[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <201802092030.GCE64020.JVOLHFFQSOtMFO@I-love.SAKURA.ne.jp>
Date: Fri, 9 Feb 2018 20:30:41 +0900
From: Tetsuo Handa <penguin-kernel@...ove.SAKURA.ne.jp>
To: chris@...is-wilson.co.uk, linux-kernel@...r.kernel.org
Cc: mingo@...nel.org, akpm@...ux-foundation.org, ak@...ux.intel.com,
jack@...e.cz, aryabinin@...tuozzo.com, dvyukov@...gle.com
Subject: Re: [PATCH] khungtaskd: Kick stuck processes
Chris Wilson wrote:
> Quoting Tetsuo Handa (2018-02-08 23:10:43)
> > Chris Wilson wrote:
> > > After spotting a stuck process, and having decided not to panic, give
> > > the task a kick to see if that helps it to recover (e.g. to paper over a
> > > missed wake up).
> >
> > Yes, we are seeing hangs at io_schedule(), but doesn't optionally allowing
> > io_schedule() be replaced with timeout version (e.g. dump_page() upon timeout
> > if io_schedule() was called for e.g. wait_on_page_bit()) give us more clue?
>
> Yes, this isn't for debugging who left the page locked (or the exact
> root cause), this is just trying to allow the system to limp along
> afterwards :) From personal experience, I know how easy it is to lose a
> wakeup and the only thing to notice is khungtaskd shouting every 120s.
Calling wake_up_process() does not sleep, does it? Then, I think you can
do it using SystemTap, for SystemTap gives you ability to call exported
functions at (almost) arbitrary line of (almost) arbitrary file. You can find
https://events.static.linuxfound.org/sites/events/files/slides/LCJ2014-en_0.pdf
for an example.
Powered by blists - more mailing lists