Re: [PATCH] KVM: x86: Provide a capability to disable APERF/MPERF read intercepts

[Paolo Bonzini <pbonzini@redhat.com>]

On 3/14/25 14:59, Sean Christopherson wrote:
> On Thu, Mar 13, 2025, Jim Mattson wrote:
>> On Mon, Feb 24, 2025 at 4:47 PM Jim Mattson <jmattson@google.com> wrote:
>>>
>>> Allow a guest to read the physical IA32_APERF and IA32_MPERF MSRs
>>> without interception.
>>>
>>> The IA32_APERF and IA32_MPERF MSRs are not virtualized. Writes are not
>>> handled at all. The MSR values are not zeroed on vCPU creation, saved
>>> on suspend, or restored on resume. No accommodation is made for
>>> processor migration or for sharing a logical processor with other
>>> tasks. No adjustments are made for non-unit TSC multipliers. The MSRs
>>> do not account for time the same way as the comparable PMU events,
>>> whether the PMU is virtualized by the traditional emulation method or
>>> the new mediated pass-through approach.
>>>
>>> Nonetheless, in a properly constrained environment, this capability
>>> can be combined with a guest CPUID table that advertises support for
>>> CPUID.6:ECX.APERFMPERF[bit 0] to induce a Linux guest to report the
>>> effective physical CPU frequency in /proc/cpuinfo. Moreover, there is
>>> no performance cost for this capability.
>>>
>>> Signed-off-by: Jim Mattson <jmattson@google.com>
>>> ---
> 
> ...
> 
>> Any thoughts?
> 
> It's absolutely absurd, but I like it.  I would much rather provide functionality
> that is flawed in obvious ways, as opposed to functionality that is flawed in
> subtle and hard-to-grok ways.  Especially when the former is orders of magnitude
> less complex.
> 
> I have no objections, so long as we add very explicit disclaimers in the docs.
> 
> FWIW, the only reason my response was delayed is because I was trying to figure
> out if there's a clean way to avoid adding a large number of a capabilities for
> things like this.

True but it's not even a capability, it's just a new bit in the existing 
KVM_CAP_X86_DISABLE_EXITS.

Just one question:

> -       u64 r = KVM_X86_DISABLE_EXITS_PAUSE;
> +       u64 r = KVM_X86_DISABLE_EXITS_PAUSE | KVM_X86_DISABLE_EXITS_APERFMPERF;

Should it be conditional on the host having the APERFMPERF feature 
itself?  As is the patch _does_ do something sensible, i.e. #GP, but 
this puts the burden on userspace of checking the host CPUID and 
figuring out whether it makes sense to expose the feature to the guest. 
It would be simpler for userspace to be able to say "if the bit is there 
then enable it and make it visible through CPUID".

Paolo