Re: [PATCH v2 1/2] VT-d: avoid NULL deref on domain_context_mapping_one() error paths

[Jason Andryuk <jandryuk@xxxxxxxxx>]

On Fri, Apr 8, 2022 at 2:03 AM Jan Beulich <jbeulich@xxxxxxxx> wrote:
>
> On 07.04.2022 18:31, Jason Andryuk wrote:
> > As I understand it, an RMRR is common with USB controllers for
> > implementing legacy mouse & keyboard support.  The Cannon Point PCH is
> > fairly modern, so I'd expect it to use PCI Express.  Xen seems to
> > identify it as DEV_TYPE_PCI given "PCI" above.  It is an integrated
> > PCI device, so it has no upstream bridge.  Maybe that is why it can
> > still be assigned?  The "unsupported assignment" is a legacy PCI
> > device, behind a bridge, with an RMRR?
>
> Ah yes - the "behind a bridge" aspect does matter (but I can't
> adjust the description of an already committed patch). That's both
> for the respective part of the XSA-400 series as well as for the
> change here.

Ok.  Thank you for confirming.  It is captured in the mailing list
archive at least.

Regards,
Jason