ARM: use high vectors if possible

ARM: use high vectors if possible

[Sascha Hauer]

Created this series to get high vector support. Using high vectors
at 0xffff0000 allow us to map in a faulting page at 0x0 with which
we can catch NULL pointer dereferences. The other patches are only
some cleanups around exceptions which I came along the way.

Sascha

Sascha Hauer (8):
      ARM: add missing volatile in get_cr()
      ARM mmu: fix arm_create_pte
      ARM mmu: use high vectors if possible
      ARM: remove unused exception
      ARM: exceptions: remove unnecessary function declarations
      ARM: remove unused irq enable/disable functions
      ARM: fix comments in interrupts.c
      ARM: some cleanup in interrupts.c

 arch/arm/cpu/exceptions.S     |    5 --
 arch/arm/cpu/interrupts.c     |  105 ++++++++++-------------------------------
 arch/arm/cpu/mmu.c            |   39 ++++++++++++---
 arch/arm/cpu/start.c          |    2 +-
 arch/arm/include/asm/system.h |    2 +-
 5 files changed, 60 insertions(+), 93 deletions(-)

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 1/8] ARM: add missing volatile in get_cr()

[Sascha Hauer]

Without it, the compiler optimizes away subsequent reads of the
control register.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 arch/arm/include/asm/system.h |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/arm/include/asm/system.h b/arch/arm/include/asm/system.h
index 77d6305..1d67492 100644
--- a/arch/arm/include/asm/system.h
+++ b/arch/arm/include/asm/system.h
@@ -59,7 +59,7 @@
 static inline unsigned int get_cr(void)
 {
 	unsigned int val;
-	asm("mrc p15, 0, %0, c1, c0, 0  @ get CR" : "=r" (val) : : "cc");
+	asm volatile ("mrc p15, 0, %0, c1, c0, 0  @ get CR" : "=r" (val) : : "cc");
 	return val;
 }
 
-- 
1.7.5.4


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 2/8] ARM mmu: fix arm_create_pte

[Sascha Hauer]

Each section is 1MiB, so we have to shift by 20 to get the ttb
entry corresponding to a section.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 arch/arm/cpu/mmu.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/arch/arm/cpu/mmu.c b/arch/arm/cpu/mmu.c
index 1f62d61..2f754c8 100644
--- a/arch/arm/cpu/mmu.c
+++ b/arch/arm/cpu/mmu.c
@@ -65,7 +65,7 @@ static u32 *arm_create_pte(unsigned long virt)
 
 	table = memalign(0x400, 0x400);
 
-	ttb[virt] = (unsigned long)table | PMD_TYPE_TABLE;
+	ttb[virt >> 20] = (unsigned long)table | PMD_TYPE_TABLE;
 
 	for (i = 0; i < 256; i++) {
 		table[i] = virt | PTE_TYPE_SMALL | PTE_FLAGS_UNCACHED;
-- 
1.7.5.4


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 3/8] ARM mmu: use high vectors if possible

[Sascha Hauer]

Using high vectors allows us to map a faulting zero page to
catch NULL pointer dereferences.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 arch/arm/cpu/mmu.c |   37 +++++++++++++++++++++++++++++++------
 1 files changed, 31 insertions(+), 6 deletions(-)

diff --git a/arch/arm/cpu/mmu.c b/arch/arm/cpu/mmu.c
index 2f754c8..b669349 100644
--- a/arch/arm/cpu/mmu.c
+++ b/arch/arm/cpu/mmu.c
@@ -4,6 +4,7 @@
 #include <errno.h>
 #include <sizes.h>
 #include <asm/memory.h>
+#include <asm/system.h>
 
 static unsigned long *ttb;
 
@@ -161,22 +162,46 @@ static int arm_mmu_remap_sdram(struct arm_memory *mem)
 #define ARM_VECTORS_SIZE	(sizeof(u32) * 8 * 2)
 
 /*
- * Allocate a page, map it to the zero page and copy our exception
- * vectors there.
+ * Map vectors and zero page
  */
 static void vectors_init(void)
 {
-	u32 *exc;
+	u32 *exc, *zero = NULL;
 	void *vectors;
 	extern unsigned long exception_vectors;
-
-	exc = arm_create_pte(0x0);
+	u32 cr;
+
+	cr = get_cr();
+	cr |= CR_V;
+	set_cr(cr);
+	cr = get_cr();
+
+	if (cr & CR_V) {
+		/*
+		 * If we can use high vectors, create the second level
+		 * page table for the high vectors and zero page
+		 */
+		exc = arm_create_pte(0xfff00000);
+		zero = arm_create_pte(0x0);
+
+		/* Set the zero page to faulting */
+		zero[0] = 0;
+	} else {
+		/*
+		 * Otherwise map the vectors to the zero page. We have to
+		 * live without being able to catch NULL pointer dereferences
+		 */
+		exc = arm_create_pte(0x0);
+	}
 
 	vectors = xmemalign(PAGE_SIZE, PAGE_SIZE);
 	memset(vectors, 0, PAGE_SIZE);
 	memcpy(vectors, &exception_vectors, ARM_VECTORS_SIZE);
 
-	exc[0] = (u32)vectors | PTE_TYPE_SMALL | PTE_FLAGS_CACHED;
+	if (cr & CR_V)
+		exc[256 - 16] = (u32)vectors | PTE_TYPE_SMALL | PTE_FLAGS_CACHED;
+	else
+		exc[0] = (u32)vectors | PTE_TYPE_SMALL | PTE_FLAGS_CACHED;
 }
 
 /*
-- 
1.7.5.4


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 4/8] ARM: remove unused exception

[Sascha Hauer]

The exception vector at 0x14 is not used on arm, so no need
to bind this address to a exception handler. Remove the
corresponding code

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 arch/arm/cpu/exceptions.S |    5 -----
 arch/arm/cpu/interrupts.c |   13 -------------
 arch/arm/cpu/start.c      |    2 +-
 3 files changed, 1 insertions(+), 19 deletions(-)

diff --git a/arch/arm/cpu/exceptions.S b/arch/arm/cpu/exceptions.S
index 4741bb8..a879755 100644
--- a/arch/arm/cpu/exceptions.S
+++ b/arch/arm/cpu/exceptions.S
@@ -136,11 +136,6 @@ data_abort:
 	bl 	do_data_abort
 
 	.align	5
-.globl not_used
-not_used:
-	get_bad_stack
-	bad_save_user_regs
-	bl 	do_not_used
 
 .globl irq
 .globl fiq
diff --git a/arch/arm/cpu/interrupts.c b/arch/arm/cpu/interrupts.c
index 5168921..68b2931 100644
--- a/arch/arm/cpu/interrupts.c
+++ b/arch/arm/cpu/interrupts.c
@@ -173,19 +173,6 @@ void do_data_abort (struct pt_regs *pt_regs)
 }
 
 /**
- * The CPU catches a not-used(?) abort.
- * @param[in] pt_regs Register set content when the accident happens
- *
- * FIXME: What does it mean, why is reset the only solution?
- */
-void do_not_used (struct pt_regs *pt_regs)
-{
-	printf ("not used\n");
-	show_regs (pt_regs);
-	bad_mode ();
-}
-
-/**
  * The CPU catches a fast interrupt request.
  * @param[in] pt_regs Register set content when the interrupt happens
  *
diff --git a/arch/arm/cpu/start.c b/arch/arm/cpu/start.c
index f20ce74..5e09300 100644
--- a/arch/arm/cpu/start.c
+++ b/arch/arm/cpu/start.c
@@ -36,7 +36,7 @@ void __naked __section(.text_entry) exception_vectors(void)
 		"ldr pc, =software_interrupt\n"		/* software interrupt (SWI) */
 		"ldr pc, =prefetch_abort\n"		/* prefetch abort */
 		"ldr pc, =data_abort\n"			/* data abort */
-		"ldr pc, =not_used\n"			/* (reserved) */
+		"1: bne 1b\n"				/* (reserved) */
 		"ldr pc, =irq\n"			/* irq (interrupt) */
 		"ldr pc, =fiq\n"			/* fiq (fast interrupt) */
 #else
-- 
1.7.5.4


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 5/8] ARM: exceptions: remove unnecessary function declarations

[Sascha Hauer]

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 arch/arm/cpu/interrupts.c |    8 --------
 1 files changed, 0 insertions(+), 8 deletions(-)

diff --git a/arch/arm/cpu/interrupts.c b/arch/arm/cpu/interrupts.c
index 68b2931..a405fae 100644
--- a/arch/arm/cpu/interrupts.c
+++ b/arch/arm/cpu/interrupts.c
@@ -29,14 +29,6 @@
 #include <asm/ptrace.h>
 #include <asm/unwind.h>
 
-void do_undefined_instruction (struct pt_regs *pt_regs);
-void do_software_interrupt (struct pt_regs *pt_regs);
-void do_prefetch_abort (struct pt_regs *pt_regs);
-void do_data_abort (struct pt_regs *pt_regs);
-void do_not_used (struct pt_regs *pt_regs);
-void do_fiq (struct pt_regs *pt_regs);
-void do_irq (struct pt_regs *pt_regs);
-
 #ifdef CONFIG_USE_IRQ
 /* enable IRQ interrupts */
 void enable_interrupts (void)
-- 
1.7.5.4


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 6/8] ARM: remove unused irq enable/disable functions

[Sascha Hauer]

We do not have irq support in barebox, so remove the unused interrupt
functions.

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 arch/arm/cpu/interrupts.c |   31 -------------------------------
 1 files changed, 0 insertions(+), 31 deletions(-)

diff --git a/arch/arm/cpu/interrupts.c b/arch/arm/cpu/interrupts.c
index a405fae..71744ac 100644
--- a/arch/arm/cpu/interrupts.c
+++ b/arch/arm/cpu/interrupts.c
@@ -29,37 +29,6 @@
 #include <asm/ptrace.h>
 #include <asm/unwind.h>
 
-#ifdef CONFIG_USE_IRQ
-/* enable IRQ interrupts */
-void enable_interrupts (void)
-{
-	unsigned long temp;
-	__asm__ __volatile__("mrs %0, cpsr\n"
-			     "bic %0, %0, #0x80\n"
-			     "msr cpsr_c, %0"
-			     : "=r" (temp)
-			     :
-			     : "memory");
-}
-
-
-/*
- * disable IRQ/FIQ interrupts
- * returns true if interrupts had been enabled before we disabled them
- */
-int disable_interrupts (void)
-{
-	unsigned long old,temp;
-	__asm__ __volatile__("mrs %0, cpsr\n"
-			     "orr %1, %0, #0xc0\n"
-			     "msr cpsr_c, %1"
-			     : "=r" (old), "=r" (temp)
-			     :
-			     : "memory");
-	return(old & 0x80) == 0;
-}
-#endif
-
 /**
  * FIXME
  */
-- 
1.7.5.4


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 7/8] ARM: fix comments in interrupts.c

[Sascha Hauer]

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 arch/arm/cpu/interrupts.c |   12 ++++++------
 1 files changed, 6 insertions(+), 6 deletions(-)

diff --git a/arch/arm/cpu/interrupts.c b/arch/arm/cpu/interrupts.c
index 71744ac..0c21bc1 100644
--- a/arch/arm/cpu/interrupts.c
+++ b/arch/arm/cpu/interrupts.c
@@ -97,8 +97,8 @@ void do_undefined_instruction (struct pt_regs *pt_regs)
  * The CPU catches a software interrupt
  * @param[in] pt_regs Register set content when the interrupt happens
  *
- * There is not functione behind this feature. So what to do else than
- * a reset? 
+ * There is no function behind this feature. So what to do else than
+ * a reset?
  */
 void do_software_interrupt (struct pt_regs *pt_regs)
 {
@@ -111,7 +111,7 @@ void do_software_interrupt (struct pt_regs *pt_regs)
  * The CPU catches a prefetch abort. That really should not happen!
  * @param[in] pt_regs Register set content when the accident happens
  *
- * FIXME: What does it mean, why is reset the only solution?
+ * instruction fetch from an unmapped area
  */
 void do_prefetch_abort (struct pt_regs *pt_regs)
 {
@@ -124,7 +124,7 @@ void do_prefetch_abort (struct pt_regs *pt_regs)
  * The CPU catches a data abort. That really should not happen!
  * @param[in] pt_regs Register set content when the accident happens
  *
- * FIXME: What does it mean, why is reset the only solution?
+ * data fetch from an unmapped area
  */
 void do_data_abort (struct pt_regs *pt_regs)
 {
@@ -137,7 +137,7 @@ void do_data_abort (struct pt_regs *pt_regs)
  * The CPU catches a fast interrupt request.
  * @param[in] pt_regs Register set content when the interrupt happens
  *
- * FIXME: What does it mean, why is reset the only solution?
+ * We never enable FIQs, so this should not happen
  */
 void do_fiq (struct pt_regs *pt_regs)
 {
@@ -150,7 +150,7 @@ void do_fiq (struct pt_regs *pt_regs)
  * The CPU catches a regular interrupt.
  * @param[in] pt_regs Register set content when the interrupt happens
  *
- * FIXME: What does it mean, why is reset the only solution?
+ * We never enable interrupts, so this should not happen
  */
 void do_irq (struct pt_regs *pt_regs)
 {
-- 
1.7.5.4


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

[PATCH 8/8] ARM: some cleanup in interrupts.c

[Sascha Hauer]

- Don't call panic with "resetting CPU...". Depending on the
  configuration the system might also hang.
- panic does not return, so no need to call reset_cpu afterwards
- bundle show_regs and panic into a seperate functions to not have
  to call both functions from each exception handler

Signed-off-by: Sascha Hauer <s.hauer@pengutronix.de>
---
 arch/arm/cpu/interrupts.c |   43 +++++++++++++++++++++----------------------
 1 files changed, 21 insertions(+), 22 deletions(-)

diff --git a/arch/arm/cpu/interrupts.c b/arch/arm/cpu/interrupts.c
index 0c21bc1..3d2077f 100644
--- a/arch/arm/cpu/interrupts.c
+++ b/arch/arm/cpu/interrupts.c
@@ -30,15 +30,6 @@
 #include <asm/unwind.h>
 
 /**
- * FIXME
- */
-static void bad_mode (void)
-{
-	panic ("Resetting CPU ...\n");
-	reset_cpu (0);
-}
-
-/**
  * Display current register set content
  * @param[in] regs Guess what
  */
@@ -82,6 +73,13 @@ void show_regs (struct pt_regs *regs)
 #endif
 }
 
+static void __noreturn do_exception(struct pt_regs *pt_regs)
+{
+	show_regs(pt_regs);
+
+	panic("");
+}
+
 /**
  * The CPU runs into an undefined instruction. That really should not happen!
  * @param[in] pt_regs Register set content when the accident happens
@@ -89,8 +87,7 @@ void show_regs (struct pt_regs *regs)
 void do_undefined_instruction (struct pt_regs *pt_regs)
 {
 	printf ("undefined instruction\n");
-	show_regs (pt_regs);
-	bad_mode ();
+	do_exception(pt_regs);
 }
 
 /**
@@ -103,8 +100,7 @@ void do_undefined_instruction (struct pt_regs *pt_regs)
 void do_software_interrupt (struct pt_regs *pt_regs)
 {
 	printf ("software interrupt\n");
-	show_regs (pt_regs);
-	bad_mode ();
+	do_exception(pt_regs);
 }
 
 /**
@@ -116,8 +112,7 @@ void do_software_interrupt (struct pt_regs *pt_regs)
 void do_prefetch_abort (struct pt_regs *pt_regs)
 {
 	printf ("prefetch abort\n");
-	show_regs (pt_regs);
-	bad_mode ();
+	do_exception(pt_regs);
 }
 
 /**
@@ -128,9 +123,15 @@ void do_prefetch_abort (struct pt_regs *pt_regs)
  */
 void do_data_abort (struct pt_regs *pt_regs)
 {
-	printf ("data abort\n");
-	show_regs (pt_regs);
-	bad_mode ();
+	u32 far;
+
+	asm volatile ("mrc     p15, 0, %0, c6, c0, 0" : "=r" (far) : : "cc");
+
+	printf("unable to handle %s at address 0x%08x\n",
+			far < PAGE_SIZE ? "NULL pointer dereference" :
+			"paging request", far);
+
+	do_exception(pt_regs);
 }
 
 /**
@@ -142,8 +143,7 @@ void do_data_abort (struct pt_regs *pt_regs)
 void do_fiq (struct pt_regs *pt_regs)
 {
 	printf ("fast interrupt request\n");
-	show_regs (pt_regs);
-	bad_mode ();
+	do_exception(pt_regs);
 }
 
 /**
@@ -155,6 +155,5 @@ void do_fiq (struct pt_regs *pt_regs)
 void do_irq (struct pt_regs *pt_regs)
 {
 	printf ("interrupt request\n");
-	show_regs (pt_regs);
-	bad_mode ();
+	do_exception(pt_regs);
 }
-- 
1.7.5.4


_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: ARM: use high vectors if possible

[Jean-Christophe PLAGNIOL-VILLARD]

On 08:46 Mon 08 Aug     , Sascha Hauer wrote:
> Created this series to get high vector support. Using high vectors
> at 0xffff0000 allow us to map in a faulting page at 0x0 with which
> we can catch NULL pointer dereferences. The other patches are only
> some cleanups around exceptions which I came along the way.
be careful on ixp4xx (which I work on 0x0 is the nor flash address)

Best Regards,
J.

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: ARM: use high vectors if possible

[Sascha Hauer]

Hi J,

On Mon, Aug 08, 2011 at 09:06:50AM +0200, Jean-Christophe PLAGNIOL-VILLARD wrote:
> On 08:46 Mon 08 Aug     , Sascha Hauer wrote:
> > Created this series to get high vector support. Using high vectors
> > at 0xffff0000 allow us to map in a faulting page at 0x0 with which
> > we can catch NULL pointer dereferences. The other patches are only
> > some cleanups around exceptions which I came along the way.
> be careful on ixp4xx (which I work on 0x0 is the nor flash address)

Well, we already map the vectors to 0x0 without this series, so ixp4xx
is broken anyway.
So we either have to add a flag somewhere to disable mapping anything
to 0x0 or we have to use the mmu to map the nor flash somewhere else.

I see this is a problem which must be solved, but this series does not
change the current situation in this regard.

Sascha


-- 
Pengutronix e.K.                           |                             |
Industrial Linux Solutions                 | http://www.pengutronix.de/  |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0    |
Amtsgericht Hildesheim, HRA 2686           | Fax:   +49-5121-206917-5555 |

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox

Re: ARM: use high vectors if possible

[Jean-Christophe PLAGNIOL-VILLARD]

On 09:32 Mon 08 Aug     , Sascha Hauer wrote:
> Hi J,
> 
> On Mon, Aug 08, 2011 at 09:06:50AM +0200, Jean-Christophe PLAGNIOL-VILLARD wrote:
> > On 08:46 Mon 08 Aug     , Sascha Hauer wrote:
> > > Created this series to get high vector support. Using high vectors
> > > at 0xffff0000 allow us to map in a faulting page at 0x0 with which
> > > we can catch NULL pointer dereferences. The other patches are only
> > > some cleanups around exceptions which I came along the way.
> > be careful on ixp4xx (which I work on 0x0 is the nor flash address)
> 
> Well, we already map the vectors to 0x0 without this series, so ixp4xx
> is broken anyway.
> So we either have to add a flag somewhere to disable mapping anything
> to 0x0 or we have to use the mmu to map the nor flash somewhere else.
> 
> I see this is a problem which must be solved, but this series does not
> change the current situation in this regard.
I'll prefer to remap the flash and keep the same scheme everywhere

can you drop a line in the doc about this

Best Regards,
J.

_______________________________________________
barebox mailing list
barebox@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/barebox