Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1757311Ab0FUDHa (ORCPT ); Sun, 20 Jun 2010 23:07:30 -0400 Received: from fgwmail7.fujitsu.co.jp ([192.51.44.37]:37820 "EHLO fgwmail7.fujitsu.co.jp" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754270Ab0FUDH2 (ORCPT ); Sun, 20 Jun 2010 23:07:28 -0400 X-SecurityPolicyCheck-FJ: OK by FujitsuOutboundMailChecker v1.3.1 Date: Mon, 21 Jun 2010 12:02:56 +0900 From: KAMEZAWA Hiroyuki To: Richard Yao Cc: Andrew Hendry , linux-kernel@vger.kernel.org Subject: Re: Kernel does strange things when compilations push memory usage above physical memory and the compilations are being done in a tmpfs, despite having ample swap Message-Id: <20100621120256.8d2a2fca.kamezawa.hiroyu@jp.fujitsu.com> In-Reply-To: References: Organization: FUJITSU Co. LTD. X-Mailer: Sylpheed 3.0.2 (GTK+ 2.10.14; i686-pc-mingw32) Mime-Version: 1.0 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 Content-Length: 5574 Lines: 118 On Sun, 20 Jun 2010 22:23:50 -0400 Richard Yao wrote: > My system is still responsive if it has not locked-up, even after the > oom-killer appears to have killed stuff. > > Does the kernel need to be compiled with any special options to have > it report to dmesg that the oom-killer activated? I cited the > oom-killer as being activated because several things would > inexplicit-ably crash when the system is under memory pressure, but > looking in my dmesg log at a crash that occurred earlier today when I > forgot to unmount my tmpfs, I do not see any references to the oom > killer, just the process that crashed: > Are your oom-killer happens on workload on tmpfs ? (I thought /var/tmp isn' tmpfs..) plz check this. = http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=d6da1a5abc2bf3a06a5bda08e0f6833409234666 = Does oom-killer happens also on 2.6.33 ? Thanks, =Kame > [ 5873.816211] chrome[18404]: segfault at 8 ip 0000000001063b9b sp > 00007fffb0a7f540 error 4 in chrome[400000+28be000] > > Could it be that a bug is causing the kernel to map the same region of > physical memory to multiple programs? > > On Sun, Jun 20, 2010 at 10:03 PM, Andrew Hendry wrote: > > After the oom killer has killed things, is your system still really > > sluggish if it doesn't lockup? > > > > I have what might be a similar issue, after a lot of compiling on a ramdisk. > > http://marc.info/?l=linux-kernel&m=127569877714937&w=2 > > > > Oom killer keeps killing processes until almost nothing is left. > > Free memory is very high, and system is still very sluggish. > > > > On Mon, Jun 21, 2010 at 10:21 AM, Richard Yao wrote: > >> Dear Everyone, > >> > >> My desktop has 4GB of RAM and it is running an unpatched Linux 2.6.34 > >> kernel. I recently migrated it from Windows 7 to Gentoo Linux and I am > >> encountering a highly peculiar problem when I build/rebuild system > >> packages in a manner that stresses memory. > >> > >> When system memory usage exceeds 4GB because I have several > >> compilations running simultaneously, all of which have had -j5 passed > >> to make, with the build scripts sharing an 8GB tmpfs directory, the > >> system typically responds by activating the kernel oom-killer, which > >> will usually kill some of the processes involved in the compilations, > >> among other things. This is with an 8GB swap partition and barely any > >> of it is touched when this happens according to KDE's system monitor. > >> Rarer, but alternative responses that the system has made to such > >> circumstances involve the system package manager failing > >> mid-compilation with "Segmentation fault" printed to the console or > >> open office failing with an obscure error message. Usually just > >> compiling open office alone is enough to have things fail, although I > >> usually see it fail with an obscure 5 digit error message that has no > >> meaning which I can derive from doing searches with Google. Unmounting > >> my tmpfs directory and doing things as I normally would do them makes > >> these issues disappear. > >> > >> I have run memtest and it has not detected any hardware issues. I > >> tried asking for help on the Gentoo Linux forums, but I received no > >> responses and this looks like a kernel issue, so I thought it would be > >> a good idea to ask for assistance on the kernel mailing list. Here is > >> a link to a copy of my kernel's .config file: > >> > >> http://paste.pocoo.org/show/227799/ > >> > >> As I was typing this, I had openoffice 3.2.1 and something else > >> compiling in the background and the system completely froze. This is > >> the first I have seen my system do this and it was about 10 minutes > >> after the oom-killer had already taken out kwin and several tabs in > >> chromium. I had SSH running in the background, but even that has been > >> rendered inaccessible by the freeze. I cannot get a response from the > >> system via arping and nmap is telling me that the system is down. > >> > >> Earlier today, I tried to reproduce this issue under simpler > >> cirumstances by doing dd bs=4096 count=2097152 if=/dev/zero > >> of=/var/tmp/portage/zero.bak. As a consequence of all of the swapping > >> that occurred, the system's X server become unresponsive, so I walked > >> away and came back a few minutes later to find that the KDE System > >> Monitor had crashed, but everything else seemed fine. > >> > >> Any help with this issue would be appreciated. I am willing to > >> recompile my system in whatever manner necessary to diagnose the cause > >> of this issue. Please CC me any responses made either directly or > >> indirectly in response to this message. > >> > >> Yours truly, > >> Richard Yao > >> -- > >> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in > >> the body of a message to majordomo@vger.kernel.org > >> More majordomo info at  http://vger.kernel.org/majordomo-info.html > >> Please read the FAQ at  http://www.tux.org/lkml/ > >> > > > -- > To unsubscribe from this list: send the line "unsubscribe linux-kernel" in > the body of a message to majordomo@vger.kernel.org > More majordomo info at http://vger.kernel.org/majordomo-info.html > Please read the FAQ at http://www.tux.org/lkml/ > -- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/