Re: 4.13: blk-mq fails to recover from suspend

[Martin Steigerwald <martin@lichtvoll.de>]

Hello Boyan,

Boyan Penkov - 01.09.17, 17:40:
> I just pulled in 4.13, with the intention of trying bfq.
> 
> To enable bfq as a scheduler, I passed "scsi_mod.use_blk_mq=1" to
> grub, and added bfq via modprobe.  So far, so good: the scheduler does
> turn up (vis-a-vis cat /sys/block/sda/queue/scheduler).
> 
> If I leave the machine and it sleeps, however, it does not wake up.
> This is documented: https://lkml.org/lkml/2017/7/30/3
> 
> It seems, based on the discussion in the thread above, that the issue
> has been resolved in 4.13-rc4 and 4.13-rc7, as the two patches
> mentioned there are in mainline.
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h
> =v4.13-rc4&id=765e40b675a9566459ddcb8358ad16f3b8344bbe
> 
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?h
> =v4.13-rc7&id=5f042e7cbd9eb
> 
> Am I missing something here?  Does 4.13 from experimental really have
> these in?  Should I be looking elsewhere to help resolve this?

Did you review the thread "I/O hangs after resuming from suspend-to-ram" on 
LKML (I think I have been epically misunderstood in this thread as it can 
indeed be the case that my issue is also related to blk-mq, I am currently 
using cfq again and machine is stable on resume from suspend)? In the end Ming 
Lei posted about seven more fundamental patches to fix the issue for good and I 
am not sure that these went in.

Its late and I wanna go sleep…, but I see if I find the Message-Id of that 
message. Hmmm, no its not in that thread. It was in a different thread, I 
think Ming posted that patch series as a new thread, but I don´t find it right 
now.

Well maybe I gave enough pointers for you to dig it up yourself. If you find 
anything, please let me know. I just don´t find the thread with that patch 
series right now. Ah, I did. Its not on LKML, but on linux-block:

[PATCH 0/9] block/scsi: safe SCSI quiescing

Message-Id: <20170831172728.15817-1-ming.lei@redhat.com>

I bet that should be enough for you to dig it out.

Goodnight,
-- 
Martin