[dm-devel] [for-4.16 PATCH v6 0/3] blk-mq: improve DM's blk-mq IO merging via blk_insert_cloned_request feedback

[Laurence Oberman]

Grabbing Mikes latest combined tree and testing it now

Thanks
Laurence

On Wed, Jan 17, 2018 at 6:57 PM, Laurence Oberman <loberman at redhat.com>
wrote:

> On Wed, 2018-01-17 at 23:53 +0000, Bart Van Assche wrote:
> > On Wed, 2018-01-17 at 18:43 -0500, Laurence Oberman wrote:
> > > On Wed, 2018-01-17 at 23:31 +0000, Bart Van Assche wrote:
> > > > On Wed, 2018-01-17 at 11:58 -0500, Mike Snitzer wrote:
> > > > > On Wed, Jan 17 2018 at 11:50am -0500,
> > > > > Jens Axboe <axboe at kernel.dk> wrote:
> > > > >
> > > > > > On 1/17/18 9:25 AM, Mike Snitzer wrote:
> > > > > > > Hi Jens,
> > > > > > >
> > > > > > > Think this finally takes care of it! ;)
> > > > > > >
> > > > > > > Thanks,
> > > > > > > Mike
> > > > > > >
> > > > > > > Mike Snitzer (2):
> > > > > > >   blk-mq: factor out a few helpers from
> > > > > > > __blk_mq_try_issue_directly
> > > > > > >   blk-mq-sched: remove unused 'can_block' arg from
> > > > > > > blk_mq_sched_insert_request
> > > > > > >
> > > > > > > Ming Lei (1):
> > > > > > >   blk-mq: improve DM's blk-mq IO merging via
> > > > > > > blk_insert_cloned_request feedback
> > > > > >
> > > > > > Applied - added actual commit message to patch 3.
> > > > >
> > > > > Great, thanks.
> > > >
> > > > Hello Mike,
> > > >
> > > > Laurence hit the following while retesting the SRP initiator
> > > > code:
> > > >
> > > > [ 2223.797129] list_add corruption. prev->next should be next
> > > > (00000000e0ddd5dd), but was 000000003defe5cd.
> > > > (prev=000000003defe5cd).
> > > > [ 2223.862168] WARNING: CPU: 14 PID: 577 at lib/list_debug.c:28
> > > > __list_add_valid+0x6a/0x70
> > > > [ 2224.481151] CPU: 14 PID: 577 Comm: kworker/14:1H Tainted:
> > > > G          I      4.15.0-rc8.bart3+ #1
> > > > [ 2224.531193] Hardware name: HP ProLiant DL380 G7, BIOS P67
> > > > 08/16/2015
> > > > [ 2224.567150] Workqueue: kblockd blk_mq_run_work_fn
> > > > [ 2224.593182] RIP: 0010:__list_add_valid+0x6a/0x70
> > > > [ 2224.967002] Call Trace:
> > > > [ 2224.980941]  blk_mq_request_bypass_insert+0x57/0xa0
> > > > [ 2225.009044]  __blk_mq_try_issue_directly+0x56/0x1e0
> > > > [ 2225.037007]  blk_mq_request_direct_issue+0x5d/0xc0
> > > > [ 2225.090608]  map_request+0x142/0x260 [dm_mod]
> > > > [ 2225.114756]  dm_mq_queue_rq+0xa4/0x120 [dm_mod]
> > > > [ 2225.140812]  blk_mq_dispatch_rq_list+0x90/0x5b0
> > > > [ 2225.211769]  blk_mq_sched_dispatch_requests+0x107/0x1a0
> > > > [ 2225.240825]  __blk_mq_run_hw_queue+0x5f/0xf0
> > > > [ 2225.264852]  process_one_work+0x141/0x340
> > > > [ 2225.287872]  worker_thread+0x47/0x3e0
> > > > [ 2225.308354]  kthread+0xf5/0x130
> > > > [ 2225.396405]  ret_from_fork+0x32/0x40
> > > >
> > > > That call trace did not show up before this patch series was
> > > > added to
> > > > Jens'
> > > > tree. This is a regression. Could this have been introduced by
> > > > this
> > > > patch
> > > > series?
> > > >
> > > > Thanks,
> > > >
> > > > Bart.
> > >
> > > Hi Bart
> > > One thing to note.
> > >
> > > I tested Mike's combined tree on the weekend fully dm4.16-block4.16
> > > and
> > > did not see this.
> > > This was with Mike combined tree and SRPT running 4.13-rc2.
> > >
> > > I also tested your tree Monday with the revert of the
> > > scatter/gather
> > > patches with both SRP and SRPT running your tree and it was fine.
> > >
> > > So its a combination of what you provided me before and that has
> > > been
> > > added to your tree.
> > >
> > > Mike combined tree seemed to be fine, I can revisit that if needed.
> > > I
> > > still have that kernel in place.
> > >
> > > I was not running latest SRPT when I tested Mike's tree
> >
> > Hello Laurence,
> >
> > The tree I sent you this morning did not only include Mike's latest
> > dm code
> > but also Jens' latest for-next branch. So what you wrote above does
> > not
> > contradict what I wrote in my e-mail, namely that I suspect that a
> > regression
> > was introduced by the patches in the series "blk-mq: improve DM's
> > blk-mq IO
> > merging via blk_insert_cloned_request feedback". These changes namely
> > went in
> > through the block tree and not through the dm tree. Additionally,
> > these
> > changes were not present in the block-scsi-for-next branch I sent you
> > on
> > Monday.
> >
> > Bart.
>
> Hi Bart
> Thank you
>
> I probably don't have latest stuff Mike added Monday in his tree.
> I have not gone back to test that, been busy with yours all week.
>
> I will get with Mike and pull his latest and test it with SRPT on 4.13-
> rc2 like before.
>
> That should be the best way to isolate this.
>
> Regards
> Laurence
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/dm-devel/attachments/20180117/6737c669/attachment.htm>