Subject: Re: wedges vs. not-quite-wedges, was > 1T filesystems, disklabels, etc
To: Bill Studenmund <wrstuden@netbsd.org>
From: David Laight <david@l8s.co.uk>
List: tech-kern
Date: 12/21/2002 11:03:36
> > The mbr extended partitions require more care.
> > Correctly formatted, they are a linked list in increasing sector numbers.
> > So if you allocate some free space (or free an extended partition)
> > then the numbers of any partitions further down the extended one
> > change.  This isn't really acceptable...
> >
> > The information required to tie this down need only be in a file
> > in the root filesystem though - it doesn't need to be on the same
> > disk as the extended partition.
> 
> You've just described (in brief) the wedges config file. While I don't
> like the fact that if you radically adjust an mbr extended partition (set)
> things will move around, I don't see what else you can do since the
> alternative, the history file, is worse.

Not really, the wedges config file would be telling you what to
expect and where it would be (all the problems of sitting on
stale info that the current disklabel has when it points outside
the netbsd partition).

I was still going to drive the identification from the disk
partitioning info (as you suggested), but validate against the
saved info as a sanity check to ensure (as suggested on another
leaf of this thread) that you don't run a ufs fsck on a msdos filesystem
or do an rm -rf on what isn't /tmp any more.

	David

-- 
David Laight: david@l8s.co.uk