Re: [Xen-devel] Re: PCI passthrough support via PV-PCI

[Keir Fraser <keir.fraser@xxxxxxxxxxxxx>]

On 2/7/08 15:28, "Samuel Thibault" <samuel.thibault@xxxxxxxxxxxxx> wrote:

>> When I added VT-d support for PV guests I made xend do the device
>> assignment iff the destination was a PV domain.  Is there any reason
>> why device assignment can not be performed by xend for HVM guests as
>> well?
>> 
>> Other resource assignment (IOMEM and IRQs) still seems to happen in
>> xend for HVM guests as well as for PV guest.  What other domctls does
>> ioemu need to perform?
> 
> ioemu needs to update the mappings, see the pt_iomem_map() function.

I suggest the domctls should set up access permissions (e.g., by use of
existing iomem_premission domctl). And then extend XENMEM_add_to_physmap to
allow (re-)mapping of permitted I/O ranges into pseudophysical address
space. The new passthru-specific domctls may well be ill conceived.

 -- Keir



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