From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from mail-pg0-f72.google.com (mail-pg0-f72.google.com [74.125.83.72]) by kanga.kvack.org (Postfix) with ESMTP id EF8A2280247 for ; Wed, 15 Nov 2017 19:46:17 -0500 (EST) Received: by mail-pg0-f72.google.com with SMTP id r12so14351084pgu.9 for ; Wed, 15 Nov 2017 16:46:17 -0800 (PST) Received: from lgeamrelo11.lge.com (LGEAMRELO11.lge.com. [156.147.23.51]) by mx.google.com with ESMTP id 9si18704568ple.456.2017.11.15.16.46.16 for ; Wed, 15 Nov 2017 16:46:16 -0800 (PST) Date: Thu, 16 Nov 2017 09:46:14 +0900 From: Minchan Kim Subject: Re: [PATCH 1/2] mm,vmscan: Kill global shrinker lock. Message-ID: <20171116004614.GB12222@bbox> References: <1510609063-3327-1-git-send-email-penguin-kernel@I-love.SAKURA.ne.jp> <20171115005602.GB23810@bbox> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: Sender: owner-linux-mm@kvack.org List-ID: To: Shakeel Butt Cc: Tetsuo Handa , Huang Ying , Mel Gorman , Vladimir Davydov , Michal Hocko , Johannes Weiner , Andrew Morton , Greg Thelen , Linux MM , LKML On Tue, Nov 14, 2017 at 10:28:10PM -0800, Shakeel Butt wrote: > On Tue, Nov 14, 2017 at 4:56 PM, Minchan Kim wrote: > > On Tue, Nov 14, 2017 at 06:37:42AM +0900, Tetsuo Handa wrote: > >> When shrinker_rwsem was introduced, it was assumed that > >> register_shrinker()/unregister_shrinker() are really unlikely paths > >> which are called during initialization and tear down. But nowadays, > >> register_shrinker()/unregister_shrinker() might be called regularly. > >> This patch prepares for allowing parallel registration/unregistration > >> of shrinkers. > >> > >> Since do_shrink_slab() can reschedule, we cannot protect shrinker_list > >> using one RCU section. But using atomic_inc()/atomic_dec() for each > >> do_shrink_slab() call will not impact so much. > >> > >> This patch uses polling loop with short sleep for unregister_shrinker() > >> rather than wait_on_atomic_t(), for we can save reader's cost (plain > >> atomic_dec() compared to atomic_dec_and_test()), we can expect that > >> do_shrink_slab() of unregistering shrinker likely returns shortly, and > >> we can avoid khungtaskd warnings when do_shrink_slab() of unregistering > >> shrinker unexpectedly took so long. > >> > >> Signed-off-by: Tetsuo Handa > > > > Before reviewing this patch, can't we solve the problem with more > > simple way? Like this. > > > > Shakeel, What do you think? > > > > Seems simple enough. I will run my test (running fork bomb in one > memcg and separately time a mount operation) and update if numbers > differ significantly. Thanks. > > > diff --git a/mm/vmscan.c b/mm/vmscan.c > > index 13d711dd8776..cbb624cb9baa 100644 > > --- a/mm/vmscan.c > > +++ b/mm/vmscan.c > > @@ -498,6 +498,14 @@ static unsigned long shrink_slab(gfp_t gfp_mask, int nid, > > sc.nid = 0; > > > > freed += do_shrink_slab(&sc, shrinker, nr_scanned, nr_eligible); > > + /* > > + * bail out if someone want to register a new shrinker to prevent > > + * long time stall by parallel ongoing shrinking. > > + */ > > + if (rwsem_is_contended(&shrinker_rwsem)) { > > + freed = 1; > > freed = freed ?: 1; Yub. -- 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: email@kvack.org