[PATCH 00/23] entry assembly cleanups

[PATCH 00/23] entry assembly cleanups

[Russell King - ARM Linux]

This patch series cleans up the entry assembly and CPU abort helpers.
Notable things about this patch series:

1. We now call the data and prefetch abort handlers with interrupts
   disabled.  This makes us more like x86 in this regard, and means
   we can move the PMU abort special case out of the assembly.

   As a result of this, we now always tell the irqs tracing code that
   irqs are turned off whenever we enter an exception handler from SVC
   mode.

2. The CPU abort helpers are no longer functions in their own right -
   they tail-call through to the main C abort handlers now, resulting
   in less entry code.

3. IRQs off tracing is implemented more accurately for user aborts.

4. Avoid reloading PSR into r5 on return in SVC exception handlers as
   we now preserve the value across the code.  There is no reason for
   PSR to be changed while processing an SVC mode interrupt, prefetch
   or data abort.

I did notice this though in abort-lv4t.S:

        tst     r8, #1 << 20                    @ L = 1 -> write?

where the comment is wrong.  L means load.  Set for load, so it's a
read.  The ARM610/ARM710 (proc-arm6_7.S) got this right, so I'm not
sure why abort-lv4t.S (which is basically a copy plus thumb handling)
decided to change this to be incorrect.  I haven't fixed this in this
series.

[PATCH 01/23] ARM: entry: remove unused irq_prio_table macro

[Russell King - ARM Linux]

Platforms provide an empty irq_prio_table macro, and as nothing uses
this macro, it can simply be removed.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/mach-bcmring/include/mach/entry-macro.S |    4 ----
 arch/arm/mach-davinci/include/mach/entry-macro.S |    3 ---
 arch/arm/mach-h720x/include/mach/entry-macro.S   |    3 ---
 arch/arm/mach-lpc32xx/include/mach/entry-macro.S |    4 ----
 arch/arm/mach-omap2/include/mach/entry-macro.S   |    3 ---
 arch/arm/mach-pnx4008/include/mach/entry-macro.S |    5 -----
 arch/arm/plat-mxc/include/mach/entry-macro.S     |    4 ----
 7 files changed, 0 insertions(+), 26 deletions(-)

diff --git a/arch/arm/mach-bcmring/include/mach/entry-macro.S b/arch/arm/mach-bcmring/include/mach/entry-macro.S
index 7d393ca..94c950d 100644
--- a/arch/arm/mach-bcmring/include/mach/entry-macro.S
+++ b/arch/arm/mach-bcmring/include/mach/entry-macro.S
@@ -80,7 +80,3 @@
 
 		.macro  arch_ret_to_user, tmp1, tmp2
 		.endm
-
-		.macro	irq_prio_table
-		.endm
-
diff --git a/arch/arm/mach-davinci/include/mach/entry-macro.S b/arch/arm/mach-davinci/include/mach/entry-macro.S
index fbdebc7..e14c0dc 100644
--- a/arch/arm/mach-davinci/include/mach/entry-macro.S
+++ b/arch/arm/mach-davinci/include/mach/entry-macro.S
@@ -46,6 +46,3 @@
 #endif
 1002:
 		.endm
-
-		.macro	irq_prio_table
-		.endm
diff --git a/arch/arm/mach-h720x/include/mach/entry-macro.S b/arch/arm/mach-h720x/include/mach/entry-macro.S
index 6d3b917..c3948e5 100644
--- a/arch/arm/mach-h720x/include/mach/entry-macro.S
+++ b/arch/arm/mach-h720x/include/mach/entry-macro.S
@@ -57,9 +57,6 @@
 		tst     \irqstat, #1		       @ bit 0 should be set
 		.endm
 
-		.macro  irq_prio_table
-		.endm
-
 #else
 #error hynix processor selection missmatch
 #endif
diff --git a/arch/arm/mach-lpc32xx/include/mach/entry-macro.S b/arch/arm/mach-lpc32xx/include/mach/entry-macro.S
index 870227c..b725f6c 100644
--- a/arch/arm/mach-lpc32xx/include/mach/entry-macro.S
+++ b/arch/arm/mach-lpc32xx/include/mach/entry-macro.S
@@ -41,7 +41,3 @@
 	rsb	\irqnr, \irqnr, #31
 	teq	\irqstat, #0
 	.endm
-
-	.macro	irq_prio_table
-	.endm
-
diff --git a/arch/arm/mach-omap2/include/mach/entry-macro.S b/arch/arm/mach-omap2/include/mach/entry-macro.S
index a48690b..ceb8b7e 100644
--- a/arch/arm/mach-omap2/include/mach/entry-macro.S
+++ b/arch/arm/mach-omap2/include/mach/entry-macro.S
@@ -165,6 +165,3 @@
 #endif
 
 #endif	/* MULTI_OMAP2 */
-
-		.macro	irq_prio_table
-		.endm
diff --git a/arch/arm/mach-pnx4008/include/mach/entry-macro.S b/arch/arm/mach-pnx4008/include/mach/entry-macro.S
index 8003037..db7eeeb 100644
--- a/arch/arm/mach-pnx4008/include/mach/entry-macro.S
+++ b/arch/arm/mach-pnx4008/include/mach/entry-macro.S
@@ -120,8 +120,3 @@
 1003:
 		.endm
 
-
-		.macro	irq_prio_table
-		.endm
-
-
diff --git a/arch/arm/plat-mxc/include/mach/entry-macro.S b/arch/arm/plat-mxc/include/mach/entry-macro.S
index 2e49e71..066d464 100644
--- a/arch/arm/plat-mxc/include/mach/entry-macro.S
+++ b/arch/arm/plat-mxc/include/mach/entry-macro.S
@@ -78,7 +78,3 @@
 	movs \irqnr, \irqnr
 #endif
 	.endm
-
-	@ irq priority table (not used)
-	.macro	irq_prio_table
-	.endm
-- 
1.7.4.4

[PATCH 02/23] ARM: entry: shark: don't directly reference registers in macros

[Russell King - ARM Linux]

Directly referencing registers in macros makes assembly code harder
to change, because the macros have side effects which are non-obvious.
Use the provided 'base' register rather than directly referencing r4.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/mach-shark/include/mach/entry-macro.S |   10 +++++-----
 1 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/arch/arm/mach-shark/include/mach/entry-macro.S b/arch/arm/mach-shark/include/mach/entry-macro.S
index e2853c0..0bb6cc6 100644
--- a/arch/arm/mach-shark/include/mach/entry-macro.S
+++ b/arch/arm/mach-shark/include/mach/entry-macro.S
@@ -11,17 +11,17 @@
 		.endm
 
 		.macro  get_irqnr_preamble, base, tmp
+		mov	\base, #0xe0000000
 		.endm
 
 		.macro  arch_ret_to_user, tmp1, tmp2
 		.endm
 
 		.macro	get_irqnr_and_base, irqnr, irqstat, base, tmp
-		mov	r4, #0xe0000000
 
 		mov	\irqstat, #0x0C
-		strb	\irqstat, [r4, #0x20]		@outb(0x0C, 0x20) /* Poll command */
-		ldrb	\irqnr, [r4, #0x20]		@irq = inb(0x20) & 7
+		strb	\irqstat, [\base, #0x20]	@outb(0x0C, 0x20) /* Poll command */
+		ldrb	\irqnr, [\base, #0x20]		@irq = inb(0x20) & 7
 		and	\irqstat, \irqnr, #0x80
 		teq	\irqstat, #0
 		beq	43f
@@ -29,8 +29,8 @@
 		teq	\irqnr, #2
 		bne	44f
 43:		mov	\irqstat, #0x0C
-		strb	\irqstat, [r4, #0xa0]		@outb(0x0C, 0xA0) /* Poll command */
-		ldrb	\irqnr, [r4, #0xa0]		@irq = (inb(0xA0) & 7) + 8
+		strb	\irqstat, [\base, #0xa0]	@outb(0x0C, 0xA0) /* Poll command */
+		ldrb	\irqnr, [\base, #0xa0]		@irq = (inb(0xA0) & 7) + 8
 		and	\irqstat, \irqnr, #0x80
 		teq	\irqstat, #0
 		beq	44f
-- 
1.7.4.4

[PATCH 03/23] ARM: entry: prefetch/data abort helpers: convert to macros

[Russell King - ARM Linux]

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |   85 +++++++++++++++++------------------------
 1 files changed, 35 insertions(+), 50 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 90c62cd..dbe9eb8 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -44,6 +44,37 @@
 9997:
 	.endm
 
+	.macro	pabt_helper
+	mov	r0, r2			@ pass address of aborted instruction.
+#ifdef MULTI_PABORT
+	ldr	r4, .LCprocfns
+	mov	lr, pc
+	ldr	pc, [r4, #PROCESSOR_PABT_FUNC]
+#else
+	bl	CPU_PABORT_HANDLER
+#endif
+	.endm
+
+	.macro	dabt_helper
+
+	@
+	@ Call the processor-specific abort handler:
+	@
+	@  r2 - aborted context pc
+	@  r3 - aborted context cpsr
+	@
+	@ The abort handler must return the aborted address in r0, and
+	@ the fault status register in r1.  r9 must be preserved.
+	@
+#ifdef MULTI_DABORT
+	ldr	r4, .LCprocfns
+	mov	lr, pc
+	ldr	pc, [r4, #PROCESSOR_DABT_FUNC]
+#else
+	bl	CPU_DABORT_HANDLER
+#endif
+	.endm
+
 #ifdef CONFIG_KPROBES
 	.section	.kprobes.text,"ax",%progbits
 #else
@@ -159,22 +190,7 @@ __dabt_svc:
 	tst	r3, #PSR_I_BIT
 	biceq	r9, r9, #PSR_I_BIT
 
-	@
-	@ Call the processor-specific abort handler:
-	@
-	@  r2 - aborted context pc
-	@  r3 - aborted context cpsr
-	@
-	@ The abort handler must return the aborted address in r0, and
-	@ the fault status register in r1.  r9 must be preserved.
-	@
-#ifdef MULTI_DABORT
-	ldr	r4, .LCprocfns
-	mov	lr, pc
-	ldr	pc, [r4, #PROCESSOR_DABT_FUNC]
-#else
-	bl	CPU_DABORT_HANDLER
-#endif
+	dabt_helper
 
 	@
 	@ set desired IRQ state, then call main handler
@@ -298,14 +314,7 @@ __pabt_svc:
 	tst	r3, #PSR_I_BIT
 	biceq	r9, r9, #PSR_I_BIT
 
-	mov	r0, r2			@ pass address of aborted instruction.
-#ifdef MULTI_PABORT
-	ldr	r4, .LCprocfns
-	mov	lr, pc
-	ldr	pc, [r4, #PROCESSOR_PABT_FUNC]
-#else
-	bl	CPU_PABORT_HANDLER
-#endif
+	pabt_helper
 	debug_entry r1
 	msr	cpsr_c, r9			@ Maybe enable interrupts
 	mov	r2, sp				@ regs
@@ -401,23 +410,7 @@ ENDPROC(__pabt_svc)
 __dabt_usr:
 	usr_entry
 	kuser_cmpxchg_check
-
-	@
-	@ Call the processor-specific abort handler:
-	@
-	@  r2 - aborted context pc
-	@  r3 - aborted context cpsr
-	@
-	@ The abort handler must return the aborted address in r0, and
-	@ the fault status register in r1.
-	@
-#ifdef MULTI_DABORT
-	ldr	r4, .LCprocfns
-	mov	lr, pc
-	ldr	pc, [r4, #PROCESSOR_DABT_FUNC]
-#else
-	bl	CPU_DABORT_HANDLER
-#endif
+	dabt_helper
 
 	@
 	@ IRQs on, then call the main handler
@@ -682,15 +675,7 @@ ENDPROC(__und_usr_unknown)
 	.align	5
 __pabt_usr:
 	usr_entry
-
-	mov	r0, r2			@ pass address of aborted instruction.
-#ifdef MULTI_PABORT
-	ldr	r4, .LCprocfns
-	mov	lr, pc
-	ldr	pc, [r4, #PROCESSOR_PABT_FUNC]
-#else
-	bl	CPU_PABORT_HANDLER
-#endif
+	pabt_helper
 	debug_entry r1
 	enable_irq				@ Enable interrupts
 	mov	r2, sp				@ regs
-- 
1.7.4.4

[PATCH 00/23] entry assembly cleanups

[Russell King - ARM Linux]

Oops, here's the overall diffstat:

 arch/arm/include/asm/entry-macro-multi.S         |   14 +-
 arch/arm/kernel/entry-armv.S                     |  271 +++++++++-------------
 arch/arm/kernel/entry-header.S                   |   19 --
 arch/arm/kernel/hw_breakpoint.c                  |    6 +-
 arch/arm/mach-bcmring/include/mach/entry-macro.S |    4 -
 arch/arm/mach-davinci/include/mach/entry-macro.S |    3 -
 arch/arm/mach-h720x/include/mach/entry-macro.S   |    3 -
 arch/arm/mach-lpc32xx/include/mach/entry-macro.S |    4 -
 arch/arm/mach-omap2/include/mach/entry-macro.S   |    3 -
 arch/arm/mach-pnx4008/include/mach/entry-macro.S |    5 -
 arch/arm/mach-shark/include/mach/entry-macro.S   |   10 +-
 arch/arm/mm/abort-ev4.S                          |   17 +-
 arch/arm/mm/abort-ev4t.S                         |   17 +-
 arch/arm/mm/abort-ev5t.S                         |   19 +-
 arch/arm/mm/abort-ev5tj.S                        |   25 +--
 arch/arm/mm/abort-ev6.S                          |   25 +--
 arch/arm/mm/abort-ev7.S                          |   25 +--
 arch/arm/mm/abort-lv4t.S                         |  141 ++++++------
 arch/arm/mm/abort-macro.S                        |   34 ++--
 arch/arm/mm/abort-nommu.S                        |   10 +-
 arch/arm/mm/alignment.c                          |    3 +
 arch/arm/mm/fault.c                              |    4 +
 arch/arm/mm/pabort-legacy.S                      |   10 +-
 arch/arm/mm/pabort-v6.S                          |   10 +-
 arch/arm/mm/pabort-v7.S                          |   11 +-
 arch/arm/mm/proc-arm6_7.S                        |   90 ++++----
 arch/arm/plat-mxc/include/mach/entry-macro.S     |    4 -
 27 files changed, 333 insertions(+), 454 deletions(-)

[PATCH 04/23] ARM: entry: prefetch/data abort helpers: avoid corrupting r4

[Russell King - ARM Linux]

Replace r4 with ip for calling abort helpers - ip is allowed to be
corrupted by called functions in the ABI, so it makes more sense to
use such a register.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |    8 ++++----
 1 files changed, 4 insertions(+), 4 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index dbe9eb8..6855f6d 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -47,9 +47,9 @@
 	.macro	pabt_helper
 	mov	r0, r2			@ pass address of aborted instruction.
 #ifdef MULTI_PABORT
-	ldr	r4, .LCprocfns
+	ldr	ip, .LCprocfns
 	mov	lr, pc
-	ldr	pc, [r4, #PROCESSOR_PABT_FUNC]
+	ldr	pc, [ip, #PROCESSOR_PABT_FUNC]
 #else
 	bl	CPU_PABORT_HANDLER
 #endif
@@ -67,9 +67,9 @@
 	@ the fault status register in r1.  r9 must be preserved.
 	@
 #ifdef MULTI_DABORT
-	ldr	r4, .LCprocfns
+	ldr	ip, .LCprocfns
 	mov	lr, pc
-	ldr	pc, [r4, #PROCESSOR_DABT_FUNC]
+	ldr	pc, [ip, #PROCESSOR_DABT_FUNC]
 #else
 	bl	CPU_DABORT_HANDLER
 #endif
-- 
1.7.4.4

[PATCH 05/23] ARM: entry: abort-macro: specify registers to be used for macros

[Russell King - ARM Linux]

Require all callers of abort macros to specify the registers to be
used.  This improves the documentation at the callsites as to which
registers are being used by this assembly code.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/mm/abort-ev4t.S  |    2 +-
 arch/arm/mm/abort-ev5t.S  |    4 ++--
 arch/arm/mm/abort-ev5tj.S |    4 ++--
 arch/arm/mm/abort-ev6.S   |    4 ++--
 arch/arm/mm/abort-macro.S |   30 +++++++++++++++---------------
 5 files changed, 22 insertions(+), 22 deletions(-)

diff --git a/arch/arm/mm/abort-ev4t.S b/arch/arm/mm/abort-ev4t.S
index b628254..9910123 100644
--- a/arch/arm/mm/abort-ev4t.S
+++ b/arch/arm/mm/abort-ev4t.S
@@ -22,7 +22,7 @@
 ENTRY(v4t_early_abort)
 	mrc	p15, 0, r1, c5, c0, 0		@ get FSR
 	mrc	p15, 0, r0, c6, c0, 0		@ get FAR
-	do_thumb_abort
+	do_thumb_abort fsr=r1, pc=r2, psr=r3, tmp=r3
 	ldreq	r3, [r2]			@ read aborted ARM instruction
 	bic	r1, r1, #1 << 11 | 1 << 10	@ clear bits 11 and 10 of FSR
 	tst	r3, #1 << 20			@ check write
diff --git a/arch/arm/mm/abort-ev5t.S b/arch/arm/mm/abort-ev5t.S
index 02251b5..800e8d4 100644
--- a/arch/arm/mm/abort-ev5t.S
+++ b/arch/arm/mm/abort-ev5t.S
@@ -22,10 +22,10 @@
 ENTRY(v5t_early_abort)
 	mrc	p15, 0, r1, c5, c0, 0		@ get FSR
 	mrc	p15, 0, r0, c6, c0, 0		@ get FAR
-	do_thumb_abort
+	do_thumb_abort fsr=r1, pc=r2, psr=r3, tmp=r3
 	ldreq	r3, [r2]			@ read aborted ARM instruction
 	bic	r1, r1, #1 << 11		@ clear bits 11 of FSR
-	do_ldrd_abort
+	do_ldrd_abort tmp=r2, insn=r3
 	tst	r3, #1 << 20			@ check write
 	orreq	r1, r1, #1 << 11
 	mov	pc, lr
diff --git a/arch/arm/mm/abort-ev5tj.S b/arch/arm/mm/abort-ev5tj.S
index bce68d6..bcb58d2 100644
--- a/arch/arm/mm/abort-ev5tj.S
+++ b/arch/arm/mm/abort-ev5tj.S
@@ -25,9 +25,9 @@ ENTRY(v5tj_early_abort)
 	bic	r1, r1, #1 << 11 | 1 << 10	@ clear bits 11 and 10 of FSR
 	tst	r3, #PSR_J_BIT			@ Java?
 	movne	pc, lr
-	do_thumb_abort
+	do_thumb_abort fsr=r1, pc=r2, psr=r3, tmp=r3
 	ldreq	r3, [r2]			@ read aborted ARM instruction
-	do_ldrd_abort
+	do_ldrd_abort tmp=r2, insn=r3
 	tst	r3, #1 << 20			@ L = 0 -> write
 	orreq	r1, r1, #1 << 11		@ yes.
 	mov	pc, lr
diff --git a/arch/arm/mm/abort-ev6.S b/arch/arm/mm/abort-ev6.S
index 1478aa5..ef526e7 100644
--- a/arch/arm/mm/abort-ev6.S
+++ b/arch/arm/mm/abort-ev6.S
@@ -35,12 +35,12 @@ ENTRY(v6_early_abort)
 	bic	r1, r1, #1 << 11		@ clear bit 11 of FSR
 	tst	r3, #PSR_J_BIT			@ Java?
 	movne	pc, lr
-	do_thumb_abort
+	do_thumb_abort fsr=r1, pc=r2, psr=r3, tmp=r3
 	ldreq	r3, [r2]			@ read aborted ARM instruction
 #ifdef CONFIG_CPU_ENDIAN_BE8
 	reveq	r3, r3
 #endif
-	do_ldrd_abort
+	do_ldrd_abort tmp=r2, insn=r3
 	tst	r3, #1 << 20			@ L = 0 -> write
 	orreq	r1, r1, #1 << 11		@ yes.
 	mov	pc, lr
diff --git a/arch/arm/mm/abort-macro.S b/arch/arm/mm/abort-macro.S
index d7cb1bf..8d3b9f9 100644
--- a/arch/arm/mm/abort-macro.S
+++ b/arch/arm/mm/abort-macro.S
@@ -9,33 +9,33 @@
  *
  */
 
-	.macro	do_thumb_abort
-	tst	r3, #PSR_T_BIT
+	.macro	do_thumb_abort, fsr, pc, psr, tmp
+	tst	\psr, #PSR_T_BIT
 	beq	not_thumb
-	ldrh	r3, [r2]			@ Read aborted Thumb instruction
-	and	r3, r3, # 0xfe00		@ Mask opcode field
-	cmp	r3, # 0x5600			@ Is it ldrsb?
-	orreq	r3, r3, #1 << 11		@ Set L-bit if yes
-	tst	r3, #1 << 11			@ L = 0 -> write
-	orreq	r1, r1, #1 << 11		@ yes.
+	ldrh	\tmp, [\pc]			@ Read aborted Thumb instruction
+	and	\tmp, \tmp, # 0xfe00		@ Mask opcode field
+	cmp	\tmp, # 0x5600			@ Is it ldrsb?
+	orreq	\tmp, \tmp, #1 << 11		@ Set L-bit if yes
+	tst	\tmp, #1 << 11			@ L = 0 -> write
+	orreq	\psr, \psr, #1 << 11		@ yes.
 	mov	pc, lr
 not_thumb:
 	.endm
 
 /*
- * We check for the following insturction encoding for LDRD.
+ * We check for the following instruction encoding for LDRD.
  *
- * [27:25] == 0
+ * [27:25] == 000
  *   [7:4] == 1101
  *    [20] == 0
  */
- 	.macro	do_ldrd_abort
- 	tst	r3, #0x0e000000			@ [27:25] == 0
+	.macro	do_ldrd_abort, tmp, insn
+	tst	\insn, #0x0e000000		@ [27:25] == 0
 	bne	not_ldrd
-	and	r2, r3, #0x000000f0		@ [7:4] == 1101
-	cmp	r2, #0x000000d0
+	and	\tmp, \insn, #0x000000f0	@ [7:4] == 1101
+	cmp	\tmp, #0x000000d0
 	bne	not_ldrd
-	tst	r3, #1 << 20			@ [20] == 0
+	tst	\insn, #1 << 20			@ [20] == 0
 	moveq	pc, lr
 not_ldrd:
 	.endm
-- 
1.7.4.4

[PATCH 06/23] ARM: entry: abort-macro: simplify do_ldrd_abort

[Russell King - ARM Linux]

We can test bits 27:25 and 20 of the instruction at the same time;
there's no need to separate out the check of bit 20.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/mm/abort-macro.S |    4 +---
 1 files changed, 1 insertions(+), 3 deletions(-)

diff --git a/arch/arm/mm/abort-macro.S b/arch/arm/mm/abort-macro.S
index 8d3b9f9..af97a10 100644
--- a/arch/arm/mm/abort-macro.S
+++ b/arch/arm/mm/abort-macro.S
@@ -30,12 +30,10 @@ not_thumb:
  *    [20] == 0
  */
 	.macro	do_ldrd_abort, tmp, insn
-	tst	\insn, #0x0e000000		@ [27:25] == 0
+	tst	\insn, #0x0e100000		@ [27:25,20] == 0
 	bne	not_ldrd
 	and	\tmp, \insn, #0x000000f0	@ [7:4] == 1101
 	cmp	\tmp, #0x000000d0
-	bne	not_ldrd
-	tst	\insn, #1 << 20			@ [20] == 0
 	moveq	pc, lr
 not_ldrd:
 	.endm
-- 
1.7.4.4

[PATCH 07/23] ARM: entry: no need to increase preempt count for IRQ handlers

[Russell King - ARM Linux]

irq_enter() and irq_exit() already take care of the preempt_count
handling for interrupts, which increment and decrement the hardirq
bits of the preempt count.  So we can remove the preempt count handing
in our IRQ entry/exit assembly, like x86 did some 9 years ago.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |   27 ++++-----------------------
 1 files changed, 4 insertions(+), 23 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 6855f6d..1e5f387 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -220,16 +220,12 @@ __irq_svc:
 #ifdef CONFIG_TRACE_IRQFLAGS
 	bl	trace_hardirqs_off
 #endif
-#ifdef CONFIG_PREEMPT
-	get_thread_info tsk
-	ldr	r8, [tsk, #TI_PREEMPT]		@ get preempt count
-	add	r7, r8, #1			@ increment it
-	str	r7, [tsk, #TI_PREEMPT]
-#endif
 
 	irq_handler
+
 #ifdef CONFIG_PREEMPT
-	str	r8, [tsk, #TI_PREEMPT]		@ restore preempt count
+	get_thread_info tsk
+	ldr	r8, [tsk, #TI_PREEMPT]		@ get preempt count
 	ldr	r0, [tsk, #TI_FLAGS]		@ get flags
 	teq	r8, #0				@ if preempt count != 0
 	movne	r0, #0				@ force flags to 0
@@ -432,23 +428,8 @@ __irq_usr:
 	bl	trace_hardirqs_off
 #endif
 
-	get_thread_info tsk
-#ifdef CONFIG_PREEMPT
-	ldr	r8, [tsk, #TI_PREEMPT]		@ get preempt count
-	add	r7, r8, #1			@ increment it
-	str	r7, [tsk, #TI_PREEMPT]
-#endif
-
 	irq_handler
-#ifdef CONFIG_PREEMPT
-	ldr	r0, [tsk, #TI_PREEMPT]
-	str	r8, [tsk, #TI_PREEMPT]
-	teq	r0, r7
- ARM(	strne	r0, [r0, -r0]	)
- THUMB(	movne	r0, #0		)
- THUMB(	strne	r0, [r0]	)
-#endif
-
+	get_thread_info tsk
 	mov	why, #0
 	b	ret_to_user_from_irq
  UNWIND(.fnend		)
-- 
1.7.4.4

[PATCH 08/23] ARM: entry: no need to check parent IRQ mask in IRQ handler return

[Russell King - ARM Linux]

There's no point checking to see whether IRQs were masked in the parent
context when returning from IRQ handling - the fact that we're handling
an IRQ means that the parent context must have had IRQs unmasked.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |    5 +++--
 1 files changed, 3 insertions(+), 2 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 1e5f387..fd42e66 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -234,8 +234,9 @@ __irq_svc:
 #endif
 	ldr	r4, [sp, #S_PSR]		@ irqs are already disabled
 #ifdef CONFIG_TRACE_IRQFLAGS
-	tst	r4, #PSR_I_BIT
-	bleq	trace_hardirqs_on
+	@ The parent context IRQs must have been enabled to get here in
+	@ the first place, so there's no point checking the PSR I bit.
+	bl	trace_hardirqs_on
 #endif
 	svc_exit r4				@ return from exception
  UNWIND(.fnend		)
-- 
1.7.4.4

[PATCH 09/23] ARM: entry: rejig register allocation in exception entry handlers

[Russell King - ARM Linux]

This allows us to avoid moving registers twice to work around the
clobbered registers when we add calls to trace_hardirqs_{on,off}.

Ensure that all SVC handlers return with SPSR in r5 for consistency.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |   75 ++++++++++++++++++++++-------------------
 1 files changed, 40 insertions(+), 35 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index fd42e66..353b639 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -45,7 +45,7 @@
 	.endm
 
 	.macro	pabt_helper
-	mov	r0, r2			@ pass address of aborted instruction.
+	mov	r0, r4			@ pass address of aborted instruction.
 #ifdef MULTI_PABORT
 	ldr	ip, .LCprocfns
 	mov	lr, pc
@@ -56,6 +56,8 @@
 	.endm
 
 	.macro	dabt_helper
+	mov	r2, r4
+	mov	r3, r5
 
 	@
 	@ Call the processor-specific abort handler:
@@ -157,26 +159,26 @@ ENDPROC(__und_invalid)
  SPFIX(	subeq	sp, sp, #4	)
 	stmia	sp, {r1 - r12}
 
-	ldmia	r0, {r1 - r3}
-	add	r5, sp, #S_SP - 4	@ here for interlock avoidance
-	mov	r4, #-1			@  ""  ""      ""       ""
-	add	r0, sp, #(S_FRAME_SIZE + \stack_hole - 4)
- SPFIX(	addeq	r0, r0, #4	)
-	str	r1, [sp, #-4]!		@ save the "real" r0 copied
+	ldmia	r0, {r3 - r5}
+	add	r7, sp, #S_SP - 4	@ here for interlock avoidance
+	mov	r6, #-1			@  ""  ""      ""       ""
+	add	r2, sp, #(S_FRAME_SIZE + \stack_hole - 4)
+ SPFIX(	addeq	r2, r2, #4	)
+	str	r3, [sp, #-4]!		@ save the "real" r0 copied
 					@ from the exception stack
 
-	mov	r1, lr
+	mov	r3, lr
 
 	@
 	@ We are now ready to fill in the remaining blanks on the stack:
 	@
-	@  r0 - sp_svc
-	@  r1 - lr_svc
-	@  r2 - lr_<exception>, already fixed up for correct return/restart
-	@  r3 - spsr_<exception>
-	@  r4 - orig_r0 (see pt_regs definition in ptrace.h)
+	@  r2 - sp_svc
+	@  r3 - lr_svc
+	@  r4 - lr_<exception>, already fixed up for correct return/restart
+	@  r5 - spsr_<exception>
+	@  r6 - orig_r0 (see pt_regs definition in ptrace.h)
 	@
-	stmia	r5, {r0 - r4}
+	stmia	r7, {r2 - r6}
 	.endm
 
 	.align	5
@@ -187,7 +189,7 @@ __dabt_svc:
 	@ get ready to re-enable interrupts if appropriate
 	@
 	mrs	r9, cpsr
-	tst	r3, #PSR_I_BIT
+	tst	r5, #PSR_I_BIT
 	biceq	r9, r9, #PSR_I_BIT
 
 	dabt_helper
@@ -208,8 +210,8 @@ __dabt_svc:
 	@
 	@ restore SPSR and restart the instruction
 	@
-	ldr	r2, [sp, #S_PSR]
-	svc_exit r2				@ return from exception
+	ldr	r5, [sp, #S_PSR]
+	svc_exit r5				@ return from exception
  UNWIND(.fnend		)
 ENDPROC(__dabt_svc)
 
@@ -232,13 +234,13 @@ __irq_svc:
 	tst	r0, #_TIF_NEED_RESCHED
 	blne	svc_preempt
 #endif
-	ldr	r4, [sp, #S_PSR]		@ irqs are already disabled
+	ldr	r5, [sp, #S_PSR]
 #ifdef CONFIG_TRACE_IRQFLAGS
 	@ The parent context IRQs must have been enabled to get here in
 	@ the first place, so there's no point checking the PSR I bit.
 	bl	trace_hardirqs_on
 #endif
-	svc_exit r4				@ return from exception
+	svc_exit r5				@ return from exception
  UNWIND(.fnend		)
 ENDPROC(__irq_svc)
 
@@ -273,15 +275,16 @@ __und_svc:
 	@  r0 - instruction
 	@
 #ifndef	CONFIG_THUMB2_KERNEL
-	ldr	r0, [r2, #-4]
+	ldr	r0, [r4, #-4]
 #else
-	ldrh	r0, [r2, #-2]			@ Thumb instruction at LR - 2
+	ldrh	r0, [r4, #-2]			@ Thumb instruction at LR - 2
 	and	r9, r0, #0xf800
 	cmp	r9, #0xe800			@ 32-bit instruction if xx >= 0
-	ldrhhs	r9, [r2]			@ bottom 16 bits
+	ldrhhs	r9, [r4]			@ bottom 16 bits
 	orrhs	r0, r9, r0, lsl #16
 #endif
 	adr	r9, BSYM(1f)
+	mov	r2, r4
 	bl	call_fpe
 
 	mov	r0, sp				@ struct pt_regs *regs
@@ -295,8 +298,8 @@ __und_svc:
 	@
 	@ restore SPSR and restart the instruction
 	@
-	ldr	r2, [sp, #S_PSR]		@ Get SVC cpsr
-	svc_exit r2				@ return from exception
+	ldr	r5, [sp, #S_PSR]		@ Get SVC cpsr
+	svc_exit r5				@ return from exception
  UNWIND(.fnend		)
 ENDPROC(__und_svc)
 
@@ -308,7 +311,7 @@ __pabt_svc:
 	@ re-enable interrupts if appropriate
 	@
 	mrs	r9, cpsr
-	tst	r3, #PSR_I_BIT
+	tst	r5, #PSR_I_BIT
 	biceq	r9, r9, #PSR_I_BIT
 
 	pabt_helper
@@ -325,8 +328,8 @@ __pabt_svc:
 	@
 	@ restore SPSR and restart the instruction
 	@
-	ldr	r2, [sp, #S_PSR]
-	svc_exit r2				@ return from exception
+	ldr	r5, [sp, #S_PSR]
+	svc_exit r5				@ return from exception
  UNWIND(.fnend		)
 ENDPROC(__pabt_svc)
 
@@ -357,23 +360,23 @@ ENDPROC(__pabt_svc)
  ARM(	stmib	sp, {r1 - r12}	)
  THUMB(	stmia	sp, {r0 - r12}	)
 
-	ldmia	r0, {r1 - r3}
+	ldmia	r0, {r3 - r5}
 	add	r0, sp, #S_PC		@ here for interlock avoidance
-	mov	r4, #-1			@  ""  ""     ""        ""
+	mov	r6, #-1			@  ""  ""     ""        ""
 
-	str	r1, [sp]		@ save the "real" r0 copied
+	str	r3, [sp]		@ save the "real" r0 copied
 					@ from the exception stack
 
 	@
 	@ We are now ready to fill in the remaining blanks on the stack:
 	@
-	@  r2 - lr_<exception>, already fixed up for correct return/restart
-	@  r3 - spsr_<exception>
-	@  r4 - orig_r0 (see pt_regs definition in ptrace.h)
+	@  r4 - lr_<exception>, already fixed up for correct return/restart
+	@  r5 - spsr_<exception>
+	@  r6 - orig_r0 (see pt_regs definition in ptrace.h)
 	@
 	@ Also, separately save sp_usr and lr_usr
 	@
-	stmia	r0, {r2 - r4}
+	stmia	r0, {r4 - r6}
  ARM(	stmdb	r0, {sp, lr}^			)
  THUMB(	store_user_sp_lr r0, r1, S_SP - S_PC	)
 
@@ -397,7 +400,7 @@ ENDPROC(__pabt_svc)
 	@ if it was interrupted in a critical region.  Here we
 	@ perform a quick test inline since it should be false
 	@ 99.9999% of the time.  The rest is done out of line.
-	cmp	r2, #TASK_SIZE
+	cmp	r4, #TASK_SIZE
 	blhs	kuser_cmpxchg_fixup
 #endif
 #endif
@@ -441,6 +444,8 @@ ENDPROC(__irq_usr)
 	.align	5
 __und_usr:
 	usr_entry
+	mov	r2, r4
+	mov	r3, r5
 
 	@
 	@ fall through to the emulation code, which returns using r9 if
-- 
1.7.4.4

[PATCH 10/23] ARM: entry: prefetch abort helper: pass aborted pc in r4 rather than r0

[Russell King - ARM Linux]

This avoids unnecessary instructions for CPUs which implement the IFAR
(instruction fault address register).

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |    2 +-
 arch/arm/mm/pabort-legacy.S  |    3 ++-
 arch/arm/mm/pabort-v6.S      |    3 ++-
 arch/arm/mm/pabort-v7.S      |    2 +-
 4 files changed, 6 insertions(+), 4 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 353b639..ee425f7 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -45,7 +45,7 @@
 	.endm
 
 	.macro	pabt_helper
-	mov	r0, r4			@ pass address of aborted instruction.
+	@ PABORT handler takes fault address in r4
 #ifdef MULTI_PABORT
 	ldr	ip, .LCprocfns
 	mov	lr, pc
diff --git a/arch/arm/mm/pabort-legacy.S b/arch/arm/mm/pabort-legacy.S
index 87970eb..8a5d8aa 100644
--- a/arch/arm/mm/pabort-legacy.S
+++ b/arch/arm/mm/pabort-legacy.S
@@ -4,7 +4,7 @@
 /*
  * Function: legacy_pabort
  *
- * Params  : r0 = address of aborted instruction
+ * Params  : r4 = address of aborted instruction
  *
  * Returns : r0 = address of abort
  *	   : r1 = Simulated IFSR with section translation fault status
@@ -14,6 +14,7 @@
 
 	.align	5
 ENTRY(legacy_pabort)
+	mov	r0, r4
 	mov	r1, #5
 	mov	pc, lr
 ENDPROC(legacy_pabort)
diff --git a/arch/arm/mm/pabort-v6.S b/arch/arm/mm/pabort-v6.S
index 06e3d1e..eaac1cb 100644
--- a/arch/arm/mm/pabort-v6.S
+++ b/arch/arm/mm/pabort-v6.S
@@ -4,7 +4,7 @@
 /*
  * Function: v6_pabort
  *
- * Params  : r0 = address of aborted instruction
+ * Params  : r4 = address of aborted instruction
  *
  * Returns : r0 = address of abort
  *	   : r1 = IFSR
@@ -14,6 +14,7 @@
 
 	.align	5
 ENTRY(v6_pabort)
+	mov	r0, r4
 	mrc	p15, 0, r1, c5, c0, 1		@ get IFSR
 	mov	pc, lr
 ENDPROC(v6_pabort)
diff --git a/arch/arm/mm/pabort-v7.S b/arch/arm/mm/pabort-v7.S
index a8b3b30..b515e0b 100644
--- a/arch/arm/mm/pabort-v7.S
+++ b/arch/arm/mm/pabort-v7.S
@@ -4,7 +4,7 @@
 /*
  * Function: v6_pabort
  *
- * Params  : r0 = address of aborted instruction
+ * Params  : r4 = address of aborted instruction
  *
  * Returns : r0 = address of abort
  *	   : r1 = IFSR
-- 
1.7.4.4

[PATCH 11/23] ARM: entry: avoid enabling interrupts in prefetch/data abort handlers

[Russell King - ARM Linux]

Avoid enabling interrupts if the parent context had interrupts enabled
in the abort handler assembly code, and move this into the breakpoint/
page/alignment fault handlers instead.

This gets rid of some special-casing for the breakpoint fault handlers
from the low level abort handler path.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S    |   43 +++++++++++++++++---------------------
 arch/arm/kernel/entry-header.S  |   19 -----------------
 arch/arm/kernel/hw_breakpoint.c |    6 +++-
 arch/arm/mm/alignment.c         |    3 ++
 arch/arm/mm/fault.c             |    4 +++
 5 files changed, 30 insertions(+), 45 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index ee425f7..8048056 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -185,20 +185,15 @@ ENDPROC(__und_invalid)
 __dabt_svc:
 	svc_entry
 
-	@
-	@ get ready to re-enable interrupts if appropriate
-	@
-	mrs	r9, cpsr
-	tst	r5, #PSR_I_BIT
-	biceq	r9, r9, #PSR_I_BIT
+#ifdef CONFIG_TRACE_IRQFLAGS
+	bl	trace_hardirqs_off
+#endif
 
 	dabt_helper
 
 	@
-	@ set desired IRQ state, then call main handler
+	@ call main handler
 	@
-	debug_entry r1
-	msr	cpsr_c, r9
 	mov	r2, sp
 	bl	do_DataAbort
 
@@ -211,6 +206,12 @@ __dabt_svc:
 	@ restore SPSR and restart the instruction
 	@
 	ldr	r5, [sp, #S_PSR]
+#ifdef CONFIG_TRACE_IRQFLAGS
+	tst	r5, #PSR_I_BIT
+	bleq	trace_hardirqs_on
+	tst	r5, #PSR_I_BIT
+	blne	trace_hardirqs_off
+#endif
 	svc_exit r5				@ return from exception
  UNWIND(.fnend		)
 ENDPROC(__dabt_svc)
@@ -307,16 +308,11 @@ ENDPROC(__und_svc)
 __pabt_svc:
 	svc_entry
 
-	@
-	@ re-enable interrupts if appropriate
-	@
-	mrs	r9, cpsr
-	tst	r5, #PSR_I_BIT
-	biceq	r9, r9, #PSR_I_BIT
+#ifdef CONFIG_TRACE_IRQFLAGS
+	bl	trace_hardirqs_off
+#endif
 
 	pabt_helper
-	debug_entry r1
-	msr	cpsr_c, r9			@ Maybe enable interrupts
 	mov	r2, sp				@ regs
 	bl	do_PrefetchAbort		@ call abort handler
 
@@ -329,6 +325,12 @@ __pabt_svc:
 	@ restore SPSR and restart the instruction
 	@
 	ldr	r5, [sp, #S_PSR]
+#ifdef CONFIG_TRACE_IRQFLAGS
+	tst	r5, #PSR_I_BIT
+	bleq	trace_hardirqs_on
+	tst	r5, #PSR_I_BIT
+	blne	trace_hardirqs_off
+#endif
 	svc_exit r5				@ return from exception
  UNWIND(.fnend		)
 ENDPROC(__pabt_svc)
@@ -412,11 +414,6 @@ __dabt_usr:
 	kuser_cmpxchg_check
 	dabt_helper
 
-	@
-	@ IRQs on, then call the main handler
-	@
-	debug_entry r1
-	enable_irq
 	mov	r2, sp
 	adr	lr, BSYM(ret_from_exception)
 	b	do_DataAbort
@@ -663,8 +660,6 @@ ENDPROC(__und_usr_unknown)
 __pabt_usr:
 	usr_entry
 	pabt_helper
-	debug_entry r1
-	enable_irq				@ Enable interrupts
 	mov	r2, sp				@ regs
 	bl	do_PrefetchAbort		@ call abort handler
  UNWIND(.fnend		)
diff --git a/arch/arm/kernel/entry-header.S b/arch/arm/kernel/entry-header.S
index 051166c..4d6ad83 100644
--- a/arch/arm/kernel/entry-header.S
+++ b/arch/arm/kernel/entry-header.S
@@ -165,25 +165,6 @@
 	.endm
 #endif	/* !CONFIG_THUMB2_KERNEL */
 
-	@
-	@ Debug exceptions are taken as prefetch or data aborts.
-	@ We must disable preemption during the handler so that
-	@ we can access the debug registers safely.
-	@
-	.macro	debug_entry, fsr
-#if defined(CONFIG_HAVE_HW_BREAKPOINT) && defined(CONFIG_PREEMPT)
-	ldr	r4, =0x40f		@ mask out fsr.fs
-	and	r5, r4, \fsr
-	cmp	r5, #2			@ debug exception
-	bne	1f
-	get_thread_info r10
-	ldr	r6, [r10, #TI_PREEMPT]	@ get preempt count
-	add	r11, r6, #1		@ increment it
-	str	r11, [r10, #TI_PREEMPT]
-1:
-#endif
-	.endm
-
 /*
  * These are the registers used in the syscall handler, and allow us to
  * have in theory up to 7 arguments to a function - r0 to r6.
diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c
index 87acc25..b813e1e 100644
--- a/arch/arm/kernel/hw_breakpoint.c
+++ b/arch/arm/kernel/hw_breakpoint.c
@@ -804,8 +804,10 @@ static int hw_breakpoint_pending(unsigned long addr, unsigned int fsr,
 	int ret = 0;
 	u32 dscr;
 
-	/* We must be called with preemption disabled. */
-	WARN_ON(preemptible());
+	preempt_disable();
+
+	if (interrupts_enabled(regs))
+		local_irq_enable();
 
 	/* We only handle watchpoints and hardware breakpoints. */
 	ARM_DBG_READ(c1, 0, dscr);
diff --git a/arch/arm/mm/alignment.c b/arch/arm/mm/alignment.c
index 724ba3b..be7c638 100644
--- a/arch/arm/mm/alignment.c
+++ b/arch/arm/mm/alignment.c
@@ -727,6 +727,9 @@ do_alignment(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
 	int isize = 4;
 	int thumb2_32b = 0;
 
+	if (interrupts_enabled(regs))
+		local_irq_enable();
+
 	instrptr = instruction_pointer(regs);
 
 	fs = get_fs();
diff --git a/arch/arm/mm/fault.c b/arch/arm/mm/fault.c
index bc0e1d8..20e5d51 100644
--- a/arch/arm/mm/fault.c
+++ b/arch/arm/mm/fault.c
@@ -285,6 +285,10 @@ do_page_fault(unsigned long addr, unsigned int fsr, struct pt_regs *regs)
 	tsk = current;
 	mm  = tsk->mm;
 
+	/* Enable interrupts if they were enabled in the parent context. */
+	if (interrupts_enabled(regs))
+		local_irq_enable();
+
 	/*
 	 * If we're in an interrupt or have no user
 	 * context, we must not take the fault..
-- 
1.7.4.4

[PATCH 12/23] ARM: entry: instrument svc undefined exception handler with irqtrace

[Russell King - ARM Linux]

Add irqtrace function calls to the undefined exception handler, so
that we get sane lockdep traces from locking problems in undefined
exception handlers.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |   10 ++++++++++
 1 files changed, 10 insertions(+), 0 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 8048056..474934c 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -268,6 +268,10 @@ __und_svc:
 	svc_entry
 #endif
 
+#ifdef CONFIG_TRACE_IRQFLAGS
+	bl	trace_hardirqs_off
+#endif
+
 	@
 	@ call emulation code, which returns using r9 if it has emulated
 	@ the instruction, or the more conventional lr if we are to treat
@@ -300,6 +304,12 @@ __und_svc:
 	@ restore SPSR and restart the instruction
 	@
 	ldr	r5, [sp, #S_PSR]		@ Get SVC cpsr
+#ifdef CONFIG_TRACE_IRQFLAGS
+	tst	r5, #PSR_I_BIT
+	bleq	trace_hardirqs_on
+	tst	r5, #PSR_I_BIT
+	blne	trace_hardirqs_off
+#endif
 	svc_exit r5				@ return from exception
  UNWIND(.fnend		)
 ENDPROC(__und_svc)
-- 
1.7.4.4

[PATCH 13/23] ARM: entry: instrument usr exception handlers with irqsoff tracing

[Russell King - ARM Linux]

As we no longer re-enable interrupts in these exception handlers, add
the irqsoff tracing calls to them so that the kernel tracks the state
more accurately.

Note that these calls are conditional on IRQSOFF_TRACER:

  kernel ----------> user ---------> kernel
          ^ irqs enabled   ^ irqs disabled

No kernel code can run on the local CPU until we've re-entered the
kernel through one of the exception handlers - and userspace can not
take any locks etc.  So, the kernel doesn't care about the IRQ mask
state while userspace is running unless we're doing IRQ off latency
tracing.  So, we can (and do) avoid the overhead of updating the IRQ
mask state on every kernel->user and user->kernel transition.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |   17 ++++++++++++++++-
 1 files changed, 16 insertions(+), 1 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 474934c..c01cdb1 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -421,6 +421,11 @@ ENDPROC(__pabt_svc)
 	.align	5
 __dabt_usr:
 	usr_entry
+
+#ifdef CONFIG_IRQSOFF_TRACER
+	bl	trace_hardirqs_off
+#endif
+
 	kuser_cmpxchg_check
 	dabt_helper
 
@@ -433,12 +438,12 @@ ENDPROC(__dabt_usr)
 	.align	5
 __irq_usr:
 	usr_entry
-	kuser_cmpxchg_check
 
 #ifdef CONFIG_IRQSOFF_TRACER
 	bl	trace_hardirqs_off
 #endif
 
+	kuser_cmpxchg_check
 	irq_handler
 	get_thread_info tsk
 	mov	why, #0
@@ -451,6 +456,11 @@ ENDPROC(__irq_usr)
 	.align	5
 __und_usr:
 	usr_entry
+
+#ifdef CONFIG_IRQSOFF_TRACER
+	bl	trace_hardirqs_off
+#endif
+
 	mov	r2, r4
 	mov	r3, r5
 
@@ -669,6 +679,11 @@ ENDPROC(__und_usr_unknown)
 	.align	5
 __pabt_usr:
 	usr_entry
+
+#ifdef CONFIG_IRQSOFF_TRACER
+	bl	trace_hardirqs_off
+#endif
+
 	pabt_helper
 	mov	r2, sp				@ regs
 	bl	do_PrefetchAbort		@ call abort handler
-- 
1.7.4.4

[PATCH 14/23] ARM: entry: consolidate trace_hardirqs_off into (svc|usr)_entry macros

[Russell King - ARM Linux]

All handlers now call trace_hardirqs_off, so move this common code into
the (svc|usr)_entry assembler macros.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |   47 +++++++----------------------------------
 1 files changed, 8 insertions(+), 39 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index c01cdb1..1da985f 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -179,16 +179,15 @@ ENDPROC(__und_invalid)
 	@  r6 - orig_r0 (see pt_regs definition in ptrace.h)
 	@
 	stmia	r7, {r2 - r6}
-	.endm
-
-	.align	5
-__dabt_svc:
-	svc_entry
 
 #ifdef CONFIG_TRACE_IRQFLAGS
 	bl	trace_hardirqs_off
 #endif
+	.endm
 
+	.align	5
+__dabt_svc:
+	svc_entry
 	dabt_helper
 
 	@
@@ -219,11 +218,6 @@ ENDPROC(__dabt_svc)
 	.align	5
 __irq_svc:
 	svc_entry
-
-#ifdef CONFIG_TRACE_IRQFLAGS
-	bl	trace_hardirqs_off
-#endif
-
 	irq_handler
 
 #ifdef CONFIG_PREEMPT
@@ -267,11 +261,6 @@ __und_svc:
 #else
 	svc_entry
 #endif
-
-#ifdef CONFIG_TRACE_IRQFLAGS
-	bl	trace_hardirqs_off
-#endif
-
 	@
 	@ call emulation code, which returns using r9 if it has emulated
 	@ the instruction, or the more conventional lr if we are to treat
@@ -317,11 +306,6 @@ ENDPROC(__und_svc)
 	.align	5
 __pabt_svc:
 	svc_entry
-
-#ifdef CONFIG_TRACE_IRQFLAGS
-	bl	trace_hardirqs_off
-#endif
-
 	pabt_helper
 	mov	r2, sp				@ regs
 	bl	do_PrefetchAbort		@ call abort handler
@@ -401,6 +385,10 @@ ENDPROC(__pabt_svc)
 	@ Clear FP to mark the first stack frame
 	@
 	zero_fp
+
+#ifdef CONFIG_IRQSOFF_TRACER
+	bl	trace_hardirqs_off
+#endif
 	.endm
 
 	.macro	kuser_cmpxchg_check
@@ -421,11 +409,6 @@ ENDPROC(__pabt_svc)
 	.align	5
 __dabt_usr:
 	usr_entry
-
-#ifdef CONFIG_IRQSOFF_TRACER
-	bl	trace_hardirqs_off
-#endif
-
 	kuser_cmpxchg_check
 	dabt_helper
 
@@ -438,11 +421,6 @@ ENDPROC(__dabt_usr)
 	.align	5
 __irq_usr:
 	usr_entry
-
-#ifdef CONFIG_IRQSOFF_TRACER
-	bl	trace_hardirqs_off
-#endif
-
 	kuser_cmpxchg_check
 	irq_handler
 	get_thread_info tsk
@@ -457,10 +435,6 @@ ENDPROC(__irq_usr)
 __und_usr:
 	usr_entry
 
-#ifdef CONFIG_IRQSOFF_TRACER
-	bl	trace_hardirqs_off
-#endif
-
 	mov	r2, r4
 	mov	r3, r5
 
@@ -679,11 +653,6 @@ ENDPROC(__und_usr_unknown)
 	.align	5
 __pabt_usr:
 	usr_entry
-
-#ifdef CONFIG_IRQSOFF_TRACER
-	bl	trace_hardirqs_off
-#endif
-
 	pabt_helper
 	mov	r2, sp				@ regs
 	bl	do_PrefetchAbort		@ call abort handler
-- 
1.7.4.4

[PATCH 15/23] ARM: entry: re-allocate registers in irq entry assembly macros

[Russell King - ARM Linux]

This avoids the irq entry assembly corrupting r5, thereby allowing it
to be preserved through to the svc exit code.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/include/asm/entry-macro-multi.S |   14 +++++++-------
 arch/arm/kernel/entry-armv.S             |   10 +++++-----
 2 files changed, 12 insertions(+), 12 deletions(-)

diff --git a/arch/arm/include/asm/entry-macro-multi.S b/arch/arm/include/asm/entry-macro-multi.S
index 2da8547..2f1e209 100644
--- a/arch/arm/include/asm/entry-macro-multi.S
+++ b/arch/arm/include/asm/entry-macro-multi.S
@@ -4,8 +4,8 @@
  * Interrupt handling.  Preserves r7, r8, r9
  */
 	.macro	arch_irq_handler_default
-	get_irqnr_preamble r5, lr
-1:	get_irqnr_and_base r0, r6, r5, lr
+	get_irqnr_preamble r6, lr
+1:	get_irqnr_and_base r0, r2, r6, lr
 	movne	r1, sp
 	@
 	@ routine called with r0 = irq number, r1 = struct pt_regs *
@@ -17,17 +17,17 @@
 	/*
 	 * XXX
 	 *
-	 * this macro assumes that irqstat (r6) and base (r5) are
+	 * this macro assumes that irqstat (r2) and base (r6) are
 	 * preserved from get_irqnr_and_base above
 	 */
-	ALT_SMP(test_for_ipi r0, r6, r5, lr)
+	ALT_SMP(test_for_ipi r0, r2, r6, lr)
 	ALT_UP_B(9997f)
 	movne	r1, sp
 	adrne	lr, BSYM(1b)
 	bne	do_IPI
 
 #ifdef CONFIG_LOCAL_TIMERS
-	test_for_ltirq r0, r6, r5, lr
+	test_for_ltirq r0, r2, r6, lr
 	movne	r0, sp
 	adrne	lr, BSYM(1b)
 	bne	do_local_timer
@@ -40,7 +40,7 @@
 	.align	5
 	.global \symbol_name
 \symbol_name:
-	mov	r4, lr
+	mov	r8, lr
 	arch_irq_handler_default
-	mov     pc, r4
+	mov     pc, r8
 	.endm
diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 1da985f..b7c4792 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -29,16 +29,16 @@
 #include <asm/entry-macro-multi.S>
 
 /*
- * Interrupt handling.  Preserves r7, r8, r9
+ * Interrupt handling.
  */
 	.macro	irq_handler
 #ifdef CONFIG_MULTI_IRQ_HANDLER
-	ldr	r5, =handle_arch_irq
+	ldr	r1, =handle_arch_irq
 	mov	r0, sp
-	ldr	r5, [r5]
+	ldr	r1, [r1]
 	adr	lr, BSYM(9997f)
-	teq	r5, #0
-	movne	pc, r5
+	teq	r1, #0
+	movne	pc, r1
 #endif
 	arch_irq_handler_default
 9997:
-- 
1.7.4.4

[PATCH 16/23] ARM: entry: prefetch abort: tail-call the main prefetch abort handler

[Russell King - ARM Linux]

Tail-call the main C prefetch abort handler code from the per-CPU
helper code.  Also note that the helper function becomes ABI
compliant in terms of the registers preserved.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |    8 +++-----
 arch/arm/mm/pabort-legacy.S  |    9 +++++----
 arch/arm/mm/pabort-v6.S      |    9 +++++----
 arch/arm/mm/pabort-v7.S      |   11 ++++++-----
 4 files changed, 19 insertions(+), 18 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index b7c4792..54c07a8 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -45,7 +45,7 @@
 	.endm
 
 	.macro	pabt_helper
-	@ PABORT handler takes fault address in r4
+	@ PABORT handler takes pt_regs in r2, fault address in r4 and psr in r5
 #ifdef MULTI_PABORT
 	ldr	ip, .LCprocfns
 	mov	lr, pc
@@ -306,9 +306,8 @@ ENDPROC(__und_svc)
 	.align	5
 __pabt_svc:
 	svc_entry
-	pabt_helper
 	mov	r2, sp				@ regs
-	bl	do_PrefetchAbort		@ call abort handler
+	pabt_helper
 
 	@
 	@ IRQs off again before pulling preserved data off the stack
@@ -653,9 +652,8 @@ ENDPROC(__und_usr_unknown)
 	.align	5
 __pabt_usr:
 	usr_entry
-	pabt_helper
 	mov	r2, sp				@ regs
-	bl	do_PrefetchAbort		@ call abort handler
+	pabt_helper
  UNWIND(.fnend		)
 	/* fall through */
 /*
diff --git a/arch/arm/mm/pabort-legacy.S b/arch/arm/mm/pabort-legacy.S
index 8a5d8aa..8bbff02 100644
--- a/arch/arm/mm/pabort-legacy.S
+++ b/arch/arm/mm/pabort-legacy.S
@@ -4,10 +4,11 @@
 /*
  * Function: legacy_pabort
  *
- * Params  : r4 = address of aborted instruction
+ * Params  : r2 = pt_regs
+ *	   : r4 = address of aborted instruction
+ *	   : r5 = psr for parent context
  *
- * Returns : r0 = address of abort
- *	   : r1 = Simulated IFSR with section translation fault status
+ * Returns : r4 - r11, r13 preserved
  *
  * Purpose : obtain information about current prefetch abort.
  */
@@ -16,5 +17,5 @@
 ENTRY(legacy_pabort)
 	mov	r0, r4
 	mov	r1, #5
-	mov	pc, lr
+	b	do_PrefetchAbort
 ENDPROC(legacy_pabort)
diff --git a/arch/arm/mm/pabort-v6.S b/arch/arm/mm/pabort-v6.S
index eaac1cb..9627646 100644
--- a/arch/arm/mm/pabort-v6.S
+++ b/arch/arm/mm/pabort-v6.S
@@ -4,10 +4,11 @@
 /*
  * Function: v6_pabort
  *
- * Params  : r4 = address of aborted instruction
+ * Params  : r2 = pt_regs
+ *	   : r4 = address of aborted instruction
+ *	   : r5 = psr for parent context
  *
- * Returns : r0 = address of abort
- *	   : r1 = IFSR
+ * Returns : r4 - r11, r13 preserved
  *
  * Purpose : obtain information about current prefetch abort.
  */
@@ -16,5 +17,5 @@
 ENTRY(v6_pabort)
 	mov	r0, r4
 	mrc	p15, 0, r1, c5, c0, 1		@ get IFSR
-	mov	pc, lr
+	b	do_PrefetchAbort
 ENDPROC(v6_pabort)
diff --git a/arch/arm/mm/pabort-v7.S b/arch/arm/mm/pabort-v7.S
index b515e0b..875761f 100644
--- a/arch/arm/mm/pabort-v7.S
+++ b/arch/arm/mm/pabort-v7.S
@@ -2,12 +2,13 @@
 #include <asm/assembler.h>
 
 /*
- * Function: v6_pabort
+ * Function: v7_pabort
  *
- * Params  : r4 = address of aborted instruction
+ * Params  : r2 = pt_regs
+ *	   : r4 = address of aborted instruction
+ *	   : r5 = psr for parent context
  *
- * Returns : r0 = address of abort
- *	   : r1 = IFSR
+ * Returns : r4 - r11, r13 preserved
  *
  * Purpose : obtain information about current prefetch abort.
  */
@@ -16,5 +17,5 @@
 ENTRY(v7_pabort)
 	mrc	p15, 0, r0, c6, c0, 2		@ get IFAR
 	mrc	p15, 0, r1, c5, c0, 1		@ get IFSR
-	mov	pc, lr
+	b	do_PrefetchAbort
 ENDPROC(v7_pabort)
-- 
1.7.4.4

[PATCH 17/23] ARM: entry: data abort: arrange for CPU abort helpers to take pc/psr in r4/r5

[Russell King - ARM Linux]

Re-jig the CPU abort helpers to take the PC/PSR in r4/r5 rather
than r2/r3.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |    6 ++----
 arch/arm/mm/abort-ev4.S      |    8 +++-----
 arch/arm/mm/abort-ev4t.S     |    8 ++++----
 arch/arm/mm/abort-ev5t.S     |    8 ++++----
 arch/arm/mm/abort-ev5tj.S    |   12 +++++-------
 arch/arm/mm/abort-ev6.S      |   12 +++++-------
 arch/arm/mm/abort-ev7.S      |    4 ++--
 arch/arm/mm/abort-lv4t.S     |   12 ++++++------
 arch/arm/mm/abort-nommu.S    |    4 ++--
 arch/arm/mm/proc-arm6_7.S    |   10 +++++-----
 10 files changed, 38 insertions(+), 46 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 54c07a8..36c9409 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -56,14 +56,12 @@
 	.endm
 
 	.macro	dabt_helper
-	mov	r2, r4
-	mov	r3, r5
 
 	@
 	@ Call the processor-specific abort handler:
 	@
-	@  r2 - aborted context pc
-	@  r3 - aborted context cpsr
+	@  r4 - aborted context pc
+	@  r5 - aborted context psr
 	@
 	@ The abort handler must return the aborted address in r0, and
 	@ the fault status register in r1.  r9 must be preserved.
diff --git a/arch/arm/mm/abort-ev4.S b/arch/arm/mm/abort-ev4.S
index 4f18f9e..beb112b 100644
--- a/arch/arm/mm/abort-ev4.S
+++ b/arch/arm/mm/abort-ev4.S
@@ -3,8 +3,8 @@
 /*
  * Function: v4_early_abort
  *
- * Params  : r2 = address of aborted instruction
- *         : r3 = saved SPSR
+ * Params  : r4 = aborted context pc
+ *	   : r5 = aborted context psr
  *
  * Returns : r0 = address of abort
  *	   : r1 = FSR, bit 11 = write
@@ -21,10 +21,8 @@
 ENTRY(v4_early_abort)
 	mrc	p15, 0, r1, c5, c0, 0		@ get FSR
 	mrc	p15, 0, r0, c6, c0, 0		@ get FAR
-	ldr	r3, [r2]			@ read aborted ARM instruction
+	ldr	r3, [r4]			@ read aborted ARM instruction
 	bic	r1, r1, #1 << 11 | 1 << 10	@ clear bits 11 and 10 of FSR
 	tst	r3, #1 << 20			@ L = 1 -> write?
 	orreq	r1, r1, #1 << 11		@ yes.
 	mov	pc, lr
-
-
diff --git a/arch/arm/mm/abort-ev4t.S b/arch/arm/mm/abort-ev4t.S
index 9910123..eaa4ac0 100644
--- a/arch/arm/mm/abort-ev4t.S
+++ b/arch/arm/mm/abort-ev4t.S
@@ -4,8 +4,8 @@
 /*
  * Function: v4t_early_abort
  *
- * Params  : r2 = address of aborted instruction
- *         : r3 = saved SPSR
+ * Params  : r4 = aborted context pc
+ *	   : r5 = aborted context psr
  *
  * Returns : r0 = address of abort
  *	   : r1 = FSR, bit 11 = write
@@ -22,8 +22,8 @@
 ENTRY(v4t_early_abort)
 	mrc	p15, 0, r1, c5, c0, 0		@ get FSR
 	mrc	p15, 0, r0, c6, c0, 0		@ get FAR
-	do_thumb_abort fsr=r1, pc=r2, psr=r3, tmp=r3
-	ldreq	r3, [r2]			@ read aborted ARM instruction
+	do_thumb_abort fsr=r1, pc=r4, psr=r5, tmp=r3
+	ldreq	r3, [r4]			@ read aborted ARM instruction
 	bic	r1, r1, #1 << 11 | 1 << 10	@ clear bits 11 and 10 of FSR
 	tst	r3, #1 << 20			@ check write
 	orreq	r1, r1, #1 << 11
diff --git a/arch/arm/mm/abort-ev5t.S b/arch/arm/mm/abort-ev5t.S
index 800e8d4..97eee7c 100644
--- a/arch/arm/mm/abort-ev5t.S
+++ b/arch/arm/mm/abort-ev5t.S
@@ -4,8 +4,8 @@
 /*
  * Function: v5t_early_abort
  *
- * Params  : r2 = address of aborted instruction
- *         : r3 = saved SPSR
+ * Params  : r4 = aborted context pc
+ *	   : r5 = aborted context psr
  *
  * Returns : r0 = address of abort
  *	   : r1 = FSR, bit 11 = write
@@ -22,8 +22,8 @@
 ENTRY(v5t_early_abort)
 	mrc	p15, 0, r1, c5, c0, 0		@ get FSR
 	mrc	p15, 0, r0, c6, c0, 0		@ get FAR
-	do_thumb_abort fsr=r1, pc=r2, psr=r3, tmp=r3
-	ldreq	r3, [r2]			@ read aborted ARM instruction
+	do_thumb_abort fsr=r1, pc=r4, psr=r5, tmp=r3
+	ldreq	r3, [r4]			@ read aborted ARM instruction
 	bic	r1, r1, #1 << 11		@ clear bits 11 of FSR
 	do_ldrd_abort tmp=r2, insn=r3
 	tst	r3, #1 << 20			@ check write
diff --git a/arch/arm/mm/abort-ev5tj.S b/arch/arm/mm/abort-ev5tj.S
index bcb58d2..9a365cf 100644
--- a/arch/arm/mm/abort-ev5tj.S
+++ b/arch/arm/mm/abort-ev5tj.S
@@ -4,8 +4,8 @@
 /*
  * Function: v5tj_early_abort
  *
- * Params  : r2 = address of aborted instruction
- *         : r3 = saved SPSR
+ * Params  : r4 = aborted context pc
+ *	   : r5 = aborted context psr
  *
  * Returns : r0 = address of abort
  *	   : r1 = FSR, bit 11 = write
@@ -23,13 +23,11 @@ ENTRY(v5tj_early_abort)
 	mrc	p15, 0, r1, c5, c0, 0		@ get FSR
 	mrc	p15, 0, r0, c6, c0, 0		@ get FAR
 	bic	r1, r1, #1 << 11 | 1 << 10	@ clear bits 11 and 10 of FSR
-	tst	r3, #PSR_J_BIT			@ Java?
+	tst	r5, #PSR_J_BIT			@ Java?
 	movne	pc, lr
-	do_thumb_abort fsr=r1, pc=r2, psr=r3, tmp=r3
-	ldreq	r3, [r2]			@ read aborted ARM instruction
+	do_thumb_abort fsr=r1, pc=r4, psr=r5, tmp=r3
+	ldreq	r3, [r4]			@ read aborted ARM instruction
 	do_ldrd_abort tmp=r2, insn=r3
 	tst	r3, #1 << 20			@ L = 0 -> write
 	orreq	r1, r1, #1 << 11		@ yes.
 	mov	pc, lr
-
-
diff --git a/arch/arm/mm/abort-ev6.S b/arch/arm/mm/abort-ev6.S
index ef526e7..52db4a3 100644
--- a/arch/arm/mm/abort-ev6.S
+++ b/arch/arm/mm/abort-ev6.S
@@ -4,8 +4,8 @@
 /*
  * Function: v6_early_abort
  *
- * Params  : r2 = address of aborted instruction
- *         : r3 = saved SPSR
+ * Params  : r4 = aborted context pc
+ *	   : r5 = aborted context psr
  *
  * Returns : r0 = address of abort
  *	   : r1 = FSR, bit 11 = write
@@ -33,10 +33,10 @@ ENTRY(v6_early_abort)
  * The test below covers all the write situations, including Java bytecodes
  */
 	bic	r1, r1, #1 << 11		@ clear bit 11 of FSR
-	tst	r3, #PSR_J_BIT			@ Java?
+	tst	r5, #PSR_J_BIT			@ Java?
 	movne	pc, lr
-	do_thumb_abort fsr=r1, pc=r2, psr=r3, tmp=r3
-	ldreq	r3, [r2]			@ read aborted ARM instruction
+	do_thumb_abort fsr=r1, pc=r4, psr=r5, tmp=r3
+	ldreq	r3, [r4]			@ read aborted ARM instruction
 #ifdef CONFIG_CPU_ENDIAN_BE8
 	reveq	r3, r3
 #endif
@@ -44,5 +44,3 @@ ENTRY(v6_early_abort)
 	tst	r3, #1 << 20			@ L = 0 -> write
 	orreq	r1, r1, #1 << 11		@ yes.
 	mov	pc, lr
-
-
diff --git a/arch/arm/mm/abort-ev7.S b/arch/arm/mm/abort-ev7.S
index ec88b15..6cb5143 100644
--- a/arch/arm/mm/abort-ev7.S
+++ b/arch/arm/mm/abort-ev7.S
@@ -3,8 +3,8 @@
 /*
  * Function: v7_early_abort
  *
- * Params  : r2 = address of aborted instruction
- *         : r3 = saved SPSR
+ * Params  : r4 = aborted context pc
+ *	   : r5 = aborted context psr
  *
  * Returns : r0 = address of abort
  *	   : r1 = FSR, bit 11 = write
diff --git a/arch/arm/mm/abort-lv4t.S b/arch/arm/mm/abort-lv4t.S
index 9fb7b0e..fea7514 100644
--- a/arch/arm/mm/abort-lv4t.S
+++ b/arch/arm/mm/abort-lv4t.S
@@ -3,8 +3,8 @@
 /*
  * Function: v4t_late_abort
  *
- * Params  : r2 = address of aborted instruction
- *         : r3 = saved SPSR
+ * Params  : r4 = aborted context pc
+ *	   : r5 = aborted context psr
  *
  * Returns : r0 = address of abort
  *	   : r1 = FSR, bit 11 = write
@@ -18,7 +18,7 @@
  * picture.  Unfortunately, this does happen.  We live with it.
  */
 ENTRY(v4t_late_abort)
-	tst	r3, #PSR_T_BIT			@ check for thumb mode
+	tst	r5, #PSR_T_BIT			@ check for thumb mode
 #ifdef CONFIG_CPU_CP15_MMU
 	mrc	p15, 0, r1, c5, c0, 0		@ get FSR
 	mrc	p15, 0, r0, c6, c0, 0		@ get FAR
@@ -28,7 +28,7 @@ ENTRY(v4t_late_abort)
 	mov	r1, #0
 #endif
 	bne	.data_thumb_abort
-	ldr	r8, [r2]			@ read arm instruction
+	ldr	r8, [r4]			@ read arm instruction
 	tst	r8, #1 << 20			@ L = 1 -> write?
 	orreq	r1, r1, #1 << 11		@ yes.
 	and	r7, r8, #15 << 24
@@ -52,7 +52,7 @@ ENTRY(v4t_late_abort)
 /* e */	b	.data_unknown
 /* f */
 .data_unknown:	@ Part of jumptable
-	mov	r0, r2
+	mov	r0, r4
 	mov	r1, r8
 	mov	r2, sp
 	bl	baddataabort
@@ -159,7 +159,7 @@ ENTRY(v4t_late_abort)
 	b	.data_unknown			@ F: MUL?
 
 .data_thumb_abort:
-	ldrh	r8, [r2]			@ read instruction
+	ldrh	r8, [r4]			@ read instruction
 	tst	r8, #1 << 11			@ L = 1 -> write?
 	orreq	r1, r1, #1 << 8			@ yes
 	and	r7, r8, #15 << 12
diff --git a/arch/arm/mm/abort-nommu.S b/arch/arm/mm/abort-nommu.S
index 625e580..9eaef6f 100644
--- a/arch/arm/mm/abort-nommu.S
+++ b/arch/arm/mm/abort-nommu.S
@@ -3,8 +3,8 @@
 /*
  * Function: nommu_early_abort
  *
- * Params  : r2 = address of aborted instruction
- *         : r3 = saved SPSR
+ * Params  : r4 = aborted context pc
+ *	   : r5 = aborted context psr
  *
  * Returns : r0 = 0 (abort address)
  *	   : r1 = 0 (FSR)
diff --git a/arch/arm/mm/proc-arm6_7.S b/arch/arm/mm/proc-arm6_7.S
index 5f79dc4..e7be700 100644
--- a/arch/arm/mm/proc-arm6_7.S
+++ b/arch/arm/mm/proc-arm6_7.S
@@ -29,8 +29,8 @@ ENTRY(cpu_arm7_dcache_clean_area)
 /*
  * Function: arm6_7_data_abort ()
  *
- * Params  : r2 = address of aborted instruction
- *	   : sp = pointer to registers
+ * Params  : r4 = aborted context pc
+ *	   : r5 = aborted context psr
  *
  * Purpose : obtain information about current aborted instruction
  *
@@ -41,7 +41,7 @@ ENTRY(cpu_arm7_dcache_clean_area)
 ENTRY(cpu_arm7_data_abort)
 	mrc	p15, 0, r1, c5, c0, 0		@ get FSR
 	mrc	p15, 0, r0, c6, c0, 0		@ get FAR
-	ldr	r8, [r2]			@ read arm instruction
+	ldr	r8, [r4]			@ read arm instruction
 	tst	r8, #1 << 20			@ L = 0 -> write?
 	orreq	r1, r1, #1 << 11		@ yes.
 	and	r7, r8, #15 << 24
@@ -65,7 +65,7 @@ ENTRY(cpu_arm7_data_abort)
 /* e */	b	.data_unknown
 /* f */
 .data_unknown:	@ Part of jumptable
-	mov	r0, r2
+	mov	r0, r4
 	mov	r1, r8
 	mov	r2, sp
 	bl	baddataabort
@@ -74,7 +74,7 @@ ENTRY(cpu_arm7_data_abort)
 ENTRY(cpu_arm6_data_abort)
 	mrc	p15, 0, r1, c5, c0, 0		@ get FSR
 	mrc	p15, 0, r0, c6, c0, 0		@ get FAR
-	ldr	r8, [r2]			@ read arm instruction
+	ldr	r8, [r4]			@ read arm instruction
 	tst	r8, #1 << 20			@ L = 0 -> write?
 	orreq	r1, r1, #1 << 11		@ yes.
 	and	r7, r8, #14 << 24
-- 
1.7.4.4

[PATCH 18/23] ARM: entry: data abort: avoid using r2 in abort helpers

[Russell King - ARM Linux]

This allows us to pass the pt_regs pointer in to these functions
ready for tail-calling the abort handler.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/mm/abort-ev5t.S  |    2 +-
 arch/arm/mm/abort-ev5tj.S |    2 +-
 arch/arm/mm/abort-ev6.S   |    2 +-
 arch/arm/mm/abort-ev7.S   |    8 ++++----
 arch/arm/mm/abort-lv4t.S  |   34 +++++++++++++++++-----------------
 arch/arm/mm/proc-arm6_7.S |   18 +++++++++---------
 6 files changed, 33 insertions(+), 33 deletions(-)

diff --git a/arch/arm/mm/abort-ev5t.S b/arch/arm/mm/abort-ev5t.S
index 97eee7c..751391a 100644
--- a/arch/arm/mm/abort-ev5t.S
+++ b/arch/arm/mm/abort-ev5t.S
@@ -25,7 +25,7 @@ ENTRY(v5t_early_abort)
 	do_thumb_abort fsr=r1, pc=r4, psr=r5, tmp=r3
 	ldreq	r3, [r4]			@ read aborted ARM instruction
 	bic	r1, r1, #1 << 11		@ clear bits 11 of FSR
-	do_ldrd_abort tmp=r2, insn=r3
+	do_ldrd_abort tmp=ip, insn=r3
 	tst	r3, #1 << 20			@ check write
 	orreq	r1, r1, #1 << 11
 	mov	pc, lr
diff --git a/arch/arm/mm/abort-ev5tj.S b/arch/arm/mm/abort-ev5tj.S
index 9a365cf..ccfbc93 100644
--- a/arch/arm/mm/abort-ev5tj.S
+++ b/arch/arm/mm/abort-ev5tj.S
@@ -27,7 +27,7 @@ ENTRY(v5tj_early_abort)
 	movne	pc, lr
 	do_thumb_abort fsr=r1, pc=r4, psr=r5, tmp=r3
 	ldreq	r3, [r4]			@ read aborted ARM instruction
-	do_ldrd_abort tmp=r2, insn=r3
+	do_ldrd_abort tmp=ip, insn=r3
 	tst	r3, #1 << 20			@ L = 0 -> write
 	orreq	r1, r1, #1 << 11		@ yes.
 	mov	pc, lr
diff --git a/arch/arm/mm/abort-ev6.S b/arch/arm/mm/abort-ev6.S
index 52db4a3..b64d886 100644
--- a/arch/arm/mm/abort-ev6.S
+++ b/arch/arm/mm/abort-ev6.S
@@ -40,7 +40,7 @@ ENTRY(v6_early_abort)
 #ifdef CONFIG_CPU_ENDIAN_BE8
 	reveq	r3, r3
 #endif
-	do_ldrd_abort tmp=r2, insn=r3
+	do_ldrd_abort tmp=ip, insn=r3
 	tst	r3, #1 << 20			@ L = 0 -> write
 	orreq	r1, r1, #1 << 11		@ yes.
 	mov	pc, lr
diff --git a/arch/arm/mm/abort-ev7.S b/arch/arm/mm/abort-ev7.S
index 6cb5143..6f98b3a 100644
--- a/arch/arm/mm/abort-ev7.S
+++ b/arch/arm/mm/abort-ev7.S
@@ -41,13 +41,13 @@ ENTRY(v7_early_abort)
 
 	mcr	p15, 0, r0, c7, c8, 0   	@ Retranslate FAR
 	isb
-	mrc	p15, 0, r2, c7, c4, 0   	@ Read the PAR
-	and	r3, r2, #0x7b   		@ On translation fault
+	mrc	p15, 0, ip, c7, c4, 0   	@ Read the PAR
+	and	r3, ip, #0x7b   		@ On translation fault
 	cmp	r3, #0x0b
 	movne	pc, lr
 	bic	r1, r1, #0xf			@ Fix up FSR FS[5:0]
-	and	r2, r2, #0x7e
-	orr	r1, r1, r2, LSR #1
+	and	ip, ip, #0x7e
+	orr	r1, r1, ip, LSR #1
 #endif
 
 	mov	pc, lr
diff --git a/arch/arm/mm/abort-lv4t.S b/arch/arm/mm/abort-lv4t.S
index fea7514..d032b1f 100644
--- a/arch/arm/mm/abort-lv4t.S
+++ b/arch/arm/mm/abort-lv4t.S
@@ -64,12 +64,12 @@ ENTRY(v4t_late_abort)
 	mov	r7, #0x11
 	orr	r7, r7, #0x1100
 	and	r6, r8, r7
-	and	r2, r8, r7, lsl #1
-	add	r6, r6, r2, lsr #1
-	and	r2, r8, r7, lsl #2
-	add	r6, r6, r2, lsr #2
-	and	r2, r8, r7, lsl #3
-	add	r6, r6, r2, lsr #3
+	and	r9, r8, r7, lsl #1
+	add	r6, r6, r9, lsr #1
+	and	r9, r8, r7, lsl #2
+	add	r6, r6, r9, lsr #2
+	and	r9, r8, r7, lsl #3
+	add	r6, r6, r9, lsr #3
 	add	r6, r6, r6, lsr #8
 	add	r6, r6, r6, lsr #4
 	and	r6, r6, #15			@ r6 = no. of registers to transfer.
@@ -103,13 +103,13 @@ ENTRY(v4t_late_abort)
 	tst	r8, #1 << 21			@ check writeback bit
 	moveq	pc, lr				@ no writeback -> no fixup
 .data_arm_lateldrpostconst:
-	movs	r2, r8, lsl #20			@ Get offset
+	movs	r9, r8, lsl #20			@ Get offset
 	moveq	pc, lr				@ zero -> no fixup
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
 	ldr	r7, [sp, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
-	subne	r7, r7, r2, lsr #20		@ Undo increment
-	addeq	r7, r7, r2, lsr #20		@ Undo decrement
+	subne	r7, r7, r9, lsr #20		@ Undo increment
+	addeq	r7, r7, r9, lsr #20		@ Undo decrement
 	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
 	mov	pc, lr
 
@@ -194,11 +194,11 @@ ENTRY(v4t_late_abort)
 	tst	r8, #1 << 10
 	beq	.data_unknown
 	and	r6, r8, #0x55			@ hweight8(r8) + R bit
-	and	r2, r8, #0xaa
-	add	r6, r6, r2, lsr #1
-	and	r2, r6, #0xcc
+	and	r9, r8, #0xaa
+	add	r6, r6, r9, lsr #1
+	and	r9, r6, #0xcc
 	and	r6, r6, #0x33
-	add	r6, r6, r2, lsr #2
+	add	r6, r6, r9, lsr #2
 	movs	r7, r8, lsr #9			@ C = r8 bit 8 (R bit)
 	adc	r6, r6, r6, lsr #4		@ high + low nibble + R bit
 	and	r6, r6, #15			@ number of regs to transfer
@@ -211,11 +211,11 @@ ENTRY(v4t_late_abort)
 
 .data_thumb_ldmstm:
 	and	r6, r8, #0x55			@ hweight8(r8)
-	and	r2, r8, #0xaa
-	add	r6, r6, r2, lsr #1
-	and	r2, r6, #0xcc
+	and	r9, r8, #0xaa
+	add	r6, r6, r9, lsr #1
+	and	r9, r6, #0xcc
 	and	r6, r6, #0x33
-	add	r6, r6, r2, lsr #2
+	add	r6, r6, r9, lsr #2
 	add	r6, r6, r6, lsr #4
 	and	r5, r8, #7 << 8
 	ldr	r7, [sp, r5, lsr #6]
diff --git a/arch/arm/mm/proc-arm6_7.S b/arch/arm/mm/proc-arm6_7.S
index e7be700..d4c328e 100644
--- a/arch/arm/mm/proc-arm6_7.S
+++ b/arch/arm/mm/proc-arm6_7.S
@@ -87,12 +87,12 @@ ENTRY(cpu_arm6_data_abort)
 	mov	r7, #0x11
 	orr	r7, r7, #0x1100
 	and	r6, r8, r7
-	and	r2, r8, r7, lsl #1
-	add	r6, r6, r2, lsr #1
-	and	r2, r8, r7, lsl #2
-	add	r6, r6, r2, lsr #2
-	and	r2, r8, r7, lsl #3
-	add	r6, r6, r2, lsr #3
+	and	r9, r8, r7, lsl #1
+	add	r6, r6, r9, lsr #1
+	and	r9, r8, r7, lsl #2
+	add	r6, r6, r9, lsr #2
+	and	r9, r8, r7, lsl #3
+	add	r6, r6, r9, lsr #3
 	add	r6, r6, r6, lsr #8
 	add	r6, r6, r6, lsr #4
 	and	r6, r6, #15			@ r6 = no. of registers to transfer.
@@ -117,13 +117,13 @@ ENTRY(cpu_arm6_data_abort)
 	tst	r8, #1 << 21			@ check writeback bit
 	moveq	pc, lr				@ no writeback -> no fixup
 .data_arm_lateldrpostconst:
-	movs	r2, r8, lsl #20			@ Get offset
+	movs	r9, r8, lsl #20			@ Get offset
 	moveq	pc, lr				@ zero -> no fixup
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
 	ldr	r7, [sp, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
-	subne	r7, r7, r2, lsr #20		@ Undo increment
-	addeq	r7, r7, r2, lsr #20		@ Undo decrement
+	subne	r7, r7, r9, lsr #20		@ Undo increment
+	addeq	r7, r7, r9, lsr #20		@ Undo decrement
 	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
 	mov	pc, lr
 
-- 
1.7.4.4

[PATCH 19/23] ARM: entry: data abort: tail-call the main data abort handler

[Russell King - ARM Linux]

Tail-call the main C data abort handler code from the per-CPU helper
code.  Update the comments in the code wrt the new calling and return
register state.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |   14 +++---------
 arch/arm/mm/abort-ev4.S      |   11 +++------
 arch/arm/mm/abort-ev4t.S     |   11 +++------
 arch/arm/mm/abort-ev5t.S     |   11 +++------
 arch/arm/mm/abort-ev5tj.S    |   13 ++++-------
 arch/arm/mm/abort-ev6.S      |   13 ++++-------
 arch/arm/mm/abort-ev7.S      |   15 +++++--------
 arch/arm/mm/abort-lv4t.S     |   43 ++++++++++++++++++++---------------------
 arch/arm/mm/abort-macro.S    |    4 +-
 arch/arm/mm/abort-nommu.S    |    8 +++---
 arch/arm/mm/proc-arm6_7.S    |   29 +++++++++++++--------------
 11 files changed, 73 insertions(+), 99 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 36c9409..3fa90ae 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -60,6 +60,7 @@
 	@
 	@ Call the processor-specific abort handler:
 	@
+	@  r2 - pt_regs
 	@  r4 - aborted context pc
 	@  r5 - aborted context psr
 	@
@@ -186,13 +187,8 @@ ENDPROC(__und_invalid)
 	.align	5
 __dabt_svc:
 	svc_entry
-	dabt_helper
-
-	@
-	@ call main handler
-	@
 	mov	r2, sp
-	bl	do_DataAbort
+	dabt_helper
 
 	@
 	@ IRQs off again before pulling preserved data off the stack
@@ -407,11 +403,9 @@ ENDPROC(__pabt_svc)
 __dabt_usr:
 	usr_entry
 	kuser_cmpxchg_check
-	dabt_helper
-
 	mov	r2, sp
-	adr	lr, BSYM(ret_from_exception)
-	b	do_DataAbort
+	dabt_helper
+	b	ret_from_exception
  UNWIND(.fnend		)
 ENDPROC(__dabt_usr)
 
diff --git a/arch/arm/mm/abort-ev4.S b/arch/arm/mm/abort-ev4.S
index beb112b..54473cd 100644
--- a/arch/arm/mm/abort-ev4.S
+++ b/arch/arm/mm/abort-ev4.S
@@ -3,14 +3,11 @@
 /*
  * Function: v4_early_abort
  *
- * Params  : r4 = aborted context pc
+ * Params  : r2 = pt_regs
+ *	   : r4 = aborted context pc
  *	   : r5 = aborted context psr
  *
- * Returns : r0 = address of abort
- *	   : r1 = FSR, bit 11 = write
- *	   : r2-r8 = corrupted
- *	   : r9 = preserved
- *	   : sp = pointer to registers
+ * Returns : r4 - r11, r13 preserved
  *
  * Purpose : obtain information about current aborted instruction.
  * Note: we read user space.  This means we might cause a data
@@ -25,4 +22,4 @@ ENTRY(v4_early_abort)
 	bic	r1, r1, #1 << 11 | 1 << 10	@ clear bits 11 and 10 of FSR
 	tst	r3, #1 << 20			@ L = 1 -> write?
 	orreq	r1, r1, #1 << 11		@ yes.
-	mov	pc, lr
+	b	do_DataAbort
diff --git a/arch/arm/mm/abort-ev4t.S b/arch/arm/mm/abort-ev4t.S
index eaa4ac0..9da704e 100644
--- a/arch/arm/mm/abort-ev4t.S
+++ b/arch/arm/mm/abort-ev4t.S
@@ -4,14 +4,11 @@
 /*
  * Function: v4t_early_abort
  *
- * Params  : r4 = aborted context pc
+ * Params  : r2 = pt_regs
+ *	   : r4 = aborted context pc
  *	   : r5 = aborted context psr
  *
- * Returns : r0 = address of abort
- *	   : r1 = FSR, bit 11 = write
- *	   : r2-r8 = corrupted
- *	   : r9 = preserved
- *	   : sp = pointer to registers
+ * Returns : r4 - r11, r13 preserved
  *
  * Purpose : obtain information about current aborted instruction.
  * Note: we read user space.  This means we might cause a data
@@ -27,4 +24,4 @@ ENTRY(v4t_early_abort)
 	bic	r1, r1, #1 << 11 | 1 << 10	@ clear bits 11 and 10 of FSR
 	tst	r3, #1 << 20			@ check write
 	orreq	r1, r1, #1 << 11
-	mov	pc, lr
+	b	do_DataAbort
diff --git a/arch/arm/mm/abort-ev5t.S b/arch/arm/mm/abort-ev5t.S
index 751391a..a0908d4 100644
--- a/arch/arm/mm/abort-ev5t.S
+++ b/arch/arm/mm/abort-ev5t.S
@@ -4,14 +4,11 @@
 /*
  * Function: v5t_early_abort
  *
- * Params  : r4 = aborted context pc
+ * Params  : r2 = pt_regs
+ *	   : r4 = aborted context pc
  *	   : r5 = aborted context psr
  *
- * Returns : r0 = address of abort
- *	   : r1 = FSR, bit 11 = write
- *	   : r2-r8 = corrupted
- *	   : r9 = preserved
- *	   : sp = pointer to registers
+ * Returns : r4 - r11, r13 preserved
  *
  * Purpose : obtain information about current aborted instruction.
  * Note: we read user space.  This means we might cause a data
@@ -28,4 +25,4 @@ ENTRY(v5t_early_abort)
 	do_ldrd_abort tmp=ip, insn=r3
 	tst	r3, #1 << 20			@ check write
 	orreq	r1, r1, #1 << 11
-	mov	pc, lr
+	b	do_DataAbort
diff --git a/arch/arm/mm/abort-ev5tj.S b/arch/arm/mm/abort-ev5tj.S
index ccfbc93..4006b7a 100644
--- a/arch/arm/mm/abort-ev5tj.S
+++ b/arch/arm/mm/abort-ev5tj.S
@@ -4,14 +4,11 @@
 /*
  * Function: v5tj_early_abort
  *
- * Params  : r4 = aborted context pc
+ * Params  : r2 = pt_regs
+ *	   : r4 = aborted context pc
  *	   : r5 = aborted context psr
  *
- * Returns : r0 = address of abort
- *	   : r1 = FSR, bit 11 = write
- *	   : r2-r8 = corrupted
- *	   : r9 = preserved
- *	   : sp = pointer to registers
+ * Returns : r4 - r11, r13 preserved
  *
  * Purpose : obtain information about current aborted instruction.
  * Note: we read user space.  This means we might cause a data
@@ -24,10 +21,10 @@ ENTRY(v5tj_early_abort)
 	mrc	p15, 0, r0, c6, c0, 0		@ get FAR
 	bic	r1, r1, #1 << 11 | 1 << 10	@ clear bits 11 and 10 of FSR
 	tst	r5, #PSR_J_BIT			@ Java?
-	movne	pc, lr
+	bne	do_DataAbort
 	do_thumb_abort fsr=r1, pc=r4, psr=r5, tmp=r3
 	ldreq	r3, [r4]			@ read aborted ARM instruction
 	do_ldrd_abort tmp=ip, insn=r3
 	tst	r3, #1 << 20			@ L = 0 -> write
 	orreq	r1, r1, #1 << 11		@ yes.
-	mov	pc, lr
+	b	do_DataAbort
diff --git a/arch/arm/mm/abort-ev6.S b/arch/arm/mm/abort-ev6.S
index b64d886..ff1f7cc 100644
--- a/arch/arm/mm/abort-ev6.S
+++ b/arch/arm/mm/abort-ev6.S
@@ -4,14 +4,11 @@
 /*
  * Function: v6_early_abort
  *
- * Params  : r4 = aborted context pc
+ * Params  : r2 = pt_regs
+ *	   : r4 = aborted context pc
  *	   : r5 = aborted context psr
  *
- * Returns : r0 = address of abort
- *	   : r1 = FSR, bit 11 = write
- *	   : r2-r8 = corrupted
- *	   : r9 = preserved
- *	   : sp = pointer to registers
+ * Returns : r4 - r11, r13 preserved
  *
  * Purpose : obtain information about current aborted instruction.
  * Note: we read user space.  This means we might cause a data
@@ -34,7 +31,7 @@ ENTRY(v6_early_abort)
  */
 	bic	r1, r1, #1 << 11		@ clear bit 11 of FSR
 	tst	r5, #PSR_J_BIT			@ Java?
-	movne	pc, lr
+	bne	do_DataAbort
 	do_thumb_abort fsr=r1, pc=r4, psr=r5, tmp=r3
 	ldreq	r3, [r4]			@ read aborted ARM instruction
 #ifdef CONFIG_CPU_ENDIAN_BE8
@@ -43,4 +40,4 @@ ENTRY(v6_early_abort)
 	do_ldrd_abort tmp=ip, insn=r3
 	tst	r3, #1 << 20			@ L = 0 -> write
 	orreq	r1, r1, #1 << 11		@ yes.
-	mov	pc, lr
+	b	do_DataAbort
diff --git a/arch/arm/mm/abort-ev7.S b/arch/arm/mm/abort-ev7.S
index 6f98b3a..7033752 100644
--- a/arch/arm/mm/abort-ev7.S
+++ b/arch/arm/mm/abort-ev7.S
@@ -3,14 +3,11 @@
 /*
  * Function: v7_early_abort
  *
- * Params  : r4 = aborted context pc
+ * Params  : r2 = pt_regs
+ *	   : r4 = aborted context pc
  *	   : r5 = aborted context psr
  *
- * Returns : r0 = address of abort
- *	   : r1 = FSR, bit 11 = write
- *	   : r2-r8 = corrupted
- *	   : r9 = preserved
- *	   : sp = pointer to registers
+ * Returns : r4 - r11, r13 preserved
  *
  * Purpose : obtain information about current aborted instruction.
  */
@@ -37,18 +34,18 @@ ENTRY(v7_early_abort)
 	ldr	r3, =0x40d			@ On permission fault
 	and	r3, r1, r3
 	cmp	r3, #0x0d
-	movne	pc, lr
+	bne	do_DataAbort
 
 	mcr	p15, 0, r0, c7, c8, 0   	@ Retranslate FAR
 	isb
 	mrc	p15, 0, ip, c7, c4, 0   	@ Read the PAR
 	and	r3, ip, #0x7b   		@ On translation fault
 	cmp	r3, #0x0b
-	movne	pc, lr
+	bne	do_DataAbort
 	bic	r1, r1, #0xf			@ Fix up FSR FS[5:0]
 	and	ip, ip, #0x7e
 	orr	r1, r1, ip, LSR #1
 #endif
 
-	mov	pc, lr
+	b	do_DataAbort
 ENDPROC(v7_early_abort)
diff --git a/arch/arm/mm/abort-lv4t.S b/arch/arm/mm/abort-lv4t.S
index d032b1f..d432f31 100644
--- a/arch/arm/mm/abort-lv4t.S
+++ b/arch/arm/mm/abort-lv4t.S
@@ -3,7 +3,8 @@
 /*
  * Function: v4t_late_abort
  *
- * Params  : r4 = aborted context pc
+ * Params  : r2 = pt_regs
+ *	   : r4 = aborted context pc
  *	   : r5 = aborted context psr
  *
  * Returns : r0 = address of abort
@@ -47,20 +48,18 @@ ENTRY(v4t_late_abort)
 /* 9 */	b	.data_arm_ldmstm		@ ldm*b	rn, <rlist>
 /* a */	b	.data_unknown
 /* b */	b	.data_unknown
-/* c */	mov	pc, lr				@ ldc	rd, [rn], #m	@ Same as ldr	rd, [rn], #m
-/* d */	mov	pc, lr				@ ldc	rd, [rn, #m]
+/* c */	b	do_DataAbort			@ ldc	rd, [rn], #m	@ Same as ldr	rd, [rn], #m
+/* d */	b	do_DataAbort			@ ldc	rd, [rn, #m]
 /* e */	b	.data_unknown
 /* f */
 .data_unknown:	@ Part of jumptable
 	mov	r0, r4
 	mov	r1, r8
-	mov	r2, sp
-	bl	baddataabort
-	b	ret_from_exception
+	b	baddataabort
 
 .data_arm_ldmstm:
 	tst	r8, #1 << 21			@ check writeback bit
-	moveq	pc, lr				@ no writeback -> no fixup
+	beq	do_DataAbort			@ no writeback -> no fixup
 	mov	r7, #0x11
 	orr	r7, r7, #0x1100
 	and	r6, r8, r7
@@ -79,11 +78,11 @@ ENTRY(v4t_late_abort)
 	subne	r7, r7, r6, lsl #2		@ Undo increment
 	addeq	r7, r7, r6, lsl #2		@ Undo decrement
 	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
-	mov	pc, lr
+	b	do_DataAbort
 
 .data_arm_lateldrhpre:
 	tst	r8, #1 << 21			@ Check writeback bit
-	moveq	pc, lr				@ No writeback -> no fixup
+	beq	do_DataAbort			@ No writeback -> no fixup
 .data_arm_lateldrhpost:
 	and	r5, r8, #0x00f			@ get Rm / low nibble of immediate value
 	tst	r8, #1 << 22			@ if (immediate offset)
@@ -97,25 +96,25 @@ ENTRY(v4t_late_abort)
 	subne	r7, r7, r6			@ Undo incrmenet
 	addeq	r7, r7, r6			@ Undo decrement
 	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
-	mov	pc, lr
+	b	do_DataAbort
 
 .data_arm_lateldrpreconst:
 	tst	r8, #1 << 21			@ check writeback bit
-	moveq	pc, lr				@ no writeback -> no fixup
+	beq	do_DataAbort			@ no writeback -> no fixup
 .data_arm_lateldrpostconst:
 	movs	r9, r8, lsl #20			@ Get offset
-	moveq	pc, lr				@ zero -> no fixup
+	beq	do_DataAbort			@ zero -> no fixup
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
 	ldr	r7, [sp, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r9, lsr #20		@ Undo increment
 	addeq	r7, r7, r9, lsr #20		@ Undo decrement
 	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
-	mov	pc, lr
+	b	do_DataAbort
 
 .data_arm_lateldrprereg:
 	tst	r8, #1 << 21			@ check writeback bit
-	moveq	pc, lr				@ no writeback -> no fixup
+	beq	do_DataAbort			@ no writeback -> no fixup
 .data_arm_lateldrpostreg:
 	and	r7, r8, #15			@ Extract 'm' from instruction
 	ldr	r6, [sp, r7, lsl #2]		@ Get register 'Rm'
@@ -172,10 +171,10 @@ ENTRY(v4t_late_abort)
 /* 3 */	b	.data_unknown
 /* 4 */	b	.data_unknown
 /* 5 */	b	.data_thumb_reg
-/* 6 */	mov	pc, lr
-/* 7 */	mov	pc, lr
-/* 8 */	mov	pc, lr
-/* 9 */	mov	pc, lr
+/* 6 */	b	do_DataAbort
+/* 7 */	b	do_DataAbort
+/* 8 */	b	do_DataAbort
+/* 9 */	b	do_DataAbort
 /* A */	b	.data_unknown
 /* B */	b	.data_thumb_pushpop
 /* C */	b	.data_thumb_ldmstm
@@ -185,10 +184,10 @@ ENTRY(v4t_late_abort)
 
 .data_thumb_reg:
 	tst	r8, #1 << 9
-	moveq	pc, lr
+	beq	do_DataAbort
 	tst	r8, #1 << 10			@ If 'S' (signed) bit is set
 	movne	r1, #0				@ it must be a load instr
-	mov	pc, lr
+	b	do_DataAbort
 
 .data_thumb_pushpop:
 	tst	r8, #1 << 10
@@ -207,7 +206,7 @@ ENTRY(v4t_late_abort)
 	addeq	r7, r7, r6, lsl #2		@ increment SP if PUSH
 	subne	r7, r7, r6, lsl #2		@ decrement SP if POP
 	str	r7, [sp, #13 << 2]
-	mov	pc, lr
+	b	do_DataAbort
 
 .data_thumb_ldmstm:
 	and	r6, r8, #0x55			@ hweight8(r8)
@@ -222,4 +221,4 @@ ENTRY(v4t_late_abort)
 	and	r6, r6, #15			@ number of regs to transfer
 	sub	r7, r7, r6, lsl #2		@ always decrement
 	str	r7, [sp, r5, lsr #6]
-	mov	pc, lr
+	b	do_DataAbort
diff --git a/arch/arm/mm/abort-macro.S b/arch/arm/mm/abort-macro.S
index af97a10..52162d5 100644
--- a/arch/arm/mm/abort-macro.S
+++ b/arch/arm/mm/abort-macro.S
@@ -18,7 +18,7 @@
 	orreq	\tmp, \tmp, #1 << 11		@ Set L-bit if yes
 	tst	\tmp, #1 << 11			@ L = 0 -> write
 	orreq	\psr, \psr, #1 << 11		@ yes.
-	mov	pc, lr
+	b	do_DataAbort
 not_thumb:
 	.endm
 
@@ -34,7 +34,7 @@ not_thumb:
 	bne	not_ldrd
 	and	\tmp, \insn, #0x000000f0	@ [7:4] == 1101
 	cmp	\tmp, #0x000000d0
-	moveq	pc, lr
+	beq	do_DataAbort
 not_ldrd:
 	.endm
 
diff --git a/arch/arm/mm/abort-nommu.S b/arch/arm/mm/abort-nommu.S
index 9eaef6f..119cb47 100644
--- a/arch/arm/mm/abort-nommu.S
+++ b/arch/arm/mm/abort-nommu.S
@@ -3,11 +3,11 @@
 /*
  * Function: nommu_early_abort
  *
- * Params  : r4 = aborted context pc
+ * Params  : r2 = pt_regs
+ *	   : r4 = aborted context pc
  *	   : r5 = aborted context psr
  *
- * Returns : r0 = 0 (abort address)
- *	   : r1 = 0 (FSR)
+ * Returns : r4 - r11, r13 preserved
  *
  * Note: There is no FSR/FAR on !CPU_CP15_MMU cores.
  *       Just fill zero into the registers.
@@ -16,5 +16,5 @@
 ENTRY(nommu_early_abort)
 	mov	r0, #0				@ clear r0, r1 (no FSR/FAR)
 	mov	r1, #0
-	mov	pc, lr
+	b	do_DataAbort
 ENDPROC(nommu_early_abort)
diff --git a/arch/arm/mm/proc-arm6_7.S b/arch/arm/mm/proc-arm6_7.S
index d4c328e..d755d5b 100644
--- a/arch/arm/mm/proc-arm6_7.S
+++ b/arch/arm/mm/proc-arm6_7.S
@@ -29,7 +29,8 @@ ENTRY(cpu_arm7_dcache_clean_area)
 /*
  * Function: arm6_7_data_abort ()
  *
- * Params  : r4 = aborted context pc
+ * Params  : r2 = pt_regs
+ *	   : r4 = aborted context pc
  *	   : r5 = aborted context psr
  *
  * Purpose : obtain information about current aborted instruction
@@ -49,7 +50,7 @@ ENTRY(cpu_arm7_data_abort)
 	nop
 
 /* 0 */	b	.data_unknown
-/* 1 */	mov	pc, lr				@ swp
+/* 1 */	b	do_DataAbort			@ swp
 /* 2 */	b	.data_unknown
 /* 3 */	b	.data_unknown
 /* 4 */	b	.data_arm_lateldrpostconst	@ ldr	rd, [rn], #m
@@ -60,16 +61,14 @@ ENTRY(cpu_arm7_data_abort)
 /* 9 */	b	.data_arm_ldmstm		@ ldm*b	rn, <rlist>
 /* a */	b	.data_unknown
 /* b */	b	.data_unknown
-/* c */	mov	pc, lr				@ ldc	rd, [rn], #m	@ Same as ldr	rd, [rn], #m
-/* d */	mov	pc, lr				@ ldc	rd, [rn, #m]
+/* c */	b	do_DataAbort			@ ldc	rd, [rn], #m	@ Same as ldr	rd, [rn], #m
+/* d */	b	do_DataAbort			@ ldc	rd, [rn, #m]
 /* e */	b	.data_unknown
 /* f */
 .data_unknown:	@ Part of jumptable
 	mov	r0, r4
 	mov	r1, r8
-	mov	r2, sp
-	bl	baddataabort
-	b	ret_from_exception
+	b	baddataabort
 
 ENTRY(cpu_arm6_data_abort)
 	mrc	p15, 0, r1, c5, c0, 0		@ get FSR
@@ -79,11 +78,11 @@ ENTRY(cpu_arm6_data_abort)
 	orreq	r1, r1, #1 << 11		@ yes.
 	and	r7, r8, #14 << 24
 	teq	r7, #8 << 24			@ was it ldm/stm
-	movne	pc, lr
+	bne	do_DataAbort
 
 .data_arm_ldmstm:
 	tst	r8, #1 << 21			@ check writeback bit
-	moveq	pc, lr				@ no writeback -> no fixup
+	beq	do_DataAbort			@ no writeback -> no fixup
 	mov	r7, #0x11
 	orr	r7, r7, #0x1100
 	and	r6, r8, r7
@@ -102,7 +101,7 @@ ENTRY(cpu_arm6_data_abort)
 	subne	r7, r7, r6, lsl #2		@ Undo increment
 	addeq	r7, r7, r6, lsl #2		@ Undo decrement
 	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
-	mov	pc, lr
+	b	do_DataAbort
 
 .data_arm_apply_r6_and_rn:
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
@@ -111,25 +110,25 @@ ENTRY(cpu_arm6_data_abort)
 	subne	r7, r7, r6			@ Undo incrmenet
 	addeq	r7, r7, r6			@ Undo decrement
 	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
-	mov	pc, lr
+	b	do_DataAbort
 
 .data_arm_lateldrpreconst:
 	tst	r8, #1 << 21			@ check writeback bit
-	moveq	pc, lr				@ no writeback -> no fixup
+	beq	do_DataAbort			@ no writeback -> no fixup
 .data_arm_lateldrpostconst:
 	movs	r9, r8, lsl #20			@ Get offset
-	moveq	pc, lr				@ zero -> no fixup
+	beq	do_DataAbort			@ zero -> no fixup
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
 	ldr	r7, [sp, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r9, lsr #20		@ Undo increment
 	addeq	r7, r7, r9, lsr #20		@ Undo decrement
 	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
-	mov	pc, lr
+	b	do_DataAbort
 
 .data_arm_lateldrprereg:
 	tst	r8, #1 << 21			@ check writeback bit
-	moveq	pc, lr				@ no writeback -> no fixup
+	beq	do_DataAbort			@ no writeback -> no fixup
 .data_arm_lateldrpostreg:
 	and	r7, r8, #15			@ Extract 'm' from instruction
 	ldr	r6, [sp, r7, lsl #2]		@ Get register 'Rm'
-- 
1.7.4.4

[PATCH 20/23] ARM: entry: data abort: use r2 as base of pt_regs rather than stack

[Russell King - ARM Linux]

Now that we pass r2 into these helper functions as the pointer to
pt_regs, use r2 as the base of the registers on the stack rather
than using the stack pointer directly.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/mm/abort-lv4t.S  |   24 ++++++++++++------------
 arch/arm/mm/proc-arm6_7.S |   14 +++++++-------
 2 files changed, 19 insertions(+), 19 deletions(-)

diff --git a/arch/arm/mm/abort-lv4t.S b/arch/arm/mm/abort-lv4t.S
index d432f31..921aaab 100644
--- a/arch/arm/mm/abort-lv4t.S
+++ b/arch/arm/mm/abort-lv4t.S
@@ -73,11 +73,11 @@ ENTRY(v4t_late_abort)
 	add	r6, r6, r6, lsr #4
 	and	r6, r6, #15			@ r6 = no. of registers to transfer.
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [sp, r5, lsr #14]		@ Get register 'Rn'
+	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r6, lsl #2		@ Undo increment
 	addeq	r7, r7, r6, lsl #2		@ Undo decrement
-	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_lateldrhpre:
@@ -88,14 +88,14 @@ ENTRY(v4t_late_abort)
 	tst	r8, #1 << 22			@ if (immediate offset)
 	andne	r6, r8, #0xf00			@ { immediate high nibble
 	orrne	r6, r5, r6, lsr #4		@   combine nibbles } else
-	ldreq	r6, [sp, r5, lsl #2]		@ { load Rm value }
+	ldreq	r6, [r2, r5, lsl #2]		@ { load Rm value }
 .data_arm_apply_r6_and_rn:
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [sp, r5, lsr #14]		@ Get register 'Rn'
+	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r6			@ Undo incrmenet
 	addeq	r7, r7, r6			@ Undo decrement
-	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_lateldrpreconst:
@@ -105,11 +105,11 @@ ENTRY(v4t_late_abort)
 	movs	r9, r8, lsl #20			@ Get offset
 	beq	do_DataAbort			@ zero -> no fixup
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [sp, r5, lsr #14]		@ Get register 'Rn'
+	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r9, lsr #20		@ Undo increment
 	addeq	r7, r7, r9, lsr #20		@ Undo decrement
-	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_lateldrprereg:
@@ -117,7 +117,7 @@ ENTRY(v4t_late_abort)
 	beq	do_DataAbort			@ no writeback -> no fixup
 .data_arm_lateldrpostreg:
 	and	r7, r8, #15			@ Extract 'm' from instruction
-	ldr	r6, [sp, r7, lsl #2]		@ Get register 'Rm'
+	ldr	r6, [r2, r7, lsl #2]		@ Get register 'Rm'
 	mov	r5, r8, lsr #7			@ get shift count
 	ands	r5, r5, #31
 	and	r7, r8, #0x70			@ get shift type
@@ -201,11 +201,11 @@ ENTRY(v4t_late_abort)
 	movs	r7, r8, lsr #9			@ C = r8 bit 8 (R bit)
 	adc	r6, r6, r6, lsr #4		@ high + low nibble + R bit
 	and	r6, r6, #15			@ number of regs to transfer
-	ldr	r7, [sp, #13 << 2]
+	ldr	r7, [r2, #13 << 2]
 	tst	r8, #1 << 11
 	addeq	r7, r7, r6, lsl #2		@ increment SP if PUSH
 	subne	r7, r7, r6, lsl #2		@ decrement SP if POP
-	str	r7, [sp, #13 << 2]
+	str	r7, [r2, #13 << 2]
 	b	do_DataAbort
 
 .data_thumb_ldmstm:
@@ -217,8 +217,8 @@ ENTRY(v4t_late_abort)
 	add	r6, r6, r9, lsr #2
 	add	r6, r6, r6, lsr #4
 	and	r5, r8, #7 << 8
-	ldr	r7, [sp, r5, lsr #6]
+	ldr	r7, [r2, r5, lsr #6]
 	and	r6, r6, #15			@ number of regs to transfer
 	sub	r7, r7, r6, lsl #2		@ always decrement
-	str	r7, [sp, r5, lsr #6]
+	str	r7, [r2, r5, lsr #6]
 	b	do_DataAbort
diff --git a/arch/arm/mm/proc-arm6_7.S b/arch/arm/mm/proc-arm6_7.S
index d755d5b..141906e 100644
--- a/arch/arm/mm/proc-arm6_7.S
+++ b/arch/arm/mm/proc-arm6_7.S
@@ -96,20 +96,20 @@ ENTRY(cpu_arm6_data_abort)
 	add	r6, r6, r6, lsr #4
 	and	r6, r6, #15			@ r6 = no. of registers to transfer.
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [sp, r5, lsr #14]		@ Get register 'Rn'
+	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r6, lsl #2		@ Undo increment
 	addeq	r7, r7, r6, lsl #2		@ Undo decrement
-	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_apply_r6_and_rn:
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [sp, r5, lsr #14]		@ Get register 'Rn'
+	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r6			@ Undo incrmenet
 	addeq	r7, r7, r6			@ Undo decrement
-	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_lateldrpreconst:
@@ -119,11 +119,11 @@ ENTRY(cpu_arm6_data_abort)
 	movs	r9, r8, lsl #20			@ Get offset
 	beq	do_DataAbort			@ zero -> no fixup
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [sp, r5, lsr #14]		@ Get register 'Rn'
+	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r9, lsr #20		@ Undo increment
 	addeq	r7, r7, r9, lsr #20		@ Undo decrement
-	str	r7, [sp, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_lateldrprereg:
@@ -131,7 +131,7 @@ ENTRY(cpu_arm6_data_abort)
 	beq	do_DataAbort			@ no writeback -> no fixup
 .data_arm_lateldrpostreg:
 	and	r7, r8, #15			@ Extract 'm' from instruction
-	ldr	r6, [sp, r7, lsl #2]		@ Get register 'Rm'
+	ldr	r6, [r2, r7, lsl #2]		@ Get register 'Rm'
 	mov	r5, r8, lsr #7			@ get shift count
 	ands	r5, r5, #31
 	and	r7, r8, #0x70			@ get shift type
-- 
1.7.4.4

[PATCH 21/23] ARM: entry: data abort: always use r6 for offset

[Russell King - ARM Linux]

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/mm/abort-lv4t.S  |    6 +++---
 arch/arm/mm/proc-arm6_7.S |    6 +++---
 2 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/arch/arm/mm/abort-lv4t.S b/arch/arm/mm/abort-lv4t.S
index 921aaab..54b6d27 100644
--- a/arch/arm/mm/abort-lv4t.S
+++ b/arch/arm/mm/abort-lv4t.S
@@ -102,13 +102,13 @@ ENTRY(v4t_late_abort)
 	tst	r8, #1 << 21			@ check writeback bit
 	beq	do_DataAbort			@ no writeback -> no fixup
 .data_arm_lateldrpostconst:
-	movs	r9, r8, lsl #20			@ Get offset
+	movs	r6, r8, lsl #20			@ Get offset
 	beq	do_DataAbort			@ zero -> no fixup
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
 	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
-	subne	r7, r7, r9, lsr #20		@ Undo increment
-	addeq	r7, r7, r9, lsr #20		@ Undo decrement
+	subne	r7, r7, r6, lsr #20		@ Undo increment
+	addeq	r7, r7, r6, lsr #20		@ Undo decrement
 	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
diff --git a/arch/arm/mm/proc-arm6_7.S b/arch/arm/mm/proc-arm6_7.S
index 141906e..4d96311 100644
--- a/arch/arm/mm/proc-arm6_7.S
+++ b/arch/arm/mm/proc-arm6_7.S
@@ -116,13 +116,13 @@ ENTRY(cpu_arm6_data_abort)
 	tst	r8, #1 << 21			@ check writeback bit
 	beq	do_DataAbort			@ no writeback -> no fixup
 .data_arm_lateldrpostconst:
-	movs	r9, r8, lsl #20			@ Get offset
+	movs	r6, r8, lsl #20			@ Get offset
 	beq	do_DataAbort			@ zero -> no fixup
 	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
 	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
-	subne	r7, r7, r9, lsr #20		@ Undo increment
-	addeq	r7, r7, r9, lsr #20		@ Undo decrement
+	subne	r7, r7, r6, lsr #20		@ Undo increment
+	addeq	r7, r7, r6, lsr #20		@ Undo decrement
 	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
-- 
1.7.4.4

[PATCH 22/23] ARM: entry: data abort: ensure r5 is preserved by abort functions

[Russell King - ARM Linux]

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/mm/abort-lv4t.S  |   48 ++++++++++++++++++++------------------------
 arch/arm/mm/proc-arm6_7.S |   33 +++++++++++++++---------------
 2 files changed, 38 insertions(+), 43 deletions(-)

diff --git a/arch/arm/mm/abort-lv4t.S b/arch/arm/mm/abort-lv4t.S
index 54b6d27..f398258 100644
--- a/arch/arm/mm/abort-lv4t.S
+++ b/arch/arm/mm/abort-lv4t.S
@@ -7,11 +7,7 @@
  *	   : r4 = aborted context pc
  *	   : r5 = aborted context psr
  *
- * Returns : r0 = address of abort
- *	   : r1 = FSR, bit 11 = write
- *	   : r2-r8 = corrupted
- *	   : r9 = preserved
- *	   : sp = pointer to registers
+ * Returns : r4-r5, r10-r11, r13 preserved
  *
  * Purpose : obtain information about current aborted instruction.
  * Note: we read user space.  This means we might cause a data
@@ -72,30 +68,30 @@ ENTRY(v4t_late_abort)
 	add	r6, r6, r6, lsr #8
 	add	r6, r6, r6, lsr #4
 	and	r6, r6, #15			@ r6 = no. of registers to transfer.
-	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
+	and	r9, r8, #15 << 16		@ Extract 'n' from instruction
+	ldr	r7, [r2, r9, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r6, lsl #2		@ Undo increment
 	addeq	r7, r7, r6, lsl #2		@ Undo decrement
-	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r9, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_lateldrhpre:
 	tst	r8, #1 << 21			@ Check writeback bit
 	beq	do_DataAbort			@ No writeback -> no fixup
 .data_arm_lateldrhpost:
-	and	r5, r8, #0x00f			@ get Rm / low nibble of immediate value
+	and	r9, r8, #0x00f			@ get Rm / low nibble of immediate value
 	tst	r8, #1 << 22			@ if (immediate offset)
 	andne	r6, r8, #0xf00			@ { immediate high nibble
-	orrne	r6, r5, r6, lsr #4		@   combine nibbles } else
-	ldreq	r6, [r2, r5, lsl #2]		@ { load Rm value }
+	orrne	r6, r9, r6, lsr #4		@   combine nibbles } else
+	ldreq	r6, [r2, r9, lsl #2]		@ { load Rm value }
 .data_arm_apply_r6_and_rn:
-	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
+	and	r9, r8, #15 << 16		@ Extract 'n' from instruction
+	ldr	r7, [r2, r9, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r6			@ Undo incrmenet
 	addeq	r7, r7, r6			@ Undo decrement
-	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r9, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_lateldrpreconst:
@@ -104,12 +100,12 @@ ENTRY(v4t_late_abort)
 .data_arm_lateldrpostconst:
 	movs	r6, r8, lsl #20			@ Get offset
 	beq	do_DataAbort			@ zero -> no fixup
-	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
+	and	r9, r8, #15 << 16		@ Extract 'n' from instruction
+	ldr	r7, [r2, r9, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r6, lsr #20		@ Undo increment
 	addeq	r7, r7, r6, lsr #20		@ Undo decrement
-	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r9, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_lateldrprereg:
@@ -118,14 +114,14 @@ ENTRY(v4t_late_abort)
 .data_arm_lateldrpostreg:
 	and	r7, r8, #15			@ Extract 'm' from instruction
 	ldr	r6, [r2, r7, lsl #2]		@ Get register 'Rm'
-	mov	r5, r8, lsr #7			@ get shift count
-	ands	r5, r5, #31
+	mov	r9, r8, lsr #7			@ get shift count
+	ands	r9, r9, #31
 	and	r7, r8, #0x70			@ get shift type
 	orreq	r7, r7, #8			@ shift count = 0
 	add	pc, pc, r7
 	nop
 
-	mov	r6, r6, lsl r5			@ 0: LSL #!0
+	mov	r6, r6, lsl r9			@ 0: LSL #!0
 	b	.data_arm_apply_r6_and_rn
 	b	.data_arm_apply_r6_and_rn	@ 1: LSL #0
 	nop
@@ -133,7 +129,7 @@ ENTRY(v4t_late_abort)
 	nop
 	b	.data_unknown			@ 3: MUL?
 	nop
-	mov	r6, r6, lsr r5			@ 4: LSR #!0
+	mov	r6, r6, lsr r9			@ 4: LSR #!0
 	b	.data_arm_apply_r6_and_rn
 	mov	r6, r6, lsr #32			@ 5: LSR #32
 	b	.data_arm_apply_r6_and_rn
@@ -141,7 +137,7 @@ ENTRY(v4t_late_abort)
 	nop
 	b	.data_unknown			@ 7: MUL?
 	nop
-	mov	r6, r6, asr r5			@ 8: ASR #!0
+	mov	r6, r6, asr r9			@ 8: ASR #!0
 	b	.data_arm_apply_r6_and_rn
 	mov	r6, r6, asr #32			@ 9: ASR #32
 	b	.data_arm_apply_r6_and_rn
@@ -149,7 +145,7 @@ ENTRY(v4t_late_abort)
 	nop
 	b	.data_unknown			@ B: MUL?
 	nop
-	mov	r6, r6, ror r5			@ C: ROR #!0
+	mov	r6, r6, ror r9			@ C: ROR #!0
 	b	.data_arm_apply_r6_and_rn
 	mov	r6, r6, rrx			@ D: RRX
 	b	.data_arm_apply_r6_and_rn
@@ -216,9 +212,9 @@ ENTRY(v4t_late_abort)
 	and	r6, r6, #0x33
 	add	r6, r6, r9, lsr #2
 	add	r6, r6, r6, lsr #4
-	and	r5, r8, #7 << 8
-	ldr	r7, [r2, r5, lsr #6]
+	and	r9, r8, #7 << 8
+	ldr	r7, [r2, r9, lsr #6]
 	and	r6, r6, #15			@ number of regs to transfer
 	sub	r7, r7, r6, lsl #2		@ always decrement
-	str	r7, [r2, r5, lsr #6]
+	str	r7, [r2, r9, lsr #6]
 	b	do_DataAbort
diff --git a/arch/arm/mm/proc-arm6_7.S b/arch/arm/mm/proc-arm6_7.S
index 4d96311..50e3543 100644
--- a/arch/arm/mm/proc-arm6_7.S
+++ b/arch/arm/mm/proc-arm6_7.S
@@ -35,8 +35,7 @@ ENTRY(cpu_arm7_dcache_clean_area)
  *
  * Purpose : obtain information about current aborted instruction
  *
- * Returns : r0 = address of abort
- *	   : r1 = FSR
+ * Returns : r4-r5, r10-r11, r13 preserved
  */
 
 ENTRY(cpu_arm7_data_abort)
@@ -95,21 +94,21 @@ ENTRY(cpu_arm6_data_abort)
 	add	r6, r6, r6, lsr #8
 	add	r6, r6, r6, lsr #4
 	and	r6, r6, #15			@ r6 = no. of registers to transfer.
-	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
+	and	r9, r8, #15 << 16		@ Extract 'n' from instruction
+	ldr	r7, [r2, r9, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r6, lsl #2		@ Undo increment
 	addeq	r7, r7, r6, lsl #2		@ Undo decrement
-	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r9, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_apply_r6_and_rn:
-	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
+	and	r9, r8, #15 << 16		@ Extract 'n' from instruction
+	ldr	r7, [r2, r9, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r6			@ Undo incrmenet
 	addeq	r7, r7, r6			@ Undo decrement
-	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r9, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_lateldrpreconst:
@@ -118,12 +117,12 @@ ENTRY(cpu_arm6_data_abort)
 .data_arm_lateldrpostconst:
 	movs	r6, r8, lsl #20			@ Get offset
 	beq	do_DataAbort			@ zero -> no fixup
-	and	r5, r8, #15 << 16		@ Extract 'n' from instruction
-	ldr	r7, [r2, r5, lsr #14]		@ Get register 'Rn'
+	and	r9, r8, #15 << 16		@ Extract 'n' from instruction
+	ldr	r7, [r2, r9, lsr #14]		@ Get register 'Rn'
 	tst	r8, #1 << 23			@ Check U bit
 	subne	r7, r7, r6, lsr #20		@ Undo increment
 	addeq	r7, r7, r6, lsr #20		@ Undo decrement
-	str	r7, [r2, r5, lsr #14]		@ Put register 'Rn'
+	str	r7, [r2, r9, lsr #14]		@ Put register 'Rn'
 	b	do_DataAbort
 
 .data_arm_lateldrprereg:
@@ -132,14 +131,14 @@ ENTRY(cpu_arm6_data_abort)
 .data_arm_lateldrpostreg:
 	and	r7, r8, #15			@ Extract 'm' from instruction
 	ldr	r6, [r2, r7, lsl #2]		@ Get register 'Rm'
-	mov	r5, r8, lsr #7			@ get shift count
-	ands	r5, r5, #31
+	mov	r9, r8, lsr #7			@ get shift count
+	ands	r9, r9, #31
 	and	r7, r8, #0x70			@ get shift type
 	orreq	r7, r7, #8			@ shift count = 0
 	add	pc, pc, r7
 	nop
 
-	mov	r6, r6, lsl r5			@ 0: LSL #!0
+	mov	r6, r6, lsl r9			@ 0: LSL #!0
 	b	.data_arm_apply_r6_and_rn
 	b	.data_arm_apply_r6_and_rn	@ 1: LSL #0
 	nop
@@ -147,7 +146,7 @@ ENTRY(cpu_arm6_data_abort)
 	nop
 	b	.data_unknown			@ 3: MUL?
 	nop
-	mov	r6, r6, lsr r5			@ 4: LSR #!0
+	mov	r6, r6, lsr r9			@ 4: LSR #!0
 	b	.data_arm_apply_r6_and_rn
 	mov	r6, r6, lsr #32			@ 5: LSR #32
 	b	.data_arm_apply_r6_and_rn
@@ -155,7 +154,7 @@ ENTRY(cpu_arm6_data_abort)
 	nop
 	b	.data_unknown			@ 7: MUL?
 	nop
-	mov	r6, r6, asr r5			@ 8: ASR #!0
+	mov	r6, r6, asr r9			@ 8: ASR #!0
 	b	.data_arm_apply_r6_and_rn
 	mov	r6, r6, asr #32			@ 9: ASR #32
 	b	.data_arm_apply_r6_and_rn
@@ -163,7 +162,7 @@ ENTRY(cpu_arm6_data_abort)
 	nop
 	b	.data_unknown			@ B: MUL?
 	nop
-	mov	r6, r6, ror r5			@ C: ROR #!0
+	mov	r6, r6, ror r9			@ C: ROR #!0
 	b	.data_arm_apply_r6_and_rn
 	mov	r6, r6, rrx			@ D: RRX
 	b	.data_arm_apply_r6_and_rn
-- 
1.7.4.4

[PATCH 23/23] ARM: entry: no need to reload the SPSR value from struct pt_regs

[Russell King - ARM Linux]

The SVC IRQ, prefetch and data abort handlers preserve the SPSR value
via r5 across the exception.  Rather than re-loading it from pt_regs,
use the preserved value instead.

Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
---
 arch/arm/kernel/entry-armv.S |   10 +---------
 1 files changed, 1 insertions(+), 9 deletions(-)

diff --git a/arch/arm/kernel/entry-armv.S b/arch/arm/kernel/entry-armv.S
index 3fa90ae..047ed5d 100644
--- a/arch/arm/kernel/entry-armv.S
+++ b/arch/arm/kernel/entry-armv.S
@@ -195,10 +195,6 @@ __dabt_svc:
 	@
 	disable_irq_notrace
 
-	@
-	@ restore SPSR and restart the instruction
-	@
-	ldr	r5, [sp, #S_PSR]
 #ifdef CONFIG_TRACE_IRQFLAGS
 	tst	r5, #PSR_I_BIT
 	bleq	trace_hardirqs_on
@@ -223,7 +219,7 @@ __irq_svc:
 	tst	r0, #_TIF_NEED_RESCHED
 	blne	svc_preempt
 #endif
-	ldr	r5, [sp, #S_PSR]
+
 #ifdef CONFIG_TRACE_IRQFLAGS
 	@ The parent context IRQs must have been enabled to get here in
 	@ the first place, so there's no point checking the PSR I bit.
@@ -308,10 +304,6 @@ __pabt_svc:
 	@
 	disable_irq_notrace
 
-	@
-	@ restore SPSR and restart the instruction
-	@
-	ldr	r5, [sp, #S_PSR]
 #ifdef CONFIG_TRACE_IRQFLAGS
 	tst	r5, #PSR_I_BIT
 	bleq	trace_hardirqs_on
-- 
1.7.4.4

[PATCH 00/23] entry assembly cleanups

[Jean-Christophe PLAGNIOL-VILLARD]

On 10:18 Wed 29 Jun     , Russell King - ARM Linux wrote:
> This patch series cleans up the entry assembly and CPU abort helpers.
> Notable things about this patch series:
> 
do you have a git repos where we could pull it?

otherwise great job tks

Best Regards,
J.

[PATCH 11/23] ARM: entry: avoid enabling interrupts in prefetch/data abort handlers

[Will Deacon]

Hi Russell,

This looks good, thanks. Minor comment inline.

On Wed, Jun 29, 2011 at 10:22:35AM +0100, Russell King - ARM Linux wrote:
> Avoid enabling interrupts if the parent context had interrupts enabled
> in the abort handler assembly code, and move this into the breakpoint/
> page/alignment fault handlers instead.
> 
> This gets rid of some special-casing for the breakpoint fault handlers
> from the low level abort handler path.
> 
> Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
> ---
>  arch/arm/kernel/entry-armv.S    |   43 +++++++++++++++++---------------------
>  arch/arm/kernel/entry-header.S  |   19 -----------------
>  arch/arm/kernel/hw_breakpoint.c |    6 +++-
>  arch/arm/mm/alignment.c         |    3 ++
>  arch/arm/mm/fault.c             |    4 +++
>  5 files changed, 30 insertions(+), 45 deletions(-)

[...]

> diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c
> index 87acc25..b813e1e 100644
> --- a/arch/arm/kernel/hw_breakpoint.c
> +++ b/arch/arm/kernel/hw_breakpoint.c
> @@ -804,8 +804,10 @@ static int hw_breakpoint_pending(unsigned long addr, unsigned int fsr,
>  	int ret = 0;
>  	u32 dscr;
>  
> -	/* We must be called with preemption disabled. */
> -	WARN_ON(preemptible());
> +	preempt_disable();
> +
> +	if (interrupts_enabled(regs))
> +		local_irq_enable();
>  
>  	/* We only handle watchpoints and hardware breakpoints. */
>  	ARM_DBG_READ(c1, 0, dscr);

Could you also update the comments for this function too please? There's one
immediately before the function that states we are called with preemption
disabled and there's another one where we re-enable preemption stating that
it was disabled by the low-level exception handling code.

With those two extra changes:

Acked-by: Will Deacon <will.deacon@arm.com>

Cheers,

Will

[PATCH 11/23] ARM: entry: avoid enabling interrupts in prefetch/data abort handlers

[Russell King - ARM Linux]

On Wed, Jun 29, 2011 at 09:05:23PM +0100, Will Deacon wrote:
> Hi Russell,
> 
> This looks good, thanks. Minor comment inline.
> 
> On Wed, Jun 29, 2011 at 10:22:35AM +0100, Russell King - ARM Linux wrote:
> > Avoid enabling interrupts if the parent context had interrupts enabled
> > in the abort handler assembly code, and move this into the breakpoint/
> > page/alignment fault handlers instead.
> > 
> > This gets rid of some special-casing for the breakpoint fault handlers
> > from the low level abort handler path.
> > 
> > Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
> > ---
> >  arch/arm/kernel/entry-armv.S    |   43 +++++++++++++++++---------------------
> >  arch/arm/kernel/entry-header.S  |   19 -----------------
> >  arch/arm/kernel/hw_breakpoint.c |    6 +++-
> >  arch/arm/mm/alignment.c         |    3 ++
> >  arch/arm/mm/fault.c             |    4 +++
> >  5 files changed, 30 insertions(+), 45 deletions(-)
> 
> [...]
> 
> > diff --git a/arch/arm/kernel/hw_breakpoint.c b/arch/arm/kernel/hw_breakpoint.c
> > index 87acc25..b813e1e 100644
> > --- a/arch/arm/kernel/hw_breakpoint.c
> > +++ b/arch/arm/kernel/hw_breakpoint.c
> > @@ -804,8 +804,10 @@ static int hw_breakpoint_pending(unsigned long addr, unsigned int fsr,
> >  	int ret = 0;
> >  	u32 dscr;
> >  
> > -	/* We must be called with preemption disabled. */
> > -	WARN_ON(preemptible());
> > +	preempt_disable();
> > +
> > +	if (interrupts_enabled(regs))
> > +		local_irq_enable();
> >  
> >  	/* We only handle watchpoints and hardware breakpoints. */
> >  	ARM_DBG_READ(c1, 0, dscr);
> 
> Could you also update the comments for this function too please? There's one
> immediately before the function that states we are called with preemption
> disabled and there's another one where we re-enable preemption stating that
> it was disabled by the low-level exception handling code.

I'll change the one before the function thusly s/preemption/interrupts/ 
and remove the other entirely because it no longer serves much purpose.
Ok?

[PATCH 11/23] ARM: entry: avoid enabling interrupts in prefetch/data abort handlers

[Will Deacon]

On Thu, Jun 30, 2011 at 10:27:02AM +0100, Russell King - ARM Linux wrote:
> On Wed, Jun 29, 2011 at 09:05:23PM +0100, Will Deacon wrote:
> > Could you also update the comments for this function too please? There's one
> > immediately before the function that states we are called with preemption
> > disabled and there's another one where we re-enable preemption stating that
> > it was disabled by the low-level exception handling code.
> 
> I'll change the one before the function thusly s/preemption/interrupts/ 
> and remove the other entirely because it no longer serves much purpose.
> Ok?

Sounds good to me, you add my ack with those changes.

Thanks,

Will