[ On Tuesday, March 11, 2003 at 15:14:58 (-0800), Simon Gerraty wrote: ]
> Subject: Re: /sbin/umount should support umount_* (PR#698) 
>
> I have no problem with that - my only real issue is that umount(8)
> should stop thinking it "knows" how to deal with something that
> appears to be an NFS mount.

That's it in a nutshell I think.

>  It could simply spot the fact that a PID
> was provided at mount time and take that as a clue that it should do
> nothing more that the unmount syscall - then as you say, the fs can
> ask the helper process to "get thee hence".

Ah, but in that case then "umount" shouldn't even have to know there's
some process associated with a mount point -- it should just call
umount(2) and be done with it.  If that's the way it's supposed to be
then there shouldn't be any special code in umount(8) to look for
anything special or do anything special -- indeed umount(8) doesn't even
have to know what type of filesystem it's unmounting so long as the user
hasn't asked to unmount only those of a certain specified type.

Alternately umount(8) could look at the type of filesystem and always
invoke a separate /sbin/umount_* program just like mount(8) does and
then each FS unmount operation could be implemented in any way desired,
whether that's killing some process, calling umount(2), sending some
packet to some server, or some combination of those, or whatever.

-- 
								Greg A. Woods

+1 416 218-0098;            <g.a.woods@ieee.org>;           <woods@robohack.ca>
Planix, Inc. <woods@planix.com>; VE3TCP; Secrets of the Weird <woods@weird.com>