On Sat, 8 Mar 2003 12:04:41 -0500, "D'Arcy J.M. Cain" writes:
>> Yep, that's exactly how it works - umount_snfs just kill's the
>> mount_snfs process that does all the real work.
>>
>> On netbsd though, umount has no way of detecting that its an snfs
>> mount rather than a vanila NFS mount, since not enough info is stored
>> in the kernel.  And since umount "knows" NFS it just tries to send an
>> RPC to the server - which fails.
>
>I'm either being unclear or dense.  What I was getting at was the idea of 

You are being neither (I think).  This amounts to the same suggestion
that Charles made and which I think would be a good solution.
Ie. for a mount type like snfs where the mount_snfs process will dag
around actually doing some magic, it could put its PID in the mount
args.

umount would then simply need to _know_ that if PID is set, the right
thing for it to do is just KILL that PID.  Rather than what it
currently does "oh NFS I know what to do about that...".

Thanks
--sjg