FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
matroskadec.c
Go to the documentation of this file.
1 /*
2  * Matroska file demuxer
3  * Copyright (c) 2003-2008 The FFmpeg Project
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * FFmpeg 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 GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 /**
23  * @file
24  * Matroska file demuxer
25  * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26  * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27  * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28  * @see specs available on the Matroska project page: http://www.matroska.org/
29  */
30 
31 #include "config.h"
32 
33 #include <inttypes.h>
34 #include <stdio.h>
35 #if CONFIG_BZLIB
36 #include <bzlib.h>
37 #endif
38 #if CONFIG_ZLIB
39 #include <zlib.h>
40 #endif
41 
42 #include "libavutil/avstring.h"
43 #include "libavutil/base64.h"
44 #include "libavutil/dict.h"
45 #include "libavutil/intfloat.h"
46 #include "libavutil/intreadwrite.h"
47 #include "libavutil/lzo.h"
48 #include "libavutil/mathematics.h"
49 
50 #include "libavcodec/bytestream.h"
51 #include "libavcodec/flac.h"
52 #include "libavcodec/mpeg4audio.h"
53 
54 #include "avformat.h"
55 #include "avio_internal.h"
56 #include "internal.h"
57 #include "isom.h"
58 #include "matroska.h"
59 #include "oggdec.h"
60 /* For ff_codec_get_id(). */
61 #include "riff.h"
62 #include "rmsipr.h"
63 
64 typedef enum {
76 } EbmlType;
77 
78 typedef const struct EbmlSyntax {
79  uint32_t id;
83  union {
84  uint64_t u;
85  double f;
86  const char *s;
87  const struct EbmlSyntax *n;
88  } def;
89 } EbmlSyntax;
90 
91 typedef struct {
92  int nb_elem;
93  void *elem;
94 } EbmlList;
95 
96 typedef struct {
97  int size;
99  int64_t pos;
100 } EbmlBin;
101 
102 typedef struct {
103  uint64_t version;
104  uint64_t max_size;
105  uint64_t id_length;
106  char *doctype;
107  uint64_t doctype_version;
108 } Ebml;
109 
110 typedef struct {
111  uint64_t algo;
114 
115 typedef struct {
116  uint64_t algo;
119 
120 typedef struct {
121  uint64_t scope;
122  uint64_t type;
126 
127 typedef struct {
128  double frame_rate;
129  uint64_t display_width;
130  uint64_t display_height;
131  uint64_t pixel_width;
132  uint64_t pixel_height;
134  uint64_t stereo_mode;
135  uint64_t alpha_mode;
137 
138 typedef struct {
139  double samplerate;
141  uint64_t bitdepth;
142  uint64_t channels;
143 
144  /* real audio header (extracted from extradata) */
150  int pkt_cnt;
151  uint64_t buf_timecode;
154 
155 typedef struct {
156  uint64_t uid;
157  uint64_t type;
159 
160 typedef struct {
163 
164 typedef struct {
165  uint64_t num;
166  uint64_t uid;
167  uint64_t type;
168  char *name;
169  char *codec_id;
171  char *language;
172  double time_scale;
174  uint64_t flag_default;
175  uint64_t flag_forced;
176  uint64_t seek_preroll;
181  uint64_t codec_delay;
182 
184  int64_t end_timecode;
187 } MatroskaTrack;
188 
189 typedef struct {
190  uint64_t uid;
191  char *filename;
192  char *mime;
194 
197 
198 typedef struct {
199  uint64_t start;
200  uint64_t end;
201  uint64_t uid;
202  char *title;
203 
206 
207 typedef struct {
208  uint64_t track;
209  uint64_t pos;
211 
212 typedef struct {
213  uint64_t time;
215 } MatroskaIndex;
216 
217 typedef struct {
218  char *name;
219  char *string;
220  char *lang;
221  uint64_t def;
223 } MatroskaTag;
224 
225 typedef struct {
226  char *type;
227  uint64_t typevalue;
228  uint64_t trackuid;
229  uint64_t chapteruid;
230  uint64_t attachuid;
232 
233 typedef struct {
236 } MatroskaTags;
237 
238 typedef struct {
239  uint64_t id;
240  uint64_t pos;
242 
243 typedef struct {
244  uint64_t start;
245  uint64_t length;
246 } MatroskaLevel;
247 
248 typedef struct {
249  uint64_t timecode;
252 
253 typedef struct {
255 
256  /* EBML stuff */
259  int level_up;
260  uint32_t current_id;
261 
262  uint64_t time_scale;
263  double duration;
264  char *title;
265  char *muxingapp;
273 
274  /* byte position of the segment inside the stream */
275  int64_t segment_start;
276 
277  /* the packet queue */
281 
282  int done;
283 
284  /* What to skip before effectively reading a packet. */
287 
288  /* File has a CUES element, but we defer parsing until it is needed. */
290 
294 
295  /* File has SSA subtitles which prevent incremental cluster parsing. */
298 
299 typedef struct {
300  uint64_t duration;
301  int64_t reference;
302  uint64_t non_simple;
304  uint64_t additional_id;
307 } MatroskaBlock;
308 
310  { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
311  { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
312  { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
313  { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
314  { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
317  { 0 }
318 };
319 
321  { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
322  { 0 }
323 };
324 
326  { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
328  { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
330  { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
331  { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
333  { 0 }
334 };
335 
337  { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
338  { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
339  { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
340  { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
341  { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
342  { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
343  { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
352  { 0 }
353 };
354 
356  { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
357  { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
359  { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
360  { 0 }
361 };
362 
366  { 0 }
367 };
368 
377  { 0 }
378 };
380  { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
381  { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
382  { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
383  { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
385  { 0 }
386 };
387 
389  { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
390  { 0 }
391 };
392 
396  { 0 }
397 };
398 
400  { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
401  { 0 }
402 };
403 
405  { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
406  { 0 }
407 };
408 
410  { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
412  { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
415  { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
416  { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
417  { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
418  { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
419  { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
420  { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
421  { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
422  { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
423  { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
424  { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
425  { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
426  { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
427  { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
436  { 0 }
437 };
438 
440  { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
441  { 0 }
442 };
443 
445  { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
446  { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
447  { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
448  { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
450  { 0 }
451 };
452 
454  { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
455  { 0 }
456 };
457 
459  { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
461  { 0 }
462 };
463 
467  { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
468  { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
473  { 0 }
474 };
475 
477  { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
482  { 0 }
483 };
484 
486  { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
487  { 0 }
488 };
489 
491  { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
496  { 0 }
497 };
498 
500  { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
501  { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
502  { 0 }
503 };
504 
506  { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
507  { 0 }
508 };
509 
511  { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
512  { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
513  { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
514  { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
515  { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
516  { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
517  { 0 }
518 };
519 
522  { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
523  { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
524  { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
525  { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
526  { 0 }
527 };
528 
530  { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
531  { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
532  { 0 }
533 };
534 
536  { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
537  { 0 }
538 };
539 
541  { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
542  { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
543  { 0 }
544 };
545 
547  { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
548  { 0 }
549 };
550 
552  { MATROSKA_ID_INFO, EBML_NEST, 0, 0, { .n = matroska_info } },
553  { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, { .n = matroska_tracks } },
554  { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, { .n = matroska_attachments } },
555  { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, { .n = matroska_chapters } },
556  { MATROSKA_ID_CUES, EBML_NEST, 0, 0, { .n = matroska_index } },
557  { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, { .n = matroska_tags } },
558  { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, { .n = matroska_seekhead } },
560  { 0 }
561 };
562 
564  { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
565  { 0 }
566 };
567 
569  { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
570  { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
571  { 0 }
572 };
573 
575  { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
576  { 0 }
577 };
578 
580  { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
581  { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
582  { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
584  { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
585  { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference) },
587  { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
588  { 0 }
589 };
590 
592  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
593  { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
594  { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
597  { 0 }
598 };
599 
601  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
606  { 0 }
607 };
608 
610  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
611  { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
612  { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
620  { 0 }
621 };
622 
624  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
629  { 0 }
630 };
631 
633  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
638  { 0 }
639 };
640 
641 static const char *const matroska_doctypes[] = { "matroska", "webm" };
642 
643 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
644 {
645  AVIOContext *pb = matroska->ctx->pb;
646  uint32_t id;
647  matroska->current_id = 0;
648  matroska->num_levels = 0;
649 
650  /* seek to next position to resync from */
651  if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
652  goto eof;
653 
654  id = avio_rb32(pb);
655 
656  // try to find a toplevel element
657  while (!avio_feof(pb)) {
658  if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
659  id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
661  id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
662  matroska->current_id = id;
663  return 0;
664  }
665  id = (id << 8) | avio_r8(pb);
666  }
667 
668 eof:
669  matroska->done = 1;
670  return AVERROR_EOF;
671 }
672 
673 /*
674  * Return: Whether we reached the end of a level in the hierarchy or not.
675  */
677 {
678  AVIOContext *pb = matroska->ctx->pb;
679  int64_t pos = avio_tell(pb);
680 
681  if (matroska->num_levels > 0) {
682  MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
683  if (pos - level->start >= level->length || matroska->current_id) {
684  matroska->num_levels--;
685  return 1;
686  }
687  }
688  return 0;
689 }
690 
691 /*
692  * Read: an "EBML number", which is defined as a variable-length
693  * array of bytes. The first byte indicates the length by giving a
694  * number of 0-bits followed by a one. The position of the first
695  * "one" bit inside the first byte indicates the length of this
696  * number.
697  * Returns: number of bytes read, < 0 on error
698  */
700  int max_size, uint64_t *number)
701 {
702  int read = 1, n = 1;
703  uint64_t total = 0;
704 
705  /* The first byte tells us the length in bytes - avio_r8() can normally
706  * return 0, but since that's not a valid first ebmlID byte, we can
707  * use it safely here to catch EOS. */
708  if (!(total = avio_r8(pb))) {
709  /* we might encounter EOS here */
710  if (!avio_feof(pb)) {
711  int64_t pos = avio_tell(pb);
712  av_log(matroska->ctx, AV_LOG_ERROR,
713  "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
714  pos, pos);
715  return pb->error ? pb->error : AVERROR(EIO);
716  }
717  return AVERROR_EOF;
718  }
719 
720  /* get the length of the EBML number */
721  read = 8 - ff_log2_tab[total];
722  if (read > max_size) {
723  int64_t pos = avio_tell(pb) - 1;
724  av_log(matroska->ctx, AV_LOG_ERROR,
725  "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
726  (uint8_t) total, pos, pos);
727  return AVERROR_INVALIDDATA;
728  }
729 
730  /* read out length */
731  total ^= 1 << ff_log2_tab[total];
732  while (n++ < read)
733  total = (total << 8) | avio_r8(pb);
734 
735  *number = total;
736 
737  return read;
738 }
739 
740 /**
741  * Read a EBML length value.
742  * This needs special handling for the "unknown length" case which has multiple
743  * encodings.
744  */
746  uint64_t *number)
747 {
748  int res = ebml_read_num(matroska, pb, 8, number);
749  if (res > 0 && *number + 1 == 1ULL << (7 * res))
750  *number = 0xffffffffffffffULL;
751  return res;
752 }
753 
754 /*
755  * Read the next element as an unsigned int.
756  * 0 is success, < 0 is failure.
757  */
758 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
759 {
760  int n = 0;
761 
762  if (size > 8)
763  return AVERROR_INVALIDDATA;
764 
765  /* big-endian ordering; build up number */
766  *num = 0;
767  while (n++ < size)
768  *num = (*num << 8) | avio_r8(pb);
769 
770  return 0;
771 }
772 
773 /*
774  * Read the next element as a signed int.
775  * 0 is success, < 0 is failure.
776  */
777 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
778 {
779  int n = 1;
780 
781  if (size > 8)
782  return AVERROR_INVALIDDATA;
783 
784  if (size == 0) {
785  *num = 0;
786  } else {
787  *num = sign_extend(avio_r8(pb), 8);
788 
789  /* big-endian ordering; build up number */
790  while (n++ < size)
791  *num = (*num << 8) | avio_r8(pb);
792  }
793 
794  return 0;
795 }
796 
797 /*
798  * Read the next element as a float.
799  * 0 is success, < 0 is failure.
800  */
801 static int ebml_read_float(AVIOContext *pb, int size, double *num)
802 {
803  if (size == 0)
804  *num = 0;
805  else if (size == 4)
806  *num = av_int2float(avio_rb32(pb));
807  else if (size == 8)
808  *num = av_int2double(avio_rb64(pb));
809  else
810  return AVERROR_INVALIDDATA;
811 
812  return 0;
813 }
814 
815 /*
816  * Read the next element as an ASCII string.
817  * 0 is success, < 0 is failure.
818  */
819 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
820 {
821  char *res;
822 
823  /* EBML strings are usually not 0-terminated, so we allocate one
824  * byte more, read the string and NULL-terminate it ourselves. */
825  if (!(res = av_malloc(size + 1)))
826  return AVERROR(ENOMEM);
827  if (avio_read(pb, (uint8_t *) res, size) != size) {
828  av_free(res);
829  return AVERROR(EIO);
830  }
831  (res)[size] = '\0';
832  av_free(*str);
833  *str = res;
834 
835  return 0;
836 }
837 
838 /*
839  * Read the next element as binary data.
840  * 0 is success, < 0 is failure.
841  */
842 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
843 {
844  av_fast_padded_malloc(&bin->data, &bin->size, length);
845  if (!bin->data)
846  return AVERROR(ENOMEM);
847 
848  bin->size = length;
849  bin->pos = avio_tell(pb);
850  if (avio_read(pb, bin->data, length) != length) {
851  av_freep(&bin->data);
852  bin->size = 0;
853  return AVERROR(EIO);
854  }
855 
856  return 0;
857 }
858 
859 /*
860  * Read the next element, but only the header. The contents
861  * are supposed to be sub-elements which can be read separately.
862  * 0 is success, < 0 is failure.
863  */
864 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
865 {
866  AVIOContext *pb = matroska->ctx->pb;
868 
869  if (matroska->num_levels >= EBML_MAX_DEPTH) {
870  av_log(matroska->ctx, AV_LOG_ERROR,
871  "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
872  return AVERROR(ENOSYS);
873  }
874 
875  level = &matroska->levels[matroska->num_levels++];
876  level->start = avio_tell(pb);
877  level->length = length;
878 
879  return 0;
880 }
881 
882 /*
883  * Read signed/unsigned "EBML" numbers.
884  * Return: number of bytes processed, < 0 on error
885  */
887  uint8_t *data, uint32_t size, uint64_t *num)
888 {
889  AVIOContext pb;
890  ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
891  return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
892 }
893 
894 /*
895  * Same as above, but signed.
896  */
898  uint8_t *data, uint32_t size, int64_t *num)
899 {
900  uint64_t unum;
901  int res;
902 
903  /* read as unsigned number first */
904  if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
905  return res;
906 
907  /* make signed (weird way) */
908  *num = unum - ((1LL << (7 * res - 1)) - 1);
909 
910  return res;
911 }
912 
913 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
914  EbmlSyntax *syntax, void *data);
915 
916 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
917  uint32_t id, void *data)
918 {
919  int i;
920  for (i = 0; syntax[i].id; i++)
921  if (id == syntax[i].id)
922  break;
923  if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
924  matroska->num_levels > 0 &&
925  matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
926  return 0; // we reached the end of an unknown size cluster
927  if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
928  av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%"PRIX32"\n", id);
929  if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
930  return AVERROR_INVALIDDATA;
931  }
932  return ebml_parse_elem(matroska, &syntax[i], data);
933 }
934 
935 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
936  void *data)
937 {
938  if (!matroska->current_id) {
939  uint64_t id;
940  int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
941  if (res < 0)
942  return res;
943  matroska->current_id = id | 1 << 7 * res;
944  }
945  return ebml_parse_id(matroska, syntax, matroska->current_id, data);
946 }
947 
948 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
949  void *data)
950 {
951  int i, res = 0;
952 
953  for (i = 0; syntax[i].id; i++)
954  switch (syntax[i].type) {
955  case EBML_UINT:
956  *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
957  break;
958  case EBML_FLOAT:
959  *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
960  break;
961  case EBML_STR:
962  case EBML_UTF8:
963  // the default may be NULL
964  if (syntax[i].def.s) {
965  uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
966  *dst = av_strdup(syntax[i].def.s);
967  if (!*dst)
968  return AVERROR(ENOMEM);
969  }
970  break;
971  }
972 
973  while (!res && !ebml_level_end(matroska))
974  res = ebml_parse(matroska, syntax, data);
975 
976  return res;
977 }
978 
980  EbmlSyntax *syntax, void *data)
981 {
982  static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
983  [EBML_UINT] = 8,
984  [EBML_FLOAT] = 8,
985  // max. 16 MB for strings
986  [EBML_STR] = 0x1000000,
987  [EBML_UTF8] = 0x1000000,
988  // max. 256 MB for binary data
989  [EBML_BIN] = 0x10000000,
990  // no limits for anything else
991  };
992  AVIOContext *pb = matroska->ctx->pb;
993  uint32_t id = syntax->id;
994  uint64_t length;
995  int res;
996  void *newelem;
997 
998  data = (char *) data + syntax->data_offset;
999  if (syntax->list_elem_size) {
1000  EbmlList *list = data;
1001  newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1002  if (!newelem)
1003  return AVERROR(ENOMEM);
1004  list->elem = newelem;
1005  data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1006  memset(data, 0, syntax->list_elem_size);
1007  list->nb_elem++;
1008  }
1009 
1010  if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1011  matroska->current_id = 0;
1012  if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1013  return res;
1014  if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1015  av_log(matroska->ctx, AV_LOG_ERROR,
1016  "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1017  length, max_lengths[syntax->type], syntax->type);
1018  return AVERROR_INVALIDDATA;
1019  }
1020  }
1021 
1022  switch (syntax->type) {
1023  case EBML_UINT:
1024  res = ebml_read_uint(pb, length, data);
1025  break;
1026  case EBML_SINT:
1027  res = ebml_read_sint(pb, length, data);
1028  break;
1029  case EBML_FLOAT:
1030  res = ebml_read_float(pb, length, data);
1031  break;
1032  case EBML_STR:
1033  case EBML_UTF8:
1034  res = ebml_read_ascii(pb, length, data);
1035  break;
1036  case EBML_BIN:
1037  res = ebml_read_binary(pb, length, data);
1038  break;
1039  case EBML_NEST:
1040  if ((res = ebml_read_master(matroska, length)) < 0)
1041  return res;
1042  if (id == MATROSKA_ID_SEGMENT)
1043  matroska->segment_start = avio_tell(matroska->ctx->pb);
1044  return ebml_parse_nest(matroska, syntax->def.n, data);
1045  case EBML_PASS:
1046  return ebml_parse_id(matroska, syntax->def.n, id, data);
1047  case EBML_STOP:
1048  return 1;
1049  default:
1050  if (ffio_limit(pb, length) != length)
1051  return AVERROR(EIO);
1052  return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
1053  }
1054  if (res == AVERROR_INVALIDDATA)
1055  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1056  else if (res == AVERROR(EIO))
1057  av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1058  return res;
1059 }
1060 
1061 static void ebml_free(EbmlSyntax *syntax, void *data)
1062 {
1063  int i, j;
1064  for (i = 0; syntax[i].id; i++) {
1065  void *data_off = (char *) data + syntax[i].data_offset;
1066  switch (syntax[i].type) {
1067  case EBML_STR:
1068  case EBML_UTF8:
1069  av_freep(data_off);
1070  break;
1071  case EBML_BIN:
1072  av_freep(&((EbmlBin *) data_off)->data);
1073  break;
1074  case EBML_NEST:
1075  if (syntax[i].list_elem_size) {
1076  EbmlList *list = data_off;
1077  char *ptr = list->elem;
1078  for (j = 0; j < list->nb_elem;
1079  j++, ptr += syntax[i].list_elem_size)
1080  ebml_free(syntax[i].def.n, ptr);
1081  av_free(list->elem);
1082  } else
1083  ebml_free(syntax[i].def.n, data_off);
1084  default:
1085  break;
1086  }
1087  }
1088 }
1089 
1090 /*
1091  * Autodetecting...
1092  */
1094 {
1095  uint64_t total = 0;
1096  int len_mask = 0x80, size = 1, n = 1, i;
1097 
1098  /* EBML header? */
1099  if (AV_RB32(p->buf) != EBML_ID_HEADER)
1100  return 0;
1101 
1102  /* length of header */
1103  total = p->buf[4];
1104  while (size <= 8 && !(total & len_mask)) {
1105  size++;
1106  len_mask >>= 1;
1107  }
1108  if (size > 8)
1109  return 0;
1110  total &= (len_mask - 1);
1111  while (n < size)
1112  total = (total << 8) | p->buf[4 + n++];
1113 
1114  /* Does the probe data contain the whole header? */
1115  if (p->buf_size < 4 + size + total)
1116  return 0;
1117 
1118  /* The header should contain a known document type. For now,
1119  * we don't parse the whole header but simply check for the
1120  * availability of that array of characters inside the header.
1121  * Not fully fool-proof, but good enough. */
1122  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1123  int probelen = strlen(matroska_doctypes[i]);
1124  if (total < probelen)
1125  continue;
1126  for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1127  if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1128  return AVPROBE_SCORE_MAX;
1129  }
1130 
1131  // probably valid EBML header but no recognized doctype
1132  return AVPROBE_SCORE_EXTENSION;
1133 }
1134 
1136  int num)
1137 {
1138  MatroskaTrack *tracks = matroska->tracks.elem;
1139  int i;
1140 
1141  for (i = 0; i < matroska->tracks.nb_elem; i++)
1142  if (tracks[i].num == num)
1143  return &tracks[i];
1144 
1145  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1146  return NULL;
1147 }
1148 
1149 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1150  MatroskaTrack *track)
1151 {
1152  MatroskaTrackEncoding *encodings = track->encodings.elem;
1153  uint8_t *data = *buf;
1154  int isize = *buf_size;
1155  uint8_t *pkt_data = NULL;
1156  uint8_t av_unused *newpktdata;
1157  int pkt_size = isize;
1158  int result = 0;
1159  int olen;
1160 
1161  if (pkt_size >= 10000000U)
1162  return AVERROR_INVALIDDATA;
1163 
1164  switch (encodings[0].compression.algo) {
1166  {
1167  int header_size = encodings[0].compression.settings.size;
1168  uint8_t *header = encodings[0].compression.settings.data;
1169 
1170  if (header_size && !header) {
1171  av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1172  return -1;
1173  }
1174 
1175  if (!header_size)
1176  return 0;
1177 
1178  pkt_size = isize + header_size;
1179  pkt_data = av_malloc(pkt_size);
1180  if (!pkt_data)
1181  return AVERROR(ENOMEM);
1182 
1183  memcpy(pkt_data, header, header_size);
1184  memcpy(pkt_data + header_size, data, isize);
1185  break;
1186  }
1187 #if CONFIG_LZO
1189  do {
1190  olen = pkt_size *= 3;
1191  newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1192  if (!newpktdata) {
1193  result = AVERROR(ENOMEM);
1194  goto failed;
1195  }
1196  pkt_data = newpktdata;
1197  result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1198  } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1199  if (result) {
1200  result = AVERROR_INVALIDDATA;
1201  goto failed;
1202  }
1203  pkt_size -= olen;
1204  break;
1205 #endif
1206 #if CONFIG_ZLIB
1208  {
1209  z_stream zstream = { 0 };
1210  if (inflateInit(&zstream) != Z_OK)
1211  return -1;
1212  zstream.next_in = data;
1213  zstream.avail_in = isize;
1214  do {
1215  pkt_size *= 3;
1216  newpktdata = av_realloc(pkt_data, pkt_size);
1217  if (!newpktdata) {
1218  inflateEnd(&zstream);
1219  goto failed;
1220  }
1221  pkt_data = newpktdata;
1222  zstream.avail_out = pkt_size - zstream.total_out;
1223  zstream.next_out = pkt_data + zstream.total_out;
1224  if (pkt_data) {
1225  result = inflate(&zstream, Z_NO_FLUSH);
1226  } else
1227  result = Z_MEM_ERROR;
1228  } while (result == Z_OK && pkt_size < 10000000);
1229  pkt_size = zstream.total_out;
1230  inflateEnd(&zstream);
1231  if (result != Z_STREAM_END) {
1232  if (result == Z_MEM_ERROR)
1233  result = AVERROR(ENOMEM);
1234  else
1235  result = AVERROR_INVALIDDATA;
1236  goto failed;
1237  }
1238  break;
1239  }
1240 #endif
1241 #if CONFIG_BZLIB
1243  {
1244  bz_stream bzstream = { 0 };
1245  if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1246  return -1;
1247  bzstream.next_in = data;
1248  bzstream.avail_in = isize;
1249  do {
1250  pkt_size *= 3;
1251  newpktdata = av_realloc(pkt_data, pkt_size);
1252  if (!newpktdata) {
1253  BZ2_bzDecompressEnd(&bzstream);
1254  goto failed;
1255  }
1256  pkt_data = newpktdata;
1257  bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1258  bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1259  if (pkt_data) {
1260  result = BZ2_bzDecompress(&bzstream);
1261  } else
1262  result = BZ_MEM_ERROR;
1263  } while (result == BZ_OK && pkt_size < 10000000);
1264  pkt_size = bzstream.total_out_lo32;
1265  BZ2_bzDecompressEnd(&bzstream);
1266  if (result != BZ_STREAM_END) {
1267  if (result == BZ_MEM_ERROR)
1268  result = AVERROR(ENOMEM);
1269  else
1270  result = AVERROR_INVALIDDATA;
1271  goto failed;
1272  }
1273  break;
1274  }
1275 #endif
1276  default:
1277  return AVERROR_INVALIDDATA;
1278  }
1279 
1280  *buf = pkt_data;
1281  *buf_size = pkt_size;
1282  return 0;
1283 
1284 failed:
1285  av_free(pkt_data);
1286  return result;
1287 }
1288 
1290  AVDictionary **metadata, char *prefix)
1291 {
1292  MatroskaTag *tags = list->elem;
1293  char key[1024];
1294  int i;
1295 
1296  for (i = 0; i < list->nb_elem; i++) {
1297  const char *lang = tags[i].lang &&
1298  strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1299 
1300  if (!tags[i].name) {
1301  av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1302  continue;
1303  }
1304  if (prefix)
1305  snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1306  else
1307  av_strlcpy(key, tags[i].name, sizeof(key));
1308  if (tags[i].def || !lang) {
1309  av_dict_set(metadata, key, tags[i].string, 0);
1310  if (tags[i].sub.nb_elem)
1311  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1312  }
1313  if (lang) {
1314  av_strlcat(key, "-", sizeof(key));
1315  av_strlcat(key, lang, sizeof(key));
1316  av_dict_set(metadata, key, tags[i].string, 0);
1317  if (tags[i].sub.nb_elem)
1318  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1319  }
1320  }
1321  ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1322 }
1323 
1325 {
1326  MatroskaDemuxContext *matroska = s->priv_data;
1327  MatroskaTags *tags = matroska->tags.elem;
1328  int i, j;
1329 
1330  for (i = 0; i < matroska->tags.nb_elem; i++) {
1331  if (tags[i].target.attachuid) {
1332  MatroskaAttachment *attachment = matroska->attachments.elem;
1333  for (j = 0; j < matroska->attachments.nb_elem; j++)
1334  if (attachment[j].uid == tags[i].target.attachuid &&
1335  attachment[j].stream)
1336  matroska_convert_tag(s, &tags[i].tag,
1337  &attachment[j].stream->metadata, NULL);
1338  } else if (tags[i].target.chapteruid) {
1339  MatroskaChapter *chapter = matroska->chapters.elem;
1340  for (j = 0; j < matroska->chapters.nb_elem; j++)
1341  if (chapter[j].uid == tags[i].target.chapteruid &&
1342  chapter[j].chapter)
1343  matroska_convert_tag(s, &tags[i].tag,
1344  &chapter[j].chapter->metadata, NULL);
1345  } else if (tags[i].target.trackuid) {
1346  MatroskaTrack *track = matroska->tracks.elem;
1347  for (j = 0; j < matroska->tracks.nb_elem; j++)
1348  if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1349  matroska_convert_tag(s, &tags[i].tag,
1350  &track[j].stream->metadata, NULL);
1351  } else {
1352  matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1353  tags[i].target.type);
1354  }
1355  }
1356 }
1357 
1359  int idx)
1360 {
1361  EbmlList *seekhead_list = &matroska->seekhead;
1362  uint32_t level_up = matroska->level_up;
1363  uint32_t saved_id = matroska->current_id;
1364  MatroskaSeekhead *seekhead = seekhead_list->elem;
1365  int64_t before_pos = avio_tell(matroska->ctx->pb);
1367  int64_t offset;
1368  int ret = 0;
1369 
1370  if (idx >= seekhead_list->nb_elem ||
1371  seekhead[idx].id == MATROSKA_ID_SEEKHEAD ||
1372  seekhead[idx].id == MATROSKA_ID_CLUSTER)
1373  return 0;
1374 
1375  /* seek */
1376  offset = seekhead[idx].pos + matroska->segment_start;
1377  if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1378  /* We don't want to lose our seekhead level, so we add
1379  * a dummy. This is a crude hack. */
1380  if (matroska->num_levels == EBML_MAX_DEPTH) {
1381  av_log(matroska->ctx, AV_LOG_INFO,
1382  "Max EBML element depth (%d) reached, "
1383  "cannot parse further.\n", EBML_MAX_DEPTH);
1384  ret = AVERROR_INVALIDDATA;
1385  } else {
1386  level.start = 0;
1387  level.length = (uint64_t) -1;
1388  matroska->levels[matroska->num_levels] = level;
1389  matroska->num_levels++;
1390  matroska->current_id = 0;
1391 
1392  ret = ebml_parse(matroska, matroska_segment, matroska);
1393 
1394  /* remove dummy level */
1395  while (matroska->num_levels) {
1396  uint64_t length = matroska->levels[--matroska->num_levels].length;
1397  if (length == (uint64_t) -1)
1398  break;
1399  }
1400  }
1401  }
1402  /* seek back */
1403  avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1404  matroska->level_up = level_up;
1405  matroska->current_id = saved_id;
1406 
1407  return ret;
1408 }
1409 
1411 {
1412  EbmlList *seekhead_list = &matroska->seekhead;
1413  int64_t before_pos = avio_tell(matroska->ctx->pb);
1414  int i;
1415 
1416  // we should not do any seeking in the streaming case
1417  if (!matroska->ctx->pb->seekable ||
1418  (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1419  return;
1420 
1421  for (i = 0; i < seekhead_list->nb_elem; i++) {
1422  MatroskaSeekhead *seekhead = seekhead_list->elem;
1423  if (seekhead[i].pos <= before_pos)
1424  continue;
1425 
1426  // defer cues parsing until we actually need cue data.
1427  if (seekhead[i].id == MATROSKA_ID_CUES) {
1428  matroska->cues_parsing_deferred = 1;
1429  continue;
1430  }
1431 
1432  if (matroska_parse_seekhead_entry(matroska, i) < 0) {
1433  // mark index as broken
1434  matroska->cues_parsing_deferred = -1;
1435  break;
1436  }
1437  }
1438 }
1439 
1441 {
1442  EbmlList *index_list;
1444  int index_scale = 1;
1445  int i, j;
1446 
1447  index_list = &matroska->index;
1448  index = index_list->elem;
1449  if (index_list->nb_elem &&
1450  index[0].time > 1E14 / matroska->time_scale) {
1451  av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1452  index_scale = matroska->time_scale;
1453  }
1454  for (i = 0; i < index_list->nb_elem; i++) {
1455  EbmlList *pos_list = &index[i].pos;
1456  MatroskaIndexPos *pos = pos_list->elem;
1457  for (j = 0; j < pos_list->nb_elem; j++) {
1458  MatroskaTrack *track = matroska_find_track_by_num(matroska,
1459  pos[j].track);
1460  if (track && track->stream)
1461  av_add_index_entry(track->stream,
1462  pos[j].pos + matroska->segment_start,
1463  index[i].time / index_scale, 0, 0,
1465  }
1466  }
1467 }
1468 
1470  EbmlList *seekhead_list = &matroska->seekhead;
1471  MatroskaSeekhead *seekhead = seekhead_list->elem;
1472  int i;
1473 
1474  for (i = 0; i < seekhead_list->nb_elem; i++)
1475  if (seekhead[i].id == MATROSKA_ID_CUES)
1476  break;
1477  av_assert1(i <= seekhead_list->nb_elem);
1478 
1479  if (matroska_parse_seekhead_entry(matroska, i) < 0)
1480  matroska->cues_parsing_deferred = -1;
1481  matroska_add_index_entries(matroska);
1482 }
1483 
1485 {
1486  static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1487  int profile;
1488 
1489  for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1490  if (strstr(codec_id, aac_profiles[profile]))
1491  break;
1492  return profile + 1;
1493 }
1494 
1495 static int matroska_aac_sri(int samplerate)
1496 {
1497  int sri;
1498 
1499  for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1500  if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1501  break;
1502  return sri;
1503 }
1504 
1505 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1506 {
1507  char buffer[32];
1508  /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1509  time_t creation_time = date_utc / 1000000000 + 978307200;
1510  struct tm *ptm = gmtime(&creation_time);
1511  if (!ptm) return;
1512  strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm);
1513  av_dict_set(metadata, "creation_time", buffer, 0);
1514 }
1515 
1517  MatroskaTrack *track,
1518  int *offset)
1519 {
1520  AVStream *st = track->stream;
1521  uint8_t *p = track->codec_priv.data;
1522  int size = track->codec_priv.size;
1523 
1524  if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1525  av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1526  track->codec_priv.size = 0;
1527  return 0;
1528  }
1529  *offset = 8;
1530  track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1531 
1532  p += track->codec_priv.size;
1533  size -= track->codec_priv.size;
1534 
1535  /* parse the remaining metadata blocks if present */
1536  while (size >= 4) {
1537  int block_last, block_type, block_size;
1538 
1539  flac_parse_block_header(p, &block_last, &block_type, &block_size);
1540 
1541  p += 4;
1542  size -= 4;
1543  if (block_size > size)
1544  return 0;
1545 
1546  /* check for the channel mask */
1547  if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1548  AVDictionary *dict = NULL;
1549  AVDictionaryEntry *chmask;
1550 
1551  ff_vorbis_comment(s, &dict, p, block_size, 0);
1552  chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1553  if (chmask) {
1554  uint64_t mask = strtol(chmask->value, NULL, 0);
1555  if (!mask || mask & ~0x3ffffULL) {
1557  "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1558  } else
1559  st->codec->channel_layout = mask;
1560  }
1561  av_dict_free(&dict);
1562  }
1563 
1564  p += block_size;
1565  size -= block_size;
1566  }
1567 
1568  return 0;
1569 }
1570 
1572 {
1573  MatroskaDemuxContext *matroska = s->priv_data;
1574  MatroskaTrack *tracks = matroska->tracks.elem;
1575  AVStream *st;
1576  int i, j, ret;
1577  int k;
1578 
1579  for (i = 0; i < matroska->tracks.nb_elem; i++) {
1580  MatroskaTrack *track = &tracks[i];
1582  EbmlList *encodings_list = &track->encodings;
1583  MatroskaTrackEncoding *encodings = encodings_list->elem;
1584  uint8_t *extradata = NULL;
1585  int extradata_size = 0;
1586  int extradata_offset = 0;
1587  uint32_t fourcc = 0;
1588  AVIOContext b;
1589  char* key_id_base64 = NULL;
1590  int bit_depth = -1;
1591 
1592  /* Apply some sanity checks. */
1593  if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1594  track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1595  track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1596  track->type != MATROSKA_TRACK_TYPE_METADATA) {
1597  av_log(matroska->ctx, AV_LOG_INFO,
1598  "Unknown or unsupported track type %"PRIu64"\n",
1599  track->type);
1600  continue;
1601  }
1602  if (!track->codec_id)
1603  continue;
1604 
1605  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1606  if (!track->default_duration && track->video.frame_rate > 0)
1607  track->default_duration = 1000000000 / track->video.frame_rate;
1608  if (track->video.display_width == -1)
1609  track->video.display_width = track->video.pixel_width;
1610  if (track->video.display_height == -1)
1611  track->video.display_height = track->video.pixel_height;
1612  if (track->video.color_space.size == 4)
1613  fourcc = AV_RL32(track->video.color_space.data);
1614  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1615  if (!track->audio.out_samplerate)
1616  track->audio.out_samplerate = track->audio.samplerate;
1617  }
1618  if (encodings_list->nb_elem > 1) {
1619  av_log(matroska->ctx, AV_LOG_ERROR,
1620  "Multiple combined encodings not supported");
1621  } else if (encodings_list->nb_elem == 1) {
1622  if (encodings[0].type) {
1623  if (encodings[0].encryption.key_id.size > 0) {
1624  /* Save the encryption key id to be stored later as a
1625  metadata tag. */
1626  const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1627  key_id_base64 = av_malloc(b64_size);
1628  if (key_id_base64 == NULL)
1629  return AVERROR(ENOMEM);
1630 
1631  av_base64_encode(key_id_base64, b64_size,
1632  encodings[0].encryption.key_id.data,
1633  encodings[0].encryption.key_id.size);
1634  } else {
1635  encodings[0].scope = 0;
1636  av_log(matroska->ctx, AV_LOG_ERROR,
1637  "Unsupported encoding type");
1638  }
1639  } else if (
1640 #if CONFIG_ZLIB
1641  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1642 #endif
1643 #if CONFIG_BZLIB
1645 #endif
1646 #if CONFIG_LZO
1648 #endif
1650  encodings[0].scope = 0;
1651  av_log(matroska->ctx, AV_LOG_ERROR,
1652  "Unsupported encoding type");
1653  } else if (track->codec_priv.size && encodings[0].scope & 2) {
1654  uint8_t *codec_priv = track->codec_priv.data;
1655  int ret = matroska_decode_buffer(&track->codec_priv.data,
1656  &track->codec_priv.size,
1657  track);
1658  if (ret < 0) {
1659  track->codec_priv.data = NULL;
1660  track->codec_priv.size = 0;
1661  av_log(matroska->ctx, AV_LOG_ERROR,
1662  "Failed to decode codec private data\n");
1663  }
1664 
1665  if (codec_priv != track->codec_priv.data)
1666  av_free(codec_priv);
1667  }
1668  }
1669 
1670  for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
1671  if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1672  strlen(ff_mkv_codec_tags[j].str))) {
1673  codec_id = ff_mkv_codec_tags[j].id;
1674  break;
1675  }
1676  }
1677 
1678  st = track->stream = avformat_new_stream(s, NULL);
1679  if (!st) {
1680  av_free(key_id_base64);
1681  return AVERROR(ENOMEM);
1682  }
1683 
1684  if (key_id_base64) {
1685  /* export encryption key id as base64 metadata tag */
1686  av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
1687  av_freep(&key_id_base64);
1688  }
1689 
1690  if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
1691  track->codec_priv.size >= 40 &&
1692  track->codec_priv.data) {
1693  track->ms_compat = 1;
1694  bit_depth = AV_RL16(track->codec_priv.data + 14);
1695  fourcc = AV_RL32(track->codec_priv.data + 16);
1697  fourcc);
1698  if (!codec_id)
1700  fourcc);
1701  extradata_offset = 40;
1702  } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
1703  track->codec_priv.size >= 14 &&
1704  track->codec_priv.data) {
1705  int ret;
1706  ffio_init_context(&b, track->codec_priv.data,
1707  track->codec_priv.size,
1708  0, NULL, NULL, NULL, NULL);
1709  ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1710  if (ret < 0)
1711  return ret;
1712  codec_id = st->codec->codec_id;
1713  extradata_offset = FFMIN(track->codec_priv.size, 18);
1714  } else if (!strcmp(track->codec_id, "A_QUICKTIME")
1715  && (track->codec_priv.size >= 86)
1716  && (track->codec_priv.data)) {
1717  fourcc = AV_RL32(track->codec_priv.data + 4);
1718  codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1720  fourcc = AV_RL32(track->codec_priv.data);
1721  codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1722  }
1723  } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
1724  (track->codec_priv.size >= 21) &&
1725  (track->codec_priv.data)) {
1726  fourcc = AV_RL32(track->codec_priv.data + 4);
1727  codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1729  fourcc = AV_RL32(track->codec_priv.data);
1730  codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1731  }
1732  if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI "))
1733  codec_id = AV_CODEC_ID_SVQ3;
1734  } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1735  switch (track->audio.bitdepth) {
1736  case 8:
1737  codec_id = AV_CODEC_ID_PCM_U8;
1738  break;
1739  case 24:
1740  codec_id = AV_CODEC_ID_PCM_S24BE;
1741  break;
1742  case 32:
1743  codec_id = AV_CODEC_ID_PCM_S32BE;
1744  break;
1745  }
1746  } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1747  switch (track->audio.bitdepth) {
1748  case 8:
1749  codec_id = AV_CODEC_ID_PCM_U8;
1750  break;
1751  case 24:
1752  codec_id = AV_CODEC_ID_PCM_S24LE;
1753  break;
1754  case 32:
1755  codec_id = AV_CODEC_ID_PCM_S32LE;
1756  break;
1757  }
1758  } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
1759  track->audio.bitdepth == 64) {
1760  codec_id = AV_CODEC_ID_PCM_F64LE;
1761  } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1762  int profile = matroska_aac_profile(track->codec_id);
1763  int sri = matroska_aac_sri(track->audio.samplerate);
1764  extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1765  if (!extradata)
1766  return AVERROR(ENOMEM);
1767  extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
1768  extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
1769  if (strstr(track->codec_id, "SBR")) {
1770  sri = matroska_aac_sri(track->audio.out_samplerate);
1771  extradata[2] = 0x56;
1772  extradata[3] = 0xE5;
1773  extradata[4] = 0x80 | (sri << 3);
1774  extradata_size = 5;
1775  } else
1776  extradata_size = 2;
1777  } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - FF_INPUT_BUFFER_PADDING_SIZE) {
1778  /* Only ALAC's magic cookie is stored in Matroska's track headers.
1779  * Create the "atom size", "tag", and "tag version" fields the
1780  * decoder expects manually. */
1781  extradata_size = 12 + track->codec_priv.size;
1782  extradata = av_mallocz(extradata_size +
1784  if (!extradata)
1785  return AVERROR(ENOMEM);
1786  AV_WB32(extradata, extradata_size);
1787  memcpy(&extradata[4], "alac", 4);
1788  AV_WB32(&extradata[8], 0);
1789  memcpy(&extradata[12], track->codec_priv.data,
1790  track->codec_priv.size);
1791  } else if (codec_id == AV_CODEC_ID_TTA) {
1792  extradata_size = 30;
1793  extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1794  if (!extradata)
1795  return AVERROR(ENOMEM);
1796  ffio_init_context(&b, extradata, extradata_size, 1,
1797  NULL, NULL, NULL, NULL);
1798  avio_write(&b, "TTA1", 4);
1799  avio_wl16(&b, 1);
1800  avio_wl16(&b, track->audio.channels);
1801  avio_wl16(&b, track->audio.bitdepth);
1802  if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
1803  return AVERROR_INVALIDDATA;
1804  avio_wl32(&b, track->audio.out_samplerate);
1805  avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
1806  track->audio.out_samplerate,
1807  AV_TIME_BASE * 1000));
1808  } else if (codec_id == AV_CODEC_ID_RV10 ||
1809  codec_id == AV_CODEC_ID_RV20 ||
1810  codec_id == AV_CODEC_ID_RV30 ||
1811  codec_id == AV_CODEC_ID_RV40) {
1812  extradata_offset = 26;
1813  } else if (codec_id == AV_CODEC_ID_RA_144) {
1814  track->audio.out_samplerate = 8000;
1815  track->audio.channels = 1;
1816  } else if ((codec_id == AV_CODEC_ID_RA_288 ||
1817  codec_id == AV_CODEC_ID_COOK ||
1818  codec_id == AV_CODEC_ID_ATRAC3 ||
1819  codec_id == AV_CODEC_ID_SIPR)
1820  && track->codec_priv.data) {
1821  int flavor;
1822 
1823  ffio_init_context(&b, track->codec_priv.data,
1824  track->codec_priv.size,
1825  0, NULL, NULL, NULL, NULL);
1826  avio_skip(&b, 22);
1827  flavor = avio_rb16(&b);
1828  track->audio.coded_framesize = avio_rb32(&b);
1829  avio_skip(&b, 12);
1830  track->audio.sub_packet_h = avio_rb16(&b);
1831  track->audio.frame_size = avio_rb16(&b);
1832  track->audio.sub_packet_size = avio_rb16(&b);
1833  if (flavor < 0 ||
1834  track->audio.coded_framesize <= 0 ||
1835  track->audio.sub_packet_h <= 0 ||
1836  track->audio.frame_size <= 0 ||
1837  track->audio.sub_packet_size <= 0)
1838  return AVERROR_INVALIDDATA;
1839  track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
1840  track->audio.frame_size);
1841  if (!track->audio.buf)
1842  return AVERROR(ENOMEM);
1843  if (codec_id == AV_CODEC_ID_RA_288) {
1844  st->codec->block_align = track->audio.coded_framesize;
1845  track->codec_priv.size = 0;
1846  } else {
1847  if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1848  static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1849  track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1850  st->codec->bit_rate = sipr_bit_rate[flavor];
1851  }
1852  st->codec->block_align = track->audio.sub_packet_size;
1853  extradata_offset = 78;
1854  }
1855  } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
1856  ret = matroska_parse_flac(s, track, &extradata_offset);
1857  if (ret < 0)
1858  return ret;
1859  } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
1860  fourcc = AV_RL32(track->codec_priv.data);
1861  }
1862  track->codec_priv.size -= extradata_offset;
1863 
1864  if (codec_id == AV_CODEC_ID_NONE)
1865  av_log(matroska->ctx, AV_LOG_INFO,
1866  "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1867 
1868  if (track->time_scale < 0.01)
1869  track->time_scale = 1.0;
1870  avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
1871  1000 * 1000 * 1000); /* 64 bit pts in ns */
1872 
1873  /* convert the delay from ns to the track timebase */
1874  track->codec_delay = av_rescale_q(track->codec_delay,
1875  (AVRational){ 1, 1000000000 },
1876  st->time_base);
1877 
1878  st->codec->codec_id = codec_id;
1879 
1880  if (strcmp(track->language, "und"))
1881  av_dict_set(&st->metadata, "language", track->language, 0);
1882  av_dict_set(&st->metadata, "title", track->name, 0);
1883 
1884  if (track->flag_default)
1886  if (track->flag_forced)
1888 
1889  if (!st->codec->extradata) {
1890  if (extradata) {
1891  st->codec->extradata = extradata;
1892  st->codec->extradata_size = extradata_size;
1893  } else if (track->codec_priv.data && track->codec_priv.size > 0) {
1894  if (ff_alloc_extradata(st->codec, track->codec_priv.size))
1895  return AVERROR(ENOMEM);
1896  memcpy(st->codec->extradata,
1897  track->codec_priv.data + extradata_offset,
1898  track->codec_priv.size);
1899  }
1900  }
1901 
1902  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1903  MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
1904 
1906  st->codec->codec_tag = fourcc;
1907  if (bit_depth >= 0)
1908  st->codec->bits_per_coded_sample = bit_depth;
1909  st->codec->width = track->video.pixel_width;
1910  st->codec->height = track->video.pixel_height;
1912  &st->sample_aspect_ratio.den,
1913  st->codec->height * track->video.display_width,
1914  st->codec->width * track->video.display_height,
1915  255);
1916  if (st->codec->codec_id != AV_CODEC_ID_HEVC)
1918 
1919  if (track->default_duration) {
1921  1000000000, track->default_duration, 30000);
1922 #if FF_API_R_FRAME_RATE
1923  if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000L
1924  && st->avg_frame_rate.num > st->avg_frame_rate.den * 5L)
1925  st->r_frame_rate = st->avg_frame_rate;
1926 #endif
1927  }
1928 
1929  /* export stereo mode flag as metadata tag */
1930  if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
1931  av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
1932 
1933  /* export alpha mode flag as metadata tag */
1934  if (track->video.alpha_mode)
1935  av_dict_set(&st->metadata, "alpha_mode", "1", 0);
1936 
1937  /* if we have virtual track, mark the real tracks */
1938  for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
1939  char buf[32];
1940  if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
1941  continue;
1942  snprintf(buf, sizeof(buf), "%s_%d",
1943  ff_matroska_video_stereo_plane[planes[j].type], i);
1944  for (k=0; k < matroska->tracks.nb_elem; k++)
1945  if (planes[j].uid == tracks[k].uid) {
1946  av_dict_set(&s->streams[k]->metadata,
1947  "stereo_mode", buf, 0);
1948  break;
1949  }
1950  }
1951  // add stream level stereo3d side data if it is a supported format
1952  if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
1953  track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
1954  int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
1955  if (ret < 0)
1956  return ret;
1957  }
1958  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1960  st->codec->sample_rate = track->audio.out_samplerate;
1961  st->codec->channels = track->audio.channels;
1962  if (!st->codec->bits_per_coded_sample)
1963  st->codec->bits_per_coded_sample = track->audio.bitdepth;
1964  if (st->codec->codec_id != AV_CODEC_ID_AAC)
1966  if (track->codec_delay > 0) {
1967  st->codec->delay = av_rescale_q(track->codec_delay,
1968  st->time_base,
1969  (AVRational){1, st->codec->sample_rate});
1970  }
1971  if (track->seek_preroll > 0) {
1973  av_rescale_q(track->seek_preroll,
1974  (AVRational){1, 1000000000},
1975  (AVRational){1, st->codec->sample_rate}));
1976  }
1977  } else if (codec_id == AV_CODEC_ID_WEBVTT) {
1978  st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1979 
1980  if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
1981  st->disposition |= AV_DISPOSITION_CAPTIONS;
1982  } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
1983  st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
1984  } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
1985  st->disposition |= AV_DISPOSITION_METADATA;
1986  }
1987  } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1988  st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1989  if (st->codec->codec_id == AV_CODEC_ID_ASS)
1990  matroska->contains_ssa = 1;
1991  }
1992  }
1993 
1994  return 0;
1995 }
1996 
1998 {
1999  MatroskaDemuxContext *matroska = s->priv_data;
2000  EbmlList *attachments_list = &matroska->attachments;
2001  EbmlList *chapters_list = &matroska->chapters;
2002  MatroskaAttachment *attachments;
2003  MatroskaChapter *chapters;
2004  uint64_t max_start = 0;
2005  int64_t pos;
2006  Ebml ebml = { 0 };
2007  int i, j, res;
2008 
2009  matroska->ctx = s;
2010 
2011  /* First read the EBML header. */
2012  if (ebml_parse(matroska, ebml_syntax, &ebml) ||
2013  ebml.version > EBML_VERSION ||
2014  ebml.max_size > sizeof(uint64_t) ||
2015  ebml.id_length > sizeof(uint32_t) ||
2016  ebml.doctype_version > 3 ||
2017  !ebml.doctype) {
2018  av_log(matroska->ctx, AV_LOG_ERROR,
2019  "EBML header using unsupported features\n"
2020  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2021  ebml.version, ebml.doctype, ebml.doctype_version);
2022  ebml_free(ebml_syntax, &ebml);
2023  return AVERROR_PATCHWELCOME;
2024  } else if (ebml.doctype_version == 3) {
2025  av_log(matroska->ctx, AV_LOG_WARNING,
2026  "EBML header using unsupported features\n"
2027  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2028  ebml.version, ebml.doctype, ebml.doctype_version);
2029  }
2030  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2031  if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2032  break;
2033  if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2034  av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2035  if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2036  ebml_free(ebml_syntax, &ebml);
2037  return AVERROR_INVALIDDATA;
2038  }
2039  }
2040  ebml_free(ebml_syntax, &ebml);
2041 
2042  /* The next thing is a segment. */
2043  pos = avio_tell(matroska->ctx->pb);
2044  res = ebml_parse(matroska, matroska_segments, matroska);
2045  // try resyncing until we find a EBML_STOP type element.
2046  while (res != 1) {
2047  res = matroska_resync(matroska, pos);
2048  if (res < 0)
2049  return res;
2050  pos = avio_tell(matroska->ctx->pb);
2051  res = ebml_parse(matroska, matroska_segment, matroska);
2052  }
2053  matroska_execute_seekhead(matroska);
2054 
2055  if (!matroska->time_scale)
2056  matroska->time_scale = 1000000;
2057  if (matroska->duration)
2058  matroska->ctx->duration = matroska->duration * matroska->time_scale *
2059  1000 / AV_TIME_BASE;
2060  av_dict_set(&s->metadata, "title", matroska->title, 0);
2061  av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2062 
2063  if (matroska->date_utc.size == 8)
2065 
2066  res = matroska_parse_tracks(s);
2067  if (res < 0)
2068  return res;
2069 
2070  attachments = attachments_list->elem;
2071  for (j = 0; j < attachments_list->nb_elem; j++) {
2072  if (!(attachments[j].filename && attachments[j].mime &&
2073  attachments[j].bin.data && attachments[j].bin.size > 0)) {
2074  av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2075  } else {
2076  AVStream *st = avformat_new_stream(s, NULL);
2077  if (!st)
2078  break;
2079  av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2080  av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2083  if (ff_alloc_extradata(st->codec, attachments[j].bin.size))
2084  break;
2085  memcpy(st->codec->extradata, attachments[j].bin.data,
2086  attachments[j].bin.size);
2087 
2088  for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2089  if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2090  strlen(ff_mkv_mime_tags[i].str))) {
2091  st->codec->codec_id = ff_mkv_mime_tags[i].id;
2092  break;
2093  }
2094  }
2095  attachments[j].stream = st;
2096  }
2097  }
2098 
2099  chapters = chapters_list->elem;
2100  for (i = 0; i < chapters_list->nb_elem; i++)
2101  if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2102  (max_start == 0 || chapters[i].start > max_start)) {
2103  chapters[i].chapter =
2104  avpriv_new_chapter(s, chapters[i].uid,
2105  (AVRational) { 1, 1000000000 },
2106  chapters[i].start, chapters[i].end,
2107  chapters[i].title);
2108  if (chapters[i].chapter) {
2109  av_dict_set(&chapters[i].chapter->metadata,
2110  "title", chapters[i].title, 0);
2111  }
2112  max_start = chapters[i].start;
2113  }
2114 
2115  matroska_add_index_entries(matroska);
2116 
2118 
2119  return 0;
2120 }
2121 
2122 /*
2123  * Put one packet in an application-supplied AVPacket struct.
2124  * Returns 0 on success or -1 on failure.
2125  */
2127  AVPacket *pkt)
2128 {
2129  if (matroska->num_packets > 0) {
2130  memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2131  av_free(matroska->packets[0]);
2132  if (matroska->num_packets > 1) {
2133  void *newpackets;
2134  memmove(&matroska->packets[0], &matroska->packets[1],
2135  (matroska->num_packets - 1) * sizeof(AVPacket *));
2136  newpackets = av_realloc(matroska->packets,
2137  (matroska->num_packets - 1) *
2138  sizeof(AVPacket *));
2139  if (newpackets)
2140  matroska->packets = newpackets;
2141  } else {
2142  av_freep(&matroska->packets);
2143  matroska->prev_pkt = NULL;
2144  }
2145  matroska->num_packets--;
2146  return 0;
2147  }
2148 
2149  return -1;
2150 }
2151 
2152 /*
2153  * Free all packets in our internal queue.
2154  */
2156 {
2157  matroska->prev_pkt = NULL;
2158  if (matroska->packets) {
2159  int n;
2160  for (n = 0; n < matroska->num_packets; n++) {
2161  av_free_packet(matroska->packets[n]);
2162  av_free(matroska->packets[n]);
2163  }
2164  av_freep(&matroska->packets);
2165  matroska->num_packets = 0;
2166  }
2167 }
2168 
2170  int *buf_size, int type,
2171  uint32_t **lace_buf, int *laces)
2172 {
2173  int res = 0, n, size = *buf_size;
2174  uint8_t *data = *buf;
2175  uint32_t *lace_size;
2176 
2177  if (!type) {
2178  *laces = 1;
2179  *lace_buf = av_mallocz(sizeof(int));
2180  if (!*lace_buf)
2181  return AVERROR(ENOMEM);
2182 
2183  *lace_buf[0] = size;
2184  return 0;
2185  }
2186 
2187  av_assert0(size > 0);
2188  *laces = *data + 1;
2189  data += 1;
2190  size -= 1;
2191  lace_size = av_mallocz(*laces * sizeof(int));
2192  if (!lace_size)
2193  return AVERROR(ENOMEM);
2194 
2195  switch (type) {
2196  case 0x1: /* Xiph lacing */
2197  {
2198  uint8_t temp;
2199  uint32_t total = 0;
2200  for (n = 0; res == 0 && n < *laces - 1; n++) {
2201  while (1) {
2202  if (size <= total) {
2203  res = AVERROR_INVALIDDATA;
2204  break;
2205  }
2206  temp = *data;
2207  total += temp;
2208  lace_size[n] += temp;
2209  data += 1;
2210  size -= 1;
2211  if (temp != 0xff)
2212  break;
2213  }
2214  }
2215  if (size <= total) {
2216  res = AVERROR_INVALIDDATA;
2217  break;
2218  }
2219 
2220  lace_size[n] = size - total;
2221  break;
2222  }
2223 
2224  case 0x2: /* fixed-size lacing */
2225  if (size % (*laces)) {
2226  res = AVERROR_INVALIDDATA;
2227  break;
2228  }
2229  for (n = 0; n < *laces; n++)
2230  lace_size[n] = size / *laces;
2231  break;
2232 
2233  case 0x3: /* EBML lacing */
2234  {
2235  uint64_t num;
2236  uint64_t total;
2237  n = matroska_ebmlnum_uint(matroska, data, size, &num);
2238  if (n < 0 || num > INT_MAX) {
2239  av_log(matroska->ctx, AV_LOG_INFO,
2240  "EBML block data error\n");
2241  res = n<0 ? n : AVERROR_INVALIDDATA;
2242  break;
2243  }
2244  data += n;
2245  size -= n;
2246  total = lace_size[0] = num;
2247  for (n = 1; res == 0 && n < *laces - 1; n++) {
2248  int64_t snum;
2249  int r;
2250  r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2251  if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2252  av_log(matroska->ctx, AV_LOG_INFO,
2253  "EBML block data error\n");
2254  res = r<0 ? r : AVERROR_INVALIDDATA;
2255  break;
2256  }
2257  data += r;
2258  size -= r;
2259  lace_size[n] = lace_size[n - 1] + snum;
2260  total += lace_size[n];
2261  }
2262  if (size <= total) {
2263  res = AVERROR_INVALIDDATA;
2264  break;
2265  }
2266  lace_size[*laces - 1] = size - total;
2267  break;
2268  }
2269  }
2270 
2271  *buf = data;
2272  *lace_buf = lace_size;
2273  *buf_size = size;
2274 
2275  return res;
2276 }
2277 
2279  MatroskaTrack *track, AVStream *st,
2280  uint8_t *data, int size, uint64_t timecode,
2281  int64_t pos)
2282 {
2283  int a = st->codec->block_align;
2284  int sps = track->audio.sub_packet_size;
2285  int cfs = track->audio.coded_framesize;
2286  int h = track->audio.sub_packet_h;
2287  int y = track->audio.sub_packet_cnt;
2288  int w = track->audio.frame_size;
2289  int x;
2290 
2291  if (!track->audio.pkt_cnt) {
2292  if (track->audio.sub_packet_cnt == 0)
2293  track->audio.buf_timecode = timecode;
2294  if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2295  if (size < cfs * h / 2) {
2296  av_log(matroska->ctx, AV_LOG_ERROR,
2297  "Corrupt int4 RM-style audio packet size\n");
2298  return AVERROR_INVALIDDATA;
2299  }
2300  for (x = 0; x < h / 2; x++)
2301  memcpy(track->audio.buf + x * 2 * w + y * cfs,
2302  data + x * cfs, cfs);
2303  } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2304  if (size < w) {
2305  av_log(matroska->ctx, AV_LOG_ERROR,
2306  "Corrupt sipr RM-style audio packet size\n");
2307  return AVERROR_INVALIDDATA;
2308  }
2309  memcpy(track->audio.buf + y * w, data, w);
2310  } else {
2311  if (size < sps * w / sps || h<=0 || w%sps) {
2312  av_log(matroska->ctx, AV_LOG_ERROR,
2313  "Corrupt generic RM-style audio packet size\n");
2314  return AVERROR_INVALIDDATA;
2315  }
2316  for (x = 0; x < w / sps; x++)
2317  memcpy(track->audio.buf +
2318  sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2319  data + x * sps, sps);
2320  }
2321 
2322  if (++track->audio.sub_packet_cnt >= h) {
2323  if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2324  ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2325  track->audio.sub_packet_cnt = 0;
2326  track->audio.pkt_cnt = h * w / a;
2327  }
2328  }
2329 
2330  while (track->audio.pkt_cnt) {
2331  int ret;
2332  AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2333  if (!pkt)
2334  return AVERROR(ENOMEM);
2335 
2336  ret = av_new_packet(pkt, a);
2337  if (ret < 0) {
2338  av_free(pkt);
2339  return ret;
2340  }
2341  memcpy(pkt->data,
2342  track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2343  a);
2344  pkt->pts = track->audio.buf_timecode;
2346  pkt->pos = pos;
2347  pkt->stream_index = st->index;
2348  dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2349  }
2350 
2351  return 0;
2352 }
2353 
2354 /* reconstruct full wavpack blocks from mangled matroska ones */
2356  uint8_t **pdst, int *size)
2357 {
2358  uint8_t *dst = NULL;
2359  int dstlen = 0;
2360  int srclen = *size;
2361  uint32_t samples;
2362  uint16_t ver;
2363  int ret, offset = 0;
2364 
2365  if (srclen < 12 || track->stream->codec->extradata_size < 2)
2366  return AVERROR_INVALIDDATA;
2367 
2368  ver = AV_RL16(track->stream->codec->extradata);
2369 
2370  samples = AV_RL32(src);
2371  src += 4;
2372  srclen -= 4;
2373 
2374  while (srclen >= 8) {
2375  int multiblock;
2376  uint32_t blocksize;
2377  uint8_t *tmp;
2378 
2379  uint32_t flags = AV_RL32(src);
2380  uint32_t crc = AV_RL32(src + 4);
2381  src += 8;
2382  srclen -= 8;
2383 
2384  multiblock = (flags & 0x1800) != 0x1800;
2385  if (multiblock) {
2386  if (srclen < 4) {
2387  ret = AVERROR_INVALIDDATA;
2388  goto fail;
2389  }
2390  blocksize = AV_RL32(src);
2391  src += 4;
2392  srclen -= 4;
2393  } else
2394  blocksize = srclen;
2395 
2396  if (blocksize > srclen) {
2397  ret = AVERROR_INVALIDDATA;
2398  goto fail;
2399  }
2400 
2401  tmp = av_realloc(dst, dstlen + blocksize + 32);
2402  if (!tmp) {
2403  ret = AVERROR(ENOMEM);
2404  goto fail;
2405  }
2406  dst = tmp;
2407  dstlen += blocksize + 32;
2408 
2409  AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2410  AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2411  AV_WL16(dst + offset + 8, ver); // version
2412  AV_WL16(dst + offset + 10, 0); // track/index_no
2413  AV_WL32(dst + offset + 12, 0); // total samples
2414  AV_WL32(dst + offset + 16, 0); // block index
2415  AV_WL32(dst + offset + 20, samples); // number of samples
2416  AV_WL32(dst + offset + 24, flags); // flags
2417  AV_WL32(dst + offset + 28, crc); // crc
2418  memcpy(dst + offset + 32, src, blocksize); // block data
2419 
2420  src += blocksize;
2421  srclen -= blocksize;
2422  offset += blocksize + 32;
2423  }
2424 
2425  *pdst = dst;
2426  *size = dstlen;
2427 
2428  return 0;
2429 
2430 fail:
2431  av_freep(&dst);
2432  return ret;
2433 }
2434 
2436  MatroskaTrack *track,
2437  AVStream *st,
2438  uint8_t *data, int data_len,
2439  uint64_t timecode,
2440  uint64_t duration,
2441  int64_t pos)
2442 {
2443  AVPacket *pkt;
2444  uint8_t *id, *settings, *text, *buf;
2445  int id_len, settings_len, text_len;
2446  uint8_t *p, *q;
2447  int err;
2448 
2449  if (data_len <= 0)
2450  return AVERROR_INVALIDDATA;
2451 
2452  p = data;
2453  q = data + data_len;
2454 
2455  id = p;
2456  id_len = -1;
2457  while (p < q) {
2458  if (*p == '\r' || *p == '\n') {
2459  id_len = p - id;
2460  if (*p == '\r')
2461  p++;
2462  break;
2463  }
2464  p++;
2465  }
2466 
2467  if (p >= q || *p != '\n')
2468  return AVERROR_INVALIDDATA;
2469  p++;
2470 
2471  settings = p;
2472  settings_len = -1;
2473  while (p < q) {
2474  if (*p == '\r' || *p == '\n') {
2475  settings_len = p - settings;
2476  if (*p == '\r')
2477  p++;
2478  break;
2479  }
2480  p++;
2481  }
2482 
2483  if (p >= q || *p != '\n')
2484  return AVERROR_INVALIDDATA;
2485  p++;
2486 
2487  text = p;
2488  text_len = q - p;
2489  while (text_len > 0) {
2490  const int len = text_len - 1;
2491  const uint8_t c = p[len];
2492  if (c != '\r' && c != '\n')
2493  break;
2494  text_len = len;
2495  }
2496 
2497  if (text_len <= 0)
2498  return AVERROR_INVALIDDATA;
2499 
2500  pkt = av_mallocz(sizeof(*pkt));
2501  err = av_new_packet(pkt, text_len);
2502  if (err < 0) {
2503  av_free(pkt);
2504  return AVERROR(err);
2505  }
2506 
2507  memcpy(pkt->data, text, text_len);
2508 
2509  if (id_len > 0) {
2510  buf = av_packet_new_side_data(pkt,
2512  id_len);
2513  if (!buf) {
2514  av_free(pkt);
2515  return AVERROR(ENOMEM);
2516  }
2517  memcpy(buf, id, id_len);
2518  }
2519 
2520  if (settings_len > 0) {
2521  buf = av_packet_new_side_data(pkt,
2523  settings_len);
2524  if (!buf) {
2525  av_free(pkt);
2526  return AVERROR(ENOMEM);
2527  }
2528  memcpy(buf, settings, settings_len);
2529  }
2530 
2531  // Do we need this for subtitles?
2532  // pkt->flags = AV_PKT_FLAG_KEY;
2533 
2534  pkt->stream_index = st->index;
2535  pkt->pts = timecode;
2536 
2537  // Do we need this for subtitles?
2538  // pkt->dts = timecode;
2539 
2540  pkt->duration = duration;
2541  pkt->pos = pos;
2542 
2543  dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2544  matroska->prev_pkt = pkt;
2545 
2546  return 0;
2547 }
2548 
2550  MatroskaTrack *track, AVStream *st,
2551  uint8_t *data, int pkt_size,
2552  uint64_t timecode, uint64_t lace_duration,
2553  int64_t pos, int is_keyframe,
2554  uint8_t *additional, uint64_t additional_id, int additional_size,
2555  int64_t discard_padding)
2556 {
2557  MatroskaTrackEncoding *encodings = track->encodings.elem;
2558  uint8_t *pkt_data = data;
2559  int offset = 0, res;
2560  AVPacket *pkt;
2561 
2562  if (encodings && !encodings->type && encodings->scope & 1) {
2563  res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2564  if (res < 0)
2565  return res;
2566  }
2567 
2568  if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
2569  uint8_t *wv_data;
2570  res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2571  if (res < 0) {
2572  av_log(matroska->ctx, AV_LOG_ERROR,
2573  "Error parsing a wavpack block.\n");
2574  goto fail;
2575  }
2576  if (pkt_data != data)
2577  av_freep(&pkt_data);
2578  pkt_data = wv_data;
2579  }
2580 
2581  if (st->codec->codec_id == AV_CODEC_ID_PRORES &&
2582  AV_RB32(&data[4]) != MKBETAG('i', 'c', 'p', 'f'))
2583  offset = 8;
2584 
2585  pkt = av_mallocz(sizeof(AVPacket));
2586  /* XXX: prevent data copy... */
2587  if (av_new_packet(pkt, pkt_size + offset) < 0) {
2588  av_free(pkt);
2589  res = AVERROR(ENOMEM);
2590  goto fail;
2591  }
2592 
2593  if (st->codec->codec_id == AV_CODEC_ID_PRORES && offset == 8) {
2594  uint8_t *buf = pkt->data;
2595  bytestream_put_be32(&buf, pkt_size);
2596  bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2597  }
2598 
2599  memcpy(pkt->data + offset, pkt_data, pkt_size);
2600 
2601  if (pkt_data != data)
2602  av_freep(&pkt_data);
2603 
2604  pkt->flags = is_keyframe;
2605  pkt->stream_index = st->index;
2606 
2607  if (additional_size > 0) {
2608  uint8_t *side_data = av_packet_new_side_data(pkt,
2610  additional_size + 8);
2611  if (!side_data) {
2612  av_free_packet(pkt);
2613  av_free(pkt);
2614  return AVERROR(ENOMEM);
2615  }
2616  AV_WB64(side_data, additional_id);
2617  memcpy(side_data + 8, additional, additional_size);
2618  }
2619 
2620  if (discard_padding) {
2621  uint8_t *side_data = av_packet_new_side_data(pkt,
2623  10);
2624  if (!side_data) {
2625  av_free_packet(pkt);
2626  av_free(pkt);
2627  return AVERROR(ENOMEM);
2628  }
2629  AV_WL32(side_data, 0);
2630  AV_WL32(side_data + 4, av_rescale_q(discard_padding,
2631  (AVRational){1, 1000000000},
2632  (AVRational){1, st->codec->sample_rate}));
2633  }
2634 
2635  if (track->ms_compat)
2636  pkt->dts = timecode;
2637  else
2638  pkt->pts = timecode;
2639  pkt->pos = pos;
2640  if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2641  /*
2642  * For backward compatibility.
2643  * Historically, we have put subtitle duration
2644  * in convergence_duration, on the off chance
2645  * that the time_scale is less than 1us, which
2646  * could result in a 32bit overflow on the
2647  * normal duration field.
2648  */
2649  pkt->convergence_duration = lace_duration;
2650  }
2651 
2652  if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
2653  lace_duration <= INT_MAX) {
2654  /*
2655  * For non subtitle tracks, just store the duration
2656  * as normal.
2657  *
2658  * If it's a subtitle track and duration value does
2659  * not overflow a uint32, then also store it normally.
2660  */
2661  pkt->duration = lace_duration;
2662  }
2663 
2664  dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2665  matroska->prev_pkt = pkt;
2666 
2667  return 0;
2668 
2669 fail:
2670  if (pkt_data != data)
2671  av_freep(&pkt_data);
2672  return res;
2673 }
2674 
2676  int size, int64_t pos, uint64_t cluster_time,
2677  uint64_t block_duration, int is_keyframe,
2678  uint8_t *additional, uint64_t additional_id, int additional_size,
2679  int64_t cluster_pos, int64_t discard_padding)
2680 {
2681  uint64_t timecode = AV_NOPTS_VALUE;
2682  MatroskaTrack *track;
2683  int res = 0;
2684  AVStream *st;
2685  int16_t block_time;
2686  uint32_t *lace_size = NULL;
2687  int n, flags, laces = 0;
2688  uint64_t num;
2689  int trust_default_duration = 1;
2690 
2691  if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2692  av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2693  return n;
2694  }
2695  data += n;
2696  size -= n;
2697 
2698  track = matroska_find_track_by_num(matroska, num);
2699  if (!track || !track->stream) {
2700  av_log(matroska->ctx, AV_LOG_INFO,
2701  "Invalid stream %"PRIu64" or size %u\n", num, size);
2702  return AVERROR_INVALIDDATA;
2703  } else if (size <= 3)
2704  return 0;
2705  st = track->stream;
2706  if (st->discard >= AVDISCARD_ALL)
2707  return res;
2708  av_assert1(block_duration != AV_NOPTS_VALUE);
2709 
2710  block_time = sign_extend(AV_RB16(data), 16);
2711  data += 2;
2712  flags = *data++;
2713  size -= 3;
2714  if (is_keyframe == -1)
2715  is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2716 
2717  if (cluster_time != (uint64_t) -1 &&
2718  (block_time >= 0 || cluster_time >= -block_time)) {
2719  timecode = cluster_time + block_time - track->codec_delay;
2720  if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2721  timecode < track->end_timecode)
2722  is_keyframe = 0; /* overlapping subtitles are not key frame */
2723  if (is_keyframe)
2724  av_add_index_entry(st, cluster_pos, timecode, 0, 0,
2726  }
2727 
2728  if (matroska->skip_to_keyframe &&
2729  track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2730  if (timecode < matroska->skip_to_timecode)
2731  return res;
2732  if (is_keyframe)
2733  matroska->skip_to_keyframe = 0;
2734  else if (!st->skip_to_keyframe) {
2735  av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2736  matroska->skip_to_keyframe = 0;
2737  }
2738  }
2739 
2740  res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2741  &lace_size, &laces);
2742 
2743  if (res)
2744  goto end;
2745 
2746  if (track->audio.samplerate == 8000) {
2747  // If this is needed for more codecs, then add them here
2748  if (st->codec->codec_id == AV_CODEC_ID_AC3) {
2749  if (track->audio.samplerate != st->codec->sample_rate || !st->codec->frame_size)
2750  trust_default_duration = 0;
2751  }
2752  }
2753 
2754  if (!block_duration && trust_default_duration)
2755  block_duration = track->default_duration * laces / matroska->time_scale;
2756 
2757  if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2758  track->end_timecode =
2759  FFMAX(track->end_timecode, timecode + block_duration);
2760 
2761  for (n = 0; n < laces; n++) {
2762  int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2763 
2764  if (lace_size[n] > size) {
2765  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2766  break;
2767  }
2768 
2769  if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2770  st->codec->codec_id == AV_CODEC_ID_COOK ||
2771  st->codec->codec_id == AV_CODEC_ID_SIPR ||
2772  st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2773  st->codec->block_align && track->audio.sub_packet_size) {
2774  res = matroska_parse_rm_audio(matroska, track, st, data,
2775  lace_size[n],
2776  timecode, pos);
2777  if (res)
2778  goto end;
2779 
2780  } else if (st->codec->codec_id == AV_CODEC_ID_WEBVTT) {
2781  res = matroska_parse_webvtt(matroska, track, st,
2782  data, lace_size[n],
2783  timecode, lace_duration,
2784  pos);
2785  if (res)
2786  goto end;
2787  } else {
2788  res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2789  timecode, lace_duration, pos,
2790  !n ? is_keyframe : 0,
2791  additional, additional_id, additional_size,
2792  discard_padding);
2793  if (res)
2794  goto end;
2795  }
2796 
2797  if (timecode != AV_NOPTS_VALUE)
2798  timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2799  data += lace_size[n];
2800  size -= lace_size[n];
2801  }
2802 
2803 end:
2804  av_free(lace_size);
2805  return res;
2806 }
2807 
2809 {
2810  EbmlList *blocks_list;
2811  MatroskaBlock *blocks;
2812  int i, res;
2813  res = ebml_parse(matroska,
2814  matroska_cluster_incremental_parsing,
2815  &matroska->current_cluster);
2816  if (res == 1) {
2817  /* New Cluster */
2818  if (matroska->current_cluster_pos)
2819  ebml_level_end(matroska);
2820  ebml_free(matroska_cluster, &matroska->current_cluster);
2821  memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2822  matroska->current_cluster_num_blocks = 0;
2823  matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2824  matroska->prev_pkt = NULL;
2825  /* sizeof the ID which was already read */
2826  if (matroska->current_id)
2827  matroska->current_cluster_pos -= 4;
2828  res = ebml_parse(matroska,
2829  matroska_clusters_incremental,
2830  &matroska->current_cluster);
2831  /* Try parsing the block again. */
2832  if (res == 1)
2833  res = ebml_parse(matroska,
2834  matroska_cluster_incremental_parsing,
2835  &matroska->current_cluster);
2836  }
2837 
2838  if (!res &&
2839  matroska->current_cluster_num_blocks <
2840  matroska->current_cluster.blocks.nb_elem) {
2841  blocks_list = &matroska->current_cluster.blocks;
2842  blocks = blocks_list->elem;
2843 
2844  matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2845  i = blocks_list->nb_elem - 1;
2846  if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2847  int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2848  uint8_t* additional = blocks[i].additional.size > 0 ?
2849  blocks[i].additional.data : NULL;
2850  if (!blocks[i].non_simple)
2851  blocks[i].duration = 0;
2852  res = matroska_parse_block(matroska, blocks[i].bin.data,
2853  blocks[i].bin.size, blocks[i].bin.pos,
2854  matroska->current_cluster.timecode,
2855  blocks[i].duration, is_keyframe,
2856  additional, blocks[i].additional_id,
2857  blocks[i].additional.size,
2858  matroska->current_cluster_pos,
2859  blocks[i].discard_padding);
2860  }
2861  }
2862 
2863  return res;
2864 }
2865 
2867 {
2868  MatroskaCluster cluster = { 0 };
2869  EbmlList *blocks_list;
2870  MatroskaBlock *blocks;
2871  int i, res;
2872  int64_t pos;
2873 
2874  if (!matroska->contains_ssa)
2875  return matroska_parse_cluster_incremental(matroska);
2876  pos = avio_tell(matroska->ctx->pb);
2877  matroska->prev_pkt = NULL;
2878  if (matroska->current_id)
2879  pos -= 4; /* sizeof the ID which was already read */
2880  res = ebml_parse(matroska, matroska_clusters, &cluster);
2881  blocks_list = &cluster.blocks;
2882  blocks = blocks_list->elem;
2883  for (i = 0; i < blocks_list->nb_elem; i++)
2884  if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2885  int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2886  res = matroska_parse_block(matroska, blocks[i].bin.data,
2887  blocks[i].bin.size, blocks[i].bin.pos,
2888  cluster.timecode, blocks[i].duration,
2889  is_keyframe, NULL, 0, 0, pos,
2890  blocks[i].discard_padding);
2891  }
2892  ebml_free(matroska_cluster, &cluster);
2893  return res;
2894 }
2895 
2897 {
2898  MatroskaDemuxContext *matroska = s->priv_data;
2899 
2900  while (matroska_deliver_packet(matroska, pkt)) {
2901  int64_t pos = avio_tell(matroska->ctx->pb);
2902  if (matroska->done)
2903  return AVERROR_EOF;
2904  if (matroska_parse_cluster(matroska) < 0)
2905  matroska_resync(matroska, pos);
2906  }
2907 
2908  return 0;
2909 }
2910 
2911 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2912  int64_t timestamp, int flags)
2913 {
2914  MatroskaDemuxContext *matroska = s->priv_data;
2915  MatroskaTrack *tracks = matroska->tracks.elem;
2916  AVStream *st = s->streams[stream_index];
2917  int i, index, index_sub, index_min;
2918 
2919  /* Parse the CUES now since we need the index data to seek. */
2920  if (matroska->cues_parsing_deferred > 0) {
2921  matroska->cues_parsing_deferred = 0;
2922  matroska_parse_cues(matroska);
2923  }
2924 
2925  if (!st->nb_index_entries)
2926  goto err;
2927  timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2928 
2929  if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
2930  avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
2931  SEEK_SET);
2932  matroska->current_id = 0;
2933  while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
2934  matroska_clear_queue(matroska);
2935  if (matroska_parse_cluster(matroska) < 0)
2936  break;
2937  }
2938  }
2939 
2940  matroska_clear_queue(matroska);
2941  if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
2942  goto err;
2943 
2944  index_min = index;
2945  for (i = 0; i < matroska->tracks.nb_elem; i++) {
2946  tracks[i].audio.pkt_cnt = 0;
2947  tracks[i].audio.sub_packet_cnt = 0;
2948  tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2949  tracks[i].end_timecode = 0;
2950  if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2951  tracks[i].stream->discard != AVDISCARD_ALL) {
2952  index_sub = av_index_search_timestamp(
2953  tracks[i].stream, st->index_entries[index].timestamp,
2955  while (index_sub >= 0 &&
2956  index_min > 0 &&
2957  tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
2958  st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
2959  index_min--;
2960  }
2961  }
2962 
2963  avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2964  matroska->current_id = 0;
2965  if (flags & AVSEEK_FLAG_ANY) {
2966  st->skip_to_keyframe = 0;
2967  matroska->skip_to_timecode = timestamp;
2968  } else {
2969  st->skip_to_keyframe = 1;
2970  matroska->skip_to_timecode = st->index_entries[index].timestamp;
2971  }
2972  matroska->skip_to_keyframe = 1;
2973  matroska->done = 0;
2974  matroska->num_levels = 0;
2975  ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2976  return 0;
2977 err:
2978  // slightly hackish but allows proper fallback to
2979  // the generic seeking code.
2980  matroska_clear_queue(matroska);
2981  matroska->current_id = 0;
2982  st->skip_to_keyframe =
2983  matroska->skip_to_keyframe = 0;
2984  matroska->done = 0;
2985  matroska->num_levels = 0;
2986  return -1;
2987 }
2988 
2990 {
2991  MatroskaDemuxContext *matroska = s->priv_data;
2992  MatroskaTrack *tracks = matroska->tracks.elem;
2993  int n;
2994 
2995  matroska_clear_queue(matroska);
2996 
2997  for (n = 0; n < matroska->tracks.nb_elem; n++)
2998  if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
2999  av_free(tracks[n].audio.buf);
3000  ebml_free(matroska_cluster, &matroska->current_cluster);
3001  ebml_free(matroska_segment, matroska);
3002 
3003  return 0;
3004 }
3005 
3006 typedef struct {
3007  int64_t start_time_ns;
3008  int64_t end_time_ns;
3009  int64_t start_offset;
3010  int64_t end_offset;
3011 } CueDesc;
3012 
3013 /* This function searches all the Cues and returns the CueDesc corresponding the
3014  * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3015  * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3016  */
3017 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3018  MatroskaDemuxContext *matroska = s->priv_data;
3019  CueDesc cue_desc;
3020  int i;
3021  int nb_index_entries = s->streams[0]->nb_index_entries;
3022  AVIndexEntry *index_entries = s->streams[0]->index_entries;
3023  if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3024  for (i = 1; i < nb_index_entries; i++) {
3025  if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3026  index_entries[i].timestamp * matroska->time_scale > ts) {
3027  break;
3028  }
3029  }
3030  --i;
3031  cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3032  cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3033  if (i != nb_index_entries - 1) {
3034  cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3035  cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3036  } else {
3037  cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3038  // FIXME: this needs special handling for files where Cues appear
3039  // before Clusters. the current logic assumes Cues appear after
3040  // Clusters.
3041  cue_desc.end_offset = cues_start - matroska->segment_start;
3042  }
3043  return cue_desc;
3044 }
3045 
3047 {
3048  MatroskaDemuxContext *matroska = s->priv_data;
3049  int64_t cluster_pos, before_pos;
3050  int index, rv = 1;
3051  if (s->streams[0]->nb_index_entries <= 0) return 0;
3052  // seek to the first cluster using cues.
3053  index = av_index_search_timestamp(s->streams[0], 0, 0);
3054  if (index < 0) return 0;
3055  cluster_pos = s->streams[0]->index_entries[index].pos;
3056  before_pos = avio_tell(s->pb);
3057  while (1) {
3058  int64_t cluster_id = 0, cluster_length = 0;
3059  AVPacket *pkt;
3060  avio_seek(s->pb, cluster_pos, SEEK_SET);
3061  // read cluster id and length
3062  ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id);
3063  ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3064  if (cluster_id != 0xF43B675) { // done with all clusters
3065  break;
3066  }
3067  avio_seek(s->pb, cluster_pos, SEEK_SET);
3068  matroska->current_id = 0;
3069  matroska_clear_queue(matroska);
3070  if (matroska_parse_cluster(matroska) < 0 ||
3071  matroska->num_packets <= 0) {
3072  break;
3073  }
3074  pkt = matroska->packets[0];
3075  cluster_pos += cluster_length + 12; // 12 is the offset of the cluster id and length.
3076  if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3077  rv = 0;
3078  break;
3079  }
3080  }
3081  avio_seek(s->pb, before_pos, SEEK_SET);
3082  return rv;
3083 }
3084 
3085 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3086  double min_buffer, double* buffer,
3087  double* sec_to_download, AVFormatContext *s,
3088  int64_t cues_start)
3089 {
3090  double nano_seconds_per_second = 1000000000.0;
3091  double time_sec = time_ns / nano_seconds_per_second;
3092  int rv = 0;
3093  int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3094  int64_t end_time_ns = time_ns + time_to_search_ns;
3095  double sec_downloaded = 0.0;
3096  CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3097  if (desc_curr.start_time_ns == -1)
3098  return -1;
3099  *sec_to_download = 0.0;
3100 
3101  // Check for non cue start time.
3102  if (time_ns > desc_curr.start_time_ns) {
3103  int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3104  double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3105  double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3106  double timeToDownload = (cueBytes * 8.0) / bps;
3107 
3108  sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3109  *sec_to_download += timeToDownload;
3110 
3111  // Check if the search ends within the first cue.
3112  if (desc_curr.end_time_ns >= end_time_ns) {
3113  double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3114  double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3115  sec_downloaded = percent_to_sub * sec_downloaded;
3116  *sec_to_download = percent_to_sub * *sec_to_download;
3117  }
3118 
3119  if ((sec_downloaded + *buffer) <= min_buffer) {
3120  return 1;
3121  }
3122 
3123  // Get the next Cue.
3124  desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3125  }
3126 
3127  while (desc_curr.start_time_ns != -1) {
3128  int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3129  int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3130  double desc_sec = desc_ns / nano_seconds_per_second;
3131  double bits = (desc_bytes * 8.0);
3132  double time_to_download = bits / bps;
3133 
3134  sec_downloaded += desc_sec - time_to_download;
3135  *sec_to_download += time_to_download;
3136 
3137  if (desc_curr.end_time_ns >= end_time_ns) {
3138  double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3139  double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3140  sec_downloaded = percent_to_sub * sec_downloaded;
3141  *sec_to_download = percent_to_sub * *sec_to_download;
3142 
3143  if ((sec_downloaded + *buffer) <= min_buffer)
3144  rv = 1;
3145  break;
3146  }
3147 
3148  if ((sec_downloaded + *buffer) <= min_buffer) {
3149  rv = 1;
3150  break;
3151  }
3152 
3153  desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3154  }
3155  *buffer = *buffer + sec_downloaded;
3156  return rv;
3157 }
3158 
3159 /* This function computes the bandwidth of the WebM file with the help of
3160  * buffer_size_after_time_downloaded() function. Both of these functions are
3161  * adapted from WebM Tools project and are adapted to work with FFmpeg's
3162  * Matroska parsing mechanism.
3163  *
3164  * Returns the bandwidth of the file on success; -1 on error.
3165  * */
3166 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3167 {
3168  MatroskaDemuxContext *matroska = s->priv_data;
3169  AVStream *st = s->streams[0];
3170  double bandwidth = 0.0;
3171  int i;
3172 
3173  for (i = 0; i < st->nb_index_entries; i++) {
3174  int64_t prebuffer_ns = 1000000000;
3175  int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3176  double nano_seconds_per_second = 1000000000.0;
3177  int64_t prebuffered_ns = time_ns + prebuffer_ns;
3178  double prebuffer_bytes = 0.0;
3179  int64_t temp_prebuffer_ns = prebuffer_ns;
3180  int64_t pre_bytes, pre_ns;
3181  double pre_sec, prebuffer, bits_per_second;
3182  CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3183 
3184  // Start with the first Cue.
3185  CueDesc desc_end = desc_beg;
3186 
3187  // Figure out how much data we have downloaded for the prebuffer. This will
3188  // be used later to adjust the bits per sample to try.
3189  while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3190  // Prebuffered the entire Cue.
3191  prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3192  temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3193  desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3194  }
3195  if (desc_end.start_time_ns == -1) {
3196  // The prebuffer is larger than the duration.
3197  if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3198  return -1;
3199  bits_per_second = 0.0;
3200  } else {
3201  // The prebuffer ends in the last Cue. Estimate how much data was
3202  // prebuffered.
3203  pre_bytes = desc_end.end_offset - desc_end.start_offset;
3204  pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3205  pre_sec = pre_ns / nano_seconds_per_second;
3206  prebuffer_bytes +=
3207  pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3208 
3209  prebuffer = prebuffer_ns / nano_seconds_per_second;
3210 
3211  // Set this to 0.0 in case our prebuffer buffers the entire video.
3212  bits_per_second = 0.0;
3213  do {
3214  int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3215  int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3216  double desc_sec = desc_ns / nano_seconds_per_second;
3217  double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3218 
3219  // Drop the bps by the percentage of bytes buffered.
3220  double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3221  double mod_bits_per_second = calc_bits_per_second * percent;
3222 
3223  if (prebuffer < desc_sec) {
3224  double search_sec =
3225  (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3226 
3227  // Add 1 so the bits per second should be a little bit greater than file
3228  // datarate.
3229  int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3230  const double min_buffer = 0.0;
3231  double buffer = prebuffer;
3232  double sec_to_download = 0.0;
3233 
3234  int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3235  min_buffer, &buffer, &sec_to_download,
3236  s, cues_start);
3237  if (rv < 0) {
3238  return -1;
3239  } else if (rv == 0) {
3240  bits_per_second = (double)(bps);
3241  break;
3242  }
3243  }
3244 
3245  desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3246  } while (desc_end.start_time_ns != -1);
3247  }
3248  if (bandwidth < bits_per_second) bandwidth = bits_per_second;
3249  }
3250  return (int64_t)bandwidth;
3251 }
3252 
3254 {
3255  MatroskaDemuxContext *matroska = s->priv_data;
3256  EbmlList *seekhead_list = &matroska->seekhead;
3257  MatroskaSeekhead *seekhead = seekhead_list->elem;
3258  char *buf;
3259  int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
3260  int i;
3261 
3262  // determine cues start and end positions
3263  for (i = 0; i < seekhead_list->nb_elem; i++)
3264  if (seekhead[i].id == MATROSKA_ID_CUES)
3265  break;
3266 
3267  if (i >= seekhead_list->nb_elem) return -1;
3268 
3269  before_pos = avio_tell(matroska->ctx->pb);
3270  cues_start = seekhead[i].pos + matroska->segment_start;
3271  if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
3272  // cues_end is computed as cues_start + cues_length + length of the
3273  // Cues element ID + EBML length of the Cues element. cues_end is
3274  // inclusive and the above sum is reduced by 1.
3275  uint64_t cues_length = 0, cues_id = 0, bytes_read = 0;
3276  bytes_read += ebml_read_num(matroska, matroska->ctx->pb, 4, &cues_id);
3277  bytes_read += ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
3278  cues_end = cues_start + cues_length + bytes_read - 1;
3279  }
3280  avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
3281  if (cues_start == -1 || cues_end == -1) return -1;
3282 
3283  // parse the cues
3284  matroska_parse_cues(matroska);
3285 
3286  // cues start
3287  av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
3288 
3289  // cues end
3290  av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
3291 
3292  // bandwidth
3293  bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
3294  if (bandwidth < 0) return -1;
3295  av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
3296 
3297  // check if all clusters start with key frames
3299 
3300  // store cue point timestamps as a comma separated list for checking subsegment alignment in
3301  // the muxer. assumes that each timestamp cannot be more than 20 characters long.
3302  buf = av_malloc(s->streams[0]->nb_index_entries * 20 * sizeof(char));
3303  if (!buf) return -1;
3304  strcpy(buf, "");
3305  for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
3306  snprintf(buf, (i + 1) * 20 * sizeof(char),
3307  "%s%" PRId64, buf, s->streams[0]->index_entries[i].timestamp);
3308  if (i != s->streams[0]->nb_index_entries - 1)
3309  strncat(buf, ",", sizeof(char));
3310  }
3311  av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
3312  av_free(buf);
3313 
3314  return 0;
3315 }
3316 
3318 {
3319  char *buf;
3320  int ret = matroska_read_header(s);
3321  MatroskaTrack *tracks;
3322  MatroskaDemuxContext *matroska = s->priv_data;
3323  if (ret) {
3324  av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
3325  return -1;
3326  }
3327 
3328  // initialization range
3329  // 5 is the offset of Cluster ID.
3331 
3332  // basename of the file
3333  buf = strrchr(s->filename, '/');
3334  av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->filename, 0);
3335 
3336  // duration
3337  buf = av_asprintf("%g", matroska->duration);
3338  if (!buf) return AVERROR(ENOMEM);
3339  av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
3340  av_free(buf);
3341 
3342  // track number
3343  tracks = matroska->tracks.elem;
3344  av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
3345 
3346  // parse the cues and populate Cue related fields
3347  return webm_dash_manifest_cues(s);
3348 }
3349 
3351 {
3352  return AVERROR_EOF;
3353 }
3354 
3356  .name = "matroska,webm",
3357  .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
3358  .extensions = "mkv,mk3d,mka,mks",
3359  .priv_data_size = sizeof(MatroskaDemuxContext),
3365  .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
3366 };
3367 
3369  .name = "webm_dash_manifest",
3370  .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
3371  .priv_data_size = sizeof(MatroskaDemuxContext),
3375 };