Re: [RFC] sched: add sched_show_task() variant that accepts log level
From: Tomasz Figa
Date: Thu Apr 24 2025 - 00:34:44 EST
On Wed, Mar 12, 2025 at 11:51 PM Tomasz Figa <tfiga@xxxxxxxxxxxx> wrote:
>
> Hi Sergey,
>
> On Mon, Mar 10, 2025 at 5:25 PM Sergey Senozhatsky
> <senozhatsky@xxxxxxxxxxxx> wrote:
> >
> > Hi,
> >
> > A request for comments, not a formal patch.
> >
> > Is there any reason why hung-task watchdog prints error header
> > with KERN_ERR log level while task's backtrace is printed with
> > KERN_INFO? Will it make sense to unify log levels and introduce
> > sched_show_task() variant that accept log level param so that
> > everything that hung-task watchdog prints becomes KERN_ERR?
>
> Thanks a lot for looking into this.
>
> Let me just add that I've been looking into ways to automatically
> analyze kernel crash reports and I noticed that when I filter out
> KERN_INFO and higher log levels, I end up losing useful information
> for hung-task watchdog-kind of problems. This change would greatly
> help in filtering out unnecessary noise from the logs.
>
> By the way, if having it as KERN_INFO by default would still be
> desirable, I suppose we could add a KConfig option to set the desired
> log level?
Gentle ping on this patch.
Best regards,
Tomasz