with_krbafs broken

[Mark W. Eichin]

> BTW, did you have any success with the xfs problems?

Not yet, though I've made several more attempts.  Looking at xfs.o
with gdb and "ptype" shows reasonable looking values for the fields of
struct xfs_attr and some others I spot checked, so that appears to be
in sync (I did switch to 0.7.2 as well, with the same results.)  I've
added some additional logging, and read more code, which taught me
that internally the XFS layer is using addresses of struct xfs_nodes
as inode numbers, but that this looks to be intentional.  (Are there
any "internals" or program flow docs, or do I just have to interrogate
you mercilessly when you get to Boston? :-)

Don't know if I mentioned before but after installing xfs, running
arlad, mounting, stat'ing /afs, unmounting, killing arlad, and
rmmod'ing xfs,  amd (user space amd, not the kernel autofs) is no
longer reacting to requests, though existing NFS mounts are still
up... 

Still, a "stat /afs" does show signs of possible structure size
problems; the device and inode value matches what xfs logs, but the
times and uid are odd...

swat+% stat /afs
  File: "/afs"
  Size: 2          Blocks: 0         Directory
Access: (0777/drwxrwxrwx)         Uid: ( 8192/ UNKNOWN)  Gid: (    0/  root)
Device: 3          Inode: 33211672   Links: 0    
Access: Wed Dec 31 19:00:00 1969
Modify: Wed Dec 31 19:00:00 1969
Change: Sun Sep  6 23:45:55 1992

though the Change time doesn't look particularly serendipitous:
swat+% date +%s --date "1992-09-06 23:45:55"
715837555
swat+% printf "%x\n" 715837555
2aaad073

			_Mark_ <eichin at kitten.gen.ma.us>
			The Herd of Kittens