From: Samuel Thibault Subject: Re: [PATCH] multiboot-like module support Date: Wed, 9 Dec 2009 15:59:15 +0100 Message-ID: <20091209145915.GD5233@const.bordeaux.inria.fr> References: <20091209140149.GF5233@const.bordeaux.inria.fr> <1260368975.23698.40977.camel@zakaz.uk.xensource.com> Mime-Version: 1.0 Content-Type: text/plain; charset=iso-8859-1 Content-Transfer-Encoding: quoted-printable Return-path: Content-Disposition: inline In-Reply-To: <1260368975.23698.40977.camel@zakaz.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: Ian Campbell Cc: "xen-devel@lists.xensource.com" List-Id: xen-devel@lists.xenproject.org Ian Campbell, le Wed 09 Dec 2009 14:29:35 +0000, a =E9crit : > On Wed, 2009-12-09 at 14:01 +0000, Samuel Thibault wrote: > > + * A multiboot module is a package containing modules very similar t= o > > a > > + * multiboot module array. The only differences are:=20 >=20 > I don't really know the multiboot layout but is there a reason not to > use it as is rather than defining a new version? It'd make the builder more tricky, as multiboot uses absolute physical addresses. That'd mean extending libxc in order to take the list of modules itself and build the table very late, during xc_dom_build_image(), when the target virtual address is known, while by using relative offsets, there is no need to, it's self-contained. Relative offsets also permit to build a package even before submitting it to xm create (that's how I'm using it atm actually). I'm not against it, it's just that it is much less simple. One good thing however is that then the code building the table would be in libxc, shared by pv-grub and xend. The other difference is how to know the number of modules. In the multiboot standard, it's simply set in the boot_info structure. That could be added to the start_info structure for instance. > I would imagine that using the standard layout could potentially lead t= o > less Xen specific code on the guest side. Well, yes, I'd just answer that there aren't so many multiboot implementations anyway :) Samuel