From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from ipmail06.adl2.internode.on.net ([150.101.137.129]:46131 "EHLO ipmail06.adl2.internode.on.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751337AbdEQAjf (ORCPT ); Tue, 16 May 2017 20:39:35 -0400 Date: Wed, 17 May 2017 10:39:31 +1000 From: Dave Chinner Subject: Re: [PATCH] xfs: avoid mount-time deadlock in CoW extent recovery Message-ID: <20170517003931.GN17542@dastard> References: <20170516192748.GQ4519@birch.djwong.org> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20170516192748.GQ4519@birch.djwong.org> Sender: linux-xfs-owner@vger.kernel.org List-ID: List-Id: xfs To: "Darrick J. Wong" Cc: xfs On Tue, May 16, 2017 at 12:27:48PM -0700, Darrick J. Wong wrote: > If a malicious user corrupts the refcount btree to cause a cycle between > different levels of the tree, the next mount attempt will deadlock in > the CoW recovery routine while grabbing buffer locks. We can use the > ability to re-grab a buffer that was previous locked to a transaction to > avoid deadlocks, so do that here. > > Signed-off-by: Darrick J. Wong Missing comments in the code explaining the deadlock being avoided by using xfs_trans_alloc_empty().... Cheers, Dave. -- Dave Chinner david@fromorbit.com