Re: [Linux 5.18-rc1] WARNING: CPU: 1 PID: 0 at kernel/sched/fair.c:3355 update_blocked_averages

[Ammar Faizi]

On 4/5/22 7:21 PM, Dietmar Eggemann wrote:
> Tried to recreate the issue but no success so far. I used you config
> file, clang-14 and a Xeon CPU E5-2690 v2 (2 sockets 40 CPUs) with 20
> two-level cgoupv1 taskgroups '/X/Y' with 'hackbench (10 groups, 40 fds)
> + idling' running in all '/X/Y/'.
>
> What userspace are you running?

HP Laptop, Intel i7-1165G7, 8 CPUs, with 16 GB of RAM. Ubuntu 21.10. Just for
daily workstation. Compiling kernel, browsing and coding stuff.

> There seemed to be some pressure on your machine when it happened?

Yeah, might be, I don't fully remember the activity at the time it
happened, though.

> > <6>[13420.623334][    C7] perf: interrupt took too long (2530 > 2500),
> > lowering kernel.perf_event_max_sample_rate to 78900
>
> Maybe you could split the SCHED_WARN_ON so we know which signal causes this?

OK, I will apply the diff on top of 5.18-rc1 and will start using it for daily
routine tomorrow morning. Let's see if I can hit this bug again. Will send an
update later...

Thank you.

> diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
> index d4bd299d67ab..0d45e09e5bfc 100644
> --- a/kernel/sched/fair.c
> +++ b/kernel/sched/fair.c
> @@ -3350,9 +3350,9 @@ static inline bool cfs_rq_is_decayed(struct cfs_rq
> *cfs_rq)
>           * Make sure that rounding and/or propagation of PELT values never
>           * break this.
>           */
> -       SCHED_WARN_ON(cfs_rq->avg.load_avg ||
> -                     cfs_rq->avg.util_avg ||
> -                     cfs_rq->avg.runnable_avg);
> +       SCHED_WARN_ON(cfs_rq->avg.load_avg);
> +       SCHED_WARN_ON(cfs_rq->avg.util_avg);
> +       SCHED_WARN_ON(cfs_rq->avg.runnable_avg);
>
>          return true;
>   }
>
> [...]


--
Ammar Faizi