support multichannel Opus (https://github.com/bluenviron/mediamtx/issues/3355) (#572)

README.md

@@ -153,6 +153,7 @@ In RTSP, media streams are routed between server and clients by using RTP packet
 |[RFC2250, RTP Payload Format for MPEG1/MPEG2 Video](https://datatracker.ietf.org/doc/html/rfc2250)|MPEG-1 video, MPEG-2 audio, MPEG-TS payload formats|
 |[RFC2435, RTP Payload Format for JPEG-compressed Video](https://datatracker.ietf.org/doc/html/rfc2435)|M-JPEG payload format|
 |[RFC7587, RTP Payload Format for the Opus Speech and Audio Codec](https://datatracker.ietf.org/doc/html/rfc7587)|Opus payload format|
+|[Multiopus in libwebrtc](https://webrtc-review.googlesource.com/c/src/+/129768)|Opus payload format|
 |[RFC5215, RTP Payload Format for Vorbis Encoded Audio](https://datatracker.ietf.org/doc/html/rfc5215)|Vorbis payload format|
 |[RFC4184, RTP Payload Format for AC-3 Audio](https://datatracker.ietf.org/doc/html/rfc4184)|AC-3 payload format|
 |[RFC6416, RTP Payload Format for MPEG-4 Audio/Visual Streams](https://datatracker.ietf.org/doc/html/rfc6416)|MPEG-4 audio payload format|

pkg/description/session_test.go

@@ -310,8 +310,9 @@ var casesSession = []struct {
 					IsBackChannel: true,
 					Formats: []format.Format{
 						&format.Opus{
-							PayloadTyp: 111,
-							IsStereo:   false,
+							PayloadTyp:   111,
+							IsStereo:     false,
+							ChannelCount: 1,
 						},
 						&format.Generic{
 							PayloadTyp: 103,
@@ -820,8 +821,9 @@ func TestSessionFindFormat(t *testing.T) {
 				Type: MediaTypeAudio,
 				Formats: []format.Format{
 					&format.Opus{
-						PayloadTyp: 111,
-						IsStereo:   true,
+						PayloadTyp:   111,
+						IsStereo:     true,
+						ChannelCount: 2,
 					},
 				},
 			},

pkg/format/format.go

@@ -112,7 +112,7 @@ func Unmarshal(mediaType string, payloadType uint8, rtpMap string, fmtp map[stri
 
 			// audio
 
-			case codec == "opus":
+			case codec == "opus", codec == "multiopus":
 				return &Opus{}
 
 			case codec == "vorbis":

pkg/format/format_test.go

@@ -645,14 +645,37 @@ var casesFormat = []struct {
 			"sprop-stereo": "1",
 		},
 		&Opus{
-			PayloadTyp: 96,
-			IsStereo:   true,
+			PayloadTyp:   96,
+			IsStereo:     true,
+			ChannelCount: 2,
 		},
 		"opus/48000/2",
 		map[string]string{
 			"sprop-stereo": "1",
 		},
 	},
+	{
+		"audio opus 5.1",
+		"audio",
+		96,
+		"multiopus/48000/6",
+		map[string]string{
+			"num_streams":     "4",
+			"coupled_streams": "2",
+			"channel_mapping": "0,4,1,2,3,5",
+		},
+		&Opus{
+			PayloadTyp:   96,
+			ChannelCount: 6,
+		},
+		"multiopus/48000/6",
+		map[string]string{
+			"channel_mapping":      "0,4,1,2,3,5",
+			"coupled_streams":      "2",
+			"num_streams":          "4",
+			"sprop-maxcapturerate": "48000",
+		},
+	},
 	{
 		"audio ac3",
 		"audio",
@@ -1250,6 +1273,14 @@ func FuzzUnmarshalOpus(f *testing.F) {
 	})
 }
 
+func FuzzUnmarshalOpusMulti(f *testing.F) {
+	f.Add("48000/a")
+
+	f.Fuzz(func(_ *testing.T, a string) {
+		Unmarshal("audio", 96, "multiopus/"+a, nil) //nolint:errcheck
+	})
+}
+
 func FuzzUnmarshalVorbis(f *testing.F) {
 	f.Fuzz(func(_ *testing.T, a, b string) {
 		Unmarshal("audio", 96, "Vorbis/"+a, map[string]string{ //nolint:errcheck

pkg/format/opus.go

@@ -12,33 +12,63 @@ import (
 
 // Opus is the RTP format for the Opus codec.
 // Specification: https://datatracker.ietf.org/doc/html/rfc7587
+// Specification: https://webrtc-review.googlesource.com/c/src/+/129768
 type Opus struct {
-	PayloadTyp uint8
-	IsStereo   bool
+	PayloadTyp   uint8
+	ChannelCount int
+
+	// Deprecated: replaced by ChannelCount.
+	IsStereo bool
 }
 
 func (f *Opus) unmarshal(ctx *unmarshalContext) error {
 	f.PayloadTyp = ctx.payloadType
 
-	tmp := strings.SplitN(ctx.clock, "/", 2)
-	if len(tmp) != 2 {
-		return fmt.Errorf("invalid clock (%v)", ctx.clock)
-	}
-
-	sampleRate, err := strconv.ParseUint(tmp[0], 10, 31)
-	if err != nil || sampleRate != 48000 {
-		return fmt.Errorf("invalid sample rate: %d", sampleRate)
-	}
-
-	channelCount, err := strconv.ParseUint(tmp[1], 10, 31)
-	if err != nil || channelCount != 2 {
-		return fmt.Errorf("invalid channel count: %d", channelCount)
-	}
-
-	for key, val := range ctx.fmtp {
-		if key == "sprop-stereo" {
-			f.IsStereo = (val == "1")
+	if ctx.codec == "opus" {
+		tmp := strings.SplitN(ctx.clock, "/", 2)
+		if len(tmp) != 2 {
+			return fmt.Errorf("invalid clock (%v)", ctx.clock)
 		}
+
+		sampleRate, err := strconv.ParseUint(tmp[0], 10, 31)
+		if err != nil || sampleRate != 48000 {
+			return fmt.Errorf("invalid sample rate: '%s", tmp[0])
+		}
+
+		channelCount, err := strconv.ParseUint(tmp[1], 10, 31)
+		if err != nil || channelCount != 2 {
+			return fmt.Errorf("invalid channel count: '%s'", tmp[1])
+		}
+
+		// assume mono
+		f.ChannelCount = 1
+		f.IsStereo = false
+
+		for key, val := range ctx.fmtp {
+			if key == "sprop-stereo" {
+				if val == "1" {
+					f.ChannelCount = 2
+					f.IsStereo = true
+				}
+			}
+		}
+	} else {
+		tmp := strings.SplitN(ctx.clock, "/", 2)
+		if len(tmp) != 2 {
+			return fmt.Errorf("invalid clock (%v)", ctx.clock)
+		}
+
+		sampleRate, err := strconv.ParseUint(tmp[0], 10, 31)
+		if err != nil || sampleRate != 48000 {
+			return fmt.Errorf("invalid sample rate: '%s'", tmp[0])
+		}
+
+		channelCount, err := strconv.ParseUint(tmp[1], 10, 31)
+		if err != nil {
+			return fmt.Errorf("invalid channel count: '%s'", tmp[1])
+		}
+
+		f.ChannelCount = int(channelCount)
 	}
 
 	return nil
@@ -63,22 +93,77 @@ func (f *Opus) PayloadType() uint8 {
 
 // RTPMap implements Format.
 func (f *Opus) RTPMap() string {
-	// RFC7587: The RTP clock rate in "a=rtpmap" MUST be 48000, and the
-	// number of channels MUST be 2.
-	return "opus/48000/2"
+	if f.ChannelCount <= 2 {
+		// RFC7587: The RTP clock rate in "a=rtpmap" MUST be 48000, and the
+		// number of channels MUST be 2.
+		return "opus/48000/2"
+	}
+
+	return "multiopus/48000/" + strconv.FormatUint(uint64(f.ChannelCount), 10)
 }
 
 // FMTP implements Format.
 func (f *Opus) FMTP() map[string]string {
-	fmtp := map[string]string{
-		"sprop-stereo": func() string {
-			if f.IsStereo {
-				return "1"
-			}
-			return "0"
-		}(),
+	if f.ChannelCount <= 2 {
+		return map[string]string{
+			"sprop-stereo": func() string {
+				if f.ChannelCount == 2 || (f.ChannelCount == 0 && f.IsStereo) {
+					return "1"
+				}
+				return "0"
+			}(),
+		}
+	}
+
+	switch f.ChannelCount {
+	case 3:
+		return map[string]string{
+			"num_streams":          "2",
+			"coupled_streams":      "1",
+			"channel_mapping":      "0,2,1",
+			"sprop-maxcapturerate": "48000",
+		}
+
+	case 4:
+		return map[string]string{
+			"num_streams":          "2",
+			"coupled_streams":      "2",
+			"channel_mapping":      "0,1,2,3",
+			"sprop-maxcapturerate": "48000",
+		}
+
+	case 5:
+		return map[string]string{
+			"num_streams":          "3",
+			"coupled_streams":      "2",
+			"channel_mapping":      "0,4,1,2,3",
+			"sprop-maxcapturerate": "48000",
+		}
+
+	case 6:
+		return map[string]string{
+			"num_streams":          "4",
+			"coupled_streams":      "2",
+			"channel_mapping":      "0,4,1,2,3,5",
+			"sprop-maxcapturerate": "48000",
+		}
+
+	case 7:
+		return map[string]string{
+			"num_streams":          "4",
+			"coupled_streams":      "3",
+			"channel_mapping":      "0,4,1,2,3,5,6",
+			"sprop-maxcapturerate": "48000",
+		}
+
+	default: // assume 8
+		return map[string]string{
+			"num_streams":          "5",
+			"coupled_streams":      "3",
+			"channel_mapping":      "0,6,1,4,5,2,3,7",
+			"sprop-maxcapturerate": "48000",
+		}
 	}
-	return fmtp
 }
 
 // PTSEqualsDTS implements Format.

pkg/format/testdata/fuzz/FuzzUnmarshalOpusMulti/771e938e4458e983

@@ -0,0 +1,2 @@
+go test fuzz v1
+string("0")

pkg/format/testdata/fuzz/FuzzUnmarshalOpusMulti/9cd9b70960b4a733

@@ -0,0 +1,2 @@
+go test fuzz v1
+string("/")