[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Message-ID: <20251003023116.GB29748@system.software.com>
Date: Fri, 3 Oct 2025 11:31:16 +0900
From: Byungchul Park <byungchul@...com>
To: David Hildenbrand <david@...hat.com>
Cc: akpm@...ux-foundation.org, ziy@...dia.com, matthew.brost@...el.com,
joshua.hahnjy@...il.com, rakie.kim@...com, gourry@...rry.net,
ying.huang@...ux.alibaba.com, apopple@...dia.com, clameter@....com,
kravetz@...ibm.com, linux-mm@...ck.org,
linux-kernel@...r.kernel.org, max.byungchul.park@...il.com,
kernel_team@...ynix.com, harry.yoo@...cle.com,
gwan-gyeong.mun@...el.com, yeoreum.yun@....com,
syzkaller@...glegroups.com, ysk@...lloc.com,
Matthew Wilcox <willy@...radead.org>
Subject: Re: [RFC] mm/migrate: make sure folio_unlock() before
folio_wait_writeback()
On Thu, Oct 02, 2025 at 01:38:59PM +0200, David Hildenbrand wrote:
> > To simplify the scenario:
> >
>
> Just curious, where is the __folio_start_writeback() to complete the
> picture?
ext4_end_io_end() was running as a wq worker after the io completion.
DEPT report can tell that the following scenario happened with
__folio_start_writeback() called far earlier, at least, before
folio_test_writeback() was seen as true, but unfortunately DEPT doesn't
capture the exact location of __folio_start_writeback().
Byungchul
> > context X (wq worker) context Y (process context)
> >
> > migrate_pages_batch()
> > ext4_end_io_end() ...
> > ... migrate_folio_unmap()
> > ext4_get_inode_loc() ...
> > ... folio_lock() // hold the folio lock
> > bdev_getblk() ...
> > ... folio_wait_writeback() // wait forever
> > __find_get_block_slow()
> > ... ...
> > folio_lock() // wait forever
> > folio_unlock() migrate_folio_undo_src()
> > ...
> > ... folio_unlock() // never reachable
> > ext4_finish_bio()
> > ...
> > folio_end_writeback() // never reachable
> >
>
> But aren't you implying that it should from this point on be disallowed
> to call folio_wait_writeback() with the folio lock held? That sounds ...
> a bit wrong.
>
> Note that it is currently explicitly allowed: folio_wait_writeback()
> documents "If the folio is not locked, writeback may start again after
> writeback has finished.". So there is no way to prevent writeback from
> immediately starting again.
>
> In particular, wouldn't we have to fixup other callsites to make this
> consistent and then VM_WARN_ON_ONCE() assert that in folio_wait_writeback()?
>
> Of course, as we've never seen this deadlock before in practice, I do
> wonder if something else prevents it?
>
> If it's a real issue, I wonder if a trylock on the writeback path could
> be an option.
>
> --
> Cheers
>
> David / dhildenb
>
Powered by blists - more mailing lists