KVM: nVMX: Do not flush TLB on L1<->L2 transitions if L1 uses VPID and EPT

Liran Alon · bonzini · commit efebf0aaec3d · 2018-10-13T12:00:55.000+02:00
If L1 uses VPID, it expects TLB to not be flushed on L1&lt;-&gt;L2
transitions. However, code currently flushes TLB nonetheless if we
didn't allocate a vpid02 for L2. As in this case,
vmcs02-&gt;vpid == vmcs01-&gt;vpid == vmx-&gt;vpid.

But, if L1 uses EPT, TLB entires populated by L2 are tagged with EPTP02
while TLB entries populated by L1 are tagged with EPTP01.
Therefore, we can also avoid TLB flush if L1 uses VPID and EPT.

Reviewed-by: Mihai Carabas &lt;mihai.carabas@oracle.com&gt;
Reviewed-by: Darren Kenny &lt;darren.kenny@oracle.com&gt;
Reviewed-by: Nikita Leshenko &lt;nikita.leshchenko@oracle.com&gt;
Signed-off-by: Liran Alon &lt;liran.alon@oracle.com&gt;
Signed-off-by: Paolo Bonzini &lt;pbonzini@redhat.com&gt;
diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
@@ -11964,6 +11964,25 @@ static int nested_vmx_load_cr3(struct kvm_vcpu *vcpu, unsigned long cr3, bool ne
 	return 0;
 }
 
+/*
+ * Returns if KVM is able to config CPU to tag TLB entries
+ * populated by L2 differently than TLB entries populated
+ * by L1.
+ *
+ * If L1 uses EPT, then TLB entries are tagged with different EPTP.
+ *
+ * If L1 uses VPID and we allocated a vpid02, TLB entries are tagged
+ * with different VPID (L1 entries are tagged with vmx->vpid
+ * while L2 entries are tagged with vmx->nested.vpid02).
+ */
+static bool nested_has_guest_tlb_tag(struct kvm_vcpu *vcpu)
+{
+	struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
+
+	return nested_cpu_has_ept(vmcs12) ||
+	       (nested_cpu_has_vpid(vmcs12) && to_vmx(vcpu)->nested.vpid02);
+}
+
 static void prepare_vmcs02_full(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12)
 {
 	struct vcpu_vmx *vmx = to_vmx(vcpu);
@@ -12292,10 +12311,10 @@ static int prepare_vmcs02(struct kvm_vcpu *vcpu, struct vmcs12 *vmcs12,
 		 * influence global bitmap(for vpid01 and vpid02 allocation)
 		 * even if spawn a lot of nested vCPUs.
 		 */
-		if (nested_cpu_has_vpid(vmcs12) && vmx->nested.vpid02) {
+		if (nested_cpu_has_vpid(vmcs12) && nested_has_guest_tlb_tag(vcpu)) {
 			if (vmcs12->virtual_processor_id != vmx->nested.last_vpid) {
 				vmx->nested.last_vpid = vmcs12->virtual_processor_id;
-				__vmx_flush_tlb(vcpu, vmx->nested.vpid02, false);
+				__vmx_flush_tlb(vcpu, nested_get_vpid02(vcpu), false);
 			}
 		} else {
 			/*
@@ -13194,23 +13213,21 @@ static void load_vmcs12_host_state(struct kvm_vcpu *vcpu,
 	load_vmcs12_mmu_host_state(vcpu, vmcs12);
 
 	/*
-	 * If vmcs01 don't use VPID, CPU flushes TLB on every
+	 * If vmcs01 doesn't use VPID, CPU flushes TLB on every
 	 * VMEntry/VMExit. Thus, no need to flush TLB.
 	 *
-	 * If vmcs12 uses VPID, TLB entries populated by L2 are
-	 * tagged with vmx->nested.vpid02 while L1 entries are tagged
-	 * with vmx->vpid. Thus, no need to flush TLB.
+	 * If vmcs12 doesn't use VPID, L1 expects TLB to be
+	 * flushed on every VMEntry/VMExit.
 	 *
-	 * Therefore, flush TLB only in case vmcs01 uses VPID and
-	 * vmcs12 don't use VPID as in this case L1 & L2 TLB entries
-	 * are both tagged with vmx->vpid.
+	 * Otherwise, we can preserve TLB entries as long as we are
+	 * able to tag L1 TLB entries differently than L2 TLB entries.
 	 *
 	 * If vmcs12 uses EPT, we need to execute this flush on EPTP01
 	 * and therefore we request the TLB flush to happen only after VMCS EPTP
 	 * has been set by KVM_REQ_LOAD_CR3.
 	 */
 	if (enable_vpid &&
-	    !(nested_cpu_has_vpid(vmcs12) && to_vmx(vcpu)->nested.vpid02)) {
+	    (!nested_cpu_has_vpid(vmcs12) || !nested_has_guest_tlb_tag(vcpu))) {
 		kvm_make_request(KVM_REQ_TLB_FLUSH, vcpu);
 	}
 
