From: Mikulas Patocka <mpatocka@redhat.com>
To: Dave Chinner <david@fromorbit.com>
Cc: "Matthew Wilcox (Oracle)" <willy@infradead.org>,
linux-kernel@vger.kernel.org, linux-mm@kvack.org,
linux-xfs@vger.kernel.org, dm-devel@redhat.com,
Jens Axboe <axboe@kernel.dk>, NeilBrown <neilb@suse.de>
Subject: Re: [PATCH 0/6] Overhaul memalloc_no*
Date: Mon, 29 Jun 2020 09:43:23 -0400 (EDT) [thread overview]
Message-ID: <alpine.LRH.2.02.2006290918030.11293@file01.intranet.prod.int.rdu2.redhat.com> (raw)
In-Reply-To: <20200629003550.GJ2005@dread.disaster.area>
On Mon, 29 Jun 2020, Dave Chinner wrote:
> On Sat, Jun 27, 2020 at 09:09:09AM -0400, Mikulas Patocka wrote:
> >
> >
> > On Sat, 27 Jun 2020, Dave Chinner wrote:
> >
> > > On Fri, Jun 26, 2020 at 11:02:19AM -0400, Mikulas Patocka wrote:
> > > > Hi
> > > >
> > > > I suggest to join memalloc_noio and memalloc_nofs into just one flag that
> > > > prevents both filesystem recursion and i/o recursion.
> > > >
> > > > Note that any I/O can recurse into a filesystem via the loop device, thus
> > > > it doesn't make much sense to have a context where PF_MEMALLOC_NOFS is set
> > > > and PF_MEMALLOC_NOIO is not set.
> > >
> > > Correct me if I'm wrong, but I think that will prevent swapping from
> > > GFP_NOFS memory reclaim contexts.
> >
> > Yes.
> >
> > > IOWs, this will substantially
> > > change the behaviour of the memory reclaim system under sustained
> > > GFP_NOFS memory pressure. Sustained GFP_NOFS memory pressure is
> > > quite common, so I really don't think we want to telling memory
> > > reclaim "you can't do IO at all" when all we are trying to do is
> > > prevent recursion back into the same filesystem.
> >
> > So, we can define __GFP_ONLY_SWAP_IO and __GFP_IO.
>
> Uh, why?
>
> Exactly what problem are you trying to solve here?
This:
1. The filesystem does a GFP_NOFS allocation.
2. The allocation calls directly a dm-bufio shrinker.
3. The dm-bufio shrinker sees that there is __GFP_IO set, so it assumes
that it can do I/O. It selects some dirty buffers, writes them back and
waits for the I/O to finish.
4. The dirty buffers belong to a loop device.
5. The loop device thread calls the filesystem that did the GFP_NOFS
allocation in step 1 (and that is still waiting for the allocation to
succeed).
Note that setting PF_MEMALLOC_NOIO on the loop thread won't help with this
deadlock.
Do you argue that this is a bug in dm-bufio? Or a bug in the kernel? Or
that it can't happen?
> > I saw this deadlock in the past in the dm-bufio subsystem - see the commit
> > 9d28eb12447ee08bb5d1e8bb3195cf20e1ecd1c0 that fixed it.
>
> 2014?
>
> /me looks closer.
>
> Hmmm. Only sent to dm-devel, no comments, no review, just merged.
> No surprise that nobody else actually knows about this commit. Well,
> time to review it ~6 years after it was merged....
>
> | dm-bufio tested for __GFP_IO. However, dm-bufio can run on a loop block
> | device that makes calls into the filesystem. If __GFP_IO is present and
> | __GFP_FS isn't, dm-bufio could still block on filesystem operations if it
> | runs on a loop block device.
>
> OK, so from an architectural POV, this commit is fundamentally
> broken - block/device layer allocation should not allow relcaim
> recursion into filesystems because filesystems are dependent on
> the block layer making forwards progress. This commit is trying to
> work around the loop device doing GFP_KERNEL/GFP_NOFS context
> allocation back end IO path of the loop device. This part of the
> loop device is a block device, so needs to run under GFP_NOIO
> context.
I agree that it is broken, but it fixes the above deadlock.
Mikulas
next prev parent reply other threads:[~2020-06-29 13:43 UTC|newest]
Thread overview: 41+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-06-25 11:31 [PATCH 0/6] Overhaul memalloc_no* Matthew Wilcox (Oracle)
2020-06-25 11:31 ` [PATCH 1/6] mm: Replace PF_MEMALLOC_NOIO with memalloc_noio Matthew Wilcox (Oracle)
2020-06-25 12:22 ` Michal Hocko
2020-06-25 12:34 ` Matthew Wilcox
2020-06-25 12:42 ` Michal Hocko
2020-06-25 11:31 ` [PATCH 2/6] mm: Add become_kswapd and restore_kswapd Matthew Wilcox (Oracle)
2020-06-25 12:31 ` Michal Hocko
2020-06-25 11:31 ` [PATCH 3/6] xfs: Convert to memalloc_nofs_save Matthew Wilcox (Oracle)
2020-06-25 11:31 ` [PATCH 4/6] mm: Replace PF_MEMALLOC_NOFS with memalloc_nofs Matthew Wilcox (Oracle)
2020-06-25 13:35 ` Michal Hocko
2020-06-25 11:31 ` [PATCH 5/6] mm: Replace PF_MEMALLOC_NOIO with memalloc_nocma Matthew Wilcox (Oracle)
2020-06-25 11:31 ` [PATCH 6/6] mm: Add memalloc_nowait Matthew Wilcox (Oracle)
2020-06-25 12:40 ` Michal Hocko
2020-06-25 13:10 ` Matthew Wilcox
2020-06-25 13:34 ` Michal Hocko
2020-06-25 19:05 ` kernel test robot
2020-06-25 23:51 ` kernel test robot
2020-06-29 5:08 ` Mike Rapoport
2020-06-29 12:18 ` Matthew Wilcox
2020-06-29 12:52 ` Michal Hocko
2020-06-29 13:45 ` Mike Rapoport
2020-06-29 21:28 ` Matthew Wilcox
2020-06-30 6:34 ` Michal Hocko
2020-07-01 4:12 ` Matthew Wilcox
2020-07-01 5:53 ` Michal Hocko
2020-07-01 7:04 ` Mike Rapoport
2020-09-24 0:39 ` Mike Snitzer
2020-09-24 1:10 ` Matthew Wilcox
2020-10-23 14:49 ` Daniel Vetter
2020-06-25 18:48 ` [PATCH 0/6] Overhaul memalloc_no* Darrick J. Wong
2020-06-25 20:34 ` Matthew Wilcox
2020-06-25 20:36 ` Michal Hocko
2020-06-25 20:40 ` Matthew Wilcox
2020-06-26 15:02 ` Mikulas Patocka
2020-06-26 23:08 ` Dave Chinner
2020-06-27 13:09 ` Mikulas Patocka
2020-06-29 0:35 ` Dave Chinner
2020-06-29 13:43 ` Mikulas Patocka [this message]
2020-06-29 22:34 ` Dave Chinner
2020-07-03 14:26 ` [PATCH] dm-bufio: do cleanup from a workqueue Mikulas Patocka
2020-06-29 8:22 ` [PATCH 0/6] Overhaul memalloc_no* Michal Hocko
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=alpine.LRH.2.02.2006290918030.11293@file01.intranet.prod.int.rdu2.redhat.com \
--to=mpatocka@redhat.com \
--cc=axboe@kernel.dk \
--cc=david@fromorbit.com \
--cc=dm-devel@redhat.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=linux-xfs@vger.kernel.org \
--cc=neilb@suse.de \
--cc=willy@infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).