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