Example #1
0
        private bool ProcessPacket(ReadOnlySpan <byte> data, ref Memory <byte> opus, ref Memory <byte> pcm, IList <ReadOnlyMemory <byte> > pcmPackets, out AudioSender voiceSender, out AudioFormat outputFormat)
        {
            voiceSender  = null;
            outputFormat = default;

            if (!this.Rtp.IsRtpHeader(data))
            {
                return(false);
            }

            this.Rtp.DecodeHeader(data, out var sequence, out var timestamp, out var ssrc, out var hasExtension);

            if (!this.TransmittingSSRCs.TryGetValue(ssrc, out var vtx))
            {
                var decoder = Opus.CreateDecoder();

                vtx = new AudioSender(ssrc, decoder)
                {
                    // user isn't present as we haven't received a speaking event yet.
                    User = null
                };
            }

            voiceSender = vtx;
            if (sequence <= vtx.LastSequence) // out-of-order packet; discard
            {
                return(false);
            }
            var gap = vtx.LastSequence != 0 ? sequence - 1 - vtx.LastSequence : 0;

            if (gap >= 5)
            {
                this.Discord.Logger.LogWarning(VoiceNextEvents.VoiceReceiveFailure, "5 or more voice packets were dropped when receiving");
            }

            Span <byte> nonce = stackalloc byte[Sodium.NonceSize];

            this.Sodium.GetNonce(data, nonce, this.SelectedEncryptionMode);
            this.Rtp.GetDataFromPacket(data, out var encryptedOpus, this.SelectedEncryptionMode);

            var opusSize = Sodium.CalculateSourceSize(encryptedOpus);

            opus = opus.Slice(0, opusSize);
            var opusSpan = opus.Span;

            try
            {
                this.Sodium.Decrypt(encryptedOpus, opusSpan, nonce);

                // Strip extensions, if any
                if (hasExtension)
                {
                    // RFC 5285, 4.2 One-Byte header
                    // http://www.rfcreader.com/#rfc5285_line186
                    if (opusSpan[0] == 0xBE && opusSpan[1] == 0xDE)
                    {
                        var headerLen = opusSpan[2] << 8 | opusSpan[3];
                        var i         = 4;
                        for (; i < headerLen + 4; i++)
                        {
                            var @byte = opusSpan[i];

                            // ID is currently unused since we skip it anyway
                            //var id = (byte)(@byte >> 4);
                            var length = (byte)(@byte & 0x0F) + 1;

                            i += length;
                        }

                        // Strip extension padding too
                        while (opusSpan[i] == 0)
                        {
                            i++;
                        }

                        opusSpan = opusSpan.Slice(i);
                    }

                    // TODO: consider implementing RFC 5285, 4.3. Two-Byte Header
                }

                if (opusSpan[0] == 0x90)
                {
                    // I'm not 100% sure what this header is/does, however removing the data causes no
                    // real issues, and has the added benefit of removing a lot of noise.
                    opusSpan = opusSpan.Slice(2);
                }

                if (gap == 1)
                {
                    var lastSampleCount = this.Opus.GetLastPacketSampleCount(vtx.Decoder);
                    var fecpcm          = new byte[this.AudioFormat.SampleCountToSampleSize(lastSampleCount)];
                    var fecpcmMem       = fecpcm.AsSpan();
                    this.Opus.Decode(vtx.Decoder, opusSpan, ref fecpcmMem, true, out _);
                    pcmPackets.Add(fecpcm.AsMemory(0, fecpcmMem.Length));
                }
                else if (gap > 1)
                {
                    var lastSampleCount = this.Opus.GetLastPacketSampleCount(vtx.Decoder);
                    for (var i = 0; i < gap; i++)
                    {
                        var fecpcm    = new byte[this.AudioFormat.SampleCountToSampleSize(lastSampleCount)];
                        var fecpcmMem = fecpcm.AsSpan();
                        this.Opus.ProcessPacketLoss(vtx.Decoder, lastSampleCount, ref fecpcmMem);
                        pcmPackets.Add(fecpcm.AsMemory(0, fecpcmMem.Length));
                    }
                }

                var pcmSpan = pcm.Span;
                this.Opus.Decode(vtx.Decoder, opusSpan, ref pcmSpan, false, out outputFormat);
                pcm = pcm.Slice(0, pcmSpan.Length);
            }
            finally
            {
                vtx.LastSequence = sequence;
            }

            return(true);
        }
Example #2
0
        private async Task HandleDispatch(JObject jo)
        {
            var opc = (int)jo["op"];
            var opp = jo["d"] as JObject;

            switch (opc)
            {
            case 2:     // READY
                Discord.DebugLogger.LogMessage(LogLevel.Debug, "VoiceNext", "OP2 received", DateTime.Now);
                var vrp = opp.ToObject <VoiceReadyPayload>();
                SSRC = vrp.SSRC;
                ConnectionEndpoint = new ConnectionEndpoint(ConnectionEndpoint.Hostname, vrp.Port);
                // this is not the valid interval
                // oh, discord
                //this.HeartbeatInterval = vrp.HeartbeatInterval;
                HeartbeatTask = Task.Run(HeartbeatAsync);
                await Stage1(vrp).ConfigureAwait(false);

                break;

            case 4:     // SESSION_DESCRIPTION
                Discord.DebugLogger.LogMessage(LogLevel.Debug, "VoiceNext", "OP4 received", DateTime.Now);
                var vsd = opp.ToObject <VoiceSessionDescriptionPayload>();
                Key    = vsd.SecretKey;
                Sodium = new DSharpPlus.VoiceNext.Codec.Sodium(Key.AsMemory());
                await Stage2(vsd).ConfigureAwait(false);

                break;

            case 5:     // SPEAKING
                // Don't spam OP5
                //this.Discord.DebugLogger.LogMessage(LogLevel.Debug, "VoiceNext", "OP5 received", DateTime.Now);
                var spd = opp.ToObject <VoiceSpeakingPayload>();
                var spk = new UserSpeakingEventArgs(Discord)
                {
                    Speaking = spd.Speaking,
                    SSRC     = spd.SSRC.Value,
                };

                if (spk.UserId != 0 && TransmittingSSRCs.TryGetValue(spk.SSRC, out var txssrc5) && txssrc5.UserId == 0)
                {
                    txssrc5.UserId = spk.UserId;
                }
                else
                {
                    var opus = Opus.CreateDecoder();
                    var vtx  = new AudioSender(spk.SSRC, opus)
                    {
                        UserId = spd.UserId.Value
                    };

                    if (!TransmittingSSRCs.TryAdd(spk.SSRC, vtx))
                    {
                        Opus.DestroyDecoder(opus);
                    }
                }

                await _userSpeaking.InvokeAsync(spk).ConfigureAwait(false);

                break;

            case 6:     // HEARTBEAT ACK
                var dt   = DateTime.Now;
                var ping = (int)(dt - LastHeartbeat).TotalMilliseconds;
                Volatile.Write(ref _wsPing, ping);
                Discord.DebugLogger.LogMessage(LogLevel.Debug, "VoiceNext", $"Received voice heartbeat ACK, ping {ping.ToString("#,##0", CultureInfo.InvariantCulture)}ms", dt);
                LastHeartbeat = dt;
                break;

            case 8:     // HELLO
                // this sends a heartbeat interval that we need to use for heartbeating
                HeartbeatInterval = opp["heartbeat_interval"].ToObject <int>();
                break;

            case 9:     // RESUMED
                Discord.DebugLogger.LogMessage(LogLevel.Debug, "VoiceNext", "OP9 received", DateTime.Now);
                HeartbeatTask = Task.Run(HeartbeatAsync);
                break;

            case 12:     // CLIENT_CONNECTED
                var ujpd = opp.ToObject <VoiceUserJoinPayload>();

                {
                    var opus = Opus.CreateDecoder();
                    var vtx  = new AudioSender(ujpd.SSRC, opus)
                    {
                        UserId = ujpd.UserId
                    };

                    if (!TransmittingSSRCs.TryAdd(vtx.SSRC, vtx))
                    {
                        Opus.DestroyDecoder(opus);
                    }
                }

                await _userJoined.InvokeAsync(new VoiceUserJoinEventArgs(Discord) { UserId = ujpd.UserId, SSRC = ujpd.SSRC }).ConfigureAwait(false);

                break;

            case 13:     // CLIENT_DISCONNECTED
                var ulpd = opp.ToObject <VoiceUserLeavePayload>();

                var txssrc = TransmittingSSRCs.FirstOrDefault(x => x.Value.Id == ulpd.UserId);
                if (TransmittingSSRCs.ContainsKey(txssrc.Key))
                {
                    TransmittingSSRCs.TryRemove(txssrc.Key, out var txssrc13);
                    Opus.DestroyDecoder(txssrc13.Decoder);
                }

                await _userLeft.InvokeAsync(new VoiceUserLeaveEventArgs(Discord)
                {
                    UserId = ulpd.UserId,
                    SSRC   = txssrc.Key
                }).ConfigureAwait(false);

                break;

            default:
                Discord.DebugLogger.LogMessage(LogLevel.Warning, "VoiceNext", $"Unknown opcode received: {opc.ToString(CultureInfo.InvariantCulture)}", DateTime.Now);
                break;
            }
        }