2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The FFmpeg Project
5 * This file is part of FFmpeg.
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.
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.
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
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/
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"
43 #include "libavutil/time_internal.h"
45 #include "libavcodec/bytestream.h"
46 #include "libavcodec/flac.h"
47 #include "libavcodec/mpeg4audio.h"
50 #include "avio_internal.h"
55 /* For ff_codec_get_id(). */
81 typedef const struct EbmlSyntax {
90 const struct EbmlSyntax *n;
94 typedef struct EbmlList {
99 typedef struct EbmlBin {
105 typedef struct Ebml {
110 uint64_t doctype_version;
113 typedef struct MatroskaTrackCompression {
116 } MatroskaTrackCompression;
118 typedef struct MatroskaTrackEncryption {
121 } MatroskaTrackEncryption;
123 typedef struct MatroskaTrackEncoding {
126 MatroskaTrackCompression compression;
127 MatroskaTrackEncryption encryption;
128 } MatroskaTrackEncoding;
130 typedef struct MatroskaTrackVideo {
132 uint64_t display_width;
133 uint64_t display_height;
134 uint64_t pixel_width;
135 uint64_t pixel_height;
137 uint64_t stereo_mode;
139 } MatroskaTrackVideo;
141 typedef struct MatroskaTrackAudio {
143 double out_samplerate;
147 /* real audio header (extracted from extradata) */
154 uint64_t buf_timecode;
156 } MatroskaTrackAudio;
158 typedef struct MatroskaTrackPlane {
161 } MatroskaTrackPlane;
163 typedef struct MatroskaTrackOperation {
164 EbmlList combine_planes;
165 } MatroskaTrackOperation;
167 typedef struct MatroskaTrack {
176 uint64_t default_duration;
177 uint64_t flag_default;
178 uint64_t flag_forced;
179 uint64_t seek_preroll;
180 MatroskaTrackVideo video;
181 MatroskaTrackAudio audio;
182 MatroskaTrackOperation operation;
184 uint64_t codec_delay;
187 int64_t end_timecode;
189 uint64_t max_block_additional_id;
192 typedef struct MatroskaAttachment {
199 } MatroskaAttachment;
201 typedef struct MatroskaChapter {
210 typedef struct MatroskaIndexPos {
215 typedef struct MatroskaIndex {
220 typedef struct MatroskaTag {
228 typedef struct MatroskaTagTarget {
236 typedef struct MatroskaTags {
237 MatroskaTagTarget target;
241 typedef struct MatroskaSeekhead {
246 typedef struct MatroskaLevel {
251 typedef struct MatroskaCluster {
256 typedef struct MatroskaLevel1Element {
260 } MatroskaLevel1Element;
262 typedef struct MatroskaDemuxContext {
263 AVFormatContext *ctx;
267 MatroskaLevel levels[EBML_MAX_DEPTH];
277 EbmlList attachments;
283 /* byte position of the segment inside the stream */
284 int64_t segment_start;
286 /* the packet queue */
293 /* What to skip before effectively reading a packet. */
294 int skip_to_keyframe;
295 uint64_t skip_to_timecode;
297 /* File has a CUES element, but we defer parsing until it is needed. */
298 int cues_parsing_deferred;
300 /* Level1 elements and whether they were read yet */
301 MatroskaLevel1Element level1_elems[64];
302 int num_level1_elems;
304 int current_cluster_num_blocks;
305 int64_t current_cluster_pos;
306 MatroskaCluster current_cluster;
308 /* File has SSA subtitles which prevent incremental cluster parsing. */
310 } MatroskaDemuxContext;
312 typedef struct MatroskaBlock {
317 uint64_t additional_id;
319 int64_t discard_padding;
322 static EbmlSyntax ebml_header[] = {
323 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
324 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
325 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
326 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
327 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
328 { EBML_ID_EBMLVERSION, EBML_NONE },
329 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
333 static EbmlSyntax ebml_syntax[] = {
334 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
338 static EbmlSyntax matroska_info[] = {
339 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
340 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
341 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
342 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
343 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
344 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
345 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
349 static EbmlSyntax matroska_track_video[] = {
350 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
351 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
352 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
353 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
354 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
355 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
356 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
357 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
358 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
359 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
360 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
361 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
362 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_NONE },
363 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
364 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
368 static EbmlSyntax matroska_track_audio[] = {
369 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
370 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
371 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
372 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
376 static EbmlSyntax matroska_track_encoding_compression[] = {
377 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
378 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
382 static EbmlSyntax matroska_track_encoding_encryption[] = {
383 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
384 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
385 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
386 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
387 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
388 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
389 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
392 static EbmlSyntax matroska_track_encoding[] = {
393 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
394 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
395 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
396 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
397 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
401 static EbmlSyntax matroska_track_encodings[] = {
402 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
406 static EbmlSyntax matroska_track_plane[] = {
407 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
408 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
412 static EbmlSyntax matroska_track_combine_planes[] = {
413 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
417 static EbmlSyntax matroska_track_operation[] = {
418 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
422 static EbmlSyntax matroska_track[] = {
423 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
424 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
425 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
426 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
427 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
428 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
429 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
430 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
431 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
432 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
433 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
434 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
435 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
436 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
437 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
438 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
439 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
440 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
441 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
442 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
443 { MATROSKA_ID_CODECNAME, EBML_NONE },
444 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
445 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
446 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
447 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
448 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
452 static EbmlSyntax matroska_tracks[] = {
453 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
457 static EbmlSyntax matroska_attachment[] = {
458 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
459 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
460 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
461 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
462 { MATROSKA_ID_FILEDESC, EBML_NONE },
466 static EbmlSyntax matroska_attachments[] = {
467 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
471 static EbmlSyntax matroska_chapter_display[] = {
472 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
473 { MATROSKA_ID_CHAPLANG, EBML_NONE },
477 static EbmlSyntax matroska_chapter_entry[] = {
478 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
479 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
480 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
481 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
482 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
483 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
484 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
485 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
489 static EbmlSyntax matroska_chapter[] = {
490 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
491 { MATROSKA_ID_EDITIONUID, EBML_NONE },
492 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
493 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
494 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
498 static EbmlSyntax matroska_chapters[] = {
499 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
503 static EbmlSyntax matroska_index_pos[] = {
504 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
505 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
506 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
507 { MATROSKA_ID_CUEDURATION, EBML_NONE },
508 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
512 static EbmlSyntax matroska_index_entry[] = {
513 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
514 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
518 static EbmlSyntax matroska_index[] = {
519 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
523 static EbmlSyntax matroska_simpletag[] = {
524 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
525 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
526 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
527 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
528 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
529 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
533 static EbmlSyntax matroska_tagtargets[] = {
534 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
535 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
536 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
537 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
538 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
542 static EbmlSyntax matroska_tag[] = {
543 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
544 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
548 static EbmlSyntax matroska_tags[] = {
549 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
553 static EbmlSyntax matroska_seekhead_entry[] = {
554 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
555 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
559 static EbmlSyntax matroska_seekhead[] = {
560 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
564 static EbmlSyntax matroska_segment[] = {
565 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
566 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
567 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
568 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
569 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
570 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
571 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
572 { MATROSKA_ID_CLUSTER, EBML_STOP },
576 static EbmlSyntax matroska_segments[] = {
577 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
581 static EbmlSyntax matroska_blockmore[] = {
582 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
583 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
587 static EbmlSyntax matroska_blockadditions[] = {
588 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
592 static EbmlSyntax matroska_blockgroup[] = {
593 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
594 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
595 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
596 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
597 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
598 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference) },
599 { MATROSKA_ID_CODECSTATE, EBML_NONE },
600 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
604 static EbmlSyntax matroska_cluster[] = {
605 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
606 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
607 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
608 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
609 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
613 static EbmlSyntax matroska_clusters[] = {
614 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
615 { MATROSKA_ID_INFO, EBML_NONE },
616 { MATROSKA_ID_CUES, EBML_NONE },
617 { MATROSKA_ID_TAGS, EBML_NONE },
618 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
622 static EbmlSyntax matroska_cluster_incremental_parsing[] = {
623 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
624 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
625 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
626 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
627 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
628 { MATROSKA_ID_INFO, EBML_NONE },
629 { MATROSKA_ID_CUES, EBML_NONE },
630 { MATROSKA_ID_TAGS, EBML_NONE },
631 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
632 { MATROSKA_ID_CLUSTER, EBML_STOP },
636 static EbmlSyntax matroska_cluster_incremental[] = {
637 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
638 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
639 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
640 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
641 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
645 static EbmlSyntax matroska_clusters_incremental[] = {
646 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
647 { MATROSKA_ID_INFO, EBML_NONE },
648 { MATROSKA_ID_CUES, EBML_NONE },
649 { MATROSKA_ID_TAGS, EBML_NONE },
650 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
654 static const char *const matroska_doctypes[] = { "matroska", "webm" };
656 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
658 AVIOContext *pb = matroska->ctx->pb;
660 matroska->current_id = 0;
661 matroska->num_levels = 0;
663 /* seek to next position to resync from */
664 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
669 // try to find a toplevel element
670 while (!avio_feof(pb)) {
671 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
672 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
673 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
674 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
675 matroska->current_id = id;
678 id = (id << 8) | avio_r8(pb);
687 * Return: Whether we reached the end of a level in the hierarchy or not.
689 static int ebml_level_end(MatroskaDemuxContext *matroska)
691 AVIOContext *pb = matroska->ctx->pb;
692 int64_t pos = avio_tell(pb);
694 if (matroska->num_levels > 0) {
695 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
696 if (pos - level->start >= level->length || matroska->current_id) {
697 matroska->num_levels--;
705 * Read: an "EBML number", which is defined as a variable-length
706 * array of bytes. The first byte indicates the length by giving a
707 * number of 0-bits followed by a one. The position of the first
708 * "one" bit inside the first byte indicates the length of this
710 * Returns: number of bytes read, < 0 on error
712 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
713 int max_size, uint64_t *number)
718 /* The first byte tells us the length in bytes - avio_r8() can normally
719 * return 0, but since that's not a valid first ebmlID byte, we can
720 * use it safely here to catch EOS. */
721 if (!(total = avio_r8(pb))) {
722 /* we might encounter EOS here */
723 if (!avio_feof(pb)) {
724 int64_t pos = avio_tell(pb);
725 av_log(matroska->ctx, AV_LOG_ERROR,
726 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
728 return pb->error ? pb->error : AVERROR(EIO);
733 /* get the length of the EBML number */
734 read = 8 - ff_log2_tab[total];
735 if (read > max_size) {
736 int64_t pos = avio_tell(pb) - 1;
737 av_log(matroska->ctx, AV_LOG_ERROR,
738 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
739 (uint8_t) total, pos, pos);
740 return AVERROR_INVALIDDATA;
743 /* read out length */
744 total ^= 1 << ff_log2_tab[total];
746 total = (total << 8) | avio_r8(pb);
754 * Read a EBML length value.
755 * This needs special handling for the "unknown length" case which has multiple
758 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
761 int res = ebml_read_num(matroska, pb, 8, number);
762 if (res > 0 && *number + 1 == 1ULL << (7 * res))
763 *number = 0xffffffffffffffULL;
768 * Read the next element as an unsigned int.
769 * 0 is success, < 0 is failure.
771 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
776 return AVERROR_INVALIDDATA;
778 /* big-endian ordering; build up number */
781 *num = (*num << 8) | avio_r8(pb);
787 * Read the next element as a signed int.
788 * 0 is success, < 0 is failure.
790 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
795 return AVERROR_INVALIDDATA;
800 *num = sign_extend(avio_r8(pb), 8);
802 /* big-endian ordering; build up number */
804 *num = (*num << 8) | avio_r8(pb);
811 * Read the next element as a float.
812 * 0 is success, < 0 is failure.
814 static int ebml_read_float(AVIOContext *pb, int size, double *num)
819 *num = av_int2float(avio_rb32(pb));
821 *num = av_int2double(avio_rb64(pb));
823 return AVERROR_INVALIDDATA;
829 * Read the next element as an ASCII string.
830 * 0 is success, < 0 is failure.
832 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
836 /* EBML strings are usually not 0-terminated, so we allocate one
837 * byte more, read the string and NULL-terminate it ourselves. */
838 if (!(res = av_malloc(size + 1)))
839 return AVERROR(ENOMEM);
840 if (avio_read(pb, (uint8_t *) res, size) != size) {
852 * Read the next element as binary data.
853 * 0 is success, < 0 is failure.
855 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
857 av_fast_padded_malloc(&bin->data, &bin->size, length);
859 return AVERROR(ENOMEM);
862 bin->pos = avio_tell(pb);
863 if (avio_read(pb, bin->data, length) != length) {
864 av_freep(&bin->data);
873 * Read the next element, but only the header. The contents
874 * are supposed to be sub-elements which can be read separately.
875 * 0 is success, < 0 is failure.
877 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
879 AVIOContext *pb = matroska->ctx->pb;
880 MatroskaLevel *level;
882 if (matroska->num_levels >= EBML_MAX_DEPTH) {
883 av_log(matroska->ctx, AV_LOG_ERROR,
884 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
885 return AVERROR(ENOSYS);
888 level = &matroska->levels[matroska->num_levels++];
889 level->start = avio_tell(pb);
890 level->length = length;
896 * Read signed/unsigned "EBML" numbers.
897 * Return: number of bytes processed, < 0 on error
899 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
900 uint8_t *data, uint32_t size, uint64_t *num)
903 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
904 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
908 * Same as above, but signed.
910 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
911 uint8_t *data, uint32_t size, int64_t *num)
916 /* read as unsigned number first */
917 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
920 /* make signed (weird way) */
921 *num = unum - ((1LL << (7 * res - 1)) - 1);
926 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
927 EbmlSyntax *syntax, void *data);
929 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
930 uint32_t id, void *data)
933 for (i = 0; syntax[i].id; i++)
934 if (id == syntax[i].id)
936 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
937 matroska->num_levels > 0 &&
938 matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
939 return 0; // we reached the end of an unknown size cluster
940 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
941 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32"\n", id);
943 return ebml_parse_elem(matroska, &syntax[i], data);
946 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
949 if (!matroska->current_id) {
951 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
954 matroska->current_id = id | 1 << 7 * res;
956 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
959 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
964 for (i = 0; syntax[i].id; i++)
965 switch (syntax[i].type) {
967 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
970 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
974 // the default may be NULL
975 if (syntax[i].def.s) {
976 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
977 *dst = av_strdup(syntax[i].def.s);
979 return AVERROR(ENOMEM);
984 while (!res && !ebml_level_end(matroska))
985 res = ebml_parse(matroska, syntax, data);
991 * Allocate and return the entry for the level1 element with the given ID. If
992 * an entry already exists, return the existing entry.
994 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
998 MatroskaLevel1Element *elem;
1000 // Some files link to all clusters; useless.
1001 if (id == MATROSKA_ID_CLUSTER)
1004 // There can be multiple seekheads.
1005 if (id != MATROSKA_ID_SEEKHEAD) {
1006 for (i = 0; i < matroska->num_level1_elems; i++) {
1007 if (matroska->level1_elems[i].id == id)
1008 return &matroska->level1_elems[i];
1012 // Only a completely broken file would have more elements.
1013 // It also provides a low-effort way to escape from circular seekheads
1014 // (every iteration will add a level1 entry).
1015 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1016 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1020 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1021 *elem = (MatroskaLevel1Element){.id = id};
1026 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
1027 EbmlSyntax *syntax, void *data)
1029 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1032 // max. 16 MB for strings
1033 [EBML_STR] = 0x1000000,
1034 [EBML_UTF8] = 0x1000000,
1035 // max. 256 MB for binary data
1036 [EBML_BIN] = 0x10000000,
1037 // no limits for anything else
1039 AVIOContext *pb = matroska->ctx->pb;
1040 uint32_t id = syntax->id;
1044 MatroskaLevel1Element *level1_elem;
1046 data = (char *) data + syntax->data_offset;
1047 if (syntax->list_elem_size) {
1048 EbmlList *list = data;
1049 newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1051 return AVERROR(ENOMEM);
1052 list->elem = newelem;
1053 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1054 memset(data, 0, syntax->list_elem_size);
1058 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1059 matroska->current_id = 0;
1060 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1062 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1063 av_log(matroska->ctx, AV_LOG_ERROR,
1064 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1065 length, max_lengths[syntax->type], syntax->type);
1066 return AVERROR_INVALIDDATA;
1070 switch (syntax->type) {
1072 res = ebml_read_uint(pb, length, data);
1075 res = ebml_read_sint(pb, length, data);
1078 res = ebml_read_float(pb, length, data);
1082 res = ebml_read_ascii(pb, length, data);
1085 res = ebml_read_binary(pb, length, data);
1089 if ((res = ebml_read_master(matroska, length)) < 0)
1091 if (id == MATROSKA_ID_SEGMENT)
1092 matroska->segment_start = avio_tell(matroska->ctx->pb);
1093 if (id == MATROSKA_ID_CUES)
1094 matroska->cues_parsing_deferred = 0;
1095 if (syntax->type == EBML_LEVEL1 &&
1096 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1097 if (level1_elem->parsed)
1098 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1099 level1_elem->parsed = 1;
1101 return ebml_parse_nest(matroska, syntax->def.n, data);
1103 return ebml_parse_id(matroska, syntax->def.n, id, data);
1107 if (ffio_limit(pb, length) != length)
1108 return AVERROR(EIO);
1109 return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
1111 if (res == AVERROR_INVALIDDATA)
1112 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1113 else if (res == AVERROR(EIO))
1114 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1118 static void ebml_free(EbmlSyntax *syntax, void *data)
1121 for (i = 0; syntax[i].id; i++) {
1122 void *data_off = (char *) data + syntax[i].data_offset;
1123 switch (syntax[i].type) {
1129 av_freep(&((EbmlBin *) data_off)->data);
1133 if (syntax[i].list_elem_size) {
1134 EbmlList *list = data_off;
1135 char *ptr = list->elem;
1136 for (j = 0; j < list->nb_elem;
1137 j++, ptr += syntax[i].list_elem_size)
1138 ebml_free(syntax[i].def.n, ptr);
1139 av_freep(&list->elem);
1141 ebml_free(syntax[i].def.n, data_off);
1151 static int matroska_probe(AVProbeData *p)
1154 int len_mask = 0x80, size = 1, n = 1, i;
1157 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1160 /* length of header */
1162 while (size <= 8 && !(total & len_mask)) {
1168 total &= (len_mask - 1);
1170 total = (total << 8) | p->buf[4 + n++];
1172 /* Does the probe data contain the whole header? */
1173 if (p->buf_size < 4 + size + total)
1176 /* The header should contain a known document type. For now,
1177 * we don't parse the whole header but simply check for the
1178 * availability of that array of characters inside the header.
1179 * Not fully fool-proof, but good enough. */
1180 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1181 int probelen = strlen(matroska_doctypes[i]);
1182 if (total < probelen)
1184 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1185 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1186 return AVPROBE_SCORE_MAX;
1189 // probably valid EBML header but no recognized doctype
1190 return AVPROBE_SCORE_EXTENSION;
1193 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1196 MatroskaTrack *tracks = matroska->tracks.elem;
1199 for (i = 0; i < matroska->tracks.nb_elem; i++)
1200 if (tracks[i].num == num)
1203 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1207 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1208 MatroskaTrack *track)
1210 MatroskaTrackEncoding *encodings = track->encodings.elem;
1211 uint8_t *data = *buf;
1212 int isize = *buf_size;
1213 uint8_t *pkt_data = NULL;
1214 uint8_t av_unused *newpktdata;
1215 int pkt_size = isize;
1219 if (pkt_size >= 10000000U)
1220 return AVERROR_INVALIDDATA;
1222 switch (encodings[0].compression.algo) {
1223 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1225 int header_size = encodings[0].compression.settings.size;
1226 uint8_t *header = encodings[0].compression.settings.data;
1228 if (header_size && !header) {
1229 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1236 pkt_size = isize + header_size;
1237 pkt_data = av_malloc(pkt_size);
1239 return AVERROR(ENOMEM);
1241 memcpy(pkt_data, header, header_size);
1242 memcpy(pkt_data + header_size, data, isize);
1246 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1248 olen = pkt_size *= 3;
1249 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1251 result = AVERROR(ENOMEM);
1254 pkt_data = newpktdata;
1255 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1256 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1258 result = AVERROR_INVALIDDATA;
1265 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1267 z_stream zstream = { 0 };
1268 if (inflateInit(&zstream) != Z_OK)
1270 zstream.next_in = data;
1271 zstream.avail_in = isize;
1274 newpktdata = av_realloc(pkt_data, pkt_size);
1276 inflateEnd(&zstream);
1279 pkt_data = newpktdata;
1280 zstream.avail_out = pkt_size - zstream.total_out;
1281 zstream.next_out = pkt_data + zstream.total_out;
1283 result = inflate(&zstream, Z_NO_FLUSH);
1285 result = Z_MEM_ERROR;
1286 } while (result == Z_OK && pkt_size < 10000000);
1287 pkt_size = zstream.total_out;
1288 inflateEnd(&zstream);
1289 if (result != Z_STREAM_END) {
1290 if (result == Z_MEM_ERROR)
1291 result = AVERROR(ENOMEM);
1293 result = AVERROR_INVALIDDATA;
1300 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1302 bz_stream bzstream = { 0 };
1303 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1305 bzstream.next_in = data;
1306 bzstream.avail_in = isize;
1309 newpktdata = av_realloc(pkt_data, pkt_size);
1311 BZ2_bzDecompressEnd(&bzstream);
1314 pkt_data = newpktdata;
1315 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1316 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1318 result = BZ2_bzDecompress(&bzstream);
1320 result = BZ_MEM_ERROR;
1321 } while (result == BZ_OK && pkt_size < 10000000);
1322 pkt_size = bzstream.total_out_lo32;
1323 BZ2_bzDecompressEnd(&bzstream);
1324 if (result != BZ_STREAM_END) {
1325 if (result == BZ_MEM_ERROR)
1326 result = AVERROR(ENOMEM);
1328 result = AVERROR_INVALIDDATA;
1335 return AVERROR_INVALIDDATA;
1339 *buf_size = pkt_size;
1347 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1348 AVDictionary **metadata, char *prefix)
1350 MatroskaTag *tags = list->elem;
1354 for (i = 0; i < list->nb_elem; i++) {
1355 const char *lang = tags[i].lang &&
1356 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1358 if (!tags[i].name) {
1359 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1363 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1365 av_strlcpy(key, tags[i].name, sizeof(key));
1366 if (tags[i].def || !lang) {
1367 av_dict_set(metadata, key, tags[i].string, 0);
1368 if (tags[i].sub.nb_elem)
1369 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1372 av_strlcat(key, "-", sizeof(key));
1373 av_strlcat(key, lang, sizeof(key));
1374 av_dict_set(metadata, key, tags[i].string, 0);
1375 if (tags[i].sub.nb_elem)
1376 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1379 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1382 static void matroska_convert_tags(AVFormatContext *s)
1384 MatroskaDemuxContext *matroska = s->priv_data;
1385 MatroskaTags *tags = matroska->tags.elem;
1388 for (i = 0; i < matroska->tags.nb_elem; i++) {
1389 if (tags[i].target.attachuid) {
1390 MatroskaAttachment *attachment = matroska->attachments.elem;
1391 for (j = 0; j < matroska->attachments.nb_elem; j++)
1392 if (attachment[j].uid == tags[i].target.attachuid &&
1393 attachment[j].stream)
1394 matroska_convert_tag(s, &tags[i].tag,
1395 &attachment[j].stream->metadata, NULL);
1396 } else if (tags[i].target.chapteruid) {
1397 MatroskaChapter *chapter = matroska->chapters.elem;
1398 for (j = 0; j < matroska->chapters.nb_elem; j++)
1399 if (chapter[j].uid == tags[i].target.chapteruid &&
1401 matroska_convert_tag(s, &tags[i].tag,
1402 &chapter[j].chapter->metadata, NULL);
1403 } else if (tags[i].target.trackuid) {
1404 MatroskaTrack *track = matroska->tracks.elem;
1405 for (j = 0; j < matroska->tracks.nb_elem; j++)
1406 if (track[j].uid == tags[i].target.trackuid && track[j].stream)
1407 matroska_convert_tag(s, &tags[i].tag,
1408 &track[j].stream->metadata, NULL);
1410 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1411 tags[i].target.type);
1416 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1419 uint32_t level_up = matroska->level_up;
1420 uint32_t saved_id = matroska->current_id;
1421 int64_t before_pos = avio_tell(matroska->ctx->pb);
1422 MatroskaLevel level;
1427 offset = pos + matroska->segment_start;
1428 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1429 /* We don't want to lose our seekhead level, so we add
1430 * a dummy. This is a crude hack. */
1431 if (matroska->num_levels == EBML_MAX_DEPTH) {
1432 av_log(matroska->ctx, AV_LOG_INFO,
1433 "Max EBML element depth (%d) reached, "
1434 "cannot parse further.\n", EBML_MAX_DEPTH);
1435 ret = AVERROR_INVALIDDATA;
1438 level.length = (uint64_t) -1;
1439 matroska->levels[matroska->num_levels] = level;
1440 matroska->num_levels++;
1441 matroska->current_id = 0;
1443 ret = ebml_parse(matroska, matroska_segment, matroska);
1445 /* remove dummy level */
1446 while (matroska->num_levels) {
1447 uint64_t length = matroska->levels[--matroska->num_levels].length;
1448 if (length == (uint64_t) -1)
1454 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1455 matroska->level_up = level_up;
1456 matroska->current_id = saved_id;
1461 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1463 EbmlList *seekhead_list = &matroska->seekhead;
1466 // we should not do any seeking in the streaming case
1467 if (!matroska->ctx->pb->seekable)
1470 for (i = 0; i < seekhead_list->nb_elem; i++) {
1471 MatroskaSeekhead *seekheads = seekhead_list->elem;
1472 uint32_t id = seekheads[i].id;
1473 uint64_t pos = seekheads[i].pos;
1475 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1476 if (!elem || elem->parsed)
1481 // defer cues parsing until we actually need cue data.
1482 if (id == MATROSKA_ID_CUES)
1485 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1486 // mark index as broken
1487 matroska->cues_parsing_deferred = -1;
1495 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1497 EbmlList *index_list;
1498 MatroskaIndex *index;
1499 int index_scale = 1;
1502 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1505 index_list = &matroska->index;
1506 index = index_list->elem;
1507 if (index_list->nb_elem &&
1508 index[0].time > 1E14 / matroska->time_scale) {
1509 av_log(matroska->ctx, AV_LOG_WARNING, "Working around broken index.\n");
1510 index_scale = matroska->time_scale;
1512 for (i = 0; i < index_list->nb_elem; i++) {
1513 EbmlList *pos_list = &index[i].pos;
1514 MatroskaIndexPos *pos = pos_list->elem;
1515 for (j = 0; j < pos_list->nb_elem; j++) {
1516 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1518 if (track && track->stream)
1519 av_add_index_entry(track->stream,
1520 pos[j].pos + matroska->segment_start,
1521 index[i].time / index_scale, 0, 0,
1527 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1530 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1533 for (i = 0; i < matroska->num_level1_elems; i++) {
1534 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1535 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1536 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1537 matroska->cues_parsing_deferred = -1;
1543 matroska_add_index_entries(matroska);
1546 static int matroska_aac_profile(char *codec_id)
1548 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1551 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1552 if (strstr(codec_id, aac_profiles[profile]))
1557 static int matroska_aac_sri(int samplerate)
1561 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1562 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1567 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1570 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1571 time_t creation_time = date_utc / 1000000000 + 978307200;
1572 struct tm tmpbuf, *ptm = gmtime_r(&creation_time, &tmpbuf);
1574 if (strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm))
1575 av_dict_set(metadata, "creation_time", buffer, 0);
1578 static int matroska_parse_flac(AVFormatContext *s,
1579 MatroskaTrack *track,
1582 AVStream *st = track->stream;
1583 uint8_t *p = track->codec_priv.data;
1584 int size = track->codec_priv.size;
1586 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1587 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1588 track->codec_priv.size = 0;
1592 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1594 p += track->codec_priv.size;
1595 size -= track->codec_priv.size;
1597 /* parse the remaining metadata blocks if present */
1599 int block_last, block_type, block_size;
1601 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1605 if (block_size > size)
1608 /* check for the channel mask */
1609 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1610 AVDictionary *dict = NULL;
1611 AVDictionaryEntry *chmask;
1613 ff_vorbis_comment(s, &dict, p, block_size, 0);
1614 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1616 uint64_t mask = strtol(chmask->value, NULL, 0);
1617 if (!mask || mask & ~0x3ffffULL) {
1618 av_log(s, AV_LOG_WARNING,
1619 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1621 st->codec->channel_layout = mask;
1623 av_dict_free(&dict);
1633 static int matroska_parse_tracks(AVFormatContext *s)
1635 MatroskaDemuxContext *matroska = s->priv_data;
1636 MatroskaTrack *tracks = matroska->tracks.elem;
1641 for (i = 0; i < matroska->tracks.nb_elem; i++) {
1642 MatroskaTrack *track = &tracks[i];
1643 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1644 EbmlList *encodings_list = &track->encodings;
1645 MatroskaTrackEncoding *encodings = encodings_list->elem;
1646 uint8_t *extradata = NULL;
1647 int extradata_size = 0;
1648 int extradata_offset = 0;
1649 uint32_t fourcc = 0;
1651 char* key_id_base64 = NULL;
1654 /* Apply some sanity checks. */
1655 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1656 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1657 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1658 track->type != MATROSKA_TRACK_TYPE_METADATA) {
1659 av_log(matroska->ctx, AV_LOG_INFO,
1660 "Unknown or unsupported track type %"PRIu64"\n",
1664 if (!track->codec_id)
1667 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1668 if (!track->default_duration && track->video.frame_rate > 0)
1669 track->default_duration = 1000000000 / track->video.frame_rate;
1670 if (track->video.display_width == -1)
1671 track->video.display_width = track->video.pixel_width;
1672 if (track->video.display_height == -1)
1673 track->video.display_height = track->video.pixel_height;
1674 if (track->video.color_space.size == 4)
1675 fourcc = AV_RL32(track->video.color_space.data);
1676 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1677 if (!track->audio.out_samplerate)
1678 track->audio.out_samplerate = track->audio.samplerate;
1680 if (encodings_list->nb_elem > 1) {
1681 av_log(matroska->ctx, AV_LOG_ERROR,
1682 "Multiple combined encodings not supported");
1683 } else if (encodings_list->nb_elem == 1) {
1684 if (encodings[0].type) {
1685 if (encodings[0].encryption.key_id.size > 0) {
1686 /* Save the encryption key id to be stored later as a
1688 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1689 key_id_base64 = av_malloc(b64_size);
1690 if (key_id_base64 == NULL)
1691 return AVERROR(ENOMEM);
1693 av_base64_encode(key_id_base64, b64_size,
1694 encodings[0].encryption.key_id.data,
1695 encodings[0].encryption.key_id.size);
1697 encodings[0].scope = 0;
1698 av_log(matroska->ctx, AV_LOG_ERROR,
1699 "Unsupported encoding type");
1703 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1706 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1709 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1711 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
1712 encodings[0].scope = 0;
1713 av_log(matroska->ctx, AV_LOG_ERROR,
1714 "Unsupported encoding type");
1715 } else if (track->codec_priv.size && encodings[0].scope & 2) {
1716 uint8_t *codec_priv = track->codec_priv.data;
1717 int ret = matroska_decode_buffer(&track->codec_priv.data,
1718 &track->codec_priv.size,
1721 track->codec_priv.data = NULL;
1722 track->codec_priv.size = 0;
1723 av_log(matroska->ctx, AV_LOG_ERROR,
1724 "Failed to decode codec private data\n");
1727 if (codec_priv != track->codec_priv.data)
1728 av_free(codec_priv);
1732 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
1733 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1734 strlen(ff_mkv_codec_tags[j].str))) {
1735 codec_id = ff_mkv_codec_tags[j].id;
1740 st = track->stream = avformat_new_stream(s, NULL);
1742 av_free(key_id_base64);
1743 return AVERROR(ENOMEM);
1746 if (key_id_base64) {
1747 /* export encryption key id as base64 metadata tag */
1748 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
1749 av_freep(&key_id_base64);
1752 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
1753 track->codec_priv.size >= 40 &&
1754 track->codec_priv.data) {
1755 track->ms_compat = 1;
1756 bit_depth = AV_RL16(track->codec_priv.data + 14);
1757 fourcc = AV_RL32(track->codec_priv.data + 16);
1758 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
1761 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
1763 extradata_offset = 40;
1764 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
1765 track->codec_priv.size >= 14 &&
1766 track->codec_priv.data) {
1768 ffio_init_context(&b, track->codec_priv.data,
1769 track->codec_priv.size,
1770 0, NULL, NULL, NULL, NULL);
1771 ret = ff_get_wav_header(&b, st->codec, track->codec_priv.size, 0);
1774 codec_id = st->codec->codec_id;
1775 extradata_offset = FFMIN(track->codec_priv.size, 18);
1776 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
1777 && (track->codec_priv.size >= 86)
1778 && (track->codec_priv.data)) {
1779 fourcc = AV_RL32(track->codec_priv.data + 4);
1780 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1781 if (ff_codec_get_id(ff_codec_movaudio_tags, AV_RL32(track->codec_priv.data))) {
1782 fourcc = AV_RL32(track->codec_priv.data);
1783 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1785 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
1786 (track->codec_priv.size >= 21) &&
1787 (track->codec_priv.data)) {
1788 fourcc = AV_RL32(track->codec_priv.data + 4);
1789 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1790 if (ff_codec_get_id(ff_codec_movvideo_tags, AV_RL32(track->codec_priv.data))) {
1791 fourcc = AV_RL32(track->codec_priv.data);
1792 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1794 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI "))
1795 codec_id = AV_CODEC_ID_SVQ3;
1796 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1797 switch (track->audio.bitdepth) {
1799 codec_id = AV_CODEC_ID_PCM_U8;
1802 codec_id = AV_CODEC_ID_PCM_S24BE;
1805 codec_id = AV_CODEC_ID_PCM_S32BE;
1808 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1809 switch (track->audio.bitdepth) {
1811 codec_id = AV_CODEC_ID_PCM_U8;
1814 codec_id = AV_CODEC_ID_PCM_S24LE;
1817 codec_id = AV_CODEC_ID_PCM_S32LE;
1820 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
1821 track->audio.bitdepth == 64) {
1822 codec_id = AV_CODEC_ID_PCM_F64LE;
1823 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1824 int profile = matroska_aac_profile(track->codec_id);
1825 int sri = matroska_aac_sri(track->audio.samplerate);
1826 extradata = av_mallocz(5 + FF_INPUT_BUFFER_PADDING_SIZE);
1828 return AVERROR(ENOMEM);
1829 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
1830 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
1831 if (strstr(track->codec_id, "SBR")) {
1832 sri = matroska_aac_sri(track->audio.out_samplerate);
1833 extradata[2] = 0x56;
1834 extradata[3] = 0xE5;
1835 extradata[4] = 0x80 | (sri << 3);
1839 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - FF_INPUT_BUFFER_PADDING_SIZE) {
1840 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1841 * Create the "atom size", "tag", and "tag version" fields the
1842 * decoder expects manually. */
1843 extradata_size = 12 + track->codec_priv.size;
1844 extradata = av_mallocz(extradata_size +
1845 FF_INPUT_BUFFER_PADDING_SIZE);
1847 return AVERROR(ENOMEM);
1848 AV_WB32(extradata, extradata_size);
1849 memcpy(&extradata[4], "alac", 4);
1850 AV_WB32(&extradata[8], 0);
1851 memcpy(&extradata[12], track->codec_priv.data,
1852 track->codec_priv.size);
1853 } else if (codec_id == AV_CODEC_ID_TTA) {
1854 extradata_size = 30;
1855 extradata = av_mallocz(extradata_size + FF_INPUT_BUFFER_PADDING_SIZE);
1857 return AVERROR(ENOMEM);
1858 ffio_init_context(&b, extradata, extradata_size, 1,
1859 NULL, NULL, NULL, NULL);
1860 avio_write(&b, "TTA1", 4);
1862 avio_wl16(&b, track->audio.channels);
1863 avio_wl16(&b, track->audio.bitdepth);
1864 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
1865 return AVERROR_INVALIDDATA;
1866 avio_wl32(&b, track->audio.out_samplerate);
1867 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
1868 track->audio.out_samplerate,
1869 AV_TIME_BASE * 1000));
1870 } else if (codec_id == AV_CODEC_ID_RV10 ||
1871 codec_id == AV_CODEC_ID_RV20 ||
1872 codec_id == AV_CODEC_ID_RV30 ||
1873 codec_id == AV_CODEC_ID_RV40) {
1874 extradata_offset = 26;
1875 } else if (codec_id == AV_CODEC_ID_RA_144) {
1876 track->audio.out_samplerate = 8000;
1877 track->audio.channels = 1;
1878 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
1879 codec_id == AV_CODEC_ID_COOK ||
1880 codec_id == AV_CODEC_ID_ATRAC3 ||
1881 codec_id == AV_CODEC_ID_SIPR)
1882 && track->codec_priv.data) {
1885 ffio_init_context(&b, track->codec_priv.data,
1886 track->codec_priv.size,
1887 0, NULL, NULL, NULL, NULL);
1889 flavor = avio_rb16(&b);
1890 track->audio.coded_framesize = avio_rb32(&b);
1892 track->audio.sub_packet_h = avio_rb16(&b);
1893 track->audio.frame_size = avio_rb16(&b);
1894 track->audio.sub_packet_size = avio_rb16(&b);
1896 track->audio.coded_framesize <= 0 ||
1897 track->audio.sub_packet_h <= 0 ||
1898 track->audio.frame_size <= 0 ||
1899 track->audio.sub_packet_size <= 0)
1900 return AVERROR_INVALIDDATA;
1901 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
1902 track->audio.frame_size);
1903 if (!track->audio.buf)
1904 return AVERROR(ENOMEM);
1905 if (codec_id == AV_CODEC_ID_RA_288) {
1906 st->codec->block_align = track->audio.coded_framesize;
1907 track->codec_priv.size = 0;
1909 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
1910 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
1911 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
1912 st->codec->bit_rate = sipr_bit_rate[flavor];
1914 st->codec->block_align = track->audio.sub_packet_size;
1915 extradata_offset = 78;
1917 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
1918 ret = matroska_parse_flac(s, track, &extradata_offset);
1921 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
1922 fourcc = AV_RL32(track->codec_priv.data);
1924 track->codec_priv.size -= extradata_offset;
1926 if (codec_id == AV_CODEC_ID_NONE)
1927 av_log(matroska->ctx, AV_LOG_INFO,
1928 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
1930 if (track->time_scale < 0.01)
1931 track->time_scale = 1.0;
1932 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
1933 1000 * 1000 * 1000); /* 64 bit pts in ns */
1935 /* convert the delay from ns to the track timebase */
1936 track->codec_delay = av_rescale_q(track->codec_delay,
1937 (AVRational){ 1, 1000000000 },
1940 st->codec->codec_id = codec_id;
1942 if (strcmp(track->language, "und"))
1943 av_dict_set(&st->metadata, "language", track->language, 0);
1944 av_dict_set(&st->metadata, "title", track->name, 0);
1946 if (track->flag_default)
1947 st->disposition |= AV_DISPOSITION_DEFAULT;
1948 if (track->flag_forced)
1949 st->disposition |= AV_DISPOSITION_FORCED;
1951 if (!st->codec->extradata) {
1953 st->codec->extradata = extradata;
1954 st->codec->extradata_size = extradata_size;
1955 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
1956 if (ff_alloc_extradata(st->codec, track->codec_priv.size))
1957 return AVERROR(ENOMEM);
1958 memcpy(st->codec->extradata,
1959 track->codec_priv.data + extradata_offset,
1960 track->codec_priv.size);
1964 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1965 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
1967 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1968 st->codec->codec_tag = fourcc;
1970 st->codec->bits_per_coded_sample = bit_depth;
1971 st->codec->width = track->video.pixel_width;
1972 st->codec->height = track->video.pixel_height;
1973 av_reduce(&st->sample_aspect_ratio.num,
1974 &st->sample_aspect_ratio.den,
1975 st->codec->height * track->video.display_width,
1976 st->codec->width * track->video.display_height,
1978 if (st->codec->codec_id != AV_CODEC_ID_HEVC)
1979 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1981 if (track->default_duration) {
1982 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1983 1000000000, track->default_duration, 30000);
1984 #if FF_API_R_FRAME_RATE
1985 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
1986 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
1987 st->r_frame_rate = st->avg_frame_rate;
1991 /* export stereo mode flag as metadata tag */
1992 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
1993 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
1995 /* export alpha mode flag as metadata tag */
1996 if (track->video.alpha_mode)
1997 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
1999 /* if we have virtual track, mark the real tracks */
2000 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2002 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2004 snprintf(buf, sizeof(buf), "%s_%d",
2005 ff_matroska_video_stereo_plane[planes[j].type], i);
2006 for (k=0; k < matroska->tracks.nb_elem; k++)
2007 if (planes[j].uid == tracks[k].uid) {
2008 av_dict_set(&s->streams[k]->metadata,
2009 "stereo_mode", buf, 0);
2013 // add stream level stereo3d side data if it is a supported format
2014 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2015 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2016 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2020 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2021 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
2022 st->codec->sample_rate = track->audio.out_samplerate;
2023 st->codec->channels = track->audio.channels;
2024 if (!st->codec->bits_per_coded_sample)
2025 st->codec->bits_per_coded_sample = track->audio.bitdepth;
2026 if (st->codec->codec_id != AV_CODEC_ID_AAC)
2027 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2028 if (track->codec_delay > 0) {
2029 st->codec->delay = av_rescale_q(track->codec_delay,
2031 (AVRational){1, st->codec->sample_rate});
2033 if (track->seek_preroll > 0) {
2034 av_codec_set_seek_preroll(st->codec,
2035 av_rescale_q(track->seek_preroll,
2036 (AVRational){1, 1000000000},
2037 (AVRational){1, st->codec->sample_rate}));
2039 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2040 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
2042 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2043 st->disposition |= AV_DISPOSITION_CAPTIONS;
2044 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2045 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2046 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2047 st->disposition |= AV_DISPOSITION_METADATA;
2049 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2050 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
2051 if (st->codec->codec_id == AV_CODEC_ID_ASS)
2052 matroska->contains_ssa = 1;
2059 static int matroska_read_header(AVFormatContext *s)
2061 MatroskaDemuxContext *matroska = s->priv_data;
2062 EbmlList *attachments_list = &matroska->attachments;
2063 EbmlList *chapters_list = &matroska->chapters;
2064 MatroskaAttachment *attachments;
2065 MatroskaChapter *chapters;
2066 uint64_t max_start = 0;
2072 matroska->cues_parsing_deferred = 1;
2074 /* First read the EBML header. */
2075 if (ebml_parse(matroska, ebml_syntax, &ebml) ||
2076 ebml.version > EBML_VERSION ||
2077 ebml.max_size > sizeof(uint64_t) ||
2078 ebml.id_length > sizeof(uint32_t) ||
2079 ebml.doctype_version > 3 ||
2081 av_log(matroska->ctx, AV_LOG_ERROR,
2082 "EBML header using unsupported features\n"
2083 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2084 ebml.version, ebml.doctype, ebml.doctype_version);
2085 ebml_free(ebml_syntax, &ebml);
2086 return AVERROR_PATCHWELCOME;
2087 } else if (ebml.doctype_version == 3) {
2088 av_log(matroska->ctx, AV_LOG_WARNING,
2089 "EBML header using unsupported features\n"
2090 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2091 ebml.version, ebml.doctype, ebml.doctype_version);
2093 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2094 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2096 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2097 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2098 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2099 ebml_free(ebml_syntax, &ebml);
2100 return AVERROR_INVALIDDATA;
2103 ebml_free(ebml_syntax, &ebml);
2105 /* The next thing is a segment. */
2106 pos = avio_tell(matroska->ctx->pb);
2107 res = ebml_parse(matroska, matroska_segments, matroska);
2108 // try resyncing until we find a EBML_STOP type element.
2110 res = matroska_resync(matroska, pos);
2113 pos = avio_tell(matroska->ctx->pb);
2114 res = ebml_parse(matroska, matroska_segment, matroska);
2116 matroska_execute_seekhead(matroska);
2118 if (!matroska->time_scale)
2119 matroska->time_scale = 1000000;
2120 if (matroska->duration)
2121 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2122 1000 / AV_TIME_BASE;
2123 av_dict_set(&s->metadata, "title", matroska->title, 0);
2124 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2126 if (matroska->date_utc.size == 8)
2127 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2129 res = matroska_parse_tracks(s);
2133 attachments = attachments_list->elem;
2134 for (j = 0; j < attachments_list->nb_elem; j++) {
2135 if (!(attachments[j].filename && attachments[j].mime &&
2136 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2137 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2139 AVStream *st = avformat_new_stream(s, NULL);
2142 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2143 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2144 st->codec->codec_id = AV_CODEC_ID_NONE;
2145 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2146 if (ff_alloc_extradata(st->codec, attachments[j].bin.size))
2148 memcpy(st->codec->extradata, attachments[j].bin.data,
2149 attachments[j].bin.size);
2151 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2152 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2153 strlen(ff_mkv_mime_tags[i].str))) {
2154 st->codec->codec_id = ff_mkv_mime_tags[i].id;
2158 attachments[j].stream = st;
2162 chapters = chapters_list->elem;
2163 for (i = 0; i < chapters_list->nb_elem; i++)
2164 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2165 (max_start == 0 || chapters[i].start > max_start)) {
2166 chapters[i].chapter =
2167 avpriv_new_chapter(s, chapters[i].uid,
2168 (AVRational) { 1, 1000000000 },
2169 chapters[i].start, chapters[i].end,
2171 if (chapters[i].chapter) {
2172 av_dict_set(&chapters[i].chapter->metadata,
2173 "title", chapters[i].title, 0);
2175 max_start = chapters[i].start;
2178 matroska_add_index_entries(matroska);
2180 matroska_convert_tags(s);
2186 * Put one packet in an application-supplied AVPacket struct.
2187 * Returns 0 on success or -1 on failure.
2189 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2192 if (matroska->num_packets > 0) {
2193 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2194 av_freep(&matroska->packets[0]);
2195 if (matroska->num_packets > 1) {
2197 memmove(&matroska->packets[0], &matroska->packets[1],
2198 (matroska->num_packets - 1) * sizeof(AVPacket *));
2199 newpackets = av_realloc(matroska->packets,
2200 (matroska->num_packets - 1) *
2201 sizeof(AVPacket *));
2203 matroska->packets = newpackets;
2205 av_freep(&matroska->packets);
2206 matroska->prev_pkt = NULL;
2208 matroska->num_packets--;
2216 * Free all packets in our internal queue.
2218 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2220 matroska->prev_pkt = NULL;
2221 if (matroska->packets) {
2223 for (n = 0; n < matroska->num_packets; n++) {
2224 av_free_packet(matroska->packets[n]);
2225 av_freep(&matroska->packets[n]);
2227 av_freep(&matroska->packets);
2228 matroska->num_packets = 0;
2232 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2233 int *buf_size, int type,
2234 uint32_t **lace_buf, int *laces)
2236 int res = 0, n, size = *buf_size;
2237 uint8_t *data = *buf;
2238 uint32_t *lace_size;
2242 *lace_buf = av_mallocz(sizeof(int));
2244 return AVERROR(ENOMEM);
2246 *lace_buf[0] = size;
2250 av_assert0(size > 0);
2254 lace_size = av_mallocz(*laces * sizeof(int));
2256 return AVERROR(ENOMEM);
2259 case 0x1: /* Xiph lacing */
2263 for (n = 0; res == 0 && n < *laces - 1; n++) {
2265 if (size <= total) {
2266 res = AVERROR_INVALIDDATA;
2271 lace_size[n] += temp;
2278 if (size <= total) {
2279 res = AVERROR_INVALIDDATA;
2283 lace_size[n] = size - total;
2287 case 0x2: /* fixed-size lacing */
2288 if (size % (*laces)) {
2289 res = AVERROR_INVALIDDATA;
2292 for (n = 0; n < *laces; n++)
2293 lace_size[n] = size / *laces;
2296 case 0x3: /* EBML lacing */
2300 n = matroska_ebmlnum_uint(matroska, data, size, &num);
2301 if (n < 0 || num > INT_MAX) {
2302 av_log(matroska->ctx, AV_LOG_INFO,
2303 "EBML block data error\n");
2304 res = n<0 ? n : AVERROR_INVALIDDATA;
2309 total = lace_size[0] = num;
2310 for (n = 1; res == 0 && n < *laces - 1; n++) {
2313 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2314 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2315 av_log(matroska->ctx, AV_LOG_INFO,
2316 "EBML block data error\n");
2317 res = r<0 ? r : AVERROR_INVALIDDATA;
2322 lace_size[n] = lace_size[n - 1] + snum;
2323 total += lace_size[n];
2325 if (size <= total) {
2326 res = AVERROR_INVALIDDATA;
2329 lace_size[*laces - 1] = size - total;
2335 *lace_buf = lace_size;
2341 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2342 MatroskaTrack *track, AVStream *st,
2343 uint8_t *data, int size, uint64_t timecode,
2346 int a = st->codec->block_align;
2347 int sps = track->audio.sub_packet_size;
2348 int cfs = track->audio.coded_framesize;
2349 int h = track->audio.sub_packet_h;
2350 int y = track->audio.sub_packet_cnt;
2351 int w = track->audio.frame_size;
2354 if (!track->audio.pkt_cnt) {
2355 if (track->audio.sub_packet_cnt == 0)
2356 track->audio.buf_timecode = timecode;
2357 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2358 if (size < cfs * h / 2) {
2359 av_log(matroska->ctx, AV_LOG_ERROR,
2360 "Corrupt int4 RM-style audio packet size\n");
2361 return AVERROR_INVALIDDATA;
2363 for (x = 0; x < h / 2; x++)
2364 memcpy(track->audio.buf + x * 2 * w + y * cfs,
2365 data + x * cfs, cfs);
2366 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2368 av_log(matroska->ctx, AV_LOG_ERROR,
2369 "Corrupt sipr RM-style audio packet size\n");
2370 return AVERROR_INVALIDDATA;
2372 memcpy(track->audio.buf + y * w, data, w);
2374 if (size < sps * w / sps || h<=0 || w%sps) {
2375 av_log(matroska->ctx, AV_LOG_ERROR,
2376 "Corrupt generic RM-style audio packet size\n");
2377 return AVERROR_INVALIDDATA;
2379 for (x = 0; x < w / sps; x++)
2380 memcpy(track->audio.buf +
2381 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2382 data + x * sps, sps);
2385 if (++track->audio.sub_packet_cnt >= h) {
2386 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2387 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2388 track->audio.sub_packet_cnt = 0;
2389 track->audio.pkt_cnt = h * w / a;
2393 while (track->audio.pkt_cnt) {
2395 AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2397 return AVERROR(ENOMEM);
2399 ret = av_new_packet(pkt, a);
2405 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2407 pkt->pts = track->audio.buf_timecode;
2408 track->audio.buf_timecode = AV_NOPTS_VALUE;
2410 pkt->stream_index = st->index;
2411 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2417 /* reconstruct full wavpack blocks from mangled matroska ones */
2418 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
2419 uint8_t **pdst, int *size)
2421 uint8_t *dst = NULL;
2426 int ret, offset = 0;
2428 if (srclen < 12 || track->stream->codec->extradata_size < 2)
2429 return AVERROR_INVALIDDATA;
2431 ver = AV_RL16(track->stream->codec->extradata);
2433 samples = AV_RL32(src);
2437 while (srclen >= 8) {
2442 uint32_t flags = AV_RL32(src);
2443 uint32_t crc = AV_RL32(src + 4);
2447 multiblock = (flags & 0x1800) != 0x1800;
2450 ret = AVERROR_INVALIDDATA;
2453 blocksize = AV_RL32(src);
2459 if (blocksize > srclen) {
2460 ret = AVERROR_INVALIDDATA;
2464 tmp = av_realloc(dst, dstlen + blocksize + 32);
2466 ret = AVERROR(ENOMEM);
2470 dstlen += blocksize + 32;
2472 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2473 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2474 AV_WL16(dst + offset + 8, ver); // version
2475 AV_WL16(dst + offset + 10, 0); // track/index_no
2476 AV_WL32(dst + offset + 12, 0); // total samples
2477 AV_WL32(dst + offset + 16, 0); // block index
2478 AV_WL32(dst + offset + 20, samples); // number of samples
2479 AV_WL32(dst + offset + 24, flags); // flags
2480 AV_WL32(dst + offset + 28, crc); // crc
2481 memcpy(dst + offset + 32, src, blocksize); // block data
2484 srclen -= blocksize;
2485 offset += blocksize + 32;
2498 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
2499 MatroskaTrack *track,
2501 uint8_t *data, int data_len,
2507 uint8_t *id, *settings, *text, *buf;
2508 int id_len, settings_len, text_len;
2513 return AVERROR_INVALIDDATA;
2516 q = data + data_len;
2521 if (*p == '\r' || *p == '\n') {
2530 if (p >= q || *p != '\n')
2531 return AVERROR_INVALIDDATA;
2537 if (*p == '\r' || *p == '\n') {
2538 settings_len = p - settings;
2546 if (p >= q || *p != '\n')
2547 return AVERROR_INVALIDDATA;
2552 while (text_len > 0) {
2553 const int len = text_len - 1;
2554 const uint8_t c = p[len];
2555 if (c != '\r' && c != '\n')
2561 return AVERROR_INVALIDDATA;
2563 pkt = av_mallocz(sizeof(*pkt));
2565 return AVERROR(ENOMEM);
2566 err = av_new_packet(pkt, text_len);
2569 return AVERROR(err);
2572 memcpy(pkt->data, text, text_len);
2575 buf = av_packet_new_side_data(pkt,
2576 AV_PKT_DATA_WEBVTT_IDENTIFIER,
2580 return AVERROR(ENOMEM);
2582 memcpy(buf, id, id_len);
2585 if (settings_len > 0) {
2586 buf = av_packet_new_side_data(pkt,
2587 AV_PKT_DATA_WEBVTT_SETTINGS,
2591 return AVERROR(ENOMEM);
2593 memcpy(buf, settings, settings_len);
2596 // Do we need this for subtitles?
2597 // pkt->flags = AV_PKT_FLAG_KEY;
2599 pkt->stream_index = st->index;
2600 pkt->pts = timecode;
2602 // Do we need this for subtitles?
2603 // pkt->dts = timecode;
2605 pkt->duration = duration;
2608 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2609 matroska->prev_pkt = pkt;
2614 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2615 MatroskaTrack *track, AVStream *st,
2616 uint8_t *data, int pkt_size,
2617 uint64_t timecode, uint64_t lace_duration,
2618 int64_t pos, int is_keyframe,
2619 uint8_t *additional, uint64_t additional_id, int additional_size,
2620 int64_t discard_padding)
2622 MatroskaTrackEncoding *encodings = track->encodings.elem;
2623 uint8_t *pkt_data = data;
2624 int offset = 0, res;
2627 if (encodings && !encodings->type && encodings->scope & 1) {
2628 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2633 if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
2635 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2637 av_log(matroska->ctx, AV_LOG_ERROR,
2638 "Error parsing a wavpack block.\n");
2641 if (pkt_data != data)
2642 av_freep(&pkt_data);
2646 if (st->codec->codec_id == AV_CODEC_ID_PRORES &&
2647 AV_RB32(&data[4]) != MKBETAG('i', 'c', 'p', 'f'))
2650 pkt = av_mallocz(sizeof(AVPacket));
2652 return AVERROR(ENOMEM);
2653 /* XXX: prevent data copy... */
2654 if (av_new_packet(pkt, pkt_size + offset) < 0) {
2656 res = AVERROR(ENOMEM);
2660 if (st->codec->codec_id == AV_CODEC_ID_PRORES && offset == 8) {
2661 uint8_t *buf = pkt->data;
2662 bytestream_put_be32(&buf, pkt_size);
2663 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2666 memcpy(pkt->data + offset, pkt_data, pkt_size);
2668 if (pkt_data != data)
2669 av_freep(&pkt_data);
2671 pkt->flags = is_keyframe;
2672 pkt->stream_index = st->index;
2674 if (additional_size > 0) {
2675 uint8_t *side_data = av_packet_new_side_data(pkt,
2676 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
2677 additional_size + 8);
2679 av_free_packet(pkt);
2681 return AVERROR(ENOMEM);
2683 AV_WB64(side_data, additional_id);
2684 memcpy(side_data + 8, additional, additional_size);
2687 if (discard_padding) {
2688 uint8_t *side_data = av_packet_new_side_data(pkt,
2689 AV_PKT_DATA_SKIP_SAMPLES,
2692 av_free_packet(pkt);
2694 return AVERROR(ENOMEM);
2696 AV_WL32(side_data, 0);
2697 AV_WL32(side_data + 4, av_rescale_q(discard_padding,
2698 (AVRational){1, 1000000000},
2699 (AVRational){1, st->codec->sample_rate}));
2702 if (track->ms_compat)
2703 pkt->dts = timecode;
2705 pkt->pts = timecode;
2707 if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2709 * For backward compatibility.
2710 * Historically, we have put subtitle duration
2711 * in convergence_duration, on the off chance
2712 * that the time_scale is less than 1us, which
2713 * could result in a 32bit overflow on the
2714 * normal duration field.
2716 pkt->convergence_duration = lace_duration;
2719 if (track->type != MATROSKA_TRACK_TYPE_SUBTITLE ||
2720 lace_duration <= INT_MAX) {
2722 * For non subtitle tracks, just store the duration
2725 * If it's a subtitle track and duration value does
2726 * not overflow a uint32, then also store it normally.
2728 pkt->duration = lace_duration;
2731 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2732 matroska->prev_pkt = pkt;
2737 if (pkt_data != data)
2738 av_freep(&pkt_data);
2742 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2743 int size, int64_t pos, uint64_t cluster_time,
2744 uint64_t block_duration, int is_keyframe,
2745 uint8_t *additional, uint64_t additional_id, int additional_size,
2746 int64_t cluster_pos, int64_t discard_padding)
2748 uint64_t timecode = AV_NOPTS_VALUE;
2749 MatroskaTrack *track;
2753 uint32_t *lace_size = NULL;
2754 int n, flags, laces = 0;
2756 int trust_default_duration = 1;
2758 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2759 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2765 track = matroska_find_track_by_num(matroska, num);
2766 if (!track || !track->stream) {
2767 av_log(matroska->ctx, AV_LOG_INFO,
2768 "Invalid stream %"PRIu64" or size %u\n", num, size);
2769 return AVERROR_INVALIDDATA;
2770 } else if (size <= 3)
2773 if (st->discard >= AVDISCARD_ALL)
2775 av_assert1(block_duration != AV_NOPTS_VALUE);
2777 block_time = sign_extend(AV_RB16(data), 16);
2781 if (is_keyframe == -1)
2782 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2784 if (cluster_time != (uint64_t) -1 &&
2785 (block_time >= 0 || cluster_time >= -block_time)) {
2786 timecode = cluster_time + block_time - track->codec_delay;
2787 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2788 timecode < track->end_timecode)
2789 is_keyframe = 0; /* overlapping subtitles are not key frame */
2791 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
2795 if (matroska->skip_to_keyframe &&
2796 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2797 if (timecode < matroska->skip_to_timecode)
2800 matroska->skip_to_keyframe = 0;
2801 else if (!st->skip_to_keyframe) {
2802 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2803 matroska->skip_to_keyframe = 0;
2807 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2808 &lace_size, &laces);
2813 if (track->audio.samplerate == 8000) {
2814 // If this is needed for more codecs, then add them here
2815 if (st->codec->codec_id == AV_CODEC_ID_AC3) {
2816 if (track->audio.samplerate != st->codec->sample_rate || !st->codec->frame_size)
2817 trust_default_duration = 0;
2821 if (!block_duration && trust_default_duration)
2822 block_duration = track->default_duration * laces / matroska->time_scale;
2824 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2825 track->end_timecode =
2826 FFMAX(track->end_timecode, timecode + block_duration);
2828 for (n = 0; n < laces; n++) {
2829 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2831 if (lace_size[n] > size) {
2832 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2836 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2837 st->codec->codec_id == AV_CODEC_ID_COOK ||
2838 st->codec->codec_id == AV_CODEC_ID_SIPR ||
2839 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2840 st->codec->block_align && track->audio.sub_packet_size) {
2841 res = matroska_parse_rm_audio(matroska, track, st, data,
2847 } else if (st->codec->codec_id == AV_CODEC_ID_WEBVTT) {
2848 res = matroska_parse_webvtt(matroska, track, st,
2850 timecode, lace_duration,
2855 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2856 timecode, lace_duration, pos,
2857 !n ? is_keyframe : 0,
2858 additional, additional_id, additional_size,
2864 if (timecode != AV_NOPTS_VALUE)
2865 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2866 data += lace_size[n];
2867 size -= lace_size[n];
2875 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2877 EbmlList *blocks_list;
2878 MatroskaBlock *blocks;
2880 res = ebml_parse(matroska,
2881 matroska_cluster_incremental_parsing,
2882 &matroska->current_cluster);
2885 if (matroska->current_cluster_pos)
2886 ebml_level_end(matroska);
2887 ebml_free(matroska_cluster, &matroska->current_cluster);
2888 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
2889 matroska->current_cluster_num_blocks = 0;
2890 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
2891 matroska->prev_pkt = NULL;
2892 /* sizeof the ID which was already read */
2893 if (matroska->current_id)
2894 matroska->current_cluster_pos -= 4;
2895 res = ebml_parse(matroska,
2896 matroska_clusters_incremental,
2897 &matroska->current_cluster);
2898 /* Try parsing the block again. */
2900 res = ebml_parse(matroska,
2901 matroska_cluster_incremental_parsing,
2902 &matroska->current_cluster);
2906 matroska->current_cluster_num_blocks <
2907 matroska->current_cluster.blocks.nb_elem) {
2908 blocks_list = &matroska->current_cluster.blocks;
2909 blocks = blocks_list->elem;
2911 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
2912 i = blocks_list->nb_elem - 1;
2913 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2914 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2915 uint8_t* additional = blocks[i].additional.size > 0 ?
2916 blocks[i].additional.data : NULL;
2917 if (!blocks[i].non_simple)
2918 blocks[i].duration = 0;
2919 res = matroska_parse_block(matroska, blocks[i].bin.data,
2920 blocks[i].bin.size, blocks[i].bin.pos,
2921 matroska->current_cluster.timecode,
2922 blocks[i].duration, is_keyframe,
2923 additional, blocks[i].additional_id,
2924 blocks[i].additional.size,
2925 matroska->current_cluster_pos,
2926 blocks[i].discard_padding);
2933 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
2935 MatroskaCluster cluster = { 0 };
2936 EbmlList *blocks_list;
2937 MatroskaBlock *blocks;
2941 if (!matroska->contains_ssa)
2942 return matroska_parse_cluster_incremental(matroska);
2943 pos = avio_tell(matroska->ctx->pb);
2944 matroska->prev_pkt = NULL;
2945 if (matroska->current_id)
2946 pos -= 4; /* sizeof the ID which was already read */
2947 res = ebml_parse(matroska, matroska_clusters, &cluster);
2948 blocks_list = &cluster.blocks;
2949 blocks = blocks_list->elem;
2950 for (i = 0; i < blocks_list->nb_elem; i++)
2951 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
2952 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
2953 res = matroska_parse_block(matroska, blocks[i].bin.data,
2954 blocks[i].bin.size, blocks[i].bin.pos,
2955 cluster.timecode, blocks[i].duration,
2956 is_keyframe, NULL, 0, 0, pos,
2957 blocks[i].discard_padding);
2959 ebml_free(matroska_cluster, &cluster);
2963 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
2965 MatroskaDemuxContext *matroska = s->priv_data;
2967 while (matroska_deliver_packet(matroska, pkt)) {
2968 int64_t pos = avio_tell(matroska->ctx->pb);
2971 if (matroska_parse_cluster(matroska) < 0)
2972 matroska_resync(matroska, pos);
2978 static int matroska_read_seek(AVFormatContext *s, int stream_index,
2979 int64_t timestamp, int flags)
2981 MatroskaDemuxContext *matroska = s->priv_data;
2982 MatroskaTrack *tracks = NULL;
2983 AVStream *st = s->streams[stream_index];
2984 int i, index, index_sub, index_min;
2986 /* Parse the CUES now since we need the index data to seek. */
2987 if (matroska->cues_parsing_deferred > 0) {
2988 matroska->cues_parsing_deferred = 0;
2989 matroska_parse_cues(matroska);
2992 if (!st->nb_index_entries)
2994 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
2996 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
2997 avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
2999 matroska->current_id = 0;
3000 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3001 matroska_clear_queue(matroska);
3002 if (matroska_parse_cluster(matroska) < 0)
3007 matroska_clear_queue(matroska);
3008 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3012 tracks = matroska->tracks.elem;
3013 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3014 tracks[i].audio.pkt_cnt = 0;
3015 tracks[i].audio.sub_packet_cnt = 0;
3016 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3017 tracks[i].end_timecode = 0;
3018 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3019 tracks[i].stream->discard != AVDISCARD_ALL) {
3020 index_sub = av_index_search_timestamp(
3021 tracks[i].stream, st->index_entries[index].timestamp,
3022 AVSEEK_FLAG_BACKWARD);
3023 while (index_sub >= 0 &&
3025 tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
3026 st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
3031 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
3032 matroska->current_id = 0;
3033 if (flags & AVSEEK_FLAG_ANY) {
3034 st->skip_to_keyframe = 0;
3035 matroska->skip_to_timecode = timestamp;
3037 st->skip_to_keyframe = 1;
3038 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3040 matroska->skip_to_keyframe = 1;
3042 matroska->num_levels = 0;
3043 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3046 // slightly hackish but allows proper fallback to
3047 // the generic seeking code.
3048 matroska_clear_queue(matroska);
3049 matroska->current_id = 0;
3050 st->skip_to_keyframe =
3051 matroska->skip_to_keyframe = 0;
3053 matroska->num_levels = 0;
3057 static int matroska_read_close(AVFormatContext *s)
3059 MatroskaDemuxContext *matroska = s->priv_data;
3060 MatroskaTrack *tracks = matroska->tracks.elem;
3063 matroska_clear_queue(matroska);
3065 for (n = 0; n < matroska->tracks.nb_elem; n++)
3066 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3067 av_freep(&tracks[n].audio.buf);
3068 ebml_free(matroska_cluster, &matroska->current_cluster);
3069 ebml_free(matroska_segment, matroska);
3075 int64_t start_time_ns;
3076 int64_t end_time_ns;
3077 int64_t start_offset;
3081 /* This function searches all the Cues and returns the CueDesc corresponding the
3082 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3083 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3085 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3086 MatroskaDemuxContext *matroska = s->priv_data;
3089 int nb_index_entries = s->streams[0]->nb_index_entries;
3090 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3091 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3092 for (i = 1; i < nb_index_entries; i++) {
3093 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3094 index_entries[i].timestamp * matroska->time_scale > ts) {
3099 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3100 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3101 if (i != nb_index_entries - 1) {
3102 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3103 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3105 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3106 // FIXME: this needs special handling for files where Cues appear
3107 // before Clusters. the current logic assumes Cues appear after
3109 cue_desc.end_offset = cues_start - matroska->segment_start;
3114 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3116 MatroskaDemuxContext *matroska = s->priv_data;
3117 int64_t cluster_pos, before_pos;
3119 if (s->streams[0]->nb_index_entries <= 0) return 0;
3120 // seek to the first cluster using cues.
3121 index = av_index_search_timestamp(s->streams[0], 0, 0);
3122 if (index < 0) return 0;
3123 cluster_pos = s->streams[0]->index_entries[index].pos;
3124 before_pos = avio_tell(s->pb);
3126 int64_t cluster_id = 0, cluster_length = 0;
3128 avio_seek(s->pb, cluster_pos, SEEK_SET);
3129 // read cluster id and length
3130 ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id);
3131 ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3132 if (cluster_id != 0xF43B675) { // done with all clusters
3135 avio_seek(s->pb, cluster_pos, SEEK_SET);
3136 matroska->current_id = 0;
3137 matroska_clear_queue(matroska);
3138 if (matroska_parse_cluster(matroska) < 0 ||
3139 matroska->num_packets <= 0) {
3142 pkt = matroska->packets[0];
3143 cluster_pos += cluster_length + 12; // 12 is the offset of the cluster id and length.
3144 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3149 avio_seek(s->pb, before_pos, SEEK_SET);
3153 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3154 double min_buffer, double* buffer,
3155 double* sec_to_download, AVFormatContext *s,
3158 double nano_seconds_per_second = 1000000000.0;
3159 double time_sec = time_ns / nano_seconds_per_second;
3161 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3162 int64_t end_time_ns = time_ns + time_to_search_ns;
3163 double sec_downloaded = 0.0;
3164 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3165 if (desc_curr.start_time_ns == -1)
3167 *sec_to_download = 0.0;
3169 // Check for non cue start time.
3170 if (time_ns > desc_curr.start_time_ns) {
3171 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3172 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3173 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3174 double timeToDownload = (cueBytes * 8.0) / bps;
3176 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3177 *sec_to_download += timeToDownload;
3179 // Check if the search ends within the first cue.
3180 if (desc_curr.end_time_ns >= end_time_ns) {
3181 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3182 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3183 sec_downloaded = percent_to_sub * sec_downloaded;
3184 *sec_to_download = percent_to_sub * *sec_to_download;
3187 if ((sec_downloaded + *buffer) <= min_buffer) {
3191 // Get the next Cue.
3192 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3195 while (desc_curr.start_time_ns != -1) {
3196 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3197 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3198 double desc_sec = desc_ns / nano_seconds_per_second;
3199 double bits = (desc_bytes * 8.0);
3200 double time_to_download = bits / bps;
3202 sec_downloaded += desc_sec - time_to_download;
3203 *sec_to_download += time_to_download;
3205 if (desc_curr.end_time_ns >= end_time_ns) {
3206 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3207 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3208 sec_downloaded = percent_to_sub * sec_downloaded;
3209 *sec_to_download = percent_to_sub * *sec_to_download;
3211 if ((sec_downloaded + *buffer) <= min_buffer)
3216 if ((sec_downloaded + *buffer) <= min_buffer) {
3221 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3223 *buffer = *buffer + sec_downloaded;
3227 /* This function computes the bandwidth of the WebM file with the help of
3228 * buffer_size_after_time_downloaded() function. Both of these functions are
3229 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3230 * Matroska parsing mechanism.
3232 * Returns the bandwidth of the file on success; -1 on error.
3234 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3236 MatroskaDemuxContext *matroska = s->priv_data;
3237 AVStream *st = s->streams[0];
3238 double bandwidth = 0.0;
3241 for (i = 0; i < st->nb_index_entries; i++) {
3242 int64_t prebuffer_ns = 1000000000;
3243 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3244 double nano_seconds_per_second = 1000000000.0;
3245 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3246 double prebuffer_bytes = 0.0;
3247 int64_t temp_prebuffer_ns = prebuffer_ns;
3248 int64_t pre_bytes, pre_ns;
3249 double pre_sec, prebuffer, bits_per_second;
3250 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3252 // Start with the first Cue.
3253 CueDesc desc_end = desc_beg;
3255 // Figure out how much data we have downloaded for the prebuffer. This will
3256 // be used later to adjust the bits per sample to try.
3257 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3258 // Prebuffered the entire Cue.
3259 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3260 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3261 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3263 if (desc_end.start_time_ns == -1) {
3264 // The prebuffer is larger than the duration.
3265 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3267 bits_per_second = 0.0;
3269 // The prebuffer ends in the last Cue. Estimate how much data was
3271 pre_bytes = desc_end.end_offset - desc_end.start_offset;
3272 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3273 pre_sec = pre_ns / nano_seconds_per_second;
3275 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3277 prebuffer = prebuffer_ns / nano_seconds_per_second;
3279 // Set this to 0.0 in case our prebuffer buffers the entire video.
3280 bits_per_second = 0.0;
3282 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3283 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3284 double desc_sec = desc_ns / nano_seconds_per_second;
3285 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3287 // Drop the bps by the percentage of bytes buffered.
3288 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3289 double mod_bits_per_second = calc_bits_per_second * percent;
3291 if (prebuffer < desc_sec) {
3293 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3295 // Add 1 so the bits per second should be a little bit greater than file
3297 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3298 const double min_buffer = 0.0;
3299 double buffer = prebuffer;
3300 double sec_to_download = 0.0;
3302 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3303 min_buffer, &buffer, &sec_to_download,
3307 } else if (rv == 0) {
3308 bits_per_second = (double)(bps);
3313 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3314 } while (desc_end.start_time_ns != -1);
3316 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
3318 return (int64_t)bandwidth;
3321 static int webm_dash_manifest_cues(AVFormatContext *s)
3323 MatroskaDemuxContext *matroska = s->priv_data;
3324 EbmlList *seekhead_list = &matroska->seekhead;
3325 MatroskaSeekhead *seekhead = seekhead_list->elem;
3327 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
3330 // determine cues start and end positions
3331 for (i = 0; i < seekhead_list->nb_elem; i++)
3332 if (seekhead[i].id == MATROSKA_ID_CUES)
3335 if (i >= seekhead_list->nb_elem) return -1;
3337 before_pos = avio_tell(matroska->ctx->pb);
3338 cues_start = seekhead[i].pos + matroska->segment_start;
3339 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
3340 // cues_end is computed as cues_start + cues_length + length of the
3341 // Cues element ID + EBML length of the Cues element. cues_end is
3342 // inclusive and the above sum is reduced by 1.
3343 uint64_t cues_length = 0, cues_id = 0, bytes_read = 0;
3344 bytes_read += ebml_read_num(matroska, matroska->ctx->pb, 4, &cues_id);
3345 bytes_read += ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
3346 cues_end = cues_start + cues_length + bytes_read - 1;
3348 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
3349 if (cues_start == -1 || cues_end == -1) return -1;
3352 matroska_parse_cues(matroska);
3355 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
3358 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
3361 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
3362 if (bandwidth < 0) return -1;
3363 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
3365 // check if all clusters start with key frames
3366 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
3368 // store cue point timestamps as a comma separated list for checking subsegment alignment in
3369 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
3370 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20 * sizeof(char));
3371 if (!buf) return -1;
3373 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
3374 snprintf(buf, (i + 1) * 20 * sizeof(char),
3375 "%s%" PRId64, buf, s->streams[0]->index_entries[i].timestamp);
3376 if (i != s->streams[0]->nb_index_entries - 1)
3377 strncat(buf, ",", sizeof(char));
3379 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
3385 static int webm_dash_manifest_read_header(AVFormatContext *s)
3388 int ret = matroska_read_header(s);
3389 MatroskaTrack *tracks;
3390 MatroskaDemuxContext *matroska = s->priv_data;
3392 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
3396 // initialization range
3397 // 5 is the offset of Cluster ID.
3398 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, avio_tell(s->pb) - 5, 0);
3400 // basename of the file
3401 buf = strrchr(s->filename, '/');
3402 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->filename, 0);
3405 buf = av_asprintf("%g", matroska->duration);
3406 if (!buf) return AVERROR(ENOMEM);
3407 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
3411 tracks = matroska->tracks.elem;
3412 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
3414 // parse the cues and populate Cue related fields
3415 return webm_dash_manifest_cues(s);
3418 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
3423 AVInputFormat ff_matroska_demuxer = {
3424 .name = "matroska,webm",
3425 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
3426 .extensions = "mkv,mk3d,mka,mks",
3427 .priv_data_size = sizeof(MatroskaDemuxContext),
3428 .read_probe = matroska_probe,
3429 .read_header = matroska_read_header,
3430 .read_packet = matroska_read_packet,
3431 .read_close = matroska_read_close,
3432 .read_seek = matroska_read_seek,
3433 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
3436 AVInputFormat ff_webm_dash_manifest_demuxer = {
3437 .name = "webm_dash_manifest",
3438 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
3439 .priv_data_size = sizeof(MatroskaDemuxContext),
3440 .read_header = webm_dash_manifest_read_header,
3441 .read_packet = webm_dash_manifest_read_packet,
3442 .read_close = matroska_read_close,