diff --git a/include/kvm/arm_vgic.h b/include/kvm/arm_vgic.h
index 24f0394..e7efe12 100644
--- a/include/kvm/arm_vgic.h
+++ b/include/kvm/arm_vgic.h
@@ -131,6 +131,7 @@ struct vgic_irq {
 		u32 mpidr;			/* GICv3 target VCPU */
 	};
 	u8 source;			/* GICv2 SGIs only */
+	u8 active_source;		/* GICv2 SGIs only */
 	u8 priority;
 	enum vgic_irq_config config;	/* Level or edge */
 
diff --git a/virt/kvm/arm/vgic/vgic-mmio.c b/virt/kvm/arm/vgic/vgic-mmio.c
index dbe99d6..ff9655c 100644
--- a/virt/kvm/arm/vgic/vgic-mmio.c
+++ b/virt/kvm/arm/vgic/vgic-mmio.c
@@ -289,10 +289,16 @@ static void vgic_mmio_change_active(struct kvm_vcpu *vcpu, struct vgic_irq *irq,
 	       irq->vcpu->cpu != -1) /* VCPU thread is running */
 		cond_resched_lock(&irq->irq_lock);
 
-	if (irq->hw)
+	if (irq->hw) {
 		vgic_hw_irq_change_active(vcpu, irq, active, !requester_vcpu);
-	else
+	} else {
+		u32 model = vcpu->kvm->arch.vgic.vgic_model;
+
 		irq->active = active;
+		if (model == KVM_DEV_TYPE_ARM_VGIC_V2 &&
+		    active && vgic_irq_is_sgi(irq->intid))
+			irq->active_source = requester_vcpu->vcpu_id;
+	}
 
 	if (irq->active)
 		vgic_queue_irq_unlock(vcpu->kvm, irq, flags);
diff --git a/virt/kvm/arm/vgic/vgic-v2.c b/virt/kvm/arm/vgic/vgic-v2.c
index 45aa433..a5f2e44 100644
--- a/virt/kvm/arm/vgic/vgic-v2.c
+++ b/virt/kvm/arm/vgic/vgic-v2.c
@@ -37,13 +37,6 @@ void vgic_v2_init_lrs(void)
 		vgic_v2_write_lr(i, 0);
 }
 
-void vgic_v2_set_npie(struct kvm_vcpu *vcpu)
-{
-	struct vgic_v2_cpu_if *cpuif = &vcpu->arch.vgic_cpu.vgic_v2;
-
-	cpuif->vgic_hcr |= GICH_HCR_NPIE;
-}
-
 void vgic_v2_set_underflow(struct kvm_vcpu *vcpu)
 {
 	struct vgic_v2_cpu_if *cpuif = &vcpu->arch.vgic_cpu.vgic_v2;
@@ -71,13 +64,18 @@ void vgic_v2_fold_lr_state(struct kvm_vcpu *vcpu)
 	int lr;
 	unsigned long flags;
 
-	cpuif->vgic_hcr &= ~(GICH_HCR_UIE | GICH_HCR_NPIE);
+	cpuif->vgic_hcr &= ~GICH_HCR_UIE;
 
 	for (lr = 0; lr < vgic_cpu->used_lrs; lr++) {
 		u32 val = cpuif->vgic_lr[lr];
-		u32 intid = val & GICH_LR_VIRTUALID;
+		u32 cpuid, intid = val & GICH_LR_VIRTUALID;
 		struct vgic_irq *irq;
 
+		/* Extract the source vCPU id from the LR */
+		cpuid = val & GICH_LR_PHYSID_CPUID;
+		cpuid >>= GICH_LR_PHYSID_CPUID_SHIFT;
+		cpuid &= 7;
+
 		/* Notify fds when the guest EOI'ed a level-triggered SPI */
 		if (lr_signals_eoi_mi(val) && vgic_valid_spi(vcpu->kvm, intid))
 			kvm_notify_acked_irq(vcpu->kvm, 0,
@@ -90,17 +88,16 @@ void vgic_v2_fold_lr_state(struct kvm_vcpu *vcpu)
 		/* Always preserve the active bit */
 		irq->active = !!(val & GICH_LR_ACTIVE_BIT);
 
+		if (irq->active && vgic_irq_is_sgi(intid))
+			irq->active_source = cpuid;
+
 		/* Edge is the only case where we preserve the pending bit */
 		if (irq->config == VGIC_CONFIG_EDGE &&
 		    (val & GICH_LR_PENDING_BIT)) {
 			irq->pending_latch = true;
 
-			if (vgic_irq_is_sgi(intid)) {
-				u32 cpuid = val & GICH_LR_PHYSID_CPUID;
-
-				cpuid >>= GICH_LR_PHYSID_CPUID_SHIFT;
+			if (vgic_irq_is_sgi(intid))
 				irq->source |= (1 << cpuid);
-			}
 		}
 
 		/*
@@ -152,8 +149,15 @@ void vgic_v2_populate_lr(struct kvm_vcpu *vcpu, struct vgic_irq *irq, int lr)
 	u32 val = irq->intid;
 	bool allow_pending = true;
 
-	if (irq->active)
+	if (irq->active) {
 		val |= GICH_LR_ACTIVE_BIT;
+		if (vgic_irq_is_sgi(irq->intid))
+			val |= irq->active_source << GICH_LR_PHYSID_CPUID_SHIFT;
+		if (vgic_irq_is_multi_sgi(irq)) {
+			allow_pending = false;
+			val |= GICH_LR_EOI;
+		}
+	}
 
 	if (irq->hw) {
 		val |= GICH_LR_HW;
@@ -190,8 +194,10 @@ void vgic_v2_populate_lr(struct kvm_vcpu *vcpu, struct vgic_irq *irq, int lr)
 			BUG_ON(!src);
 			val |= (src - 1) << GICH_LR_PHYSID_CPUID_SHIFT;
 			irq->source &= ~(1 << (src - 1));
-			if (irq->source)
+			if (irq->source) {
 				irq->pending_latch = true;
+				val |= GICH_LR_EOI;
+			}
 		}
 	}
 
diff --git a/virt/kvm/arm/vgic/vgic-v3.c b/virt/kvm/arm/vgic/vgic-v3.c
index 8195f52..c7423f3 100644
--- a/virt/kvm/arm/vgic/vgic-v3.c
+++ b/virt/kvm/arm/vgic/vgic-v3.c
@@ -27,13 +27,6 @@ static bool group1_trap;
 static bool common_trap;
 static bool gicv4_enable;
 
-void vgic_v3_set_npie(struct kvm_vcpu *vcpu)
-{
-	struct vgic_v3_cpu_if *cpuif = &vcpu->arch.vgic_cpu.vgic_v3;
-
-	cpuif->vgic_hcr |= ICH_HCR_NPIE;
-}
-
 void vgic_v3_set_underflow(struct kvm_vcpu *vcpu)
 {
 	struct vgic_v3_cpu_if *cpuif = &vcpu->arch.vgic_cpu.vgic_v3;
@@ -55,17 +48,23 @@ void vgic_v3_fold_lr_state(struct kvm_vcpu *vcpu)
 	int lr;
 	unsigned long flags;
 
-	cpuif->vgic_hcr &= ~(ICH_HCR_UIE | ICH_HCR_NPIE);
+	cpuif->vgic_hcr &= ~ICH_HCR_UIE;
 
 	for (lr = 0; lr < vgic_cpu->used_lrs; lr++) {
 		u64 val = cpuif->vgic_lr[lr];
-		u32 intid;
+		u32 intid, cpuid;
 		struct vgic_irq *irq;
+		bool is_v2_sgi = false;
 
-		if (model == KVM_DEV_TYPE_ARM_VGIC_V3)
+		cpuid = val & GICH_LR_PHYSID_CPUID;
+		cpuid >>= GICH_LR_PHYSID_CPUID_SHIFT;
+
+		if (model == KVM_DEV_TYPE_ARM_VGIC_V3) {
 			intid = val & ICH_LR_VIRTUAL_ID_MASK;
-		else
+		} else {
 			intid = val & GICH_LR_VIRTUALID;
+			is_v2_sgi = vgic_irq_is_sgi(intid);
+		}
 
 		/* Notify fds when the guest EOI'ed a level-triggered IRQ */
 		if (lr_signals_eoi_mi(val) && vgic_valid_spi(vcpu->kvm, intid))
@@ -81,18 +80,16 @@ void vgic_v3_fold_lr_state(struct kvm_vcpu *vcpu)
 		/* Always preserve the active bit */
 		irq->active = !!(val & ICH_LR_ACTIVE_BIT);
 
+		if (irq->active && is_v2_sgi)
+			irq->active_source = cpuid;
+
 		/* Edge is the only case where we preserve the pending bit */
 		if (irq->config == VGIC_CONFIG_EDGE &&
 		    (val & ICH_LR_PENDING_BIT)) {
 			irq->pending_latch = true;
 
-			if (vgic_irq_is_sgi(intid) &&
-			    model == KVM_DEV_TYPE_ARM_VGIC_V2) {
-				u32 cpuid = val & GICH_LR_PHYSID_CPUID;
-
-				cpuid >>= GICH_LR_PHYSID_CPUID_SHIFT;
+			if (is_v2_sgi)
 				irq->source |= (1 << cpuid);
-			}
 		}
 
 		/*
@@ -133,10 +130,20 @@ void vgic_v3_populate_lr(struct kvm_vcpu *vcpu, struct vgic_irq *irq, int lr)
 {
 	u32 model = vcpu->kvm->arch.vgic.vgic_model;
 	u64 val = irq->intid;
-	bool allow_pending = true;
+	bool allow_pending = true, is_v2_sgi;
 
-	if (irq->active)
+	is_v2_sgi = (vgic_irq_is_sgi(irq->intid) &&
+		     model == KVM_DEV_TYPE_ARM_VGIC_V2);
+
+	if (irq->active) {
 		val |= ICH_LR_ACTIVE_BIT;
+		if (is_v2_sgi)
+			val |= irq->active_source << GICH_LR_PHYSID_CPUID_SHIFT;
+		if (vgic_irq_is_multi_sgi(irq)) {
+			allow_pending = false;
+			val |= ICH_LR_EOI;
+		}
+	}
 
 	if (irq->hw) {
 		val |= ICH_LR_HW;
@@ -174,8 +181,10 @@ void vgic_v3_populate_lr(struct kvm_vcpu *vcpu, struct vgic_irq *irq, int lr)
 			BUG_ON(!src);
 			val |= (src - 1) << GICH_LR_PHYSID_CPUID_SHIFT;
 			irq->source &= ~(1 << (src - 1));
-			if (irq->source)
+			if (irq->source) {
 				irq->pending_latch = true;
+				val |= ICH_LR_EOI;
+			}
 		}
 	}
 
diff --git a/virt/kvm/arm/vgic/vgic.c b/virt/kvm/arm/vgic/vgic.c
index 4b6d729..568c65f 100644
--- a/virt/kvm/arm/vgic/vgic.c
+++ b/virt/kvm/arm/vgic/vgic.c
@@ -719,14 +719,6 @@ static inline void vgic_set_underflow(struct kvm_vcpu *vcpu)
 		vgic_v3_set_underflow(vcpu);
 }
 
-static inline void vgic_set_npie(struct kvm_vcpu *vcpu)
-{
-	if (kvm_vgic_global_state.type == VGIC_V2)
-		vgic_v2_set_npie(vcpu);
-	else
-		vgic_v3_set_npie(vcpu);
-}
-
 /* Requires the ap_list_lock to be held. */
 static int compute_ap_list_depth(struct kvm_vcpu *vcpu,
 				 bool *multi_sgi)
@@ -740,17 +732,15 @@ static int compute_ap_list_depth(struct kvm_vcpu *vcpu,
 	DEBUG_SPINLOCK_BUG_ON(!spin_is_locked(&vgic_cpu->ap_list_lock));
 
 	list_for_each_entry(irq, &vgic_cpu->ap_list_head, ap_list) {
+		int w;
+
 		spin_lock(&irq->irq_lock);
 		/* GICv2 SGIs can count for more than one... */
-		if (vgic_irq_is_sgi(irq->intid) && irq->source) {
-			int w = hweight8(irq->source);
-
-			count += w;
-			*multi_sgi |= (w > 1);
-		} else {
-			count++;
-		}
+		w = vgic_irq_get_lr_count(irq);
 		spin_unlock(&irq->irq_lock);
+
+		count += w;
+		*multi_sgi |= (w > 1);
 	}
 	return count;
 }
@@ -761,7 +751,6 @@ static void vgic_flush_lr_state(struct kvm_vcpu *vcpu)
 	struct vgic_cpu *vgic_cpu = &vcpu->arch.vgic_cpu;
 	struct vgic_irq *irq;
 	int count;
-	bool npie = false;
 	bool multi_sgi;
 	u8 prio = 0xff;
 
@@ -791,10 +780,8 @@ static void vgic_flush_lr_state(struct kvm_vcpu *vcpu)
 		if (likely(vgic_target_oracle(irq) == vcpu)) {
 			vgic_populate_lr(vcpu, irq, count++);
 
-			if (irq->source) {
-				npie = true;
+			if (irq->source)
 				prio = irq->priority;
-			}
 		}
 
 		spin_unlock(&irq->irq_lock);
@@ -807,9 +794,6 @@ static void vgic_flush_lr_state(struct kvm_vcpu *vcpu)
 		}
 	}
 
-	if (npie)
-		vgic_set_npie(vcpu);
-
 	vcpu->arch.vgic_cpu.used_lrs = count;
 
 	/* Nuke remaining LRs */
diff --git a/virt/kvm/arm/vgic/vgic.h b/virt/kvm/arm/vgic/vgic.h
index 830e815..32c25d4 100644
--- a/virt/kvm/arm/vgic/vgic.h
+++ b/virt/kvm/arm/vgic/vgic.h
@@ -110,6 +110,20 @@ static inline bool vgic_irq_is_mapped_level(struct vgic_irq *irq)
 	return irq->config == VGIC_CONFIG_LEVEL && irq->hw;
 }
 
+static inline int vgic_irq_get_lr_count(struct vgic_irq *irq)
+{
+	/* Account for the active state as an interrupt */
+	if (vgic_irq_is_sgi(irq->intid) && irq->source)
+		return hweight8(irq->source) + irq->active;
+
+	return irq_is_pending(irq) || irq->active;
+}
+
+static inline bool vgic_irq_is_multi_sgi(struct vgic_irq *irq)
+{
+	return vgic_irq_get_lr_count(irq) > 1;
+}
+
 /*
  * This struct provides an intermediate representation of the fields contained
  * in the GICH_VMCR and ICH_VMCR registers, such that code exporting the GIC