[Fedora-xen] Xen, Fedora, and UEFI Secure Boot

[Konrad Rzeszutek Wilk]

On Mon, Jul 16, 2012 at 10:11:50AM +0300, Pasi Kärkkäinen wrote:
> On Sat, Jul 14, 2012 at 12:50:35PM +0100, M A Young wrote:
> > I contacted the people behind the the Fedora Seure Boot feature and
> > got the following responses, from Peter Jones:
> > 
> 
> Thanks for doing this!
> 
> > Okay, to be honest I don't remember much about Xen's layout - dom0 is the
> > management kernel the hypervisor starts? So, depending on how xen works,
> > there
> > are probably more things that need to be done in the hypervisor than
> > in the
> > kernel, because the hypervisor is the part that does most physical memory
> > accesses, and that's where there's a worry about faking SB=0 and launching
> > windows.
> > 
> > At the very least, the hypervisor will a) need to be an efi binary, and b)
> > need to be signed with the fedora kernel-signing key. It may also
> > need to be
> > audited for any command line options that allow physical memory access or
> > other similar things, analogous to Matthew's kernel patch for linux.
> > 
> > We're still working out with rel-eng how getting things signed with
> > that is
> > going to work. I don't think there's really any necessity that it's
> > announced
> > in a proper Feature, but if you feel like going that way, that's fine too.
> > 
> > and from Matthew Garrett:
> > 
> > Right. We can conceivably sign Xen as long as it's an EFI binary, but
> > I'd expect that it would have to enforce secure boot itself using the
> > host databases.
> > 
> > ------
> > 
> > So we need to get xen working with EFI, to lock xen down so it can't
> > be used to get around Secure Boot, and probably need to do some
> > enforcement of secure boot as well.
> > 
> 
> I think you already know this, but Suse guys (Jan) made xen.efi working,
> the patches are in xen-unstable (so in upcoming Xen 4.2), and they've
> backported the efi patches to Xen 4.1 in SLES11SP2, and also 
> Suse's (traditional) Linux 3.0 dom0 kernel has some patches for EFI.


<nods> I think MA Young already has an Xen 4.2 RPM - so the functionality
for that is there. It is just the matter of compiling it with a 
GCC compiler that can do PE.

Then there is the Linux upstream kernel thing. Daniel Kiper has volunteered
to look at this - but his schedule is a bit busy with kexec upstreaming.

I've no idea what 'enforce secure boot itself using the host database'
means. It probably means making some extra EFI calls maybe?