Return-path: Received: from gir.skynet.ie ([193.1.99.77]:39371 "EHLO gir.skynet.ie" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754269AbZIHKyL (ORCPT ); Tue, 8 Sep 2009 06:54:11 -0400 Date: Tue, 8 Sep 2009 11:54:15 +0100 From: Mel Gorman To: Pekka Enberg Cc: Frans Pop , linux-kernel@vger.kernel.org, linux-wireless@vger.kernel.org, ipw3945-devel@lists.sourceforge.net, Andrew Morton , cl@linux-foundation.org Subject: Re: iwlagn: order 2 page allocation failures Message-ID: <20090908105415.GD28127@csn.ul.ie> References: <200909060941.01810.elendil@planet.nl> <84144f020909060114s74de2d2y850745dd82ece753@mail.gmail.com> <200909061028.48442.elendil@planet.nl> <1252226116.11274.10.camel@penberg-laptop> MIME-Version: 1.0 Content-Type: text/plain; charset=iso-8859-15 In-Reply-To: <1252226116.11274.10.camel@penberg-laptop> Sender: linux-wireless-owner@vger.kernel.org List-ID: On Sun, Sep 06, 2009 at 11:35:16AM +0300, Pekka Enberg wrote: > On Sun, 2009-09-06 at 10:28 +0200, Frans Pop wrote: > > On Sunday 06 September 2009, Pekka Enberg wrote: > > > On Sun, Sep 6, 2009 at 10:40 AM, Frans Pop wrote: > > > > Got a couple of page allocation failures today while viewing fairly > > > > large images. System was struggling for a bit to reorganize memory > > > > and swap, but nothing really serious. Everything recovered fairly > > > > quickly. > > > > > > > > Anything to look into? > > > > > > > > System: HP 2510p; 2.6.31-rc7-56-g7c0a57d; Debian stable, KDE desktop > > > > > > > > 10:00.0 Network controller [0280]: Intel Corporation PRO/Wireless > > > > 4965 AG or AGN [Kedron] Network Connection [8086:4229] (rev 61) > > > > > > Can you post your .config, please? > > > > Attached. Thanks. > > OK, so SLUB_DEBUG_ON is disabled so it's probably not a SLUB problem. > Mel, there's quite a few page allocation failure reports recently which > makes me wonder if we broke something with the page allocator > optimization patches? Do you think the anti-fragmentation fixlet you did > for nommu would help here? > No, because that change only affected the case where the system had very little memory. The last time that there was a sudden major increase in allocation failures, it was actually was page reclaim was broken - specifically kswapd was no longer doing the job that was expected of it. The symptoms where applications stalling because they were entering direct reclaim. I haven't looked very closely at this bug report yet (catching up from being offline the last 5 days). Have there been any reclaim changes that might account for something like this? My feeling is also that a number of these page allocation failures have been related to wireless drivers. Is that accurate? If so, have there been changes made to the wireless stack in this cycle that would have increased the order of pages allocated? -- Mel Gorman Part-time Phd Student Linux Technology Center University of Limerick IBM Dublin Software Lab