From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756203Ab0C3KRT (ORCPT ); Tue, 30 Mar 2010 06:17:19 -0400 Received: from mail-pv0-f174.google.com ([74.125.83.174]:50697 "EHLO mail-pv0-f174.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752794Ab0C3KRS convert rfc822-to-8bit (ORCPT ); Tue, 30 Mar 2010 06:17:18 -0400 DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:in-reply-to:references:date:message-id:subject:from:to :cc:content-type:content-transfer-encoding; b=vubHK3Jvgxffwn1YkUQLBd11DgLBNTpMp1lsWrVWm5720xmlxx0kfvOGl5ilPqOKq9 WlCy/xDbSJKsxjT8xUrXcxRbeWPb8U1SonKtMgtuinCUGwLyUd1yKPmgfLafsqdh1ETr qHRRHUPY+qzc7WpVJ1FJ2pMqNkBd+8bdevtuU= MIME-Version: 1.0 In-Reply-To: <20100330055304.GA2983@sli10-desk.sh.intel.com> References: <20100330055304.GA2983@sli10-desk.sh.intel.com> Date: Tue, 30 Mar 2010 19:17:17 +0900 Message-ID: <28c262361003300317g6df68fc6m4385cfbe3e8a1b04@mail.gmail.com> Subject: Re: [PATCH]vmscan: handle underflow for get_scan_ratio From: Minchan Kim To: Shaohua Li Cc: linux-mm@kvack.org, linux-kernel@vger.kernel.org, akpm@linux-foundation.org, fengguang.wu@intel.com, kosaki.motohiro@jp.fujitsu.com Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8BIT Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Tue, Mar 30, 2010 at 2:53 PM, Shaohua Li wrote: > Commit 84b18490d1f1bc7ed5095c929f78bc002eb70f26 introduces a regression. > With it, our tmpfs test always oom. The test has a lot of rotated anon > pages and cause percent[0] zero. Actually the percent[0] is a very small > value, but our calculation round it to zero. The commit makes vmscan > completely skip anon pages and cause oops. > An option is if percent[x] is zero in get_scan_ratio(), forces it > to 1. See below patch. > But the offending commit still changes behavior. Without the commit, we scan > all pages if priority is zero, below patch doesn't fix this. Don't know if > It's required to fix this too. > > Signed-off-by: Shaohua Li > > diff --git a/mm/vmscan.c b/mm/vmscan.c > index 79c8098..d5cc34e 100644 > --- a/mm/vmscan.c > +++ b/mm/vmscan.c > @@ -1604,6 +1604,18 @@ static void get_scan_ratio(struct zone *zone, struct scan_control *sc, >        /* Normalize to percentages */ >        percent[0] = 100 * ap / (ap + fp + 1); >        percent[1] = 100 - percent[0]; > +       /* > +        * if percent[x] is small and rounded to 0, this case doesn't mean we > +        * should skip scan. Give it at least 1% share. > +        */ > +       if (percent[0] == 0) { > +               percent[0] = 1; > +               percent[1] = 99; > +       } > +       if (percent[1] == 0) { > +               percent[0] = 99; > +               percent[1] = 1; > +       } >  } > >  /* > Yes. It made subtle change. But we should not depend that change. Current logic seems to be good and clear than old. I think you were lucky at that time by not-good and not-clear logic. BTW, How about this? diff --git a/mm/vmscan.c b/mm/vmscan.c index 79c8098..f0df563 100644 --- a/mm/vmscan.c +++ b/mm/vmscan.c @@ -1646,11 +1646,6 @@ static void shrink_zone(int priority, struct zone *zone, int file = is_file_lru(l); unsigned long scan; - if (percent[file] == 0) { - nr[l] = 0; - continue; - } - scan = zone_nr_lru_pages(zone, sc, l); if (priority) { scan >>= priority; -- Kind regards, Minchan Kim