On 2020-01-14 05:44:33 -0800, Josef Bacik wrote:
> On 1/14/20 12:53 AM, Kusanagi Kouichi wrote:
> > I tested with xfstests and lazytime didn't cause any new failures.
> >
> > Signed-off-by: Kusanagi Kouichi <slash@xxxxxxxxxxxxxxx>
> > ---
>
> We don't use the I_DIRTY flags for tracking our inodes, and .write_inode was
> removed because we didn't need it and it deadlocks. Thanks,
>
> Josef
Did you apply the patch and deadlock occur? According to commit 3c4276936f6f
("Btrfs: fix btrfs_write_inode vs delayed iput deadlock"), which removed
.write_inode, .write_inode calls btrfs_run_delayed_iputs and deadlock occurs.
But .write_inode in this patch doesn't seem to call btrfs_run_delayed_iputs.
