Re: Implementation choices, XEN or KVM?

[Andre Przywara <andre.przywara@amd.com>]

On 06/27/2011 12:40 PM, Tim Deegan wrote:
> At 18:30 +0800 on 27 Jun (1309199414), Wei Liu wrote:
>> On Mon, Jun 27, 2011 at 4:46 PM, Tim Deegan<Tim.Deegan@citrix.com>  wrote:
>>> At 16:56 -0400 on 24 Jun (1308934602), Liu John wrote:
>>>> I will do research experiment using hypervisor. But I have not decided the
>>>> platform; both XEN and KVM work for me(my work is about MMU, specifically
>>>> shadow page tables). Though I prefer XEN, it seems XEN development and
>>>> debugging are not easy. For example, every time you add some code in
>>>> hypervisor, you have to restart computer.
>>>
>>> That's true.   However I usually find that when I make a mistake in
>>> hypervisor code the computer needs to be rebooted anyway. :)
>>> AIUI Xen can run under Qemu so the reboot time is not that long.
>>>
>>
>> Just out of my curiosity.
>>
>> Does Xen inside QEMU supports HVM guest?
>
> I haven't tried it.  ISTR seeing patches for SVM support, at least, on
> qemu-devel a while back.  So: maybe. :)

As Tim said, QEMU implements some basic SVM emulation. But as this is 
quite old and nobody seemed to care, it is currently broken, AFAICT. At 
best it is not complete anyway and so probably unusable.

But you can use the nested SVM support in KVM. Though the primary test 
vehicle is KVM on KVM ;-), we also tests Xen as a guest from time to 
time and it worked. So just use QEMU and say --enable-kvm or use 
qemu-kvm on an AMD machine and you can run Xen with HVM guests inside KVM.

Regards,
Andre.

-- 
Andre Przywara
AMD-Operating System Research Center (OSRC), Dresden, Germany