Send pending routing events

router/peer_aurp.go

@@ -115,7 +115,8 @@ type AURPPeer struct {
 	RouteTable *RouteTable
 
 	// Event tuples yet to be sent to this peer in an RI-Upd.
-	pendingEvents chan aurp.EventTuple
+	pendingEventsMu sync.Mutex
+	pendingEvents   aurp.EventTuples
 
 	// The internal states below are only set within the Handle loop, but can
 	// be read concurrently from outside.
@@ -144,7 +145,6 @@ func NewAURPPeer(routes *RouteTable, udpConn *net.UDPConn, peerAddr string, radd
 		RemoteAddr:     raddr,
 		ReceiveCh:      make(chan aurp.Packet, 1024),
 		RouteTable:     routes,
-		pendingEvents:  make(chan aurp.EventTuple, 1024),
 	}
 }
 
@@ -164,7 +164,9 @@ func (p *AURPPeer) addPendingEvent(ec aurp.EventCode, route *Route) {
 	case aurp.EventCodeND, aurp.EventCodeNRC:
 		et.Distance = 0 // "The distance field does not apply to ND or NRC event tuples and should be set to 0."
 	}
-	p.pendingEvents <- et
+	p.pendingEventsMu.Lock()
+	defer p.pendingEventsMu.Unlock()
+	p.pendingEvents = append(p.pendingEvents, et)
 }
 
 func (p *AURPPeer) RouteAdded(route *Route) {
@@ -300,6 +302,9 @@ func (p *AURPPeer) send(pkt aurp.Packet) (int, error) {
 }
 
 func (p *AURPPeer) Handle(ctx context.Context) error {
+	// Stop listening to events if the goroutine exits
+	defer p.RouteTable.RemoveObserver(p)
+
 	rticker := time.NewTicker(1 * time.Second)
 	defer rticker.Stop()
 	sticker := time.NewTicker(1 * time.Second)
@@ -468,8 +473,29 @@ func (p *AURPPeer) Handle(ctx context.Context) error {
 				if time.Since(p.lastUpdate) <= updateTimer {
 					break
 				}
-				// TODO: is there a routing update to send?
+
+				// Are there routing updates to send?
+				p.pendingEventsMu.Lock()
+				if len(p.pendingEvents) == 0 {
+					p.pendingEventsMu.Unlock()
+					break
+				}
+				// Yes - swap the slices, release the mutex, then send them
+				pending := p.pendingEvents
+				p.pendingEvents = make(aurp.EventTuples, 0, cap(pending))
+				p.pendingEventsMu.Unlock()
+
+				// TODO: eliminate events that cancel out (e.g. NA then ND)
+				// TODO: split pending events to fit within a packet
+
 				p.bumpLastUpdate()
+				aurp.Inc(&p.Transport.LocalSeq)
+				lastRISent = p.Transport.NewRIUpdPacket(pending)
+				if _, err := p.send(lastRISent); err != nil {
+					log.Printf("AURP Peer: Couldn't send RI-Upd packet: %v", err)
+					return err
+				}
+				p.setSState(SenderWaitForRIUpdAck)
 
 			case SenderWaitForRIRspAck, SenderWaitForRIUpdAck:
 				if time.Since(p.lastSend) <= sendRetryTimer {
@@ -482,6 +508,7 @@ func (p *AURPPeer) Handle(ctx context.Context) error {
 				if p.sendRetries >= sendRetryLimit {
 					log.Printf("AURP Peer: Send retry limit reached, closing connection")
 					p.setSState(SenderUnconnected)
+					p.RouteTable.RemoveObserver(p)
 					continue
 				}
 				p.incSendRetries()
@@ -496,6 +523,7 @@ func (p *AURPPeer) Handle(ctx context.Context) error {
 					break
 				}
 				p.setSState(SenderUnconnected)
+				p.RouteTable.RemoveObserver(p)
 			}
 
 		case pkt := <-p.ReceiveCh:
@@ -531,7 +559,9 @@ func (p *AURPPeer) Handle(ctx context.Context) error {
 					return err
 				}
 				if orsp.RateOrErrCode >= 0 {
+					// Data sender is successfully in connected state
 					p.setSState(SenderConnected)
+					p.RouteTable.AddObserver(p)
 				}
 
 				// If receiver is unconnected, commence connecting
@@ -635,6 +665,7 @@ func (p *AURPPeer) Handle(ctx context.Context) error {
 
 				p.setSState(SenderConnected)
 				p.resetSendRetries()
+				p.RouteTable.AddObserver(p)
 
 				// If SZI flag is set, send ZI-Rsp (transaction)
 				if pkt.Flags&aurp.RoutingFlagSendZoneInfo != 0 {