Skip to contentSkip to conversation

KVM: X86: avoid meaningless kvm_apicv_activated() check

PATCH linmiaohe 1 participant 1 message Feb 26, 2020 KVM

lore
Share & tools
lore

linmiaohe

author
#0 lore raw
Quoted 13 lines · Vitaly Kuznetsov
Quoted 11 lines · linmiaohe

From: Miaohe Lin <[email protected]>

After test_and_set_bit() for kvm->arch.apicv_inhibit_reasons, we will always get false when calling kvm_apicv_activated() because it's sure apicv_inhibit_reasons do not equal to 0.

---
 arch/x86/kvm/x86.c | 3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)
Quoted 15 lines

arch/x86/kvm/x86.c void kvm_request_apicv_update(struct kvm *kvm, bool activate, ulong bit) · line 8023

Quoted 13 lines
arch/x86/kvm/x86.c index renamed from arch/x86/kvm/x86.c +0 -0

ddcc51b89e2c..fa62dcb0ed0c 100644

arch/x86/kvm/x86.c +1 -2
@@ -8018,8 +8018,7 @@ void kvm_request_apicv_update(struct kvm *kvm, bool activate, ulong bit)
 		    !kvm_apicv_activated(kvm))
 			return;
 	} else {
-		if (test_and_set_bit(bit, &kvm->arch.apicv_inhibit_reasons) ||
-		    kvm_apicv_activated(kvm))
+		if (test_and_set_bit(bit, &kvm->arch.apicv_inhibit_reasons))
 			return;
 	}

This seems to be correct in a sense that we are not really protected against concurrent modifications of 'apicv_inhibit_reasons' (like what if 'apicv_inhibit_reasons' gets modified right after we've checked 'kvm_apicv_activated(kvm)').

Yes, there might be a race window. But this looks benign as we recalculate kvm_apicv_activated() when we proceed with KVM_REQ_APICV_UPDATE.

Quoted 3 lines

The function, however, still gives a flase impression it is somewhat protected against concurent modifications. Like what are these test_and_{set,clear}_bit() for?

Yes, I think so too. And also test_and_{set,clear}_bit() checks wheather the requested bit is {set,clear} to the requested state.

Quoted 37 lines

If I'm not mistaken, the logic this function was supposed to implement
is: change the requested bit to the requested state and, if
kvm_apicv_activated() changed (we set the first bit or cleared the last), proceed with KVM_REQ_APICV_UPDATE. What if we re-write it like

arch/x86/kvm/x86.c +9 -9
@@ -8027,19 +8027,19 @@ EXPORT_SYMBOL_GPL(kvm_vcpu_update_apicv);
 */
 void kvm_request_apicv_update(struct kvm *kvm, bool activate, ulong bit)  {
+       bool apicv_was_activated = kvm_apicv_activated(kvm);
+
       if (!kvm_x86_ops->check_apicv_inhibit_reasons ||
           !kvm_x86_ops->check_apicv_inhibit_reasons(bit))
               return;
 
-       if (activate) {
-               if (!test_and_clear_bit(bit, &kvm->arch.apicv_inhibit_reasons) ||
-                   !kvm_apicv_activated(kvm))
-                       return;
-       } else {
-               if (test_and_set_bit(bit, &kvm->arch.apicv_inhibit_reasons) ||
-                   kvm_apicv_activated(kvm))
-                       return;
-       }
+       if (activate)
+               clear_bit(bit, &kvm->arch.apicv_inhibit_reasons);
+       else
+               set_bit(bit, &kvm->arch.apicv_inhibit_reasons);
+
+       if (kvm_apicv_activated(kvm) == apicv_was_activated)
+               return;
 
       trace_kvm_apicv_update_request(activate, bit);
       if (kvm_x86_ops->pre_update_apicv_exec_ctrl)
 
Quoted 2 lines

Is this equal?

Looks good. I think this version also improves the readability. Many thanks for your advice and review!