Re: [PATCH] xen/vm_event: Clean up control-register-write vm_events

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

On 12/05/15 17:52, Razvan Cojocaru wrote:
> On 05/12/2015 06:53 PM, Andrew Cooper wrote:
>> On 12/05/15 16:50, Andrew Cooper wrote:
>>> On 12/05/15 16:49, Razvan Cojocaru wrote:
>>>> On 05/12/2015 06:35 PM, Jan Beulich wrote:
>>>>>>>> On 12.05.15 at 16:58, <rcojocaru@bitdefender.com> wrote:
>>>>>> +/* Supported values for the vm_event_write_ctrlreg index. */
>>>>>> +#define X86_CR0     (1 << 0)
>>>>>> +#define X86_CR3     (1 << 1)
>>>>>> +#define X86_CR4     (1 << 2)
>>>>>> +#define X86_XCR0    (1 << 3)
>>>>> These names, being put in the public interface, are way too generic.
>>>> I've copied them from Andrew Cooper's suggestion in the previous thread,
>>>> hopefully he'll chime in. I'm happy to rename / move them if so desired.
>>> I intended something like VM_EVENT_ARCH_X86_CR0 etc.
>> or perhaps slightly more specific to ctrl_reg read/write.
> VM_EVENT_ARCH_X86_MOV_TO_CR0? Or maybe VM_EVENT_WRITE_ARCH_X86_CR0?

Are there any other events which could plausibly use the same indicies? 
Monitoring ctrl_reg reads perhaps ?

>
> There's something else that should also be clarified: I've renamed the
> vm_event.h struct to vm_event_write_ctrlreg, and changed
> monitor.mov_to_cr* to monitor.write_ctrlreg*, but left
> xc_monitor_mov_to_cr() in libxc, and mov_to_cr in the domctl part (I
> thought it fit with mov_to_msr(), etc.). Should I change that to
> write_ctrlreg as well? On the one hand, mov_to_cr seems to be consistent
> with the previous interface, and on the other it's not consistent with
> the name changes the patch has made.

Any API/ABI changes in this area are fine until 4.6 is released, and a
consistent interface is much preferred.

~Andrew