[PATCH] KVM: fix avic_set_running for preemptable kernels

[PATCH] KVM: fix avic_set_running for preemptable kernels

[Paolo Bonzini]

avic_set_running() passes the current CPU to avic_vcpu_load(), albeit
via vcpu->cpu rather than smp_processor_id().  If the thread is migrated
while avic_set_running runs, the call to avic_vcpu_load() can use a stale
value for the processor id.  Avoid this by blocking preemption over the
entire execution of avic_set_running().

Reported-by: Sean Christopherson <seanjc@google.com>
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
---
 arch/x86/kvm/svm/avic.c | 16 +++++++++-------
 1 file changed, 9 insertions(+), 7 deletions(-)

diff --git a/arch/x86/kvm/svm/avic.c b/arch/x86/kvm/svm/avic.c
index 0a58283005f3..560807a2edd4 100644
--- a/arch/x86/kvm/svm/avic.c
+++ b/arch/x86/kvm/svm/avic.c
@@ -1000,16 +1000,18 @@ void avic_vcpu_put(struct kvm_vcpu *vcpu)
 static void avic_set_running(struct kvm_vcpu *vcpu, bool is_run)
 {
 	struct vcpu_svm *svm = to_svm(vcpu);
+	int cpu = get_cpu();
 
+	WARN_ON(cpu != vcpu->cpu);
 	svm->avic_is_running = is_run;
 
-	if (!kvm_vcpu_apicv_active(vcpu))
-		return;
-
-	if (is_run)
-		avic_vcpu_load(vcpu, vcpu->cpu);
-	else
-		avic_vcpu_put(vcpu);
+	if (kvm_vcpu_apicv_active(vcpu)) {
+		if (is_run)
+			avic_vcpu_load(vcpu, cpu);
+		else
+			avic_vcpu_put(vcpu);
+	}
+	put_cpu();
 }
 
 void svm_vcpu_blocking(struct kvm_vcpu *vcpu)
-- 
2.31.1

Re: [PATCH] KVM: fix avic_set_running for preemptable kernels

[Maxim Levitsky]

On Tue, 2021-11-30 at 03:46 -0500, Paolo Bonzini wrote:
> avic_set_running() passes the current CPU to avic_vcpu_load(), albeit
> via vcpu->cpu rather than smp_processor_id().  If the thread is migrated
> while avic_set_running runs, the call to avic_vcpu_load() can use a stale
> value for the processor id.  Avoid this by blocking preemption over the
> entire execution of avic_set_running().
> 
> Reported-by: Sean Christopherson <seanjc@google.com>
> Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>
> ---
>  arch/x86/kvm/svm/avic.c | 16 +++++++++-------
>  1 file changed, 9 insertions(+), 7 deletions(-)
> 
> diff --git a/arch/x86/kvm/svm/avic.c b/arch/x86/kvm/svm/avic.c
> index 0a58283005f3..560807a2edd4 100644
> --- a/arch/x86/kvm/svm/avic.c
> +++ b/arch/x86/kvm/svm/avic.c
> @@ -1000,16 +1000,18 @@ void avic_vcpu_put(struct kvm_vcpu *vcpu)
>  static void avic_set_running(struct kvm_vcpu *vcpu, bool is_run)
>  {
>  	struct vcpu_svm *svm = to_svm(vcpu);
> +	int cpu = get_cpu();
>  
> +	WARN_ON(cpu != vcpu->cpu);
>  	svm->avic_is_running = is_run;
>  
> -	if (!kvm_vcpu_apicv_active(vcpu))
> -		return;
> -
> -	if (is_run)
> -		avic_vcpu_load(vcpu, vcpu->cpu);
> -	else
> -		avic_vcpu_put(vcpu);
> +	if (kvm_vcpu_apicv_active(vcpu)) {
> +		if (is_run)
> +			avic_vcpu_load(vcpu, cpu);
> +		else
> +			avic_vcpu_put(vcpu);
> +	}
> +	put_cpu();
>  }
>  
>  void svm_vcpu_blocking(struct kvm_vcpu *vcpu)

To be honest, I was thinking to remove vcpu parameter from
avic_vcpu_load and let it figure it out, but this can be
always done later.

Reviewed-by: Maxim Levitsky <mlevitsk@redhat.com>

PS: this patch as expected didn't help with the 'is_running' AVIC bug
I am stuck with.

Best regards,
	Maxim Levitsky