From: Andres Lagar-Cavilla Subject: Re: [PATCH 14/14]: libxenlight, checks before device deletion and addition Date: Tue, 01 Dec 2009 09:24:38 -0500 Message-ID: <4B152726.3070607@lagarcavilla.com> References: <4B141A7A.3090308@lagarcavilla.com> <20091201062935.GE15791@gwig.uk.xensource.com> Mime-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1; format=flowed Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20091201062935.GE15791@gwig.uk.xensource.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xensource.com Errors-To: xen-devel-bounces@lists.xensource.com To: Vincent Hanquez Cc: "xen-devel@lists.xensource.com" , Stefano Stabellini List-Id: xen-devel@lists.xenproject.org Agreed in that it is a weak check being non-transactional. But some check should be there, shouldn't it? Just throwing it out there. The other checks are kosher I assume? Andres Vincent Hanquez wrote: > On Mon, Nov 30, 2009 at 07:18:18PM +0000, Andres Lagar-Cavilla wrote: > >> Proper existential checks before adding or deleting a >> device, i.e. return error if: >> >> - Adding a device to a non-existent domain >> - Adding a device twice >> - Deleting a non-existent device >> - Deleting a device from a non-existent domain >> >> Signed-off-by: Andres Lagar-Cavilla >> > > > you should not check the non-existency of a domain. this is pointless > since just after you check the domain might have gone away. also you might > think of a mode where you populate the xenstore tree before creating the domain > .. nobody does that, but I don't think there's any major stopper to do that or > that it would be a bad idea either. > > Same reason that should it not check for device existency outside of the > transaction, otherwise someome might have added it in the meantime. > > PS: please don't call libxl_free, that's really unnecessary at this stage, > until we sort memory handling completly. (there's a "GC" for now) > >