On Sep 30, John F. Woods wrote
> kern/3021 describes poor handling of UNIT ATTENTION in the scsi
> driver.  The sense code handling recently got churned, but I don't
> think this has been changed.  At any rate, the bug doesn't seem to
> have been looked at.  (This one includes only skeletal code.)
> 

Hum, this one describe in fact 2 problems:
- poor handling of UNIT ATTENTION, which makes the device invalidated
  when it shouldn't.
- Umount can't umount an invalidated disk, even with -f. In any case,
  if one do umount -f, the FS should be umounted, discarding the datas
  in caches if they can't be written.

Regarding the media change detection: This could be done at FS level,
at last for some of them: We could detect a media change by checking,
for example, the date of '/'. It is really unlikely that 2 file systems have
the same cration date, at a second precision. This would work at last for ffs,
ext2fs and I think cd9660. What about ados and MS-DOS ?

Do we need media change detection for other purpose that filesystem cache
concistency ?

--
Manuel Bouyer, LIP6, Universite Paris VI.           Manuel.Bouyer@lip6.fr
--