From: Nick Piggin Subject: Re: ext3: slow symlink corruption on umount... Date: Fri, 31 Oct 2008 20:47:20 +1100 Message-ID: <200810312047.20306.nickpiggin@yahoo.com.au> References: <20081024183733.GA25797@ajones-laptop.nbttech.com> <20081029210949.GA27274@mit.edu> <4909B8D6.7070206@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Content-Transfer-Encoding: 7bit Cc: Theodore Tso , Arthur Jones , "linux-ext4@vger.kernel.org" , sct@redhat.com, akpm@linux-foundation.org, linux-kernel@vger.kernel.org To: Eric Sandeen Return-path: In-Reply-To: <4909B8D6.7070206@redhat.com> Content-Disposition: inline Sender: linux-kernel-owner@vger.kernel.org List-Id: linux-ext4.vger.kernel.org On Friday 31 October 2008 00:38, Eric Sandeen wrote: > Theodore Tso wrote: > > On Wed, Oct 29, 2008 at 03:36:33PM -0500, Eric Sandeen wrote: > >> Sorry for the silence, this is a nice bug you've found :) > >> > >> I'll look into it ... > > > > You may want to take a quick look at this thread: > > > > http://lkml.org/lkml/2008/10/28/413 > > > > - Ted > > I thought about that, but at least at first glance I don't see how the > gfp mask change would cause this behavior...? At least, I don't think > we're seeing recursion back into the filesystem... but I'll ponder that. That's definitely a bug which I'll have to fix for 2.6.28, but I agree it's unlikely to recurse frequently like this (would only happen under high memory pressure, and only when writeout from page reclaim happens). > (Also, Arthur reports seeing this as long ago as 2.6.9...) OK, well that would confirm it.