sched_ext: BPF scheduler "lavd" errored, disabling

[anh0516]

Built from the current main branch, with Linux 6.10.3. It failed while playing Honkai: Star Rail.

[ 2768.754205] sched_ext: BPF scheduler "lavd" errored, disabling
[ 2768.754207] sched_ext: runnable task stall (kworker/11:1[86] failed to run for 37.632s)
[ 2768.754208]    process_scheduled_works+0x1ba/0x3d0
[ 2768.754212]    worker_thread+0x37f/0x670
[ 2768.754213]    kthread+0x281/0x2a0
[ 2768.754213]    ret_from_fork+0x2e/0x40
[ 2768.754215]    ret_from_fork_asm+0x1a/0x30

The thread in question with ID 86 is kworker/11:1-mm_percpu_wq, which says absolutely nothing about what exactly went wrong.

I stopped and started the scheduler and it's happy now. I do not have a reliable way to reproduce this and it hasn't happened again, but I figured I would report it anyway as it indicates lavd is not 100% stable in its current condition.

[anh0516]

Never mind. It did in fact fail again a few minutes later.

[ 6101.881328] sched_ext: BPF scheduler "lavd" errored, disabling
[ 6101.881330] sched_ext: runnable task stall (kworker/7:2[8986] failed to run for 30.784s)
[ 6101.881331]    process_scheduled_works+0x1ba/0x3d0
[ 6101.881336]    worker_thread+0x37f/0x670
[ 6101.881338]    kthread+0x281/0x2a0
[ 6101.881339]    ret_from_fork+0x2e/0x40
[ 6101.881342]    ret_from_fork_asm+0x1a/0x30

8986 is another generic workqueue thread, [kworker/7:2-mm_percpu_wq].

[multics69]

Thank you, @anh0516, for reporting the problem. Could you share your rough scenario of how to trigger the watchdog error? I will try to reproduce the problem from my side.

[anh0516]

It just failed randomly while playing 2.4's story content twice. Not reproducible at all, but if I play some more, it could hopefully happen again. What I put in this issue was all the info I have.

Do you have any recommendations on debugging or profiling tools that I could set up so I can try to better capture the issue?

[multics69]

@anh0516 Do you use either multi-CCX or multi-NUMA processor?

[anh0516]

I'm running an R5 5600G, so single-CCX.

…

On Sun, Aug 18, 2024, 11:41 PM Changwoo Min ***@***.***> wrote: @anh0516 <https://github.com/anh0516> Do you use either multi-CCX or multi-NUMA processor? — Reply to this email directly, view it on GitHub <#476 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ATSSZDFUM6OSQDSL7ZLYV73ZSFSPVAVCNFSM6AAAAABMG6FELWVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDEOJVGYYDMOBZGY> . You are receiving this because you were mentioned.Message ID: ***@***.***>