Subject: Re: /sbin/umount should support umount_* (PR#698)
To: None <>
From: Simon Gerraty <>
List: current-users
Date: 02/26/2003 01:35:04
As I mentioned previously I think "suspended" is the most appropriate
state for this PR.   I may never get around to it, but someone
else might - and that would be a Good Thing[TM]

The issue still exists, eg. I can create a wonderous new filesystem
(actually I did - sNFS ;-) that has all sorts of magic properties but
can present itself to the local system as an NFS mount.  Due to
the /sbin/mound_<type> support it is trivial to get these
things mounted.  Getting them unmounted isn't so easy.

But the kernel does not have enough info available (and we don't store
anything anywhere else) to know what the original "type" of the fs
was.  Since the kernel only sees an NFS mount and it "knows" NFS, it
just tries to send a UMOUNT_RPC to the server - which will fail
(because in this case the server is only talking over an SSL session).

Fixing this requires adding a few items to the mount info stored in
the kernel.  I never got around to it (if I'd been a committer way
back then I may well have done it) but most people using this thing
were running Solaris (where umount_<type> works just as well as