Re: mutex_lock issues during poweroff

[Maxime Ripard <maxime.ripard@free-electrons.com>]

[-- Attachment #1: Type: text/plain, Size: 1906 bytes --]

Hi Peter,

Thanks for your answer.

On Thu, Sep 07, 2017 at 02:40:33PM +0200, Peter Zijlstra wrote:
> On Thu, Sep 07, 2017 at 02:16:19PM +0200, Maxime Ripard wrote:
> > One thing worth noting is that we couldn't reproduce the issue with a
> > 4.13. We can't bisect really easily due to the amount of patches that
> > we still have on 4.9 and have all been merged since, but it seems like
> > the bug was fixed (either on purpose or as a side effect), and was
> > never sent to stable. Looking at the history of kernel/locking/mutex.c
> > during that window didn't really show anything obvious though.
> > 
> > If you have any ideas or spot something very wrong, I'd be happy to
> > hear about. Thanks!
> 
> Well, we did a _complete_ rewrite of the mutex primitive in v4.10-rc1.

Ok.

What commit happened to be the rewrite? 9d659ae14b54 ("locking/mutex:
Add lock handoff to avoid starvation") ? We backported this one and
3ca0ff571b09 ("locking/mutex: Rework mutex::owner"), and still can
reproduce the issue. Is there any other?

> Part of the reason for that rewrite was fixing a starvation case, but
> for that you'd need to actually have contending usage, which you claim
> not to have.

Yeah, we're close to the opposite case :)

> Aside from that I really can't remember any specific issues with the old
> code (4.9 is such a long time ago). You could try to disable the
> optimistic spinning code, see if that helps.
> 
> You did also say you were running on an ARM64, there were a few memory
> ordering fixes like for example commit:
> 
>   50972fe78f24 ("locking/osq_lock: Fix osq_lock queue corruption")

We're running on ARM, not ARM64 (they still are separate architectures
under arch/, unlike x86), but I'll look into them too.

Thanks!
Maxime

-- 
Maxime Ripard, Free Electrons
Embedded Linux and Kernel engineering
http://free-electrons.com

[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 801 bytes --]