Re: [PATCH v5 09/26] KVM: VMX: nVMX: Support TSC scaling and PERF_GLOBAL_CTRL with enlightened VMCS

[Vitaly Kuznetsov <vkuznets@redhat.com>]

Sean Christopherson <seanjc@google.com> writes:

> On Fri, Aug 19, 2022, Vitaly Kuznetsov wrote:
>> Sean Christopherson <seanjc@google.com> writes:
>> 
>> > On Tue, Aug 02, 2022, Vitaly Kuznetsov wrote:
>> >> diff --git a/arch/x86/kvm/vmx/evmcs.h b/arch/x86/kvm/vmx/evmcs.h
>> >> index f886a8ff0342..4b809c79ae63 100644
>> >> --- a/arch/x86/kvm/vmx/evmcs.h
>> >> +++ b/arch/x86/kvm/vmx/evmcs.h
>> >> @@ -37,16 +37,9 @@ DECLARE_STATIC_KEY_FALSE(enable_evmcs);
>> >>   *	EPTP_LIST_ADDRESS               = 0x00002024,
>> >>   *	VMREAD_BITMAP                   = 0x00002026,
>> >>   *	VMWRITE_BITMAP                  = 0x00002028,
>> >> - *
>> >> - *	TSC_MULTIPLIER                  = 0x00002032,
>> >>   *	PLE_GAP                         = 0x00004020,
>> >>   *	PLE_WINDOW                      = 0x00004022,
>> >>   *	VMX_PREEMPTION_TIMER_VALUE      = 0x0000482E,
>> >> - *      GUEST_IA32_PERF_GLOBAL_CTRL     = 0x00002808,
>> >> - *      HOST_IA32_PERF_GLOBAL_CTRL      = 0x00002c04,
>> >> - *
>> >> - * Currently unsupported in KVM:
>> >> - *	GUEST_IA32_RTIT_CTL		= 0x00002814,
>> >
>> > Almost forgot: is deleting this chunk of the comment intentional?
>> >
>> 
>> Intentional or not (I forgot :-), GUEST_IA32_RTIT_CTL is supported/used
>> by KVM since
>> 
>> commit f99e3daf94ff35dd4a878d32ff66e1fd35223ad6
>> Author: Chao Peng <chao.p.peng@linux.intel.com>
>> Date:   Wed Oct 24 16:05:10 2018 +0800
>> 
>>     KVM: x86: Add Intel PT virtualization work mode
>> 
>> ...
>>  
>> commit bf8c55d8dc094c85a3f98cd302a4dddb720dd63f
>> Author: Chao Peng <chao.p.peng@linux.intel.com>
>> Date:   Wed Oct 24 16:05:14 2018 +0800
>> 
>>     KVM: x86: Implement Intel PT MSRs read/write emulation
>> 
>> but there's no corresponding field in eVMCS. It would probably be better
>> to remove "Currently unsupported in KVM:" line leaving
>> 
>> "GUEST_IA32_RTIT_CTL             = 0x00002814" 
>> 
>> in place. 
>
> GUEST_IA32_RTIT_CTL isn't supported for nested VMX though, which is how I
> interpreted the "Currently unsupported in KVM:".  Would it be accurate to extend
> that part of the comment to "Currently unsupported in KVM for nested VMX:"?

Yea, sounds good to me. 

FWIW, there are other controls which are currently missing in KVM,
e.g. 'guest_ia32_lbr_ctl' (VMX field 0x2816) and we have no 'macro
shenanigans' to catch the moment when support for these gets introduced
in KVM. When this happens, we need to extend VMCS-to-eVMCS mapping
(vmcs_field_to_evmcs_1) to support KVM-on-Hyper-V case and, when the
corresponding field gets added to 'struct vmcs12',
copy_enlightened_to_vmcs12()/copy_vmcs12_to_enlightened(). The whole
process is manual and thus error prone...

-- 
Vitaly