libmpdemux/demux_rtp_codec.cpp

   1 /*
   2  * codec-specific routines used to interface between MPlayer
   3  * and the "LIVE555 Streaming Media" libraries
   4  *
   5  * This file is part of MPlayer.
   6  *
   7  * MPlayer is free software; you can redistribute it and/or modify
   8  * it under the terms of the GNU General Public License as published by
   9  * the Free Software Foundation; either version 2 of the License, or
  10  * (at your option) any later version.
  11  *
  12  * MPlayer is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15  * GNU General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU General Public License along
  18  * with MPlayer; if not, write to the Free Software Foundation, Inc.,
  19  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  20  */
  21
  22 #include "demux_rtp_internal.h"
  23 extern "C" {
  24 #include <limits.h>
  25 #include <math.h>
  26 #include "stheader.h"
  27 #include "libavutil/base64.h"
  28 }
  29
  30 AVCodecParserContext * h264parserctx;
  31 AVCodecContext *avcctx;
  32
  33 // Copied from vlc
  34 static unsigned char* parseH264ConfigStr( char const* configStr,
  35                                           unsigned int& configSize )
  36 {
  37
  38     char *dup, *psz;
  39     int i, i_records = 1;
  40
  41     if( configSize )
  42     configSize = 0;
  43     if( configStr == NULL || *configStr == '\0' )
  44         return NULL;
  45     psz = dup = strdup( configStr );
  46
  47  /* Count the number of comma's */
  48     for( psz = dup; *psz != '\0'; ++psz )
  49     {
  50         if( *psz == ',')
  51         {
  52             ++i_records;
  53             *psz = '\0';
  54         }
  55     }
  56
  57     unsigned char *cfg = new unsigned char[5 * strlen(dup)];
  58     psz = dup;
  59     for( i = 0; i < i_records; i++ )
  60     {
  61
  62         cfg[configSize++] = 0x00;
  63         cfg[configSize++] = 0x00;
  64         cfg[configSize++] = 0x01;
  65         configSize += av_base64_decode( (uint8_t*)&cfg[configSize],
  66                                         psz,
  67                                         5 * strlen(dup) - 3 );
  68
  69     psz += strlen(psz)+1;
  70     }
  71     free( dup );
  72
  73     return cfg;
  74 }
  75
  76 static void
  77 needVideoFrameRate(demuxer_t* demuxer, MediaSubsession* subsession); // forward
  78 static Boolean
  79 parseQTState_video(QuickTimeGenericRTPSource::QTState const& qtState,
  80                    unsigned& fourcc); // forward
  81 static Boolean
  82 parseQTState_audio(QuickTimeGenericRTPSource::QTState const& qtState,
  83                    unsigned& fourcc, unsigned& numChannels); // forward
  84
  85 static BITMAPINFOHEADER * insertVideoExtradata(BITMAPINFOHEADER *bih,
  86                                                unsigned char * extraData,
  87                                                unsigned size)
  88 {
  89     BITMAPINFOHEADER * original = bih;
  90     if (!size || size > INT_MAX - sizeof(BITMAPINFOHEADER))
  91         return bih;
  92     bih = (BITMAPINFOHEADER*)realloc(bih, sizeof(BITMAPINFOHEADER) + size);
  93     if (!bih)
  94         return original;
  95     bih->biSize = sizeof(BITMAPINFOHEADER) + size;
  96     memcpy(bih+1, extraData, size);
  97     return bih;
  98 }
  99
 100 void rtpCodecInitialize_video(demuxer_t* demuxer,
 101                               MediaSubsession* subsession,
 102                               unsigned& flags) {
 103   flags = 0;
 104   // Create a dummy video stream header
 105   // to make the main MPlayer code happy:
 106   sh_video_t* sh_video = new_sh_video(demuxer,0);
 107   BITMAPINFOHEADER* bih
 108     = (BITMAPINFOHEADER*)calloc(1,sizeof(BITMAPINFOHEADER));
 109   bih->biSize = sizeof(BITMAPINFOHEADER);
 110   sh_video->bih = bih;
 111   demux_stream_t* d_video = demuxer->video;
 112   d_video->sh = sh_video; sh_video->ds = d_video;
 113
 114   // Map known video MIME types to the BITMAPINFOHEADER parameters
 115   // that this program uses.  (Note that not all types need all
 116   // of the parameters to be set.)
 117   if (strcmp(subsession->codecName(), "MPV") == 0) {
 118     flags |= RTPSTATE_IS_MPEG12_VIDEO;
 119   } else if (strcmp(subsession->codecName(), "MP1S") == 0 ||
 120              strcmp(subsession->codecName(), "MP2T") == 0) {
 121     flags |= RTPSTATE_IS_MPEG12_VIDEO|RTPSTATE_IS_MULTIPLEXED;
 122   } else if (strcmp(subsession->codecName(), "H263") == 0 ||
 123              strcmp(subsession->codecName(), "H263-2000") == 0 ||
 124              strcmp(subsession->codecName(), "H263-1998") == 0) {
 125     bih->biCompression = sh_video->format
 126       = mmioFOURCC('H','2','6','3');
 127     needVideoFrameRate(demuxer, subsession);
 128   } else if (strcmp(subsession->codecName(), "H264") == 0) {
 129     bih->biCompression = sh_video->format
 130       = mmioFOURCC('H','2','6','4');
 131     unsigned int configLen = 0;
 132     unsigned char* configData
 133       = parseH264ConfigStr(subsession->fmtp_spropparametersets(), configLen);
 134     sh_video->bih = bih = insertVideoExtradata(bih, configData, configLen);
 135     int fooLen;
 136     const uint8_t* fooData;
 137     h264parserctx = av_parser_init(CODEC_ID_H264);
 138     avcctx = avcodec_alloc_context3(NULL);
 139     // Pass the config to the parser
 140     h264parserctx->parser->parser_parse(h264parserctx, avcctx,
 141                   &fooData, &fooLen, configData, configLen);
 142     delete[] configData;
 143     needVideoFrameRate(demuxer, subsession);
 144   } else if (strcmp(subsession->codecName(), "H261") == 0) {
 145     bih->biCompression = sh_video->format
 146       = mmioFOURCC('H','2','6','1');
 147     needVideoFrameRate(demuxer, subsession);
 148   } else if (strcmp(subsession->codecName(), "JPEG") == 0) {
 149     bih->biCompression = sh_video->format
 150       = mmioFOURCC('M','J','P','G');
 151     needVideoFrameRate(demuxer, subsession);
 152   } else if (strcmp(subsession->codecName(), "MP4V-ES") == 0) {
 153     bih->biCompression = sh_video->format
 154       = mmioFOURCC('m','p','4','v');
 155     // For the codec to work correctly, it may need a 'VOL Header' to be
 156     // inserted at the front of the data stream.  Construct this from the
 157     // "config" MIME parameter, which was present (hopefully) in the
 158     // session's SDP description:
 159     unsigned configLen;
 160     unsigned char* configData
 161       = parseGeneralConfigStr(subsession->fmtp_config(), configLen);
 162     sh_video->bih = bih = insertVideoExtradata(bih, configData, configLen);
 163     needVideoFrameRate(demuxer, subsession);
 164   } else if (strcmp(subsession->codecName(), "X-QT") == 0 ||
 165              strcmp(subsession->codecName(), "X-QUICKTIME") == 0) {
 166     // QuickTime generic RTP format, as described in
 167     // http://developer.apple.com/quicktime/icefloe/dispatch026.html
 168
 169     // We can't initialize this stream until we've received the first packet
 170     // that has QuickTime "sdAtom" information in the header.  So, keep
 171     // reading packets until we get one:
 172     unsigned char* packetData; unsigned packetDataLen; float pts;
 173     QuickTimeGenericRTPSource* qtRTPSource
 174       = (QuickTimeGenericRTPSource*)(subsession->rtpSource());
 175     unsigned fourcc;
 176     do {
 177       if (!awaitRTPPacket(demuxer, demuxer->video,
 178                           packetData, packetDataLen, pts)) {
 179         return;
 180       }
 181     } while (!parseQTState_video(qtRTPSource->qtState, fourcc));
 182
 183     bih->biCompression = sh_video->format = fourcc;
 184     bih->biWidth = qtRTPSource->qtState.width;
 185     bih->biHeight = qtRTPSource->qtState.height;
 186       if (qtRTPSource->qtState.sdAtomSize > 83)
 187         bih->biBitCount = qtRTPSource->qtState.sdAtom[83];
 188       uint8_t *pos = (uint8_t*)qtRTPSource->qtState.sdAtom + 86;
 189       uint8_t *endpos = (uint8_t*)qtRTPSource->qtState.sdAtom
 190                         + qtRTPSource->qtState.sdAtomSize;
 191       while (pos+8 < endpos) {
 192         unsigned atomLength = pos[0]<<24 | pos[1]<<16 | pos[2]<<8 | pos[3];
 193         if (atomLength == 0 || atomLength > endpos-pos) break;
 194         if (((!memcmp(pos+4, "avcC", 4) && fourcc==mmioFOURCC('a','v','c','1')) ||
 195              !memcmp(pos+4, "esds", 4) ||
 196              (!memcmp(pos+4, "SMI ", 4) && fourcc==mmioFOURCC('S','V','Q','3'))) &&
 197             atomLength > 8) {
 198           sh_video->bih = bih =
 199               insertVideoExtradata(bih, pos+8, atomLength-8);
 200           break;
 201         }
 202         pos += atomLength;
 203       }
 204     needVideoFrameRate(demuxer, subsession);
 205   } else {
 206     fprintf(stderr,
 207             "Unknown MPlayer format code for MIME type \"video/%s\"\n",
 208             subsession->codecName());
 209   }
 210 }
 211
 212 void rtpCodecInitialize_audio(demuxer_t* demuxer,
 213                               MediaSubsession* subsession,
 214                               unsigned& flags) {
 215   flags = 0;
 216   // Create a dummy audio stream header
 217   // to make the main MPlayer code happy:
 218   sh_audio_t* sh_audio = new_sh_audio(demuxer,0);
 219   WAVEFORMATEX* wf = (WAVEFORMATEX*)calloc(1,sizeof(WAVEFORMATEX));
 220   sh_audio->wf = wf;
 221   demux_stream_t* d_audio = demuxer->audio;
 222   d_audio->sh = sh_audio; sh_audio->ds = d_audio;
 223   d_audio->id = sh_audio->aid;
 224
 225   wf->nChannels = subsession->numChannels();
 226
 227   // Map known audio MIME types to the WAVEFORMATEX parameters
 228   // that this program uses.  (Note that not all types need all
 229   // of the parameters to be set.)
 230   wf->nSamplesPerSec
 231     = subsession->rtpSource()->timestampFrequency(); // by default
 232   if (strcmp(subsession->codecName(), "MPA") == 0 ||
 233       strcmp(subsession->codecName(), "MPA-ROBUST") == 0 ||
 234       strcmp(subsession->codecName(), "X-MP3-DRAFT-00") == 0) {
 235     wf->wFormatTag = sh_audio->format = 0x55;
 236     // Note: 0x55 is for layer III, but should work for I,II also
 237     wf->nSamplesPerSec = 0; // sample rate is deduced from the data
 238   } else if (strcmp(subsession->codecName(), "AC3") == 0) {
 239     wf->wFormatTag = sh_audio->format = 0x2000;
 240     wf->nSamplesPerSec = 0; // sample rate is deduced from the data
 241   } else if (strcmp(subsession->codecName(), "L16") == 0) {
 242     wf->wFormatTag = sh_audio->format = 0x736f7774; // "twos"
 243     wf->nBlockAlign = 1;
 244     wf->wBitsPerSample = 16;
 245     wf->cbSize = 0;
 246   } else if (strcmp(subsession->codecName(), "L8") == 0) {
 247     wf->wFormatTag = sh_audio->format = 0x20776172; // "raw "
 248     wf->nBlockAlign = 1;
 249     wf->wBitsPerSample = 8;
 250     wf->cbSize = 0;
 251   } else if (strcmp(subsession->codecName(), "PCMU") == 0) {
 252     wf->wFormatTag = sh_audio->format = 0x7;
 253     wf->nAvgBytesPerSec = 8000;
 254     wf->nBlockAlign = 1;
 255     wf->wBitsPerSample = 8;
 256     wf->cbSize = 0;
 257   } else if (strcmp(subsession->codecName(), "PCMA") == 0) {
 258     wf->wFormatTag = sh_audio->format = 0x6;
 259     wf->nAvgBytesPerSec = 8000;
 260     wf->nBlockAlign = 1;
 261     wf->wBitsPerSample = 8;
 262     wf->cbSize = 0;
 263   } else if (strcmp(subsession->codecName(), "AMR") == 0) {
 264     wf->wFormatTag = sh_audio->format = mmioFOURCC('s','a','m','r');
 265   } else if (strcmp(subsession->codecName(), "AMR-WB") == 0) {
 266     wf->wFormatTag = sh_audio->format = mmioFOURCC('s','a','w','b');
 267   } else if (strcmp(subsession->codecName(), "GSM") == 0) {
 268     wf->wFormatTag = sh_audio->format = mmioFOURCC('a','g','s','m');
 269     wf->nAvgBytesPerSec = 1650;
 270     wf->nBlockAlign = 33;
 271     wf->wBitsPerSample = 16;
 272     wf->cbSize = 0;
 273   } else if (strcmp(subsession->codecName(), "QCELP") == 0) {
 274     wf->wFormatTag = sh_audio->format = mmioFOURCC('Q','c','l','p');
 275     wf->nAvgBytesPerSec = 1750;
 276     wf->nBlockAlign = 35;
 277     wf->wBitsPerSample = 16;
 278     wf->cbSize = 0;
 279   } else if (strcmp(subsession->codecName(), "MP4A-LATM") == 0) {
 280     wf->wFormatTag = sh_audio->format = mmioFOURCC('m','p','4','a');
 281     // For the codec to work correctly, it needs "AudioSpecificConfig"
 282     // data, which is parsed from the "StreamMuxConfig" string that
 283     // was present (hopefully) in the SDP description:
 284     unsigned codecdata_len;
 285     sh_audio->codecdata
 286       = parseStreamMuxConfigStr(subsession->fmtp_config(),
 287                                 codecdata_len);
 288     sh_audio->codecdata_len = codecdata_len;
 289     //faad doesn't understand LATM's data length field, so omit it
 290     ((MPEG4LATMAudioRTPSource*)subsession->rtpSource())->omitLATMDataLengthField();
 291   } else if (strcmp(subsession->codecName(), "MPEG4-GENERIC") == 0) {
 292     wf->wFormatTag = sh_audio->format = mmioFOURCC('m','p','4','a');
 293     // For the codec to work correctly, it needs "AudioSpecificConfig"
 294     // data, which was present (hopefully) in the SDP description:
 295     unsigned codecdata_len;
 296     sh_audio->codecdata
 297       = parseGeneralConfigStr(subsession->fmtp_config(),
 298                               codecdata_len);
 299     sh_audio->codecdata_len = codecdata_len;
 300   } else if (strcmp(subsession->codecName(), "X-QT") == 0 ||
 301              strcmp(subsession->codecName(), "X-QUICKTIME") == 0) {
 302     // QuickTime generic RTP format, as described in
 303     // http://developer.apple.com/quicktime/icefloe/dispatch026.html
 304
 305     // We can't initialize this stream until we've received the first packet
 306     // that has QuickTime "sdAtom" information in the header.  So, keep
 307     // reading packets until we get one:
 308     unsigned char* packetData; unsigned packetDataLen; float pts;
 309     QuickTimeGenericRTPSource* qtRTPSource
 310       = (QuickTimeGenericRTPSource*)(subsession->rtpSource());
 311     unsigned fourcc, numChannels;
 312     do {
 313       if (!awaitRTPPacket(demuxer, demuxer->audio,
 314                           packetData, packetDataLen, pts)) {
 315         return;
 316       }
 317     } while (!parseQTState_audio(qtRTPSource->qtState, fourcc, numChannels));
 318
 319     wf->wFormatTag = sh_audio->format = fourcc;
 320     wf->nChannels = numChannels;
 321
 322       if (qtRTPSource->qtState.sdAtomSize > 33) {
 323         wf->wBitsPerSample = qtRTPSource->qtState.sdAtom[27];
 324         wf->nSamplesPerSec = qtRTPSource->qtState.sdAtom[32]<<8|qtRTPSource->qtState.sdAtom[33];
 325       }
 326     uint8_t *pos = (uint8_t*)qtRTPSource->qtState.sdAtom + 52;
 327     uint8_t *endpos = (uint8_t*)qtRTPSource->qtState.sdAtom
 328                       + qtRTPSource->qtState.sdAtomSize;
 329     while (pos+8 < endpos) {
 330       unsigned atomLength = pos[0]<<24 | pos[1]<<16 | pos[2]<<8 | pos[3];
 331       if (atomLength == 0 || atomLength > endpos-pos) break;
 332       if (!memcmp(pos+4, "wave", 4) && fourcc==mmioFOURCC('Q','D','M','2') &&
 333           atomLength > 8 &&
 334           atomLength <= INT_MAX) {
 335         sh_audio->codecdata = (unsigned char*) malloc(atomLength-8);
 336         if (sh_audio->codecdata) {
 337           memcpy(sh_audio->codecdata, pos+8, atomLength-8);
 338           sh_audio->codecdata_len = atomLength-8;
 339         }
 340         break;
 341       }
 342       pos += atomLength;
 343     }
 344   } else {
 345     fprintf(stderr,
 346             "Unknown MPlayer format code for MIME type \"audio/%s\"\n",
 347             subsession->codecName());
 348   }
 349 }
 350
 351 static void needVideoFrameRate(demuxer_t* demuxer,
 352                                MediaSubsession* subsession) {
 353   // For some codecs, MPlayer's decoding software can't (or refuses to :-)
 354   // figure out the frame rate by itself, so (unless the user specifies
 355   // it manually, using "-fps") we figure it out ourselves here, using the
 356   // presentation timestamps in successive packets,
 357   extern double force_fps; if (force_fps != 0.0) return; // user used "-fps"
 358
 359   demux_stream_t* d_video = demuxer->video;
 360   sh_video_t* sh_video = (sh_video_t*)(d_video->sh);
 361
 362   // If we already know the subsession's video frame rate, use it:
 363   int fps = (int)(subsession->videoFPS());
 364   if (fps != 0) {
 365     sh_video->fps = fps;
 366     sh_video->frametime = 1.0f/fps;
 367     return;
 368   }
 369
 370   // Keep looking at incoming frames until we see two with different,
 371   // non-zero "pts" timestamps:
 372   unsigned char* packetData; unsigned packetDataLen;
 373   float lastPTS = 0.0, curPTS;
 374   unsigned const maxNumFramesToWaitFor = 300;
 375   int lastfps = 0;
 376   for (unsigned i = 0; i < maxNumFramesToWaitFor; ++i) {
 377     if (!awaitRTPPacket(demuxer, d_video, packetData, packetDataLen, curPTS)) {
 378       break;
 379     }
 380
 381     if (curPTS != lastPTS && lastPTS != 0.0) {
 382       // Use the difference between these two "pts"s to guess the frame rate.
 383       // (should really check that there were no missing frames inbetween)#####
 384       // Guess the frame rate as an integer.  If it's not, use "-fps" instead.
 385       fps = (int)(1/fabs(curPTS-lastPTS) + 0.5); // rounding
 386         if (fps == lastfps) {
 387       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);
 388       sh_video->fps = fps;
 389       sh_video->frametime=1.0f/fps;
 390       return;
 391         }
 392       if (fps>lastfps) lastfps = fps;
 393     }
 394     lastPTS = curPTS;
 395   }
 396   fprintf(stderr, "demux_rtp: Failed to guess the video frame rate\n");
 397 }
 398
 399 static Boolean
 400 parseQTState_video(QuickTimeGenericRTPSource::QTState const& qtState,
 401                    unsigned& fourcc) {
 402   // qtState's "sdAtom" field is supposed to contain a QuickTime video
 403   // 'sample description' atom.  This atom's name is the 'fourcc' that we want:
 404   char const* sdAtom = qtState.sdAtom;
 405   if (sdAtom == NULL || qtState.sdAtomSize < 2*4) return False;
 406
 407   fourcc = *(unsigned*)(&sdAtom[4]); // put in host order
 408   return True;
 409 }
 410
 411 static Boolean
 412 parseQTState_audio(QuickTimeGenericRTPSource::QTState const& qtState,
 413                    unsigned& fourcc, unsigned& numChannels) {
 414   // qtState's "sdAtom" field is supposed to contain a QuickTime audio
 415   // 'sample description' atom.  This atom's name is the 'fourcc' that we want.
 416   // Also, the top half of the 5th word following the atom name should
 417   // contain the number of channels ("numChannels") that we want:
 418   char const* sdAtom = qtState.sdAtom;
 419   if (sdAtom == NULL || qtState.sdAtomSize < 7*4) return False;
 420
 421   fourcc = *(unsigned*)(&sdAtom[4]); // put in host order
 422
 423   char const* word7Ptr = &sdAtom[6*4];
 424   numChannels = (word7Ptr[0]<<8)|(word7Ptr[1]);
 425   return True;
 426 }