Ian Campbell, le Fri 11 Dec 2009 09:59:44 +0000, a écrit :
> On Fri, 2009-12-11 at 08:09 +0000, Keir Fraser wrote:
> > On 11/12/2009 02:44, "Samuel Thibault" <samuel.thibault@xxxxxxxxxxxx> wrote:
> > > Samuel Thibault, le Wed 09 Dec 2009 15:59:15 +0100, a écrit :
> > >> It'd make the builder more tricky, as multiboot uses absolute
> > >> physical addresses.
> > >
> > > Here is a patch.
> > Using my metric of "which patch is way shorter, and doesn't touch
> > start_info", I like the old patch rather more. :-)
> I think I agree. I initially asked because I wondered what the reasons
> were for the difference in layout, but Samuel's reasons seemed, well,
> reasonable ;-)
> This version doesn't seem too bad though.
Yes, it didn't go as badly as I thought. The addition of mods_count
could perhaps be avoided by saying that the multiboot module table
always ends with a NULL entry. That is still compliant with multiboot,
it just makes the PV guest have to count the modules itself (but not
have to rebase all the pointers).
About "which patch is way shorter", it could actually be useful, for
the first version, to move the code building the package into libxc, to
share it with future xm support, and we end up with basically the same
Xen-devel mailing list