Re: [PATCH] mm: Use WQ_HIGHPRI for mm_percpu_wq.

From: Tetsuo Handa <penguin-kernel@I-love.SAKURA.ne.jp>
To: tj@kernel.org, mhocko@kernel.org
Cc: akpm@linux-foundation.org, linux-mm@kvack.org, mgorman@suse.de,
	vbabka@suse.cz
Subject: Re: [PATCH] mm: Use WQ_HIGHPRI for mm_percpu_wq.
Date: Wed, 30 Aug 2017 05:29:26 +0900	[thread overview]
Message-ID: <201708300529.HEB00599.VHtOFOLFSJOMFQ@I-love.SAKURA.ne.jp> (raw)
In-Reply-To: <20170829143319.GJ491396@devbig577.frc2.facebook.com>

Tejun Heo wrote:
> Hello,
> 
> On Tue, Aug 29, 2017 at 03:33:25PM +0200, Michal Hocko wrote:
> > Hmm, we have this in should_reclaim_retry
> > 			/*
> > 			 * Memory allocation/reclaim might be called from a WQ
> > 			 * context and the current implementation of the WQ
> > 			 * concurrency control doesn't recognize that
> > 			 * a particular WQ is congested if the worker thread is
> > 			 * looping without ever sleeping. Therefore we have to
> > 			 * do a short sleep here rather than calling
> > 			 * cond_resched().
> > 			 */
> > 			if (current->flags & PF_WQ_WORKER)
> > 				schedule_timeout_uninterruptible(1);
> > 
> > And I thought it would be susfficient for kworkers for concurrency WQ
> > congestion thingy to jump in. Or do we need something more generic. E.g.
> > make cond_resched special for kworkers?
> 
> I actually think we're hitting a bug somewhere.  Tetsuo's trace with
> the patch applies doesn't add up.
> 
> Thanks.

If we are under memory pressure, __zone_watermark_ok() can return false.
If __zone_watermark_ok() == false, when is schedule_timeout_*() called explicitly?

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>