Fix memory leak in danglingEndpoints.

Endpoints which were being terminated in an ERROR state or were moved to CLOSED
by the worker goroutine do not run cleanupLocked() as that should already be run
by the worker termination. But when making that change we made the mistake of
not removing the endpoint from the danglingEndpoints which is normally done in
cleanupLocked().

As a result these endpoints are leaked since a reference is held to them in the
danglingEndpoints array forever till Stack is torn down.

PiperOrigin-RevId: 300438426

pkg/tcpip/transport/tcp/connect.go

@@ -1639,6 +1639,7 @@ func (e *endpoint) doTimeWait() (twReuse func()) {
 	const timeWaitDone = 3
 
 	s := sleep.Sleeper{}
+	defer s.Done()
 	s.AddWaker(&e.newSegmentWaker, newSegment)
 	s.AddWaker(&e.notificationWaker, notification)
 

pkg/tcpip/transport/tcp/endpoint.go

@@ -862,7 +862,6 @@ func (e *endpoint) closeNoShutdown() {
 	e.closed = true
 	// Either perform the local cleanup or kick the worker to make sure it
 	// knows it needs to cleanup.
-	tcpip.AddDanglingEndpoint(e)
 	switch e.EndpointState() {
 	// Sockets in StateSynRecv state(passive connections) are closed when
 	// the handshake fails or if the listening socket is closed while
@@ -876,6 +875,9 @@ func (e *endpoint) closeNoShutdown() {
 		// do nothing.
 	default:
 		e.workerCleanup = true
+		tcpip.AddDanglingEndpoint(e)
+		// Worker will remove the dangling endpoint when the endpoint
+		// goroutine terminates.
 		e.notifyProtocolGoroutine(notifyClose)
 	}