Message-ID: <87pm9ubnih.fsf@yhuang6-desk2.ccr.corp.intel.com>
Date:   Tue, 28 Feb 2023 09:13:26 +0800
From:   "Huang, Ying" <ying.huang@...el.com>
To:     Jan Kara <jack@...e.cz>
Cc:     Hugh Dickins <hughd@...gle.com>,
        Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
        linux-kernel@...r.kernel.org, linux-fsdevel@...r.kernel.org,
        Zi Yan <ziy@...dia.com>, Yang Shi <shy828301@...il.com>,
        Baolin Wang <baolin.wang@...ux.alibaba.com>,
        Oscar Salvador <osalvador@...e.de>,
        Matthew Wilcox <willy@...radead.org>,
        Bharata B Rao <bharata@....com>,
        Alistair Popple <apopple@...dia.com>,
        Xin Hao <xhao@...ux.alibaba.com>,
        Minchan Kim <minchan@...nel.org>,
        Mike Kravetz <mike.kravetz@...cle.com>,
        Hyeonggon Yoo <42.hyeyoo@...il.com>
Subject: Re: [PATCH -v5 0/9] migrate_pages(): batch TLB flushing

Hi, Honza,

Jan Kara <jack@...e.cz> writes:

> On Fri 17-02-23 13:47:48, Hugh Dickins wrote:
>> On Mon, 13 Feb 2023, Huang Ying wrote:
>> 
>> > From: "Huang, Ying" <ying.huang@...el.com>
>> > 
>> > Now, migrate_pages() migrate folios one by one, like the fake code as
>> > follows,
>> > 
>> >   for each folio
>> >     unmap
>> >     flush TLB
>> >     copy
>> >     restore map
>> > 
>> > If multiple folios are passed to migrate_pages(), there are
>> > opportunities to batch the TLB flushing and copying.  That is, we can
>> > change the code to something as follows,
>> > 
>> >   for each folio
>> >     unmap
>> >   for each folio
>> >     flush TLB
>> >   for each folio
>> >     copy
>> >   for each folio
>> >     restore map
>> > 
>> > The total number of TLB flushing IPI can be reduced considerably.  And
>> > we may use some hardware accelerator such as DSA to accelerate the
>> > folio copying.
>> > 
>> > So in this patch, we refactor the migrate_pages() implementation and
>> > implement the TLB flushing batching.  Base on this, hardware
>> > accelerated folio copying can be implemented.
>> > 
>> > If too many folios are passed to migrate_pages(), in the naive batched
>> > implementation, we may unmap too many folios at the same time.  The
>> > possibility for a task to wait for the migrated folios to be mapped
>> > again increases.  So the latency may be hurt.  To deal with this
>> > issue, the max number of folios be unmapped in batch is restricted to
>> > no more than HPAGE_PMD_NR in the unit of page.  That is, the influence
>> > is at the same level of THP migration.
>> > 
>> > We use the following test to measure the performance impact of the
>> > patchset,
>> > 
>> > On a 2-socket Intel server,
>> > 
>> >  - Run pmbench memory accessing benchmark
>> > 
>> >  - Run `migratepages` to migrate pages of pmbench between node 0 and
>> >    node 1 back and forth.
>> > 
>> > With the patch, the TLB flushing IPI reduces 99.1% during the test and
>> > the number of pages migrated successfully per second increases 291.7%.
>> > 
>> > Xin Hao helped to test the patchset on an ARM64 server with 128 cores,
>> > 2 NUMA nodes.  Test results show that the page migration performance
>> > increases up to 78%.
>> > 
>> > This patchset is based on mm-unstable 2023-02-10.
>> 
>> And back in linux-next this week: I tried next-20230217 overnight.
>> 
>> There is a deadlock in this patchset (and in previous versions: sorry
>> it's taken me so long to report), but I think one that's easily solved.
>> 
>> I've not bisected to precisely which patch (load can take several hours
>> to hit the deadlock), but it doesn't really matter, and I expect that
>> you can guess.
>> 
>> My root and home filesystems are ext4 (4kB blocks with 4kB PAGE_SIZE),
>> and so is the filesystem I'm testing, ext4 on /dev/loop0 on tmpfs.
>> So, plenty of ext4 page cache and buffer_heads.
>> 
>> Again and again, the deadlock is seen with buffer_migrate_folio_norefs(),
>> either in kcompactd0 or in khugepaged trying to compact, or in both:
>> it ends up calling __lock_buffer(), and that schedules away, waiting
>> forever to get BH_lock.  I have not identified who is holding BH_lock,
>> but I imagine a jbd2 journalling thread, and presume that it wants one
>> of the folio locks which migrate_pages_batch() is already holding; or
>> maybe it's all more convoluted than that.  Other tasks then back up
>> waiting on those folio locks held in the batch.
>> 
>> Never a problem with buffer_migrate_folio(), always with the "more
>> careful" buffer_migrate_folio_norefs().  And the patch below fixes
>> it for me: I've had enough hours with it now, on enough occasions,
>> to be confident of that.
>> 
>> 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!

Best Regards,
Huang, Ying

Powered by blists - more mailing lists