Re: [PATCH 3/3] thermal/intel_powerclamp: Convert to CPU hotplug state

From: Petr Mladek
Date: Thu Oct 27 2016 - 10:55:33 EST


On Mon 2016-10-24 09:55:29, Jacob Pan wrote:
> On Mon, 24 Oct 2016 17:48:07 +0200
> Petr Mladek <pmladek@xxxxxxxx> wrote:
>
> > On Fri 2016-10-21 13:21:18, Jacob Pan wrote:
> > > On Mon, 17 Oct 2016 14:32:52 +0200
> > > Petr Mladek <pmladek@xxxxxxxx> wrote:
> > >
> > > > From: Sebastian Andrzej Siewior <bigeasy@xxxxxxxxxxxxx>
> > > >
> > > Hi Sebastian,
> > > I applied this patchset on 4.9-rc1 and run some cpu online/offline
> > > loops test while injecting idle, e.g. 25%. I got system hang after a
> > > few cycles. Still looking into root cause.
> >
> > You might need the patch
> > ("sched/fair: Fix sched domains NULL deference in
> > select_idle_sibling()") from linux-tip, see
> > https://lkml.kernel.org/r/tip-9cfb38a7ba5a9c27c1af8093fb1af4b699c0a441@xxxxxxxxxxxxxx
> >
> > I have mentioned it in the cover letter. I am sorry if it was not
> > visible enough.
> Thanks for point it out again, I missed it.
>
> Now rc2 has this fix so I tried again cpuhp seems to work well now.
>
> However, I found suspend to ram does not work with this patchset. Have
> you tested it with "echo mem > /sys/power/state" while doing idle
> injection?
> It fails to enter s3 on my system.

Hmm, I haven't tested the system suspend. I am sorry but I am snowed
under some other tasks this week and will be on the Plumbers
conference the next week. I hope that I will have some time to look
at it then.

In each case, I wonder if the problem is caused by the conversion
to the kthread worker or by the CPU hotplug state conversion.

Best Regards,
Petr