Henrik Grindal Bakken <hgb@ifi.uio.no> writes:

> Charles Manning <manningc2@actrix.gen.nz> writes:
>
>> If you are using the kernel exporting then killing the daemon with
>> -9 might not be immediately cleaning up since some state is in the
>> kernel and some in the process. If that is happening then it might
>> be taking a while for the kernel to detect the problem and clean
>> up.
>>
>> Since the fs is busy while it is exported, you can't unmount it.
>>
>> If you do this cleanly with exportfs -u ... then you should be able
>> to unmount yaffs.
>
> I shut down nfs first, but I don't do exportfs -u, which I probably
> should.  I'm merely shutting down nfs first (along with the other
> running processes), and then (somewhat later) umounting.  I'll try
> to add exportfs -u to my nfs shutdown script.

Trying to add 'exportfs -au' to my shutdown script doesn't help
matters.  After running

    exportfs -au
    [ -f /var/run/sm-notify.pid ] && kill `cat /var/run/sm-notify.pid`
    [ -f /var/run/rpc.statd.pid ] && kill `cat /var/run/rpc.statd.pid`
    pkill rpc.mountd
    umount /proc/fs/nfsd

I'm left with the following troubling processes in 'ps -ef':

    root      1414     2  0 09:39 ?        00:00:00 [lockd]
    root      1415     2  4 09:39 ?        00:00:08 [nfsd]

If I give these two the kill -9 treatment, they die within the space
of some seconds (but not instantly, which is annoying), and I can
umount the fs.

I added 'pkill nfsd; pkill lockd', however, and while the processes
still exist, I can cleanly umount.  It feels a bit weird, but I'll
just go with it.

Thanks for your help, Charles.

-- 
Henrik Grindal Bakken <hgb@ifi.uio.no>
PGP ID: 8D436E52
Fingerprint: 131D 9590 F0CF 47EF 7963  02AF 9236 D25A 8D43 6E52