From mboxrd@z Thu Jan 1 00:00:00 1970 From: Eric Sandeen Subject: Re: howto downgrade ext4 to ext3 Date: Fri, 18 Sep 2009 13:24:32 -0500 Message-ID: <4AB3D060.2020101@redhat.com> References: <4AB39EC5.6070107@it-sudparis.eu> <20090918175851.GH26991@mit.edu> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Cc: jehan procaccia , linux-ext4@vger.kernel.org To: Theodore Tso Return-path: Received: from mx1.redhat.com ([209.132.183.28]:21380 "EHLO mx1.redhat.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757242AbZIRSYq (ORCPT ); Fri, 18 Sep 2009 14:24:46 -0400 In-Reply-To: <20090918175851.GH26991@mit.edu> Sender: linux-ext4-owner@vger.kernel.org List-ID: Theodore Tso wrote: > On Fri, Sep 18, 2009 at 04:52:53PM +0200, jehan procaccia wrote: >> some of you might have followed my miseries about quota support and ext4 ... >> http://marc.info/?l=linux-ext4&m=125320312905576&w=2 >> >> Apparently the only sure and secure solution offered to me is to >> downgrade my ext4 FS to ext3 :-( . >> now , is there a procedure to do that ? > > There isn't a procedure to do this, other than backup, reformat, and > restore. Sorry about that; it's just one of those things we haven't > had the resources to create --- and it would not be a trivial task. > > Looking at the problems you had, have you tried simply mounting your > ext4 filesystem with -o nodelalloc? The RHEL5.4 kernel doesn't have > various delalloc patches, and I can certainly understand why Red Hat > might not be willing to try to engage in backporting all of the > delalloc fixes to its kernel, but you might be able to use the -o > nodelalloc feature to simply disable the ext4 feature which appears to > be causing the problem. This will cause you to give up some > performance, but not as much as dropping back to ext3. Ah, that's a good idea I hadn't thought of, thanks. Delalloc should be mostly ok in that kernel, but quota doesn't handle it. So dropping quota -or- dropping delalloc should suffice. OTOH, the -o nodelalloc paths have not gotten a lot of test coverage, so caveats apply. -Eric > Regards, > > - Ted