Subject: Re: Removing tmpfs' experimental status
To: Brett Lymn <>
From: Julio M. Merino Vidal <>
List: tech-kern
Date: 10/23/2006 14:21:37
On 10/23/06, Brett Lymn <> wrote:
> On Mon, Oct 23, 2006 at 12:43:55PM +0200, Julio M. Merino Vidal wrote:
> >
> > But do we really need to specify that a given file system cannot be
> > exported?  I mean, is the lack of NFS exportability for a given file
> > system a bug?
> >
> Not a bug but it means that fileassoc will not work on that
> filesystem.  Really, generating a fh is not that hard, it just has to
> be something that can uniquely identify the file on the file system.
> At a guess, in your case, it could be the address of the file node in
> memory... it really does not matter as long as it's unique :)

Hmm, I don't think I explained myself correctly.

tmpfs _already_ supports file handles.  But exporting a tmpfs over NFS
does not work for whatever reason -- can't remember the exact details.

So, tmpfs should have a way to tell the NFS server code that it has to
disallow exports of these filesystems.  Now the question is: do we
really need to be able to do this?  I mean, may other file systems
benefit from such a feature (supporting file handles but not wanting
to be ever exported)?  Or we simply want to fix those file systems
(tmpfs in my case) to support NFS exportability?

Julio M. Merino Vidal <>
The Julipedia -