Re: [RFC PATCH 04/13] KVM: SVM: Only call vcpu_(un)blocking when AVIC is enabled.

[Suravee Suthikulpanit]

Hi Maxim,

On 2/24/22 11:54 PM, Maxim Levitsky wrote:
> On Sun, 2022-02-20 at 20:19 -0600, Suravee Suthikulpanit wrote:
> > The kvm_x86_ops.vcpu_(un)blocking are needed by AVIC only.
> > Therefore, set the ops only when AVIC is enabled.
> >
> > Suggested-by: Sean Christopherson<seanjc@xxxxxxxxxx>
> > Signed-off-by: Suravee Suthikulpanit<suravee.suthikulpanit@xxxxxxx>
> > ---
> >   arch/x86/kvm/svm/avic.c | 12 ++++++++++--
> >   arch/x86/kvm/svm/svm.c  |  7 -------
> >   arch/x86/kvm/svm/svm.h  |  2 --
> >   3 files changed, 10 insertions(+), 11 deletions(-)
> >
> > diff --git a/arch/x86/kvm/svm/avic.c b/arch/x86/kvm/svm/avic.c
> > index abde08ca23ab..0040824e4376 100644
> > --- a/arch/x86/kvm/svm/avic.c
> > +++ b/arch/x86/kvm/svm/avic.c
> > @@ -996,7 +996,7 @@ void avic_vcpu_put(struct kvm_vcpu *vcpu)
> >   	WRITE_ONCE(*(svm->avic_physical_id_cache), entry);
> >   }
> >   
> > -void avic_vcpu_blocking(struct kvm_vcpu *vcpu)
> > +static void avic_vcpu_blocking(struct kvm_vcpu *vcpu)
> >   {
> >   	if (!kvm_vcpu_apicv_active(vcpu))
> >   		return;
> > @@ -1021,7 +1021,7 @@ void avic_vcpu_blocking(struct kvm_vcpu *vcpu)
> >   	preempt_enable();
> >   }
> >   
> > -void avic_vcpu_unblocking(struct kvm_vcpu *vcpu)
> > +static void avic_vcpu_unblocking(struct kvm_vcpu *vcpu)
> >   {
> >   	int cpu;
> >   
> > @@ -1057,6 +1057,14 @@ bool avic_hardware_setup(struct kvm_x86_ops *x86_ops)
> >   		pr_info("x2AVIC enabled\n");
> >   	}
> >   
> > +	if (avic_mode) {
> > +		x86_ops->vcpu_blocking = avic_vcpu_blocking;
> > +		x86_ops->vcpu_unblocking = avic_vcpu_unblocking;
> > +	} else {
> > +		x86_ops->vcpu_blocking = NULL;
> > +		x86_ops->vcpu_unblocking = NULL;
> > +	}
> > +
> >   	amd_iommu_register_ga_log_notifier(&avic_ga_log_notifier);
> >   	return !!avic_mode;
> >   }
> > diff --git a/arch/x86/kvm/svm/svm.c b/arch/x86/kvm/svm/svm.c
> > index 3048f4b758d6..3687026f2859 100644
> > --- a/arch/x86/kvm/svm/svm.c
> > +++ b/arch/x86/kvm/svm/svm.c
> > @@ -4531,8 +4531,6 @@ static struct kvm_x86_ops svm_x86_ops __initdata = {
> >   	.prepare_guest_switch = svm_prepare_guest_switch,
> >   	.vcpu_load = svm_vcpu_load,
> >   	.vcpu_put = svm_vcpu_put,
> > -	.vcpu_blocking = avic_vcpu_blocking,
> > -	.vcpu_unblocking = avic_vcpu_unblocking,
> >   
> >   	.update_exception_bitmap = svm_update_exception_bitmap,
> >   	.get_msr_feature = svm_get_msr_feature,
> > @@ -4819,11 +4817,6 @@ static __init int svm_hardware_setup(void)
> >   
> >   	enable_apicv = avic = avic && avic_hardware_setup(&svm_x86_ops);
> >   
> > -	if (!enable_apicv) {
> > -		svm_x86_ops.vcpu_blocking = NULL;
> > -		svm_x86_ops.vcpu_unblocking = NULL;
> > -	}
> Isn't this code already zeros these callbacks when avic is not enabled?

Ah, right. I'll remove the setting to NULL.

> I am not sure why this patch is needed to be honest.

It's not related to x2AVIC. It was recommended by Sean earlier
in another patch series:

  https://lore.kernel.org/lkml/Yc3r1U6WFVDtJCZn@xxxxxxxxxx/

Since this series introduces the helper function avic_hardware_setup(),
and re-factor the AVIC setup code into the function. So, I am including
his recommendation in this series instead..

Regards,
Suravee