REGRESSION: "perf: Add PM notifiers ..." breaks resume

REGRESSION: "perf: Add PM notifiers ..." breaks resume

[Andi Kleen]

Resume stopped working on my Acer Aspire 1830T post 3.1. The screen
just stays black after wakeup.

I bisected it down to 

commit 144060fee07e9c22e179d00819c83c86fbcbf82c
Author: Peter Zijlstra <a.p.zijlstra@chello.nl>
Date:   Mon Aug 1 12:49:14 2011 +0200

    perf: Add PM notifiers to fix CPU hotplug races
    
Reverting this patch fixes the problem for me.

-Andi

-- 
ak@linux.intel.com -- Speaking for myself only.

Re: REGRESSION: "perf: Add PM notifiers ..." breaks resume

[Linus Torvalds]

Peter, Ingo, ping?

Quite frankly, I'm inclined to revert that patch just for being
horribly ugly and apparently broken regardless of the fact that it
also causes a regression.

*Notifiers* for doing something like this? smp_call_function? That's
just utter sh*t, guys.

If it is CPU state, then it should damn well be suspended and resumed
as part of CPU suspend and resume. No IPI cross-calls to ask other
CPU's to do random work, no crap like that.

Comments? Should I just revert the thing (silence is obviously "yes"
in cases like this where we are talking about regressions anyway)?

                   Linus

On Wed, Nov 2, 2011 at 6:49 AM, Andi Kleen <andi@firstfloor.org> wrote:
>
> Resume stopped working on my Acer Aspire 1830T post 3.1. The screen
> just stays black after wakeup.
>
> I bisected it down to
>
> commit 144060fee07e9c22e179d00819c83c86fbcbf82c
> Author: Peter Zijlstra <a.p.zijlstra@chello.nl>
> Date:   Mon Aug 1 12:49:14 2011 +0200
>
>    perf: Add PM notifiers to fix CPU hotplug races
>
> Reverting this patch fixes the problem for me.
>
> -Andi
>
> --
> ak@linux.intel.com -- Speaking for myself only.
>

Re: REGRESSION: "perf: Add PM notifiers ..." breaks resume

[Ingo Molnar]

* Linus Torvalds <torvalds@linux-foundation.org> wrote:

> Peter, Ingo, ping?
> 
> Quite frankly, I'm inclined to revert that patch just for being 
> horribly ugly and apparently broken regardless of the fact that it 
> also causes a regression.
> 
> *Notifiers* for doing something like this? smp_call_function? 
> That's just utter sh*t, guys.
> 
> If it is CPU state, then it should damn well be suspended and 
> resumed as part of CPU suspend and resume. No IPI cross-calls to 
> ask other CPU's to do random work, no crap like that.

Agreed, that's quite crappy.

> Comments? Should I just revert the thing (silence is obviously 
> "yes" in cases like this where we are talking about regressions 
> anyway)?

Yep, please revert it until we do the right fix. (I can queue up the 
revert too)

Thanks,

	Ingo

Re: REGRESSION: "perf: Add PM notifiers ..." breaks resume

[Linus Torvalds]

On Thu, Nov 3, 2011 at 12:37 AM, Ingo Molnar <mingo@elte.hu> wrote:
>
> Yep, please revert it until we do the right fix. (I can queue up the
> revert too)

Reverted,

                Linus

Re: REGRESSION: "perf: Add PM notifiers ..." breaks resume

[Ingo Molnar]

* Linus Torvalds <torvalds@linux-foundation.org> wrote:

> On Thu, Nov 3, 2011 at 12:37 AM, Ingo Molnar <mingo@elte.hu> wrote:
> >
> > Yep, please revert it until we do the right fix. (I can queue up the
> > revert too)
> 
> Reverted,

Thanks!

	Ingo

Re: REGRESSION: "perf: Add PM notifiers ..." breaks resume

[Peter Zijlstra]

On Wed, 2011-11-02 at 17:00 -0700, Linus Torvalds wrote:
> Peter, Ingo, ping?
> 
> 
Shoot it.. I'll poke at it later and try and solve the thing
differently.