Re: [Xen-devel] Experimental results for VGA passthrough

[Jean Guyader <jean.guyader@xxxxxxxxx>]

Yes, here are the value we need to fake and the raison why:
  - GMCH, when the vgabios execute at actually check this register in
the host bridge (bdf 0:0.0, offset 0x52). On the intel platform this
register is there to identify the amount of stolen memory and
therefore the supported resolution in legacy mode (VGA/VESA).
  - VID/DID. Linux check that. If it doesn't match with the graphic
card it doesn't reconise the chipset and doesn't create the agpgart.
  - OpRegion address. Not sure about this one, but when you
shutdown/sleep vista if this value is present the driver will try to
save some information in a ACPI NVS region.

All those register are public you can find some more information about
those here:
www.intel.com/Assets/PDF/datasheet/319970.pdf

Jean

2009/6/8 Beng Heng, Ng <bengheng@xxxxxxxxxxxxxx>:
> Also, you mentioned earlier that we need to return some fake values to
> the host bridge. Would appreciate if you could explain, or refer me to
> some writeups on this. Thanks.
>
> - Beng Heng
>
> Beng Heng, Ng wrote:
>> Could you elaborate? Thanks.
>>
>> - Beng Heng
>>
>> Jean Guyader wrote:
>>> 2009/6/8 Beng Heng, Ng <bengheng@xxxxxxxxxxxxxx>:
>>>> Uses the driver included in the patch to initialize the graphics card.
>>>> This is my own flag. I'm not sure what Jean's plans are.
>>> The intel code for qemu you've found is not made for that.
>>>
>>> Jean
>>>
>>> _______________________________________________
>>> Xen-devel mailing list
>>> Xen-devel@xxxxxxxxxxxxxxxxxxx
>>> http://lists.xensource.com/xen-devel
>>
>>
>> _______________________________________________
>> Xen-devel mailing list
>> Xen-devel@xxxxxxxxxxxxxxxxxxx
>> http://lists.xensource.com/xen-devel
>
>

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel