From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ted Ts'o Subject: Re: [PATCH 2/2] jbd: fix fsync() tid wraparound bug Date: Mon, 2 May 2011 14:29:45 -0400 Message-ID: <20110502182945.GG2819@thunk.org> References: <20110425231454.GB9486@thunk.org> <1303777411-1316-2-git-send-email-tytso@mit.edu> <20110430171711.GA2819@thunk.org> <20110502150758.GH4556@quack.suse.cz> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Cc: Ext4 Developers List , Martin_Zielinski@McAfee.com To: Jan Kara Return-path: Received: from li9-11.members.linode.com ([67.18.176.11]:38512 "EHLO test.thunk.org" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1752287Ab1EBS3v (ORCPT ); Mon, 2 May 2011 14:29:51 -0400 Content-Disposition: inline In-Reply-To: <20110502150758.GH4556@quack.suse.cz> Sender: linux-ext4-owner@vger.kernel.org List-ID: On Mon, May 02, 2011 at 05:07:58PM +0200, Jan Kara wrote: > The patch looks OK in any case. I'll take it in my tree. Great, thanks. > It would take > about 24 days of constant 1000 trans/s load to trigger this. That's a quite > heavy load but not so unrealistic with today's HW. True; but one of the reason why I'm not sure I believe that is this bug is showing up on some Android devices, where (a) 1000 trans/second *is* a not just a huge load, but almost impossible to believe, and (b) if we really are doing 2**31 commits, we would be wearing out the flash storage being used on the Android devices! (And we're seeing a handful of failures every week from the testers who participate in a kerneloops-like reporting system.) So while I would very much like to believe that it's caused by a tid wrap, I'm worried there is another bug hiding here.... - Ted