On Sat, Sep 12, 2020 at 12:14 PM Jeff Layton <jlayton@xxxxxxxxxx> wrote:
>
> ceph open-codes this around some other activity and the rationale
> for it isn't clear. There is no need to delay free_anon_bdev until
> the end of kill_sb.
>
> Signed-off-by: Jeff Layton <jlayton@xxxxxxxxxx>
> ---
> fs/ceph/super.c | 4 +---
> 1 file changed, 1 insertion(+), 3 deletions(-)
>
> diff --git a/fs/ceph/super.c b/fs/ceph/super.c
> index 7ec0e6d03d10..b3fc9bb61afc 100644
> --- a/fs/ceph/super.c
> +++ b/fs/ceph/super.c
> @@ -1205,14 +1205,13 @@ static int ceph_init_fs_context(struct fs_context *fc)
> static void ceph_kill_sb(struct super_block *s)
> {
> struct ceph_fs_client *fsc = ceph_sb_to_client(s);
> - dev_t dev = s->s_dev;
>
> dout("kill_sb %p\n", s);
>
> ceph_mdsc_pre_umount(fsc->mdsc);
> flush_fs_workqueues(fsc);
>
> - generic_shutdown_super(s);
> + kill_anon_super(s);
>
> fsc->client->extra_mon_dispatch = NULL;
> ceph_fs_debugfs_cleanup(fsc);
> @@ -1220,7 +1219,6 @@ static void ceph_kill_sb(struct super_block *s)
> ceph_fscache_unregister_fs(fsc);
>
> destroy_fs_client(fsc);
> - free_anon_bdev(dev);
> }
>
> static struct file_system_type ceph_fs_type = {
> --
> 2.26.2
>
Hi Jeff,
Just curious, did you attempt to figure out why it used to be
necessary? Looks like it goes back to a very old commit 5dfc589a8467
("ceph: unregister bdi before kill_anon_super releases device name"),
but it's not obvious to me at first sight...
A lot has changed in the bdi area since then, so if not, it probably
doesn't matter much -- AFAICT bdi would still get unregistered before
the minor number is released with this patch.
Thanks,
Ilya
