Revert "svm: Fix AVIC incomplete IPI emulation"

This reverts commit bb218fb.

As Oren Twaig pointed out the old discussion:

  https://patchwork.kernel.org/patch/8292231/

that the change coud potentially cause an extra IPI to be sent to
the destination vcpu because the AVIC hardware already set the IRR bit
before the incomplete IPI #VMEXIT with id=1 (target vcpu is not running).
Since writting to ICR and ICR2 will also set the IRR. If something triggers
the destination vcpu to get scheduled before the emulation finishes, then
this could result in an additional IPI.

Also, the issue mentioned in the commit bb218fb was misdiagnosed.

Cc: Radim Krčmář <[email protected]>
Cc: Paolo Bonzini <[email protected]>
Reported-by: Oren Twaig <[email protected]>
Signed-off-by: Suravee Suthikulpanit <[email protected]>
Cc: [email protected]
Signed-off-by: Paolo Bonzini <[email protected]>

arch/x86/kvm/svm.c

@@ -4517,14 +4517,25 @@ static int avic_incomplete_ipi_interception(struct vcpu_svm *svm)
 		kvm_lapic_reg_write(apic, APIC_ICR, icrl);
 		break;
 	case AVIC_IPI_FAILURE_TARGET_NOT_RUNNING: {
+		int i;
+		struct kvm_vcpu *vcpu;
+		struct kvm *kvm = svm->vcpu.kvm;
 		struct kvm_lapic *apic = svm->vcpu.arch.apic;
 
 		/*
-		 * Update ICR high and low, then emulate sending IPI,
-		 * which is handled when writing APIC_ICR.
+		 * At this point, we expect that the AVIC HW has already
+		 * set the appropriate IRR bits on the valid target
+		 * vcpus. So, we just need to kick the appropriate vcpu.
 		 */
-		kvm_lapic_reg_write(apic, APIC_ICR2, icrh);
-		kvm_lapic_reg_write(apic, APIC_ICR, icrl);
+		kvm_for_each_vcpu(i, vcpu, kvm) {
+			bool m = kvm_apic_match_dest(vcpu, apic,
+						     icrl & KVM_APIC_SHORT_MASK,
+						     GET_APIC_DEST_FIELD(icrh),
+						     icrl & KVM_APIC_DEST_MASK);
+
+			if (m && !avic_vcpu_is_running(vcpu))
+				kvm_vcpu_wake_up(vcpu);
+		}
 		break;
 	}
 	case AVIC_IPI_FAILURE_INVALID_TARGET: