v2_EN_SrsLibrtmp
librtmp is a client side library, seems from rtmpdump.
The use scenarios of librtmp:
- Play or suck RTMP stream: For example rtmpdump, dvr RTMP stream to flv file.
- Publish RTMP stream: Publish RTMP stream to server.
- Use sync block socket: It's ok for client.
- ARM: Can used for linux arm, for some embed device, to publish stream to server.
- Publish h.264 raw stream: SRS2.0 supports this feature, read publish-h264-raw-data
Note: About the openssl, complex and simple handshake, read RTMP protocol
Note: To cross build srs-librtmp for ARM cpu, read srs-arm
librtmp or srs-librtmp only for client side application, impossible for server side.
SRS provides different librtmp:
- The code of librtmp is hard to maintain.
- The interface of librtmp is hard to use.
- No example, while srs-librtmp provides lots of examples at trunk/research/librtmp.
- Min depends, SRS extract core/kernel/rtmp modules for srs-librtmp.
- Min library requires, srs-librtmp only depends on stdc++.
- NO ST, srs-librtmp does not depends on st.
- Provides bandwidth api, to get the bandwidth data to server, read Bandwidth Test
- Provides tracable log, to get the information on server of client, read Tracable log
- Supports directly publish h.264 raw stream, read publish-h264-raw-data
- Exports SRS to srs-librtmp as single project which can be make to .h and .a, or exports SRS to a single .h and .cpp file, read export srs librtmp
In a word, SRS provides more efficient and simple client library srs-librtmp.
SRS2.0 provides options for configure to export srs-librtmp to single project to make .h and .a, or to single .h and .cpp file.
Usage for export project:
dir=/home/winlin/srs-librtmp &&
rm -rf $dir &&
./configure --export-librtmp-project=$dir &&
cd $dir && make &&
./objs/research/librtmp/srs_play rtmp://ossrs.net/live/livestream
SRS export srs-librtmp project can be make to .h and .a, and compile all release/librtmp examples.
SRS can also export srs-librtmp to a single .h and .cpp file, and generate a simple example:
dir=/home/winlin/srs-librtmp &&
rm -rf $dir &&
./configure --export-librtmp-single=$dir &&
cd $dir && gcc example.c srs_librtmp.cpp -g -O0 -lstdc++ -o example &&
strip example && ./example
Note: The export librtmp support both relative and absolute dir path.
When make SRS, the srs-librtmp will auto generated when configure with librtmp:
./configure --with-librtmp --without-ssl && makeAll examples are built, read Examples.
Note: Recomment to disable ssl, for librtmp does not depends on ssl.
Note: srs-librtmp provides only simple handshake, without complex handshake, eventhough configure with ssl.
When build ok, user can use .h and .a library to build client application.
srs-librtmp only depends on libc++, so can be build on windows.
SRS 2.0 can export srs-librtmp to single project, or a .h and a .cpp file, read export srs librtmp.
Need to port some linux header files.
Note: Donot need ssl and st.
This section descrips the RTMP packet specification, for the srs-librtmp api to read or write data.
The api about data:
- Read RTMP packet from server: int srs_read_packet(int* type, u_int32_t* timestamp, char** data, int* size)
- Write RTMP packet to server: int srs_write_packet(int type, u_int32_t timestamp, char* data, int size)
- Write h.264 raw data to server, read publish-h264-raw-data
The RTMP packet(char* data) for api, is format in flv Video/Audio, read the trunk/doc video_file_format_spec_v10_1.pdf
- Audio data, read
E.4.2.1 AUDIODATA，p76, for example, the aac codec audio data. - Video data, read
E.4.3.1 VIDEODATA，p78, for example, the h.264 video data. - Script data, read
E.4.4.1 SCRIPTDATA，p80, for example, onMetadata call.
The RTMP packet type(int type) defines (in E.4.1 FLV Tag，page 75)：
- Audio: 8, the macro SRS_RTMP_TYPE_AUDIO
- Video: 9, the macro SRS_RTMP_TYPE_VIDEO
- Script: 18, the macro SRS_RTMP_TYPE_SCRIPT
Other parameters, for instance, the timestamp, pass by args.
About the flv specification:
- flv header format:
E.2 The FLV header，p74。 - flv body format:
E.3 The FLV File Body，p74。 - tag tag header:
E.4.1 FLV Tag，p75。
Why use flv format as srs-librtmp api data format:
- Flv is simple enough.
- FFMPEG also use flv for rtmp.
- Only need to add flv tag header, then we can write to flv file.
- When publish flv file to server, only need to parse the tag header, the tag body is the data.
srs-librtmp provides api to publish h.264 raw stream to RTMP server.
Please read http://blog.csdn.net/win_lin/article/details/41170653
When convert h.264 raw stream to RTMP packet:
- The h.264 raw stream does not specifies the dts and pts, which is calculated by encoder.
- The RTMP sequence header, always sent in the first video packet.
- The RTMP-packet = 5bytes(RTMP-header) + h.264-header + h.264-NALU-data. Refer to SrsAvcAacCodec::video_avc_demux
- The srs-librtmp provides api to directly send h.264 raw stream, while the raw stream should starts with annexb header N[00] 00 00 01, where N>=0.
The api of srs-librtmp to send h.264 raw stream:
/**
* write h.264 raw frame over RTMP to rtmp server.
* @param frames the input h264 raw data, encoded h.264 I/P/B frames data.
* frames can be one or more than one frame,
* each frame prefixed h.264 annexb header, by N[00] 00 00 01, where N>=0,
* for instance, frame = header(00 00 00 01) + payload(67 42 80 29 95 A0 14 01 6E 40)
* about annexb, @see H.264-AVC-ISO_IEC_14496-10.pdf, page 211.
* @paam frames_size the size of h264 raw data.
* assert frames_size > 0, at least has 1 bytes header.
* @param dts the dts of h.264 raw data.
* @param pts the pts of h.264 raw data.
*
* @remark, user should free the frames.
* @remark, the tbn of dts/pts is 1/1000 for RTMP, that is, in ms.
* @remark, cts = pts - dts
*
* @return 0, success; otherswise, failed.
*/
extern int srs_h264_write_raw_frames(srs_rtmp_t rtmp,
char* frames, int frames_size, u_int32_t dts, u_int32_t pts
);
For the sample h.264 file: http://ossrs.net/srs.release/3rdparty/720p.h264.raw
The data is:
// SPS
000000016742802995A014016E40
// PPS
0000000168CE3880
// IFrame
0000000165B8041014C038008B0D0D3A071.....
// PFrame
0000000141E02041F8CDDC562BBDEFAD2F.....
The sps and pps can be sent togother, or not:
// SPS+PPS
srs_h264_write_raw_frame('000000016742802995A014016E400000000168CE3880', size, dts, pts)
// IFrame
srs_h264_write_raw_frame('0000000165B8041014C038008B0D0D3A071......', size, dts, pts)
// PFrame
srs_h264_write_raw_frame('0000000141E02041F8CDDC562BBDEFAD2F......', size, dts, pts)
The NALU can be sent together, or not:
// SPS
srs_h264_write_raw_frame('000000016742802995A014016E4', size, dts, pts)
// PPS
srs_h264_write_raw_frame('00000000168CE3880', size, dts, pts)
// IFrame
srs_h264_write_raw_frame('0000000165B8041014C038008B0D0D3A071......', size, dts, pts)
// PFrame
srs_h264_write_raw_frame('0000000141E02041F8CDDC562BBDEFAD2F......', size, dts, pts)
About the api, read https://github.com/ossrs/srs/issues/66#issuecomment-62240521
About to use the api, read https://github.com/ossrs/srs/issues/66#issuecomment-62245512
srs-librtmp provides api to publish raw audio frame to SRS, and supports aac adts format stream.
The api:
/**
* write an audio raw frame to srs.
* not similar to h.264 video, the audio never aggregated, always
* encoded one frame by one, so this api is used to write a frame.
*
* @param sound_format Format of SoundData. The following values are defined:
* 0 = Linear PCM, platform endian
* 1 = ADPCM
* 2 = MP3
* 3 = Linear PCM, little endian
* 4 = Nellymoser 16 kHz mono
* 5 = Nellymoser 8 kHz mono
* 6 = Nellymoser
* 7 = G.711 A-law logarithmic PCM
* 8 = G.711 mu-law logarithmic PCM
* 9 = reserved
* 10 = AAC
* 11 = Speex
* 14 = MP3 8 kHz
* 15 = Device-specific sound
* Formats 7, 8, 14, and 15 are reserved.
* AAC is supported in Flash Player 9,0,115,0 and higher.
* Speex is supported in Flash Player 10 and higher.
* @param sound_rate Sampling rate. The following values are defined:
* 0 = 5.5 kHz
* 1 = 11 kHz
* 2 = 22 kHz
* 3 = 44 kHz
* @param sound_size Size of each audio sample. This parameter only pertains to
* uncompressed formats. Compressed formats always decode
* to 16 bits internally.
* 0 = 8-bit samples
* 1 = 16-bit samples
* @param sound_type Mono or stereo sound
* 0 = Mono sound
* 1 = Stereo sound
* @param timestamp The timestamp of audio.
*
* @example /trunk/research/librtmp/srs_aac_raw_publish.c
* @example /trunk/research/librtmp/srs_audio_raw_publish.c
*
* @remark for aac, the frame must be in ADTS format.
* @see aac-mp4a-format-ISO_IEC_14496-3+2001.pdf, page 75, 1.A.2.2 ADTS
* @remark for aac, only support profile 1-4, AAC main/LC/SSR/LTP,
* @see aac-mp4a-format-ISO_IEC_14496-3+2001.pdf, page 23, 1.5.1.1 Audio object type
*
* @see https://github.com/ossrs/srs/issues/212
* @see E.4.2.1 AUDIODATA of video_file_format_spec_v10_1.pdf
*
* @return 0, success; otherswise, failed.
*/
extern int srs_audio_write_raw_frame(srs_rtmp_t rtmp,
char sound_format, char sound_rate, char sound_size, char sound_type,
char* frame, int frame_size, u_int32_t timestamp
);
/**
* whether aac raw data is in adts format,
* which bytes sequence matches '1111 1111 1111'B, that is 0xFFF.
* @param aac_raw_data the input aac raw data, a encoded aac frame data.
* @param ac_raw_size the size of aac raw data.
*
* @reamrk used to check whether current frame is in adts format.
* @see aac-mp4a-format-ISO_IEC_14496-3+2001.pdf, page 75, 1.A.2.2 ADTS
* @example /trunk/research/librtmp/srs_aac_raw_publish.c
*
* @return 0 false; otherwise, true.
*/
extern srs_bool srs_aac_is_adts(char* aac_raw_data, int ac_raw_size);
/**
* parse the adts header to get the frame size,
* which bytes sequence matches '1111 1111 1111'B, that is 0xFFF.
* @param aac_raw_data the input aac raw data, a encoded aac frame data.
* @param ac_raw_size the size of aac raw data.
*
* @return failed when <=0 failed; otherwise, ok.
*/
extern int srs_aac_adts_frame_size(char* aac_raw_data, int ac_raw_size);
The example for bug #212 is srs_audio_raw_publish.c and srs_aac_raw_publish.c, read examples.
About the api, read https://github.com/ossrs/srs/issues/212#issuecomment-63755405
About to use the example, read https://github.com/ossrs/srs/issues/212#issuecomment-64164018
The examples for srs-librtmp, automatically build when build SRS:
- research/librtmp/srs_play.c: Use srs-librtmp to play RTMP stream.
- research/librtmp/srs_publish.c: Use srs-librtmp to publish RTMP stream.
- research/librtmp/srs_ingest_flv.c: Use srs-librtmp to read local flv to publish as RTMP stream.
- research/librtmp/srs_ingest_rtmp.c: Use srs-librtmp to read RTMP then publish as RTMP stream.
- research/librtmp/srs_bandwidth_check.c: Use srs-librtmp to check bandwidth to server.
- research/librtmp/srs_flv_injecter.c: Use srs-librtmp to inject flv keyframes offset for flv vod stream.
- research/librtmp/srs_flv_parser.c: Use srs-librtmp to show the flv file.
- research/librtmp/srs_detect_rtmp.c: Use srs-librtmp to detect the RTMP stream status.
- research/librtmp/srs_h264_raw_publish.c: Use srs-librtmp to publish h.264 raw stream to RTMP server.
- research/librtmp/srs_audio_raw_publish.c: Use srs-librtmp to publish audio raw stream to RTMP server.
- research/librtmp/srs_aac_raw_publish.c: Use srs-librtmp to publish audio aac adts raw stream to RTMP server.
- research/librtmp/srs_rtmp_dump.c: Use srs-librtmp to dump rtmp stream to flv file.
- ./objs/srs_ingest_hls: Ingest the hls live stream to RTMP to SRS.
Start SRS:
make && ./objs/srs -c srs.conf The publish example:
make && ./objs/research/librtmp/objs/srs_publish rtmp://127.0.0.1:1935/live/livestreamNote: the publish stream send random data, cannot play by player.
The play example:
make && ./objs/research/librtmp/objs/srs_play rtmp://ossrs.net/live/livestreamsuck rtmp stream like rtmpdumpWinlin 2014.11