Date:   Mon, 27 Feb 2023 21:59:17 -0800 (PST)
From:   Hugh Dickins <hughd@google.com>
To:     "Huang, Ying" <ying.huang@intel.com>
cc:     Jan Kara <jack@suse.cz>, Hugh Dickins <hughd@google.com>,
        Andrew Morton <akpm@linux-foundation.org>, linux-mm@kvack.org,
        linux-kernel@vger.kernel.org, linux-fsdevel@vger.kernel.org,
        Zi Yan <ziy@nvidia.com>, Yang Shi <shy828301@gmail.com>,
        Baolin Wang <baolin.wang@linux.alibaba.com>,
        Oscar Salvador <osalvador@suse.de>,
        Matthew Wilcox <willy@infradead.org>,
        Bharata B Rao <bharata@amd.com>,
        Alistair Popple <apopple@nvidia.com>,
        Xin Hao <xhao@linux.alibaba.com>,
        Minchan Kim <minchan@kernel.org>,
        Mike Kravetz <mike.kravetz@oracle.com>,
        Hyeonggon Yoo <42.hyeyoo@gmail.com>
Subject: Re: [PATCH -v5 0/9] migrate_pages(): batch TLB flushing
In-Reply-To: <87pm9ubnih.fsf@yhuang6-desk2.ccr.corp.intel.com>
Message-ID: <46f82df2-c925-e3f9-b185-69b3f0b6e366@google.com>
References: <20230213123444.155149-1-ying.huang@intel.com> <87a6c8c-c5c1-67dc-1e32-eb30831d6e3d@google.com> <20230227110614.dngdub2j3exr6dfp@quack3> <87pm9ubnih.fsf@yhuang6-desk2.ccr.corp.intel.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=UTF-8
Precedence: bulk

On Tue, 28 Feb 2023, Huang, Ying wrote:
> Jan Kara <jack@suse.cz> writes:
> > On Fri 17-02-23 13:47:48, Hugh Dickins wrote:
> >> 
> >> Cc'ing Jan Kara, who knows buffer_migrate_folio_norefs() and jbd2
> >> very well, and I hope can assure us that there is an understandable
> >> deadlock here, from holding several random folio locks, then trying
> >> to lock buffers.  Cc'ing fsdevel, because there's a risk that mm
> >> folk think something is safe, when it's not sufficient to cope with
> >> the diversity of filesystems.  I hope nothing more than the below is
> >> needed (and I've had no other problems with the patchset: good job),
> >> but cannot be sure.
> >
> > I suspect it can indeed be caused by the presence of the loop device as
> > Huang Ying has suggested. What filesystems using buffer_heads do is a
> > pattern like:
> >
> > bh = page_buffers(loop device page cache page);
> > lock_buffer(bh);
> > submit_bh(bh);
> > - now on loop dev this ends up doing:
> >   lo_write_bvec()
> >     vfs_iter_write()
> >       ...
> >       folio_lock(backing file folio);
> >
> > So if migration code holds "backing file folio" lock and at the same time
> > waits for 'bh' lock (while trying to migrate loop device page cache page), it
> > is a deadlock.
> >
> > Proposed solution of never waiting for locks in batched mode looks like a
> > sensible one to me...
> 
> Thank you very much for detail explanation!

Yes, thanks a lot, Jan, for elucidating the deadlocks.

I was running with the 1/3,2/3,3/3 series for 48 hours on two machines
at the weekend, and hit no problems with all of them on.

I did try to review them this evening, but there's too much for me to
take in there to give an Acked-by: but I'll ask a couple of questions.

Hugh