Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1756120Ab0FUAV1 (ORCPT ); Sun, 20 Jun 2010 20:21:27 -0400 Received: from mail-bw0-f46.google.com ([209.85.214.46]:57287 "EHLO mail-bw0-f46.google.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1754797Ab0FUAVY (ORCPT ); Sun, 20 Jun 2010 20:21:24 -0400 DomainKey-Signature: a=rsa-sha1; c=nofws; d=gmail.com; s=gamma; h=mime-version:date:message-id:subject:from:to:content-type; b=OFV+CQubvE/yLdyV7Jjwz3lAFCGXiRKe7/gwkYvqzuwBVg3mkZUJkn+GFYAqu5H7nT jrAUSK+fyTUVWXruICjS/zSolyhGYZu7Vm4DYsMuO2OKbHezWG/tKoXPyynBbBYd7NRb BA7WJVjqJ9f3dB+SjJb/ALdUdinaAiGS3lV+E= MIME-Version: 1.0 Date: Sun, 20 Jun 2010 20:21:22 -0400 Message-ID: Subject: 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 From: Richard Yao To: linux-kernel@vger.kernel.org Content-Type: text/plain; charset=ISO-8859-1 Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org Content-Length: 3035 Lines: 59 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/