Re: [GIT PULL] scheduler changes for v4.11

From: Matt Fleming
Date: Wed Feb 22 2017 - 15:52:41 EST


On Wed, 22 Feb, at 12:41:01PM, Linus Torvalds wrote:
> Hmm. The scheduler changes seem to show problems with suspend/resume.
>
> I now get his when suspending:
>
> ...
> Disabling non-boot CPUs ...
> Cannot set affinity for irq 285
> smpboot: CPU 1 is now offline
> ------------[ cut here ]------------
> WARNING: CPU: 2 PID: 25 at kernel/sched/sched.h:812
> update_load_avg+0x779/0xa70
> rq->clock_update_flags < RQCF_ACT_SKIP
> ...
> Call Trace:
> update_load_avg+0x779/0xa70
> set_next_entity+0xd9/0x220
> pick_next_task_fair+0x88/0x550
> sched_cpu_dying+0x1cf/0x2a0
> cpuhp_invoke_callback+0x80/0x3b0
> take_cpu_down+0x5e/0xa0
> multi_cpu_stop+0xb6/0xe0
>
> the issue may have been going on for a while, but the new WARN_ON()
> was added in this merge window.
>
> (Not sure how consistent this warning is, it definitely doesn't happen
> every time, I've only seen it once)

I think Peter has the fix for this somewhere, but it doesn't look like
it made it to tip/sched/core.

Peter, this is the original patch,

https://lkml.kernel.org/r/20170202155506.GX6515@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx