Re: Deadlocks due to per-process plugging

[Mike Galbraith <mgalbraith@suse.de>]

On Thu, 2012-07-12 at 00:12 +0200, Thomas Gleixner wrote: 
> On Wed, 11 Jul 2012, Jan Kara wrote:
> > On Wed 11-07-12 12:05:51, Jeff Moyer wrote:
> > > This eventually ends in a call to blk_run_queue_async(q) after
> > > submitting the I/O from the plug list.  Right?  So is the question
> > > really why doesn't the kblockd workqueue get scheduled?
> >   Ah, I didn't know this. Thanks for the hint. So in the kdump I have I can
> > see requests queued in tsk->plug despite the process is sleeping in
> > TASK_UNINTERRUPTIBLE state.  So the only way how unplug could have been
> > omitted is if tsk_is_pi_blocked() was true. Rummaging through the dump...
> > indeed task has pi_blocked_on = 0xffff8802717d79c8. The dump is from an -rt
> > kernel (I just didn't originally thought that makes any difference) so
> > actually any mutex is rtmutex and thus tsk_is_pi_blocked() is true whenever
> > we are sleeping on a mutex. So this seems like a bug in rtmutex code.
> > Thomas, you seemed to have added that condition... Any idea how to avoid
> > the deadlock?
> 
> Mike has sent out a fix related to the plug stuff, which I just posted
> for the rt stable series. Can you verify against that ?

btw, I called io_schedule() instead of a plain unplug thinking we're
going to schedule anyway, but if we unplug and schedule, and we're not
leftmost (non-rt task 'course), while we're away, likely contended mutex
we're about to take may be released or at least become less contended.
What a we won't be doing is accruing sleep time to help trigger yet more
preemption.  Anyone more deserving can move smartly rightward, and thus
out of our way for a bit.

If we're leftmost or rt, all was for naught, but it seemed worth a shot.

-Mike