From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: (majordomo@vger.kernel.org) by vger.kernel.org via listexpand id S1752818Ab1AUDrR (ORCPT ); Thu, 20 Jan 2011 22:47:17 -0500 Received: from ipmail06.adl2.internode.on.net ([150.101.137.129]:21066 "EHLO ipmail06.adl2.internode.on.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751273Ab1AUDrQ (ORCPT ); Thu, 20 Jan 2011 22:47:16 -0500 X-Greylist: delayed 303 seconds by postgrey-1.27 at vger.kernel.org; Thu, 20 Jan 2011 22:47:15 EST X-IronPort-Anti-Spam-Filtered: true X-IronPort-Anti-Spam-Result: AvsEAIaMOE15LN5m/2dsb2JhbACkYHS/HQ2FQwQ Date: Fri, 21 Jan 2011 14:42:08 +1100 From: Dave Chinner To: Alex Romosan Cc: linux-kernel@vger.kernel.org, xfs@oss.sgi.com Subject: Re: xfs used disk space bug in 2.6.38-rc1 (commit 6e857567dbbfe14dd6cc3f7414671b047b1ff5c7) Message-ID: <20110121034208.GZ16267@dastard> References: <87zkqvtet2.fsf@sycorax.lbl.gov> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <87zkqvtet2.fsf@sycorax.lbl.gov> User-Agent: Mutt/1.5.20 (2009-06-14) Sender: linux-kernel-owner@vger.kernel.org List-ID: X-Mailing-List: linux-kernel@vger.kernel.org On Thu, Jan 20, 2011 at 02:01:45PM -0800, Alex Romosan wrote: > my home directory is using xfs and i noticed that when i try to compile > a program or do any kind of disk access (like git pull, etc) the amount > of space used jumps by tens of gigabytes causing the disk to seemingly > become full (eventually). on reboot the used disk space goes back to the > correct value. i did a 'git bisect start -- fs/xfs' and i tracked it > down to commit 6e857567dbbfe14dd6cc3f7414671b047b1ff5c7. if i revert When you quote a commit, can you please include the name of the commit. e.g: 6e85756 ("xfs: don't truncate prealloc from frequently accessed inodes") > this patch then disk usage doesn't increase unexpectedly (i have 270GB > free and the disk became full by simply trying to compile wine). i can > try patches if somebody comes up with one. I don't think you tracked the problem down to the correct commit. The previous commit: 055388a xfs: dynamic speculative EOF preallocation Is the one with the known bug that causes excessive preallocation when files are zero sized. The candidate fix is here: http://oss.sgi.com/archives/xfs/2011-01/msg00281.html Can you see if that solves your problem? Cheers, Dave. -- Dave Chinner david@fromorbit.com