Subject: Re: FFS reliability problems
To: Olaf Seibert <rhialto@polderland.nl>
From: Bill Studenmund <wrstuden@netbsd.org>
List: tech-kern
Date: 06/09/2002 18:45:20
On Sat, 8 Jun 2002, Olaf Seibert wrote:

> On Sat 08 Jun 2002 at 20:09:30 +0100, David Laight wrote:
> > True - now sort out how to export:
> > - FAT filesystems
> > - overlay mounts
>
> Well, I suppose that at least the second should not be exportable, but
> just can be re-created at the client side. Just like nfs-mounted
> filesystems cannot be re-exported. I would think this is for the same
> reason.

Uhm, why should overlay (layered) mounts not be exportable? While
null/overlay themselves serve little use being exported, umapfs is a GREAT
one to export. That way you control on the server what uids can do what,
rather than relying on a umap on the client to do the right thing.

The reason you can't export an nfs mount is different. It's because you
can get into a network buffer resource deadlock, where say you need free
buffers (to do nfs client ops) to free up buffers to satisfy nfs server
ops.

Take care,

Bill