From mboxrd@z Thu Jan 1 00:00:00 1970 From: Mike Snitzer Subject: Re: v4.8 dm-mpath Date: Fri, 26 Aug 2016 10:26:47 -0400 Message-ID: <20160826142647.GA1876@redhat.com> References: <20160816191241.GA16948@redhat.com> <20160817024302.GA29089@redhat.com> <20160818015433.GA8972@redhat.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: Content-Disposition: inline In-Reply-To: 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 Thu, Aug 25 2016 at 1:40pm -0400, Bart Van Assche wrote: > On 08/17/2016 06:54 PM, Mike Snitzer wrote: > >On Wed, Aug 17 2016 at 8:29pm -0400, > >Bart Van Assche wrote: > >>Earlier today I came up with three patches that avoid the hang in > >>truncate_inode_pages_range() that I had reported before. It would be > >>appreciated if you could have a look at these patches. Please keep > >>in mind that I'm not a dm expert. > > > >I don't recall where you mentioned a hang in > >truncate_inode_pages_range().. ah I see it here: > >https://patchwork.kernel.org/patch/9202331/ > > Hello Mike, > > As usual, thanks for the quick feedback. But it seems like I sent my > e-mail too soon: after I had sent my e-mail I ran again into the > truncate_inode_pages_range() hang. I was skeptical your 3 earlier patches (particularly the __dm_destroy to use internel suspend patch) would fix anything you care about in your testing. __dm_destroy is only used once all references on the DM mpath device are dropped. When you do your fio + cable pull tests you're just bouncing underlying paths around. You aren't _ever_ destroying the multipath device. That is why your __dm_destroy patch seemed off the mark to me. > So far I have run the fio + > simulated cable pulling test with the following configurations: > * scsi-mq + single queue dm: hang occurs after a few iterations. > * scsi-mq + dm-mq: hang occurs after a few iterations. > * single queue scsi + dm-mq: test passes. > > I will have another look at the blk-mq core and scsi-mq core layers. OK.