WIP: push down hangcheck into adreno_gpu

Signed-off-by: Lucas Stach <l.stach@pengutronix.de>

6 files changed, 74 insertions, 71 deletions

diff --git a/drivers/gpu/drm/msm/adreno/a3xx_gpu.c b/drivers/gpu/drm/msm/adreno/a3xx_gpu.c
index 2483a6b9225f..386993aec7e2 100644
--- a/drivers/gpu/drm/msm/adreno/a3xx_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/a3xx_gpu.c
@@ -379,8 +379,8 @@ static const struct adreno_gpu_funcs funcs = {
 #endif
 };
 
-struct msm_gpu *a3xx_gpu_init(struct drm_device *dev,
-			      struct adreno_gem *gem)
+struct msm_gpu *a3xx_gpu_init(struct drm_device *dev, struct adreno_gem *gem,
+		struct workqueue_struct *wq)
 {
 	struct adreno_gpu *adreno_gpu = NULL;
 	struct msm_gpu *gpu;
@@ -414,7 +414,7 @@ struct msm_gpu *a3xx_gpu_init(struct drm_device *dev,
 			gpu->fast_rate, gpu->slow_rate, gpu->bus_freq);
 
 	ret = adreno_gpu_init(dev, pdev, adreno_gpu,
-			&funcs, gem, config->rev);
+			&funcs, gem, wq, config->rev);
 	if (ret)
 		goto fail;
 
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.c b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
index c0bf75b47ef6..5e10446f7798 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.c
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.c
@@ -60,6 +60,50 @@ const char *adreno_get_name(struct adreno_gpu *gpu)
 	return gpu->info->name;
 }
 
+/*
+ * Hangcheck detection for locked gpu:
+ */
+
+static void recover_worker(struct work_struct *work)
+{
+	struct adreno_gpu *gpu = container_of(work, struct adreno_gpu, recover_work);
+	struct drm_device *dev = gpu->drm;
+
+	dev_err(dev->dev, "%s: hangcheck recover!\n", adreno_get_name(gpu));
+
+	mutex_lock(&dev->struct_mutex);
+	gpu->funcs->recover(gpu);
+	mutex_unlock(&dev->struct_mutex);
+
+	gpu->gem->gem_retire(gpu->gem->priv);
+}
+
+static void hangcheck_timer_reset(struct adreno_gpu *gpu)
+{
+	DBG("%s", adreno_get_name(gpu));
+	mod_timer(&gpu->hangcheck_timer,
+			round_jiffies_up(jiffies + DRM_MSM_HANGCHECK_JIFFIES));
+}
+
+static void hangcheck_handler(unsigned long data)
+{
+	struct adreno_gpu *gpu = (struct adreno_gpu *)data;
+	uint32_t fence = gpu->funcs->last_fence(gpu);
+
+	if (fence != gpu->hangcheck_fence) {
+		/* some progress has been made.. ya! */
+		gpu->hangcheck_fence = fence;
+	} else if (fence < gpu->submitted_fence) {
+		/* no progress and not done.. hung! */
+		gpu->hangcheck_fence = fence;
+		queue_work(gpu->shared_wq, &gpu->recover_work);
+	}
+
+	/* if still more pending work, reset the hangcheck timer: */
+	if (gpu->submitted_fence > gpu->hangcheck_fence)
+		hangcheck_timer_reset(gpu);
+}
+
 int adreno_get_param(struct adreno_gpu *gpu, uint32_t param, uint64_t *value)
 {
 	switch (param) {
@@ -120,7 +164,7 @@ void adreno_recover(struct adreno_gpu *gpu)
 	gpu->rb->cur = gpu->rb->start;
 
 	/* reset completed fence seqno, just discard anything pending: */
-	gpu->memptrs->fence = gpu->base.submitted_fence;
+	gpu->memptrs->fence = gpu->submitted_fence;
 
 	gpu->funcs->pm_resume(gpu);
 	ret = gpu->funcs->hw_init(gpu);
@@ -196,6 +240,9 @@ int adreno_submit(struct adreno_gpu *gpu, struct adreno_submit *submit,
 
 	gpu->funcs->flush(gpu);
 
+	gpu->submitted_fence = submit->fence;
+	hangcheck_timer_reset(gpu);
+
 	return 0;
 }
 
@@ -238,7 +285,7 @@ void adreno_show(struct adreno_gpu *gpu, struct seq_file *m)
 			gpu->rev.patchid);
 
 	seq_printf(m, "fence:    %d/%d\n", gpu->memptrs->fence,
-			gpu->base.submitted_fence);
+			gpu->submitted_fence);
 	seq_printf(m, "rptr:     %d\n", gpu->memptrs->rptr);
 	seq_printf(m, "wptr:     %d\n", gpu->memptrs->wptr);
 	seq_printf(m, "rb wptr:  %d\n", get_wptr(gpu->rb));
@@ -268,7 +315,8 @@ static inline bool _rev_match(uint8_t entry, uint8_t id)
 
 int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 		struct adreno_gpu *gpu, const struct adreno_gpu_funcs *funcs,
-		struct adreno_gem *gem, struct adreno_rev rev)
+		struct adreno_gem *gem, struct workqueue_struct *wq,
+		struct adreno_rev rev)
 {
 	int i, ret;
 
@@ -297,6 +345,7 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 	gpu->drm = drm;
 	gpu->funcs = funcs;
 	gpu->gem = gem;
+	gpu->shared_wq = wq;
 	gpu->rev = rev;
 
 	/* XXX: urgh - pointers to self are fun! */
@@ -368,6 +417,10 @@ int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 		return ret;
 	}
 
+	INIT_WORK(&gpu->recover_work, recover_worker);
+	setup_timer(&gpu->hangcheck_timer, hangcheck_handler,
+		    (unsigned long)gpu);
+
 	return 0;
 }
 
diff --git a/drivers/gpu/drm/msm/adreno/adreno_gpu.h b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
index 8d7f26e7d950..a7fe0e5e0c62 100644
--- a/drivers/gpu/drm/msm/adreno/adreno_gpu.h
+++ b/drivers/gpu/drm/msm/adreno/adreno_gpu.h
@@ -54,6 +54,7 @@ struct adreno_gpu {
 	uint32_t revn;  /* numeric revision name */
 	const struct adreno_gpu_funcs *funcs;
 	struct adreno_gem *gem;
+	struct workqueue_struct *shared_wq;
 
 	/* firmware: */
 	const struct firmware *pm4, *pfp;
@@ -65,10 +66,19 @@ struct adreno_gpu {
 	struct drm_gem_object *memptrs_bo;
 	uint32_t memptrs_iova;
 
+	uint32_t submitted_fence;
+
 	struct adreno_ringbuffer *rb;
 	uint32_t rb_iova;
 
 	struct adreno_context *lastctx;
+
+	/* Hang Detction: */
+#define DRM_MSM_HANGCHECK_PERIOD 500 /* in ms */
+#define DRM_MSM_HANGCHECK_JIFFIES msecs_to_jiffies(DRM_MSM_HANGCHECK_PERIOD)
+	struct timer_list hangcheck_timer;
+	uint32_t hangcheck_fence;
+	struct work_struct recover_work;
 };
 #define to_adreno_gpu(x) container_of(x, struct adreno_gpu, base)
 
@@ -115,7 +125,8 @@ void adreno_wait_ring(struct adreno_gpu *gpu, uint32_t ndwords);
 
 int adreno_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 		struct adreno_gpu *gpu, const struct adreno_gpu_funcs *funcs,
-		struct adreno_gem *gem, struct adreno_rev rev);
+		struct adreno_gem *gem, struct workqueue_struct *wq,
+		struct adreno_rev rev);
 void adreno_gpu_cleanup(struct adreno_gpu *gpu);
 
 static inline void gpu_write(struct adreno_gpu *gpu, u32 reg, u32 data)
diff --git a/drivers/gpu/drm/msm/msm_drv.c b/drivers/gpu/drm/msm/msm_drv.c
index 9f3c45c95f13..a4db194c57e9 100644
--- a/drivers/gpu/drm/msm/msm_drv.c
+++ b/drivers/gpu/drm/msm/msm_drv.c
@@ -315,7 +315,7 @@ static void load_gpu(struct drm_device *dev)
 		return;
 
 	mutex_lock(&dev->struct_mutex);
-	gpu = a3xx_gpu_init(dev, &msm_gem_funcs);
+	gpu = a3xx_gpu_init(dev, &msm_gem_funcs, priv->wq);
 	if (IS_ERR(gpu)) {
 		dev_warn(dev->dev, "failed to load a3xx gpu\n");
 		gpu = NULL;
diff --git a/drivers/gpu/drm/msm/msm_gpu.c b/drivers/gpu/drm/msm/msm_gpu.c
index 96214168c91c..a18646503ecc 100644
--- a/drivers/gpu/drm/msm/msm_gpu.c
+++ b/drivers/gpu/drm/msm/msm_gpu.c
@@ -203,52 +203,6 @@ int msm_gpu_pm_suspend(struct adreno_gpu *gpu)
 }
 
 /*
- * Hangcheck detection for locked gpu:
- */
-
-static void recover_worker(struct work_struct *work)
-{
-	struct msm_gpu *gpu = container_of(work, struct msm_gpu, recover_work);
-	struct drm_device *dev = gpu->dev;
-
-	dev_err(dev->dev, "%s: hangcheck recover!\n",
-		adreno_get_name(gpu->gpu));
-
-	mutex_lock(&dev->struct_mutex);
-	gpu->funcs->recover(gpu->gpu);
-	mutex_unlock(&dev->struct_mutex);
-
-	msm_gpu_retire(gpu);
-}
-
-static void hangcheck_timer_reset(struct msm_gpu *gpu)
-{
-	DBG("%s", adreno_get_name(gpu->gpu));
-	mod_timer(&gpu->hangcheck_timer,
-			round_jiffies_up(jiffies + DRM_MSM_HANGCHECK_JIFFIES));
-}
-
-static void hangcheck_handler(unsigned long data)
-{
-	struct msm_gpu *gpu = (struct msm_gpu *)data;
-	uint32_t fence = gpu->funcs->last_fence(gpu->gpu);
-
-	if (fence != gpu->hangcheck_fence) {
-		/* some progress has been made.. ya! */
-		gpu->hangcheck_fence = fence;
-	} else if (fence < gpu->submitted_fence) {
-		/* no progress and not done.. hung! */
-		struct msm_drm_private *priv = gpu->dev->dev_private;
-		gpu->hangcheck_fence = fence;
-		queue_work(priv->wq, &gpu->recover_work);
-	}
-
-	/* if still more pending work, reset the hangcheck timer: */
-	if (gpu->submitted_fence > gpu->hangcheck_fence)
-		hangcheck_timer_reset(gpu);
-}
-
-/*
  * Cmdstream submission/retirement:
  */
 
@@ -300,8 +254,6 @@ int msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 
 	submit->base.fence = ++priv->next_fence;
 
-	gpu->submitted_fence = submit->base.fence;
-
 	ret = gpu->funcs->submit(gpu->gpu, &submit->base, &filp->ctx);
 
 	for (i = 0; i < submit->nr_bos; i++) {
@@ -323,7 +275,7 @@ int msm_gpu_submit(struct msm_gpu *gpu, struct msm_gem_submit *submit,
 
 		msm_gem_move_to_active(&msm_obj->base, gpu, submit->base.fence);
 	}
-	hangcheck_timer_reset(gpu);
+
 	mutex_unlock(&dev->struct_mutex);
 
 	return ret;
@@ -354,10 +306,6 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 
 	INIT_LIST_HEAD(&gpu->active_list);
 	INIT_WORK(&gpu->retire_work, retire_worker);
-	INIT_WORK(&gpu->recover_work, recover_worker);
-
-	setup_timer(&gpu->hangcheck_timer, hangcheck_handler,
-			(unsigned long)gpu);
 
 	BUG_ON(ARRAY_SIZE(clk_names) != ARRAY_SIZE(gpu->grp_clks));
 
diff --git a/drivers/gpu/drm/msm/msm_gpu.h b/drivers/gpu/drm/msm/msm_gpu.h
index 2dc686e05c54..35e45b865e44 100644
--- a/drivers/gpu/drm/msm/msm_gpu.h
+++ b/drivers/gpu/drm/msm/msm_gpu.h
@@ -46,8 +46,6 @@ struct msm_gpu {
 	/* list of GEM active objects: */
 	struct list_head active_list;
 
-	uint32_t submitted_fence;
-
 	/* worker for handling active-list retiring: */
 	struct work_struct retire_work;
 
@@ -62,13 +60,6 @@ struct msm_gpu {
 	struct clk *ebi1_clk, *grp_clks[5];
 	uint32_t fast_rate, slow_rate, bus_freq;
 	uint32_t bsc;
-
-	/* Hang Detction: */
-#define DRM_MSM_HANGCHECK_PERIOD 500 /* in ms */
-#define DRM_MSM_HANGCHECK_JIFFIES msecs_to_jiffies(DRM_MSM_HANGCHECK_PERIOD)
-	struct timer_list hangcheck_timer;
-	uint32_t hangcheck_fence;
-	struct work_struct recover_work;
 };
 
 int msm_gpu_pm_suspend(struct adreno_gpu *gpu);
@@ -85,7 +76,7 @@ int msm_gpu_init(struct drm_device *drm, struct platform_device *pdev,
 void msm_gpu_cleanup(struct msm_gpu *gpu);
 
 struct msm_gpu *a3xx_gpu_init(struct drm_device *dev,
-		struct adreno_gem *gem);
+		struct adreno_gem *gem, struct workqueue_struct *wq);
 void __init a3xx_register(void);
 void __exit a3xx_unregister(void);