Re: [PATCH v3 2/6] KVM: arm64: PMU: Assume PMU presence in pmu-emul.c

[Oliver Upton <oliver.upton@linux.dev>]

Hi Akihiko,

On Wed, Mar 12, 2025 at 08:55:56PM +0900, Akihiko Odaki wrote:

[...]

> diff --git a/arch/arm64/kvm/guest.c b/arch/arm64/kvm/guest.c
> index 962f985977c2..fc09eec3fd94 100644
> --- a/arch/arm64/kvm/guest.c
> +++ b/arch/arm64/kvm/guest.c
> @@ -951,6 +951,10 @@ int kvm_arm_vcpu_arch_set_attr(struct kvm_vcpu *vcpu,
>  
>  	switch (attr->group) {
>  	case KVM_ARM_VCPU_PMU_V3_CTRL:
> +		if (!kvm_vcpu_has_pmu(vcpu)) {
> +			ret = -ENODEV;
> +			break;
> +		}
>  		mutex_lock(&vcpu->kvm->arch.config_lock);
>  		ret = kvm_arm_pmu_v3_set_attr(vcpu, attr);
>  		mutex_unlock(&vcpu->kvm->arch.config_lock);
> @@ -976,6 +980,10 @@ int kvm_arm_vcpu_arch_get_attr(struct kvm_vcpu *vcpu,
>  
>  	switch (attr->group) {
>  	case KVM_ARM_VCPU_PMU_V3_CTRL:
> +		if (!kvm_vcpu_has_pmu(vcpu)) {
> +			ret = -ENODEV;
> +			break;
> +		}
>  		ret = kvm_arm_pmu_v3_get_attr(vcpu, attr);
>  		break;
>  	case KVM_ARM_VCPU_TIMER_CTRL:
> @@ -999,6 +1007,10 @@ int kvm_arm_vcpu_arch_has_attr(struct kvm_vcpu *vcpu,
>  
>  	switch (attr->group) {
>  	case KVM_ARM_VCPU_PMU_V3_CTRL:
> +		if (!kvm_vcpu_has_pmu(vcpu)) {
> +			ret = -ENXIO;
> +			break;
> +		}
>  		ret = kvm_arm_pmu_v3_has_attr(vcpu, attr);
>  		break;
>  	case KVM_ARM_VCPU_TIMER_CTRL:

I agree with you for the most part on this patch, but I prefer we keep
the kvm_vcpu_has_pmu() with the ioctl implemementations rather than the
spot at which we demux the ioctl.

Thanks,
Oliver