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 "libmkv.h"
  25 #include "matroska.h"
  26 #include "config.h"
  27
  28 int mk_seekFile(mk_Writer *w, uint64_t pos) {
  29   if (fseek(w->fp, pos, SEEK_SET))
  30     return -1;
  31
  32   w->f_pos = pos;
  33
  34   if (pos > w->f_eof)
  35     w->f_eof = pos;
  36
  37   return 0;
  38 }
  39
  40 int    mk_writeVoid(mk_Context *c, uint64_t length) {
  41   char *c_void = calloc(length, sizeof(char));
  42
  43   CHECK(mk_writeID(c, 0xec));
  44   CHECK(mk_writeSize(c, length));
  45   CHECK(mk_appendContextData(c, c_void, length));
  46   free(c_void);
  47   return 0;
  48 }
  49
  50 char  *mk_laceXiph(uint64_t *sizes, uint8_t num_frames, uint64_t *output_size) {
  51   unsigned i, j;
  52   uint64_t offset = 0;
  53   uint64_t alloc_size = num_frames * 6;  // Complete guess. We'll realloc if we need more space, though.
  54   char *laced = calloc(alloc_size, sizeof(char));
  55   if (laced == NULL)
  56     return NULL;
  57
  58   laced[offset++] = num_frames;
  59   for (i = 0; i < num_frames; i++)
  60   {
  61     for (j = sizes[i]; j >= 255 ; j -= 255)
  62     {
  63       laced[offset++] = 255;
  64       if (offset + 1 >= alloc_size) {
  65         int avg_sz = offset / (i - 1);  // Compute approximate average bytes/frame
  66         alloc_size += avg_sz * (num_frames - i);  // Add our average + number of frames left to size
  67         if ((laced = realloc(laced, alloc_size)) == NULL)
  68           return NULL;
  69       }
  70     }
  71     laced[offset++] = j;
  72   }
  73
  74   if (output_size != NULL)
  75     *output_size = offset;
  76
  77   return laced;
  78 }
  79
  80 mk_Writer *mk_createWriter(const char *filename, int64_t timescale, uint8_t vlc_compat) {
  81   mk_Writer *w = calloc(1, sizeof(*w));
  82   if (w == NULL)
  83     return NULL;
  84
  85   w->root = mk_createContext(w, NULL, 0);
  86   if (w->root == NULL) {
  87     free(w);
  88     return NULL;
  89   }
  90
  91   if ((w->cues = mk_createContext(w, w->root, 0x1c53bb6b)) == NULL) // Cues
  92   {
  93     mk_destroyContexts(w);
  94     free(w);
  95     return NULL;
  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
 114   if (w->wrote_header)
 115     return -1;
 116
 117   if ((c = mk_createContext(w, w->root, 0x1a45dfa3)) == NULL) // EBML
 118     return -1;
 119   CHECK(mk_writeUInt(c, 0x4286, 1)); // EBMLVersion
 120   CHECK(mk_writeUInt(c, 0x42f7, 1)); // EBMLReadVersion
 121   CHECK(mk_writeUInt(c, 0x42f2, 4)); // EBMLMaxIDLength
 122   CHECK(mk_writeUInt(c, 0x42f3, 8)); // EBMLMaxSizeLength
 123   CHECK(mk_writeStr(c, 0x4282, "matroska")); // DocType
 124   CHECK(mk_writeUInt(c, 0x4287, 1)); // DocTypeVersion
 125   CHECK(mk_writeUInt(c, 0x4285, 1)); // DocTypeReadversion
 126   CHECK(mk_closeContext(c, 0));
 127
 128   if ((c = mk_createContext(w, w->root, 0x18538067)) == NULL) // Segment
 129     return -1;
 130   CHECK(mk_flushContextID(c));
 131   w->segment_ptr = c->d_cur;
 132   CHECK(mk_closeContext(c, &w->segment_ptr));
 133
 134   if (w->vlc_compat)
 135   {
 136     CHECK(mk_writeVoid(w->root, 0x100));  // 256 bytes should be enough room for our Seek entries.
 137     CHECK(mk_writeVoid(w->root, 0x800)); // 2048 bytes for Chapters.
 138     CHECK(mk_writeVoid(w->root, 0x1000)); // 4096 bytes for Cues.
 139   } else
 140   {
 141     w->seek_data.seekhead = 0x80000000;
 142     CHECK(mk_writeSeek(w, &w->seekhead_ptr));
 143     w->seek_data.seekhead = 0;
 144   }
 145
 146   if ((c = mk_createContext(w, w->root, 0x1549a966)) == NULL) // SegmentInfo
 147     return -1;
 148   w->seek_data.segmentinfo = w->root->d_cur - w->segment_ptr;
 149   CHECK(mk_writeStr(c, 0x4d80, PACKAGE_STRING)); // MuxingApp
 150   CHECK(mk_writeStr(c, 0x5741, writingApp)); // WritingApp
 151   CHECK(mk_writeUInt(c, 0x2ad7b1, w->timescale)); // TimecodeScale
 152   CHECK(mk_writeFloat(c, 0x4489, 0)); // Duration
 153   w->duration_ptr = c->d_cur - 4;
 154   CHECK(mk_closeContext(c, &w->duration_ptr));
 155
 156   w->seek_data.tracks = w->root->d_cur - w->segment_ptr;
 157
 158   if (w->tracks) {
 159     CHECK(mk_closeContext(w->tracks, 0));
 160   }
 161
 162   CHECK(mk_flushContextData(w->root));
 163
 164   w->wrote_header = 1;
 165   w->def_duration = w->tracks_arr[0]->default_duration;
 166   return 0;
 167 }
 168
 169 static int mk_closeCluster(mk_Writer *w) {
 170   if (w->cluster.context == NULL)
 171     return 0;
 172   w->cluster.count++;
 173   CHECK(mk_closeContext(w->cluster.context, 0));
 174   w->cluster.context = NULL;
 175   CHECK(mk_flushContextData(w->root));
 176   return 0;
 177 }
 178
 179 int   mk_flushFrame(mk_Writer *w, mk_Track *track) {
 180   mk_Context *c;
 181   int64_t   delta, ref = 0;
 182   unsigned  fsize, bgsize;
 183   uint8_t   flags, c_delta_flags[2];
 184   int i;
 185   char *laced = NULL;
 186   uint64_t  length = 0;
 187
 188   if (!track->in_frame)
 189     return 0;
 190
 191   delta = track->frame.timecode/w->timescale - w->cluster.tc_scaled;
 192   if (delta > 32767ll || delta < -32768ll)
 193     CHECK(mk_closeCluster(w));
 194
 195   if (w->cluster.context == NULL) {
 196     w->cluster.tc_scaled = track->frame.timecode / w->timescale;
 197     w->cluster.context = mk_createContext(w, w->root, 0x1f43b675); // Cluster
 198     if (w->cluster.context == NULL)
 199       return -1;
 200
 201     w->cluster.pointer = w->f_pos - w->segment_ptr;
 202
 203     CHECK(mk_writeUInt(w->cluster.context, 0xe7, w->cluster.tc_scaled)); // Cluster Timecode
 204
 205     delta = 0;
 206     w->cluster.block_count = 0;
 207   }
 208
 209   fsize = track->frame.data ? track->frame.data->d_cur : 0;
 210   bgsize = fsize + 4 + mk_ebmlSizeSize(fsize + 4) + 1;
 211   if (!track->frame.keyframe) {
 212     ref = track->prev_frame_tc_scaled - w->cluster.tc_scaled - delta;
 213     bgsize += 1 + 1 + mk_ebmlSIntSize(ref);
 214   }
 215
 216   CHECK(mk_writeID(w->cluster.context, 0xa0)); // BlockGroup
 217   CHECK(mk_writeSize(w->cluster.context, bgsize));
 218   CHECK(mk_writeID(w->cluster.context, 0xa1)); // Block
 219
 220   switch (track->frame.lacing) {
 221     case MK_LACING_XIPH:
 222       laced = mk_laceXiph(track->frame.lacing_sizes, track->frame.lacing_num_frames, &length);
 223       break;
 224     case MK_LACING_EBML:
 225       length += mk_ebmlSizeSize(track->frame.lacing_sizes[0]) + 1;
 226       for (i = 1; i < track->frame.lacing_num_frames; i++)
 227         length += mk_ebmlSizeSize(track->frame.lacing_sizes[i] << 1);
 228       break;
 229     case MK_LACING_FIXED:
 230     {
 231       laced = calloc(1, sizeof(char));
 232       laced[0] = track->frame.lacing_num_frames;
 233       ++length;
 234     }
 235     break;
 236   }
 237
 238   CHECK(mk_writeSize(w->cluster.context, fsize + 4 + length));
 239   CHECK(mk_writeSize(w->cluster.context, track->track_id)); // track number
 240
 241   w->cluster.block_count++;
 242
 243   c_delta_flags[0] = delta >> 8;
 244   c_delta_flags[1] = delta;
 245   CHECK(mk_appendContextData(w->cluster.context, c_delta_flags, 2));
 246
 247   flags = ( track->frame.keyframe << 8 ) | track->frame.lacing;
 248   CHECK(mk_appendContextData(w->cluster.context, &flags, 1));
 249   if (track->frame.lacing) {
 250     if (track->frame.lacing == MK_LACING_EBML) {
 251       CHECK(mk_appendContextData(w->cluster.context, &track->frame.lacing_num_frames, 1));
 252       CHECK(mk_writeSize(w->cluster.context, track->frame.lacing_sizes[0]));
 253       for (i = 1; i < track->frame.lacing_num_frames; i++)
 254       {
 255         CHECK(mk_writeSSize(w->cluster.context, track->frame.lacing_sizes[i] - track->frame.lacing_sizes[i-1]));
 256       }
 257     } else if (length > 0 && laced != NULL) {
 258       CHECK(mk_appendContextData(w->cluster.context, laced, length));
 259       free(laced);
 260       laced = NULL;
 261     }
 262   }
 263
 264   if (track->frame.data) {
 265     CHECK(mk_appendContextData(w->cluster.context, track->frame.data->data, track->frame.data->d_cur));
 266     track->frame.data->d_cur = 0;
 267   }
 268   if (!track->frame.keyframe)
 269     CHECK(mk_writeSInt(w->cluster.context, 0xfb, ref)); // ReferenceBlock
 270
 271   if ((track->cue_flag || (track->track_type & MK_TRACK_VIDEO)) && track->frame.keyframe) {
 272     for(i = 0; i < w->num_tracks; i++)
 273       w->tracks_arr[i]->cue_flag = 1;
 274     if (w->cue_point.timecode != track->frame.timecode) {
 275       if (w->cue_point.context != NULL) {
 276         CHECK(mk_closeContext(w->cue_point.context, 0));
 277         w->cue_point.context = NULL;
 278       }
 279     }
 280     if (w->cue_point.context == NULL) {
 281       if ((w->cue_point.context = mk_createContext(w, w->cues, 0xbb)) == NULL)  // CuePoint
 282         return -1;
 283       CHECK(mk_writeUInt(w->cue_point.context, 0xb3, track->frame.timecode)); // CueTime
 284       w->cue_point.timecode = track->frame.timecode;
 285     }
 286
 287     if ((c = mk_createContext(w, w->cue_point.context, 0xb7)) == NULL)  // CueTrackPositions
 288       return -1;
 289     CHECK(mk_writeUInt(c, 0xf7, track->track_id)); // CueTrack
 290     CHECK(mk_writeUInt(c, 0xf1, w->cluster.pointer));  // CueClusterPosition
 291 //     CHECK(mk_writeUInt(c, 0x5378, w->cluster.block_count));  // CueBlockNumber
 292     CHECK(mk_closeContext(c, 0));
 293     track->cue_flag = 0;
 294   }
 295
 296   track->in_frame = 0;
 297   track->prev_frame_tc_scaled = w->cluster.tc_scaled + delta;
 298
 299   if (w->cluster.context->d_cur > CLSIZE)
 300     CHECK(mk_closeCluster(w));
 301
 302   return 0;
 303 }
 304
 305 int   mk_startFrame(mk_Writer *w, mk_Track *track) {
 306   if (mk_flushFrame(w, track) < 0)
 307     return -1;
 308
 309   track->in_frame = 1;
 310   track->frame.keyframe = 0;
 311   track->frame.lacing = MK_LACING_NONE;
 312   track->frame.lacing_num_frames = 0;
 313   track->frame.lacing_sizes = NULL;
 314
 315   return 0;
 316 }
 317
 318 int   mk_setFrameFlags(mk_Writer *w, mk_Track *track, int64_t timestamp, unsigned keyframe) {
 319   if (!track->in_frame)
 320     return -1;
 321
 322   track->frame.timecode = timestamp;
 323   track->frame.keyframe = keyframe != 0;
 324
 325   if (track->max_frame_tc < timestamp)
 326     track->max_frame_tc = timestamp;
 327
 328   return 0;
 329 }
 330
 331 int   mk_setFrameLacing(mk_Writer *w, mk_Track *track, uint8_t lacing, uint8_t num_frames, uint32_t sizes[]) {
 332   if (!track->in_frame)
 333     return -1;
 334   track->frame.lacing_sizes = calloc(num_frames, sizeof(uint32_t));
 335
 336   track->frame.lacing = lacing;
 337   track->frame.lacing_num_frames = num_frames;
 338   memcpy(track->frame.lacing_sizes, sizes, num_frames);
 339
 340   return 0;
 341 }
 342
 343 int   mk_addFrameData(mk_Writer *w, mk_Track *track, const void *data, unsigned size) {
 344   if (!track->in_frame)
 345     return -1;
 346
 347   if (track->frame.data == NULL)
 348     if ((track->frame.data = mk_createContext(w, NULL, 0)) == NULL)
 349       return -1;
 350
 351   return mk_appendContextData(track->frame.data, data, size);
 352 }
 353
 354 /* The offset of the SeekHead is returned in *pointer. */
 355 int mk_writeSeek(mk_Writer *w, int64_t *pointer) {
 356   mk_Context  *c, *s;
 357   int64_t   seekhead_ptr;
 358
 359   if ((c = mk_createContext(w, w->root, 0x114d9b74)) == NULL) // SeekHead
 360     return -1;
 361   if (pointer != NULL)
 362     seekhead_ptr = w->f_pos;
 363   if (w->seek_data.seekhead) {
 364     if ((s = mk_createContext(w, c, 0x4dbb)) == NULL) // Seek
 365       return -1;
 366     CHECK(mk_writeUInt(s, 0x53ab, 0x114d9b74)); // SeekID
 367     CHECK(mk_writeUInt(s, 0x53ac, w->seek_data.seekhead)); // SeekPosition
 368     CHECK(mk_closeContext(s, 0));
 369   }
 370   if (w->seek_data.segmentinfo) {
 371     if ((s = mk_createContext(w, c, 0x4dbb)) == NULL) // Seek
 372       return -1;
 373     CHECK(mk_writeUInt(s, 0x53ab, 0x1549a966)); // SeekID
 374     CHECK(mk_writeUInt(s, 0x53ac, w->seek_data.segmentinfo)); // SeekPosition
 375     CHECK(mk_closeContext(s, 0));
 376   }
 377   if (w->seek_data.tracks) {
 378     if ((s = mk_createContext(w, c, 0x4dbb)) == NULL) // Seek
 379       return -1;
 380     CHECK(mk_writeUInt(s, 0x53ab, 0x1654ae6b)); // SeekID
 381     CHECK(mk_writeUInt(s, 0x53ac, w->seek_data.tracks)); // SeekPosition
 382     CHECK(mk_closeContext(s, 0));
 383   }
 384   if (w->seek_data.cues) {
 385     if ((s = mk_createContext(w, c, 0x4dbb)) == NULL) // Seek
 386       return -1;
 387     CHECK(mk_writeUInt(s, 0x53ab, 0x1c53bb6b)); // SeekID
 388     CHECK(mk_writeUInt(s, 0x53ac, w->seek_data.cues)); // SeekPosition
 389     CHECK(mk_closeContext(s, 0));
 390   }
 391   if (w->seek_data.attachments) {
 392     if ((s = mk_createContext(w, c, 0x4dbb)) == NULL) // Seek
 393       return -1;
 394     CHECK(mk_writeUInt(s, 0x53ab, 0x1941a469)); // SeekID
 395     CHECK(mk_writeUInt(s, 0x53ac, w->seek_data.attachments)); // SeekPosition
 396     CHECK(mk_closeContext(s, 0));
 397   }
 398   if (w->seek_data.chapters) {
 399     if ((s = mk_createContext(w, c, 0x4dbb)) == NULL) // Seek
 400       return -1;
 401     CHECK(mk_writeUInt(s, 0x53ab, 0x1043a770)); // SeekID
 402     CHECK(mk_writeUInt(s, 0x53ac, w->seek_data.chapters)); // SeekPosition
 403     CHECK(mk_closeContext(s, 0));
 404   }
 405   if (w->seek_data.tags) {
 406     if ((s = mk_createContext(w, c, 0x4dbb)) == NULL) // Seek
 407       return -1;
 408     CHECK(mk_writeUInt(s, 0x53ab, 0x1254c367)); // SeekID
 409     CHECK(mk_writeUInt(s, 0x53ac, w->seek_data.tags)); // SeekPosition
 410     CHECK(mk_closeContext(s, 0));
 411   }
 412   CHECK(mk_closeContext(c, 0));
 413
 414   if (pointer != NULL)
 415     *pointer = seekhead_ptr;
 416
 417   return 0;
 418 }
 419
 420 int   mk_close(mk_Writer *w) {
 421   int   i, ret = 0;
 422   mk_Track *tk;
 423   int64_t max_frame_tc = w->tracks_arr[0]->max_frame_tc;
 424   uint64_t segment_size = 0;
 425   unsigned char c_size[8];
 426
 427   for (i = w->num_tracks - 1; i >= 0; i--)
 428   {
 429     tk = w->tracks_arr[i];
 430     w->tracks_arr[i] = NULL;
 431     if (mk_flushFrame(w, tk) < 0)
 432       ret = -1;
 433     free(tk);
 434     tk = NULL;
 435   }
 436
 437   if (mk_closeCluster(w) < 0)
 438     ret = -1;
 439
 440   if (w->chapters != NULL)
 441   {
 442     if (w->vlc_compat) {
 443       if (mk_seekFile(w, w->segment_ptr + 0x103) < 0)
 444         ret = -1;
 445     }
 446     w->seek_data.chapters = w->f_pos - w->segment_ptr;
 447     mk_writeChapters(w);
 448     if (w->vlc_compat) {
 449       if (mk_flushContextData(w->root) < 0)
 450         ret = -1;
 451       if (mk_writeVoid(w->root, (0x800 - (w->f_pos - w->segment_ptr))) < 0)
 452         ret = -1;
 453     }
 454     if (mk_flushContextData(w->root) < 0)
 455       ret = -1;
 456   }
 457
 458   w->seek_data.cues = w->f_pos - w->segment_ptr;
 459   if (w->cue_point.context != NULL)
 460     if (mk_closeContext(w->cue_point.context, 0) < 0)
 461       ret = -1;
 462 //   if (w->vlc_compat) {
 463 //     if (mk_seekFile(w, w->segment_ptr + 259 + 2051) < 0)
 464 //       ret = -1;
 465 //   }
 466   if (mk_closeContext(w->cues, 0) < 0)
 467     ret = -1;
 468   if (w->vlc_compat) {
 469     if (mk_flushContextData(w->root) < 0)
 470       ret = -1;
 471     if (mk_writeVoid(w->root, (0x1000 - (w->f_pos - w->segment_ptr))) < 0)
 472       ret = -1;
 473   }
 474   if (mk_flushContextData(w->root) < 0)
 475     ret = -1;
 476
 477   if (w->wrote_header) {
 478     if (w->vlc_compat) {
 479       if (mk_seekFile(w, w->segment_ptr) < 0)
 480         ret = -1;
 481     }
 482
 483     if (mk_writeSeek(w, &w->seek_data.seekhead) < 0)
 484       ret = -1;
 485     w->seek_data.seekhead -= w->segment_ptr;
 486
 487     if (w->vlc_compat)
 488     {
 489       if (mk_flushContextData(w->root) < 0)
 490         ret = -1;
 491       if (mk_writeVoid(w->root, (256 - (w->f_pos - w->segment_ptr))) < 0)
 492         ret = -1;
 493     }
 494
 495     if (mk_flushContextData(w->root) < 0)
 496       ret = -1;
 497
 498     if (!w->vlc_compat)
 499     {
 500       int i = w->seek_data.segmentinfo;
 501       w->seek_data.segmentinfo = 0;
 502       w->seek_data.tracks = 0;
 503       w->seek_data.cues = 0;
 504       w->seek_data.chapters = 0;
 505       w->seek_data.attachments = 0;
 506       w->seek_data.tags = 0;
 507       if (mk_seekFile(w, w->segment_ptr) < 0)
 508         ret = -1;
 509       if (mk_writeSeek(w, NULL) < 0 ||
 510           mk_flushContextData(w->root) < 0)
 511         ret = -1;
 512       if (((i + w->segment_ptr) - w->f_pos - 2) > 1)
 513         if (mk_writeVoid(w->root, (i + w->segment_ptr) - w->f_pos - 2) < 0 ||
 514             mk_flushContextData(w->root) < 0)
 515           ret = -1;
 516     }
 517
 518     if (mk_seekFile(w, w->duration_ptr) < 0)
 519       ret = -1;
 520     if (mk_writeFloatRaw(w->root, (float)((double)(max_frame_tc+w->def_duration) / w->timescale)) < 0 ||
 521         mk_flushContextData(w->root) < 0)
 522       ret = -1;
 523     if (mk_seekFile(w, w->segment_ptr - 8) < 0)
 524       ret = -1;
 525     segment_size = w->f_eof - w->segment_ptr;
 526     for (i = 7; i > 0; --i)
 527       c_size[i] = segment_size >> (8 * (7-i));
 528     c_size[i] = 0x01;
 529     if (mk_appendContextData(w->root, &c_size, 8) < 0 ||
 530         mk_flushContextData(w->root) < 0)
 531       ret = -1;
 532   }
 533
 534   if (mk_closeContext(w->root, 0) < 0)
 535     ret = -1;
 536   mk_destroyContexts(w);
 537   fclose(w->fp);
 538   free(w->tracks_arr);
 539   free(w);
 540
 541   return ret;
 542 }