On Fri, Aug 28, 2020 at 7:51 PM Xiaoyao Li <xiaoyao.li@xxxxxxxxx> wrote:
>
> On 8/29/2020 9:49 AM, Chenyi Qiang wrote:
> >
> >
> > On 8/29/2020 1:43 AM, Jim Mattson wrote:
> >> On Fri, Aug 28, 2020 at 1:54 AM Chenyi Qiang <chenyi.qiang@xxxxxxxxx>
> >> wrote:
> >>>
> >>> KVM supports the nested VM_{EXIT, ENTRY}_LOAD_IA32_PERF_GLOBAL_CTRL and
> >>> VM_{ENTRY_LOAD, EXIT_CLEAR}_BNDCFGS, but they doesn't expose during
> >>> the setup of nested VMX controls MSR.
> >>>
> >>
> >> Aren't these features added conditionally in
> >> nested_vmx_entry_exit_ctls_update() and
> >> nested_vmx_pmu_entry_exit_ctls_update()?
> >>
> >
> > Yes, but I assume vmcs_config.nested should reflect the global
> > capability of VMX MSR. KVM supports these two controls, so should be
> > exposed here.
>
> No. I prefer to say they are removed conditionally in
> nested_vmx_entry_exit_ctls_update() and
> nested_vmx_pmu_entry_exit_ctls_update().
>
> Userspace calls vmx_get_msr_feature() to query what KVM supports for
> these VMX MSR. In vmx_get_msr_feature(), it returns the value of
> vmcs_config.nested. As KVM supports these two bits, we should advertise
> them in vmcs_config.nested and report to userspace.
It would be nice if there was an API to query what MSR values KVM
supports for a specific VCPU configuration, but given that this is a
system ioctl, I agree with you.
