Re: [Xen-devel] PCI Pass-through in Xen ARM: Draft 4

[Stefano Stabellini <stefano.stabellini@xxxxxxxxxxxxx>]

On Fri, 14 Aug 2015, Jan Beulich wrote:
> >>> On 14.08.15 at 16:45, <julien.grall@xxxxxxxxxx> wrote:
> > On 14/08/15 15:37, Stefano Stabellini wrote:
> >> If you are thinking of Windows for ARM64, there isn't one yet.  When/if
> >> it will become available, there are going to be a number of issues to
> >> address before we can run it as a guest on Xen on ARM with the current
> >> architecture which doesn't do emulation. I am hopeful that we'll be able
> >> to have a discussion on ABI issues such as this one.
> > 
> > It may be worth to read [1] where the Xen ARM architecture is explained.
> > 
> > Regards,
> > 
> > [1]
> > http://wiki.xen.org/wiki/Xen_ARM_with_Virtualization_Extensions_whitepaper 
> 
> Neither Stefano's nor your reply really make a lot of sense to me: I
> realize you currently require a cooperating guest, but when
> designing something like pass-through it would seem natural to me to
> try to allow for other guest kinds in the future.

Xen x86 was designed to work around existing guests. In that context you
had un-cooperative guests, such as Windows, and cooperative guests, such
as Linux, which could be extensively modified to run on Xen.

This is not the case on ARM. Please scratch all preconceptions.

On ARM we only have one kind of guests: operating systems that have been
ported to the published Xen ABIs. There are no legacy OSes to work
around. If a new OS comes around, free or proprietary, I expect it to be
ported to the Xen ABI. And that is not much of an effort, because we
designed the interface from the start to be clean, nice and to "make
sense". Linux and FreeBSD didn't have to be extensively modified to run
on Xen on ARM. This is why I am also opposed to some of your comments on
ACPI for ARM, for example.

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel