[PATCH 1/2] arm: omap: sram: fix OMAP4 errata handling

[PATCH 1/2] arm: omap: sram: fix OMAP4 errata handling

[Aaro Koskinen]

OMAP4-specific code should be executed only if we are running on
OMAP4. Otherwise it may break multi-OMAP kernels. Found by reading
the code.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
---
 arch/arm/plat-omap/sram.c |    2 ++
 1 files changed, 2 insertions(+), 0 deletions(-)

diff --git a/arch/arm/plat-omap/sram.c b/arch/arm/plat-omap/sram.c
index 766181c..85c23db 100644
--- a/arch/arm/plat-omap/sram.c
+++ b/arch/arm/plat-omap/sram.c
@@ -175,8 +175,10 @@ static void __init omap_map_sram(void)
 		return;
 
 #ifdef CONFIG_OMAP4_ERRATA_I688
+	if (cpu_is_omap44xx()) {
 		omap_sram_start += PAGE_SIZE;
 		omap_sram_size -= SZ_16K;
+	}
 #endif
 	if (cpu_is_omap34xx()) {
 		/*
-- 
1.7.2.5

[PATCH 2/2] arm: omap: sram: skip the first 16K on OMAP3 HS

[Aaro Koskinen]

In some OMAP3 HS devices (at least Nokia N9 and N950), the public SRAM
seems to conflict with secure portition of SRAM. When booting the 3.6-rc3
kernel (and also earlier) on these devices, the kernel gets tainted with
tons of the following warnings:

	[    6.894348] In-band Error seen by MPU  at address 0
	[...]
	[    6.894378] WARNING: at arch/arm/mach-omap2/omap_l3_smx.c:162

Fix this by skipping the first 16K of the public SRAM. (Note that the
mapping could not be changed, as it resulted in secure monitor call
failure in save_secure_sram().)

This will leave 12K SRAM available that should be still sufficient. The
patch has been boot tested with vanilla 3.6-rc3 on N900, N950 and N9.

Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
---
 arch/arm/plat-omap/sram.c |    9 ++++++---
 1 files changed, 6 insertions(+), 3 deletions(-)

diff --git a/arch/arm/plat-omap/sram.c b/arch/arm/plat-omap/sram.c
index 85c23db..024f3b0 100644
--- a/arch/arm/plat-omap/sram.c
+++ b/arch/arm/plat-omap/sram.c
@@ -68,6 +68,7 @@
 
 static unsigned long omap_sram_start;
 static void __iomem *omap_sram_base;
+static unsigned long omap_sram_skip;
 static unsigned long omap_sram_size;
 static void __iomem *omap_sram_ceil;
 
@@ -106,6 +107,7 @@ static int is_sram_locked(void)
  */
 static void __init omap_detect_sram(void)
 {
+	omap_sram_skip = SRAM_BOOTLOADER_SZ;
 	if (cpu_class_is_omap2()) {
 		if (is_sram_locked()) {
 			if (cpu_is_omap34xx()) {
@@ -113,6 +115,7 @@ static void __init omap_detect_sram(void)
 				if ((omap_type() == OMAP2_DEVICE_TYPE_EMU) ||
 				    (omap_type() == OMAP2_DEVICE_TYPE_SEC)) {
 					omap_sram_size = 0x7000; /* 28K */
+					omap_sram_skip += SZ_16K;
 				} else {
 					omap_sram_size = 0x8000; /* 32K */
 				}
@@ -205,8 +208,8 @@ static void __init omap_map_sram(void)
 	 * Looks like we need to preserve some bootloader code at the
 	 * beginning of SRAM for jumping to flash for reboot to work...
 	 */
-	memset_io(omap_sram_base + SRAM_BOOTLOADER_SZ, 0,
-		  omap_sram_size - SRAM_BOOTLOADER_SZ);
+	memset_io(omap_sram_base + omap_sram_skip, 0,
+		  omap_sram_size - omap_sram_skip);
 }
 
 /*
@@ -220,7 +223,7 @@ void *omap_sram_push_address(unsigned long size)
 {
 	unsigned long available, new_ceil = (unsigned long)omap_sram_ceil;
 
-	available = omap_sram_ceil - (omap_sram_base + SRAM_BOOTLOADER_SZ);
+	available = omap_sram_ceil - (omap_sram_base + omap_sram_skip);
 
 	if (size > available) {
 		pr_err("Not enough space in SRAM\n");
-- 
1.7.2.5

Re: [PATCH 1/2] arm: omap: sram: fix OMAP4 errata handling

[Shilimkar, Santosh]

On Wed, Aug 29, 2012 at 8:24 AM, Aaro Koskinen <aaro.koskinen@iki.fi> wrote:
> OMAP4-specific code should be executed only if we are running on
> OMAP4. Otherwise it may break multi-OMAP kernels. Found by reading
> the code.
>
> Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
> ---
Looks good.

Acked-by: Santosh Shilimkar <santosh.shilimkar@ti.com>

Re: [PATCH 2/2] arm: omap: sram: skip the first 16K on OMAP3 HS

[Shilimkar, Santosh]

On Wed, Aug 29, 2012 at 8:24 AM, Aaro Koskinen <aaro.koskinen@iki.fi> wrote:
> In some OMAP3 HS devices (at least Nokia N9 and N950), the public SRAM
> seems to conflict with secure portition of SRAM. When booting the 3.6-rc3
> kernel (and also earlier) on these devices, the kernel gets tainted with
> tons of the following warnings:
>
>         [    6.894348] In-band Error seen by MPU  at address 0
>         [...]
>         [    6.894378] WARNING: at arch/arm/mach-omap2/omap_l3_smx.c:162
>
> Fix this by skipping the first 16K of the public SRAM. (Note that the
> mapping could not be changed, as it resulted in secure monitor call
> failure in save_secure_sram().)
>
> This will leave 12K SRAM available that should be still sufficient. The
> patch has been boot tested with vanilla 3.6-rc3 on N900, N950 and N9.
>
> Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
> ---
>  arch/arm/plat-omap/sram.c |    9 ++++++---
>  1 files changed, 6 insertions(+), 3 deletions(-)
>
Looks good.

Acked-by: Santosh Shilimkar <santosh.shilimkar@ti.com>

Re: [PATCH 2/2] arm: omap: sram: skip the first 16K on OMAP3 HS

[Tony Lindgren]

* Shilimkar, Santosh <santosh.shilimkar@ti.com> [120829 08:30]:
> On Wed, Aug 29, 2012 at 8:24 AM, Aaro Koskinen <aaro.koskinen@iki.fi> wrote:
> > In some OMAP3 HS devices (at least Nokia N9 and N950), the public SRAM
> > seems to conflict with secure portition of SRAM. When booting the 3.6-rc3
> > kernel (and also earlier) on these devices, the kernel gets tainted with
> > tons of the following warnings:
> >
> >         [    6.894348] In-band Error seen by MPU  at address 0
> >         [...]
> >         [    6.894378] WARNING: at arch/arm/mach-omap2/omap_l3_smx.c:162
> >
> > Fix this by skipping the first 16K of the public SRAM. (Note that the
> > mapping could not be changed, as it resulted in secure monitor call
> > failure in save_secure_sram().)
> >
> > This will leave 12K SRAM available that should be still sufficient. The
> > patch has been boot tested with vanilla 3.6-rc3 on N900, N950 and N9.
> >
> > Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
> > ---
> >  arch/arm/plat-omap/sram.c |    9 ++++++---
> >  1 files changed, 6 insertions(+), 3 deletions(-)
> >
> Looks good.
> 
> Acked-by: Santosh Shilimkar <santosh.shilimkar@ti.com>

Thanks applying both into fixes. Aaro, note that now that we're merging
patches via the arm soc tree, let's use "ARM: OMAP: " for the subject so
I don't need to fix it up ;)

Tony

Re: [PATCH 2/2] arm: omap: sram: skip the first 16K on OMAP3 HS

[Jarkko Nikula]

Hi Aaro

On Thu, 30 Aug 2012 13:40:30 -0700
Tony Lindgren <tony@atomide.com> wrote:

> * Shilimkar, Santosh <santosh.shilimkar@ti.com> [120829 08:30]:
> > On Wed, Aug 29, 2012 at 8:24 AM, Aaro Koskinen <aaro.koskinen@iki.fi> wrote:
> > > In some OMAP3 HS devices (at least Nokia N9 and N950), the public SRAM
> > > seems to conflict with secure portition of SRAM. When booting the 3.6-rc3
> > > kernel (and also earlier) on these devices, the kernel gets tainted with
> > > tons of the following warnings:
> > >
> > >         [    6.894348] In-band Error seen by MPU  at address 0
> > >         [...]
> > >         [    6.894378] WARNING: at arch/arm/mach-omap2/omap_l3_smx.c:162
> > >
> > > Fix this by skipping the first 16K of the public SRAM. (Note that the
> > > mapping could not be changed, as it resulted in secure monitor call
> > > failure in save_secure_sram().)
> > >
> > > This will leave 12K SRAM available that should be still sufficient. The
> > > patch has been boot tested with vanilla 3.6-rc3 on N900, N950 and N9.
> > >
> > > Signed-off-by: Aaro Koskinen <aaro.koskinen@iki.fi>
> > > ---
> > >  arch/arm/plat-omap/sram.c |    9 ++++++---
> > >  1 files changed, 6 insertions(+), 3 deletions(-)
> > >
> > Looks good.
> > 
> > Acked-by: Santosh Shilimkar <santosh.shilimkar@ti.com>
> 
> Thanks applying both into fixes. Aaro, note that now that we're merging
> patches via the arm soc tree, let's use "ARM: OMAP: " for the subject so
> I don't need to fix it up ;)
> 
This issue appears to be in 3.5 too (I didn't check others). Should you commit id of this patch and affected kernel versions to stable?

-- 
Jarkko