Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1753316AbcDZTkw (ORCPT ); Tue, 26 Apr 2016 15:40:52 -0400 Received: from mx2.suse.de ([195.135.220.15]:37111 "EHLO mx2.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752879AbcDZTkt (ORCPT ); Tue, 26 Apr 2016 15:40:49 -0400 Subject: Re: [PATCH mmotm 3/3] mm, compaction: prevent nr_isolated_* from going negative To: Joonsoo Kim References: <1461591269-28615-1-git-send-email-vbabka@suse.cz> <1461591350-28700-1-git-send-email-vbabka@suse.cz> <1461591350-28700-4-git-send-email-vbabka@suse.cz> <20160426005503.GC2707@js1304-P5Q-DELUXE> Cc: Andrew Morton , linux-kernel@vger.kernel.org, linux-mm@kvack.org, Michal Hocko , Hugh Dickins , Rik van Riel From: Vlastimil Babka Message-ID: <571FC43D.6010102@suse.cz> Date: Tue, 26 Apr 2016 21:40:45 +0200 User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:38.0) Gecko/20100101 Thunderbird/38.7.1 MIME-Version: 1.0 In-Reply-To: <20160426005503.GC2707@js1304-P5Q-DELUXE> Content-Type: text/plain; charset=windows-1252; format=flowed Content-Transfer-Encoding: 7bit Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 1494 Lines: 35 On 04/26/2016 02:55 AM, Joonsoo Kim wrote: > On Mon, Apr 25, 2016 at 03:35:50PM +0200, Vlastimil Babka wrote: >> @@ -846,9 +845,11 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn, >> spin_unlock_irqrestore(&zone->lru_lock, flags); >> locked = false; >> } >> - putback_movable_pages(migratelist); >> - nr_isolated = 0; >> + acct_isolated(zone, cc); >> + putback_movable_pages(&cc->migratepages); >> + cc->nr_migratepages = 0; >> cc->last_migrated_pfn = 0; >> + nr_isolated = 0; > > Is it better to use separate list and merge it cc->migratepages when > finishing instead of using cc->migratepages directly? If > isolate_migratepages() try to isolate more than one page block and keep > isolated page on previous pageblock, this putback all will invalidate > all the previous work. It would be beyond of the scope of this > function. Now, isolate_migratepages() try to isolate the page in one > pageblock so this code is safe. But, I think that removing such > dependency will be helpful in the future. I'm not strongly insisting it > so if you think it's not useful thing, please ignore this comment. migratelist was merely a reference to cc->migratepages, so it wouldn't prevent the situation you are suggesting. A truly separate list would need to be appended to cc->migratepages when leaving isolate_migratepages_block() and there's no need to do that right now. BTW, can you check patch 1/3? Thanks! Vlastimil > Thanks. >