From mboxrd@z Thu Jan 1 00:00:00 1970 From: "SourceForge.net" Subject: [ kvm-Bugs-2487340 ] Possible memory corruption when file > RAM Date: Sat, 10 Jan 2009 13:02:24 +0000 Message-ID: Mime-Version: 1.0 Content-Type: text/plain; charset="UTF-8" To: noreply@sourceforge.net Return-path: Received: from ch3.sourceforge.net ([216.34.181.60]:62029 "EHLO ch3.sourceforge.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1750813AbZAJNC1 (ORCPT ); Sat, 10 Jan 2009 08:02:27 -0500 Sender: kvm-owner@vger.kernel.org List-ID: Bugs item #2487340, was opened at 2009-01-05 21:09 Message generated for change (Comment added) made by jaketmuss You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=893831&aid=2487340&group_id=180599 Please note that this message will contain a full copy of the comment thread, including the initial issue submission, for this request, not just the latest update. Category: None Group: None Status: Pending Resolution: None Priority: 5 Private: No Submitted By: Jake Muss (jaketmuss) Assigned to: Nobody/Anonymous (nobody) Summary: Possible memory corruption when file > RAM Initial Comment: Host: Debian Lenny AMD64, Intel Q6600 , kvm-82 Guest, Debian Lenny AMD64 I've been having issues with file corruption and I've simplified the issue down to the following: create a 1G file: dd if=/dev/urandom of=1G bs=1G count=1 md5sum the file twice and the results will be different. I've tested this on ext2, ext3, backing device lvm,qcow2,raw If I change the RAM to 2G, I can then correctly md5sum the original file, however if I create a new file that's 2G the issue reappears. The only time the test has not failed was for the first few tests after the odd reboot. ---------------------------------------------------------------------- Comment By: Jake Muss (jaketmuss) Date: 2009-01-11 02:02 Message: I've tested with -no-kvm-irqchip, -no-kvm-pit switch and -no-kvm, all have the same result. ---------------------------------------------------------------------- Comment By: Amit Shah (amitshah) Date: 2009-01-11 01:47 Message: I can't reproduce this on a Lenny guest on an F10 host. Both are 64-bit. Can you try reproducing this with -no-kvm and report the results? ---------------------------------------------------------------------- Comment By: Jake Muss (jaketmuss) Date: 2009-01-10 01:18 Message: Guest RAM. I have also now tried using Fedora 10 as a guest and using Debian Lenny i386 (32 bit) as a guest, still have the same issue. I ram memcheck for one pass in fedora and that didnlt come up with anything, and also one pass on the host. I'm not able to take the host down for long periods. Lastly, I seem to be able to get the same md5 sum if nothing else is causing a big change the page cache, however the md5sum is still wrong and won't be repeated later on. So I guess it's either getting into the cache incorrectly, or something to do with the way is accessed. I did some other tests with a binary diff and in a 1 gig file I had about 8 bits different. ---------------------------------------------------------------------- Comment By: Amit Shah (amitshah) Date: 2009-01-10 00:56 Message: Which RAM do you mean? Host RAM or guest RAM? ---------------------------------------------------------------------- You can respond by visiting: https://sourceforge.net/tracker/?func=detail&atid=893831&aid=2487340&group_id=180599