libmpdemux/demux_rtp_codec.cpp

   1 ////////// Codec-specific routines used to interface between "MPlayer"
   2 ////////// and the "LIVE.COM Streaming Media" libraries:
   3
   4 #include "demux_rtp_internal.h"
   5 extern "C" {
   6 #include "stheader.h"
   7 }
   8
   9 static void
  10 needVideoFrameRate(demuxer_t* demuxer, MediaSubsession* subsession); // forward
  11 static Boolean
  12 parseQTState_video(QuickTimeGenericRTPSource::QTState const& qtState,
  13                    unsigned& fourcc); // forward
  14 static Boolean
  15 parseQTState_audio(QuickTimeGenericRTPSource::QTState const& qtState,
  16                    unsigned& fourcc, unsigned& numChannels); // forward
  17
  18 void rtpCodecInitialize_video(demuxer_t* demuxer,
  19                               MediaSubsession* subsession,
  20                               unsigned& flags) {
  21   flags = 0;
  22   // Create a dummy video stream header
  23   // to make the main MPlayer code happy:
  24   sh_video_t* sh_video = new_sh_video(demuxer,0);
  25   BITMAPINFOHEADER* bih
  26     = (BITMAPINFOHEADER*)calloc(1,sizeof(BITMAPINFOHEADER));
  27   bih->biSize = sizeof(BITMAPINFOHEADER);
  28   sh_video->bih = bih;
  29   demux_stream_t* d_video = demuxer->video;
  30   d_video->sh = sh_video; sh_video->ds = d_video;
  31
  32   // Map known video MIME types to the BITMAPINFOHEADER parameters
  33   // that this program uses.  (Note that not all types need all
  34   // of the parameters to be set.)
  35   if (strcmp(subsession->codecName(), "MPV") == 0) {
  36     flags |= RTPSTATE_IS_MPEG12_VIDEO;
  37   } else if (strcmp(subsession->codecName(), "MP1S") == 0 ||
  38              strcmp(subsession->codecName(), "MP2T") == 0) {
  39     flags |= RTPSTATE_IS_MPEG12_VIDEO|RTPSTATE_IS_MULTIPLEXED;
  40   } else if (strcmp(subsession->codecName(), "H263") == 0 ||
  41              strcmp(subsession->codecName(), "H263-1998") == 0) {
  42     bih->biCompression = sh_video->format
  43       = mmioFOURCC('H','2','6','3');
  44     needVideoFrameRate(demuxer, subsession);
  45   } else if (strcmp(subsession->codecName(), "H261") == 0) {
  46     bih->biCompression = sh_video->format
  47       = mmioFOURCC('H','2','6','1');
  48     needVideoFrameRate(demuxer, subsession);
  49   } else if (strcmp(subsession->codecName(), "JPEG") == 0) {
  50     bih->biCompression = sh_video->format
  51       = mmioFOURCC('M','J','P','G');
  52     needVideoFrameRate(demuxer, subsession);
  53   } else if (strcmp(subsession->codecName(), "MP4V-ES") == 0) {
  54     bih->biCompression = sh_video->format
  55       = mmioFOURCC('m','p','4','v');
  56     // For the codec to work correctly, it may need a 'VOL Header' to be
  57     // inserted at the front of the data stream.  Construct this from the
  58     // "config" MIME parameter, which was present (hopefully) in the
  59     // session's SDP description:
  60     unsigned configLen;
  61     unsigned char* configData
  62       = parseGeneralConfigStr(subsession->fmtp_config(), configLen);
  63     insertRTPData(demuxer, demuxer->video, configData, configLen);
  64     needVideoFrameRate(demuxer, subsession);
  65   } else if (strcmp(subsession->codecName(), "X-QT") == 0 ||
  66              strcmp(subsession->codecName(), "X-QUICKTIME") == 0) {
  67     // QuickTime generic RTP format, as described in
  68     // http://developer.apple.com/quicktime/icefloe/dispatch026.html
  69
  70     // We can't initialize this stream until we've received the first packet
  71     // that has QuickTime "sdAtom" information in the header.  So, keep
  72     // reading packets until we get one:
  73     unsigned char* packetData; unsigned packetDataLen; float pts;
  74     QuickTimeGenericRTPSource* qtRTPSource
  75       = (QuickTimeGenericRTPSource*)(subsession->rtpSource());
  76     unsigned fourcc;
  77     do {
  78       if (!awaitRTPPacket(demuxer, demuxer->video,
  79                           packetData, packetDataLen, pts)) {
  80         return;
  81       }
  82     } while (!parseQTState_video(qtRTPSource->qtState, fourcc));
  83
  84     bih->biCompression = sh_video->format = fourcc;
  85   } else {
  86     fprintf(stderr,
  87             "Unknown MPlayer format code for MIME type \"video/%s\"\n",
  88             subsession->codecName());
  89   }
  90 }
  91
  92 void rtpCodecInitialize_audio(demuxer_t* demuxer,
  93                               MediaSubsession* subsession,
  94                               unsigned& flags) {
  95   flags = 0;
  96   // Create a dummy audio stream header
  97   // to make the main MPlayer code happy:
  98   sh_audio_t* sh_audio = new_sh_audio(demuxer,0);
  99   WAVEFORMATEX* wf = (WAVEFORMATEX*)calloc(1,sizeof(WAVEFORMATEX));
 100   sh_audio->wf = wf;
 101   demux_stream_t* d_audio = demuxer->audio;
 102   d_audio->sh = sh_audio; sh_audio->ds = d_audio;
 103
 104   wf->nChannels = subsession->numChannels();
 105
 106   // Map known audio MIME types to the WAVEFORMATEX parameters
 107   // that this program uses.  (Note that not all types need all
 108   // of the parameters to be set.)
 109   wf->nSamplesPerSec
 110     = subsession->rtpSource()->timestampFrequency(); // by default
 111   if (strcmp(subsession->codecName(), "MPA") == 0 ||
 112       strcmp(subsession->codecName(), "MPA-ROBUST") == 0 ||
 113       strcmp(subsession->codecName(), "X-MP3-DRAFT-00") == 0) {
 114     wf->wFormatTag = sh_audio->format = 0x55;
 115     // Note: 0x55 is for layer III, but should work for I,II also
 116     wf->nSamplesPerSec = 0; // sample rate is deduced from the data
 117   } else if (strcmp(subsession->codecName(), "AC3") == 0) {
 118     wf->wFormatTag = sh_audio->format = 0x2000;
 119     wf->nSamplesPerSec = 0; // sample rate is deduced from the data
 120   } else if (strcmp(subsession->codecName(), "L16") == 0) {
 121     wf->wFormatTag = sh_audio->format = 0x736f7774; // "twos"
 122     wf->nBlockAlign = 1;
 123     wf->wBitsPerSample = 16;
 124     wf->cbSize = 0;
 125   } else if (strcmp(subsession->codecName(), "L8") == 0) {
 126     wf->wFormatTag = sh_audio->format = 0x20776172; // "raw "
 127     wf->nBlockAlign = 1;
 128     wf->wBitsPerSample = 8;
 129     wf->cbSize = 0;
 130   } else if (strcmp(subsession->codecName(), "PCMU") == 0) {
 131     wf->wFormatTag = sh_audio->format = 0x7;
 132     wf->nAvgBytesPerSec = 8000;
 133     wf->nBlockAlign = 1;
 134     wf->wBitsPerSample = 8;
 135     wf->cbSize = 0;
 136   } else if (strcmp(subsession->codecName(), "PCMA") == 0) {
 137     wf->wFormatTag = sh_audio->format = 0x6;
 138     wf->nAvgBytesPerSec = 8000;
 139     wf->nBlockAlign = 1;
 140     wf->wBitsPerSample = 8;
 141     wf->cbSize = 0;
 142   } else if (strcmp(subsession->codecName(), "GSM") == 0) {
 143     wf->wFormatTag = sh_audio->format = mmioFOURCC('a','g','s','m');
 144     wf->nAvgBytesPerSec = 1650;
 145     wf->nBlockAlign = 33;
 146     wf->wBitsPerSample = 16;
 147     wf->cbSize = 0;
 148   } else if (strcmp(subsession->codecName(), "QCELP") == 0) {
 149     wf->wFormatTag = sh_audio->format = mmioFOURCC('Q','c','l','p');
 150     wf->nAvgBytesPerSec = 1750;
 151     wf->nBlockAlign = 35;
 152     wf->wBitsPerSample = 16;
 153     wf->cbSize = 0;
 154   } else if (strcmp(subsession->codecName(), "MP4A-LATM") == 0) {
 155     wf->wFormatTag = sh_audio->format = mmioFOURCC('m','p','4','a');
 156     // For the codec to work correctly, it needs "AudioSpecificConfig"
 157     // data, which is parsed from the "StreamMuxConfig" string that
 158     // was present (hopefully) in the SDP description:
 159     unsigned codecdata_len;
 160     sh_audio->codecdata
 161       = parseStreamMuxConfigStr(subsession->fmtp_config(),
 162                                 codecdata_len);
 163     sh_audio->codecdata_len = codecdata_len;
 164   } else if (strcmp(subsession->codecName(), "MPEG4-GENERIC") == 0) {
 165     wf->wFormatTag = sh_audio->format = mmioFOURCC('m','p','4','a');
 166     // For the codec to work correctly, it needs "AudioSpecificConfig"
 167     // data, which was present (hopefully) in the SDP description:
 168     unsigned codecdata_len;
 169     sh_audio->codecdata
 170       = parseGeneralConfigStr(subsession->fmtp_config(),
 171                               codecdata_len);
 172     sh_audio->codecdata_len = codecdata_len;
 173   } else if (strcmp(subsession->codecName(), "X-QT") == 0 ||
 174              strcmp(subsession->codecName(), "X-QUICKTIME") == 0) {
 175     // QuickTime generic RTP format, as described in
 176     // http://developer.apple.com/quicktime/icefloe/dispatch026.html
 177
 178     // We can't initialize this stream until we've received the first packet
 179     // that has QuickTime "sdAtom" information in the header.  So, keep
 180     // reading packets until we get one:
 181     unsigned char* packetData; unsigned packetDataLen; float pts;
 182     QuickTimeGenericRTPSource* qtRTPSource
 183       = (QuickTimeGenericRTPSource*)(subsession->rtpSource());
 184     unsigned fourcc, numChannels;
 185     do {
 186       if (!awaitRTPPacket(demuxer, demuxer->audio,
 187                           packetData, packetDataLen, pts)) {
 188         return;
 189       }
 190     } while (!parseQTState_audio(qtRTPSource->qtState, fourcc, numChannels));
 191
 192     wf->wFormatTag = sh_audio->format = fourcc;
 193     wf->nChannels = numChannels;
 194   } else {
 195     fprintf(stderr,
 196             "Unknown MPlayer format code for MIME type \"audio/%s\"\n",
 197             subsession->codecName());
 198   }
 199 }
 200
 201 static void needVideoFrameRate(demuxer_t* demuxer,
 202                                MediaSubsession* subsession) {
 203   // For some codecs, MPlayer's decoding software can't (or refuses to :-)
 204   // figure out the frame rate by itself, so (unless the user specifies
 205   // it manually, using "-fps") we figure it out ourselves here, using the
 206   // presentation timestamps in successive packets,
 207   extern float force_fps; if (force_fps != 0.0) return; // user used "-fps"
 208
 209   demux_stream_t* d_video = demuxer->video;
 210   sh_video_t* sh_video = (sh_video_t*)(d_video->sh);
 211
 212   // If we already know the subsession's video frame rate, use it:
 213   int fps = (int)(subsession->videoFPS());
 214   if (fps != 0) {
 215     sh_video->fps = fps;
 216     return;
 217   }
 218
 219   // Keep looking at incoming frames until we see two with different,
 220   // non-zero "pts" timestamps:
 221   unsigned char* packetData; unsigned packetDataLen;
 222   float lastPTS = 0.0, curPTS;
 223   unsigned const maxNumFramesToWaitFor = 300;
 224   for (unsigned i = 0; i < maxNumFramesToWaitFor; ++i) {
 225     if (!awaitRTPPacket(demuxer, d_video, packetData, packetDataLen, curPTS)) {
 226       break;
 227     }
 228
 229     if (curPTS > lastPTS && lastPTS != 0.0) {
 230       // Use the difference between these two "pts"s to guess the frame rate.
 231       // (should really check that there were no missing frames inbetween)#####
 232       // Guess the frame rate as an integer.  If it's not, use "-fps" instead.
 233       fps = (int)(1/(curPTS-lastPTS) + 0.5); // rounding
 234       fprintf(stderr, "demux_rtp: Guessed the video frame rate as %d frames-per-second.\n\t(If this is wrong, use the \"-fps <frame-rate>\" option instead.)\n", fps);
 235       sh_video->fps = fps;
 236       return;
 237     }
 238     lastPTS = curPTS;
 239   }
 240   fprintf(stderr, "demux_rtp: Failed to guess the video frame rate\n");
 241 }
 242
 243 static Boolean
 244 parseQTState_video(QuickTimeGenericRTPSource::QTState const& qtState,
 245                    unsigned& fourcc) {
 246   // qtState's "sdAtom" field is supposed to contain a QuickTime video
 247   // 'sample description' atom.  This atom's name is the 'fourcc' that we want:
 248   char const* sdAtom = qtState.sdAtom;
 249   if (sdAtom == NULL || qtState.sdAtomSize < 2*4) return False;
 250
 251   fourcc = *(unsigned*)(&sdAtom[4]); // put in host order
 252   return True;
 253 }
 254
 255 static Boolean
 256 parseQTState_audio(QuickTimeGenericRTPSource::QTState const& qtState,
 257                    unsigned& fourcc, unsigned& numChannels) {
 258   // qtState's "sdAtom" field is supposed to contain a QuickTime audio
 259   // 'sample description' atom.  This atom's name is the 'fourcc' that we want.
 260   // Also, the top half of the 5th word following the atom name should
 261   // contain the number of channels ("numChannels") that we want:
 262   char const* sdAtom = qtState.sdAtom;
 263   if (sdAtom == NULL || qtState.sdAtomSize < 7*4) return False;
 264
 265   fourcc = *(unsigned*)(&sdAtom[4]); // put in host order
 266
 267   char const* word7Ptr = &sdAtom[6*4];
 268   numChannels = (word7Ptr[0]<<8)|(word7Ptr[1]);
 269   return True;
 270 }