Re: [Xen-devel] [RFC][PATCH 03/10] xen/hybrid: Xen Hybrid Extension initialization

[Jeremy Fitzhardinge <jeremy@goop.org>]

On 09/16/09 23:22, Keir Fraser wrote:
>> I think having an option to put PV guests into an HVM container is a
>> good one, but as I mentioned in the other mail, I don't think this is
>> the right approach.
>>
>> It would be much better to make it so that an unmodified guest works in
>> such a mode; even with no specific optimisations the guest would get
>> benefit from faster kernel<->usermode switches.
>>     
> By unmodified you mean ordinary PV guest?

Right.

>  It's an interesting comparison --
> PVing an HVM guest, versus HVMing (to some extent) a PV guest.
>   

KVM is basically using the model of starting with a fully emulated hvm
domain, then adding paravirtualizations as incremental extensions to
that.  If you want to go that route, then we may as well just adopt
their interfaces and use the existing kernel support as-is (though their
most useful paravirtualization - time - is adopted from Xen's ABI).

If we want to get a PV kernel which makes use of hvm features, then we
should do the analogous thing in the other direction: use the current PV
ABI as baseline, then add small optional extensions to take advantage of
the HVM container's features.

    J