Re: [V3 PATCH 7/9] x86/hvm: pkeys, add pkeys support for guest_walk_tables

[Andrew Cooper <andrew.cooper3@citrix.com>]

On 11/12/2015 07:18, Han, Huaitong wrote:
> On Thu, 2015-12-10 at 18:59 +0000, Andrew Cooper wrote:
>> On 07/12/15 09:16, Huaitong Han wrote:
>>> +
>>> +    /* PKRU dom0 is always zero */
>>> +    if ( likely(!pte_pkeys) )
>>> +        return 0;
>> This is not an architectural restriction (as far as I can tell).  Xen
>> must never make assumptions about how a guest chooses to use a
>> feature.
> Yes, I will remove the code section.
>
> On my test machine, writing non-zero value to PKRU domain 0 will lead
> to a GP fault, this is also conform to the usage scenario, and I will
> ask the intel architect the reason why SDM does not describe.

If this is (or is going to be documented as) an architectural
restriction, then the code can stay, although I would adjust the comment to

/* Architecturally, pkey 0 is always zero. */

Thinking about it, the restriction does make sense.  It equates to "the
default behaviour of things before the PKRU feature came along".

~Andrew