[U-Boot] ti, am3517: errata 430973 workaround

[U-Boot] ti, am3517: errata 430973 workaround

[Jeroen Hofstee]

Hi,

A while ago [1], a RFC was posted to disable workaround for
besides others, errata 430973. It is a bit unclear to me which
revision actually need this workaround, but as suggested in
[2] also enabling this workaround in Linux seem to make some
weird problems go away in linux (signal 4, bad instruction,
11 segfaults etc).

As said, I am a bit in doubt why this works. The board in question
is a tam3517 derived one:

cat /proc/cpuinfo
Processor    : ARMv7 Processor rev 7 (v7l)
BogoMIPS    : 397.57
Features    : swp half thumb fastmult vfp edsp neon vfpv3 tls
CPU implementer    : 0x41
CPU architecture: 7
CPU variant    : 0x1
CPU part    : 0xc08
CPU revision    : 7

Which makes this a r1p7 I assume, and hence the workaround
of linux, CONFIG_ARM_ERRATA_430973, "This option enables the
workaround for the 430973 Cortex-A8 (r1p0..r1p2) erratum",
should not be needed it seems.

On the other hand Andreas Bie?man, wrote at [3]
"I have rev 20.0 from 13-Apr-10. The three mentioned errata
should be fixed in r2p1." note, this mentions r2p1 not r1p2!

Since I don't have access to "ARM Core Cortex-A8 (AT400/AT401)
errata", I cannot look this up. Hence the question, is u-boot
wrong by enabling this workaround for a r1p7 revision or is the
comment in the kernel flawed? (or am I missing something else..)

If someone could shed some light on this it would be appreciated.

Regards,
Jeroen

[1] http://lists.denx.de/pipermail/u-boot/2013-July/158377.html
[2] http://lists.denx.de/pipermail/u-boot/2013-July/158404.html
[3] http://lists.denx.de/pipermail/u-boot/2013-July/158386.html

[U-Boot] ti, am3517: errata 430973 workaround

[Andreas Bießmann]

Hi,

On 12/08/2014 11:19 PM, Jeroen Hofstee wrote:
> Hi,
> 
> A while ago [1], a RFC was posted to disable workaround for
> besides others, errata 430973. It is a bit unclear to me which
> revision actually need this workaround, but as suggested in
> [2] also enabling this workaround in Linux seem to make some
> weird problems go away in linux (signal 4, bad instruction,
> 11 segfaults etc).
> 
> As said, I am a bit in doubt why this works. The board in question
> is a tam3517 derived one:
> 
> cat /proc/cpuinfo
> Processor    : ARMv7 Processor rev 7 (v7l)
> BogoMIPS    : 397.57
> Features    : swp half thumb fastmult vfp edsp neon vfpv3 tls
> CPU implementer    : 0x41
> CPU architecture: 7
> CPU variant    : 0x1
> CPU part    : 0xc08
> CPU revision    : 7
> 
> Which makes this a r1p7 I assume, and hence the workaround
> of linux, CONFIG_ARM_ERRATA_430973, "This option enables the
> workaround for the 430973 Cortex-A8 (r1p0..r1p2) erratum",
> should not be needed it seems.
> 
> On the other hand Andreas Bie?man, wrote at [3]
> "I have rev 20.0 from 13-Apr-10. The three mentioned errata
> should be fixed in r2p1." note, this mentions r2p1 not r1p2!
> 
> Since I don't have access to "ARM Core Cortex-A8 (AT400/AT401)
> errata", I cannot look this up. Hence the question, is u-boot
> wrong by enabling this workaround for a r1p7 revision or is the
> comment in the kernel flawed? (or am I missing something else..)

my AT400/AT401/AT490 r20.0 (13-Apr-10) says #430973 affects r1p1,m r1p2,
r1p3, r1p7 but none since r2p1. So it seems the kernel is wrong here.

Best regards

Andreas Bie?mann

> [1] http://lists.denx.de/pipermail/u-boot/2013-July/158377.html
> [2] http://lists.denx.de/pipermail/u-boot/2013-July/158404.html
> [3] http://lists.denx.de/pipermail/u-boot/2013-July/158386.html

[U-Boot] ti, am3517: errata 430973 workaround

[Robert Nelson]

On Mon, Dec 8, 2014 at 4:19 PM, Jeroen Hofstee <jeroen@myspectrum.nl> wrote:
> Hi,
>
> A while ago [1], a RFC was posted to disable workaround for
> besides others, errata 430973. It is a bit unclear to me which
> revision actually need this workaround, but as suggested in
> [2] also enabling this workaround in Linux seem to make some
> weird problems go away in linux (signal 4, bad instruction,
> 11 segfaults etc).
>
> As said, I am a bit in doubt why this works. The board in question
> is a tam3517 derived one:
>
> cat /proc/cpuinfo
> Processor    : ARMv7 Processor rev 7 (v7l)
> BogoMIPS    : 397.57
> Features    : swp half thumb fastmult vfp edsp neon vfpv3 tls
> CPU implementer    : 0x41
> CPU architecture: 7
> CPU variant    : 0x1
> CPU part    : 0xc08
> CPU revision    : 7
>
> Which makes this a r1p7 I assume, and hence the workaround
> of linux, CONFIG_ARM_ERRATA_430973, "This option enables the
> workaround for the 430973 Cortex-A8 (r1p0..r1p2) erratum",
> should not be needed it seems.

Digging thru my old beagle notes 430973 is also needed for "r1p3"
(dm3730/bb-xm), so that config option was never updated since the
errata was first discovered in r1p2 devices..

Fixed in r2p1 sounds about right, as i know for sure it works fine in
'r3p2' (am335x/bbb)

Regards,

-- 
Robert Nelson
http://www.rcn-ee.com/

[U-Boot] ti, am3517: errata 430973 workaround

[Jeroen Hofstee]

Hello Robert, Andreas,

On 09-12-14 22:45, Robert Nelson wrote:
> On Mon, Dec 8, 2014 at 4:19 PM, Jeroen Hofstee <jeroen@myspectrum.nl> wrote:
>> Hi,
>>
>> A while ago [1], a RFC was posted to disable workaround for
>> besides others, errata 430973. It is a bit unclear to me which
>> revision actually need this workaround, but as suggested in
>> [2] also enabling this workaround in Linux seem to make some
>> weird problems go away in linux (signal 4, bad instruction,
>> 11 segfaults etc).
>>
>> As said, I am a bit in doubt why this works. The board in question
>> is a tam3517 derived one:
>>
>> cat /proc/cpuinfo
>> Processor    : ARMv7 Processor rev 7 (v7l)
>> BogoMIPS    : 397.57
>> Features    : swp half thumb fastmult vfp edsp neon vfpv3 tls
>> CPU implementer    : 0x41
>> CPU architecture: 7
>> CPU variant    : 0x1
>> CPU part    : 0xc08
>> CPU revision    : 7
>>
>> Which makes this a r1p7 I assume, and hence the workaround
>> of linux, CONFIG_ARM_ERRATA_430973, "This option enables the
>> workaround for the 430973 Cortex-A8 (r1p0..r1p2) erratum",
>> should not be needed it seems.
> Digging thru my old beagle notes 430973 is also needed for "r1p3"
> (dm3730/bb-xm), so that config option was never updated since the
> errata was first discovered in r1p2 devices..
>
> Fixed in r2p1 sounds about right, as i know for sure it works fine in
> 'r3p2' (am335x/bbb)

Thanks for the clarification. I sent a patch to the linux folks
to update the help text.

Regards,
Jeroen