linux-kernel - Re: [PATCH] sched/core: Print out straggler tasks in sched_cpu

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Thu, 14 Jan 2021 10:37:35 +0000
From:   Valentin Schneider <valentin.schneider@....com>
To:     paulmck@...nel.org
Cc:     linux-kernel@...r.kernel.org, peterz@...radead.org,
        mingo@...nel.org, tglx@...utronix.de, jiangshanlai@...il.com,
        cai@...hat.com, vincent.donnefort@....com, decui@...rosoft.com,
        vincent.guittot@...aro.org, rostedt@...dmis.org, tj@...nel.org
Subject: Re: [PATCH] sched/core: Print out straggler tasks in sched_cpu_dying()

On 13/01/21 16:36, Paul E. McKenney wrote:
> On Thu, Jan 14, 2021 at 12:15:24AM +0000, Valentin Schneider wrote:
>> On 13/01/21 14:02, Paul E. McKenney wrote:
>> Thanks for giving it a spin! I think with the current series (either
>> Lai's or Peter's) sched_cpu_dying() should go smoothly, but you never
>> know.
>
> I was running the patch set having one of Lai's and three of Peter's,
> which sounds like Peter's.

That's how I was seeing it :)

> If I understand which series is which,
> Peter's has the advantage of not requiring rcutorture changes.  ;-)
>
>> > However, it did produce the following new-to-me splat, which will
>> > hopefully be of some help.
>> >
>> >                                                       Thanx, Paul
>> >
>> > ------------------------------------------------------------------------
>> >
>> > WARNING: CPU: 2 PID: 23 at kernel/kthread.c:508 kthread_set_per_cpu+0x3b/0x50
>>
>> Aha, so that's that warning I was expecting to see [1].
>> Did you also get the process_one_work() one?
>
> Yes.  Of 112 one-hour runs, there were five process_one_work() splats
> and two kthread_set_per_cpu() splats.  Each splat-ridden run had exactly
> one splat.
>

I was expecting to see both in one run, so am still somewhat confused.