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_FFMPEG
  31 AVCodecParserContext * h264parserctx;
  32 AVCodecContext *avcctx;
  33 #endif
  34
  35 // Copied from vlc
  36 static unsigned char* parseH264ConfigStr( char const* configStr,
  37                                           unsigned int& configSize )
  38 {
  39
  40     char *dup, *psz;
  41     int i, i_records = 1;
  42
  43     if( configSize )
  44     configSize = 0;
  45     if( configStr == NULL || *configStr == '\0' )
  46         return NULL;
  47     psz = dup = strdup( configStr );
  48
  49  /* Count the number of comma's */
  50     for( psz = dup; *psz != '\0'; ++psz )
  51     {
  52         if( *psz == ',')
  53         {
  54             ++i_records;
  55             *psz = '\0';
  56         }
  57     }
  58
  59     unsigned char *cfg = new unsigned char[5 * strlen(dup)];
  60     psz = dup;
  61     for( i = 0; i < i_records; i++ )
  62     {
  63
  64         cfg[configSize++] = 0x00;
  65         cfg[configSize++] = 0x00;
  66         cfg[configSize++] = 0x01;
  67         configSize += av_base64_decode( (uint8_t*)&cfg[configSize],
  68                                         psz,
  69                                         5 * strlen(dup) - 3 );
  70
  71     psz += strlen(psz)+1;
  72     }
  73     if( dup ) free( dup );
  74
  75     return cfg;
  76 }
  77
  78 static void
  79 needVideoFrameRate(demuxer_t* demuxer, MediaSubsession* subsession); // forward
  80 static Boolean
  81 parseQTState_video(QuickTimeGenericRTPSource::QTState const& qtState,
  82                    unsigned& fourcc); // forward
  83 static Boolean
  84 parseQTState_audio(QuickTimeGenericRTPSource::QTState const& qtState,
  85                    unsigned& fourcc, unsigned& numChannels); // forward
  86
  87 static BITMAPINFOHEADER * insertVideoExtradata(BITMAPINFOHEADER *bih,
  88                                                unsigned char * extraData,
  89                                                unsigned size)
  90 {
  91     BITMAPINFOHEADER * original = bih;
  92     if (!size || size > INT_MAX - sizeof(BITMAPINFOHEADER))
  93         return bih;
  94     bih = (BITMAPINFOHEADER*)realloc(bih, sizeof(BITMAPINFOHEADER) + size);
  95     if (!bih)
  96         return original;
  97     bih->biSize = sizeof(BITMAPINFOHEADER) + size;
  98     memcpy(bih+1, extraData, size);
  99     return bih;
 100 }
 101
 102 void rtpCodecInitialize_video(demuxer_t* demuxer,
 103                               MediaSubsession* subsession,
 104                               unsigned& flags) {
 105   flags = 0;
 106   // Create a dummy video stream header
 107   // to make the main MPlayer code happy:
 108   sh_video_t* sh_video = new_sh_video(demuxer,0);
 109   BITMAPINFOHEADER* bih
 110     = (BITMAPINFOHEADER*)calloc(1,sizeof(BITMAPINFOHEADER));
 111   bih->biSize = sizeof(BITMAPINFOHEADER);
 112   sh_video->bih = bih;
 113   demux_stream_t* d_video = demuxer->video;
 114   d_video->sh = sh_video; sh_video->ds = d_video;
 115
 116   // Map known video MIME types to the BITMAPINFOHEADER parameters
 117   // that this program uses.  (Note that not all types need all
 118   // of the parameters to be set.)
 119   if (strcmp(subsession->codecName(), "MPV") == 0) {
 120     flags |= RTPSTATE_IS_MPEG12_VIDEO;
 121   } else if (strcmp(subsession->codecName(), "MP1S") == 0 ||
 122              strcmp(subsession->codecName(), "MP2T") == 0) {
 123     flags |= RTPSTATE_IS_MPEG12_VIDEO|RTPSTATE_IS_MULTIPLEXED;
 124   } else if (strcmp(subsession->codecName(), "H263") == 0 ||
 125              strcmp(subsession->codecName(), "H263-2000") == 0 ||
 126              strcmp(subsession->codecName(), "H263-1998") == 0) {
 127     bih->biCompression = sh_video->format
 128       = mmioFOURCC('H','2','6','3');
 129     needVideoFrameRate(demuxer, subsession);
 130   } else if (strcmp(subsession->codecName(), "H264") == 0) {
 131     bih->biCompression = sh_video->format
 132       = mmioFOURCC('H','2','6','4');
 133     unsigned int configLen = 0;
 134     unsigned char* configData
 135       = parseH264ConfigStr(subsession->fmtp_spropparametersets(), configLen);
 136     sh_video->bih = bih = insertVideoExtradata(bih, configData, configLen);
 137 #ifdef CONFIG_FFMPEG
 138     int fooLen;
 139     const uint8_t* fooData;
 140     avcodec_register_all();
 141     h264parserctx = av_parser_init(CODEC_ID_H264);
 142     avcctx = avcodec_alloc_context();
 143     // Pass the config to the parser
 144     h264parserctx->parser->parser_parse(h264parserctx, avcctx,
 145                   &fooData, &fooLen, configData, configLen);
 146 #endif
 147     delete[] configData;
 148     needVideoFrameRate(demuxer, subsession);
 149   } else if (strcmp(subsession->codecName(), "H261") == 0) {
 150     bih->biCompression = sh_video->format
 151       = mmioFOURCC('H','2','6','1');
 152     needVideoFrameRate(demuxer, subsession);
 153   } else if (strcmp(subsession->codecName(), "JPEG") == 0) {
 154     bih->biCompression = sh_video->format
 155       = mmioFOURCC('M','J','P','G');
 156     needVideoFrameRate(demuxer, subsession);
 157   } else if (strcmp(subsession->codecName(), "MP4V-ES") == 0) {
 158     bih->biCompression = sh_video->format
 159       = mmioFOURCC('m','p','4','v');
 160     // For the codec to work correctly, it may need a 'VOL Header' to be
 161     // inserted at the front of the data stream.  Construct this from the
 162     // "config" MIME parameter, which was present (hopefully) in the
 163     // session's SDP description:
 164     unsigned configLen;
 165     unsigned char* configData
 166       = parseGeneralConfigStr(subsession->fmtp_config(), configLen);
 167     sh_video->bih = bih = insertVideoExtradata(bih, configData, configLen);
 168     needVideoFrameRate(demuxer, subsession);
 169   } else if (strcmp(subsession->codecName(), "X-QT") == 0 ||
 170              strcmp(subsession->codecName(), "X-QUICKTIME") == 0) {
 171     // QuickTime generic RTP format, as described in
 172     // http://developer.apple.com/quicktime/icefloe/dispatch026.html
 173
 174     // We can't initialize this stream until we've received the first packet
 175     // that has QuickTime "sdAtom" information in the header.  So, keep
 176     // reading packets until we get one:
 177     unsigned char* packetData; unsigned packetDataLen; float pts;
 178     QuickTimeGenericRTPSource* qtRTPSource
 179       = (QuickTimeGenericRTPSource*)(subsession->rtpSource());
 180     unsigned fourcc;
 181     do {
 182       if (!awaitRTPPacket(demuxer, demuxer->video,
 183                           packetData, packetDataLen, pts)) {
 184         return;
 185       }
 186     } while (!parseQTState_video(qtRTPSource->qtState, fourcc));
 187
 188     bih->biCompression = sh_video->format = fourcc;
 189     bih->biWidth = qtRTPSource->qtState.width;
 190     bih->biHeight = qtRTPSource->qtState.height;
 191       if (qtRTPSource->qtState.sdAtomSize > 83)
 192         bih->biBitCount = qtRTPSource->qtState.sdAtom[83];
 193       uint8_t *pos = (uint8_t*)qtRTPSource->qtState.sdAtom + 86;
 194       uint8_t *endpos = (uint8_t*)qtRTPSource->qtState.sdAtom
 195                         + qtRTPSource->qtState.sdAtomSize;
 196       while (pos+8 < endpos) {
 197         unsigned atomLength = pos[0]<<24 | pos[1]<<16 | pos[2]<<8 | pos[3];
 198         if (atomLength == 0 || atomLength > endpos-pos) break;
 199         if (((!memcmp(pos+4, "avcC", 4) && fourcc==mmioFOURCC('a','v','c','1')) ||
 200              !memcmp(pos+4, "esds", 4) ||
 201              (!memcmp(pos+4, "SMI ", 4) && fourcc==mmioFOURCC('S','V','Q','3'))) &&
 202             atomLength > 8) {
 203           sh_video->bih = bih =
 204               insertVideoExtradata(bih, pos+8, atomLength-8);
 205           break;
 206         }
 207         pos += atomLength;
 208       }
 209     needVideoFrameRate(demuxer, subsession);
 210   } else {
 211     fprintf(stderr,
 212             "Unknown MPlayer format code for MIME type \"video/%s\"\n",
 213             subsession->codecName());
 214   }
 215 }
 216
 217 void rtpCodecInitialize_audio(demuxer_t* demuxer,
 218                               MediaSubsession* subsession,
 219                               unsigned& flags) {
 220   flags = 0;
 221   // Create a dummy audio stream header
 222   // to make the main MPlayer code happy:
 223   sh_audio_t* sh_audio = new_sh_audio(demuxer,0);
 224   WAVEFORMATEX* wf = (WAVEFORMATEX*)calloc(1,sizeof(WAVEFORMATEX));
 225   sh_audio->wf = wf;
 226   demux_stream_t* d_audio = demuxer->audio;
 227   d_audio->sh = sh_audio; sh_audio->ds = d_audio;
 228   d_audio->id = sh_audio->aid;
 229
 230   wf->nChannels = subsession->numChannels();
 231
 232   // Map known audio MIME types to the WAVEFORMATEX parameters
 233   // that this program uses.  (Note that not all types need all
 234   // of the parameters to be set.)
 235   wf->nSamplesPerSec
 236     = subsession->rtpSource()->timestampFrequency(); // by default
 237   if (strcmp(subsession->codecName(), "MPA") == 0 ||
 238       strcmp(subsession->codecName(), "MPA-ROBUST") == 0 ||
 239       strcmp(subsession->codecName(), "X-MP3-DRAFT-00") == 0) {
 240     wf->wFormatTag = sh_audio->format = 0x55;
 241     // Note: 0x55 is for layer III, but should work for I,II also
 242     wf->nSamplesPerSec = 0; // sample rate is deduced from the data
 243   } else if (strcmp(subsession->codecName(), "AC3") == 0) {
 244     wf->wFormatTag = sh_audio->format = 0x2000;
 245     wf->nSamplesPerSec = 0; // sample rate is deduced from the data
 246   } else if (strcmp(subsession->codecName(), "L16") == 0) {
 247     wf->wFormatTag = sh_audio->format = 0x736f7774; // "twos"
 248     wf->nBlockAlign = 1;
 249     wf->wBitsPerSample = 16;
 250     wf->cbSize = 0;
 251   } else if (strcmp(subsession->codecName(), "L8") == 0) {
 252     wf->wFormatTag = sh_audio->format = 0x20776172; // "raw "
 253     wf->nBlockAlign = 1;
 254     wf->wBitsPerSample = 8;
 255     wf->cbSize = 0;
 256   } else if (strcmp(subsession->codecName(), "PCMU") == 0) {
 257     wf->wFormatTag = sh_audio->format = 0x7;
 258     wf->nAvgBytesPerSec = 8000;
 259     wf->nBlockAlign = 1;
 260     wf->wBitsPerSample = 8;
 261     wf->cbSize = 0;
 262   } else if (strcmp(subsession->codecName(), "PCMA") == 0) {
 263     wf->wFormatTag = sh_audio->format = 0x6;
 264     wf->nAvgBytesPerSec = 8000;
 265     wf->nBlockAlign = 1;
 266     wf->wBitsPerSample = 8;
 267     wf->cbSize = 0;
 268   } else if (strcmp(subsession->codecName(), "AMR") == 0) {
 269     wf->wFormatTag = sh_audio->format = mmioFOURCC('s','a','m','r');
 270   } else if (strcmp(subsession->codecName(), "AMR-WB") == 0) {
 271     wf->wFormatTag = sh_audio->format = mmioFOURCC('s','a','w','b');
 272   } else if (strcmp(subsession->codecName(), "GSM") == 0) {
 273     wf->wFormatTag = sh_audio->format = mmioFOURCC('a','g','s','m');
 274     wf->nAvgBytesPerSec = 1650;
 275     wf->nBlockAlign = 33;
 276     wf->wBitsPerSample = 16;
 277     wf->cbSize = 0;
 278   } else if (strcmp(subsession->codecName(), "QCELP") == 0) {
 279     wf->wFormatTag = sh_audio->format = mmioFOURCC('Q','c','l','p');
 280     wf->nAvgBytesPerSec = 1750;
 281     wf->nBlockAlign = 35;
 282     wf->wBitsPerSample = 16;
 283     wf->cbSize = 0;
 284   } else if (strcmp(subsession->codecName(), "MP4A-LATM") == 0) {
 285     wf->wFormatTag = sh_audio->format = mmioFOURCC('m','p','4','a');
 286     // For the codec to work correctly, it needs "AudioSpecificConfig"
 287     // data, which is parsed from the "StreamMuxConfig" string that
 288     // was present (hopefully) in the SDP description:
 289     unsigned codecdata_len;
 290     sh_audio->codecdata
 291       = parseStreamMuxConfigStr(subsession->fmtp_config(),
 292                                 codecdata_len);
 293     sh_audio->codecdata_len = codecdata_len;
 294     //faad doesn't understand LATM's data length field, so omit it
 295     ((MPEG4LATMAudioRTPSource*)subsession->rtpSource())->omitLATMDataLengthField();
 296   } else if (strcmp(subsession->codecName(), "MPEG4-GENERIC") == 0) {
 297     wf->wFormatTag = sh_audio->format = mmioFOURCC('m','p','4','a');
 298     // For the codec to work correctly, it needs "AudioSpecificConfig"
 299     // data, which was present (hopefully) in the SDP description:
 300     unsigned codecdata_len;
 301     sh_audio->codecdata
 302       = parseGeneralConfigStr(subsession->fmtp_config(),
 303                               codecdata_len);
 304     sh_audio->codecdata_len = codecdata_len;
 305   } else if (strcmp(subsession->codecName(), "X-QT") == 0 ||
 306              strcmp(subsession->codecName(), "X-QUICKTIME") == 0) {
 307     // QuickTime generic RTP format, as described in
 308     // http://developer.apple.com/quicktime/icefloe/dispatch026.html
 309
 310     // We can't initialize this stream until we've received the first packet
 311     // that has QuickTime "sdAtom" information in the header.  So, keep
 312     // reading packets until we get one:
 313     unsigned char* packetData; unsigned packetDataLen; float pts;
 314     QuickTimeGenericRTPSource* qtRTPSource
 315       = (QuickTimeGenericRTPSource*)(subsession->rtpSource());
 316     unsigned fourcc, numChannels;
 317     do {
 318       if (!awaitRTPPacket(demuxer, demuxer->audio,
 319                           packetData, packetDataLen, pts)) {
 320         return;
 321       }
 322     } while (!parseQTState_audio(qtRTPSource->qtState, fourcc, numChannels));
 323
 324     wf->wFormatTag = sh_audio->format = fourcc;
 325     wf->nChannels = numChannels;
 326
 327       if (qtRTPSource->qtState.sdAtomSize > 33) {
 328         wf->wBitsPerSample = qtRTPSource->qtState.sdAtom[27];
 329         wf->nSamplesPerSec = qtRTPSource->qtState.sdAtom[32]<<8|qtRTPSource->qtState.sdAtom[33];
 330       }
 331     uint8_t *pos = (uint8_t*)qtRTPSource->qtState.sdAtom + 52;
 332     uint8_t *endpos = (uint8_t*)qtRTPSource->qtState.sdAtom
 333                       + qtRTPSource->qtState.sdAtomSize;
 334     while (pos+8 < endpos) {
 335       unsigned atomLength = pos[0]<<24 | pos[1]<<16 | pos[2]<<8 | pos[3];
 336       if (atomLength == 0 || atomLength > endpos-pos) break;
 337       if (!memcmp(pos+4, "wave", 4) && fourcc==mmioFOURCC('Q','D','M','2') &&
 338           atomLength > 8 &&
 339           atomLength <= INT_MAX) {
 340         sh_audio->codecdata = (unsigned char*) malloc(atomLength-8);
 341         if (sh_audio->codecdata) {
 342           memcpy(sh_audio->codecdata, pos+8, atomLength-8);
 343           sh_audio->codecdata_len = atomLength-8;
 344         }
 345         break;
 346       }
 347       pos += atomLength;
 348     }
 349   } else {
 350     fprintf(stderr,
 351             "Unknown MPlayer format code for MIME type \"audio/%s\"\n",
 352             subsession->codecName());
 353   }
 354 }
 355
 356 static void needVideoFrameRate(demuxer_t* demuxer,
 357                                MediaSubsession* subsession) {
 358   // For some codecs, MPlayer's decoding software can't (or refuses to :-)
 359   // figure out the frame rate by itself, so (unless the user specifies
 360   // it manually, using "-fps") we figure it out ourselves here, using the
 361   // presentation timestamps in successive packets,
 362   extern double force_fps; if (force_fps != 0.0) return; // user used "-fps"
 363
 364   demux_stream_t* d_video = demuxer->video;
 365   sh_video_t* sh_video = (sh_video_t*)(d_video->sh);
 366
 367   // If we already know the subsession's video frame rate, use it:
 368   int fps = (int)(subsession->videoFPS());
 369   if (fps != 0) {
 370     sh_video->fps = fps;
 371     sh_video->frametime = 1.0f/fps;
 372     return;
 373   }
 374
 375   // Keep looking at incoming frames until we see two with different,
 376   // non-zero "pts" timestamps:
 377   unsigned char* packetData; unsigned packetDataLen;
 378   float lastPTS = 0.0, curPTS;
 379   unsigned const maxNumFramesToWaitFor = 300;
 380   int lastfps = 0;
 381   for (unsigned i = 0; i < maxNumFramesToWaitFor; ++i) {
 382     if (!awaitRTPPacket(demuxer, d_video, packetData, packetDataLen, curPTS)) {
 383       break;
 384     }
 385
 386     if (curPTS != lastPTS && lastPTS != 0.0) {
 387       // Use the difference between these two "pts"s to guess the frame rate.
 388       // (should really check that there were no missing frames inbetween)#####
 389       // Guess the frame rate as an integer.  If it's not, use "-fps" instead.
 390       fps = (int)(1/fabs(curPTS-lastPTS) + 0.5); // rounding
 391         if (fps == lastfps) {
 392       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);
 393       sh_video->fps = fps;
 394       sh_video->frametime=1.0f/fps;
 395       return;
 396         }
 397       if (fps>lastfps) lastfps = fps;
 398     }
 399     lastPTS = curPTS;
 400   }
 401   fprintf(stderr, "demux_rtp: Failed to guess the video frame rate\n");
 402 }
 403
 404 static Boolean
 405 parseQTState_video(QuickTimeGenericRTPSource::QTState const& qtState,
 406                    unsigned& fourcc) {
 407   // qtState's "sdAtom" field is supposed to contain a QuickTime video
 408   // 'sample description' atom.  This atom's name is the 'fourcc' that we want:
 409   char const* sdAtom = qtState.sdAtom;
 410   if (sdAtom == NULL || qtState.sdAtomSize < 2*4) return False;
 411
 412   fourcc = *(unsigned*)(&sdAtom[4]); // put in host order
 413   return True;
 414 }
 415
 416 static Boolean
 417 parseQTState_audio(QuickTimeGenericRTPSource::QTState const& qtState,
 418                    unsigned& fourcc, unsigned& numChannels) {
 419   // qtState's "sdAtom" field is supposed to contain a QuickTime audio
 420   // 'sample description' atom.  This atom's name is the 'fourcc' that we want.
 421   // Also, the top half of the 5th word following the atom name should
 422   // contain the number of channels ("numChannels") that we want:
 423   char const* sdAtom = qtState.sdAtom;
 424   if (sdAtom == NULL || qtState.sdAtomSize < 7*4) return False;
 425
 426   fourcc = *(unsigned*)(&sdAtom[4]); // put in host order
 427
 428   char const* word7Ptr = &sdAtom[6*4];
 429   numChannels = (word7Ptr[0]<<8)|(word7Ptr[1]);
 430   return True;
 431 }