> Yeah, I noticed this recently while restoring a snapshot.  Mount_ufs is a
> link to mount_ffs.  I -did- manage to get my "g" mounted, but then I tried
> to use an "i", which Allen has told me won't work w/o kernel changes.  I
> suspect this would be a minimal change, but I've got to get things stable
> enough to trust the disk structure first, which I don't.

Supporting the "i" partition won't be easy. A better thing to do would be
to make it so that disklabel gave you all the partitions you wanted in the
a-h range. Being able to access the "i" partition means changing /dev in
a manner NOT compatable with other kernels. You could do it, but then
you'd always have to compile your own kernels.

> When I, running mkfs 1.45, do the newfs thing from the MacOS, I find that
> the results are less than stable.  I can, from MacBSD, run newfs and clean
> up my "g" partition, but cleaning up the "a" partition is a problem I
> haven't solved yet.
> 
> To explain: when I first looked at the "g" partition after mounting it for
> the first time, I found it was 40% used.  This was according to "df".
> After dismounting, "fsck"ing, and remounting, I found the numbers went to
> what they should be, namely 0% used.  However, fsck'ing the root partition
> gives various numbers at various times.  Obviously, I can't newfs it from
> MacBSD since I'm -running- from it, and I have no other disks I'm willing
> to partition for this software considering the feeling I have for the file
> system software we're using.  I'm seeing if I can get a better look at
> what's going on.

That's what single user's for. :-) Boot single user, then fsck your
root partition. Being able to fsck the root partition is why the root
partition is mounted read-only (so that the buffer cache won't get dirty
& then written to disk over the fixed file system).

Take care,

Bill