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 
36 #include "libavutil/avstring.h"
37 #include "libavutil/base64.h"
38 #include "libavutil/dict.h"
39 #include "libavutil/intfloat.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/lzo.h"
42 #include "libavutil/mathematics.h"
44 
45 #include "libavcodec/bytestream.h"
46 #include "libavcodec/flac.h"
47 #include "libavcodec/mpeg4audio.h"
48 
49 #include "avformat.h"
50 #include "avio_internal.h"
51 #include "internal.h"
52 #include "isom.h"
53 #include "matroska.h"
54 #include "oggdec.h"
55 /* For ff_codec_get_id(). */
56 #include "riff.h"
57 #include "rmsipr.h"
58 
59 #if CONFIG_BZLIB
60 #include <bzlib.h>
61 #endif
62 #if CONFIG_ZLIB
63 #include <zlib.h>
64 #endif
65 
66 typedef enum {
78 } EbmlType;
79 
80 typedef const struct EbmlSyntax {
81  uint32_t id;
85  union {
86  uint64_t u;
87  double f;
88  const char *s;
89  const struct EbmlSyntax *n;
90  } def;
91 } EbmlSyntax;
92 
93 typedef struct {
94  int nb_elem;
95  void *elem;
96 } EbmlList;
97 
98 typedef struct {
99  int size;
101  int64_t pos;
102 } EbmlBin;
103 
104 typedef struct {
105  uint64_t version;
106  uint64_t max_size;
107  uint64_t id_length;
108  char *doctype;
109  uint64_t doctype_version;
110 } Ebml;
111 
112 typedef struct {
113  uint64_t algo;
116 
117 typedef struct {
118  uint64_t algo;
121 
122 typedef struct {
123  uint64_t scope;
124  uint64_t type;
128 
129 typedef struct {
130  double frame_rate;
131  uint64_t display_width;
132  uint64_t display_height;
133  uint64_t pixel_width;
134  uint64_t pixel_height;
136  uint64_t stereo_mode;
137  uint64_t alpha_mode;
139 
140 typedef struct {
141  double samplerate;
143  uint64_t bitdepth;
144  uint64_t channels;
145 
146  /* real audio header (extracted from extradata) */
152  int pkt_cnt;
153  uint64_t buf_timecode;
156 
157 typedef struct {
158  uint64_t uid;
159  uint64_t type;
161 
162 typedef struct {
165 
166 typedef struct {
167  uint64_t num;
168  uint64_t uid;
169  uint64_t type;
170  char *name;
171  char *codec_id;
173  char *language;
174  double time_scale;
176  uint64_t flag_default;
177  uint64_t flag_forced;
178  uint64_t seek_preroll;
183  uint64_t codec_delay;
184 
186  int64_t end_timecode;
189 } MatroskaTrack;
190 
191 typedef struct {
192  uint64_t uid;
193  char *filename;
194  char *mime;
196 
199 
200 typedef struct {
201  uint64_t start;
202  uint64_t end;
203  uint64_t uid;
204  char *title;
205 
208 
209 typedef struct {
210  uint64_t track;
211  uint64_t pos;
213 
214 typedef struct {
215  uint64_t time;
217 } MatroskaIndex;
218 
219 typedef struct {
220  char *name;
221  char *string;
222  char *lang;
223  uint64_t def;
225 } MatroskaTag;
226 
227 typedef struct {
228  char *type;
229  uint64_t typevalue;
230  uint64_t trackuid;
231  uint64_t chapteruid;
232  uint64_t attachuid;
234 
235 typedef struct {
238 } MatroskaTags;
239 
240 typedef struct {
241  uint64_t id;
242  uint64_t pos;
244 
245 typedef struct {
246  uint64_t start;
247  uint64_t length;
248 } MatroskaLevel;
249 
250 typedef struct {
251  uint64_t timecode;
254 
255 typedef struct {
257 
258  /* EBML stuff */
261  int level_up;
262  uint32_t current_id;
263 
264  uint64_t time_scale;
265  double duration;
266  char *title;
267  char *muxingapp;
275 
276  /* byte position of the segment inside the stream */
277  int64_t segment_start;
278 
279  /* the packet queue */
283 
284  int done;
285 
286  /* What to skip before effectively reading a packet. */
289 
290  /* File has a CUES element, but we defer parsing until it is needed. */
292 
296 
297  /* File has SSA subtitles which prevent incremental cluster parsing. */
300 
301 typedef struct {
302  uint64_t duration;
303  int64_t reference;
304  uint64_t non_simple;
306  uint64_t additional_id;
309 } MatroskaBlock;
310 
312  { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
313  { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
314  { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
315  { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
316  { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
319  { 0 }
320 };
321 
323  { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
324  { 0 }
325 };
326 
328  { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
330  { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
332  { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
333  { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
335  { 0 }
336 };
337 
339  { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
340  { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
341  { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
342  { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
343  { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
344  { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
345  { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
354  { 0 }
355 };
356 
358  { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
359  { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
361  { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
362  { 0 }
363 };
364 
368  { 0 }
369 };
370 
379  { 0 }
380 };
382  { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
383  { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
384  { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
385  { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
387  { 0 }
388 };
389 
391  { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
392  { 0 }
393 };
394 
398  { 0 }
399 };
400 
402  { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
403  { 0 }
404 };
405 
407  { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
408  { 0 }
409 };
410 
412  { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
414  { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
417  { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
418  { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
419  { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
420  { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
421  { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
422  { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
423  { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
424  { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
425  { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
426  { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
427  { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
428  { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
429  { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
438  { 0 }
439 };
440 
442  { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
443  { 0 }
444 };
445 
448  { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
449  { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
450  { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
452  { 0 }
453 };
454 
456  { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
457  { 0 }
458 };
459 
461  { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
463  { 0 }
464 };
465 
470  { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
475  { 0 }
476 };
477 
479  { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
484  { 0 }
485 };
486 
488  { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
489  { 0 }
490 };
491 
493  { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
498  { 0 }
499 };
500 
502  { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
503  { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
504  { 0 }
505 };
506 
508  { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
509  { 0 }
510 };
511 
513  { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
514  { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
515  { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
516  { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
517  { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
518  { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
519  { 0 }
520 };
521 
524  { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
525  { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
526  { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
527  { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
528  { 0 }
529 };
530 
532  { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
533  { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
534  { 0 }
535 };
536 
538  { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
539  { 0 }
540 };
541 
543  { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
544  { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
545  { 0 }
546 };
547 
549  { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
550  { 0 }
551 };
552 
554  { MATROSKA_ID_INFO, EBML_NEST, 0, 0, { .n = matroska_info } },
555  { MATROSKA_ID_TRACKS, EBML_NEST, 0, 0, { .n = matroska_tracks } },
556  { MATROSKA_ID_ATTACHMENTS, EBML_NEST, 0, 0, { .n = matroska_attachments } },
557  { MATROSKA_ID_CHAPTERS, EBML_NEST, 0, 0, { .n = matroska_chapters } },
558  { MATROSKA_ID_CUES, EBML_NEST, 0, 0, { .n = matroska_index } },
559  { MATROSKA_ID_TAGS, EBML_NEST, 0, 0, { .n = matroska_tags } },
560  { MATROSKA_ID_SEEKHEAD, EBML_NEST, 0, 0, { .n = matroska_seekhead } },
562  { 0 }
563 };
564 
566  { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
567  { 0 }
568 };
569 
571  { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
572  { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
573  { 0 }
574 };
575 
577  { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
578  { 0 }
579 };
580 
582  { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
583  { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
584  { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
586  { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
587  { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference) },
589  { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
590  { 0 }
591 };
592 
594  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
595  { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
596  { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
599  { 0 }
600 };
601 
603  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
608  { 0 }
609 };
610 
612  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
613  { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
614  { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
622  { 0 }
623 };
624 
626  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
631  { 0 }
632 };
633 
635  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
640  { 0 }
641 };
642 
643 static const char *const matroska_doctypes[] = { "matroska", "webm" };
644 
645 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
646 {
647  AVIOContext *pb = matroska->ctx->pb;
648  uint32_t id;
649  matroska->current_id = 0;
650  matroska->num_levels = 0;
651 
652  /* seek to next position to resync from */
653  if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
654  goto eof;
655 
656  id = avio_rb32(pb);
657 
658  // try to find a toplevel element
659  while (!avio_feof(pb)) {
660  if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
661  id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
663  id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
664  matroska->current_id = id;
665  return 0;
666  }
667  id = (id << 8) | avio_r8(pb);
668  }
669 
670 eof:
671  matroska->done = 1;
672  return AVERROR_EOF;
673 }
674 
675 /*
676  * Return: Whether we reached the end of a level in the hierarchy or not.
677  */
679 {
680  AVIOContext *pb = matroska->ctx->pb;
681  int64_t pos = avio_tell(pb);
682 
683  if (matroska->num_levels > 0) {
684  MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
685  if (pos - level->start >= level->length || matroska->current_id) {
686  matroska->num_levels--;
687  return 1;
688  }
689  }
690  return 0;
691 }
692 
693 /*
694  * Read: an "EBML number", which is defined as a variable-length
695  * array of bytes. The first byte indicates the length by giving a
696  * number of 0-bits followed by a one. The position of the first
697  * "one" bit inside the first byte indicates the length of this
698  * number.
699  * Returns: number of bytes read, < 0 on error
700  */
702  int max_size, uint64_t *number)
703 {
704  int read = 1, n = 1;
705  uint64_t total = 0;
706 
707  /* The first byte tells us the length in bytes - avio_r8() can normally
708  * return 0, but since that's not a valid first ebmlID byte, we can
709  * use it safely here to catch EOS. */
710  if (!(total = avio_r8(pb))) {
711  /* we might encounter EOS here */
712  if (!avio_feof(pb)) {
713  int64_t pos = avio_tell(pb);
714  av_log(matroska->ctx, AV_LOG_ERROR,
715  "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
716  pos, pos);
717  return pb->error ? pb->error : AVERROR(EIO);
718  }
719  return AVERROR_EOF;
720  }
721 
722  /* get the length of the EBML number */
723  read = 8 - ff_log2_tab[total];
724  if (read > max_size) {
725  int64_t pos = avio_tell(pb) - 1;
726  av_log(matroska->ctx, AV_LOG_ERROR,
727  "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
728  (uint8_t) total, pos, pos);
729  return AVERROR_INVALIDDATA;
730  }
731 
732  /* read out length */
733  total ^= 1 << ff_log2_tab[total];
734  while (n++ < read)
735  total = (total << 8) | avio_r8(pb);
736 
737  *number = total;
738 
739  return read;
740 }
741 
742 /**
743  * Read a EBML length value.
744  * This needs special handling for the "unknown length" case which has multiple
745  * encodings.
746  */
748  uint64_t *number)
749 {
750  int res = ebml_read_num(matroska, pb, 8, number);
751  if (res > 0 && *number + 1 == 1ULL << (7 * res))
752  *number = 0xffffffffffffffULL;
753  return res;
754 }
755 
756 /*
757  * Read the next element as an unsigned int.
758  * 0 is success, < 0 is failure.
759  */
760 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
761 {
762  int n = 0;
763 
764  if (size > 8)
765  return AVERROR_INVALIDDATA;
766 
767  /* big-endian ordering; build up number */
768  *num = 0;
769  while (n++ < size)
770  *num = (*num << 8) | avio_r8(pb);
771 
772  return 0;
773 }
774 
775 /*
776  * Read the next element as a signed int.
777  * 0 is success, < 0 is failure.
778  */
779 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
780 {
781  int n = 1;
782 
783  if (size > 8)
784  return AVERROR_INVALIDDATA;
785 
786  if (size == 0) {
787  *num = 0;
788  } else {
789  *num = sign_extend(avio_r8(pb), 8);
790 
791  /* big-endian ordering; build up number */
792  while (n++ < size)
793  *num = (*num << 8) | avio_r8(pb);
794  }
795 
796  return 0;
797 }
798 
799 /*
800  * Read the next element as a float.
801  * 0 is success, < 0 is failure.
802  */
803 static int ebml_read_float(AVIOContext *pb, int size, double *num)
804 {
805  if (size == 0)
806  *num = 0;
807  else if (size == 4)
808  *num = av_int2float(avio_rb32(pb));
809  else if (size == 8)
810  *num = av_int2double(avio_rb64(pb));
811  else
812  return AVERROR_INVALIDDATA;
813 
814  return 0;
815 }
816 
817 /*
818  * Read the next element as an ASCII string.
819  * 0 is success, < 0 is failure.
820  */
821 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
822 {
823  char *res;
824 
825  /* EBML strings are usually not 0-terminated, so we allocate one
826  * byte more, read the string and NULL-terminate it ourselves. */
827  if (!(res = av_malloc(size + 1)))
828  return AVERROR(ENOMEM);
829  if (avio_read(pb, (uint8_t *) res, size) != size) {
830  av_free(res);
831  return AVERROR(EIO);
832  }
833  (res)[size] = '\0';
834  av_free(*str);
835  *str = res;
836 
837  return 0;
838 }
839 
840 /*
841  * Read the next element as binary data.
842  * 0 is success, < 0 is failure.
843  */
844 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
845 {
846  av_fast_padded_malloc(&bin->data, &bin->size, length);
847  if (!bin->data)
848  return AVERROR(ENOMEM);
849 
850  bin->size = length;
851  bin->pos = avio_tell(pb);
852  if (avio_read(pb, bin->data, length) != length) {
853  av_freep(&bin->data);
854  bin->size = 0;
855  return AVERROR(EIO);
856  }
857 
858  return 0;
859 }
860 
861 /*
862  * Read the next element, but only the header. The contents
863  * are supposed to be sub-elements which can be read separately.
864  * 0 is success, < 0 is failure.
865  */
866 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
867 {
868  AVIOContext *pb = matroska->ctx->pb;
870 
871  if (matroska->num_levels >= EBML_MAX_DEPTH) {
872  av_log(matroska->ctx, AV_LOG_ERROR,
873  "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
874  return AVERROR(ENOSYS);
875  }
876 
877  level = &matroska->levels[matroska->num_levels++];
878  level->start = avio_tell(pb);
879  level->length = length;
880 
881  return 0;
882 }
883 
884 /*
885  * Read signed/unsigned "EBML" numbers.
886  * Return: number of bytes processed, < 0 on error
887  */
889  uint8_t *data, uint32_t size, uint64_t *num)
890 {
891  AVIOContext pb;
892  ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
893  return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
894 }
895 
896 /*
897  * Same as above, but signed.
898  */
900  uint8_t *data, uint32_t size, int64_t *num)
901 {
902  uint64_t unum;
903  int res;
904 
905  /* read as unsigned number first */
906  if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
907  return res;
908 
909  /* make signed (weird way) */
910  *num = unum - ((1LL << (7 * res - 1)) - 1);
911 
912  return res;
913 }
914 
915 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
916  EbmlSyntax *syntax, void *data);
917 
918 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
919  uint32_t id, void *data)
920 {
921  int i;
922  for (i = 0; syntax[i].id; i++)
923  if (id == syntax[i].id)
924  break;
925  if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
926  matroska->num_levels > 0 &&
927  matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
928  return 0; // we reached the end of an unknown size cluster
929  if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
930  av_log(matroska->ctx, AV_LOG_INFO, "Unknown entry 0x%"PRIX32"\n", id);
931  if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
932  return AVERROR_INVALIDDATA;
933  }
934  return ebml_parse_elem(matroska, &syntax[i], data);
935 }
936 
937 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
938  void *data)
939 {
940  if (!matroska->current_id) {
941  uint64_t id;
942  int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
943  if (res < 0)
944  return res;
945  matroska->current_id = id | 1 << 7 * res;
946  }
947  return ebml_parse_id(matroska, syntax, matroska->current_id, data);
948 }
949 
950 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
951  void *data)
952 {
953  int i, res = 0;
954 
955  for (i = 0; syntax[i].id; i++)
956  switch (syntax[i].type) {
957  case EBML_UINT:
958  *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
959  break;
960  case EBML_FLOAT:
961  *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
962  break;
963  case EBML_STR:
964  case EBML_UTF8:
965  // the default may be NULL
966  if (syntax[i].def.s) {
967  uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
968  *dst = av_strdup(syntax[i].def.s);
969  if (!*dst)
970  return AVERROR(ENOMEM);
971  }
972  break;
973  }
974 
975  while (!res && !ebml_level_end(matroska))
976  res = ebml_parse(matroska, syntax, data);
977 
978  return res;
979 }
980 
982  EbmlSyntax *syntax, void *data)
983 {
984  static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
985  [EBML_UINT] = 8,
986  [EBML_FLOAT] = 8,
987  // max. 16 MB for strings
988  [EBML_STR] = 0x1000000,
989  [EBML_UTF8] = 0x1000000,
990  // max. 256 MB for binary data
991  [EBML_BIN] = 0x10000000,
992  // no limits for anything else
993  };
994  AVIOContext *pb = matroska->ctx->pb;
995  uint32_t id = syntax->id;
996  uint64_t length;
997  int res;
998  void *newelem;
999 
1000  data = (char *) data + syntax->data_offset;
1001  if (syntax->list_elem_size) {
1002  EbmlList *list = data;
1003  newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1004  if (!newelem)
1005  return AVERROR(ENOMEM);
1006  list->elem = newelem;
1007  data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1008  memset(data, 0, syntax->list_elem_size);
1009  list->nb_elem++;
1010  }
1011 
1012  if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1013  matroska->current_id = 0;
1014  if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1015  return res;
1016  if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1017  av_log(matroska->ctx, AV_LOG_ERROR,
1018  "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1019  length, max_lengths[syntax->type], syntax->type);
1020  return AVERROR_INVALIDDATA;
1021  }
1022  }
1023 
1024  switch (syntax->type) {
1025  case EBML_UINT:
1026  res = ebml_read_uint(pb, length, data);
1027  break;
1028  case EBML_SINT:
1029  res = ebml_read_sint(pb, length, data);
1030  break;
1031  case EBML_FLOAT:
1032  res = ebml_read_float(pb, length, data);
1033  break;
1034  case EBML_STR:
1035  case EBML_UTF8:
1036  res = ebml_read_ascii(pb, length, data);
1037  break;
1038  case EBML_BIN:
1039  res = ebml_read_binary(pb, length, data);
1040  break;
1041  case EBML_NEST:
1042  if ((res = ebml_read_master(matroska, length)) < 0)
1043  return res;
1044  if (id == MATROSKA_ID_SEGMENT)
1045  matroska->segment_start = avio_tell(matroska->ctx->pb);
1046  return ebml_parse_nest(matroska, syntax->def.n, data);
1047  case EBML_PASS:
1048  return ebml_parse_id(matroska, syntax->def.n, id, data);
1049  case EBML_STOP:
1050  return 1;
1051  default:
1052  if (ffio_limit(pb, length) != length)
1053  return AVERROR(EIO);
1054  return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
1055  }
1056  if (res == AVERROR_INVALIDDATA)
1057  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1058  else if (res == AVERROR(EIO))
1059  av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1060  return res;
1061 }
1062 
1063 static void ebml_free(EbmlSyntax *syntax, void *data)
1064 {
1065  int i, j;
1066  for (i = 0; syntax[i].id; i++) {
1067  void *data_off = (char *) data + syntax[i].data_offset;
1068  switch (syntax[i].type) {
1069  case EBML_STR:
1070  case EBML_UTF8:
1071  av_freep(data_off);
1072  break;
1073  case EBML_BIN:
1074  av_freep(&((EbmlBin *) data_off)->data);
1075  break;
1076  case EBML_NEST:
1077  if (syntax[i].list_elem_size) {
1078  EbmlList *list = data_off;
1079  char *ptr = list->elem;
1080  for (j = 0; j < list->nb_elem;
1081  j++, ptr += syntax[i].list_elem_size)
1082  ebml_free(syntax[i].def.n, ptr);
1083  av_free(list->elem);
1084  } else
1085  ebml_free(syntax[i].def.n, data_off);
1086  default:
1087  break;
1088  }
1089  }
1090 }
1091 
1092 /*
1093  * Autodetecting...
1094  */
1096 {
1097  uint64_t total = 0;
1098  int len_mask = 0x80, size = 1, n = 1, i;
1099 
1100  /* EBML header? */
1101  if (AV_RB32(p->buf) != EBML_ID_HEADER)
1102  return 0;
1103 
1104  /* length of header */
1105  total = p->buf[4];
1106  while (size <= 8 && !(total & len_mask)) {
1107  size++;
1108  len_mask >>= 1;
1109  }
1110  if (size > 8)
1111  return 0;
1112  total &= (len_mask - 1);
1113  while (n < size)
1114  total = (total << 8) | p->buf[4 + n++];
1115 
1116  /* Does the probe data contain the whole header? */
1117  if (p->buf_size < 4 + size + total)
1118  return 0;
1119 
1120  /* The header should contain a known document type. For now,
1121  * we don't parse the whole header but simply check for the
1122  * availability of that array of characters inside the header.
1123  * Not fully fool-proof, but good enough. */
1124  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1125  int probelen = strlen(matroska_doctypes[i]);
1126  if (total < probelen)
1127  continue;
1128  for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1129  if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1130  return AVPROBE_SCORE_MAX;
1131  }
1132 
1133  // probably valid EBML header but no recognized doctype
1134  return AVPROBE_SCORE_EXTENSION;
1135 }
1136 
1138  int num)
1139 {
1140  MatroskaTrack *tracks = matroska->tracks.elem;
1141  int i;
1142 
1143  for (i = 0; i < matroska->tracks.nb_elem; i++)
1144  if (tracks[i].num == num)
1145  return &tracks[i];
1146 
1147  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1148  return NULL;
1149 }
1150 
1151 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1152  MatroskaTrack *track)
1153 {
1154  MatroskaTrackEncoding *encodings = track->encodings.elem;
1155  uint8_t *data = *buf;
1156  int isize = *buf_size;
1157  uint8_t *pkt_data = NULL;
1158  uint8_t av_unused *newpktdata;
1159  int pkt_size = isize;
1160  int result = 0;
1161  int olen;
1162 
1163  if (pkt_size >= 10000000U)
1164  return AVERROR_INVALIDDATA;
1165 
1166  switch (encodings[0].compression.algo) {
1168  {
1169  int header_size = encodings[0].compression.settings.size;
1170  uint8_t *header = encodings[0].compression.settings.data;
1171 
1172  if (header_size && !header) {
1173  av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1174  return -1;
1175  }
1176 
1177  if (!header_size)
1178  return 0;
1179 
1180  pkt_size = isize + header_size;
1181  pkt_data = av_malloc(pkt_size);
1182  if (!pkt_data)
1183  return AVERROR(ENOMEM);
1184 
1185  memcpy(pkt_data, header, header_size);
1186  memcpy(pkt_data + header_size, data, isize);
1187  break;
1188  }
1189 #if CONFIG_LZO
1191  do {
1192  olen = pkt_size *= 3;
1193  newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1194  if (!newpktdata) {
1195  result = AVERROR(ENOMEM);
1196  goto failed;
1197  }
1198  pkt_data = newpktdata;
1199  result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1200  } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1201  if (result) {
1202  result = AVERROR_INVALIDDATA;
1203  goto failed;
1204  }
1205  pkt_size -= olen;
1206  break;
1207 #endif
1208 #if CONFIG_ZLIB
1210  {
1211  z_stream zstream = { 0 };
1212  if (inflateInit(&zstream) != Z_OK)
1213  return -1;
1214  zstream.next_in = data;
1215  zstream.avail_in = isize;
1216  do {
1217  pkt_size *= 3;
1218  newpktdata = av_realloc(pkt_data, pkt_size);
1219  if (!newpktdata) {
1220  inflateEnd(&zstream);
1221  goto failed;
1222  }
1223  pkt_data = newpktdata;
1224  zstream.avail_out = pkt_size - zstream.total_out;
1225  zstream.next_out = pkt_data + zstream.total_out;
1226  if (pkt_data) {
1227  result = inflate(&zstream, Z_NO_FLUSH);
1228  } else
1229  result = Z_MEM_ERROR;
1230  } while (result == Z_OK && pkt_size < 10000000);
1231  pkt_size = zstream.total_out;
1232  inflateEnd(&zstream);
1233  if (result != Z_STREAM_END) {
1234  if (result == Z_MEM_ERROR)
1235  result = AVERROR(ENOMEM);
1236  else
1237  result = AVERROR_INVALIDDATA;
1238  goto failed;
1239  }
1240  break;
1241  }
1242 #endif
1243 #if CONFIG_BZLIB
1245  {
1246  bz_stream bzstream = { 0 };
1247  if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1248  return -1;
1249  bzstream.next_in = data;
1250  bzstream.avail_in = isize;
1251  do {
1252  pkt_size *= 3;
1253  newpktdata = av_realloc(pkt_data, pkt_size);
1254  if (!newpktdata) {
1255  BZ2_bzDecompressEnd(&bzstream);
1256  goto failed;
1257  }
1258  pkt_data = newpktdata;
1259  bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1260  bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1261  if (pkt_data) {
1262  result = BZ2_bzDecompress(&bzstream);
1263  } else
1264  result = BZ_MEM_ERROR;
1265  } while (result == BZ_OK && pkt_size < 10000000);
1266  pkt_size = bzstream.total_out_lo32;
1267  BZ2_bzDecompressEnd(&bzstream);
1268  if (result != BZ_STREAM_END) {
1269  if (result == BZ_MEM_ERROR)
1270  result = AVERROR(ENOMEM);
1271  else
1272  result = AVERROR_INVALIDDATA;
1273  goto failed;
1274  }
1275  break;
1276  }
1277 #endif
1278  default:
1279  return AVERROR_INVALIDDATA;
1280  }
1281 
1282  *buf = pkt_data;
1283  *buf_size = pkt_size;
1284  return 0;
1285 
1286 failed:
1287  av_free(pkt_data);
1288  return result;
1289 }
1290 
1292  AVDictionary **metadata, char *prefix)
1293 {
1294  MatroskaTag *tags = list->elem;
1295  char key[1024];
1296  int i;
1297 
1298  for (i = 0; i < list->nb_elem; i++) {
1299  const char *lang = tags[i].lang &&
1300  strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1301 
1302  if (!tags[i].name) {
1303  av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1304  continue;
1305  }
1306  if (prefix)
1307  snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1308  else
1309  av_strlcpy(key, tags[i].name, sizeof(key));
1310  if (tags[i].def || !lang) {
1311  av_dict_set(metadata, key, tags[i].string, 0);
1312  if (tags[i].sub.nb_elem)
1313  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1314  }
1315  if (lang) {
1316  av_strlcat(key, "-", sizeof(key));
1317  av_strlcat(key, lang, sizeof(key));
1318  av_dict_set(metadata, key, tags[i].string, 0);
1319  if (tags[i].sub.nb_elem)
1320  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1321  }
1322  }
1323  ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1324 }
1325 
1327 {
1328  MatroskaDemuxContext *matroska = s->priv_data;
1329  MatroskaTags *tags = matroska->tags.elem;
1330  int i, j;
1331 
1332  for (i = 0; i < matroska->tags.nb_elem; i++) {
1333  if (tags[i].target.attachuid) {
1334  MatroskaAttachment *attachment = matroska->attachments.elem;
1335  for (j = 0; j < matroska->attachments.nb_elem; j++)
1336  if (attachment[j].uid == tags[i].target.attachuid &&
1337  attachment[j].stream)
1338  matroska_convert_tag(s, &tags[i].tag,
1339  &attachment[j].stream->metadata, NULL);
1340  } else if (tags[i].target.chapteruid) {
1341  MatroskaChapter *chapter = matroska->chapters.elem;
1342  for (j = 0; j < matroska->chapters.nb_elem; j++)
1343  if (chapter[j].uid == tags[i].target.chapteruid &&
1344  chapter[j].chapter)
1345  matroska_convert_tag(s, &tags[i].tag,
1346  &chapter[j].chapter->metadata, NULL);
1347  } else if (tags[i].target.trackuid) {
1348  MatroskaTrack *track = matroska->tracks.elem;
1349  for (j = 0; j < matroska->tracks.nb_elem; j++)
1350  if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1351  matroska_convert_tag(s, &tags[i].tag,
1352  &track[j].stream->metadata, NULL);
1353  } else {
1354  matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1355  tags[i].target.type);
1356  }
1357  }
1358 }
1359 
1361  int idx)
1362 {
1363  EbmlList *seekhead_list = &matroska->seekhead;
1364  uint32_t level_up = matroska->level_up;
1365  uint32_t saved_id = matroska->current_id;
1366  MatroskaSeekhead *seekhead = seekhead_list->elem;
1367  int64_t before_pos = avio_tell(matroska->ctx->pb);
1369  int64_t offset;
1370  int ret = 0;
1371 
1372  if (idx >= seekhead_list->nb_elem ||
1373  seekhead[idx].id == MATROSKA_ID_SEEKHEAD ||
1374  seekhead[idx].id == MATROSKA_ID_CLUSTER)
1375  return 0;
1376 
1377  /* seek */
1378  offset = seekhead[idx].pos + matroska->segment_start;
1379  if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1380  /* We don't want to lose our seekhead level, so we add
1381  * a dummy. This is a crude hack. */
1382  if (matroska->num_levels == EBML_MAX_DEPTH) {
1383  av_log(matroska->ctx, AV_LOG_INFO,
1384  "Max EBML element depth (%d) reached, "
1385  "cannot parse further.\n", EBML_MAX_DEPTH);
1386  ret = AVERROR_INVALIDDATA;
1387  } else {
1388  level.start = 0;
1389  level.length = (uint64_t) -1;
1390  matroska->levels[matroska->num_levels] = level;
1391  matroska->num_levels++;
1392  matroska->current_id = 0;
1393 
1394  ret = ebml_parse(matroska, matroska_segment, matroska);
1395 
1396  /* remove dummy level */
1397  while (matroska->num_levels) {
1398  uint64_t length = matroska->levels[--matroska->num_levels].length;
1399  if (length == (uint64_t) -1)
1400  break;
1401  }
1402  }
1403  }
1404  /* seek back */
1405  avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1406  matroska->level_up = level_up;
1407  matroska->current_id = saved_id;
1408 
1409  return ret;
1410 }
1411 
1413 {
1414  EbmlList *seekhead_list = &matroska->seekhead;
1415  int64_t before_pos = avio_tell(matroska->ctx->pb);
1416  int i;
1417 
1418  // we should not do any seeking in the streaming case
1419  if (!matroska->ctx->pb->seekable ||
1420  (matroska->ctx->flags & AVFMT_FLAG_IGNIDX))
1421  return;
1422 
1423  for (i = 0; i < seekhead_list->nb_elem; i++) {
1424  MatroskaSeekhead *seekhead = seekhead_list->elem;
1425  if (seekhead[i].pos <= before_pos)
1426  continue;
1427 
1428  // defer cues parsing until we actually need cue data.
1429  if (seekhead[i].id == MATROSKA_ID_CUES) {
1430  matroska->cues_parsing_deferred = 1;
1431  continue;
1432  }
1433 
1434  if (matroska_parse_seekhead_entry(matroska, i) < 0) {
1435  // mark index as broken
1436  matroska->cues_parsing_deferred = -1;
1437  break;
1438  }
1439  }
1440 }
1441 
1443 {
1444  EbmlList *index_list;
1446  int index_scale = 1;
1447  int i, j;
1448 
1449  index_list = &matroska->index;
1450  index = index_list->elem;
1451  if (index_list->nb_elem &&
1452  index[0].time > 1E14 / matroska->time_scale) {
1453  av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1454  index_scale = matroska->time_scale;
1455  }
1456  for (i = 0; i < index_list->nb_elem; i++) {
1457  EbmlList *pos_list = &index[i].pos;
1458  MatroskaIndexPos *pos = pos_list->elem;
1459  for (j = 0; j < pos_list->nb_elem; j++) {
1460  MatroskaTrack *track = matroska_find_track_by_num(matroska,
1461  pos[j].track);
1462  if (track && track->stream)
1463  av_add_index_entry(track->stream,
1464  pos[j].pos + matroska->segment_start,
1465  index[i].time / index_scale, 0, 0,
1467  }
1468  }
1469 }
1470 
1472  EbmlList *seekhead_list = &matroska->seekhead;
1473  MatroskaSeekhead *seekhead = seekhead_list->elem;
1474  int i;
1475 
1476  for (i = 0; i < seekhead_list->nb_elem; i++)
1477  if (seekhead[i].id == MATROSKA_ID_CUES)
1478  break;
1479  av_assert1(i <= seekhead_list->nb_elem);
1480 
1481  if (matroska_parse_seekhead_entry(matroska, i) < 0)
1482  matroska->cues_parsing_deferred = -1;
1483  matroska_add_index_entries(matroska);
1484 }
1485 
1487 {
1488  static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1489  int profile;
1490 
1491  for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1492  if (strstr(codec_id, aac_profiles[profile]))
1493  break;
1494  return profile + 1;
1495 }
1496 
1497 static int matroska_aac_sri(int samplerate)
1498 {
1499  int sri;
1500 
1501  for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1502  if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1503  break;
1504  return sri;
1505 }
1506 
1507 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1508 {
1509  char buffer[32];
1510  /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1511  time_t creation_time = date_utc / 1000000000 + 978307200;
1512  struct tm tmpbuf, *ptm = gmtime_r(&creation_time, &tmpbuf);
1513  if (!ptm) return;
1514  if (strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm))
1515  av_dict_set(metadata, "creation_time", buffer, 0);
1516 }
1517 
1519  MatroskaTrack *track,
1520  int *offset)
1521 {
1522  AVStream *st = track->stream;
1523  uint8_t *p = track->codec_priv.data;
1524  int size = track->codec_priv.size;
1525 
1526  if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1527  av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1528  track->codec_priv.size = 0;
1529  return 0;
1530  }
1531  *offset = 8;
1532  track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1533 
1534  p += track->codec_priv.size;
1535  size -= track->codec_priv.size;
1536 
1537  /* parse the remaining metadata blocks if present */
1538  while (size >= 4) {
1539  int block_last, block_type, block_size;
1540 
1541  flac_parse_block_header(p, &block_last, &block_type, &block_size);
1542 
1543  p += 4;
1544  size -= 4;
1545  if (block_size > size)
1546  return 0;
1547 
1548  /* check for the channel mask */
1549  if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1550  AVDictionary *dict = NULL;
1551  AVDictionaryEntry *chmask;
1552 
1553  ff_vorbis_comment(s, &dict, p, block_size, 0);
1554  chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1555  if (chmask) {
1556  uint64_t mask = strtol(chmask->value, NULL, 0);
1557  if (!mask || mask & ~0x3ffffULL) {
1559  "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1560  } else
1561  st->codec->channel_layout = mask;
1562  }
1563  av_dict_free(&dict);
1564  }
1565 
1566  p += block_size;
1567  size -= block_size;
1568  }
1569 
1570  return 0;
1571 }
1572 
1574 {
1575  MatroskaDemuxContext *matroska = s->priv_data;
1576  MatroskaTrack *tracks = matroska->tracks.elem;
1577  AVStream *st;
1578  int i, j, ret;
1579  int k;
1580 
1581  for (i = 0; i < matroska->tracks.nb_elem; i++) {
1582  MatroskaTrack *track = &tracks[i];
1584  EbmlList *encodings_list = &track->encodings;
1585  MatroskaTrackEncoding *encodings = encodings_list->elem;
1586  uint8_t *extradata = NULL;
1587  int extradata_size = 0;
1588  int extradata_offset = 0;
1589  uint32_t fourcc = 0;
1590  AVIOContext b;
1591  char* key_id_base64 = NULL;
1592  int bit_depth = -1;
1593 
1594  /* Apply some sanity checks. */
1595  if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1596  track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1597  track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1598  track->type != MATROSKA_TRACK_TYPE_METADATA) {
1599  av_log(matroska->ctx, AV_LOG_INFO,
1600  "Unknown or unsupported track type %"PRIu64"\n",
1601  track->type);
1602  continue;
1603  }
1604  if (!track->codec_id)
1605  continue;
1606 
1607  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1608  if (!track->default_duration && track->video.frame_rate > 0)
1609  track->default_duration = 1000000000 / track->video.frame_rate;
1610  if (track->video.display_width == -1)
1611  track->video.display_width = track->video.pixel_width;
1612  if (track->video.display_height == -1)
1613  track->video.display_height = track->video.pixel_height;
1614  if (track->video.color_space.size == 4)
1615  fourcc = AV_RL32(track->video.color_space.data);
1616  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1617  if (!track->audio.out_samplerate)
1618  track->audio.out_samplerate = track->audio.samplerate;
1619  }
1620  if (encodings_list->nb_elem > 1) {
1621  av_log(matroska->ctx, AV_LOG_ERROR,
1622  "Multiple combined encodings not supported");
1623  } else if (encodings_list->nb_elem == 1) {
1624  if (encodings[0].type) {
1625  if (encodings[0].encryption.key_id.size > 0) {
1626  /* Save the encryption key id to be stored later as a
1627  metadata tag. */
1628  const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1629  key_id_base64 = av_malloc(b64_size);
1630  if (key_id_base64 == NULL)
1631  return AVERROR(ENOMEM);
1632 
1633  av_base64_encode(key_id_base64, b64_size,
1634  encodings[0].encryption.key_id.data,
1635  encodings[0].encryption.key_id.size);
1636  } else {
1637  encodings[0].scope = 0;
1638  av_log(matroska->ctx, AV_LOG_ERROR,
1639  "Unsupported encoding type");
1640  }
1641  } else if (
1642 #if CONFIG_ZLIB
1643  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1644 #endif
1645 #if CONFIG_BZLIB
1647 #endif
1648 #if CONFIG_LZO
1650 #endif
1652  encodings[0].scope = 0;
1653  av_log(matroska->ctx, AV_LOG_ERROR,
1654  "Unsupported encoding type");
1655  } else if (track->codec_priv.size && encodings[0].scope & 2) {
1656  uint8_t *codec_priv = track->codec_priv.data;
1657  int ret = matroska_decode_buffer(&track->codec_priv.data,
1658  &track->codec_priv.size,
1659  track);
1660  if (ret < 0) {
1661  track->codec_priv.data = NULL;
1662  track->codec_priv.size = 0;
1663  av_log(matroska->ctx, AV_LOG_ERROR,
1664  "Failed to decode codec private data\n");
1665  }
1666 
1667  if (codec_priv != track->codec_priv.data)
1668  av_free(codec_priv);
1669  }
1670  }
1671 
1672  for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
1673  if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1674  strlen(ff_mkv_codec_tags[j].str))) {
1675  codec_id = ff_mkv_codec_tags[j].id;
1676  break;
1677  }
1678  }
1679 
1680  st = track->stream = avformat_new_stream(s, NULL);
1681  if (!st) {
1682  av_free(key_id_base64);
1683  return AVERROR(ENOMEM);
1684  }
1685 
1686  if (key_id_base64) {
1687  /* export encryption key id as base64 metadata tag */
1688  av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
1689  av_freep(&key_id_base64);
1690  }
1691 
1692  if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
1693  track->codec_priv.size >= 40 &&
1694  track->codec_priv.data) {
1695  track->ms_compat = 1;
1696  bit_depth = AV_RL16(track->codec_priv.data + 14);
1697  fourcc = AV_RL32(track->codec_priv.data + 16);
1699  fourcc);
1700  if (!codec_id)
1702  fourcc);
1703  extradata_offset = 40;
1704  } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
1705  track->codec_priv.size >= 14 &&
1706  track->codec_priv.data) {
1707  int ret;
1708  ffio_init_context(&b, track->codec_priv.data,
1709  track->codec_priv.size,
1710  0, NULL, NULL, NULL, NULL);
1711  ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size);
1712  if (ret < 0)
1713  return ret;
1714  codec_id = st->codec->codec_id;
1715  extradata_offset = FFMIN(track->codec_priv.size, 18);
1716  } else if (!strcmp(track->codec_id, "A_QUICKTIME")
1717  && (track->codec_priv.size >= 86)
1718  && (track->codec_priv.data)) {
1719  fourcc = AV_RL32(track->codec_priv.data + 4);
1720  codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1722  fourcc = AV_RL32(track->codec_priv.data);
1723  codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1724  }
1725  } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
1726  (track->codec_priv.size >= 21) &&
1727  (track->codec_priv.data)) {
1728  fourcc = AV_RL32(track->codec_priv.data + 4);
1729  codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1731  fourcc = AV_RL32(track->codec_priv.data);
1732  codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1733  }
1734  if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI "))
1735  codec_id = AV_CODEC_ID_SVQ3;
1736  } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1737  switch (track->audio.bitdepth) {
1738  case 8:
1739  codec_id = AV_CODEC_ID_PCM_U8;
1740  break;
1741  case 24:
1742  codec_id = AV_CODEC_ID_PCM_S24BE;
1743  break;
1744  case 32:
1745  codec_id = AV_CODEC_ID_PCM_S32BE;
1746  break;
1747  }
1748  } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1749  switch (track->audio.bitdepth) {
1750  case 8:
1751  codec_id = AV_CODEC_ID_PCM_U8;
1752  break;
1753  case 24:
1754  codec_id = AV_CODEC_ID_PCM_S24LE;
1755  break;
1756  case 32:
1757  codec_id = AV_CODEC_ID_PCM_S32LE;
1758  break;
1759  }
1760  } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
1761  track->audio.bitdepth == 64) {
1762  codec_id = AV_CODEC_ID_PCM_F64LE;
1763  } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1764  int profile = matroska_aac_profile(track->codec_id);
1765  int sri = matroska_aac_sri(track->audio.samplerate);
1766  extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1767  if (!extradata)
1768  return AVERROR(ENOMEM);
1769  extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
1770  extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
1771  if (strstr(track->codec_id, "SBR")) {
1772  sri = matroska_aac_sri(track->audio.out_samplerate);
1773  extradata[2] = 0x56;
1774  extradata[3] = 0xE5;
1775  extradata[4] = 0x80 | (sri << 3);
1776  extradata_size = 5;
1777  } else
1778  extradata_size = 2;
1779  } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - FF_INPUT_BUFFER_PADDING_SIZE) {
1780  /* Only ALAC's magic cookie is stored in Matroska's track headers.
1781  * Create the "atom size", "tag", and "tag version" fields the
1782  * decoder expects manually. */
1783  extradata_size = 12 + track->codec_priv.size;
1784  extradata = av_mallocz(extradata_size +
1786  if (!extradata)
1787  return AVERROR(ENOMEM);
1788  AV_WB32(extradata, extradata_size);
1789  memcpy(&extradata[4], "alac", 4);
1790  AV_WB32(&extradata[8], 0);
1791  memcpy(&extradata[12], track->codec_priv.data,
1792  track->codec_priv.size);
1793  } else if (codec_id == AV_CODEC_ID_TTA) {
1794  extradata_size = 30;
1795  extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1796  if (!extradata)
1797  return AVERROR(ENOMEM);
1798  ffio_init_context(&b, extradata, extradata_size, 1,
1799  NULL, NULL, NULL, NULL);
1800  avio_write(&b, "TTA1", 4);
1801  avio_wl16(&b, 1);
1802  avio_wl16(&b, track->audio.channels);
1803  avio_wl16(&b, track->audio.bitdepth);
1804  if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
1805  return AVERROR_INVALIDDATA;
1806  avio_wl32(&b, track->audio.out_samplerate);
1807  avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
1808  track->audio.out_samplerate,
1809  AV_TIME_BASE * 1000));
1810  } else if (codec_id == AV_CODEC_ID_RV10 ||
1811  codec_id == AV_CODEC_ID_RV20 ||
1812  codec_id == AV_CODEC_ID_RV30 ||
1813  codec_id == AV_CODEC_ID_RV40) {
1814  extradata_offset = 26;
1815  } else if (codec_id == AV_CODEC_ID_RA_144) {
1816  track->audio.out_samplerate = 8000;
1817  track->audio.channels = 1;
1818  } else if ((codec_id == AV_CODEC_ID_RA_288 ||
1819  codec_id == AV_CODEC_ID_COOK ||
1820  codec_id == AV_CODEC_ID_ATRAC3 ||
1821  codec_id == AV_CODEC_ID_SIPR)
1822  && track->codec_priv.data) {
1823  int flavor;
1824 
1825  ffio_init_context(&b, track->codec_priv.data,
1826  track->codec_priv.size,
1827  0, NULL, NULL, NULL, NULL);
1828  avio_skip(&b, 22);
1829  flavor = avio_rb16(&b);
1830  track->audio.coded_framesize = avio_rb32(&b);
1831  avio_skip(&b, 12);
1832  track->audio.sub_packet_h = avio_rb16(&b);
1833  track->audio.frame_size = avio_rb16(&b);
1834  track->audio.sub_packet_size = avio_rb16(&b);
1835  if (flavor < 0 ||
1836  track->audio.coded_framesize <= 0 ||
1837  track->audio.sub_packet_h <= 0 ||
1838  track->audio.frame_size <= 0 ||
1839  track->audio.sub_packet_size <= 0)
1840  return AVERROR_INVALIDDATA;
1841  track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
1842  track->audio.frame_size);
1843  if (!track->audio.buf)
1844  return AVERROR(ENOMEM);
1845  if (codec_id == AV_CODEC_ID_RA_288) {
1846  st->codec->block_align = track->audio.coded_framesize;
1847  track->codec_priv.size = 0;
1848  } else {
1849  if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1850  static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1851  track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1852  st->codec->bit_rate = sipr_bit_rate[flavor];
1853  }
1854  st->codec->block_align = track->audio.sub_packet_size;
1855  extradata_offset = 78;
1856  }
1857  } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
1858  ret = matroska_parse_flac(s, track, &extradata_offset);
1859  if (ret < 0)
1860  return ret;
1861  } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
1862  fourcc = AV_RL32(track->codec_priv.data);
1863  }
1864  track->codec_priv.size -= extradata_offset;
1865 
1866  if (codec_id == AV_CODEC_ID_NONE)
1867  av_log(matroska->ctx, AV_LOG_INFO,
1868  "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1869 
1870  if (track->time_scale < 0.01)
1871  track->time_scale = 1.0;
1872  avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
1873  1000 * 1000 * 1000); /* 64 bit pts in ns */
1874 
1875  /* convert the delay from ns to the track timebase */
1876  track->codec_delay = av_rescale_q(track->codec_delay,
1877  (AVRational){ 1, 1000000000 },
1878  st->time_base);
1879 
1880  st->codec->codec_id = codec_id;
1881 
1882  if (strcmp(track->language, "und"))
1883  av_dict_set(&st->metadata, "language", track->language, 0);
1884  av_dict_set(&st->metadata, "title", track->name, 0);
1885 
1886  if (track->flag_default)
1888  if (track->flag_forced)
1890 
1891  if (!st->codec->extradata) {
1892  if (extradata) {
1893  st->codec->extradata = extradata;
1894  st->codec->extradata_size = extradata_size;
1895  } else if (track->codec_priv.data && track->codec_priv.size > 0) {
1896  if (ff_alloc_extradata(st->codec, track->codec_priv.size))
1897  return AVERROR(ENOMEM);
1898  memcpy(st->codec->extradata,
1899  track->codec_priv.data + extradata_offset,
1900  track->codec_priv.size);
1901  }
1902  }
1903 
1904  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1905  MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
1906 
1908  st->codec->codec_tag = fourcc;
1909  if (bit_depth >= 0)
1910  st->codec->bits_per_coded_sample = bit_depth;
1911  st->codec->width = track->video.pixel_width;
1912  st->codec->height = track->video.pixel_height;
1914  &st->sample_aspect_ratio.den,
1915  st->codec->height * track->video.display_width,
1916  st->codec->width * track->video.display_height,
1917  255);
1918  if (st->codec->codec_id != AV_CODEC_ID_HEVC)
1920 
1921  if (track->default_duration) {
1923  1000000000, track->default_duration, 30000);
1924 #if FF_API_R_FRAME_RATE
1925  if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000L
1926  && st->avg_frame_rate.num > st->avg_frame_rate.den * 5L)
1927  st->r_frame_rate = st->avg_frame_rate;
1928 #endif
1929  }
1930 
1931  /* export stereo mode flag as metadata tag */
1932  if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
1933  av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
1934 
1935  /* export alpha mode flag as metadata tag */
1936  if (track->video.alpha_mode)
1937  av_dict_set(&st->metadata, "alpha_mode", "1", 0);
1938 
1939  /* if we have virtual track, mark the real tracks */
1940  for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
1941  char buf[32];
1942  if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
1943  continue;
1944  snprintf(buf, sizeof(buf), "%s_%d",
1945  ff_matroska_video_stereo_plane[planes[j].type], i);
1946  for (k=0; k < matroska->tracks.nb_elem; k++)
1947  if (planes[j].uid == tracks[k].uid) {
1948  av_dict_set(&s->streams[k]->metadata,
1949  "stereo_mode", buf, 0);
1950  break;
1951  }
1952  }
1953  // add stream level stereo3d side data if it is a supported format
1954  if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
1955  track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
1956  int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
1957  if (ret < 0)
1958  return ret;
1959  }
1960  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1962  st->codec->sample_rate = track->audio.out_samplerate;
1963  st->codec->channels = track->audio.channels;
1964  if (!st->codec->bits_per_coded_sample)
1965  st->codec->bits_per_coded_sample = track->audio.bitdepth;
1966  if (st->codec->codec_id != AV_CODEC_ID_AAC)
1968  if (track->codec_delay > 0) {
1969  st->codec->delay = av_rescale_q(track->codec_delay,
1970  st->time_base,
1971  (AVRational){1, st->codec->sample_rate});
1972  }
1973  if (track->seek_preroll > 0) {
1975  av_rescale_q(track->seek_preroll,
1976  (AVRational){1, 1000000000},
1977  (AVRational){1, st->codec->sample_rate}));
1978  }
1979  } else if (codec_id == AV_CODEC_ID_WEBVTT) {
1980  st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1981 
1982  if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
1983  st->disposition |= AV_DISPOSITION_CAPTIONS;
1984  } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
1985  st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
1986  } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
1987  st->disposition |= AV_DISPOSITION_METADATA;
1988  }
1989  } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
1990  st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1991  if (st->codec->codec_id == AV_CODEC_ID_ASS)
1992  matroska->contains_ssa = 1;
1993  }
1994  }
1995 
1996  return 0;
1997 }
1998 
2000 {
2001  MatroskaDemuxContext *matroska = s->priv_data;
2002  EbmlList *attachments_list = &matroska->attachments;
2003  EbmlList *chapters_list = &matroska->chapters;
2004  MatroskaAttachment *attachments;
2005  MatroskaChapter *chapters;
2006  uint64_t max_start = 0;
2007  int64_t pos;
2008  Ebml ebml = { 0 };
2009  int i, j, res;
2010 
2011  matroska->ctx = s;
2012 
2013  /* First read the EBML header. */
2014  if (ebml_parse(matroska, ebml_syntax, &ebml) ||
2015  ebml.version > EBML_VERSION ||
2016  ebml.max_size > sizeof(uint64_t) ||
2017  ebml.id_length > sizeof(uint32_t) ||
2018  ebml.doctype_version > 3 ||
2019  !ebml.doctype) {
2020  av_log(matroska->ctx, AV_LOG_ERROR,
2021  "EBML header using unsupported features\n"
2022  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2023  ebml.version, ebml.doctype, ebml.doctype_version);
2024  ebml_free(ebml_syntax, &ebml);
2025  return AVERROR_PATCHWELCOME;
2026  } else if (ebml.doctype_version == 3) {
2027  av_log(matroska->ctx, AV_LOG_WARNING,
2028  "EBML header using unsupported features\n"
2029  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2030  ebml.version, ebml.doctype, ebml.doctype_version);
2031  }
2032  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2033  if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2034  break;
2035  if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2036  av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2037  if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2038  ebml_free(ebml_syntax, &ebml);
2039  return AVERROR_INVALIDDATA;
2040  }
2041  }
2042  ebml_free(ebml_syntax, &ebml);
2043 
2044  /* The next thing is a segment. */
2045  pos = avio_tell(matroska->ctx->pb);
2046  res = ebml_parse(matroska, matroska_segments, matroska);
2047  // try resyncing until we find a EBML_STOP type element.
2048  while (res != 1) {
2049  res = matroska_resync(matroska, pos);
2050  if (res < 0)
2051  return res;
2052  pos = avio_tell(matroska->ctx->pb);
2053  res = ebml_parse(matroska, matroska_segment, matroska);
2054  }
2055  matroska_execute_seekhead(matroska);
2056 
2057  if (!matroska->time_scale)
2058  matroska->time_scale = 1000000;
2059  if (matroska->duration)
2060  matroska->ctx->duration = matroska->duration * matroska->time_scale *
2061  1000 / AV_TIME_BASE;
2062  av_dict_set(&s->metadata, "title", matroska->title, 0);
2063  av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2064 
2065  if (matroska->date_utc.size == 8)
2067 
2068  res = matroska_parse_tracks(s);
2069  if (res < 0)
2070  return res;
2071 
2072  attachments = attachments_list->elem;
2073  for (j = 0; j < attachments_list->nb_elem; j++) {
2074  if (!(attachments[j].filename && attachments[j].mime &&
2075  attachments[j].bin.data && attachments[j].bin.size > 0)) {
2076  av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2077  } else {
2078  AVStream *st = avformat_new_stream(s, NULL);
2079  if (!st)
2080  break;
2081  av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2082  av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2085  if (ff_alloc_extradata(st->codec, attachments[j].bin.size))
2086  break;
2087  memcpy(st->codec->extradata, attachments[j].bin.data,
2088  attachments[j].bin.size);
2089 
2090  for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2091  if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2092  strlen(ff_mkv_mime_tags[i].str))) {
2093  st->codec->codec_id = ff_mkv_mime_tags[i].id;
2094  break;
2095  }
2096  }
2097  attachments[j].stream = st;
2098  }
2099  }
2100 
2101  chapters = chapters_list->elem;
2102  for (i = 0; i < chapters_list->nb_elem; i++)
2103  if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2104  (max_start == 0 || chapters[i].start > max_start)) {
2105  chapters[i].chapter =
2106  avpriv_new_chapter(s, chapters[i].uid,
2107  (AVRational) { 1, 1000000000 },
2108  chapters[i].start, chapters[i].end,
2109  chapters[i].title);
2110  if (chapters[i].chapter) {
2111  av_dict_set(&chapters[i].chapter->metadata,
2112  "title", chapters[i].title, 0);
2113  }
2114  max_start = chapters[i].start;
2115  }
2116 
2117  matroska_add_index_entries(matroska);
2118 
2120 
2121  return 0;
2122 }
2123 
2124 /*
2125  * Put one packet in an application-supplied AVPacket struct.
2126  * Returns 0 on success or -1 on failure.
2127  */
2129  AVPacket *pkt)
2130 {
2131  if (matroska->num_packets > 0) {
2132  memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2133  av_free(matroska->packets[0]);
2134  if (matroska->num_packets > 1) {
2135  void *newpackets;
2136  memmove(&matroska->packets[0], &matroska->packets[1],
2137  (matroska->num_packets - 1) * sizeof(AVPacket *));
2138  newpackets = av_realloc(matroska->packets,
2139  (matroska->num_packets - 1) *
2140  sizeof(AVPacket *));
2141  if (newpackets)
2142  matroska->packets = newpackets;
2143  } else {
2144  av_freep(&matroska->packets);
2145  matroska->prev_pkt = NULL;
2146  }
2147  matroska->num_packets--;
2148  return 0;
2149  }
2150 
2151  return -1;
2152 }
2153 
2154 /*
2155  * Free all packets in our internal queue.
2156  */
2158 {
2159  matroska->prev_pkt = NULL;
2160  if (matroska->packets) {
2161  int n;
2162  for (n = 0; n < matroska->num_packets; n++) {
2163  av_free_packet(matroska->packets[n]);
2164  av_free(matroska->packets[n]);
2165  }
2166  av_freep(&matroska->packets);
2167  matroska->num_packets = 0;
2168  }
2169 }
2170 
2172  int *buf_size, int type,
2173  uint32_t **lace_buf, int *laces)
2174 {
2175  int res = 0, n, size = *buf_size;
2176  uint8_t *data = *buf;
2177  uint32_t *lace_size;
2178 
2179  if (!type) {
2180  *laces = 1;
2181  *lace_buf = av_mallocz(sizeof(int));
2182  if (!*lace_buf)
2183  return AVERROR(ENOMEM);
2184 
2185  *lace_buf[0] = size;
2186  return 0;
2187  }
2188 
2189  av_assert0(size > 0);
2190  *laces = *data + 1;
2191  data += 1;
2192  size -= 1;
2193  lace_size = av_mallocz(*laces * sizeof(int));
2194  if (!lace_size)
2195  return AVERROR(ENOMEM);
2196 
2197  switch (type) {
2198  case 0x1: /* Xiph lacing */
2199  {
2200  uint8_t temp;
2201  uint32_t total = 0;
2202  for (n = 0; res == 0 && n < *laces - 1; n++) {
2203  while (1) {
2204  if (size <= total) {
2205  res = AVERROR_INVALIDDATA;
2206  break;
2207  }
2208  temp = *data;
2209  total += temp;
2210  lace_size[n] += temp;
2211  data += 1;
2212  size -= 1;
2213  if (temp != 0xff)
2214  break;
2215  }
2216  }
2217  if (size <= total) {
2218  res = AVERROR_INVALIDDATA;
2219  break;
2220  }
2221 
2222  lace_size[n] = size - total;
2223  break;
2224  }
2225 
2226  case 0x2: /* fixed-size lacing */
2227  if (size % (*laces)) {
2228  res = AVERROR_INVALIDDATA;
2229  break;
2230  }
2231  for (n = 0; n < *laces; n++)
2232  lace_size[n] = size / *laces;
2233  break;
2234 
2235  case 0x3: /* EBML lacing */
2236  {
2237  uint64_t num;
2238  uint64_t total;
2239  n = matroska_ebmlnum_uint(matroska, data, size, &num);
2240  if (n < 0 || num > INT_MAX) {
2241  av_log(matroska->ctx, AV_LOG_INFO,
2242  "EBML block data error\n");
2243  res = n<0 ? n : AVERROR_INVALIDDATA;
2244  break;
2245  }
2246  data += n;
2247  size -= n;
2248  total = lace_size[0] = num;
2249  for (n = 1; res == 0 && n < *laces - 1; n++) {
2250  int64_t snum;
2251  int r;
2252  r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2253  if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2254  av_log(matroska->ctx, AV_LOG_INFO,
2255  "EBML block data error\n");
2256  res = r<0 ? r : AVERROR_INVALIDDATA;
2257  break;
2258  }
2259  data += r;
2260  size -= r;
2261  lace_size[n] = lace_size[n - 1] + snum;
2262  total += lace_size[n];
2263  }
2264  if (size <= total) {
2265  res = AVERROR_INVALIDDATA;
2266  break;
2267  }
2268  lace_size[*laces - 1] = size - total;
2269  break;
2270  }
2271  }
2272 
2273  *buf = data;
2274  *lace_buf = lace_size;
2275  *buf_size = size;
2276 
2277  return res;
2278 }
2279 
2281  MatroskaTrack *track, AVStream *st,
2282  uint8_t *data, int size, uint64_t timecode,
2283  int64_t pos)
2284 {
2285  int a = st->codec->block_align;
2286  int sps = track->audio.sub_packet_size;
2287  int cfs = track->audio.coded_framesize;
2288  int h = track->audio.sub_packet_h;
2289  int y = track->audio.sub_packet_cnt;
2290  int w = track->audio.frame_size;
2291  int x;
2292 
2293  if (!track->audio.pkt_cnt) {
2294  if (track->audio.sub_packet_cnt == 0)
2295  track->audio.buf_timecode = timecode;
2296  if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2297  if (size < cfs * h / 2) {
2298  av_log(matroska->ctx, AV_LOG_ERROR,
2299  "Corrupt int4 RM-style audio packet size\n");
2300  return AVERROR_INVALIDDATA;
2301  }
2302  for (x = 0; x < h / 2; x++)
2303  memcpy(track->audio.buf + x * 2 * w + y * cfs,
2304  data + x * cfs, cfs);
2305  } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2306  if (size < w) {
2307  av_log(matroska->ctx, AV_LOG_ERROR,
2308  "Corrupt sipr RM-style audio packet size\n");
2309  return AVERROR_INVALIDDATA;
2310  }
2311  memcpy(track->audio.buf + y * w, data, w);
2312  } else {
2313  if (size < sps * w / sps || h<=0 || w%sps) {
2314  av_log(matroska->ctx, AV_LOG_ERROR,
2315  "Corrupt generic RM-style audio packet size\n");
2316  return AVERROR_INVALIDDATA;
2317  }
2318  for (x = 0; x < w / sps; x++)
2319  memcpy(track->audio.buf +
2320  sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2321  data + x * sps, sps);
2322  }
2323 
2324  if (++track->audio.sub_packet_cnt >= h) {
2325  if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2326  ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2327  track->audio.sub_packet_cnt = 0;
2328  track->audio.pkt_cnt = h * w / a;
2329  }
2330  }
2331 
2332  while (track->audio.pkt_cnt) {
2333  int ret;
2334  AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2335  if (!pkt)
2336  return AVERROR(ENOMEM);
2337 
2338  ret = av_new_packet(pkt, a);
2339  if (ret < 0) {
2340  av_free(pkt);
2341  return ret;
2342  }
2343  memcpy(pkt->data,
2344  track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2345  a);
2346  pkt->pts = track->audio.buf_timecode;
2348  pkt->pos = pos;
2349  pkt->stream_index = st->index;
2350  dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2351  }
2352 
2353  return 0;
2354 }
2355 
2356 /* reconstruct full wavpack blocks from mangled matroska ones */
2358  uint8_t **pdst, int *size)
2359 {
2360  uint8_t *dst = NULL;
2361  int dstlen = 0;
2362  int srclen = *size;
2363  uint32_t samples;
2364  uint16_t ver;
2365  int ret, offset = 0;
2366 
2367  if (srclen < 12 || track->stream->codec->extradata_size < 2)
2368  return AVERROR_INVALIDDATA;
2369 
2370  ver = AV_RL16(track->stream->codec->extradata);
2371 
2372  samples = AV_RL32(src);
2373  src += 4;
2374  srclen -= 4;
2375 
2376  while (srclen >= 8) {
2377  int multiblock;
2378  uint32_t blocksize;
2379  uint8_t *tmp;
2380 
2381  uint32_t flags = AV_RL32(src);
2382  uint32_t crc = AV_RL32(src + 4);
2383  src += 8;
2384  srclen -= 8;
2385 
2386  multiblock = (flags & 0x1800) != 0x1800;
2387  if (multiblock) {
2388  if (srclen < 4) {
2389  ret = AVERROR_INVALIDDATA;
2390  goto fail;
2391  }
2392  blocksize = AV_RL32(src);
2393  src += 4;
2394  srclen -= 4;
2395  } else
2396  blocksize = srclen;
2397 
2398  if (blocksize > srclen) {
2399  ret = AVERROR_INVALIDDATA;
2400  goto fail;
2401  }
2402 
2403  tmp = av_realloc(dst, dstlen + blocksize + 32);
2404  if (!tmp) {
2405  ret = AVERROR(ENOMEM);
2406  goto fail;
2407  }
2408  dst = tmp;
2409  dstlen += blocksize + 32;
2410 
2411  AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2412  AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2413  AV_WL16(dst + offset + 8, ver); // version
2414  AV_WL16(dst + offset + 10, 0); // track/index_no
2415  AV_WL32(dst + offset + 12, 0); // total samples
2416  AV_WL32(dst + offset + 16, 0); // block index
2417  AV_WL32(dst + offset + 20, samples); // number of samples
2418  AV_WL32(dst + offset + 24, flags); // flags
2419  AV_WL32(dst + offset + 28, crc); // crc
2420  memcpy(dst + offset + 32, src, blocksize); // block data
2421 
2422  src += blocksize;
2423  srclen -= blocksize;
2424  offset += blocksize + 32;
2425  }
2426 
2427  *pdst = dst;
2428  *size = dstlen;
2429 
2430  return 0;
2431 
2432 fail:
2433  av_freep(&dst);
2434  return ret;
2435 }
2436 
2438  MatroskaTrack *track,
2439  AVStream *st,
2440  uint8_t *data, int data_len,
2441  uint64_t timecode,
2442  uint64_t duration,
2443  int64_t pos)
2444 {
2445  AVPacket *pkt;
2446  uint8_t *id, *settings, *text, *buf;
2447  int id_len, settings_len, text_len;
2448  uint8_t *p, *q;
2449  int err;
2450 
2451  if (data_len <= 0)
2452  return AVERROR_INVALIDDATA;
2453 
2454  p = data;
2455  q = data + data_len;
2456 
2457  id = p;
2458  id_len = -1;
2459  while (p < q) {
2460  if (*p == '\r' || *p == '\n') {
2461  id_len = p - id;
2462  if (*p == '\r')
2463  p++;
2464  break;
2465  }
2466  p++;
2467  }
2468 
2469  if (p >= q || *p != '\n')
2470  return AVERROR_INVALIDDATA;
2471  p++;
2472 
2473  settings = p;
2474  settings_len = -1;
2475  while (p < q) {
2476  if (*p == '\r' || *p == '\n') {
2477  settings_len = p - settings;
2478  if (*p == '\r')
2479  p++;
2480  break;
2481  }
2482  p++;
2483  }
2484 
2485  if (p >= q || *p != '\n')
2486  return AVERROR_INVALIDDATA;
2487  p++;
2488 
2489  text = p;
2490  text_len = q - p;
2491  while (text_len > 0) {
2492  const int len = text_len - 1;
2493  const uint8_t c = p[len];
2494  if (c != '\r' && c != '\n')
2495  break;
2496  text_len = len;
2497  }
2498 
2499  if (text_len <= 0)
2500  return AVERROR_INVALIDDATA;
2501 
2502  pkt = av_mallocz(sizeof(*pkt));
2503  err = av_new_packet(pkt, text_len);
2504  if (err < 0) {
2505  av_free(pkt);
2506  return AVERROR(err);
2507  }
2508 
2509  memcpy(pkt->data, text, text_len);
2510 
2511  if (id_len > 0) {
2512  buf = av_packet_new_side_data(pkt,
2514  id_len);
2515  if (!buf) {
2516  av_free(pkt);
2517  return AVERROR(ENOMEM);
2518  }
2519  memcpy(buf, id, id_len);
2520  }
2521 
2522  if (settings_len > 0) {
2523  buf = av_packet_new_side_data(pkt,
2525  settings_len);
2526  if (!buf) {
2527  av_free(pkt);
2528  return AVERROR(ENOMEM);
2529  }
2530  memcpy(buf, settings, settings_len);
2531  }
2532 
2533  // Do we need this for subtitles?
2534  // pkt->flags = AV_PKT_FLAG_KEY;
2535 
2536  pkt->stream_index = st->index;
2537  pkt->pts = timecode;
2538 
2539  // Do we need this for subtitles?
2540  // pkt->dts = timecode;
2541 
2542  pkt->duration = duration;
2543  pkt->pos = pos;
2544 
2545  dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2546  matroska->prev_pkt = pkt;
2547 
2548  return 0;
2549 }
2550 
2552  MatroskaTrack *track, AVStream *st,
2553  uint8_t *data, int pkt_size,
2554  uint64_t timecode, uint64_t lace_duration,
2555  int64_t pos, int is_keyframe,
2556  uint8_t *additional, uint64_t additional_id, int additional_size,
2557  int64_t discard_padding)
2558 {
2559  MatroskaTrackEncoding *encodings = track->encodings.elem;
2560  uint8_t *pkt_data = data;
2561  int offset = 0, res;
2562  AVPacket *pkt;
2563 
2564  if (encodings && !encodings->type && encodings->scope & 1) {
2565  res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2566  if (res < 0)
2567  return res;
2568  }
2569 
2570  if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
2571  uint8_t *wv_data;
2572  res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2573  if (res < 0) {
2574  av_log(matroska->ctx, AV_LOG_ERROR,
2575  "Error parsing a wavpack block.\n");
2576  goto fail;
2577  }
2578  if (pkt_data != data)
2579  av_freep(&pkt_data);
2580  pkt_data = wv_data;
2581  }
2582 
2583  if (st->codec->codec_id == AV_CODEC_ID_PRORES &&
2584  AV_RB32(&data[4]) != MKBETAG('i', 'c', 'p', 'f'))
2585  offset = 8;
2586 
2587  pkt = av_mallocz(sizeof(AVPacket));
2588  /* XXX: prevent data copy... */
2589  if (av_new_packet(pkt, pkt_size + offset) < 0) {
2590  av_free(pkt);
2591  res = AVERROR(ENOMEM);
2592  goto fail;
2593  }
2594 
2595  if (st->codec->codec_id == AV_CODEC_ID_PRORES && offset == 8) {
2596  uint8_t *buf = pkt->data;
2597  bytestream_put_be32(&buf, pkt_size);
2598  bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2599  }
2600 
2601  memcpy(pkt->data + offset, pkt_data, pkt_size);
2602 
2603  if (pkt_data != data)
2604  av_freep(&pkt_data);
2605 
2606  pkt->flags = is_keyframe;
2607  pkt->stream_index = st->index;
2608 
2609  if (additional_size > 0) {
2610  uint8_t *side_data = av_packet_new_side_data(pkt,
2612  additional_size + 8);
2613  if (!side_data) {
2614  av_free_packet(pkt);
2615  av_free(pkt);
2616  return AVERROR(ENOMEM);
2617  }
2618  AV_WB64(side_data, additional_id);
2619  memcpy(side_data + 8, additional, additional_size);
2620  }
2621 
2622  if (discard_padding) {
2623  uint8_t *side_data = av_packet_new_side_data(pkt,
2625  10);
2626  if (!side_data) {
2627  av_free_packet(pkt);
2628  av_free(pkt);
2629  return AVERROR(ENOMEM);
2630  }
2631  AV_WL32(side_data, 0);
2632  AV_WL32(side_data + 4, av_rescale_q(discard_padding,
2633  (AVRational){1, 1000000000},
2634  (AVRational){1, st->codec->sample_rate}));
2635  }
2636 
2637  if (track->ms_compat)
2638  pkt->dts = timecode;
2639  else
2640  pkt->pts = timecode;
2641  pkt->pos = pos;
2642  if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2643  /*
2644  * For backward compatibility.
2645  * Historically, we have put subtitle duration
2646  * in convergence_duration, on the off chance
2647  * that the time_scale is less than 1us, which
2648  * could result in a 32bit overflow on the
2649  * normal duration field.
2650  */
2651  pkt->convergence_duration = lace_duration;
2652  }
2653 
2654  if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
2655  lace_duration <= INT_MAX) {
2656  /*
2657  * For non subtitle tracks, just store the duration
2658  * as normal.
2659  *
2660  * If it's a subtitle track and duration value does
2661  * not overflow a uint32, then also store it normally.
2662  */
2663  pkt->duration = lace_duration;
2664  }
2665 
2666  dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2667  matroska->prev_pkt = pkt;
2668 
2669  return 0;
2670 
2671 fail:
2672  if (pkt_data != data)
2673  av_freep(&pkt_data);
2674  return res;
2675 }
2676 
2678  int size, int64_t pos, uint64_t cluster_time,
2679  uint64_t block_duration, int is_keyframe,
2680  uint8_t *additional, uint64_t additional_id, int additional_size,
2681  int64_t cluster_pos, int64_t discard_padding)
2682 {
2683  uint64_t timecode = AV_NOPTS_VALUE;
2684  MatroskaTrack *track;
2685  int res = 0;
2686  AVStream *st;
2687  int16_t block_time;
2688  uint32_t *lace_size = NULL;
2689  int n, flags, laces = 0;
2690  uint64_t num;
2691  int trust_default_duration = 1;
2692 
2693  if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2694  av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2695  return n;
2696  }
2697  data += n;
2698  size -= n;
2699 
2700  track = matroska_find_track_by_num(matroska, num);
2701  if (!track || !track->stream) {
2702  av_log(matroska->ctx, AV_LOG_INFO,
2703  "Invalid stream %"PRIu64" or size %u\n", num, size);
2704  return AVERROR_INVALIDDATA;
2705  } else if (size <= 3)
2706  return 0;
2707  st = track->stream;
2708  if (st->discard >= AVDISCARD_ALL)
2709  return res;
2710  av_assert1(block_duration != AV_NOPTS_VALUE);
2711 
2712  block_time = sign_extend(AV_RB16(data), 16);
2713  data += 2;
2714  flags = *data++;
2715  size -= 3;
2716  if (is_keyframe == -1)
2717  is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2718 
2719  if (cluster_time != (uint64_t) -1 &&
2720  (block_time >= 0 || cluster_time >= -block_time)) {
2721  timecode = cluster_time + block_time - track->codec_delay;
2722  if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2723  timecode < track->end_timecode)
2724  is_keyframe = 0; /* overlapping subtitles are not key frame */
2725  if (is_keyframe)
2726  av_add_index_entry(st, cluster_pos, timecode, 0, 0,
2728  }
2729 
2730  if (matroska->skip_to_keyframe &&
2731  track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2732  if (timecode < matroska->skip_to_timecode)
2733  return res;
2734  if (is_keyframe)
2735  matroska->skip_to_keyframe = 0;
2736  else if (!st->skip_to_keyframe) {
2737  av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2738  matroska->skip_to_keyframe = 0;
2739  }
2740  }
2741 
2742  res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2743  &lace_size, &laces);
2744 
2745  if (res)
2746  goto end;
2747 
2748  if (track->audio.samplerate == 8000) {
2749  // If this is needed for more codecs, then add them here
2750  if (st->codec->codec_id == AV_CODEC_ID_AC3) {
2751  if (track->audio.samplerate != st->codec->sample_rate || !st->codec->frame_size)
2752  trust_default_duration = 0;
2753  }
2754  }
2755 
2756  if (!block_duration && trust_default_duration)
2757  block_duration = track->default_duration * laces / matroska->time_scale;
2758 
2759  if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2760  track->end_timecode =
2761  FFMAX(track->end_timecode, timecode + block_duration);
2762 
2763  for (n = 0; n < laces; n++) {
2764  int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2765 
2766  if (lace_size[n] > size) {
2767  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2768  break;
2769  }
2770 
2771  if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2772  st->codec->codec_id == AV_CODEC_ID_COOK ||
2773  st->codec->codec_id == AV_CODEC_ID_SIPR ||
2774  st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2775  st->codec->block_align && track->audio.sub_packet_size) {
2776  res = matroska_parse_rm_audio(matroska, track, st, data,
2777  lace_size[n],
2778  timecode, pos);
2779  if (res)
2780  goto end;
2781 
2782  } else if (st->codec->codec_id == AV_CODEC_ID_WEBVTT) {
2783  res = matroska_parse_webvtt(matroska, track, st,
2784  data, lace_size[n],
2785  timecode, lace_duration,
2786  pos);
2787  if (res)
2788  goto end;
2789  } else {
2790  res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2791  timecode, lace_duration, pos,
2792  !n ? is_keyframe : 0,
2793  additional, additional_id, additional_size,
2794  discard_padding);
2795  if (res)
2796  goto end;
2797  }
2798 
2799  if (timecode != AV_NOPTS_VALUE)
2800  timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2801  data += lace_size[n];
2802  size -= lace_size[n];
2803  }
2804 
2805 end:
2806  av_free(lace_size);
2807  return res;
2808 }
2809 
2811 {
2812  EbmlList *blocks_list;
2813  MatroskaBlock *blocks;
2814  int i, res;
2815  res = ebml_parse(matroska,
2816  matroska_cluster_incremental_parsing,
2817  &matroska->current_cluster);
2818  if (res == 1) {
2819  /* New Cluster */
2820  if (matroska->current_cluster_pos)
2821  ebml_level_end(matroska);
2822  ebml_free(matroska_cluster, &matroska->current_cluster);
2823  memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2824  matroska->current_cluster_num_blocks = 0;
2825  matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2826  matroska->prev_pkt = NULL;
2827  /* sizeof the ID which was already read */
2828  if (matroska->current_id)
2829  matroska->current_cluster_pos -= 4;
2830  res = ebml_parse(matroska,
2831  matroska_clusters_incremental,
2832  &matroska->current_cluster);
2833  /* Try parsing the block again. */
2834  if (res == 1)
2835  res = ebml_parse(matroska,
2836  matroska_cluster_incremental_parsing,
2837  &matroska->current_cluster);
2838  }
2839 
2840  if (!res &&
2841  matroska->current_cluster_num_blocks <
2842  matroska->current_cluster.blocks.nb_elem) {
2843  blocks_list = &matroska->current_cluster.blocks;
2844  blocks = blocks_list->elem;
2845 
2846  matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2847  i = blocks_list->nb_elem - 1;
2848  if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2849  int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2850  uint8_t* additional = blocks[i].additional.size > 0 ?
2851  blocks[i].additional.data : NULL;
2852  if (!blocks[i].non_simple)
2853  blocks[i].duration = 0;
2854  res = matroska_parse_block(matroska, blocks[i].bin.data,
2855  blocks[i].bin.size, blocks[i].bin.pos,
2856  matroska->current_cluster.timecode,
2857  blocks[i].duration, is_keyframe,
2858  additional, blocks[i].additional_id,
2859  blocks[i].additional.size,
2860  matroska->current_cluster_pos,
2861  blocks[i].discard_padding);
2862  }
2863  }
2864 
2865  return res;
2866 }
2867 
2869 {
2870  MatroskaCluster cluster = { 0 };
2871  EbmlList *blocks_list;
2872  MatroskaBlock *blocks;
2873  int i, res;
2874  int64_t pos;
2875 
2876  if (!matroska->contains_ssa)
2877  return matroska_parse_cluster_incremental(matroska);
2878  pos = avio_tell(matroska->ctx->pb);
2879  matroska->prev_pkt = NULL;
2880  if (matroska->current_id)
2881  pos -= 4; /* sizeof the ID which was already read */
2882  res = ebml_parse(matroska, matroska_clusters, &cluster);
2883  blocks_list = &cluster.blocks;
2884  blocks = blocks_list->elem;
2885  for (i = 0; i < blocks_list->nb_elem; i++)
2886  if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2887  int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2888  res = matroska_parse_block(matroska, blocks[i].bin.data,
2889  blocks[i].bin.size, blocks[i].bin.pos,
2890  cluster.timecode, blocks[i].duration,
2891  is_keyframe, NULL, 0, 0, pos,
2892  blocks[i].discard_padding);
2893  }
2894  ebml_free(matroska_cluster, &cluster);
2895  return res;
2896 }
2897 
2899 {
2900  MatroskaDemuxContext *matroska = s->priv_data;
2901 
2902  while (matroska_deliver_packet(matroska, pkt)) {
2903  int64_t pos = avio_tell(matroska->ctx->pb);
2904  if (matroska->done)
2905  return AVERROR_EOF;
2906  if (matroska_parse_cluster(matroska) < 0)
2907  matroska_resync(matroska, pos);
2908  }
2909 
2910  return 0;
2911 }
2912 
2913 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2914  int64_t timestamp, int flags)
2915 {
2916  MatroskaDemuxContext *matroska = s->priv_data;
2917  MatroskaTrack *tracks = NULL;
2918  AVStream *st = s->streams[stream_index];
2919  int i, index, index_sub, index_min;
2920 
2921  /* Parse the CUES now since we need the index data to seek. */
2922  if (matroska->cues_parsing_deferred > 0) {
2923  matroska->cues_parsing_deferred = 0;
2924  matroska_parse_cues(matroska);
2925  }
2926 
2927  if (!st->nb_index_entries)
2928  goto err;
2929  timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2930 
2931  if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
2932  avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
2933  SEEK_SET);
2934  matroska->current_id = 0;
2935  while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
2936  matroska_clear_queue(matroska);
2937  if (matroska_parse_cluster(matroska) < 0)
2938  break;
2939  }
2940  }
2941 
2942  matroska_clear_queue(matroska);
2943  if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
2944  goto err;
2945 
2946  index_min = index;
2947  tracks = matroska->tracks.elem;
2948  for (i = 0; i < matroska->tracks.nb_elem; i++) {
2949  tracks[i].audio.pkt_cnt = 0;
2950  tracks[i].audio.sub_packet_cnt = 0;
2951  tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
2952  tracks[i].end_timecode = 0;
2953  if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2954  tracks[i].stream->discard != AVDISCARD_ALL) {
2955  index_sub = av_index_search_timestamp(
2956  tracks[i].stream, st->index_entries[index].timestamp,
2958  while (index_sub >= 0 &&
2959  index_min > 0 &&
2960  tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
2961  st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
2962  index_min--;
2963  }
2964  }
2965 
2966  avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
2967  matroska->current_id = 0;
2968  if (flags & AVSEEK_FLAG_ANY) {
2969  st->skip_to_keyframe = 0;
2970  matroska->skip_to_timecode = timestamp;
2971  } else {
2972  st->skip_to_keyframe = 1;
2973  matroska->skip_to_timecode = st->index_entries[index].timestamp;
2974  }
2975  matroska->skip_to_keyframe = 1;
2976  matroska->done = 0;
2977  matroska->num_levels = 0;
2978  ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
2979  return 0;
2980 err:
2981  // slightly hackish but allows proper fallback to
2982  // the generic seeking code.
2983  matroska_clear_queue(matroska);
2984  matroska->current_id = 0;
2985  st->skip_to_keyframe =
2986  matroska->skip_to_keyframe = 0;
2987  matroska->done = 0;
2988  matroska->num_levels = 0;
2989  return -1;
2990 }
2991 
2993 {
2994  MatroskaDemuxContext *matroska = s->priv_data;
2995  MatroskaTrack *tracks = matroska->tracks.elem;
2996  int n;
2997 
2998  matroska_clear_queue(matroska);
2999 
3000  for (n = 0; n < matroska->tracks.nb_elem; n++)
3001  if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3002  av_free(tracks[n].audio.buf);
3003  ebml_free(matroska_cluster, &matroska->current_cluster);
3004  ebml_free(matroska_segment, matroska);
3005 
3006  return 0;
3007 }
3008 
3009 typedef struct {
3010  int64_t start_time_ns;
3011  int64_t end_time_ns;
3012  int64_t start_offset;
3013  int64_t end_offset;
3014 } CueDesc;
3015 
3016 /* This function searches all the Cues and returns the CueDesc corresponding the
3017  * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3018  * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3019  */
3020 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3021  MatroskaDemuxContext *matroska = s->priv_data;
3022  CueDesc cue_desc;
3023  int i;
3024  int nb_index_entries = s->streams[0]->nb_index_entries;
3025  AVIndexEntry *index_entries = s->streams[0]->index_entries;
3026  if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3027  for (i = 1; i < nb_index_entries; i++) {
3028  if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3029  index_entries[i].timestamp * matroska->time_scale > ts) {
3030  break;
3031  }
3032  }
3033  --i;
3034  cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3035  cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3036  if (i != nb_index_entries - 1) {
3037  cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3038  cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3039  } else {
3040  cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3041  // FIXME: this needs special handling for files where Cues appear
3042  // before Clusters. the current logic assumes Cues appear after
3043  // Clusters.
3044  cue_desc.end_offset = cues_start - matroska->segment_start;
3045  }
3046  return cue_desc;
3047 }
3048 
3050 {
3051  MatroskaDemuxContext *matroska = s->priv_data;
3052  int64_t cluster_pos, before_pos;
3053  int index, rv = 1;
3054  if (s->streams[0]->nb_index_entries <= 0) return 0;
3055  // seek to the first cluster using cues.
3056  index = av_index_search_timestamp(s->streams[0], 0, 0);
3057  if (index < 0) return 0;
3058  cluster_pos = s->streams[0]->index_entries[index].pos;
3059  before_pos = avio_tell(s->pb);
3060  while (1) {
3061  int64_t cluster_id = 0, cluster_length = 0;
3062  AVPacket *pkt;
3063  avio_seek(s->pb, cluster_pos, SEEK_SET);
3064  // read cluster id and length
3065  ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id);
3066  ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3067  if (cluster_id != 0xF43B675) { // done with all clusters
3068  break;
3069  }
3070  avio_seek(s->pb, cluster_pos, SEEK_SET);
3071  matroska->current_id = 0;
3072  matroska_clear_queue(matroska);
3073  if (matroska_parse_cluster(matroska) < 0 ||
3074  matroska->num_packets <= 0) {
3075  break;
3076  }
3077  pkt = matroska->packets[0];
3078  cluster_pos += cluster_length + 12; // 12 is the offset of the cluster id and length.
3079  if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3080  rv = 0;
3081  break;
3082  }
3083  }
3084  avio_seek(s->pb, before_pos, SEEK_SET);
3085  return rv;
3086 }
3087 
3088 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3089  double min_buffer, double* buffer,
3090  double* sec_to_download, AVFormatContext *s,
3091  int64_t cues_start)
3092 {
3093  double nano_seconds_per_second = 1000000000.0;
3094  double time_sec = time_ns / nano_seconds_per_second;
3095  int rv = 0;
3096  int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3097  int64_t end_time_ns = time_ns + time_to_search_ns;
3098  double sec_downloaded = 0.0;
3099  CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3100  if (desc_curr.start_time_ns == -1)
3101  return -1;
3102  *sec_to_download = 0.0;
3103 
3104  // Check for non cue start time.
3105  if (time_ns > desc_curr.start_time_ns) {
3106  int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3107  double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3108  double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3109  double timeToDownload = (cueBytes * 8.0) / bps;
3110 
3111  sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3112  *sec_to_download += timeToDownload;
3113 
3114  // Check if the search ends within the first cue.
3115  if (desc_curr.end_time_ns >= end_time_ns) {
3116  double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3117  double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3118  sec_downloaded = percent_to_sub * sec_downloaded;
3119  *sec_to_download = percent_to_sub * *sec_to_download;
3120  }
3121 
3122  if ((sec_downloaded + *buffer) <= min_buffer) {
3123  return 1;
3124  }
3125 
3126  // Get the next Cue.
3127  desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3128  }
3129 
3130  while (desc_curr.start_time_ns != -1) {
3131  int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3132  int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3133  double desc_sec = desc_ns / nano_seconds_per_second;
3134  double bits = (desc_bytes * 8.0);
3135  double time_to_download = bits / bps;
3136 
3137  sec_downloaded += desc_sec - time_to_download;
3138  *sec_to_download += time_to_download;
3139 
3140  if (desc_curr.end_time_ns >= end_time_ns) {
3141  double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3142  double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3143  sec_downloaded = percent_to_sub * sec_downloaded;
3144  *sec_to_download = percent_to_sub * *sec_to_download;
3145 
3146  if ((sec_downloaded + *buffer) <= min_buffer)
3147  rv = 1;
3148  break;
3149  }
3150 
3151  if ((sec_downloaded + *buffer) <= min_buffer) {
3152  rv = 1;
3153  break;
3154  }
3155 
3156  desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3157  }
3158  *buffer = *buffer + sec_downloaded;
3159  return rv;
3160 }
3161 
3162 /* This function computes the bandwidth of the WebM file with the help of
3163  * buffer_size_after_time_downloaded() function. Both of these functions are
3164  * adapted from WebM Tools project and are adapted to work with FFmpeg's
3165  * Matroska parsing mechanism.
3166  *
3167  * Returns the bandwidth of the file on success; -1 on error.
3168  * */
3169 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3170 {
3171  MatroskaDemuxContext *matroska = s->priv_data;
3172  AVStream *st = s->streams[0];
3173  double bandwidth = 0.0;
3174  int i;
3175 
3176  for (i = 0; i < st->nb_index_entries; i++) {
3177  int64_t prebuffer_ns = 1000000000;
3178  int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3179  double nano_seconds_per_second = 1000000000.0;
3180  int64_t prebuffered_ns = time_ns + prebuffer_ns;
3181  double prebuffer_bytes = 0.0;
3182  int64_t temp_prebuffer_ns = prebuffer_ns;
3183  int64_t pre_bytes, pre_ns;
3184  double pre_sec, prebuffer, bits_per_second;
3185  CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3186 
3187  // Start with the first Cue.
3188  CueDesc desc_end = desc_beg;
3189 
3190  // Figure out how much data we have downloaded for the prebuffer. This will
3191  // be used later to adjust the bits per sample to try.
3192  while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3193  // Prebuffered the entire Cue.
3194  prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3195  temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3196  desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3197  }
3198  if (desc_end.start_time_ns == -1) {
3199  // The prebuffer is larger than the duration.
3200  if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3201  return -1;
3202  bits_per_second = 0.0;
3203  } else {
3204  // The prebuffer ends in the last Cue. Estimate how much data was
3205  // prebuffered.
3206  pre_bytes = desc_end.end_offset - desc_end.start_offset;
3207  pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3208  pre_sec = pre_ns / nano_seconds_per_second;
3209  prebuffer_bytes +=
3210  pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3211 
3212  prebuffer = prebuffer_ns / nano_seconds_per_second;
3213 
3214  // Set this to 0.0 in case our prebuffer buffers the entire video.
3215  bits_per_second = 0.0;
3216  do {
3217  int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3218  int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3219  double desc_sec = desc_ns / nano_seconds_per_second;
3220  double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3221 
3222  // Drop the bps by the percentage of bytes buffered.
3223  double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3224  double mod_bits_per_second = calc_bits_per_second * percent;
3225 
3226  if (prebuffer < desc_sec) {
3227  double search_sec =
3228  (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3229 
3230  // Add 1 so the bits per second should be a little bit greater than file
3231  // datarate.
3232  int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3233  const double min_buffer = 0.0;
3234  double buffer = prebuffer;
3235  double sec_to_download = 0.0;
3236 
3237  int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3238  min_buffer, &buffer, &sec_to_download,
3239  s, cues_start);
3240  if (rv < 0) {
3241  return -1;
3242  } else if (rv == 0) {
3243  bits_per_second = (double)(bps);
3244  break;
3245  }
3246  }
3247 
3248  desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3249  } while (desc_end.start_time_ns != -1);
3250  }
3251  if (bandwidth < bits_per_second) bandwidth = bits_per_second;
3252  }
3253  return (int64_t)bandwidth;
3254 }
3255 
3257 {
3258  MatroskaDemuxContext *matroska = s->priv_data;
3259  EbmlList *seekhead_list = &matroska->seekhead;
3260  MatroskaSeekhead *seekhead = seekhead_list->elem;
3261  char *buf;
3262  int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
3263  int i;
3264 
3265  // determine cues start and end positions
3266  for (i = 0; i < seekhead_list->nb_elem; i++)
3267  if (seekhead[i].id == MATROSKA_ID_CUES)
3268  break;
3269 
3270  if (i >= seekhead_list->nb_elem) return -1;
3271 
3272  before_pos = avio_tell(matroska->ctx->pb);
3273  cues_start = seekhead[i].pos + matroska->segment_start;
3274  if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
3275  // cues_end is computed as cues_start + cues_length + length of the
3276  // Cues element ID + EBML length of the Cues element. cues_end is
3277  // inclusive and the above sum is reduced by 1.
3278  uint64_t cues_length = 0, cues_id = 0, bytes_read = 0;
3279  bytes_read += ebml_read_num(matroska, matroska->ctx->pb, 4, &cues_id);
3280  bytes_read += ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
3281  cues_end = cues_start + cues_length + bytes_read - 1;
3282  }
3283  avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
3284  if (cues_start == -1 || cues_end == -1) return -1;
3285 
3286  // parse the cues
3287  matroska_parse_cues(matroska);
3288 
3289  // cues start
3290  av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
3291 
3292  // cues end
3293  av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
3294 
3295  // bandwidth
3296  bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
3297  if (bandwidth < 0) return -1;
3298  av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
3299 
3300  // check if all clusters start with key frames
3302 
3303  // store cue point timestamps as a comma separated list for checking subsegment alignment in
3304  // the muxer. assumes that each timestamp cannot be more than 20 characters long.
3305  buf = av_malloc(s->streams[0]->nb_index_entries * 20 * sizeof(char));
3306  if (!buf) return -1;
3307  strcpy(buf, "");
3308  for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
3309  snprintf(buf, (i + 1) * 20 * sizeof(char),
3310  "%s%" PRId64, buf, s->streams[0]->index_entries[i].timestamp);
3311  if (i != s->streams[0]->nb_index_entries - 1)
3312  strncat(buf, ",", sizeof(char));
3313  }
3314  av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
3315  av_free(buf);
3316 
3317  return 0;
3318 }
3319 
3321 {
3322  char *buf;
3323  int ret = matroska_read_header(s);
3324  MatroskaTrack *tracks;
3325  MatroskaDemuxContext *matroska = s->priv_data;
3326  if (ret) {
3327  av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
3328  return -1;
3329  }
3330 
3331  // initialization range
3332  // 5 is the offset of Cluster ID.
3334 
3335  // basename of the file
3336  buf = strrchr(s->filename, '/');
3337  av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->filename, 0);
3338 
3339  // duration
3340  buf = av_asprintf("%g", matroska->duration);
3341  if (!buf) return AVERROR(ENOMEM);
3342  av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
3343  av_free(buf);
3344 
3345  // track number
3346  tracks = matroska->tracks.elem;
3347  av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
3348 
3349  // parse the cues and populate Cue related fields
3350  return webm_dash_manifest_cues(s);
3351 }
3352 
3354 {
3355  return AVERROR_EOF;
3356 }
3357 
3359  .name = "matroska,webm",
3360  .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
3361  .extensions = "mkv,mk3d,mka,mks",
3362  .priv_data_size = sizeof(MatroskaDemuxContext),
3368  .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
3369 };
3370 
3372  .name = "webm_dash_manifest",
3373  .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
3374  .priv_data_size = sizeof(MatroskaDemuxContext),
3378 };