From mboxrd@z Thu Jan 1 00:00:00 1970 From: George Dunlap Subject: Re: [PATCH 0/2] Handle xen_platform_pci=0 case Date: Fri, 22 Nov 2013 17:24:54 +0000 Message-ID: <528F9366.1050601@eu.citrix.com> References: <1385133191-23033-1-git-send-email-anthony.perard@citrix.com> <1385135773.25845.99.camel@kazak.uk.xensource.com> <20131122171834.GC10855@perard.uk.xensource.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; Format="flowed" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <20131122171834.GC10855@perard.uk.xensource.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Anthony PERARD , Ian Campbell Cc: Stefano Stabellini , Ian Jackson , Xen Devel List-Id: xen-devel@lists.xenproject.org On 22/11/13 17:18, Anthony PERARD wrote: > On Fri, Nov 22, 2013 at 03:56:13PM +0000, Ian Campbell wrote: >> On Fri, 2013-11-22 at 15:13 +0000, Anthony PERARD wrote: >>> There are two patches, the first one adds an option to specifies the QEMU >>> machine that a user wants and we handle the xen_platform_pci=0 case using the >>> new option. >> I'm thinking we should make this new option an enum of actually >> supported machine types, rather than just a free reign. Makes more sense >> from a "what do we support" PoV. > And start with the following list ? > pc > pc-i440fx-1.7 > pc-i440fx-1.6 > pc-i440fx-1.5 > pc-i440fx-1.4 > pc-1.3 > pc-1.2 > pc-1.1 > pc-1.0 > pc-0.15 > pc-0.14 > pc-0.13 > pc-0.12 > pc-0.11 > pc-0.10 > xenfv > > I wonder how many of this will work. I suspect most of them. The only > down side is that it will be necessary to patch libxl if one want to > try, let say q35. In a way I think that's the point -- by allowing the user to specify an arbitrary sting, we are implicitly supporting a gigantic range of things which may work in one release or in one particular configuration and then break in a different one. By making an enum, we bake in the exact ones which we consider to be valid. Also, we've been trying to get away from exposing a bunch of qemu-isms to the user via libxl and the xl config file. What might the *user* be trying to do in specifying this option? We should make sure the interface specifies things the user actually wants, and then use qemu's options to implement that, rather than asking the user to specify qemu stuff directly. -George