From mboxrd@z Thu Jan 1 00:00:00 1970 Return-Path: Received: from ipmail06.adl2.internode.on.net ([150.101.137.129]:59535 "EHLO ipmail06.adl2.internode.on.net" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1751493AbcFOGv5 (ORCPT ); Wed, 15 Jun 2016 02:51:57 -0400 Date: Wed, 15 Jun 2016 16:51:21 +1000 From: Dave Chinner To: Willy Tarreau Cc: Greg KH , "Thomas D." , xfs@oss.sgi.com, stable@vger.kernel.org, spender@grsecurity.net Subject: Re: Something badly broken with the latest XFS changeset in all stable kernels? Message-ID: <20160615065121.GE26977@dastard> References: <75808782-835f-4bc9-5243-b25cab00d6f3@whissi.de> <20160615000241.GC26977@dastard> <20160615013056.GA23074@kroah.com> <20160615052831.GA3452@1wt.eu> MIME-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Disposition: inline In-Reply-To: <20160615052831.GA3452@1wt.eu> Sender: stable-owner@vger.kernel.org List-ID: On Wed, Jun 15, 2016 at 07:28:31AM +0200, Willy Tarreau wrote: > On Tue, Jun 14, 2016 at 06:30:56PM -0700, Greg KH wrote: > > As the patch was tagged with 3.10+, I assumed that it was safe to be > > merged to those older kernels, otherwise I would never have done so. We > > do have ways to mark external things like this for stable patches, it's > > a great help when doing backports. > > I guess only 3.14.72 was affected in the end. 3.18.35 has the fix but my > understanding is that after 3.17 it's OK. Correct. > Apparently it's not yet queued > for 3.16. The queue for 3.16 has a different bunch of XFS stuff that I've previous said is pretty risky and not advisable to back port (the mmaplock stuff). I'm waiting for that to get to users and a new bunch of whacky problems to be reported... > Jiri has queued it for 3.12 but not yet released it. For 3.10 > I only pick patches that are already in a 3.14 release so I didn't have > the time to backport it to 3.10 yet as the preview started before the > release. Just lucky, eh? > Overall the regression lived only 8 days in a single branch, I > guess it shows that our process works rather well and limits the exposure > to regressions. That it got as far as release and it took so long to get to the upstream maintainers shows the process could do with being improved. Cheers, Dave. -- Dave Chinner david@fromorbit.com