Re: [PATCH v4 3/4] KVM: nSVM: implement the nested idle halt intercept

[Manali Shukla <manali.shukla@amd.com>]

On 12/20/2024 6:31 AM, Sean Christopherson wrote:

>> diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
>> index e86b79e975d3..38d546788fc6 100644
>> --- a/arch/x86/kvm/svm/svm.c
>> +++ b/arch/x86/kvm/svm/svm.c
>> @@ -4425,6 +4425,7 @@ static void svm_vcpu_after_set_cpuid(struct kvm_vcpu *vcpu)
>>  	kvm_governed_feature_check_and_set(vcpu, X86_FEATURE_PFTHRESHOLD);
>>  	kvm_governed_feature_check_and_set(vcpu, X86_FEATURE_VGIF);
>>  	kvm_governed_feature_check_and_set(vcpu, X86_FEATURE_VNMI);
>> +	kvm_governed_feature_check_and_set(vcpu, X86_FEATURE_IDLE_HLT);
>>  
>>  	svm_recalc_instruction_intercepts(vcpu, svm);
>>  
>> @@ -5228,6 +5229,9 @@ static __init void svm_set_cpu_caps(void)
>>  		if (vnmi)
>>  			kvm_cpu_cap_set(X86_FEATURE_VNMI);
>>  
>> +		if (cpu_feature_enabled(X86_FEATURE_IDLE_HLT))
>> +			kvm_cpu_cap_set(X86_FEATURE_IDLE_HLT);
> 
> kvm_cpu_cap_check_and_set() does this for you.
> 

Sure. I will use kvm_cpu_cap_check_and_set() in V5.

>> +
>>  		/* Nested VM can receive #VMEXIT instead of triggering #GP */
>>  		kvm_cpu_cap_set(X86_FEATURE_SVME_ADDR_CHK);
>>  	}
>> -- 
>> 2.34.1
>>

- Manali