On 9/16/2015 12:19 PM, Alexandra Yates wrote:
> Hi Jaroslav,
>
> On 05/11/2015 09:54 AM, Arjan van de Ven wrote:
>> but it's really not the right thing.
>> the reason is that the statistics code still uses the "native intel msr" model,
>> so it gets all zeroes... it should use "just trust the kernel" model in
> Based on Arjan's feedback do you have a change for this patch? Otherwise I would have to reject it.

we do need to solve this problem... it's real.
just we need to really solve it, not just a workaround.

is not insane hard (we'd need to do a test read msr, if that fails because of broken virtualization, fall back to the degraded mode automatically)