Re: [PATCH V2] xen: vmx: Use an INT 2 call to process real NMI's instead of self_nmi() in VMEXIT handler

[Mats Petersson <mats.petersson@citrix.com>]

On 16/11/12 08:17, Jan Beulich wrote:
>>>> On 15.11.12 at 18:25, Tim Deegan <tim@xen.org> wrote:
>> At 16:52 +0000 on 15 Nov (1352998340), Andrew Cooper wrote:
>>> It is also possible to get a reentrant NMI if there is a pagefault (or
>>> handful of other possible faults) when trying to execute the iret of
>>> the NMI itself; NMIs can get re-enabled from the iret of the
>>> pagefault, and we take a new NMI before attempting to retry the iret
>>> from the original NMI.
>> Yes, I hadn't thought of that case.
> But what would make a fault happen on that IRET? Oh, yes,
> there is one case - the guest having its previous instruction end
> exactly at the canonical/non-canonical boundary. But for the
> sake of correctness, that's a #GP then. I would suppose this
> would better be filtered (manually injecting a #GP into the guest)
> than allowed to actually cause a #GP.
Or, if for some reason the address we return to is "not present". Now, 
in the current Xen, Xen itself doesn't get paged out, but in a PV guest, 
I'm pretty certain the guest could decide to page out some code-page, 
which just happens to be the one we were about to return to?

--
Mats
>
> Jan
>
>
> _______________________________________________
> Xen-devel mailing list
> Xen-devel@lists.xen.org
> http://lists.xen.org/xen-devel
>
>