From mboxrd@z Thu Jan 1 00:00:00 1970 From: Neil Brown Subject: Re: + knfsd-exportfs-add-exportfsh-header-fix.patch added to -mm tree Date: Thu, 17 May 2007 10:05:29 +1000 Message-ID: <17995.40009.645694.142225@notabene.brown> References: <20070516065721.GA9884@lst.de> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Content-Transfer-Encoding: 7bit Cc: Christoph Hellwig , akpm@linux-foundation.org, hch@infradead.org, linux-fsdevel@vger.kernel.org To: Steven French Return-path: Received: from ns1.suse.de ([195.135.220.2]:41744 "EHLO mx1.suse.de" rhost-flags-OK-OK-OK-OK) by vger.kernel.org with ESMTP id S1757634AbXEQAFt (ORCPT ); Wed, 16 May 2007 20:05:49 -0400 In-Reply-To: message from Steven French on Wednesday May 16 Sender: linux-fsdevel-owner@vger.kernel.org List-Id: linux-fsdevel.vger.kernel.org On Wednesday May 16, sfrench@us.ibm.com wrote: > Any ideas what are the minimum export operation(s) that cifs would need to > add to export under nfsd? It was not clear to me after reading the > Exporting document in Documentation directory. You need to be able to map a dentry to a filehandle (you get about 20 bytes) and back again. If CIFS provides some fix-length identifier for files, then you might be able to do it. If not, cannot really do it at all. And I suspect the later. (There are other requirements, like get_parent, but we could probably work around those if we really needed to). Theoretically, you could make it work with NFSv4 and volatile file handles, but I doubt it would really work in practice. I don't think the "volatile" concept quite stretch as far as you would need. Probably the best way to nfs-export a CIFS filesystem is to use the user-space nfs server. It caches recently used filenames and uses a filehandle which is a hash of the name. It works on a best-effort basis, and if, for example, the server restarts, you will lose connections to open files. While it is not perfect, it can be very useful in some situations. NeilBrown