Skip to contentSkip to conversation

ARM: mmu: Set domain permissions to client access

PATCH New R Sricharan 5 participants 25 messages Jan 8, 2013 – Mar 28, 2013 Arm (32-bit)

lore
Share & tools
lore
#0 lore raw

cover ARM: mmu: Set domain permissions to client access

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by
<Vincent Stehle <[email protected]>

[1] http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

 arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
 arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
 arch/arm/include/asm/cache.h                   |    1 +
 arch/arm/include/asm/system.h                  |   14 ++++++++++
 arch/arm/lib/cache-cp15.c                      |   13 ++++++++-
 5 files changed, 65 insertions(+), 1 deletion(-)
Signature · 1 line
1.7.9.5
#1 lore raw

PATCH 1/2 ARM: mmu: Introduce weak dram_bank_setup function

Introduce a weak version of dram_bank_setup function to allow a platform specific function.

This is used in the subsequent patch to setup dram region without 'XN' attribute in order to enable the region under client permissions.

Cc Vincent Stehle <[email protected]>
Cc Tom Rini <[email protected]>
---
 arch/arm/include/asm/cache.h |    1 +
 arch/arm/lib/cache-cp15.c    |    4 +++-
 2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/include/asm/cache.h +1 -0
index eef6a5a..93811d2 100644
@@ -42,6 +42,7 @@ static inline void invalidate_l2_cache(void)
 void l2_cache_enable(void);
 void l2_cache_disable(void);
 
+void dram_bank_mmu_setup(int bank);
 /*
  * The current upper bound for ARM L1 data cache line sizes is 64 bytes.  We
  * use that value for aligning DMA buffers unless the board config has specified
arch/arm/lib/cache-cp15.c +3 -1
index 6edf815..3e5de55 100644
@@ -23,6 +23,8 @@
 
 #include <common.h>
 #include <asm/system.h>
+#include <asm/cache.h>
+#include <linux/compiler.h>
 
 #if !(defined(CONFIG_SYS_ICACHE_OFF) && defined(CONFIG_SYS_DCACHE_OFF))
 
@@ -77,7 +79,7 @@ void mmu_set_region_dcache_behaviour(u32 start, int size,
 	mmu_page_table_flush((u32)&page_table[start], (u32)&page_table[end]);
 }
 
-static inline void dram_bank_mmu_setup(int bank)
+__weak void dram_bank_mmu_setup(int bank)
 {
 	bd_t *bd = gd->bd;
 	int	i;
Signature · 1 line
1.7.9.5
#2 lore raw

PATCH 1/2 ARM: mmu: Introduce weak dram_bank_setup function

In reply to R Sricharan

Introduce a weak version of dram_bank_setup function to allow a platform specific function.

This is used in the subsequent patch to setup dram region without 'XN' attribute in order to enable the region under client permissions.

Cc Vincent Stehle <[email protected]>
Cc Tom Rini <[email protected]>
---
 arch/arm/include/asm/cache.h |    1 +
 arch/arm/lib/cache-cp15.c    |    4 +++-
 2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/include/asm/cache.h +1 -0
index eef6a5a..93811d2 100644
@@ -42,6 +42,7 @@ static inline void invalidate_l2_cache(void)
 void l2_cache_enable(void);
 void l2_cache_disable(void);
 
+void dram_bank_mmu_setup(int bank);
 /*
  * The current upper bound for ARM L1 data cache line sizes is 64 bytes.  We
  * use that value for aligning DMA buffers unless the board config has specified
arch/arm/lib/cache-cp15.c +3 -1
index 6edf815..3e5de55 100644
@@ -23,6 +23,8 @@
 
 #include <common.h>
 #include <asm/system.h>
+#include <asm/cache.h>
+#include <linux/compiler.h>
 
 #if !(defined(CONFIG_SYS_ICACHE_OFF) && defined(CONFIG_SYS_DCACHE_OFF))
 
@@ -77,7 +79,7 @@ void mmu_set_region_dcache_behaviour(u32 start, int size,
 	mmu_page_table_flush((u32)&page_table[start], (u32)&page_table[end]);
 }
 
-static inline void dram_bank_mmu_setup(int bank)
+__weak void dram_bank_mmu_setup(int bank)
 {
 	bd_t *bd = gd->bd;
 	int	i;
Signature · 1 line
1.7.9.5
#3 lore raw

PATCH 2/2 ARM: mmu: Set domain permissions to client access

In reply to R Sricharan

The 'XN' execute never bit is set in the pagetables. This will prevent speculative prefetches to non executable regions. But the domain permissions are set as master in the DACR register. So the pagetable attribute for 'XN' is not effective. Change the permissions to client.

This fixes lot of speculative prefetch aborts seen on OMAP5 secure devices.

---
 arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
 arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
 arch/arm/include/asm/system.h                  |   14 ++++++++++
 arch/arm/lib/cache-cp15.c                      |    7 +++++
 4 files changed, 59 insertions(+)
arch/arm/cpu/armv7/cache_v7.c +3 -0
index 5f6d039..8748c14 100644
@@ -340,6 +340,9 @@ void mmu_page_table_flush(unsigned long start, unsigned long stop)
 {
 }
 
+void arm_init_domains(void)
+{
+}
 #endif /* #ifndef CONFIG_SYS_DCACHE_OFF */
 
 #ifndef CONFIG_SYS_ICACHE_OFF
arch/arm/cpu/armv7/omap-common/hwinit-common.c +35 -0
index 9ef10bd..4eaf75b 100644
@@ -32,6 +32,12 @@
 #include <asm/arch/sys_proto.h>
 #include <asm/sizes.h>
 #include <asm/emif.h>
+#include <asm/cache.h>
+#include <asm/system.h>
+
+#define ARMV7_DCACHE_WRITEBACK  0xe
+#define	ARMV7_DOMAIN_CLIENT	1
+#define ARMV7_DOMAIN_MASK	(0x3 << 0)
 
 DECLARE_GLOBAL_DATA_PTR;
 
@@ -258,4 +264,33 @@ void enable_caches(void)
 	/* Enable D-cache. I-cache is already enabled in start.S */
 	dcache_enable();
 }
+
+void dram_bank_mmu_setup(int bank)
+{
+	bd_t *bd = gd->bd;
+	int	i;
+
+	u32 start = bd->bi_dram[bank].start >> 20;
+	u32 size = bd->bi_dram[bank].size >> 20;
+	u32 end = start + size;
+
+	debug("%s: bank: %d\n", __func__, bank);
+	for (i = start; i < end; i++)
+		set_section_dcache(i, ARMV7_DCACHE_WRITEBACK);
+
+}
+
+void arm_init_domains(void)
+{
+	u32 reg;
+
+	reg = get_dacr();
+	/*
+	* Set DOMAIN to client access so that all permissions
+	* set in pagetables are validated by the mmu.
+	*/
+	reg &= ~ARMV7_DOMAIN_MASK;
+	reg |= ARMV7_DOMAIN_CLIENT;
+	set_dacr(reg);
+}
 #endif
arch/arm/include/asm/system.h +14 -0
index 78ca8e0..9cfbb17 100644
@@ -75,6 +75,20 @@ static inline void set_cr(unsigned int val)
 	isb();
 }
 
+static inline unsigned int get_dacr(void)
+{
+	unsigned int val;
+	asm("mrc p15, 0, %0, c3, c0, 0	@ get DACR" : "=r" (val) : : "cc");
+	return val;
+}
+
+static inline void set_dacr(unsigned int val)
+{
+	asm volatile("mcr p15, 0, %0, c3, c0, 0	@ set DACR"
+	  : : "r" (val) : "cc");
+	isb();
+}
+
 /* options available for data cache on each page */
 enum dcache_option {
 	DCACHE_OFF = 0x12,
arch/arm/lib/cache-cp15.c +7 -0
index 3e5de55..0852f40 100644
@@ -36,6 +36,10 @@ void __arm_init_before_mmu(void)
 void arm_init_before_mmu(void)
 	__attribute__((weak, alias("__arm_init_before_mmu")));
 
+__weak void arm_init_domains(void)
+{
+}
+
 static void cp_delay (void)
 {
 	volatile int i;
@@ -117,6 +121,9 @@ static inline void mmu_setup(void)
 	/* Set the access control to all-supervisor */
 	asm volatile("mcr p15, 0, %0, c3, c0, 0"
 		     : : "r" (~0));
+
+	arm_init_domains();
+
 	/* and enable the mmu */
 	reg = get_cr();	/* get control reg. */
 	cp_delay();
Signature · 1 line
1.7.9.5
#4 lore raw

cover ARM: mmu: Set domain permissions to client access

In reply to R Sricharan

Hi R Sicharan, Vincent,

Quoted 20 lines · R Sricharan

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by
<Vincent Stehle <[email protected]>

[1] http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

Quoted 6 lines
 arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
 arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
 arch/arm/include/asm/cache.h                   |    1 +
 arch/arm/include/asm/system.h                  |   14 ++++++++++
 arch/arm/lib/cache-cp15.c                      |   13 ++++++++-
 5 files changed, 65 insertions(+), 1 deletion(-)

Does that mean the rest of Vincent's patch set is not needed any more?

Amicalement,

Signature · 1 line
Albert.
#5 lore raw

cover ARM: mmu: Set domain permissions to client access

Hi,

Quoted 25 lines · Albert ARIBAUD

Hi R Sicharan, Vincent,

Quoted 20 lines · R Sricharan

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by
<Vincent Stehle <[email protected]>

[1] http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

Quoted 8 lines
Quoted 6 lines
  arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
  arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
  arch/arm/include/asm/cache.h                   |    1 +
  arch/arm/include/asm/system.h                  |   14 ++++++++++
  arch/arm/lib/cache-cp15.c                      |   13 ++++++++-
  5 files changed, 65 insertions(+), 1 deletion(-)

Does that mean the rest of Vincent's patch set is not needed any more?

Except for one patch that i have mentioned, rest are needed

http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

Regards,
Sricharan

#6 lore raw

cover ARM: mmu: Set domain permissions to client access

Hi,

Quoted 36 lines · R Sricharan

Hi,

Quoted 33 lines · Albert ARIBAUD

Hi R Sicharan, Vincent,

On Tue, 8 Jan 2013 23:38:22 +0530, R Sricharan <[email protected]>
wrote: i meant

Quoted 27 lines

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by
<Vincent Stehle <[email protected]>

[1] http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

  arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
  arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35
++++++++++++++++++++++++
  arch/arm/include/asm/cache.h                   |    1 +
  arch/arm/include/asm/system.h                  |   14 ++++++++++
  arch/arm/lib/cache-cp15.c                      |   13 ++++++++-
  5 files changed, 65 insertions(+), 1 deletion(-)
Quoted 5 lines
Quoted 1 line

Does that mean the rest of Vincent's patch set is not needed any more?

Except for one patch that i have mentioned, rest are needed

   http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

Sorry..
i meant Except for one patch, rest are "not needed"

Regards,
Sricharan

#7 lore raw

cover ARM: mmu: Set domain permissions to client access

Hi R,

Quoted 38 lines · R Sricharan

Hi,

Quoted 36 lines · R Sricharan

Hi,

Quoted 33 lines · Albert ARIBAUD

Hi R Sicharan, Vincent,

On Tue, 8 Jan 2013 23:38:22 +0530, R Sricharan <[email protected]>
wrote: i meant

Quoted 27 lines

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by
<Vincent Stehle <[email protected]>

[1] http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

  arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
  arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35
++++++++++++++++++++++++
  arch/arm/include/asm/cache.h                   |    1 +
  arch/arm/include/asm/system.h                  |   14 ++++++++++
  arch/arm/lib/cache-cp15.c                      |   13 ++++++++-
  5 files changed, 65 insertions(+), 1 deletion(-)
Quoted 7 lines
Quoted 5 lines
Quoted 1 line

Does that mean the rest of Vincent's patch set is not needed any more?

Except for one patch that i have mentioned, rest are needed

   http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

Sorry..
i meant Except for one patch, rest are "not needed"

Sorry, I may not have made myself clear enough. Does your patchset replace Vincent's patches 2/3 and 3/3 and should I consider them "superseded"?

Quoted 2 lines

Regards,
Sricharan

Amicalement,

Signature · 1 line
Albert.
#8 lore raw

cover ARM: mmu: Set domain permissions to client access

Quoted 46 lines · Albert ARIBAUD

Hi R,

Quoted 41 lines · R Sricharan

Hi,

Quoted 39 lines · R Sricharan

Hi,

Quoted 33 lines · Albert ARIBAUD

Hi R Sicharan, Vincent,

On Tue, 8 Jan 2013 23:38:22 +0530, R Sricharan <[email protected]>
wrote: i meant

Quoted 20 lines

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by
<Vincent Stehle <[email protected]>

[1] http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

Quoted 7 lines
   arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
   arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35
++++++++++++++++++++++++
   arch/arm/include/asm/cache.h                   |    1 +
   arch/arm/include/asm/system.h                  |   14 ++++++++++
   arch/arm/lib/cache-cp15.c                      |   13 ++++++++-
   5 files changed, 65 insertions(+), 1 deletion(-)
Quoted 1 line

Does that mean the rest of Vincent's patch set is not needed any more?

Except for one patch that i have mentioned, rest are needed

Quoted 8 lines
Quoted 4 lines
Quoted 2 lines
    http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

Sorry..
i meant Except for one patch, rest are "not needed"

Sorry, I may not have made myself clear enough. Does your patchset replace Vincent's patches 2/3 and 3/3 and should I consider them "superseded"?

Yes, you are correct.

Regards,
Sricharan

#9 lore raw

cover ARM: mmu: Set domain permissions to client access

In reply to R Sricharan

(sorry for any duplicate of this mail)

Hi R Sricharan,

Quoted 20 lines · R Sricharan

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by
<Vincent Stehle <[email protected]>

[1] http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

Quoted 7 lines
 arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
 arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
 arch/arm/include/asm/cache.h                   |    1 +
 arch/arm/include/asm/system.h                  |   14 ++++++++++
 arch/arm/lib/cache-cp15.c                      |   13 ++++++++-
 5 files changed, 65 insertions(+), 1 deletion(-)

Series applied to u-boot-arm/master, thanks!

Amicalement,

Signature · 1 line
Albert.
#10 lore raw

cover ARM: mmu: Set domain permissions to client access - build warnings!

Quoted 35 lines · Albert ARIBAUD

(sorry for any duplicate of this mail)

Hi R Sricharan,

Quoted 20 lines · R Sricharan

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by
<Vincent Stehle <[email protected]>

[1] http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

Quoted 7 lines
 arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
 arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
 arch/arm/include/asm/cache.h                   |    1 +
 arch/arm/include/asm/system.h                  |   14 ++++++++++
 arch/arm/lib/cache-cp15.c                      |   13 ++++++++-
 5 files changed, 65 insertions(+), 1 deletion(-)
Quoted 1 line

Series applied to u-boot-arm/master, thanks!

Correction: I'd wrongly inferred from the testing done on OMAP5 that the series would build fine. However it appears that for targets omap4_sdp4430, omap4_panda and omap5_evm, patch 2/2 has warnings:

hwinit-common.c: In function 'dram_bank_mmu_setup':
hwinit-common.c:279:3: warning: implicit declaration of function 'set_section_dcache' [-Wimplicit-function-declaration]

Luckily, I had not committed to the git repo yet.

R Sicharan, can you please check and fix this warning?

Amicalement,

Signature · 1 line
Albert.
#11 lore raw

cover ARM: mmu: Set domain permissions to client access - build warnings!

Hi Albert,

Quoted 46 lines · Albert ARIBAUD
Quoted 27 lines · Albert ARIBAUD

(sorry for any duplicate of this mail)

Hi R Sricharan,

Quoted 20 lines · R Sricharan

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by
<Vincent Stehle<[email protected]>

[1] http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

Quoted 8 lines
Quoted 7 lines
  arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
  arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
  arch/arm/include/asm/cache.h                   |    1 +
  arch/arm/include/asm/system.h                  |   14 ++++++++++
  arch/arm/lib/cache-cp15.c                      |   13 ++++++++-
  5 files changed, 65 insertions(+), 1 deletion(-)

Series applied to u-boot-arm/master, thanks!

Correction: I'd wrongly inferred from the testing done on OMAP5 that the series would build fine. However it appears that for targets omap4_sdp4430, omap4_panda and omap5_evm, patch 2/2 has warnings:

hwinit-common.c: In function 'dram_bank_mmu_setup':
hwinit-common.c:279:3: warning: implicit declaration of function 'set_section_dcache' [-Wimplicit-function-declaration]

Quoted 5 lines

Luckily, I had not committed to the git repo yet.

R Sicharan, can you please check and fix this warning?

Amicalement,

Oh, thats because as i mentioned above, it depends on the below patch from vincent .

http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

I did not repost it though. I will do it now for clarity.

Regards,
Sricharan

#12 lore raw

PATCH RESEND 1/1 ARM: cache: declare set_section_dcache

From: Vincent Stehl? <[email protected]>

We declare the set_section_dcache function globally in the cache header, for later use by e.g. machine specific code.

Signed-off-by Vincent Stehl? <v-stehle <at> ti.com>
Cc Tom Rini <trini <at> ti.com>
Cc Albert ARIBAUD <[email protected]>
---
 arch/arm/include/asm/cache.h |    1 +
 1 file changed, 1 insertion(+)
arch/arm/include/asm/cache.h +1 -0
index eef6a5a..416d2c8 100644
@@ -41,6 +41,7 @@ static inline void invalidate_l2_cache(void)
 
 void l2_cache_enable(void);
 void l2_cache_disable(void);
+void set_section_dcache(int section, enum dcache_option option);
 
 /*
  * The current upper bound for ARM L1 data cache line sizes is 64 bytes.  We
Signature · 1 line
1.7.9.5
#13 lore raw

cover ARM: mmu: Set domain permissions to client access - build warnings!

In reply to Sricharan R

Hi Sricharan,

Quoted 53 lines · Sricharan R

Hi Albert,

Quoted 50 lines · Albert ARIBAUD
Quoted 27 lines · Albert ARIBAUD

(sorry for any duplicate of this mail)

Hi R Sricharan,

Quoted 20 lines · R Sricharan

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by
<Vincent Stehle<[email protected]>

[1] http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

Quoted 8 lines
Quoted 7 lines
  arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
  arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
  arch/arm/include/asm/cache.h                   |    1 +
  arch/arm/include/asm/system.h                  |   14 ++++++++++
  arch/arm/lib/cache-cp15.c                      |   13 ++++++++-
  5 files changed, 65 insertions(+), 1 deletion(-)

Series applied to u-boot-arm/master, thanks!

Correction: I'd wrongly inferred from the testing done on OMAP5 that the series would build fine. However it appears that for targets omap4_sdp4430, omap4_panda and omap5_evm, patch 2/2 has warnings:

hwinit-common.c: In function 'dram_bank_mmu_setup':
hwinit-common.c:279:3: warning: implicit declaration of function 'set_section_dcache' [-Wimplicit-function-declaration]

Luckily, I had not committed to the git repo yet.

R Sicharan, can you please check and fix this warning?

Quoted 7 lines
Quoted 1 line

Amicalement,

Oh, thats because as i mentioned above, it depends on the below patch from vincent .

   http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

I did not repost it though. I will do it now for clarity.

Ok -- Vincent's patch was not in my Patchwork TODO and I had read patches 1/2 and 2/2 in Patchwork, but I had not read the cover letter thoroughly enough so I had missed the dependency.

(Cc:ing Vincent and Tom)

Vincent's patch series is delegated to Tom and marked "superseded" in Patchwork but I don't see any later, superseding, series. What can/should I make of it?

Quoted 2 lines

Regards,
Sricharan

Amicalement,

Signature · 1 line
Albert.
#14 lore raw

cover ARM: mmu: Set domain permissions to client access - build warnings!

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 39 lines · Albert ARIBAUD

Hi Sricharan,

Quoted 34 lines · Sricharan R

Hi Albert,

Quoted 31 lines · Albert ARIBAUD
Quoted 28 lines · Albert ARIBAUD

(sorry for any duplicate of this mail)

Hi R Sricharan,

Quoted 21 lines · R Sricharan

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by <Vincent Stehle<[email protected]>

[1]
http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):

Quoted 38 lines
Quoted 26 lines
Quoted 26 lines
Quoted 11 lines
Quoted 11 lines

ARM: mmu: Introduce weak dram_bank_setup function ARM: mmu:
Set domain permissions to client access

arch/arm/cpu/armv7/cache_v7.c | 3 ++
arch/arm/cpu/armv7/omap-common/hwinit-common.c | 35
++++++++++++++++++++++++ arch/arm/include/asm/cache.h
| 1 + arch/arm/include/asm/system.h |
14 ++++++++++ arch/arm/lib/cache-cp15.c
| 13 ++++++++- 5 files changed, 65 insertions(+), 1
deletion(-)

Quoted 1 line

Series applied to u-boot-arm/master, thanks!

Correction: I'd wrongly inferred from the testing done on OMAP5 that the series would build fine. However it appears that for targets omap4_sdp4430, omap4_panda and omap5_evm, patch 2/2 has warnings:

hwinit-common.c: In function 'dram_bank_mmu_setup':
hwinit-common.c:279:3: warning: implicit declaration of function 'set_section_dcache' [-Wimplicit-function-declaration]

Luckily, I had not committed to the git repo yet.

R Sicharan, can you please check and fix this warning?

Quoted 7 lines
Quoted 1 line

Amicalement,

Oh, thats because as i mentioned above, it depends on the below patch from vincent .

http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

I did not repost it though. I will do it now for clarity.

Ok -- Vincent's patch was not in my Patchwork TODO and I had read patches 1/2 and 2/2 in Patchwork, but I had not read the cover letter thoroughly enough so I had missed the dependency.

Quoted 5 lines

(Cc:ing Vincent and Tom)

Vincent's patch series is delegated to Tom and marked "superseded" in Patchwork but I don't see any later, superseding, series. What can/should I make of it?

It's possible I mis-read things and it shouldn't be superseded. Can you list all of the patches that Albert should apply please? Thanks!

- --
Tom
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.11 (GNU/Linux)
Comment: Using GnuPG with Thunderbird - http://www.enigmail.net/

iQIcBAEBAgAGBQJRMLl9AAoJENk4IS6UOR1WPX4P+wRRwsSgPm5edbbBN/TkBi6a x1LIMEKYC2QN73ea8czz4oeOR76MXidQUdYRmimJ2qzYRaEq9Y9ulFjIs0zc/6Gu ydnO4eeb8keU27ItlmKubnEOwsev/wtZNIQhvwJJNvuSiG9rBvKAGDh9L8UQrq3t cwhqWpNhL4isoYG/rB2BDHKrwB62DboOzmaNHACpVaCZNg606QH+eHnget1CWBCD p2O88TSkOVoQHVAj76mdgm8+9WE/mlqn5h4RUU81U3DHiAuTDg0N+J1IJPd3zjhP 86zvj1q4i3JuZySg+IdxouiPTqnw88ZYJ2XHtQvAeDGr5Q680Gb1k18vMIRGlNtv AVx9ipQHaOFXSbTQzlmTNT4EutseaGVa5Ioj3vUkRozBONIfuDmEZAnE+YXFfgQV OQZzidbWLOSFbfDPSi6XMUgtXVEtUPP6eGPV0sv5HUvfYr+3zUMtFwdGqe47X1lC z5V1xpug2mOvU0+6l1nO1egxdRM1JlHroV33oRBXMFoUZUowpobGWeJubjw7WB86 39qoR4xwnyPup3+HHlKVBXmRg6j7jayOUdU0o69kJYO7M0QB/CjADePpSDTrcb9n lkk5KSYVQRzbOar7n9J6jiYoL5WFdj7uSCLpfQ5u/b26r/U2I/YGWPj/w6oGR9+v +3tp/XeRDuXUQkplQEd7
=jW2y
-----END PGP SIGNATURE-----

#15 lore raw

cover ARM: mmu: Set domain permissions to client access - build warnings!

Hi Tom/Albert,
________________________________________
From: Rini, Tom
Sent: Friday, March 01, 2013 7:51 PM
To: Albert ARIBAUD
Cc: R, Sricharan; U-Boot; Stehle, Vincent
Subject: Re: [U-Boot] [PATCH 0/2] ARM: mmu: Set domain permissions to client access - build warnings!

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 39 lines · Albert ARIBAUD

Hi Sricharan,

Quoted 34 lines · Sricharan R

Hi Albert,

Quoted 31 lines · Albert ARIBAUD
Quoted 28 lines · Albert ARIBAUD

(sorry for any duplicate of this mail)

Hi R Sricharan,

Quoted 21 lines · R Sricharan

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by <Vincent Stehle<[email protected]>

[1]
http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):

Quoted 38 lines
Quoted 26 lines
Quoted 26 lines
Quoted 11 lines
Quoted 11 lines

ARM: mmu: Introduce weak dram_bank_setup function ARM: mmu:
Set domain permissions to client access

arch/arm/cpu/armv7/cache_v7.c | 3 ++
arch/arm/cpu/armv7/omap-common/hwinit-common.c | 35
++++++++++++++++++++++++ arch/arm/include/asm/cache.h
| 1 + arch/arm/include/asm/system.h |
14 ++++++++++ arch/arm/lib/cache-cp15.c
| 13 ++++++++- 5 files changed, 65 insertions(+), 1
deletion(-)

Quoted 1 line

Series applied to u-boot-arm/master, thanks!

Correction: I'd wrongly inferred from the testing done on OMAP5 that the series would build fine. However it appears that for targets omap4_sdp4430, omap4_panda and omap5_evm, patch 2/2 has warnings:

hwinit-common.c: In function 'dram_bank_mmu_setup':
hwinit-common.c:279:3: warning: implicit declaration of function 'set_section_dcache' [-Wimplicit-function-declaration]

Luckily, I had not committed to the git repo yet.

R Sicharan, can you please check and fix this warning?

Quoted 7 lines
Quoted 1 line

Amicalement,

Oh, thats because as i mentioned above, it depends on the below patch from vincent .

http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

I did not repost it though. I will do it now for clarity.

Ok -- Vincent's patch was not in my Patchwork TODO and I had read patches 1/2 and 2/2 in Patchwork, but I had not read the cover letter thoroughly enough so I had missed the dependency.

Quoted 8 lines

(Cc:ing Vincent and Tom)

Vincent's patch series is delegated to Tom and marked "superseded" in Patchwork but I don't see any later, superseding, series. What can/should I make of it?

It's possible I mis-read things and it shouldn't be superseded. Can you list all of the patches that Albert should apply please? Thanks!

Sorry, i should have mentioned this clearly previously.

I reposted it again. (ie) only the below patch from Vincent should be applied first.

    http://www.mail-archive.com/u-boot at lists.denx.de/msg107190.html

Apologies for the confusion..

Regards,
Sricharan

#16 lore raw

cover ARM: mmu: Set domain permissions to client access - build warnings!

Hi Sricharan,

Quoted 99 lines · Sricharan

Hi Tom/Albert,
________________________________________
From: Rini, Tom
Sent: Friday, March 01, 2013 7:51 PM
To: Albert ARIBAUD
Cc: R, Sricharan; U-Boot; Stehle, Vincent
Subject: Re: [U-Boot] [PATCH 0/2] ARM: mmu: Set domain permissions to client access - build warnings!

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Quoted 39 lines · Albert ARIBAUD

Hi Sricharan,

Quoted 34 lines · Sricharan R

Hi Albert,

Quoted 31 lines · Albert ARIBAUD
Quoted 28 lines · Albert ARIBAUD

(sorry for any duplicate of this mail)

Hi R Sricharan,

Quoted 21 lines · R Sricharan

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This series depends on [1] the patch sent by <Vincent Stehle<[email protected]>

[1]
http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

R Sricharan (2):

Quoted 38 lines
Quoted 26 lines
Quoted 26 lines
Quoted 11 lines
Quoted 11 lines

ARM: mmu: Introduce weak dram_bank_setup function ARM: mmu:
Set domain permissions to client access

arch/arm/cpu/armv7/cache_v7.c | 3 ++
arch/arm/cpu/armv7/omap-common/hwinit-common.c | 35
++++++++++++++++++++++++ arch/arm/include/asm/cache.h
| 1 + arch/arm/include/asm/system.h |
14 ++++++++++ arch/arm/lib/cache-cp15.c
| 13 ++++++++- 5 files changed, 65 insertions(+), 1
deletion(-)

Quoted 1 line

Series applied to u-boot-arm/master, thanks!

Correction: I'd wrongly inferred from the testing done on OMAP5 that the series would build fine. However it appears that for targets omap4_sdp4430, omap4_panda and omap5_evm, patch 2/2 has warnings:

hwinit-common.c: In function 'dram_bank_mmu_setup':
hwinit-common.c:279:3: warning: implicit declaration of function 'set_section_dcache' [-Wimplicit-function-declaration]

Luckily, I had not committed to the git repo yet.

R Sicharan, can you please check and fix this warning?

Quoted 7 lines
Quoted 1 line

Amicalement,

Oh, thats because as i mentioned above, it depends on the below patch from vincent .

http://www.mail-archive.com/u-boot at lists.denx.de/msg102709.html

I did not repost it though. I will do it now for clarity.

Ok -- Vincent's patch was not in my Patchwork TODO and I had read patches 1/2 and 2/2 in Patchwork, but I had not read the cover letter thoroughly enough so I had missed the dependency.

Quoted 8 lines

(Cc:ing Vincent and Tom)

Vincent's patch series is delegated to Tom and marked "superseded" in Patchwork but I don't see any later, superseding, series. What can/should I make of it?

It's possible I mis-read things and it shouldn't be superseded. Can you list all of the patches that Albert should apply please? Thanks!

Quoted 6 lines

Sorry, i should have mentioned this clearly previously.

I reposted it again. (ie) only the below patch from Vincent should be applied first.

    http://www.mail-archive.com/u-boot at lists.denx.de/msg107190.html

Apologies for the confusion..

Just to be clear: to apply Sricharan's series, patch 1/3 of Vincent's series is reqired, but I think Vincent's series should be applied in one go, not patch by patch. Basically, this means we need Vincent's series to be applied first, then we can apply Sricharan's.

Quoted 2 lines

Regards,
Sricharan

Amicalement,

Signature · 1 line
Albert.
#17 lore raw

cover ARM: mmu: Set domain permissions to client access - build warnings!

Quoted 2 lines · Albert ARIBAUD

(..) Basically, this means we need Vincent's series to be applied first, then we can apply Sricharan's.

Hi,

I think this is too much trouble for a "one liner". Please feel free to squash.

Best regards,

V.

#18 lore raw

cover ARM: mmu: Set domain permissions to client access

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This is a repost of the older series to include Vincent's patch in the same one.

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

Vincent Stehl? (1):
ARM: cache: declare set_section_dcache

 arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
 arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
 arch/arm/include/asm/cache.h                   |    2 ++
 arch/arm/include/asm/system.h                  |   14 ++++++++++
 arch/arm/lib/cache-cp15.c                      |   11 +++++++-
 5 files changed, 64 insertions(+), 1 deletion(-)
Signature · 1 line
1.7.9.5
#19 lore raw

PATCH RESEND 1/3 ARM: cache: declare set_section_dcache

From: Vincent Stehl? <[email protected]>

We declare the set_section_dcache function globally in the cache header, for later use by e.g. machine specific code.

Signed-off-by Vincent Stehl? <v-stehle <at> ti.com>
Cc Tom Rini <trini <at> ti.com>
Cc Albert ARIBAUD <[email protected]>
---
 arch/arm/include/asm/cache.h |    1 +
 1 file changed, 1 insertion(+)
arch/arm/include/asm/cache.h +1 -0
index eef6a5a..416d2c8 100644
@@ -41,6 +41,7 @@ static inline void invalidate_l2_cache(void)
 
 void l2_cache_enable(void);
 void l2_cache_disable(void);
+void set_section_dcache(int section, enum dcache_option option);
 
 /*
  * The current upper bound for ARM L1 data cache line sizes is 64 bytes.  We
Signature · 1 line
1.7.9.5
#20 lore raw

PATCH RESEND 2/3 ARM: mmu: Introduce weak dram_bank_setup function

In reply to Sricharan R

From: R Sricharan <[email protected]>

Introduce a weak version of dram_bank_setup function to allow a platform specific function.

This is used in the subsequent patch to setup dram region without 'XN' attribute in order to enable the region under client permissions.

Cc Vincent Stehle <[email protected]>
Cc Tom Rini <[email protected]>
Cc Albert ARIBAUD <[email protected]>
---
 arch/arm/include/asm/cache.h |    1 +
 arch/arm/lib/cache-cp15.c    |    4 +++-
 2 files changed, 4 insertions(+), 1 deletion(-)
arch/arm/include/asm/cache.h +1 -0
index 416d2c8..8153484 100644
@@ -43,6 +43,7 @@ void l2_cache_enable(void);
 void l2_cache_disable(void);
 void set_section_dcache(int section, enum dcache_option option);
 
+void dram_bank_mmu_setup(int bank);
 /*
  * The current upper bound for ARM L1 data cache line sizes is 64 bytes.  We
  * use that value for aligning DMA buffers unless the board config has specified
arch/arm/lib/cache-cp15.c +3 -1
index b6e5e95..6ecbedf 100644
@@ -23,6 +23,8 @@
 
 #include <common.h>
 #include <asm/system.h>
+#include <asm/cache.h>
+#include <linux/compiler.h>
 
 #if !(defined(CONFIG_SYS_ICACHE_OFF) && defined(CONFIG_SYS_DCACHE_OFF))
 
@@ -77,7 +79,7 @@ void mmu_set_region_dcache_behaviour(u32 start, int size,
 	mmu_page_table_flush((u32)&page_table[start], (u32)&page_table[end]);
 }
 
-static inline void dram_bank_mmu_setup(int bank)
+__weak void dram_bank_mmu_setup(int bank)
 {
 	bd_t *bd = gd->bd;
 	int	i;
Signature · 1 line
1.7.9.5
#21 lore raw

PATCH RESEND 3/3 ARM: mmu: Set domain permissions to client access

In reply to Sricharan R

From: R Sricharan <[email protected]>

The 'XN' execute never bit is set in the pagetables. This will prevent speculative prefetches to non executable regions. But the domain permissions are set as master in the DACR register. So the pagetable attribute for 'XN' is not effective. Change the permissions to client.

This fixes lot of speculative prefetch aborts seen on OMAP5 secure devices.

---
 arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
 arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
 arch/arm/include/asm/system.h                  |   14 ++++++++++
 arch/arm/lib/cache-cp15.c                      |    7 +++++
 4 files changed, 59 insertions(+)
arch/arm/cpu/armv7/cache_v7.c +3 -0
index 5f6d039..8748c14 100644
@@ -340,6 +340,9 @@ void mmu_page_table_flush(unsigned long start, unsigned long stop)
 {
 }
 
+void arm_init_domains(void)
+{
+}
 #endif /* #ifndef CONFIG_SYS_DCACHE_OFF */
 
 #ifndef CONFIG_SYS_ICACHE_OFF
arch/arm/cpu/armv7/omap-common/hwinit-common.c +35 -0
index 9ef10bd..4eaf75b 100644
@@ -32,6 +32,12 @@
 #include <asm/arch/sys_proto.h>
 #include <asm/sizes.h>
 #include <asm/emif.h>
+#include <asm/cache.h>
+#include <asm/system.h>
+
+#define ARMV7_DCACHE_WRITEBACK  0xe
+#define	ARMV7_DOMAIN_CLIENT	1
+#define ARMV7_DOMAIN_MASK	(0x3 << 0)
 
 DECLARE_GLOBAL_DATA_PTR;
 
@@ -258,4 +264,33 @@ void enable_caches(void)
 	/* Enable D-cache. I-cache is already enabled in start.S */
 	dcache_enable();
 }
+
+void dram_bank_mmu_setup(int bank)
+{
+	bd_t *bd = gd->bd;
+	int	i;
+
+	u32 start = bd->bi_dram[bank].start >> 20;
+	u32 size = bd->bi_dram[bank].size >> 20;
+	u32 end = start + size;
+
+	debug("%s: bank: %d\n", __func__, bank);
+	for (i = start; i < end; i++)
+		set_section_dcache(i, ARMV7_DCACHE_WRITEBACK);
+
+}
+
+void arm_init_domains(void)
+{
+	u32 reg;
+
+	reg = get_dacr();
+	/*
+	* Set DOMAIN to client access so that all permissions
+	* set in pagetables are validated by the mmu.
+	*/
+	reg &= ~ARMV7_DOMAIN_MASK;
+	reg |= ARMV7_DOMAIN_CLIENT;
+	set_dacr(reg);
+}
 #endif
arch/arm/include/asm/system.h +14 -0
index 1918492..760345f 100644
@@ -81,6 +81,20 @@ static inline void set_cr(unsigned int val)
 	isb();
 }
 
+static inline unsigned int get_dacr(void)
+{
+	unsigned int val;
+	asm("mrc p15, 0, %0, c3, c0, 0	@ get DACR" : "=r" (val) : : "cc");
+	return val;
+}
+
+static inline void set_dacr(unsigned int val)
+{
+	asm volatile("mcr p15, 0, %0, c3, c0, 0	@ set DACR"
+	  : : "r" (val) : "cc");
+	isb();
+}
+
 /* options available for data cache on each page */
 enum dcache_option {
 	DCACHE_OFF = 0x12,
arch/arm/lib/cache-cp15.c +7 -0
index 6ecbedf..4abe1cf 100644
@@ -36,6 +36,10 @@ void __arm_init_before_mmu(void)
 void arm_init_before_mmu(void)
 	__attribute__((weak, alias("__arm_init_before_mmu")));
 
+__weak void arm_init_domains(void)
+{
+}
+
 static void cp_delay (void)
 {
 	volatile int i;
@@ -117,6 +121,9 @@ static inline void mmu_setup(void)
 	/* Set the access control to all-supervisor */
 	asm volatile("mcr p15, 0, %0, c3, c0, 0"
 		     : : "r" (~0));
+
+	arm_init_domains();
+
 	/* and enable the mmu */
 	reg = get_cr();	/* get control reg. */
 	cp_delay();
Signature · 1 line
1.7.9.5
#22 lore raw

cover ARM: mmu: Set domain permissions to client access

In reply to Sricharan R

Hi Albert,

Quoted 23 lines · Sricharan R

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This is a repost of the older series to include Vincent's patch in the same one.

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

Vincent Stehl? (1):
ARM: cache: declare set_section_dcache

Quoted 7 lines
 arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
 arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
 arch/arm/include/asm/cache.h                   |    2 ++
 arch/arm/include/asm/system.h                  |   14 ++++++++++
 arch/arm/lib/cache-cp15.c                      |   11 +++++++-
 5 files changed, 64 insertions(+), 1 deletion(-)

Is this ok to get in now ?

Regards,
Sricharan

#23 lore raw

cover ARM: mmu: Set domain permissions to client access

In reply to Sricharan R

Hi Sricharan,

Quoted 23 lines · Sricharan R

Currently for ARM based cpu's, mmu pagetable attributes are set with manager permissions for all 4GB address space. Because of this the 'execute never (XN)' permission is never checked on read sensitive regions which results in speculative aborts.

This series changes the domain permissions of the full 4GB space to client access for OMAP socs. This avoids all the speculative aborts that are currently seen on OMAP5 secure devices.

Tested on OMAP5 SDP (HS) soc.

This is a repost of the older series to include Vincent's patch in the same one.

R Sricharan (2):
ARM: mmu: Introduce weak dram_bank_setup function
ARM: mmu: Set domain permissions to client access

Vincent Stehl? (1):
ARM: cache: declare set_section_dcache

Quoted 7 lines
 arch/arm/cpu/armv7/cache_v7.c                  |    3 ++
 arch/arm/cpu/armv7/omap-common/hwinit-common.c |   35 ++++++++++++++++++++++++
 arch/arm/include/asm/cache.h                   |    2 ++
 arch/arm/include/asm/system.h                  |   14 ++++++++++
 arch/arm/lib/cache-cp15.c                      |   11 +++++++-
 5 files changed, 64 insertions(+), 1 deletion(-)

Series applied to u-boot-arm/master, thanks!

Amicalement,

Signature · 1 line
Albert.
#24 lore raw

cover ARM: mmu: Set domain permissions to client access - build warnings!

In reply to Vincent Stehlé

Hi,

Quoted 12 lines · Vincent Stehl?
Quoted 2 lines · Albert ARIBAUD

(..) Basically, this means we need Vincent's series to be applied first, then we can apply Sricharan's.

Hi,

I think this is too much trouble for a "one liner". Please feel free to squash.

Best regards,

V.

Ok, i reposted the series again after squashing.

http://www.mail-archive.com/u-boot at lists.denx.de/msg107487.html

Regards,
Sricharan