Re: [PATCH 2/3] btrfs: support to find missing device by path

[Anand Jain <Anand.Jain@oracle.com>]

>>>    Critically we don't need this patch. right ?
>>>    Anyway user of replace cli can use devid if device read fails.

 > Yes, I agree it.

>>>    I think David is talking about:
>>>      [PATCH] device delete by devid
>>>
>>>    it was critical for device delete. since there wasn't device
>>>    delete by devid. I used device delete by devid instead of
>>>    device path strcmp mainly because to maintain consistency
>>>    between device replace and delete.
>>>    the above patch set also provides code cleanups between
>>>    device replace and delete codes.

 > We can delete the device by devid on behalf of "btrfs_find_device".

  yes. patch-set (above) uses btrfs_find_device for device delete now.
  replace was already using it.

 > In my opinion, the dev path is easier and humanized to use.

  yes. good to have. in the long run. But not a regression/critical.
  this will conflict with my patch, can you rebase on top of
  above path which has some cleanups in this area as well.

 > This was OK before, but now I can not replace offline device
 > by path.
 > So I consider it as a regression.

  You mean to say you could replace the offline device using the
  device path before (not devid) and now you can't ?

  Then what patch introduced the regression ? Do you see any
  older version replace working with offline device using the
  device path ?

Thanks, Anand