HVM guests are currently permitted to modify the memory and I/O
decode bits in the PCI command register of devices passed through to
them. Unless the device is an SR-IOV virtual function, after
disabling one or both of these bits subsequent accesses to the MMIO
or I/O port ranges would - on PCI Express devices - lead to
Unsupported Request responses. The treatment of such errors is
platform specific.
Furthermore (at least) devices under control of the Linux pciback
driver in the host are handed to guests with the aforementioned bits
turned off. This means that such accesses can similarly lead to
Unsupported Request responses until these flags are set as needed by
the guest.
In the event that the platform surfaces aforementioned UR responses
as Non-Maskable Interrupts, and either the OS is configured to treat
NMIs as fatal or (e.g. via ACPI's APEI) the platform tells the OS to
treat these errors as fatal, the host would crash, leading to a
Denial of Service.