Re: [PATCH] btrfs: fix deadlocks with transaction commit when cloning inline extents

[Qu Wenruo]

在 2026/8/30 20:12, Melon Liu 写道:
> On Sat, Aug 29, 2026 at 06:25:11PM +0930, Qu Wenruo wrote:
> > 在 2026/8/29 17:09, Melon Liu 写道:
> > > Commits b48c980b6a7e ("btrfs: fix deadlock between reflink and transaction
> > > commit when using flushoncommit") and 532085d00eb5 ("btrfs: fix deadlock
> > > cloning inline extent when using flushoncommit") moved the destination
> > > i_size update before btrfs_start_transaction() to prevent
> > > extent_writepage() from invalidating a folio beyond EOF.  However, two
> > > deadlock windows remain.
> > >
> > > First, after the folio is dirtied but before i_size is updated,
> > > extent_writepage() can still see the folio beyond EOF.  It calls
> > > folio_invalidate(), and btrfs_invalidate_folio() waits for the destination
> > > extent range lock held by the clone task.
> > >
> > > Second, if extent_writepage() sees the folio within EOF, it proceeds to
> > > writepage_delalloc(), which also waits for the destination extent range
> > > lock.
> >
> > What about letting the agent to create a reproducer, trigger it and paste
> > the lockdep/hang calltrace?
> >
> > Also an ASCII chart showing the window between involved threads?
[...]
> > > Fixes: 05a5a7621ce6 ("Btrfs: implement full reflink support for inline extents")
> > > Cc: stable@xxxxxxxxxxxxxxx
> > > Signed-off-by: Melon Liu <melon1335@xxxxxxx>
> >
> > And AI disclosure.
> Thanks for the review. I'll address your comments and send a v2.

And also check the sashiko review:

https://sashiko.dev/#/patchset/a74ov5nvs5u4nkqhrs5lp2jaj664wewkqqsjxzhuemzazzmzvx%40eeof6sjb3xdc