Samsung Exynos4 ("Orion") doesn't have proper PMU definitions

[Wade Cherry]

This issue was found with Streamline, and later confirmed with perf. arch/arm/plat-s5p/dev-pmu.c is incomplete. Even worse, it pulls old "arm-pmu" device from A8-based SOC (I think) so Streamline/perf think it should work, but it doesn't. The end result is no proper error message for the user and the PMU interrupts are not firing so the data is invalid.

[Scott Bambrough (scottb)]

This has the potential to disrupt PMWG work. Can we fix this please.

[Vinod Krishnamoni (vinod-krishnamoni)]

We are keen to see this fixed as well from ARM, as this breaks the performance analysis tools.

[Angus Ainslie (angus-akkea)]

Inder has found a fix and will post a patch in the next few days

On Thu, Sep 29, 2011 at 7:44 AM, Vinod Krishnamoni
<email address hidden> wrote:
> We are keen to see this fixed as well from ARM, as this breaks the
> performance analysis tools.
>
> --
> You received this bug notification because you are a member of Linaro
> Landing Team: Samsung, which is subscribed to linaro-landing-team-
> samsung.
> https://bugs.launchpad.net/bugs/859825
>
> Title:
>  Samsung Exynos4 ("Orion") doesn't have proper PMU definitions
>
> Status in Linaro Samsung Landing Team Project:
>  Triaged
>
> Bug description:
>  This issue was found with Streamline, and later confirmed with perf.
>  arch/arm/plat-s5p/dev-pmu.c is incomplete. Even worse, it pulls old
>  "arm-pmu" device from A8-based SOC (I think) so Streamline/perf think
>  it should work, but it doesn't. The end result is no proper error
>  message for the user and the PMU interrupts are not firing so the data
>  is invalid.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/linaro-landing-team-samsung/+bug/859825/+subscriptions
>

--
Angus Ainslie <email address hidden>
Team Lead, Samsung Landing Team

[Vinod Krishnamoni (vinod-krishnamoni)]

Thanks Angus, please update the issue as soon the patch is posted so that we can pick that up.

[Angus Ainslie (angus-akkea)]

The linux-image-3.0.0-1000-origen_3.0.0-1000.0samsung15_armel.deb contains a patch for the PMU irq'a and enables tracing and monitoring options.

The changes are available in hwpacks for 20111005 onward

[Vinod Krishnamoni (vinod-krishnamoni)]

Hi Angus,

Please could you point me at the souce tree where the fix was made ? ( We do not have an Origen, we'll need to rebuild it for an Orion platfrom )

Thanks
Vinod

[Angus Ainslie (angus-akkea)]

Hi Vinod,

The patch is here

http://git.linaro.org/gitweb?p=people/angus/linux-linaro-3.0.git;a=commit;h=7ca7902200de17dd572524c8c66a3eba03e4fe34

Angus

On Wed, Oct 5, 2011 at 8:41 AM, Vinod Krishnamoni
<email address hidden> wrote:
> Hi Angus,
>
> Please could you point me at the souce tree where the fix was made ? (
> We do not have an Origen, we'll need to rebuild it for an Orion platfrom
> )
>
> Thanks
> Vinod
>
> --
> You received this bug notification because you are a member of Linaro
> Landing Team: Samsung, which is subscribed to linaro-landing-team-
> samsung.
> https://bugs.launchpad.net/bugs/859825
>
> Title:
>  Samsung Exynos4 ("Orion") doesn't have proper PMU definitions
>
> Status in Linaro Samsung Landing Team Project:
>  Fix Committed
>
> Bug description:
>  This issue was found with Streamline, and later confirmed with perf.
>  arch/arm/plat-s5p/dev-pmu.c is incomplete. Even worse, it pulls old
>  "arm-pmu" device from A8-based SOC (I think) so Streamline/perf think
>  it should work, but it doesn't. The end result is no proper error
>  message for the user and the PMU interrupts are not firing so the data
>  is invalid.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/linaro-landing-team-samsung/+bug/859825/+subscriptions
>

--
Angus Ainslie <email address hidden>
Team Lead, Samsung Landing Team

[Angus Ainslie (angus-akkea)]

Hi Vinod,

Did you get a chance to test the patch ? Does it fox this bug ?

Thanks
Angus

On Wed, Oct 5, 2011 at 9:33 AM, Angus Ainslie <email address hidden> wrote:
> Hi Vinod,
>
> The patch is here
>
> http://git.linaro.org/gitweb?p=people/angus/linux-linaro-3.0.git;a=commit;h=7ca7902200de17dd572524c8c66a3eba03e4fe34
>
> Angus
>
> On Wed, Oct 5, 2011 at 8:41 AM, Vinod Krishnamoni
> <email address hidden> wrote:
>> Hi Angus,
>>
>> Please could you point me at the souce tree where the fix was made ? (
>> We do not have an Origen, we'll need to rebuild it for an Orion platfrom
>> )
>>
>> Thanks
>> Vinod
>>
>> --
>> You received this bug notification because you are a member of Linaro
>> Landing Team: Samsung, which is subscribed to linaro-landing-team-
>> samsung.
>> https://bugs.launchpad.net/bugs/859825
>>
>> Title:
>>  Samsung Exynos4 ("Orion") doesn't have proper PMU definitions
>>
>> Status in Linaro Samsung Landing Team Project:
>>  Fix Committed
>>
>> Bug description:
>>  This issue was found with Streamline, and later confirmed with perf.
>>  arch/arm/plat-s5p/dev-pmu.c is incomplete. Even worse, it pulls old
>>  "arm-pmu" device from A8-based SOC (I think) so Streamline/perf think
>>  it should work, but it doesn't. The end result is no proper error
>>  message for the user and the PMU interrupts are not firing so the data
>>  is invalid.
>>
>> To manage notifications about this bug go to:
>> https://bugs.launchpad.net/linaro-landing-team-samsung/+bug/859825/+subscriptions
>>
>
>
>
> --
> Angus Ainslie <email address hidden>
> Team Lead, Samsung Landing Team
>

--
Angus Ainslie <email address hidden>
Team Lead, Samsung Landing Team

[Vinod Krishnamoni (vinod-krishnamoni)]

Hi Angus,
I have passed this to our development team, waiting to hear from them. Shall let you know as soon as I hear from them.

Thanks
Vinod

[Vinod Krishnamoni (vinod-krishnamoni)]

Update from Wade Cherry, who's working on the defect at ARM:

"After incorporating the changes by Inderpal Singh from the Linaro branch into my code base for Orion, the Orion board has the pmu interrupt for core 0 set to irq 146 and core 1 set to irq 154. However, when calling init_pmu() an error is returned indicating that associating the irq to a certain core failed. Thus, I was not able to verify the changes, though there may be something else at odds here: 1) I'm not using Exynos4 but Orion s5p v310, what are the differences? 2) I did not use the same code base."

[Angus Ainslie (angus-akkea)]

Hi Wade

On Mon, Oct 24, 2011 at 8:19 AM, Vinod Krishnamoni
<email address hidden> wrote:
> Update from Wade Cherry, who's working on the defect at ARM:
>
> "After incorporating the changes by Inderpal Singh from the Linaro
> branch into my code base for Orion, the Orion board has the pmu
> interrupt for core 0 set to irq 146 and core 1 set to irq 154. However,
> when calling init_pmu() an error is returned indicating that associating
> the irq to a certain core failed. Thus, I was not able to verify the
> changes, though there may be something else at odds here: 1) I'm not
> using Exynos4 but Orion s5p v310, what are the differences? 2) I did not
> use the same code base."
>

1) Exynos4(210) and Orion s5pv310 are the same part. However there a
re a few revisions of the chip could you tell us what CPU id is
reported during boot, or better yet add a bootlog to the bug.

2) is the code base you used available somewhere ?

Thanks
Angus

> --
> You received this bug notification because you are a member of Linaro
> Landing Team: Samsung, which is subscribed to linaro-landing-team-
> samsung.
> https://bugs.launchpad.net/bugs/859825
>
> Title:
>  Samsung Exynos4 ("Orion") doesn't have proper PMU definitions
>
> Status in Linaro Samsung Landing Team Project:
>  Fix Committed
>
> Bug description:
>  This issue was found with Streamline, and later confirmed with perf.
>  arch/arm/plat-s5p/dev-pmu.c is incomplete. Even worse, it pulls old
>  "arm-pmu" device from A8-based SOC (I think) so Streamline/perf think
>  it should work, but it doesn't. The end result is no proper error
>  message for the user and the PMU interrupts are not firing so the data
>  is invalid.
>
> To manage notifications about this bug go to:
> https://bugs.launchpad.net/linaro-landing-team-samsung/+bug/859825/+subscriptions
>

--
Angus Ainslie <email address hidden>
Team Lead, Samsung Landing Team

[Wade Cherry (wade-cherry-j-deactivatedaccount)]

The Samsung Orion boot log is attached, it looks like the cpu id is
CPU S5PV310 (id 0x43200000)

[Wade Cherry (wade-cherry-j-deactivatedaccount)]

The code base I am using is a snapshot of http://git.kernel.org/?p=linux/kernel/git/kki_ap/linux-2.6-samsung.git;a=shortlog;h=refs/heads/2.6.36-samsung obtained on 2010-11-18. Though this repository seems out of date and is gone now.

[Inderpal Singh (ips)]

Its very old code, moreover from the cpu id, it seems its a very old EVT0 based SMDK board.
We have been using Origen which is EVT1.1 based.

I checked in origen, s5p_pmu_init function does not return any error. So both irqs are getting associated with their respective cores.