Re: PATCH: setup_vmcs_config: disable TSC scaling on unlike processors

["Radim Krčmář" <rkrcmar@redhat.com>]

2016-12-02 15:58-0500, Don Bowman:
> On 2 December 2016 at 14:10, Don Bowman <db@donbowman.ca> wrote:
>> On 2 December 2016 at 10:07, Radim Krčmář <rkrcmar@redhat.com> wrote:
>>> 2016-12-01 21:32-0500, Don Bowman:
>>>> My system has what i thought were two identical processors (same
>>>> stepping ID etc).
>>>> However, bafflingly, one of them has the ability to do TSC scaling,
>>>> and one does not (as reported in the vmcs).
>>>
> OK, how about this? The check has to be in setup_vmcs_config() not
> elsewhere I think. This is where the rdmsr occurs, and immediately
> following that is the compare against the other processor(s). Unless
> I'm missing something I don't see how vmx_secondary_exec_control()
> could work. For the enable I was following the 'enable_pml' which is
> already there, but have changed it below. vmx_check_processor_compat()
> calls setup_vmcs_config() and then the memcmp() immediately
> afterwards.

Right, KVM checks this early ... I don't like that the patch treats
tsc_scaling specially while the same could happen with some other
feature.  What about warning in vmx_check_processor_compat() if features
that won't be used don't match, but letting the check pass?

(I'd even prefer an unsafe option to disable the check than to treat
 tsc_scaling differently.)