diff --git a/arch/arm/include/asm/kvm_asm.h b/arch/arm/include/asm/kvm_asm.h
index 8ef05381984b..dd16044b34b6 100644
--- a/arch/arm/include/asm/kvm_asm.h
+++ b/arch/arm/include/asm/kvm_asm.h
@@ -75,7 +75,10 @@ extern void __init_stage2_translation(void);
 extern void __kvm_hyp_reset(unsigned long);
 
 extern u64 __vgic_v3_get_ich_vtr_el2(void);
+extern u64 __vgic_v3_read_vmcr(void);
+extern void __vgic_v3_write_vmcr(u32 vmcr);
 extern void __vgic_v3_init_lrs(void);
+
 #endif
 
 #endif /* __ARM_KVM_ASM_H__ */
diff --git a/arch/arm/kvm/arm.c b/arch/arm/kvm/arm.c
index 96dba7cd8be7..46fd37578693 100644
--- a/arch/arm/kvm/arm.c
+++ b/arch/arm/kvm/arm.c
@@ -351,15 +351,14 @@ void kvm_arch_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
 	vcpu->arch.host_cpu_context = this_cpu_ptr(kvm_host_cpu_state);
 
 	kvm_arm_set_running_vcpu(vcpu);
+
+	kvm_vgic_load(vcpu);
 }
 
 void kvm_arch_vcpu_put(struct kvm_vcpu *vcpu)
 {
-	/*
-	 * The arch-generic KVM code expects the cpu field of a vcpu to be -1
-	 * if the vcpu is no longer assigned to a cpu.  This is used for the
-	 * optimized make_all_cpus_request path.
-	 */
+	kvm_vgic_put(vcpu);
+
 	vcpu->cpu = -1;
 
 	kvm_arm_set_running_vcpu(NULL);
@@ -633,7 +632,9 @@ int kvm_arch_vcpu_ioctl_run(struct kvm_vcpu *vcpu, struct kvm_run *run)
 		 * non-preemptible context.
 		 */
 		preempt_disable();
+
 		kvm_pmu_flush_hwstate(vcpu);
+
 		kvm_timer_flush_hwstate(vcpu);
 		kvm_vgic_flush_hwstate(vcpu);
 
diff --git a/arch/arm64/include/asm/kvm_asm.h b/arch/arm64/include/asm/kvm_asm.h
index ec3553eb9349..49f99cd02613 100644
--- a/arch/arm64/include/asm/kvm_asm.h
+++ b/arch/arm64/include/asm/kvm_asm.h
@@ -59,6 +59,8 @@ extern void __kvm_tlb_flush_local_vmid(struct kvm_vcpu *vcpu);
 extern int __kvm_vcpu_run(struct kvm_vcpu *vcpu);
 
 extern u64 __vgic_v3_get_ich_vtr_el2(void);
+extern u64 __vgic_v3_read_vmcr(void);
+extern void __vgic_v3_write_vmcr(u32 vmcr);
 extern void __vgic_v3_init_lrs(void);
 
 extern u32 __kvm_get_mdcr_el2(void);
diff --git a/include/kvm/arm_vgic.h b/include/kvm/arm_vgic.h
index b72dd2ad5f44..f7a2e31eb4c1 100644
--- a/include/kvm/arm_vgic.h
+++ b/include/kvm/arm_vgic.h
@@ -306,6 +306,9 @@ bool kvm_vgic_map_is_active(struct kvm_vcpu *vcpu, unsigned int virt_irq);
 
 int kvm_vgic_vcpu_pending_irq(struct kvm_vcpu *vcpu);
 
+void kvm_vgic_load(struct kvm_vcpu *vcpu);
+void kvm_vgic_put(struct kvm_vcpu *vcpu);
+
 #define irqchip_in_kernel(k)	(!!((k)->arch.vgic.in_kernel))
 #define vgic_initialized(k)	((k)->arch.vgic.initialized)
 #define vgic_ready(k)		((k)->arch.vgic.ready)
diff --git a/virt/kvm/arm/hyp/vgic-v2-sr.c b/virt/kvm/arm/hyp/vgic-v2-sr.c
index c8aeb7b91ec8..d3d3b9b0c2c3 100644
--- a/virt/kvm/arm/hyp/vgic-v2-sr.c
+++ b/virt/kvm/arm/hyp/vgic-v2-sr.c
@@ -114,8 +114,6 @@ void __hyp_text __vgic_v2_save_state(struct kvm_vcpu *vcpu)
 	if (!base)
 		return;
 
-	cpu_if->vgic_vmcr = readl_relaxed(base + GICH_VMCR);
-
 	if (vcpu->arch.vgic_cpu.live_lrs) {
 		cpu_if->vgic_apr = readl_relaxed(base + GICH_APR);
 
@@ -165,7 +163,6 @@ void __hyp_text __vgic_v2_restore_state(struct kvm_vcpu *vcpu)
 		}
 	}
 
-	writel_relaxed(cpu_if->vgic_vmcr, base + GICH_VMCR);
 	vcpu->arch.vgic_cpu.live_lrs = live_lrs;
 }
 
diff --git a/virt/kvm/arm/hyp/vgic-v3-sr.c b/virt/kvm/arm/hyp/vgic-v3-sr.c
index 3947095cc0a1..e51ee7edf953 100644
--- a/virt/kvm/arm/hyp/vgic-v3-sr.c
+++ b/virt/kvm/arm/hyp/vgic-v3-sr.c
@@ -159,8 +159,6 @@ void __hyp_text __vgic_v3_save_state(struct kvm_vcpu *vcpu)
 	if (!cpu_if->vgic_sre)
 		dsb(st);
 
-	cpu_if->vgic_vmcr  = read_gicreg(ICH_VMCR_EL2);
-
 	if (vcpu->arch.vgic_cpu.live_lrs) {
 		int i;
 		u32 max_lr_idx, nr_pri_bits;
@@ -261,8 +259,6 @@ void __hyp_text __vgic_v3_restore_state(struct kvm_vcpu *vcpu)
 			live_lrs |= (1 << i);
 	}
 
-	write_gicreg(cpu_if->vgic_vmcr, ICH_VMCR_EL2);
-
 	if (live_lrs) {
 		write_gicreg(cpu_if->vgic_hcr, ICH_HCR_EL2);
 
@@ -326,3 +322,13 @@ u64 __hyp_text __vgic_v3_get_ich_vtr_el2(void)
 {
 	return read_gicreg(ICH_VTR_EL2);
 }
+
+u64 __hyp_text __vgic_v3_read_vmcr(void)
+{
+	return read_gicreg(ICH_VMCR_EL2);
+}
+
+void __hyp_text __vgic_v3_write_vmcr(u32 vmcr)
+{
+	write_gicreg(vmcr, ICH_VMCR_EL2);
+}
diff --git a/virt/kvm/arm/vgic/vgic-init.c b/virt/kvm/arm/vgic/vgic-init.c
index 276139a24e6f..e8e973b72ca5 100644
--- a/virt/kvm/arm/vgic/vgic-init.c
+++ b/virt/kvm/arm/vgic/vgic-init.c
@@ -262,6 +262,18 @@ int vgic_init(struct kvm *kvm)
 	vgic_debug_init(kvm);
 
 	dist->initialized = true;
+
+	/*
+	 * If we're initializing GICv2 on-demand when first running the VCPU
+	 * then we need to load the VGIC state onto the CPU.  We can detect
+	 * this easily by checking if we are in between vcpu_load and vcpu_put
+	 * when we just initialized the VGIC.
+	 */
+	preempt_disable();
+	vcpu = kvm_arm_get_running_vcpu();
+	if (vcpu)
+		kvm_vgic_load(vcpu);
+	preempt_enable();
 out:
 	return ret;
 }
diff --git a/virt/kvm/arm/vgic/vgic-v2.c b/virt/kvm/arm/vgic/vgic-v2.c
index b834ecdf3225..2f241e026c8f 100644
--- a/virt/kvm/arm/vgic/vgic-v2.c
+++ b/virt/kvm/arm/vgic/vgic-v2.c
@@ -184,6 +184,7 @@ void vgic_v2_clear_lr(struct kvm_vcpu *vcpu, int lr)
 
 void vgic_v2_set_vmcr(struct kvm_vcpu *vcpu, struct vgic_vmcr *vmcrp)
 {
+	struct vgic_v2_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v2;
 	u32 vmcr;
 
 	vmcr  = (vmcrp->ctlr << GICH_VMCR_CTRL_SHIFT) & GICH_VMCR_CTRL_MASK;
@@ -194,12 +195,15 @@ void vgic_v2_set_vmcr(struct kvm_vcpu *vcpu, struct vgic_vmcr *vmcrp)
 	vmcr |= (vmcrp->pmr << GICH_VMCR_PRIMASK_SHIFT) &
 		GICH_VMCR_PRIMASK_MASK;
 
-	vcpu->arch.vgic_cpu.vgic_v2.vgic_vmcr = vmcr;
+	cpu_if->vgic_vmcr = vmcr;
 }
 
 void vgic_v2_get_vmcr(struct kvm_vcpu *vcpu, struct vgic_vmcr *vmcrp)
 {
-	u32 vmcr = vcpu->arch.vgic_cpu.vgic_v2.vgic_vmcr;
+	struct vgic_v2_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v2;
+	u32 vmcr;
+
+	vmcr = cpu_if->vgic_vmcr;
 
 	vmcrp->ctlr = (vmcr & GICH_VMCR_CTRL_MASK) >>
 			GICH_VMCR_CTRL_SHIFT;
@@ -375,3 +379,19 @@ int vgic_v2_probe(const struct gic_kvm_info *info)
 
 	return ret;
 }
+
+void vgic_v2_load(struct kvm_vcpu *vcpu)
+{
+	struct vgic_v2_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v2;
+	struct vgic_dist *vgic = &vcpu->kvm->arch.vgic;
+
+	writel_relaxed(cpu_if->vgic_vmcr, vgic->vctrl_base + GICH_VMCR);
+}
+
+void vgic_v2_put(struct kvm_vcpu *vcpu)
+{
+	struct vgic_v2_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v2;
+	struct vgic_dist *vgic = &vcpu->kvm->arch.vgic;
+
+	cpu_if->vgic_vmcr = readl_relaxed(vgic->vctrl_base + GICH_VMCR);
+}
diff --git a/virt/kvm/arm/vgic/vgic-v3.c b/virt/kvm/arm/vgic/vgic-v3.c
index be0f4c3e0142..99213d744e4f 100644
--- a/virt/kvm/arm/vgic/vgic-v3.c
+++ b/virt/kvm/arm/vgic/vgic-v3.c
@@ -173,6 +173,7 @@ void vgic_v3_clear_lr(struct kvm_vcpu *vcpu, int lr)
 
 void vgic_v3_set_vmcr(struct kvm_vcpu *vcpu, struct vgic_vmcr *vmcrp)
 {
+	struct vgic_v3_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v3;
 	u32 vmcr;
 
 	/*
@@ -188,12 +189,15 @@ void vgic_v3_set_vmcr(struct kvm_vcpu *vcpu, struct vgic_vmcr *vmcrp)
 	vmcr |= (vmcrp->grpen0 << ICH_VMCR_ENG0_SHIFT) & ICH_VMCR_ENG0_MASK;
 	vmcr |= (vmcrp->grpen1 << ICH_VMCR_ENG1_SHIFT) & ICH_VMCR_ENG1_MASK;
 
-	vcpu->arch.vgic_cpu.vgic_v3.vgic_vmcr = vmcr;
+	cpu_if->vgic_vmcr = vmcr;
 }
 
 void vgic_v3_get_vmcr(struct kvm_vcpu *vcpu, struct vgic_vmcr *vmcrp)
 {
-	u32 vmcr = vcpu->arch.vgic_cpu.vgic_v3.vgic_vmcr;
+	struct vgic_v3_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v3;
+	u32 vmcr;
+
+	vmcr = cpu_if->vgic_vmcr;
 
 	/*
 	 * Ignore the FIQen bit, because GIC emulation always implies
@@ -386,3 +390,17 @@ int vgic_v3_probe(const struct gic_kvm_info *info)
 
 	return 0;
 }
+
+void vgic_v3_load(struct kvm_vcpu *vcpu)
+{
+	struct vgic_v3_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v3;
+
+	kvm_call_hyp(__vgic_v3_write_vmcr, cpu_if->vgic_vmcr);
+}
+
+void vgic_v3_put(struct kvm_vcpu *vcpu)
+{
+	struct vgic_v3_cpu_if *cpu_if = &vcpu->arch.vgic_cpu.vgic_v3;
+
+	cpu_if->vgic_vmcr = kvm_call_hyp(__vgic_v3_read_vmcr);
+}
diff --git a/virt/kvm/arm/vgic/vgic.c b/virt/kvm/arm/vgic/vgic.c
index 654dfd40e449..2ac0def57424 100644
--- a/virt/kvm/arm/vgic/vgic.c
+++ b/virt/kvm/arm/vgic/vgic.c
@@ -656,6 +656,28 @@ void kvm_vgic_flush_hwstate(struct kvm_vcpu *vcpu)
 	spin_unlock(&vcpu->arch.vgic_cpu.ap_list_lock);
 }
 
+void kvm_vgic_load(struct kvm_vcpu *vcpu)
+{
+	if (unlikely(!vgic_initialized(vcpu->kvm)))
+		return;
+
+	if (kvm_vgic_global_state.type == VGIC_V2)
+		vgic_v2_load(vcpu);
+	else
+		vgic_v3_load(vcpu);
+}
+
+void kvm_vgic_put(struct kvm_vcpu *vcpu)
+{
+	if (unlikely(!vgic_initialized(vcpu->kvm)))
+		return;
+
+	if (kvm_vgic_global_state.type == VGIC_V2)
+		vgic_v2_put(vcpu);
+	else
+		vgic_v3_put(vcpu);
+}
+
 int kvm_vgic_vcpu_pending_irq(struct kvm_vcpu *vcpu)
 {
 	struct vgic_cpu *vgic_cpu = &vcpu->arch.vgic_cpu;
diff --git a/virt/kvm/arm/vgic/vgic.h b/virt/kvm/arm/vgic/vgic.h
index db28f7cadab2..9afb4557c7e8 100644
--- a/virt/kvm/arm/vgic/vgic.h
+++ b/virt/kvm/arm/vgic/vgic.h
@@ -130,6 +130,9 @@ int vgic_v2_map_resources(struct kvm *kvm);
 int vgic_register_dist_iodev(struct kvm *kvm, gpa_t dist_base_address,
 			     enum vgic_type);
 
+void vgic_v2_load(struct kvm_vcpu *vcpu);
+void vgic_v2_put(struct kvm_vcpu *vcpu);
+
 static inline void vgic_get_irq_kref(struct vgic_irq *irq)
 {
 	if (irq->intid < VGIC_MIN_LPI)
@@ -150,6 +153,9 @@ int vgic_v3_probe(const struct gic_kvm_info *info);
 int vgic_v3_map_resources(struct kvm *kvm);
 int vgic_register_redist_iodevs(struct kvm *kvm, gpa_t dist_base_address);
 
+void vgic_v3_load(struct kvm_vcpu *vcpu);
+void vgic_v3_put(struct kvm_vcpu *vcpu);
+
 int vgic_register_its_iodevs(struct kvm *kvm);
 bool vgic_has_its(struct kvm *kvm);
 int kvm_vgic_register_its_device(void);