feat: use native jetkvm-audio

Author: lqs

audio.go

@@ -1,81 +1,14 @@
 package kvm
 
 import (
-	"fmt"
-	"net"
 	"os/exec"
-	"sync"
-	"syscall"
-	"time"
 )
 
-func startFFmpeg() (cmd *exec.Cmd, err error) {
-	binaryPath := "/userdata/jetkvm/bin/ffmpeg"
-	// Run the binary in the background
-	cmd = exec.Command(binaryPath,
-		"-f", "alsa",
-		"-channels", "2",
-		"-sample_rate", "48000",
-		"-i", "hw:1,0",
-		"-c:a", "libopus",
-		"-b:a", "64k", // ought to be enough for anybody
-		"-vbr", "off",
-		"-frame_duration", "20",
-		"-compression_level", "2",
-		"-f", "rtp",
-		"rtp://127.0.0.1:3333")
-
-	nativeOutputLock := sync.Mutex{}
-	nativeStdout := &nativeOutput{
-		mu:     &nativeOutputLock,
-		logger: nativeLogger.Info().Str("pipe", "stdout"),
-	}
-	nativeStderr := &nativeOutput{
-		mu:     &nativeOutputLock,
-		logger: nativeLogger.Info().Str("pipe", "stderr"),
-	}
-
-	// Redirect stdout and stderr to the current process
-	cmd.Stdout = nativeStdout
-	cmd.Stderr = nativeStderr
-
-	// Set the process group ID so we can kill the process and its children when this process exits
-	cmd.SysProcAttr = &syscall.SysProcAttr{
-		Setpgid:   true,
-		Pdeathsig: syscall.SIGKILL,
-	}
-
-	// Start the command
-	if err := cmd.Start(); err != nil {
-		return nil, fmt.Errorf("failed to start binary: %w", err)
-	}
-
-	return
+func runAudioClient() (cmd *exec.Cmd, err error) {
+	return startNativeBinary("/userdata/jetkvm/bin/jetkvm_audio")
 }
 
-func StartRtpAudioServer(handleClient func(net.Conn)) {
-	scopedLogger := nativeLogger.With().
-		Logger()
-
-	listener, err := net.ListenUDP("udp", &net.UDPAddr{IP: net.ParseIP("127.0.0.1"), Port: 3333})
-	if err != nil {
-		scopedLogger.Warn().Err(err).Msg("failed to start server")
-		return
-	}
-
-	scopedLogger.Info().Msg("server listening")
-
-	go func() {
-		for {
-			cmd, err := startFFmpeg()
-			if err != nil {
-				scopedLogger.Error().Err(err).Msg("failed to start ffmpeg")
-			}
-			err = cmd.Wait()
-			scopedLogger.Error().Err(err).Msg("ffmpeg exited, restarting")
-			time.Sleep(2 * time.Second)
-		}
-	}()
-
-	go handleClient(listener)
+func StartAudioServer() {
+	nativeAudioSocketListener = StartNativeSocketServer("/var/run/jetkvm_audio.sock", handleAudioClient, false)
+	nativeLogger.Debug().Msg("native app audio sock started")
 }

main.go

@@ -77,7 +77,11 @@ func Main() {
 
 	// initialize usb gadget
 	initUsbGadget()
-	StartRtpAudioServer(handleAudioClient)
+
+	StartAudioServer()
+	if _, err := runAudioClient(); err != nil {
+		logger.Warn().Err(err).Msg("failed to run audio client")
+	}
 
 	if err := setInitialVirtualMediaState(); err != nil {
 		logger.Warn().Err(err).Msg("failed to set initial virtual media state")

native.go

@@ -13,8 +13,6 @@ import (
 	"time"
 
 	"github.com/jetkvm/kvm/resource"
-	"github.com/pion/rtp"
-
 	"github.com/pion/webrtc/v4/pkg/media"
 )
 
@@ -107,6 +105,7 @@ func WriteCtrlMessage(message []byte) error {
 
 var nativeCtrlSocketListener net.Listener  //nolint:unused
 var nativeVideoSocketListener net.Listener //nolint:unused
+var nativeAudioSocketListener net.Listener //nolint:unused
 
 var ctrlClientConnected = make(chan struct{})
 
@@ -260,8 +259,6 @@ func handleAudioClient(conn net.Conn) {
 
 	scopedLogger.Info().Msg("native audio socket client connected")
 	inboundPacket := make([]byte, maxAudioFrameSize)
-	var timestamp uint32
-	var packet rtp.Packet
 	for {
 		n, err := conn.Read(inboundPacket)
 		if err != nil {
@@ -270,20 +267,10 @@ func handleAudioClient(conn net.Conn) {
 		}
 
 		if currentSession != nil {
-			if err := packet.Unmarshal(inboundPacket[:n]); err != nil {
-				scopedLogger.Warn().Err(err).Msg("error unmarshalling audio socket packet")
-				continue
-			}
-
-			timestamp += 960
-			packet.Header.Timestamp = timestamp
-			buf, err := packet.Marshal()
-			if err != nil {
-				scopedLogger.Warn().Err(err).Msg("error marshalling packet")
-				continue
-			}
-
-			if _, err := currentSession.AudioTrack.Write(buf); err != nil {
+			if err := currentSession.AudioTrack.WriteSample(media.Sample{
+				Data:     inboundPacket[:n],
+				Duration: 20 * time.Millisecond,
+			}); err != nil {
 				scopedLogger.Warn().Err(err).Msg("error writing sample")
 			}
 		}

webrtc.go

@@ -18,7 +18,7 @@ import (
 type Session struct {
 	peerConnection           *webrtc.PeerConnection
 	VideoTrack               *webrtc.TrackLocalStaticSample
-	AudioTrack               *webrtc.TrackLocalStaticRTP
+	AudioTrack               *webrtc.TrackLocalStaticSample
 	ControlChannel           *webrtc.DataChannel
 	RPCChannel               *webrtc.DataChannel
 	HidChannel               *webrtc.DataChannel
@@ -137,7 +137,7 @@ func newSession(config SessionConfig) (*Session, error) {
 		return nil, err
 	}
 
-	session.AudioTrack, err = webrtc.NewTrackLocalStaticRTP(webrtc.RTPCodecCapability{MimeType: webrtc.MimeTypeOpus}, "audio", "kvm")
+	session.AudioTrack, err = webrtc.NewTrackLocalStaticSample(webrtc.RTPCodecCapability{MimeType: webrtc.MimeTypeOpus}, "audio", "kvm")
 	if err != nil {
 		return nil, err
 	}