On Sat, Feb 14, 2004 at 12:41:14AM +1100, matthew green wrote:
> 
> 
> can you ktrace this and see what getcwd() is returning?

BTW, mountd is also confused by this. I ktraced it, and here is what it
shows:
   160 mountd   CALL  chdir(0x1fffff1c0)
   160 mountd   NAMI  "/users/disco3"
   160 mountd   RET   chdir 0
   160 mountd   CALL  __lstat13(0x1fffff1ce,0x1ffffded8)
   160 mountd   NAMI  "guests"
   160 mountd   RET   __lstat13 0
   160 mountd   CALL  chdir(0x1fffff1ce)
   160 mountd   NAMI  "guests"
   160 mountd   RET   chdir 0
   160 mountd   CALL  __getcwd(0x1fffff1c0,0x400)
   160 mountd   RET   __getcwd 26/0x1a
   160 mountd   CALL  fchdir(0x5)
   160 mountd   RET   fchdir 0
   160 mountd   CALL  close(0x5)
   160 mountd   RET   close 0
   160 mountd   CALL  __stat13(0x1fffff1c0,0x1ffffe3f0)
   160 mountd   NAMI  "/usersusers/disco3/guests"

(the real path is /users/disco3/guests, of course).
A /usersusers -> /users symlink works around the problem.

Any idea on how to track this down welcome. I can keep the host in this
state until sunday.
But it looks like the problem is in the kenrel. It seems that only one
partition (/) is affected.

-- 
Manuel Bouyer <bouyer@antioche.eu.org>
     NetBSD: 26 ans d'experience feront toujours la difference
--