On 07/26/11 15:19, Tim Deegan wrote:
> At 14:11 +0200 on 26 Jul (1311689501), Christoph Egger wrote:
>> On 07/25/11 12:30, Tim Deegan wrote:
>>> At 11:25 +0100 on 25 Jul (1311593146), Tim Deegan wrote:
>>>> Also this function should be called from the CPUID trap handler to make
>>>> sure we never advertise PSE1GB when we're not going to support it.
>>>
>>> Er, not this function exactly, since CPUID should report the feature
>>> even when the guest's not in long_mode.  I think it needs a
>>> hvm_pse1G_supported that can be called from CPUID, and then
>>> guest_supports_1G_superpages() boils down to
>>> "(GUEST_PAGING_LEVEL>= 4)&&   hvm_pse1G_supported(v)"
>>
>> New version attached. I removed the fake l1e calculation.
>
> If you just remove it, you need to update hap_p2m_ga_to_gfn to figure
> out the GFN some other way!  I expect that's what's causing your
> problem.  You should either provide the fake l1e, and say why in the
> comment, or audit all callers of the function to make sure they don't
> need it.

Ah, I see now. I misunderstood you.

I forgot to increase the pfn by the right number of 2m pages for
the l1e calculation.

Now it works for me: The l2 guest boots and passes my tests.



-- 
---to satisfy European Law for business letters:
Advanced Micro Devices GmbH
Einsteinring 24, 85689 Dornach b. Muenchen
Geschaeftsfuehrer: Alberto Bozzo, Andrew Bowd
Sitz: Dornach, Gemeinde Aschheim, Landkreis Muenchen
Registergericht Muenchen, HRB Nr. 43632