Re: [XEN PATCH] tools/libs/light/libxl_pci.c: explicitly grant access to Intel IGD opregion

[Chuck Zmudzinski <brchuckz@xxxxxxxxxxxx>]

On 3/15/22 7:38 AM, Jan Beulich wrote:
> On 14.03.2022 04:41, Chuck Zmudzinski wrote:
>
> > +        ret = xc_domain_iomem_permission(CTX->xch, stubdom_domid,
> > +                                         vga_iomem_start,
> > +                                         IGD_OPREGION_PAGES, 1);
> > +        if (ret < 0) {
> > +            LOGED(ERROR, domid,
> > +                  "failed to give stubdom%d access to iomem range "
> > +                  "%"PRIx64"-%"PRIx64" for IGD passthru",
> > +                  stubdom_domid, vga_iomem_start, (vga_iomem_start +
> > +                                                IGD_OPREGION_PAGES - 1));
> > +            return ret;
> > +        }
> I have to admit that I find it odd that this is done unconditionally,
> but I notice the same is done in pre-existing code. I would have
> expected this to happen only when there actually is a device model
> stub domain.
>
> Jan
I dumped the value of stubdom_id for my tests with the
device model running in dom0:

libxl: info: libxl_pci.c:2556:libxl__grant_vga_iomem_permission: Domain 
3: stubdom id: 0
As I expected, when there is not a device model stub domain
and the device model runs in dom0, the stubdom_id is 0.

I will now do some tests to see if this is necessary when the
device model runs in dom0. I would like to know if the device
model running in dom0 needs to have access granted here
or not. When there is a device model stub domain, I presume
it is necessary, and I can check that also and write the
next version of the patch accordingly.

Chuck