From: Eric Sandeen Subject: Re: Bug in delayed allocation: really bad block layouts! Date: Sun, 10 Aug 2008 13:54:50 -0500 Message-ID: <489F397A.1040709@redhat.com> References: <489F2B38.3060104@redhat.com> <20080810182231.GB15353@mit.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 7bit Cc: linux-ext4@vger.kernel.org To: Theodore Tso Return-path: Received: from mx1.redhat.com ([66.187.233.31]:47209 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1753291AbYHJSyy (ORCPT ); Sun, 10 Aug 2008 14:54:54 -0400 In-Reply-To: <20080810182231.GB15353@mit.edu> Sender: linux-ext4-owner@vger.kernel.org List-ID: Theodore Tso wrote: > On Sun, Aug 10, 2008 at 12:54:00PM -0500, Eric Sandeen wrote: >> Hm, and I tried writing out 10 files in order as a simple test but >> umount/remount brought me back many 0-byte files, I need to update my >> patchset I guess. :) >> > > One of the questions in my mind is whether this is a regression > triggered by the some of our most recent patches.... since I only > have 2.2% files reported a fragmented by e2fsck, and if this problem > had always been there, I would have expected a much higher > fragmentation number. So if you have some older kernels, you might > want to see if you can replicate the problem. I've since found that > just doing a copy via "(tar -cf - -C / usr/include ) | tar -C /mnt -xf -)" > is sufficient to see the problem. Just add a "sync; sleep 5" before > the umount. :-) It may be; I tried this and then a quick filefrag run: # filefrag usr/include/*.h | grep -v extents | awk -F : '{print $2}' | sort | uniq -c 146 1 extent found so everything came out contiguous. This was with 2.6.27-0.186.rc0.git15.fc10.x86_64 -Eric