From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mike Snitzer Subject: Re: [PATCH 0/7] dm-mpath: Fix a race condition in the blk-mq path Date: Tue, 15 Nov 2016 20:08:02 -0500 Message-ID: <20161116010802.GF19059@redhat.com> References: <20161116004752.GD19059@redhat.com> <51cae932-470e-aa27-8c7d-e33ed978f37e@sandisk.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Disposition: inline In-Reply-To: <51cae932-470e-aa27-8c7d-e33ed978f37e@sandisk.com> List-Unsubscribe: , List-Archive: List-Post: List-Help: List-Subscribe: , Sender: dm-devel-bounces@redhat.com Errors-To: dm-devel-bounces@redhat.com To: Bart Van Assche Cc: device-mapper development List-Id: dm-devel.ids On Tue, Nov 15 2016 at 7:57pm -0500, Bart Van Assche wrote: > On 11/15/2016 04:47 PM, Mike Snitzer wrote: > >On Tue, Nov 15 2016 at 6:31pm -0500, > >Bart Van Assche wrote: > >>The seven patches in this series is what I came up with while > >>reviewing and testing the dm-mpath single queue and multiqueue code. > >>It would be appreciated if these patches would be considered for > >>inclusion in the upstream kernel. > > > >This series seems like it is not a product of need. But that of changes > >that fell out from code review. > > > >If not, what test was failing that now passes with this patchset? > > Hello Mike, > > Without this patch series I see sporadic I/O errors when running I/O > on top of dm-mq-on-mq. With this patch series my dm-mq-on-mq tests > pass. However, I still see sporadic I/O errors being reported when I > run I/O on top of dm-sq-on-mq and very sporadic I/O errors with my > dm-sq-on-sq tests. It is not yet clear to me what is causing these > I/O errors but it's probably something in either the dm core or the > dm-mpath driver. My tests scripts are available at > https://github.com/bvanassche/srp-test in case you would like to > have a look. I'm getting very tired of this. Last I knew those tests pass. Do you keep changing the tests or something? There is no change in this entire series that seems needed. Exception possibly being the patch 1/7 -- given you put so much pressure on DM device teardown vs concurrent IO. Please drop all but patch 1/7 and see if your tests pass.