KVM: x86: Replace late check_nested_events() hack with more precise fix

Add an argument to interrupt_allowed and nmi_allowed, to checking if
interrupt injection is blocked.  Use the hook to handle the case where
an interrupt arrives between check_nested_events() and the injection
logic.  Drop the retry of check_nested_events() that hack-a-fixed the
same condition.

Blocking injection is also a bit of a hack, e.g. KVM should do exiting
and non-exiting interrupt processing in a single pass, but it's a more
precise hack.  The old comment is also misleading, e.g. KVM_REQ_EVENT is
purely an optimization, setting it on every run loop (which KVM doesn't
do) should not affect functionality, only performance.
Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
Message-Id: <20200423022550.15113-13-sean.j.christopherson@intel.com>
[Extend to SVM, add SMI and NMI.  Even though NMI and SMI cannot come
 asynchronously right now, making the fix generic is easy and removes a
 special case. - Paolo]
Signed-off-by: Paolo Bonzini <pbonzini@redhat.com>

arch/x86/include/asm/kvm_host.h

@@ -1138,8 +1138,8 @@ struct kvm_x86_ops {
 	void (*set_nmi)(struct kvm_vcpu *vcpu);
 	void (*queue_exception)(struct kvm_vcpu *vcpu);
 	void (*cancel_injection)(struct kvm_vcpu *vcpu);
-	bool (*interrupt_allowed)(struct kvm_vcpu *vcpu);
-	bool (*nmi_allowed)(struct kvm_vcpu *vcpu);
+	bool (*interrupt_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
+	bool (*nmi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
 	bool (*get_nmi_mask)(struct kvm_vcpu *vcpu);
 	void (*set_nmi_mask)(struct kvm_vcpu *vcpu, bool masked);
 	void (*enable_nmi_window)(struct kvm_vcpu *vcpu);
@@ -1237,7 +1237,7 @@ struct kvm_x86_ops {
 
 	void (*setup_mce)(struct kvm_vcpu *vcpu);
 
-	bool (*smi_allowed)(struct kvm_vcpu *vcpu);
+	bool (*smi_allowed)(struct kvm_vcpu *vcpu, bool for_injection);
 	int (*pre_enter_smm)(struct kvm_vcpu *vcpu, char *smstate);
 	int (*pre_leave_smm)(struct kvm_vcpu *vcpu, const char *smstate);
 	int (*enable_smi_window)(struct kvm_vcpu *vcpu);

arch/x86/kvm/svm/svm.c

@@ -3083,13 +3083,17 @@ bool svm_nmi_blocked(struct kvm_vcpu *vcpu)
 	return ret;
 }
 
-static bool svm_nmi_allowed(struct kvm_vcpu *vcpu)
+static bool svm_nmi_allowed(struct kvm_vcpu *vcpu, bool for_injection)
 {
 	struct vcpu_svm *svm = to_svm(vcpu);
 	if (svm->nested.nested_run_pending)
 		return false;
 
-        return !svm_nmi_blocked(vcpu);
+	/* An NMI must not be injected into L2 if it's supposed to VM-Exit.  */
+	if (for_injection && is_guest_mode(vcpu) && nested_exit_on_nmi(svm))
+		return false;
+
+	return !svm_nmi_blocked(vcpu);
 }
 
 static bool svm_get_nmi_mask(struct kvm_vcpu *vcpu)
@@ -3138,13 +3142,20 @@ bool svm_interrupt_blocked(struct kvm_vcpu *vcpu)
 	return (vmcb->control.int_state & SVM_INTERRUPT_SHADOW_MASK);
 }
 
-static bool svm_interrupt_allowed(struct kvm_vcpu *vcpu)
+static bool svm_interrupt_allowed(struct kvm_vcpu *vcpu, bool for_injection)
 {
 	struct vcpu_svm *svm = to_svm(vcpu);
 	if (svm->nested.nested_run_pending)
 		return false;
 
-        return !svm_interrupt_blocked(vcpu);
+	/*
+	 * An IRQ must not be injected into L2 if it's supposed to VM-Exit,
+	 * e.g. if the IRQ arrived asynchronously after checking nested events.
+	 */
+	if (for_injection && is_guest_mode(vcpu) && nested_exit_on_intr(svm))
+		return false;
+
+	return !svm_interrupt_blocked(vcpu);
 }
 
 static void enable_irq_window(struct kvm_vcpu *vcpu)
@@ -3812,12 +3823,16 @@ bool svm_smi_blocked(struct kvm_vcpu *vcpu)
 	return is_smm(vcpu);
 }
 
-static bool svm_smi_allowed(struct kvm_vcpu *vcpu)
+static bool svm_smi_allowed(struct kvm_vcpu *vcpu, bool for_injection)
 {
 	struct vcpu_svm *svm = to_svm(vcpu);
 	if (svm->nested.nested_run_pending)
 		return false;
 
+	/* An SMI must not be injected into L2 if it's supposed to VM-Exit.  */
+	if (for_injection && is_guest_mode(vcpu) && nested_exit_on_smi(svm))
+		return false;
+
 	return !svm_smi_blocked(vcpu);
 }
 

arch/x86/kvm/vmx/vmx.c

@@ -4523,11 +4523,15 @@ bool vmx_nmi_blocked(struct kvm_vcpu *vcpu)
 		 GUEST_INTR_STATE_NMI));
 }
 
-static bool vmx_nmi_allowed(struct kvm_vcpu *vcpu)
+static bool vmx_nmi_allowed(struct kvm_vcpu *vcpu, bool for_injection)
 {
 	if (to_vmx(vcpu)->nested.nested_run_pending)
 		return false;
 
+	/* An NMI must not be injected into L2 if it's supposed to VM-Exit.  */
+	if (for_injection && is_guest_mode(vcpu) && nested_exit_on_nmi(vcpu))
+		return false;
+
 	return !vmx_nmi_blocked(vcpu);
 }
 
@@ -4541,11 +4545,18 @@ bool vmx_interrupt_blocked(struct kvm_vcpu *vcpu)
 		(GUEST_INTR_STATE_STI | GUEST_INTR_STATE_MOV_SS));
 }
 
-static bool vmx_interrupt_allowed(struct kvm_vcpu *vcpu)
+static bool vmx_interrupt_allowed(struct kvm_vcpu *vcpu, bool for_injection)
 {
 	if (to_vmx(vcpu)->nested.nested_run_pending)
 		return false;
 
+       /*
+        * An IRQ must not be injected into L2 if it's supposed to VM-Exit,
+        * e.g. if the IRQ arrived asynchronously after checking nested events.
+        */
+	if (for_injection && is_guest_mode(vcpu) && nested_exit_on_intr(vcpu))
+		return false;
+
 	return !vmx_interrupt_blocked(vcpu);
 }
 
@@ -7688,7 +7699,7 @@ static void vmx_setup_mce(struct kvm_vcpu *vcpu)
 			~FEAT_CTL_LMCE_ENABLED;
 }
 
-static bool vmx_smi_allowed(struct kvm_vcpu *vcpu)
+static bool vmx_smi_allowed(struct kvm_vcpu *vcpu, bool for_injection)
 {
 	/* we need a nested vmexit to enter SMM, postpone if run is pending */
 	if (to_vmx(vcpu)->nested.nested_run_pending)

arch/x86/kvm/x86.c

@@ -7766,32 +7766,20 @@ static int inject_pending_event(struct kvm_vcpu *vcpu)
 	if (kvm_event_needs_reinjection(vcpu))
 		return 0;
 
-	if (vcpu->arch.smi_pending && kvm_x86_ops.smi_allowed(vcpu)) {
+	if (vcpu->arch.smi_pending &&
+	    kvm_x86_ops.smi_allowed(vcpu, true)) {
 		vcpu->arch.smi_pending = false;
 		++vcpu->arch.smi_count;
 		enter_smm(vcpu);
-	} else if (vcpu->arch.nmi_pending && kvm_x86_ops.nmi_allowed(vcpu)) {
+	} else if (vcpu->arch.nmi_pending &&
+		   kvm_x86_ops.nmi_allowed(vcpu, true)) {
 		--vcpu->arch.nmi_pending;
 		vcpu->arch.nmi_injected = true;
 		kvm_x86_ops.set_nmi(vcpu);
-	} else if (kvm_cpu_has_injectable_intr(vcpu)) {
-		/*
-		 * Because interrupts can be injected asynchronously, we are
-		 * calling check_nested_events again here to avoid a race condition.
-		 * See https://lkml.org/lkml/2014/7/2/60 for discussion about this
-		 * proposal and current concerns.  Perhaps we should be setting
-		 * KVM_REQ_EVENT only on certain events and not unconditionally?
-		 */
-		if (is_guest_mode(vcpu)) {
-			r = kvm_x86_ops.nested_ops->check_events(vcpu);
-			if (r != 0)
-				return r;
-		}
-		if (kvm_x86_ops.interrupt_allowed(vcpu)) {
-			kvm_queue_interrupt(vcpu, kvm_cpu_get_interrupt(vcpu),
-					    false);
-			kvm_x86_ops.set_irq(vcpu);
-		}
+	} else if (kvm_cpu_has_injectable_intr(vcpu) &&
+		   kvm_x86_ops.interrupt_allowed(vcpu, true)) {
+		kvm_queue_interrupt(vcpu, kvm_cpu_get_interrupt(vcpu), false);
+		kvm_x86_ops.set_irq(vcpu);
 	}
 
 	return 0;
@@ -10203,12 +10191,12 @@ static inline bool kvm_vcpu_has_events(struct kvm_vcpu *vcpu)
 
 	if (kvm_test_request(KVM_REQ_NMI, vcpu) ||
 	    (vcpu->arch.nmi_pending &&
-	     kvm_x86_ops.nmi_allowed(vcpu)))
+	     kvm_x86_ops.nmi_allowed(vcpu, false)))
 		return true;
 
 	if (kvm_test_request(KVM_REQ_SMI, vcpu) ||
 	    (vcpu->arch.smi_pending &&
-	     kvm_x86_ops.smi_allowed(vcpu)))
+	     kvm_x86_ops.smi_allowed(vcpu, false)))
 		return true;
 
 	if (kvm_arch_interrupt_allowed(vcpu) &&
@@ -10260,7 +10248,7 @@ int kvm_arch_vcpu_should_kick(struct kvm_vcpu *vcpu)
 
 int kvm_arch_interrupt_allowed(struct kvm_vcpu *vcpu)
 {
-	return kvm_x86_ops.interrupt_allowed(vcpu);
+	return kvm_x86_ops.interrupt_allowed(vcpu, false);
 }
 
 unsigned long kvm_get_linear_rip(struct kvm_vcpu *vcpu)
@@ -10425,7 +10413,7 @@ bool kvm_can_do_async_pf(struct kvm_vcpu *vcpu)
 	 * If interrupts are off we cannot even use an artificial
 	 * halt state.
 	 */
-	return kvm_x86_ops.interrupt_allowed(vcpu);
+	return kvm_arch_interrupt_allowed(vcpu);
 }
 
 void kvm_arch_async_page_not_present(struct kvm_vcpu *vcpu,