Re: [Xen-devel] [PATCH 3/3] x86/HVM: make hvm_efer_valid() honor guest features

[Andrew Cooper <andrew.cooper3@xxxxxxxxxx>]

On 09/01/15 15:33, Jan Beulich wrote:
>>>> On 09.01.15 at 16:09, <andrew.cooper3@xxxxxxxxxx> wrote:
>> On 09/01/15 11:20, Jan Beulich wrote:
>>>>>> On 08.01.15 at 19:49, <andrew.cooper3@xxxxxxxxxx> wrote:
>>>> I think hvm_efer_valid() also needs the current EFER and CR0 to work out
>>>> what the current LMA should be, and reject any attempt to change it.
>>> I.e. this would be needed only for the restore path (and it not being
>>> done hasn't caused problems for me because guests don't normally
>>> get migrated before they fully enabled long mode).
>> Urgh.  In the restore case we don't have a current EFER to use.  In this
>> case I think it is acceptable to check that new_lma is (new_lme &&
>> !cr4.pg), which allows us to declare that the restore state is
>> consistent (presuming we have already validated cr4).
> I suppose you mean cr0 here, and also new_lma == new_lme && cr0.pg?

Oops yes - I do on both counts.

~Andrew

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