新增支持enhanced-rtmp h265 推流 (#2694)

src/Extension/Factory.cpp

@@ -211,7 +211,7 @@ static CodecId getVideoCodecIdByAmf(const AMFValue &val){
     return CodecInvalid;
 }
 
-Track::Ptr getTrackByCodecId(CodecId codecId, int sample_rate = 0, int channels = 0, int sample_bit = 0) {
+Track::Ptr Factory::getTrackByCodecId(CodecId codecId, int sample_rate, int channels, int sample_bit) {
     switch (codecId){
         case CodecH264 : return std::make_shared<H264Track>();
         case CodecH265 : return std::make_shared<H265Track>();

src/Extension/Factory.h

@@ -21,6 +21,16 @@ namespace mediakit{
 
 class Factory {
 public:
+
+    /**
+     * 根据codec_id 获取track
+     * @param codecId 编码id
+     * @param sample_rate 采样率，视频固定为90000
+     * @param channels 音频通道数
+     * @param sample_bit 音频采样位数
+     */
+    static Track::Ptr getTrackByCodecId(CodecId codecId, int sample_rate = 0, int channels = 0, int sample_bit = 0);
+
     ////////////////////////////////rtsp相关//////////////////////////////////
     /**
      * 根据sdp生成Track对象

src/Extension/H264Rtmp.cpp

@@ -164,12 +164,8 @@ bool H264RtmpEncoder::inputFrame(const Frame::Ptr &frame) {
         //not config
         _rtmp_packet->buffer[1] = true;
         int32_t cts = pts - dts;
-        if (cts < 0) {
-            cts = 0;
-        }
         //cts
         set_be24(&_rtmp_packet->buffer[2], cts);
-
         _rtmp_packet->time_stamp = dts;
         _rtmp_packet->body_size = _rtmp_packet->buffer.size();
         _rtmp_packet->chunk_id = CHUNK_VIDEO;

src/Extension/H265Rtmp.h

@@ -15,7 +15,7 @@
 #include "Extension/Track.h"
 #include "Extension/H265.h"
 
-namespace mediakit{
+namespace mediakit {
 /**
  * h265 Rtmp解码类
  * 将 h265 over rtmp 解复用出 h265-Frame
@@ -25,7 +25,7 @@ public:
     using Ptr = std::shared_ptr<H265RtmpDecoder>;
 
     H265RtmpDecoder();
-    ~H265RtmpDecoder() {}
+    ~H265RtmpDecoder() = default;
 
     /**
      * 输入265 Rtmp包
@@ -33,22 +33,23 @@ public:
      */
     void inputRtmp(const RtmpPacket::Ptr &rtmp) override;
 
-    CodecId getCodecId() const override{
-        return CodecH265;
-    }
+    CodecId getCodecId() const override { return CodecH265; }
 
 protected:
-    void onGetH265(const char *pcData, size_t iLen, uint32_t dts,uint32_t pts);
     H265Frame::Ptr obtainFrame();
 
+    void onGetH265(const char *data, size_t size, uint32_t dts, uint32_t pts);
+    void splitFrame(const uint8_t *data, size_t size, uint32_t dts, uint32_t pts);
+
 protected:
+    RtmpPacketInfo _info;
     H265Frame::Ptr _h265frame;
 };
 
 /**
  * 265 Rtmp打包类
  */
-class H265RtmpEncoder : public H265RtmpDecoder{
+class H265RtmpEncoder : public H265RtmpDecoder {
 public:
     using Ptr = std::shared_ptr<H265RtmpEncoder>;
 
@@ -87,9 +88,9 @@ private:
     std::string _pps;
     H265Track::Ptr _track;
     RtmpPacket::Ptr _rtmp_packet;
-    FrameMerger _merger{FrameMerger::mp4_nal_size};
+    FrameMerger _merger { FrameMerger::mp4_nal_size };
 };
 
-}//namespace mediakit
+} // namespace mediakit
 
-#endif //ZLMEDIAKIT_H265RTMPCODEC_H
+#endif // ZLMEDIAKIT_H265RTMPCODEC_H

src/Rtmp/Rtmp.cpp

@@ -258,6 +258,55 @@ void RtmpHandshake::random_generate(char *bytes, int size)
     }
 }
 
+CodecId parseVideoRtmpPacket(const uint8_t *data, size_t size, RtmpPacketInfo *info) {
+    RtmpPacketInfo save;
+    info = info ? info : &save;
+    info->codec = CodecInvalid;
+
+    CHECK(size > 0);
+    if (data[0] >> 7 == 1) {
+        // IsExHeader == 1
+        CHECK(size >= 5, "Invalid rtmp buffer size: ", size);
+        info->is_enhanced = true;
+        info->video.frame_type = (RtmpFrameType)((data[0] >> 4) & 0x07);
+        info->video.pkt_type = (RtmpPacketType)(data[0] & 0x0f);
+        if (memcmp(data + 1, "av01", 4) == 0) {
+            // AV1
+            info->codec = CodecAV1;
+        } else if (memcmp(data + 1, "vp09", 4) == 0) {
+            // VP9
+            info->codec = CodecVP9;
+        } else if (memcmp(data + 1, "hvc1", 4) == 0) {
+            // HEVC(H265)
+            info->codec = CodecH265;
+        } else {
+            WarnL << "Rtmp video codec not supported: " << std::string((char *)data + 1, 4);
+        }
+    } else {
+        // IsExHeader == 0
+        info->is_enhanced = false;
+        info->video.frame_type = (RtmpFrameType)(data[0] >> 4);
+        info->video.rtmp_codec = (RtmpVideoCodec)(data[0] & 0x0f);
+
+        switch (info->video.rtmp_codec) {
+            case RtmpVideoCodec::h264: {
+                CHECK(size >= 1, "Invalid rtmp buffer size: ", size);
+                info->codec = CodecH264;
+                info->video.h264_pkt_type = (RtmpH264PacketType)data[1];
+                break;
+            }
+            case RtmpVideoCodec::h265: {
+                CHECK(size >= 1, "Invalid rtmp buffer size: ", size);
+                info->codec = CodecH265;
+                info->video.h264_pkt_type = (RtmpH264PacketType)data[1];
+                break;
+            }
+            default: WarnL << "Rtmp video codec not supported: " << (int)info->video.rtmp_codec; break;
+        }
+    }
+    return info->codec;
+}
+
 }//namespace mediakit
 
 namespace toolkit {

src/Rtmp/Rtmp.h

@@ -269,5 +269,72 @@ private:
 //根据音频track获取flags
 uint8_t getAudioRtmpFlags(const Track::Ptr &track);
 
+enum class RtmpFrameType : uint8_t {
+    reserved = 0,
+    key_frame = 1, // key frame (for AVC, a seekable frame)
+    inter_frame = 2, // inter frame (for AVC, a non-seekable frame)
+    disposable_inter_frame = 3, // disposable inter frame (H.263 only)
+    generated_key_frame = 4, // generated key frame (reserved for server use only)
+    video_info_frame = 5, // video info/command frame
+};
+
+enum class RtmpVideoCodec : uint8_t {
+    h263 = 2, // Sorenson H.263
+    screen_video = 3, // Screen video
+    vp6 = 4, // On2 VP6
+    vp6_alpha = 5, // On2 VP6 with alpha channel
+    screen_video2 = 6, // Screen video version 2
+    h264 = 7, // avc
+    h265 = 12, // 国内扩展
+};
+
+enum class RtmpH264PacketType : uint8_t {
+    h264_config_header = 0, // AVC sequence header(sps/pps)
+    h264_nalu = 1, // AVC NALU
+    h264_end_seq = 2, // AVC end of sequence (lower level NALU sequence ender is not REQUIRED or supported)
+};
+
+enum class RtmpPacketType : uint8_t {
+    PacketTypeSequenceStart = 0,
+    PacketTypeCodedFrames = 1,
+    PacketTypeSequenceEnd = 2,
+
+    // CompositionTime Offset is implied to equal zero. This is
+    // an optimization to save putting SI24 composition time value of zero on
+    // the wire. See pseudo code below in the VideoTagBody section
+    PacketTypeCodedFramesX = 3,
+
+    // VideoTagBody does not contain video data. VideoTagBody
+    // instead contains an AMF encoded metadata. See Metadata Frame
+    // section for an illustration of its usage. As an example, the metadata
+    // can be HDR information. This is a good way to signal HDR
+    // information. This also opens up future ways to express additional
+    // metadata that is meant for the next video sequence.
+    //
+    // note: presence of PacketTypeMetadata means that FrameType
+    // flags at the top of this table should be ignored
+    PacketTypeMetadata = 4,
+
+    // Carriage of bitstream in MPEG-2 TS format
+    // note: PacketTypeSequenceStart and PacketTypeMPEG2TSSequenceStart
+    // are mutually exclusive
+    PacketTypeMPEG2TSSequenceStart = 5,
+};
+
+struct RtmpPacketInfo {
+    CodecId codec = CodecInvalid;
+    bool is_enhanced;
+    union {
+        struct {
+            RtmpFrameType frame_type;
+            RtmpVideoCodec rtmp_codec;
+            RtmpPacketType pkt_type;
+            RtmpH264PacketType h264_pkt_type;
+        } video;
+    };
+};
+// https://github.com/veovera/enhanced-rtmp/blob/main/enhanced-rtmp-v1.pdf
+CodecId parseVideoRtmpPacket(const uint8_t *data, size_t size, RtmpPacketInfo *info = nullptr);
+
 }//namespace mediakit
 #endif//__rtmp_h

src/Rtmp/RtmpDemuxer.cpp

@@ -19,12 +19,12 @@ size_t RtmpDemuxer::trackCount(const AMFValue &metadata) {
     size_t ret = 0;
     metadata.object_for_each([&](const string &key, const AMFValue &val) {
         if (key == "videocodecid") {
-            //找到视频
+            // 找到视频
             ++ret;
             return;
         }
         if (key == "audiocodecid") {
-            //找到音频
+            // 找到音频
             ++ret;
             return;
         }
@@ -32,7 +32,7 @@ size_t RtmpDemuxer::trackCount(const AMFValue &metadata) {
     return ret;
 }
 
-bool RtmpDemuxer::loadMetaData(const AMFValue &val){
+bool RtmpDemuxer::loadMetaData(const AMFValue &val) {
     bool ret = false;
     try {
         int audiosamplerate = 0;
@@ -60,12 +60,12 @@ bool RtmpDemuxer::loadMetaData(const AMFValue &val){
                 return;
             }
             if (key == "videocodecid") {
-                //找到视频
+                // 找到视频
                 videocodecid = &val;
                 return;
             }
             if (key == "audiocodecid") {
-                //找到音频
+                // 找到音频
                 audiocodecid = &val;
                 return;
             }
@@ -75,16 +75,22 @@ bool RtmpDemuxer::loadMetaData(const AMFValue &val){
             }
             if (key == "videodatarate") {
                 videodatarate = val.as_integer();
+                _videodatarate = videodatarate * 1024;
                 return;
             }
         });
         if (videocodecid) {
-            //有视频
+            // 有视频
             ret = true;
+            if (videocodecid->type() == AMF_NUMBER && videocodecid->as_integer() == (int)RtmpVideoCodec::h264) {
+                // https://github.com/veovera/enhanced-rtmp/issues/8
+                _complete_delay = true;
+            } else {
                 makeVideoTrack(*videocodecid, videodatarate * 1024);
             }
+        }
         if (audiocodecid) {
-            //有音频
+            // 有音频
             ret = true;
             makeAudioTrack(*audiocodecid, audiosamplerate, audiochannels, audiosamplesize, audiodatarate * 1024);
         }
@@ -92,8 +98,8 @@ bool RtmpDemuxer::loadMetaData(const AMFValue &val){
         WarnL << ex.what();
     }
 
-    if (ret) {
-        //metadata中存在track相关的描述，那么我们根据metadata判断有多少个track
+    if (ret && !_complete_delay) {
+        // metadata中存在track相关的描述，那么我们根据metadata判断有多少个track
         addTrackCompleted();
     }
     return ret;
@@ -108,8 +114,14 @@ void RtmpDemuxer::inputRtmp(const RtmpPacket::Ptr &pkt) {
         case MSG_VIDEO: {
             if (!_try_get_video_track) {
                 _try_get_video_track = true;
-                auto codec = AMFValue(pkt->getMediaType());
-                makeVideoTrack(codec, 0);
+                RtmpPacketInfo info;
+                auto codec_id = parseVideoRtmpPacket((uint8_t *)pkt->data(), pkt->size(), &info);
+                if (codec_id != CodecInvalid) {
+                    makeVideoTrack(Factory::getTrackByCodecId(codec_id), _videodatarate);
+                    if (_complete_delay) {
+                        addTrackCompleted();
+                    }
+                }
             }
             if (_video_rtmp_decoder) {
                 _video_rtmp_decoder->inputRtmp(pkt);
@@ -128,51 +140,55 @@ void RtmpDemuxer::inputRtmp(const RtmpPacket::Ptr &pkt) {
             }
             break;
         }
-        default : break;
+        default: break;
     }
 }
 
 void RtmpDemuxer::makeVideoTrack(const AMFValue &videoCodec, int bit_rate) {
+    makeVideoTrack(Factory::getVideoTrackByAmf(videoCodec), bit_rate);
+}
+
+void RtmpDemuxer::makeVideoTrack(const Track::Ptr &track, int bit_rate) {
     if (_video_rtmp_decoder) {
         return;
     }
-    //生成Track对象
-    _video_track = dynamic_pointer_cast<VideoTrack>(Factory::getVideoTrackByAmf(videoCodec));
+    // 生成Track对象
+    _video_track = dynamic_pointer_cast<VideoTrack>(track);
     if (!_video_track) {
         return;
     }
-    //生成rtmpCodec对象以便解码rtmp
+    // 生成rtmpCodec对象以便解码rtmp
     _video_rtmp_decoder = Factory::getRtmpCodecByTrack(_video_track, false);
     if (!_video_rtmp_decoder) {
-        //找不到相应的rtmp解码器，该track无效
+        // 找不到相应的rtmp解码器，该track无效
         _video_track.reset();
         return;
     }
     _video_track->setBitRate(bit_rate);
-    //设置rtmp解码器代理，生成的frame写入该Track
+    // 设置rtmp解码器代理，生成的frame写入该Track
     _video_rtmp_decoder->addDelegate(_video_track);
     addTrack(_video_track);
     _try_get_video_track = true;
 }
 
-void RtmpDemuxer::makeAudioTrack(const AMFValue &audioCodec,int sample_rate, int channels, int sample_bit, int bit_rate) {
+void RtmpDemuxer::makeAudioTrack(const AMFValue &audioCodec, int sample_rate, int channels, int sample_bit, int bit_rate) {
     if (_audio_rtmp_decoder) {
         return;
     }
-    //生成Track对象
+    // 生成Track对象
     _audio_track = dynamic_pointer_cast<AudioTrack>(Factory::getAudioTrackByAmf(audioCodec, sample_rate, channels, sample_bit));
     if (!_audio_track) {
         return;
     }
-    //生成rtmpCodec对象以便解码rtmp
+    // 生成rtmpCodec对象以便解码rtmp
     _audio_rtmp_decoder = Factory::getRtmpCodecByTrack(_audio_track, false);
     if (!_audio_rtmp_decoder) {
-        //找不到相应的rtmp解码器，该track无效
+        // 找不到相应的rtmp解码器，该track无效
         _audio_track.reset();
         return;
     }
     _audio_track->setBitRate(bit_rate);
-    //设置rtmp解码器代理，生成的frame写入该Track
+    // 设置rtmp解码器代理，生成的frame写入该Track
     _audio_rtmp_decoder->addDelegate(_audio_track);
     addTrack(_audio_track);
     _try_get_audio_track = true;

src/Rtmp/RtmpDemuxer.h

@@ -45,12 +45,15 @@ public:
 
 private:
     void makeVideoTrack(const AMFValue &val, int bit_rate);
+    void makeVideoTrack(const Track::Ptr &val, int bit_rate);
     void makeAudioTrack(const AMFValue &val, int sample_rate, int channels, int sample_bit, int bit_rate);
 
 private:
     bool _try_get_video_track = false;
     bool _try_get_audio_track = false;
+    bool _complete_delay = false;
     float _duration = 0;
+    int _videodatarate = 0;
     AudioTrack::Ptr _audio_track;
     VideoTrack::Ptr _video_track;
     RtmpCodec::Ptr _audio_rtmp_decoder;

src/Rtmp/RtmpPlayer.cpp

@@ -191,6 +191,13 @@ void RtmpPlayer::send_connect() {
     obj.set("audioCodecs", (double) (0x0400));
     //只支持H264
     obj.set("videoCodecs", (double) (0x0080));
+
+    AMFValue fourCcList(AMF_STRICT_ARRAY);
+    fourCcList.add("av01");
+    fourCcList.add("vp09");
+    fourCcList.add("hvc1");
+    obj.set("fourCcList", fourCcList);
+
     sendInvoke("connect", obj);
     addOnResultCB([this](AMFDecoder &dec) {
         //TraceL << "connect result";

src/Rtmp/RtmpPusher.cpp

@@ -135,6 +135,13 @@ void RtmpPusher::send_connect() {
     obj.set("type", "nonprivate");
     obj.set("tcUrl", _tc_url);
     obj.set("swfUrl", _tc_url);
+
+    AMFValue fourCcList(AMF_STRICT_ARRAY);
+    fourCcList.add("av01");
+    fourCcList.add("vp09");
+    fourCcList.add("hvc1");
+    obj.set("fourCcList", fourCcList);
+
     sendInvoke("connect", obj);
     addOnResultCB([this](AMFDecoder &dec) {
         //TraceL << "connect result";