diff --git a/arch/x86/include/asm/kvm_host.h b/arch/x86/include/asm/kvm_host.h
index 262b0bc..dcdc915 100644
--- a/arch/x86/include/asm/kvm_host.h
+++ b/arch/x86/include/asm/kvm_host.h
@@ -985,6 +985,14 @@ struct kvm_x86_ops {
 
 	void (*tlb_flush)(struct kvm_vcpu *vcpu, bool invalidate_gpa);
 
+	/*
+	 * Flush any TLB entries associated with the given GVA.
+	 * Does not need to flush GPA->HPA mappings.
+	 * Can potentially get non-canonical addresses through INVLPGs, which
+	 * the implementation may choose to ignore if appropriate.
+	 */
+	void (*tlb_flush_gva)(struct kvm_vcpu *vcpu, gva_t addr);
+
 	void (*run)(struct kvm_vcpu *vcpu);
 	int (*handle_exit)(struct kvm_vcpu *vcpu);
 	void (*skip_emulated_instruction)(struct kvm_vcpu *vcpu);
diff --git a/arch/x86/kvm/mmu.c b/arch/x86/kvm/mmu.c
index 9446a36..f84c194 100644
--- a/arch/x86/kvm/mmu.c
+++ b/arch/x86/kvm/mmu.c
@@ -5226,6 +5226,10 @@ void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
 {
 	struct kvm_mmu *mmu = &vcpu->arch.mmu;
 
+	/* INVLPG on a * non-canonical address is a NOP according to the SDM.  */
+	if (is_noncanonical_address(gva, vcpu))
+		return;
+
 	mmu->invlpg(vcpu, gva, mmu->root_hpa);
 
 	/*
@@ -5242,7 +5246,7 @@ void kvm_mmu_invlpg(struct kvm_vcpu *vcpu, gva_t gva)
 	if (VALID_PAGE(mmu->prev_root.hpa))
 		mmu->invlpg(vcpu, gva, mmu->prev_root.hpa);
 
-	kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
+	kvm_x86_ops->tlb_flush_gva(vcpu, gva);
 	++vcpu->stat.invlpg;
 }
 EXPORT_SYMBOL_GPL(kvm_mmu_invlpg);
@@ -5250,18 +5254,22 @@ EXPORT_SYMBOL_GPL(kvm_mmu_invlpg);
 void kvm_mmu_invpcid_gva(struct kvm_vcpu *vcpu, gva_t gva, unsigned long pcid)
 {
 	struct kvm_mmu *mmu = &vcpu->arch.mmu;
+	bool tlb_flush = false;
 
 	if (pcid == kvm_get_active_pcid(vcpu)) {
 		mmu->invlpg(vcpu, gva, mmu->root_hpa);
-		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
+		tlb_flush = true;
 	}
 
 	if (VALID_PAGE(mmu->prev_root.hpa) &&
 	    pcid == kvm_get_pcid(vcpu, mmu->prev_root.cr3)) {
 		mmu->invlpg(vcpu, gva, mmu->prev_root.hpa);
-		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
+		tlb_flush = true;
 	}
 
+	if (tlb_flush)
+		kvm_x86_ops->tlb_flush_gva(vcpu, gva);
+
 	++vcpu->stat.invlpg;
 
 	/*
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c
index be9931c..73e27a9 100644
--- a/arch/x86/kvm/svm.c
+++ b/arch/x86/kvm/svm.c
@@ -5434,6 +5434,13 @@ static void svm_flush_tlb(struct kvm_vcpu *vcpu, bool invalidate_gpa)
 		svm->asid_generation--;
 }
 
+static void svm_flush_tlb_gva(struct kvm_vcpu *vcpu, gva_t gva)
+{
+	struct vcpu_svm *svm = to_svm(vcpu);
+
+	invlpga(gva, svm->vmcb->control.asid);
+}
+
 static void svm_prepare_guest_switch(struct kvm_vcpu *vcpu)
 {
 }
@@ -7086,6 +7093,7 @@ static struct kvm_x86_ops svm_x86_ops __ro_after_init = {
 	.set_rflags = svm_set_rflags,
 
 	.tlb_flush = svm_flush_tlb,
+	.tlb_flush_gva = svm_flush_tlb_gva,
 
 	.run = svm_vcpu_run,
 	.handle_exit = handle_exit,
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index b812100..5aea5af 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -1992,6 +1992,19 @@ static void loaded_vmcs_clear(struct loaded_vmcs *loaded_vmcs)
 			 __loaded_vmcs_clear, loaded_vmcs, 1);
 }
 
+static inline bool vpid_sync_vcpu_addr(int vpid, gva_t addr)
+{
+	if (vpid == 0)
+		return true;
+
+	if (cpu_has_vmx_invvpid_individual_addr()) {
+		__invvpid(VMX_VPID_EXTENT_INDIVIDUAL_ADDR, vpid, addr);
+		return true;
+	}
+
+	return false;
+}
+
 static inline void vpid_sync_vcpu_single(int vpid)
 {
 	if (vpid == 0)
@@ -4833,6 +4846,20 @@ static void vmx_flush_tlb(struct kvm_vcpu *vcpu, bool invalidate_gpa)
 	__vmx_flush_tlb(vcpu, to_vmx(vcpu)->vpid, invalidate_gpa);
 }
 
+static void vmx_flush_tlb_gva(struct kvm_vcpu *vcpu, gva_t addr)
+{
+	int vpid = to_vmx(vcpu)->vpid;
+
+	if (!vpid_sync_vcpu_addr(vpid, addr))
+		vpid_sync_context(vpid);
+
+	/*
+	 * If VPIDs are not supported or enabled, then the above is a no-op.
+	 * But we don't really need a TLB flush in that case anyway, because
+	 * each VM entry/exit includes an implicit flush when VPID is 0.
+	 */
+}
+
 static void vmx_decache_cr0_guest_bits(struct kvm_vcpu *vcpu)
 {
 	ulong cr0_guest_owned_bits = vcpu->arch.cr0_guest_owned_bits;
@@ -13603,6 +13630,7 @@ static struct kvm_x86_ops vmx_x86_ops __ro_after_init = {
 	.set_rflags = vmx_set_rflags,
 
 	.tlb_flush = vmx_flush_tlb,
+	.tlb_flush_gva = vmx_flush_tlb_gva,
 
 	.run = vmx_vcpu_run,
 	.handle_exit = vmx_handle_exit,