/home: Stale NFS file handle

[Dave Sherohman <esper@sherohman.org>]

I took the NFS server down to replace a couple hard drives on
Saturday and now the clients are all complaining that the NFS handle
for /home is stale.  mounting and umounting the same export elsewhere
doesn't work, despite some claims on lkml that this might do some
magic.  The other things I've tried:

# umount -f /home
umount2: Device or resource busy
umount: /home: device is busy

# fuser -km /home
/home: Stale NFS file handle

# lsof | grep home
lsof: WARNING: can't stat() nfs file system /home
      Output information may be incomplete.

# mount /home -o remount,ro
mount: /home is busy

# echo 255 > /proc/sys/sunrpc/nfs_debug
# less /var/log/syslog
kernel: NFS: revalidating (3/2)
(Followed by multiple iterations of:)
kernel: NFS: lookup(//username)
kernel: NFS call  lookup username
kernel: NFS reply lookup: -116

/# cd /home
/home# sync
/home# ls
ls: .: Stale NFS file handle

Have I missed anything?  Is there any way of recovering from this
short of rebooting?  (I've already rebooted 5 of 8 clients, but the
other three already had users logged in before I got to them...)

-- 
When we reduce our own liberties to stop terrorism, the terrorists
have already won. - reverius

Innocence is no protection when governments go bad. - Tom Swiss