Re: [Xen-devel] [PATCH v2] x86/HVM: don't #GP/#SS on wrapping virt->linear translations

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 06.07.17 at 11:21, <JBeulich@xxxxxxxx> wrote:
> Real hardware wraps silently, so we should behave the same. Also split
> real and VM86 mode handling, as the latter really ought to have limit
> checks applied.

AMD 32-bit behavior is more complicated, even beyond what is
being stated in the fix for XSA-186: For two-part accesses, the
first part wrapping will cause #GP/#SS, while the wrap being at
the boundary of the two parts or in the second part,
segmentation checks don't signal an exception (in the draft XTF
test sent yesterday #PF is being observed instead). I have to
admit that I don't consider this sane behavior to emulate, but of
course we could do so (but that's going to be clumsy afaict, as
the emulator would need to tell the ->read() hook that no wrap
check is to be performed on those second part reads).

64-bit behavior matches Intel's. (All from looking at a Fam15
CPU.)

Thoughts?

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
https://lists.xen.org/xen-devel