_id{ LIQUID_COOLING = 0x01 }; +struct atom_firmware_info_v3_2 { + struct atom_common_table_header table_header; + uint32_t firmware_revision; + uint32_t bootup_sclk_in10khz; + uint32_t bootup_mclk_in10khz; + uint32_t firmware_capability; // enum atombios_firmware_capability + uint32_t main_call_parser_entry; /* direct address of main parser call in VBIOS binary. */ + uint32_t bios_scratch_reg_startaddr; // 1st bios scratch register dword address + uint16_t bootup_vddc_mv; + uint16_t bootup_vddci_mv; + uint16_t bootup_mvddc_mv; + uint16_t bootup_vddgfx_mv; + uint8_t mem_module_id; + uint8_t coolingsolution_id; /*0: Air cooling; 1: Liquid cooling ... */ + uint8_t reserved1[2]; + uint32_t mc_baseaddr_high; + uint32_t mc_baseaddr_low; + uint8_t board_i2c_feature_id; // enum of atom_board_i2c_feature_id_def + uint8_t board_i2c_feature_gpio_id; // i2c id find in gpio_lut data table gpio_id + uint8_t board_i2c_feature_slave_addr; + uint8_t reserved3; + uint16_t bootup_mvddq_mv; + uint16_t bootup_mvpp_mv; + uint32_t zfbstartaddrin16mb; + uint32_t reserved2[3]; +}; /* *************************************************************************** @@ -1169,7 +1195,64 @@ struct atom_gfx_info_v2_2 uint32_t rlc_gpu_timer_refclk; }; +struct atom_gfx_info_v2_3 { + struct atom_common_table_header table_header; + uint8_t gfxip_min_ver; + uint8_t gfxip_max_ver; + uint8_t max_shader_engines; + uint8_t max_tile_pipes; + uint8_t max_cu_per_sh; + uint8_t max_sh_per_se; + uint8_t max_backends_per_se; + uint8_t max_texture_channel_caches; + uint32_t regaddr_cp_dma_src_addr; + uint32_t regaddr_cp_dma_src_addr_hi; + uint32_t regaddr_cp_dma_dst_addr; + uint32_t regaddr_cp_dma_dst_addr_hi; + uint32_t regaddr_cp_dma_command; + uint32_t regaddr_cp_status; + uint32_t regaddr_rlc_gpu_clock_32; + uint32_t rlc_gpu_timer_refclk; + uint8_t active_cu_per_sh; + uint8_t active_rb_per_se; + uint16_t gcgoldenoffset; + uint32_t rm21_sram_vmin_value; +}; +struct atom_gfx_info_v2_4 { + struct atom_common_table_header table_header; + uint8_t gfxip_min_ver; + uint8_t gfxip_max_ver; + uint8_t gc_num_se; + uint8_t max_tile_pipes; + uint8_t gc_num_cu_per_sh; + uint8_t gc_num_sh_per_se; + uint8_t gc_num_rb_per_se; + uint8_t gc_num_tccs; + uint32_t regaddr_cp_dma_src_addr; + uint32_t regaddr_cp_dma_src_addr_hi; + uint32_t regaddr_cp_dma_dst_addr; + uint32_t regaddr_cp_dma_dst_addr_hi; + uint32_t regaddr_cp_dma_command; + uint32_t regaddr_cp_status; + uint32_t regaddr_rlc_gpu_clock_32; + uint32_t rlc_gpu_timer_refclk; + uint8_t active_cu_per_sh; + uint8_t active_rb_per_se; + uint16_t gcgoldenoffset; + uint16_t gc_num_gprs; + uint16_t gc_gsprim_buff_depth; + uint16_t gc_parameter_cache_depth; + uint16_t gc_wave_size; + uint16_t gc_max_waves_per_simd; + uint16_t gc_lds_size; + uint8_t gc_num_max_gs_thds; + uint8_t gc_gs_table_depth; + uint8_t gc_double_offchip_lds_buffer; + uint8_t gc_max_scratch_slots_per_cu; + uint32_t sram_rm_fuses_val; + uint32_t sram_custom_rm_fuses_val; +}; /* *************************************************************************** @@ -1198,6 +1281,76 @@ struct atom_smu_info_v3_1 uint8_t fw_ctf_polarity; // GPIO polarity for CTF }; +struct atom_smu_info_v3_2 { + struct atom_common_table_header table_header; + uint8_t smuip_min_ver; + uint8_t smuip_max_ver; + uint8_t smu_rsd1; + uint8_t gpuclk_ss_mode; + uint16_t sclk_ss_percentage; + uint16_t sclk_ss_rate_10hz; + uint16_t gpuclk_ss_percentage; // in unit of 0.001% + uint16_t gpuclk_ss_rate_10hz; + uint32_t core_refclk_10khz; + uint8_t ac_dc_gpio_bit; // GPIO bit shift in SMU_GPIOPAD_A configured for AC/DC switching, =0xff means invalid + uint8_t ac_dc_polarity; // GPIO polarity for AC/DC switching + uint8_t vr0hot_gpio_bit; // GPIO bit shift in SMU_GPIOPAD_A configured for VR0 HOT event, =0xff means invalid + uint8_t vr0hot_polarity; // GPIO polarity for VR0 HOT event + uint8_t vr1hot_gpio_bit; // GPIO bit shift in SMU_GPIOPAD_A configured for VR1 HOT event , =0xff means invalid + uint8_t vr1hot_polarity; // GPIO polarity for VR1 HOT event + uint8_t fw_ctf_gpio_bit; // GPIO bit shift in SMU_GPIOPAD_A configured for CTF, =0xff means invalid + uint8_t fw_ctf_polarity; // GPIO polarity for CTF + uint8_t pcc_gpio_bit; // GPIO bit shift in SMU_GPIOPAD_A configured for PCC, =0xff means invalid + uint8_t pcc_gpio_polarity; // GPIO polarity for CTF + uint16_t smugoldenoffset; + uint32_t gpupll_vco_freq_10khz; + uint32_t bootup_smnclk_10khz; + uint32_t bootup_socclk_10khz; + uint32_t bootup_mp0clk_10khz; + uint32_t bootup_mp1clk_10khz; + uint32_t bootup_lclk_10khz; + uint32_t bootup_dcefclk_10khz; + uint32_t ctf_threshold_override_value; + uint32_t reserved[5]; +}; + +struct atom_smu_info_v3_3 { + struct atom_common_table_header table_header; + uint8_t smuip_min_ver; + uint8_t smuip_max_ver; + uint8_t smu_rsd1; + uint8_t gpuclk_ss_mode; + uint16_t sclk_ss_percentage; + uint16_t sclk_ss_rate_10hz; + uint16_t gpuclk_ss_percentage; // in unit of 0.001% + uint16_t gpuclk_ss_rate_10hz; + uint32_t core_refclk_10khz; + uint8_t ac_dc_gpio_bit; // GPIO bit shift in SMU_GPIOPAD_A configured for AC/DC switching, =0xff means invalid + uint8_t ac_dc_polarity; // GPIO polarity for AC/DC switching + uint8_t vr0hot_gpio_bit; // GPIO bit shift in SMU_GPIOPAD_A configured for VR0 HOT event, =0xff means invalid + uint8_t vr0hot_polarity; // GPIO polarity for VR0 HOT event + uint8_t vr1hot_gpio_bit; // GPIO bit shift in SMU_GPIOPAD_A configured for VR1 HOT event , =0xff means invalid + uint8_t vr1hot_polarity; // GPIO polarity for VR1 HOT event + uint8_t fw_ctf_gpio_bit; // GPIO bit shift in SMU_GPIOPAD_A configured for CTF, =0xff means invalid + uint8_t fw_ctf_polarity; // GPIO polarity for CTF + uint8_t pcc_gpio_bit; // GPIO bit shift in SMU_GPIOPAD_A configured for PCC, =0xff means invalid + uint8_t pcc_gpio_polarity; // GPIO polarity for CTF + uint16_t smugoldenoffset; + uint32_t gpupll_vco_freq_10khz; + uint32_t bootup_smnclk_10khz; + uint32_t bootup_socclk_10khz; + uint32_t bootup_mp0clk_10khz; + uint32_t bootup_mp1clk_10khz; + uint32_t bootup_lclk_10khz; + uint32_t bootup_dcefclk_10khz; + uint32_t ctf_threshold_override_value; + uint32_t syspll3_0_vco_freq_10khz; + uint32_t syspll3_1_vco_freq_10khz; + uint32_t bootup_fclk_10khz; + uint32_t bootup_waflclk_10khz; + uint32_t reserved[3]; +}; + /* *************************************************************************** Data Table smc_dpm_info structure @@ -1283,7 +1436,6 @@ struct atom_smc_dpm_info_v4_1 uint32_t boardreserved[10]; }; - /* *************************************************************************** Data Table asic_profiling_info structure @@ -1864,6 +2016,55 @@ enum atom_smu9_syspll0_clock_id SMU9_SYSPLL0_DISPCLK_ID = 11, // DISPCLK }; +enum atom_smu11_syspll_id { + SMU11_SYSPLL0_ID = 0, + SMU11_SYSPLL1_0_ID = 1, + SMU11_SYSPLL1_1_ID = 2, + SMU11_SYSPLL1_2_ID = 3, + SMU11_SYSPLL2_ID = 4, + SMU11_SYSPLL3_0_ID = 5, + SMU11_SYSPLL3_1_ID = 6, +}; + + +enum atom_smu11_syspll0_clock_id { + SMU11_SYSPLL0_SOCCLK_ID = 0, // SOCCLK + SMU11_SYSPLL0_MP0CLK_ID = 1, // MP0CLK + SMU11_SYSPLL0_DCLK_ID = 2, // DCLK + SMU11_SYSPLL0_VCLK_ID = 3, // VCLK + SMU11_SYSPLL0_ECLK_ID = 4, // ECLK + SMU11_SYSPLL0_DCEFCLK_ID = 5, // DCEFCLK +}; + + +enum atom_smu11_syspll1_0_clock_id { + SMU11_SYSPLL1_0_UCLKA_ID = 0, // UCLK_a +}; + +enum atom_smu11_syspll1_1_clock_id { + SMU11_SYSPLL1_0_UCLKB_ID = 0, // UCLK_b +}; + +enum atom_smu11_syspll1_2_clock_id { + SMU11_SYSPLL1_0_FCLK_ID = 0, // FCLK +}; + +enum atom_smu11_syspll2_clock_id { + SMU11_SYSPLL2_GFXCLK_ID = 0, // GFXCLK +}; + +enum atom_smu11_syspll3_0_clock_id { + SMU11_SYSPLL3_0_WAFCLK_ID = 0, // WAFCLK + SMU11_SYSPLL3_0_DISPCLK_ID = 1, // DISPCLK + SMU11_SYSPLL3_0_DPREFCLK_ID = 2, // DPREFCLK +}; + +enum atom_smu11_syspll3_1_clock_id { + SMU11_SYSPLL3_1_MP1CLK_ID = 0, // MP1CLK + SMU11_SYSPLL3_1_SMNCLK_ID = 1, // SMNCLK + SMU11_SYSPLL3_1_LCLK_ID = 2, // LCLK +}; + struct atom_get_smu_clock_info_output_parameters_v3_1 { union { diff --git a/drivers/gpu/drm/amd/include/cgs_common.h b/drivers/gpu/drm/amd/include/cgs_common.h index f2814ae7ecddb..a69deb3a2ac07 100644 --- a/drivers/gpu/drm/amd/include/cgs_common.h +++ b/drivers/gpu/drm/amd/include/cgs_common.h @@ -42,20 +42,6 @@ enum cgs_ind_reg { CGS_IND_REG__AUDIO_ENDPT }; -/** - * enum cgs_engine - Engines that can be statically power-gated - */ -enum cgs_engine { - CGS_ENGINE__UVD, - CGS_ENGINE__VCE, - CGS_ENGINE__VP8, - CGS_ENGINE__ACP_DMA, - CGS_ENGINE__ACP_DSP0, - CGS_ENGINE__ACP_DSP1, - CGS_ENGINE__ISP, - /* ... */ -}; - /* * enum cgs_ucode_id - Firmware types for different IPs */ @@ -76,17 +62,6 @@ enum cgs_ucode_id { CGS_UCODE_ID_MAXIMUM, }; -/* - * enum cgs_resource_type - GPU resource type - */ -enum cgs_resource_type { - CGS_RESOURCE_TYPE_MMIO = 0, - CGS_RESOURCE_TYPE_FB, - CGS_RESOURCE_TYPE_IO, - CGS_RESOURCE_TYPE_DOORBELL, - CGS_RESOURCE_TYPE_ROM, -}; - /** * struct cgs_firmware_info - Firmware information */ @@ -104,17 +79,6 @@ struct cgs_firmware_info { bool is_kicker; }; -struct cgs_mode_info { - uint32_t refresh_rate; - uint32_t vblank_time_us; -}; - -struct cgs_display_info { - uint32_t display_count; - uint32_t active_display_mask; - struct cgs_mode_info *mode_info; -}; - typedef unsigned long cgs_handle_t; /** @@ -170,119 +134,18 @@ typedef void (*cgs_write_ind_register_t)(struct cgs_device *cgs_device, enum cgs #define CGS_WREG32_FIELD_IND(device, space, reg, field, val) \ cgs_write_ind_register(device, space, ix##reg, (cgs_read_ind_register(device, space, ix##reg) & ~CGS_REG_FIELD_MASK(reg, field)) | (val) << CGS_REG_FIELD_SHIFT(reg, field)) -/** - * cgs_get_pci_resource() - provide access to a device resource (PCI BAR) - * @cgs_device: opaque device handle - * @resource_type: Type of Resource (MMIO, IO, ROM, FB, DOORBELL) - * @size: size of the region - * @offset: offset from the start of the region - * @resource_base: base address (not including offset) returned - * - * Return: 0 on success, -errno otherwise - */ -typedef int (*cgs_get_pci_resource_t)(struct cgs_device *cgs_device, - enum cgs_resource_type resource_type, - uint64_t size, - uint64_t offset, - uint64_t *resource_base); - -/** - * cgs_atom_get_data_table() - Get a pointer to an ATOM BIOS data table - * @cgs_device: opaque device handle - * @table: data table index - * @size: size of the table (output, may be NULL) - * @frev: table format revision (output, may be NULL) - * @crev: table content revision (output, may be NULL) - * - * Return: Pointer to start of the table, or NULL on failure - */ -typedef const void *(*cgs_atom_get_data_table_t)( - struct cgs_device *cgs_device, unsigned table, - uint16_t *size, uint8_t *frev, uint8_t *crev); - -/** - * cgs_atom_get_cmd_table_revs() - Get ATOM BIOS command table revisions - * @cgs_device: opaque device handle - * @table: data table index - * @frev: table format revision (output, may be NULL) - * @crev: table content revision (output, may be NULL) - * - * Return: 0 on success, -errno otherwise - */ -typedef int (*cgs_atom_get_cmd_table_revs_t)(struct cgs_device *cgs_device, unsigned table, - uint8_t *frev, uint8_t *crev); - -/** - * cgs_atom_exec_cmd_table() - Execute an ATOM BIOS command table - * @cgs_device: opaque device handle - * @table: command table index - * @args: arguments - * - * Return: 0 on success, -errno otherwise - */ -typedef int (*cgs_atom_exec_cmd_table_t)(struct cgs_device *cgs_device, - unsigned table, void *args); - -/** - * cgs_get_firmware_info - Get the firmware information from core driver - * @cgs_device: opaque device handle - * @type: the firmware type - * @info: returend firmware information - * - * Return: 0 on success, -errno otherwise - */ typedef int (*cgs_get_firmware_info)(struct cgs_device *cgs_device, enum cgs_ucode_id type, struct cgs_firmware_info *info); -typedef int (*cgs_rel_firmware)(struct cgs_device *cgs_device, - enum cgs_ucode_id type); - -typedef int(*cgs_set_powergating_state)(struct cgs_device *cgs_device, - enum amd_ip_block_type block_type, - enum amd_powergating_state state); - -typedef int(*cgs_set_clockgating_state)(struct cgs_device *cgs_device, - enum amd_ip_block_type block_type, - enum amd_clockgating_state state); - -typedef int(*cgs_get_active_displays_info)( - struct cgs_device *cgs_device, - struct cgs_display_info *info); - -typedef int (*cgs_notify_dpm_enabled)(struct cgs_device *cgs_device, bool enabled); - -typedef int (*cgs_is_virtualization_enabled_t)(void *cgs_device); - -typedef int (*cgs_enter_safe_mode)(struct cgs_device *cgs_device, bool en); - -typedef void (*cgs_lock_grbm_idx)(struct cgs_device *cgs_device, bool lock); - struct cgs_ops { /* MMIO access */ cgs_read_register_t read_register; cgs_write_register_t write_register; cgs_read_ind_register_t read_ind_register; cgs_write_ind_register_t write_ind_register; - /* PCI resources */ - cgs_get_pci_resource_t get_pci_resource; - /* ATOM BIOS */ - cgs_atom_get_data_table_t atom_get_data_table; - cgs_atom_get_cmd_table_revs_t atom_get_cmd_table_revs; - cgs_atom_exec_cmd_table_t atom_exec_cmd_table; /* Firmware Info */ cgs_get_firmware_info get_firmware_info; - cgs_rel_firmware rel_firmware; - /* cg pg interface*/ - cgs_set_powergating_state set_powergating_state; - cgs_set_clockgating_state set_clockgating_state; - /* display manager */ - cgs_get_active_displays_info get_active_displays_info; - /* notify dpm enabled */ - cgs_notify_dpm_enabled notify_dpm_enabled; - cgs_is_virtualization_enabled_t is_virtualization_enabled; - cgs_enter_safe_mode enter_safe_mode; - cgs_lock_grbm_idx lock_grbm_idx; }; struct cgs_os_ops; /* To be define in OS-specific CGS header */ @@ -309,40 +172,7 @@ struct cgs_device #define cgs_write_ind_register(dev,space,index,value) \ CGS_CALL(write_ind_register,dev,space,index,value) -#define cgs_atom_get_data_table(dev,table,size,frev,crev) \ - CGS_CALL(atom_get_data_table,dev,table,size,frev,crev) -#define cgs_atom_get_cmd_table_revs(dev,table,frev,crev) \ - CGS_CALL(atom_get_cmd_table_revs,dev,table,frev,crev) -#define cgs_atom_exec_cmd_table(dev,table,args) \ - CGS_CALL(atom_exec_cmd_table,dev,table,args) - #define cgs_get_firmware_info(dev, type, info) \ CGS_CALL(get_firmware_info, dev, type, info) -#define cgs_rel_firmware(dev, type) \ - CGS_CALL(rel_firmware, dev, type) -#define cgs_set_powergating_state(dev, block_type, state) \ - CGS_CALL(set_powergating_state, dev, block_type, state) -#define cgs_set_clockgating_state(dev, block_type, state) \ - CGS_CALL(set_clockgating_state, dev, block_type, state) -#define cgs_notify_dpm_enabled(dev, enabled) \ - CGS_CALL(notify_dpm_enabled, dev, enabled) - -#define cgs_get_active_displays_info(dev, info) \ - CGS_CALL(get_active_displays_info, dev, info) - -#define cgs_get_pci_resource(cgs_device, resource_type, size, offset, \ - resource_base) \ - CGS_CALL(get_pci_resource, cgs_device, resource_type, size, offset, \ - resource_base) - -#define cgs_is_virtualization_enabled(cgs_device) \ - CGS_CALL(is_virtualization_enabled, cgs_device) - -#define cgs_enter_safe_mode(cgs_device, en) \ - CGS_CALL(enter_safe_mode, cgs_device, en) - -#define cgs_lock_grbm_idx(cgs_device, lock) \ - CGS_CALL(lock_grbm_idx, cgs_device, lock) - #endif /* _CGS_COMMON_H */ diff --git a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h index 237289a72bb7b..5733fbee07f7f 100644 --- a/drivers/gpu/drm/amd/include/kgd_kfd_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_kfd_interface.h @@ -100,6 +100,21 @@ struct kgd2kfd_shared_resources { /* Bit n == 1 means Queue n is available for KFD */ DECLARE_BITMAP(queue_bitmap, KGD_MAX_QUEUES); + /* Doorbell assignments (SOC15 and later chips only). Only + * specific doorbells are routed to each SDMA engine. Others + * are routed to IH and VCN. They are not usable by the CP. + * + * Any doorbell number D that satisfies the following condition + * is reserved: (D & reserved_doorbell_mask) == reserved_doorbell_val + * + * KFD currently uses 1024 (= 0x3ff) doorbells per process. If + * doorbells 0x0f0-0x0f7 and 0x2f-0x2f7 are reserved, that means + * mask would be set to 0x1f8 and val set to 0x0f0. + */ + unsigned int sdma_doorbell[2][2]; + unsigned int reserved_doorbell_mask; + unsigned int reserved_doorbell_val; + /* Base address of doorbell aperture. */ phys_addr_t doorbell_physical_address; @@ -173,8 +188,6 @@ struct tile_config { * @set_pasid_vmid_mapping: Exposes pasid/vmid pair to the H/W for no cp * scheduling mode. Only used for no cp scheduling mode. * - * @init_pipeline: Initialized the compute pipelines. - * * @hqd_load: Loads the mqd structure to a H/W hqd slot. used only for no cp * sceduling mode. * @@ -274,9 +287,6 @@ struct kfd2kgd_calls { int (*set_pasid_vmid_mapping)(struct kgd_dev *kgd, unsigned int pasid, unsigned int vmid); - int (*init_pipeline)(struct kgd_dev *kgd, uint32_t pipe_id, - uint32_t hpd_size, uint64_t hpd_gpu_addr); - int (*init_interrupts)(struct kgd_dev *kgd, uint32_t pipe_id); int (*hqd_load)(struct kgd_dev *kgd, void *mqd, uint32_t pipe_id, @@ -382,6 +392,10 @@ struct kfd2kgd_calls { * * @resume: Notifies amdkfd about a resume action done to a kgd device * + * @quiesce_mm: Quiesce all user queue access to specified MM address space + * + * @resume_mm: Resume user queue access to specified MM address space + * * @schedule_evict_and_restore_process: Schedules work queue that will prepare * for safe eviction of KFD BOs that belong to the specified process. * @@ -399,6 +413,8 @@ struct kgd2kfd_calls { void (*interrupt)(struct kfd_dev *kfd, const void *ih_ring_entry); void (*suspend)(struct kfd_dev *kfd); int (*resume)(struct kfd_dev *kfd); + int (*quiesce_mm)(struct mm_struct *mm); + int (*resume_mm)(struct mm_struct *mm); int (*schedule_evict_and_restore_process)(struct mm_struct *mm, struct dma_fence *fence); }; diff --git a/drivers/gpu/drm/amd/include/kgd_pp_interface.h b/drivers/gpu/drm/amd/include/kgd_pp_interface.h index 5c840c022b52d..06f08f34a110d 100644 --- a/drivers/gpu/drm/amd/include/kgd_pp_interface.h +++ b/drivers/gpu/drm/amd/include/kgd_pp_interface.h @@ -94,6 +94,7 @@ enum pp_clock_type { PP_PCIE, OD_SCLK, OD_MCLK, + OD_RANGE, }; enum amd_pp_sensors { @@ -149,13 +150,6 @@ struct pp_states_info { uint32_t states[16]; }; -struct pp_gpu_power { - uint32_t vddc_power; - uint32_t vddci_power; - uint32_t max_gpu_power; - uint32_t average_gpu_power; -}; - #define PP_GROUP_MASK 0xF0000000 #define PP_GROUP_SHIFT 28 @@ -246,11 +240,6 @@ struct amd_pm_funcs { int (*load_firmware)(void *handle); int (*wait_for_fw_loading_complete)(void *handle); int (*set_clockgating_by_smu)(void *handle, uint32_t msg_id); - int (*notify_smu_memory_info)(void *handle, uint32_t virtual_addr_low, - uint32_t virtual_addr_hi, - uint32_t mc_addr_low, - uint32_t mc_addr_hi, - uint32_t size); int (*set_power_limit)(void *handle, uint32_t n); int (*get_power_limit)(void *handle, uint32_t *limit, bool default_limit); /* export to DC */ diff --git a/drivers/gpu/drm/amd/include/soc15_ih_clientid.h b/drivers/gpu/drm/amd/include/soc15_ih_clientid.h index a12d4f27cfa43..12e196c15bbe0 100644 --- a/drivers/gpu/drm/amd/include/soc15_ih_clientid.h +++ b/drivers/gpu/drm/amd/include/soc15_ih_clientid.h @@ -43,6 +43,7 @@ enum soc15_ih_clientid { SOC15_IH_CLIENTID_SE2SH = 0x0c, SOC15_IH_CLIENTID_SE3SH = 0x0d, SOC15_IH_CLIENTID_SYSHUB = 0x0e, + SOC15_IH_CLIENTID_UVD1 = 0x0e, SOC15_IH_CLIENTID_THM = 0x0f, SOC15_IH_CLIENTID_UVD = 0x10, SOC15_IH_CLIENTID_VCE0 = 0x11, diff --git a/drivers/gpu/drm/amd/include/v9_structs.h b/drivers/gpu/drm/amd/include/v9_structs.h index 2fb25abaf7c8b..ceaf4932258da 100644 --- a/drivers/gpu/drm/amd/include/v9_structs.h +++ b/drivers/gpu/drm/amd/include/v9_structs.h @@ -29,10 +29,10 @@ struct v9_sdma_mqd { uint32_t sdmax_rlcx_rb_base; uint32_t sdmax_rlcx_rb_base_hi; uint32_t sdmax_rlcx_rb_rptr; + uint32_t sdmax_rlcx_rb_rptr_hi; uint32_t sdmax_rlcx_rb_wptr; + uint32_t sdmax_rlcx_rb_wptr_hi; uint32_t sdmax_rlcx_rb_wptr_poll_cntl; - uint32_t sdmax_rlcx_rb_wptr_poll_addr_hi; - uint32_t sdmax_rlcx_rb_wptr_poll_addr_lo; uint32_t sdmax_rlcx_rb_rptr_addr_hi; uint32_t sdmax_rlcx_rb_rptr_addr_lo; uint32_t sdmax_rlcx_ib_cntl; @@ -44,29 +44,29 @@ struct v9_sdma_mqd { uint32_t sdmax_rlcx_skip_cntl; uint32_t sdmax_rlcx_context_status; uint32_t sdmax_rlcx_doorbell; - uint32_t sdmax_rlcx_virtual_addr; - uint32_t sdmax_rlcx_ape1_cntl; + uint32_t sdmax_rlcx_status; uint32_t sdmax_rlcx_doorbell_log; - uint32_t reserved_22; - uint32_t reserved_23; - uint32_t reserved_24; - uint32_t reserved_25; - uint32_t reserved_26; - uint32_t reserved_27; - uint32_t reserved_28; - uint32_t reserved_29; - uint32_t reserved_30; - uint32_t reserved_31; - uint32_t reserved_32; - uint32_t reserved_33; - uint32_t reserved_34; - uint32_t reserved_35; - uint32_t reserved_36; - uint32_t reserved_37; - uint32_t reserved_38; - uint32_t reserved_39; - uint32_t reserved_40; - uint32_t reserved_41; + uint32_t sdmax_rlcx_watermark; + uint32_t sdmax_rlcx_doorbell_offset; + uint32_t sdmax_rlcx_csa_addr_lo; + uint32_t sdmax_rlcx_csa_addr_hi; + uint32_t sdmax_rlcx_ib_sub_remain; + uint32_t sdmax_rlcx_preempt; + uint32_t sdmax_rlcx_dummy_reg; + uint32_t sdmax_rlcx_rb_wptr_poll_addr_hi; + uint32_t sdmax_rlcx_rb_wptr_poll_addr_lo; + uint32_t sdmax_rlcx_rb_aql_cntl; + uint32_t sdmax_rlcx_minor_ptr_update; + uint32_t sdmax_rlcx_midcmd_data0; + uint32_t sdmax_rlcx_midcmd_data1; + uint32_t sdmax_rlcx_midcmd_data2; + uint32_t sdmax_rlcx_midcmd_data3; + uint32_t sdmax_rlcx_midcmd_data4; + uint32_t sdmax_rlcx_midcmd_data5; + uint32_t sdmax_rlcx_midcmd_data6; + uint32_t sdmax_rlcx_midcmd_data7; + uint32_t sdmax_rlcx_midcmd_data8; + uint32_t sdmax_rlcx_midcmd_cntl; uint32_t reserved_42; uint32_t reserved_43; uint32_t reserved_44; diff --git a/drivers/gpu/drm/amd/include/vega20_ip_offset.h b/drivers/gpu/drm/amd/include/vega20_ip_offset.h new file mode 100644 index 0000000000000..2a2a9cc8bedb6 --- /dev/null +++ b/drivers/gpu/drm/amd/include/vega20_ip_offset.h @@ -0,0 +1,1051 @@ +/* + * Copyright (C) 2018 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included + * in all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS + * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN + * AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN + * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. + */ +#ifndef _vega20_ip_offset_HEADER +#define _vega20_ip_offset_HEADER + +#define MAX_INSTANCE 6 +#define MAX_SEGMENT 6 + + +struct IP_BASE_INSTANCE +{ + unsigned int segment[MAX_SEGMENT]; +}; + +struct IP_BASE +{ + struct IP_BASE_INSTANCE instance[MAX_INSTANCE]; +}; + + +static const struct IP_BASE ATHUB_BASE ={ { { { 0x00000C20, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE CLK_BASE ={ { { { 0x00016C00, 0x00016E00, 0x00017000, 0x00017200, 0x0001B000, 0x0001B200 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE DCE_BASE ={ { { { 0x00000012, 0x000000C0, 0x000034C0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE DF_BASE ={ { { { 0x00007000, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE FUSE_BASE ={ { { { 0x00017400, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE GC_BASE ={ { { { 0x00002000, 0x0000A000, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE HDP_BASE ={ { { { 0x00000F20, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE MMHUB_BASE ={ { { { 0x0001A000, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE MP0_BASE ={ { { { 0x00016000, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE MP1_BASE ={ { { { 0x00016000, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE NBIO_BASE ={ { { { 0x00000000, 0x00000014, 0x00000D20, 0x00010400, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE OSSSYS_BASE ={ { { { 0x000010A0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE SDMA0_BASE ={ { { { 0x00001260, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE SDMA1_BASE ={ { { { 0x00001860, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE SMUIO_BASE ={ { { { 0x00016800, 0x00016A00, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE THM_BASE ={ { { { 0x00016600, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE UMC_BASE ={ { { { 0x00014000, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE UVD_BASE ={ { { { 0x00007800, 0x00007E00, 0, 0, 0, 0 } }, + { { 0, 0x00009000, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +/* Adjust VCE_BASE to make vce_4_1 use vce_4_0 offset header files*/ +static const struct IP_BASE VCE_BASE ={ { { { 0x00007E00/* 0x00008800 */, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE XDMA_BASE ={ { { { 0x00003400, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; +static const struct IP_BASE RSMU_BASE ={ { { { 0x00012000, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } }, + { { 0, 0, 0, 0, 0, 0 } } } }; + + +#define ATHUB_BASE__INST0_SEG0 0x00000C20 +#define ATHUB_BASE__INST0_SEG1 0 +#define ATHUB_BASE__INST0_SEG2 0 +#define ATHUB_BASE__INST0_SEG3 0 +#define ATHUB_BASE__INST0_SEG4 0 +#define ATHUB_BASE__INST0_SEG5 0 + +#define ATHUB_BASE__INST1_SEG0 0 +#define ATHUB_BASE__INST1_SEG1 0 +#define ATHUB_BASE__INST1_SEG2 0 +#define ATHUB_BASE__INST1_SEG3 0 +#define ATHUB_BASE__INST1_SEG4 0 +#define ATHUB_BASE__INST1_SEG5 0 + +#define ATHUB_BASE__INST2_SEG0 0 +#define ATHUB_BASE__INST2_SEG1 0 +#define ATHUB_BASE__INST2_SEG2 0 +#define ATHUB_BASE__INST2_SEG3 0 +#define ATHUB_BASE__INST2_SEG4 0 +#define ATHUB_BASE__INST2_SEG5 0 + +#define ATHUB_BASE__INST3_SEG0 0 +#define ATHUB_BASE__INST3_SEG1 0 +#define ATHUB_BASE__INST3_SEG2 0 +#define ATHUB_BASE__INST3_SEG3 0 +#define ATHUB_BASE__INST3_SEG4 0 +#define ATHUB_BASE__INST3_SEG5 0 + +#define ATHUB_BASE__INST4_SEG0 0 +#define ATHUB_BASE__INST4_SEG1 0 +#define ATHUB_BASE__INST4_SEG2 0 +#define ATHUB_BASE__INST4_SEG3 0 +#define ATHUB_BASE__INST4_SEG4 0 +#define ATHUB_BASE__INST4_SEG5 0 + +#define ATHUB_BASE__INST5_SEG0 0 +#define ATHUB_BASE__INST5_SEG1 0 +#define ATHUB_BASE__INST5_SEG2 0 +#define ATHUB_BASE__INST5_SEG3 0 +#define ATHUB_BASE__INST5_SEG4 0 +#define ATHUB_BASE__INST5_SEG5 0 + +#define CLK_BASE__INST0_SEG0 0x00016C00 +#define CLK_BASE__INST0_SEG1 0x00016E00 +#define CLK_BASE__INST0_SEG2 0x00017000 +#define CLK_BASE__INST0_SEG3 0x00017200 +#define CLK_BASE__INST0_SEG4 0x0001B000 +#define CLK_BASE__INST0_SEG5 0x0001B200 + +#define CLK_BASE__INST1_SEG0 0 +#define CLK_BASE__INST1_SEG1 0 +#define CLK_BASE__INST1_SEG2 0 +#define CLK_BASE__INST1_SEG3 0 +#define CLK_BASE__INST1_SEG4 0 +#define CLK_BASE__INST1_SEG5 0 + +#define CLK_BASE__INST2_SEG0 0 +#define CLK_BASE__INST2_SEG1 0 +#define CLK_BASE__INST2_SEG2 0 +#define CLK_BASE__INST2_SEG3 0 +#define CLK_BASE__INST2_SEG4 0 +#define CLK_BASE__INST2_SEG5 0 + +#define CLK_BASE__INST3_SEG0 0 +#define CLK_BASE__INST3_SEG1 0 +#define CLK_BASE__INST3_SEG2 0 +#define CLK_BASE__INST3_SEG3 0 +#define CLK_BASE__INST3_SEG4 0 +#define CLK_BASE__INST3_SEG5 0 + +#define CLK_BASE__INST4_SEG0 0 +#define CLK_BASE__INST4_SEG1 0 +#define CLK_BASE__INST4_SEG2 0 +#define CLK_BASE__INST4_SEG3 0 +#define CLK_BASE__INST4_SEG4 0 +#define CLK_BASE__INST4_SEG5 0 + +#define CLK_BASE__INST5_SEG0 0 +#define CLK_BASE__INST5_SEG1 0 +#define CLK_BASE__INST5_SEG2 0 +#define CLK_BASE__INST5_SEG3 0 +#define CLK_BASE__INST5_SEG4 0 +#define CLK_BASE__INST5_SEG5 0 + +#define DCE_BASE__INST0_SEG0 0x00000012 +#define DCE_BASE__INST0_SEG1 0x000000C0 +#define DCE_BASE__INST0_SEG2 0x000034C0 +#define DCE_BASE__INST0_SEG3 0 +#define DCE_BASE__INST0_SEG4 0 +#define DCE_BASE__INST0_SEG5 0 + +#define DCE_BASE__INST1_SEG0 0 +#define DCE_BASE__INST1_SEG1 0 +#define DCE_BASE__INST1_SEG2 0 +#define DCE_BASE__INST1_SEG3 0 +#define DCE_BASE__INST1_SEG4 0 +#define DCE_BASE__INST1_SEG5 0 + +#define DCE_BASE__INST2_SEG0 0 +#define DCE_BASE__INST2_SEG1 0 +#define DCE_BASE__INST2_SEG2 0 +#define DCE_BASE__INST2_SEG3 0 +#define DCE_BASE__INST2_SEG4 0 +#define DCE_BASE__INST2_SEG5 0 + +#define DCE_BASE__INST3_SEG0 0 +#define DCE_BASE__INST3_SEG1 0 +#define DCE_BASE__INST3_SEG2 0 +#define DCE_BASE__INST3_SEG3 0 +#define DCE_BASE__INST3_SEG4 0 +#define DCE_BASE__INST3_SEG5 0 + +#define DCE_BASE__INST4_SEG0 0 +#define DCE_BASE__INST4_SEG1 0 +#define DCE_BASE__INST4_SEG2 0 +#define DCE_BASE__INST4_SEG3 0 +#define DCE_BASE__INST4_SEG4 0 +#define DCE_BASE__INST4_SEG5 0 + +#define DCE_BASE__INST5_SEG0 0 +#define DCE_BASE__INST5_SEG1 0 +#define DCE_BASE__INST5_SEG2 0 +#define DCE_BASE__INST5_SEG3 0 +#define DCE_BASE__INST5_SEG4 0 +#define DCE_BASE__INST5_SEG5 0 + +#define DF_BASE__INST0_SEG0 0x00007000 +#define DF_BASE__INST0_SEG1 0 +#define DF_BASE__INST0_SEG2 0 +#define DF_BASE__INST0_SEG3 0 +#define DF_BASE__INST0_SEG4 0 +#define DF_BASE__INST0_SEG5 0 + +#define DF_BASE__INST1_SEG0 0 +#define DF_BASE__INST1_SEG1 0 +#define DF_BASE__INST1_SEG2 0 +#define DF_BASE__INST1_SEG3 0 +#define DF_BASE__INST1_SEG4 0 +#define DF_BASE__INST1_SEG5 0 + +#define DF_BASE__INST2_SEG0 0 +#define DF_BASE__INST2_SEG1 0 +#define DF_BASE__INST2_SEG2 0 +#define DF_BASE__INST2_SEG3 0 +#define DF_BASE__INST2_SEG4 0 +#define DF_BASE__INST2_SEG5 0 + +#define DF_BASE__INST3_SEG0 0 +#define DF_BASE__INST3_SEG1 0 +#define DF_BASE__INST3_SEG2 0 +#define DF_BASE__INST3_SEG3 0 +#define DF_BASE__INST3_SEG4 0 +#define DF_BASE__INST3_SEG5 0 + +#define DF_BASE__INST4_SEG0 0 +#define DF_BASE__INST4_SEG1 0 +#define DF_BASE__INST4_SEG2 0 +#define DF_BASE__INST4_SEG3 0 +#define DF_BASE__INST4_SEG4 0 +#define DF_BASE__INST4_SEG5 0 + +#define DF_BASE__INST5_SEG0 0 +#define DF_BASE__INST5_SEG1 0 +#define DF_BASE__INST5_SEG2 0 +#define DF_BASE__INST5_SEG3 0 +#define DF_BASE__INST5_SEG4 0 +#define DF_BASE__INST5_SEG5 0 + +#define FUSE_BASE__INST0_SEG0 0x00017400 +#define FUSE_BASE__INST0_SEG1 0 +#define FUSE_BASE__INST0_SEG2 0 +#define FUSE_BASE__INST0_SEG3 0 +#define FUSE_BASE__INST0_SEG4 0 +#define FUSE_BASE__INST0_SEG5 0 + +#define FUSE_BASE__INST1_SEG0 0 +#define FUSE_BASE__INST1_SEG1 0 +#define FUSE_BASE__INST1_SEG2 0 +#define FUSE_BASE__INST1_SEG3 0 +#define FUSE_BASE__INST1_SEG4 0 +#define FUSE_BASE__INST1_SEG5 0 + +#define FUSE_BASE__INST2_SEG0 0 +#define FUSE_BASE__INST2_SEG1 0 +#define FUSE_BASE__INST2_SEG2 0 +#define FUSE_BASE__INST2_SEG3 0 +#define FUSE_BASE__INST2_SEG4 0 +#define FUSE_BASE__INST2_SEG5 0 + +#define FUSE_BASE__INST3_SEG0 0 +#define FUSE_BASE__INST3_SEG1 0 +#define FUSE_BASE__INST3_SEG2 0 +#define FUSE_BASE__INST3_SEG3 0 +#define FUSE_BASE__INST3_SEG4 0 +#define FUSE_BASE__INST3_SEG5 0 + +#define FUSE_BASE__INST4_SEG0 0 +#define FUSE_BASE__INST4_SEG1 0 +#define FUSE_BASE__INST4_SEG2 0 +#define FUSE_BASE__INST4_SEG3 0 +#define FUSE_BASE__INST4_SEG4 0 +#define FUSE_BASE__INST4_SEG5 0 + +#define FUSE_BASE__INST5_SEG0 0 +#define FUSE_BASE__INST5_SEG1 0 +#define FUSE_BASE__INST5_SEG2 0 +#define FUSE_BASE__INST5_SEG3 0 +#define FUSE_BASE__INST5_SEG4 0 +#define FUSE_BASE__INST5_SEG5 0 + +#define GC_BASE__INST0_SEG0 0x00002000 +#define GC_BASE__INST0_SEG1 0x0000A000 +#define GC_BASE__INST0_SEG2 0 +#define GC_BASE__INST0_SEG3 0 +#define GC_BASE__INST0_SEG4 0 +#define GC_BASE__INST0_SEG5 0 + +#define GC_BASE__INST1_SEG0 0 +#define GC_BASE__INST1_SEG1 0 +#define GC_BASE__INST1_SEG2 0 +#define GC_BASE__INST1_SEG3 0 +#define GC_BASE__INST1_SEG4 0 +#define GC_BASE__INST1_SEG5 0 + +#define GC_BASE__INST2_SEG0 0 +#define GC_BASE__INST2_SEG1 0 +#define GC_BASE__INST2_SEG2 0 +#define GC_BASE__INST2_SEG3 0 +#define GC_BASE__INST2_SEG4 0 +#define GC_BASE__INST2_SEG5 0 + +#define GC_BASE__INST3_SEG0 0 +#define GC_BASE__INST3_SEG1 0 +#define GC_BASE__INST3_SEG2 0 +#define GC_BASE__INST3_SEG3 0 +#define GC_BASE__INST3_SEG4 0 +#define GC_BASE__INST3_SEG5 0 + +#define GC_BASE__INST4_SEG0 0 +#define GC_BASE__INST4_SEG1 0 +#define GC_BASE__INST4_SEG2 0 +#define GC_BASE__INST4_SEG3 0 +#define GC_BASE__INST4_SEG4 0 +#define GC_BASE__INST4_SEG5 0 + +#define GC_BASE__INST5_SEG0 0 +#define GC_BASE__INST5_SEG1 0 +#define GC_BASE__INST5_SEG2 0 +#define GC_BASE__INST5_SEG3 0 +#define GC_BASE__INST5_SEG4 0 +#define GC_BASE__INST5_SEG5 0 + +#define HDP_BASE__INST0_SEG0 0x00000F20 +#define HDP_BASE__INST0_SEG1 0 +#define HDP_BASE__INST0_SEG2 0 +#define HDP_BASE__INST0_SEG3 0 +#define HDP_BASE__INST0_SEG4 0 +#define HDP_BASE__INST0_SEG5 0 + +#define HDP_BASE__INST1_SEG0 0 +#define HDP_BASE__INST1_SEG1 0 +#define HDP_BASE__INST1_SEG2 0 +#define HDP_BASE__INST1_SEG3 0 +#define HDP_BASE__INST1_SEG4 0 +#define HDP_BASE__INST1_SEG5 0 + +#define HDP_BASE__INST2_SEG0 0 +#define HDP_BASE__INST2_SEG1 0 +#define HDP_BASE__INST2_SEG2 0 +#define HDP_BASE__INST2_SEG3 0 +#define HDP_BASE__INST2_SEG4 0 +#define HDP_BASE__INST2_SEG5 0 + +#define HDP_BASE__INST3_SEG0 0 +#define HDP_BASE__INST3_SEG1 0 +#define HDP_BASE__INST3_SEG2 0 +#define HDP_BASE__INST3_SEG3 0 +#define HDP_BASE__INST3_SEG4 0 +#define HDP_BASE__INST3_SEG5 0 + +#define HDP_BASE__INST4_SEG0 0 +#define HDP_BASE__INST4_SEG1 0 +#define HDP_BASE__INST4_SEG2 0 +#define HDP_BASE__INST4_SEG3 0 +#define HDP_BASE__INST4_SEG4 0 +#define HDP_BASE__INST4_SEG5 0 + +#define HDP_BASE__INST5_SEG0 0 +#define HDP_BASE__INST5_SEG1 0 +#define HDP_BASE__INST5_SEG2 0 +#define HDP_BASE__INST5_SEG3 0 +#define HDP_BASE__INST5_SEG4 0 +#define HDP_BASE__INST5_SEG5 0 + +#define MMHUB_BASE__INST0_SEG0 0x0001A000 +#define MMHUB_BASE__INST0_SEG1 0 +#define MMHUB_BASE__INST0_SEG2 0 +#define MMHUB_BASE__INST0_SEG3 0 +#define MMHUB_BASE__INST0_SEG4 0 +#define MMHUB_BASE__INST0_SEG5 0 + +#define MMHUB_BASE__INST1_SEG0 0 +#define MMHUB_BASE__INST1_SEG1 0 +#define MMHUB_BASE__INST1_SEG2 0 +#define MMHUB_BASE__INST1_SEG3 0 +#define MMHUB_BASE__INST1_SEG4 0 +#define MMHUB_BASE__INST1_SEG5 0 + +#define MMHUB_BASE__INST2_SEG0 0 +#define MMHUB_BASE__INST2_SEG1 0 +#define MMHUB_BASE__INST2_SEG2 0 +#define MMHUB_BASE__INST2_SEG3 0 +#define MMHUB_BASE__INST2_SEG4 0 +#define MMHUB_BASE__INST2_SEG5 0 + +#define MMHUB_BASE__INST3_SEG0 0 +#define MMHUB_BASE__INST3_SEG1 0 +#define MMHUB_BASE__INST3_SEG2 0 +#define MMHUB_BASE__INST3_SEG3 0 +#define MMHUB_BASE__INST3_SEG4 0 +#define MMHUB_BASE__INST3_SEG5 0 + +#define MMHUB_BASE__INST4_SEG0 0 +#define MMHUB_BASE__INST4_SEG1 0 +#define MMHUB_BASE__INST4_SEG2 0 +#define MMHUB_BASE__INST4_SEG3 0 +#define MMHUB_BASE__INST4_SEG4 0 +#define MMHUB_BASE__INST4_SEG5 0 + +#define MMHUB_BASE__INST5_SEG0 0 +#define MMHUB_BASE__INST5_SEG1 0 +#define MMHUB_BASE__INST5_SEG2 0 +#define MMHUB_BASE__INST5_SEG3 0 +#define MMHUB_BASE__INST5_SEG4 0 +#define MMHUB_BASE__INST5_SEG5 0 + +#define MP0_BASE__INST0_SEG0 0x00016000 +#define MP0_BASE__INST0_SEG1 0 +#define MP0_BASE__INST0_SEG2 0 +#define MP0_BASE__INST0_SEG3 0 +#define MP0_BASE__INST0_SEG4 0 +#define MP0_BASE__INST0_SEG5 0 + +#define MP0_BASE__INST1_SEG0 0 +#define MP0_BASE__INST1_SEG1 0 +#define MP0_BASE__INST1_SEG2 0 +#define MP0_BASE__INST1_SEG3 0 +#define MP0_BASE__INST1_SEG4 0 +#define MP0_BASE__INST1_SEG5 0 + +#define MP0_BASE__INST2_SEG0 0 +#define MP0_BASE__INST2_SEG1 0 +#define MP0_BASE__INST2_SEG2 0 +#define MP0_BASE__INST2_SEG3 0 +#define MP0_BASE__INST2_SEG4 0 +#define MP0_BASE__INST2_SEG5 0 + +#define MP0_BASE__INST3_SEG0 0 +#define MP0_BASE__INST3_SEG1 0 +#define MP0_BASE__INST3_SEG2 0 +#define MP0_BASE__INST3_SEG3 0 +#define MP0_BASE__INST3_SEG4 0 +#define MP0_BASE__INST3_SEG5 0 + +#define MP0_BASE__INST4_SEG0 0 +#define MP0_BASE__INST4_SEG1 0 +#define MP0_BASE__INST4_SEG2 0 +#define MP0_BASE__INST4_SEG3 0 +#define MP0_BASE__INST4_SEG4 0 +#define MP0_BASE__INST4_SEG5 0 + +#define MP0_BASE__INST5_SEG0 0 +#define MP0_BASE__INST5_SEG1 0 +#define MP0_BASE__INST5_SEG2 0 +#define MP0_BASE__INST5_SEG3 0 +#define MP0_BASE__INST5_SEG4 0 +#define MP0_BASE__INST5_SEG5 0 + +#define MP1_BASE__INST0_SEG0 0x00016000 +#define MP1_BASE__INST0_SEG1 0 +#define MP1_BASE__INST0_SEG2 0 +#define MP1_BASE__INST0_SEG3 0 +#define MP1_BASE__INST0_SEG4 0 +#define MP1_BASE__INST0_SEG5 0 + +#define MP1_BASE__INST1_SEG0 0 +#define MP1_BASE__INST1_SEG1 0 +#define MP1_BASE__INST1_SEG2 0 +#define MP1_BASE__INST1_SEG3 0 +#define MP1_BASE__INST1_SEG4 0 +#define MP1_BASE__INST1_SEG5 0 + +#define MP1_BASE__INST2_SEG0 0 +#define MP1_BASE__INST2_SEG1 0 +#define MP1_BASE__INST2_SEG2 0 +#define MP1_BASE__INST2_SEG3 0 +#define MP1_BASE__INST2_SEG4 0 +#define MP1_BASE__INST2_SEG5 0 + +#define MP1_BASE__INST3_SEG0 0 +#define MP1_BASE__INST3_SEG1 0 +#define MP1_BASE__INST3_SEG2 0 +#define MP1_BASE__INST3_SEG3 0 +#define MP1_BASE__INST3_SEG4 0 +#define MP1_BASE__INST3_SEG5 0 + +#define MP1_BASE__INST4_SEG0 0 +#define MP1_BASE__INST4_SEG1 0 +#define MP1_BASE__INST4_SEG2 0 +#define MP1_BASE__INST4_SEG3 0 +#define MP1_BASE__INST4_SEG4 0 +#define MP1_BASE__INST4_SEG5 0 + +#define MP1_BASE__INST5_SEG0 0 +#define MP1_BASE__INST5_SEG1 0 +#define MP1_BASE__INST5_SEG2 0 +#define MP1_BASE__INST5_SEG3 0 +#define MP1_BASE__INST5_SEG4 0 +#define MP1_BASE__INST5_SEG5 0 + +#define NBIO_BASE__INST0_SEG0 0x00000000 +#define NBIO_BASE__INST0_SEG1 0x00000014 +#define NBIO_BASE__INST0_SEG2 0x00000D20 +#define NBIO_BASE__INST0_SEG3 0x00010400 +#define NBIO_BASE__INST0_SEG4 0 +#define NBIO_BASE__INST0_SEG5 0 + +#define NBIO_BASE__INST1_SEG0 0 +#define NBIO_BASE__INST1_SEG1 0 +#define NBIO_BASE__INST1_SEG2 0 +#define NBIO_BASE__INST1_SEG3 0 +#define NBIO_BASE__INST1_SEG4 0 +#define NBIO_BASE__INST1_SEG5 0 + +#define NBIO_BASE__INST2_SEG0 0 +#define NBIO_BASE__INST2_SEG1 0 +#define NBIO_BASE__INST2_SEG2 0 +#define NBIO_BASE__INST2_SEG3 0 +#define NBIO_BASE__INST2_SEG4 0 +#define NBIO_BASE__INST2_SEG5 0 + +#define NBIO_BASE__INST3_SEG0 0 +#define NBIO_BASE__INST3_SEG1 0 +#define NBIO_BASE__INST3_SEG2 0 +#define NBIO_BASE__INST3_SEG3 0 +#define NBIO_BASE__INST3_SEG4 0 +#define NBIO_BASE__INST3_SEG5 0 + +#define NBIO_BASE__INST4_SEG0 0 +#define NBIO_BASE__INST4_SEG1 0 +#define NBIO_BASE__INST4_SEG2 0 +#define NBIO_BASE__INST4_SEG3 0 +#define NBIO_BASE__INST4_SEG4 0 +#define NBIO_BASE__INST4_SEG5 0 + +#define NBIO_BASE__INST5_SEG0 0 +#define NBIO_BASE__INST5_SEG1 0 +#define NBIO_BASE__INST5_SEG2 0 +#define NBIO_BASE__INST5_SEG3 0 +#define NBIO_BASE__INST5_SEG4 0 +#define NBIO_BASE__INST5_SEG5 0 + +#define OSSSYS_BASE__INST0_SEG0 0x000010A0 +#define OSSSYS_BASE__INST0_SEG1 0 +#define OSSSYS_BASE__INST0_SEG2 0 +#define OSSSYS_BASE__INST0_SEG3 0 +#define OSSSYS_BASE__INST0_SEG4 0 +#define OSSSYS_BASE__INST0_SEG5 0 + +#define OSSSYS_BASE__INST1_SEG0 0 +#define OSSSYS_BASE__INST1_SEG1 0 +#define OSSSYS_BASE__INST1_SEG2 0 +#define OSSSYS_BASE__INST1_SEG3 0 +#define OSSSYS_BASE__INST1_SEG4 0 +#define OSSSYS_BASE__INST1_SEG5 0 + +#define OSSSYS_BASE__INST2_SEG0 0 +#define OSSSYS_BASE__INST2_SEG1 0 +#define OSSSYS_BASE__INST2_SEG2 0 +#define OSSSYS_BASE__INST2_SEG3 0 +#define OSSSYS_BASE__INST2_SEG4 0 +#define OSSSYS_BASE__INST2_SEG5 0 + +#define OSSSYS_BASE__INST3_SEG0 0 +#define OSSSYS_BASE__INST3_SEG1 0 +#define OSSSYS_BASE__INST3_SEG2 0 +#define OSSSYS_BASE__INST3_SEG3 0 +#define OSSSYS_BASE__INST3_SEG4 0 +#define OSSSYS_BASE__INST3_SEG5 0 + +#define OSSSYS_BASE__INST4_SEG0 0 +#define OSSSYS_BASE__INST4_SEG1 0 +#define OSSSYS_BASE__INST4_SEG2 0 +#define OSSSYS_BASE__INST4_SEG3 0 +#define OSSSYS_BASE__INST4_SEG4 0 +#define OSSSYS_BASE__INST4_SEG5 0 + +#define OSSSYS_BASE__INST5_SEG0 0 +#define OSSSYS_BASE__INST5_SEG1 0 +#define OSSSYS_BASE__INST5_SEG2 0 +#define OSSSYS_BASE__INST5_SEG3 0 +#define OSSSYS_BASE__INST5_SEG4 0 +#define OSSSYS_BASE__INST5_SEG5 0 + +#define SDMA0_BASE__INST0_SEG0 0x00001260 +#define SDMA0_BASE__INST0_SEG1 0 +#define SDMA0_BASE__INST0_SEG2 0 +#define SDMA0_BASE__INST0_SEG3 0 +#define SDMA0_BASE__INST0_SEG4 0 +#define SDMA0_BASE__INST0_SEG5 0 + +#define SDMA0_BASE__INST1_SEG0 0 +#define SDMA0_BASE__INST1_SEG1 0 +#define SDMA0_BASE__INST1_SEG2 0 +#define SDMA0_BASE__INST1_SEG3 0 +#define SDMA0_BASE__INST1_SEG4 0 +#define SDMA0_BASE__INST1_SEG5 0 + +#define SDMA0_BASE__INST2_SEG0 0 +#define SDMA0_BASE__INST2_SEG1 0 +#define SDMA0_BASE__INST2_SEG2 0 +#define SDMA0_BASE__INST2_SEG3 0 +#define SDMA0_BASE__INST2_SEG4 0 +#define SDMA0_BASE__INST2_SEG5 0 + +#define SDMA0_BASE__INST3_SEG0 0 +#define SDMA0_BASE__INST3_SEG1 0 +#define SDMA0_BASE__INST3_SEG2 0 +#define SDMA0_BASE__INST3_SEG3 0 +#define SDMA0_BASE__INST3_SEG4 0 +#define SDMA0_BASE__INST3_SEG5 0 + +#define SDMA0_BASE__INST4_SEG0 0 +#define SDMA0_BASE__INST4_SEG1 0 +#define SDMA0_BASE__INST4_SEG2 0 +#define SDMA0_BASE__INST4_SEG3 0 +#define SDMA0_BASE__INST4_SEG4 0 +#define SDMA0_BASE__INST4_SEG5 0 + +#define SDMA0_BASE__INST5_SEG0 0 +#define SDMA0_BASE__INST5_SEG1 0 +#define SDMA0_BASE__INST5_SEG2 0 +#define SDMA0_BASE__INST5_SEG3 0 +#define SDMA0_BASE__INST5_SEG4 0 +#define SDMA0_BASE__INST5_SEG5 0 + +#define SDMA1_BASE__INST0_SEG0 0x00001860 +#define SDMA1_BASE__INST0_SEG1 0 +#define SDMA1_BASE__INST0_SEG2 0 +#define SDMA1_BASE__INST0_SEG3 0 +#define SDMA1_BASE__INST0_SEG4 0 +#define SDMA1_BASE__INST0_SEG5 0 + +#define SDMA1_BASE__INST1_SEG0 0 +#define SDMA1_BASE__INST1_SEG1 0 +#define SDMA1_BASE__INST1_SEG2 0 +#define SDMA1_BASE__INST1_SEG3 0 +#define SDMA1_BASE__INST1_SEG4 0 +#define SDMA1_BASE__INST1_SEG5 0 + +#define SDMA1_BASE__INST2_SEG0 0 +#define SDMA1_BASE__INST2_SEG1 0 +#define SDMA1_BASE__INST2_SEG2 0 +#define SDMA1_BASE__INST2_SEG3 0 +#define SDMA1_BASE__INST2_SEG4 0 +#define SDMA1_BASE__INST2_SEG5 0 + +#define SDMA1_BASE__INST3_SEG0 0 +#define SDMA1_BASE__INST3_SEG1 0 +#define SDMA1_BASE__INST3_SEG2 0 +#define SDMA1_BASE__INST3_SEG3 0 +#define SDMA1_BASE__INST3_SEG4 0 +#define SDMA1_BASE__INST3_SEG5 0 + +#define SDMA1_BASE__INST4_SEG0 0 +#define SDMA1_BASE__INST4_SEG1 0 +#define SDMA1_BASE__INST4_SEG2 0 +#define SDMA1_BASE__INST4_SEG3 0 +#define SDMA1_BASE__INST4_SEG4 0 +#define SDMA1_BASE__INST4_SEG5 0 + +#define SDMA1_BASE__INST5_SEG0 0 +#define SDMA1_BASE__INST5_SEG1 0 +#define SDMA1_BASE__INST5_SEG2 0 +#define SDMA1_BASE__INST5_SEG3 0 +#define SDMA1_BASE__INST5_SEG4 0 +#define SDMA1_BASE__INST5_SEG5 0 + +#define SMUIO_BASE__INST0_SEG0 0x00016800 +#define SMUIO_BASE__INST0_SEG1 0x00016A00 +#define SMUIO_BASE__INST0_SEG2 0 +#define SMUIO_BASE__INST0_SEG3 0 +#define SMUIO_BASE__INST0_SEG4 0 +#define SMUIO_BASE__INST0_SEG5 0 + +#define SMUIO_BASE__INST1_SEG0 0 +#define SMUIO_BASE__INST1_SEG1 0 +#define SMUIO_BASE__INST1_SEG2 0 +#define SMUIO_BASE__INST1_SEG3 0 +#define SMUIO_BASE__INST1_SEG4 0 +#define SMUIO_BASE__INST1_SEG5 0 + +#define SMUIO_BASE__INST2_SEG0 0 +#define SMUIO_BASE__INST2_SEG1 0 +#define SMUIO_BASE__INST2_SEG2 0 +#define SMUIO_BASE__INST2_SEG3 0 +#define SMUIO_BASE__INST2_SEG4 0 +#define SMUIO_BASE__INST2_SEG5 0 + +#define SMUIO_BASE__INST3_SEG0 0 +#define SMUIO_BASE__INST3_SEG1 0 +#define SMUIO_BASE__INST3_SEG2 0 +#define SMUIO_BASE__INST3_SEG3 0 +#define SMUIO_BASE__INST3_SEG4 0 +#define SMUIO_BASE__INST3_SEG5 0 + +#define SMUIO_BASE__INST4_SEG0 0 +#define SMUIO_BASE__INST4_SEG1 0 +#define SMUIO_BASE__INST4_SEG2 0 +#define SMUIO_BASE__INST4_SEG3 0 +#define SMUIO_BASE__INST4_SEG4 0 +#define SMUIO_BASE__INST4_SEG5 0 + +#define SMUIO_BASE__INST5_SEG0 0 +#define SMUIO_BASE__INST5_SEG1 0 +#define SMUIO_BASE__INST5_SEG2 0 +#define SMUIO_BASE__INST5_SEG3 0 +#define SMUIO_BASE__INST5_SEG4 0 +#define SMUIO_BASE__INST5_SEG5 0 + +#define THM_BASE__INST0_SEG0 0x00016600 +#define THM_BASE__INST0_SEG1 0 +#define THM_BASE__INST0_SEG2 0 +#define THM_BASE__INST0_SEG3 0 +#define THM_BASE__INST0_SEG4 0 +#define THM_BASE__INST0_SEG5 0 + +#define THM_BASE__INST1_SEG0 0 +#define THM_BASE__INST1_SEG1 0 +#define THM_BASE__INST1_SEG2 0 +#define THM_BASE__INST1_SEG3 0 +#define THM_BASE__INST1_SEG4 0 +#define THM_BASE__INST1_SEG5 0 + +#define THM_BASE__INST2_SEG0 0 +#define THM_BASE__INST2_SEG1 0 +#define THM_BASE__INST2_SEG2 0 +#define THM_BASE__INST2_SEG3 0 +#define THM_BASE__INST2_SEG4 0 +#define THM_BASE__INST2_SEG5 0 + +#define THM_BASE__INST3_SEG0 0 +#define THM_BASE__INST3_SEG1 0 +#define THM_BASE__INST3_SEG2 0 +#define THM_BASE__INST3_SEG3 0 +#define THM_BASE__INST3_SEG4 0 +#define THM_BASE__INST3_SEG5 0 + +#define THM_BASE__INST4_SEG0 0 +#define THM_BASE__INST4_SEG1 0 +#define THM_BASE__INST4_SEG2 0 +#define THM_BASE__INST4_SEG3 0 +#define THM_BASE__INST4_SEG4 0 +#define THM_BASE__INST4_SEG5 0 + +#define THM_BASE__INST5_SEG0 0 +#define THM_BASE__INST5_SEG1 0 +#define THM_BASE__INST5_SEG2 0 +#define THM_BASE__INST5_SEG3 0 +#define THM_BASE__INST5_SEG4 0 +#define THM_BASE__INST5_SEG5 0 + +#define UMC_BASE__INST0_SEG0 0x00014000 +#define UMC_BASE__INST0_SEG1 0 +#define UMC_BASE__INST0_SEG2 0 +#define UMC_BASE__INST0_SEG3 0 +#define UMC_BASE__INST0_SEG4 0 +#define UMC_BASE__INST0_SEG5 0 + +#define UMC_BASE__INST1_SEG0 0 +#define UMC_BASE__INST1_SEG1 0 +#define UMC_BASE__INST1_SEG2 0 +#define UMC_BASE__INST1_SEG3 0 +#define UMC_BASE__INST1_SEG4 0 +#define UMC_BASE__INST1_SEG5 0 + +#define UMC_BASE__INST2_SEG0 0 +#define UMC_BASE__INST2_SEG1 0 +#define UMC_BASE__INST2_SEG2 0 +#define UMC_BASE__INST2_SEG3 0 +#define UMC_BASE__INST2_SEG4 0 +#define UMC_BASE__INST2_SEG5 0 + +#define UMC_BASE__INST3_SEG0 0 +#define UMC_BASE__INST3_SEG1 0 +#define UMC_BASE__INST3_SEG2 0 +#define UMC_BASE__INST3_SEG3 0 +#define UMC_BASE__INST3_SEG4 0 +#define UMC_BASE__INST3_SEG5 0 + +#define UMC_BASE__INST4_SEG0 0 +#define UMC_BASE__INST4_SEG1 0 +#define UMC_BASE__INST4_SEG2 0 +#define UMC_BASE__INST4_SEG3 0 +#define UMC_BASE__INST4_SEG4 0 +#define UMC_BASE__INST4_SEG5 0 + +#define UMC_BASE__INST5_SEG0 0 +#define UMC_BASE__INST5_SEG1 0 +#define UMC_BASE__INST5_SEG2 0 +#define UMC_BASE__INST5_SEG3 0 +#define UMC_BASE__INST5_SEG4 0 +#define UMC_BASE__INST5_SEG5 0 + +#define UVD_BASE__INST0_SEG0 0x00007800 +#define UVD_BASE__INST0_SEG1 0x00007E00 +#define UVD_BASE__INST0_SEG2 0 +#define UVD_BASE__INST0_SEG3 0 +#define UVD_BASE__INST0_SEG4 0 +#define UVD_BASE__INST0_SEG5 0 + +#define UVD_BASE__INST1_SEG0 0 +#define UVD_BASE__INST1_SEG1 0x00009000 +#define UVD_BASE__INST1_SEG2 0 +#define UVD_BASE__INST1_SEG3 0 +#define UVD_BASE__INST1_SEG4 0 +#define UVD_BASE__INST1_SEG5 0 + +#define UVD_BASE__INST2_SEG0 0 +#define UVD_BASE__INST2_SEG1 0 +#define UVD_BASE__INST2_SEG2 0 +#define UVD_BASE__INST2_SEG3 0 +#define UVD_BASE__INST2_SEG4 0 +#define UVD_BASE__INST2_SEG5 0 + +#define UVD_BASE__INST3_SEG0 0 +#define UVD_BASE__INST3_SEG1 0 +#define UVD_BASE__INST3_SEG2 0 +#define UVD_BASE__INST3_SEG3 0 +#define UVD_BASE__INST3_SEG4 0 +#define UVD_BASE__INST3_SEG5 0 + +#define UVD_BASE__INST4_SEG0 0 +#define UVD_BASE__INST4_SEG1 0 +#define UVD_BASE__INST4_SEG2 0 +#define UVD_BASE__INST4_SEG3 0 +#define UVD_BASE__INST4_SEG4 0 +#define UVD_BASE__INST4_SEG5 0 + +#define UVD_BASE__INST5_SEG0 0 +#define UVD_BASE__INST5_SEG1 0 +#define UVD_BASE__INST5_SEG2 0 +#define UVD_BASE__INST5_SEG3 0 +#define UVD_BASE__INST5_SEG4 0 +#define UVD_BASE__INST5_SEG5 0 + +#define VCE_BASE__INST0_SEG0 0x00008800 +#define VCE_BASE__INST0_SEG1 0 +#define VCE_BASE__INST0_SEG2 0 +#define VCE_BASE__INST0_SEG3 0 +#define VCE_BASE__INST0_SEG4 0 +#define VCE_BASE__INST0_SEG5 0 + +#define VCE_BASE__INST1_SEG0 0 +#define VCE_BASE__INST1_SEG1 0 +#define VCE_BASE__INST1_SEG2 0 +#define VCE_BASE__INST1_SEG3 0 +#define VCE_BASE__INST1_SEG4 0 +#define VCE_BASE__INST1_SEG5 0 + +#define VCE_BASE__INST2_SEG0 0 +#define VCE_BASE__INST2_SEG1 0 +#define VCE_BASE__INST2_SEG2 0 +#define VCE_BASE__INST2_SEG3 0 +#define VCE_BASE__INST2_SEG4 0 +#define VCE_BASE__INST2_SEG5 0 + +#define VCE_BASE__INST3_SEG0 0 +#define VCE_BASE__INST3_SEG1 0 +#define VCE_BASE__INST3_SEG2 0 +#define VCE_BASE__INST3_SEG3 0 +#define VCE_BASE__INST3_SEG4 0 +#define VCE_BASE__INST3_SEG5 0 + +#define VCE_BASE__INST4_SEG0 0 +#define VCE_BASE__INST4_SEG1 0 +#define VCE_BASE__INST4_SEG2 0 +#define VCE_BASE__INST4_SEG3 0 +#define VCE_BASE__INST4_SEG4 0 +#define VCE_BASE__INST4_SEG5 0 + +#define VCE_BASE__INST5_SEG0 0 +#define VCE_BASE__INST5_SEG1 0 +#define VCE_BASE__INST5_SEG2 0 +#define VCE_BASE__INST5_SEG3 0 +#define VCE_BASE__INST5_SEG4 0 +#define VCE_BASE__INST5_SEG5 0 + +#define XDMA_BASE__INST0_SEG0 0x00003400 +#define XDMA_BASE__INST0_SEG1 0 +#define XDMA_BASE__INST0_SEG2 0 +#define XDMA_BASE__INST0_SEG3 0 +#define XDMA_BASE__INST0_SEG4 0 +#define XDMA_BASE__INST0_SEG5 0 + +#define XDMA_BASE__INST1_SEG0 0 +#define XDMA_BASE__INST1_SEG1 0 +#define XDMA_BASE__INST1_SEG2 0 +#define XDMA_BASE__INST1_SEG3 0 +#define XDMA_BASE__INST1_SEG4 0 +#define XDMA_BASE__INST1_SEG5 0 + +#define XDMA_BASE__INST2_SEG0 0 +#define XDMA_BASE__INST2_SEG1 0 +#define XDMA_BASE__INST2_SEG2 0 +#define XDMA_BASE__INST2_SEG3 0 +#define XDMA_BASE__INST2_SEG4 0 +#define XDMA_BASE__INST2_SEG5 0 + +#define XDMA_BASE__INST3_SEG0 0 +#define XDMA_BASE__INST3_SEG1 0 +#define XDMA_BASE__INST3_SEG2 0 +#define XDMA_BASE__INST3_SEG3 0 +#define XDMA_BASE__INST3_SEG4 0 +#define XDMA_BASE__INST3_SEG5 0 + +#define XDMA_BASE__INST4_SEG0 0 +#define XDMA_BASE__INST4_SEG1 0 +#define XDMA_BASE__INST4_SEG2 0 +#define XDMA_BASE__INST4_SEG3 0 +#define XDMA_BASE__INST4_SEG4 0 +#define XDMA_BASE__INST4_SEG5 0 + +#define XDMA_BASE__INST5_SEG0 0 +#define XDMA_BASE__INST5_SEG1 0 +#define XDMA_BASE__INST5_SEG2 0 +#define XDMA_BASE__INST5_SEG3 0 +#define XDMA_BASE__INST5_SEG4 0 +#define XDMA_BASE__INST5_SEG5 0 + +#define RSMU_BASE__INST0_SEG0 0x00012000 +#define RSMU_BASE__INST0_SEG1 0 +#define RSMU_BASE__INST0_SEG2 0 +#define RSMU_BASE__INST0_SEG3 0 +#define RSMU_BASE__INST0_SEG4 0 +#define RSMU_BASE__INST0_SEG5 0 + +#define RSMU_BASE__INST1_SEG0 0 +#define RSMU_BASE__INST1_SEG1 0 +#define RSMU_BASE__INST1_SEG2 0 +#define RSMU_BASE__INST1_SEG3 0 +#define RSMU_BASE__INST1_SEG4 0 +#define RSMU_BASE__INST1_SEG5 0 + +#define RSMU_BASE__INST2_SEG0 0 +#define RSMU_BASE__INST2_SEG1 0 +#define RSMU_BASE__INST2_SEG2 0 +#define RSMU_BASE__INST2_SEG3 0 +#define RSMU_BASE__INST2_SEG4 0 +#define RSMU_BASE__INST2_SEG5 0 + +#define RSMU_BASE__INST3_SEG0 0 +#define RSMU_BASE__INST3_SEG1 0 +#define RSMU_BASE__INST3_SEG2 0 +#define RSMU_BASE__INST3_SEG3 0 +#define RSMU_BASE__INST3_SEG4 0 +#define RSMU_BASE__INST3_SEG5 0 + +#define RSMU_BASE__INST4_SEG0 0 +#define RSMU_BASE__INST4_SEG1 0 +#define RSMU_BASE__INST4_SEG2 0 +#define RSMU_BASE__INST4_SEG3 0 +#define RSMU_BASE__INST4_SEG4 0 +#define RSMU_BASE__INST4_SEG5 0 + +#define RSMU_BASE__INST5_SEG0 0 +#define RSMU_BASE__INST5_SEG1 0 +#define RSMU_BASE__INST5_SEG2 0 +#define RSMU_BASE__INST5_SEG3 0 +#define RSMU_BASE__INST5_SEG4 0 +#define RSMU_BASE__INST5_SEG5 0 + +#endif + diff --git a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c index 7e8ad30d98e2a..b493369e6d0f9 100644 --- a/drivers/gpu/drm/amd/powerplay/amd_powerplay.c +++ b/drivers/gpu/drm/amd/powerplay/amd_powerplay.c @@ -25,30 +25,16 @@ #include <linux/kernel.h> #include <linux/gfp.h> #include <linux/slab.h> +#include <linux/firmware.h> #include "amd_shared.h" #include "amd_powerplay.h" #include "power_state.h" #include "amdgpu.h" #include "hwmgr.h" -#define PP_DPM_DISABLED 0xCCCC - -static int pp_dpm_dispatch_tasks(void *handle, enum amd_pp_task task_id, - enum amd_pm_state_type *user_state); static const struct amd_pm_funcs pp_dpm_funcs; -static inline int pp_check(struct pp_hwmgr *hwmgr) -{ - if (hwmgr == NULL || hwmgr->smumgr_funcs == NULL) - return -EINVAL; - - if (hwmgr->pm_en == 0 || hwmgr->hwmgr_func == NULL) - return PP_DPM_DISABLED; - - return 0; -} - static int amd_powerplay_create(struct amdgpu_device *adev) { struct pp_hwmgr *hwmgr; @@ -61,19 +47,21 @@ static int amd_powerplay_create(struct amdgpu_device *adev) return -ENOMEM; hwmgr->adev = adev; - hwmgr->pm_en = (amdgpu_dpm != 0 && !amdgpu_sriov_vf(adev)) ? true : false; + hwmgr->not_vf = !amdgpu_sriov_vf(adev); + hwmgr->pm_en = (amdgpu_dpm && hwmgr->not_vf) ? true : false; hwmgr->device = amdgpu_cgs_create_device(adev); mutex_init(&hwmgr->smu_lock); hwmgr->chip_family = adev->family; hwmgr->chip_id = adev->asic_type; - hwmgr->feature_mask = amdgpu_pp_feature_mask; + hwmgr->feature_mask = adev->powerplay.pp_feature; + hwmgr->display_config = &adev->pm.pm_display_cfg; adev->powerplay.pp_handle = hwmgr; adev->powerplay.pp_funcs = &pp_dpm_funcs; return 0; } -static int amd_powerplay_destroy(struct amdgpu_device *adev) +static void amd_powerplay_destroy(struct amdgpu_device *adev) { struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; @@ -82,8 +70,6 @@ static int amd_powerplay_destroy(struct amdgpu_device *adev) kfree(hwmgr); hwmgr = NULL; - - return 0; } static int pp_early_init(void *handle) @@ -109,18 +95,9 @@ static int pp_sw_init(void *handle) struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret >= 0) { - if (hwmgr->smumgr_funcs->smu_init == NULL) - return -EINVAL; - - ret = hwmgr->smumgr_funcs->smu_init(hwmgr); + ret = hwmgr_sw_init(hwmgr); - phm_register_irq_handlers(hwmgr); - - pr_debug("amdgpu: powerplay sw initialized\n"); - } + pr_debug("powerplay sw init %s\n", ret ? "failed" : "successfully"); return ret; } @@ -129,16 +106,14 @@ static int pp_sw_fini(void *handle) { struct amdgpu_device *adev = handle; struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; - int ret = 0; - ret = pp_check(hwmgr); - if (ret >= 0) { - if (hwmgr->smumgr_funcs->smu_fini != NULL) - hwmgr->smumgr_funcs->smu_fini(hwmgr); - } + hwmgr_sw_fini(hwmgr); - if (adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) + if (adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) { + release_firmware(adev->pm.fw); + adev->pm.fw = NULL; amdgpu_ucode_fini_bo(adev); + } return 0; } @@ -152,55 +127,76 @@ static int pp_hw_init(void *handle) if (adev->firmware.load_type == AMDGPU_FW_LOAD_SMU) amdgpu_ucode_init_bo(adev); - ret = pp_check(hwmgr); + ret = hwmgr_hw_init(hwmgr); - if (ret >= 0) { - if (hwmgr->smumgr_funcs->start_smu == NULL) - return -EINVAL; + if (ret) + pr_err("powerplay hw init failed\n"); - if (hwmgr->smumgr_funcs->start_smu(hwmgr)) { - pr_err("smc start failed\n"); - hwmgr->smumgr_funcs->smu_fini(hwmgr); - return -EINVAL; - } - if (ret == PP_DPM_DISABLED) - goto exit; - ret = hwmgr_hw_init(hwmgr); - if (ret) - goto exit; - } return ret; -exit: - hwmgr->pm_en = 0; - cgs_notify_dpm_enabled(hwmgr->device, false); - return 0; - } static int pp_hw_fini(void *handle) { struct amdgpu_device *adev = handle; struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; - int ret = 0; - ret = pp_check(hwmgr); - if (ret == 0) - hwmgr_hw_fini(hwmgr); + hwmgr_hw_fini(hwmgr); return 0; } +static void pp_reserve_vram_for_smu(struct amdgpu_device *adev) +{ + int r = -EINVAL; + void *cpu_ptr = NULL; + uint64_t gpu_addr; + struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; + + if (amdgpu_bo_create_kernel(adev, adev->pm.smu_prv_buffer_size, + PAGE_SIZE, AMDGPU_GEM_DOMAIN_GTT, + &adev->pm.smu_prv_buffer, + &gpu_addr, + &cpu_ptr)) { + DRM_ERROR("amdgpu: failed to create smu prv buffer\n"); + return; + } + + if (hwmgr->hwmgr_func->notify_cac_buffer_info) + r = hwmgr->hwmgr_func->notify_cac_buffer_info(hwmgr, + lower_32_bits((unsigned long)cpu_ptr), + upper_32_bits((unsigned long)cpu_ptr), + lower_32_bits(gpu_addr), + upper_32_bits(gpu_addr), + adev->pm.smu_prv_buffer_size); + + if (r) { + amdgpu_bo_free_kernel(&adev->pm.smu_prv_buffer, NULL, NULL); + adev->pm.smu_prv_buffer = NULL; + DRM_ERROR("amdgpu: failed to notify SMU buffer address\n"); + } +} + static int pp_late_init(void *handle) { struct amdgpu_device *adev = handle; struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; - int ret = 0; - - ret = pp_check(hwmgr); + int ret; - if (ret == 0) - pp_dpm_dispatch_tasks(hwmgr, + if (hwmgr && hwmgr->pm_en) { + mutex_lock(&hwmgr->smu_lock); + hwmgr_handle_task(hwmgr, AMD_PP_TASK_COMPLETE_INIT, NULL); + mutex_unlock(&hwmgr->smu_lock); + } + if (adev->pm.smu_prv_buffer_size != 0) + pp_reserve_vram_for_smu(adev); + + if (hwmgr->hwmgr_func->gfx_off_control && + (hwmgr->feature_mask & PP_GFXOFF_MASK)) { + ret = hwmgr->hwmgr_func->gfx_off_control(hwmgr, true); + if (ret) + pr_err("gfx off enabling failed!\n"); + } return 0; } @@ -209,6 +205,8 @@ static void pp_late_fini(void *handle) { struct amdgpu_device *adev = handle; + if (adev->pm.smu_prv_buffer) + amdgpu_bo_free_kernel(&adev->pm.smu_prv_buffer, NULL, NULL); amd_powerplay_destroy(adev); } @@ -233,12 +231,18 @@ static int pp_set_powergating_state(void *handle, { struct amdgpu_device *adev = handle; struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; - int ret = 0; + int ret; - ret = pp_check(hwmgr); + if (!hwmgr || !hwmgr->pm_en) + return 0; - if (ret) - return ret; + if (hwmgr->hwmgr_func->gfx_off_control) { + /* Enable/disable GFX off through SMU */ + ret = hwmgr->hwmgr_func->gfx_off_control(hwmgr, + state == AMD_PG_STATE_GATE); + if (ret) + pr_err("gfx off control failed!\n"); + } if (hwmgr->hwmgr_func->enable_per_cu_power_gating == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -254,38 +258,16 @@ static int pp_suspend(void *handle) { struct amdgpu_device *adev = handle; struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; - int ret = 0; - ret = pp_check(hwmgr); - if (ret == 0) - hwmgr_hw_suspend(hwmgr); - return 0; + return hwmgr_suspend(hwmgr); } static int pp_resume(void *handle) { struct amdgpu_device *adev = handle; struct pp_hwmgr *hwmgr = adev->powerplay.pp_handle; - int ret; - - ret = pp_check(hwmgr); - - if (ret < 0) - return ret; - - if (hwmgr->smumgr_funcs->start_smu == NULL) - return -EINVAL; - - if (hwmgr->smumgr_funcs->start_smu(hwmgr)) { - pr_err("smc start failed\n"); - hwmgr->smumgr_funcs->smu_fini(hwmgr); - return -EINVAL; - } - - if (ret == PP_DPM_DISABLED) - return 0; - return hwmgr_hw_resume(hwmgr); + return hwmgr_resume(hwmgr); } static int pp_set_clockgating_state(void *handle, @@ -334,12 +316,9 @@ static int pp_dpm_fw_loading_complete(void *handle) static int pp_set_clockgating_by_smu(void *handle, uint32_t msg_id) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; - - ret = pp_check(hwmgr); - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->update_clock_gatings == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -362,10 +341,10 @@ static void pp_dpm_en_umd_pstate(struct pp_hwmgr *hwmgr, if (*level & profile_mode_mask) { hwmgr->saved_dpm_level = hwmgr->dpm_level; hwmgr->en_umd_pstate = true; - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_GFX, AMD_CG_STATE_UNGATE); - cgs_set_powergating_state(hwmgr->device, + amdgpu_device_ip_set_powergating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_GFX, AMD_PG_STATE_UNGATE); } @@ -375,10 +354,10 @@ static void pp_dpm_en_umd_pstate(struct pp_hwmgr *hwmgr, if (*level == AMD_DPM_FORCED_LEVEL_PROFILE_EXIT) *level = hwmgr->saved_dpm_level; hwmgr->en_umd_pstate = false; - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_GFX, AMD_CG_STATE_GATE); - cgs_set_powergating_state(hwmgr->device, + amdgpu_device_ip_set_powergating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_GFX, AMD_PG_STATE_GATE); } @@ -389,12 +368,9 @@ static int pp_dpm_force_performance_level(void *handle, enum amd_dpm_forced_level level) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (level == hwmgr->dpm_level) return 0; @@ -412,13 +388,10 @@ static enum amd_dpm_forced_level pp_dpm_get_performance_level( void *handle) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; enum amd_dpm_forced_level level; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; mutex_lock(&hwmgr->smu_lock); level = hwmgr->dpm_level; @@ -429,13 +402,10 @@ static enum amd_dpm_forced_level pp_dpm_get_performance_level( static uint32_t pp_dpm_get_sclk(void *handle, bool low) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; uint32_t clk = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return 0; if (hwmgr->hwmgr_func->get_sclk == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -450,13 +420,10 @@ static uint32_t pp_dpm_get_sclk(void *handle, bool low) static uint32_t pp_dpm_get_mclk(void *handle, bool low) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; uint32_t clk = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return 0; if (hwmgr->hwmgr_func->get_mclk == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -471,11 +438,8 @@ static uint32_t pp_dpm_get_mclk(void *handle, bool low) static void pp_dpm_powergate_vce(void *handle, bool gate) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; - ret = pp_check(hwmgr); - - if (ret) + if (!hwmgr || !hwmgr->pm_en) return; if (hwmgr->hwmgr_func->powergate_vce == NULL) { @@ -490,11 +454,8 @@ static void pp_dpm_powergate_vce(void *handle, bool gate) static void pp_dpm_powergate_uvd(void *handle, bool gate) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; - - ret = pp_check(hwmgr); - if (ret) + if (!hwmgr || !hwmgr->pm_en) return; if (hwmgr->hwmgr_func->powergate_uvd == NULL) { @@ -512,10 +473,8 @@ static int pp_dpm_dispatch_tasks(void *handle, enum amd_pp_task task_id, int ret = 0; struct pp_hwmgr *hwmgr = handle; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; mutex_lock(&hwmgr->smu_lock); ret = hwmgr_handle_task(hwmgr, task_id, user_state); @@ -528,15 +487,9 @@ static enum amd_pm_state_type pp_dpm_get_current_power_state(void *handle) { struct pp_hwmgr *hwmgr = handle; struct pp_power_state *state; - int ret = 0; enum amd_pm_state_type pm_type; - ret = pp_check(hwmgr); - - if (ret) - return ret; - - if (hwmgr->current_ps == NULL) + if (!hwmgr || !hwmgr->pm_en || !hwmgr->current_ps) return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -568,11 +521,8 @@ static enum amd_pm_state_type pp_dpm_get_current_power_state(void *handle) static void pp_dpm_set_fan_control_mode(void *handle, uint32_t mode) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; - - ret = pp_check(hwmgr); - if (ret) + if (!hwmgr || !hwmgr->pm_en) return; if (hwmgr->hwmgr_func->set_fan_control_mode == NULL) { @@ -587,13 +537,10 @@ static void pp_dpm_set_fan_control_mode(void *handle, uint32_t mode) static uint32_t pp_dpm_get_fan_control_mode(void *handle) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; uint32_t mode = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return 0; if (hwmgr->hwmgr_func->get_fan_control_mode == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -610,10 +557,8 @@ static int pp_dpm_set_fan_speed_percent(void *handle, uint32_t percent) struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->set_fan_speed_percent == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -630,10 +575,8 @@ static int pp_dpm_get_fan_speed_percent(void *handle, uint32_t *speed) struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->get_fan_speed_percent == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -651,10 +594,8 @@ static int pp_dpm_get_fan_speed_rpm(void *handle, uint32_t *rpm) struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->get_fan_speed_rpm == NULL) return -EINVAL; @@ -670,16 +611,10 @@ static int pp_dpm_get_pp_num_states(void *handle, { struct pp_hwmgr *hwmgr = handle; int i; - int ret = 0; memset(data, 0, sizeof(*data)); - ret = pp_check(hwmgr); - - if (ret) - return ret; - - if (hwmgr->ps == NULL) + if (!hwmgr || !hwmgr->pm_en ||!hwmgr->ps) return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -713,15 +648,9 @@ static int pp_dpm_get_pp_num_states(void *handle, static int pp_dpm_get_pp_table(void *handle, char **table) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; int size = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; - - if (!hwmgr->soft_pp_table) + if (!hwmgr || !hwmgr->pm_en ||!hwmgr->soft_pp_table) return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -736,10 +665,6 @@ static int amd_powerplay_reset(void *handle) struct pp_hwmgr *hwmgr = handle; int ret; - ret = pp_check(hwmgr); - if (ret) - return ret; - ret = hwmgr_hw_fini(hwmgr); if (ret) return ret; @@ -754,40 +679,38 @@ static int amd_powerplay_reset(void *handle) static int pp_dpm_set_pp_table(void *handle, const char *buf, size_t size) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; + int ret = -ENOMEM; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; mutex_lock(&hwmgr->smu_lock); if (!hwmgr->hardcode_pp_table) { hwmgr->hardcode_pp_table = kmemdup(hwmgr->soft_pp_table, hwmgr->soft_pp_table_size, GFP_KERNEL); - if (!hwmgr->hardcode_pp_table) { - mutex_unlock(&hwmgr->smu_lock); - return -ENOMEM; - } + if (!hwmgr->hardcode_pp_table) + goto err; } memcpy(hwmgr->hardcode_pp_table, buf, size); hwmgr->soft_pp_table = hwmgr->hardcode_pp_table; - mutex_unlock(&hwmgr->smu_lock); ret = amd_powerplay_reset(handle); if (ret) - return ret; + goto err; if (hwmgr->hwmgr_func->avfs_control) { ret = hwmgr->hwmgr_func->avfs_control(hwmgr, false); if (ret) - return ret; + goto err; } - + mutex_unlock(&hwmgr->smu_lock); return 0; +err: + mutex_unlock(&hwmgr->smu_lock); + return ret; } static int pp_dpm_force_clock_level(void *handle, @@ -796,10 +719,8 @@ static int pp_dpm_force_clock_level(void *handle, struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->force_clock_level == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -820,10 +741,8 @@ static int pp_dpm_print_clock_levels(void *handle, struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->print_clock_levels == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -840,10 +759,8 @@ static int pp_dpm_get_sclk_od(void *handle) struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->get_sclk_od == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -860,10 +777,8 @@ static int pp_dpm_set_sclk_od(void *handle, uint32_t value) struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->set_sclk_od == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -881,10 +796,8 @@ static int pp_dpm_get_mclk_od(void *handle) struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->get_mclk_od == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -901,10 +814,8 @@ static int pp_dpm_set_mclk_od(void *handle, uint32_t value) struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->set_mclk_od == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -922,11 +833,7 @@ static int pp_dpm_read_sensor(void *handle, int idx, struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - if (ret) - return ret; - - if (value == NULL) + if (!hwmgr || !hwmgr->pm_en || !value) return -EINVAL; switch (idx) { @@ -948,14 +855,11 @@ static struct amd_vce_state* pp_dpm_get_vce_clock_state(void *handle, unsigned idx) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; - - ret = pp_check(hwmgr); - if (ret) + if (!hwmgr || !hwmgr->pm_en) return NULL; - if (hwmgr && idx < hwmgr->num_vce_state_tables) + if (idx < hwmgr->num_vce_state_tables) return &hwmgr->vce_states[idx]; return NULL; } @@ -964,7 +868,7 @@ static int pp_get_power_profile_mode(void *handle, char *buf) { struct pp_hwmgr *hwmgr = handle; - if (!buf || pp_check(hwmgr)) + if (!hwmgr || !hwmgr->pm_en || !buf) return -EINVAL; if (hwmgr->hwmgr_func->get_power_profile_mode == NULL) { @@ -980,12 +884,12 @@ static int pp_set_power_profile_mode(void *handle, long *input, uint32_t size) struct pp_hwmgr *hwmgr = handle; int ret = -EINVAL; - if (pp_check(hwmgr)) - return -EINVAL; + if (!hwmgr || !hwmgr->pm_en) + return ret; if (hwmgr->hwmgr_func->set_power_profile_mode == NULL) { pr_info("%s was not implemented.\n", __func__); - return -EINVAL; + return ret; } mutex_lock(&hwmgr->smu_lock); if (hwmgr->dpm_level == AMD_DPM_FORCED_LEVEL_MANUAL) @@ -998,7 +902,7 @@ static int pp_odn_edit_dpm_table(void *handle, uint32_t type, long *input, uint3 { struct pp_hwmgr *hwmgr = handle; - if (pp_check(hwmgr)) + if (!hwmgr || !hwmgr->pm_en) return -EINVAL; if (hwmgr->hwmgr_func->odn_edit_dpm_table == NULL) { @@ -1016,7 +920,7 @@ static int pp_dpm_switch_power_profile(void *handle, long workload; uint32_t index; - if (pp_check(hwmgr)) + if (!hwmgr || !hwmgr->pm_en) return -EINVAL; if (hwmgr->hwmgr_func->set_power_profile_mode == NULL) { @@ -1048,46 +952,12 @@ static int pp_dpm_switch_power_profile(void *handle, return 0; } -static int pp_dpm_notify_smu_memory_info(void *handle, - uint32_t virtual_addr_low, - uint32_t virtual_addr_hi, - uint32_t mc_addr_low, - uint32_t mc_addr_hi, - uint32_t size) -{ - struct pp_hwmgr *hwmgr = handle; - int ret = 0; - - ret = pp_check(hwmgr); - - if (ret) - return ret; - - if (hwmgr->hwmgr_func->notify_cac_buffer_info == NULL) { - pr_info("%s was not implemented.\n", __func__); - return -EINVAL; - } - - mutex_lock(&hwmgr->smu_lock); - - ret = hwmgr->hwmgr_func->notify_cac_buffer_info(hwmgr, virtual_addr_low, - virtual_addr_hi, mc_addr_low, mc_addr_hi, - size); - - mutex_unlock(&hwmgr->smu_lock); - - return ret; -} - static int pp_set_power_limit(void *handle, uint32_t limit) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->set_power_limit == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -1104,20 +974,14 @@ static int pp_set_power_limit(void *handle, uint32_t limit) hwmgr->hwmgr_func->set_power_limit(hwmgr, limit); hwmgr->power_limit = limit; mutex_unlock(&hwmgr->smu_lock); - return ret; + return 0; } static int pp_get_power_limit(void *handle, uint32_t *limit, bool default_limit) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; - - ret = pp_check(hwmgr); - if (ret) - return ret; - - if (limit == NULL) + if (!hwmgr || !hwmgr->pm_en ||!limit) return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -1129,19 +993,16 @@ static int pp_get_power_limit(void *handle, uint32_t *limit, bool default_limit) mutex_unlock(&hwmgr->smu_lock); - return ret; + return 0; } static int pp_display_configuration_change(void *handle, const struct amd_pp_display_configuration *display_config) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; mutex_lock(&hwmgr->smu_lock); phm_store_dal_configuration_data(hwmgr, display_config); @@ -1155,12 +1016,7 @@ static int pp_get_display_power_level(void *handle, struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; - - if (output == NULL) + if (!hwmgr || !hwmgr->pm_en ||!output) return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -1177,10 +1033,8 @@ static int pp_get_current_clocks(void *handle, struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -1225,10 +1079,8 @@ static int pp_get_clock_by_type(void *handle, enum amd_pp_clock_type type, struc struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (clocks == NULL) return -EINVAL; @@ -1246,11 +1098,7 @@ static int pp_get_clock_by_type_with_latency(void *handle, struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - if (ret) - return ret; - - if (!clocks) + if (!hwmgr || !hwmgr->pm_en ||!clocks) return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -1266,11 +1114,7 @@ static int pp_get_clock_by_type_with_voltage(void *handle, struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - if (ret) - return ret; - - if (!clocks) + if (!hwmgr || !hwmgr->pm_en ||!clocks) return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -1287,11 +1131,7 @@ static int pp_set_watermarks_for_clocks_ranges(void *handle, struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - if (ret) - return ret; - - if (!wm_with_clock_ranges) + if (!hwmgr || !hwmgr->pm_en ||!wm_with_clock_ranges) return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -1308,11 +1148,7 @@ static int pp_display_clock_voltage_request(void *handle, struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - if (ret) - return ret; - - if (!clock) + if (!hwmgr || !hwmgr->pm_en ||!clock) return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -1328,12 +1164,7 @@ static int pp_get_display_mode_validation_clocks(void *handle, struct pp_hwmgr *hwmgr = handle; int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; - - if (clocks == NULL) + if (!hwmgr || !hwmgr->pm_en ||!clocks) return -EINVAL; mutex_lock(&hwmgr->smu_lock); @@ -1348,12 +1179,9 @@ static int pp_get_display_mode_validation_clocks(void *handle, static int pp_set_mmhub_powergating_by_smu(void *handle) { struct pp_hwmgr *hwmgr = handle; - int ret = 0; - ret = pp_check(hwmgr); - - if (ret) - return ret; + if (!hwmgr || !hwmgr->pm_en) + return -EINVAL; if (hwmgr->hwmgr_func->set_mmhub_powergating_by_smu == NULL) { pr_info("%s was not implemented.\n", __func__); @@ -1390,7 +1218,6 @@ static const struct amd_pm_funcs pp_dpm_funcs = { .get_vce_clock_state = pp_dpm_get_vce_clock_state, .switch_power_profile = pp_dpm_switch_power_profile, .set_clockgating_by_smu = pp_set_clockgating_by_smu, - .notify_smu_memory_info = pp_dpm_notify_smu_memory_info, .get_power_profile_mode = pp_get_power_profile_mode, .set_power_profile_mode = pp_set_power_profile_mode, .odn_edit_dpm_table = pp_odn_edit_dpm_table, diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c b/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c index ae2e9339dd6b7..a0bb921fac228 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/hardwaremanager.c @@ -75,8 +75,7 @@ int phm_set_power_state(struct pp_hwmgr *hwmgr, int phm_enable_dynamic_state_management(struct pp_hwmgr *hwmgr) { - int ret = 1; - bool enabled; + int ret = -EINVAL;; PHM_FUNC_CHECK(hwmgr); if (smum_is_dpm_running(hwmgr)) { @@ -87,17 +86,12 @@ int phm_enable_dynamic_state_management(struct pp_hwmgr *hwmgr) if (NULL != hwmgr->hwmgr_func->dynamic_state_management_enable) ret = hwmgr->hwmgr_func->dynamic_state_management_enable(hwmgr); - enabled = ret == 0; - - cgs_notify_dpm_enabled(hwmgr->device, enabled); - return ret; } int phm_disable_dynamic_state_management(struct pp_hwmgr *hwmgr) { - int ret = -1; - bool enabled; + int ret = -EINVAL; PHM_FUNC_CHECK(hwmgr); @@ -109,10 +103,6 @@ int phm_disable_dynamic_state_management(struct pp_hwmgr *hwmgr) if (hwmgr->hwmgr_func->dynamic_state_management_disable) ret = hwmgr->hwmgr_func->dynamic_state_management_disable(hwmgr); - enabled = ret == 0 ? false : true; - - cgs_notify_dpm_enabled(hwmgr->device, enabled); - return ret; } @@ -142,6 +132,15 @@ int phm_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, return 0; } +int phm_apply_clock_adjust_rules(struct pp_hwmgr *hwmgr) +{ + PHM_FUNC_CHECK(hwmgr); + + if (hwmgr->hwmgr_func->apply_clocks_adjust_rules != NULL) + return hwmgr->hwmgr_func->apply_clocks_adjust_rules(hwmgr); + return 0; +} + int phm_powerdown_uvd(struct pp_hwmgr *hwmgr) { PHM_FUNC_CHECK(hwmgr); @@ -171,6 +170,16 @@ int phm_disable_clock_power_gatings(struct pp_hwmgr *hwmgr) return 0; } +int phm_pre_display_configuration_changed(struct pp_hwmgr *hwmgr) +{ + PHM_FUNC_CHECK(hwmgr); + + if (NULL != hwmgr->hwmgr_func->pre_display_config_changed) + hwmgr->hwmgr_func->pre_display_config_changed(hwmgr); + + return 0; + +} int phm_display_configuration_changed(struct pp_hwmgr *hwmgr) { @@ -275,13 +284,11 @@ int phm_store_dal_configuration_data(struct pp_hwmgr *hwmgr, if (display_config == NULL) return -EINVAL; - hwmgr->display_config = *display_config; - if (NULL != hwmgr->hwmgr_func->set_deep_sleep_dcefclk) - hwmgr->hwmgr_func->set_deep_sleep_dcefclk(hwmgr, hwmgr->display_config.min_dcef_deep_sleep_set_clk); + hwmgr->hwmgr_func->set_deep_sleep_dcefclk(hwmgr, display_config->min_dcef_deep_sleep_set_clk); - for (index = 0; index < hwmgr->display_config.num_path_including_non_display; index++) { - if (hwmgr->display_config.displays[index].controller_id != 0) + for (index = 0; index < display_config->num_path_including_non_display; index++) { + if (display_config->displays[index].controller_id != 0) number_of_active_display++; } diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c index 42982055b1618..e63bc47dc715f 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/hwmgr.c @@ -40,6 +40,7 @@ extern const struct pp_smumgr_func iceland_smu_funcs; extern const struct pp_smumgr_func tonga_smu_funcs; extern const struct pp_smumgr_func fiji_smu_funcs; extern const struct pp_smumgr_func polaris10_smu_funcs; +extern const struct pp_smumgr_func vegam_smu_funcs; extern const struct pp_smumgr_func vega10_smu_funcs; extern const struct pp_smumgr_func vega12_smu_funcs; extern const struct pp_smumgr_func smu10_smu_funcs; @@ -76,7 +77,7 @@ static void hwmgr_init_workload_prority(struct pp_hwmgr *hwmgr) int hwmgr_early_init(struct pp_hwmgr *hwmgr) { - if (hwmgr == NULL) + if (!hwmgr) return -EINVAL; hwmgr->usec_timeout = AMD_MAX_USEC_TIMEOUT; @@ -95,7 +96,8 @@ int hwmgr_early_init(struct pp_hwmgr *hwmgr) hwmgr->smumgr_funcs = &ci_smu_funcs; ci_set_asic_special_caps(hwmgr); hwmgr->feature_mask &= ~(PP_VBI_TIME_SUPPORT_MASK | - PP_ENABLE_GFX_CG_THRU_SMU); + PP_ENABLE_GFX_CG_THRU_SMU | + PP_GFXOFF_MASK); hwmgr->pp_table_version = PP_TABLE_V0; hwmgr->od_enabled = false; smu7_init_function_pointers(hwmgr); @@ -103,9 +105,11 @@ int hwmgr_early_init(struct pp_hwmgr *hwmgr) case AMDGPU_FAMILY_CZ: hwmgr->od_enabled = false; hwmgr->smumgr_funcs = &smu8_smu_funcs; + hwmgr->feature_mask &= ~PP_GFXOFF_MASK; smu8_init_function_pointers(hwmgr); break; case AMDGPU_FAMILY_VI: + hwmgr->feature_mask &= ~PP_GFXOFF_MASK; switch (hwmgr->chip_id) { case CHIP_TOPAZ: hwmgr->smumgr_funcs = &iceland_smu_funcs; @@ -133,14 +137,21 @@ int hwmgr_early_init(struct pp_hwmgr *hwmgr) polaris_set_asic_special_caps(hwmgr); hwmgr->feature_mask &= ~(PP_UVD_HANDSHAKE_MASK); break; + case CHIP_VEGAM: + hwmgr->smumgr_funcs = &vegam_smu_funcs; + polaris_set_asic_special_caps(hwmgr); + hwmgr->feature_mask &= ~(PP_UVD_HANDSHAKE_MASK); + break; default: return -EINVAL; } smu7_init_function_pointers(hwmgr); break; case AMDGPU_FAMILY_AI: + hwmgr->feature_mask &= ~PP_GFXOFF_MASK; switch (hwmgr->chip_id) { case CHIP_VEGA10: + case CHIP_VEGA20: hwmgr->smumgr_funcs = &vega10_smu_funcs; vega10_hwmgr_init(hwmgr); break; @@ -170,22 +181,58 @@ int hwmgr_early_init(struct pp_hwmgr *hwmgr) return 0; } +int hwmgr_sw_init(struct pp_hwmgr *hwmgr) +{ + if (!hwmgr|| !hwmgr->smumgr_funcs || !hwmgr->smumgr_funcs->smu_init) + return -EINVAL; + + phm_register_irq_handlers(hwmgr); + + return hwmgr->smumgr_funcs->smu_init(hwmgr); +} + + +int hwmgr_sw_fini(struct pp_hwmgr *hwmgr) +{ + if (hwmgr && hwmgr->smumgr_funcs && hwmgr->smumgr_funcs->smu_fini) + hwmgr->smumgr_funcs->smu_fini(hwmgr); + + return 0; +} + int hwmgr_hw_init(struct pp_hwmgr *hwmgr) { int ret = 0; - if (hwmgr == NULL) + if (!hwmgr || !hwmgr->smumgr_funcs) return -EINVAL; - if (hwmgr->pptable_func == NULL || - hwmgr->pptable_func->pptable_init == NULL || - hwmgr->hwmgr_func->backend_init == NULL) - return -EINVAL; + if (hwmgr->smumgr_funcs->start_smu) { + ret = hwmgr->smumgr_funcs->start_smu(hwmgr); + if (ret) { + pr_err("smc start failed\n"); + return -EINVAL; + } + } + + if (!hwmgr->pm_en) + return 0; + + if (!hwmgr->pptable_func || + !hwmgr->pptable_func->pptable_init || + !hwmgr->hwmgr_func->backend_init) { + hwmgr->pm_en = false; + pr_info("dpm not supported \n"); + return 0; + } ret = hwmgr->pptable_func->pptable_init(hwmgr); if (ret) goto err; + ((struct amdgpu_device *)hwmgr->adev)->pm.no_fan = + hwmgr->thermal_controller.fanInfo.bNoFan; + ret = hwmgr->hwmgr_func->backend_init(hwmgr); if (ret) goto err1; @@ -206,6 +253,8 @@ int hwmgr_hw_init(struct pp_hwmgr *hwmgr) if (ret) goto err2; + ((struct amdgpu_device *)hwmgr->adev)->pm.dpm_enabled = true; + return 0; err2: if (hwmgr->hwmgr_func->backend_fini) @@ -214,14 +263,13 @@ err1: if (hwmgr->pptable_func->pptable_fini) hwmgr->pptable_func->pptable_fini(hwmgr); err: - pr_err("amdgpu: powerplay initialization failed\n"); return ret; } int hwmgr_hw_fini(struct pp_hwmgr *hwmgr) { - if (hwmgr == NULL) - return -EINVAL; + if (!hwmgr || !hwmgr->pm_en) + return 0; phm_stop_thermal_controller(hwmgr); psm_set_boot_states(hwmgr); @@ -236,12 +284,12 @@ int hwmgr_hw_fini(struct pp_hwmgr *hwmgr) return psm_fini_power_state_table(hwmgr); } -int hwmgr_hw_suspend(struct pp_hwmgr *hwmgr) +int hwmgr_suspend(struct pp_hwmgr *hwmgr) { int ret = 0; - if (hwmgr == NULL) - return -EINVAL; + if (!hwmgr || !hwmgr->pm_en) + return 0; phm_disable_smc_firmware_ctf(hwmgr); ret = psm_set_boot_states(hwmgr); @@ -255,13 +303,23 @@ int hwmgr_hw_suspend(struct pp_hwmgr *hwmgr) return ret; } -int hwmgr_hw_resume(struct pp_hwmgr *hwmgr) +int hwmgr_resume(struct pp_hwmgr *hwmgr) { int ret = 0; - if (hwmgr == NULL) + if (!hwmgr) return -EINVAL; + if (hwmgr->smumgr_funcs && hwmgr->smumgr_funcs->start_smu) { + if (hwmgr->smumgr_funcs->start_smu(hwmgr)) { + pr_err("smc start failed\n"); + return -EINVAL; + } + } + + if (!hwmgr->pm_en) + return 0; + ret = phm_setup_asic(hwmgr); if (ret) return ret; @@ -270,9 +328,6 @@ int hwmgr_hw_resume(struct pp_hwmgr *hwmgr) if (ret) return ret; ret = phm_start_thermal_controller(hwmgr); - if (ret) - return ret; - ret |= psm_set_performance_states(hwmgr); if (ret) return ret; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c b/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c index 0f2851b5b368e..0af13c1543287 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/pp_psm.c @@ -46,7 +46,7 @@ int psm_init_power_state_table(struct pp_hwmgr *hwmgr) sizeof(struct pp_power_state); if (table_entries == 0 || size == 0) { - pr_warn("Please check whether power state management is suppported on this asic\n"); + pr_warn("Please check whether power state management is supported on this asic\n"); return 0; } @@ -265,6 +265,15 @@ int psm_adjust_power_state_dynamic(struct pp_hwmgr *hwmgr, bool skip, if (skip) return 0; + if (!hwmgr->ps) + /* + * for vega12/vega20 which does not support power state manager + * DAL clock limits should also be honoured + */ + phm_apply_clock_adjust_rules(hwmgr); + + phm_pre_display_configuration_changed(hwmgr); + phm_display_configuration_changed(hwmgr); if (hwmgr->ps) diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.c b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.c index c6febbf0bf693..7047e29755c35 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.c @@ -23,7 +23,8 @@ #include "pp_debug.h" #include <linux/module.h> #include <linux/slab.h> - +#include <linux/delay.h> +#include "atom.h" #include "ppatomctrl.h" #include "atombios.h" #include "cgs_common.h" @@ -128,7 +129,6 @@ static int atomctrl_set_mc_reg_address_table( return 0; } - int atomctrl_initialize_mc_reg_table( struct pp_hwmgr *hwmgr, uint8_t module_index, @@ -141,7 +141,7 @@ int atomctrl_initialize_mc_reg_table( u16 size; vram_info = (ATOM_VRAM_INFO_HEADER_V2_1 *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, GetIndexIntoMasterTable(DATA, VRAM_Info), &size, &frev, &crev); if (module_index >= vram_info->ucNumOfVRAMModule) { @@ -174,6 +174,8 @@ int atomctrl_set_engine_dram_timings_rv770( uint32_t engine_clock, uint32_t memory_clock) { + struct amdgpu_device *adev = hwmgr->adev; + SET_ENGINE_CLOCK_PS_ALLOCATION engine_clock_parameters; /* They are both in 10KHz Units. */ @@ -184,9 +186,10 @@ int atomctrl_set_engine_dram_timings_rv770( /* in 10 khz units.*/ engine_clock_parameters.sReserved.ulClock = cpu_to_le32(memory_clock & SET_CLOCK_FREQ_MASK); - return cgs_atom_exec_cmd_table(hwmgr->device, + + return amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, DynamicMemorySettings), - &engine_clock_parameters); + (uint32_t *)&engine_clock_parameters); } /** @@ -203,7 +206,7 @@ static ATOM_VOLTAGE_OBJECT_INFO *get_voltage_info_table(void *device) union voltage_object_info *voltage_info; voltage_info = (union voltage_object_info *) - cgs_atom_get_data_table(device, index, + smu_atom_get_data_table(device, index, &size, &frev, &crev); if (voltage_info != NULL) @@ -247,16 +250,16 @@ int atomctrl_get_memory_pll_dividers_si( pp_atomctrl_memory_clock_param *mpll_param, bool strobe_mode) { + struct amdgpu_device *adev = hwmgr->adev; COMPUTE_MEMORY_CLOCK_PARAM_PARAMETERS_V2_1 mpll_parameters; int result; mpll_parameters.ulClock = cpu_to_le32(clock_value); mpll_parameters.ucInputFlag = (uint8_t)((strobe_mode) ? 1 : 0); - result = cgs_atom_exec_cmd_table - (hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ComputeMemoryClockParam), - &mpll_parameters); + (uint32_t *)&mpll_parameters); if (0 == result) { mpll_param->mpll_fb_divider.clk_frac = @@ -295,14 +298,15 @@ int atomctrl_get_memory_pll_dividers_si( int atomctrl_get_memory_pll_dividers_vi(struct pp_hwmgr *hwmgr, uint32_t clock_value, pp_atomctrl_memory_clock_param *mpll_param) { + struct amdgpu_device *adev = hwmgr->adev; COMPUTE_MEMORY_CLOCK_PARAM_PARAMETERS_V2_2 mpll_parameters; int result; mpll_parameters.ulClock.ulClock = cpu_to_le32(clock_value); - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ComputeMemoryClockParam), - &mpll_parameters); + (uint32_t *)&mpll_parameters); if (!result) mpll_param->mpll_post_divider = @@ -311,19 +315,49 @@ int atomctrl_get_memory_pll_dividers_vi(struct pp_hwmgr *hwmgr, return result; } +int atomctrl_get_memory_pll_dividers_ai(struct pp_hwmgr *hwmgr, + uint32_t clock_value, + pp_atomctrl_memory_clock_param_ai *mpll_param) +{ + struct amdgpu_device *adev = hwmgr->adev; + COMPUTE_MEMORY_CLOCK_PARAM_PARAMETERS_V2_3 mpll_parameters = {{0}, 0, 0}; + int result; + + mpll_parameters.ulClock.ulClock = cpu_to_le32(clock_value); + + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, + GetIndexIntoMasterTable(COMMAND, ComputeMemoryClockParam), + (uint32_t *)&mpll_parameters); + + /* VEGAM's mpll takes sometime to finish computing */ + udelay(10); + + if (!result) { + mpll_param->ulMclk_fcw_int = + le16_to_cpu(mpll_parameters.usMclk_fcw_int); + mpll_param->ulMclk_fcw_frac = + le16_to_cpu(mpll_parameters.usMclk_fcw_frac); + mpll_param->ulClock = + le32_to_cpu(mpll_parameters.ulClock.ulClock); + mpll_param->ulPostDiv = mpll_parameters.ulClock.ucPostDiv; + } + + return result; +} + int atomctrl_get_engine_pll_dividers_kong(struct pp_hwmgr *hwmgr, uint32_t clock_value, pp_atomctrl_clock_dividers_kong *dividers) { + struct amdgpu_device *adev = hwmgr->adev; COMPUTE_MEMORY_ENGINE_PLL_PARAMETERS_V4 pll_parameters; int result; pll_parameters.ulClock = cpu_to_le32(clock_value); - result = cgs_atom_exec_cmd_table - (hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ComputeMemoryEnginePLL), - &pll_parameters); + (uint32_t *)&pll_parameters); if (0 == result) { dividers->pll_post_divider = pll_parameters.ucPostDiv; @@ -338,16 +372,16 @@ int atomctrl_get_engine_pll_dividers_vi( uint32_t clock_value, pp_atomctrl_clock_dividers_vi *dividers) { + struct amdgpu_device *adev = hwmgr->adev; COMPUTE_GPU_CLOCK_OUTPUT_PARAMETERS_V1_6 pll_patameters; int result; pll_patameters.ulClock.ulClock = cpu_to_le32(clock_value); pll_patameters.ulClock.ucPostDiv = COMPUTE_GPUCLK_INPUT_FLAG_SCLK; - result = cgs_atom_exec_cmd_table - (hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ComputeMemoryEnginePLL), - &pll_patameters); + (uint32_t *)&pll_patameters); if (0 == result) { dividers->pll_post_divider = @@ -375,16 +409,16 @@ int atomctrl_get_engine_pll_dividers_ai(struct pp_hwmgr *hwmgr, uint32_t clock_value, pp_atomctrl_clock_dividers_ai *dividers) { + struct amdgpu_device *adev = hwmgr->adev; COMPUTE_GPU_CLOCK_OUTPUT_PARAMETERS_V1_7 pll_patameters; int result; pll_patameters.ulClock.ulClock = cpu_to_le32(clock_value); pll_patameters.ulClock.ucPostDiv = COMPUTE_GPUCLK_INPUT_FLAG_SCLK; - result = cgs_atom_exec_cmd_table - (hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ComputeMemoryEnginePLL), - &pll_patameters); + (uint32_t *)&pll_patameters); if (0 == result) { dividers->usSclk_fcw_frac = le16_to_cpu(pll_patameters.usSclk_fcw_frac); @@ -407,6 +441,7 @@ int atomctrl_get_dfs_pll_dividers_vi( uint32_t clock_value, pp_atomctrl_clock_dividers_vi *dividers) { + struct amdgpu_device *adev = hwmgr->adev; COMPUTE_GPU_CLOCK_OUTPUT_PARAMETERS_V1_6 pll_patameters; int result; @@ -414,10 +449,9 @@ int atomctrl_get_dfs_pll_dividers_vi( pll_patameters.ulClock.ucPostDiv = COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK; - result = cgs_atom_exec_cmd_table - (hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ComputeMemoryEnginePLL), - &pll_patameters); + (uint32_t *)&pll_patameters); if (0 == result) { dividers->pll_post_divider = @@ -452,7 +486,7 @@ uint32_t atomctrl_get_reference_clock(struct pp_hwmgr *hwmgr) uint32_t clock; fw_info = (ATOM_FIRMWARE_INFO *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, GetIndexIntoMasterTable(DATA, FirmwareInfo), &size, &frev, &crev); @@ -476,7 +510,7 @@ bool atomctrl_is_voltage_controlled_by_gpio_v3( uint8_t voltage_mode) { ATOM_VOLTAGE_OBJECT_INFO_V3_1 *voltage_info = - (ATOM_VOLTAGE_OBJECT_INFO_V3_1 *)get_voltage_info_table(hwmgr->device); + (ATOM_VOLTAGE_OBJECT_INFO_V3_1 *)get_voltage_info_table(hwmgr->adev); bool ret; PP_ASSERT_WITH_CODE((NULL != voltage_info), @@ -495,7 +529,7 @@ int atomctrl_get_voltage_table_v3( pp_atomctrl_voltage_table *voltage_table) { ATOM_VOLTAGE_OBJECT_INFO_V3_1 *voltage_info = - (ATOM_VOLTAGE_OBJECT_INFO_V3_1 *)get_voltage_info_table(hwmgr->device); + (ATOM_VOLTAGE_OBJECT_INFO_V3_1 *)get_voltage_info_table(hwmgr->adev); const ATOM_VOLTAGE_OBJECT_V3 *voltage_object; unsigned int i; @@ -572,7 +606,7 @@ static ATOM_GPIO_PIN_LUT *get_gpio_lookup_table(void *device) void *table_address; table_address = (ATOM_GPIO_PIN_LUT *) - cgs_atom_get_data_table(device, + smu_atom_get_data_table(device, GetIndexIntoMasterTable(DATA, GPIO_Pin_LUT), &size, &frev, &crev); @@ -592,7 +626,7 @@ bool atomctrl_get_pp_assign_pin( { bool bRet = false; ATOM_GPIO_PIN_LUT *gpio_lookup_table = - get_gpio_lookup_table(hwmgr->device); + get_gpio_lookup_table(hwmgr->adev); PP_ASSERT_WITH_CODE((NULL != gpio_lookup_table), "Could not find GPIO lookup Table in BIOS.", return false); @@ -613,7 +647,7 @@ int atomctrl_calculate_voltage_evv_on_sclk( bool debug) { ATOM_ASIC_PROFILING_INFO_V3_4 *getASICProfilingInfo; - + struct amdgpu_device *adev = hwmgr->adev; EFUSE_LINEAR_FUNC_PARAM sRO_fuse; EFUSE_LINEAR_FUNC_PARAM sCACm_fuse; EFUSE_LINEAR_FUNC_PARAM sCACb_fuse; @@ -640,7 +674,7 @@ int atomctrl_calculate_voltage_evv_on_sclk( int result; getASICProfilingInfo = (ATOM_ASIC_PROFILING_INFO_V3_4 *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, GetIndexIntoMasterTable(DATA, ASIC_ProfilingInfo), NULL, NULL, NULL); @@ -706,9 +740,9 @@ int atomctrl_calculate_voltage_evv_on_sclk( sOutput_FuseValues.sEfuse = sInput_FuseValues; - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ReadEfuseValue), - &sOutput_FuseValues); + (uint32_t *)&sOutput_FuseValues); if (result) return result; @@ -727,9 +761,9 @@ int atomctrl_calculate_voltage_evv_on_sclk( sOutput_FuseValues.sEfuse = sInput_FuseValues; - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ReadEfuseValue), - &sOutput_FuseValues); + (uint32_t *)&sOutput_FuseValues); if (result) return result; @@ -747,9 +781,9 @@ int atomctrl_calculate_voltage_evv_on_sclk( sInput_FuseValues.ucBitLength = sCACb_fuse.ucEfuseLength; sOutput_FuseValues.sEfuse = sInput_FuseValues; - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ReadEfuseValue), - &sOutput_FuseValues); + (uint32_t *)&sOutput_FuseValues); if (result) return result; @@ -768,9 +802,9 @@ int atomctrl_calculate_voltage_evv_on_sclk( sOutput_FuseValues.sEfuse = sInput_FuseValues; - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ReadEfuseValue), - &sOutput_FuseValues); + (uint32_t *)&sOutput_FuseValues); if (result) return result; @@ -790,9 +824,9 @@ int atomctrl_calculate_voltage_evv_on_sclk( sOutput_FuseValues.sEfuse = sInput_FuseValues; - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ReadEfuseValue), - &sOutput_FuseValues); + (uint32_t *)&sOutput_FuseValues); if (result) return result; @@ -811,9 +845,9 @@ int atomctrl_calculate_voltage_evv_on_sclk( sInput_FuseValues.ucBitLength = sKv_b_fuse.ucEfuseLength; sOutput_FuseValues.sEfuse = sInput_FuseValues; - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ReadEfuseValue), - &sOutput_FuseValues); + (uint32_t *)&sOutput_FuseValues); if (result) return result; @@ -842,9 +876,9 @@ int atomctrl_calculate_voltage_evv_on_sclk( sOutput_FuseValues.sEfuse = sInput_FuseValues; - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ReadEfuseValue), - &sOutput_FuseValues); + (uint32_t *)&sOutput_FuseValues); if (result) return result; @@ -1053,8 +1087,9 @@ int atomctrl_get_voltage_evv_on_sclk( uint32_t sclk, uint16_t virtual_voltage_Id, uint16_t *voltage) { - int result; + struct amdgpu_device *adev = hwmgr->adev; GET_VOLTAGE_INFO_INPUT_PARAMETER_V1_2 get_voltage_info_param_space; + int result; get_voltage_info_param_space.ucVoltageType = voltage_type; @@ -1065,14 +1100,12 @@ int atomctrl_get_voltage_evv_on_sclk( get_voltage_info_param_space.ulSCLKFreq = cpu_to_le32(sclk); - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, GetVoltageInfo), - &get_voltage_info_param_space); - - if (0 != result) - return result; + (uint32_t *)&get_voltage_info_param_space); - *voltage = le16_to_cpu(((GET_EVV_VOLTAGE_INFO_OUTPUT_PARAMETER_V1_2 *) + *voltage = result ? 0 : + le16_to_cpu(((GET_EVV_VOLTAGE_INFO_OUTPUT_PARAMETER_V1_2 *) (&get_voltage_info_param_space))->usVoltageLevel); return result; @@ -1088,9 +1121,10 @@ int atomctrl_get_voltage_evv(struct pp_hwmgr *hwmgr, uint16_t virtual_voltage_id, uint16_t *voltage) { + struct amdgpu_device *adev = hwmgr->adev; + GET_VOLTAGE_INFO_INPUT_PARAMETER_V1_2 get_voltage_info_param_space; int result; int entry_id; - GET_VOLTAGE_INFO_INPUT_PARAMETER_V1_2 get_voltage_info_param_space; /* search for leakage voltage ID 0xff01 ~ 0xff08 and sckl */ for (entry_id = 0; entry_id < hwmgr->dyn_state.vddc_dependency_on_sclk->count; entry_id++) { @@ -1111,9 +1145,9 @@ int atomctrl_get_voltage_evv(struct pp_hwmgr *hwmgr, get_voltage_info_param_space.ulSCLKFreq = cpu_to_le32(hwmgr->dyn_state.vddc_dependency_on_sclk->entries[entry_id].clk); - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, GetVoltageInfo), - &get_voltage_info_param_space); + (uint32_t *)&get_voltage_info_param_space); if (0 != result) return result; @@ -1135,7 +1169,7 @@ uint32_t atomctrl_get_mpll_reference_clock(struct pp_hwmgr *hwmgr) u16 size; fw_info = (ATOM_COMMON_TABLE_HEADER *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, GetIndexIntoMasterTable(DATA, FirmwareInfo), &size, &frev, &crev); @@ -1167,7 +1201,7 @@ static ATOM_ASIC_INTERNAL_SS_INFO *asic_internal_ss_get_ss_table(void *device) u16 size; table = (ATOM_ASIC_INTERNAL_SS_INFO *) - cgs_atom_get_data_table(device, + smu_atom_get_data_table(device, GetIndexIntoMasterTable(DATA, ASIC_InternalSS_Info), &size, &frev, &crev); @@ -1188,7 +1222,7 @@ static int asic_internal_ss_get_ss_asignment(struct pp_hwmgr *hwmgr, memset(ssEntry, 0x00, sizeof(pp_atomctrl_internal_ss_info)); - table = asic_internal_ss_get_ss_table(hwmgr->device); + table = asic_internal_ss_get_ss_table(hwmgr->adev); if (NULL == table) return -1; @@ -1260,9 +1294,10 @@ int atomctrl_get_engine_clock_spread_spectrum( ASIC_INTERNAL_ENGINE_SS, engine_clock, ssInfo); } -int atomctrl_read_efuse(void *device, uint16_t start_index, +int atomctrl_read_efuse(struct pp_hwmgr *hwmgr, uint16_t start_index, uint16_t end_index, uint32_t mask, uint32_t *efuse) { + struct amdgpu_device *adev = hwmgr->adev; int result; READ_EFUSE_VALUE_PARAMETER efuse_param; @@ -1272,11 +1307,10 @@ int atomctrl_read_efuse(void *device, uint16_t start_index, efuse_param.sEfuse.ucBitLength = (uint8_t) ((end_index - start_index) + 1); - result = cgs_atom_exec_cmd_table(device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, ReadEfuseValue), - &efuse_param); - if (!result) - *efuse = le32_to_cpu(efuse_param.ulEfuseValue) & mask; + (uint32_t *)&efuse_param); + *efuse = result ? 0 : le32_to_cpu(efuse_param.ulEfuseValue) & mask; return result; } @@ -1284,6 +1318,7 @@ int atomctrl_read_efuse(void *device, uint16_t start_index, int atomctrl_set_ac_timing_ai(struct pp_hwmgr *hwmgr, uint32_t memory_clock, uint8_t level) { + struct amdgpu_device *adev = hwmgr->adev; DYNAMICE_MEMORY_SETTINGS_PARAMETER_V2_1 memory_clock_parameters; int result; @@ -1293,10 +1328,9 @@ int atomctrl_set_ac_timing_ai(struct pp_hwmgr *hwmgr, uint32_t memory_clock, ADJUST_MC_SETTING_PARAM; memory_clock_parameters.asDPMMCReg.ucMclkDPMState = level; - result = cgs_atom_exec_cmd_table - (hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, DynamicMemorySettings), - &memory_clock_parameters); + (uint32_t *)&memory_clock_parameters); return result; } @@ -1304,7 +1338,7 @@ int atomctrl_set_ac_timing_ai(struct pp_hwmgr *hwmgr, uint32_t memory_clock, int atomctrl_get_voltage_evv_on_sclk_ai(struct pp_hwmgr *hwmgr, uint8_t voltage_type, uint32_t sclk, uint16_t virtual_voltage_Id, uint32_t *voltage) { - + struct amdgpu_device *adev = hwmgr->adev; int result; GET_VOLTAGE_INFO_INPUT_PARAMETER_V1_3 get_voltage_info_param_space; @@ -1313,15 +1347,12 @@ int atomctrl_get_voltage_evv_on_sclk_ai(struct pp_hwmgr *hwmgr, uint8_t voltage_ get_voltage_info_param_space.usVoltageLevel = cpu_to_le16(virtual_voltage_Id); get_voltage_info_param_space.ulSCLKFreq = cpu_to_le32(sclk); - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, GetVoltageInfo), - &get_voltage_info_param_space); + (uint32_t *)&get_voltage_info_param_space); - if (0 != result) - return result; - - *voltage = le32_to_cpu(((GET_EVV_VOLTAGE_INFO_OUTPUT_PARAMETER_V1_3 *) - (&get_voltage_info_param_space))->ulVoltageLevel); + *voltage = result ? 0 : + le32_to_cpu(((GET_EVV_VOLTAGE_INFO_OUTPUT_PARAMETER_V1_3 *)(&get_voltage_info_param_space))->ulVoltageLevel); return result; } @@ -1334,7 +1365,7 @@ int atomctrl_get_smc_sclk_range_table(struct pp_hwmgr *hwmgr, struct pp_atom_ctr u16 size; ATOM_SMU_INFO_V2_1 *psmu_info = - (ATOM_SMU_INFO_V2_1 *)cgs_atom_get_data_table(hwmgr->device, + (ATOM_SMU_INFO_V2_1 *)smu_atom_get_data_table(hwmgr->adev, GetIndexIntoMasterTable(DATA, SMU_Info), &size, &frev, &crev); @@ -1362,7 +1393,7 @@ int atomctrl_get_avfs_information(struct pp_hwmgr *hwmgr, return -EINVAL; profile = (ATOM_ASIC_PROFILING_INFO_V3_6 *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, GetIndexIntoMasterTable(DATA, ASIC_ProfilingInfo), NULL, NULL, NULL); if (!profile) @@ -1402,7 +1433,7 @@ int atomctrl_get_svi2_info(struct pp_hwmgr *hwmgr, uint8_t voltage_type, uint16_t *load_line) { ATOM_VOLTAGE_OBJECT_INFO_V3_1 *voltage_info = - (ATOM_VOLTAGE_OBJECT_INFO_V3_1 *)get_voltage_info_table(hwmgr->device); + (ATOM_VOLTAGE_OBJECT_INFO_V3_1 *)get_voltage_info_table(hwmgr->adev); const ATOM_VOLTAGE_OBJECT_V3 *voltage_object; @@ -1421,16 +1452,17 @@ int atomctrl_get_svi2_info(struct pp_hwmgr *hwmgr, uint8_t voltage_type, int atomctrl_get_leakage_id_from_efuse(struct pp_hwmgr *hwmgr, uint16_t *virtual_voltage_id) { - int result; + struct amdgpu_device *adev = hwmgr->adev; SET_VOLTAGE_PS_ALLOCATION allocation; SET_VOLTAGE_PARAMETERS_V1_3 *voltage_parameters = (SET_VOLTAGE_PARAMETERS_V1_3 *)&allocation.sASICSetVoltage; + int result; voltage_parameters->ucVoltageMode = ATOM_GET_LEAKAGE_ID; - result = cgs_atom_exec_cmd_table(hwmgr->device, + result = amdgpu_atom_execute_table(adev->mode_info.atom_context, GetIndexIntoMasterTable(COMMAND, SetVoltage), - voltage_parameters); + (uint32_t *)voltage_parameters); *virtual_voltage_id = voltage_parameters->usVoltageLevel; @@ -1453,7 +1485,7 @@ int atomctrl_get_leakage_vddc_base_on_leakage(struct pp_hwmgr *hwmgr, ix = GetIndexIntoMasterTable(DATA, ASIC_ProfilingInfo); profile = (ATOM_ASIC_PROFILING_INFO_V2_1 *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, ix, NULL, NULL, NULL); if (!profile) @@ -1498,3 +1530,33 @@ int atomctrl_get_leakage_vddc_base_on_leakage(struct pp_hwmgr *hwmgr, return 0; } + +void atomctrl_get_voltage_range(struct pp_hwmgr *hwmgr, uint32_t *max_vddc, + uint32_t *min_vddc) +{ + void *profile; + + profile = smu_atom_get_data_table(hwmgr->adev, + GetIndexIntoMasterTable(DATA, ASIC_ProfilingInfo), + NULL, NULL, NULL); + + if (profile) { + switch (hwmgr->chip_id) { + case CHIP_TONGA: + case CHIP_FIJI: + *max_vddc = le32_to_cpu(((ATOM_ASIC_PROFILING_INFO_V3_3 *)profile)->ulMaxVddc/4); + *min_vddc = le32_to_cpu(((ATOM_ASIC_PROFILING_INFO_V3_3 *)profile)->ulMinVddc/4); + return; + case CHIP_POLARIS11: + case CHIP_POLARIS10: + case CHIP_POLARIS12: + *max_vddc = le32_to_cpu(((ATOM_ASIC_PROFILING_INFO_V3_6 *)profile)->ulMaxVddc/100); + *min_vddc = le32_to_cpu(((ATOM_ASIC_PROFILING_INFO_V3_6 *)profile)->ulMinVddc/100); + return; + default: + break; + } + } + *max_vddc = 0; + *min_vddc = 0; +} diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.h b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.h index c44a92064cf1b..3ee54f1829432 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomctrl.h @@ -146,6 +146,14 @@ struct pp_atomctrl_memory_clock_param { }; typedef struct pp_atomctrl_memory_clock_param pp_atomctrl_memory_clock_param; +struct pp_atomctrl_memory_clock_param_ai { + uint32_t ulClock; + uint32_t ulPostDiv; + uint16_t ulMclk_fcw_frac; + uint16_t ulMclk_fcw_int; +}; +typedef struct pp_atomctrl_memory_clock_param_ai pp_atomctrl_memory_clock_param_ai; + struct pp_atomctrl_internal_ss_info { uint32_t speed_spectrum_percentage; /* in 1/100 percentage */ uint32_t speed_spectrum_rate; /* in KHz */ @@ -295,10 +303,12 @@ extern bool atomctrl_is_voltage_controlled_by_gpio_v3(struct pp_hwmgr *hwmgr, ui extern int atomctrl_get_voltage_table_v3(struct pp_hwmgr *hwmgr, uint8_t voltage_type, uint8_t voltage_mode, pp_atomctrl_voltage_table *voltage_table); extern int atomctrl_get_memory_pll_dividers_vi(struct pp_hwmgr *hwmgr, uint32_t clock_value, pp_atomctrl_memory_clock_param *mpll_param); +extern int atomctrl_get_memory_pll_dividers_ai(struct pp_hwmgr *hwmgr, + uint32_t clock_value, pp_atomctrl_memory_clock_param_ai *mpll_param); extern int atomctrl_get_engine_pll_dividers_kong(struct pp_hwmgr *hwmgr, uint32_t clock_value, pp_atomctrl_clock_dividers_kong *dividers); -extern int atomctrl_read_efuse(void *device, uint16_t start_index, +extern int atomctrl_read_efuse(struct pp_hwmgr *hwmgr, uint16_t start_index, uint16_t end_index, uint32_t mask, uint32_t *efuse); extern int atomctrl_calculate_voltage_evv_on_sclk(struct pp_hwmgr *hwmgr, uint8_t voltage_type, uint32_t sclk, uint16_t virtual_voltage_Id, uint16_t *voltage, uint16_t dpm_level, bool debug); @@ -320,5 +330,8 @@ extern int atomctrl_get_leakage_vddc_base_on_leakage(struct pp_hwmgr *hwmgr, uint16_t virtual_voltage_id, uint16_t efuse_voltage_id); extern int atomctrl_get_leakage_id_from_efuse(struct pp_hwmgr *hwmgr, uint16_t *virtual_voltage_id); + +extern void atomctrl_get_voltage_range(struct pp_hwmgr *hwmgr, uint32_t *max_vddc, + uint32_t *min_vddc); #endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.c b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.c index ad42caac033e4..c97b0e5ba43b6 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.c @@ -23,9 +23,9 @@ #include "ppatomfwctrl.h" #include "atomfirmware.h" +#include "atom.h" #include "pp_debug.h" - static const union atom_voltage_object_v4 *pp_atomfwctrl_lookup_voltage_type_v4( const struct atom_voltage_objects_info_v4_1 *voltage_object_info_table, uint8_t voltage_type, uint8_t voltage_mode) @@ -38,35 +38,34 @@ static const union atom_voltage_object_v4 *pp_atomfwctrl_lookup_voltage_type_v4( while (offset < size) { const union atom_voltage_object_v4 *voltage_object = - (const union atom_voltage_object_v4 *)(start + offset); + (const union atom_voltage_object_v4 *)(start + offset); - if (voltage_type == voltage_object->gpio_voltage_obj.header.voltage_type && - voltage_mode == voltage_object->gpio_voltage_obj.header.voltage_mode) - return voltage_object; + if (voltage_type == voltage_object->gpio_voltage_obj.header.voltage_type && + voltage_mode == voltage_object->gpio_voltage_obj.header.voltage_mode) + return voltage_object; - offset += le16_to_cpu(voltage_object->gpio_voltage_obj.header.object_size); + offset += le16_to_cpu(voltage_object->gpio_voltage_obj.header.object_size); - } + } - return NULL; + return NULL; } static struct atom_voltage_objects_info_v4_1 *pp_atomfwctrl_get_voltage_info_table( struct pp_hwmgr *hwmgr) { - const void *table_address; - uint16_t idx; + const void *table_address; + uint16_t idx; - idx = GetIndexIntoMasterDataTable(voltageobject_info); - table_address = cgs_atom_get_data_table(hwmgr->device, - idx, NULL, NULL, NULL); + idx = GetIndexIntoMasterDataTable(voltageobject_info); + table_address = smu_atom_get_data_table(hwmgr->adev, + idx, NULL, NULL, NULL); - PP_ASSERT_WITH_CODE( - table_address, - "Error retrieving BIOS Table Address!", - return NULL); + PP_ASSERT_WITH_CODE(table_address, + "Error retrieving BIOS Table Address!", + return NULL); - return (struct atom_voltage_objects_info_v4_1 *)table_address; + return (struct atom_voltage_objects_info_v4_1 *)table_address; } /** @@ -167,7 +166,7 @@ static struct atom_gpio_pin_lut_v2_1 *pp_atomfwctrl_get_gpio_lookup_table( uint16_t idx; idx = GetIndexIntoMasterDataTable(gpio_pin_lut); - table_address = cgs_atom_get_data_table(hwmgr->device, + table_address = smu_atom_get_data_table(hwmgr->adev, idx, NULL, NULL, NULL); PP_ASSERT_WITH_CODE(table_address, "Error retrieving BIOS Table Address!", @@ -248,28 +247,30 @@ int pp_atomfwctrl_get_gpu_pll_dividers_vega10(struct pp_hwmgr *hwmgr, uint32_t clock_type, uint32_t clock_value, struct pp_atomfwctrl_clock_dividers_soc15 *dividers) { + struct amdgpu_device *adev = hwmgr->adev; struct compute_gpu_clock_input_parameter_v1_8 pll_parameters; struct compute_gpu_clock_output_parameter_v1_8 *pll_output; - int result; uint32_t idx; pll_parameters.gpuclock_10khz = (uint32_t)clock_value; pll_parameters.gpu_clock_type = clock_type; idx = GetIndexIntoMasterCmdTable(computegpuclockparam); - result = cgs_atom_exec_cmd_table(hwmgr->device, idx, &pll_parameters); - - if (!result) { - pll_output = (struct compute_gpu_clock_output_parameter_v1_8 *) - &pll_parameters; - dividers->ulClock = le32_to_cpu(pll_output->gpuclock_10khz); - dividers->ulDid = le32_to_cpu(pll_output->dfs_did); - dividers->ulPll_fb_mult = le32_to_cpu(pll_output->pll_fb_mult); - dividers->ulPll_ss_fbsmult = le32_to_cpu(pll_output->pll_ss_fbsmult); - dividers->usPll_ss_slew_frac = le16_to_cpu(pll_output->pll_ss_slew_frac); - dividers->ucPll_ss_enable = pll_output->pll_ss_enable; - } - return result; + + if (amdgpu_atom_execute_table( + adev->mode_info.atom_context, idx, (uint32_t *)&pll_parameters)) + return -EINVAL; + + pll_output = (struct compute_gpu_clock_output_parameter_v1_8 *) + &pll_parameters; + dividers->ulClock = le32_to_cpu(pll_output->gpuclock_10khz); + dividers->ulDid = le32_to_cpu(pll_output->dfs_did); + dividers->ulPll_fb_mult = le32_to_cpu(pll_output->pll_fb_mult); + dividers->ulPll_ss_fbsmult = le32_to_cpu(pll_output->pll_ss_fbsmult); + dividers->usPll_ss_slew_frac = le16_to_cpu(pll_output->pll_ss_slew_frac); + dividers->ucPll_ss_enable = pll_output->pll_ss_enable; + + return 0; } int pp_atomfwctrl_get_avfs_information(struct pp_hwmgr *hwmgr, @@ -283,7 +284,7 @@ int pp_atomfwctrl_get_avfs_information(struct pp_hwmgr *hwmgr, idx = GetIndexIntoMasterDataTable(asic_profiling_info); profile = (struct atom_asic_profiling_info_v4_1 *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, idx, NULL, NULL, NULL); if (!profile) @@ -467,7 +468,7 @@ int pp_atomfwctrl_get_gpio_information(struct pp_hwmgr *hwmgr, idx = GetIndexIntoMasterDataTable(smu_info); info = (struct atom_smu_info_v3_1 *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, idx, NULL, NULL, NULL); if (!info) { @@ -487,8 +488,9 @@ int pp_atomfwctrl_get_gpio_information(struct pp_hwmgr *hwmgr, return 0; } -int pp_atomfwctrl__get_clk_information_by_clkid(struct pp_hwmgr *hwmgr, BIOS_CLKID id, uint32_t *frequency) +int pp_atomfwctrl_get_clk_information_by_clkid(struct pp_hwmgr *hwmgr, BIOS_CLKID id, uint32_t *frequency) { + struct amdgpu_device *adev = hwmgr->adev; struct atom_get_smu_clock_info_parameters_v3_1 parameters; struct atom_get_smu_clock_info_output_parameters_v3_1 *output; uint32_t ix; @@ -497,13 +499,13 @@ int pp_atomfwctrl__get_clk_information_by_clkid(struct pp_hwmgr *hwmgr, BIOS_CLK parameters.command = GET_SMU_CLOCK_INFO_V3_1_GET_CLOCK_FREQ; ix = GetIndexIntoMasterCmdTable(getsmuclockinfo); - if (!cgs_atom_exec_cmd_table(hwmgr->device, ix, ¶meters)) { - output = (struct atom_get_smu_clock_info_output_parameters_v3_1 *)¶meters; - *frequency = output->atom_smu_outputclkfreq.smu_clock_freq_hz / 10000; - } else { - pr_info("Error execute_table getsmuclockinfo!"); - return -1; - } + + if (amdgpu_atom_execute_table( + adev->mode_info.atom_context, ix, (uint32_t *)¶meters)) + return -EINVAL; + + output = (struct atom_get_smu_clock_info_output_parameters_v3_1 *)¶meters; + *frequency = output->atom_smu_outputclkfreq.smu_clock_freq_hz / 10000; return 0; } @@ -513,11 +515,10 @@ int pp_atomfwctrl_get_vbios_bootup_values(struct pp_hwmgr *hwmgr, { struct atom_firmware_info_v3_1 *info = NULL; uint16_t ix; - uint32_t frequency = 0; ix = GetIndexIntoMasterDataTable(firmwareinfo); info = (struct atom_firmware_info_v3_1 *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, ix, NULL, NULL, NULL); if (!info) { @@ -536,12 +537,6 @@ int pp_atomfwctrl_get_vbios_bootup_values(struct pp_hwmgr *hwmgr, boot_values->ulSocClk = 0; boot_values->ulDCEFClk = 0; - if (!pp_atomfwctrl__get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_SOCCLK_ID, &frequency)) - boot_values->ulSocClk = frequency; - - if (!pp_atomfwctrl__get_clk_information_by_clkid(hwmgr, SMU9_SYSPLL0_DCEFCLK_ID, &frequency)) - boot_values->ulDCEFClk = frequency; - return 0; } @@ -553,7 +548,7 @@ int pp_atomfwctrl_get_smc_dpm_information(struct pp_hwmgr *hwmgr, ix = GetIndexIntoMasterDataTable(smc_dpm_info); info = (struct atom_smc_dpm_info_v4_1 *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, ix, NULL, NULL, NULL); if (!info) { pr_info("Error retrieving BIOS Table Address!"); diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.h b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.h index 8df1e84f27c91..fe10aa4db5e64 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/ppatomfwctrl.h @@ -230,6 +230,8 @@ int pp_atomfwctrl_get_vbios_bootup_values(struct pp_hwmgr *hwmgr, struct pp_atomfwctrl_bios_boot_up_values *boot_values); int pp_atomfwctrl_get_smc_dpm_information(struct pp_hwmgr *hwmgr, struct pp_atomfwctrl_smc_dpm_parameters *param); +int pp_atomfwctrl_get_clk_information_by_clkid(struct pp_hwmgr *hwmgr, + BIOS_CLKID id, uint32_t *frequency); #endif diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/process_pptables_v1_0.c b/drivers/gpu/drm/amd/powerplay/hwmgr/process_pptables_v1_0.c index c9eecce5683f7..f0d48b183d22c 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/process_pptables_v1_0.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/process_pptables_v1_0.c @@ -141,7 +141,7 @@ static const void *get_powerplay_table(struct pp_hwmgr *hwmgr) if (!table_address) { table_address = (ATOM_Tonga_POWERPLAYTABLE *) - cgs_atom_get_data_table(hwmgr->device, + smu_atom_get_data_table(hwmgr->adev, index, &size, &frev, &crev); hwmgr->soft_pp_table = table_address; /*Cache the result in RAM.*/ hwmgr->soft_pp_table_size = size; @@ -728,6 +728,32 @@ static int get_mm_clock_voltage_table( return 0; } +static int get_gpio_table(struct pp_hwmgr *hwmgr, + struct phm_ppt_v1_gpio_table **pp_tonga_gpio_table, + const ATOM_Tonga_GPIO_Table *atom_gpio_table) +{ + uint32_t table_size; + struct phm_ppt_v1_gpio_table *pp_gpio_table; + struct phm_ppt_v1_information *pp_table_information = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + + table_size = sizeof(struct phm_ppt_v1_gpio_table); + pp_gpio_table = kzalloc(table_size, GFP_KERNEL); + if (!pp_gpio_table) + return -ENOMEM; + + if (pp_table_information->vdd_dep_on_sclk->count < + atom_gpio_table->ucVRHotTriggeredSclkDpmIndex) + PP_ASSERT_WITH_CODE(false, + "SCLK DPM index for VRHot cannot exceed the total sclk level count!",); + else + pp_gpio_table->vrhot_triggered_sclk_dpm_index = + atom_gpio_table->ucVRHotTriggeredSclkDpmIndex; + + *pp_tonga_gpio_table = pp_gpio_table; + + return 0; +} /** * Private Function used during initialization. * Initialize clock voltage dependency @@ -761,11 +787,15 @@ static int init_clock_voltage_dependency( const PPTable_Generic_SubTable_Header *pcie_table = (const PPTable_Generic_SubTable_Header *)(((unsigned long) powerplay_table) + le16_to_cpu(powerplay_table->usPCIETableOffset)); + const ATOM_Tonga_GPIO_Table *gpio_table = + (const ATOM_Tonga_GPIO_Table *)(((unsigned long) powerplay_table) + + le16_to_cpu(powerplay_table->usGPIOTableOffset)); pp_table_information->vdd_dep_on_sclk = NULL; pp_table_information->vdd_dep_on_mclk = NULL; pp_table_information->mm_dep_table = NULL; pp_table_information->pcie_table = NULL; + pp_table_information->gpio_table = NULL; if (powerplay_table->usMMDependencyTableOffset != 0) result = get_mm_clock_voltage_table(hwmgr, @@ -810,6 +840,10 @@ static int init_clock_voltage_dependency( result = get_valid_clk(hwmgr, &pp_table_information->valid_sclk_values, pp_table_information->vdd_dep_on_sclk); + if (!result && gpio_table) + result = get_gpio_table(hwmgr, &pp_table_information->gpio_table, + gpio_table); + return result; } @@ -1116,6 +1150,9 @@ static int pp_tables_v1_0_uninitialize(struct pp_hwmgr *hwmgr) kfree(pp_table_information->pcie_table); pp_table_information->pcie_table = NULL; + kfree(pp_table_information->gpio_table); + pp_table_information->gpio_table = NULL; + kfree(hwmgr->pptable); hwmgr->pptable = NULL; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/processpptables.c b/drivers/gpu/drm/amd/powerplay/hwmgr/processpptables.c index 36ca7c419c90a..ce64dfabd34bd 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/processpptables.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/processpptables.c @@ -837,7 +837,7 @@ static const ATOM_PPLIB_POWERPLAYTABLE *get_powerplay_table( hwmgr->soft_pp_table = &soft_dummy_pp_table[0]; hwmgr->soft_pp_table_size = sizeof(soft_dummy_pp_table); } else { - table_addr = cgs_atom_get_data_table(hwmgr->device, + table_addr = smu_atom_get_data_table(hwmgr->adev, GetIndexIntoMasterTable(DATA, PowerPlayInfo), &size, &frev, &crev); hwmgr->soft_pp_table = table_addr; @@ -1058,7 +1058,7 @@ static int init_overdrive_limits(struct pp_hwmgr *hwmgr, return 0; /* We assume here that fw_info is unchanged if this call fails.*/ - fw_info = cgs_atom_get_data_table(hwmgr->device, + fw_info = smu_atom_get_data_table(hwmgr->adev, GetIndexIntoMasterTable(DATA, FirmwareInfo), &size, &frev, &crev); diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c index 10253b89b3d87..85f84f4d8be57 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.c @@ -34,7 +34,7 @@ #include "rv_ppsmc.h" #include "smu10_hwmgr.h" #include "power_state.h" -#include "pp_soc15.h" +#include "soc15_common.h" #define SMU10_MAX_DEEPSLEEP_DIVIDER_ID 5 #define SMU10_MINIMUM_ENGINE_CLOCK 800 /* 8Mhz, the low boundary of engine clock allowed on this chip */ @@ -42,6 +42,13 @@ #define SMU10_DISPCLK_BYPASS_THRESHOLD 10000 /* 100Mhz */ #define SMC_RAM_END 0x40000 +#define mmPWR_MISC_CNTL_STATUS 0x0183 +#define mmPWR_MISC_CNTL_STATUS_BASE_IDX 0 +#define PWR_MISC_CNTL_STATUS__PWR_GFX_RLC_CGPG_EN__SHIFT 0x0 +#define PWR_MISC_CNTL_STATUS__PWR_GFXOFF_STATUS__SHIFT 0x1 +#define PWR_MISC_CNTL_STATUS__PWR_GFX_RLC_CGPG_EN_MASK 0x00000001L +#define PWR_MISC_CNTL_STATUS__PWR_GFXOFF_STATUS_MASK 0x00000006L + static const unsigned long SMU10_Magic = (unsigned long) PHM_Rv_Magic; @@ -74,11 +81,15 @@ static int smu10_initialize_dpm_defaults(struct pp_hwmgr *hwmgr) smu10_data->thermal_auto_throttling_treshold = 0; smu10_data->is_nb_dpm_enabled = 1; smu10_data->dpm_flags = 1; - smu10_data->gfx_off_controled_by_driver = false; smu10_data->need_min_deep_sleep_dcefclk = true; smu10_data->num_active_display = 0; smu10_data->deep_sleep_dcefclk = 0; + if (hwmgr->feature_mask & PP_GFXOFF_MASK) + smu10_data->gfx_off_controled_by_driver = true; + else + smu10_data->gfx_off_controled_by_driver = false; + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_SclkDeepSleep); @@ -161,7 +172,7 @@ static int smu10_set_clock_limit(struct pp_hwmgr *hwmgr, const void *input) struct PP_Clocks clocks = {0}; struct pp_display_clock_request clock_req; - clocks.dcefClock = hwmgr->display_config.min_dcef_set_clk; + clocks.dcefClock = hwmgr->display_config->min_dcef_set_clk; clock_req.clock_type = amd_pp_dcf_clock; clock_req.clock_freq_in_khz = clocks.dcefClock * 10; @@ -206,12 +217,18 @@ static int smu10_set_power_state_tasks(struct pp_hwmgr *hwmgr, const void *input static int smu10_init_power_gate_state(struct pp_hwmgr *hwmgr) { struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); + struct amdgpu_device *adev = hwmgr->adev; smu10_data->vcn_power_gated = true; smu10_data->isp_tileA_power_gated = true; smu10_data->isp_tileB_power_gated = true; - return 0; + if (adev->pg_flags & AMD_PG_SUPPORT_GFX_PG) + return smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetGfxCGPG, + true); + else + return 0; } @@ -237,13 +254,31 @@ static int smu10_power_off_asic(struct pp_hwmgr *hwmgr) return smu10_reset_cc6_data(hwmgr); } +static bool smu10_is_gfx_on(struct pp_hwmgr *hwmgr) +{ + uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; + + reg = RREG32_SOC15(PWR, 0, mmPWR_MISC_CNTL_STATUS); + if ((reg & PWR_MISC_CNTL_STATUS__PWR_GFXOFF_STATUS_MASK) == + (0x2 << PWR_MISC_CNTL_STATUS__PWR_GFXOFF_STATUS__SHIFT)) + return true; + + return false; +} + static int smu10_disable_gfx_off(struct pp_hwmgr *hwmgr) { struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); - if (smu10_data->gfx_off_controled_by_driver) + if (smu10_data->gfx_off_controled_by_driver) { smum_send_msg_to_smc(hwmgr, PPSMC_MSG_DisableGfxOff); + /* confirm gfx is back to "on" state */ + while (!smu10_is_gfx_on(hwmgr)) + msleep(1); + } + return 0; } @@ -267,6 +302,14 @@ static int smu10_enable_dpm_tasks(struct pp_hwmgr *hwmgr) return smu10_enable_gfx_off(hwmgr); } +static int smu10_gfx_off_control(struct pp_hwmgr *hwmgr, bool enable) +{ + if (enable) + return smu10_enable_gfx_off(hwmgr); + else + return smu10_disable_gfx_off(hwmgr); +} + static int smu10_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, struct pp_power_state *prequest_ps, const struct pp_power_state *pcurrent_ps) @@ -340,7 +383,7 @@ static int smu10_get_clock_voltage_dependency_table(struct pp_hwmgr *hwmgr, static int smu10_populate_clock_table(struct pp_hwmgr *hwmgr) { - int result; + uint32_t result; struct smu10_hwmgr *smu10_data = (struct smu10_hwmgr *)(hwmgr->backend); DpmClocks_t *table = &(smu10_data->clock_table); @@ -386,11 +429,11 @@ static int smu10_populate_clock_table(struct pp_hwmgr *hwmgr) smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMinGfxclkFrequency); result = smum_get_argument(hwmgr); - smu10_data->gfx_min_freq_limit = result * 100; + smu10_data->gfx_min_freq_limit = result / 10 * 1000; smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetMaxGfxclkFrequency); result = smum_get_argument(hwmgr); - smu10_data->gfx_max_freq_limit = result * 100; + smu10_data->gfx_max_freq_limit = result / 10 * 1000; return 0; } @@ -436,8 +479,8 @@ static int smu10_hwmgr_backend_init(struct pp_hwmgr *hwmgr) hwmgr->platform_descriptor.minimumClocksReductionPercentage = 50; - hwmgr->pstate_sclk = SMU10_UMD_PSTATE_GFXCLK; - hwmgr->pstate_mclk = SMU10_UMD_PSTATE_FCLK; + hwmgr->pstate_sclk = SMU10_UMD_PSTATE_GFXCLK * 100; + hwmgr->pstate_mclk = SMU10_UMD_PSTATE_FCLK * 100; return result; } @@ -472,6 +515,8 @@ static int smu10_hwmgr_backend_fini(struct pp_hwmgr *hwmgr) static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, enum amd_dpm_forced_level level) { + struct smu10_hwmgr *data = hwmgr->backend; + if (hwmgr->smu_version < 0x1E3700) { pr_info("smu firmware version too old, can not set dpm level\n"); return 0; @@ -482,7 +527,7 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, case AMD_DPM_FORCED_LEVEL_PROFILE_PEAK: smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinGfxClk, - SMU10_UMD_PSTATE_PEAK_GFXCLK); + data->gfx_max_freq_limit/100); smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinFclkByFreq, SMU10_UMD_PSTATE_PEAK_FCLK); @@ -495,7 +540,7 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetSoftMaxGfxClk, - SMU10_UMD_PSTATE_PEAK_GFXCLK); + data->gfx_max_freq_limit/100); smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetSoftMaxFclkByFreq, SMU10_UMD_PSTATE_PEAK_FCLK); @@ -509,10 +554,10 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_SCLK: smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinGfxClk, - SMU10_UMD_PSTATE_MIN_GFXCLK); + data->gfx_min_freq_limit/100); smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetSoftMaxGfxClk, - SMU10_UMD_PSTATE_MIN_GFXCLK); + data->gfx_min_freq_limit/100); break; case AMD_DPM_FORCED_LEVEL_PROFILE_MIN_MCLK: smum_send_msg_to_smc_with_parameter(hwmgr, @@ -552,10 +597,13 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, case AMD_DPM_FORCED_LEVEL_AUTO: smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinGfxClk, - SMU10_UMD_PSTATE_MIN_GFXCLK); + data->gfx_min_freq_limit/100); smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinFclkByFreq, + hwmgr->display_config->num_display > 3 ? + SMU10_UMD_PSTATE_PEAK_FCLK : SMU10_UMD_PSTATE_MIN_FCLK); + smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinSocclkByFreq, SMU10_UMD_PSTATE_MIN_SOCCLK); @@ -565,7 +613,7 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetSoftMaxGfxClk, - SMU10_UMD_PSTATE_PEAK_GFXCLK); + data->gfx_max_freq_limit/100); smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetSoftMaxFclkByFreq, SMU10_UMD_PSTATE_PEAK_FCLK); @@ -579,10 +627,10 @@ static int smu10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, case AMD_DPM_FORCED_LEVEL_LOW: smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinGfxClk, - SMU10_UMD_PSTATE_MIN_GFXCLK); + data->gfx_min_freq_limit/100); smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetSoftMaxGfxClk, - SMU10_UMD_PSTATE_MIN_GFXCLK); + data->gfx_min_freq_limit/100); smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SetHardMinFclkByFreq, SMU10_UMD_PSTATE_MIN_FCLK); @@ -699,6 +747,16 @@ static int smu10_set_cpu_power_state(struct pp_hwmgr *hwmgr) static int smu10_store_cc6_data(struct pp_hwmgr *hwmgr, uint32_t separation_time, bool cc6_disable, bool pstate_disable, bool pstate_switch_disable) { + struct smu10_hwmgr *data = (struct smu10_hwmgr *)(hwmgr->backend); + + if (separation_time != data->separation_time || + cc6_disable != data->cc6_disable || + pstate_disable != data->pstate_disable) { + data->separation_time = separation_time; + data->cc6_disable = cc6_disable; + data->pstate_disable = pstate_disable; + data->cc6_setting_changed = true; + } return 0; } @@ -711,6 +769,51 @@ static int smu10_get_dal_power_level(struct pp_hwmgr *hwmgr, static int smu10_force_clock_level(struct pp_hwmgr *hwmgr, enum pp_clock_type type, uint32_t mask) { + struct smu10_hwmgr *data = hwmgr->backend; + struct smu10_voltage_dependency_table *mclk_table = + data->clock_vol_info.vdd_dep_on_fclk; + uint32_t low, high; + + low = mask ? (ffs(mask) - 1) : 0; + high = mask ? (fls(mask) - 1) : 0; + + switch (type) { + case PP_SCLK: + if (low > 2 || high > 2) { + pr_info("Currently sclk only support 3 levels on RV\n"); + return -EINVAL; + } + + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetHardMinGfxClk, + low == 2 ? data->gfx_max_freq_limit/100 : + low == 1 ? SMU10_UMD_PSTATE_GFXCLK : + data->gfx_min_freq_limit/100); + + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMaxGfxClk, + high == 0 ? data->gfx_min_freq_limit/100 : + high == 1 ? SMU10_UMD_PSTATE_GFXCLK : + data->gfx_max_freq_limit/100); + break; + + case PP_MCLK: + if (low > mclk_table->count - 1 || high > mclk_table->count - 1) + return -EINVAL; + + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetHardMinFclkByFreq, + mclk_table->entries[low].clk/100); + + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMaxFclkByFreq, + mclk_table->entries[high].clk/100); + break; + + case PP_PCIE: + default: + break; + } return 0; } @@ -720,21 +823,30 @@ static int smu10_print_clock_levels(struct pp_hwmgr *hwmgr, struct smu10_hwmgr *data = (struct smu10_hwmgr *)(hwmgr->backend); struct smu10_voltage_dependency_table *mclk_table = data->clock_vol_info.vdd_dep_on_fclk; - int i, now, size = 0; + uint32_t i, now, size = 0; switch (type) { case PP_SCLK: smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetGfxclkFrequency); now = smum_get_argument(hwmgr); + /* driver only know min/max gfx_clk, Add level 1 for all other gfx clks */ + if (now == data->gfx_max_freq_limit/100) + i = 2; + else if (now == data->gfx_min_freq_limit/100) + i = 0; + else + i = 1; + size += sprintf(buf + size, "0: %uMhz %s\n", - data->gfx_min_freq_limit / 100, - ((data->gfx_min_freq_limit / 100) - == now) ? "*" : ""); + data->gfx_min_freq_limit/100, + i == 0 ? "*" : ""); size += sprintf(buf + size, "1: %uMhz %s\n", - data->gfx_max_freq_limit / 100, - ((data->gfx_max_freq_limit / 100) - == now) ? "*" : ""); + i == 1 ? now : SMU10_UMD_PSTATE_GFXCLK, + i == 1 ? "*" : ""); + size += sprintf(buf + size, "2: %uMhz %s\n", + data->gfx_max_freq_limit/100, + i == 2 ? "*" : ""); break; case PP_MCLK: smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetFclkFrequency); @@ -947,9 +1059,8 @@ static int smu10_get_max_high_clocks(struct pp_hwmgr *hwmgr, struct amd_pp_simpl static int smu10_thermal_get_temperature(struct pp_hwmgr *hwmgr) { - uint32_t reg_offset = soc15_get_register_offset(THM_HWID, 0, - mmTHM_TCON_CUR_TMP_BASE_IDX, mmTHM_TCON_CUR_TMP); - uint32_t reg_value = cgs_read_register(hwmgr->device, reg_offset); + struct amdgpu_device *adev = hwmgr->adev; + uint32_t reg_value = RREG32_SOC15(THM, 0, mmTHM_TCON_CUR_TMP); int cur_temp = (reg_value & THM_TCON_CUR_TMP__CUR_TEMP_MASK) >> THM_TCON_CUR_TMP__CUR_TEMP__SHIFT; @@ -993,11 +1104,47 @@ static int smu10_read_sensor(struct pp_hwmgr *hwmgr, int idx, return ret; } +static int smu10_set_watermarks_for_clocks_ranges(struct pp_hwmgr *hwmgr, + struct pp_wm_sets_with_clock_ranges_soc15 *wm_with_clock_ranges) +{ + struct smu10_hwmgr *data = hwmgr->backend; + Watermarks_t *table = &(data->water_marks_table); + int result = 0; + + smu_set_watermarks_for_clocks_ranges(table,wm_with_clock_ranges); + smum_smc_table_manager(hwmgr, (uint8_t *)table, (uint16_t)SMU10_WMTABLE, false); + data->water_marks_exist = true; + return result; +} + +static int smu10_smus_notify_pwe(struct pp_hwmgr *hwmgr) +{ + + return smum_send_msg_to_smc(hwmgr, PPSMC_MSG_SetRccPfcPmeRestoreRegister); +} + static int smu10_set_mmhub_powergating_by_smu(struct pp_hwmgr *hwmgr) { return smum_send_msg_to_smc(hwmgr, PPSMC_MSG_PowerGateMmHub); } +static void smu10_powergate_vcn(struct pp_hwmgr *hwmgr, bool bgate) +{ + if (bgate) { + amdgpu_device_ip_set_powergating_state(hwmgr->adev, + AMD_IP_BLOCK_TYPE_VCN, + AMD_PG_STATE_GATE); + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_PowerDownVcn, 0); + } else { + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_PowerUpVcn, 0); + amdgpu_device_ip_set_powergating_state(hwmgr->adev, + AMD_IP_BLOCK_TYPE_VCN, + AMD_PG_STATE_UNGATE); + } +} + static const struct pp_hwmgr_func smu10_hwmgr_funcs = { .backend_init = smu10_hwmgr_backend_init, .backend_fini = smu10_hwmgr_backend_fini, @@ -1006,7 +1153,7 @@ static const struct pp_hwmgr_func smu10_hwmgr_funcs = { .force_dpm_level = smu10_dpm_force_dpm_level, .get_power_state_size = smu10_get_power_state_size, .powerdown_uvd = NULL, - .powergate_uvd = NULL, + .powergate_uvd = smu10_powergate_vcn, .powergate_vce = NULL, .get_mclk = smu10_dpm_get_mclk, .get_sclk = smu10_dpm_get_sclk, @@ -1022,6 +1169,7 @@ static const struct pp_hwmgr_func smu10_hwmgr_funcs = { .get_current_shallow_sleep_clocks = smu10_get_current_shallow_sleep_clocks, .get_clock_by_type_with_latency = smu10_get_clock_by_type_with_latency, .get_clock_by_type_with_voltage = smu10_get_clock_by_type_with_voltage, + .set_watermarks_for_clocks_ranges = smu10_set_watermarks_for_clocks_ranges, .get_max_high_clocks = smu10_get_max_high_clocks, .read_sensor = smu10_read_sensor, .set_active_display_count = smu10_set_active_display_count, @@ -1032,6 +1180,8 @@ static const struct pp_hwmgr_func smu10_hwmgr_funcs = { .power_state_set = smu10_set_power_state_tasks, .dynamic_state_management_disable = smu10_disable_dpm_tasks, .set_mmhub_powergating_by_smu = smu10_set_mmhub_powergating_by_smu, + .smus_notify_pwe = smu10_smus_notify_pwe, + .gfx_off_control = smu10_gfx_off_control, }; int smu10_init_function_pointers(struct pp_hwmgr *hwmgr) diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.h b/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.h index 175c3a592b6c5..1fb296a996f3a 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu10_hwmgr.h @@ -290,6 +290,7 @@ struct smu10_hwmgr { bool vcn_dpg_mode; bool gfx_off_controled_by_driver; + bool water_marks_exist; Watermarks_t water_marks_table; struct smu10_clock_voltage_information clock_vol_info; DpmClocks_t clock_table; @@ -310,11 +311,9 @@ int smu10_init_function_pointers(struct pp_hwmgr *hwmgr); #define SMU10_UMD_PSTATE_FCLK 933 #define SMU10_UMD_PSTATE_VCE 0x03C00320 -#define SMU10_UMD_PSTATE_PEAK_GFXCLK 1100 #define SMU10_UMD_PSTATE_PEAK_SOCCLK 757 #define SMU10_UMD_PSTATE_PEAK_FCLK 1200 -#define SMU10_UMD_PSTATE_MIN_GFXCLK 200 #define SMU10_UMD_PSTATE_MIN_FCLK 400 #define SMU10_UMD_PSTATE_MIN_SOCCLK 200 #define SMU10_UMD_PSTATE_MIN_VCE 0x0190012C diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_clockpowergating.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_clockpowergating.c index f4cbaee4e2ca6..6d72a5600917c 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_clockpowergating.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_clockpowergating.c @@ -147,20 +147,20 @@ void smu7_powergate_uvd(struct pp_hwmgr *hwmgr, bool bgate) data->uvd_power_gated = bgate; if (bgate) { - cgs_set_powergating_state(hwmgr->device, + amdgpu_device_ip_set_powergating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_UVD, AMD_PG_STATE_GATE); - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_UVD, AMD_CG_STATE_GATE); smu7_update_uvd_dpm(hwmgr, true); smu7_powerdown_uvd(hwmgr); } else { smu7_powerup_uvd(hwmgr); - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_UVD, AMD_CG_STATE_UNGATE); - cgs_set_powergating_state(hwmgr->device, + amdgpu_device_ip_set_powergating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_UVD, AMD_PG_STATE_UNGATE); smu7_update_uvd_dpm(hwmgr, false); @@ -175,20 +175,20 @@ void smu7_powergate_vce(struct pp_hwmgr *hwmgr, bool bgate) data->vce_power_gated = bgate; if (bgate) { - cgs_set_powergating_state(hwmgr->device, + amdgpu_device_ip_set_powergating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_VCE, AMD_PG_STATE_GATE); - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_VCE, AMD_CG_STATE_GATE); smu7_update_vce_dpm(hwmgr, true); smu7_powerdown_vce(hwmgr); } else { smu7_powerup_vce(hwmgr); - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_VCE, AMD_CG_STATE_UNGATE); - cgs_set_powergating_state(hwmgr->device, + amdgpu_device_ip_set_powergating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_VCE, AMD_PG_STATE_UNGATE); smu7_update_vce_dpm(hwmgr, false); diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c index 26fbeafc3c965..45e9b8cb169d8 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.c @@ -61,10 +61,6 @@ #define SMC_CG_IND_START 0xc0030000 #define SMC_CG_IND_END 0xc0040000 -#define VOLTAGE_SCALE 4 -#define VOLTAGE_VID_OFFSET_SCALE1 625 -#define VOLTAGE_VID_OFFSET_SCALE2 100 - #define MEM_FREQ_LOW_LATENCY 25000 #define MEM_FREQ_HIGH_LATENCY 80000 @@ -79,14 +75,23 @@ #define PCIE_BUS_CLK 10000 #define TCLK (PCIE_BUS_CLK / 10) -static const struct profile_mode_setting smu7_profiling[5] = +static const struct profile_mode_setting smu7_profiling[6] = {{1, 0, 100, 30, 1, 0, 100, 10}, {1, 10, 0, 30, 0, 0, 0, 0}, {0, 0, 0, 0, 1, 10, 16, 31}, {1, 0, 11, 50, 1, 0, 100, 10}, {1, 0, 5, 30, 0, 0, 0, 0}, + {0, 0, 0, 0, 0, 0, 0, 0}, }; +#define PPSMC_MSG_SetVBITimeout_VEGAM ((uint16_t) 0x310) + +#define ixPWR_SVI2_PLANE1_LOAD 0xC0200280 +#define PWR_SVI2_PLANE1_LOAD__PSI1_MASK 0x00000020L +#define PWR_SVI2_PLANE1_LOAD__PSI0_EN_MASK 0x00000040L +#define PWR_SVI2_PLANE1_LOAD__PSI1__SHIFT 0x00000005 +#define PWR_SVI2_PLANE1_LOAD__PSI0_EN__SHIFT 0x00000006 + /** Values for the CG_THERMAL_CTRL::DPM_EVENT_SRC field. */ enum DPM_EVENT_SRC { DPM_EVENT_SRC_ANALOG = 0, @@ -168,6 +173,13 @@ static int smu7_get_current_pcie_lane_number(struct pp_hwmgr *hwmgr) */ static int smu7_enable_smc_voltage_controller(struct pp_hwmgr *hwmgr) { + if (hwmgr->chip_id == CHIP_VEGAM) { + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, + CGS_IND_REG__SMC, PWR_SVI2_PLANE1_LOAD, PSI1, 0); + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, + CGS_IND_REG__SMC, PWR_SVI2_PLANE1_LOAD, PSI0_EN, 0); + } + if (hwmgr->feature_mask & PP_SMC_VOLTAGE_CONTROL_MASK) smum_send_msg_to_smc(hwmgr, PPSMC_MSG_Voltage_Cntl_Enable); @@ -797,32 +809,6 @@ static int smu7_setup_dpm_tables_v1(struct pp_hwmgr *hwmgr) return 0; } -static int smu7_get_voltage_dependency_table( - const struct phm_ppt_v1_clock_voltage_dependency_table *allowed_dep_table, - struct phm_ppt_v1_clock_voltage_dependency_table *dep_table) -{ - uint8_t i = 0; - PP_ASSERT_WITH_CODE((0 != allowed_dep_table->count), - "Voltage Lookup Table empty", - return -EINVAL); - - dep_table->count = allowed_dep_table->count; - for (i=0; i<dep_table->count; i++) { - dep_table->entries[i].clk = allowed_dep_table->entries[i].clk; - dep_table->entries[i].vddInd = allowed_dep_table->entries[i].vddInd; - dep_table->entries[i].vdd_offset = allowed_dep_table->entries[i].vdd_offset; - dep_table->entries[i].vddc = allowed_dep_table->entries[i].vddc; - dep_table->entries[i].vddgfx = allowed_dep_table->entries[i].vddgfx; - dep_table->entries[i].vddci = allowed_dep_table->entries[i].vddci; - dep_table->entries[i].mvdd = allowed_dep_table->entries[i].mvdd; - dep_table->entries[i].phases = allowed_dep_table->entries[i].phases; - dep_table->entries[i].cks_enable = allowed_dep_table->entries[i].cks_enable; - dep_table->entries[i].cks_voffset = allowed_dep_table->entries[i].cks_voffset; - } - - return 0; -} - static int smu7_odn_initial_default_setting(struct pp_hwmgr *hwmgr) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); @@ -850,7 +836,7 @@ static int smu7_odn_initial_default_setting(struct pp_hwmgr *hwmgr) entries[i].vddc = dep_sclk_table->entries[i].vddc; } - smu7_get_voltage_dependency_table(dep_sclk_table, + smu_get_voltage_dependency_table_ppt_v1(dep_sclk_table, (struct phm_ppt_v1_clock_voltage_dependency_table *)&(odn_table->vdd_dependency_on_sclk)); odn_table->odn_memory_clock_dpm_levels.num_of_pl = @@ -862,12 +848,40 @@ static int smu7_odn_initial_default_setting(struct pp_hwmgr *hwmgr) entries[i].vddc = dep_mclk_table->entries[i].vddc; } - smu7_get_voltage_dependency_table(dep_mclk_table, + smu_get_voltage_dependency_table_ppt_v1(dep_mclk_table, (struct phm_ppt_v1_clock_voltage_dependency_table *)&(odn_table->vdd_dependency_on_mclk)); return 0; } +static void smu7_setup_voltage_range_from_vbios(struct pp_hwmgr *hwmgr) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_clock_voltage_dependency_table *dep_sclk_table; + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + uint32_t min_vddc = 0; + uint32_t max_vddc = 0; + + if (!table_info) + return; + + dep_sclk_table = table_info->vdd_dep_on_sclk; + + atomctrl_get_voltage_range(hwmgr, &max_vddc, &min_vddc); + + if (min_vddc == 0 || min_vddc > 2000 + || min_vddc > dep_sclk_table->entries[0].vddc) + min_vddc = dep_sclk_table->entries[0].vddc; + + if (max_vddc == 0 || max_vddc > 2000 + || max_vddc < dep_sclk_table->entries[dep_sclk_table->count-1].vddc) + max_vddc = dep_sclk_table->entries[dep_sclk_table->count-1].vddc; + + data->odn_dpm_table.min_vddc = min_vddc; + data->odn_dpm_table.max_vddc = max_vddc; +} + static int smu7_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); @@ -886,8 +900,10 @@ static int smu7_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) sizeof(struct smu7_dpm_table)); /* initialize ODN table */ - if (hwmgr->od_enabled) + if (hwmgr->od_enabled) { + smu7_setup_voltage_range_from_vbios(hwmgr); smu7_odn_initial_default_setting(hwmgr); + } return 0; } @@ -965,6 +981,22 @@ static int smu7_disable_deep_sleep_master_switch(struct pp_hwmgr *hwmgr) return 0; } +static int smu7_disable_sclk_vce_handshake(struct pp_hwmgr *hwmgr) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + uint32_t soft_register_value = 0; + uint32_t handshake_disables_offset = data->soft_regs_start + + smum_get_offsetof(hwmgr, + SMU_SoftRegisters, HandshakeDisables); + + soft_register_value = cgs_read_ind_register(hwmgr->device, + CGS_IND_REG__SMC, handshake_disables_offset); + soft_register_value |= SMU7_VCE_SCLK_HANDSHAKE_DISABLE; + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + handshake_disables_offset, soft_register_value); + return 0; +} + static int smu7_disable_handshake_uvd(struct pp_hwmgr *hwmgr) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); @@ -987,23 +1019,29 @@ static int smu7_enable_sclk_mclk_dpm(struct pp_hwmgr *hwmgr) struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); /* enable SCLK dpm */ - if (!data->sclk_dpm_key_disabled) + if (!data->sclk_dpm_key_disabled) { + if (hwmgr->chip_id == CHIP_VEGAM) + smu7_disable_sclk_vce_handshake(hwmgr); + PP_ASSERT_WITH_CODE( (0 == smum_send_msg_to_smc(hwmgr, PPSMC_MSG_DPM_Enable)), "Failed to enable SCLK DPM during DPM Start Function!", return -EINVAL); + } /* enable MCLK dpm */ if (0 == data->mclk_dpm_key_disabled) { if (!(hwmgr->feature_mask & PP_UVD_HANDSHAKE_MASK)) smu7_disable_handshake_uvd(hwmgr); + PP_ASSERT_WITH_CODE( (0 == smum_send_msg_to_smc(hwmgr, PPSMC_MSG_MCLKDPM_Enable)), "Failed to enable MCLK DPM during DPM Start Function!", return -EINVAL); - PHM_WRITE_FIELD(hwmgr->device, MC_SEQ_CNTL_3, CAC_EN, 0x1); + if (hwmgr->chip_family != CHIP_VEGAM) + PHM_WRITE_FIELD(hwmgr->device, MC_SEQ_CNTL_3, CAC_EN, 0x1); if (hwmgr->chip_family == AMDGPU_FAMILY_CI) { @@ -1019,8 +1057,13 @@ static int smu7_enable_sclk_mclk_dpm(struct pp_hwmgr *hwmgr) cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixLCAC_MC1_CNTL, 0x5); cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixLCAC_CPL_CNTL, 0x100005); udelay(10); - cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixLCAC_MC0_CNTL, 0x400005); - cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixLCAC_MC1_CNTL, 0x400005); + if (hwmgr->chip_id == CHIP_VEGAM) { + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixLCAC_MC0_CNTL, 0x400009); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixLCAC_MC1_CNTL, 0x400009); + } else { + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixLCAC_MC0_CNTL, 0x400005); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixLCAC_MC1_CNTL, 0x400005); + } cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixLCAC_CPL_CNTL, 0x500005); } } @@ -1229,7 +1272,7 @@ static int smu7_enable_dpm_tasks(struct pp_hwmgr *hwmgr) tmp_result = smu7_construct_voltage_tables(hwmgr); PP_ASSERT_WITH_CODE((0 == tmp_result), - "Failed to contruct voltage tables!", + "Failed to construct voltage tables!", result = tmp_result); } smum_initialize_mc_reg_table(hwmgr); @@ -1261,10 +1304,12 @@ static int smu7_enable_dpm_tasks(struct pp_hwmgr *hwmgr) PP_ASSERT_WITH_CODE((0 == tmp_result), "Failed to process firmware header!", result = tmp_result); - tmp_result = smu7_initial_switch_from_arbf0_to_f1(hwmgr); - PP_ASSERT_WITH_CODE((0 == tmp_result), - "Failed to initialize switch from ArbF0 to F1!", - result = tmp_result); + if (hwmgr->chip_id != CHIP_VEGAM) { + tmp_result = smu7_initial_switch_from_arbf0_to_f1(hwmgr); + PP_ASSERT_WITH_CODE((0 == tmp_result), + "Failed to initialize switch from ArbF0 to F1!", + result = tmp_result); + } result = smu7_setup_default_dpm_tables(hwmgr); PP_ASSERT_WITH_CODE(0 == result, @@ -2754,6 +2799,9 @@ static int smu7_vblank_too_short(struct pp_hwmgr *hwmgr, case CHIP_POLARIS12: switch_limit_us = data->is_memory_gddr5 ? 190 : 150; break; + case CHIP_VEGAM: + switch_limit_us = 30; + break; default: switch_limit_us = data->is_memory_gddr5 ? 450 : 150; break; @@ -2777,8 +2825,6 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, struct PP_Clocks minimum_clocks = {0}; bool disable_mclk_switching; bool disable_mclk_switching_for_frame_lock; - struct cgs_display_info info = {0}; - struct cgs_mode_info mode_info = {0}; const struct phm_clock_and_voltage_limits *max_limits; uint32_t i; struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); @@ -2787,7 +2833,6 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, int32_t count; int32_t stable_pstate_sclk = 0, stable_pstate_mclk = 0; - info.mode_info = &mode_info; data->battery_state = (PP_StateUILabel_Battery == request_ps->classification.ui_label); @@ -2809,10 +2854,8 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, } } - cgs_get_active_displays_info(hwmgr->device, &info); - - minimum_clocks.engineClock = hwmgr->display_config.min_core_set_clock; - minimum_clocks.memoryClock = hwmgr->display_config.min_mem_set_clock; + minimum_clocks.engineClock = hwmgr->display_config->min_core_set_clock; + minimum_clocks.memoryClock = hwmgr->display_config->min_mem_set_clock; if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_StablePState)) { @@ -2843,12 +2886,12 @@ static int smu7_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, PHM_PlatformCaps_DisableMclkSwitchingForFrameLock); - if (info.display_count == 0) + if (hwmgr->display_config->num_display == 0) disable_mclk_switching = false; else - disable_mclk_switching = ((1 < info.display_count) || + disable_mclk_switching = ((1 < hwmgr->display_config->num_display) || disable_mclk_switching_for_frame_lock || - smu7_vblank_too_short(hwmgr, mode_info.vblank_time_us)); + smu7_vblank_too_short(hwmgr, hwmgr->display_config->min_vblank_time)); sclk = smu7_ps->performance_levels[0].engine_clock; mclk = smu7_ps->performance_levels[0].memory_clock; @@ -2957,8 +3000,7 @@ static int smu7_dpm_patch_boot_state(struct pp_hwmgr *hwmgr, /* First retrieve the Boot clocks and VDDC from the firmware info table. * We assume here that fw_info is unchanged if this call fails. */ - fw_info = (ATOM_FIRMWARE_INFO_V2_2 *)cgs_atom_get_data_table( - hwmgr->device, index, + fw_info = (ATOM_FIRMWARE_INFO_V2_2 *)smu_atom_get_data_table(hwmgr->adev, index, &size, &frev, &crev); if (!fw_info) /* During a test, there is no firmware info table. */ @@ -3366,34 +3408,35 @@ static int smu7_get_pp_table_entry(struct pp_hwmgr *hwmgr, return 0; } -static int smu7_get_gpu_power(struct pp_hwmgr *hwmgr, - struct pp_gpu_power *query) +static int smu7_get_gpu_power(struct pp_hwmgr *hwmgr, u32 *query) { - PP_ASSERT_WITH_CODE(!smum_send_msg_to_smc(hwmgr, - PPSMC_MSG_PmStatusLogStart), - "Failed to start pm status log!", - return -1); + int i; + u32 tmp = 0; - /* Sampling period from 50ms to 4sec */ - msleep_interruptible(200); + if (!query) + return -EINVAL; - PP_ASSERT_WITH_CODE(!smum_send_msg_to_smc(hwmgr, - PPSMC_MSG_PmStatusLogSample), - "Failed to sample pm status log!", - return -1); + smum_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_GetCurrPkgPwr, 0); + tmp = cgs_read_register(hwmgr->device, mmSMC_MSG_ARG_0); + *query = tmp; - query->vddc_power = cgs_read_ind_register(hwmgr->device, - CGS_IND_REG__SMC, - ixSMU_PM_STATUS_40); - query->vddci_power = cgs_read_ind_register(hwmgr->device, - CGS_IND_REG__SMC, - ixSMU_PM_STATUS_49); - query->max_gpu_power = cgs_read_ind_register(hwmgr->device, - CGS_IND_REG__SMC, - ixSMU_PM_STATUS_94); - query->average_gpu_power = cgs_read_ind_register(hwmgr->device, - CGS_IND_REG__SMC, - ixSMU_PM_STATUS_95); + if (tmp != 0) + return 0; + + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_PmStatusLogStart); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixSMU_PM_STATUS_94, 0); + + for (i = 0; i < 10; i++) { + mdelay(1); + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_PmStatusLogSample); + tmp = cgs_read_ind_register(hwmgr->device, + CGS_IND_REG__SMC, + ixSMU_PM_STATUS_94); + if (tmp != 0) + break; + } + *query = tmp; return 0; } @@ -3446,10 +3489,7 @@ static int smu7_read_sensor(struct pp_hwmgr *hwmgr, int idx, *size = 4; return 0; case AMDGPU_PP_SENSOR_GPU_POWER: - if (*size < sizeof(struct pp_gpu_power)) - return -EINVAL; - *size = sizeof(struct pp_gpu_power); - return smu7_get_gpu_power(hwmgr, (struct pp_gpu_power *)value); + return smu7_get_gpu_power(hwmgr, (uint32_t *)value); case AMDGPU_PP_SENSOR_VDDGFX: if ((data->vr_config & 0xff) == 0x2) val_vid = PHM_READ_INDIRECT_FIELD(hwmgr->device, @@ -3480,7 +3520,6 @@ static int smu7_find_dpm_states_clocks_in_dpm_table(struct pp_hwmgr *hwmgr, cons [smu7_ps->performance_level_count - 1].memory_clock; struct PP_Clocks min_clocks = {0}; uint32_t i; - struct cgs_display_info info = {0}; for (i = 0; i < sclk_table->count; i++) { if (sclk == sclk_table->dpm_levels[i].value) @@ -3507,9 +3546,8 @@ static int smu7_find_dpm_states_clocks_in_dpm_table(struct pp_hwmgr *hwmgr, cons if (i >= mclk_table->count) data->need_update_smu7_dpm_table |= DPMTABLE_OD_UPDATE_MCLK; - cgs_get_active_displays_info(hwmgr->device, &info); - if (data->display_timing.num_existing_displays != info.display_count) + if (data->display_timing.num_existing_displays != hwmgr->display_config->num_display) data->need_update_smu7_dpm_table |= DPMTABLE_UPDATE_MCLK; return 0; @@ -3812,9 +3850,14 @@ static int smu7_notify_smc_display(struct pp_hwmgr *hwmgr) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); - if (hwmgr->feature_mask & PP_VBI_TIME_SUPPORT_MASK) - smum_send_msg_to_smc_with_parameter(hwmgr, - (PPSMC_Msg)PPSMC_MSG_SetVBITimeout, data->frame_time_x2); + if (hwmgr->feature_mask & PP_VBI_TIME_SUPPORT_MASK) { + if (hwmgr->chip_id == CHIP_VEGAM) + smum_send_msg_to_smc_with_parameter(hwmgr, + (PPSMC_Msg)PPSMC_MSG_SetVBITimeout_VEGAM, data->frame_time_x2); + else + smum_send_msg_to_smc_with_parameter(hwmgr, + (PPSMC_Msg)PPSMC_MSG_SetVBITimeout, data->frame_time_x2); + } return (smum_send_msg_to_smc(hwmgr, (PPSMC_Msg)PPSMC_HasDisplay) == 0) ? 0 : -EINVAL; } @@ -3908,15 +3951,8 @@ smu7_notify_smc_display_change(struct pp_hwmgr *hwmgr, bool has_display) static int smu7_notify_smc_display_config_after_ps_adjustment(struct pp_hwmgr *hwmgr) { - uint32_t num_active_displays = 0; - struct cgs_display_info info = {0}; - - info.mode_info = NULL; - cgs_get_active_displays_info(hwmgr->device, &info); - - num_active_displays = info.display_count; - - if (num_active_displays > 1 && hwmgr->display_config.multi_monitor_in_sync != true) + if (hwmgr->display_config->num_display > 1 && + !hwmgr->display_config->multi_monitor_in_sync) smu7_notify_smc_display_change(hwmgr, false); return 0; @@ -3931,33 +3967,24 @@ smu7_notify_smc_display_config_after_ps_adjustment(struct pp_hwmgr *hwmgr) static int smu7_program_display_gap(struct pp_hwmgr *hwmgr) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); - uint32_t num_active_displays = 0; uint32_t display_gap = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_DISPLAY_GAP_CNTL); uint32_t display_gap2; uint32_t pre_vbi_time_in_us; uint32_t frame_time_in_us; - uint32_t ref_clock; - uint32_t refresh_rate = 0; - struct cgs_display_info info = {0}; - struct cgs_mode_info mode_info = {0}; + uint32_t ref_clock, refresh_rate; - info.mode_info = &mode_info; - cgs_get_active_displays_info(hwmgr->device, &info); - num_active_displays = info.display_count; - - display_gap = PHM_SET_FIELD(display_gap, CG_DISPLAY_GAP_CNTL, DISP_GAP, (num_active_displays > 0) ? DISPLAY_GAP_VBLANK_OR_WM : DISPLAY_GAP_IGNORE); + display_gap = PHM_SET_FIELD(display_gap, CG_DISPLAY_GAP_CNTL, DISP_GAP, (hwmgr->display_config->num_display > 0) ? DISPLAY_GAP_VBLANK_OR_WM : DISPLAY_GAP_IGNORE); cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixCG_DISPLAY_GAP_CNTL, display_gap); ref_clock = amdgpu_asic_get_xclk((struct amdgpu_device *)hwmgr->adev); - - refresh_rate = mode_info.refresh_rate; + refresh_rate = hwmgr->display_config->vrefresh; if (0 == refresh_rate) refresh_rate = 60; frame_time_in_us = 1000000 / refresh_rate; - pre_vbi_time_in_us = frame_time_in_us - 200 - mode_info.vblank_time_us; + pre_vbi_time_in_us = frame_time_in_us - 200 - hwmgr->display_config->min_vblank_time; data->frame_time_x2 = frame_time_in_us * 2 / 100; @@ -4037,17 +4064,14 @@ smu7_check_smc_update_required_for_display_configuration(struct pp_hwmgr *hwmgr) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); bool is_update_required = false; - struct cgs_display_info info = {0, 0, NULL}; - cgs_get_active_displays_info(hwmgr->device, &info); - - if (data->display_timing.num_existing_displays != info.display_count) + if (data->display_timing.num_existing_displays != hwmgr->display_config->num_display) is_update_required = true; if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_SclkDeepSleep)) { - if (data->display_timing.min_clock_in_sr != hwmgr->display_config.min_core_set_clock_in_sr && + if (data->display_timing.min_clock_in_sr != hwmgr->display_config->min_core_set_clock_in_sr && (data->display_timing.min_clock_in_sr >= SMU7_MINIMUM_ENGINE_CLOCK || - hwmgr->display_config.min_core_set_clock_in_sr >= SMU7_MINIMUM_ENGINE_CLOCK)) + hwmgr->display_config->min_core_set_clock_in_sr >= SMU7_MINIMUM_ENGINE_CLOCK)) is_update_required = true; } return is_update_required; @@ -4102,7 +4126,7 @@ static int smu7_check_states_equal(struct pp_hwmgr *hwmgr, return 0; } -static int smu7_upload_mc_firmware(struct pp_hwmgr *hwmgr) +static int smu7_check_mc_firmware(struct pp_hwmgr *hwmgr) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); @@ -4181,13 +4205,9 @@ static int smu7_read_clock_registers(struct pp_hwmgr *hwmgr) static int smu7_get_memory_type(struct pp_hwmgr *hwmgr) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); - uint32_t temp; - - temp = cgs_read_register(hwmgr->device, mmMC_SEQ_MISC0); + struct amdgpu_device *adev = hwmgr->adev; - data->is_memory_gddr5 = (MC_SEQ_MISC0_GDDR5_VALUE == - ((temp & MC_SEQ_MISC0_GDDR5_MASK) >> - MC_SEQ_MISC0_GDDR5_SHIFT)); + data->is_memory_gddr5 = (adev->gmc.vram_type == AMDGPU_VRAM_TYPE_GDDR5); return 0; } @@ -4235,7 +4255,7 @@ static int smu7_setup_asic_task(struct pp_hwmgr *hwmgr) { int tmp_result, result = 0; - smu7_upload_mc_firmware(hwmgr); + smu7_check_mc_firmware(hwmgr); tmp_result = smu7_read_clock_registers(hwmgr); PP_ASSERT_WITH_CODE((0 == tmp_result), @@ -4370,22 +4390,36 @@ static int smu7_print_clock_levels(struct pp_hwmgr *hwmgr, break; case OD_SCLK: if (hwmgr->od_enabled) { - size = sprintf(buf, "%s: \n", "OD_SCLK"); + size = sprintf(buf, "%s:\n", "OD_SCLK"); for (i = 0; i < odn_sclk_table->num_of_pl; i++) - size += sprintf(buf + size, "%d: %10uMhz %10u mV\n", - i, odn_sclk_table->entries[i].clock / 100, + size += sprintf(buf + size, "%d: %10uMHz %10umV\n", + i, odn_sclk_table->entries[i].clock/100, odn_sclk_table->entries[i].vddc); } break; case OD_MCLK: if (hwmgr->od_enabled) { - size = sprintf(buf, "%s: \n", "OD_MCLK"); + size = sprintf(buf, "%s:\n", "OD_MCLK"); for (i = 0; i < odn_mclk_table->num_of_pl; i++) - size += sprintf(buf + size, "%d: %10uMhz %10u mV\n", - i, odn_mclk_table->entries[i].clock / 100, + size += sprintf(buf + size, "%d: %10uMHz %10umV\n", + i, odn_mclk_table->entries[i].clock/100, odn_mclk_table->entries[i].vddc); } break; + case OD_RANGE: + if (hwmgr->od_enabled) { + size = sprintf(buf, "%s:\n", "OD_RANGE"); + size += sprintf(buf + size, "SCLK: %7uMHz %10uMHz\n", + data->golden_dpm_table.sclk_table.dpm_levels[0].value/100, + hwmgr->platform_descriptor.overdriveLimit.engineClock/100); + size += sprintf(buf + size, "MCLK: %7uMHz %10uMHz\n", + data->golden_dpm_table.mclk_table.dpm_levels[0].value/100, + hwmgr->platform_descriptor.overdriveLimit.memoryClock/100); + size += sprintf(buf + size, "VDDC: %7umV %11umV\n", + data->odn_dpm_table.min_vddc, + data->odn_dpm_table.max_vddc); + } + break; default: break; } @@ -4669,36 +4703,27 @@ static bool smu7_check_clk_voltage_valid(struct pp_hwmgr *hwmgr, { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); - struct phm_ppt_v1_information *table_info = - (struct phm_ppt_v1_information *)(hwmgr->pptable); - uint32_t min_vddc; - struct phm_ppt_v1_clock_voltage_dependency_table *dep_sclk_table; - - if (table_info == NULL) - return false; - - dep_sclk_table = table_info->vdd_dep_on_sclk; - min_vddc = dep_sclk_table->entries[0].vddc; - - if (voltage < min_vddc || voltage > 2000) { - pr_info("OD voltage is out of range [%d - 2000] mV\n", min_vddc); + if (voltage < data->odn_dpm_table.min_vddc || voltage > data->odn_dpm_table.max_vddc) { + pr_info("OD voltage is out of range [%d - %d] mV\n", + data->odn_dpm_table.min_vddc, + data->odn_dpm_table.max_vddc); return false; } if (type == PP_OD_EDIT_SCLK_VDDC_TABLE) { - if (data->vbios_boot_state.sclk_bootup_value > clk || + if (data->golden_dpm_table.sclk_table.dpm_levels[0].value > clk || hwmgr->platform_descriptor.overdriveLimit.engineClock < clk) { pr_info("OD engine clock is out of range [%d - %d] MHz\n", - data->vbios_boot_state.sclk_bootup_value, - hwmgr->platform_descriptor.overdriveLimit.engineClock / 100); + data->golden_dpm_table.sclk_table.dpm_levels[0].value/100, + hwmgr->platform_descriptor.overdriveLimit.engineClock/100); return false; } } else if (type == PP_OD_EDIT_MCLK_VDDC_TABLE) { - if (data->vbios_boot_state.mclk_bootup_value > clk || + if (data->golden_dpm_table.mclk_table.dpm_levels[0].value > clk || hwmgr->platform_descriptor.overdriveLimit.memoryClock < clk) { pr_info("OD memory clock is out of range [%d - %d] MHz\n", - data->vbios_boot_state.mclk_bootup_value/100, - hwmgr->platform_descriptor.overdriveLimit.memoryClock / 100); + data->golden_dpm_table.mclk_table.dpm_levels[0].value/100, + hwmgr->platform_descriptor.overdriveLimit.memoryClock/100); return false; } } else { @@ -4747,10 +4772,6 @@ static void smu7_check_dpm_table_updated(struct pp_hwmgr *hwmgr) return; } } - if (i == dep_table->count && data->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_VDDC) { - data->need_update_smu7_dpm_table &= ~DPMTABLE_OD_UPDATE_VDDC; - data->need_update_smu7_dpm_table |= DPMTABLE_OD_UPDATE_MCLK; - } dep_table = table_info->vdd_dep_on_sclk; odn_dep_table = (struct phm_ppt_v1_clock_voltage_dependency_table *)&(odn_table->vdd_dependency_on_sclk); @@ -4760,9 +4781,9 @@ static void smu7_check_dpm_table_updated(struct pp_hwmgr *hwmgr) return; } } - if (i == dep_table->count && data->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_VDDC) { + if (data->need_update_smu7_dpm_table & DPMTABLE_OD_UPDATE_VDDC) { data->need_update_smu7_dpm_table &= ~DPMTABLE_OD_UPDATE_VDDC; - data->need_update_smu7_dpm_table |= DPMTABLE_OD_UPDATE_SCLK; + data->need_update_smu7_dpm_table |= DPMTABLE_OD_UPDATE_SCLK | DPMTABLE_OD_UPDATE_MCLK; } } @@ -4864,6 +4885,17 @@ static int smu7_get_power_profile_mode(struct pp_hwmgr *hwmgr, char *buf) len = sizeof(smu7_profiling) / sizeof(struct profile_mode_setting); for (i = 0; i < len; i++) { + if (i == hwmgr->power_profile_mode) { + size += sprintf(buf + size, "%3d %14s %s: %8d %16d %16d %16d %16d %16d\n", + i, profile_name[i], "*", + data->current_profile_setting.sclk_up_hyst, + data->current_profile_setting.sclk_down_hyst, + data->current_profile_setting.sclk_activity, + data->current_profile_setting.mclk_up_hyst, + data->current_profile_setting.mclk_down_hyst, + data->current_profile_setting.mclk_activity); + continue; + } if (smu7_profiling[i].bupdate_sclk) size += sprintf(buf + size, "%3d %16s: %8d %16d %16d ", i, profile_name[i], smu7_profiling[i].sclk_up_hyst, @@ -4883,24 +4915,6 @@ static int smu7_get_power_profile_mode(struct pp_hwmgr *hwmgr, char *buf) "-", "-", "-"); } - size += sprintf(buf + size, "%3d %16s: %8d %16d %16d %16d %16d %16d\n", - i, profile_name[i], - data->custom_profile_setting.sclk_up_hyst, - data->custom_profile_setting.sclk_down_hyst, - data->custom_profile_setting.sclk_activity, - data->custom_profile_setting.mclk_up_hyst, - data->custom_profile_setting.mclk_down_hyst, - data->custom_profile_setting.mclk_activity); - - size += sprintf(buf + size, "%3s %16s: %8d %16d %16d %16d %16d %16d\n", - "*", "CURRENT", - data->current_profile_setting.sclk_up_hyst, - data->current_profile_setting.sclk_down_hyst, - data->current_profile_setting.sclk_activity, - data->current_profile_setting.mclk_up_hyst, - data->current_profile_setting.mclk_down_hyst, - data->current_profile_setting.mclk_activity); - return size; } @@ -4939,16 +4953,16 @@ static int smu7_set_power_profile_mode(struct pp_hwmgr *hwmgr, long *input, uint if (size < 8) return -EINVAL; - data->custom_profile_setting.bupdate_sclk = input[0]; - data->custom_profile_setting.sclk_up_hyst = input[1]; - data->custom_profile_setting.sclk_down_hyst = input[2]; - data->custom_profile_setting.sclk_activity = input[3]; - data->custom_profile_setting.bupdate_mclk = input[4]; - data->custom_profile_setting.mclk_up_hyst = input[5]; - data->custom_profile_setting.mclk_down_hyst = input[6]; - data->custom_profile_setting.mclk_activity = input[7]; - if (!smum_update_dpm_settings(hwmgr, &data->custom_profile_setting)) { - memcpy(&data->current_profile_setting, &data->custom_profile_setting, sizeof(struct profile_mode_setting)); + tmp.bupdate_sclk = input[0]; + tmp.sclk_up_hyst = input[1]; + tmp.sclk_down_hyst = input[2]; + tmp.sclk_activity = input[3]; + tmp.bupdate_mclk = input[4]; + tmp.mclk_up_hyst = input[5]; + tmp.mclk_down_hyst = input[6]; + tmp.mclk_activity = input[7]; + if (!smum_update_dpm_settings(hwmgr, &tmp)) { + memcpy(&data->current_profile_setting, &tmp, sizeof(struct profile_mode_setting)); hwmgr->power_profile_mode = mode; } break; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.h b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.h index f40179c9ca977..c91e75db6a8e5 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_hwmgr.h @@ -184,6 +184,8 @@ struct smu7_odn_dpm_table { struct smu7_odn_clock_voltage_dependency_table vdd_dependency_on_sclk; struct smu7_odn_clock_voltage_dependency_table vdd_dependency_on_mclk; uint32_t odn_mclk_min_limit; + uint32_t min_vddc; + uint32_t max_vddc; }; struct profile_mode_setting { @@ -325,7 +327,6 @@ struct smu7_hwmgr { uint16_t mem_latency_high; uint16_t mem_latency_low; uint32_t vr_config; - struct profile_mode_setting custom_profile_setting; struct profile_mode_setting current_profile_setting; }; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_powertune.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_powertune.c index 03bc7453f3b1e..c952845833d72 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_powertune.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu7_powertune.c @@ -623,6 +623,190 @@ static const struct gpu_pt_config_reg DIDTConfig_Polaris11_Kicker[] = { 0xFFFFFFFF } /* End of list */ }; +static const struct gpu_pt_config_reg GCCACConfig_VegaM[] = +{ +// --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- +// Offset Mask Shift Value Type +// --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- + // DIDT_SQ + // + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x00060013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x00860013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x01060013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x01860013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x02060013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x02860013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x03060013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x03860013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x04060013, GPU_CONFIGREG_GC_CAC_IND }, + + // DIDT_TD + // + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x000E0013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x008E0013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x010E0013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x018E0013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x020E0013, GPU_CONFIGREG_GC_CAC_IND }, + + // DIDT_TCP + // + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x00100013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x00900013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x01100013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x01900013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x02100013, GPU_CONFIGREG_GC_CAC_IND }, + { ixGC_CAC_CNTL, 0xFFFFFFFF, 0, 0x02900013, GPU_CONFIGREG_GC_CAC_IND }, + + { 0xFFFFFFFF } // End of list +}; + +static const struct gpu_pt_config_reg DIDTConfig_VegaM[] = +{ +// --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- +// Offset Mask Shift Value Type +// --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- + // DIDT_SQ + // + { ixDIDT_SQ_WEIGHT0_3, DIDT_SQ_WEIGHT0_3__WEIGHT0_MASK, DIDT_SQ_WEIGHT0_3__WEIGHT0__SHIFT, 0x0073, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_WEIGHT0_3, DIDT_SQ_WEIGHT0_3__WEIGHT1_MASK, DIDT_SQ_WEIGHT0_3__WEIGHT1__SHIFT, 0x00ab, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_WEIGHT0_3, DIDT_SQ_WEIGHT0_3__WEIGHT2_MASK, DIDT_SQ_WEIGHT0_3__WEIGHT2__SHIFT, 0x0084, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_WEIGHT0_3, DIDT_SQ_WEIGHT0_3__WEIGHT3_MASK, DIDT_SQ_WEIGHT0_3__WEIGHT3__SHIFT, 0x005a, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_SQ_WEIGHT4_7, DIDT_SQ_WEIGHT4_7__WEIGHT4_MASK, DIDT_SQ_WEIGHT4_7__WEIGHT4__SHIFT, 0x0067, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_WEIGHT4_7, DIDT_SQ_WEIGHT4_7__WEIGHT5_MASK, DIDT_SQ_WEIGHT4_7__WEIGHT5__SHIFT, 0x0084, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_WEIGHT4_7, DIDT_SQ_WEIGHT4_7__WEIGHT6_MASK, DIDT_SQ_WEIGHT4_7__WEIGHT6__SHIFT, 0x0027, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_WEIGHT4_7, DIDT_SQ_WEIGHT4_7__WEIGHT7_MASK, DIDT_SQ_WEIGHT4_7__WEIGHT7__SHIFT, 0x0046, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_SQ_WEIGHT8_11, DIDT_SQ_WEIGHT8_11__WEIGHT8_MASK, DIDT_SQ_WEIGHT8_11__WEIGHT8__SHIFT, 0x00aa, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_WEIGHT8_11, DIDT_SQ_WEIGHT8_11__WEIGHT9_MASK, DIDT_SQ_WEIGHT8_11__WEIGHT9__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_WEIGHT8_11, DIDT_SQ_WEIGHT8_11__WEIGHT10_MASK, DIDT_SQ_WEIGHT8_11__WEIGHT10__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_WEIGHT8_11, DIDT_SQ_WEIGHT8_11__WEIGHT11_MASK, DIDT_SQ_WEIGHT8_11__WEIGHT11__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_SQ_CTRL1, DIDT_SQ_CTRL1__MIN_POWER_MASK, DIDT_SQ_CTRL1__MIN_POWER__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL1, DIDT_SQ_CTRL1__MAX_POWER_MASK, DIDT_SQ_CTRL1__MAX_POWER__SHIFT, 0xffff, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_SQ_CTRL_OCP, DIDT_SQ_CTRL_OCP__UNUSED_0_MASK, DIDT_SQ_CTRL_OCP__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL_OCP, DIDT_SQ_CTRL_OCP__OCP_MAX_POWER_MASK, DIDT_SQ_CTRL_OCP__OCP_MAX_POWER__SHIFT, 0xffff, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_SQ_CTRL2, DIDT_SQ_CTRL2__MAX_POWER_DELTA_MASK, DIDT_SQ_CTRL2__MAX_POWER_DELTA__SHIFT, 0x3853, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL2, DIDT_SQ_CTRL2__UNUSED_0_MASK, DIDT_SQ_CTRL2__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL2, DIDT_SQ_CTRL2__SHORT_TERM_INTERVAL_SIZE_MASK, DIDT_SQ_CTRL2__SHORT_TERM_INTERVAL_SIZE__SHIFT, 0x005a, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL2, DIDT_SQ_CTRL2__UNUSED_1_MASK, DIDT_SQ_CTRL2__UNUSED_1__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL2, DIDT_SQ_CTRL2__LONG_TERM_INTERVAL_RATIO_MASK, DIDT_SQ_CTRL2__LONG_TERM_INTERVAL_RATIO__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL2, DIDT_SQ_CTRL2__UNUSED_2_MASK, DIDT_SQ_CTRL2__UNUSED_2__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_SQ_STALL_CTRL, DIDT_SQ_STALL_CTRL__DIDT_STALL_CTRL_ENABLE_MASK, DIDT_SQ_STALL_CTRL__DIDT_STALL_CTRL_ENABLE__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_STALL_CTRL, DIDT_SQ_STALL_CTRL__DIDT_STALL_DELAY_HI_MASK, DIDT_SQ_STALL_CTRL__DIDT_STALL_DELAY_HI__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_STALL_CTRL, DIDT_SQ_STALL_CTRL__DIDT_STALL_DELAY_LO_MASK, DIDT_SQ_STALL_CTRL__DIDT_STALL_DELAY_LO__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_STALL_CTRL, DIDT_SQ_STALL_CTRL__DIDT_HI_POWER_THRESHOLD_MASK, DIDT_SQ_STALL_CTRL__DIDT_HI_POWER_THRESHOLD__SHIFT, 0x0ebb, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_STALL_CTRL, DIDT_SQ_STALL_CTRL__UNUSED_0_MASK, DIDT_SQ_STALL_CTRL__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_SQ_TUNING_CTRL, DIDT_SQ_TUNING_CTRL__DIDT_TUNING_ENABLE_MASK, DIDT_SQ_TUNING_CTRL__DIDT_TUNING_ENABLE__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_TUNING_CTRL, DIDT_SQ_TUNING_CTRL__MAX_POWER_DELTA_HI_MASK, DIDT_SQ_TUNING_CTRL__MAX_POWER_DELTA_HI__SHIFT, 0x3853, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_TUNING_CTRL, DIDT_SQ_TUNING_CTRL__MAX_POWER_DELTA_LO_MASK, DIDT_SQ_TUNING_CTRL__MAX_POWER_DELTA_LO__SHIFT, 0x3153, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_TUNING_CTRL, DIDT_SQ_TUNING_CTRL__UNUSED_0_MASK, DIDT_SQ_TUNING_CTRL__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_SQ_CTRL0, DIDT_SQ_CTRL0__DIDT_CTRL_EN_MASK, DIDT_SQ_CTRL0__DIDT_CTRL_EN__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL0, DIDT_SQ_CTRL0__USE_REF_CLOCK_MASK, DIDT_SQ_CTRL0__USE_REF_CLOCK__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL0, DIDT_SQ_CTRL0__PHASE_OFFSET_MASK, DIDT_SQ_CTRL0__PHASE_OFFSET__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL0, DIDT_SQ_CTRL0__DIDT_CTRL_RST_MASK, DIDT_SQ_CTRL0__DIDT_CTRL_RST__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL0, DIDT_SQ_CTRL0__DIDT_CLK_EN_OVERRIDE_MASK, DIDT_SQ_CTRL0__DIDT_CLK_EN_OVERRIDE__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL0, DIDT_SQ_CTRL0__DIDT_MAX_STALLS_ALLOWED_HI_MASK, DIDT_SQ_CTRL0__DIDT_MAX_STALLS_ALLOWED_HI__SHIFT, 0x0010, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL0, DIDT_SQ_CTRL0__DIDT_MAX_STALLS_ALLOWED_LO_MASK, DIDT_SQ_CTRL0__DIDT_MAX_STALLS_ALLOWED_LO__SHIFT, 0x0010, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_SQ_CTRL0, DIDT_SQ_CTRL0__UNUSED_0_MASK, DIDT_SQ_CTRL0__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + // DIDT_TD + // + { ixDIDT_TD_WEIGHT0_3, DIDT_TD_WEIGHT0_3__WEIGHT0_MASK, DIDT_TD_WEIGHT0_3__WEIGHT0__SHIFT, 0x000a, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_WEIGHT0_3, DIDT_TD_WEIGHT0_3__WEIGHT1_MASK, DIDT_TD_WEIGHT0_3__WEIGHT1__SHIFT, 0x0010, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_WEIGHT0_3, DIDT_TD_WEIGHT0_3__WEIGHT2_MASK, DIDT_TD_WEIGHT0_3__WEIGHT2__SHIFT, 0x0017, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_WEIGHT0_3, DIDT_TD_WEIGHT0_3__WEIGHT3_MASK, DIDT_TD_WEIGHT0_3__WEIGHT3__SHIFT, 0x002f, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TD_WEIGHT4_7, DIDT_TD_WEIGHT4_7__WEIGHT4_MASK, DIDT_TD_WEIGHT4_7__WEIGHT4__SHIFT, 0x0046, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_WEIGHT4_7, DIDT_TD_WEIGHT4_7__WEIGHT5_MASK, DIDT_TD_WEIGHT4_7__WEIGHT5__SHIFT, 0x005d, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_WEIGHT4_7, DIDT_TD_WEIGHT4_7__WEIGHT6_MASK, DIDT_TD_WEIGHT4_7__WEIGHT6__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_WEIGHT4_7, DIDT_TD_WEIGHT4_7__WEIGHT7_MASK, DIDT_TD_WEIGHT4_7__WEIGHT7__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TD_CTRL1, DIDT_TD_CTRL1__MIN_POWER_MASK, DIDT_TD_CTRL1__MIN_POWER__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL1, DIDT_TD_CTRL1__MAX_POWER_MASK, DIDT_TD_CTRL1__MAX_POWER__SHIFT, 0xffff, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TD_CTRL_OCP, DIDT_TD_CTRL_OCP__UNUSED_0_MASK, DIDT_TD_CTRL_OCP__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL_OCP, DIDT_TD_CTRL_OCP__OCP_MAX_POWER_MASK, DIDT_TD_CTRL_OCP__OCP_MAX_POWER__SHIFT, 0x00ff, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TD_CTRL2, DIDT_TD_CTRL2__MAX_POWER_DELTA_MASK, DIDT_TD_CTRL2__MAX_POWER_DELTA__SHIFT, 0x3fff, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL2, DIDT_TD_CTRL2__UNUSED_0_MASK, DIDT_TD_CTRL2__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL2, DIDT_TD_CTRL2__SHORT_TERM_INTERVAL_SIZE_MASK, DIDT_TD_CTRL2__SHORT_TERM_INTERVAL_SIZE__SHIFT, 0x000f, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL2, DIDT_TD_CTRL2__UNUSED_1_MASK, DIDT_TD_CTRL2__UNUSED_1__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL2, DIDT_TD_CTRL2__LONG_TERM_INTERVAL_RATIO_MASK, DIDT_TD_CTRL2__LONG_TERM_INTERVAL_RATIO__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL2, DIDT_TD_CTRL2__UNUSED_2_MASK, DIDT_TD_CTRL2__UNUSED_2__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TD_STALL_CTRL, DIDT_TD_STALL_CTRL__DIDT_STALL_CTRL_ENABLE_MASK, DIDT_TD_STALL_CTRL__DIDT_STALL_CTRL_ENABLE__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_STALL_CTRL, DIDT_TD_STALL_CTRL__DIDT_STALL_DELAY_HI_MASK, DIDT_TD_STALL_CTRL__DIDT_STALL_DELAY_HI__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_STALL_CTRL, DIDT_TD_STALL_CTRL__DIDT_STALL_DELAY_LO_MASK, DIDT_TD_STALL_CTRL__DIDT_STALL_DELAY_LO__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_STALL_CTRL, DIDT_TD_STALL_CTRL__DIDT_HI_POWER_THRESHOLD_MASK, DIDT_TD_STALL_CTRL__DIDT_HI_POWER_THRESHOLD__SHIFT, 0x01aa, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_STALL_CTRL, DIDT_TD_STALL_CTRL__UNUSED_0_MASK, DIDT_TD_STALL_CTRL__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TD_TUNING_CTRL, DIDT_TD_TUNING_CTRL__DIDT_TUNING_ENABLE_MASK, DIDT_TD_TUNING_CTRL__DIDT_TUNING_ENABLE__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_TUNING_CTRL, DIDT_TD_TUNING_CTRL__MAX_POWER_DELTA_HI_MASK, DIDT_TD_TUNING_CTRL__MAX_POWER_DELTA_HI__SHIFT, 0x0dde, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_TUNING_CTRL, DIDT_TD_TUNING_CTRL__MAX_POWER_DELTA_LO_MASK, DIDT_TD_TUNING_CTRL__MAX_POWER_DELTA_LO__SHIFT, 0x0dde, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_TUNING_CTRL, DIDT_TD_TUNING_CTRL__UNUSED_0_MASK, DIDT_TD_TUNING_CTRL__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TD_CTRL0, DIDT_TD_CTRL0__DIDT_CTRL_EN_MASK, DIDT_TD_CTRL0__DIDT_CTRL_EN__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL0, DIDT_TD_CTRL0__USE_REF_CLOCK_MASK, DIDT_TD_CTRL0__USE_REF_CLOCK__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL0, DIDT_TD_CTRL0__PHASE_OFFSET_MASK, DIDT_TD_CTRL0__PHASE_OFFSET__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL0, DIDT_TD_CTRL0__DIDT_CTRL_RST_MASK, DIDT_TD_CTRL0__DIDT_CTRL_RST__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL0, DIDT_TD_CTRL0__DIDT_CLK_EN_OVERRIDE_MASK, DIDT_TD_CTRL0__DIDT_CLK_EN_OVERRIDE__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL0, DIDT_TD_CTRL0__DIDT_MAX_STALLS_ALLOWED_HI_MASK, DIDT_TD_CTRL0__DIDT_MAX_STALLS_ALLOWED_HI__SHIFT, 0x0009, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL0, DIDT_TD_CTRL0__DIDT_MAX_STALLS_ALLOWED_LO_MASK, DIDT_TD_CTRL0__DIDT_MAX_STALLS_ALLOWED_LO__SHIFT, 0x0009, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TD_CTRL0, DIDT_TD_CTRL0__UNUSED_0_MASK, DIDT_TD_CTRL0__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + // DIDT_TCP + // + { ixDIDT_TCP_WEIGHT0_3, DIDT_TCP_WEIGHT0_3__WEIGHT0_MASK, DIDT_TCP_WEIGHT0_3__WEIGHT0__SHIFT, 0x0004, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_WEIGHT0_3, DIDT_TCP_WEIGHT0_3__WEIGHT1_MASK, DIDT_TCP_WEIGHT0_3__WEIGHT1__SHIFT, 0x0037, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_WEIGHT0_3, DIDT_TCP_WEIGHT0_3__WEIGHT2_MASK, DIDT_TCP_WEIGHT0_3__WEIGHT2__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_WEIGHT0_3, DIDT_TCP_WEIGHT0_3__WEIGHT3_MASK, DIDT_TCP_WEIGHT0_3__WEIGHT3__SHIFT, 0x00ff, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TCP_WEIGHT4_7, DIDT_TCP_WEIGHT4_7__WEIGHT4_MASK, DIDT_TCP_WEIGHT4_7__WEIGHT4__SHIFT, 0x0054, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_WEIGHT4_7, DIDT_TCP_WEIGHT4_7__WEIGHT5_MASK, DIDT_TCP_WEIGHT4_7__WEIGHT5__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_WEIGHT4_7, DIDT_TCP_WEIGHT4_7__WEIGHT6_MASK, DIDT_TCP_WEIGHT4_7__WEIGHT6__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_WEIGHT4_7, DIDT_TCP_WEIGHT4_7__WEIGHT7_MASK, DIDT_TCP_WEIGHT4_7__WEIGHT7__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TCP_CTRL1, DIDT_TCP_CTRL1__MIN_POWER_MASK, DIDT_TCP_CTRL1__MIN_POWER__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL1, DIDT_TCP_CTRL1__MAX_POWER_MASK, DIDT_TCP_CTRL1__MAX_POWER__SHIFT, 0xffff, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TCP_CTRL_OCP, DIDT_TCP_CTRL_OCP__UNUSED_0_MASK, DIDT_TCP_CTRL_OCP__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL_OCP, DIDT_TCP_CTRL_OCP__OCP_MAX_POWER_MASK, DIDT_TCP_CTRL_OCP__OCP_MAX_POWER__SHIFT, 0xffff, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TCP_CTRL2, DIDT_TCP_CTRL2__MAX_POWER_DELTA_MASK, DIDT_TCP_CTRL2__MAX_POWER_DELTA__SHIFT, 0x3dde, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL2, DIDT_TCP_CTRL2__UNUSED_0_MASK, DIDT_TCP_CTRL2__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL2, DIDT_TCP_CTRL2__SHORT_TERM_INTERVAL_SIZE_MASK, DIDT_TCP_CTRL2__SHORT_TERM_INTERVAL_SIZE__SHIFT, 0x0032, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL2, DIDT_TCP_CTRL2__UNUSED_1_MASK, DIDT_TCP_CTRL2__UNUSED_1__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL2, DIDT_TCP_CTRL2__LONG_TERM_INTERVAL_RATIO_MASK, DIDT_TCP_CTRL2__LONG_TERM_INTERVAL_RATIO__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL2, DIDT_TCP_CTRL2__UNUSED_2_MASK, DIDT_TCP_CTRL2__UNUSED_2__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TCP_STALL_CTRL, DIDT_TCP_STALL_CTRL__DIDT_STALL_CTRL_ENABLE_MASK, DIDT_TCP_STALL_CTRL__DIDT_STALL_CTRL_ENABLE__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_STALL_CTRL, DIDT_TCP_STALL_CTRL__DIDT_STALL_DELAY_HI_MASK, DIDT_TCP_STALL_CTRL__DIDT_STALL_DELAY_HI__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_STALL_CTRL, DIDT_TCP_STALL_CTRL__DIDT_STALL_DELAY_LO_MASK, DIDT_TCP_STALL_CTRL__DIDT_STALL_DELAY_LO__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_STALL_CTRL, DIDT_TCP_STALL_CTRL__DIDT_HI_POWER_THRESHOLD_MASK, DIDT_TCP_STALL_CTRL__DIDT_HI_POWER_THRESHOLD__SHIFT,0x01aa, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_STALL_CTRL, DIDT_TCP_STALL_CTRL__UNUSED_0_MASK, DIDT_TCP_STALL_CTRL__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TCP_TUNING_CTRL, DIDT_TCP_TUNING_CTRL__DIDT_TUNING_ENABLE_MASK, DIDT_TCP_TUNING_CTRL__DIDT_TUNING_ENABLE__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_TUNING_CTRL, DIDT_TCP_TUNING_CTRL__MAX_POWER_DELTA_HI_MASK, DIDT_TCP_TUNING_CTRL__MAX_POWER_DELTA_HI__SHIFT, 0x3dde, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_TUNING_CTRL, DIDT_TCP_TUNING_CTRL__MAX_POWER_DELTA_LO_MASK, DIDT_TCP_TUNING_CTRL__MAX_POWER_DELTA_LO__SHIFT, 0x3dde, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_TUNING_CTRL, DIDT_TCP_TUNING_CTRL__UNUSED_0_MASK, DIDT_TCP_TUNING_CTRL__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { ixDIDT_TCP_CTRL0, DIDT_TCP_CTRL0__DIDT_CTRL_EN_MASK, DIDT_TCP_CTRL0__DIDT_CTRL_EN__SHIFT, 0x0001, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL0, DIDT_TCP_CTRL0__USE_REF_CLOCK_MASK, DIDT_TCP_CTRL0__USE_REF_CLOCK__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL0, DIDT_TCP_CTRL0__PHASE_OFFSET_MASK, DIDT_TCP_CTRL0__PHASE_OFFSET__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL0, DIDT_TCP_CTRL0__DIDT_CTRL_RST_MASK, DIDT_TCP_CTRL0__DIDT_CTRL_RST__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL0, DIDT_TCP_CTRL0__DIDT_CLK_EN_OVERRIDE_MASK, DIDT_TCP_CTRL0__DIDT_CLK_EN_OVERRIDE__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL0, DIDT_TCP_CTRL0__DIDT_MAX_STALLS_ALLOWED_HI_MASK, DIDT_TCP_CTRL0__DIDT_MAX_STALLS_ALLOWED_HI__SHIFT, 0x0010, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL0, DIDT_TCP_CTRL0__DIDT_MAX_STALLS_ALLOWED_LO_MASK, DIDT_TCP_CTRL0__DIDT_MAX_STALLS_ALLOWED_LO__SHIFT, 0x0010, GPU_CONFIGREG_DIDT_IND }, + { ixDIDT_TCP_CTRL0, DIDT_TCP_CTRL0__UNUSED_0_MASK, DIDT_TCP_CTRL0__UNUSED_0__SHIFT, 0x0000, GPU_CONFIGREG_DIDT_IND }, + + { 0xFFFFFFFF } // End of list +}; static int smu7_enable_didt(struct pp_hwmgr *hwmgr, const bool enable) { uint32_t en = enable ? 1 : 0; @@ -740,8 +924,8 @@ int smu7_enable_didt_config(struct pp_hwmgr *hwmgr) PP_CAP(PHM_PlatformCaps_TDRamping) || PP_CAP(PHM_PlatformCaps_TCPRamping)) { - cgs_enter_safe_mode(hwmgr->device, true); - cgs_lock_grbm_idx(hwmgr->device, true); + adev->gfx.rlc.funcs->enter_safe_mode(adev); + mutex_lock(&adev->grbm_idx_mutex); value = 0; value2 = cgs_read_register(hwmgr->device, mmGRBM_GFX_INDEX); for (count = 0; count < num_se; count++) { @@ -752,67 +936,80 @@ int smu7_enable_didt_config(struct pp_hwmgr *hwmgr) if (hwmgr->chip_id == CHIP_POLARIS10) { result = smu7_program_pt_config_registers(hwmgr, GCCACConfig_Polaris10); - PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", return result); + PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", goto error); result = smu7_program_pt_config_registers(hwmgr, DIDTConfig_Polaris10); - PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", return result); + PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", goto error); } else if (hwmgr->chip_id == CHIP_POLARIS11) { result = smu7_program_pt_config_registers(hwmgr, GCCACConfig_Polaris11); - PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", return result); + PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", goto error); if (hwmgr->is_kicker) result = smu7_program_pt_config_registers(hwmgr, DIDTConfig_Polaris11_Kicker); else result = smu7_program_pt_config_registers(hwmgr, DIDTConfig_Polaris11); - PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", return result); + PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", goto error); } else if (hwmgr->chip_id == CHIP_POLARIS12) { result = smu7_program_pt_config_registers(hwmgr, GCCACConfig_Polaris11); - PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", return result); + PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", goto error); result = smu7_program_pt_config_registers(hwmgr, DIDTConfig_Polaris12); - PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", return result); + PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", goto error); + } else if (hwmgr->chip_id == CHIP_VEGAM) { + result = smu7_program_pt_config_registers(hwmgr, GCCACConfig_VegaM); + PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", goto error); + result = smu7_program_pt_config_registers(hwmgr, DIDTConfig_VegaM); + PP_ASSERT_WITH_CODE((result == 0), "DIDT Config failed.", goto error); } } cgs_write_register(hwmgr->device, mmGRBM_GFX_INDEX, value2); result = smu7_enable_didt(hwmgr, true); - PP_ASSERT_WITH_CODE((result == 0), "EnableDiDt failed.", return result); + PP_ASSERT_WITH_CODE((result == 0), "EnableDiDt failed.", goto error); if (hwmgr->chip_id == CHIP_POLARIS11) { result = smum_send_msg_to_smc(hwmgr, (uint16_t)(PPSMC_MSG_EnableDpmDidt)); PP_ASSERT_WITH_CODE((0 == result), - "Failed to enable DPM DIDT.", return result); + "Failed to enable DPM DIDT.", goto error); } - cgs_lock_grbm_idx(hwmgr->device, false); - cgs_enter_safe_mode(hwmgr->device, false); + mutex_unlock(&adev->grbm_idx_mutex); + adev->gfx.rlc.funcs->exit_safe_mode(adev); } return 0; +error: + mutex_unlock(&adev->grbm_idx_mutex); + adev->gfx.rlc.funcs->exit_safe_mode(adev); + return result; } int smu7_disable_didt_config(struct pp_hwmgr *hwmgr) { int result; + struct amdgpu_device *adev = hwmgr->adev; if (PP_CAP(PHM_PlatformCaps_SQRamping) || PP_CAP(PHM_PlatformCaps_DBRamping) || PP_CAP(PHM_PlatformCaps_TDRamping) || PP_CAP(PHM_PlatformCaps_TCPRamping)) { - cgs_enter_safe_mode(hwmgr->device, true); + adev->gfx.rlc.funcs->enter_safe_mode(adev); result = smu7_enable_didt(hwmgr, false); PP_ASSERT_WITH_CODE((result == 0), "Post DIDT enable clock gating failed.", - return result); + goto error); if (hwmgr->chip_id == CHIP_POLARIS11) { result = smum_send_msg_to_smc(hwmgr, (uint16_t)(PPSMC_MSG_DisableDpmDidt)); PP_ASSERT_WITH_CODE((0 == result), - "Failed to disable DPM DIDT.", return result); + "Failed to disable DPM DIDT.", goto error); } - cgs_enter_safe_mode(hwmgr->device, false); + adev->gfx.rlc.funcs->exit_safe_mode(adev); } return 0; +error: + adev->gfx.rlc.funcs->exit_safe_mode(adev); + return result; } int smu7_enable_smc_cac(struct pp_hwmgr *hwmgr) @@ -852,12 +1049,10 @@ int smu7_set_power_limit(struct pp_hwmgr *hwmgr, uint32_t n) { struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); - n = (n & 0xff) << 8; - if (data->power_containment_features & POWERCONTAINMENT_FEATURE_PkgPwrLimit) return smum_send_msg_to_smc_with_parameter(hwmgr, - PPSMC_MSG_PkgPwrSetLimit, n); + PPSMC_MSG_PkgPwrSetLimit, n<<8); return 0; } diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu8_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu8_hwmgr.c index 7b26607c646a1..50690c72b2ea1 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu8_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu8_hwmgr.c @@ -314,8 +314,7 @@ static int smu8_get_system_info_data(struct pp_hwmgr *hwmgr) uint8_t frev, crev; uint16_t size; - info = (ATOM_INTEGRATED_SYSTEM_INFO_V1_9 *) cgs_atom_get_data_table( - hwmgr->device, + info = (ATOM_INTEGRATED_SYSTEM_INFO_V1_9 *)smu_atom_get_data_table(hwmgr->adev, GetIndexIntoMasterTable(DATA, IntegratedSystemInfo), &size, &frev, &crev); @@ -694,7 +693,7 @@ static int smu8_update_sclk_limit(struct pp_hwmgr *hwmgr) else data->sclk_dpm.soft_max_clk = table->entries[table->count - 1].clk; - clock = hwmgr->display_config.min_core_set_clock; + clock = hwmgr->display_config->min_core_set_clock; if (clock == 0) pr_debug("min_core_set_clock not set\n"); @@ -749,7 +748,7 @@ static int smu8_set_deep_sleep_sclk_threshold(struct pp_hwmgr *hwmgr) { if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_SclkDeepSleep)) { - uint32_t clks = hwmgr->display_config.min_core_set_clock_in_sr; + uint32_t clks = hwmgr->display_config->min_core_set_clock_in_sr; if (clks == 0) clks = SMU8_MIN_DEEP_SLEEP_SCLK; @@ -1041,25 +1040,21 @@ static int smu8_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, struct smu8_hwmgr *data = hwmgr->backend; struct PP_Clocks clocks = {0, 0, 0, 0}; bool force_high; - uint32_t num_of_active_displays = 0; - struct cgs_display_info info = {0}; smu8_ps->need_dfs_bypass = true; data->battery_state = (PP_StateUILabel_Battery == prequest_ps->classification.ui_label); - clocks.memoryClock = hwmgr->display_config.min_mem_set_clock != 0 ? - hwmgr->display_config.min_mem_set_clock : + clocks.memoryClock = hwmgr->display_config->min_mem_set_clock != 0 ? + hwmgr->display_config->min_mem_set_clock : data->sys_info.nbp_memory_clock[1]; - cgs_get_active_displays_info(hwmgr->device, &info); - num_of_active_displays = info.display_count; if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_StablePState)) clocks.memoryClock = hwmgr->dyn_state.max_clock_voltage_on_ac.mclk; force_high = (clocks.memoryClock > data->sys_info.nbp_memory_clock[SMU8_NUM_NBPMEMORYCLOCK - 1]) - || (num_of_active_displays >= 3); + || (hwmgr->display_config->num_display >= 3); smu8_ps->action = smu8_current_ps->action; @@ -1897,20 +1892,20 @@ static void smu8_dpm_powergate_uvd(struct pp_hwmgr *hwmgr, bool bgate) data->uvd_power_gated = bgate; if (bgate) { - cgs_set_powergating_state(hwmgr->device, + amdgpu_device_ip_set_powergating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_UVD, AMD_PG_STATE_GATE); - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_UVD, AMD_CG_STATE_GATE); smu8_dpm_update_uvd_dpm(hwmgr, true); smu8_dpm_powerdown_uvd(hwmgr); } else { smu8_dpm_powerup_uvd(hwmgr); - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_UVD, AMD_CG_STATE_UNGATE); - cgs_set_powergating_state(hwmgr->device, + amdgpu_device_ip_set_powergating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_UVD, AMD_PG_STATE_UNGATE); smu8_dpm_update_uvd_dpm(hwmgr, false); @@ -1923,12 +1918,10 @@ static void smu8_dpm_powergate_vce(struct pp_hwmgr *hwmgr, bool bgate) struct smu8_hwmgr *data = hwmgr->backend; if (bgate) { - cgs_set_powergating_state( - hwmgr->device, + amdgpu_device_ip_set_powergating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_VCE, AMD_PG_STATE_GATE); - cgs_set_clockgating_state( - hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_VCE, AMD_CG_STATE_GATE); smu8_enable_disable_vce_dpm(hwmgr, false); @@ -1937,12 +1930,10 @@ static void smu8_dpm_powergate_vce(struct pp_hwmgr *hwmgr, bool bgate) } else { smu8_dpm_powerup_vce(hwmgr); data->vce_power_gated = false; - cgs_set_clockgating_state( - hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_VCE, AMD_CG_STATE_UNGATE); - cgs_set_powergating_state( - hwmgr->device, + amdgpu_device_ip_set_powergating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_VCE, AMD_PG_STATE_UNGATE); smu8_dpm_update_vce_dpm(hwmgr); diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu_helper.c b/drivers/gpu/drm/amd/powerplay/hwmgr/smu_helper.c index 598122854ab58..93a3d022ba47a 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu_helper.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu_helper.c @@ -24,6 +24,7 @@ #include "pp_debug.h" #include "ppatomctrl.h" #include "ppsmc.h" +#include "atom.h" uint8_t convert_to_vid(uint16_t vddc) { @@ -608,3 +609,100 @@ int smu9_register_irq_handlers(struct pp_hwmgr *hwmgr) return 0; } + +void *smu_atom_get_data_table(void *dev, uint32_t table, uint16_t *size, + uint8_t *frev, uint8_t *crev) +{ + struct amdgpu_device *adev = dev; + uint16_t data_start; + + if (amdgpu_atom_parse_data_header( + adev->mode_info.atom_context, table, size, + frev, crev, &data_start)) + return (uint8_t *)adev->mode_info.atom_context->bios + + data_start; + + return NULL; +} + +int smu_get_voltage_dependency_table_ppt_v1( + const struct phm_ppt_v1_clock_voltage_dependency_table *allowed_dep_table, + struct phm_ppt_v1_clock_voltage_dependency_table *dep_table) +{ + uint8_t i = 0; + PP_ASSERT_WITH_CODE((0 != allowed_dep_table->count), + "Voltage Lookup Table empty", + return -EINVAL); + + dep_table->count = allowed_dep_table->count; + for (i=0; i<dep_table->count; i++) { + dep_table->entries[i].clk = allowed_dep_table->entries[i].clk; + dep_table->entries[i].vddInd = allowed_dep_table->entries[i].vddInd; + dep_table->entries[i].vdd_offset = allowed_dep_table->entries[i].vdd_offset; + dep_table->entries[i].vddc = allowed_dep_table->entries[i].vddc; + dep_table->entries[i].vddgfx = allowed_dep_table->entries[i].vddgfx; + dep_table->entries[i].vddci = allowed_dep_table->entries[i].vddci; + dep_table->entries[i].mvdd = allowed_dep_table->entries[i].mvdd; + dep_table->entries[i].phases = allowed_dep_table->entries[i].phases; + dep_table->entries[i].cks_enable = allowed_dep_table->entries[i].cks_enable; + dep_table->entries[i].cks_voffset = allowed_dep_table->entries[i].cks_voffset; + } + + return 0; +} + +int smu_set_watermarks_for_clocks_ranges(void *wt_table, + struct pp_wm_sets_with_clock_ranges_soc15 *wm_with_clock_ranges) +{ + uint32_t i; + struct watermarks *table = wt_table; + + if (!table || !wm_with_clock_ranges) + return -EINVAL; + + if (wm_with_clock_ranges->num_wm_sets_dmif > 4 || wm_with_clock_ranges->num_wm_sets_mcif > 4) + return -EINVAL; + + for (i = 0; i < wm_with_clock_ranges->num_wm_sets_dmif; i++) { + table->WatermarkRow[1][i].MinClock = + cpu_to_le16((uint16_t) + (wm_with_clock_ranges->wm_sets_dmif[i].wm_min_dcefclk_in_khz) / + 100); + table->WatermarkRow[1][i].MaxClock = + cpu_to_le16((uint16_t) + (wm_with_clock_ranges->wm_sets_dmif[i].wm_max_dcefclk_in_khz) / + 100); + table->WatermarkRow[1][i].MinUclk = + cpu_to_le16((uint16_t) + (wm_with_clock_ranges->wm_sets_dmif[i].wm_min_memclk_in_khz) / + 100); + table->WatermarkRow[1][i].MaxUclk = + cpu_to_le16((uint16_t) + (wm_with_clock_ranges->wm_sets_dmif[i].wm_max_memclk_in_khz) / + 100); + table->WatermarkRow[1][i].WmSetting = (uint8_t) + wm_with_clock_ranges->wm_sets_dmif[i].wm_set_id; + } + + for (i = 0; i < wm_with_clock_ranges->num_wm_sets_mcif; i++) { + table->WatermarkRow[0][i].MinClock = + cpu_to_le16((uint16_t) + (wm_with_clock_ranges->wm_sets_mcif[i].wm_min_socclk_in_khz) / + 100); + table->WatermarkRow[0][i].MaxClock = + cpu_to_le16((uint16_t) + (wm_with_clock_ranges->wm_sets_mcif[i].wm_max_socclk_in_khz) / + 100); + table->WatermarkRow[0][i].MinUclk = + cpu_to_le16((uint16_t) + (wm_with_clock_ranges->wm_sets_mcif[i].wm_min_memclk_in_khz) / + 100); + table->WatermarkRow[0][i].MaxUclk = + cpu_to_le16((uint16_t) + (wm_with_clock_ranges->wm_sets_mcif[i].wm_max_memclk_in_khz) / + 100); + table->WatermarkRow[0][i].WmSetting = (uint8_t) + wm_with_clock_ranges->wm_sets_mcif[i].wm_set_id; + } + return 0; +} diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/smu_helper.h b/drivers/gpu/drm/amd/powerplay/hwmgr/smu_helper.h index d37d16e4b613a..916cc01e76528 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/smu_helper.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/smu_helper.h @@ -26,10 +26,27 @@ struct pp_atomctrl_voltage_table; struct pp_hwmgr; struct phm_ppt_v1_voltage_lookup_table; +struct Watermarks_t; +struct pp_wm_sets_with_clock_ranges_soc15; uint8_t convert_to_vid(uint16_t vddc); uint16_t convert_to_vddc(uint8_t vid); +struct watermark_row_generic_t { + uint16_t MinClock; + uint16_t MaxClock; + uint16_t MinUclk; + uint16_t MaxUclk; + + uint8_t WmSetting; + uint8_t Padding[3]; +}; + +struct watermarks { + struct watermark_row_generic_t WatermarkRow[2][4]; + uint32_t padding[7]; +}; + extern int phm_wait_for_register_unequal(struct pp_hwmgr *hwmgr, uint32_t index, uint32_t value, uint32_t mask); @@ -82,6 +99,16 @@ int phm_irq_process(struct amdgpu_device *adev, int smu9_register_irq_handlers(struct pp_hwmgr *hwmgr); +void *smu_atom_get_data_table(void *dev, uint32_t table, uint16_t *size, + uint8_t *frev, uint8_t *crev); + +int smu_get_voltage_dependency_table_ppt_v1( + const struct phm_ppt_v1_clock_voltage_dependency_table *allowed_dep_table, + struct phm_ppt_v1_clock_voltage_dependency_table *dep_table); + +int smu_set_watermarks_for_clocks_ranges(void *wt_table, + struct pp_wm_sets_with_clock_ranges_soc15 *wm_with_clock_ranges); + #define PHM_FIELD_SHIFT(reg, field) reg##__##field##__SHIFT #define PHM_FIELD_MASK(reg, field) reg##__##field##_MASK diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c index 7cbb56ba6fab9..d156b7bb92ae7 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.c @@ -36,7 +36,7 @@ #include "smu9.h" #include "smu9_driver_if.h" #include "vega10_inc.h" -#include "pp_soc15.h" +#include "soc15_common.h" #include "pppcielanes.h" #include "vega10_hwmgr.h" #include "vega10_processpptables.h" @@ -51,10 +51,6 @@ #include "smuio/smuio_9_0_offset.h" #include "smuio/smuio_9_0_sh_mask.h" -#define VOLTAGE_SCALE 4 -#define VOLTAGE_VID_OFFSET_SCALE1 625 -#define VOLTAGE_VID_OFFSET_SCALE2 100 - #define HBM_MEMORY_CHANNEL_WIDTH 128 static const uint32_t channel_number[] = {1, 2, 0, 4, 0, 8, 0, 16, 2}; @@ -79,8 +75,6 @@ static const uint32_t channel_number[] = {1, 2, 0, 4, 0, 8, 0, 16, 2}; #define DF_CS_AON0_DramBaseAddress0__IntLvNumChan_MASK 0x000000F0L #define DF_CS_AON0_DramBaseAddress0__IntLvAddrSel_MASK 0x00000700L #define DF_CS_AON0_DramBaseAddress0__DramBaseAddr_MASK 0xFFFFF000L -static int vega10_force_clock_level(struct pp_hwmgr *hwmgr, - enum pp_clock_type type, uint32_t mask); static const ULONG PhwVega10_Magic = (ULONG)(PHM_VIslands_Magic); @@ -291,6 +285,48 @@ static int vega10_set_features_platform_caps(struct pp_hwmgr *hwmgr) return 0; } +static int vega10_odn_initial_default_setting(struct pp_hwmgr *hwmgr) +{ + struct vega10_hwmgr *data = hwmgr->backend; + struct phm_ppt_v2_information *table_info = + (struct phm_ppt_v2_information *)(hwmgr->pptable); + struct vega10_odn_dpm_table *odn_table = &(data->odn_dpm_table); + struct vega10_odn_vddc_lookup_table *od_lookup_table; + struct phm_ppt_v1_voltage_lookup_table *vddc_lookup_table; + struct phm_ppt_v1_clock_voltage_dependency_table *dep_table[3]; + struct phm_ppt_v1_clock_voltage_dependency_table *od_table[3]; + uint32_t i; + + od_lookup_table = &odn_table->vddc_lookup_table; + vddc_lookup_table = table_info->vddc_lookup_table; + + for (i = 0; i < vddc_lookup_table->count; i++) + od_lookup_table->entries[i].us_vdd = vddc_lookup_table->entries[i].us_vdd; + + od_lookup_table->count = vddc_lookup_table->count; + + dep_table[0] = table_info->vdd_dep_on_sclk; + dep_table[1] = table_info->vdd_dep_on_mclk; + dep_table[2] = table_info->vdd_dep_on_socclk; + od_table[0] = (struct phm_ppt_v1_clock_voltage_dependency_table *)&odn_table->vdd_dep_on_sclk; + od_table[1] = (struct phm_ppt_v1_clock_voltage_dependency_table *)&odn_table->vdd_dep_on_mclk; + od_table[2] = (struct phm_ppt_v1_clock_voltage_dependency_table *)&odn_table->vdd_dep_on_socclk; + + for (i = 0; i < 3; i++) + smu_get_voltage_dependency_table_ppt_v1(dep_table[i], od_table[i]); + + if (odn_table->max_vddc == 0 || odn_table->max_vddc > 2000) + odn_table->max_vddc = dep_table[0]->entries[dep_table[0]->count - 1].vddc; + if (odn_table->min_vddc == 0 || odn_table->min_vddc > 2000) + odn_table->min_vddc = dep_table[0]->entries[0].vddc; + + i = od_table[2]->count - 1; + od_table[2]->entries[i].clk = hwmgr->platform_descriptor.overdriveLimit.memoryClock; + od_table[2]->entries[i].vddc = odn_table->max_vddc; + + return 0; +} + static void vega10_init_dpm_defaults(struct pp_hwmgr *hwmgr) { struct vega10_hwmgr *data = hwmgr->backend; @@ -427,7 +463,6 @@ static void vega10_init_dpm_defaults(struct pp_hwmgr *hwmgr) /* ACG firmware has major version 5 */ if ((hwmgr->smu_version & 0xff000000) == 0x5000000) data->smu_features[GNLD_ACG].supported = true; - if (data->registry_data.didt_support) data->smu_features[GNLD_DIDT].supported = true; @@ -754,7 +789,6 @@ static int vega10_hwmgr_backend_init(struct pp_hwmgr *hwmgr) uint32_t config_telemetry = 0; struct pp_atomfwctrl_voltage_table vol_table; struct amdgpu_device *adev = hwmgr->adev; - uint32_t reg; data = kzalloc(sizeof(struct vega10_hwmgr), GFP_KERNEL); if (data == NULL) @@ -860,10 +894,7 @@ static int vega10_hwmgr_backend_init(struct pp_hwmgr *hwmgr) advanceFanControlParameters.usFanPWMMinLimit * hwmgr->thermal_controller.fanInfo.ulMaxRPM / 100; - reg = soc15_get_register_offset(DF_HWID, 0, - mmDF_CS_AON0_DramBaseAddress0_BASE_IDX, - mmDF_CS_AON0_DramBaseAddress0); - data->mem_channels = (cgs_read_register(hwmgr->device, reg) & + data->mem_channels = (RREG32_SOC15(DF, 0, mmDF_CS_AON0_DramBaseAddress0) & DF_CS_AON0_DramBaseAddress0__IntLvNumChan_MASK) >> DF_CS_AON0_DramBaseAddress0__IntLvNumChan__SHIFT; PP_ASSERT_WITH_CODE(data->mem_channels < ARRAY_SIZE(channel_number), @@ -1370,48 +1401,6 @@ static int vega10_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) memcpy(&(data->golden_dpm_table), &(data->dpm_table), sizeof(struct vega10_dpm_table)); - if (PP_CAP(PHM_PlatformCaps_ODNinACSupport) || - PP_CAP(PHM_PlatformCaps_ODNinDCSupport)) { - data->odn_dpm_table.odn_core_clock_dpm_levels.num_of_pl = - data->dpm_table.gfx_table.count; - for (i = 0; i < data->dpm_table.gfx_table.count; i++) { - data->odn_dpm_table.odn_core_clock_dpm_levels.entries[i].clock = - data->dpm_table.gfx_table.dpm_levels[i].value; - data->odn_dpm_table.odn_core_clock_dpm_levels.entries[i].enabled = true; - } - - data->odn_dpm_table.vdd_dependency_on_sclk.count = - dep_gfx_table->count; - for (i = 0; i < dep_gfx_table->count; i++) { - data->odn_dpm_table.vdd_dependency_on_sclk.entries[i].clk = - dep_gfx_table->entries[i].clk; - data->odn_dpm_table.vdd_dependency_on_sclk.entries[i].vddInd = - dep_gfx_table->entries[i].vddInd; - data->odn_dpm_table.vdd_dependency_on_sclk.entries[i].cks_enable = - dep_gfx_table->entries[i].cks_enable; - data->odn_dpm_table.vdd_dependency_on_sclk.entries[i].cks_voffset = - dep_gfx_table->entries[i].cks_voffset; - } - - data->odn_dpm_table.odn_memory_clock_dpm_levels.num_of_pl = - data->dpm_table.mem_table.count; - for (i = 0; i < data->dpm_table.mem_table.count; i++) { - data->odn_dpm_table.odn_memory_clock_dpm_levels.entries[i].clock = - data->dpm_table.mem_table.dpm_levels[i].value; - data->odn_dpm_table.odn_memory_clock_dpm_levels.entries[i].enabled = true; - } - - data->odn_dpm_table.vdd_dependency_on_mclk.count = dep_mclk_table->count; - for (i = 0; i < dep_mclk_table->count; i++) { - data->odn_dpm_table.vdd_dependency_on_mclk.entries[i].clk = - dep_mclk_table->entries[i].clk; - data->odn_dpm_table.vdd_dependency_on_mclk.entries[i].vddInd = - dep_mclk_table->entries[i].vddInd; - data->odn_dpm_table.vdd_dependency_on_mclk.entries[i].vddci = - dep_mclk_table->entries[i].vddci; - } - } - return 0; } @@ -1514,18 +1503,18 @@ static int vega10_populate_single_gfx_level(struct pp_hwmgr *hwmgr, { struct phm_ppt_v2_information *table_info = (struct phm_ppt_v2_information *)(hwmgr->pptable); - struct phm_ppt_v1_clock_voltage_dependency_table *dep_on_sclk = - table_info->vdd_dep_on_sclk; + struct phm_ppt_v1_clock_voltage_dependency_table *dep_on_sclk; struct vega10_hwmgr *data = hwmgr->backend; struct pp_atomfwctrl_clock_dividers_soc15 dividers; uint32_t gfx_max_clock = hwmgr->platform_descriptor.overdriveLimit.engineClock; uint32_t i = 0; - if (data->apply_overdrive_next_settings_mask & - DPMTABLE_OD_UPDATE_VDDC) + if (hwmgr->od_enabled) dep_on_sclk = (struct phm_ppt_v1_clock_voltage_dependency_table *) - &(data->odn_dpm_table.vdd_dependency_on_sclk); + &(data->odn_dpm_table.vdd_dep_on_sclk); + else + dep_on_sclk = table_info->vdd_dep_on_sclk; PP_ASSERT_WITH_CODE(dep_on_sclk, "Invalid SOC_VDD-GFX_CLK Dependency Table!", @@ -1577,23 +1566,32 @@ static int vega10_populate_single_soc_level(struct pp_hwmgr *hwmgr, uint32_t soc_clock, uint8_t *current_soc_did, uint8_t *current_vol_index) { + struct vega10_hwmgr *data = hwmgr->backend; struct phm_ppt_v2_information *table_info = (struct phm_ppt_v2_information *)(hwmgr->pptable); - struct phm_ppt_v1_clock_voltage_dependency_table *dep_on_soc = - table_info->vdd_dep_on_socclk; + struct phm_ppt_v1_clock_voltage_dependency_table *dep_on_soc; struct pp_atomfwctrl_clock_dividers_soc15 dividers; uint32_t i; - PP_ASSERT_WITH_CODE(dep_on_soc, - "Invalid SOC_VDD-SOC_CLK Dependency Table!", - return -EINVAL); - for (i = 0; i < dep_on_soc->count; i++) { - if (dep_on_soc->entries[i].clk == soc_clock) - break; + if (hwmgr->od_enabled) { + dep_on_soc = (struct phm_ppt_v1_clock_voltage_dependency_table *) + &data->odn_dpm_table.vdd_dep_on_socclk; + for (i = 0; i < dep_on_soc->count; i++) { + if (dep_on_soc->entries[i].clk >= soc_clock) + break; + } + } else { + dep_on_soc = table_info->vdd_dep_on_socclk; + for (i = 0; i < dep_on_soc->count; i++) { + if (dep_on_soc->entries[i].clk == soc_clock) + break; + } } + PP_ASSERT_WITH_CODE(dep_on_soc->count > i, "Cannot find SOC_CLK in SOC_VDD-SOC_CLK Dependency Table", return -EINVAL); + PP_ASSERT_WITH_CODE(!pp_atomfwctrl_get_gpu_pll_dividers_vega10(hwmgr, COMPUTE_GPUCLK_INPUT_FLAG_DEFAULT_GPUCLK, soc_clock, ÷rs), @@ -1602,22 +1600,6 @@ static int vega10_populate_single_soc_level(struct pp_hwmgr *hwmgr, *current_soc_did = (uint8_t)dividers.ulDid; *current_vol_index = (uint8_t)(dep_on_soc->entries[i].vddInd); - - return 0; -} - -uint16_t vega10_locate_vddc_given_clock(struct pp_hwmgr *hwmgr, - uint32_t clk, - struct phm_ppt_v1_clock_voltage_dependency_table *dep_table) -{ - uint16_t i; - - for (i = 0; i < dep_table->count; i++) { - if (dep_table->entries[i].clk == clk) - return dep_table->entries[i].vddc; - } - - pr_info("[LocateVddcGivenClock] Cannot locate SOC Vddc for this clock!"); return 0; } @@ -1631,8 +1613,6 @@ static int vega10_populate_all_graphic_levels(struct pp_hwmgr *hwmgr) struct vega10_hwmgr *data = hwmgr->backend; struct phm_ppt_v2_information *table_info = (struct phm_ppt_v2_information *)(hwmgr->pptable); - struct phm_ppt_v1_clock_voltage_dependency_table *dep_table = - table_info->vdd_dep_on_socclk; PPTable_t *pp_table = &(data->smc_state_table.pp_table); struct vega10_single_dpm_table *dpm_table = &(data->dpm_table.gfx_table); int result = 0; @@ -1663,11 +1643,6 @@ static int vega10_populate_all_graphic_levels(struct pp_hwmgr *hwmgr) dpm_table = &(data->dpm_table.soc_table); for (i = 0; i < dpm_table->count; i++) { - pp_table->SocVid[i] = - (uint8_t)convert_to_vid( - vega10_locate_vddc_given_clock(hwmgr, - dpm_table->dpm_levels[i].value, - dep_table)); result = vega10_populate_single_soc_level(hwmgr, dpm_table->dpm_levels[i].value, &(pp_table->SocclkDid[i]), @@ -1678,7 +1653,6 @@ static int vega10_populate_all_graphic_levels(struct pp_hwmgr *hwmgr) j = i - 1; while (i < NUM_SOCCLK_DPM_LEVELS) { - pp_table->SocVid[i] = pp_table->SocVid[j]; result = vega10_populate_single_soc_level(hwmgr, dpm_table->dpm_levels[j].value, &(pp_table->SocclkDid[i]), @@ -1691,6 +1665,32 @@ static int vega10_populate_all_graphic_levels(struct pp_hwmgr *hwmgr) return result; } +static void vega10_populate_vddc_soc_levels(struct pp_hwmgr *hwmgr) +{ + struct vega10_hwmgr *data = hwmgr->backend; + PPTable_t *pp_table = &(data->smc_state_table.pp_table); + struct phm_ppt_v2_information *table_info = hwmgr->pptable; + struct phm_ppt_v1_voltage_lookup_table *vddc_lookup_table; + + uint8_t soc_vid = 0; + uint32_t i, max_vddc_level; + + if (hwmgr->od_enabled) + vddc_lookup_table = (struct phm_ppt_v1_voltage_lookup_table *)&data->odn_dpm_table.vddc_lookup_table; + else + vddc_lookup_table = table_info->vddc_lookup_table; + + max_vddc_level = vddc_lookup_table->count; + for (i = 0; i < max_vddc_level; i++) { + soc_vid = (uint8_t)convert_to_vid(vddc_lookup_table->entries[i].us_vdd); + pp_table->SocVid[i] = soc_vid; + } + while (i < MAX_REGULAR_DPM_NUMBER) { + pp_table->SocVid[i] = soc_vid; + i++; + } +} + /** * @brief Populates single SMC GFXCLK structure using the provided clock. * @@ -1705,25 +1705,25 @@ static int vega10_populate_single_memory_level(struct pp_hwmgr *hwmgr, struct vega10_hwmgr *data = hwmgr->backend; struct phm_ppt_v2_information *table_info = (struct phm_ppt_v2_information *)(hwmgr->pptable); - struct phm_ppt_v1_clock_voltage_dependency_table *dep_on_mclk = - table_info->vdd_dep_on_mclk; + struct phm_ppt_v1_clock_voltage_dependency_table *dep_on_mclk; struct pp_atomfwctrl_clock_dividers_soc15 dividers; uint32_t mem_max_clock = hwmgr->platform_descriptor.overdriveLimit.memoryClock; uint32_t i = 0; - if (data->apply_overdrive_next_settings_mask & - DPMTABLE_OD_UPDATE_VDDC) + if (hwmgr->od_enabled) dep_on_mclk = (struct phm_ppt_v1_clock_voltage_dependency_table *) - &data->odn_dpm_table.vdd_dependency_on_mclk; + &data->odn_dpm_table.vdd_dep_on_mclk; + else + dep_on_mclk = table_info->vdd_dep_on_mclk; PP_ASSERT_WITH_CODE(dep_on_mclk, "Invalid SOC_VDD-UCLK Dependency Table!", return -EINVAL); - if (data->need_update_dpm_table & DPMTABLE_OD_UPDATE_MCLK) + if (data->need_update_dpm_table & DPMTABLE_OD_UPDATE_MCLK) { mem_clock = mem_clock > mem_max_clock ? mem_max_clock : mem_clock; - else { + } else { for (i = 0; i < dep_on_mclk->count; i++) { if (dep_on_mclk->entries[i].clk == mem_clock) break; @@ -2067,6 +2067,9 @@ static int vega10_populate_avfs_parameters(struct pp_hwmgr *hwmgr) if (data->smu_features[GNLD_AVFS].supported) { result = pp_atomfwctrl_get_avfs_information(hwmgr, &avfs_params); if (!result) { + data->odn_dpm_table.max_vddc = avfs_params.ulMaxVddc; + data->odn_dpm_table.min_vddc = avfs_params.ulMinVddc; + pp_table->MinVoltageVid = (uint8_t) convert_to_vid((uint16_t)(avfs_params.ulMinVddc)); pp_table->MaxVoltageVid = (uint8_t) @@ -2345,6 +2348,22 @@ static int vega10_avfs_enable(struct pp_hwmgr *hwmgr, bool enable) return 0; } +static int vega10_update_avfs(struct pp_hwmgr *hwmgr) +{ + struct vega10_hwmgr *data = hwmgr->backend; + + if (data->need_update_dpm_table & DPMTABLE_OD_UPDATE_VDDC) { + vega10_avfs_enable(hwmgr, false); + } else if (data->need_update_dpm_table) { + vega10_avfs_enable(hwmgr, false); + vega10_avfs_enable(hwmgr, true); + } else { + vega10_avfs_enable(hwmgr, true); + } + + return 0; +} + static int vega10_populate_and_upload_avfs_fuse_override(struct pp_hwmgr *hwmgr) { int result = 0; @@ -2406,6 +2425,10 @@ static int vega10_init_smc_table(struct pp_hwmgr *hwmgr) "Failed to setup default DPM tables!", return result); + /* initialize ODN table */ + if (hwmgr->od_enabled) + vega10_odn_initial_default_setting(hwmgr); + pp_atomfwctrl_get_voltage_table_v4(hwmgr, VOLTAGE_TYPE_VDDC, VOLTAGE_OBJ_SVID2, &voltage_table); pp_table->MaxVidStep = voltage_table.max_vid_step; @@ -2452,6 +2475,8 @@ static int vega10_init_smc_table(struct pp_hwmgr *hwmgr) "Failed to initialize Memory Level!", return result); + vega10_populate_vddc_soc_levels(hwmgr); + result = vega10_populate_all_display_clock_levels(hwmgr); PP_ASSERT_WITH_CODE(!result, "Failed to initialize Display Level!", @@ -2481,6 +2506,12 @@ static int vega10_init_smc_table(struct pp_hwmgr *hwmgr) data->vbios_boot_state.mvddc = boot_up_values.usMvddc; data->vbios_boot_state.gfx_clock = boot_up_values.ulGfxClk; data->vbios_boot_state.mem_clock = boot_up_values.ulUClk; + pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, + SMU9_SYSPLL0_SOCCLK_ID, &boot_up_values.ulSocClk); + + pp_atomfwctrl_get_clk_information_by_clkid(hwmgr, + SMU9_SYSPLL0_DCEFCLK_ID, &boot_up_values.ulDCEFClk); + data->vbios_boot_state.soc_clock = boot_up_values.ulSocClk; data->vbios_boot_state.dcef_clock = boot_up_values.ulDCEFClk; if (0 != boot_up_values.usVddc) { @@ -2829,7 +2860,7 @@ static int vega10_enable_dpm_tasks(struct pp_hwmgr *hwmgr) tmp_result = vega10_construct_voltage_tables(hwmgr); PP_ASSERT_WITH_CODE(!tmp_result, - "Failed to contruct voltage tables!", + "Failed to construct voltage tables!", result = tmp_result); tmp_result = vega10_init_smc_table(hwmgr); @@ -3028,7 +3059,6 @@ static int vega10_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, bool disable_mclk_switching_for_frame_lock; bool disable_mclk_switching_for_vr; bool force_mclk_high; - struct cgs_display_info info = {0}; const struct phm_clock_and_voltage_limits *max_limits; uint32_t i; struct vega10_hwmgr *data = hwmgr->backend; @@ -3063,11 +3093,9 @@ static int vega10_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, } } - cgs_get_active_displays_info(hwmgr->device, &info); - /* result = PHM_CheckVBlankTime(hwmgr, &vblankTooShort);*/ - minimum_clocks.engineClock = hwmgr->display_config.min_core_set_clock; - minimum_clocks.memoryClock = hwmgr->display_config.min_mem_set_clock; + minimum_clocks.engineClock = hwmgr->display_config->min_core_set_clock; + minimum_clocks.memoryClock = hwmgr->display_config->min_mem_set_clock; if (PP_CAP(PHM_PlatformCaps_StablePState)) { stable_pstate_sclk_dpm_percentage = @@ -3107,10 +3135,10 @@ static int vega10_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, PP_CAP(PHM_PlatformCaps_DisableMclkSwitchForVR); force_mclk_high = PP_CAP(PHM_PlatformCaps_ForceMclkHigh); - if (info.display_count == 0) + if (hwmgr->display_config->num_display == 0) disable_mclk_switching = false; else - disable_mclk_switching = (info.display_count > 1) || + disable_mclk_switching = (hwmgr->display_config->num_display > 1) || disable_mclk_switching_for_frame_lock || disable_mclk_switching_for_vr || force_mclk_high; @@ -3171,87 +3199,11 @@ static int vega10_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, static int vega10_find_dpm_states_clocks_in_dpm_table(struct pp_hwmgr *hwmgr, const void *input) { - const struct phm_set_power_state_input *states = - (const struct phm_set_power_state_input *)input; - const struct vega10_power_state *vega10_ps = - cast_const_phw_vega10_power_state(states->pnew_state); struct vega10_hwmgr *data = hwmgr->backend; - struct vega10_single_dpm_table *sclk_table = - &(data->dpm_table.gfx_table); - uint32_t sclk = vega10_ps->performance_levels - [vega10_ps->performance_level_count - 1].gfx_clock; - struct vega10_single_dpm_table *mclk_table = - &(data->dpm_table.mem_table); - uint32_t mclk = vega10_ps->performance_levels - [vega10_ps->performance_level_count - 1].mem_clock; - struct PP_Clocks min_clocks = {0}; - uint32_t i; - struct cgs_display_info info = {0}; - - data->need_update_dpm_table = 0; - - if (PP_CAP(PHM_PlatformCaps_ODNinACSupport) || - PP_CAP(PHM_PlatformCaps_ODNinDCSupport)) { - for (i = 0; i < sclk_table->count; i++) { - if (sclk == sclk_table->dpm_levels[i].value) - break; - } - - if (!(data->apply_overdrive_next_settings_mask & - DPMTABLE_OD_UPDATE_SCLK) && i >= sclk_table->count) { - /* Check SCLK in DAL's minimum clocks - * in case DeepSleep divider update is required. - */ - if (data->display_timing.min_clock_in_sr != - min_clocks.engineClockInSR && - (min_clocks.engineClockInSR >= - VEGA10_MINIMUM_ENGINE_CLOCK || - data->display_timing.min_clock_in_sr >= - VEGA10_MINIMUM_ENGINE_CLOCK)) - data->need_update_dpm_table |= DPMTABLE_UPDATE_SCLK; - } - - cgs_get_active_displays_info(hwmgr->device, &info); - - if (data->display_timing.num_existing_displays != - info.display_count) - data->need_update_dpm_table |= DPMTABLE_UPDATE_MCLK; - } else { - for (i = 0; i < sclk_table->count; i++) { - if (sclk == sclk_table->dpm_levels[i].value) - break; - } - - if (i >= sclk_table->count) - data->need_update_dpm_table |= DPMTABLE_OD_UPDATE_SCLK; - else { - /* Check SCLK in DAL's minimum clocks - * in case DeepSleep divider update is required. - */ - if (data->display_timing.min_clock_in_sr != - min_clocks.engineClockInSR && - (min_clocks.engineClockInSR >= - VEGA10_MINIMUM_ENGINE_CLOCK || - data->display_timing.min_clock_in_sr >= - VEGA10_MINIMUM_ENGINE_CLOCK)) - data->need_update_dpm_table |= DPMTABLE_UPDATE_SCLK; - } - for (i = 0; i < mclk_table->count; i++) { - if (mclk == mclk_table->dpm_levels[i].value) - break; - } - - cgs_get_active_displays_info(hwmgr->device, &info); + if (data->display_timing.num_existing_displays != hwmgr->display_config->num_display) + data->need_update_dpm_table |= DPMTABLE_UPDATE_MCLK; - if (i >= mclk_table->count) - data->need_update_dpm_table |= DPMTABLE_OD_UPDATE_MCLK; - - if (data->display_timing.num_existing_displays != - info.display_count || - i >= mclk_table->count) - data->need_update_dpm_table |= DPMTABLE_UPDATE_MCLK; - } return 0; } @@ -3259,194 +3211,29 @@ static int vega10_populate_and_upload_sclk_mclk_dpm_levels( struct pp_hwmgr *hwmgr, const void *input) { int result = 0; - const struct phm_set_power_state_input *states = - (const struct phm_set_power_state_input *)input; - const struct vega10_power_state *vega10_ps = - cast_const_phw_vega10_power_state(states->pnew_state); struct vega10_hwmgr *data = hwmgr->backend; - uint32_t sclk = vega10_ps->performance_levels - [vega10_ps->performance_level_count - 1].gfx_clock; - uint32_t mclk = vega10_ps->performance_levels - [vega10_ps->performance_level_count - 1].mem_clock; - struct vega10_dpm_table *dpm_table = &data->dpm_table; - struct vega10_dpm_table *golden_dpm_table = - &data->golden_dpm_table; - uint32_t dpm_count, clock_percent; - uint32_t i; - if (PP_CAP(PHM_PlatformCaps_ODNinACSupport) || - PP_CAP(PHM_PlatformCaps_ODNinDCSupport)) { - - if (!data->need_update_dpm_table && - !data->apply_optimized_settings && - !data->apply_overdrive_next_settings_mask) - return 0; - - if (data->apply_overdrive_next_settings_mask & - DPMTABLE_OD_UPDATE_SCLK) { - for (dpm_count = 0; - dpm_count < dpm_table->gfx_table.count; - dpm_count++) { - dpm_table->gfx_table.dpm_levels[dpm_count].enabled = - data->odn_dpm_table.odn_core_clock_dpm_levels.entries[dpm_count].enabled; - dpm_table->gfx_table.dpm_levels[dpm_count].value = - data->odn_dpm_table.odn_core_clock_dpm_levels.entries[dpm_count].clock; - } - } - - if (data->apply_overdrive_next_settings_mask & - DPMTABLE_OD_UPDATE_MCLK) { - for (dpm_count = 0; - dpm_count < dpm_table->mem_table.count; - dpm_count++) { - dpm_table->mem_table.dpm_levels[dpm_count].enabled = - data->odn_dpm_table.odn_memory_clock_dpm_levels.entries[dpm_count].enabled; - dpm_table->mem_table.dpm_levels[dpm_count].value = - data->odn_dpm_table.odn_memory_clock_dpm_levels.entries[dpm_count].clock; - } - } - - if ((data->need_update_dpm_table & DPMTABLE_UPDATE_SCLK) || - data->apply_optimized_settings || - (data->apply_overdrive_next_settings_mask & - DPMTABLE_OD_UPDATE_SCLK)) { - result = vega10_populate_all_graphic_levels(hwmgr); - PP_ASSERT_WITH_CODE(!result, - "Failed to populate SCLK during PopulateNewDPMClocksStates Function!", - return result); - } - - if ((data->need_update_dpm_table & DPMTABLE_UPDATE_MCLK) || - (data->apply_overdrive_next_settings_mask & - DPMTABLE_OD_UPDATE_MCLK)){ - result = vega10_populate_all_memory_levels(hwmgr); - PP_ASSERT_WITH_CODE(!result, - "Failed to populate MCLK during PopulateNewDPMClocksStates Function!", - return result); - } - } else { - if (!data->need_update_dpm_table && - !data->apply_optimized_settings) - return 0; - - if (data->need_update_dpm_table & DPMTABLE_OD_UPDATE_SCLK && - data->smu_features[GNLD_DPM_GFXCLK].supported) { - dpm_table-> - gfx_table.dpm_levels[dpm_table->gfx_table.count - 1]. - value = sclk; - if (hwmgr->od_enabled) { - /* Need to do calculation based on the golden DPM table - * as the Heatmap GPU Clock axis is also based on - * the default values - */ - PP_ASSERT_WITH_CODE( - golden_dpm_table->gfx_table.dpm_levels - [golden_dpm_table->gfx_table.count - 1].value, - "Divide by 0!", - return -1); - - dpm_count = dpm_table->gfx_table.count < 2 ? - 0 : dpm_table->gfx_table.count - 2; - for (i = dpm_count; i > 1; i--) { - if (sclk > golden_dpm_table->gfx_table.dpm_levels - [golden_dpm_table->gfx_table.count - 1].value) { - clock_percent = - ((sclk - golden_dpm_table->gfx_table.dpm_levels - [golden_dpm_table->gfx_table.count - 1].value) * - 100) / - golden_dpm_table->gfx_table.dpm_levels - [golden_dpm_table->gfx_table.count - 1].value; - - dpm_table->gfx_table.dpm_levels[i].value = - golden_dpm_table->gfx_table.dpm_levels[i].value + - (golden_dpm_table->gfx_table.dpm_levels[i].value * - clock_percent) / 100; - } else if (golden_dpm_table-> - gfx_table.dpm_levels[dpm_table->gfx_table.count-1].value > - sclk) { - clock_percent = - ((golden_dpm_table->gfx_table.dpm_levels - [golden_dpm_table->gfx_table.count - 1].value - - sclk) * 100) / - golden_dpm_table->gfx_table.dpm_levels - [golden_dpm_table->gfx_table.count-1].value; - - dpm_table->gfx_table.dpm_levels[i].value = - golden_dpm_table->gfx_table.dpm_levels[i].value - - (golden_dpm_table->gfx_table.dpm_levels[i].value * - clock_percent) / 100; - } else - dpm_table->gfx_table.dpm_levels[i].value = - golden_dpm_table->gfx_table.dpm_levels[i].value; - } - } - } + if (!data->need_update_dpm_table) + return 0; - if (data->need_update_dpm_table & DPMTABLE_OD_UPDATE_MCLK && - data->smu_features[GNLD_DPM_UCLK].supported) { - dpm_table-> - mem_table.dpm_levels[dpm_table->mem_table.count - 1]. - value = mclk; + if (data->need_update_dpm_table & + (DPMTABLE_OD_UPDATE_SCLK + DPMTABLE_UPDATE_SCLK + DPMTABLE_UPDATE_SOCCLK)) { + result = vega10_populate_all_graphic_levels(hwmgr); + PP_ASSERT_WITH_CODE((0 == result), + "Failed to populate SCLK during PopulateNewDPMClocksStates Function!", + return result); + } - if (hwmgr->od_enabled) { - PP_ASSERT_WITH_CODE( - golden_dpm_table->mem_table.dpm_levels - [golden_dpm_table->mem_table.count - 1].value, - "Divide by 0!", - return -1); + if (data->need_update_dpm_table & + (DPMTABLE_OD_UPDATE_MCLK + DPMTABLE_UPDATE_MCLK)) { + result = vega10_populate_all_memory_levels(hwmgr); + PP_ASSERT_WITH_CODE((0 == result), + "Failed to populate MCLK during PopulateNewDPMClocksStates Function!", + return result); + } - dpm_count = dpm_table->mem_table.count < 2 ? - 0 : dpm_table->mem_table.count - 2; - for (i = dpm_count; i > 1; i--) { - if (mclk > golden_dpm_table->mem_table.dpm_levels - [golden_dpm_table->mem_table.count-1].value) { - clock_percent = ((mclk - - golden_dpm_table->mem_table.dpm_levels - [golden_dpm_table->mem_table.count-1].value) * - 100) / - golden_dpm_table->mem_table.dpm_levels - [golden_dpm_table->mem_table.count-1].value; - - dpm_table->mem_table.dpm_levels[i].value = - golden_dpm_table->mem_table.dpm_levels[i].value + - (golden_dpm_table->mem_table.dpm_levels[i].value * - clock_percent) / 100; - } else if (golden_dpm_table->mem_table.dpm_levels - [dpm_table->mem_table.count-1].value > mclk) { - clock_percent = ((golden_dpm_table->mem_table.dpm_levels - [golden_dpm_table->mem_table.count-1].value - mclk) * - 100) / - golden_dpm_table->mem_table.dpm_levels - [golden_dpm_table->mem_table.count-1].value; - - dpm_table->mem_table.dpm_levels[i].value = - golden_dpm_table->mem_table.dpm_levels[i].value - - (golden_dpm_table->mem_table.dpm_levels[i].value * - clock_percent) / 100; - } else - dpm_table->mem_table.dpm_levels[i].value = - golden_dpm_table->mem_table.dpm_levels[i].value; - } - } - } + vega10_populate_vddc_soc_levels(hwmgr); - if ((data->need_update_dpm_table & - (DPMTABLE_OD_UPDATE_SCLK + DPMTABLE_UPDATE_SCLK)) || - data->apply_optimized_settings) { - result = vega10_populate_all_graphic_levels(hwmgr); - PP_ASSERT_WITH_CODE(!result, - "Failed to populate SCLK during PopulateNewDPMClocksStates Function!", - return result); - } - - if (data->need_update_dpm_table & - (DPMTABLE_OD_UPDATE_MCLK + DPMTABLE_UPDATE_MCLK)) { - result = vega10_populate_all_memory_levels(hwmgr); - PP_ASSERT_WITH_CODE(!result, - "Failed to populate MCLK during PopulateNewDPMClocksStates Function!", - return result); - } - } return result; } @@ -3742,8 +3529,9 @@ static int vega10_set_power_state_tasks(struct pp_hwmgr *hwmgr, PP_ASSERT_WITH_CODE(!result, "Failed to upload PPtable!", return result); - data->apply_optimized_settings = false; - data->apply_overdrive_next_settings_mask = 0; + vega10_update_avfs(hwmgr); + + data->need_update_dpm_table &= DPMTABLE_OD_UPDATE_VDDC; return 0; } @@ -3793,16 +3581,18 @@ static uint32_t vega10_dpm_get_mclk(struct pp_hwmgr *hwmgr, bool low) } static int vega10_get_gpu_power(struct pp_hwmgr *hwmgr, - struct pp_gpu_power *query) + uint32_t *query) { uint32_t value; + if (!query) + return -EINVAL; + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetCurrPkgPwr); value = smum_get_argument(hwmgr); - /* power value is an integer */ - memset(query, 0, sizeof *query); - query->average_gpu_power = value << 8; + /* SMC returning actual watts, keep consistent with legacy asics, low 8 bit as 8 fractional bits */ + *query = value << 8; return 0; } @@ -3810,22 +3600,18 @@ static int vega10_get_gpu_power(struct pp_hwmgr *hwmgr, static int vega10_read_sensor(struct pp_hwmgr *hwmgr, int idx, void *value, int *size) { - uint32_t sclk_idx, mclk_idx, activity_percent = 0; + struct amdgpu_device *adev = hwmgr->adev; + uint32_t sclk_mhz, mclk_idx, activity_percent = 0; struct vega10_hwmgr *data = hwmgr->backend; struct vega10_dpm_table *dpm_table = &data->dpm_table; int ret = 0; - uint32_t reg, val_vid; + uint32_t val_vid; switch (idx) { case AMDGPU_PP_SENSOR_GFX_SCLK: - smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetCurrentGfxclkIndex); - sclk_idx = smum_get_argument(hwmgr); - if (sclk_idx < dpm_table->gfx_table.count) { - *((uint32_t *)value) = dpm_table->gfx_table.dpm_levels[sclk_idx].value; - *size = 4; - } else { - ret = -EINVAL; - } + smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetAverageGfxclkActualFrequency); + sclk_mhz = smum_get_argument(hwmgr); + *((uint32_t *)value) = sclk_mhz * 100; break; case AMDGPU_PP_SENSOR_GFX_MCLK: smum_send_msg_to_smc(hwmgr, PPSMC_MSG_GetCurrentUclkIndex); @@ -3856,18 +3642,10 @@ static int vega10_read_sensor(struct pp_hwmgr *hwmgr, int idx, *size = 4; break; case AMDGPU_PP_SENSOR_GPU_POWER: - if (*size < sizeof(struct pp_gpu_power)) - ret = -EINVAL; - else { - *size = sizeof(struct pp_gpu_power); - ret = vega10_get_gpu_power(hwmgr, (struct pp_gpu_power *)value); - } + ret = vega10_get_gpu_power(hwmgr, (uint32_t *)value); break; case AMDGPU_PP_SENSOR_VDDGFX: - reg = soc15_get_register_offset(SMUIO_HWID, 0, - mmSMUSVI0_PLANE0_CURRENTVID_BASE_IDX, - mmSMUSVI0_PLANE0_CURRENTVID); - val_vid = (cgs_read_register(hwmgr->device, reg) & + val_vid = (RREG32_SOC15(SMUIO, 0, mmSMUSVI0_PLANE0_CURRENTVID) & SMUSVI0_PLANE0_CURRENTVID__CURRENT_SVI0_PLANE0_VID_MASK) >> SMUSVI0_PLANE0_CURRENTVID__CURRENT_SVI0_PLANE0_VID__SHIFT; *((uint32_t *)value) = (uint32_t)convert_to_vddc((uint8_t)val_vid); @@ -3956,26 +3734,18 @@ static int vega10_notify_smc_display_config_after_ps_adjustment( (struct phm_ppt_v2_information *)hwmgr->pptable; struct phm_ppt_v1_clock_voltage_dependency_table *mclk_table = table_info->vdd_dep_on_mclk; uint32_t idx; - uint32_t num_active_disps = 0; - struct cgs_display_info info = {0}; struct PP_Clocks min_clocks = {0}; uint32_t i; struct pp_display_clock_request clock_req; - info.mode_info = NULL; - - cgs_get_active_displays_info(hwmgr->device, &info); - - num_active_disps = info.display_count; - - if (num_active_disps > 1) + if (hwmgr->display_config->num_display > 1) vega10_notify_smc_display_change(hwmgr, false); else vega10_notify_smc_display_change(hwmgr, true); - min_clocks.dcefClock = hwmgr->display_config.min_dcef_set_clk; - min_clocks.dcefClockInSR = hwmgr->display_config.min_dcef_deep_sleep_set_clk; - min_clocks.memoryClock = hwmgr->display_config.min_mem_set_clock; + min_clocks.dcefClock = hwmgr->display_config->min_dcef_set_clk; + min_clocks.dcefClockInSR = hwmgr->display_config->min_dcef_deep_sleep_set_clk; + min_clocks.memoryClock = hwmgr->display_config->min_mem_set_clock; for (i = 0; i < dpm_table->count; i++) { if (dpm_table->dpm_levels[i].value == min_clocks.dcefClock) @@ -4120,6 +3890,47 @@ static void vega10_set_fan_control_mode(struct pp_hwmgr *hwmgr, uint32_t mode) } } +static int vega10_force_clock_level(struct pp_hwmgr *hwmgr, + enum pp_clock_type type, uint32_t mask) +{ + struct vega10_hwmgr *data = hwmgr->backend; + + switch (type) { + case PP_SCLK: + data->smc_state_table.gfx_boot_level = mask ? (ffs(mask) - 1) : 0; + data->smc_state_table.gfx_max_level = mask ? (fls(mask) - 1) : 0; + + PP_ASSERT_WITH_CODE(!vega10_upload_dpm_bootup_level(hwmgr), + "Failed to upload boot level to lowest!", + return -EINVAL); + + PP_ASSERT_WITH_CODE(!vega10_upload_dpm_max_level(hwmgr), + "Failed to upload dpm max level to highest!", + return -EINVAL); + break; + + case PP_MCLK: + data->smc_state_table.mem_boot_level = mask ? (ffs(mask) - 1) : 0; + data->smc_state_table.mem_max_level = mask ? (fls(mask) - 1) : 0; + + PP_ASSERT_WITH_CODE(!vega10_upload_dpm_bootup_level(hwmgr), + "Failed to upload boot level to lowest!", + return -EINVAL); + + PP_ASSERT_WITH_CODE(!vega10_upload_dpm_max_level(hwmgr), + "Failed to upload dpm max level to highest!", + return -EINVAL); + + break; + + case PP_PCIE: + default: + break; + } + + return 0; +} + static int vega10_dpm_force_dpm_level(struct pp_hwmgr *hwmgr, enum amd_dpm_forced_level level) { @@ -4356,97 +4167,15 @@ static int vega10_set_watermarks_for_clocks_ranges(struct pp_hwmgr *hwmgr, struct vega10_hwmgr *data = hwmgr->backend; Watermarks_t *table = &(data->smc_state_table.water_marks_table); int result = 0; - uint32_t i; if (!data->registry_data.disable_water_mark) { - for (i = 0; i < wm_with_clock_ranges->num_wm_sets_dmif; i++) { - table->WatermarkRow[WM_DCEFCLK][i].MinClock = - cpu_to_le16((uint16_t) - (wm_with_clock_ranges->wm_sets_dmif[i].wm_min_dcefclk_in_khz) / - 100); - table->WatermarkRow[WM_DCEFCLK][i].MaxClock = - cpu_to_le16((uint16_t) - (wm_with_clock_ranges->wm_sets_dmif[i].wm_max_dcefclk_in_khz) / - 100); - table->WatermarkRow[WM_DCEFCLK][i].MinUclk = - cpu_to_le16((uint16_t) - (wm_with_clock_ranges->wm_sets_dmif[i].wm_min_memclk_in_khz) / - 100); - table->WatermarkRow[WM_DCEFCLK][i].MaxUclk = - cpu_to_le16((uint16_t) - (wm_with_clock_ranges->wm_sets_dmif[i].wm_max_memclk_in_khz) / - 100); - table->WatermarkRow[WM_DCEFCLK][i].WmSetting = (uint8_t) - wm_with_clock_ranges->wm_sets_dmif[i].wm_set_id; - } - - for (i = 0; i < wm_with_clock_ranges->num_wm_sets_mcif; i++) { - table->WatermarkRow[WM_SOCCLK][i].MinClock = - cpu_to_le16((uint16_t) - (wm_with_clock_ranges->wm_sets_mcif[i].wm_min_socclk_in_khz) / - 100); - table->WatermarkRow[WM_SOCCLK][i].MaxClock = - cpu_to_le16((uint16_t) - (wm_with_clock_ranges->wm_sets_mcif[i].wm_max_socclk_in_khz) / - 100); - table->WatermarkRow[WM_SOCCLK][i].MinUclk = - cpu_to_le16((uint16_t) - (wm_with_clock_ranges->wm_sets_mcif[i].wm_min_memclk_in_khz) / - 100); - table->WatermarkRow[WM_SOCCLK][i].MaxUclk = - cpu_to_le16((uint16_t) - (wm_with_clock_ranges->wm_sets_mcif[i].wm_max_memclk_in_khz) / - 100); - table->WatermarkRow[WM_SOCCLK][i].WmSetting = (uint8_t) - wm_with_clock_ranges->wm_sets_mcif[i].wm_set_id; - } + smu_set_watermarks_for_clocks_ranges(table, wm_with_clock_ranges); data->water_marks_bitmap = WaterMarksExist; } return result; } -static int vega10_force_clock_level(struct pp_hwmgr *hwmgr, - enum pp_clock_type type, uint32_t mask) -{ - struct vega10_hwmgr *data = hwmgr->backend; - - switch (type) { - case PP_SCLK: - data->smc_state_table.gfx_boot_level = mask ? (ffs(mask) - 1) : 0; - data->smc_state_table.gfx_max_level = mask ? (fls(mask) - 1) : 0; - - PP_ASSERT_WITH_CODE(!vega10_upload_dpm_bootup_level(hwmgr), - "Failed to upload boot level to lowest!", - return -EINVAL); - - PP_ASSERT_WITH_CODE(!vega10_upload_dpm_max_level(hwmgr), - "Failed to upload dpm max level to highest!", - return -EINVAL); - break; - - case PP_MCLK: - data->smc_state_table.mem_boot_level = mask ? (ffs(mask) - 1) : 0; - data->smc_state_table.mem_max_level = mask ? (fls(mask) - 1) : 0; - - PP_ASSERT_WITH_CODE(!vega10_upload_dpm_bootup_level(hwmgr), - "Failed to upload boot level to lowest!", - return -EINVAL); - - PP_ASSERT_WITH_CODE(!vega10_upload_dpm_max_level(hwmgr), - "Failed to upload dpm max level to highest!", - return -EINVAL); - - break; - - case PP_PCIE: - default: - break; - } - - return 0; -} - static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, enum pp_clock_type type, char *buf) { @@ -4454,6 +4183,8 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, struct vega10_single_dpm_table *sclk_table = &(data->dpm_table.gfx_table); struct vega10_single_dpm_table *mclk_table = &(data->dpm_table.mem_table); struct vega10_pcie_table *pcie_table = &(data->dpm_table.pcie_table); + struct vega10_odn_clock_voltage_dependency_table *podn_vdd_dep = NULL; + int i, now, size = 0; switch (type) { @@ -4492,6 +4223,40 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, (pcie_table->pcie_gen[i] == 2) ? "8.0GT/s, x16" : "", (i == now) ? "*" : ""); break; + case OD_SCLK: + if (hwmgr->od_enabled) { + size = sprintf(buf, "%s:\n", "OD_SCLK"); + podn_vdd_dep = &data->odn_dpm_table.vdd_dep_on_sclk; + for (i = 0; i < podn_vdd_dep->count; i++) + size += sprintf(buf + size, "%d: %10uMhz %10umV\n", + i, podn_vdd_dep->entries[i].clk / 100, + podn_vdd_dep->entries[i].vddc); + } + break; + case OD_MCLK: + if (hwmgr->od_enabled) { + size = sprintf(buf, "%s:\n", "OD_MCLK"); + podn_vdd_dep = &data->odn_dpm_table.vdd_dep_on_mclk; + for (i = 0; i < podn_vdd_dep->count; i++) + size += sprintf(buf + size, "%d: %10uMhz %10umV\n", + i, podn_vdd_dep->entries[i].clk/100, + podn_vdd_dep->entries[i].vddc); + } + break; + case OD_RANGE: + if (hwmgr->od_enabled) { + size = sprintf(buf, "%s:\n", "OD_RANGE"); + size += sprintf(buf + size, "SCLK: %7uMHz %10uMHz\n", + data->golden_dpm_table.gfx_table.dpm_levels[0].value/100, + hwmgr->platform_descriptor.overdriveLimit.engineClock/100); + size += sprintf(buf + size, "MCLK: %7uMHz %10uMHz\n", + data->golden_dpm_table.mem_table.dpm_levels[0].value/100, + hwmgr->platform_descriptor.overdriveLimit.memoryClock/100); + size += sprintf(buf + size, "VDDC: %7umV %11umV\n", + data->odn_dpm_table.min_vddc, + data->odn_dpm_table.max_vddc); + } + break; default: break; } @@ -4501,10 +4266,8 @@ static int vega10_print_clock_levels(struct pp_hwmgr *hwmgr, static int vega10_display_configuration_changed_task(struct pp_hwmgr *hwmgr) { struct vega10_hwmgr *data = hwmgr->backend; - int result = 0; - uint32_t num_turned_on_displays = 1; Watermarks_t *wm_table = &(data->smc_state_table.water_marks_table); - struct cgs_display_info info = {0}; + int result = 0; if ((data->water_marks_bitmap & WaterMarksExist) && !(data->water_marks_bitmap & WaterMarksLoaded)) { @@ -4514,10 +4277,8 @@ static int vega10_display_configuration_changed_task(struct pp_hwmgr *hwmgr) } if (data->water_marks_bitmap & WaterMarksLoaded) { - cgs_get_active_displays_info(hwmgr->device, &info); - num_turned_on_displays = info.display_count; smum_send_msg_to_smc_with_parameter(hwmgr, - PPSMC_MSG_NumOfDisplays, num_turned_on_displays); + PPSMC_MSG_NumOfDisplays, hwmgr->display_config->num_display); } return result; @@ -4603,15 +4364,12 @@ vega10_check_smc_update_required_for_display_configuration(struct pp_hwmgr *hwmg { struct vega10_hwmgr *data = hwmgr->backend; bool is_update_required = false; - struct cgs_display_info info = {0, 0, NULL}; - cgs_get_active_displays_info(hwmgr->device, &info); - - if (data->display_timing.num_existing_displays != info.display_count) + if (data->display_timing.num_existing_displays != hwmgr->display_config->num_display) is_update_required = true; if (PP_CAP(PHM_PlatformCaps_SclkDeepSleep)) { - if (data->display_timing.min_clock_in_sr != hwmgr->display_config.min_core_set_clock_in_sr) + if (data->display_timing.min_clock_in_sr != hwmgr->display_config->min_core_set_clock_in_sr) is_update_required = true; } @@ -4886,6 +4644,200 @@ static int vega10_set_power_profile_mode(struct pp_hwmgr *hwmgr, long *input, ui return 0; } + +static bool vega10_check_clk_voltage_valid(struct pp_hwmgr *hwmgr, + enum PP_OD_DPM_TABLE_COMMAND type, + uint32_t clk, + uint32_t voltage) +{ + struct vega10_hwmgr *data = hwmgr->backend; + struct vega10_odn_dpm_table *odn_table = &(data->odn_dpm_table); + struct vega10_single_dpm_table *golden_table; + + if (voltage < odn_table->min_vddc || voltage > odn_table->max_vddc) { + pr_info("OD voltage is out of range [%d - %d] mV\n", odn_table->min_vddc, odn_table->max_vddc); + return false; + } + + if (type == PP_OD_EDIT_SCLK_VDDC_TABLE) { + golden_table = &(data->golden_dpm_table.gfx_table); + if (golden_table->dpm_levels[0].value > clk || + hwmgr->platform_descriptor.overdriveLimit.engineClock < clk) { + pr_info("OD engine clock is out of range [%d - %d] MHz\n", + golden_table->dpm_levels[0].value/100, + hwmgr->platform_descriptor.overdriveLimit.engineClock/100); + return false; + } + } else if (type == PP_OD_EDIT_MCLK_VDDC_TABLE) { + golden_table = &(data->golden_dpm_table.mem_table); + if (golden_table->dpm_levels[0].value > clk || + hwmgr->platform_descriptor.overdriveLimit.memoryClock < clk) { + pr_info("OD memory clock is out of range [%d - %d] MHz\n", + golden_table->dpm_levels[0].value/100, + hwmgr->platform_descriptor.overdriveLimit.memoryClock/100); + return false; + } + } else { + return false; + } + + return true; +} + +static void vega10_check_dpm_table_updated(struct pp_hwmgr *hwmgr) +{ + struct vega10_hwmgr *data = hwmgr->backend; + struct vega10_odn_dpm_table *odn_table = &(data->odn_dpm_table); + struct phm_ppt_v2_information *table_info = hwmgr->pptable; + struct phm_ppt_v1_clock_voltage_dependency_table *dep_table; + struct phm_ppt_v1_clock_voltage_dependency_table *odn_dep_table; + uint32_t i; + + dep_table = table_info->vdd_dep_on_mclk; + odn_dep_table = (struct phm_ppt_v1_clock_voltage_dependency_table *)&(odn_table->vdd_dep_on_mclk); + + for (i = 0; i < dep_table->count; i++) { + if (dep_table->entries[i].vddc != odn_dep_table->entries[i].vddc) { + data->need_update_dpm_table |= DPMTABLE_OD_UPDATE_VDDC | DPMTABLE_OD_UPDATE_MCLK; + return; + } + } + + dep_table = table_info->vdd_dep_on_sclk; + odn_dep_table = (struct phm_ppt_v1_clock_voltage_dependency_table *)&(odn_table->vdd_dep_on_sclk); + for (i = 0; i < dep_table->count; i++) { + if (dep_table->entries[i].vddc != odn_dep_table->entries[i].vddc) { + data->need_update_dpm_table |= DPMTABLE_OD_UPDATE_VDDC | DPMTABLE_OD_UPDATE_SCLK; + return; + } + } + + if (data->need_update_dpm_table & DPMTABLE_OD_UPDATE_VDDC) { + data->need_update_dpm_table &= ~DPMTABLE_OD_UPDATE_VDDC; + data->need_update_dpm_table |= DPMTABLE_OD_UPDATE_SCLK | DPMTABLE_OD_UPDATE_MCLK; + } +} + +static void vega10_odn_update_soc_table(struct pp_hwmgr *hwmgr, + enum PP_OD_DPM_TABLE_COMMAND type) +{ + struct vega10_hwmgr *data = hwmgr->backend; + struct phm_ppt_v2_information *table_info = hwmgr->pptable; + struct phm_ppt_v1_clock_voltage_dependency_table *dep_table = table_info->vdd_dep_on_socclk; + struct vega10_single_dpm_table *dpm_table = &data->golden_dpm_table.soc_table; + + struct vega10_odn_clock_voltage_dependency_table *podn_vdd_dep_on_socclk = + &data->odn_dpm_table.vdd_dep_on_socclk; + struct vega10_odn_vddc_lookup_table *od_vddc_lookup_table = &data->odn_dpm_table.vddc_lookup_table; + + struct vega10_odn_clock_voltage_dependency_table *podn_vdd_dep; + uint8_t i, j; + + if (type == PP_OD_EDIT_SCLK_VDDC_TABLE) { + podn_vdd_dep = &data->odn_dpm_table.vdd_dep_on_sclk; + for (i = 0; i < podn_vdd_dep->count - 1; i++) + od_vddc_lookup_table->entries[i].us_vdd = podn_vdd_dep->entries[i].vddc; + if (od_vddc_lookup_table->entries[i].us_vdd < podn_vdd_dep->entries[i].vddc) + od_vddc_lookup_table->entries[i].us_vdd = podn_vdd_dep->entries[i].vddc; + } else if (type == PP_OD_EDIT_MCLK_VDDC_TABLE) { + podn_vdd_dep = &data->odn_dpm_table.vdd_dep_on_mclk; + for (i = 0; i < dpm_table->count; i++) { + for (j = 0; j < od_vddc_lookup_table->count; j++) { + if (od_vddc_lookup_table->entries[j].us_vdd > + podn_vdd_dep->entries[i].vddc) + break; + } + if (j == od_vddc_lookup_table->count) { + od_vddc_lookup_table->entries[j-1].us_vdd = + podn_vdd_dep->entries[i].vddc; + data->need_update_dpm_table |= DPMTABLE_OD_UPDATE_VDDC; + } + podn_vdd_dep->entries[i].vddInd = j; + } + dpm_table = &data->dpm_table.soc_table; + for (i = 0; i < dep_table->count; i++) { + if (dep_table->entries[i].vddInd == podn_vdd_dep->entries[dep_table->count-1].vddInd && + dep_table->entries[i].clk < podn_vdd_dep->entries[dep_table->count-1].clk) { + data->need_update_dpm_table |= DPMTABLE_UPDATE_SOCCLK; + podn_vdd_dep_on_socclk->entries[i].clk = podn_vdd_dep->entries[dep_table->count-1].clk; + dpm_table->dpm_levels[i].value = podn_vdd_dep_on_socclk->entries[i].clk; + } + } + if (podn_vdd_dep_on_socclk->entries[podn_vdd_dep_on_socclk->count - 1].clk < + podn_vdd_dep->entries[dep_table->count-1].clk) { + data->need_update_dpm_table |= DPMTABLE_UPDATE_SOCCLK; + podn_vdd_dep_on_socclk->entries[podn_vdd_dep_on_socclk->count - 1].clk = podn_vdd_dep->entries[dep_table->count-1].clk; + dpm_table->dpm_levels[podn_vdd_dep_on_socclk->count - 1].value = podn_vdd_dep->entries[dep_table->count-1].clk; + } + if (podn_vdd_dep_on_socclk->entries[podn_vdd_dep_on_socclk->count - 1].vddInd < + podn_vdd_dep->entries[dep_table->count-1].vddInd) { + data->need_update_dpm_table |= DPMTABLE_UPDATE_SOCCLK; + podn_vdd_dep_on_socclk->entries[podn_vdd_dep_on_socclk->count - 1].vddInd = podn_vdd_dep->entries[dep_table->count-1].vddInd; + } + } +} + +static int vega10_odn_edit_dpm_table(struct pp_hwmgr *hwmgr, + enum PP_OD_DPM_TABLE_COMMAND type, + long *input, uint32_t size) +{ + struct vega10_hwmgr *data = hwmgr->backend; + struct vega10_odn_clock_voltage_dependency_table *podn_vdd_dep_table; + struct vega10_single_dpm_table *dpm_table; + + uint32_t input_clk; + uint32_t input_vol; + uint32_t input_level; + uint32_t i; + + PP_ASSERT_WITH_CODE(input, "NULL user input for clock and voltage", + return -EINVAL); + + if (!hwmgr->od_enabled) { + pr_info("OverDrive feature not enabled\n"); + return -EINVAL; + } + + if (PP_OD_EDIT_SCLK_VDDC_TABLE == type) { + dpm_table = &data->dpm_table.gfx_table; + podn_vdd_dep_table = &data->odn_dpm_table.vdd_dep_on_sclk; + data->need_update_dpm_table |= DPMTABLE_OD_UPDATE_SCLK; + } else if (PP_OD_EDIT_MCLK_VDDC_TABLE == type) { + dpm_table = &data->dpm_table.mem_table; + podn_vdd_dep_table = &data->odn_dpm_table.vdd_dep_on_mclk; + data->need_update_dpm_table |= DPMTABLE_OD_UPDATE_MCLK; + } else if (PP_OD_RESTORE_DEFAULT_TABLE == type) { + memcpy(&(data->dpm_table), &(data->golden_dpm_table), sizeof(struct vega10_dpm_table)); + vega10_odn_initial_default_setting(hwmgr); + return 0; + } else if (PP_OD_COMMIT_DPM_TABLE == type) { + vega10_check_dpm_table_updated(hwmgr); + return 0; + } else { + return -EINVAL; + } + + for (i = 0; i < size; i += 3) { + if (i + 3 > size || input[i] >= podn_vdd_dep_table->count) { + pr_info("invalid clock voltage input\n"); + return 0; + } + input_level = input[i]; + input_clk = input[i+1] * 100; + input_vol = input[i+2]; + + if (vega10_check_clk_voltage_valid(hwmgr, type, input_clk, input_vol)) { + dpm_table->dpm_levels[input_level].value = input_clk; + podn_vdd_dep_table->entries[input_level].clk = input_clk; + podn_vdd_dep_table->entries[input_level].vddc = input_vol; + } else { + return -EINVAL; + } + } + vega10_odn_update_soc_table(hwmgr, type); + return 0; +} + static const struct pp_hwmgr_func vega10_hwmgr_funcs = { .backend_init = vega10_hwmgr_backend_init, .backend_fini = vega10_hwmgr_backend_fini, @@ -4944,6 +4896,7 @@ static const struct pp_hwmgr_func vega10_hwmgr_funcs = { .get_power_profile_mode = vega10_get_power_profile_mode, .set_power_profile_mode = vega10_set_power_profile_mode, .set_power_limit = vega10_set_power_limit, + .odn_edit_dpm_table = vega10_odn_edit_dpm_table, }; int vega10_enable_smc_features(struct pp_hwmgr *hwmgr, diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.h b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.h index 5339ea1f3dce1..aadd6cbc7e85d 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_hwmgr.h @@ -282,15 +282,21 @@ struct vega10_registry_data { struct vega10_odn_clock_voltage_dependency_table { uint32_t count; - struct phm_ppt_v1_clock_voltage_dependency_record - entries[MAX_REGULAR_DPM_NUMBER]; + struct phm_ppt_v1_clock_voltage_dependency_record entries[MAX_REGULAR_DPM_NUMBER]; +}; + +struct vega10_odn_vddc_lookup_table { + uint32_t count; + struct phm_ppt_v1_voltage_lookup_record entries[MAX_REGULAR_DPM_NUMBER]; }; struct vega10_odn_dpm_table { - struct phm_odn_clock_levels odn_core_clock_dpm_levels; - struct phm_odn_clock_levels odn_memory_clock_dpm_levels; - struct vega10_odn_clock_voltage_dependency_table vdd_dependency_on_sclk; - struct vega10_odn_clock_voltage_dependency_table vdd_dependency_on_mclk; + struct vega10_odn_clock_voltage_dependency_table vdd_dep_on_sclk; + struct vega10_odn_clock_voltage_dependency_table vdd_dep_on_mclk; + struct vega10_odn_clock_voltage_dependency_table vdd_dep_on_socclk; + struct vega10_odn_vddc_lookup_table vddc_lookup_table; + uint32_t max_vddc; + uint32_t min_vddc; }; struct vega10_odn_fan_table { @@ -301,8 +307,8 @@ struct vega10_odn_fan_table { }; struct vega10_hwmgr { - struct vega10_dpm_table dpm_table; - struct vega10_dpm_table golden_dpm_table; + struct vega10_dpm_table dpm_table; + struct vega10_dpm_table golden_dpm_table; struct vega10_registry_data registry_data; struct vega10_vbios_boot_state vbios_boot_state; struct vega10_mclk_latency_table mclk_latency_table; @@ -368,12 +374,8 @@ struct vega10_hwmgr { bool need_long_memory_training; /* Internal settings to apply the application power optimization parameters */ - bool apply_optimized_settings; uint32_t disable_dpm_mask; - /* ---- Overdrive next setting ---- */ - uint32_t apply_overdrive_next_settings_mask; - /* ---- SMU9 ---- */ struct smu_features smu_features[GNLD_FEATURES_MAX]; struct vega10_smc_state_table smc_state_table; diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_powertune.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_powertune.c index ba63faefc61fd..a9efd8554fbc2 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_powertune.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_powertune.c @@ -27,7 +27,7 @@ #include "vega10_ppsmc.h" #include "vega10_inc.h" #include "pp_debug.h" -#include "pp_soc15.h" +#include "soc15_common.h" static const struct vega10_didt_config_reg SEDiDtTuningCtrlConfig_Vega10[] = { @@ -888,36 +888,36 @@ static void vega10_didt_set_mask(struct pp_hwmgr *hwmgr, const bool enable) if (PP_CAP(PHM_PlatformCaps_DiDtEDCEnable)) { if (PP_CAP(PHM_PlatformCaps_SQRamping)) { data = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__DIDT, ixDIDT_SQ_EDC_CTRL); - data = CGS_REG_SET_FIELD(data, DIDT_SQ_EDC_CTRL, EDC_EN, en); - data = CGS_REG_SET_FIELD(data, DIDT_SQ_EDC_CTRL, EDC_SW_RST, ~en); + data = REG_SET_FIELD(data, DIDT_SQ_EDC_CTRL, EDC_EN, en); + data = REG_SET_FIELD(data, DIDT_SQ_EDC_CTRL, EDC_SW_RST, ~en); cgs_write_ind_register(hwmgr->device, CGS_IND_REG__DIDT, ixDIDT_SQ_EDC_CTRL, data); } if (PP_CAP(PHM_PlatformCaps_DBRamping)) { data = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__DIDT, ixDIDT_DB_EDC_CTRL); - data = CGS_REG_SET_FIELD(data, DIDT_DB_EDC_CTRL, EDC_EN, en); - data = CGS_REG_SET_FIELD(data, DIDT_DB_EDC_CTRL, EDC_SW_RST, ~en); + data = REG_SET_FIELD(data, DIDT_DB_EDC_CTRL, EDC_EN, en); + data = REG_SET_FIELD(data, DIDT_DB_EDC_CTRL, EDC_SW_RST, ~en); cgs_write_ind_register(hwmgr->device, CGS_IND_REG__DIDT, ixDIDT_DB_EDC_CTRL, data); } if (PP_CAP(PHM_PlatformCaps_TDRamping)) { data = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__DIDT, ixDIDT_TD_EDC_CTRL); - data = CGS_REG_SET_FIELD(data, DIDT_TD_EDC_CTRL, EDC_EN, en); - data = CGS_REG_SET_FIELD(data, DIDT_TD_EDC_CTRL, EDC_SW_RST, ~en); + data = REG_SET_FIELD(data, DIDT_TD_EDC_CTRL, EDC_EN, en); + data = REG_SET_FIELD(data, DIDT_TD_EDC_CTRL, EDC_SW_RST, ~en); cgs_write_ind_register(hwmgr->device, CGS_IND_REG__DIDT, ixDIDT_TD_EDC_CTRL, data); } if (PP_CAP(PHM_PlatformCaps_TCPRamping)) { data = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__DIDT, ixDIDT_TCP_EDC_CTRL); - data = CGS_REG_SET_FIELD(data, DIDT_TCP_EDC_CTRL, EDC_EN, en); - data = CGS_REG_SET_FIELD(data, DIDT_TCP_EDC_CTRL, EDC_SW_RST, ~en); + data = REG_SET_FIELD(data, DIDT_TCP_EDC_CTRL, EDC_EN, en); + data = REG_SET_FIELD(data, DIDT_TCP_EDC_CTRL, EDC_SW_RST, ~en); cgs_write_ind_register(hwmgr->device, CGS_IND_REG__DIDT, ixDIDT_TCP_EDC_CTRL, data); } if (PP_CAP(PHM_PlatformCaps_DBRRamping)) { data = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__DIDT, ixDIDT_DBR_EDC_CTRL); - data = CGS_REG_SET_FIELD(data, DIDT_DBR_EDC_CTRL, EDC_EN, en); - data = CGS_REG_SET_FIELD(data, DIDT_DBR_EDC_CTRL, EDC_SW_RST, ~en); + data = REG_SET_FIELD(data, DIDT_DBR_EDC_CTRL, EDC_EN, en); + data = REG_SET_FIELD(data, DIDT_DBR_EDC_CTRL, EDC_SW_RST, ~en); cgs_write_ind_register(hwmgr->device, CGS_IND_REG__DIDT, ixDIDT_DBR_EDC_CTRL, data); } } @@ -930,20 +930,18 @@ static void vega10_didt_set_mask(struct pp_hwmgr *hwmgr, const bool enable) static int vega10_enable_cac_driving_se_didt_config(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; int result; uint32_t num_se = 0, count, data; - struct amdgpu_device *adev = hwmgr->adev; - uint32_t reg; num_se = adev->gfx.config.max_shader_engines; - cgs_enter_safe_mode(hwmgr->device, true); + adev->gfx.rlc.funcs->enter_safe_mode(adev); - cgs_lock_grbm_idx(hwmgr->device, true); - reg = soc15_get_register_offset(GC_HWID, 0, mmGRBM_GFX_INDEX_BASE_IDX, mmGRBM_GFX_INDEX); + mutex_lock(&adev->grbm_idx_mutex); for (count = 0; count < num_se; count++) { data = GRBM_GFX_INDEX__INSTANCE_BROADCAST_WRITES_MASK | GRBM_GFX_INDEX__SH_BROADCAST_WRITES_MASK | ( count << GRBM_GFX_INDEX__SE_INDEX__SHIFT); - cgs_write_register(hwmgr->device, reg, data); + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, data); result = vega10_program_didt_config_registers(hwmgr, SEDiDtStallCtrlConfig_vega10, VEGA10_CONFIGREG_DIDT); result |= vega10_program_didt_config_registers(hwmgr, SEDiDtStallPatternConfig_vega10, VEGA10_CONFIGREG_DIDT); @@ -958,43 +956,43 @@ static int vega10_enable_cac_driving_se_didt_config(struct pp_hwmgr *hwmgr) if (0 != result) break; } - cgs_write_register(hwmgr->device, reg, 0xE0000000); - cgs_lock_grbm_idx(hwmgr->device, false); + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, 0xE0000000); + mutex_unlock(&adev->grbm_idx_mutex); vega10_didt_set_mask(hwmgr, true); - cgs_enter_safe_mode(hwmgr->device, false); + adev->gfx.rlc.funcs->exit_safe_mode(adev); return 0; } static int vega10_disable_cac_driving_se_didt_config(struct pp_hwmgr *hwmgr) { - cgs_enter_safe_mode(hwmgr->device, true); + struct amdgpu_device *adev = hwmgr->adev; + + adev->gfx.rlc.funcs->enter_safe_mode(adev); vega10_didt_set_mask(hwmgr, false); - cgs_enter_safe_mode(hwmgr->device, false); + adev->gfx.rlc.funcs->exit_safe_mode(adev); return 0; } static int vega10_enable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; int result; uint32_t num_se = 0, count, data; - struct amdgpu_device *adev = hwmgr->adev; - uint32_t reg; num_se = adev->gfx.config.max_shader_engines; - cgs_enter_safe_mode(hwmgr->device, true); + adev->gfx.rlc.funcs->enter_safe_mode(adev); - cgs_lock_grbm_idx(hwmgr->device, true); - reg = soc15_get_register_offset(GC_HWID, 0, mmGRBM_GFX_INDEX_BASE_IDX, mmGRBM_GFX_INDEX); + mutex_lock(&adev->grbm_idx_mutex); for (count = 0; count < num_se; count++) { data = GRBM_GFX_INDEX__INSTANCE_BROADCAST_WRITES_MASK | GRBM_GFX_INDEX__SH_BROADCAST_WRITES_MASK | ( count << GRBM_GFX_INDEX__SE_INDEX__SHIFT); - cgs_write_register(hwmgr->device, reg, data); + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, data); result = vega10_program_didt_config_registers(hwmgr, SEDiDtStallCtrlConfig_vega10, VEGA10_CONFIGREG_DIDT); result |= vega10_program_didt_config_registers(hwmgr, SEDiDtStallPatternConfig_vega10, VEGA10_CONFIGREG_DIDT); @@ -1003,12 +1001,12 @@ static int vega10_enable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) if (0 != result) break; } - cgs_write_register(hwmgr->device, reg, 0xE0000000); - cgs_lock_grbm_idx(hwmgr->device, false); + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, 0xE0000000); + mutex_unlock(&adev->grbm_idx_mutex); vega10_didt_set_mask(hwmgr, true); - cgs_enter_safe_mode(hwmgr->device, false); + adev->gfx.rlc.funcs->exit_safe_mode(adev); vega10_program_gc_didt_config_registers(hwmgr, GCDiDtDroopCtrlConfig_vega10); if (PP_CAP(PHM_PlatformCaps_GCEDC)) @@ -1022,13 +1020,14 @@ static int vega10_enable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) static int vega10_disable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; uint32_t data; - cgs_enter_safe_mode(hwmgr->device, true); + adev->gfx.rlc.funcs->enter_safe_mode(adev); vega10_didt_set_mask(hwmgr, false); - cgs_enter_safe_mode(hwmgr->device, false); + adev->gfx.rlc.funcs->exit_safe_mode(adev); if (PP_CAP(PHM_PlatformCaps_GCEDC)) { data = 0x00000000; @@ -1043,20 +1042,18 @@ static int vega10_disable_psm_gc_didt_config(struct pp_hwmgr *hwmgr) static int vega10_enable_se_edc_config(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; int result; uint32_t num_se = 0, count, data; - struct amdgpu_device *adev = hwmgr->adev; - uint32_t reg; num_se = adev->gfx.config.max_shader_engines; - cgs_enter_safe_mode(hwmgr->device, true); + adev->gfx.rlc.funcs->enter_safe_mode(adev); - cgs_lock_grbm_idx(hwmgr->device, true); - reg = soc15_get_register_offset(GC_HWID, 0, mmGRBM_GFX_INDEX_BASE_IDX, mmGRBM_GFX_INDEX); + mutex_lock(&adev->grbm_idx_mutex); for (count = 0; count < num_se; count++) { data = GRBM_GFX_INDEX__INSTANCE_BROADCAST_WRITES_MASK | GRBM_GFX_INDEX__SH_BROADCAST_WRITES_MASK | ( count << GRBM_GFX_INDEX__SE_INDEX__SHIFT); - cgs_write_register(hwmgr->device, reg, data); + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, data); result = vega10_program_didt_config_registers(hwmgr, SEDiDtWeightConfig_Vega10, VEGA10_CONFIGREG_DIDT); result |= vega10_program_didt_config_registers(hwmgr, SEEDCStallPatternConfig_Vega10, VEGA10_CONFIGREG_DIDT); result |= vega10_program_didt_config_registers(hwmgr, SEEDCStallDelayConfig_Vega10, VEGA10_CONFIGREG_DIDT); @@ -1067,46 +1064,46 @@ static int vega10_enable_se_edc_config(struct pp_hwmgr *hwmgr) if (0 != result) break; } - cgs_write_register(hwmgr->device, reg, 0xE0000000); - cgs_lock_grbm_idx(hwmgr->device, false); + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, 0xE0000000); + mutex_unlock(&adev->grbm_idx_mutex); vega10_didt_set_mask(hwmgr, true); - cgs_enter_safe_mode(hwmgr->device, false); + adev->gfx.rlc.funcs->exit_safe_mode(adev); return 0; } static int vega10_disable_se_edc_config(struct pp_hwmgr *hwmgr) { - cgs_enter_safe_mode(hwmgr->device, true); + struct amdgpu_device *adev = hwmgr->adev; + + adev->gfx.rlc.funcs->enter_safe_mode(adev); vega10_didt_set_mask(hwmgr, false); - cgs_enter_safe_mode(hwmgr->device, false); + adev->gfx.rlc.funcs->exit_safe_mode(adev); return 0; } static int vega10_enable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; int result; uint32_t num_se = 0; uint32_t count, data; - struct amdgpu_device *adev = hwmgr->adev; - uint32_t reg; num_se = adev->gfx.config.max_shader_engines; - cgs_enter_safe_mode(hwmgr->device, true); + adev->gfx.rlc.funcs->enter_safe_mode(adev); vega10_program_gc_didt_config_registers(hwmgr, AvfsPSMResetConfig_vega10); - cgs_lock_grbm_idx(hwmgr->device, true); - reg = soc15_get_register_offset(GC_HWID, 0, mmGRBM_GFX_INDEX_BASE_IDX, mmGRBM_GFX_INDEX); + mutex_lock(&adev->grbm_idx_mutex); for (count = 0; count < num_se; count++) { data = GRBM_GFX_INDEX__INSTANCE_BROADCAST_WRITES_MASK | GRBM_GFX_INDEX__SH_BROADCAST_WRITES_MASK | ( count << GRBM_GFX_INDEX__SE_INDEX__SHIFT); - cgs_write_register(hwmgr->device, reg, data); + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, data); result |= vega10_program_didt_config_registers(hwmgr, PSMSEEDCStallPatternConfig_Vega10, VEGA10_CONFIGREG_DIDT); result |= vega10_program_didt_config_registers(hwmgr, PSMSEEDCStallDelayConfig_Vega10, VEGA10_CONFIGREG_DIDT); result |= vega10_program_didt_config_registers(hwmgr, PSMSEEDCCtrlResetConfig_Vega10, VEGA10_CONFIGREG_DIDT); @@ -1115,12 +1112,12 @@ static int vega10_enable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) if (0 != result) break; } - cgs_write_register(hwmgr->device, reg, 0xE0000000); - cgs_lock_grbm_idx(hwmgr->device, false); + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, 0xE0000000); + mutex_unlock(&adev->grbm_idx_mutex); vega10_didt_set_mask(hwmgr, true); - cgs_enter_safe_mode(hwmgr->device, false); + adev->gfx.rlc.funcs->exit_safe_mode(adev); vega10_program_gc_didt_config_registers(hwmgr, PSMGCEDCDroopCtrlConfig_vega10); @@ -1137,13 +1134,14 @@ static int vega10_enable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) static int vega10_disable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; uint32_t data; - cgs_enter_safe_mode(hwmgr->device, true); + adev->gfx.rlc.funcs->enter_safe_mode(adev); vega10_didt_set_mask(hwmgr, false); - cgs_enter_safe_mode(hwmgr->device, false); + adev->gfx.rlc.funcs->exit_safe_mode(adev); if (PP_CAP(PHM_PlatformCaps_GCEDC)) { data = 0x00000000; @@ -1158,15 +1156,14 @@ static int vega10_disable_psm_gc_edc_config(struct pp_hwmgr *hwmgr) static int vega10_enable_se_edc_force_stall_config(struct pp_hwmgr *hwmgr) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; int result; - cgs_enter_safe_mode(hwmgr->device, true); + adev->gfx.rlc.funcs->enter_safe_mode(adev); - cgs_lock_grbm_idx(hwmgr->device, true); - reg = soc15_get_register_offset(GC_HWID, 0, mmGRBM_GFX_INDEX_BASE_IDX, mmGRBM_GFX_INDEX); - cgs_write_register(hwmgr->device, reg, 0xE0000000); - cgs_lock_grbm_idx(hwmgr->device, false); + mutex_lock(&adev->grbm_idx_mutex); + WREG32_SOC15(GC, 0, mmGRBM_GFX_INDEX, 0xE0000000); + mutex_unlock(&adev->grbm_idx_mutex); result = vega10_program_didt_config_registers(hwmgr, SEEDCForceStallPatternConfig_Vega10, VEGA10_CONFIGREG_DIDT); result |= vega10_program_didt_config_registers(hwmgr, SEEDCCtrlForceStallConfig_Vega10, VEGA10_CONFIGREG_DIDT); @@ -1175,7 +1172,7 @@ static int vega10_enable_se_edc_force_stall_config(struct pp_hwmgr *hwmgr) vega10_didt_set_mask(hwmgr, false); - cgs_enter_safe_mode(hwmgr->device, false); + adev->gfx.rlc.funcs->exit_safe_mode(adev); return 0; } diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c index c61d0744860d4..0768d259c07c3 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_processpptables.c @@ -52,7 +52,7 @@ static const void *get_powerplay_table(struct pp_hwmgr *hwmgr) if (!table_address) { table_address = (ATOM_Vega10_POWERPLAYTABLE *) - cgs_atom_get_data_table(hwmgr->device, index, + smu_atom_get_data_table(hwmgr->adev, index, &size, &frev, &crev); hwmgr->soft_pp_table = table_address; /*Cache the result in RAM.*/ diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.c index 9f18226a56ea3..aa044c1955fe0 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega10_thermal.c @@ -25,7 +25,7 @@ #include "vega10_hwmgr.h" #include "vega10_ppsmc.h" #include "vega10_inc.h" -#include "pp_soc15.h" +#include "soc15_common.h" #include "pp_debug.h" static int vega10_get_current_rpm(struct pp_hwmgr *hwmgr, uint32_t *current_rpm) @@ -89,6 +89,7 @@ int vega10_fan_ctrl_get_fan_speed_percent(struct pp_hwmgr *hwmgr, int vega10_fan_ctrl_get_fan_speed_rpm(struct pp_hwmgr *hwmgr, uint32_t *speed) { + struct amdgpu_device *adev = hwmgr->adev; struct vega10_hwmgr *data = hwmgr->backend; uint32_t tach_period; uint32_t crystal_clock_freq; @@ -100,10 +101,8 @@ int vega10_fan_ctrl_get_fan_speed_rpm(struct pp_hwmgr *hwmgr, uint32_t *speed) if (data->smu_features[GNLD_FAN_CONTROL].supported) { result = vega10_get_current_rpm(hwmgr, speed); } else { - uint32_t reg = soc15_get_register_offset(THM_HWID, 0, - mmCG_TACH_STATUS_BASE_IDX, mmCG_TACH_STATUS); tach_period = - CGS_REG_GET_FIELD(cgs_read_register(hwmgr->device, reg), + REG_GET_FIELD(RREG32_SOC15(THM, 0, mmCG_TACH_STATUS), CG_TACH_STATUS, TACH_PERIOD); @@ -127,26 +126,23 @@ int vega10_fan_ctrl_get_fan_speed_rpm(struct pp_hwmgr *hwmgr, uint32_t *speed) */ int vega10_fan_ctrl_set_static_mode(struct pp_hwmgr *hwmgr, uint32_t mode) { - uint32_t reg; - - reg = soc15_get_register_offset(THM_HWID, 0, - mmCG_FDO_CTRL2_BASE_IDX, mmCG_FDO_CTRL2); + struct amdgpu_device *adev = hwmgr->adev; if (hwmgr->fan_ctrl_is_in_default_mode) { hwmgr->fan_ctrl_default_mode = - CGS_REG_GET_FIELD(cgs_read_register(hwmgr->device, reg), + REG_GET_FIELD(RREG32_SOC15(THM, 0, mmCG_FDO_CTRL2), CG_FDO_CTRL2, FDO_PWM_MODE); hwmgr->tmin = - CGS_REG_GET_FIELD(cgs_read_register(hwmgr->device, reg), + REG_GET_FIELD(RREG32_SOC15(THM, 0, mmCG_FDO_CTRL2), CG_FDO_CTRL2, TMIN); hwmgr->fan_ctrl_is_in_default_mode = false; } - cgs_write_register(hwmgr->device, reg, - CGS_REG_SET_FIELD(cgs_read_register(hwmgr->device, reg), + WREG32_SOC15(THM, 0, mmCG_FDO_CTRL2, + REG_SET_FIELD(RREG32_SOC15(THM, 0, mmCG_FDO_CTRL2), CG_FDO_CTRL2, TMIN, 0)); - cgs_write_register(hwmgr->device, reg, - CGS_REG_SET_FIELD(cgs_read_register(hwmgr->device, reg), + WREG32_SOC15(THM, 0, mmCG_FDO_CTRL2, + REG_SET_FIELD(RREG32_SOC15(THM, 0, mmCG_FDO_CTRL2), CG_FDO_CTRL2, FDO_PWM_MODE, mode)); return 0; @@ -159,18 +155,15 @@ int vega10_fan_ctrl_set_static_mode(struct pp_hwmgr *hwmgr, uint32_t mode) */ int vega10_fan_ctrl_set_default_mode(struct pp_hwmgr *hwmgr) { - uint32_t reg; - - reg = soc15_get_register_offset(THM_HWID, 0, - mmCG_FDO_CTRL2_BASE_IDX, mmCG_FDO_CTRL2); + struct amdgpu_device *adev = hwmgr->adev; if (!hwmgr->fan_ctrl_is_in_default_mode) { - cgs_write_register(hwmgr->device, reg, - CGS_REG_SET_FIELD(cgs_read_register(hwmgr->device, reg), + WREG32_SOC15(THM, 0, mmCG_FDO_CTRL2, + REG_SET_FIELD(RREG32_SOC15(THM, 0, mmCG_FDO_CTRL2), CG_FDO_CTRL2, FDO_PWM_MODE, hwmgr->fan_ctrl_default_mode)); - cgs_write_register(hwmgr->device, reg, - CGS_REG_SET_FIELD(cgs_read_register(hwmgr->device, reg), + WREG32_SOC15(THM, 0, mmCG_FDO_CTRL2, + REG_SET_FIELD(RREG32_SOC15(THM, 0, mmCG_FDO_CTRL2), CG_FDO_CTRL2, TMIN, hwmgr->tmin << CG_FDO_CTRL2__TMIN__SHIFT)); hwmgr->fan_ctrl_is_in_default_mode = true; @@ -257,10 +250,10 @@ int vega10_fan_ctrl_stop_smc_fan_control(struct pp_hwmgr *hwmgr) int vega10_fan_ctrl_set_fan_speed_percent(struct pp_hwmgr *hwmgr, uint32_t speed) { + struct amdgpu_device *adev = hwmgr->adev; uint32_t duty100; uint32_t duty; uint64_t tmp64; - uint32_t reg; if (hwmgr->thermal_controller.fanInfo.bNoFan) return 0; @@ -271,10 +264,7 @@ int vega10_fan_ctrl_set_fan_speed_percent(struct pp_hwmgr *hwmgr, if (PP_CAP(PHM_PlatformCaps_MicrocodeFanControl)) vega10_fan_ctrl_stop_smc_fan_control(hwmgr); - reg = soc15_get_register_offset(THM_HWID, 0, - mmCG_FDO_CTRL1_BASE_IDX, mmCG_FDO_CTRL1); - - duty100 = CGS_REG_GET_FIELD(cgs_read_register(hwmgr->device, reg), + duty100 = REG_GET_FIELD(RREG32_SOC15(THM, 0, mmCG_FDO_CTRL1), CG_FDO_CTRL1, FMAX_DUTY100); if (duty100 == 0) @@ -284,10 +274,8 @@ int vega10_fan_ctrl_set_fan_speed_percent(struct pp_hwmgr *hwmgr, do_div(tmp64, 100); duty = (uint32_t)tmp64; - reg = soc15_get_register_offset(THM_HWID, 0, - mmCG_FDO_CTRL0_BASE_IDX, mmCG_FDO_CTRL0); - cgs_write_register(hwmgr->device, reg, - CGS_REG_SET_FIELD(cgs_read_register(hwmgr->device, reg), + WREG32_SOC15(THM, 0, mmCG_FDO_CTRL0, + REG_SET_FIELD(RREG32_SOC15(THM, 0, mmCG_FDO_CTRL0), CG_FDO_CTRL0, FDO_STATIC_DUTY, duty)); return vega10_fan_ctrl_set_static_mode(hwmgr, FDO_PWM_MODE_STATIC); @@ -317,10 +305,10 @@ int vega10_fan_ctrl_reset_fan_speed_to_default(struct pp_hwmgr *hwmgr) */ int vega10_fan_ctrl_set_fan_speed_rpm(struct pp_hwmgr *hwmgr, uint32_t speed) { + struct amdgpu_device *adev = hwmgr->adev; uint32_t tach_period; uint32_t crystal_clock_freq; int result = 0; - uint32_t reg; if (hwmgr->thermal_controller.fanInfo.bNoFan || (speed < hwmgr->thermal_controller.fanInfo.ulMinRPM) || @@ -333,10 +321,8 @@ int vega10_fan_ctrl_set_fan_speed_rpm(struct pp_hwmgr *hwmgr, uint32_t speed) if (!result) { crystal_clock_freq = amdgpu_asic_get_xclk((struct amdgpu_device *)hwmgr->adev); tach_period = 60 * crystal_clock_freq * 10000 / (8 * speed); - reg = soc15_get_register_offset(THM_HWID, 0, - mmCG_TACH_STATUS_BASE_IDX, mmCG_TACH_STATUS); - cgs_write_register(hwmgr->device, reg, - CGS_REG_SET_FIELD(cgs_read_register(hwmgr->device, reg), + WREG32_SOC15(THM, 0, mmCG_TACH_STATUS, + REG_SET_FIELD(RREG32_SOC15(THM, 0, mmCG_TACH_STATUS), CG_TACH_STATUS, TACH_PERIOD, tach_period)); } @@ -350,13 +336,10 @@ int vega10_fan_ctrl_set_fan_speed_rpm(struct pp_hwmgr *hwmgr, uint32_t speed) */ int vega10_thermal_get_temperature(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; int temp; - uint32_t reg; - reg = soc15_get_register_offset(THM_HWID, 0, - mmCG_MULT_THERMAL_STATUS_BASE_IDX, mmCG_MULT_THERMAL_STATUS); - - temp = cgs_read_register(hwmgr->device, reg); + temp = RREG32_SOC15(THM, 0, mmCG_MULT_THERMAL_STATUS); temp = (temp & CG_MULT_THERMAL_STATUS__CTF_TEMP_MASK) >> CG_MULT_THERMAL_STATUS__CTF_TEMP__SHIFT; @@ -379,11 +362,12 @@ int vega10_thermal_get_temperature(struct pp_hwmgr *hwmgr) static int vega10_thermal_set_temperature_range(struct pp_hwmgr *hwmgr, struct PP_TemperatureRange *range) { + struct amdgpu_device *adev = hwmgr->adev; int low = VEGA10_THERMAL_MINIMUM_ALERT_TEMP * PP_TEMPERATURE_UNITS_PER_CENTIGRADES; int high = VEGA10_THERMAL_MAXIMUM_ALERT_TEMP * PP_TEMPERATURE_UNITS_PER_CENTIGRADES; - uint32_t val, reg; + uint32_t val; if (low < range->min) low = range->min; @@ -393,20 +377,17 @@ static int vega10_thermal_set_temperature_range(struct pp_hwmgr *hwmgr, if (low > high) return -EINVAL; - reg = soc15_get_register_offset(THM_HWID, 0, - mmTHM_THERMAL_INT_CTRL_BASE_IDX, mmTHM_THERMAL_INT_CTRL); - - val = cgs_read_register(hwmgr->device, reg); + val = RREG32_SOC15(THM, 0, mmTHM_THERMAL_INT_CTRL); - val = CGS_REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, MAX_IH_CREDIT, 5); - val = CGS_REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, THERM_IH_HW_ENA, 1); - val = CGS_REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, DIG_THERM_INTH, (high / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); - val = CGS_REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, DIG_THERM_INTL, (low / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); + val = REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, MAX_IH_CREDIT, 5); + val = REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, THERM_IH_HW_ENA, 1); + val = REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, DIG_THERM_INTH, (high / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); + val = REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, DIG_THERM_INTL, (low / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); val &= (~THM_THERMAL_INT_CTRL__THERM_TRIGGER_MASK_MASK) & (~THM_THERMAL_INT_CTRL__THERM_INTH_MASK_MASK) & (~THM_THERMAL_INT_CTRL__THERM_INTL_MASK_MASK); - cgs_write_register(hwmgr->device, reg, val); + WREG32_SOC15(THM, 0, mmTHM_THERMAL_INT_CTRL, val); return 0; } @@ -418,21 +399,17 @@ static int vega10_thermal_set_temperature_range(struct pp_hwmgr *hwmgr, */ static int vega10_thermal_initialize(struct pp_hwmgr *hwmgr) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; if (hwmgr->thermal_controller.fanInfo.ucTachometerPulsesPerRevolution) { - reg = soc15_get_register_offset(THM_HWID, 0, - mmCG_TACH_CTRL_BASE_IDX, mmCG_TACH_CTRL); - cgs_write_register(hwmgr->device, reg, - CGS_REG_SET_FIELD(cgs_read_register(hwmgr->device, reg), + WREG32_SOC15(THM, 0, mmCG_TACH_CTRL, + REG_SET_FIELD(RREG32_SOC15(THM, 0, mmCG_TACH_CTRL), CG_TACH_CTRL, EDGE_PER_REV, hwmgr->thermal_controller.fanInfo.ucTachometerPulsesPerRevolution - 1)); } - reg = soc15_get_register_offset(THM_HWID, 0, - mmCG_FDO_CTRL2_BASE_IDX, mmCG_FDO_CTRL2); - cgs_write_register(hwmgr->device, reg, - CGS_REG_SET_FIELD(cgs_read_register(hwmgr->device, reg), + WREG32_SOC15(THM, 0, mmCG_FDO_CTRL2, + REG_SET_FIELD(RREG32_SOC15(THM, 0, mmCG_FDO_CTRL2), CG_FDO_CTRL2, TACH_PWM_RESP_RATE, 0x28)); return 0; @@ -445,9 +422,9 @@ static int vega10_thermal_initialize(struct pp_hwmgr *hwmgr) */ static int vega10_thermal_enable_alert(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; struct vega10_hwmgr *data = hwmgr->backend; uint32_t val = 0; - uint32_t reg; if (data->smu_features[GNLD_FW_CTF].supported) { if (data->smu_features[GNLD_FW_CTF].enabled) @@ -465,8 +442,7 @@ static int vega10_thermal_enable_alert(struct pp_hwmgr *hwmgr) val |= (1 << THM_THERMAL_INT_ENA__THERM_INTL_CLR__SHIFT); val |= (1 << THM_THERMAL_INT_ENA__THERM_TRIGGER_CLR__SHIFT); - reg = soc15_get_register_offset(THM_HWID, 0, mmTHM_THERMAL_INT_ENA_BASE_IDX, mmTHM_THERMAL_INT_ENA); - cgs_write_register(hwmgr->device, reg, val); + WREG32_SOC15(THM, 0, mmTHM_THERMAL_INT_ENA, val); return 0; } @@ -477,8 +453,8 @@ static int vega10_thermal_enable_alert(struct pp_hwmgr *hwmgr) */ int vega10_thermal_disable_alert(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; struct vega10_hwmgr *data = hwmgr->backend; - uint32_t reg; if (data->smu_features[GNLD_FW_CTF].supported) { if (!data->smu_features[GNLD_FW_CTF].enabled) @@ -493,8 +469,7 @@ int vega10_thermal_disable_alert(struct pp_hwmgr *hwmgr) data->smu_features[GNLD_FW_CTF].enabled = false; } - reg = soc15_get_register_offset(THM_HWID, 0, mmTHM_THERMAL_INT_ENA_BASE_IDX, mmTHM_THERMAL_INT_ENA); - cgs_write_register(hwmgr->device, reg, 0); + WREG32_SOC15(THM, 0, mmTHM_THERMAL_INT_ENA, 0); return 0; } diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.c index 200de46bd06b2..782e2098824df 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.c @@ -34,7 +34,6 @@ #include "atomfirmware.h" #include "cgs_common.h" #include "vega12_inc.h" -#include "pp_soc15.h" #include "pppcielanes.h" #include "vega12_hwmgr.h" #include "vega12_processpptables.h" @@ -546,6 +545,7 @@ static int vega12_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) return -EINVAL); dpm_table->dpm_levels[i].value = clock; + dpm_table->dpm_levels[i].enabled = true; } vega12_init_dpm_state(&(dpm_table->dpm_state)); @@ -565,6 +565,7 @@ static int vega12_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) return -EINVAL); dpm_table->dpm_levels[i].value = clock; + dpm_table->dpm_levels[i].enabled = true; } vega12_init_dpm_state(&(dpm_table->dpm_state)); @@ -585,6 +586,7 @@ static int vega12_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) return -EINVAL); dpm_table->dpm_levels[i].value = clock; + dpm_table->dpm_levels[i].enabled = true; } vega12_init_dpm_state(&(dpm_table->dpm_state)); @@ -605,6 +607,7 @@ static int vega12_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) return -EINVAL); dpm_table->dpm_levels[i].value = clock; + dpm_table->dpm_levels[i].enabled = true; } vega12_init_dpm_state(&(dpm_table->dpm_state)); @@ -625,6 +628,7 @@ static int vega12_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) return -EINVAL); dpm_table->dpm_levels[i].value = clock; + dpm_table->dpm_levels[i].enabled = true; } vega12_init_dpm_state(&(dpm_table->dpm_state)); @@ -645,6 +649,7 @@ static int vega12_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) return -EINVAL); dpm_table->dpm_levels[i].value = clock; + dpm_table->dpm_levels[i].enabled = true; } vega12_init_dpm_state(&(dpm_table->dpm_state)); @@ -666,6 +671,7 @@ static int vega12_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) return -EINVAL); dpm_table->dpm_levels[i].value = clock; + dpm_table->dpm_levels[i].enabled = true; } vega12_init_dpm_state(&(dpm_table->dpm_state)); @@ -686,6 +692,7 @@ static int vega12_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) return -EINVAL); dpm_table->dpm_levels[i].value = clock; + dpm_table->dpm_levels[i].enabled = true; } vega12_init_dpm_state(&(dpm_table->dpm_state)); @@ -706,6 +713,7 @@ static int vega12_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) return -EINVAL); dpm_table->dpm_levels[i].value = clock; + dpm_table->dpm_levels[i].enabled = true; } vega12_init_dpm_state(&(dpm_table->dpm_state)); @@ -726,6 +734,7 @@ static int vega12_setup_default_dpm_tables(struct pp_hwmgr *hwmgr) return -EINVAL); dpm_table->dpm_levels[i].value = clock; + dpm_table->dpm_levels[i].enabled = true; } vega12_init_dpm_state(&(dpm_table->dpm_state)); @@ -992,15 +1001,55 @@ static uint32_t vega12_find_highest_dpm_level( static int vega12_upload_dpm_min_level(struct pp_hwmgr *hwmgr) { + struct vega12_hwmgr *data = hwmgr->backend; + if (data->smc_state_table.gfx_boot_level != + data->dpm_table.gfx_table.dpm_state.soft_min_level) { + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMinByFreq, + PPCLK_GFXCLK<<16 | data->dpm_table.gfx_table.dpm_levels[data->smc_state_table.gfx_boot_level].value); + data->dpm_table.gfx_table.dpm_state.soft_min_level = + data->smc_state_table.gfx_boot_level; + } + + if (data->smc_state_table.mem_boot_level != + data->dpm_table.mem_table.dpm_state.soft_min_level) { + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMinByFreq, + PPCLK_UCLK<<16 | data->dpm_table.mem_table.dpm_levels[data->smc_state_table.mem_boot_level].value); + data->dpm_table.mem_table.dpm_state.soft_min_level = + data->smc_state_table.mem_boot_level; + } + return 0; + } static int vega12_upload_dpm_max_level(struct pp_hwmgr *hwmgr) { + struct vega12_hwmgr *data = hwmgr->backend; + if (data->smc_state_table.gfx_max_level != + data->dpm_table.gfx_table.dpm_state.soft_max_level) { + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMaxByFreq, + /* plus the vale by 1 to align the resolution */ + PPCLK_GFXCLK<<16 | (data->dpm_table.gfx_table.dpm_levels[data->smc_state_table.gfx_max_level].value + 1)); + data->dpm_table.gfx_table.dpm_state.soft_max_level = + data->smc_state_table.gfx_max_level; + } + + if (data->smc_state_table.mem_max_level != + data->dpm_table.mem_table.dpm_state.soft_max_level) { + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SetSoftMaxByFreq, + /* plus the vale by 1 to align the resolution */ + PPCLK_UCLK<<16 | (data->dpm_table.mem_table.dpm_levels[data->smc_state_table.mem_max_level].value + 1)); + data->dpm_table.mem_table.dpm_state.soft_max_level = + data->smc_state_table.mem_max_level; + } + return 0; } - int vega12_enable_disable_vce_dpm(struct pp_hwmgr *hwmgr, bool enable) { struct vega12_hwmgr *data = @@ -1064,8 +1113,7 @@ static uint32_t vega12_dpm_get_mclk(struct pp_hwmgr *hwmgr, bool low) return (mem_clk * 100); } -static int vega12_get_gpu_power(struct pp_hwmgr *hwmgr, - struct pp_gpu_power *query) +static int vega12_get_gpu_power(struct pp_hwmgr *hwmgr, uint32_t *query) { #if 0 uint32_t value; @@ -1077,7 +1125,7 @@ static int vega12_get_gpu_power(struct pp_hwmgr *hwmgr, vega12_read_arg_from_smc(hwmgr, &value); /* power value is an integer */ - query->average_gpu_power = value << 8; + *query = value << 8; #endif return 0; } @@ -1186,12 +1234,8 @@ static int vega12_read_sensor(struct pp_hwmgr *hwmgr, int idx, *size = 4; break; case AMDGPU_PP_SENSOR_GPU_POWER: - if (*size < sizeof(struct pp_gpu_power)) - ret = -EINVAL; - else { - *size = sizeof(struct pp_gpu_power); - ret = vega12_get_gpu_power(hwmgr, (struct pp_gpu_power *)value); - } + ret = vega12_get_gpu_power(hwmgr, (uint32_t *)value); + break; default: ret = -EINVAL; @@ -1260,23 +1304,18 @@ static int vega12_notify_smc_display_config_after_ps_adjustment( { struct vega12_hwmgr *data = (struct vega12_hwmgr *)(hwmgr->backend); - uint32_t num_active_disps = 0; - struct cgs_display_info info = {0}; struct PP_Clocks min_clocks = {0}; struct pp_display_clock_request clock_req; uint32_t clk_request; - info.mode_info = NULL; - cgs_get_active_displays_info(hwmgr->device, &info); - num_active_disps = info.display_count; - if (num_active_disps > 1) + if (hwmgr->display_config->num_display > 1) vega12_notify_smc_display_change(hwmgr, false); else vega12_notify_smc_display_change(hwmgr, true); - min_clocks.dcefClock = hwmgr->display_config.min_dcef_set_clk; - min_clocks.dcefClockInSR = hwmgr->display_config.min_dcef_deep_sleep_set_clk; - min_clocks.memoryClock = hwmgr->display_config.min_mem_set_clock; + min_clocks.dcefClock = hwmgr->display_config->min_dcef_set_clk; + min_clocks.dcefClockInSR = hwmgr->display_config->min_dcef_deep_sleep_set_clk; + min_clocks.memoryClock = hwmgr->display_config->min_mem_set_clock; if (data->smu_features[GNLD_DPM_DCEFCLK].supported) { clock_req.clock_type = amd_pp_dcef_clock; @@ -1832,9 +1871,7 @@ static int vega12_display_configuration_changed_task(struct pp_hwmgr *hwmgr) { struct vega12_hwmgr *data = (struct vega12_hwmgr *)(hwmgr->backend); int result = 0; - uint32_t num_turned_on_displays = 1; Watermarks_t *wm_table = &(data->smc_state_table.water_marks_table); - struct cgs_display_info info = {0}; if ((data->water_marks_bitmap & WaterMarksExist) && !(data->water_marks_bitmap & WaterMarksLoaded)) { @@ -1846,12 +1883,9 @@ static int vega12_display_configuration_changed_task(struct pp_hwmgr *hwmgr) if ((data->water_marks_bitmap & WaterMarksExist) && data->smu_features[GNLD_DPM_DCEFCLK].supported && - data->smu_features[GNLD_DPM_SOCCLK].supported) { - cgs_get_active_displays_info(hwmgr->device, &info); - num_turned_on_displays = info.display_count; + data->smu_features[GNLD_DPM_SOCCLK].supported) smum_send_msg_to_smc_with_parameter(hwmgr, - PPSMC_MSG_NumOfDisplays, num_turned_on_displays); - } + PPSMC_MSG_NumOfDisplays, hwmgr->display_config->num_display); return result; } @@ -1894,15 +1928,12 @@ vega12_check_smc_update_required_for_display_configuration(struct pp_hwmgr *hwmg { struct vega12_hwmgr *data = (struct vega12_hwmgr *)(hwmgr->backend); bool is_update_required = false; - struct cgs_display_info info = {0, 0, NULL}; - - cgs_get_active_displays_info(hwmgr->device, &info); - if (data->display_timing.num_existing_displays != info.display_count) + if (data->display_timing.num_existing_displays != hwmgr->display_config->num_display) is_update_required = true; if (data->registry_data.gfx_clk_deep_sleep_support) { - if (data->display_timing.min_clock_in_sr != hwmgr->display_config.min_core_set_clock_in_sr) + if (data->display_timing.min_clock_in_sr != hwmgr->display_config->min_core_set_clock_in_sr) is_update_required = true; } diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.h b/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.h index bc98b1df3b653..e81ded1ec1982 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_hwmgr.h @@ -33,7 +33,7 @@ #define WaterMarksExist 1 #define WaterMarksLoaded 2 -#define VG12_PSUEDO_NUM_GFXCLK_DPM_LEVELS 8 +#define VG12_PSUEDO_NUM_GFXCLK_DPM_LEVELS 16 #define VG12_PSUEDO_NUM_SOCCLK_DPM_LEVELS 8 #define VG12_PSUEDO_NUM_DCEFCLK_DPM_LEVELS 8 #define VG12_PSUEDO_NUM_UCLK_DPM_LEVELS 4 diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_processpptables.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_processpptables.c index b34113f459044..888ddca902d89 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_processpptables.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_processpptables.c @@ -51,7 +51,7 @@ static const void *get_powerplay_table(struct pp_hwmgr *hwmgr) if (!table_address) { table_address = (ATOM_Vega12_POWERPLAYTABLE *) - cgs_atom_get_data_table(hwmgr->device, index, + smu_atom_get_data_table(hwmgr->adev, index, &size, &frev, &crev); hwmgr->soft_pp_table = table_address; /*Cache the result in RAM.*/ @@ -224,6 +224,11 @@ static int append_vbios_pptable(struct pp_hwmgr *hwmgr, PPTable_t *ppsmc_pptable ppsmc_pptable->AcgGfxclkSpreadPercent = smc_dpm_table.acggfxclkspreadpercent; ppsmc_pptable->AcgGfxclkSpreadFreq = smc_dpm_table.acggfxclkspreadfreq; + /* 0xFFFF will disable the ACG feature */ + if (!(hwmgr->feature_mask & PP_ACG_MASK)) { + ppsmc_pptable->AcgThresholdFreqHigh = 0xFFFF; + ppsmc_pptable->AcgThresholdFreqLow = 0xFFFF; + } return 0; } diff --git a/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_thermal.c b/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_thermal.c index df0fa815cd6ed..cfd9e6ccb7901 100644 --- a/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_thermal.c +++ b/drivers/gpu/drm/amd/powerplay/hwmgr/vega12_thermal.c @@ -26,7 +26,7 @@ #include "vega12_smumgr.h" #include "vega12_ppsmc.h" #include "vega12_inc.h" -#include "pp_soc15.h" +#include "soc15_common.h" #include "pp_debug.h" static int vega12_get_current_rpm(struct pp_hwmgr *hwmgr, uint32_t *current_rpm) @@ -147,13 +147,10 @@ int vega12_fan_ctrl_reset_fan_speed_to_default(struct pp_hwmgr *hwmgr) */ int vega12_thermal_get_temperature(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; int temp = 0; - uint32_t reg; - reg = soc15_get_register_offset(THM_HWID, 0, - mmCG_MULT_THERMAL_STATUS_BASE_IDX, mmCG_MULT_THERMAL_STATUS); - - temp = cgs_read_register(hwmgr->device, reg); + temp = RREG32_SOC15(THM, 0, mmCG_MULT_THERMAL_STATUS); temp = (temp & CG_MULT_THERMAL_STATUS__CTF_TEMP_MASK) >> CG_MULT_THERMAL_STATUS__CTF_TEMP__SHIFT; @@ -175,11 +172,12 @@ int vega12_thermal_get_temperature(struct pp_hwmgr *hwmgr) static int vega12_thermal_set_temperature_range(struct pp_hwmgr *hwmgr, struct PP_TemperatureRange *range) { + struct amdgpu_device *adev = hwmgr->adev; int low = VEGA12_THERMAL_MINIMUM_ALERT_TEMP * PP_TEMPERATURE_UNITS_PER_CENTIGRADES; int high = VEGA12_THERMAL_MAXIMUM_ALERT_TEMP * PP_TEMPERATURE_UNITS_PER_CENTIGRADES; - uint32_t val, reg; + uint32_t val; if (low < range->min) low = range->min; @@ -189,18 +187,15 @@ static int vega12_thermal_set_temperature_range(struct pp_hwmgr *hwmgr, if (low > high) return -EINVAL; - reg = soc15_get_register_offset(THM_HWID, 0, - mmTHM_THERMAL_INT_CTRL_BASE_IDX, mmTHM_THERMAL_INT_CTRL); - - val = cgs_read_register(hwmgr->device, reg); + val = RREG32_SOC15(THM, 0, mmTHM_THERMAL_INT_CTRL); - val = CGS_REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, MAX_IH_CREDIT, 5); - val = CGS_REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, THERM_IH_HW_ENA, 1); - val = CGS_REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, DIG_THERM_INTH, (high / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); - val = CGS_REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, DIG_THERM_INTL, (low / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); + val = REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, MAX_IH_CREDIT, 5); + val = REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, THERM_IH_HW_ENA, 1); + val = REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, DIG_THERM_INTH, (high / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); + val = REG_SET_FIELD(val, THM_THERMAL_INT_CTRL, DIG_THERM_INTL, (low / PP_TEMPERATURE_UNITS_PER_CENTIGRADES)); val = val & (~THM_THERMAL_INT_CTRL__THERM_TRIGGER_MASK_MASK); - cgs_write_register(hwmgr->device, reg, val); + WREG32_SOC15(THM, 0, mmTHM_THERMAL_INT_CTRL, val); return 0; } @@ -212,15 +207,14 @@ static int vega12_thermal_set_temperature_range(struct pp_hwmgr *hwmgr, */ static int vega12_thermal_enable_alert(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; uint32_t val = 0; - uint32_t reg; val |= (1 << THM_THERMAL_INT_ENA__THERM_INTH_CLR__SHIFT); val |= (1 << THM_THERMAL_INT_ENA__THERM_INTL_CLR__SHIFT); val |= (1 << THM_THERMAL_INT_ENA__THERM_TRIGGER_CLR__SHIFT); - reg = soc15_get_register_offset(THM_HWID, 0, mmTHM_THERMAL_INT_ENA_BASE_IDX, mmTHM_THERMAL_INT_ENA); - cgs_write_register(hwmgr->device, reg, val); + WREG32_SOC15(THM, 0, mmTHM_THERMAL_INT_ENA, val); return 0; } @@ -231,10 +225,9 @@ static int vega12_thermal_enable_alert(struct pp_hwmgr *hwmgr) */ int vega12_thermal_disable_alert(struct pp_hwmgr *hwmgr) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; - reg = soc15_get_register_offset(THM_HWID, 0, mmTHM_THERMAL_INT_ENA_BASE_IDX, mmTHM_THERMAL_INT_ENA); - cgs_write_register(hwmgr->device, reg, 0); + WREG32_SOC15(THM, 0, mmTHM_THERMAL_INT_ENA, 0); return 0; } diff --git a/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h b/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h index 8b78bbecd1bc5..a202247c98944 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h +++ b/drivers/gpu/drm/amd/powerplay/inc/hardwaremanager.h @@ -377,11 +377,7 @@ struct phm_clocks { #define DPMTABLE_UPDATE_SCLK 0x00000004 #define DPMTABLE_UPDATE_MCLK 0x00000008 #define DPMTABLE_OD_UPDATE_VDDC 0x00000010 - -/* To determine if sclk and mclk are in overdrive state */ -#define SCLK_OVERDRIVE_ENABLED 0x00000001 -#define MCLK_OVERDRIVE_ENABLED 0x00000002 -#define VDDC_OVERDRIVE_ENABLED 0x00000010 +#define DPMTABLE_UPDATE_SOCCLK 0x00000020 struct phm_odn_performance_level { uint32_t clock; @@ -414,7 +410,10 @@ extern int phm_apply_state_adjust_rules(struct pp_hwmgr *hwmgr, struct pp_power_state *adjusted_ps, const struct pp_power_state *current_ps); +extern int phm_apply_clock_adjust_rules(struct pp_hwmgr *hwmgr); + extern int phm_force_dpm_levels(struct pp_hwmgr *hwmgr, enum amd_dpm_forced_level level); +extern int phm_pre_display_configuration_changed(struct pp_hwmgr *hwmgr); extern int phm_display_configuration_changed(struct pp_hwmgr *hwmgr); extern int phm_notify_smc_display_config_after_ps_adjustment(struct pp_hwmgr *hwmgr); extern int phm_register_irq_handlers(struct pp_hwmgr *hwmgr); diff --git a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h index 17f811d181c88..b99fb8ac822c8 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h +++ b/drivers/gpu/drm/amd/powerplay/inc/hwmgr.h @@ -38,6 +38,8 @@ struct phm_fan_speed_info; struct pp_atomctrl_voltage_table; #define VOLTAGE_SCALE 4 +#define VOLTAGE_VID_OFFSET_SCALE1 625 +#define VOLTAGE_VID_OFFSET_SCALE2 100 enum DISPLAY_GAP { DISPLAY_GAP_VBLANK_OR_WM = 0, /* Wait for vblank or MCHG watermark. */ @@ -64,24 +66,6 @@ struct vi_dpm_table { #define PCIE_PERF_REQ_GEN2 3 #define PCIE_PERF_REQ_GEN3 4 -enum PP_FEATURE_MASK { - PP_SCLK_DPM_MASK = 0x1, - PP_MCLK_DPM_MASK = 0x2, - PP_PCIE_DPM_MASK = 0x4, - PP_SCLK_DEEP_SLEEP_MASK = 0x8, - PP_POWER_CONTAINMENT_MASK = 0x10, - PP_UVD_HANDSHAKE_MASK = 0x20, - PP_SMC_VOLTAGE_CONTROL_MASK = 0x40, - PP_VBI_TIME_SUPPORT_MASK = 0x80, - PP_ULV_MASK = 0x100, - PP_ENABLE_GFX_CG_THRU_SMU = 0x200, - PP_CLOCK_STRETCH_MASK = 0x400, - PP_OD_FUZZY_FAN_CONTROL_MASK = 0x800, - PP_SOCCLK_DPM_MASK = 0x1000, - PP_DCEFCLK_DPM_MASK = 0x2000, - PP_OVERDRIVE_MASK = 0x4000, -}; - enum PHM_BackEnd_Magic { PHM_Dummy_Magic = 0xAA5555AA, PHM_RV770_Magic = 0xDCBAABCD, @@ -245,6 +229,8 @@ struct pp_hwmgr_func { struct pp_power_state *prequest_ps, const struct pp_power_state *pcurrent_ps); + int (*apply_clocks_adjust_rules)(struct pp_hwmgr *hwmgr); + int (*force_dpm_level)(struct pp_hwmgr *hw_mgr, enum amd_dpm_forced_level level); @@ -268,6 +254,7 @@ struct pp_hwmgr_func { const void *state); int (*enable_clock_power_gating)(struct pp_hwmgr *hwmgr); int (*notify_smc_display_config_after_ps_adjustment)(struct pp_hwmgr *hwmgr); + int (*pre_display_config_changed)(struct pp_hwmgr *hwmgr); int (*display_config_changed)(struct pp_hwmgr *hwmgr); int (*disable_clock_power_gating)(struct pp_hwmgr *hwmgr); int (*update_clock_gatings)(struct pp_hwmgr *hwmgr, @@ -312,6 +299,7 @@ struct pp_hwmgr_func { int (*display_clock_voltage_request)(struct pp_hwmgr *hwmgr, struct pp_display_clock_request *clock); int (*get_max_high_clocks)(struct pp_hwmgr *hwmgr, struct amd_pp_simple_clock_info *clocks); + int (*gfx_off_control)(struct pp_hwmgr *hwmgr, bool enable); int (*power_off_asic)(struct pp_hwmgr *hwmgr); int (*force_clock_level)(struct pp_hwmgr *hwmgr, enum pp_clock_type type, uint32_t mask); int (*print_clock_levels)(struct pp_hwmgr *hwmgr, enum pp_clock_type type, char *buf); @@ -341,6 +329,7 @@ struct pp_hwmgr_func { long *input, uint32_t size); int (*set_power_limit)(struct pp_hwmgr *hwmgr, uint32_t n); int (*set_mmhub_powergating_by_smu)(struct pp_hwmgr *hwmgr); + int (*smus_notify_pwe)(struct pp_hwmgr *hwmgr); }; struct pp_table_func { @@ -718,6 +707,7 @@ struct pp_hwmgr { uint32_t chip_family; uint32_t chip_id; uint32_t smu_version; + bool not_vf; bool pm_en; struct mutex smu_lock; @@ -764,7 +754,7 @@ struct pp_hwmgr { struct pp_power_state *request_ps; struct pp_power_state *boot_ps; struct pp_power_state *uvd_ps; - struct amd_pp_display_configuration display_config; + const struct amd_pp_display_configuration *display_config; uint32_t feature_mask; bool avfs_supported; /* UMD Pstate */ @@ -782,10 +772,13 @@ struct pp_hwmgr { }; int hwmgr_early_init(struct pp_hwmgr *hwmgr); +int hwmgr_sw_init(struct pp_hwmgr *hwmgr); +int hwmgr_sw_fini(struct pp_hwmgr *hwmgr); int hwmgr_hw_init(struct pp_hwmgr *hwmgr); int hwmgr_hw_fini(struct pp_hwmgr *hwmgr); -int hwmgr_hw_suspend(struct pp_hwmgr *hwmgr); -int hwmgr_hw_resume(struct pp_hwmgr *hwmgr); +int hwmgr_suspend(struct pp_hwmgr *hwmgr); +int hwmgr_resume(struct pp_hwmgr *hwmgr); + int hwmgr_handle_task(struct pp_hwmgr *hwmgr, enum amd_pp_task task_id, enum amd_pm_state_type *user_state); diff --git a/drivers/gpu/drm/amd/powerplay/inc/rv_ppsmc.h b/drivers/gpu/drm/amd/powerplay/inc/rv_ppsmc.h index 426bff2aad2b1..a2991fa2e6f8e 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/rv_ppsmc.h +++ b/drivers/gpu/drm/amd/powerplay/inc/rv_ppsmc.h @@ -75,13 +75,15 @@ #define PPSMC_MSG_GetMinGfxclkFrequency 0x2C #define PPSMC_MSG_GetMaxGfxclkFrequency 0x2D #define PPSMC_MSG_SoftReset 0x2E +#define PPSMC_MSG_SetGfxCGPG 0x2F #define PPSMC_MSG_SetSoftMaxGfxClk 0x30 #define PPSMC_MSG_SetHardMinGfxClk 0x31 #define PPSMC_MSG_SetSoftMaxSocclkByFreq 0x32 #define PPSMC_MSG_SetSoftMaxFclkByFreq 0x33 #define PPSMC_MSG_SetSoftMaxVcn 0x34 #define PPSMC_MSG_PowerGateMmHub 0x35 -#define PPSMC_Message_Count 0x36 +#define PPSMC_MSG_SetRccPfcPmeRestoreRegister 0x36 +#define PPSMC_Message_Count 0x37 typedef uint16_t PPSMC_Result; diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu75.h b/drivers/gpu/drm/amd/powerplay/inc/smu75.h new file mode 100644 index 0000000000000..7715230015330 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/inc/smu75.h @@ -0,0 +1,760 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#ifndef SMU75_H +#define SMU75_H + +#pragma pack(push, 1) + +typedef struct { + uint32_t high; + uint32_t low; +} data_64_t; + +typedef struct { + data_64_t high; + data_64_t low; +} data_128_t; + +#define SMU__DGPU_ONLY + +#define SMU__NUM_SCLK_DPM_STATE 8 +#define SMU__NUM_MCLK_DPM_LEVELS 4 +#define SMU__NUM_LCLK_DPM_LEVELS 8 +#define SMU__NUM_PCIE_DPM_LEVELS 8 + +#define SMU7_CONTEXT_ID_SMC 1 +#define SMU7_CONTEXT_ID_VBIOS 2 + +#define SMU75_MAX_LEVELS_VDDC 16 +#define SMU75_MAX_LEVELS_VDDGFX 16 +#define SMU75_MAX_LEVELS_VDDCI 8 +#define SMU75_MAX_LEVELS_MVDD 4 + +#define SMU_MAX_SMIO_LEVELS 4 + +#define SMU75_MAX_LEVELS_GRAPHICS SMU__NUM_SCLK_DPM_STATE +#define SMU75_MAX_LEVELS_MEMORY SMU__NUM_MCLK_DPM_LEVELS +#define SMU75_MAX_LEVELS_GIO SMU__NUM_LCLK_DPM_LEVELS +#define SMU75_MAX_LEVELS_LINK SMU__NUM_PCIE_DPM_LEVELS +#define SMU75_MAX_LEVELS_UVD 8 +#define SMU75_MAX_LEVELS_VCE 8 +#define SMU75_MAX_LEVELS_ACP 8 +#define SMU75_MAX_LEVELS_SAMU 8 +#define SMU75_MAX_ENTRIES_SMIO 32 + +#define DPM_NO_LIMIT 0 +#define DPM_NO_UP 1 +#define DPM_GO_DOWN 2 +#define DPM_GO_UP 3 + +#define SMU7_FIRST_DPM_GRAPHICS_LEVEL 0 +#define SMU7_FIRST_DPM_MEMORY_LEVEL 0 + +#define GPIO_CLAMP_MODE_VRHOT 1 +#define GPIO_CLAMP_MODE_THERM 2 +#define GPIO_CLAMP_MODE_DC 4 + +#define SCRATCH_B_TARG_PCIE_INDEX_SHIFT 0 +#define SCRATCH_B_TARG_PCIE_INDEX_MASK (0x7<<SCRATCH_B_TARG_PCIE_INDEX_SHIFT) +#define SCRATCH_B_CURR_PCIE_INDEX_SHIFT 3 +#define SCRATCH_B_CURR_PCIE_INDEX_MASK (0x7<<SCRATCH_B_CURR_PCIE_INDEX_SHIFT) +#define SCRATCH_B_TARG_UVD_INDEX_SHIFT 6 +#define SCRATCH_B_TARG_UVD_INDEX_MASK (0x7<<SCRATCH_B_TARG_UVD_INDEX_SHIFT) +#define SCRATCH_B_CURR_UVD_INDEX_SHIFT 9 +#define SCRATCH_B_CURR_UVD_INDEX_MASK (0x7<<SCRATCH_B_CURR_UVD_INDEX_SHIFT) +#define SCRATCH_B_TARG_VCE_INDEX_SHIFT 12 +#define SCRATCH_B_TARG_VCE_INDEX_MASK (0x7<<SCRATCH_B_TARG_VCE_INDEX_SHIFT) +#define SCRATCH_B_CURR_VCE_INDEX_SHIFT 15 +#define SCRATCH_B_CURR_VCE_INDEX_MASK (0x7<<SCRATCH_B_CURR_VCE_INDEX_SHIFT) +#define SCRATCH_B_TARG_ACP_INDEX_SHIFT 18 +#define SCRATCH_B_TARG_ACP_INDEX_MASK (0x7<<SCRATCH_B_TARG_ACP_INDEX_SHIFT) +#define SCRATCH_B_CURR_ACP_INDEX_SHIFT 21 +#define SCRATCH_B_CURR_ACP_INDEX_MASK (0x7<<SCRATCH_B_CURR_ACP_INDEX_SHIFT) +#define SCRATCH_B_TARG_SAMU_INDEX_SHIFT 24 +#define SCRATCH_B_TARG_SAMU_INDEX_MASK (0x7<<SCRATCH_B_TARG_SAMU_INDEX_SHIFT) +#define SCRATCH_B_CURR_SAMU_INDEX_SHIFT 27 +#define SCRATCH_B_CURR_SAMU_INDEX_MASK (0x7<<SCRATCH_B_CURR_SAMU_INDEX_SHIFT) + +/* Virtualization Defines */ +#define CG_XDMA_MASK 0x1 +#define CG_XDMA_SHIFT 0 +#define CG_UVD_MASK 0x2 +#define CG_UVD_SHIFT 1 +#define CG_VCE_MASK 0x4 +#define CG_VCE_SHIFT 2 +#define CG_SAMU_MASK 0x8 +#define CG_SAMU_SHIFT 3 +#define CG_GFX_MASK 0x10 +#define CG_GFX_SHIFT 4 +#define CG_SDMA_MASK 0x20 +#define CG_SDMA_SHIFT 5 +#define CG_HDP_MASK 0x40 +#define CG_HDP_SHIFT 6 +#define CG_MC_MASK 0x80 +#define CG_MC_SHIFT 7 +#define CG_DRM_MASK 0x100 +#define CG_DRM_SHIFT 8 +#define CG_ROM_MASK 0x200 +#define CG_ROM_SHIFT 9 +#define CG_BIF_MASK 0x400 +#define CG_BIF_SHIFT 10 + +#if defined SMU__DGPU_ONLY +#define SMU75_DTE_ITERATIONS 5 +#define SMU75_DTE_SOURCES 3 +#define SMU75_DTE_SINKS 1 +#define SMU75_NUM_CPU_TES 0 +#define SMU75_NUM_GPU_TES 1 +#define SMU75_NUM_NON_TES 2 +#define SMU75_DTE_FAN_SCALAR_MIN 0x100 +#define SMU75_DTE_FAN_SCALAR_MAX 0x166 +#define SMU75_DTE_FAN_TEMP_MAX 93 +#define SMU75_DTE_FAN_TEMP_MIN 83 +#endif +#define SMU75_THERMAL_INPUT_LOOP_COUNT 2 +#define SMU75_THERMAL_CLAMP_MODE_COUNT 2 + +#define EXP_M1_1 93 +#define EXP_M2_1 195759 +#define EXP_B_1 111176531 + +#define EXP_M1_2 67 +#define EXP_M2_2 153720 +#define EXP_B_2 94415767 + +#define EXP_M1_3 48 +#define EXP_M2_3 119796 +#define EXP_B_3 79195279 + +#define EXP_M1_4 550 +#define EXP_M2_4 1484190 +#define EXP_B_4 1051432828 + +#define EXP_M1_5 394 +#define EXP_M2_5 1143049 +#define EXP_B_5 864288432 + +struct SMU7_HystController_Data { + uint16_t waterfall_up; + uint16_t waterfall_down; + uint16_t waterfall_limit; + uint16_t release_cnt; + uint16_t release_limit; + uint16_t spare; +}; + +typedef struct SMU7_HystController_Data SMU7_HystController_Data; + +struct SMU75_PIDController { + uint32_t Ki; + int32_t LFWindupUpperLim; + int32_t LFWindupLowerLim; + uint32_t StatePrecision; + uint32_t LfPrecision; + uint32_t LfOffset; + uint32_t MaxState; + uint32_t MaxLfFraction; + uint32_t StateShift; +}; + +typedef struct SMU75_PIDController SMU75_PIDController; + +struct SMU7_LocalDpmScoreboard { + uint32_t PercentageBusy; + + int32_t PIDError; + int32_t PIDIntegral; + int32_t PIDOutput; + + uint32_t SigmaDeltaAccum; + uint32_t SigmaDeltaOutput; + uint32_t SigmaDeltaLevel; + + uint32_t UtilizationSetpoint; + + uint8_t TdpClampMode; + uint8_t TdcClampMode; + uint8_t ThermClampMode; + uint8_t VoltageBusy; + + int8_t CurrLevel; + int8_t TargLevel; + uint8_t LevelChangeInProgress; + uint8_t UpHyst; + + uint8_t DownHyst; + uint8_t VoltageDownHyst; + uint8_t DpmEnable; + uint8_t DpmRunning; + + uint8_t DpmForce; + uint8_t DpmForceLevel; + uint8_t DisplayWatermark; + uint8_t McArbIndex; + + uint32_t MinimumPerfSclk; + + uint8_t AcpiReq; + uint8_t AcpiAck; + uint8_t GfxClkSlow; + uint8_t GpioClampMode; + + uint8_t EnableModeSwitchRLCNotification; + uint8_t EnabledLevelsChange; + uint8_t DteClampMode; + uint8_t FpsClampMode; + + uint16_t LevelResidencyCounters [SMU75_MAX_LEVELS_GRAPHICS]; + uint16_t LevelSwitchCounters [SMU75_MAX_LEVELS_GRAPHICS]; + + void (*TargetStateCalculator)(uint8_t); + void (*SavedTargetStateCalculator)(uint8_t); + + uint16_t AutoDpmInterval; + uint16_t AutoDpmRange; + + uint8_t FpsEnabled; + uint8_t MaxPerfLevel; + uint8_t AllowLowClkInterruptToHost; + uint8_t FpsRunning; + + uint32_t MaxAllowedFrequency; + + uint32_t FilteredSclkFrequency; + uint32_t LastSclkFrequency; + uint32_t FilteredSclkFrequencyCnt; + + uint8_t MinPerfLevel; +#ifdef SMU__FIRMWARE_SCKS_PRESENT__1 + uint8_t ScksClampMode; + uint8_t padding[2]; +#else + uint8_t padding[3]; +#endif + + uint16_t FpsAlpha; + uint16_t DeltaTime; + uint32_t CurrentFps; + uint32_t FilteredFps; + uint32_t FrameCount; + uint32_t FrameCountLast; + uint16_t FpsTargetScalar; + uint16_t FpsWaterfallLimitScalar; + uint16_t FpsAlphaScalar; + uint16_t spare8; + SMU7_HystController_Data HystControllerData; +}; + +typedef struct SMU7_LocalDpmScoreboard SMU7_LocalDpmScoreboard; + +#define SMU7_MAX_VOLTAGE_CLIENTS 12 + +typedef uint8_t (*VoltageChangeHandler_t)(uint16_t, uint8_t); + +#define VDDC_MASK 0x00007FFF +#define VDDC_SHIFT 0 +#define VDDCI_MASK 0x3FFF8000 +#define VDDCI_SHIFT 15 +#define PHASES_MASK 0xC0000000 +#define PHASES_SHIFT 30 + +typedef uint32_t SMU_VoltageLevel; + +struct SMU7_VoltageScoreboard { + SMU_VoltageLevel TargetVoltage; + uint16_t MaxVid; + uint8_t HighestVidOffset; + uint8_t CurrentVidOffset; + + uint16_t CurrentVddc; + uint16_t CurrentVddci; + + uint8_t ControllerBusy; + uint8_t CurrentVid; + uint8_t CurrentVddciVid; + uint8_t padding; + + SMU_VoltageLevel RequestedVoltage[SMU7_MAX_VOLTAGE_CLIENTS]; + SMU_VoltageLevel TargetVoltageState; + uint8_t EnabledRequest[SMU7_MAX_VOLTAGE_CLIENTS]; + + uint8_t padding2; + uint8_t padding3; + uint8_t ControllerEnable; + uint8_t ControllerRunning; + uint16_t CurrentStdVoltageHiSidd; + uint16_t CurrentStdVoltageLoSidd; + uint8_t OverrideVoltage; + uint8_t padding4; + uint8_t padding5; + uint8_t CurrentPhases; + + VoltageChangeHandler_t ChangeVddc; + VoltageChangeHandler_t ChangeVddci; + VoltageChangeHandler_t ChangePhase; + VoltageChangeHandler_t ChangeMvdd; + + VoltageChangeHandler_t functionLinks[6]; + + uint16_t * VddcFollower1; + int16_t Driver_OD_RequestedVidOffset1; + int16_t Driver_OD_RequestedVidOffset2; +}; + +typedef struct SMU7_VoltageScoreboard SMU7_VoltageScoreboard; + +#define SMU7_MAX_PCIE_LINK_SPEEDS 3 + +struct SMU7_PCIeLinkSpeedScoreboard { + uint8_t DpmEnable; + uint8_t DpmRunning; + uint8_t DpmForce; + uint8_t DpmForceLevel; + + uint8_t CurrentLinkSpeed; + uint8_t EnabledLevelsChange; + uint16_t AutoDpmInterval; + + uint16_t AutoDpmRange; + uint16_t AutoDpmCount; + + uint8_t DpmMode; + uint8_t AcpiReq; + uint8_t AcpiAck; + uint8_t CurrentLinkLevel; +}; + +typedef struct SMU7_PCIeLinkSpeedScoreboard SMU7_PCIeLinkSpeedScoreboard; + +#define SMU7_LKGE_LUT_NUM_OF_TEMP_ENTRIES 16 +#define SMU7_LKGE_LUT_NUM_OF_VOLT_ENTRIES 16 + +#define SMU7_SCALE_I 7 +#define SMU7_SCALE_R 12 + +struct SMU7_PowerScoreboard { + uint32_t GpuPower; + + uint32_t VddcPower; + uint32_t VddcVoltage; + uint32_t VddcCurrent; + + uint32_t VddciPower; + uint32_t VddciVoltage; + uint32_t VddciCurrent; + + uint32_t RocPower; + + uint16_t Telemetry_1_slope; + uint16_t Telemetry_2_slope; + int32_t Telemetry_1_offset; + int32_t Telemetry_2_offset; + + uint8_t MCLK_patch_flag; + uint8_t reserved[3]; +}; + +typedef struct SMU7_PowerScoreboard SMU7_PowerScoreboard; + +#define SMU7_SCLK_DPM_CONFIG_MASK 0x01 +#define SMU7_VOLTAGE_CONTROLLER_CONFIG_MASK 0x02 +#define SMU7_THERMAL_CONTROLLER_CONFIG_MASK 0x04 +#define SMU7_MCLK_DPM_CONFIG_MASK 0x08 +#define SMU7_UVD_DPM_CONFIG_MASK 0x10 +#define SMU7_VCE_DPM_CONFIG_MASK 0x20 +#define SMU7_ACP_DPM_CONFIG_MASK 0x40 +#define SMU7_SAMU_DPM_CONFIG_MASK 0x80 +#define SMU7_PCIEGEN_DPM_CONFIG_MASK 0x100 + +#define SMU7_ACP_MCLK_HANDSHAKE_DISABLE 0x00000001 +#define SMU7_ACP_SCLK_HANDSHAKE_DISABLE 0x00000002 +#define SMU7_UVD_MCLK_HANDSHAKE_DISABLE 0x00000100 +#define SMU7_UVD_SCLK_HANDSHAKE_DISABLE 0x00000200 +#define SMU7_VCE_MCLK_HANDSHAKE_DISABLE 0x00010000 +#define SMU7_VCE_SCLK_HANDSHAKE_DISABLE 0x00020000 + +struct SMU75_SoftRegisters { + uint32_t RefClockFrequency; + uint32_t PmTimerPeriod; + uint32_t FeatureEnables; +#if defined (SMU__DGPU_ONLY) + uint32_t PreVBlankGap; + uint32_t VBlankTimeout; + uint32_t TrainTimeGap; + uint32_t MvddSwitchTime; + uint32_t LongestAcpiTrainTime; + uint32_t AcpiDelay; + uint32_t G5TrainTime; + uint32_t DelayMpllPwron; + uint32_t VoltageChangeTimeout; +#endif + uint32_t HandshakeDisables; + + uint8_t DisplayPhy1Config; + uint8_t DisplayPhy2Config; + uint8_t DisplayPhy3Config; + uint8_t DisplayPhy4Config; + + uint8_t DisplayPhy5Config; + uint8_t DisplayPhy6Config; + uint8_t DisplayPhy7Config; + uint8_t DisplayPhy8Config; + + uint32_t AverageGraphicsActivity; + uint32_t AverageMemoryActivity; + uint32_t AverageGioActivity; + + uint8_t SClkDpmEnabledLevels; + uint8_t MClkDpmEnabledLevels; + uint8_t LClkDpmEnabledLevels; + uint8_t PCIeDpmEnabledLevels; + + uint8_t UVDDpmEnabledLevels; + uint8_t SAMUDpmEnabledLevels; + uint8_t ACPDpmEnabledLevels; + uint8_t VCEDpmEnabledLevels; + + uint32_t DRAM_LOG_ADDR_H; + uint32_t DRAM_LOG_ADDR_L; + uint32_t DRAM_LOG_PHY_ADDR_H; + uint32_t DRAM_LOG_PHY_ADDR_L; + uint32_t DRAM_LOG_BUFF_SIZE; + uint32_t UlvEnterCount; + uint32_t UlvTime; + uint32_t UcodeLoadStatus; + uint32_t AllowMvddSwitch; + uint8_t Activity_Weight; + uint8_t Reserved8[3]; +}; + +typedef struct SMU75_SoftRegisters SMU75_SoftRegisters; + +struct SMU75_Firmware_Header { + uint32_t Digest[5]; + uint32_t Version; + uint32_t HeaderSize; + uint32_t Flags; + uint32_t EntryPoint; + uint32_t CodeSize; + uint32_t ImageSize; + + uint32_t Rtos; + uint32_t SoftRegisters; + uint32_t DpmTable; + uint32_t FanTable; + uint32_t CacConfigTable; + uint32_t CacStatusTable; + uint32_t mcRegisterTable; + uint32_t mcArbDramTimingTable; + uint32_t PmFuseTable; + uint32_t Globals; + uint32_t ClockStretcherTable; + uint32_t VftTable; + uint32_t Reserved1; + uint32_t AvfsCksOff_AvfsGbvTable; + uint32_t AvfsCksOff_BtcGbvTable; + uint32_t MM_AvfsTable; + uint32_t PowerSharingTable; + uint32_t AvfsTable; + uint32_t AvfsCksOffGbvTable; + uint32_t AvfsMeanNSigma; + uint32_t AvfsSclkOffsetTable; + uint32_t Reserved[12]; + uint32_t Signature; +}; + +typedef struct SMU75_Firmware_Header SMU75_Firmware_Header; + +#define SMU7_FIRMWARE_HEADER_LOCATION 0x20000 + +enum DisplayConfig { + PowerDown = 1, + DP54x4, + DP54x2, + DP54x1, + DP27x4, + DP27x2, + DP27x1, + HDMI297, + HDMI162, + LVDS, + DP324x4, + DP324x2, + DP324x1 +}; + +#define MC_BLOCK_COUNT 1 +#define CPL_BLOCK_COUNT 5 +#define SE_BLOCK_COUNT 15 +#define GC_BLOCK_COUNT 24 + +struct SMU7_Local_Cac { + uint8_t BlockId; + uint8_t SignalId; + uint8_t Threshold; + uint8_t Padding; +}; + +typedef struct SMU7_Local_Cac SMU7_Local_Cac; + +struct SMU7_Local_Cac_Table { + SMU7_Local_Cac CplLocalCac[CPL_BLOCK_COUNT]; + SMU7_Local_Cac McLocalCac[MC_BLOCK_COUNT]; + SMU7_Local_Cac SeLocalCac[SE_BLOCK_COUNT]; + SMU7_Local_Cac GcLocalCac[GC_BLOCK_COUNT]; +}; + +typedef struct SMU7_Local_Cac_Table SMU7_Local_Cac_Table; + +#pragma pack(pop) + +#define CG_SYS_BITMASK_FIRST_BIT 0 +#define CG_SYS_BITMASK_LAST_BIT 10 +#define CG_SYS_BIF_MGLS_SHIFT 0 +#define CG_SYS_ROM_SHIFT 1 +#define CG_SYS_MC_MGCG_SHIFT 2 +#define CG_SYS_MC_MGLS_SHIFT 3 +#define CG_SYS_SDMA_MGCG_SHIFT 4 +#define CG_SYS_SDMA_MGLS_SHIFT 5 +#define CG_SYS_DRM_MGCG_SHIFT 6 +#define CG_SYS_HDP_MGCG_SHIFT 7 +#define CG_SYS_HDP_MGLS_SHIFT 8 +#define CG_SYS_DRM_MGLS_SHIFT 9 +#define CG_SYS_BIF_MGCG_SHIFT 10 + +#define CG_SYS_BIF_MGLS_MASK 0x1 +#define CG_SYS_ROM_MASK 0x2 +#define CG_SYS_MC_MGCG_MASK 0x4 +#define CG_SYS_MC_MGLS_MASK 0x8 +#define CG_SYS_SDMA_MGCG_MASK 0x10 +#define CG_SYS_SDMA_MGLS_MASK 0x20 +#define CG_SYS_DRM_MGCG_MASK 0x40 +#define CG_SYS_HDP_MGCG_MASK 0x80 +#define CG_SYS_HDP_MGLS_MASK 0x100 +#define CG_SYS_DRM_MGLS_MASK 0x200 +#define CG_SYS_BIF_MGCG_MASK 0x400 + +#define CG_GFX_BITMASK_FIRST_BIT 16 +#define CG_GFX_BITMASK_LAST_BIT 24 + +#define CG_GFX_CGCG_SHIFT 16 +#define CG_GFX_CGLS_SHIFT 17 +#define CG_CPF_MGCG_SHIFT 18 +#define CG_RLC_MGCG_SHIFT 19 +#define CG_GFX_OTHERS_MGCG_SHIFT 20 +#define CG_GFX_3DCG_SHIFT 21 +#define CG_GFX_3DLS_SHIFT 22 +#define CG_GFX_RLC_LS_SHIFT 23 +#define CG_GFX_CP_LS_SHIFT 24 + +#define CG_GFX_CGCG_MASK 0x00010000 +#define CG_GFX_CGLS_MASK 0x00020000 +#define CG_CPF_MGCG_MASK 0x00040000 +#define CG_RLC_MGCG_MASK 0x00080000 +#define CG_GFX_OTHERS_MGCG_MASK 0x00100000 +#define CG_GFX_3DCG_MASK 0x00200000 +#define CG_GFX_3DLS_MASK 0x00400000 +#define CG_GFX_RLC_LS_MASK 0x00800000 +#define CG_GFX_CP_LS_MASK 0x01000000 + + +#define VRCONF_VDDC_MASK 0x000000FF +#define VRCONF_VDDC_SHIFT 0 +#define VRCONF_VDDGFX_MASK 0x0000FF00 +#define VRCONF_VDDGFX_SHIFT 8 +#define VRCONF_VDDCI_MASK 0x00FF0000 +#define VRCONF_VDDCI_SHIFT 16 +#define VRCONF_MVDD_MASK 0xFF000000 +#define VRCONF_MVDD_SHIFT 24 + +#define VR_MERGED_WITH_VDDC 0 +#define VR_SVI2_PLANE_1 1 +#define VR_SVI2_PLANE_2 2 +#define VR_SMIO_PATTERN_1 3 +#define VR_SMIO_PATTERN_2 4 +#define VR_STATIC_VOLTAGE 5 + +#define CLOCK_STRETCHER_MAX_ENTRIES 0x4 +#define CKS_LOOKUPTable_MAX_ENTRIES 0x4 + +#define CLOCK_STRETCHER_SETTING_DDT_MASK 0x01 +#define CLOCK_STRETCHER_SETTING_DDT_SHIFT 0x0 +#define CLOCK_STRETCHER_SETTING_STRETCH_AMOUNT_MASK 0x1E +#define CLOCK_STRETCHER_SETTING_STRETCH_AMOUNT_SHIFT 0x1 +#define CLOCK_STRETCHER_SETTING_ENABLE_MASK 0x80 +#define CLOCK_STRETCHER_SETTING_ENABLE_SHIFT 0x7 + +struct SMU_ClockStretcherDataTableEntry { + uint8_t minVID; + uint8_t maxVID; + + uint16_t setting; +}; +typedef struct SMU_ClockStretcherDataTableEntry SMU_ClockStretcherDataTableEntry; + +struct SMU_ClockStretcherDataTable { + SMU_ClockStretcherDataTableEntry ClockStretcherDataTableEntry[CLOCK_STRETCHER_MAX_ENTRIES]; +}; +typedef struct SMU_ClockStretcherDataTable SMU_ClockStretcherDataTable; + +struct SMU_CKS_LOOKUPTableEntry { + uint16_t minFreq; + uint16_t maxFreq; + + uint8_t setting; + uint8_t padding[3]; +}; +typedef struct SMU_CKS_LOOKUPTableEntry SMU_CKS_LOOKUPTableEntry; + +struct SMU_CKS_LOOKUPTable { + SMU_CKS_LOOKUPTableEntry CKS_LOOKUPTableEntry[CKS_LOOKUPTable_MAX_ENTRIES]; +}; +typedef struct SMU_CKS_LOOKUPTable SMU_CKS_LOOKUPTable; + +struct AgmAvfsData_t { + uint16_t avgPsmCount[28]; + uint16_t minPsmCount[28]; +}; +typedef struct AgmAvfsData_t AgmAvfsData_t; + +enum VFT_COLUMNS { + SCLK0, + SCLK1, + SCLK2, + SCLK3, + SCLK4, + SCLK5, + SCLK6, + SCLK7, + + NUM_VFT_COLUMNS +}; +enum { + SCS_FUSE_T0, + SCS_FUSE_T1, + NUM_SCS_FUSE_TEMPERATURE +}; +enum { + SCKS_ON, + SCKS_OFF, + NUM_SCKS_STATE_TYPES +}; + +#define VFT_TABLE_DEFINED + +#define TEMP_RANGE_MAXSTEPS 12 +struct VFT_CELL_t { + uint16_t Voltage; +}; + +typedef struct VFT_CELL_t VFT_CELL_t; +#ifdef SMU__FIRMWARE_SCKS_PRESENT__1 +struct SCS_CELL_t { + uint16_t PsmCnt[NUM_SCKS_STATE_TYPES]; +}; +typedef struct SCS_CELL_t SCS_CELL_t; +#endif + +struct VFT_TABLE_t { + VFT_CELL_t Cell[TEMP_RANGE_MAXSTEPS][NUM_VFT_COLUMNS]; + uint16_t AvfsGbv [NUM_VFT_COLUMNS]; + uint16_t BtcGbv [NUM_VFT_COLUMNS]; + int16_t Temperature [TEMP_RANGE_MAXSTEPS]; + +#ifdef SMU__FIRMWARE_SCKS_PRESENT__1 + SCS_CELL_t ScksCell[TEMP_RANGE_MAXSTEPS][NUM_VFT_COLUMNS]; +#endif + + uint8_t NumTemperatureSteps; + uint8_t padding[3]; +}; +typedef struct VFT_TABLE_t VFT_TABLE_t; + +#define BTCGB_VDROOP_TABLE_MAX_ENTRIES 2 +#define AVFSGB_VDROOP_TABLE_MAX_ENTRIES 2 + +struct GB_VDROOP_TABLE_t { + int32_t a0; + int32_t a1; + int32_t a2; + uint32_t spare; +}; +typedef struct GB_VDROOP_TABLE_t GB_VDROOP_TABLE_t; + +struct SMU_QuadraticCoeffs { + int32_t m1; + int32_t b; + + int16_t m2; + uint8_t m1_shift; + uint8_t m2_shift; +}; +typedef struct SMU_QuadraticCoeffs SMU_QuadraticCoeffs; + +struct AVFS_Margin_t { + VFT_CELL_t Cell[NUM_VFT_COLUMNS]; +}; +typedef struct AVFS_Margin_t AVFS_Margin_t; + +struct AVFS_CksOff_Gbv_t { + VFT_CELL_t Cell[NUM_VFT_COLUMNS]; +}; +typedef struct AVFS_CksOff_Gbv_t AVFS_CksOff_Gbv_t; + +struct AVFS_CksOff_AvfsGbv_t { + VFT_CELL_t Cell[NUM_VFT_COLUMNS]; +}; +typedef struct AVFS_CksOff_AvfsGbv_t AVFS_CksOff_AvfsGbv_t; + +struct AVFS_CksOff_BtcGbv_t { + VFT_CELL_t Cell[NUM_VFT_COLUMNS]; +}; +typedef struct AVFS_CksOff_BtcGbv_t AVFS_CksOff_BtcGbv_t; + +struct AVFS_meanNsigma_t { + uint32_t Aconstant[3]; + uint16_t DC_tol_sigma; + uint16_t Platform_mean; + uint16_t Platform_sigma; + uint16_t PSM_Age_CompFactor; + uint8_t Static_Voltage_Offset[NUM_VFT_COLUMNS]; +}; +typedef struct AVFS_meanNsigma_t AVFS_meanNsigma_t; + +struct AVFS_Sclk_Offset_t { + uint16_t Sclk_Offset[8]; +}; +typedef struct AVFS_Sclk_Offset_t AVFS_Sclk_Offset_t; + +struct Power_Sharing_t { + uint32_t EnergyCounter; + uint32_t EngeryThreshold; + uint64_t AM_SCLK_CNT; + uint64_t AM_0_BUSY_CNT; +}; +typedef struct Power_Sharing_t Power_Sharing_t; + + +#endif + + diff --git a/drivers/gpu/drm/amd/powerplay/inc/smu75_discrete.h b/drivers/gpu/drm/amd/powerplay/inc/smu75_discrete.h new file mode 100644 index 0000000000000..b64e58a22ddf6 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/inc/smu75_discrete.h @@ -0,0 +1,886 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef SMU75_DISCRETE_H +#define SMU75_DISCRETE_H + +#include "smu75.h" + +#pragma pack(push, 1) + +#define NUM_SCLK_RANGE 8 + +#define VCO_3_6 1 +#define VCO_2_4 3 + +#define POSTDIV_DIV_BY_1 0 +#define POSTDIV_DIV_BY_2 1 +#define POSTDIV_DIV_BY_4 2 +#define POSTDIV_DIV_BY_8 3 +#define POSTDIV_DIV_BY_16 4 + +struct sclkFcwRange_t { + uint8_t vco_setting; /* 1: 3-6GHz, 3: 2-4GHz */ + uint8_t postdiv; /* divide by 2^n */ + uint16_t fcw_pcc; + uint16_t fcw_trans_upper; + uint16_t fcw_trans_lower; +}; +typedef struct sclkFcwRange_t sclkFcwRange_t; + +struct SMIO_Pattern { + uint16_t Voltage; + uint8_t Smio; + uint8_t padding; +}; + +typedef struct SMIO_Pattern SMIO_Pattern; + +struct SMIO_Table { + SMIO_Pattern Pattern[SMU_MAX_SMIO_LEVELS]; +}; + +typedef struct SMIO_Table SMIO_Table; + +struct SMU_SclkSetting { + uint32_t SclkFrequency; + uint16_t Fcw_int; + uint16_t Fcw_frac; + uint16_t Pcc_fcw_int; + uint8_t PllRange; + uint8_t SSc_En; + uint16_t Sclk_slew_rate; + uint16_t Pcc_up_slew_rate; + uint16_t Pcc_down_slew_rate; + uint16_t Fcw1_int; + uint16_t Fcw1_frac; + uint16_t Sclk_ss_slew_rate; +}; +typedef struct SMU_SclkSetting SMU_SclkSetting; + +struct SMU75_Discrete_GraphicsLevel { + SMU_VoltageLevel MinVoltage; + + uint8_t pcieDpmLevel; + uint8_t DeepSleepDivId; + uint16_t ActivityLevel; + + uint32_t CgSpllFuncCntl3; + uint32_t CgSpllFuncCntl4; + uint32_t CcPwrDynRm; + uint32_t CcPwrDynRm1; + + uint8_t SclkDid; + uint8_t padding; + uint8_t EnabledForActivity; + uint8_t EnabledForThrottle; + uint8_t UpHyst; + uint8_t DownHyst; + uint8_t VoltageDownHyst; + uint8_t PowerThrottle; + + SMU_SclkSetting SclkSetting; + + uint8_t ScksStretchThreshVid[NUM_SCKS_STATE_TYPES]; + uint16_t Padding; +}; + +typedef struct SMU75_Discrete_GraphicsLevel SMU75_Discrete_GraphicsLevel; + +struct SMU75_Discrete_ACPILevel { + uint32_t Flags; + SMU_VoltageLevel MinVoltage; + uint32_t SclkFrequency; + uint8_t SclkDid; + uint8_t DisplayWatermark; + uint8_t DeepSleepDivId; + uint8_t padding; + uint32_t CcPwrDynRm; + uint32_t CcPwrDynRm1; + + SMU_SclkSetting SclkSetting; +}; + +typedef struct SMU75_Discrete_ACPILevel SMU75_Discrete_ACPILevel; + +struct SMU75_Discrete_Ulv { + uint32_t CcPwrDynRm; + uint32_t CcPwrDynRm1; + uint16_t VddcOffset; + uint8_t VddcOffsetVid; + uint8_t VddcPhase; + uint16_t BifSclkDfs; + uint16_t Reserved; +}; + +typedef struct SMU75_Discrete_Ulv SMU75_Discrete_Ulv; + +struct SMU75_Discrete_MemoryLevel { + SMU_VoltageLevel MinVoltage; + uint32_t MinMvdd; + + uint32_t MclkFrequency; + + uint8_t StutterEnable; + uint8_t EnabledForThrottle; + uint8_t EnabledForActivity; + uint8_t padding_0; + + uint8_t UpHyst; + uint8_t DownHyst; + uint8_t VoltageDownHyst; + uint8_t padding_1; + + uint16_t ActivityLevel; + uint8_t DisplayWatermark; + uint8_t padding_2; + + uint16_t Fcw_int; + uint16_t Fcw_frac; + uint8_t Postdiv; + uint8_t padding_3[3]; +}; + +typedef struct SMU75_Discrete_MemoryLevel SMU75_Discrete_MemoryLevel; + +struct SMU75_Discrete_LinkLevel { + uint8_t PcieGenSpeed; + uint8_t PcieLaneCount; + uint8_t EnabledForActivity; + uint8_t SPC; + uint32_t DownThreshold; + uint32_t UpThreshold; + uint16_t BifSclkDfs; + uint16_t Reserved; +}; + +typedef struct SMU75_Discrete_LinkLevel SMU75_Discrete_LinkLevel; + + +/* MC ARB DRAM Timing registers. */ +struct SMU75_Discrete_MCArbDramTimingTableEntry { + uint32_t McArbDramTiming; + uint32_t McArbDramTiming2; + uint32_t McArbBurstTime; + uint32_t McArbRfshRate; + uint32_t McArbMisc3; +}; + +typedef struct SMU75_Discrete_MCArbDramTimingTableEntry SMU75_Discrete_MCArbDramTimingTableEntry; + +struct SMU75_Discrete_MCArbDramTimingTable { + SMU75_Discrete_MCArbDramTimingTableEntry entries[SMU__NUM_SCLK_DPM_STATE][SMU__NUM_MCLK_DPM_LEVELS]; +}; + +typedef struct SMU75_Discrete_MCArbDramTimingTable SMU75_Discrete_MCArbDramTimingTable; + +/* UVD VCLK/DCLK state (level) definition. */ +struct SMU75_Discrete_UvdLevel { + uint32_t VclkFrequency; + uint32_t DclkFrequency; + SMU_VoltageLevel MinVoltage; + uint8_t VclkDivider; + uint8_t DclkDivider; + uint8_t padding[2]; +}; + +typedef struct SMU75_Discrete_UvdLevel SMU75_Discrete_UvdLevel; + +/* Clocks for other external blocks (VCE, ACP, SAMU). */ +struct SMU75_Discrete_ExtClkLevel { + uint32_t Frequency; + SMU_VoltageLevel MinVoltage; + uint8_t Divider; + uint8_t padding[3]; +}; + +typedef struct SMU75_Discrete_ExtClkLevel SMU75_Discrete_ExtClkLevel; + +struct SMU75_Discrete_StateInfo { + uint32_t SclkFrequency; + uint32_t MclkFrequency; + uint32_t VclkFrequency; + uint32_t DclkFrequency; + uint32_t SamclkFrequency; + uint32_t AclkFrequency; + uint32_t EclkFrequency; + uint16_t MvddVoltage; + uint16_t padding16; + uint8_t DisplayWatermark; + uint8_t McArbIndex; + uint8_t McRegIndex; + uint8_t SeqIndex; + uint8_t SclkDid; + int8_t SclkIndex; + int8_t MclkIndex; + uint8_t PCIeGen; +}; + +typedef struct SMU75_Discrete_StateInfo SMU75_Discrete_StateInfo; + +struct SMU75_Discrete_DpmTable { + SMU75_PIDController GraphicsPIDController; + SMU75_PIDController MemoryPIDController; + SMU75_PIDController LinkPIDController; + + uint32_t SystemFlags; + + uint32_t VRConfig; + uint32_t SmioMask1; + uint32_t SmioMask2; + SMIO_Table SmioTable1; + SMIO_Table SmioTable2; + + uint32_t MvddLevelCount; + + uint8_t BapmVddcVidHiSidd [SMU75_MAX_LEVELS_VDDC]; + uint8_t BapmVddcVidLoSidd [SMU75_MAX_LEVELS_VDDC]; + uint8_t BapmVddcVidHiSidd2 [SMU75_MAX_LEVELS_VDDC]; + + uint8_t GraphicsDpmLevelCount; + uint8_t MemoryDpmLevelCount; + uint8_t LinkLevelCount; + uint8_t MasterDeepSleepControl; + + uint8_t UvdLevelCount; + uint8_t VceLevelCount; + uint8_t AcpLevelCount; + uint8_t SamuLevelCount; + + uint8_t ThermOutGpio; + uint8_t ThermOutPolarity; + uint8_t ThermOutMode; + uint8_t BootPhases; + + uint8_t VRHotLevel; + uint8_t LdoRefSel; + + uint8_t Reserved1[2]; + + uint16_t FanStartTemperature; + uint16_t FanStopTemperature; + + uint16_t MaxVoltage; + uint16_t Reserved2; + uint32_t Reserved; + + SMU75_Discrete_GraphicsLevel GraphicsLevel [SMU75_MAX_LEVELS_GRAPHICS]; + SMU75_Discrete_MemoryLevel MemoryACPILevel; + SMU75_Discrete_MemoryLevel MemoryLevel [SMU75_MAX_LEVELS_MEMORY]; + SMU75_Discrete_LinkLevel LinkLevel [SMU75_MAX_LEVELS_LINK]; + SMU75_Discrete_ACPILevel ACPILevel; + SMU75_Discrete_UvdLevel UvdLevel [SMU75_MAX_LEVELS_UVD]; + SMU75_Discrete_ExtClkLevel VceLevel [SMU75_MAX_LEVELS_VCE]; + SMU75_Discrete_ExtClkLevel AcpLevel [SMU75_MAX_LEVELS_ACP]; + SMU75_Discrete_ExtClkLevel SamuLevel [SMU75_MAX_LEVELS_SAMU]; + SMU75_Discrete_Ulv Ulv; + + uint8_t DisplayWatermark [SMU75_MAX_LEVELS_MEMORY][SMU75_MAX_LEVELS_GRAPHICS]; + + uint32_t SclkStepSize; + uint32_t Smio [SMU75_MAX_ENTRIES_SMIO]; + + uint8_t UvdBootLevel; + uint8_t VceBootLevel; + uint8_t AcpBootLevel; + uint8_t SamuBootLevel; + + uint8_t GraphicsBootLevel; + uint8_t GraphicsVoltageChangeEnable; + uint8_t GraphicsThermThrottleEnable; + uint8_t GraphicsInterval; + + uint8_t VoltageInterval; + uint8_t ThermalInterval; + uint16_t TemperatureLimitHigh; + + uint16_t TemperatureLimitLow; + uint8_t MemoryBootLevel; + uint8_t MemoryVoltageChangeEnable; + + uint16_t BootMVdd; + uint8_t MemoryInterval; + uint8_t MemoryThermThrottleEnable; + + uint16_t VoltageResponseTime; + uint16_t PhaseResponseTime; + + uint8_t PCIeBootLinkLevel; + uint8_t PCIeGenInterval; + uint8_t DTEInterval; + uint8_t DTEMode; + + uint8_t SVI2Enable; + uint8_t VRHotGpio; + uint8_t AcDcGpio; + uint8_t ThermGpio; + + uint16_t PPM_PkgPwrLimit; + uint16_t PPM_TemperatureLimit; + + uint16_t DefaultTdp; + uint16_t TargetTdp; + + uint16_t FpsHighThreshold; + uint16_t FpsLowThreshold; + + uint16_t BAPMTI_R [SMU75_DTE_ITERATIONS][SMU75_DTE_SOURCES][SMU75_DTE_SINKS]; + uint16_t BAPMTI_RC [SMU75_DTE_ITERATIONS][SMU75_DTE_SOURCES][SMU75_DTE_SINKS]; + + uint16_t TemperatureLimitEdge; + uint16_t TemperatureLimitHotspot; + + uint16_t BootVddc; + uint16_t BootVddci; + + uint16_t FanGainEdge; + uint16_t FanGainHotspot; + + uint32_t LowSclkInterruptThreshold; + uint32_t VddGfxReChkWait; + + uint8_t ClockStretcherAmount; + uint8_t Sclk_CKS_masterEn0_7; + uint8_t Sclk_CKS_masterEn8_15; + uint8_t DPMFreezeAndForced; + + uint8_t Sclk_voltageOffset[8]; + + SMU_ClockStretcherDataTable ClockStretcherDataTable; + SMU_CKS_LOOKUPTable CKS_LOOKUPTable; + + uint32_t CurrSclkPllRange; + sclkFcwRange_t SclkFcwRangeTable[NUM_SCLK_RANGE]; + + GB_VDROOP_TABLE_t BTCGB_VDROOP_TABLE[BTCGB_VDROOP_TABLE_MAX_ENTRIES]; + SMU_QuadraticCoeffs AVFSGB_FUSE_TABLE[AVFSGB_VDROOP_TABLE_MAX_ENTRIES]; +}; + +typedef struct SMU75_Discrete_DpmTable SMU75_Discrete_DpmTable; + +struct SMU75_Discrete_FanTable { + uint16_t FdoMode; + int16_t TempMin; + int16_t TempMed; + int16_t TempMax; + int16_t Slope1; + int16_t Slope2; + int16_t FdoMin; + int16_t HystUp; + int16_t HystDown; + int16_t HystSlope; + int16_t TempRespLim; + int16_t TempCurr; + int16_t SlopeCurr; + int16_t PwmCurr; + uint32_t RefreshPeriod; + int16_t FdoMax; + uint8_t TempSrc; + int8_t Padding; +}; + +typedef struct SMU75_Discrete_FanTable SMU75_Discrete_FanTable; + +#define SMU7_DISCRETE_GPIO_SCLK_DEBUG 4 +#define SMU7_DISCRETE_GPIO_SCLK_DEBUG_BIT (0x1 << SMU7_DISCRETE_GPIO_SCLK_DEBUG) + + + +struct SMU7_MclkDpmScoreboard { + uint32_t PercentageBusy; + + int32_t PIDError; + int32_t PIDIntegral; + int32_t PIDOutput; + + uint32_t SigmaDeltaAccum; + uint32_t SigmaDeltaOutput; + uint32_t SigmaDeltaLevel; + + uint32_t UtilizationSetpoint; + + uint8_t TdpClampMode; + uint8_t TdcClampMode; + uint8_t ThermClampMode; + uint8_t VoltageBusy; + + int8_t CurrLevel; + int8_t TargLevel; + uint8_t LevelChangeInProgress; + uint8_t UpHyst; + + uint8_t DownHyst; + uint8_t VoltageDownHyst; + uint8_t DpmEnable; + uint8_t DpmRunning; + + uint8_t DpmForce; + uint8_t DpmForceLevel; + uint8_t padding2; + uint8_t McArbIndex; + + uint32_t MinimumPerfMclk; + + uint8_t AcpiReq; + uint8_t AcpiAck; + uint8_t MclkSwitchInProgress; + uint8_t MclkSwitchCritical; + + uint8_t IgnoreVBlank; + uint8_t TargetMclkIndex; + uint8_t TargetMvddIndex; + uint8_t MclkSwitchResult; + + uint16_t VbiFailureCount; + uint8_t VbiWaitCounter; + uint8_t EnabledLevelsChange; + + uint16_t LevelResidencyCounters [SMU75_MAX_LEVELS_MEMORY]; + uint16_t LevelSwitchCounters [SMU75_MAX_LEVELS_MEMORY]; + + void (*TargetStateCalculator)(uint8_t); + void (*SavedTargetStateCalculator)(uint8_t); + + uint16_t AutoDpmInterval; + uint16_t AutoDpmRange; + + uint16_t VbiTimeoutCount; + uint16_t MclkSwitchingTime; + + uint8_t fastSwitch; + uint8_t Save_PIC_VDDGFX_EXIT; + uint8_t Save_PIC_VDDGFX_ENTER; + uint8_t VbiTimeout; + + uint32_t HbmTempRegBackup; +}; + +typedef struct SMU7_MclkDpmScoreboard SMU7_MclkDpmScoreboard; + +struct SMU7_UlvScoreboard { + uint8_t EnterUlv; + uint8_t ExitUlv; + uint8_t UlvActive; + uint8_t WaitingForUlv; + uint8_t UlvEnable; + uint8_t UlvRunning; + uint8_t UlvMasterEnable; + uint8_t padding; + uint32_t UlvAbortedCount; + uint32_t UlvTimeStamp; +}; + +typedef struct SMU7_UlvScoreboard SMU7_UlvScoreboard; + +struct VddgfxSavedRegisters { + uint32_t GPU_DBG[3]; + uint32_t MEC_BaseAddress_Hi; + uint32_t MEC_BaseAddress_Lo; + uint32_t THM_TMON0_CTRL2__RDIR_PRESENT; + uint32_t THM_TMON1_CTRL2__RDIR_PRESENT; + uint32_t CP_INT_CNTL; +}; + +typedef struct VddgfxSavedRegisters VddgfxSavedRegisters; + +struct SMU7_VddGfxScoreboard { + uint8_t VddGfxEnable; + uint8_t VddGfxActive; + uint8_t VPUResetOccured; + uint8_t padding; + + uint32_t VddGfxEnteredCount; + uint32_t VddGfxAbortedCount; + + uint32_t VddGfxVid; + + VddgfxSavedRegisters SavedRegisters; +}; + +typedef struct SMU7_VddGfxScoreboard SMU7_VddGfxScoreboard; + +struct SMU7_TdcLimitScoreboard { + uint8_t Enable; + uint8_t Running; + uint16_t Alpha; + uint32_t FilteredIddc; + uint32_t IddcLimit; + uint32_t IddcHyst; + SMU7_HystController_Data HystControllerData; +}; + +typedef struct SMU7_TdcLimitScoreboard SMU7_TdcLimitScoreboard; + +struct SMU7_PkgPwrLimitScoreboard { + uint8_t Enable; + uint8_t Running; + uint16_t Alpha; + uint32_t FilteredPkgPwr; + uint32_t Limit; + uint32_t Hyst; + uint32_t LimitFromDriver; + uint8_t PowerSharingEnabled; + uint8_t PowerSharingCounter; + uint8_t PowerSharingINTEnabled; + uint8_t GFXActivityCounterEnabled; + uint32_t EnergyCount; + uint32_t PSACTCount; + uint8_t RollOverRequired; + uint8_t RollOverCount; + uint8_t padding[2]; + SMU7_HystController_Data HystControllerData; +}; + +typedef struct SMU7_PkgPwrLimitScoreboard SMU7_PkgPwrLimitScoreboard; + +struct SMU7_BapmScoreboard { + uint32_t source_powers[SMU75_DTE_SOURCES]; + uint32_t source_powers_last[SMU75_DTE_SOURCES]; + int32_t entity_temperatures[SMU75_NUM_GPU_TES]; + int32_t initial_entity_temperatures[SMU75_NUM_GPU_TES]; + int32_t Limit; + int32_t Hyst; + int32_t therm_influence_coeff_table[SMU75_DTE_ITERATIONS * SMU75_DTE_SOURCES * SMU75_DTE_SINKS * 2]; + int32_t therm_node_table[SMU75_DTE_ITERATIONS * SMU75_DTE_SOURCES * SMU75_DTE_SINKS]; + uint16_t ConfigTDPPowerScalar; + uint16_t FanSpeedPowerScalar; + uint16_t OverDrivePowerScalar; + uint16_t OverDriveLimitScalar; + uint16_t FinalPowerScalar; + uint8_t VariantID; + uint8_t spare997; + + SMU7_HystController_Data HystControllerData; + + int32_t temperature_gradient_slope; + int32_t temperature_gradient; + uint32_t measured_temperature; +}; + + +typedef struct SMU7_BapmScoreboard SMU7_BapmScoreboard; + +struct SMU7_AcpiScoreboard { + uint32_t SavedInterruptMask[2]; + uint8_t LastACPIRequest; + uint8_t CgBifResp; + uint8_t RequestType; + uint8_t Padding; + SMU75_Discrete_ACPILevel D0Level; +}; + +typedef struct SMU7_AcpiScoreboard SMU7_AcpiScoreboard; + +struct SMU75_Discrete_PmFuses { + uint8_t BapmVddCVidHiSidd[8]; + + uint8_t BapmVddCVidLoSidd[8]; + + uint8_t VddCVid[8]; + + uint8_t SviLoadLineEn; + uint8_t SviLoadLineVddC; + uint8_t SviLoadLineTrimVddC; + uint8_t SviLoadLineOffsetVddC; + + uint16_t TDC_VDDC_PkgLimit; + uint8_t TDC_VDDC_ThrottleReleaseLimitPerc; + uint8_t TDC_MAWt; + + uint8_t TdcWaterfallCtl; + uint8_t LPMLTemperatureMin; + uint8_t LPMLTemperatureMax; + uint8_t Reserved; + + uint8_t LPMLTemperatureScaler[16]; + + int16_t FuzzyFan_ErrorSetDelta; + int16_t FuzzyFan_ErrorRateSetDelta; + int16_t FuzzyFan_PwmSetDelta; + uint16_t Reserved6; + + uint8_t GnbLPML[16]; + + uint8_t GnbLPMLMaxVid; + uint8_t GnbLPMLMinVid; + uint8_t Reserved1[2]; + + uint16_t BapmVddCBaseLeakageHiSidd; + uint16_t BapmVddCBaseLeakageLoSidd; + + uint16_t VFT_Temp[3]; + uint8_t Version; + uint8_t padding; + + SMU_QuadraticCoeffs VFT_ATE[3]; + + SMU_QuadraticCoeffs AVFS_GB; + SMU_QuadraticCoeffs ATE_ACBTC_GB; + + SMU_QuadraticCoeffs P2V; + + uint32_t PsmCharzFreq; + + uint16_t InversionVoltage; + uint16_t PsmCharzTemp; + + uint32_t EnabledAvfsModules; + + SMU_QuadraticCoeffs BtcGbv_CksOff; +}; + +typedef struct SMU75_Discrete_PmFuses SMU75_Discrete_PmFuses; + +struct SMU7_Discrete_Log_Header_Table { + uint32_t version; + uint32_t asic_id; + uint16_t flags; + uint16_t entry_size; + uint32_t total_size; + uint32_t num_of_entries; + uint8_t type; + uint8_t mode; + uint8_t filler_0[2]; + uint32_t filler_1[2]; +}; + +typedef struct SMU7_Discrete_Log_Header_Table SMU7_Discrete_Log_Header_Table; + +struct SMU7_Discrete_Log_Cntl { + uint8_t Enabled; + uint8_t Type; + uint8_t padding[2]; + uint32_t BufferSize; + uint32_t SamplesLogged; + uint32_t SampleSize; + uint32_t AddrL; + uint32_t AddrH; +}; + +typedef struct SMU7_Discrete_Log_Cntl SMU7_Discrete_Log_Cntl; + +#if defined SMU__DGPU_ONLY +#define CAC_ACC_NW_NUM_OF_SIGNALS 87 +#endif + + +struct SMU7_Discrete_Cac_Collection_Table { + uint32_t temperature; + uint32_t cac_acc_nw[CAC_ACC_NW_NUM_OF_SIGNALS]; +}; + +typedef struct SMU7_Discrete_Cac_Collection_Table SMU7_Discrete_Cac_Collection_Table; + +struct SMU7_Discrete_Cac_Verification_Table { + uint32_t VddcTotalPower; + uint32_t VddcLeakagePower; + uint32_t VddcConstantPower; + uint32_t VddcGfxDynamicPower; + uint32_t VddcUvdDynamicPower; + uint32_t VddcVceDynamicPower; + uint32_t VddcAcpDynamicPower; + uint32_t VddcPcieDynamicPower; + uint32_t VddcDceDynamicPower; + uint32_t VddcCurrent; + uint32_t VddcVoltage; + uint32_t VddciTotalPower; + uint32_t VddciLeakagePower; + uint32_t VddciConstantPower; + uint32_t VddciDynamicPower; + uint32_t Vddr1TotalPower; + uint32_t Vddr1LeakagePower; + uint32_t Vddr1ConstantPower; + uint32_t Vddr1DynamicPower; + uint32_t spare[4]; + uint32_t temperature; +}; + +typedef struct SMU7_Discrete_Cac_Verification_Table SMU7_Discrete_Cac_Verification_Table; + +struct SMU7_Discrete_Pm_Status_Table { + int32_t T_meas_max[SMU75_THERMAL_INPUT_LOOP_COUNT]; + int32_t T_meas_acc[SMU75_THERMAL_INPUT_LOOP_COUNT]; + + uint32_t I_calc_max; + uint32_t I_calc_acc; + uint32_t P_meas_acc; + uint32_t V_meas_load_acc; + uint32_t I_meas_acc; + uint32_t P_meas_acc_vddci; + uint32_t V_meas_load_acc_vddci; + uint32_t I_meas_acc_vddci; + + uint16_t Sclk_dpm_residency[8]; + uint16_t Uvd_dpm_residency[8]; + uint16_t Vce_dpm_residency[8]; + uint16_t Mclk_dpm_residency[4]; + + uint32_t P_roc_acc; + uint32_t PkgPwr_max; + uint32_t PkgPwr_acc; + uint32_t MclkSwitchingTime_max; + uint32_t MclkSwitchingTime_acc; + uint32_t FanPwm_acc; + uint32_t FanRpm_acc; + uint32_t Gfx_busy_acc; + uint32_t Mc_busy_acc; + uint32_t Fps_acc; + + uint32_t AccCnt; +}; + +typedef struct SMU7_Discrete_Pm_Status_Table SMU7_Discrete_Pm_Status_Table; + +struct SMU7_Discrete_AutoWattMan_Status_Table { + int32_t T_meas_acc[SMU75_THERMAL_INPUT_LOOP_COUNT]; + uint16_t Sclk_dpm_residency[8]; + uint16_t Mclk_dpm_residency[4]; + uint32_t TgpPwr_acc; + uint32_t Gfx_busy_acc; + uint32_t Mc_busy_acc; + uint32_t AccCnt; +}; + +typedef struct SMU7_Discrete_AutoWattMan_Status_Table SMU7_Discrete_AutoWattMan_Status_Table; + +#define SMU7_MAX_GFX_CU_COUNT 24 +#define SMU7_MIN_GFX_CU_COUNT 8 +#define SMU7_GFX_CU_PG_ENABLE_DC_MAX_CU_SHIFT 0 +#define SMU7_GFX_CU_PG_ENABLE_DC_MAX_CU_MASK (0xFFFF << SMU7_GFX_CU_PG_ENABLE_DC_MAX_CU_SHIFT) +#define SMU7_GFX_CU_PG_ENABLE_AC_MAX_CU_SHIFT 16 +#define SMU7_GFX_CU_PG_ENABLE_AC_MAX_CU_MASK (0xFFFF << SMU7_GFX_CU_PG_ENABLE_AC_MAX_CU_SHIFT) + +struct SMU7_GfxCuPgScoreboard { + uint8_t Enabled; + uint8_t WaterfallUp; + uint8_t WaterfallDown; + uint8_t WaterfallLimit; + uint8_t CurrMaxCu; + uint8_t TargMaxCu; + uint8_t ClampMode; + uint8_t Active; + uint8_t MaxSupportedCu; + uint8_t MinSupportedCu; + uint8_t PendingGfxCuHostInterrupt; + uint8_t LastFilteredMaxCuInteger; + uint16_t FilteredMaxCu; + uint16_t FilteredMaxCuAlpha; + uint16_t FilterResetCount; + uint16_t FilterResetCountLimit; + uint8_t ForceCu; + uint8_t ForceCuCount; + uint8_t AcModeMaxCu; + uint8_t DcModeMaxCu; +}; + +typedef struct SMU7_GfxCuPgScoreboard SMU7_GfxCuPgScoreboard; + +#define SMU7_SCLK_CAC 0x561 +#define SMU7_MCLK_CAC 0xF9 +#define SMU7_VCLK_CAC 0x2DE +#define SMU7_DCLK_CAC 0x2DE +#define SMU7_ECLK_CAC 0x25E +#define SMU7_ACLK_CAC 0x25E +#define SMU7_SAMCLK_CAC 0x25E +#define SMU7_DISPCLK_CAC 0x100 +#define SMU7_CAC_CONSTANT 0x2EE3430 +#define SMU7_CAC_CONSTANT_SHIFT 18 + +#define SMU7_VDDCI_MCLK_CONST 1765 +#define SMU7_VDDCI_MCLK_CONST_SHIFT 16 +#define SMU7_VDDCI_VDDCI_CONST 50958 +#define SMU7_VDDCI_VDDCI_CONST_SHIFT 14 +#define SMU7_VDDCI_CONST 11781 +#define SMU7_VDDCI_STROBE_PWR 1331 + +#define SMU7_VDDR1_CONST 693 +#define SMU7_VDDR1_CAC_WEIGHT 20 +#define SMU7_VDDR1_CAC_WEIGHT_SHIFT 19 +#define SMU7_VDDR1_STROBE_PWR 512 + +#define SMU7_AREA_COEFF_UVD 0xA78 +#define SMU7_AREA_COEFF_VCE 0x190A +#define SMU7_AREA_COEFF_ACP 0x22D1 +#define SMU7_AREA_COEFF_SAMU 0x534 + +#define SMU7_THERM_OUT_MODE_DISABLE 0x0 +#define SMU7_THERM_OUT_MODE_THERM_ONLY 0x1 +#define SMU7_THERM_OUT_MODE_THERM_VRHOT 0x2 + +#define SQ_Enable_MASK 0x1 +#define SQ_IR_MASK 0x2 +#define SQ_PCC_MASK 0x4 +#define SQ_EDC_MASK 0x8 + +#define TCP_Enable_MASK 0x100 +#define TCP_IR_MASK 0x200 +#define TCP_PCC_MASK 0x400 +#define TCP_EDC_MASK 0x800 + +#define TD_Enable_MASK 0x10000 +#define TD_IR_MASK 0x20000 +#define TD_PCC_MASK 0x40000 +#define TD_EDC_MASK 0x80000 + +#define DB_Enable_MASK 0x1000000 +#define DB_IR_MASK 0x2000000 +#define DB_PCC_MASK 0x4000000 +#define DB_EDC_MASK 0x8000000 + +#define SQ_Enable_SHIFT 0 +#define SQ_IR_SHIFT 1 +#define SQ_PCC_SHIFT 2 +#define SQ_EDC_SHIFT 3 + +#define TCP_Enable_SHIFT 8 +#define TCP_IR_SHIFT 9 +#define TCP_PCC_SHIFT 10 +#define TCP_EDC_SHIFT 11 + +#define TD_Enable_SHIFT 16 +#define TD_IR_SHIFT 17 +#define TD_PCC_SHIFT 18 +#define TD_EDC_SHIFT 19 + +#define DB_Enable_SHIFT 24 +#define DB_IR_SHIFT 25 +#define DB_PCC_SHIFT 26 +#define DB_EDC_SHIFT 27 + +#define PMFUSES_AVFSSIZE 104 + +#define BTCGB0_Vdroop_Enable_MASK 0x1 +#define BTCGB1_Vdroop_Enable_MASK 0x2 +#define AVFSGB0_Vdroop_Enable_MASK 0x4 +#define AVFSGB1_Vdroop_Enable_MASK 0x8 + +#define BTCGB0_Vdroop_Enable_SHIFT 0 +#define BTCGB1_Vdroop_Enable_SHIFT 1 +#define AVFSGB0_Vdroop_Enable_SHIFT 2 +#define AVFSGB1_Vdroop_Enable_SHIFT 3 + +#pragma pack(pop) + + +#endif + diff --git a/drivers/gpu/drm/amd/powerplay/inc/vega10_ppsmc.h b/drivers/gpu/drm/amd/powerplay/inc/vega10_ppsmc.h index c3ed737ab9514..715b5a1688312 100644 --- a/drivers/gpu/drm/amd/powerplay/inc/vega10_ppsmc.h +++ b/drivers/gpu/drm/amd/powerplay/inc/vega10_ppsmc.h @@ -131,6 +131,7 @@ typedef uint16_t PPSMC_Result; #define PPSMC_MSG_RunAcgInOpenLoop 0x5E #define PPSMC_MSG_InitializeAcg 0x5F #define PPSMC_MSG_GetCurrPkgPwr 0x61 +#define PPSMC_MSG_GetAverageGfxclkActualFrequency 0x63 #define PPSMC_MSG_SetPccThrottleLevel 0x67 #define PPSMC_MSG_UpdatePkgPwrPidAlpha 0x68 #define PPSMC_Message_Count 0x69 diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/Makefile b/drivers/gpu/drm/amd/powerplay/smumgr/Makefile index 9587550754211..0a200406a1ec2 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/Makefile +++ b/drivers/gpu/drm/amd/powerplay/smumgr/Makefile @@ -26,7 +26,7 @@ SMU_MGR = smumgr.o smu8_smumgr.o tonga_smumgr.o fiji_smumgr.o \ polaris10_smumgr.o iceland_smumgr.o \ smu7_smumgr.o vega10_smumgr.o smu10_smumgr.o ci_smumgr.o \ - vega12_smumgr.o + vega12_smumgr.o vegam_smumgr.o AMD_PP_SMUMGR = $(addprefix $(AMD_PP_PATH)/smumgr/,$(SMU_MGR)) diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/ci_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/ci_smumgr.c index 08d000140ecab..2d4ec8ac3a088 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/ci_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/ci_smumgr.c @@ -61,9 +61,6 @@ #define SMC_RAM_END 0x40000 -#define VOLTAGE_SCALE 4 -#define VOLTAGE_VID_OFFSET_SCALE1 625 -#define VOLTAGE_VID_OFFSET_SCALE2 100 #define CISLAND_MINIMUM_ENGINE_CLOCK 800 #define CISLAND_MAX_DEEPSLEEP_DIVIDER_ID 5 @@ -211,9 +208,7 @@ static int ci_send_msg_to_smc(struct pp_hwmgr *hwmgr, uint16_t msg) { int ret; - if (!ci_is_smc_ram_running(hwmgr)) - return -EINVAL; - + cgs_write_register(hwmgr->device, mmSMC_RESP_0, 0); cgs_write_register(hwmgr->device, mmSMC_MESSAGE_0, msg); PHM_WAIT_FIELD_UNEQUAL(hwmgr, SMC_RESP_0, SMC_RESP, 0); @@ -1182,7 +1177,6 @@ static int ci_populate_single_memory_level( struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); int result = 0; bool dll_state_on; - struct cgs_display_info info = {0}; uint32_t mclk_edc_wr_enable_threshold = 40000; uint32_t mclk_edc_enable_threshold = 40000; uint32_t mclk_strobe_mode_threshold = 40000; @@ -1236,8 +1230,7 @@ static int ci_populate_single_memory_level( /* default set to low watermark. Highest level will be set to high later.*/ memory_level->DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; - cgs_get_active_displays_info(hwmgr->device, &info); - data->display_timing.num_existing_displays = info.display_count; + data->display_timing.num_existing_displays = hwmgr->display_config->num_display; /* stutter mode not support on ci */ @@ -2784,7 +2777,6 @@ static int ci_smu_fini(struct pp_hwmgr *hwmgr) { kfree(hwmgr->smu_backend); hwmgr->smu_backend = NULL; - cgs_rel_firmware(hwmgr->device, CGS_UCODE_ID_SMU); return 0; } diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/fiji_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/fiji_smumgr.c index faef78321446f..53df9405f43a3 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/fiji_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/fiji_smumgr.c @@ -53,10 +53,7 @@ #define FIJI_SMC_SIZE 0x20000 -#define VOLTAGE_SCALE 4 #define POWERTUNE_DEFAULT_SET_MAX 1 -#define VOLTAGE_VID_OFFSET_SCALE1 625 -#define VOLTAGE_VID_OFFSET_SCALE2 100 #define VDDC_VDDCI_DELTA 300 #define MC_CG_ARB_FREQ_F1 0x0b @@ -288,8 +285,7 @@ static int fiji_start_smu(struct pp_hwmgr *hwmgr) struct fiji_smumgr *priv = (struct fiji_smumgr *)(hwmgr->smu_backend); /* Only start SMC if SMC RAM is not running */ - if (!(smu7_is_smc_ram_running(hwmgr) - || cgs_is_virtualization_enabled(hwmgr->device))) { + if (!smu7_is_smc_ram_running(hwmgr) && hwmgr->not_vf) { /* Check if SMU is running in protected mode */ if (0 == PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, @@ -307,13 +303,13 @@ static int fiji_start_smu(struct pp_hwmgr *hwmgr) } /* To initialize all clock gating before RLC loaded and running.*/ - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_GFX, AMD_CG_STATE_GATE); - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_GMC, AMD_CG_STATE_GATE); - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_SDMA, AMD_CG_STATE_GATE); - cgs_set_clockgating_state(hwmgr->device, + amdgpu_device_ip_set_clockgating_state(hwmgr->adev, AMD_IP_BLOCK_TYPE_COMMON, AMD_CG_STATE_GATE); /* Setup SoftRegsStart here for register lookup in case @@ -335,10 +331,10 @@ static bool fiji_is_hw_avfs_present(struct pp_hwmgr *hwmgr) uint32_t efuse = 0; uint32_t mask = (1 << ((AVFS_EN_MSB - AVFS_EN_LSB) + 1)) - 1; - if (cgs_is_virtualization_enabled(hwmgr->device)) - return 0; + if (!hwmgr->not_vf) + return false; - if (!atomctrl_read_efuse(hwmgr->device, AVFS_EN_LSB, AVFS_EN_MSB, + if (!atomctrl_read_efuse(hwmgr, AVFS_EN_LSB, AVFS_EN_MSB, mask, &efuse)) { if (efuse) return true; @@ -989,11 +985,11 @@ static int fiji_populate_single_graphic_level(struct pp_hwmgr *hwmgr, threshold = clock * data->fast_watermark_threshold / 100; - data->display_timing.min_clock_in_sr = hwmgr->display_config.min_core_set_clock_in_sr; + data->display_timing.min_clock_in_sr = hwmgr->display_config->min_core_set_clock_in_sr; if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_SclkDeepSleep)) level->DeepSleepDivId = smu7_get_sleep_divider_id_from_clock(clock, - hwmgr->display_config.min_core_set_clock_in_sr); + hwmgr->display_config->min_core_set_clock_in_sr); /* Default to slow, highest DPM level will be diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/iceland_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/iceland_smumgr.c index d4bb934e73343..415f691c3fa90 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/iceland_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/iceland_smumgr.c @@ -60,10 +60,7 @@ #define ICELAND_SMC_SIZE 0x20000 -#define VOLTAGE_SCALE 4 #define POWERTUNE_DEFAULT_SET_MAX 1 -#define VOLTAGE_VID_OFFSET_SCALE1 625 -#define VOLTAGE_VID_OFFSET_SCALE2 100 #define MC_CG_ARB_FREQ_F1 0x0b #define VDDC_VDDCI_DELTA 200 @@ -932,7 +929,7 @@ static int iceland_populate_single_graphic_level(struct pp_hwmgr *hwmgr, graphic_level->PowerThrottle = 0; data->display_timing.min_clock_in_sr = - hwmgr->display_config.min_core_set_clock_in_sr; + hwmgr->display_config->min_core_set_clock_in_sr; if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_SclkDeepSleep)) @@ -1236,7 +1233,6 @@ static int iceland_populate_single_memory_level( struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); int result = 0; bool dll_state_on; - struct cgs_display_info info = {0}; uint32_t mclk_edc_wr_enable_threshold = 40000; uint32_t mclk_edc_enable_threshold = 40000; uint32_t mclk_strobe_mode_threshold = 40000; @@ -1283,8 +1279,7 @@ static int iceland_populate_single_memory_level( /* default set to low watermark. Highest level will be set to high later.*/ memory_level->DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; - cgs_get_active_displays_info(hwmgr->device, &info); - data->display_timing.num_existing_displays = info.display_count; + data->display_timing.num_existing_displays = hwmgr->display_config->num_display; /* stutter mode not support on iceland */ diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c index 997a777dd35b2..a8c6524f07e40 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/polaris10_smumgr.c @@ -52,8 +52,6 @@ #include "dce/dce_10_0_sh_mask.h" #define POLARIS10_SMC_SIZE 0x20000 -#define VOLTAGE_VID_OFFSET_SCALE1 625 -#define VOLTAGE_VID_OFFSET_SCALE2 100 #define POWERTUNE_DEFAULT_SET_MAX 1 #define VDDC_VDDCI_DELTA 200 #define MC_CG_ARB_FREQ_F1 0x0b @@ -295,25 +293,16 @@ static int polaris10_start_smu(struct pp_hwmgr *hwmgr) struct polaris10_smumgr *smu_data = (struct polaris10_smumgr *)(hwmgr->smu_backend); /* Only start SMC if SMC RAM is not running */ - if (!(smu7_is_smc_ram_running(hwmgr) - || cgs_is_virtualization_enabled(hwmgr->device))) { + if (!smu7_is_smc_ram_running(hwmgr) && hwmgr->not_vf) { smu_data->protected_mode = (uint8_t) (PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, SMU_FIRMWARE, SMU_MODE)); smu_data->smu7_data.security_hard_key = (uint8_t) (PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, SMU_FIRMWARE, SMU_SEL)); /* Check if SMU is running in protected mode */ - if (smu_data->protected_mode == 0) { + if (smu_data->protected_mode == 0) result = polaris10_start_smu_in_non_protection_mode(hwmgr); - } else { + else result = polaris10_start_smu_in_protection_mode(hwmgr); - /* If failed, try with different security Key. */ - if (result != 0) { - smu_data->smu7_data.security_hard_key ^= 1; - cgs_rel_firmware(hwmgr->device, CGS_UCODE_ID_SMU); - result = polaris10_start_smu_in_protection_mode(hwmgr); - } - } - if (result != 0) PP_ASSERT_WITH_CODE(0, "Failed to load SMU ucode.", return result); @@ -951,11 +940,11 @@ static int polaris10_populate_single_graphic_level(struct pp_hwmgr *hwmgr, level->DownHyst = data->current_profile_setting.sclk_down_hyst; level->VoltageDownHyst = 0; level->PowerThrottle = 0; - data->display_timing.min_clock_in_sr = hwmgr->display_config.min_core_set_clock_in_sr; + data->display_timing.min_clock_in_sr = hwmgr->display_config->min_core_set_clock_in_sr; if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_SclkDeepSleep)) level->DeepSleepDivId = smu7_get_sleep_divider_id_from_clock(clock, - hwmgr->display_config.min_core_set_clock_in_sr); + hwmgr->display_config->min_core_set_clock_in_sr); /* Default to slow, highest DPM level will be * set to PPSMC_DISPLAY_WATERMARK_LOW later. @@ -1085,11 +1074,9 @@ static int polaris10_populate_single_memory_level(struct pp_hwmgr *hwmgr, struct phm_ppt_v1_information *table_info = (struct phm_ppt_v1_information *)(hwmgr->pptable); int result = 0; - struct cgs_display_info info = {0, 0, NULL}; uint32_t mclk_stutter_mode_threshold = 40000; phm_ppt_v1_clock_voltage_dependency_table *vdd_dep_table = NULL; - cgs_get_active_displays_info(hwmgr->device, &info); if (hwmgr->od_enabled) vdd_dep_table = (phm_ppt_v1_clock_voltage_dependency_table *)&data->odn_dpm_table.vdd_dependency_on_mclk; @@ -1115,7 +1102,7 @@ static int polaris10_populate_single_memory_level(struct pp_hwmgr *hwmgr, mem_level->StutterEnable = false; mem_level->DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; - data->display_timing.num_existing_displays = info.display_count; + data->display_timing.num_existing_displays = hwmgr->display_config->num_display; if (mclk_stutter_mode_threshold && (clock <= mclk_stutter_mode_threshold) && diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/smu10_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/smu10_smumgr.c index bc53f2beda30f..0a563f6fe9ea6 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/smu10_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/smu10_smumgr.c @@ -23,7 +23,7 @@ #include "smumgr.h" #include "smu10_inc.h" -#include "pp_soc15.h" +#include "soc15_common.h" #include "smu10_smumgr.h" #include "ppatomctrl.h" #include "rv_ppsmc.h" @@ -33,8 +33,6 @@ #include "pp_debug.h" -#define VOLTAGE_SCALE 4 - #define BUFFER_SIZE 80000 #define MAX_STRING_SIZE 15 #define BUFFER_SIZETWO 131072 @@ -49,48 +47,41 @@ static uint32_t smu10_wait_for_response(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; uint32_t reg; - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_90_BASE_IDX, mmMP1_SMN_C2PMSG_90); + reg = SOC15_REG_OFFSET(MP1, 0, mmMP1_SMN_C2PMSG_90); phm_wait_for_register_unequal(hwmgr, reg, 0, MP1_C2PMSG_90__CONTENT_MASK); - return cgs_read_register(hwmgr->device, reg); + return RREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_90); } static int smu10_send_msg_to_smc_without_waiting(struct pp_hwmgr *hwmgr, uint16_t msg) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_66_BASE_IDX, mmMP1_SMN_C2PMSG_66); - cgs_write_register(hwmgr->device, reg, msg); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_66, msg); return 0; } static int smu10_read_arg_from_smc(struct pp_hwmgr *hwmgr) { - uint32_t reg; - - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_82_BASE_IDX, mmMP1_SMN_C2PMSG_82); + struct amdgpu_device *adev = hwmgr->adev; - return cgs_read_register(hwmgr->device, reg); + return RREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_82); } static int smu10_send_msg_to_smc(struct pp_hwmgr *hwmgr, uint16_t msg) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; smu10_wait_for_response(hwmgr); - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_90_BASE_IDX, mmMP1_SMN_C2PMSG_90); - cgs_write_register(hwmgr->device, reg, 0); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_90, 0); smu10_send_msg_to_smc_without_waiting(hwmgr, msg); @@ -104,17 +95,13 @@ static int smu10_send_msg_to_smc(struct pp_hwmgr *hwmgr, uint16_t msg) static int smu10_send_msg_to_smc_with_parameter(struct pp_hwmgr *hwmgr, uint16_t msg, uint32_t parameter) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; smu10_wait_for_response(hwmgr); - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_90_BASE_IDX, mmMP1_SMN_C2PMSG_90); - cgs_write_register(hwmgr->device, reg, 0); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_90, 0); - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_82_BASE_IDX, mmMP1_SMN_C2PMSG_82); - cgs_write_register(hwmgr->device, reg, parameter); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_82, parameter); smu10_send_msg_to_smc_without_waiting(hwmgr, msg); diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/smu7_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/smu7_smumgr.c index 0399c10d2be05..d644a9bb9078d 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/smu7_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/smu7_smumgr.c @@ -167,24 +167,25 @@ int smu7_send_msg_to_smc(struct pp_hwmgr *hwmgr, uint16_t msg) { int ret; - if (!smu7_is_smc_ram_running(hwmgr)) - return -EINVAL; - - PHM_WAIT_FIELD_UNEQUAL(hwmgr, SMC_RESP_0, SMC_RESP, 0); ret = PHM_READ_FIELD(hwmgr->device, SMC_RESP_0, SMC_RESP); - if (ret != 1) - pr_info("\n failed to send pre message %x ret is %d \n", msg, ret); + if (ret == 0xFE) + pr_debug("last message was not supported\n"); + else if (ret != 1) + pr_info("\n last message was failed ret is %d\n", ret); + cgs_write_register(hwmgr->device, mmSMC_RESP_0, 0); cgs_write_register(hwmgr->device, mmSMC_MESSAGE_0, msg); PHM_WAIT_FIELD_UNEQUAL(hwmgr, SMC_RESP_0, SMC_RESP, 0); ret = PHM_READ_FIELD(hwmgr->device, SMC_RESP_0, SMC_RESP); - if (ret != 1) + if (ret == 0xFE) + pr_debug("message %x was not supported\n", msg); + else if (ret != 1) pr_info("\n failed to send message %x ret is %d \n", msg, ret); return 0; @@ -199,10 +200,6 @@ int smu7_send_msg_to_smc_without_waiting(struct pp_hwmgr *hwmgr, uint16_t msg) int smu7_send_msg_to_smc_with_parameter(struct pp_hwmgr *hwmgr, uint16_t msg, uint32_t parameter) { - if (!smu7_is_smc_ram_running(hwmgr)) { - return -EINVAL; - } - PHM_WAIT_FIELD_UNEQUAL(hwmgr, SMC_RESP_0, SMC_RESP, 0); cgs_write_register(hwmgr->device, mmSMC_MSG_ARG_0, parameter); @@ -231,16 +228,6 @@ int smu7_send_msg_to_smc_offset(struct pp_hwmgr *hwmgr) return 0; } -int smu7_wait_for_smc_inactive(struct pp_hwmgr *hwmgr) -{ - if (!smu7_is_smc_ram_running(hwmgr)) - return -EINVAL; - - PHM_WAIT_VFPF_INDIRECT_FIELD(hwmgr, SMC_IND, SMC_SYSCON_CLOCK_CNTL_0, cken, 0); - return 0; -} - - enum cgs_ucode_id smu7_convert_fw_type_to_cgs(uint32_t fw_type) { enum cgs_ucode_id result = CGS_UCODE_ID_MAXIMUM; @@ -296,11 +283,9 @@ int smu7_read_smc_sram_dword(struct pp_hwmgr *hwmgr, uint32_t smc_addr, uint32_t result = smu7_set_smc_sram_address(hwmgr, smc_addr, limit); - if (result) - return result; + *value = result ? 0 : cgs_read_register(hwmgr->device, mmSMC_IND_DATA_11); - *value = cgs_read_register(hwmgr->device, mmSMC_IND_DATA_11); - return 0; + return result; } int smu7_write_smc_sram_dword(struct pp_hwmgr *hwmgr, uint32_t smc_addr, uint32_t value, uint32_t limit) @@ -375,7 +360,7 @@ static int smu7_populate_single_firmware_entry(struct pp_hwmgr *hwmgr, entry->meta_data_addr_low = 0; /* digest need be excluded out */ - if (cgs_is_virtualization_enabled(hwmgr->device)) + if (!hwmgr->not_vf) info.image_size -= 20; entry->data_size_byte = info.image_size; entry->num_register_entries = 0; @@ -409,7 +394,7 @@ int smu7_request_smu_load_fw(struct pp_hwmgr *hwmgr) 0x0); if (hwmgr->chip_id > CHIP_TOPAZ) { /* add support for Topaz */ - if (!cgs_is_virtualization_enabled(hwmgr->device)) { + if (hwmgr->not_vf) { smu7_send_msg_to_smc_with_parameter(hwmgr, PPSMC_MSG_SMU_DRAM_ADDR_HI, upper_32_bits(smu_data->smu_buffer.mc_addr)); @@ -467,7 +452,7 @@ int smu7_request_smu_load_fw(struct pp_hwmgr *hwmgr) PP_ASSERT_WITH_CODE(0 == smu7_populate_single_firmware_entry(hwmgr, UCODE_ID_SDMA1, &toc->entry[toc->num_entries++]), "Failed to Get Firmware Entry.", return -EINVAL); - if (cgs_is_virtualization_enabled(hwmgr->device)) + if (!hwmgr->not_vf) PP_ASSERT_WITH_CODE(0 == smu7_populate_single_firmware_entry(hwmgr, UCODE_ID_MEC_STORAGE, &toc->entry[toc->num_entries++]), "Failed to Get Firmware Entry.", return -EINVAL); @@ -608,7 +593,7 @@ int smu7_init(struct pp_hwmgr *hwmgr) smu_data->header = smu_data->header_buffer.kaddr; smu_data->header_buffer.mc_addr = mc_addr; - if (cgs_is_virtualization_enabled(hwmgr->device)) + if (!hwmgr->not_vf) return 0; smu_data->smu_buffer.data_size = 200*4096; @@ -643,13 +628,12 @@ int smu7_smu_fini(struct pp_hwmgr *hwmgr) &smu_data->header_buffer.mc_addr, &smu_data->header_buffer.kaddr); - if (!cgs_is_virtualization_enabled(hwmgr->device)) + if (hwmgr->not_vf) amdgpu_bo_free_kernel(&smu_data->smu_buffer.handle, &smu_data->smu_buffer.mc_addr, &smu_data->smu_buffer.kaddr); kfree(hwmgr->smu_backend); hwmgr->smu_backend = NULL; - cgs_rel_firmware(hwmgr->device, CGS_UCODE_ID_SMU); return 0; } diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/smu7_smumgr.h b/drivers/gpu/drm/amd/powerplay/smumgr/smu7_smumgr.h index 126d300259baa..39c9bfda0ab41 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/smu7_smumgr.h +++ b/drivers/gpu/drm/amd/powerplay/smumgr/smu7_smumgr.h @@ -67,7 +67,6 @@ int smu7_send_msg_to_smc_with_parameter(struct pp_hwmgr *hwmgr, uint16_t msg, int smu7_send_msg_to_smc_with_parameter_without_waiting(struct pp_hwmgr *hwmgr, uint16_t msg, uint32_t parameter); int smu7_send_msg_to_smc_offset(struct pp_hwmgr *hwmgr); -int smu7_wait_for_smc_inactive(struct pp_hwmgr *hwmgr); enum cgs_ucode_id smu7_convert_fw_type_to_cgs(uint32_t fw_type); int smu7_read_smc_sram_dword(struct pp_hwmgr *hwmgr, uint32_t smc_addr, diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c index c28b60aae5f82..c9837935f0f5e 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/smumgr.c @@ -41,9 +41,11 @@ MODULE_FIRMWARE("amdgpu/polaris11_smc.bin"); MODULE_FIRMWARE("amdgpu/polaris11_smc_sk.bin"); MODULE_FIRMWARE("amdgpu/polaris11_k_smc.bin"); MODULE_FIRMWARE("amdgpu/polaris12_smc.bin"); +MODULE_FIRMWARE("amdgpu/vegam_smc.bin"); MODULE_FIRMWARE("amdgpu/vega10_smc.bin"); MODULE_FIRMWARE("amdgpu/vega10_acg_smc.bin"); MODULE_FIRMWARE("amdgpu/vega12_smc.bin"); +MODULE_FIRMWARE("amdgpu/vega20_smc.bin"); int smum_thermal_avfs_enable(struct pp_hwmgr *hwmgr) { diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/tonga_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/tonga_smumgr.c index b51d7468c3e75..782b19fc2e701 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/tonga_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/tonga_smumgr.c @@ -55,11 +55,7 @@ #include "dce/dce_10_0_d.h" #include "dce/dce_10_0_sh_mask.h" - -#define VOLTAGE_SCALE 4 #define POWERTUNE_DEFAULT_SET_MAX 1 -#define VOLTAGE_VID_OFFSET_SCALE1 625 -#define VOLTAGE_VID_OFFSET_SCALE2 100 #define MC_CG_ARB_FREQ_F1 0x0b #define VDDC_VDDCI_DELTA 200 @@ -199,8 +195,7 @@ static int tonga_start_smu(struct pp_hwmgr *hwmgr) int result; /* Only start SMC if SMC RAM is not running */ - if (!(smu7_is_smc_ram_running(hwmgr) || - cgs_is_virtualization_enabled(hwmgr->device))) { + if (!smu7_is_smc_ram_running(hwmgr) && hwmgr->not_vf) { /*Check if SMU is running in protected mode*/ if (0 == PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, SMU_FIRMWARE, SMU_MODE)) { @@ -651,7 +646,7 @@ static int tonga_populate_single_graphic_level(struct pp_hwmgr *hwmgr, graphic_level->PowerThrottle = 0; data->display_timing.min_clock_in_sr = - hwmgr->display_config.min_core_set_clock_in_sr; + hwmgr->display_config->min_core_set_clock_in_sr; if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_SclkDeepSleep)) @@ -957,18 +952,17 @@ static int tonga_populate_single_memory_level( SMU72_Discrete_MemoryLevel *memory_level ) { - uint32_t mvdd = 0; struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); struct phm_ppt_v1_information *pptable_info = (struct phm_ppt_v1_information *)(hwmgr->pptable); - int result = 0; - bool dll_state_on; - struct cgs_display_info info = {0}; uint32_t mclk_edc_wr_enable_threshold = 40000; uint32_t mclk_stutter_mode_threshold = 30000; uint32_t mclk_edc_enable_threshold = 40000; uint32_t mclk_strobe_mode_threshold = 40000; phm_ppt_v1_clock_voltage_dependency_table *vdd_dep_table = NULL; + int result = 0; + bool dll_state_on; + uint32_t mvdd = 0; if (hwmgr->od_enabled) vdd_dep_table = (phm_ppt_v1_clock_voltage_dependency_table *)&data->odn_dpm_table.vdd_dependency_on_mclk; @@ -1009,8 +1003,7 @@ static int tonga_populate_single_memory_level( /* default set to low watermark. Highest level will be set to high later.*/ memory_level->DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; - cgs_get_active_displays_info(hwmgr->device, &info); - data->display_timing.num_existing_displays = info.display_count; + data->display_timing.num_existing_displays = hwmgr->display_config->num_display; if ((mclk_stutter_mode_threshold != 0) && (memory_clock <= mclk_stutter_mode_threshold) && diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/vega10_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/vega10_smumgr.c index 4aafb043bcb08..e84669c448a30 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/vega10_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/vega10_smumgr.c @@ -23,7 +23,7 @@ #include "smumgr.h" #include "vega10_inc.h" -#include "pp_soc15.h" +#include "soc15_common.h" #include "vega10_smumgr.h" #include "vega10_hwmgr.h" #include "vega10_ppsmc.h" @@ -35,8 +35,6 @@ #define AVFS_EN_MSB 1568 #define AVFS_EN_LSB 1568 -#define VOLTAGE_SCALE 4 - /* Microcode file is stored in this buffer */ #define BUFFER_SIZE 80000 #define MAX_STRING_SIZE 15 @@ -54,18 +52,13 @@ static bool vega10_is_smc_ram_running(struct pp_hwmgr *hwmgr) { - uint32_t mp1_fw_flags, reg; - - reg = soc15_get_register_offset(NBIF_HWID, 0, - mmPCIE_INDEX2_BASE_IDX, mmPCIE_INDEX2); + struct amdgpu_device *adev = hwmgr->adev; + uint32_t mp1_fw_flags; - cgs_write_register(hwmgr->device, reg, + WREG32_SOC15(NBIF, 0, mmPCIE_INDEX2, (MP1_Public | (smnMP1_FIRMWARE_FLAGS & 0xffffffff))); - reg = soc15_get_register_offset(NBIF_HWID, 0, - mmPCIE_DATA2_BASE_IDX, mmPCIE_DATA2); - - mp1_fw_flags = cgs_read_register(hwmgr->device, reg); + mp1_fw_flags = RREG32_SOC15(NBIF, 0, mmPCIE_DATA2); if (mp1_fw_flags & MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED_MASK) return true; @@ -81,11 +74,11 @@ static bool vega10_is_smc_ram_running(struct pp_hwmgr *hwmgr) */ static uint32_t vega10_wait_for_response(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; uint32_t reg; uint32_t ret; - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_90_BASE_IDX, mmMP1_SMN_C2PMSG_90); + reg = SOC15_REG_OFFSET(MP1, 0, mmMP1_SMN_C2PMSG_90); ret = phm_wait_for_register_unequal(hwmgr, reg, 0, MP1_C2PMSG_90__CONTENT_MASK); @@ -93,7 +86,7 @@ static uint32_t vega10_wait_for_response(struct pp_hwmgr *hwmgr) if (ret) pr_err("No response from smu\n"); - return cgs_read_register(hwmgr->device, reg); + return RREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_90); } /* @@ -105,11 +98,9 @@ static uint32_t vega10_wait_for_response(struct pp_hwmgr *hwmgr) static int vega10_send_msg_to_smc_without_waiting(struct pp_hwmgr *hwmgr, uint16_t msg) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_66_BASE_IDX, mmMP1_SMN_C2PMSG_66); - cgs_write_register(hwmgr->device, reg, msg); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_66, msg); return 0; } @@ -122,14 +113,12 @@ static int vega10_send_msg_to_smc_without_waiting(struct pp_hwmgr *hwmgr, */ static int vega10_send_msg_to_smc(struct pp_hwmgr *hwmgr, uint16_t msg) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; uint32_t ret; vega10_wait_for_response(hwmgr); - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_90_BASE_IDX, mmMP1_SMN_C2PMSG_90); - cgs_write_register(hwmgr->device, reg, 0); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_90, 0); vega10_send_msg_to_smc_without_waiting(hwmgr, msg); @@ -150,18 +139,14 @@ static int vega10_send_msg_to_smc(struct pp_hwmgr *hwmgr, uint16_t msg) static int vega10_send_msg_to_smc_with_parameter(struct pp_hwmgr *hwmgr, uint16_t msg, uint32_t parameter) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; uint32_t ret; vega10_wait_for_response(hwmgr); - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_90_BASE_IDX, mmMP1_SMN_C2PMSG_90); - cgs_write_register(hwmgr->device, reg, 0); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_90, 0); - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_82_BASE_IDX, mmMP1_SMN_C2PMSG_82); - cgs_write_register(hwmgr->device, reg, parameter); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_82, parameter); vega10_send_msg_to_smc_without_waiting(hwmgr, msg); @@ -174,12 +159,9 @@ static int vega10_send_msg_to_smc_with_parameter(struct pp_hwmgr *hwmgr, static int vega10_get_argument(struct pp_hwmgr *hwmgr) { - uint32_t reg; - - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_82_BASE_IDX, mmMP1_SMN_C2PMSG_82); + struct amdgpu_device *adev = hwmgr->adev; - return cgs_read_register(hwmgr->device, reg); + return RREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_82); } static int vega10_copy_table_from_smc(struct pp_hwmgr *hwmgr, diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/vega12_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/vega12_smumgr.c index 651a3f28734bc..7d9b40e8b1bf0 100644 --- a/drivers/gpu/drm/amd/powerplay/smumgr/vega12_smumgr.c +++ b/drivers/gpu/drm/amd/powerplay/smumgr/vega12_smumgr.c @@ -23,7 +23,7 @@ #include "smumgr.h" #include "vega12_inc.h" -#include "pp_soc15.h" +#include "soc15_common.h" #include "vega12_smumgr.h" #include "vega12_ppsmc.h" #include "vega12/smu9_driver_if.h" @@ -44,18 +44,13 @@ static bool vega12_is_smc_ram_running(struct pp_hwmgr *hwmgr) { - uint32_t mp1_fw_flags, reg; + struct amdgpu_device *adev = hwmgr->adev; + uint32_t mp1_fw_flags; - reg = soc15_get_register_offset(NBIF_HWID, 0, - mmPCIE_INDEX2_BASE_IDX, mmPCIE_INDEX2); - - cgs_write_register(hwmgr->device, reg, + WREG32_SOC15(NBIF, 0, mmPCIE_INDEX2, (MP1_Public | (smnMP1_FIRMWARE_FLAGS & 0xffffffff))); - reg = soc15_get_register_offset(NBIF_HWID, 0, - mmPCIE_DATA2_BASE_IDX, mmPCIE_DATA2); - - mp1_fw_flags = cgs_read_register(hwmgr->device, reg); + mp1_fw_flags = RREG32_SOC15(NBIF, 0, mmPCIE_DATA2); if ((mp1_fw_flags & MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED_MASK) >> MP1_FIRMWARE_FLAGS__INTERRUPTS_ENABLED__SHIFT) @@ -72,15 +67,15 @@ static bool vega12_is_smc_ram_running(struct pp_hwmgr *hwmgr) */ static uint32_t vega12_wait_for_response(struct pp_hwmgr *hwmgr) { + struct amdgpu_device *adev = hwmgr->adev; uint32_t reg; - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_90_BASE_IDX, mmMP1_SMN_C2PMSG_90); + reg = SOC15_REG_OFFSET(MP1, 0, mmMP1_SMN_C2PMSG_90); phm_wait_for_register_unequal(hwmgr, reg, 0, MP1_C2PMSG_90__CONTENT_MASK); - return cgs_read_register(hwmgr->device, reg); + return RREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_90); } /* @@ -92,11 +87,9 @@ static uint32_t vega12_wait_for_response(struct pp_hwmgr *hwmgr) int vega12_send_msg_to_smc_without_waiting(struct pp_hwmgr *hwmgr, uint16_t msg) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_66_BASE_IDX, mmMP1_SMN_C2PMSG_66); - cgs_write_register(hwmgr->device, reg, msg); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_66, msg); return 0; } @@ -109,13 +102,11 @@ int vega12_send_msg_to_smc_without_waiting(struct pp_hwmgr *hwmgr, */ int vega12_send_msg_to_smc(struct pp_hwmgr *hwmgr, uint16_t msg) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; vega12_wait_for_response(hwmgr); - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_90_BASE_IDX, mmMP1_SMN_C2PMSG_90); - cgs_write_register(hwmgr->device, reg, 0); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_90, 0); vega12_send_msg_to_smc_without_waiting(hwmgr, msg); @@ -135,17 +126,13 @@ int vega12_send_msg_to_smc(struct pp_hwmgr *hwmgr, uint16_t msg) int vega12_send_msg_to_smc_with_parameter(struct pp_hwmgr *hwmgr, uint16_t msg, uint32_t parameter) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; vega12_wait_for_response(hwmgr); - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_90_BASE_IDX, mmMP1_SMN_C2PMSG_90); - cgs_write_register(hwmgr->device, reg, 0); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_90, 0); - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_82_BASE_IDX, mmMP1_SMN_C2PMSG_82); - cgs_write_register(hwmgr->device, reg, parameter); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_82, parameter); vega12_send_msg_to_smc_without_waiting(hwmgr, msg); @@ -166,11 +153,9 @@ int vega12_send_msg_to_smc_with_parameter(struct pp_hwmgr *hwmgr, int vega12_send_msg_to_smc_with_parameter_without_waiting( struct pp_hwmgr *hwmgr, uint16_t msg, uint32_t parameter) { - uint32_t reg; + struct amdgpu_device *adev = hwmgr->adev; - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_66_BASE_IDX, mmMP1_SMN_C2PMSG_66); - cgs_write_register(hwmgr->device, reg, parameter); + WREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_66, parameter); return vega12_send_msg_to_smc_without_waiting(hwmgr, msg); } @@ -183,12 +168,9 @@ int vega12_send_msg_to_smc_with_parameter_without_waiting( */ int vega12_read_arg_from_smc(struct pp_hwmgr *hwmgr, uint32_t *arg) { - uint32_t reg; - - reg = soc15_get_register_offset(MP1_HWID, 0, - mmMP1_SMN_C2PMSG_82_BASE_IDX, mmMP1_SMN_C2PMSG_82); + struct amdgpu_device *adev = hwmgr->adev; - *arg = cgs_read_register(hwmgr->device, reg); + *arg = RREG32_SOC15(MP1, 0, mmMP1_SMN_C2PMSG_82); return 0; } diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/vegam_smumgr.c b/drivers/gpu/drm/amd/powerplay/smumgr/vegam_smumgr.c new file mode 100644 index 0000000000000..2de48959ac935 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/smumgr/vegam_smumgr.c @@ -0,0 +1,2383 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ +#include "pp_debug.h" +#include "smumgr.h" +#include "smu_ucode_xfer_vi.h" +#include "vegam_smumgr.h" +#include "smu/smu_7_1_3_d.h" +#include "smu/smu_7_1_3_sh_mask.h" +#include "gmc/gmc_8_1_d.h" +#include "gmc/gmc_8_1_sh_mask.h" +#include "oss/oss_3_0_d.h" +#include "gca/gfx_8_0_d.h" +#include "bif/bif_5_0_d.h" +#include "bif/bif_5_0_sh_mask.h" +#include "ppatomctrl.h" +#include "cgs_common.h" +#include "smu7_ppsmc.h" + +#include "smu7_dyn_defaults.h" + +#include "smu7_hwmgr.h" +#include "hardwaremanager.h" +#include "ppatomctrl.h" +#include "atombios.h" +#include "pppcielanes.h" + +#include "dce/dce_11_2_d.h" +#include "dce/dce_11_2_sh_mask.h" + +#define PPVEGAM_TARGETACTIVITY_DFLT 50 + +#define VOLTAGE_VID_OFFSET_SCALE1 625 +#define VOLTAGE_VID_OFFSET_SCALE2 100 +#define POWERTUNE_DEFAULT_SET_MAX 1 +#define VDDC_VDDCI_DELTA 200 +#define MC_CG_ARB_FREQ_F1 0x0b + +#define STRAP_ASIC_RO_LSB 2168 +#define STRAP_ASIC_RO_MSB 2175 + +#define PPSMC_MSG_ApplyAvfsCksOffVoltage ((uint16_t) 0x415) +#define PPSMC_MSG_EnableModeSwitchRLCNotification ((uint16_t) 0x305) + +static const struct vegam_pt_defaults +vegam_power_tune_data_set_array[POWERTUNE_DEFAULT_SET_MAX] = { + /* sviLoadLIneEn, SviLoadLineVddC, TDC_VDDC_ThrottleReleaseLimitPerc, TDC_MAWt, + * TdcWaterfallCtl, DTEAmbientTempBase, DisplayCac, BAPM_TEMP_GRADIENT */ + { 1, 0xF, 0xFD, 0x19, 5, 45, 0, 0xB0000, + { 0x79, 0x253, 0x25D, 0xAE, 0x72, 0x80, 0x83, 0x86, 0x6F, 0xC8, 0xC9, 0xC9, 0x2F, 0x4D, 0x61}, + { 0x17C, 0x172, 0x180, 0x1BC, 0x1B3, 0x1BD, 0x206, 0x200, 0x203, 0x25D, 0x25A, 0x255, 0x2C3, 0x2C5, 0x2B4 } }, +}; + +static const sclkFcwRange_t Range_Table[NUM_SCLK_RANGE] = { + {VCO_2_4, POSTDIV_DIV_BY_16, 75, 160, 112}, + {VCO_3_6, POSTDIV_DIV_BY_16, 112, 224, 160}, + {VCO_2_4, POSTDIV_DIV_BY_8, 75, 160, 112}, + {VCO_3_6, POSTDIV_DIV_BY_8, 112, 224, 160}, + {VCO_2_4, POSTDIV_DIV_BY_4, 75, 160, 112}, + {VCO_3_6, POSTDIV_DIV_BY_4, 112, 216, 160}, + {VCO_2_4, POSTDIV_DIV_BY_2, 75, 160, 108}, + {VCO_3_6, POSTDIV_DIV_BY_2, 112, 216, 160} }; + +static int vegam_smu_init(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data; + + smu_data = kzalloc(sizeof(struct vegam_smumgr), GFP_KERNEL); + if (smu_data == NULL) + return -ENOMEM; + + hwmgr->smu_backend = smu_data; + + if (smu7_init(hwmgr)) { + kfree(smu_data); + return -EINVAL; + } + + return 0; +} + +static int vegam_start_smu_in_protection_mode(struct pp_hwmgr *hwmgr) +{ + int result = 0; + + /* Wait for smc boot up */ + /* PHM_WAIT_VFPF_INDIRECT_FIELD_UNEQUAL(smumgr, SMC_IND, RCU_UC_EVENTS, boot_seq_done, 0) */ + + /* Assert reset */ + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_RESET_CNTL, rst_reg, 1); + + result = smu7_upload_smu_firmware_image(hwmgr); + if (result != 0) + return result; + + /* Clear status */ + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixSMU_STATUS, 0); + + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_CLOCK_CNTL_0, ck_disable, 0); + + /* De-assert reset */ + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_RESET_CNTL, rst_reg, 0); + + + PHM_WAIT_VFPF_INDIRECT_FIELD(hwmgr, SMC_IND, RCU_UC_EVENTS, INTERRUPTS_ENABLED, 1); + + + /* Call Test SMU message with 0x20000 offset to trigger SMU start */ + smu7_send_msg_to_smc_offset(hwmgr); + + /* Wait done bit to be set */ + /* Check pass/failed indicator */ + + PHM_WAIT_VFPF_INDIRECT_FIELD_UNEQUAL(hwmgr, SMC_IND, SMU_STATUS, SMU_DONE, 0); + + if (1 != PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SMU_STATUS, SMU_PASS)) + PP_ASSERT_WITH_CODE(false, "SMU Firmware start failed!", return -1); + + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixFIRMWARE_FLAGS, 0); + + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_RESET_CNTL, rst_reg, 1); + + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_RESET_CNTL, rst_reg, 0); + + /* Wait for firmware to initialize */ + PHM_WAIT_VFPF_INDIRECT_FIELD(hwmgr, SMC_IND, FIRMWARE_FLAGS, INTERRUPTS_ENABLED, 1); + + return result; +} + +static int vegam_start_smu_in_non_protection_mode(struct pp_hwmgr *hwmgr) +{ + int result = 0; + + /* wait for smc boot up */ + PHM_WAIT_VFPF_INDIRECT_FIELD_UNEQUAL(hwmgr, SMC_IND, RCU_UC_EVENTS, boot_seq_done, 0); + + /* Clear firmware interrupt enable flag */ + /* PHM_WRITE_VFPF_INDIRECT_FIELD(pSmuMgr, SMC_IND, SMC_SYSCON_MISC_CNTL, pre_fetcher_en, 1); */ + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixFIRMWARE_FLAGS, 0); + + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_RESET_CNTL, + rst_reg, 1); + + result = smu7_upload_smu_firmware_image(hwmgr); + if (result != 0) + return result; + + /* Set smc instruct start point at 0x0 */ + smu7_program_jump_on_start(hwmgr); + + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_CLOCK_CNTL_0, ck_disable, 0); + + PHM_WRITE_VFPF_INDIRECT_FIELD(hwmgr->device, CGS_IND_REG__SMC, + SMC_SYSCON_RESET_CNTL, rst_reg, 0); + + /* Wait for firmware to initialize */ + + PHM_WAIT_VFPF_INDIRECT_FIELD(hwmgr, SMC_IND, + FIRMWARE_FLAGS, INTERRUPTS_ENABLED, 1); + + return result; +} + +static int vegam_start_smu(struct pp_hwmgr *hwmgr) +{ + int result = 0; + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + + /* Only start SMC if SMC RAM is not running */ + if (!smu7_is_smc_ram_running(hwmgr) && hwmgr->not_vf) { + smu_data->protected_mode = (uint8_t)(PHM_READ_VFPF_INDIRECT_FIELD(hwmgr->device, + CGS_IND_REG__SMC, SMU_FIRMWARE, SMU_MODE)); + smu_data->smu7_data.security_hard_key = (uint8_t)(PHM_READ_VFPF_INDIRECT_FIELD( + hwmgr->device, CGS_IND_REG__SMC, SMU_FIRMWARE, SMU_SEL)); + + /* Check if SMU is running in protected mode */ + if (smu_data->protected_mode == 0) + result = vegam_start_smu_in_non_protection_mode(hwmgr); + else + result = vegam_start_smu_in_protection_mode(hwmgr); + + if (result != 0) + PP_ASSERT_WITH_CODE(0, "Failed to load SMU ucode.", return result); + } + + /* Setup SoftRegsStart here for register lookup in case DummyBackEnd is used and ProcessFirmwareHeader is not executed */ + smu7_read_smc_sram_dword(hwmgr, + SMU7_FIRMWARE_HEADER_LOCATION + offsetof(SMU75_Firmware_Header, SoftRegisters), + &(smu_data->smu7_data.soft_regs_start), + 0x40000); + + result = smu7_request_smu_load_fw(hwmgr); + + return result; +} + +static int vegam_process_firmware_header(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + uint32_t tmp; + int result; + bool error = false; + + result = smu7_read_smc_sram_dword(hwmgr, + SMU7_FIRMWARE_HEADER_LOCATION + + offsetof(SMU75_Firmware_Header, DpmTable), + &tmp, SMC_RAM_END); + + if (0 == result) + smu_data->smu7_data.dpm_table_start = tmp; + + error |= (0 != result); + + result = smu7_read_smc_sram_dword(hwmgr, + SMU7_FIRMWARE_HEADER_LOCATION + + offsetof(SMU75_Firmware_Header, SoftRegisters), + &tmp, SMC_RAM_END); + + if (!result) { + data->soft_regs_start = tmp; + smu_data->smu7_data.soft_regs_start = tmp; + } + + error |= (0 != result); + + result = smu7_read_smc_sram_dword(hwmgr, + SMU7_FIRMWARE_HEADER_LOCATION + + offsetof(SMU75_Firmware_Header, mcRegisterTable), + &tmp, SMC_RAM_END); + + if (!result) + smu_data->smu7_data.mc_reg_table_start = tmp; + + result = smu7_read_smc_sram_dword(hwmgr, + SMU7_FIRMWARE_HEADER_LOCATION + + offsetof(SMU75_Firmware_Header, FanTable), + &tmp, SMC_RAM_END); + + if (!result) + smu_data->smu7_data.fan_table_start = tmp; + + error |= (0 != result); + + result = smu7_read_smc_sram_dword(hwmgr, + SMU7_FIRMWARE_HEADER_LOCATION + + offsetof(SMU75_Firmware_Header, mcArbDramTimingTable), + &tmp, SMC_RAM_END); + + if (!result) + smu_data->smu7_data.arb_table_start = tmp; + + error |= (0 != result); + + result = smu7_read_smc_sram_dword(hwmgr, + SMU7_FIRMWARE_HEADER_LOCATION + + offsetof(SMU75_Firmware_Header, Version), + &tmp, SMC_RAM_END); + + if (!result) + hwmgr->microcode_version_info.SMC = tmp; + + error |= (0 != result); + + return error ? -1 : 0; +} + +static bool vegam_is_dpm_running(struct pp_hwmgr *hwmgr) +{ + return (1 == PHM_READ_INDIRECT_FIELD(hwmgr->device, + CGS_IND_REG__SMC, FEATURE_STATUS, VOLTAGE_CONTROLLER_ON)) + ? true : false; +} + +static uint32_t vegam_get_mac_definition(uint32_t value) +{ + switch (value) { + case SMU_MAX_LEVELS_GRAPHICS: + return SMU75_MAX_LEVELS_GRAPHICS; + case SMU_MAX_LEVELS_MEMORY: + return SMU75_MAX_LEVELS_MEMORY; + case SMU_MAX_LEVELS_LINK: + return SMU75_MAX_LEVELS_LINK; + case SMU_MAX_ENTRIES_SMIO: + return SMU75_MAX_ENTRIES_SMIO; + case SMU_MAX_LEVELS_VDDC: + return SMU75_MAX_LEVELS_VDDC; + case SMU_MAX_LEVELS_VDDGFX: + return SMU75_MAX_LEVELS_VDDGFX; + case SMU_MAX_LEVELS_VDDCI: + return SMU75_MAX_LEVELS_VDDCI; + case SMU_MAX_LEVELS_MVDD: + return SMU75_MAX_LEVELS_MVDD; + case SMU_UVD_MCLK_HANDSHAKE_DISABLE: + return SMU7_UVD_MCLK_HANDSHAKE_DISABLE | + SMU7_VCE_MCLK_HANDSHAKE_DISABLE; + } + + pr_warn("can't get the mac of %x\n", value); + return 0; +} + +static int vegam_update_uvd_smc_table(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + uint32_t mm_boot_level_offset, mm_boot_level_value; + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + + smu_data->smc_state_table.UvdBootLevel = 0; + if (table_info->mm_dep_table->count > 0) + smu_data->smc_state_table.UvdBootLevel = + (uint8_t) (table_info->mm_dep_table->count - 1); + mm_boot_level_offset = smu_data->smu7_data.dpm_table_start + offsetof(SMU75_Discrete_DpmTable, + UvdBootLevel); + mm_boot_level_offset /= 4; + mm_boot_level_offset *= 4; + mm_boot_level_value = cgs_read_ind_register(hwmgr->device, + CGS_IND_REG__SMC, mm_boot_level_offset); + mm_boot_level_value &= 0x00FFFFFF; + mm_boot_level_value |= smu_data->smc_state_table.UvdBootLevel << 24; + cgs_write_ind_register(hwmgr->device, + CGS_IND_REG__SMC, mm_boot_level_offset, mm_boot_level_value); + + if (!phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_UVDDPM) || + phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_StablePState)) + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_UVDDPM_SetEnabledMask, + (uint32_t)(1 << smu_data->smc_state_table.UvdBootLevel)); + return 0; +} + +static int vegam_update_vce_smc_table(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + uint32_t mm_boot_level_offset, mm_boot_level_value; + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_StablePState)) + smu_data->smc_state_table.VceBootLevel = + (uint8_t) (table_info->mm_dep_table->count - 1); + else + smu_data->smc_state_table.VceBootLevel = 0; + + mm_boot_level_offset = smu_data->smu7_data.dpm_table_start + + offsetof(SMU75_Discrete_DpmTable, VceBootLevel); + mm_boot_level_offset /= 4; + mm_boot_level_offset *= 4; + mm_boot_level_value = cgs_read_ind_register(hwmgr->device, + CGS_IND_REG__SMC, mm_boot_level_offset); + mm_boot_level_value &= 0xFF00FFFF; + mm_boot_level_value |= smu_data->smc_state_table.VceBootLevel << 16; + cgs_write_ind_register(hwmgr->device, + CGS_IND_REG__SMC, mm_boot_level_offset, mm_boot_level_value); + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_StablePState)) + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_VCEDPM_SetEnabledMask, + (uint32_t)1 << smu_data->smc_state_table.VceBootLevel); + return 0; +} + +static int vegam_update_samu_smc_table(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + uint32_t mm_boot_level_offset, mm_boot_level_value; + + + smu_data->smc_state_table.SamuBootLevel = 0; + mm_boot_level_offset = smu_data->smu7_data.dpm_table_start + + offsetof(SMU75_Discrete_DpmTable, SamuBootLevel); + + mm_boot_level_offset /= 4; + mm_boot_level_offset *= 4; + mm_boot_level_value = cgs_read_ind_register(hwmgr->device, + CGS_IND_REG__SMC, mm_boot_level_offset); + mm_boot_level_value &= 0xFFFFFF00; + mm_boot_level_value |= smu_data->smc_state_table.SamuBootLevel << 0; + cgs_write_ind_register(hwmgr->device, + CGS_IND_REG__SMC, mm_boot_level_offset, mm_boot_level_value); + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_StablePState)) + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_SAMUDPM_SetEnabledMask, + (uint32_t)(1 << smu_data->smc_state_table.SamuBootLevel)); + return 0; +} + + +static int vegam_update_bif_smc_table(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct phm_ppt_v1_pcie_table *pcie_table = table_info->pcie_table; + int max_entry, i; + + max_entry = (SMU75_MAX_LEVELS_LINK < pcie_table->count) ? + SMU75_MAX_LEVELS_LINK : + pcie_table->count; + /* Setup BIF_SCLK levels */ + for (i = 0; i < max_entry; i++) + smu_data->bif_sclk_table[i] = pcie_table->entries[i].pcie_sclk; + return 0; +} + +static int vegam_update_smc_table(struct pp_hwmgr *hwmgr, uint32_t type) +{ + switch (type) { + case SMU_UVD_TABLE: + vegam_update_uvd_smc_table(hwmgr); + break; + case SMU_VCE_TABLE: + vegam_update_vce_smc_table(hwmgr); + break; + case SMU_SAMU_TABLE: + vegam_update_samu_smc_table(hwmgr); + break; + case SMU_BIF_TABLE: + vegam_update_bif_smc_table(hwmgr); + break; + default: + break; + } + return 0; +} + +static void vegam_initialize_power_tune_defaults(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + + if (table_info && + table_info->cac_dtp_table->usPowerTuneDataSetID <= POWERTUNE_DEFAULT_SET_MAX && + table_info->cac_dtp_table->usPowerTuneDataSetID) + smu_data->power_tune_defaults = + &vegam_power_tune_data_set_array + [table_info->cac_dtp_table->usPowerTuneDataSetID - 1]; + else + smu_data->power_tune_defaults = &vegam_power_tune_data_set_array[0]; + +} + +static int vegam_populate_smc_mvdd_table(struct pp_hwmgr *hwmgr, + SMU75_Discrete_DpmTable *table) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + uint32_t count, level; + + if (SMU7_VOLTAGE_CONTROL_BY_GPIO == data->mvdd_control) { + count = data->mvdd_voltage_table.count; + if (count > SMU_MAX_SMIO_LEVELS) + count = SMU_MAX_SMIO_LEVELS; + for (level = 0; level < count; level++) { + table->SmioTable2.Pattern[level].Voltage = PP_HOST_TO_SMC_US( + data->mvdd_voltage_table.entries[count].value * VOLTAGE_SCALE); + /* Index into DpmTable.Smio. Drive bits from Smio entry to get this voltage level.*/ + table->SmioTable2.Pattern[level].Smio = + (uint8_t) level; + table->Smio[level] |= + data->mvdd_voltage_table.entries[level].smio_low; + } + table->SmioMask2 = data->mvdd_voltage_table.mask_low; + + table->MvddLevelCount = (uint32_t) PP_HOST_TO_SMC_UL(count); + } + + return 0; +} + +static int vegam_populate_smc_vddci_table(struct pp_hwmgr *hwmgr, + struct SMU75_Discrete_DpmTable *table) +{ + uint32_t count, level; + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + + count = data->vddci_voltage_table.count; + + if (SMU7_VOLTAGE_CONTROL_BY_GPIO == data->vddci_control) { + if (count > SMU_MAX_SMIO_LEVELS) + count = SMU_MAX_SMIO_LEVELS; + for (level = 0; level < count; ++level) { + table->SmioTable1.Pattern[level].Voltage = PP_HOST_TO_SMC_US( + data->vddci_voltage_table.entries[level].value * VOLTAGE_SCALE); + table->SmioTable1.Pattern[level].Smio = (uint8_t) level; + + table->Smio[level] |= data->vddci_voltage_table.entries[level].smio_low; + } + } + + table->SmioMask1 = data->vddci_voltage_table.mask_low; + + return 0; +} + +static int vegam_populate_cac_table(struct pp_hwmgr *hwmgr, + struct SMU75_Discrete_DpmTable *table) +{ + uint32_t count; + uint8_t index; + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct phm_ppt_v1_voltage_lookup_table *lookup_table = + table_info->vddc_lookup_table; + /* tables is already swapped, so in order to use the value from it, + * we need to swap it back. + * We are populating vddc CAC data to BapmVddc table + * in split and merged mode + */ + for (count = 0; count < lookup_table->count; count++) { + index = phm_get_voltage_index(lookup_table, + data->vddc_voltage_table.entries[count].value); + table->BapmVddcVidLoSidd[count] = + convert_to_vid(lookup_table->entries[index].us_cac_low); + table->BapmVddcVidHiSidd[count] = + convert_to_vid(lookup_table->entries[index].us_cac_mid); + table->BapmVddcVidHiSidd2[count] = + convert_to_vid(lookup_table->entries[index].us_cac_high); + } + + return 0; +} + +static int vegam_populate_smc_voltage_tables(struct pp_hwmgr *hwmgr, + struct SMU75_Discrete_DpmTable *table) +{ + vegam_populate_smc_vddci_table(hwmgr, table); + vegam_populate_smc_mvdd_table(hwmgr, table); + vegam_populate_cac_table(hwmgr, table); + + return 0; +} + +static int vegam_populate_ulv_level(struct pp_hwmgr *hwmgr, + struct SMU75_Discrete_Ulv *state) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + + state->CcPwrDynRm = 0; + state->CcPwrDynRm1 = 0; + + state->VddcOffset = (uint16_t) table_info->us_ulv_voltage_offset; + state->VddcOffsetVid = (uint8_t)(table_info->us_ulv_voltage_offset * + VOLTAGE_VID_OFFSET_SCALE2 / VOLTAGE_VID_OFFSET_SCALE1); + + state->VddcPhase = data->vddc_phase_shed_control ^ 0x3; + + CONVERT_FROM_HOST_TO_SMC_UL(state->CcPwrDynRm); + CONVERT_FROM_HOST_TO_SMC_UL(state->CcPwrDynRm1); + CONVERT_FROM_HOST_TO_SMC_US(state->VddcOffset); + + return 0; +} + +static int vegam_populate_ulv_state(struct pp_hwmgr *hwmgr, + struct SMU75_Discrete_DpmTable *table) +{ + return vegam_populate_ulv_level(hwmgr, &table->Ulv); +} + +static int vegam_populate_smc_link_level(struct pp_hwmgr *hwmgr, + struct SMU75_Discrete_DpmTable *table) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct vegam_smumgr *smu_data = + (struct vegam_smumgr *)(hwmgr->smu_backend); + struct smu7_dpm_table *dpm_table = &data->dpm_table; + int i; + + /* Index (dpm_table->pcie_speed_table.count) + * is reserved for PCIE boot level. */ + for (i = 0; i <= dpm_table->pcie_speed_table.count; i++) { + table->LinkLevel[i].PcieGenSpeed = + (uint8_t)dpm_table->pcie_speed_table.dpm_levels[i].value; + table->LinkLevel[i].PcieLaneCount = (uint8_t)encode_pcie_lane_width( + dpm_table->pcie_speed_table.dpm_levels[i].param1); + table->LinkLevel[i].EnabledForActivity = 1; + table->LinkLevel[i].SPC = (uint8_t)(data->pcie_spc_cap & 0xff); + table->LinkLevel[i].DownThreshold = PP_HOST_TO_SMC_UL(5); + table->LinkLevel[i].UpThreshold = PP_HOST_TO_SMC_UL(30); + } + + smu_data->smc_state_table.LinkLevelCount = + (uint8_t)dpm_table->pcie_speed_table.count; + +/* To Do move to hwmgr */ + data->dpm_level_enable_mask.pcie_dpm_enable_mask = + phm_get_dpm_level_enable_mask_value(&dpm_table->pcie_speed_table); + + return 0; +} + +static int vegam_get_dependency_volt_by_clk(struct pp_hwmgr *hwmgr, + struct phm_ppt_v1_clock_voltage_dependency_table *dep_table, + uint32_t clock, SMU_VoltageLevel *voltage, uint32_t *mvdd) +{ + uint32_t i; + uint16_t vddci; + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + + *voltage = *mvdd = 0; + + /* clock - voltage dependency table is empty table */ + if (dep_table->count == 0) + return -EINVAL; + + for (i = 0; i < dep_table->count; i++) { + /* find first sclk bigger than request */ + if (dep_table->entries[i].clk >= clock) { + *voltage |= (dep_table->entries[i].vddc * + VOLTAGE_SCALE) << VDDC_SHIFT; + if (SMU7_VOLTAGE_CONTROL_NONE == data->vddci_control) + *voltage |= (data->vbios_boot_state.vddci_bootup_value * + VOLTAGE_SCALE) << VDDCI_SHIFT; + else if (dep_table->entries[i].vddci) + *voltage |= (dep_table->entries[i].vddci * + VOLTAGE_SCALE) << VDDCI_SHIFT; + else { + vddci = phm_find_closest_vddci(&(data->vddci_voltage_table), + (dep_table->entries[i].vddc - + (uint16_t)VDDC_VDDCI_DELTA)); + *voltage |= (vddci * VOLTAGE_SCALE) << VDDCI_SHIFT; + } + + if (SMU7_VOLTAGE_CONTROL_NONE == data->mvdd_control) + *mvdd = data->vbios_boot_state.mvdd_bootup_value * + VOLTAGE_SCALE; + else if (dep_table->entries[i].mvdd) + *mvdd = (uint32_t) dep_table->entries[i].mvdd * + VOLTAGE_SCALE; + + *voltage |= 1 << PHASES_SHIFT; + return 0; + } + } + + /* sclk is bigger than max sclk in the dependence table */ + *voltage |= (dep_table->entries[i - 1].vddc * VOLTAGE_SCALE) << VDDC_SHIFT; + vddci = phm_find_closest_vddci(&(data->vddci_voltage_table), + (dep_table->entries[i - 1].vddc - + (uint16_t)VDDC_VDDCI_DELTA)); + + if (SMU7_VOLTAGE_CONTROL_NONE == data->vddci_control) + *voltage |= (data->vbios_boot_state.vddci_bootup_value * + VOLTAGE_SCALE) << VDDCI_SHIFT; + else if (dep_table->entries[i - 1].vddci) + *voltage |= (dep_table->entries[i - 1].vddci * + VOLTAGE_SCALE) << VDDC_SHIFT; + else + *voltage |= (vddci * VOLTAGE_SCALE) << VDDCI_SHIFT; + + if (SMU7_VOLTAGE_CONTROL_NONE == data->mvdd_control) + *mvdd = data->vbios_boot_state.mvdd_bootup_value * VOLTAGE_SCALE; + else if (dep_table->entries[i].mvdd) + *mvdd = (uint32_t) dep_table->entries[i - 1].mvdd * VOLTAGE_SCALE; + + return 0; +} + +static void vegam_get_sclk_range_table(struct pp_hwmgr *hwmgr, + SMU75_Discrete_DpmTable *table) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + uint32_t i, ref_clk; + + struct pp_atom_ctrl_sclk_range_table range_table_from_vbios = { { {0} } }; + + ref_clk = amdgpu_asic_get_xclk((struct amdgpu_device *)hwmgr->adev); + + if (0 == atomctrl_get_smc_sclk_range_table(hwmgr, &range_table_from_vbios)) { + for (i = 0; i < NUM_SCLK_RANGE; i++) { + table->SclkFcwRangeTable[i].vco_setting = + range_table_from_vbios.entry[i].ucVco_setting; + table->SclkFcwRangeTable[i].postdiv = + range_table_from_vbios.entry[i].ucPostdiv; + table->SclkFcwRangeTable[i].fcw_pcc = + range_table_from_vbios.entry[i].usFcw_pcc; + + table->SclkFcwRangeTable[i].fcw_trans_upper = + range_table_from_vbios.entry[i].usFcw_trans_upper; + table->SclkFcwRangeTable[i].fcw_trans_lower = + range_table_from_vbios.entry[i].usRcw_trans_lower; + + CONVERT_FROM_HOST_TO_SMC_US(table->SclkFcwRangeTable[i].fcw_pcc); + CONVERT_FROM_HOST_TO_SMC_US(table->SclkFcwRangeTable[i].fcw_trans_upper); + CONVERT_FROM_HOST_TO_SMC_US(table->SclkFcwRangeTable[i].fcw_trans_lower); + } + return; + } + + for (i = 0; i < NUM_SCLK_RANGE; i++) { + smu_data->range_table[i].trans_lower_frequency = + (ref_clk * Range_Table[i].fcw_trans_lower) >> Range_Table[i].postdiv; + smu_data->range_table[i].trans_upper_frequency = + (ref_clk * Range_Table[i].fcw_trans_upper) >> Range_Table[i].postdiv; + + table->SclkFcwRangeTable[i].vco_setting = Range_Table[i].vco_setting; + table->SclkFcwRangeTable[i].postdiv = Range_Table[i].postdiv; + table->SclkFcwRangeTable[i].fcw_pcc = Range_Table[i].fcw_pcc; + + table->SclkFcwRangeTable[i].fcw_trans_upper = Range_Table[i].fcw_trans_upper; + table->SclkFcwRangeTable[i].fcw_trans_lower = Range_Table[i].fcw_trans_lower; + + CONVERT_FROM_HOST_TO_SMC_US(table->SclkFcwRangeTable[i].fcw_pcc); + CONVERT_FROM_HOST_TO_SMC_US(table->SclkFcwRangeTable[i].fcw_trans_upper); + CONVERT_FROM_HOST_TO_SMC_US(table->SclkFcwRangeTable[i].fcw_trans_lower); + } +} + +static int vegam_calculate_sclk_params(struct pp_hwmgr *hwmgr, + uint32_t clock, SMU_SclkSetting *sclk_setting) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + const SMU75_Discrete_DpmTable *table = &(smu_data->smc_state_table); + struct pp_atomctrl_clock_dividers_ai dividers; + uint32_t ref_clock; + uint32_t pcc_target_percent, pcc_target_freq, ss_target_percent, ss_target_freq; + uint8_t i; + int result; + uint64_t temp; + + sclk_setting->SclkFrequency = clock; + /* get the engine clock dividers for this clock value */ + result = atomctrl_get_engine_pll_dividers_ai(hwmgr, clock, ÷rs); + if (result == 0) { + sclk_setting->Fcw_int = dividers.usSclk_fcw_int; + sclk_setting->Fcw_frac = dividers.usSclk_fcw_frac; + sclk_setting->Pcc_fcw_int = dividers.usPcc_fcw_int; + sclk_setting->PllRange = dividers.ucSclkPllRange; + sclk_setting->Sclk_slew_rate = 0x400; + sclk_setting->Pcc_up_slew_rate = dividers.usPcc_fcw_slew_frac; + sclk_setting->Pcc_down_slew_rate = 0xffff; + sclk_setting->SSc_En = dividers.ucSscEnable; + sclk_setting->Fcw1_int = dividers.usSsc_fcw1_int; + sclk_setting->Fcw1_frac = dividers.usSsc_fcw1_frac; + sclk_setting->Sclk_ss_slew_rate = dividers.usSsc_fcw_slew_frac; + return result; + } + + ref_clock = amdgpu_asic_get_xclk((struct amdgpu_device *)hwmgr->adev); + + for (i = 0; i < NUM_SCLK_RANGE; i++) { + if (clock > smu_data->range_table[i].trans_lower_frequency + && clock <= smu_data->range_table[i].trans_upper_frequency) { + sclk_setting->PllRange = i; + break; + } + } + + sclk_setting->Fcw_int = (uint16_t) + ((clock << table->SclkFcwRangeTable[sclk_setting->PllRange].postdiv) / + ref_clock); + temp = clock << table->SclkFcwRangeTable[sclk_setting->PllRange].postdiv; + temp <<= 0x10; + do_div(temp, ref_clock); + sclk_setting->Fcw_frac = temp & 0xffff; + + pcc_target_percent = 10; /* Hardcode 10% for now. */ + pcc_target_freq = clock - (clock * pcc_target_percent / 100); + sclk_setting->Pcc_fcw_int = (uint16_t) + ((pcc_target_freq << table->SclkFcwRangeTable[sclk_setting->PllRange].postdiv) / + ref_clock); + + ss_target_percent = 2; /* Hardcode 2% for now. */ + sclk_setting->SSc_En = 0; + if (ss_target_percent) { + sclk_setting->SSc_En = 1; + ss_target_freq = clock - (clock * ss_target_percent / 100); + sclk_setting->Fcw1_int = (uint16_t) + ((ss_target_freq << table->SclkFcwRangeTable[sclk_setting->PllRange].postdiv) / + ref_clock); + temp = ss_target_freq << table->SclkFcwRangeTable[sclk_setting->PllRange].postdiv; + temp <<= 0x10; + do_div(temp, ref_clock); + sclk_setting->Fcw1_frac = temp & 0xffff; + } + + return 0; +} + +static uint8_t vegam_get_sleep_divider_id_from_clock(uint32_t clock, + uint32_t clock_insr) +{ + uint8_t i; + uint32_t temp; + uint32_t min = max(clock_insr, (uint32_t)SMU7_MINIMUM_ENGINE_CLOCK); + + PP_ASSERT_WITH_CODE((clock >= min), + "Engine clock can't satisfy stutter requirement!", + return 0); + for (i = 31; ; i--) { + temp = clock / (i + 1); + + if (temp >= min || i == 0) + break; + } + return i; +} + +static int vegam_populate_single_graphic_level(struct pp_hwmgr *hwmgr, + uint32_t clock, struct SMU75_Discrete_GraphicsLevel *level) +{ + int result; + /* PP_Clocks minClocks; */ + uint32_t mvdd; + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + SMU_SclkSetting curr_sclk_setting = { 0 }; + + result = vegam_calculate_sclk_params(hwmgr, clock, &curr_sclk_setting); + + /* populate graphics levels */ + result = vegam_get_dependency_volt_by_clk(hwmgr, + table_info->vdd_dep_on_sclk, clock, + &level->MinVoltage, &mvdd); + + PP_ASSERT_WITH_CODE((0 == result), + "can not find VDDC voltage value for " + "VDDC engine clock dependency table", + return result); + level->ActivityLevel = (uint16_t)(SclkDPMTuning_VEGAM >> DPMTuning_Activity_Shift); + + level->CcPwrDynRm = 0; + level->CcPwrDynRm1 = 0; + level->EnabledForActivity = 0; + level->EnabledForThrottle = 1; + level->VoltageDownHyst = 0; + level->PowerThrottle = 0; + data->display_timing.min_clock_in_sr = hwmgr->display_config->min_core_set_clock_in_sr; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, PHM_PlatformCaps_SclkDeepSleep)) + level->DeepSleepDivId = vegam_get_sleep_divider_id_from_clock(clock, + hwmgr->display_config->min_core_set_clock_in_sr); + + level->SclkSetting = curr_sclk_setting; + + CONVERT_FROM_HOST_TO_SMC_UL(level->MinVoltage); + CONVERT_FROM_HOST_TO_SMC_UL(level->CcPwrDynRm); + CONVERT_FROM_HOST_TO_SMC_UL(level->CcPwrDynRm1); + CONVERT_FROM_HOST_TO_SMC_US(level->ActivityLevel); + CONVERT_FROM_HOST_TO_SMC_UL(level->SclkSetting.SclkFrequency); + CONVERT_FROM_HOST_TO_SMC_US(level->SclkSetting.Fcw_int); + CONVERT_FROM_HOST_TO_SMC_US(level->SclkSetting.Fcw_frac); + CONVERT_FROM_HOST_TO_SMC_US(level->SclkSetting.Pcc_fcw_int); + CONVERT_FROM_HOST_TO_SMC_US(level->SclkSetting.Sclk_slew_rate); + CONVERT_FROM_HOST_TO_SMC_US(level->SclkSetting.Pcc_up_slew_rate); + CONVERT_FROM_HOST_TO_SMC_US(level->SclkSetting.Pcc_down_slew_rate); + CONVERT_FROM_HOST_TO_SMC_US(level->SclkSetting.Fcw1_int); + CONVERT_FROM_HOST_TO_SMC_US(level->SclkSetting.Fcw1_frac); + CONVERT_FROM_HOST_TO_SMC_US(level->SclkSetting.Sclk_ss_slew_rate); + return 0; +} + +static int vegam_populate_all_graphic_levels(struct pp_hwmgr *hwmgr) +{ + struct smu7_hwmgr *hw_data = (struct smu7_hwmgr *)(hwmgr->backend); + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + struct smu7_dpm_table *dpm_table = &hw_data->dpm_table; + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct phm_ppt_v1_pcie_table *pcie_table = table_info->pcie_table; + uint8_t pcie_entry_cnt = (uint8_t) hw_data->dpm_table.pcie_speed_table.count; + int result = 0; + uint32_t array = smu_data->smu7_data.dpm_table_start + + offsetof(SMU75_Discrete_DpmTable, GraphicsLevel); + uint32_t array_size = sizeof(struct SMU75_Discrete_GraphicsLevel) * + SMU75_MAX_LEVELS_GRAPHICS; + struct SMU75_Discrete_GraphicsLevel *levels = + smu_data->smc_state_table.GraphicsLevel; + uint32_t i, max_entry; + uint8_t hightest_pcie_level_enabled = 0, + lowest_pcie_level_enabled = 0, + mid_pcie_level_enabled = 0, + count = 0; + + vegam_get_sclk_range_table(hwmgr, &(smu_data->smc_state_table)); + + for (i = 0; i < dpm_table->sclk_table.count; i++) { + + result = vegam_populate_single_graphic_level(hwmgr, + dpm_table->sclk_table.dpm_levels[i].value, + &(smu_data->smc_state_table.GraphicsLevel[i])); + if (result) + return result; + + levels[i].UpHyst = (uint8_t) + (SclkDPMTuning_VEGAM >> DPMTuning_Uphyst_Shift); + levels[i].DownHyst = (uint8_t) + (SclkDPMTuning_VEGAM >> DPMTuning_Downhyst_Shift); + /* Making sure only DPM level 0-1 have Deep Sleep Div ID populated. */ + if (i > 1) + levels[i].DeepSleepDivId = 0; + } + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SPLLShutdownSupport)) + smu_data->smc_state_table.GraphicsLevel[0].SclkSetting.SSc_En = 0; + + smu_data->smc_state_table.GraphicsDpmLevelCount = + (uint8_t)dpm_table->sclk_table.count; + hw_data->dpm_level_enable_mask.sclk_dpm_enable_mask = + phm_get_dpm_level_enable_mask_value(&dpm_table->sclk_table); + + for (i = 0; i < dpm_table->sclk_table.count; i++) + levels[i].EnabledForActivity = + (hw_data->dpm_level_enable_mask.sclk_dpm_enable_mask >> i) & 0x1; + + if (pcie_table != NULL) { + PP_ASSERT_WITH_CODE((1 <= pcie_entry_cnt), + "There must be 1 or more PCIE levels defined in PPTable.", + return -EINVAL); + max_entry = pcie_entry_cnt - 1; + for (i = 0; i < dpm_table->sclk_table.count; i++) + levels[i].pcieDpmLevel = + (uint8_t) ((i < max_entry) ? i : max_entry); + } else { + while (hw_data->dpm_level_enable_mask.pcie_dpm_enable_mask && + ((hw_data->dpm_level_enable_mask.pcie_dpm_enable_mask & + (1 << (hightest_pcie_level_enabled + 1))) != 0)) + hightest_pcie_level_enabled++; + + while (hw_data->dpm_level_enable_mask.pcie_dpm_enable_mask && + ((hw_data->dpm_level_enable_mask.pcie_dpm_enable_mask & + (1 << lowest_pcie_level_enabled)) == 0)) + lowest_pcie_level_enabled++; + + while ((count < hightest_pcie_level_enabled) && + ((hw_data->dpm_level_enable_mask.pcie_dpm_enable_mask & + (1 << (lowest_pcie_level_enabled + 1 + count))) == 0)) + count++; + + mid_pcie_level_enabled = (lowest_pcie_level_enabled + 1 + count) < + hightest_pcie_level_enabled ? + (lowest_pcie_level_enabled + 1 + count) : + hightest_pcie_level_enabled; + + /* set pcieDpmLevel to hightest_pcie_level_enabled */ + for (i = 2; i < dpm_table->sclk_table.count; i++) + levels[i].pcieDpmLevel = hightest_pcie_level_enabled; + + /* set pcieDpmLevel to lowest_pcie_level_enabled */ + levels[0].pcieDpmLevel = lowest_pcie_level_enabled; + + /* set pcieDpmLevel to mid_pcie_level_enabled */ + levels[1].pcieDpmLevel = mid_pcie_level_enabled; + } + /* level count will send to smc once at init smc table and never change */ + result = smu7_copy_bytes_to_smc(hwmgr, array, (uint8_t *)levels, + (uint32_t)array_size, SMC_RAM_END); + + return result; +} + +static int vegam_calculate_mclk_params(struct pp_hwmgr *hwmgr, + uint32_t clock, struct SMU75_Discrete_MemoryLevel *mem_level) +{ + struct pp_atomctrl_memory_clock_param_ai mpll_param; + + PP_ASSERT_WITH_CODE(!atomctrl_get_memory_pll_dividers_ai(hwmgr, + clock, &mpll_param), + "Failed to retrieve memory pll parameter.", + return -EINVAL); + + mem_level->MclkFrequency = (uint32_t)mpll_param.ulClock; + mem_level->Fcw_int = (uint16_t)mpll_param.ulMclk_fcw_int; + mem_level->Fcw_frac = (uint16_t)mpll_param.ulMclk_fcw_frac; + mem_level->Postdiv = (uint8_t)mpll_param.ulPostDiv; + + return 0; +} + +static int vegam_populate_single_memory_level(struct pp_hwmgr *hwmgr, + uint32_t clock, struct SMU75_Discrete_MemoryLevel *mem_level) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + int result = 0; + uint32_t mclk_stutter_mode_threshold = 60000; + + + if (table_info->vdd_dep_on_mclk) { + result = vegam_get_dependency_volt_by_clk(hwmgr, + table_info->vdd_dep_on_mclk, clock, + &mem_level->MinVoltage, &mem_level->MinMvdd); + PP_ASSERT_WITH_CODE(!result, + "can not find MinVddc voltage value from memory " + "VDDC voltage dependency table", return result); + } + + result = vegam_calculate_mclk_params(hwmgr, clock, mem_level); + PP_ASSERT_WITH_CODE(!result, + "Failed to calculate mclk params.", + return -EINVAL); + + mem_level->EnabledForThrottle = 1; + mem_level->EnabledForActivity = 0; + mem_level->VoltageDownHyst = 0; + mem_level->ActivityLevel = (uint16_t) + (MemoryDPMTuning_VEGAM >> DPMTuning_Activity_Shift); + mem_level->StutterEnable = false; + mem_level->DisplayWatermark = PPSMC_DISPLAY_WATERMARK_LOW; + + data->display_timing.num_existing_displays = hwmgr->display_config->num_display; + + if (mclk_stutter_mode_threshold && + (clock <= mclk_stutter_mode_threshold) && + (PHM_READ_FIELD(hwmgr->device, DPG_PIPE_STUTTER_CONTROL, + STUTTER_ENABLE) & 0x1)) + mem_level->StutterEnable = true; + + if (!result) { + CONVERT_FROM_HOST_TO_SMC_UL(mem_level->MinMvdd); + CONVERT_FROM_HOST_TO_SMC_UL(mem_level->MclkFrequency); + CONVERT_FROM_HOST_TO_SMC_US(mem_level->Fcw_int); + CONVERT_FROM_HOST_TO_SMC_US(mem_level->Fcw_frac); + CONVERT_FROM_HOST_TO_SMC_US(mem_level->ActivityLevel); + CONVERT_FROM_HOST_TO_SMC_UL(mem_level->MinVoltage); + } + + return result; +} + +static int vegam_populate_all_memory_levels(struct pp_hwmgr *hwmgr) +{ + struct smu7_hwmgr *hw_data = (struct smu7_hwmgr *)(hwmgr->backend); + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + struct smu7_dpm_table *dpm_table = &hw_data->dpm_table; + int result; + /* populate MCLK dpm table to SMU7 */ + uint32_t array = smu_data->smu7_data.dpm_table_start + + offsetof(SMU75_Discrete_DpmTable, MemoryLevel); + uint32_t array_size = sizeof(SMU75_Discrete_MemoryLevel) * + SMU75_MAX_LEVELS_MEMORY; + struct SMU75_Discrete_MemoryLevel *levels = + smu_data->smc_state_table.MemoryLevel; + uint32_t i; + + for (i = 0; i < dpm_table->mclk_table.count; i++) { + PP_ASSERT_WITH_CODE((0 != dpm_table->mclk_table.dpm_levels[i].value), + "can not populate memory level as memory clock is zero", + return -EINVAL); + result = vegam_populate_single_memory_level(hwmgr, + dpm_table->mclk_table.dpm_levels[i].value, + &levels[i]); + + if (result) + return result; + + levels[i].UpHyst = (uint8_t) + (MemoryDPMTuning_VEGAM >> DPMTuning_Uphyst_Shift); + levels[i].DownHyst = (uint8_t) + (MemoryDPMTuning_VEGAM >> DPMTuning_Downhyst_Shift); + } + + smu_data->smc_state_table.MemoryDpmLevelCount = + (uint8_t)dpm_table->mclk_table.count; + hw_data->dpm_level_enable_mask.mclk_dpm_enable_mask = + phm_get_dpm_level_enable_mask_value(&dpm_table->mclk_table); + + for (i = 0; i < dpm_table->mclk_table.count; i++) + levels[i].EnabledForActivity = + (hw_data->dpm_level_enable_mask.mclk_dpm_enable_mask >> i) & 0x1; + + levels[dpm_table->mclk_table.count - 1].DisplayWatermark = + PPSMC_DISPLAY_WATERMARK_HIGH; + + /* level count will send to smc once at init smc table and never change */ + result = smu7_copy_bytes_to_smc(hwmgr, array, (uint8_t *)levels, + (uint32_t)array_size, SMC_RAM_END); + + return result; +} + +static int vegam_populate_mvdd_value(struct pp_hwmgr *hwmgr, + uint32_t mclk, SMIO_Pattern *smio_pat) +{ + const struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + uint32_t i = 0; + + if (SMU7_VOLTAGE_CONTROL_NONE != data->mvdd_control) { + /* find mvdd value which clock is more than request */ + for (i = 0; i < table_info->vdd_dep_on_mclk->count; i++) { + if (mclk <= table_info->vdd_dep_on_mclk->entries[i].clk) { + smio_pat->Voltage = data->mvdd_voltage_table.entries[i].value; + break; + } + } + PP_ASSERT_WITH_CODE(i < table_info->vdd_dep_on_mclk->count, + "MVDD Voltage is outside the supported range.", + return -EINVAL); + } else + return -EINVAL; + + return 0; +} + +static int vegam_populate_smc_acpi_level(struct pp_hwmgr *hwmgr, + SMU75_Discrete_DpmTable *table) +{ + int result = 0; + uint32_t sclk_frequency; + const struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + SMIO_Pattern vol_level; + uint32_t mvdd; + uint16_t us_mvdd; + + table->ACPILevel.Flags &= ~PPSMC_SWSTATE_FLAG_DC; + + /* Get MinVoltage and Frequency from DPM0, + * already converted to SMC_UL */ + sclk_frequency = data->vbios_boot_state.sclk_bootup_value; + result = vegam_get_dependency_volt_by_clk(hwmgr, + table_info->vdd_dep_on_sclk, + sclk_frequency, + &table->ACPILevel.MinVoltage, &mvdd); + PP_ASSERT_WITH_CODE(!result, + "Cannot find ACPI VDDC voltage value " + "in Clock Dependency Table", + ); + + result = vegam_calculate_sclk_params(hwmgr, sclk_frequency, + &(table->ACPILevel.SclkSetting)); + PP_ASSERT_WITH_CODE(!result, + "Error retrieving Engine Clock dividers from VBIOS.", + return result); + + table->ACPILevel.DeepSleepDivId = 0; + table->ACPILevel.CcPwrDynRm = 0; + table->ACPILevel.CcPwrDynRm1 = 0; + + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.Flags); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.MinVoltage); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.CcPwrDynRm); + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.CcPwrDynRm1); + + CONVERT_FROM_HOST_TO_SMC_UL(table->ACPILevel.SclkSetting.SclkFrequency); + CONVERT_FROM_HOST_TO_SMC_US(table->ACPILevel.SclkSetting.Fcw_int); + CONVERT_FROM_HOST_TO_SMC_US(table->ACPILevel.SclkSetting.Fcw_frac); + CONVERT_FROM_HOST_TO_SMC_US(table->ACPILevel.SclkSetting.Pcc_fcw_int); + CONVERT_FROM_HOST_TO_SMC_US(table->ACPILevel.SclkSetting.Sclk_slew_rate); + CONVERT_FROM_HOST_TO_SMC_US(table->ACPILevel.SclkSetting.Pcc_up_slew_rate); + CONVERT_FROM_HOST_TO_SMC_US(table->ACPILevel.SclkSetting.Pcc_down_slew_rate); + CONVERT_FROM_HOST_TO_SMC_US(table->ACPILevel.SclkSetting.Fcw1_int); + CONVERT_FROM_HOST_TO_SMC_US(table->ACPILevel.SclkSetting.Fcw1_frac); + CONVERT_FROM_HOST_TO_SMC_US(table->ACPILevel.SclkSetting.Sclk_ss_slew_rate); + + + /* Get MinVoltage and Frequency from DPM0, already converted to SMC_UL */ + table->MemoryACPILevel.MclkFrequency = data->vbios_boot_state.mclk_bootup_value; + result = vegam_get_dependency_volt_by_clk(hwmgr, + table_info->vdd_dep_on_mclk, + table->MemoryACPILevel.MclkFrequency, + &table->MemoryACPILevel.MinVoltage, &mvdd); + PP_ASSERT_WITH_CODE((0 == result), + "Cannot find ACPI VDDCI voltage value " + "in Clock Dependency Table", + ); + + us_mvdd = 0; + if ((SMU7_VOLTAGE_CONTROL_NONE == data->mvdd_control) || + (data->mclk_dpm_key_disabled)) + us_mvdd = data->vbios_boot_state.mvdd_bootup_value; + else { + if (!vegam_populate_mvdd_value(hwmgr, + data->dpm_table.mclk_table.dpm_levels[0].value, + &vol_level)) + us_mvdd = vol_level.Voltage; + } + + if (!vegam_populate_mvdd_value(hwmgr, 0, &vol_level)) + table->MemoryACPILevel.MinMvdd = PP_HOST_TO_SMC_UL(vol_level.Voltage); + else + table->MemoryACPILevel.MinMvdd = 0; + + table->MemoryACPILevel.StutterEnable = false; + + table->MemoryACPILevel.EnabledForThrottle = 0; + table->MemoryACPILevel.EnabledForActivity = 0; + table->MemoryACPILevel.UpHyst = 0; + table->MemoryACPILevel.DownHyst = 100; + table->MemoryACPILevel.VoltageDownHyst = 0; + table->MemoryACPILevel.ActivityLevel = + PP_HOST_TO_SMC_US(data->current_profile_setting.mclk_activity); + + CONVERT_FROM_HOST_TO_SMC_UL(table->MemoryACPILevel.MclkFrequency); + CONVERT_FROM_HOST_TO_SMC_UL(table->MemoryACPILevel.MinVoltage); + + return result; +} + +static int vegam_populate_smc_vce_level(struct pp_hwmgr *hwmgr, + SMU75_Discrete_DpmTable *table) +{ + int result = -EINVAL; + uint8_t count; + struct pp_atomctrl_clock_dividers_vi dividers; + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct phm_ppt_v1_mm_clock_voltage_dependency_table *mm_table = + table_info->mm_dep_table; + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + uint32_t vddci; + + table->VceLevelCount = (uint8_t)(mm_table->count); + table->VceBootLevel = 0; + + for (count = 0; count < table->VceLevelCount; count++) { + table->VceLevel[count].Frequency = mm_table->entries[count].eclk; + table->VceLevel[count].MinVoltage = 0; + table->VceLevel[count].MinVoltage |= + (mm_table->entries[count].vddc * VOLTAGE_SCALE) << VDDC_SHIFT; + + if (SMU7_VOLTAGE_CONTROL_BY_GPIO == data->vddci_control) + vddci = (uint32_t)phm_find_closest_vddci(&(data->vddci_voltage_table), + mm_table->entries[count].vddc - VDDC_VDDCI_DELTA); + else if (SMU7_VOLTAGE_CONTROL_BY_SVID2 == data->vddci_control) + vddci = mm_table->entries[count].vddc - VDDC_VDDCI_DELTA; + else + vddci = (data->vbios_boot_state.vddci_bootup_value * VOLTAGE_SCALE) << VDDCI_SHIFT; + + + table->VceLevel[count].MinVoltage |= + (vddci * VOLTAGE_SCALE) << VDDCI_SHIFT; + table->VceLevel[count].MinVoltage |= 1 << PHASES_SHIFT; + + /*retrieve divider value for VBIOS */ + result = atomctrl_get_dfs_pll_dividers_vi(hwmgr, + table->VceLevel[count].Frequency, ÷rs); + PP_ASSERT_WITH_CODE((0 == result), + "can not find divide id for VCE engine clock", + return result); + + table->VceLevel[count].Divider = (uint8_t)dividers.pll_post_divider; + + CONVERT_FROM_HOST_TO_SMC_UL(table->VceLevel[count].Frequency); + CONVERT_FROM_HOST_TO_SMC_UL(table->VceLevel[count].MinVoltage); + } + return result; +} + +static int vegam_populate_smc_samu_level(struct pp_hwmgr *hwmgr, + SMU75_Discrete_DpmTable *table) +{ + int result = -EINVAL; + uint8_t count; + struct pp_atomctrl_clock_dividers_vi dividers; + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct phm_ppt_v1_mm_clock_voltage_dependency_table *mm_table = + table_info->mm_dep_table; + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + uint32_t vddci; + + table->SamuBootLevel = 0; + table->SamuLevelCount = (uint8_t)(mm_table->count); + + for (count = 0; count < table->SamuLevelCount; count++) { + /* not sure whether we need evclk or not */ + table->SamuLevel[count].MinVoltage = 0; + table->SamuLevel[count].Frequency = mm_table->entries[count].samclock; + table->SamuLevel[count].MinVoltage |= (mm_table->entries[count].vddc * + VOLTAGE_SCALE) << VDDC_SHIFT; + + if (SMU7_VOLTAGE_CONTROL_BY_GPIO == data->vddci_control) + vddci = (uint32_t)phm_find_closest_vddci(&(data->vddci_voltage_table), + mm_table->entries[count].vddc - VDDC_VDDCI_DELTA); + else if (SMU7_VOLTAGE_CONTROL_BY_SVID2 == data->vddci_control) + vddci = mm_table->entries[count].vddc - VDDC_VDDCI_DELTA; + else + vddci = (data->vbios_boot_state.vddci_bootup_value * VOLTAGE_SCALE) << VDDCI_SHIFT; + + table->SamuLevel[count].MinVoltage |= (vddci * VOLTAGE_SCALE) << VDDCI_SHIFT; + table->SamuLevel[count].MinVoltage |= 1 << PHASES_SHIFT; + + /* retrieve divider value for VBIOS */ + result = atomctrl_get_dfs_pll_dividers_vi(hwmgr, + table->SamuLevel[count].Frequency, ÷rs); + PP_ASSERT_WITH_CODE((0 == result), + "can not find divide id for samu clock", return result); + + table->SamuLevel[count].Divider = (uint8_t)dividers.pll_post_divider; + + CONVERT_FROM_HOST_TO_SMC_UL(table->SamuLevel[count].Frequency); + CONVERT_FROM_HOST_TO_SMC_UL(table->SamuLevel[count].MinVoltage); + } + return result; +} + +static int vegam_populate_memory_timing_parameters(struct pp_hwmgr *hwmgr, + int32_t eng_clock, int32_t mem_clock, + SMU75_Discrete_MCArbDramTimingTableEntry *arb_regs) +{ + uint32_t dram_timing; + uint32_t dram_timing2; + uint32_t burst_time; + uint32_t rfsh_rate; + uint32_t misc3; + + int result; + + result = atomctrl_set_engine_dram_timings_rv770(hwmgr, + eng_clock, mem_clock); + PP_ASSERT_WITH_CODE(result == 0, + "Error calling VBIOS to set DRAM_TIMING.", + return result); + + dram_timing = cgs_read_register(hwmgr->device, mmMC_ARB_DRAM_TIMING); + dram_timing2 = cgs_read_register(hwmgr->device, mmMC_ARB_DRAM_TIMING2); + burst_time = cgs_read_register(hwmgr->device, mmMC_ARB_BURST_TIME); + rfsh_rate = cgs_read_register(hwmgr->device, mmMC_ARB_RFSH_RATE); + misc3 = cgs_read_register(hwmgr->device, mmMC_ARB_MISC3); + + arb_regs->McArbDramTiming = PP_HOST_TO_SMC_UL(dram_timing); + arb_regs->McArbDramTiming2 = PP_HOST_TO_SMC_UL(dram_timing2); + arb_regs->McArbBurstTime = PP_HOST_TO_SMC_UL(burst_time); + arb_regs->McArbRfshRate = PP_HOST_TO_SMC_UL(rfsh_rate); + arb_regs->McArbMisc3 = PP_HOST_TO_SMC_UL(misc3); + + return 0; +} + +static int vegam_program_memory_timing_parameters(struct pp_hwmgr *hwmgr) +{ + struct smu7_hwmgr *hw_data = (struct smu7_hwmgr *)(hwmgr->backend); + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + struct SMU75_Discrete_MCArbDramTimingTable arb_regs; + uint32_t i, j; + int result = 0; + + memset(&arb_regs, 0, sizeof(SMU75_Discrete_MCArbDramTimingTable)); + + for (i = 0; i < hw_data->dpm_table.sclk_table.count; i++) { + for (j = 0; j < hw_data->dpm_table.mclk_table.count; j++) { + result = vegam_populate_memory_timing_parameters(hwmgr, + hw_data->dpm_table.sclk_table.dpm_levels[i].value, + hw_data->dpm_table.mclk_table.dpm_levels[j].value, + &arb_regs.entries[i][j]); + if (result) + return result; + } + } + + result = smu7_copy_bytes_to_smc( + hwmgr, + smu_data->smu7_data.arb_table_start, + (uint8_t *)&arb_regs, + sizeof(SMU75_Discrete_MCArbDramTimingTable), + SMC_RAM_END); + return result; +} + +static int vegam_populate_smc_uvd_level(struct pp_hwmgr *hwmgr, + struct SMU75_Discrete_DpmTable *table) +{ + int result = -EINVAL; + uint8_t count; + struct pp_atomctrl_clock_dividers_vi dividers; + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct phm_ppt_v1_mm_clock_voltage_dependency_table *mm_table = + table_info->mm_dep_table; + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + uint32_t vddci; + + table->UvdLevelCount = (uint8_t)(mm_table->count); + table->UvdBootLevel = 0; + + for (count = 0; count < table->UvdLevelCount; count++) { + table->UvdLevel[count].MinVoltage = 0; + table->UvdLevel[count].VclkFrequency = mm_table->entries[count].vclk; + table->UvdLevel[count].DclkFrequency = mm_table->entries[count].dclk; + table->UvdLevel[count].MinVoltage |= + (mm_table->entries[count].vddc * VOLTAGE_SCALE) << VDDC_SHIFT; + + if (SMU7_VOLTAGE_CONTROL_BY_GPIO == data->vddci_control) + vddci = (uint32_t)phm_find_closest_vddci(&(data->vddci_voltage_table), + mm_table->entries[count].vddc - VDDC_VDDCI_DELTA); + else if (SMU7_VOLTAGE_CONTROL_BY_SVID2 == data->vddci_control) + vddci = mm_table->entries[count].vddc - VDDC_VDDCI_DELTA; + else + vddci = (data->vbios_boot_state.vddci_bootup_value * VOLTAGE_SCALE) << VDDCI_SHIFT; + + table->UvdLevel[count].MinVoltage |= (vddci * VOLTAGE_SCALE) << VDDCI_SHIFT; + table->UvdLevel[count].MinVoltage |= 1 << PHASES_SHIFT; + + /* retrieve divider value for VBIOS */ + result = atomctrl_get_dfs_pll_dividers_vi(hwmgr, + table->UvdLevel[count].VclkFrequency, ÷rs); + PP_ASSERT_WITH_CODE((0 == result), + "can not find divide id for Vclk clock", return result); + + table->UvdLevel[count].VclkDivider = (uint8_t)dividers.pll_post_divider; + + result = atomctrl_get_dfs_pll_dividers_vi(hwmgr, + table->UvdLevel[count].DclkFrequency, ÷rs); + PP_ASSERT_WITH_CODE((0 == result), + "can not find divide id for Dclk clock", return result); + + table->UvdLevel[count].DclkDivider = (uint8_t)dividers.pll_post_divider; + + CONVERT_FROM_HOST_TO_SMC_UL(table->UvdLevel[count].VclkFrequency); + CONVERT_FROM_HOST_TO_SMC_UL(table->UvdLevel[count].DclkFrequency); + CONVERT_FROM_HOST_TO_SMC_UL(table->UvdLevel[count].MinVoltage); + } + + return result; +} + +static int vegam_populate_smc_boot_level(struct pp_hwmgr *hwmgr, + struct SMU75_Discrete_DpmTable *table) +{ + int result = 0; + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + + table->GraphicsBootLevel = 0; + table->MemoryBootLevel = 0; + + /* find boot level from dpm table */ + result = phm_find_boot_level(&(data->dpm_table.sclk_table), + data->vbios_boot_state.sclk_bootup_value, + (uint32_t *)&(table->GraphicsBootLevel)); + + result = phm_find_boot_level(&(data->dpm_table.mclk_table), + data->vbios_boot_state.mclk_bootup_value, + (uint32_t *)&(table->MemoryBootLevel)); + + table->BootVddc = data->vbios_boot_state.vddc_bootup_value * + VOLTAGE_SCALE; + table->BootVddci = data->vbios_boot_state.vddci_bootup_value * + VOLTAGE_SCALE; + table->BootMVdd = data->vbios_boot_state.mvdd_bootup_value * + VOLTAGE_SCALE; + + CONVERT_FROM_HOST_TO_SMC_US(table->BootVddc); + CONVERT_FROM_HOST_TO_SMC_US(table->BootVddci); + CONVERT_FROM_HOST_TO_SMC_US(table->BootMVdd); + + return 0; +} + +static int vegam_populate_smc_initial_state(struct pp_hwmgr *hwmgr) +{ + struct smu7_hwmgr *hw_data = (struct smu7_hwmgr *)(hwmgr->backend); + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + uint8_t count, level; + + count = (uint8_t)(table_info->vdd_dep_on_sclk->count); + + for (level = 0; level < count; level++) { + if (table_info->vdd_dep_on_sclk->entries[level].clk >= + hw_data->vbios_boot_state.sclk_bootup_value) { + smu_data->smc_state_table.GraphicsBootLevel = level; + break; + } + } + + count = (uint8_t)(table_info->vdd_dep_on_mclk->count); + for (level = 0; level < count; level++) { + if (table_info->vdd_dep_on_mclk->entries[level].clk >= + hw_data->vbios_boot_state.mclk_bootup_value) { + smu_data->smc_state_table.MemoryBootLevel = level; + break; + } + } + + return 0; +} + +static uint16_t scale_fan_gain_settings(uint16_t raw_setting) +{ + uint32_t tmp; + tmp = raw_setting * 4096 / 100; + return (uint16_t)tmp; +} + +static int vegam_populate_bapm_parameters_in_dpm_table(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + + const struct vegam_pt_defaults *defaults = smu_data->power_tune_defaults; + SMU75_Discrete_DpmTable *table = &(smu_data->smc_state_table); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct phm_cac_tdp_table *cac_dtp_table = table_info->cac_dtp_table; + struct pp_advance_fan_control_parameters *fan_table = + &hwmgr->thermal_controller.advanceFanControlParameters; + int i, j, k; + const uint16_t *pdef1; + const uint16_t *pdef2; + + table->DefaultTdp = PP_HOST_TO_SMC_US((uint16_t)(cac_dtp_table->usTDP * 128)); + table->TargetTdp = PP_HOST_TO_SMC_US((uint16_t)(cac_dtp_table->usTDP * 128)); + + PP_ASSERT_WITH_CODE(cac_dtp_table->usTargetOperatingTemp <= 255, + "Target Operating Temp is out of Range!", + ); + + table->TemperatureLimitEdge = PP_HOST_TO_SMC_US( + cac_dtp_table->usTargetOperatingTemp * 256); + table->TemperatureLimitHotspot = PP_HOST_TO_SMC_US( + cac_dtp_table->usTemperatureLimitHotspot * 256); + table->FanGainEdge = PP_HOST_TO_SMC_US( + scale_fan_gain_settings(fan_table->usFanGainEdge)); + table->FanGainHotspot = PP_HOST_TO_SMC_US( + scale_fan_gain_settings(fan_table->usFanGainHotspot)); + + pdef1 = defaults->BAPMTI_R; + pdef2 = defaults->BAPMTI_RC; + + for (i = 0; i < SMU75_DTE_ITERATIONS; i++) { + for (j = 0; j < SMU75_DTE_SOURCES; j++) { + for (k = 0; k < SMU75_DTE_SINKS; k++) { + table->BAPMTI_R[i][j][k] = PP_HOST_TO_SMC_US(*pdef1); + table->BAPMTI_RC[i][j][k] = PP_HOST_TO_SMC_US(*pdef2); + pdef1++; + pdef2++; + } + } + } + + return 0; +} + +static int vegam_populate_clock_stretcher_data_table(struct pp_hwmgr *hwmgr) +{ + uint32_t ro, efuse, volt_without_cks, volt_with_cks, value, max, min; + struct vegam_smumgr *smu_data = + (struct vegam_smumgr *)(hwmgr->smu_backend); + + uint8_t i, stretch_amount, stretch_amount2, volt_offset = 0; + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct phm_ppt_v1_clock_voltage_dependency_table *sclk_table = + table_info->vdd_dep_on_sclk; + uint32_t mask = (1 << ((STRAP_ASIC_RO_MSB - STRAP_ASIC_RO_LSB) + 1)) - 1; + + stretch_amount = (uint8_t)table_info->cac_dtp_table->usClockStretchAmount; + + atomctrl_read_efuse(hwmgr, STRAP_ASIC_RO_LSB, STRAP_ASIC_RO_MSB, + mask, &efuse); + + min = 1200; + max = 2500; + + ro = efuse * (max - min) / 255 + min; + + /* Populate Sclk_CKS_masterEn0_7 and Sclk_voltageOffset */ + for (i = 0; i < sclk_table->count; i++) { + smu_data->smc_state_table.Sclk_CKS_masterEn0_7 |= + sclk_table->entries[i].cks_enable << i; + volt_without_cks = (uint32_t)((2753594000U + (sclk_table->entries[i].clk/100) * + 136418 - (ro - 70) * 1000000) / + (2424180 - (sclk_table->entries[i].clk/100) * 1132925/1000)); + volt_with_cks = (uint32_t)((2797202000U + sclk_table->entries[i].clk/100 * + 3232 - (ro - 65) * 1000000) / + (2522480 - sclk_table->entries[i].clk/100 * 115764/100)); + + if (volt_without_cks >= volt_with_cks) + volt_offset = (uint8_t)(((volt_without_cks - volt_with_cks + + sclk_table->entries[i].cks_voffset) * 100 + 624) / 625); + + smu_data->smc_state_table.Sclk_voltageOffset[i] = volt_offset; + } + + smu_data->smc_state_table.LdoRefSel = + (table_info->cac_dtp_table->ucCKS_LDO_REFSEL != 0) ? + table_info->cac_dtp_table->ucCKS_LDO_REFSEL : 5; + /* Populate CKS Lookup Table */ + if (stretch_amount == 1 || stretch_amount == 2 || stretch_amount == 5) + stretch_amount2 = 0; + else if (stretch_amount == 3 || stretch_amount == 4) + stretch_amount2 = 1; + else { + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_ClockStretcher); + PP_ASSERT_WITH_CODE(false, + "Stretch Amount in PPTable not supported\n", + return -EINVAL); + } + + value = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixPWR_CKS_CNTL); + value &= 0xFFFFFFFE; + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, ixPWR_CKS_CNTL, value); + + return 0; +} + +static bool vegam_is_hw_avfs_present(struct pp_hwmgr *hwmgr) +{ + uint32_t efuse; + + efuse = cgs_read_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixSMU_EFUSE_0 + (49 * 4)); + efuse &= 0x00000001; + + if (efuse) + return true; + + return false; +} + +static int vegam_populate_avfs_parameters(struct pp_hwmgr *hwmgr) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + + SMU75_Discrete_DpmTable *table = &(smu_data->smc_state_table); + int result = 0; + struct pp_atom_ctrl__avfs_parameters avfs_params = {0}; + AVFS_meanNsigma_t AVFS_meanNsigma = { {0} }; + AVFS_Sclk_Offset_t AVFS_SclkOffset = { {0} }; + uint32_t tmp, i; + + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)hwmgr->pptable; + struct phm_ppt_v1_clock_voltage_dependency_table *sclk_table = + table_info->vdd_dep_on_sclk; + + if (!hwmgr->avfs_supported) + return 0; + + result = atomctrl_get_avfs_information(hwmgr, &avfs_params); + + if (0 == result) { + table->BTCGB_VDROOP_TABLE[0].a0 = + PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSON_a0); + table->BTCGB_VDROOP_TABLE[0].a1 = + PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSON_a1); + table->BTCGB_VDROOP_TABLE[0].a2 = + PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSON_a2); + table->BTCGB_VDROOP_TABLE[1].a0 = + PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a0); + table->BTCGB_VDROOP_TABLE[1].a1 = + PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a1); + table->BTCGB_VDROOP_TABLE[1].a2 = + PP_HOST_TO_SMC_UL(avfs_params.ulGB_VDROOP_TABLE_CKSOFF_a2); + table->AVFSGB_FUSE_TABLE[0].m1 = + PP_HOST_TO_SMC_UL(avfs_params.ulAVFSGB_FUSE_TABLE_CKSON_m1); + table->AVFSGB_FUSE_TABLE[0].m2 = + PP_HOST_TO_SMC_US(avfs_params.usAVFSGB_FUSE_TABLE_CKSON_m2); + table->AVFSGB_FUSE_TABLE[0].b = + PP_HOST_TO_SMC_UL(avfs_params.ulAVFSGB_FUSE_TABLE_CKSON_b); + table->AVFSGB_FUSE_TABLE[0].m1_shift = 24; + table->AVFSGB_FUSE_TABLE[0].m2_shift = 12; + table->AVFSGB_FUSE_TABLE[1].m1 = + PP_HOST_TO_SMC_UL(avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_m1); + table->AVFSGB_FUSE_TABLE[1].m2 = + PP_HOST_TO_SMC_US(avfs_params.usAVFSGB_FUSE_TABLE_CKSOFF_m2); + table->AVFSGB_FUSE_TABLE[1].b = + PP_HOST_TO_SMC_UL(avfs_params.ulAVFSGB_FUSE_TABLE_CKSOFF_b); + table->AVFSGB_FUSE_TABLE[1].m1_shift = 24; + table->AVFSGB_FUSE_TABLE[1].m2_shift = 12; + table->MaxVoltage = PP_HOST_TO_SMC_US(avfs_params.usMaxVoltage_0_25mv); + AVFS_meanNsigma.Aconstant[0] = + PP_HOST_TO_SMC_UL(avfs_params.ulAVFS_meanNsigma_Acontant0); + AVFS_meanNsigma.Aconstant[1] = + PP_HOST_TO_SMC_UL(avfs_params.ulAVFS_meanNsigma_Acontant1); + AVFS_meanNsigma.Aconstant[2] = + PP_HOST_TO_SMC_UL(avfs_params.ulAVFS_meanNsigma_Acontant2); + AVFS_meanNsigma.DC_tol_sigma = + PP_HOST_TO_SMC_US(avfs_params.usAVFS_meanNsigma_DC_tol_sigma); + AVFS_meanNsigma.Platform_mean = + PP_HOST_TO_SMC_US(avfs_params.usAVFS_meanNsigma_Platform_mean); + AVFS_meanNsigma.PSM_Age_CompFactor = + PP_HOST_TO_SMC_US(avfs_params.usPSM_Age_ComFactor); + AVFS_meanNsigma.Platform_sigma = + PP_HOST_TO_SMC_US(avfs_params.usAVFS_meanNsigma_Platform_sigma); + + for (i = 0; i < sclk_table->count; i++) { + AVFS_meanNsigma.Static_Voltage_Offset[i] = + (uint8_t)(sclk_table->entries[i].cks_voffset * 100 / 625); + AVFS_SclkOffset.Sclk_Offset[i] = + PP_HOST_TO_SMC_US((uint16_t) + (sclk_table->entries[i].sclk_offset) / 100); + } + + result = smu7_read_smc_sram_dword(hwmgr, + SMU7_FIRMWARE_HEADER_LOCATION + + offsetof(SMU75_Firmware_Header, AvfsMeanNSigma), + &tmp, SMC_RAM_END); + smu7_copy_bytes_to_smc(hwmgr, + tmp, + (uint8_t *)&AVFS_meanNsigma, + sizeof(AVFS_meanNsigma_t), + SMC_RAM_END); + + result = smu7_read_smc_sram_dword(hwmgr, + SMU7_FIRMWARE_HEADER_LOCATION + + offsetof(SMU75_Firmware_Header, AvfsSclkOffsetTable), + &tmp, SMC_RAM_END); + smu7_copy_bytes_to_smc(hwmgr, + tmp, + (uint8_t *)&AVFS_SclkOffset, + sizeof(AVFS_Sclk_Offset_t), + SMC_RAM_END); + + data->avfs_vdroop_override_setting = + (avfs_params.ucEnableGB_VDROOP_TABLE_CKSON << BTCGB0_Vdroop_Enable_SHIFT) | + (avfs_params.ucEnableGB_VDROOP_TABLE_CKSOFF << BTCGB1_Vdroop_Enable_SHIFT) | + (avfs_params.ucEnableGB_FUSE_TABLE_CKSON << AVFSGB0_Vdroop_Enable_SHIFT) | + (avfs_params.ucEnableGB_FUSE_TABLE_CKSOFF << AVFSGB1_Vdroop_Enable_SHIFT); + data->apply_avfs_cks_off_voltage = + (avfs_params.ucEnableApplyAVFS_CKS_OFF_Voltage == 1) ? true : false; + } + return result; +} + +static int vegam_populate_vr_config(struct pp_hwmgr *hwmgr, + struct SMU75_Discrete_DpmTable *table) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct vegam_smumgr *smu_data = + (struct vegam_smumgr *)(hwmgr->smu_backend); + uint16_t config; + + config = VR_MERGED_WITH_VDDC; + table->VRConfig |= (config << VRCONF_VDDGFX_SHIFT); + + /* Set Vddc Voltage Controller */ + if (SMU7_VOLTAGE_CONTROL_BY_SVID2 == data->voltage_control) { + config = VR_SVI2_PLANE_1; + table->VRConfig |= config; + } else { + PP_ASSERT_WITH_CODE(false, + "VDDC should be on SVI2 control in merged mode!", + ); + } + /* Set Vddci Voltage Controller */ + if (SMU7_VOLTAGE_CONTROL_BY_SVID2 == data->vddci_control) { + config = VR_SVI2_PLANE_2; /* only in merged mode */ + table->VRConfig |= (config << VRCONF_VDDCI_SHIFT); + } else if (SMU7_VOLTAGE_CONTROL_BY_GPIO == data->vddci_control) { + config = VR_SMIO_PATTERN_1; + table->VRConfig |= (config << VRCONF_VDDCI_SHIFT); + } else { + config = VR_STATIC_VOLTAGE; + table->VRConfig |= (config << VRCONF_VDDCI_SHIFT); + } + /* Set Mvdd Voltage Controller */ + if (SMU7_VOLTAGE_CONTROL_BY_SVID2 == data->mvdd_control) { + if (config != VR_SVI2_PLANE_2) { + config = VR_SVI2_PLANE_2; + table->VRConfig |= (config << VRCONF_MVDD_SHIFT); + cgs_write_ind_register(hwmgr->device, + CGS_IND_REG__SMC, + smu_data->smu7_data.soft_regs_start + + offsetof(SMU75_SoftRegisters, AllowMvddSwitch), + 0x1); + } else { + PP_ASSERT_WITH_CODE(false, + "SVI2 Plane 2 is already taken, set MVDD as Static",); + config = VR_STATIC_VOLTAGE; + table->VRConfig = (config << VRCONF_MVDD_SHIFT); + } + } else if (SMU7_VOLTAGE_CONTROL_BY_GPIO == data->mvdd_control) { + config = VR_SMIO_PATTERN_2; + table->VRConfig = (config << VRCONF_MVDD_SHIFT); + cgs_write_ind_register(hwmgr->device, + CGS_IND_REG__SMC, + smu_data->smu7_data.soft_regs_start + + offsetof(SMU75_SoftRegisters, AllowMvddSwitch), + 0x1); + } else { + config = VR_STATIC_VOLTAGE; + table->VRConfig |= (config << VRCONF_MVDD_SHIFT); + } + + return 0; +} + +static int vegam_populate_svi_load_line(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + const struct vegam_pt_defaults *defaults = smu_data->power_tune_defaults; + + smu_data->power_tune_table.SviLoadLineEn = defaults->SviLoadLineEn; + smu_data->power_tune_table.SviLoadLineVddC = defaults->SviLoadLineVddC; + smu_data->power_tune_table.SviLoadLineTrimVddC = 3; + smu_data->power_tune_table.SviLoadLineOffsetVddC = 0; + + return 0; +} + +static int vegam_populate_tdc_limit(struct pp_hwmgr *hwmgr) +{ + uint16_t tdc_limit; + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + const struct vegam_pt_defaults *defaults = smu_data->power_tune_defaults; + + tdc_limit = (uint16_t)(table_info->cac_dtp_table->usTDC * 128); + smu_data->power_tune_table.TDC_VDDC_PkgLimit = + CONVERT_FROM_HOST_TO_SMC_US(tdc_limit); + smu_data->power_tune_table.TDC_VDDC_ThrottleReleaseLimitPerc = + defaults->TDC_VDDC_ThrottleReleaseLimitPerc; + smu_data->power_tune_table.TDC_MAWt = defaults->TDC_MAWt; + + return 0; +} + +static int vegam_populate_dw8(struct pp_hwmgr *hwmgr, uint32_t fuse_table_offset) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + const struct vegam_pt_defaults *defaults = smu_data->power_tune_defaults; + uint32_t temp; + + if (smu7_read_smc_sram_dword(hwmgr, + fuse_table_offset + + offsetof(SMU75_Discrete_PmFuses, TdcWaterfallCtl), + (uint32_t *)&temp, SMC_RAM_END)) + PP_ASSERT_WITH_CODE(false, + "Attempt to read PmFuses.DW6 (SviLoadLineEn) from SMC Failed!", + return -EINVAL); + else { + smu_data->power_tune_table.TdcWaterfallCtl = defaults->TdcWaterfallCtl; + smu_data->power_tune_table.LPMLTemperatureMin = + (uint8_t)((temp >> 16) & 0xff); + smu_data->power_tune_table.LPMLTemperatureMax = + (uint8_t)((temp >> 8) & 0xff); + smu_data->power_tune_table.Reserved = (uint8_t)(temp & 0xff); + } + return 0; +} + +static int vegam_populate_temperature_scaler(struct pp_hwmgr *hwmgr) +{ + int i; + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + + /* Currently not used. Set all to zero. */ + for (i = 0; i < 16; i++) + smu_data->power_tune_table.LPMLTemperatureScaler[i] = 0; + + return 0; +} + +static int vegam_populate_fuzzy_fan(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + +/* TO DO move to hwmgr */ + if ((hwmgr->thermal_controller.advanceFanControlParameters.usFanOutputSensitivity & (1 << 15)) + || 0 == hwmgr->thermal_controller.advanceFanControlParameters.usFanOutputSensitivity) + hwmgr->thermal_controller.advanceFanControlParameters.usFanOutputSensitivity = + hwmgr->thermal_controller.advanceFanControlParameters.usDefaultFanOutputSensitivity; + + smu_data->power_tune_table.FuzzyFan_PwmSetDelta = PP_HOST_TO_SMC_US( + hwmgr->thermal_controller.advanceFanControlParameters.usFanOutputSensitivity); + return 0; +} + +static int vegam_populate_gnb_lpml(struct pp_hwmgr *hwmgr) +{ + int i; + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + + /* Currently not used. Set all to zero. */ + for (i = 0; i < 16; i++) + smu_data->power_tune_table.GnbLPML[i] = 0; + + return 0; +} + +static int vegam_populate_bapm_vddc_base_leakage_sidd(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + uint16_t hi_sidd = smu_data->power_tune_table.BapmVddCBaseLeakageHiSidd; + uint16_t lo_sidd = smu_data->power_tune_table.BapmVddCBaseLeakageLoSidd; + struct phm_cac_tdp_table *cac_table = table_info->cac_dtp_table; + + hi_sidd = (uint16_t)(cac_table->usHighCACLeakage / 100 * 256); + lo_sidd = (uint16_t)(cac_table->usLowCACLeakage / 100 * 256); + + smu_data->power_tune_table.BapmVddCBaseLeakageHiSidd = + CONVERT_FROM_HOST_TO_SMC_US(hi_sidd); + smu_data->power_tune_table.BapmVddCBaseLeakageLoSidd = + CONVERT_FROM_HOST_TO_SMC_US(lo_sidd); + + return 0; +} + +static int vegam_populate_pm_fuses(struct pp_hwmgr *hwmgr) +{ + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + uint32_t pm_fuse_table_offset; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_PowerContainment)) { + if (smu7_read_smc_sram_dword(hwmgr, + SMU7_FIRMWARE_HEADER_LOCATION + + offsetof(SMU75_Firmware_Header, PmFuseTable), + &pm_fuse_table_offset, SMC_RAM_END)) + PP_ASSERT_WITH_CODE(false, + "Attempt to get pm_fuse_table_offset Failed!", + return -EINVAL); + + if (vegam_populate_svi_load_line(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate SviLoadLine Failed!", + return -EINVAL); + + if (vegam_populate_tdc_limit(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate TDCLimit Failed!", return -EINVAL); + + if (vegam_populate_dw8(hwmgr, pm_fuse_table_offset)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate TdcWaterfallCtl, " + "LPMLTemperature Min and Max Failed!", + return -EINVAL); + + if (0 != vegam_populate_temperature_scaler(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate LPMLTemperatureScaler Failed!", + return -EINVAL); + + if (vegam_populate_fuzzy_fan(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate Fuzzy Fan Control parameters Failed!", + return -EINVAL); + + if (vegam_populate_gnb_lpml(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate GnbLPML Failed!", + return -EINVAL); + + if (vegam_populate_bapm_vddc_base_leakage_sidd(hwmgr)) + PP_ASSERT_WITH_CODE(false, + "Attempt to populate BapmVddCBaseLeakage Hi and Lo " + "Sidd Failed!", return -EINVAL); + + if (smu7_copy_bytes_to_smc(hwmgr, pm_fuse_table_offset, + (uint8_t *)&smu_data->power_tune_table, + (sizeof(struct SMU75_Discrete_PmFuses) - PMFUSES_AVFSSIZE), + SMC_RAM_END)) + PP_ASSERT_WITH_CODE(false, + "Attempt to download PmFuseTable Failed!", + return -EINVAL); + } + return 0; +} + +static int vegam_enable_reconfig_cus(struct pp_hwmgr *hwmgr) +{ + struct amdgpu_device *adev = hwmgr->adev; + + smum_send_msg_to_smc_with_parameter(hwmgr, + PPSMC_MSG_EnableModeSwitchRLCNotification, + adev->gfx.cu_info.number); + + return 0; +} + +static int vegam_init_smc_table(struct pp_hwmgr *hwmgr) +{ + int result; + struct smu7_hwmgr *hw_data = (struct smu7_hwmgr *)(hwmgr->backend); + struct vegam_smumgr *smu_data = (struct vegam_smumgr *)(hwmgr->smu_backend); + + struct phm_ppt_v1_information *table_info = + (struct phm_ppt_v1_information *)(hwmgr->pptable); + struct SMU75_Discrete_DpmTable *table = &(smu_data->smc_state_table); + uint8_t i; + struct pp_atomctrl_gpio_pin_assignment gpio_pin; + struct phm_ppt_v1_gpio_table *gpio_table = + (struct phm_ppt_v1_gpio_table *)table_info->gpio_table; + pp_atomctrl_clock_dividers_vi dividers; + + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_AutomaticDCTransition); + + vegam_initialize_power_tune_defaults(hwmgr); + + if (SMU7_VOLTAGE_CONTROL_NONE != hw_data->voltage_control) + vegam_populate_smc_voltage_tables(hwmgr, table); + + table->SystemFlags = 0; + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_AutomaticDCTransition)) + table->SystemFlags |= PPSMC_SYSTEMFLAG_GPIO_DC; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_StepVddc)) + table->SystemFlags |= PPSMC_SYSTEMFLAG_STEPVDDC; + + if (hw_data->is_memory_gddr5) + table->SystemFlags |= PPSMC_SYSTEMFLAG_GDDR5; + + if (hw_data->ulv_supported && table_info->us_ulv_voltage_offset) { + result = vegam_populate_ulv_state(hwmgr, table); + PP_ASSERT_WITH_CODE(!result, + "Failed to initialize ULV state!", return result); + cgs_write_ind_register(hwmgr->device, CGS_IND_REG__SMC, + ixCG_ULV_PARAMETER, SMU7_CGULVPARAMETER_DFLT); + } + + result = vegam_populate_smc_link_level(hwmgr, table); + PP_ASSERT_WITH_CODE(!result, + "Failed to initialize Link Level!", return result); + + result = vegam_populate_all_graphic_levels(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "Failed to initialize Graphics Level!", return result); + + result = vegam_populate_all_memory_levels(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "Failed to initialize Memory Level!", return result); + + result = vegam_populate_smc_acpi_level(hwmgr, table); + PP_ASSERT_WITH_CODE(!result, + "Failed to initialize ACPI Level!", return result); + + result = vegam_populate_smc_vce_level(hwmgr, table); + PP_ASSERT_WITH_CODE(!result, + "Failed to initialize VCE Level!", return result); + + result = vegam_populate_smc_samu_level(hwmgr, table); + PP_ASSERT_WITH_CODE(!result, + "Failed to initialize SAMU Level!", return result); + + /* Since only the initial state is completely set up at this point + * (the other states are just copies of the boot state) we only + * need to populate the ARB settings for the initial state. + */ + result = vegam_program_memory_timing_parameters(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "Failed to Write ARB settings for the initial state.", return result); + + result = vegam_populate_smc_uvd_level(hwmgr, table); + PP_ASSERT_WITH_CODE(!result, + "Failed to initialize UVD Level!", return result); + + result = vegam_populate_smc_boot_level(hwmgr, table); + PP_ASSERT_WITH_CODE(!result, + "Failed to initialize Boot Level!", return result); + + result = vegam_populate_smc_initial_state(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "Failed to initialize Boot State!", return result); + + result = vegam_populate_bapm_parameters_in_dpm_table(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "Failed to populate BAPM Parameters!", return result); + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_ClockStretcher)) { + result = vegam_populate_clock_stretcher_data_table(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "Failed to populate Clock Stretcher Data Table!", + return result); + } + + result = vegam_populate_avfs_parameters(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "Failed to populate AVFS Parameters!", return result;); + + table->CurrSclkPllRange = 0xff; + table->GraphicsVoltageChangeEnable = 1; + table->GraphicsThermThrottleEnable = 1; + table->GraphicsInterval = 1; + table->VoltageInterval = 1; + table->ThermalInterval = 1; + table->TemperatureLimitHigh = + table_info->cac_dtp_table->usTargetOperatingTemp * + SMU7_Q88_FORMAT_CONVERSION_UNIT; + table->TemperatureLimitLow = + (table_info->cac_dtp_table->usTargetOperatingTemp - 1) * + SMU7_Q88_FORMAT_CONVERSION_UNIT; + table->MemoryVoltageChangeEnable = 1; + table->MemoryInterval = 1; + table->VoltageResponseTime = 0; + table->PhaseResponseTime = 0; + table->MemoryThermThrottleEnable = 1; + + PP_ASSERT_WITH_CODE(hw_data->dpm_table.pcie_speed_table.count >= 1, + "There must be 1 or more PCIE levels defined in PPTable.", + return -EINVAL); + table->PCIeBootLinkLevel = + hw_data->dpm_table.pcie_speed_table.count; + table->PCIeGenInterval = 1; + table->VRConfig = 0; + + result = vegam_populate_vr_config(hwmgr, table); + PP_ASSERT_WITH_CODE(!result, + "Failed to populate VRConfig setting!", return result); + + table->ThermGpio = 17; + table->SclkStepSize = 0x4000; + + if (atomctrl_get_pp_assign_pin(hwmgr, + VDDC_VRHOT_GPIO_PINID, &gpio_pin)) { + table->VRHotGpio = gpio_pin.uc_gpio_pin_bit_shift; + if (gpio_table) + table->VRHotLevel = + table_info->gpio_table->vrhot_triggered_sclk_dpm_index; + } else { + table->VRHotGpio = SMU7_UNUSED_GPIO_PIN; + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_RegulatorHot); + } + + if (atomctrl_get_pp_assign_pin(hwmgr, + PP_AC_DC_SWITCH_GPIO_PINID, &gpio_pin)) { + table->AcDcGpio = gpio_pin.uc_gpio_pin_bit_shift; + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_AutomaticDCTransition) && + !smum_send_msg_to_smc(hwmgr, PPSMC_MSG_UseNewGPIOScheme)) + phm_cap_set(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SMCtoPPLIBAcdcGpioScheme); + } else { + table->AcDcGpio = SMU7_UNUSED_GPIO_PIN; + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_AutomaticDCTransition); + } + + /* Thermal Output GPIO */ + if (atomctrl_get_pp_assign_pin(hwmgr, + THERMAL_INT_OUTPUT_GPIO_PINID, &gpio_pin)) { + table->ThermOutGpio = gpio_pin.uc_gpio_pin_bit_shift; + + /* For porlarity read GPIOPAD_A with assigned Gpio pin + * since VBIOS will program this register to set 'inactive state', + * driver can then determine 'active state' from this and + * program SMU with correct polarity + */ + table->ThermOutPolarity = + (0 == (cgs_read_register(hwmgr->device, mmGPIOPAD_A) & + (1 << gpio_pin.uc_gpio_pin_bit_shift))) ? 1:0; + table->ThermOutMode = SMU7_THERM_OUT_MODE_THERM_ONLY; + + /* if required, combine VRHot/PCC with thermal out GPIO */ + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_RegulatorHot) && + phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_CombinePCCWithThermalSignal)) + table->ThermOutMode = SMU7_THERM_OUT_MODE_THERM_VRHOT; + } else { + table->ThermOutGpio = 17; + table->ThermOutPolarity = 1; + table->ThermOutMode = SMU7_THERM_OUT_MODE_DISABLE; + } + + /* Populate BIF_SCLK levels into SMC DPM table */ + for (i = 0; i <= hw_data->dpm_table.pcie_speed_table.count; i++) { + result = atomctrl_get_dfs_pll_dividers_vi(hwmgr, + smu_data->bif_sclk_table[i], ÷rs); + PP_ASSERT_WITH_CODE(!result, + "Can not find DFS divide id for Sclk", + return result); + + if (i == 0) + table->Ulv.BifSclkDfs = + PP_HOST_TO_SMC_US((uint16_t)(dividers.pll_post_divider)); + else + table->LinkLevel[i - 1].BifSclkDfs = + PP_HOST_TO_SMC_US((uint16_t)(dividers.pll_post_divider)); + } + + for (i = 0; i < SMU75_MAX_ENTRIES_SMIO; i++) + table->Smio[i] = PP_HOST_TO_SMC_UL(table->Smio[i]); + + CONVERT_FROM_HOST_TO_SMC_UL(table->SystemFlags); + CONVERT_FROM_HOST_TO_SMC_UL(table->VRConfig); + CONVERT_FROM_HOST_TO_SMC_UL(table->SmioMask1); + CONVERT_FROM_HOST_TO_SMC_UL(table->SmioMask2); + CONVERT_FROM_HOST_TO_SMC_UL(table->SclkStepSize); + CONVERT_FROM_HOST_TO_SMC_UL(table->CurrSclkPllRange); + CONVERT_FROM_HOST_TO_SMC_US(table->TemperatureLimitHigh); + CONVERT_FROM_HOST_TO_SMC_US(table->TemperatureLimitLow); + CONVERT_FROM_HOST_TO_SMC_US(table->VoltageResponseTime); + CONVERT_FROM_HOST_TO_SMC_US(table->PhaseResponseTime); + + /* Upload all dpm data to SMC memory.(dpm level, dpm level count etc) */ + result = smu7_copy_bytes_to_smc(hwmgr, + smu_data->smu7_data.dpm_table_start + + offsetof(SMU75_Discrete_DpmTable, SystemFlags), + (uint8_t *)&(table->SystemFlags), + sizeof(SMU75_Discrete_DpmTable) - 3 * sizeof(SMU75_PIDController), + SMC_RAM_END); + PP_ASSERT_WITH_CODE(!result, + "Failed to upload dpm data to SMC memory!", return result); + + result = vegam_populate_pm_fuses(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "Failed to populate PM fuses to SMC memory!", return result); + + result = vegam_enable_reconfig_cus(hwmgr); + PP_ASSERT_WITH_CODE(!result, + "Failed to enable reconfigurable CUs!", return result); + + return 0; +} + +static uint32_t vegam_get_offsetof(uint32_t type, uint32_t member) +{ + switch (type) { + case SMU_SoftRegisters: + switch (member) { + case HandshakeDisables: + return offsetof(SMU75_SoftRegisters, HandshakeDisables); + case VoltageChangeTimeout: + return offsetof(SMU75_SoftRegisters, VoltageChangeTimeout); + case AverageGraphicsActivity: + return offsetof(SMU75_SoftRegisters, AverageGraphicsActivity); + case PreVBlankGap: + return offsetof(SMU75_SoftRegisters, PreVBlankGap); + case VBlankTimeout: + return offsetof(SMU75_SoftRegisters, VBlankTimeout); + case UcodeLoadStatus: + return offsetof(SMU75_SoftRegisters, UcodeLoadStatus); + case DRAM_LOG_ADDR_H: + return offsetof(SMU75_SoftRegisters, DRAM_LOG_ADDR_H); + case DRAM_LOG_ADDR_L: + return offsetof(SMU75_SoftRegisters, DRAM_LOG_ADDR_L); + case DRAM_LOG_PHY_ADDR_H: + return offsetof(SMU75_SoftRegisters, DRAM_LOG_PHY_ADDR_H); + case DRAM_LOG_PHY_ADDR_L: + return offsetof(SMU75_SoftRegisters, DRAM_LOG_PHY_ADDR_L); + case DRAM_LOG_BUFF_SIZE: + return offsetof(SMU75_SoftRegisters, DRAM_LOG_BUFF_SIZE); + } + case SMU_Discrete_DpmTable: + switch (member) { + case UvdBootLevel: + return offsetof(SMU75_Discrete_DpmTable, UvdBootLevel); + case VceBootLevel: + return offsetof(SMU75_Discrete_DpmTable, VceBootLevel); + case SamuBootLevel: + return offsetof(SMU75_Discrete_DpmTable, SamuBootLevel); + case LowSclkInterruptThreshold: + return offsetof(SMU75_Discrete_DpmTable, LowSclkInterruptThreshold); + } + } + pr_warn("can't get the offset of type %x member %x\n", type, member); + return 0; +} + +static int vegam_program_mem_timing_parameters(struct pp_hwmgr *hwmgr) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + + if (data->need_update_smu7_dpm_table & + (DPMTABLE_OD_UPDATE_SCLK + + DPMTABLE_UPDATE_SCLK + + DPMTABLE_UPDATE_MCLK)) + return vegam_program_memory_timing_parameters(hwmgr); + + return 0; +} + +static int vegam_update_sclk_threshold(struct pp_hwmgr *hwmgr) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + struct vegam_smumgr *smu_data = + (struct vegam_smumgr *)(hwmgr->smu_backend); + int result = 0; + uint32_t low_sclk_interrupt_threshold = 0; + + if (phm_cap_enabled(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_SclkThrottleLowNotification) + && (data->low_sclk_interrupt_threshold != 0)) { + low_sclk_interrupt_threshold = + data->low_sclk_interrupt_threshold; + + CONVERT_FROM_HOST_TO_SMC_UL(low_sclk_interrupt_threshold); + + result = smu7_copy_bytes_to_smc( + hwmgr, + smu_data->smu7_data.dpm_table_start + + offsetof(SMU75_Discrete_DpmTable, + LowSclkInterruptThreshold), + (uint8_t *)&low_sclk_interrupt_threshold, + sizeof(uint32_t), + SMC_RAM_END); + } + PP_ASSERT_WITH_CODE((result == 0), + "Failed to update SCLK threshold!", return result); + + result = vegam_program_mem_timing_parameters(hwmgr); + PP_ASSERT_WITH_CODE((result == 0), + "Failed to program memory timing parameters!", + ); + + return result; +} + +int vegam_thermal_avfs_enable(struct pp_hwmgr *hwmgr) +{ + struct smu7_hwmgr *data = (struct smu7_hwmgr *)(hwmgr->backend); + int ret; + + if (!hwmgr->avfs_supported) + return 0; + + ret = smum_send_msg_to_smc(hwmgr, PPSMC_MSG_EnableAvfs); + if (!ret) { + if (data->apply_avfs_cks_off_voltage) + ret = smum_send_msg_to_smc(hwmgr, PPSMC_MSG_ApplyAvfsCksOffVoltage); + } + + return ret; +} + +static int vegam_thermal_setup_fan_table(struct pp_hwmgr *hwmgr) +{ + PP_ASSERT_WITH_CODE(hwmgr->thermal_controller.fanInfo.bNoFan, + "VBIOS fan info is not correct!", + ); + phm_cap_unset(hwmgr->platform_descriptor.platformCaps, + PHM_PlatformCaps_MicrocodeFanControl); + return 0; +} + +const struct pp_smumgr_func vegam_smu_funcs = { + .smu_init = vegam_smu_init, + .smu_fini = smu7_smu_fini, + .start_smu = vegam_start_smu, + .check_fw_load_finish = smu7_check_fw_load_finish, + .request_smu_load_fw = smu7_reload_firmware, + .request_smu_load_specific_fw = NULL, + .send_msg_to_smc = smu7_send_msg_to_smc, + .send_msg_to_smc_with_parameter = smu7_send_msg_to_smc_with_parameter, + .process_firmware_header = vegam_process_firmware_header, + .is_dpm_running = vegam_is_dpm_running, + .get_mac_definition = vegam_get_mac_definition, + .update_smc_table = vegam_update_smc_table, + .init_smc_table = vegam_init_smc_table, + .get_offsetof = vegam_get_offsetof, + .populate_all_graphic_levels = vegam_populate_all_graphic_levels, + .populate_all_memory_levels = vegam_populate_all_memory_levels, + .update_sclk_threshold = vegam_update_sclk_threshold, + .is_hw_avfs_present = vegam_is_hw_avfs_present, + .thermal_avfs_enable = vegam_thermal_avfs_enable, + .thermal_setup_fan_table = vegam_thermal_setup_fan_table, +}; diff --git a/drivers/gpu/drm/amd/powerplay/smumgr/vegam_smumgr.h b/drivers/gpu/drm/amd/powerplay/smumgr/vegam_smumgr.h new file mode 100644 index 0000000000000..2b65582385008 --- /dev/null +++ b/drivers/gpu/drm/amd/powerplay/smumgr/vegam_smumgr.h @@ -0,0 +1,75 @@ +/* + * Copyright 2017 Advanced Micro Devices, Inc. + * + * Permission is hereby granted, free of charge, to any person obtaining a + * copy of this software and associated documentation files (the "Software"), + * to deal in the Software without restriction, including without limitation + * the rights to use, copy, modify, merge, publish, distribute, sublicense, + * and/or sell copies of the Software, and to permit persons to whom the + * Software is furnished to do so, subject to the following conditions: + * + * The above copyright notice and this permission notice shall be included in + * all copies or substantial portions of the Software. + * + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL + * THE COPYRIGHT HOLDER(S) OR AUTHOR(S) BE LIABLE FOR ANY CLAIM, DAMAGES OR + * OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, + * ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR + * OTHER DEALINGS IN THE SOFTWARE. + * + */ + +#ifndef _VEGAM_SMUMANAGER_H +#define _VEGAM_SMUMANAGER_H + + +#include <pp_endian.h> +#include "smu75_discrete.h" +#include "smu7_smumgr.h" + +#define SMC_RAM_END 0x40000 + +#define DPMTuning_Uphyst_Shift 0 +#define DPMTuning_Downhyst_Shift 8 +#define DPMTuning_Activity_Shift 16 + +#define GraphicsDPMTuning_VEGAM 0x001e6400 +#define MemoryDPMTuning_VEGAM 0x000f3c0a +#define SclkDPMTuning_VEGAM 0x002d000a +#define MclkDPMTuning_VEGAM 0x001f100a + + +struct vegam_pt_defaults { + uint8_t SviLoadLineEn; + uint8_t SviLoadLineVddC; + uint8_t TDC_VDDC_ThrottleReleaseLimitPerc; + uint8_t TDC_MAWt; + uint8_t TdcWaterfallCtl; + uint8_t DTEAmbientTempBase; + + uint32_t DisplayCac; + uint32_t BAPM_TEMP_GRADIENT; + uint16_t BAPMTI_R[SMU75_DTE_ITERATIONS * SMU75_DTE_SOURCES * SMU75_DTE_SINKS]; + uint16_t BAPMTI_RC[SMU75_DTE_ITERATIONS * SMU75_DTE_SOURCES * SMU75_DTE_SINKS]; +}; + +struct vegam_range_table { + uint32_t trans_lower_frequency; /* in 10khz */ + uint32_t trans_upper_frequency; +}; + +struct vegam_smumgr { + struct smu7_smumgr smu7_data; + uint8_t protected_mode; + SMU75_Discrete_DpmTable smc_state_table; + struct SMU75_Discrete_Ulv ulv_setting; + struct SMU75_Discrete_PmFuses power_tune_table; + struct vegam_range_table range_table[NUM_SCLK_RANGE]; + const struct vegam_pt_defaults *power_tune_defaults; + uint32_t bif_sclk_table[SMU75_MAX_LEVELS_LINK]; +}; + + +#endif |