From mboxrd@z Thu Jan 1 00:00:00 1970 From: Anthony Liguori Subject: Re: [PATCH] Add an ioctl interface for simple xenstore access Date: Sat, 18 Mar 2006 11:56:27 -0600 Message-ID: <441C49CB.2090308@us.ibm.com> References: <87r750pupr.fsf@caffeine.beaverton.ibm.com> <20060318071405.GD23816@granada.merseine.nu> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20060318071405.GD23816@granada.merseine.nu> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Muli Ben-Yehuda Cc: Dan Smith , Xen Developers List-Id: xen-devel@lists.xenproject.org Muli Ben-Yehuda wrote: > On Fri, Mar 17, 2006 at 02:33:36PM -0800, Dan Smith wrote: > > >> This patch adds an ioctl interface to /proc/xen/xenbus to allow simple >> access to xenstore from domU. This patch introduces only >> xenbus_read() support, but write, remove, etc could be easily added. >> Also, this interface can be easily moved to /dev/xen/xenbus (or >> something similar) later. >> > > Hmm, wouldn't a virtual file system be a much better fit for xenstore > access? the kernel community considers ioctls Evil with a capital E. > I'm not sure how one would express the transaction semantics of xenstore via a virtual filesystem. There is also an awkward mapping between some of the attributes of xenstore (for instance, there's no way to determine if an attribute returned by a list is a directory or not) that makes a filesystem non-trivial. If you remember, I had written a fuse filesystem front-end for xenstore at one point. It's definitely neat, but I think we would have to make some changes to Xenstore for it to really work out well. Regards, Anthony Liguori > Cheers, > Muli >