On 3/21/23 1:21?PM, Linus Torvalds wrote: > On Tue, Mar 21, 2023 at 12:16?PM Jens Axboe <axboe@xxxxxxxxx> wrote: >> >> I haven't seen the patch yet as it hasn't been pushed, > > Well, it went out a couple of minutes before your email, so it's out now. Yep I see it now, looks as expected. >> It may make sense to add some debug check for >> PF_KTHREAD having TIF_NOTIFY_RESUME set, or task_work pending for that >> matter, as that is generally not workable without doing something to >> handle it explicitly. > > Yeah, I guess we could have some generic check for that. I'm not sure > where it would be. Scheduler? Off the top of my head, two options, both in kernel/sched/core.c: 1) Add it to schedule_debug() 2) Add it to sched_submit_work(), adding PF_KTHREAD to the flags checked for PF_IO_WORKER | PF_WQ_WORKER to avoid adding any extra fast-path overhead. Alternatively, I guess it could go in kthread_exit() as well. But for workloads with a persistent kthread that doesn't really go away, that won't catch it. -- Jens Axboe
