[fix] race condition when trying to reconnect (#5547)

client/service.go

@@ -275,7 +275,15 @@ func (svr *Service) Run(ctx context.Context) error {
 }
 
 func (svr *Service) keepControllerWorking() {
-	<-svr.ctl.Done()
+	// ctl is swapped under ctlMu in loopLoginUntilSuccess / stop; take a
+	// snapshot before waiting so stop's nil write does not race this read.
+	svr.ctlMu.RLock()
+	ctl := svr.ctl
+	svr.ctlMu.RUnlock()
+	if ctl == nil {
+		return
+	}
+	<-ctl.Done()
 
 	// There is a situation where the login is successful but due to certain reasons,
 	// the control immediately exits. It is necessary to limit the frequency of reconnection in this case.
@@ -285,8 +293,11 @@ func (svr *Service) keepControllerWorking() {
 		// loopLoginUntilSuccess is another layer of loop that will continuously attempt to
 		// login to the server until successful.
 		svr.loopLoginUntilSuccess(20*time.Second, false)
-		if svr.ctl != nil {
-			<-svr.ctl.Done()
+		svr.ctlMu.RLock()
+		ctl := svr.ctl
+		svr.ctlMu.RUnlock()
+		if ctl != nil {
+			<-ctl.Done()
 			return false, errors.New("control is closed and try another loop")
 		}
 		// If the control is nil, it means that the login failed and the service is also closed.
@@ -377,7 +388,7 @@ func (svr *Service) UpdateAllConfigurer(proxyCfgs []v1.ProxyConfigurer, visitorC
 	svr.ctlMu.RUnlock()
 
 	if ctl != nil {
-		return svr.ctl.UpdateAllConfigurer(proxyCfgs, visitorCfgs)
+		return ctl.UpdateAllConfigurer(proxyCfgs, visitorCfgs)
 	}
 	return nil
 }