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, 0x1c53bb6b)) == 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, 0x114d9b74)) == 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   if ((c = mk_createContext(w, w->root, 0x1a45dfa3)) == NULL) // EBML
 121     return -1;
 122   CHECK(mk_writeUInt(c, 0x4286, 1)); // EBMLVersion
 123   CHECK(mk_writeUInt(c, 0x42f7, 1)); // EBMLReadVersion
 124   CHECK(mk_writeUInt(c, 0x42f2, 4)); // EBMLMaxIDLength
 125   CHECK(mk_writeUInt(c, 0x42f3, 8)); // EBMLMaxSizeLength
 126   CHECK(mk_writeStr(c, 0x4282, "matroska")); // DocType
 127   CHECK(mk_writeUInt(c, 0x4287, 1)); // DocTypeVersion
 128   CHECK(mk_writeUInt(c, 0x4285, 1)); // DocTypeReadversion
 129   CHECK(mk_closeContext(c, 0));
 130
 131   if ((c = mk_createContext(w, w->root, 0x18538067)) == NULL) // Segment
 132     return -1;
 133   CHECK(mk_flushContextID(c));
 134   w->segment_ptr = c->d_cur;
 135   CHECK(mk_closeContext(c, &w->segment_ptr));
 136
 137   if (w->vlc_compat) {
 138     CHECK(mk_writeVoid(w->root, 0x100));  // 256 bytes should be enough room for our Seek entries.
 139     CHECK(mk_writeVoid(w->root, 0x800)); // 2048 bytes for Chapters.
 140   }
 141   else {
 142     w->seek_data.seekhead = 0x80000000;
 143     CHECK(mk_writeSeekHead(w, &w->seekhead_ptr));
 144     w->seek_data.seekhead = 0;
 145   }
 146
 147   if ((c = mk_createContext(w, w->root, 0x1549a966)) == NULL) // SegmentInfo
 148     return -1;
 149   w->seek_data.segmentinfo = w->root->d_cur - w->segment_ptr;
 150   CHECK(mk_writeVoid(c, 16));   /* Reserve space for a SegmentUID, we'll write the it later. */
 151   CHECK(mk_writeStr(c, 0x4d80, PACKAGE_STRING)); // MuxingApp
 152   CHECK(mk_writeStr(c, 0x5741, writingApp)); // WritingApp
 153   CHECK(mk_writeUInt(c, 0x2ad7b1, w->timescale)); // TimecodeScale
 154   CHECK(mk_writeFloat(c, 0x4489, 0)); // Duration
 155   w->duration_ptr = c->d_cur - 4;
 156   CHECK(mk_closeContext(c, &offset));
 157   w->duration_ptr += offset;
 158   w->segmentuid_ptr = offset;
 159
 160   w->seek_data.tracks = w->root->d_cur - w->segment_ptr;
 161
 162   if (w->tracks)
 163     CHECK(mk_closeContext(w->tracks, 0));
 164
 165   CHECK(mk_flushContextData(w->root));
 166
 167   w->wrote_header = 1;
 168   w->def_duration = w->tracks_arr[0]->default_duration;
 169   return 0;
 170 }
 171
 172 static int mk_closeCluster(mk_Writer *w) {
 173   if (w->cluster.context == NULL)
 174     return 0;
 175   w->cluster.count++;
 176   CHECK(mk_closeContext(w->cluster.context, 0));
 177   w->cluster.context = NULL;
 178   CHECK(mk_flushContextData(w->root));
 179   return 0;
 180 }
 181
 182 int   mk_flushFrame(mk_Writer *w, mk_Track *track) {
 183   mk_Context *c, *tp;
 184   int64_t   delta, ref = 0;
 185   unsigned  fsize, bgsize;
 186   uint8_t   flags, c_delta[2];
 187   int i;
 188   char *laced = NULL;
 189   uint64_t  length = 0;
 190
 191   if (!track->in_frame)
 192     return 0;
 193
 194   delta = track->frame.timecode/w->timescale - w->cluster.tc_scaled;
 195   if (delta > 2000ll || delta < -2000ll)
 196     CHECK(mk_closeCluster(w));
 197
 198   if (w->cluster.context == NULL) {
 199     w->cluster.tc_scaled = track->frame.timecode / w->timescale;
 200     w->cluster.context = mk_createContext(w, w->root, 0x1f43b675); // Cluster
 201     if (w->cluster.context == NULL)
 202       return -1;
 203
 204     w->cluster.pointer = w->f_pos - w->segment_ptr;
 205
 206     if (w->vlc_compat)
 207       CHECK(mk_writeSeek(w, w->cluster.seekhead, 0x1f43b675, w->cluster.pointer));
 208
 209     CHECK(mk_writeUInt(w->cluster.context, 0xe7, w->cluster.tc_scaled)); // Cluster Timecode
 210
 211     delta = 0;
 212     w->cluster.block_count = 0;
 213   }
 214
 215   /* Calculate the encoded lacing sizes. */
 216   switch (track->frame.lacing) {
 217           case MK_LACING_XIPH:
 218                   laced = mk_laceXiph(track->frame.lacing_sizes, track->frame.lacing_num_frames, &length);
 219                   break;
 220           case MK_LACING_EBML:
 221           {
 222                   uint64_t u_size = 0;
 223                   length += mk_ebmlSizeSize(track->frame.lacing_sizes[0]) + 1;  // Add one for the frame count.
 224                   for (i = 1; i < track->frame.lacing_num_frames; i++)
 225                   {
 226                           u_size = llabs(track->frame.lacing_sizes[i] - track->frame.lacing_sizes[i-1]);
 227                           length += mk_ebmlSizeSize((u_size) << 1);             // Shift by one so we get the right size for a signed number.
 228                   }
 229                   break;
 230           }
 231           case MK_LACING_FIXED:
 232           {
 233                   laced = calloc(1, sizeof(char));
 234                   laced[0] = track->frame.lacing_num_frames;
 235                   ++length;
 236                   break;
 237           }
 238   }
 239
 240   fsize = track->frame.data ? track->frame.data->d_cur : 0;
 241   bgsize = fsize + 4 + mk_ebmlSizeSize(fsize + 4 + length) + 1 + length;
 242   if (!track->frame.keyframe) {
 243     ref = track->prev_frame_tc_scaled - w->cluster.tc_scaled - delta;
 244     bgsize += 1 + 1 + mk_ebmlSIntSize(ref);
 245   }
 246
 247   CHECK(mk_writeID(w->cluster.context, 0xa0)); // BlockGroup
 248   CHECK(mk_writeSize(w->cluster.context, bgsize));
 249   CHECK(mk_writeID(w->cluster.context, 0xa1)); // Block
 250   CHECK(mk_writeSize(w->cluster.context, fsize + 4 + length));  // Block size
 251   CHECK(mk_writeSize(w->cluster.context, track->track_id)); // track number
 252
 253   w->cluster.block_count++;
 254
 255   c_delta[0] = delta >> 8;
 256   c_delta[1] = delta;
 257   CHECK(mk_appendContextData(w->cluster.context, c_delta, 2));  // Timecode relative to Cluster.
 258
 259 //   flags = ( track->frame.keyframe << 8 ) | track->frame.lacing;
 260   flags = track->frame.lacing << 1;     // Flags: Bit 5-6 describe what type of lacing to use.
 261   CHECK(mk_appendContextData(w->cluster.context, &flags, 1));
 262   if (track->frame.lacing) {
 263     if (track->frame.lacing == MK_LACING_EBML) {
 264       CHECK(mk_appendContextData(w->cluster.context, &track->frame.lacing_num_frames, 1));      // Number of frames in lace-1
 265       CHECK(mk_writeSize(w->cluster.context, track->frame.lacing_sizes[0]));    // Size of 1st frame.
 266       for (i = 1; i < track->frame.lacing_num_frames; i++)
 267       {
 268         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.
 269       }
 270     } else if (length > 0 && laced != NULL) {
 271       CHECK(mk_appendContextData(w->cluster.context, laced, length));
 272       free(laced);
 273       laced = NULL;
 274     }
 275   }
 276
 277   if (track->frame.data) {
 278     CHECK(mk_appendContextData(w->cluster.context, track->frame.data->data, track->frame.data->d_cur));
 279     track->frame.data->d_cur = 0;
 280   }
 281   if (!track->frame.keyframe)
 282     CHECK(mk_writeSInt(w->cluster.context, 0xfb, ref)); // ReferenceBlock
 283
 284   if (track->frame.keyframe && (track->track_type & MK_TRACK_VIDEO) && ((track->prev_cue_pos + 3*CLSIZE) <= w->f_pos || track->frame.timecode == 0)) {
 285     if ((c = mk_createContext(w, w->cues, 0xbb)) == NULL)  // CuePoint
 286       return -1;
 287     CHECK(mk_writeUInt(c, 0xb3, track->frame.timecode)); // CueTime
 288
 289     if ((tp = mk_createContext(w, c, 0xb7)) == NULL)  // CueTrackPositions
 290       return -1;
 291     CHECK(mk_writeUInt(tp, 0xf7, track->track_id)); // CueTrack
 292     CHECK(mk_writeUInt(tp, 0xf1, w->cluster.pointer));  // CueClusterPosition
 293 //     CHECK(mk_writeUInt(c, 0x5378, w->cluster.block_count));  // CueBlockNumber
 294     CHECK(mk_closeContext(tp, 0));
 295     CHECK(mk_closeContext(c, 0));
 296     track->prev_cue_pos = w->f_pos;
 297   }
 298
 299   track->in_frame = 0;
 300   track->prev_frame_tc_scaled = w->cluster.tc_scaled + delta;
 301
 302   if (w->cluster.context->d_cur > CLSIZE)
 303     CHECK(mk_closeCluster(w));
 304
 305   return 0;
 306 }
 307
 308 int   mk_startFrame(mk_Writer *w, mk_Track *track) {
 309   if (mk_flushFrame(w, track) < 0)
 310     return -1;
 311
 312   track->in_frame = 1;
 313   track->frame.keyframe = 0;
 314   track->frame.lacing = MK_LACING_NONE;
 315   track->frame.lacing_num_frames = 0;
 316   track->frame.lacing_sizes = NULL;
 317
 318   return 0;
 319 }
 320
 321 int   mk_setFrameFlags(mk_Writer *w, mk_Track *track, int64_t timestamp, unsigned keyframe) {
 322   if (!track->in_frame)
 323     return -1;
 324
 325   track->frame.timecode = timestamp;
 326   track->frame.keyframe = keyframe != 0;
 327
 328   if (track->max_frame_tc < timestamp)
 329     track->max_frame_tc = timestamp;
 330
 331   return 0;
 332 }
 333
 334 int   mk_setFrameLacing(mk_Writer *w, mk_Track *track, uint8_t lacing, uint8_t num_frames, uint64_t sizes[]) {
 335   if (!track->in_frame)
 336     return -1;
 337   track->frame.lacing_sizes = calloc(num_frames, sizeof(uint64_t));
 338
 339   track->frame.lacing = lacing;
 340   track->frame.lacing_num_frames = num_frames;
 341   memcpy(track->frame.lacing_sizes, sizes, num_frames * sizeof(uint64_t));
 342
 343   return 0;
 344 }
 345
 346 int   mk_addFrameData(mk_Writer *w, mk_Track *track, const void *data, unsigned size) {
 347   if (!track->in_frame)
 348     return -1;
 349
 350   if (track->frame.data == NULL)
 351     if ((track->frame.data = mk_createContext(w, NULL, 0)) == NULL)
 352       return -1;
 353
 354   md5_update(&w->segment_md5, (unsigned char *)data, size);
 355
 356   return mk_appendContextData(track->frame.data, data, size);
 357 }
 358
 359 int   mk_writeSeek(mk_Writer *w, mk_Context *c, unsigned seek_id, uint64_t seek_pos) {
 360   mk_Context  *s;
 361
 362   if ((s = mk_createContext(w, c, 0x4dbb)) == NULL) // Seek
 363     return -1;
 364   CHECK(mk_writeUInt(s, 0x53ab, seek_id));  // SeekID
 365   CHECK(mk_writeUInt(s, 0x53ac, seek_pos)); // SeekPosition
 366   CHECK(mk_closeContext(s, 0));
 367
 368   return 0;
 369 }
 370
 371 /* The offset of the SeekHead is returned in *pointer. */
 372 int mk_writeSeekHead(mk_Writer *w, int64_t *pointer) {
 373   mk_Context  *c;
 374   int64_t   seekhead_ptr;
 375
 376   if ((c = mk_createContext(w, w->root, 0x114d9b74)) == NULL) // SeekHead
 377     return -1;
 378   if (pointer != NULL)
 379     seekhead_ptr = w->f_pos;
 380   if (w->seek_data.seekhead)
 381     CHECK(mk_writeSeek(w, c, 0x114d9b74, w->seek_data.seekhead));
 382   if (w->seek_data.segmentinfo)
 383     CHECK(mk_writeSeek(w, c, 0x1549a966, w->seek_data.segmentinfo));
 384   if (w->seek_data.tracks)
 385     CHECK(mk_writeSeek(w, c, 0x1654ae6b, w->seek_data.tracks));
 386   if (w->seek_data.cues)
 387     CHECK(mk_writeSeek(w, c, 0x1c53bb6b, w->seek_data.cues));
 388   if (w->seek_data.attachments)
 389     CHECK(mk_writeSeek(w, c, 0x1941a469, w->seek_data.attachments));
 390   if (w->seek_data.chapters)
 391     CHECK(mk_writeSeek(w, c, 0x1043a770, w->seek_data.chapters));
 392   if (w->seek_data.tags)
 393     CHECK(mk_writeSeek(w, c, 0x1254c367, w->seek_data.tags));
 394   CHECK(mk_closeContext(c, 0));
 395
 396   if (pointer != NULL)
 397     *pointer = seekhead_ptr;
 398
 399   return 0;
 400 }
 401
 402 int   mk_close(mk_Writer *w) {
 403   int   i, ret = 0;
 404   mk_Track *tk;
 405   int64_t max_frame_tc = w->tracks_arr[0]->max_frame_tc;
 406   uint64_t segment_size = 0;
 407   unsigned char c_size[8];
 408   unsigned char segment_uid[16];
 409
 410   md5_finish(&w->segment_md5, segment_uid);
 411
 412   for (i = w->num_tracks - 1; i >= 0; i--)
 413   {
 414     tk = w->tracks_arr[i];
 415     w->tracks_arr[i] = NULL;
 416     --w->num_tracks;
 417     if (mk_flushFrame(w, tk) < 0)
 418       ret = -1;
 419     free(tk);
 420     tk = NULL;
 421   }
 422
 423   if (mk_closeCluster(w) < 0)
 424     ret = -1;
 425
 426   w->seek_data.cues = w->f_pos - w->segment_ptr;
 427   if (mk_closeContext(w->cues, 0) < 0)
 428     ret = -1;
 429   if (mk_flushContextData(w->root) < 0)
 430     ret = -1;
 431
 432   if (w->vlc_compat && w->cluster.seekhead) {
 433     w->seek_data.seekhead = w->f_pos - w->segment_ptr;
 434     if (mk_closeContext(w->cluster.seekhead, 0) < 0)
 435       ret = -1;
 436     if (mk_flushContextData(w->root) < 0)
 437       ret = -1;
 438   }
 439
 440   if (w->chapters != NULL)
 441   {
 442     if (w->vlc_compat) {
 443       if (mk_flushContextData(w->root) < 0)
 444         ret = -1;
 445       if (mk_seekFile(w, w->segment_ptr + 0x100 + 3) < 0)
 446         ret = -1;
 447     }
 448     w->seek_data.chapters = w->f_pos - w->segment_ptr;
 449     mk_writeChapters(w);
 450     if (mk_flushContextData(w->root) < 0)
 451       ret = -1;
 452     if (w->vlc_compat) {
 453       if (mk_writeVoid(w->root, (0x800 - (w->f_pos - w->segment_ptr - 0x100 - 3))) < 0)
 454         ret = -1;
 455       if (mk_flushContextData(w->root) < 0)
 456         ret = -1;
 457     }
 458   }
 459
 460   if (w->wrote_header) {
 461     if (w->vlc_compat) {
 462       if (mk_seekFile(w, w->segment_ptr) < 0)
 463         ret = -1;
 464     }
 465
 466     if (mk_writeSeekHead(w, &w->seek_data.seekhead) < 0)
 467       ret = -1;
 468     w->seek_data.seekhead -= w->segment_ptr;
 469
 470     if (w->vlc_compat)
 471     {
 472       if (mk_flushContextData(w->root) < 0)
 473         ret = -1;
 474       if (mk_writeVoid(w->root, (0x100 - (w->f_pos - w->segment_ptr))) < 0)
 475         ret = -1;
 476     }
 477
 478     if (mk_flushContextData(w->root) < 0)
 479       ret = -1;
 480
 481     if (!w->vlc_compat)
 482     {
 483       int i = w->seek_data.segmentinfo;
 484       w->seek_data.segmentinfo = 0;
 485       w->seek_data.tracks = 0;
 486       w->seek_data.cues = 0;
 487       w->seek_data.chapters = 0;
 488       w->seek_data.attachments = 0;
 489       w->seek_data.tags = 0;
 490       if (mk_seekFile(w, w->segment_ptr) < 0)
 491         ret = -1;
 492       if (mk_writeSeekHead(w, NULL) < 0 ||
 493           mk_flushContextData(w->root) < 0)
 494         ret = -1;
 495       if (((i + w->segment_ptr) - w->f_pos - 2) > 1)
 496         if (mk_writeVoid(w->root, (i + w->segment_ptr) - w->f_pos - 2) < 0 ||
 497             mk_flushContextData(w->root) < 0)
 498           ret = -1;
 499     }
 500
 501     if (mk_seekFile(w, w->duration_ptr) < 0)
 502       ret = -1;
 503     if (mk_writeFloatRaw(w->root, (float)((double)(max_frame_tc+w->def_duration) / w->timescale)) < 0 ||
 504         mk_flushContextData(w->root) < 0)
 505       ret = -1;
 506     if (mk_seekFile(w, w->segment_ptr - 8) < 0)
 507       ret = -1;
 508     segment_size = w->f_eof - w->segment_ptr;
 509     for (i = 7; i > 0; --i)
 510       c_size[i] = segment_size >> (8 * (7-i));
 511     c_size[i] = 0x01;
 512     if (mk_appendContextData(w->root, &c_size, 8) < 0 ||
 513         mk_flushContextData(w->root) < 0)
 514       ret = -1;
 515         if (mk_seekFile(w, w->segmentuid_ptr) < 0)
 516                 ret = -1;
 517         if (mk_writeBin(w->root, 0x73a4, segment_uid, sizeof(segment_uid)) < 0 ||
 518                 mk_flushContextData(w->root) < 0)
 519           ret = -1;
 520   }
 521
 522   if (mk_closeContext(w->root, 0) < 0)
 523     ret = -1;
 524   mk_destroyContexts(w);
 525   fclose(w->fp);
 526   free(w->tracks_arr);
 527   free(w);
 528
 529   return ret;
 530 }