Merge branch 'work/thumb2' into next

22 files changed, 164 insertions, 36 deletions

diff --git a/arch/arm/Kconfig b/arch/arm/Kconfig
index 67a1009777..8a4e1a2b41 100644
--- a/arch/arm/Kconfig
+++ b/arch/arm/Kconfig
@@ -102,6 +102,18 @@ config AEABI
 
 	  To use this you need GCC version 4.0.0 or later.
 
+config THUMB2_BAREBOX
+	select ARM_ASM_UNIFIED
+	depends on CPU_V7
+	bool "Compile barebox in thumb-2 mode (read help)"
+	help
+	  This enables compilation of barebox in thumb-2 mode which generates
+	  ~25% smaller binaries. Arm Assembly code needs some fixups to be able
+	  to work correctly in thumb-2 mode. the barebox core should have these
+	  fixups since most assembly code is derived from the Kernel. However,
+	  your board lowlevel init code may break in thumb-2 mode. You have been
+	  warned.
+
 endmenu
 
 menu "Arm specific settings         "
diff --git a/arch/arm/Makefile b/arch/arm/Makefile
index b28a5b2263..4c6a566ff2 100644
--- a/arch/arm/Makefile
+++ b/arch/arm/Makefile
@@ -39,8 +39,15 @@ ifeq ($(CONFIG_ARM_UNWIND),y)
 CFLAGS_ABI	+=-funwind-tables
 endif
 
-CPPFLAGS += $(CFLAGS_ABI) $(arch-y) $(tune-y) -msoft-float
-AFLAGS   += -include asm/unified.h -msoft-float
+ifeq ($(CONFIG_THUMB2_BAREBOX),y)
+AFLAGS_AUTOIT	:=$(call as-option,-Wa$(comma)-mimplicit-it=always,-Wa$(comma)-mauto-it)
+AFLAGS_NOWARN	:=$(call as-option,-Wa$(comma)-mno-warn-deprecated,-Wa$(comma)-W)
+CFLAGS_THUMB2	:=-mthumb $(AFLAGS_AUTOIT) $(AFLAGS_NOWARN)
+AFLAGS_THUMB2	:=$(CFLAGS_THUMB2) -Wa$(comma)-mthumb
+endif
+
+CPPFLAGS += $(CFLAGS_ABI) $(arch-y) $(tune-y) -msoft-float $(CFLAGS_THUMB2)
+AFLAGS   += -include asm/unified.h -msoft-float $(AFLAGS_THUMB2)
 
 # Machine directory name.  This list is sorted alphanumerically
 # by CONFIG_* macro name.
diff --git a/arch/arm/boards/eukrea_cpuimx25/eukrea_cpuimx25.c b/arch/arm/boards/eukrea_cpuimx25/eukrea_cpuimx25.c
index 73e54f2f54..695e75f288 100644
--- a/arch/arm/boards/eukrea_cpuimx25/eukrea_cpuimx25.c
+++ b/arch/arm/boards/eukrea_cpuimx25/eukrea_cpuimx25.c
@@ -46,13 +46,14 @@
 #include <usb/fsl_usb2.h>
 #include <mach/usb.h>
 #include <mach/devices-imx25.h>
+#include <asm/barebox-arm-head.h>
 
 extern unsigned long _stext;
 extern void exception_vectors(void);
 
 void __naked __flash_header_start go(void)
 {
-	__asm__ __volatile__("b exception_vectors\n");
+	barebox_arm_head();
 }
 
 struct imx_dcd_entry __dcd_entry_section dcd_entry[] = {
diff --git a/arch/arm/boards/eukrea_cpuimx35/flash_header.c b/arch/arm/boards/eukrea_cpuimx35/flash_header.c
index 93c8348209..f8ed5d483e 100644
--- a/arch/arm/boards/eukrea_cpuimx35/flash_header.c
+++ b/arch/arm/boards/eukrea_cpuimx35/flash_header.c
@@ -1,12 +1,13 @@
 #include <common.h>
 #include <mach/imx-flash-header.h>
 #include <mach/imx-regs.h>
+#include <asm/barebox-arm-head.h>
 
 extern void exception_vectors(void);
 
 void __naked __flash_header_start go(void)
 {
-	__asm__ __volatile__("b exception_vectors\n");
+	barebox_arm_head();
 }
 
 struct imx_dcd_entry __dcd_entry_section dcd_entry[] = {
diff --git a/arch/arm/boards/eukrea_cpuimx51/flash_header.c b/arch/arm/boards/eukrea_cpuimx51/flash_header.c
index f953b09bde..ac6bbdcc0f 100644
--- a/arch/arm/boards/eukrea_cpuimx51/flash_header.c
+++ b/arch/arm/boards/eukrea_cpuimx51/flash_header.c
@@ -1,11 +1,12 @@
 #include <common.h>
 #include <mach/imx-flash-header.h>
+#include <asm/barebox-arm-head.h>
 
 extern unsigned long _stext;
 
 void __naked __flash_header_start go(void)
 {
-	__asm__ __volatile__("b exception_vectors\n");
+	barebox_arm_head();
 }
 
 struct imx_dcd_entry __dcd_entry_section dcd_entry[] = {
diff --git a/arch/arm/boards/freescale-mx25-3-stack/3stack.c b/arch/arm/boards/freescale-mx25-3-stack/3stack.c
index 5aa54e43ff..d7e24cf3e7 100644
--- a/arch/arm/boards/freescale-mx25-3-stack/3stack.c
+++ b/arch/arm/boards/freescale-mx25-3-stack/3stack.c
@@ -42,13 +42,14 @@
 #include <i2c/i2c.h>
 #include <mfd/mc34704.h>
 #include <mach/devices-imx25.h>
+#include <asm/barebox-arm-head.h>
 
 extern unsigned long _stext;
 extern void exception_vectors(void);
 
 void __naked __flash_header_start go(void)
 {
-	__asm__ __volatile__("b exception_vectors\n");
+	barebox_arm_head();
 }
 
 struct imx_dcd_entry __dcd_entry_section dcd_entry[] = {
diff --git a/arch/arm/boards/freescale-mx35-3-stack/flash_header.c b/arch/arm/boards/freescale-mx35-3-stack/flash_header.c
index 92f214202e..85b52a25ca 100644
--- a/arch/arm/boards/freescale-mx35-3-stack/flash_header.c
+++ b/arch/arm/boards/freescale-mx35-3-stack/flash_header.c
@@ -1,12 +1,13 @@
 #include <common.h>
 #include <mach/imx-flash-header.h>
 #include <mach/imx-regs.h>
+#include <asm/barebox-arm-head.h>
 
 extern void exception_vectors(void);
 
 void __naked __flash_header_start go(void)
 {
-	__asm__ __volatile__("b exception_vectors\n");
+	barebox_arm_head();
 }
 
 struct imx_dcd_entry __dcd_entry_section dcd_entry[] = {
diff --git a/arch/arm/boards/freescale-mx51-pdk/flash_header.c b/arch/arm/boards/freescale-mx51-pdk/flash_header.c
index 5f94506b69..297dca1c29 100644
--- a/arch/arm/boards/freescale-mx51-pdk/flash_header.c
+++ b/arch/arm/boards/freescale-mx51-pdk/flash_header.c
@@ -1,11 +1,12 @@
 #include <common.h>
 #include <mach/imx-flash-header.h>
+#include <asm/barebox-arm-head.h>
 
 extern unsigned long _stext;
 
 void __naked __flash_header_start go(void)
 {
-	__asm__ __volatile__("b exception_vectors\n");
+	barebox_arm_head();
 }
 
 struct imx_dcd_entry __dcd_entry_section dcd_entry[] = {
diff --git a/arch/arm/boards/freescale-mx53-loco/flash_header.c b/arch/arm/boards/freescale-mx53-loco/flash_header.c
index 490e223f0e..b45964059a 100644
--- a/arch/arm/boards/freescale-mx53-loco/flash_header.c
+++ b/arch/arm/boards/freescale-mx53-loco/flash_header.c
@@ -16,10 +16,11 @@
 #include <common.h>
 #include <asm/byteorder.h>
 #include <mach/imx-flash-header.h>
+#include <asm/barebox-arm-head.h>
 
 void __naked __flash_header_start go(void)
 {
-	__asm__ __volatile__("b exception_vectors\n");
+	barebox_arm_head();
 }
 
 struct imx_dcd_v2_entry __dcd_entry_section dcd_entry[] = {
diff --git a/arch/arm/boards/freescale-mx53-smd/flash_header.c b/arch/arm/boards/freescale-mx53-smd/flash_header.c
index 490e223f0e..b45964059a 100644
--- a/arch/arm/boards/freescale-mx53-smd/flash_header.c
+++ b/arch/arm/boards/freescale-mx53-smd/flash_header.c
@@ -16,10 +16,11 @@
 #include <common.h>
 #include <asm/byteorder.h>
 #include <mach/imx-flash-header.h>
+#include <asm/barebox-arm-head.h>
 
 void __naked __flash_header_start go(void)
 {
-	__asm__ __volatile__("b exception_vectors\n");
+	barebox_arm_head();
 }
 
 struct imx_dcd_v2_entry __dcd_entry_section dcd_entry[] = {
diff --git a/arch/arm/cpu/cpu.c b/arch/arm/cpu/cpu.c
index cf31e8bb2e..71ef8c06d6 100644
--- a/arch/arm/cpu/cpu.c
+++ b/arch/arm/cpu/cpu.c
@@ -26,6 +26,7 @@
  */
 
 #include <common.h>
+#include <init.h>
 #include <command.h>
 #include <cache.h>
 #include <asm/mmu.h>
@@ -89,3 +90,29 @@ void arch_shutdown(void)
 	);
 #endif
 }
+
+#ifdef CONFIG_THUMB2_BAREBOX
+static void thumb2_execute(void *func, int argc, char *argv[])
+{
+	/*
+	 * Switch back to arm mode before executing external
+	 * programs.
+	 */
+	__asm__ __volatile__ (
+		"mov r0, #0\n"
+		"mov r1, %0\n"
+		"mov r2, %1\n"
+		"bx %2\n"
+		:
+		: "r" (argc - 1), "r" (&argv[1]), "r" (func)
+		: "r0", "r1", "r2"
+	);
+}
+
+static int execute_init(void)
+{
+	do_execute = thumb2_execute;
+	return 0;
+}
+postcore_initcall(execute_init);
+#endif
diff --git a/arch/arm/cpu/exceptions.S b/arch/arm/cpu/exceptions.S
index 6f35d408f4..c08537a405 100644
--- a/arch/arm/cpu/exceptions.S
+++ b/arch/arm/cpu/exceptions.S
@@ -106,6 +106,7 @@ _STACK_START:
  * exception handlers
  */
 	.section ".text","ax"
+	.arm
 
 	.align  5
 .globl undefined_instruction
diff --git a/arch/arm/cpu/mmu.c b/arch/arm/cpu/mmu.c
index 6d9b1e0231..c19f9311cd 100644
--- a/arch/arm/cpu/mmu.c
+++ b/arch/arm/cpu/mmu.c
@@ -4,6 +4,7 @@
 #include <errno.h>
 #include <sizes.h>
 #include <asm/memory.h>
+#include <asm/barebox-arm.h>
 #include <asm/system.h>
 #include <memory.h>
 
@@ -182,7 +183,6 @@ static void vectors_init(void)
 {
 	u32 *exc, *zero = NULL;
 	void *vectors;
-	extern unsigned long exception_vectors;
 	u32 cr;
 
 	cr = get_cr();
@@ -210,7 +210,7 @@ static void vectors_init(void)
 
 	vectors = xmemalign(PAGE_SIZE, PAGE_SIZE);
 	memset(vectors, 0, PAGE_SIZE);
-	memcpy(vectors, &exception_vectors, ARM_VECTORS_SIZE);
+	memcpy(vectors, __exceptions_start, __exceptions_stop - __exceptions_start);
 
 	if (cr & CR_V)
 		exc[256 - 16] = (u32)vectors | PTE_TYPE_SMALL | PTE_FLAGS_CACHED;
diff --git a/arch/arm/cpu/start.c b/arch/arm/cpu/start.c
index 61d7e3e5d5..ede8fd1f09 100644
--- a/arch/arm/cpu/start.c
+++ b/arch/arm/cpu/start.c
@@ -23,37 +23,38 @@
 #include <common.h>
 #include <init.h>
 #include <asm/barebox-arm.h>
+#include <asm/barebox-arm-head.h>
 #include <asm/system.h>
 #include <asm-generic/memory_layout.h>
 #include <asm/sections.h>
 
-void __naked __section(.text_entry) exception_vectors(void)
+void __naked __section(.text_entry) start(void)
+{
+	barebox_arm_head();
+}
+
+void __naked __section(.text_exceptions) exception_vectors(void)
 {
 	__asm__ __volatile__ (
+		".arm\n"
 		"b reset\n"				/* reset */
 #ifdef CONFIG_ARM_EXCEPTIONS
 		"ldr pc, =undefined_instruction\n"	/* undefined instruction */
 		"ldr pc, =software_interrupt\n"		/* software interrupt (SWI) */
 		"ldr pc, =prefetch_abort\n"		/* prefetch abort */
 		"ldr pc, =data_abort\n"			/* data abort */
-		"1: bne 1b\n"				/* (reserved) */
+		"1: b 1b\n"				/* (reserved) */
 		"ldr pc, =irq\n"			/* irq (interrupt) */
 		"ldr pc, =fiq\n"			/* fiq (fast interrupt) */
 #else
-		"1: bne 1b\n"				/* undefined instruction */
-		"1: bne 1b\n"				/* software interrupt (SWI) */
-		"1: bne 1b\n"				/* prefetch abort */
-		"1: bne 1b\n"				/* data abort */
-		"1: bne 1b\n"				/* (reserved) */
-		"1: bne 1b\n"				/* irq (interrupt) */
-		"1: bne 1b\n"				/* fiq (fast interrupt) */
+		"1: b 1b\n"				/* undefined instruction */
+		"1: b 1b\n"				/* software interrupt (SWI) */
+		"1: b 1b\n"				/* prefetch abort */
+		"1: b 1b\n"				/* data abort */
+		"1: b 1b\n"				/* (reserved) */
+		"1: b 1b\n"				/* irq (interrupt) */
+		"1: b 1b\n"				/* fiq (fast interrupt) */
 #endif
-		".word 0x65726162\n"			/* 'bare' */
-		".word 0x00786f62\n"			/* 'box' */
-		".word _text\n"				/* text base. If copied there,
-							 * barebox can skip relocation
-							 */
-		".word _barebox_image_size\n"		/* image size to copy */
 	);
 }
 
@@ -109,7 +110,7 @@ void __naked __bare_init reset(void)
  * Board code can jump here by either returning from board_init_lowlevel
  * or by calling this funtion directly.
  */
-void __naked __bare_init board_init_lowlevel_return(void)
+void __naked __section(.text_ll_return) board_init_lowlevel_return(void)
 {
 	uint32_t r, addr;
 
@@ -124,7 +125,7 @@ void __naked __bare_init board_init_lowlevel_return(void)
 	__asm__ __volatile__("mov sp, %0" : : "r"(r));
 
 	/* Get start of binary image */
-	addr -= (uint32_t)&board_init_lowlevel_return - TEXT_BASE;
+	addr -= (uint32_t)&__ll_return - TEXT_BASE;
 
 	/* relocate to link address if necessary */
 	if (addr != TEXT_BASE)
diff --git a/arch/arm/include/asm/barebox-arm-head.h b/arch/arm/include/asm/barebox-arm-head.h
new file mode 100644
index 0000000000..0dc3074a27
--- /dev/null
+++ b/arch/arm/include/asm/barebox-arm-head.h
@@ -0,0 +1,36 @@
+#ifndef __ASM_ARM_HEAD_H
+#define __ASM_ARM_HEAD_H
+
+static inline void barebox_arm_head(void)
+{
+	__asm__ __volatile__ (
+#ifdef CONFIG_THUMB2_BAREBOX
+		".arm\n"
+		"adr r9, 1f + 1\n"
+		"bx r9\n"
+		".thumb\n"
+		"1:\n"
+		"bl reset\n"
+		".rept 10\n"
+		"1: b 1b\n"
+		".endr\n"
+#else
+		"b reset\n"
+		"1: b 1b\n"
+		"1: b 1b\n"
+		"1: b 1b\n"
+		"1: b 1b\n"
+		"1: b 1b\n"
+		"1: b 1b\n"
+		"1: b 1b\n"
+#endif
+		".word 0x65726162\n"			/* 'bare' */
+		".word 0x00786f62\n"			/* 'box' */
+		".word _text\n"				/* text base. If copied there,
+							 * barebox can skip relocation
+							 */
+		".word _barebox_image_size\n"		/* image size to copy */
+	);
+}
+
+#endif /* __ASM_ARM_HEAD_H */
diff --git a/arch/arm/include/asm/barebox-arm.h b/arch/arm/include/asm/barebox-arm.h
index 7bb1af1606..b880dd49b3 100644
--- a/arch/arm/include/asm/barebox-arm.h
+++ b/arch/arm/include/asm/barebox-arm.h
@@ -36,6 +36,9 @@ int	cleanup_before_linux(void);
 int	board_init(void);
 int	dram_init (void);
 
+extern char __exceptions_start[], __exceptions_stop[];
+extern char __ll_return[];
+
 void board_init_lowlevel(void);
 void board_init_lowlevel_return(void);
 void arch_init_lowlevel(void);
diff --git a/arch/arm/include/asm/unified.h b/arch/arm/include/asm/unified.h
index bc631161e9..4d855c8352 100644
--- a/arch/arm/include/asm/unified.h
+++ b/arch/arm/include/asm/unified.h
@@ -24,10 +24,10 @@
 	.syntax unified
 #endif
 
-#ifdef CONFIG_THUMB2_KERNEL
+#ifdef CONFIG_THUMB2_BAREBOX
 
 #if __GNUC__ < 4
-#error Thumb-2 kernel requires gcc >= 4
+#error Thumb-2 barebox requires gcc >= 4
 #endif
 
 /* The CPSR bit describing the instruction set (Thumb) */
@@ -40,7 +40,7 @@
 #endif
 #define BSYM(sym)	sym + 1
 
-#else	/* !CONFIG_THUMB2_KERNEL */
+#else	/* !CONFIG_THUMB2_BAREBOX */
 
 /* The CPSR bit describing the instruction set (ARM) */
 #define PSR_ISETSTATE	0
@@ -52,7 +52,7 @@
 #endif
 #define BSYM(sym)	sym
 
-#endif	/* CONFIG_THUMB2_KERNEL */
+#endif	/* CONFIG_THUMB2_BAREBOX */
 
 #ifndef CONFIG_ARM_ASM_UNIFIED
 
diff --git a/arch/arm/lib/armlinux.c b/arch/arm/lib/armlinux.c
index 85fe2b9f66..a167036db5 100644
--- a/arch/arm/lib/armlinux.c
+++ b/arch/arm/lib/armlinux.c
@@ -255,6 +255,7 @@ void start_linux(void *adr, int swap, unsigned long initrd_address,
 {
 	void (*kernel)(int zero, int arch, void *params) = adr;
 	void *params = NULL;
+	int architecture;
 
 	if (oftree) {
 		printf("booting Linux kernel with devicetree\n");
@@ -272,5 +273,19 @@ void start_linux(void *adr, int swap, unsigned long initrd_address,
 		__asm__ __volatile__("mcr p15, 0, %0, c1, c0" :: "r" (reg));
 	}
 
-	kernel(0, armlinux_get_architecture(), params);
+	architecture = armlinux_get_architecture();
+
+#ifdef CONFIG_THUMB2_BAREBOX
+	__asm__ __volatile__ (
+		"mov r0, #0\n"
+		"mov r1, %0\n"
+		"mov r2, %1\n"
+		"bx %2\n"
+		:
+		: "r" (architecture), "r" (params), "r" (kernel)
+		: "r0", "r1", "r2"
+	);
+#else
+	kernel(0, architecture, params);
+#endif
 }
diff --git a/arch/arm/lib/barebox.lds.S b/arch/arm/lib/barebox.lds.S
index e9f62107f9..bc20694945 100644
--- a/arch/arm/lib/barebox.lds.S
+++ b/arch/arm/lib/barebox.lds.S
@@ -26,7 +26,7 @@
 
 OUTPUT_FORMAT("elf32-littlearm", "elf32-littlearm", "elf32-littlearm")
 OUTPUT_ARCH(arm)
-ENTRY(exception_vectors)
+ENTRY(start)
 SECTIONS
 {
 	. = TEXT_BASE;
@@ -39,6 +39,8 @@ SECTIONS
 		_stext = .;
 		_text = .;
 		*(.text_entry*)
+		__ll_return = .;
+		*(.text_ll_return*)
 #ifdef CONFIG_ARCH_EP93XX
 		/* the EP93xx expects to find the pattern 'CRUS' at 0x1000 */
 	  . = 0x1000;
@@ -47,6 +49,9 @@ SECTIONS
 		__bare_init_start = .;
 		*(.text_bare_init*)
 		__bare_init_end = .;
+		__exceptions_start = .;
+		KEEP(*(.text_exceptions*))
+		__exceptions_stop = .;
 		*(.text*)
 	}
 	BAREBOX_BARE_INIT_SIZE
diff --git a/commands/go.c b/commands/go.c
index b97fcef32a..c821207a6f 100644
--- a/commands/go.c
+++ b/commands/go.c
@@ -62,7 +62,11 @@ static int do_go(int argc, char *argv[])
 	func = addr;
 
 	shutdown_barebox();
-	func(argc - 1, &argv[1]);
+
+	if (do_execute)
+		do_execute(func, argc - 1, &argv[1]);
+	else
+		func(argc - 1, &argv[1]);
 
 	/*
 	 * The application returned. Since we have shutdown barebox and
diff --git a/common/misc.c b/common/misc.c
index a0059c1b79..b31a45c1a4 100644
--- a/common/misc.c
+++ b/common/misc.c
@@ -125,3 +125,6 @@ void perror(const char *s)
 #endif
 }
 EXPORT_SYMBOL(perror);
+
+void (*do_execute)(void *func, int argc, char *argv[]);
+EXPORT_SYMBOL(do_execute);
diff --git a/include/common.h b/include/common.h
index d22787542e..d2347f8629 100644
--- a/include/common.h
+++ b/include/common.h
@@ -146,6 +146,12 @@ unsigned long strtoul_suffix(const char *str, char **endp, int base);
 void start_barebox(void);
 void shutdown_barebox(void);
 
+/*
+ * architectures which have special calling conventions for
+ * executing programs should set this. Used by the 'go' command
+ */
+extern void (*do_execute)(void *func, int argc, char *argv[]);
+
 void arch_shutdown(void);
 
 int run_shell(void);