fix: lds wasmplugin cache issue

pilot/pkg/networking/core/v1alpha3/gateway.go

@@ -310,6 +310,8 @@ func (configgen *ConfigGeneratorImpl) buildGatewayListeners(builder *ListenerBui
 			}
 		}
 	}
+
+	unpatchedListeners := make([]*listener.Listener, 0)
 	for _, ml := range mutableopts {
 		ml.mutable.Listener = buildGatewayListener(*ml.opts, ml.transport)
 		log.Debugf("buildGatewayListeners: marshaling listener %q with %d filter chains",
@@ -320,22 +322,28 @@ func (configgen *ConfigGeneratorImpl) buildGatewayListeners(builder *ListenerBui
 			errs = multierror.Append(errs, fmt.Errorf("gateway omitting listener %q due to: %v", ml.mutable.Listener.Name, err.Error()))
 			continue
 		}
-		listeners = append(listeners, ml.mutable.Listener)
-
-		if features.EnableLDSCaching {
-			listenerCache := &ListenerCache{
-				ListenerName:    ml.mutable.Listener.Name,
-				Gateways:        gatewaysByListenerName[ml.mutable.Listener.Name],
-				EnvoyFilterKeys: efKeys,
-				WasmPlugins:     listenerWasmPlugins,
-			}
-			resource := &discovery.Resource{
-				Name:     ml.mutable.Listener.Name,
-				Resource: protoconv.MessageToAny(ml.mutable.Listener),
+		unpatchedListeners = append(unpatchedListeners, ml.mutable.Listener)
+	}
+	if len(unpatchedListeners) > 0 {
+		builder.gatewayListeners = unpatchedListeners
+		builder.patchListeners()
+		for _, l := range builder.gatewayListeners {
+			if features.EnableLDSCaching {
+				listenerCache := &ListenerCache{
+					ListenerName:    l.Name,
+					Gateways:        gatewaysByListenerName[l.Name],
+					EnvoyFilterKeys: efKeys,
+					WasmPlugins:     listenerWasmPlugins,
+				}
+				resource := &discovery.Resource{
+					Name:     l.Name,
+					Resource: protoconv.MessageToAny(l),
+				}
+				configgen.Cache.Add(listenerCache, req, resource)
 			}
-			configgen.Cache.Add(listenerCache, req, resource)
 		}
 	}
+
 	// We'll try to return any listeners we successfully marshaled; if we have none, we'll emit the error we built up
 	err := errs.ErrorOrNil()
 	if err != nil {
@@ -348,7 +356,7 @@ func (configgen *ConfigGeneratorImpl) buildGatewayListeners(builder *ListenerBui
 		return builder, cacheStats{}
 	}
 
-	builder.gatewayListeners = listeners
+	builder.gatewayListeners = append(builder.gatewayListeners, listeners...)
 	return builder, cacheStats{hits: hit, miss: miss}
 }
 

pilot/pkg/networking/core/v1alpha3/listener.go

@@ -113,31 +113,14 @@ func (configgen *ConfigGeneratorImpl) BuildListeners(node *model.Proxy,
 		builder, cacheStats = configgen.buildGatewayListeners(builder, req, efKeys)
 	}
 
-	var l []*listener.Listener
-	// Because the EnvoyFilter still needs to be patched before the final LDS data is generated,
-	// it is determined here that as long as all LDS caches are hit, the cached listeners will be
-	// used to avoid repeatedly patching the EnvoyFilter.
-	var useCached bool
-	if !cacheStats.empty() && cacheStats.miss == 0 {
-		node.RLock()
-		if len(node.CachedListeners) > 0 {
-			l = node.CachedListeners
-			useCached = true
-			log.Debugf("node %s use cached listeners: %+v", node.ID, node.CachedListeners)
-		}
-		node.RUnlock()
-	}
-	if !useCached {
+	if builder.node.Type != model.Router {
 		builder.patchListeners()
-		l = builder.getListeners()
-		if builder.node.EnableHBONE() && !builder.node.IsAmbient() {
-			l = append(l, buildConnectOriginateListener())
-		}
-		node.Lock()
-		node.CachedListeners = l
-		log.Debugf("node %s cached listeners: %+v", node.ID, node.CachedListeners)
-		node.Unlock()
 	}
+	l := builder.getListeners()
+	if builder.node.EnableHBONE() && !builder.node.IsAmbient() {
+		l = append(l, buildConnectOriginateListener())
+	}
+
 	return l, model.XdsLogDetails{AdditionalInfo: fmt.Sprintf("cached:%v/%v", cacheStats.hits, cacheStats.hits+cacheStats.miss)}
 }