Use spatial scalability when simulcast is not available.

If the client requested a low-resolution stream and there is none,
switch to SID=0.

rtpconn/rtpconn.go

@@ -129,8 +129,12 @@ func (down *rtpDownTrack) SetCname(cname string) {
 }
 
 type layerInfo struct {
+	// current sid, desired sid, and max sid seen
 	sid, wantedSid, maxSid uint8
+	// current tid, desired tid, and max tid seen
 	tid, wantedTid, maxTid uint8
+	// if true, stick to sid 0
+	limitSid bool
 }
 
 func (down *rtpDownTrack) getLayerInfo() layerInfo {
@@ -139,6 +143,7 @@ func (down *rtpDownTrack) getLayerInfo() layerInfo {
 		sid:       uint8((info & 0xF)),
 		wantedSid: uint8((info >> 4) & 0xF),
 		maxSid:    uint8((info >> 8) & 0xF),
+		limitSid:  ((info >> 12) & 1) != 0,
 		tid:       uint8((info >> 16) & 0xF),
 		wantedTid: uint8((info >> 20) & 0xF),
 		maxTid:    uint8((info >> 24) & 0xF),
@@ -146,10 +151,15 @@ func (down *rtpDownTrack) getLayerInfo() layerInfo {
 }
 
 func (down *rtpDownTrack) setLayerInfo(info layerInfo) {
+	var l uint32
+	if info.limitSid {
+		l = 1 << 12
+	}
 	atomic.StoreUint32(&down.atomics.layerInfo,
 		uint32(info.sid&0xF)|
 			uint32(info.wantedSid&0xF)<<4|
 			uint32(info.maxSid&0xF)<<8|
+			l|
 			uint32(info.tid&0xF)<<16|
 			uint32(info.wantedTid&0xF)<<20|
 			uint32(info.maxTid&0xF)<<24,
@@ -221,6 +231,7 @@ func (down *rtpDownTrack) Write(buf []byte) (int, error) {
 
 	layer := down.getLayerInfo()
 
+	// increase eagerly if this is the first time we see a given layer
 	if flags.Tid > layer.maxTid || flags.Sid > layer.maxSid {
 		if flags.Tid > layer.maxTid {
 			if layer.tid == layer.maxTid {
@@ -230,7 +241,7 @@ func (down *rtpDownTrack) Write(buf []byte) (int, error) {
 			layer.maxTid = flags.Tid
 		}
 		if flags.Sid > layer.maxSid {
-			if layer.sid == layer.maxSid {
+			if layer.sid == layer.maxSid && !layer.limitSid {
 				layer.wantedSid = flags.Sid
 				layer.sid = flags.Sid
 			}
@@ -240,6 +251,7 @@ func (down *rtpDownTrack) Write(buf []byte) (int, error) {
 		down.adjustLayer()
 		layer = down.getLayerInfo()
 	}
+
 	if flags.Start && (layer.tid != layer.wantedTid) {
 		if layer.wantedTid < layer.tid || flags.TidUpSync {
 			layer.tid = layer.wantedTid
@@ -331,7 +343,10 @@ func (t *rtpDownTrack) adjustLayer() {
 	if rate < max*7/8 {
 		// switch up
 		layer := t.getLayerInfo()
-		if layer.sid < layer.maxSid {
+		if layer.limitSid && layer.wantedSid != 0 {
+			layer.wantedSid = 0
+			t.setLayerInfo(layer)
+		} else if !layer.limitSid && layer.sid < layer.maxSid {
 			layer.wantedSid = layer.sid + 1
 			t.setLayerInfo(layer)
 		} else if layer.tid < layer.maxTid {
@@ -345,7 +360,11 @@ func (t *rtpDownTrack) adjustLayer() {
 			layer.wantedTid = layer.tid - 1
 			t.setLayerInfo(layer)
 		} else if layer.sid > 0 {
+			if layer.limitSid {
+				layer.wantedSid = 0
+			} else {
 				layer.wantedSid = layer.sid - 1
+			}
 			t.setLayerInfo(layer)
 		}
 	}

rtpconn/rtpconn_test.go

@@ -18,7 +18,7 @@ func TestDownTrackAtomics(t *testing.T) {
 	down.setSRTime(4, 5)
 	down.maxBitrate.Set(6, rtptime.Jiffies())
 	down.maxREMBBitrate.Set(7, rtptime.Jiffies())
-	info := layerInfo{8, 9, 10, 11, 12, 13}
+	info := layerInfo{8, 9, 10, 11, 12, 13, true}
 	down.setLayerInfo(info)
 	ntp, rtp := down.getTimeOffset()
 	rtt := down.getRTT()

rtpconn/webclient.go

@@ -446,7 +446,7 @@ func delDownTrackUnlocked(conn *rtpDownConnection, track *rtpDownTrack) error {
 	return os.ErrNotExist
 }
 
-func replaceTracks(conn *rtpDownConnection, remote []conn.UpTrack) (bool, error) {
+func replaceTracks(conn *rtpDownConnection, remote []conn.UpTrack, limitSid bool) (bool, error) {
 	conn.mu.Lock()
 	defer conn.mu.Unlock()
 
@@ -489,6 +489,17 @@ outer2:
 		del = append(del, track)
 	}
 
+	defer func() {
+		for _, t := range conn.tracks {
+			layer := t.getLayerInfo()
+			layer.limitSid = limitSid
+			if limitSid {
+				layer.wantedSid = 0
+			}
+			t.setLayerInfo(layer)
+		}
+	}()
+
 	if len(del) == 0 && len(add) == 0 {
 		return false, nil
 	}
@@ -729,7 +740,7 @@ func requestConns(target group.Client, g *group.Group, id string) {
 	}
 }
 
-func requestedTracks(c *webClient, override []string, up conn.Up, tracks []conn.UpTrack) []conn.UpTrack {
+func requestedTracks(c *webClient, override []string, up conn.Up, tracks []conn.UpTrack) ([]conn.UpTrack, bool) {
 	r := override
 	if r == nil {
 		var ok bool
@@ -738,7 +749,7 @@ func requestedTracks(c *webClient, override []string, up conn.Up, tracks []conn.
 			r, ok = c.requested[""]
 		}
 		if !ok || len(r) == 0 {
-			return nil
+			return nil, false
 		}
 	}
 
@@ -777,6 +788,7 @@ func requestedTracks(c *webClient, override []string, up conn.Up, tracks []conn.
 	}
 
 	var ts []conn.UpTrack
+	limitSid := false
 	if audio {
 		t := find(webrtc.RTPCodecTypeAudio)
 		if t != nil {
@@ -796,10 +808,13 @@ func requestedTracks(c *webClient, override []string, up conn.Up, tracks []conn.
 		)
 		if t != nil {
 			ts = append(ts, t)
+			if t.Label() != "l" {
+				limitSid = true
+			}
 		}
 	}
 
-	return ts
+	return ts, limitSid
 }
 
 func (c *webClient) PushConn(g *group.Group, id string, up conn.Up, tracks []conn.UpTrack, replace string) error {
@@ -991,6 +1006,7 @@ func clientLoop(c *webClient, ws *websocket.Conn) error {
 
 func pushDownConn(c *webClient, id string, up conn.Up, tracks []conn.UpTrack, replace string) error {
 	var requested []conn.UpTrack
+	limitSid := false
 	if up != nil {
 		var old *rtpDownConnection
 		if replace != "" {
@@ -1002,7 +1018,7 @@ func pushDownConn(c *webClient, id string, up conn.Up, tracks []conn.UpTrack, re
 		if old != nil {
 			override = old.requested
 		}
-		requested = requestedTracks(c, override, up, tracks)
+		requested, limitSid = requestedTracks(c, override, up, tracks)
 	}
 
 	if replace != "" {
@@ -1031,7 +1047,7 @@ func pushDownConn(c *webClient, id string, up conn.Up, tracks []conn.UpTrack, re
 		}
 		return err
 	}
-	done, err := replaceTracks(down, requested)
+	done, err := replaceTracks(down, requested, limitSid)
 	if err != nil || !done {
 		return err
 	}