From: Andreas Dilger Subject: Re: Fw: [BUG -mm] ext3_orphan_add() accessing corrupted list on a corrupted ext3fs Date: Thu, 1 Feb 2007 18:19:38 -0700 Message-ID: <20070202011938.GC5717@schatzie.adilger.int> References: <20070201010836.31a63ef2.akpm@osdl.org> <20070201102534.GF5404@schatzie.adilger.int> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Andrew Morton , "linux-ext4@vger.kernel.org" , Fengguang Wu To: Alex Tomas Return-path: Received: from mail.clusterfs.com ([206.168.112.78]:46236 "EHLO mail.clusterfs.com" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1423058AbXBBB0o (ORCPT ); Thu, 1 Feb 2007 20:26:44 -0500 Content-Disposition: inline In-Reply-To: Sender: linux-ext4-owner@vger.kernel.org List-Id: linux-ext4.vger.kernel.org On Feb 01, 2007 20:28 +0300, Alex Tomas wrote: > >>>>> Andreas Dilger (AD) writes: > AD> I don't have a comment on the actual bug here, but this is another case > AD> where it would be nice to have multi-mount protection built into ext3... > AD> When I last proposed this it was refused on the grounds that an external > AD> HA manager should be doing this job but I don't think that is realistic. > > can we use JBD-like approach? export some inode and MMP/whatever > would 'ping' 1st block of the inode. I'd be happy enough to implement the engine for this in a generic kernel layer instead of being ext3-specific. That said, ext3 also has the ability to mark the filesystem incompatible to older kernels that do not support the MMP protocol, so this avoids the requirement that both kernels are up-to-date in order to not corrupt shared images. Cheers, Andreas -- Andreas Dilger Principal Software Engineer Cluster File Systems, Inc.