Date:        Tue, 14 Feb 2006 19:59:31 +0100 (CET)
    From:        Havard Eidnes <he@NetBSD.org>
    Message-ID:  <20060214.195931.85739192.he@uninett.no>

  | Hm, isn't the low-tech solution to this particular problem to
  | deviate from the comments in fstab(5), and use different pass
  | numbers in fstab for the file systems which have different
  | devices but actually use the same underlying spindles?

Actually, no, it isn't.   Certainly fixing fsck to return to the
old days of honouring the pass numbers more than it does now
would be possible, and a smaller change to fsck than I'm planning
(which will probably give those things some meaning again,
though not as much as they used to have).

But you really don't want to rely upon humans knowing which
real hardware underlies logical devices (like raids, etc).
The hardware can alter (a raid can migrate from one set of
drives to another over its lifetime).   What's wanted to
be known is what hardware is actually being used now to
implement the software device, not what it was called
a year ago when the device was created and added to fstab.

I can even imagine some future universe where it might be
possible to tell a raid which devices it should read from
and which to leave idle (obviously for raid1, maybe for
some of the others) for a particular filesystem on the
raid (writes naturally would work as normal, but for
fsck, and several other, purposes, writes are insignificant).
This could then allow 2 (maybe more) filesystems on the same
raid set to be checked in parallel - each using a different
(set of) drive(s)..   That is way beyond any current
plan however.

kre