On Sun, Jun 09, 2002 at 06:36:20PM -0700, Bill Studenmund wrote:
> 
> Uhm, what's the problem for layered file systems? When I was at NAS and
> developed overlay mounts, we exported them all the time.
> 
> The trick to getting the mount point the same is have all the mounts in
> /etc/fstab. That way they get mounted in the same order, so their mount
> ids are the same. Look at how the code works now.

Erm dynamically generated overlays!
You also need to ensure that the file handle is persistent
- at least some implementations of overlay FS use(d) an
in kernel table...

I've not looked at the netbsd overlay (etc) fs code.
However I have done a large rewrite of a different one
that could have NFS components and be exported itself.
Without dumping data into files it is (more or less)
impossible to have persistent file handles (unless you
generate very large ones)

I also suspect that the main reason NFS filesystems are
not exportable is that NFS (v2) requires that the underlying
filesystem have 8 byte file handles - but NFS ones somewhat
larger.  IIRC v3? allows long handles?

	David

-- 
David Laight: david@l8s.co.uk