src/matroska.c

   1 /*****************************************************************************
   2  * matroska.c:
   3  *****************************************************************************
   4  * Copyright (C) 2005 x264 project
   5  * $Id: $
   6  *
   7  * Authors: Mike Matsnev
   8  *          Nathan Caldwell
   9  *
  10  * This program is free software; you can redistribute it and/or modify
  11  * it under the terms of the GNU General Public License as published by
  12  * the Free Software Foundation; either version 2 of the License, or
  13  * (at your option) any later version.
  14  *
  15  * This program is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  18  * GNU General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU General Public License
  21  * along with this program; if not, write to the Free Software
  22  * Foundation, Inc., 59 Temple Place - Suite 330, Boston, MA  02111, USA.
  23  *****************************************************************************/
  24 #include "config.h"
  25 #include "libmkv.h"
  26 #include "matroska.h"
  27 #include "md5.h"
  28
  29 int mk_seekFile(mk_Writer *w, uint64_t pos) {
  30   if (fseek(w->fp, pos, SEEK_SET))
  31     return -1;
  32
  33   w->f_pos = pos;
  34
  35   if (pos > w->f_eof)
  36     w->f_eof = pos;
  37
  38   return 0;
  39 }
  40
  41 char  *mk_laceXiph(uint64_t *sizes, uint8_t num_frames, uint64_t *output_size) {
  42   unsigned i, j;
  43   uint64_t offset = 0;
  44   uint64_t alloc_size = num_frames * 6;  // Complete guess. We'll realloc if we need more space, though.
  45   char *laced = calloc(alloc_size, sizeof(char));
  46   if (laced == NULL)
  47     return NULL;
  48
  49   laced[offset++] = num_frames;
  50   for (i = 0; i < num_frames; i++)
  51   {
  52     for (j = sizes[i]; j >= 255 ; j -= 255)
  53     {
  54       laced[offset++] = 255;
  55       if (offset + 1 >= alloc_size) {
  56         int avg_sz = offset / (i - 1);  // Compute approximate average bytes/frame
  57         alloc_size += avg_sz * (num_frames - i);  // Add our average + number of frames left to size
  58         if ((laced = realloc(laced, alloc_size)) == NULL)
  59           return NULL;
  60       }
  61     }
  62     laced[offset++] = j;
  63   }
  64
  65   if (output_size != NULL)
  66     *output_size = offset;
  67
  68   return laced;
  69 }
  70
  71 mk_Writer *mk_createWriter(const char *filename, int64_t timescale, uint8_t vlc_compat) {
  72   mk_Writer *w = calloc(1, sizeof(*w));
  73   if (w == NULL)
  74     return NULL;
  75
  76   w->root = mk_createContext(w, NULL, 0);
  77   if (w->root == NULL) {
  78     free(w);
  79     return NULL;
  80   }
  81
  82   if ((w->cues = mk_createContext(w, w->root, MATROSKA_ID_CUES)) == NULL) // Cues
  83   {
  84     mk_destroyContexts(w);
  85     free(w);
  86     return NULL;
  87   }
  88
  89   if (vlc_compat) {
  90     if ((w->cluster.seekhead = mk_createContext(w, w->root, MATROSKA_ID_SEEKHEAD)) == NULL) // SeekHead
  91     {
  92       mk_destroyContexts(w);
  93       free(w);
  94       return NULL;
  95     }
  96   }
  97
  98   w->fp = fopen(filename, "wb");
  99   if (w->fp == NULL) {
 100     mk_destroyContexts(w);
 101     free(w);
 102     return NULL;
 103   }
 104
 105   w->timescale = timescale;
 106   w->vlc_compat = vlc_compat;
 107
 108   return w;
 109 }
 110
 111 int   mk_writeHeader(mk_Writer *w, const char *writingApp) {
 112   mk_Context  *c;
 113   int64_t offset = 0;
 114
 115   if (w->wrote_header)
 116     return -1;
 117
 118   md5_starts(&w->segment_md5);  /* Initalize MD5 */
 119
 120   CHECK(mk_writeEbmlHeader(w, "matroska", MATROSKA_VERSION, MATROSKA_VERSION));
 121
 122   if ((c = mk_createContext(w, w->root, MATROSKA_ID_SEGMENT)) == NULL) // Segment
 123     return -1;
 124   CHECK(mk_flushContextID(c));
 125   w->segment_ptr = c->d_cur;
 126   CHECK(mk_closeContext(c, &w->segment_ptr));
 127
 128   if (w->vlc_compat) {
 129     CHECK(mk_writeVoid(w->root, 0x100));  // 256 bytes should be enough room for our Seek entries.
 130     CHECK(mk_writeVoid(w->root, 0x800)); // 2048 bytes for Chapters.
 131   }
 132   else {
 133     w->seek_data.seekhead = 0x80000000;
 134     CHECK(mk_writeSeekHead(w, &w->seekhead_ptr));
 135     w->seek_data.seekhead = 0;
 136   }
 137
 138   if ((c = mk_createContext(w, w->root, MATROSKA_ID_INFO)) == NULL) // SegmentInfo
 139     return -1;
 140   w->seek_data.segmentinfo = w->root->d_cur - w->segment_ptr;
 141   CHECK(mk_writeVoid(c, 16));   /* Reserve space for a SegmentUID, we'll write the it later. */
 142   CHECK(mk_writeStr(c, MATROSKA_ID_MUXINGAPP, PACKAGE_STRING)); // MuxingApp
 143   CHECK(mk_writeStr(c, MATROSKA_ID_WRITINGAPP, writingApp)); // WritingApp
 144   CHECK(mk_writeUInt(c, MATROSKA_ID_TIMECODESCALE, w->timescale)); // TimecodeScale
 145   CHECK(mk_writeFloat(c, MATROSKA_ID_DURATION, 0)); // Duration
 146   w->duration_ptr = c->d_cur - 4;
 147   CHECK(mk_closeContext(c, &offset));
 148   w->duration_ptr += offset;
 149   w->segmentuid_ptr = offset;
 150
 151   w->seek_data.tracks = w->root->d_cur - w->segment_ptr;
 152
 153   if (w->tracks)
 154     CHECK(mk_closeContext(w->tracks, 0));
 155
 156   CHECK(mk_flushContextData(w->root));
 157
 158   w->wrote_header = 1;
 159   w->def_duration = w->tracks_arr[0]->default_duration;
 160   return 0;
 161 }
 162
 163 static int mk_closeCluster(mk_Writer *w) {
 164   if (w->cluster.context == NULL)
 165     return 0;
 166   w->cluster.count++;
 167   CHECK(mk_closeContext(w->cluster.context, 0));
 168   w->cluster.context = NULL;
 169   CHECK(mk_flushContextData(w->root));
 170   return 0;
 171 }
 172
 173 int   mk_flushFrame(mk_Writer *w, mk_Track *track) {
 174   mk_Context *c, *tp;
 175   int64_t   delta, ref = 0;
 176   unsigned  fsize, bgsize;
 177   uint8_t   flags, c_delta[2];
 178   int i;
 179   char *laced = NULL;
 180   uint64_t  length = 0;
 181
 182   if (!track->in_frame)
 183     return 0;
 184
 185   delta = track->frame.timecode/w->timescale - w->cluster.tc_scaled;
 186   if (delta > 2000ll || delta < -2000ll)
 187     CHECK(mk_closeCluster(w));
 188
 189   if (w->cluster.context == NULL) {
 190     w->cluster.tc_scaled = track->frame.timecode / w->timescale;
 191     w->cluster.context = mk_createContext(w, w->root, MATROSKA_ID_CLUSTER); // Cluster
 192     if (w->cluster.context == NULL)
 193       return -1;
 194
 195     w->cluster.pointer = w->f_pos - w->segment_ptr;
 196
 197         /*
 198          * Dirty HACK: This is for HandBrake/VLC. The first timecode we get isnt't
 199          * always 0, if not we set it to 0.
 200          */
 201         if (w->cluster.count == 0 && w->vlc_compat && track->frame.timecode != 0) {
 202                 w->cluster.tc_scaled = 0;
 203                 delta = track->frame.timecode/w->timescale;
 204         } else {
 205                 delta = 0;
 206         }
 207
 208     if (w->vlc_compat)
 209       CHECK(mk_writeSeek(w, w->cluster.seekhead, MATROSKA_ID_CLUSTER, w->cluster.pointer));
 210
 211     CHECK(mk_writeUInt(w->cluster.context, MATROSKA_ID_CLUSTERTIMECODE, w->cluster.tc_scaled)); // Cluster Timecode
 212
 213     w->cluster.block_count = 0;
 214   }
 215
 216   /* Calculate the encoded lacing sizes. */
 217   switch (track->frame.lacing) {
 218           case MK_LACING_XIPH:
 219                   laced = mk_laceXiph(track->frame.lacing_sizes, track->frame.lacing_num_frames, &length);
 220                   break;
 221           case MK_LACING_EBML:
 222           {
 223                   uint64_t u_size = 0;
 224                   length += mk_ebmlSizeSize(track->frame.lacing_sizes[0]) + 1;  // Add one for the frame count.
 225                   for (i = 1; i < track->frame.lacing_num_frames; i++)
 226                   {
 227                           u_size = llabs(track->frame.lacing_sizes[i] - track->frame.lacing_sizes[i-1]);
 228                           length += mk_ebmlSizeSize((u_size) << 1);             // Shift by one so we get the right size for a signed number.
 229                   }
 230                   break;
 231           }
 232           case MK_LACING_FIXED:
 233           {
 234                   laced = calloc(1, sizeof(char));
 235                   laced[0] = track->frame.lacing_num_frames;
 236                   ++length;
 237                   break;
 238           }
 239   }
 240
 241   fsize = track->frame.data ? track->frame.data->d_cur : 0;
 242   bgsize = fsize + 4 + mk_ebmlSizeSize(fsize + 4 + length) + 1 + length;
 243   if (!track->frame.keyframe) {
 244     ref = track->prev_frame_tc_scaled - w->cluster.tc_scaled - delta;
 245     bgsize += 1 + 1 + mk_ebmlSIntSize(ref);
 246   }
 247
 248   CHECK(mk_writeID(w->cluster.context, MATROSKA_ID_BLOCKGROUP)); // BlockGroup
 249   CHECK(mk_writeSize(w->cluster.context, bgsize));
 250   CHECK(mk_writeID(w->cluster.context, MATROSKA_ID_BLOCK)); // Block
 251   CHECK(mk_writeSize(w->cluster.context, fsize + 4 + length));  // Block size
 252   CHECK(mk_writeSize(w->cluster.context, track->track_id)); // track number
 253
 254   w->cluster.block_count++;
 255
 256   c_delta[0] = delta >> 8;
 257   c_delta[1] = delta;
 258   CHECK(mk_appendContextData(w->cluster.context, c_delta, 2));  // Timecode relative to Cluster.
 259
 260 //   flags = ( track->frame.keyframe << 8 ) | track->frame.lacing;
 261   flags = track->frame.lacing << 1;     // Flags: Bit 5-6 describe what type of lacing to use.
 262   CHECK(mk_appendContextData(w->cluster.context, &flags, 1));
 263   if (track->frame.lacing) {
 264     if (track->frame.lacing == MK_LACING_EBML) {
 265       CHECK(mk_appendContextData(w->cluster.context, &track->frame.lacing_num_frames, 1));      // Number of frames in lace-1
 266       CHECK(mk_writeSize(w->cluster.context, track->frame.lacing_sizes[0]));    // Size of 1st frame.
 267       for (i = 1; i < track->frame.lacing_num_frames; i++)
 268       {
 269         CHECK(mk_writeSSize(w->cluster.context, track->frame.lacing_sizes[i] - track->frame.lacing_sizes[i-1]));        // Size difference between previous size and this size.
 270       }
 271     } else if (length > 0 && laced != NULL) {
 272       CHECK(mk_appendContextData(w->cluster.context, laced, length));
 273       free(laced);
 274       laced = NULL;
 275     }
 276   }
 277
 278   if (track->frame.data) {
 279     CHECK(mk_appendContextData(w->cluster.context, track->frame.data->data, track->frame.data->d_cur));
 280     track->frame.data->d_cur = 0;
 281   }
 282   if (!track->frame.keyframe)
 283     CHECK(mk_writeSInt(w->cluster.context, MATROSKA_ID_REFERENCEBLOCK, ref)); // ReferenceBlock
 284
 285   /* This may get a little out of hand, but it seems sane enough for now. */
 286   if (track->frame.keyframe && (track->track_type == MK_TRACK_VIDEO)) {
 287 //   if (track->frame.keyframe && (track->track_type & MK_TRACK_VIDEO) && ((track->prev_cue_pos + 3*CLSIZE) <= w->f_pos || track->frame.timecode == 0)) {
 288     if ((c = mk_createContext(w, w->cues, MATROSKA_ID_CUEPOINT)) == NULL)  // CuePoint
 289       return -1;
 290     CHECK(mk_writeUInt(c, MATROSKA_ID_CUETIME, (track->frame.timecode / w->timescale))); // CueTime
 291
 292     if ((tp = mk_createContext(w, c, MATROSKA_ID_CUETRACKPOSITIONS)) == NULL)  // CueTrackPositions
 293       return -1;
 294     CHECK(mk_writeUInt(tp, MATROSKA_ID_CUETRACK, track->track_id)); // CueTrack
 295     CHECK(mk_writeUInt(tp, MATROSKA_ID_CUECLUSTERPOSITION, w->cluster.pointer));  // CueClusterPosition
 296 //     CHECK(mk_writeUInt(c, MATROSKA_ID_CUEBLOCKNUMBER, w->cluster.block_count));  // CueBlockNumber
 297     CHECK(mk_closeContext(tp, 0));
 298     CHECK(mk_closeContext(c, 0));
 299     track->prev_cue_pos = w->f_pos;
 300   }
 301
 302   track->in_frame = 0;
 303   track->prev_frame_tc_scaled = w->cluster.tc_scaled + delta;
 304
 305   if (w->cluster.context->d_cur > CLSIZE)
 306     CHECK(mk_closeCluster(w));
 307
 308   return 0;
 309 }
 310
 311 int   mk_startFrame(mk_Writer *w, mk_Track *track) {
 312   if (mk_flushFrame(w, track) < 0)
 313     return -1;
 314
 315   track->in_frame = 1;
 316   track->frame.keyframe = 0;
 317   track->frame.lacing = MK_LACING_NONE;
 318   track->frame.lacing_num_frames = 0;
 319   track->frame.lacing_sizes = NULL;
 320
 321   return 0;
 322 }
 323
 324 int   mk_setFrameFlags(mk_Writer *w, mk_Track *track, int64_t timestamp, unsigned keyframe) {
 325   if (!track->in_frame)
 326     return -1;
 327
 328   track->frame.timecode = timestamp;
 329   track->frame.keyframe = keyframe != 0;
 330
 331   if (track->max_frame_tc < timestamp)
 332     track->max_frame_tc = timestamp;
 333
 334   return 0;
 335 }
 336
 337 int   mk_setFrameLacing(mk_Writer *w, mk_Track *track, uint8_t lacing, uint8_t num_frames, uint64_t sizes[]) {
 338   if (!track->in_frame)
 339     return -1;
 340   track->frame.lacing_sizes = calloc(num_frames, sizeof(uint64_t));
 341
 342   track->frame.lacing = lacing;
 343   track->frame.lacing_num_frames = num_frames;
 344   memcpy(track->frame.lacing_sizes, sizes, num_frames * sizeof(uint64_t));
 345
 346   return 0;
 347 }
 348
 349 int   mk_addFrameData(mk_Writer *w, mk_Track *track, const void *data, unsigned size) {
 350   if (!track->in_frame)
 351     return -1;
 352
 353   if (track->frame.data == NULL)
 354     if ((track->frame.data = mk_createContext(w, NULL, 0)) == NULL)
 355       return -1;
 356
 357   md5_update(&w->segment_md5, (unsigned char *)data, size);
 358
 359   return mk_appendContextData(track->frame.data, data, size);
 360 }
 361
 362 int   mk_writeSeek(mk_Writer *w, mk_Context *c, unsigned seek_id, uint64_t seek_pos) {
 363   mk_Context  *s;
 364
 365   if ((s = mk_createContext(w, c, MATROSKA_ID_SEEKENTRY)) == NULL) // Seek
 366     return -1;
 367   CHECK(mk_writeUInt(s, MATROSKA_ID_SEEKID, seek_id));  // SeekID
 368   CHECK(mk_writeUInt(s, MATROSKA_ID_SEEKPOSITION, seek_pos)); // SeekPosition
 369   CHECK(mk_closeContext(s, 0));
 370
 371   return 0;
 372 }
 373
 374 /* The offset of the SeekHead is returned in *pointer. */
 375 int mk_writeSeekHead(mk_Writer *w, int64_t *pointer) {
 376   mk_Context  *c;
 377   int64_t   seekhead_ptr;
 378
 379   if ((c = mk_createContext(w, w->root, MATROSKA_ID_SEEKHEAD)) == NULL) // SeekHead
 380     return -1;
 381   if (pointer != NULL)
 382     seekhead_ptr = w->f_pos;
 383   if (w->seek_data.seekhead)
 384     CHECK(mk_writeSeek(w, c, MATROSKA_ID_SEEKHEAD, w->seek_data.seekhead));
 385   if (w->seek_data.segmentinfo)
 386     CHECK(mk_writeSeek(w, c, MATROSKA_ID_INFO, w->seek_data.segmentinfo));
 387   if (w->seek_data.tracks)
 388     CHECK(mk_writeSeek(w, c, MATROSKA_ID_TRACKS, w->seek_data.tracks));
 389   if (w->seek_data.cues)
 390     CHECK(mk_writeSeek(w, c, MATROSKA_ID_CUES, w->seek_data.cues));
 391   if (w->seek_data.attachments)
 392     CHECK(mk_writeSeek(w, c, MATROSKA_ID_ATTACHMENTS, w->seek_data.attachments));
 393   if (w->seek_data.chapters)
 394     CHECK(mk_writeSeek(w, c, MATROSKA_ID_CHAPTERS, w->seek_data.chapters));
 395   if (w->seek_data.tags)
 396     CHECK(mk_writeSeek(w, c, MATROSKA_ID_TAGS, w->seek_data.tags));
 397   CHECK(mk_closeContext(c, 0));
 398
 399   if (pointer != NULL)
 400     *pointer = seekhead_ptr;
 401
 402   return 0;
 403 }
 404
 405 int   mk_close(mk_Writer *w) {
 406   int   i, ret = 0;
 407   mk_Track *tk;
 408   int64_t max_frame_tc = w->tracks_arr[0]->max_frame_tc;
 409   uint64_t segment_size = 0;
 410   unsigned char c_size[8];
 411   unsigned char segment_uid[16];
 412
 413   md5_finish(&w->segment_md5, segment_uid);
 414
 415   for (i = w->num_tracks - 1; i >= 0; i--)
 416   {
 417     tk = w->tracks_arr[i];
 418     w->tracks_arr[i] = NULL;
 419     --w->num_tracks;
 420     if (mk_flushFrame(w, tk) < 0)
 421       ret = -1;
 422     free(tk);
 423     tk = NULL;
 424   }
 425
 426   if (mk_closeCluster(w) < 0)
 427     ret = -1;
 428
 429   w->seek_data.cues = w->f_pos - w->segment_ptr;
 430   if (mk_closeContext(w->cues, 0) < 0)
 431     ret = -1;
 432   if (mk_flushContextData(w->root) < 0)
 433     ret = -1;
 434
 435   if (w->vlc_compat && w->cluster.seekhead) {
 436     w->seek_data.seekhead = w->f_pos - w->segment_ptr;
 437     if (mk_closeContext(w->cluster.seekhead, 0) < 0)
 438       ret = -1;
 439     if (mk_flushContextData(w->root) < 0)
 440       ret = -1;
 441   }
 442
 443   if (w->chapters != NULL)
 444   {
 445     if (w->vlc_compat) {
 446       if (mk_flushContextData(w->root) < 0)
 447         ret = -1;
 448       if (mk_seekFile(w, w->segment_ptr + 0x100 + 3) < 0)
 449         ret = -1;
 450     }
 451     w->seek_data.chapters = w->f_pos - w->segment_ptr;
 452     mk_writeChapters(w);
 453     if (mk_flushContextData(w->root) < 0)
 454       ret = -1;
 455     if (w->vlc_compat) {
 456       if (mk_writeVoid(w->root, (0x800 - (w->f_pos - w->segment_ptr - 0x100 - 3))) < 0)
 457         ret = -1;
 458       if (mk_flushContextData(w->root) < 0)
 459         ret = -1;
 460     }
 461   }
 462
 463   if (w->wrote_header) {
 464     if (w->vlc_compat) {
 465       if (mk_seekFile(w, w->segment_ptr) < 0)
 466         ret = -1;
 467     }
 468
 469     if (mk_writeSeekHead(w, &w->seek_data.seekhead) < 0)
 470       ret = -1;
 471     w->seek_data.seekhead -= w->segment_ptr;
 472
 473     if (w->vlc_compat)
 474     {
 475       if (mk_flushContextData(w->root) < 0)
 476         ret = -1;
 477       if (mk_writeVoid(w->root, (0x100 - (w->f_pos - w->segment_ptr))) < 0)
 478         ret = -1;
 479     }
 480
 481     if (mk_flushContextData(w->root) < 0)
 482       ret = -1;
 483
 484     if (!w->vlc_compat)
 485     {
 486       int i = w->seek_data.segmentinfo;
 487       w->seek_data.segmentinfo = 0;
 488       w->seek_data.tracks = 0;
 489       w->seek_data.cues = 0;
 490       w->seek_data.chapters = 0;
 491       w->seek_data.attachments = 0;
 492       w->seek_data.tags = 0;
 493       if (mk_seekFile(w, w->segment_ptr) < 0)
 494         ret = -1;
 495       if (mk_writeSeekHead(w, NULL) < 0 ||
 496           mk_flushContextData(w->root) < 0)
 497         ret = -1;
 498       if (((i + w->segment_ptr) - w->f_pos - 2) > 1)
 499         if (mk_writeVoid(w->root, (i + w->segment_ptr) - w->f_pos - 2) < 0 ||
 500             mk_flushContextData(w->root) < 0)
 501           ret = -1;
 502     }
 503
 504     if (mk_seekFile(w, w->duration_ptr) < 0)
 505       ret = -1;
 506     if (mk_writeFloatRaw(w->root, (float)((double)(max_frame_tc+w->def_duration) / w->timescale)) < 0 ||
 507         mk_flushContextData(w->root) < 0)
 508       ret = -1;
 509     if (mk_seekFile(w, w->segment_ptr - 8) < 0)
 510       ret = -1;
 511     segment_size = w->f_eof - w->segment_ptr;
 512     for (i = 7; i > 0; --i)
 513       c_size[i] = segment_size >> (8 * (7-i));
 514     c_size[i] = 0x01;
 515     if (mk_appendContextData(w->root, &c_size, 8) < 0 ||
 516         mk_flushContextData(w->root) < 0)
 517       ret = -1;
 518         if (mk_seekFile(w, w->segmentuid_ptr) < 0)
 519                 ret = -1;
 520     if (mk_writeBin(w->root, MATROSKA_ID_SEGMENTUID, segment_uid, sizeof(segment_uid)) < 0 ||
 521                 mk_flushContextData(w->root) < 0)
 522           ret = -1;
 523   }
 524
 525   if (mk_closeContext(w->root, 0) < 0)
 526     ret = -1;
 527   mk_destroyContexts(w);
 528   fclose(w->fp);
 529   free(w->tracks_arr);
 530   free(w);
 531
 532   return ret;
 533 }