Fix svc encoding for chrome mobile on iOS (#3751)

The browser could send rtp packets of svc encoding without
DD extension while the sdp negotiates it, sfu detects extension
in rtp packet for this case.

pkg/clientconfiguration/conf.go

@@ -49,21 +49,6 @@ var StaticConfigurations = []ConfigurationItem{
 		},
 		Merge: true,
 	},
-	// disable advanced codecs when publishing using JS SDK from iOS,
-	// seeing publish failures (no DD header extension found) when Chrome Mobile publishes VP9,
-	// being defensive and disabling advanced codecs
-	{
-		Match: must.Get(NewScriptMatch(`c.os == "ios" && c.sdk == "js"`)),
-		Configuration: &livekit.ClientConfiguration{
-			DisabledCodecs: &livekit.DisabledCodecs{
-				Publish: []*livekit.Codec{
-					{Mime: mime.MimeTypeVP9.String()},
-					{Mime: mime.MimeTypeAV1.String()},
-				},
-			},
-		},
-		Merge: true,
-	},
 	{
 		Match: must.Get(NewScriptMatch(`(c.device_model == "xiaomi 2201117ti" && c.os == "android") ||
 		  ((c.browser == "firefox" || c.browser == "firefox mobile") && (c.os == "linux" || c.os == "android"))`)),

pkg/rtc/mediatrack.go

@@ -316,7 +316,6 @@ func (t *MediaTrack) AddReceiver(receiver *webrtc.RTPReceiver, track sfu.TrackRe
 			sfu.WithPliThrottleConfig(t.params.PLIThrottleConfig),
 			sfu.WithAudioConfig(t.params.AudioConfig),
 			sfu.WithLoadBalanceThreshold(20),
-			sfu.WithStreamTrackers(),
 			sfu.WithForwardStats(t.params.ForwardStats),
 		)
 		newWR.OnCloseHandler(func() {

pkg/sfu/buffer/buffer.go

@@ -890,8 +890,8 @@ func (b *Buffer) getExtPacket(rtpPacket *rtp.Packet, arrivalTime int64, flowStat
 		ddVal, videoLayer, err := b.ddParser.Parse(ep.Packet)
 		if err != nil {
 			if errors.Is(err, ErrDDExtentionNotFound) {
-				if b.mime == mime.MimeTypeVP8 {
-					b.logger.Infow("dd extension not found for vp8 packet, disable dd parser")
+				if b.mime == mime.MimeTypeVP8 || b.mime == mime.MimeTypeVP9 {
+					b.logger.Infow("dd extension not found,  disable dd parser")
 					b.ddParser = nil
 					b.createFrameRateCalculator()
 				}

pkg/sfu/forwarder.go

@@ -231,6 +231,7 @@ type Forwarder struct {
 	dummyStartTSOffset       uint64
 	refInfos                 [buffer.DefaultMaxLayerSpatial + 1]refInfo
 	refIsSVC                 bool
+	isDDAvailable            bool
 
 	provisional *VideoAllocationProvisional
 
@@ -341,8 +342,8 @@ func (f *Forwarder) DetermineCodec(codec webrtc.RTPCodecCapability, extensions [
 
 	case mime.MimeTypeVP9:
 		// DD-TODO : we only enable dd layer selector for av1/vp9 now, in the future we can enable it for vp8 too
-		isDDAvailable := ddAvailable(extensions)
-		if isDDAvailable {
+		f.isDDAvailable = ddAvailable(extensions)
+		if f.isDDAvailable {
 			if f.vls != nil {
 				f.vls = videolayerselector.NewDependencyDescriptorFromOther(f.vls)
 			} else {
@@ -359,8 +360,8 @@ func (f *Forwarder) DetermineCodec(codec webrtc.RTPCodecCapability, extensions [
 
 	case mime.MimeTypeAV1:
 		// DD-TODO : we only enable dd layer selector for av1/vp9 now, in the future we can enable it for vp8 too
-		isDDAvailable := ddAvailable(extensions)
-		if isDDAvailable {
+		f.isDDAvailable = ddAvailable(extensions)
+		if f.isDDAvailable {
 			if f.vls != nil {
 				f.vls = videolayerselector.NewDependencyDescriptorFromOther(f.vls)
 			} else {
@@ -2002,6 +2003,15 @@ func (f *Forwarder) getTranslationParamsVideo(extPkt *buffer.ExtPacket, layer in
 
 	result := f.vls.Select(extPkt, layer)
 	if !result.IsSelected {
+		if f.isDDAvailable && extPkt.DependencyDescriptor == nil {
+			f.isDDAvailable = false
+			switch f.mime {
+			case mime.MimeTypeVP9:
+				f.vls = videolayerselector.NewVP9FromOther(f.vls)
+			case mime.MimeTypeAV1:
+				f.vls = videolayerselector.NewSimulcastFromOther(f.vls)
+			}
+		}
 		tp.shouldDrop = true
 		if f.started && result.IsRelevant {
 			// call to update highest incoming sequence number and other internal structures

pkg/sfu/receiver.go

@@ -34,7 +34,6 @@ import (
 	"github.com/livekit/livekit-server/pkg/sfu/buffer"
 	"github.com/livekit/livekit-server/pkg/sfu/connectionquality"
 	"github.com/livekit/livekit-server/pkg/sfu/mime"
-	dd "github.com/livekit/livekit-server/pkg/sfu/rtpextension/dependencydescriptor"
 	"github.com/livekit/livekit-server/pkg/sfu/rtpstats"
 	"github.com/livekit/livekit-server/pkg/sfu/streamtracker"
 )
@@ -175,7 +174,6 @@ type WebRTCReceiver struct {
 	onCloseHandler     func()
 	closeOnce          sync.Once
 	closed             atomic.Bool
-	useTrackers        bool
 	trackInfo          atomic.Pointer[livekit.TrackInfo]
 
 	onRTCP func([]rtcp.Packet)
@@ -219,14 +217,6 @@ func WithAudioConfig(audioConfig AudioConfig) ReceiverOpts {
 	}
 }
 
-// WithStreamTrackers enables StreamTracker use for simulcast
-func WithStreamTrackers() ReceiverOpts {
-	return func(w *WebRTCReceiver) *WebRTCReceiver {
-		w.useTrackers = true
-		return w
-	}
-}
-
 // WithLoadBalanceThreshold enables parallelization of packet writes when downTracks exceeds threshold
 // Value should be between 3 and 150.
 // For a server handling a few large rooms, use a smaller value (required to handle very large (250+ participant) rooms).
@@ -296,15 +286,6 @@ func NewWebRTCReceiver(
 
 	w.streamTrackerManager = NewStreamTrackerManager(logger, trackInfo, w.isSVC, w.codec.ClockRate, streamTrackerManagerConfig)
 	w.streamTrackerManager.SetListener(w)
-	// SVC-TODO: Handle DD for non-SVC cases???
-	if w.isSVC {
-		for _, ext := range receiver.GetParameters().HeaderExtensions {
-			if ext.URI == dd.ExtensionURI {
-				w.streamTrackerManager.AddDependencyDescriptorTrackers()
-				break
-			}
-		}
-	}
 
 	return w
 }
@@ -452,10 +433,6 @@ func (w *WebRTCReceiver) AddUpTrack(track TrackRemote, buff *buffer.Buffer) erro
 	buff.SetRTT(rtt)
 	buff.SetPaused(w.streamTrackerManager.IsPaused())
 
-	if w.Kind() == webrtc.RTPCodecTypeVideo && w.useTrackers {
-		w.streamTrackerManager.AddTracker(layer)
-	}
-
 	go w.forwardRTP(layer, buff)
 	return nil
 }
@@ -769,7 +746,6 @@ func (w *WebRTCReceiver) forwardRTP(layer int32, buff *buffer.Buffer) {
 		w.logger.Errorw("invalid layer", nil, "layer", layer)
 		return
 	}
-	spatialTrackers[layer] = w.streamTrackerManager.GetTracker(layer)
 
 	pktBuf := make([]byte, bucket.MaxPktSize)
 	for {
@@ -822,6 +798,9 @@ func (w *WebRTCReceiver) forwardRTP(layer int32, buff *buffer.Buffer) {
 			if spatialTrackers[spatialLayer] == nil {
 				spatialTrackers[spatialLayer] = w.streamTrackerManager.GetTracker(spatialLayer)
 				if spatialTrackers[spatialLayer] == nil {
+					if w.isSVC && pkt.DependencyDescriptor != nil {
+						w.streamTrackerManager.AddDependencyDescriptorTrackers()
+					}
 					spatialTrackers[spatialLayer] = w.streamTrackerManager.AddTracker(spatialLayer)
 				}
 			}