3.10.1 cpufreq suspend/resume regression still present in 3.10.2

3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Kiko Piris]

Hi,

linux-3.10.1 introduced a regression in cpufreq breaking suspend/resume
cycle for some people [1].

There were also some other threads about it in lkml.

I see 3.10.2-stable was released some days ago. I couldn’t see anything
about fixing this regression reported in the changelog.

And to be 110% certain, I compiled it and tried suspending/resuming;
it’s still broken.

Is this going to be fixed in 3.10 stable branch?

Thank you very much in advance.


[1] first commit that bisect informed as introducing the regression:

| commit c02527487f0c8feb578c0394ad481a97f26d3bd2
| Author: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
| Date:   Mon Jul 1 00:40:55 2013 +0200
|
|     cpufreq: Fix cpufreq regression after suspend/resume
|
|     commit f51e1eb63d9c28cec188337ee656a13be6980cfd upstream.

https://groups.google.com/forum/#!topic/linux.kernel/LUi1pFCxbSI

-- 
Kiko

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Srivatsa S. Bhat]

On 07/22/2013 08:46 PM, Kiko Piris wrote:
> Hi,
> 
> linux-3.10.1 introduced a regression in cpufreq breaking suspend/resume
> cycle for some people [1].
> 
> There were also some other threads about it in lkml.
> 
> I see 3.10.2-stable was released some days ago. I couldn’t see anything
> about fixing this regression reported in the changelog.
> 
> And to be 110% certain, I compiled it and tried suspending/resuming;
> it’s still broken.
> 
> Is this going to be fixed in 3.10 stable branch?
> 

The patches that fix that regression went into mainline just a few days
ago as these commits: 

commit aae760ed21cd690fe8a6db9f3a177ad55d7e12ab
Author: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Date:   Fri Jul 12 03:45:37 2013 +0530

    cpufreq: Revert commit a66b2e to fix suspend/resume regression


commit e8d05276f236ee6435e78411f62be9714e0b9377
Author: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
Date:   Tue Jul 16 22:46:48 2013 +0200

    cpufreq: Revert commit 2f7021a8 to fix CPU hotplug regression


And both of them have been CC'ed to -stable. So they should be hitting
the stable tree soon.

Hmm, that reminds me.. whenever a patch cc'ed to stable hit the mainline,
the patch signers used to receive an automatic email from Greg. I didn't
get that for the above two patches.. Did the process change due to the
recent discussions around -stable tree maintenance?


Regards,
Srivatsa S. Bhat

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Greg KH]

On Mon, Jul 22, 2013 at 08:52:21PM +0530, Srivatsa S. Bhat wrote:
> And both of them have been CC'ed to -stable. So they should be hitting
> the stable tree soon.
> 
> Hmm, that reminds me.. whenever a patch cc'ed to stable hit the mainline,
> the patch signers used to receive an automatic email from Greg. I didn't
> get that for the above two patches.. Did the process change due to the
> recent discussions around -stable tree maintenance?

No, it did not change, I just have 150+ pending patches for the
3.11-stable tree that I have yet to get through, give me a few weeks
please.

thanks,

greg k-h

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Srivatsa S. Bhat]

On 07/22/2013 10:23 PM, Greg KH wrote:
> On Mon, Jul 22, 2013 at 08:52:21PM +0530, Srivatsa S. Bhat wrote:
>> And both of them have been CC'ed to -stable. So they should be hitting
>> the stable tree soon.
>>
>> Hmm, that reminds me.. whenever a patch cc'ed to stable hit the mainline,
>> the patch signers used to receive an automatic email from Greg. I didn't
>> get that for the above two patches.. Did the process change due to the
>> recent discussions around -stable tree maintenance?
> 
> No, it did not change, I just have 150+ pending patches for the
> 3.11-stable tree that I have yet to get through, give me a few weeks
> please.
> 

Oh! Sure, no problem.

Thank you!

Regards,
Srivatsa S. Bhat

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Paul Bolle]

On Mon, 2013-07-22 at 17:16 +0200, Kiko Piris wrote:
> linux-3.10.1 introduced a regression in cpufreq breaking suspend/resume
> cycle for some people [1].

The upstream commit that probably will fix the regression that Kiko ran
into is aae760ed21cd690fe8a6db9f3a177ad55d7e12ab ("cpufreq: Revert
commit a66b2e to fix suspend/resume regression"). It's only part of
v3.11-rc2 and was committed with a
    Cc: 3.10+ <stable@vger.kernel.org>

tag. So I'd guess that fix will be part of a future v3.10.y (stable)
releases.

> There were also some other threads about it in lkml.
> 
> I see 3.10.2-stable was released some days ago. I couldn’t see anything
> about fixing this regression reported in the changelog.
> 
> And to be 110% certain, I compiled it and tried suspending/resuming;
> it’s still broken.
> 
> Is this going to be fixed in 3.10 stable branch?
> 
> Thank you very much in advance.

Note that I already suggested to Kiko to try that patch (see
https://lkml.org/lkml/2013/7/14/35 ). I don't know whether Kiko tried
it, as I never received a reply to my suggestion. So, for now, I can
only say that that patch will probably fix the issue.

> [1] first commit that bisect informed as introducing the regression:
> 
> | commit c02527487f0c8feb578c0394ad481a97f26d3bd2
> | Author: Srivatsa S. Bhat <srivatsa.bhat@linux.vnet.ibm.com>
> | Date:   Mon Jul 1 00:40:55 2013 +0200
> |
> |     cpufreq: Fix cpufreq regression after suspend/resume
> |
> |     commit f51e1eb63d9c28cec188337ee656a13be6980cfd upstream.
> 
> https://groups.google.com/forum/#!topic/linux.kernel/LUi1pFCxbSI


Paul Bolle

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Kiko Piris]

On 22/07/2013 at 17:34 +0200, Paul Bolle wrote:

> Note that I already suggested to Kiko to try that patch (see
> https://lkml.org/lkml/2013/7/14/35 ). I don't know whether Kiko tried
> it, as I never received a reply to my suggestion. So, for now, I can
> only say that that patch will probably fix the issue.

Hi,

Aww! :(

I saw your message and was going to apply it, but I didn’t fully
understand that thread and wasn’t sure exactly what patch I was supposed
to try and above what version should I apply it (my “kernel skills” are
very very limited).

Then I just went back to 3.10.0 and forgot about it (until I saw the
3.10.2 release announcement). Sorry about that!

If you want me to test any patch, I would be more than happy to do so; I
would be grafeful if you told me exactly what patch to apply and above
which kernel version.

Cheers!

-- 
Kiko

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Srivatsa S. Bhat]

On 07/22/2013 09:24 PM, Kiko Piris wrote:
> On 22/07/2013 at 17:34 +0200, Paul Bolle wrote:
> 
>> Note that I already suggested to Kiko to try that patch (see
>> https://lkml.org/lkml/2013/7/14/35 ). I don't know whether Kiko tried
>> it, as I never received a reply to my suggestion. So, for now, I can
>> only say that that patch will probably fix the issue.
> 
> Hi,
> 
> Aww! :(
> 
> I saw your message and was going to apply it, but I didn’t fully
> understand that thread and wasn’t sure exactly what patch I was supposed
> to try and above what version should I apply it (my “kernel skills” are
> very very limited).
> 
> Then I just went back to 3.10.0 and forgot about it (until I saw the
> 3.10.2 release announcement). Sorry about that!
> 
> If you want me to test any patch, I would be more than happy to do so; I
> would be grafeful if you told me exactly what patch to apply and above
> which kernel version.
> 

Try applying the two mainline commits that I mentioned in my previous
mail, on top of 3.10.2 and check if it fixes your problem.

Regards,
Srivatsa S. Bhat

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Paul Bolle]

Kiko,

On Mon, 2013-07-22 at 21:29 +0530, Srivatsa S. Bhat wrote:
> On 07/22/2013 09:24 PM, Kiko Piris wrote: 
> > I saw your message and was going to apply it, but I didn’t fully
> > understand that thread and wasn’t sure exactly what patch I was supposed
> > to try and above what version should I apply it (my “kernel skills” are
> > very very limited).

Perhaps my message was unclear, too terse, whatever. I wouldn't have
minded to clarify my message (ie, "apply path x on top of release y). 

> > If you want me to test any patch, I would be more than happy to do so; I
> > would be grafeful if you told me exactly what patch to apply and above
> > which kernel version. 
> 
> Try applying the two mainline commits that I mentioned in my previous
> mail, on top of 3.10.2 and check if it fixes your problem.

Please try what Srivatsa suggested, as Srivatsa understands this issue
much better than I do!


Paul Bolle

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Srivatsa S. Bhat]

On 07/22/2013 09:42 PM, Paul Bolle wrote:
> Kiko,
> 
> On Mon, 2013-07-22 at 21:29 +0530, Srivatsa S. Bhat wrote:
>> On 07/22/2013 09:24 PM, Kiko Piris wrote: 
>>> I saw your message and was going to apply it, but I didn’t fully
>>> understand that thread and wasn’t sure exactly what patch I was supposed
>>> to try and above what version should I apply it (my “kernel skills” are
>>> very very limited).
> 
> Perhaps my message was unclear, too terse, whatever. I wouldn't have
> minded to clarify my message (ie, "apply path x on top of release y). 
> 
>>> If you want me to test any patch, I would be more than happy to do so; I
>>> would be grafeful if you told me exactly what patch to apply and above
>>> which kernel version. 
>>
>> Try applying the two mainline commits that I mentioned in my previous
>> mail, on top of 3.10.2 and check if it fixes your problem.
> 
> Please try what Srivatsa suggested, as Srivatsa understands this issue
> much better than I do!
> 

Nevertheless, thank you very much for following up on the bug-reports
and suggesting to apply the right fixes. I'm really grateful for that! :-)

Regards,
Srivatsa S. Bhat

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Kiko Piris]

On Mon, Jul 22, 2013 at 6:12 PM, Paul Bolle <pebolle@tiscali.nl> wrote:

> Perhaps my message was unclear, too terse, whatever. I wouldn't have
> minded to clarify my message (ie, "apply path x on top of release y).

You are absolutely right, as I said it’s my fault for not following
through it :-)

> Please try what Srivatsa suggested, as Srivatsa understands this issue
> much better than I do!

I will do it.

I don’t think I will have time to do it today; I probably won’t be
able to do it until tomorrow evening (CEST). And I also have to learn
how to extract a patch from a given commit with git (I'm very very
newbie with this git thing :-P ).

Thank you!

--
Kiko

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Kiko Piris]

On 22/07/2013 at 21:29 +0530, Srivatsa S. Bhat wrote:

> Try applying the two mainline commits that I mentioned in my previous
> mail, on top of 3.10.2 and check if it fixes your problem.

It does fix it.

Running 3.10.2 with your two patches, suspended and resumed a couple of
times without a problem (previously it didn’t wake up 100% of the
times).

Thank you both for your help!

Regards.

-- 
Kiko

Re: 3.10.1 cpufreq suspend/resume regression still present in 3.10.2

[Srivatsa S. Bhat]

On 07/23/2013 02:22 AM, Kiko Piris wrote:
> On 22/07/2013 at 21:29 +0530, Srivatsa S. Bhat wrote:
> 
>> Try applying the two mainline commits that I mentioned in my previous
>> mail, on top of 3.10.2 and check if it fixes your problem.
> 
> It does fix it.
> 
> Running 3.10.2 with your two patches, suspended and resumed a couple of
> times without a problem (previously it didn’t wake up 100% of the
> times).
> 
> Thank you both for your help!
> 

Thank you for your testing efforts and for confirming that the
patches fix your issue.

 
Regards,
Srivatsa S. Bhat