Re: [PATCH 1/5] xen/vm_event: Added support for XSETBV events

[Razvan Cojocaru <rcojocaru@bitdefender.com>]

On 05/08/2015 02:52 PM, Jan Beulich wrote:
>>>> On 08.05.15 at 13:05, <tamas.lengyel@zentific.com> wrote:
>> On Fri, May 8, 2015 at 12:55 PM, Andrew Cooper
>> <andrew.cooper3@citrix.com> wrote:
>>> On 08/05/15 11:53, Tamas K Lengyel wrote:
>>>> On Fri, May 8, 2015 at 11:10 AM, Andrew Cooper
>>>> <andrew.cooper3@citrix.com> wrote:
>>>>> On 08/05/15 10:06, Razvan Cojocaru wrote:
>>>>>> On 05/07/2015 09:03 PM, Andrew Cooper wrote:
>>>>>>> In an effort to be architecture neutral, it might be an idea to have
>>>>>>> something like
>>>>>>>
>>>>>>> struct vm_event_write_cr {
>>>>>>>     uint64_t index;
>>>>>>>     uint64_t old_val, new_val;
>>>>>>> };
>>>>>>>
>>>>>>> And have a per-arch index of control registers, such as
>>>>>>>
>>>>>>> X86_CR0
>>>>>>> X86_CR3
>>>>>>> X86_CR4
>>>>>>> X86_XCR0
>>>>>>> ...
>>>>>>> ARM32_$foo
>>>> On ARM there are no "cr" registers so IMHO it would be better to
>>>> rename the struct vm_event_write_register. Other than that this sounds
>>>> like a good addition to the interface.
>>>
>>> But there are surely the concept of "control registers" ?
>>>
>>> (I have no knowledge in this area)
>>>
>>> ~Andrew
>>
>> (Re-adding xen-devel)
>>
>> Certainly, they are just not (necessarily) called "CR". For example,
>> CR3 equivalent on ARM is TTBR1. So what I meant here is that naming
>> the struct should not be x86 specific.
> 
> In which case - vm_event_write_ctrlreg?

Looks good. Of course, the underlying footwork will need to stay just as
complicated - sync / enabled flags for each supported register, but the
interface will be cleaner and there will be less repetition for
xc_monitor_*() and hvm_event_*().


Thanks,
Razvan