Re: [PATCH v2] btrfs: do not skip re-registration for the mounted device

[David Sterba <dsterba@suse.cz>]

On Wed, Feb 21, 2024 at 10:09:59PM +0530, Anand Jain wrote:
> On 2/20/24 23:42, David Sterba wrote:
> > On Tue, Feb 20, 2024 at 02:08:00PM +0000, Filipe Manana wrote:
> >> On Wed, Feb 14, 2024 at 7:17 AM David Sterba <dsterba@suse.cz> wrote:
> >>> On Tue, Feb 13, 2024 at 09:13:56AM +0800, Anand Jain wrote:
> >>> https://btrfs.readthedocs.io/en/latest/dev/Developer-s-FAQ.html#ordering
> >>
> >> So this introduces a regression.
> >>
> >> $ ./check btrfs/14[6-9] btrfs/15[8-9]
> > 
> > Thanks, with this I can reproduce it and have some ideas what could go
> > wrong.
> 
> Thanks indeed.

I tested the following, it fixes the fsid problems and has passed full
fstests run. The temp-fsid test coverage needs to be done still.

@@ -1388,6 +1388,10 @@ struct btrfs_device *btrfs_scan_one_device(const char *path, blk_mode_t flags,
        if (ret)
                btrfs_warn(NULL, "lookup bdev failed for path %s: %d",
                           path, ret);
+       if (devt) {
+               printk(KERN_ERR "free stale devt (for path %s)\n", path);
+               btrfs_free_stale_devices(devt, NULL);
+       }