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/opt.h"
44 #include "libavutil/time_internal.h"
46 #include "libavcodec/bytestream.h"
47 #include "libavcodec/flac.h"
48 #include "libavcodec/mpeg4audio.h"
51 #include "avio_internal.h"
56 /* For ff_codec_get_id(). */
82 typedef const struct EbmlSyntax {
91 const struct EbmlSyntax *n;
95 typedef struct EbmlList {
100 typedef struct EbmlBin {
106 typedef struct Ebml {
111 uint64_t doctype_version;
114 typedef struct MatroskaTrackCompression {
117 } MatroskaTrackCompression;
119 typedef struct MatroskaTrackEncryption {
122 } MatroskaTrackEncryption;
124 typedef struct MatroskaTrackEncoding {
127 MatroskaTrackCompression compression;
128 MatroskaTrackEncryption encryption;
129 } MatroskaTrackEncoding;
131 typedef struct MatroskaTrackVideo {
133 uint64_t display_width;
134 uint64_t display_height;
135 uint64_t pixel_width;
136 uint64_t pixel_height;
138 uint64_t stereo_mode;
140 } MatroskaTrackVideo;
142 typedef struct MatroskaTrackAudio {
144 double out_samplerate;
148 /* real audio header (extracted from extradata) */
155 uint64_t buf_timecode;
157 } MatroskaTrackAudio;
159 typedef struct MatroskaTrackPlane {
162 } MatroskaTrackPlane;
164 typedef struct MatroskaTrackOperation {
165 EbmlList combine_planes;
166 } MatroskaTrackOperation;
168 typedef struct MatroskaTrack {
177 uint64_t default_duration;
178 uint64_t flag_default;
179 uint64_t flag_forced;
180 uint64_t seek_preroll;
181 MatroskaTrackVideo video;
182 MatroskaTrackAudio audio;
183 MatroskaTrackOperation operation;
185 uint64_t codec_delay;
188 int64_t end_timecode;
190 uint64_t max_block_additional_id;
193 typedef struct MatroskaAttachment {
200 } MatroskaAttachment;
202 typedef struct MatroskaChapter {
211 typedef struct MatroskaIndexPos {
216 typedef struct MatroskaIndex {
221 typedef struct MatroskaTag {
229 typedef struct MatroskaTagTarget {
237 typedef struct MatroskaTags {
238 MatroskaTagTarget target;
242 typedef struct MatroskaSeekhead {
247 typedef struct MatroskaLevel {
252 typedef struct MatroskaCluster {
257 typedef struct MatroskaLevel1Element {
261 } MatroskaLevel1Element;
263 typedef struct MatroskaDemuxContext {
264 const AVClass *class;
265 AVFormatContext *ctx;
269 MatroskaLevel levels[EBML_MAX_DEPTH];
279 EbmlList attachments;
285 /* byte position of the segment inside the stream */
286 int64_t segment_start;
288 /* the packet queue */
295 /* What to skip before effectively reading a packet. */
296 int skip_to_keyframe;
297 uint64_t skip_to_timecode;
299 /* File has a CUES element, but we defer parsing until it is needed. */
300 int cues_parsing_deferred;
302 /* Level1 elements and whether they were read yet */
303 MatroskaLevel1Element level1_elems[64];
304 int num_level1_elems;
306 int current_cluster_num_blocks;
307 int64_t current_cluster_pos;
308 MatroskaCluster current_cluster;
310 /* File has SSA subtitles which prevent incremental cluster parsing. */
313 /* WebM DASH Manifest live flag/ */
315 } MatroskaDemuxContext;
317 typedef struct MatroskaBlock {
322 uint64_t additional_id;
324 int64_t discard_padding;
327 static const EbmlSyntax ebml_header[] = {
328 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
329 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
330 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
331 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
332 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
333 { EBML_ID_EBMLVERSION, EBML_NONE },
334 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
338 static const EbmlSyntax ebml_syntax[] = {
339 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
343 static const EbmlSyntax matroska_info[] = {
344 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
345 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
346 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
347 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
348 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
349 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
350 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
354 static const EbmlSyntax matroska_track_video[] = {
355 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
356 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
357 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
358 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
359 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
360 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
361 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
362 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
363 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
364 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
365 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
366 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
367 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_NONE },
368 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
369 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
373 static const EbmlSyntax matroska_track_audio[] = {
374 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
375 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
376 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
377 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
381 static const EbmlSyntax matroska_track_encoding_compression[] = {
382 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
383 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
387 static const EbmlSyntax matroska_track_encoding_encryption[] = {
388 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
389 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
390 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
391 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
392 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
393 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
394 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
397 static const EbmlSyntax matroska_track_encoding[] = {
398 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
399 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
400 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
401 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
402 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
406 static const EbmlSyntax matroska_track_encodings[] = {
407 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
411 static const EbmlSyntax matroska_track_plane[] = {
412 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
413 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
417 static const EbmlSyntax matroska_track_combine_planes[] = {
418 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
422 static const EbmlSyntax matroska_track_operation[] = {
423 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
427 static const EbmlSyntax matroska_track[] = {
428 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
429 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
430 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
431 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
432 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
433 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
434 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
435 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
436 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
437 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
438 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
439 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
440 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
441 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
442 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
443 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
444 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
445 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
446 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
447 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
448 { MATROSKA_ID_CODECNAME, EBML_NONE },
449 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
450 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
451 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
452 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
453 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
457 static const EbmlSyntax matroska_tracks[] = {
458 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
462 static const EbmlSyntax matroska_attachment[] = {
463 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
464 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
465 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
466 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
467 { MATROSKA_ID_FILEDESC, EBML_NONE },
471 static const EbmlSyntax matroska_attachments[] = {
472 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
476 static const EbmlSyntax matroska_chapter_display[] = {
477 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
478 { MATROSKA_ID_CHAPLANG, EBML_NONE },
479 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
483 static const EbmlSyntax matroska_chapter_entry[] = {
484 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
485 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
486 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
487 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
488 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
489 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
490 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
491 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
495 static const EbmlSyntax matroska_chapter[] = {
496 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
497 { MATROSKA_ID_EDITIONUID, EBML_NONE },
498 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
499 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
500 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
504 static const EbmlSyntax matroska_chapters[] = {
505 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
509 static const EbmlSyntax matroska_index_pos[] = {
510 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
511 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
512 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
513 { MATROSKA_ID_CUEDURATION, EBML_NONE },
514 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
518 static const EbmlSyntax matroska_index_entry[] = {
519 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
520 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
524 static const EbmlSyntax matroska_index[] = {
525 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
529 static const EbmlSyntax matroska_simpletag[] = {
530 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
531 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
532 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
533 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
534 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
535 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
539 static const EbmlSyntax matroska_tagtargets[] = {
540 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
541 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
542 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
543 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
544 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
548 static const EbmlSyntax matroska_tag[] = {
549 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
550 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
554 static const EbmlSyntax matroska_tags[] = {
555 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
559 static const EbmlSyntax matroska_seekhead_entry[] = {
560 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
561 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
565 static const EbmlSyntax matroska_seekhead[] = {
566 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
570 static const EbmlSyntax matroska_segment[] = {
571 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
572 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
573 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
574 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
575 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
576 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
577 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
578 { MATROSKA_ID_CLUSTER, EBML_STOP },
582 static const EbmlSyntax matroska_segments[] = {
583 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
587 static const EbmlSyntax matroska_blockmore[] = {
588 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
589 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
593 static const EbmlSyntax matroska_blockadditions[] = {
594 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
598 static const EbmlSyntax matroska_blockgroup[] = {
599 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
600 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
601 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
602 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
603 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
604 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference) },
605 { MATROSKA_ID_CODECSTATE, EBML_NONE },
606 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
610 static const EbmlSyntax matroska_cluster[] = {
611 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
612 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
613 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
614 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
615 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
619 static const EbmlSyntax matroska_clusters[] = {
620 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
621 { MATROSKA_ID_INFO, EBML_NONE },
622 { MATROSKA_ID_CUES, EBML_NONE },
623 { MATROSKA_ID_TAGS, EBML_NONE },
624 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
628 static const EbmlSyntax matroska_cluster_incremental_parsing[] = {
629 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
630 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
631 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
632 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
633 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
634 { MATROSKA_ID_INFO, EBML_NONE },
635 { MATROSKA_ID_CUES, EBML_NONE },
636 { MATROSKA_ID_TAGS, EBML_NONE },
637 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
638 { MATROSKA_ID_CLUSTER, EBML_STOP },
642 static const EbmlSyntax matroska_cluster_incremental[] = {
643 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
644 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
645 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
646 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
647 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
651 static const EbmlSyntax matroska_clusters_incremental[] = {
652 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
653 { MATROSKA_ID_INFO, EBML_NONE },
654 { MATROSKA_ID_CUES, EBML_NONE },
655 { MATROSKA_ID_TAGS, EBML_NONE },
656 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
660 static const char *const matroska_doctypes[] = { "matroska", "webm" };
662 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
664 AVIOContext *pb = matroska->ctx->pb;
666 matroska->current_id = 0;
667 matroska->num_levels = 0;
669 /* seek to next position to resync from */
670 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
675 // try to find a toplevel element
676 while (!avio_feof(pb)) {
677 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
678 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
679 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
680 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
681 matroska->current_id = id;
684 id = (id << 8) | avio_r8(pb);
693 * Return: Whether we reached the end of a level in the hierarchy or not.
695 static int ebml_level_end(MatroskaDemuxContext *matroska)
697 AVIOContext *pb = matroska->ctx->pb;
698 int64_t pos = avio_tell(pb);
700 if (matroska->num_levels > 0) {
701 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
702 if (pos - level->start >= level->length || matroska->current_id) {
703 matroska->num_levels--;
707 return (matroska->is_live && matroska->ctx->pb->eof_reached) ? 1 : 0;
711 * Read: an "EBML number", which is defined as a variable-length
712 * array of bytes. The first byte indicates the length by giving a
713 * number of 0-bits followed by a one. The position of the first
714 * "one" bit inside the first byte indicates the length of this
716 * Returns: number of bytes read, < 0 on error
718 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
719 int max_size, uint64_t *number)
724 /* The first byte tells us the length in bytes - avio_r8() can normally
725 * return 0, but since that's not a valid first ebmlID byte, we can
726 * use it safely here to catch EOS. */
727 if (!(total = avio_r8(pb))) {
728 /* we might encounter EOS here */
729 if (!avio_feof(pb)) {
730 int64_t pos = avio_tell(pb);
731 av_log(matroska->ctx, AV_LOG_ERROR,
732 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
734 return pb->error ? pb->error : AVERROR(EIO);
739 /* get the length of the EBML number */
740 read = 8 - ff_log2_tab[total];
741 if (read > max_size) {
742 int64_t pos = avio_tell(pb) - 1;
743 av_log(matroska->ctx, AV_LOG_ERROR,
744 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
745 (uint8_t) total, pos, pos);
746 return AVERROR_INVALIDDATA;
749 /* read out length */
750 total ^= 1 << ff_log2_tab[total];
752 total = (total << 8) | avio_r8(pb);
760 * Read a EBML length value.
761 * This needs special handling for the "unknown length" case which has multiple
764 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
767 int res = ebml_read_num(matroska, pb, 8, number);
768 if (res > 0 && *number + 1 == 1ULL << (7 * res))
769 *number = 0xffffffffffffffULL;
774 * Read the next element as an unsigned int.
775 * 0 is success, < 0 is failure.
777 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
782 return AVERROR_INVALIDDATA;
784 /* big-endian ordering; build up number */
787 *num = (*num << 8) | avio_r8(pb);
793 * Read the next element as a signed int.
794 * 0 is success, < 0 is failure.
796 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
801 return AVERROR_INVALIDDATA;
806 *num = sign_extend(avio_r8(pb), 8);
808 /* big-endian ordering; build up number */
810 *num = ((uint64_t)*num << 8) | avio_r8(pb);
817 * Read the next element as a float.
818 * 0 is success, < 0 is failure.
820 static int ebml_read_float(AVIOContext *pb, int size, double *num)
825 *num = av_int2float(avio_rb32(pb));
827 *num = av_int2double(avio_rb64(pb));
829 return AVERROR_INVALIDDATA;
835 * Read the next element as an ASCII string.
836 * 0 is success, < 0 is failure.
838 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
842 /* EBML strings are usually not 0-terminated, so we allocate one
843 * byte more, read the string and NULL-terminate it ourselves. */
844 if (!(res = av_malloc(size + 1)))
845 return AVERROR(ENOMEM);
846 if (avio_read(pb, (uint8_t *) res, size) != size) {
858 * Read the next element as binary data.
859 * 0 is success, < 0 is failure.
861 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
863 av_fast_padded_malloc(&bin->data, &bin->size, length);
865 return AVERROR(ENOMEM);
868 bin->pos = avio_tell(pb);
869 if (avio_read(pb, bin->data, length) != length) {
870 av_freep(&bin->data);
879 * Read the next element, but only the header. The contents
880 * are supposed to be sub-elements which can be read separately.
881 * 0 is success, < 0 is failure.
883 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
885 AVIOContext *pb = matroska->ctx->pb;
886 MatroskaLevel *level;
888 if (matroska->num_levels >= EBML_MAX_DEPTH) {
889 av_log(matroska->ctx, AV_LOG_ERROR,
890 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
891 return AVERROR(ENOSYS);
894 level = &matroska->levels[matroska->num_levels++];
895 level->start = avio_tell(pb);
896 level->length = length;
902 * Read signed/unsigned "EBML" numbers.
903 * Return: number of bytes processed, < 0 on error
905 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
906 uint8_t *data, uint32_t size, uint64_t *num)
909 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
910 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
914 * Same as above, but signed.
916 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
917 uint8_t *data, uint32_t size, int64_t *num)
922 /* read as unsigned number first */
923 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
926 /* make signed (weird way) */
927 *num = unum - ((1LL << (7 * res - 1)) - 1);
932 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
933 EbmlSyntax *syntax, void *data);
935 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
936 uint32_t id, void *data)
939 for (i = 0; syntax[i].id; i++)
940 if (id == syntax[i].id)
942 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
943 matroska->num_levels > 0 &&
944 matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
945 return 0; // we reached the end of an unknown size cluster
946 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
947 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32"\n", id);
949 return ebml_parse_elem(matroska, &syntax[i], data);
952 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
955 if (!matroska->current_id) {
957 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
959 // in live mode, finish parsing if EOF is reached.
960 return (matroska->is_live && matroska->ctx->pb->eof_reached &&
961 res == AVERROR_EOF) ? 1 : res;
963 matroska->current_id = id | 1 << 7 * res;
965 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
968 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
973 for (i = 0; syntax[i].id; i++)
974 switch (syntax[i].type) {
976 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
979 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
983 // the default may be NULL
984 if (syntax[i].def.s) {
985 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
986 *dst = av_strdup(syntax[i].def.s);
988 return AVERROR(ENOMEM);
993 while (!res && !ebml_level_end(matroska))
994 res = ebml_parse(matroska, syntax, data);
999 static int is_ebml_id_valid(uint32_t id)
1001 // Due to endian nonsense in Matroska, the highest byte with any bits set
1002 // will contain the leading length bit. This bit in turn identifies the
1003 // total byte length of the element by its position within the byte.
1004 unsigned int bits = av_log2(id);
1005 return id && (bits + 7) / 8 == (8 - bits % 8);
1009 * Allocate and return the entry for the level1 element with the given ID. If
1010 * an entry already exists, return the existing entry.
1012 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1016 MatroskaLevel1Element *elem;
1018 if (!is_ebml_id_valid(id))
1021 // Some files link to all clusters; useless.
1022 if (id == MATROSKA_ID_CLUSTER)
1025 // There can be multiple seekheads.
1026 if (id != MATROSKA_ID_SEEKHEAD) {
1027 for (i = 0; i < matroska->num_level1_elems; i++) {
1028 if (matroska->level1_elems[i].id == id)
1029 return &matroska->level1_elems[i];
1033 // Only a completely broken file would have more elements.
1034 // It also provides a low-effort way to escape from circular seekheads
1035 // (every iteration will add a level1 entry).
1036 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1037 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1041 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1042 *elem = (MatroskaLevel1Element){.id = id};
1047 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
1048 EbmlSyntax *syntax, void *data)
1050 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1053 // max. 16 MB for strings
1054 [EBML_STR] = 0x1000000,
1055 [EBML_UTF8] = 0x1000000,
1056 // max. 256 MB for binary data
1057 [EBML_BIN] = 0x10000000,
1058 // no limits for anything else
1060 AVIOContext *pb = matroska->ctx->pb;
1061 uint32_t id = syntax->id;
1065 MatroskaLevel1Element *level1_elem;
1067 data = (char *) data + syntax->data_offset;
1068 if (syntax->list_elem_size) {
1069 EbmlList *list = data;
1070 newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1072 return AVERROR(ENOMEM);
1073 list->elem = newelem;
1074 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1075 memset(data, 0, syntax->list_elem_size);
1079 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1080 matroska->current_id = 0;
1081 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1083 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1084 av_log(matroska->ctx, AV_LOG_ERROR,
1085 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1086 length, max_lengths[syntax->type], syntax->type);
1087 return AVERROR_INVALIDDATA;
1091 switch (syntax->type) {
1093 res = ebml_read_uint(pb, length, data);
1096 res = ebml_read_sint(pb, length, data);
1099 res = ebml_read_float(pb, length, data);
1103 res = ebml_read_ascii(pb, length, data);
1106 res = ebml_read_binary(pb, length, data);
1110 if ((res = ebml_read_master(matroska, length)) < 0)
1112 if (id == MATROSKA_ID_SEGMENT)
1113 matroska->segment_start = avio_tell(matroska->ctx->pb);
1114 if (id == MATROSKA_ID_CUES)
1115 matroska->cues_parsing_deferred = 0;
1116 if (syntax->type == EBML_LEVEL1 &&
1117 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1118 if (level1_elem->parsed)
1119 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1120 level1_elem->parsed = 1;
1122 return ebml_parse_nest(matroska, syntax->def.n, data);
1124 return ebml_parse_id(matroska, syntax->def.n, id, data);
1128 if (ffio_limit(pb, length) != length)
1129 return AVERROR(EIO);
1130 return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
1132 if (res == AVERROR_INVALIDDATA)
1133 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1134 else if (res == AVERROR(EIO))
1135 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1139 static void ebml_free(EbmlSyntax *syntax, void *data)
1142 for (i = 0; syntax[i].id; i++) {
1143 void *data_off = (char *) data + syntax[i].data_offset;
1144 switch (syntax[i].type) {
1150 av_freep(&((EbmlBin *) data_off)->data);
1154 if (syntax[i].list_elem_size) {
1155 EbmlList *list = data_off;
1156 char *ptr = list->elem;
1157 for (j = 0; j < list->nb_elem;
1158 j++, ptr += syntax[i].list_elem_size)
1159 ebml_free(syntax[i].def.n, ptr);
1160 av_freep(&list->elem);
1162 ebml_free(syntax[i].def.n, data_off);
1172 static int matroska_probe(AVProbeData *p)
1175 int len_mask = 0x80, size = 1, n = 1, i;
1178 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1181 /* length of header */
1183 while (size <= 8 && !(total & len_mask)) {
1189 total &= (len_mask - 1);
1191 total = (total << 8) | p->buf[4 + n++];
1193 /* Does the probe data contain the whole header? */
1194 if (p->buf_size < 4 + size + total)
1197 /* The header should contain a known document type. For now,
1198 * we don't parse the whole header but simply check for the
1199 * availability of that array of characters inside the header.
1200 * Not fully fool-proof, but good enough. */
1201 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1202 size_t probelen = strlen(matroska_doctypes[i]);
1203 if (total < probelen)
1205 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1206 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1207 return AVPROBE_SCORE_MAX;
1210 // probably valid EBML header but no recognized doctype
1211 return AVPROBE_SCORE_EXTENSION;
1214 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1217 MatroskaTrack *tracks = matroska->tracks.elem;
1220 for (i = 0; i < matroska->tracks.nb_elem; i++)
1221 if (tracks[i].num == num)
1224 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1228 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1229 MatroskaTrack *track)
1231 MatroskaTrackEncoding *encodings = track->encodings.elem;
1232 uint8_t *data = *buf;
1233 int isize = *buf_size;
1234 uint8_t *pkt_data = NULL;
1235 uint8_t av_unused *newpktdata;
1236 int pkt_size = isize;
1240 if (pkt_size >= 10000000U)
1241 return AVERROR_INVALIDDATA;
1243 switch (encodings[0].compression.algo) {
1244 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1246 int header_size = encodings[0].compression.settings.size;
1247 uint8_t *header = encodings[0].compression.settings.data;
1249 if (header_size && !header) {
1250 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1257 pkt_size = isize + header_size;
1258 pkt_data = av_malloc(pkt_size);
1260 return AVERROR(ENOMEM);
1262 memcpy(pkt_data, header, header_size);
1263 memcpy(pkt_data + header_size, data, isize);
1267 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1269 olen = pkt_size *= 3;
1270 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1272 result = AVERROR(ENOMEM);
1275 pkt_data = newpktdata;
1276 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1277 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1279 result = AVERROR_INVALIDDATA;
1286 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1288 z_stream zstream = { 0 };
1289 if (inflateInit(&zstream) != Z_OK)
1291 zstream.next_in = data;
1292 zstream.avail_in = isize;
1295 newpktdata = av_realloc(pkt_data, pkt_size);
1297 inflateEnd(&zstream);
1298 result = AVERROR(ENOMEM);
1301 pkt_data = newpktdata;
1302 zstream.avail_out = pkt_size - zstream.total_out;
1303 zstream.next_out = pkt_data + zstream.total_out;
1304 result = inflate(&zstream, Z_NO_FLUSH);
1305 } while (result == Z_OK && pkt_size < 10000000);
1306 pkt_size = zstream.total_out;
1307 inflateEnd(&zstream);
1308 if (result != Z_STREAM_END) {
1309 if (result == Z_MEM_ERROR)
1310 result = AVERROR(ENOMEM);
1312 result = AVERROR_INVALIDDATA;
1319 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1321 bz_stream bzstream = { 0 };
1322 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1324 bzstream.next_in = data;
1325 bzstream.avail_in = isize;
1328 newpktdata = av_realloc(pkt_data, pkt_size);
1330 BZ2_bzDecompressEnd(&bzstream);
1331 result = AVERROR(ENOMEM);
1334 pkt_data = newpktdata;
1335 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1336 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1337 result = BZ2_bzDecompress(&bzstream);
1338 } while (result == BZ_OK && pkt_size < 10000000);
1339 pkt_size = bzstream.total_out_lo32;
1340 BZ2_bzDecompressEnd(&bzstream);
1341 if (result != BZ_STREAM_END) {
1342 if (result == BZ_MEM_ERROR)
1343 result = AVERROR(ENOMEM);
1345 result = AVERROR_INVALIDDATA;
1352 return AVERROR_INVALIDDATA;
1356 *buf_size = pkt_size;
1364 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1365 AVDictionary **metadata, char *prefix)
1367 MatroskaTag *tags = list->elem;
1371 for (i = 0; i < list->nb_elem; i++) {
1372 const char *lang = tags[i].lang &&
1373 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1375 if (!tags[i].name) {
1376 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1380 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1382 av_strlcpy(key, tags[i].name, sizeof(key));
1383 if (tags[i].def || !lang) {
1384 av_dict_set(metadata, key, tags[i].string, 0);
1385 if (tags[i].sub.nb_elem)
1386 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1389 av_strlcat(key, "-", sizeof(key));
1390 av_strlcat(key, lang, sizeof(key));
1391 av_dict_set(metadata, key, tags[i].string, 0);
1392 if (tags[i].sub.nb_elem)
1393 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1396 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1399 static void matroska_convert_tags(AVFormatContext *s)
1401 MatroskaDemuxContext *matroska = s->priv_data;
1402 MatroskaTags *tags = matroska->tags.elem;
1405 for (i = 0; i < matroska->tags.nb_elem; i++) {
1406 if (tags[i].target.attachuid) {
1407 MatroskaAttachment *attachment = matroska->attachments.elem;
1409 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1410 if (attachment[j].uid == tags[i].target.attachuid &&
1411 attachment[j].stream) {
1412 matroska_convert_tag(s, &tags[i].tag,
1413 &attachment[j].stream->metadata, NULL);
1418 av_log(NULL, AV_LOG_WARNING,
1419 "The tags at index %d refer to a "
1420 "non-existent attachment %"PRId64".\n",
1421 i, tags[i].target.attachuid);
1423 } else if (tags[i].target.chapteruid) {
1424 MatroskaChapter *chapter = matroska->chapters.elem;
1426 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1427 if (chapter[j].uid == tags[i].target.chapteruid &&
1428 chapter[j].chapter) {
1429 matroska_convert_tag(s, &tags[i].tag,
1430 &chapter[j].chapter->metadata, NULL);
1435 av_log(NULL, AV_LOG_WARNING,
1436 "The tags at index %d refer to a non-existent chapter "
1438 i, tags[i].target.chapteruid);
1440 } else if (tags[i].target.trackuid) {
1441 MatroskaTrack *track = matroska->tracks.elem;
1443 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1444 if (track[j].uid == tags[i].target.trackuid &&
1446 matroska_convert_tag(s, &tags[i].tag,
1447 &track[j].stream->metadata, NULL);
1452 av_log(NULL, AV_LOG_WARNING,
1453 "The tags at index %d refer to a non-existent track "
1455 i, tags[i].target.trackuid);
1458 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1459 tags[i].target.type);
1464 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1467 uint32_t level_up = matroska->level_up;
1468 uint32_t saved_id = matroska->current_id;
1469 int64_t before_pos = avio_tell(matroska->ctx->pb);
1470 MatroskaLevel level;
1475 offset = pos + matroska->segment_start;
1476 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1477 /* We don't want to lose our seekhead level, so we add
1478 * a dummy. This is a crude hack. */
1479 if (matroska->num_levels == EBML_MAX_DEPTH) {
1480 av_log(matroska->ctx, AV_LOG_INFO,
1481 "Max EBML element depth (%d) reached, "
1482 "cannot parse further.\n", EBML_MAX_DEPTH);
1483 ret = AVERROR_INVALIDDATA;
1486 level.length = (uint64_t) -1;
1487 matroska->levels[matroska->num_levels] = level;
1488 matroska->num_levels++;
1489 matroska->current_id = 0;
1491 ret = ebml_parse(matroska, matroska_segment, matroska);
1493 /* remove dummy level */
1494 while (matroska->num_levels) {
1495 uint64_t length = matroska->levels[--matroska->num_levels].length;
1496 if (length == (uint64_t) -1)
1502 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1503 matroska->level_up = level_up;
1504 matroska->current_id = saved_id;
1509 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1511 EbmlList *seekhead_list = &matroska->seekhead;
1514 // we should not do any seeking in the streaming case
1515 if (!matroska->ctx->pb->seekable)
1518 for (i = 0; i < seekhead_list->nb_elem; i++) {
1519 MatroskaSeekhead *seekheads = seekhead_list->elem;
1520 uint32_t id = seekheads[i].id;
1521 uint64_t pos = seekheads[i].pos;
1523 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1524 if (!elem || elem->parsed)
1529 // defer cues parsing until we actually need cue data.
1530 if (id == MATROSKA_ID_CUES)
1533 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1534 // mark index as broken
1535 matroska->cues_parsing_deferred = -1;
1543 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1545 EbmlList *index_list;
1546 MatroskaIndex *index;
1547 uint64_t index_scale = 1;
1550 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1553 index_list = &matroska->index;
1554 index = index_list->elem;
1555 if (index_list->nb_elem < 2)
1557 if (index[1].time > 1E14 / matroska->time_scale) {
1558 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1561 for (i = 0; i < index_list->nb_elem; i++) {
1562 EbmlList *pos_list = &index[i].pos;
1563 MatroskaIndexPos *pos = pos_list->elem;
1564 for (j = 0; j < pos_list->nb_elem; j++) {
1565 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1567 if (track && track->stream)
1568 av_add_index_entry(track->stream,
1569 pos[j].pos + matroska->segment_start,
1570 index[i].time / index_scale, 0, 0,
1576 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1579 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1582 for (i = 0; i < matroska->num_level1_elems; i++) {
1583 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1584 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1585 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1586 matroska->cues_parsing_deferred = -1;
1592 matroska_add_index_entries(matroska);
1595 static int matroska_aac_profile(char *codec_id)
1597 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1600 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1601 if (strstr(codec_id, aac_profiles[profile]))
1606 static int matroska_aac_sri(int samplerate)
1610 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1611 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1616 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1619 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1620 time_t creation_time = date_utc / 1000000000 + 978307200;
1621 struct tm tmpbuf, *ptm = gmtime_r(&creation_time, &tmpbuf);
1623 if (strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm))
1624 av_dict_set(metadata, "creation_time", buffer, 0);
1627 static int matroska_parse_flac(AVFormatContext *s,
1628 MatroskaTrack *track,
1631 AVStream *st = track->stream;
1632 uint8_t *p = track->codec_priv.data;
1633 int size = track->codec_priv.size;
1635 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1636 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1637 track->codec_priv.size = 0;
1641 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1643 p += track->codec_priv.size;
1644 size -= track->codec_priv.size;
1646 /* parse the remaining metadata blocks if present */
1648 int block_last, block_type, block_size;
1650 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1654 if (block_size > size)
1657 /* check for the channel mask */
1658 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1659 AVDictionary *dict = NULL;
1660 AVDictionaryEntry *chmask;
1662 ff_vorbis_comment(s, &dict, p, block_size, 0);
1663 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1665 uint64_t mask = strtol(chmask->value, NULL, 0);
1666 if (!mask || mask & ~0x3ffffULL) {
1667 av_log(s, AV_LOG_WARNING,
1668 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1670 st->codec->channel_layout = mask;
1672 av_dict_free(&dict);
1682 static void mkv_stereo_mode_display_mul(int stereo_mode, int *h_width, int *h_height)
1684 switch (stereo_mode) {
1685 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
1686 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
1687 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
1688 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
1689 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
1691 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
1692 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
1693 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
1694 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
1697 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
1698 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
1699 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
1700 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
1706 static int matroska_parse_tracks(AVFormatContext *s)
1708 MatroskaDemuxContext *matroska = s->priv_data;
1709 MatroskaTrack *tracks = matroska->tracks.elem;
1714 for (i = 0; i < matroska->tracks.nb_elem; i++) {
1715 MatroskaTrack *track = &tracks[i];
1716 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1717 EbmlList *encodings_list = &track->encodings;
1718 MatroskaTrackEncoding *encodings = encodings_list->elem;
1719 uint8_t *extradata = NULL;
1720 int extradata_size = 0;
1721 int extradata_offset = 0;
1722 uint32_t fourcc = 0;
1724 char* key_id_base64 = NULL;
1727 /* Apply some sanity checks. */
1728 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1729 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1730 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1731 track->type != MATROSKA_TRACK_TYPE_METADATA) {
1732 av_log(matroska->ctx, AV_LOG_INFO,
1733 "Unknown or unsupported track type %"PRIu64"\n",
1737 if (!track->codec_id)
1740 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
1741 isnan(track->audio.samplerate)) {
1742 av_log(matroska->ctx, AV_LOG_WARNING,
1743 "Invalid sample rate %f, defaulting to 8000 instead.\n",
1744 track->audio.samplerate);
1745 track->audio.samplerate = 8000;
1748 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1749 if (!track->default_duration && track->video.frame_rate > 0)
1750 track->default_duration = 1000000000 / track->video.frame_rate;
1751 if (track->video.display_width == -1)
1752 track->video.display_width = track->video.pixel_width;
1753 if (track->video.display_height == -1)
1754 track->video.display_height = track->video.pixel_height;
1755 if (track->video.color_space.size == 4)
1756 fourcc = AV_RL32(track->video.color_space.data);
1757 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1758 if (!track->audio.out_samplerate)
1759 track->audio.out_samplerate = track->audio.samplerate;
1761 if (encodings_list->nb_elem > 1) {
1762 av_log(matroska->ctx, AV_LOG_ERROR,
1763 "Multiple combined encodings not supported");
1764 } else if (encodings_list->nb_elem == 1) {
1765 if (encodings[0].type) {
1766 if (encodings[0].encryption.key_id.size > 0) {
1767 /* Save the encryption key id to be stored later as a
1769 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1770 key_id_base64 = av_malloc(b64_size);
1771 if (key_id_base64 == NULL)
1772 return AVERROR(ENOMEM);
1774 av_base64_encode(key_id_base64, b64_size,
1775 encodings[0].encryption.key_id.data,
1776 encodings[0].encryption.key_id.size);
1778 encodings[0].scope = 0;
1779 av_log(matroska->ctx, AV_LOG_ERROR,
1780 "Unsupported encoding type");
1784 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1787 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1790 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1792 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
1793 encodings[0].scope = 0;
1794 av_log(matroska->ctx, AV_LOG_ERROR,
1795 "Unsupported encoding type");
1796 } else if (track->codec_priv.size && encodings[0].scope & 2) {
1797 uint8_t *codec_priv = track->codec_priv.data;
1798 int ret = matroska_decode_buffer(&track->codec_priv.data,
1799 &track->codec_priv.size,
1802 track->codec_priv.data = NULL;
1803 track->codec_priv.size = 0;
1804 av_log(matroska->ctx, AV_LOG_ERROR,
1805 "Failed to decode codec private data\n");
1808 if (codec_priv != track->codec_priv.data)
1809 av_free(codec_priv);
1813 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
1814 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1815 strlen(ff_mkv_codec_tags[j].str))) {
1816 codec_id = ff_mkv_codec_tags[j].id;
1821 st = track->stream = avformat_new_stream(s, NULL);
1823 av_free(key_id_base64);
1824 return AVERROR(ENOMEM);
1827 if (key_id_base64) {
1828 /* export encryption key id as base64 metadata tag */
1829 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
1830 av_freep(&key_id_base64);
1833 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
1834 track->codec_priv.size >= 40 &&
1835 track->codec_priv.data) {
1836 track->ms_compat = 1;
1837 bit_depth = AV_RL16(track->codec_priv.data + 14);
1838 fourcc = AV_RL32(track->codec_priv.data + 16);
1839 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
1842 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
1844 extradata_offset = 40;
1845 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
1846 track->codec_priv.size >= 14 &&
1847 track->codec_priv.data) {
1849 ffio_init_context(&b, track->codec_priv.data,
1850 track->codec_priv.size,
1851 0, NULL, NULL, NULL, NULL);
1852 ret = ff_get_wav_header(s, &b, st->codec, track->codec_priv.size, 0);
1855 codec_id = st->codec->codec_id;
1856 extradata_offset = FFMIN(track->codec_priv.size, 18);
1857 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
1858 && (track->codec_priv.size >= 86)
1859 && (track->codec_priv.data)) {
1860 fourcc = AV_RL32(track->codec_priv.data + 4);
1861 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1862 if (ff_codec_get_id(ff_codec_movaudio_tags, AV_RL32(track->codec_priv.data))) {
1863 fourcc = AV_RL32(track->codec_priv.data);
1864 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1866 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
1867 (track->codec_priv.size >= 21) &&
1868 (track->codec_priv.data)) {
1869 fourcc = AV_RL32(track->codec_priv.data + 4);
1870 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1871 if (ff_codec_get_id(ff_codec_movvideo_tags, AV_RL32(track->codec_priv.data))) {
1872 fourcc = AV_RL32(track->codec_priv.data);
1873 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1875 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI "))
1876 codec_id = AV_CODEC_ID_SVQ3;
1877 if (codec_id == AV_CODEC_ID_NONE) {
1879 av_get_codec_tag_string(buf, sizeof(buf), fourcc);
1880 av_log(matroska->ctx, AV_LOG_ERROR,
1881 "mov FourCC not found %s.\n", buf);
1883 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1884 switch (track->audio.bitdepth) {
1886 codec_id = AV_CODEC_ID_PCM_U8;
1889 codec_id = AV_CODEC_ID_PCM_S24BE;
1892 codec_id = AV_CODEC_ID_PCM_S32BE;
1895 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1896 switch (track->audio.bitdepth) {
1898 codec_id = AV_CODEC_ID_PCM_U8;
1901 codec_id = AV_CODEC_ID_PCM_S24LE;
1904 codec_id = AV_CODEC_ID_PCM_S32LE;
1907 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
1908 track->audio.bitdepth == 64) {
1909 codec_id = AV_CODEC_ID_PCM_F64LE;
1910 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1911 int profile = matroska_aac_profile(track->codec_id);
1912 int sri = matroska_aac_sri(track->audio.samplerate);
1913 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
1915 return AVERROR(ENOMEM);
1916 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
1917 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
1918 if (strstr(track->codec_id, "SBR")) {
1919 sri = matroska_aac_sri(track->audio.out_samplerate);
1920 extradata[2] = 0x56;
1921 extradata[3] = 0xE5;
1922 extradata[4] = 0x80 | (sri << 3);
1926 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
1927 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1928 * Create the "atom size", "tag", and "tag version" fields the
1929 * decoder expects manually. */
1930 extradata_size = 12 + track->codec_priv.size;
1931 extradata = av_mallocz(extradata_size +
1932 AV_INPUT_BUFFER_PADDING_SIZE);
1934 return AVERROR(ENOMEM);
1935 AV_WB32(extradata, extradata_size);
1936 memcpy(&extradata[4], "alac", 4);
1937 AV_WB32(&extradata[8], 0);
1938 memcpy(&extradata[12], track->codec_priv.data,
1939 track->codec_priv.size);
1940 } else if (codec_id == AV_CODEC_ID_TTA) {
1941 extradata_size = 30;
1942 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
1944 return AVERROR(ENOMEM);
1945 ffio_init_context(&b, extradata, extradata_size, 1,
1946 NULL, NULL, NULL, NULL);
1947 avio_write(&b, "TTA1", 4);
1949 if (track->audio.channels > UINT16_MAX ||
1950 track->audio.bitdepth > UINT16_MAX) {
1951 av_log(matroska->ctx, AV_LOG_WARNING,
1952 "Too large audio channel number %"PRIu64
1953 " or bitdepth %"PRIu64". Skipping track.\n",
1954 track->audio.channels, track->audio.bitdepth);
1955 av_freep(&extradata);
1956 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
1957 return AVERROR_INVALIDDATA;
1961 avio_wl16(&b, track->audio.channels);
1962 avio_wl16(&b, track->audio.bitdepth);
1963 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
1964 return AVERROR_INVALIDDATA;
1965 avio_wl32(&b, track->audio.out_samplerate);
1966 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
1967 track->audio.out_samplerate,
1968 AV_TIME_BASE * 1000));
1969 } else if (codec_id == AV_CODEC_ID_RV10 ||
1970 codec_id == AV_CODEC_ID_RV20 ||
1971 codec_id == AV_CODEC_ID_RV30 ||
1972 codec_id == AV_CODEC_ID_RV40) {
1973 extradata_offset = 26;
1974 } else if (codec_id == AV_CODEC_ID_RA_144) {
1975 track->audio.out_samplerate = 8000;
1976 track->audio.channels = 1;
1977 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
1978 codec_id == AV_CODEC_ID_COOK ||
1979 codec_id == AV_CODEC_ID_ATRAC3 ||
1980 codec_id == AV_CODEC_ID_SIPR)
1981 && track->codec_priv.data) {
1984 ffio_init_context(&b, track->codec_priv.data,
1985 track->codec_priv.size,
1986 0, NULL, NULL, NULL, NULL);
1988 flavor = avio_rb16(&b);
1989 track->audio.coded_framesize = avio_rb32(&b);
1991 track->audio.sub_packet_h = avio_rb16(&b);
1992 track->audio.frame_size = avio_rb16(&b);
1993 track->audio.sub_packet_size = avio_rb16(&b);
1995 track->audio.coded_framesize <= 0 ||
1996 track->audio.sub_packet_h <= 0 ||
1997 track->audio.frame_size <= 0 ||
1998 track->audio.sub_packet_size <= 0)
1999 return AVERROR_INVALIDDATA;
2000 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2001 track->audio.frame_size);
2002 if (!track->audio.buf)
2003 return AVERROR(ENOMEM);
2004 if (codec_id == AV_CODEC_ID_RA_288) {
2005 st->codec->block_align = track->audio.coded_framesize;
2006 track->codec_priv.size = 0;
2008 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2009 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2010 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2011 st->codec->bit_rate = sipr_bit_rate[flavor];
2013 st->codec->block_align = track->audio.sub_packet_size;
2014 extradata_offset = 78;
2016 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2017 ret = matroska_parse_flac(s, track, &extradata_offset);
2020 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2021 fourcc = AV_RL32(track->codec_priv.data);
2023 track->codec_priv.size -= extradata_offset;
2025 if (codec_id == AV_CODEC_ID_NONE)
2026 av_log(matroska->ctx, AV_LOG_INFO,
2027 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2029 if (track->time_scale < 0.01)
2030 track->time_scale = 1.0;
2031 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2032 1000 * 1000 * 1000); /* 64 bit pts in ns */
2034 /* convert the delay from ns to the track timebase */
2035 track->codec_delay = av_rescale_q(track->codec_delay,
2036 (AVRational){ 1, 1000000000 },
2039 st->codec->codec_id = codec_id;
2041 if (strcmp(track->language, "und"))
2042 av_dict_set(&st->metadata, "language", track->language, 0);
2043 av_dict_set(&st->metadata, "title", track->name, 0);
2045 if (track->flag_default)
2046 st->disposition |= AV_DISPOSITION_DEFAULT;
2047 if (track->flag_forced)
2048 st->disposition |= AV_DISPOSITION_FORCED;
2050 if (!st->codec->extradata) {
2052 st->codec->extradata = extradata;
2053 st->codec->extradata_size = extradata_size;
2054 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2055 if (ff_alloc_extradata(st->codec, track->codec_priv.size))
2056 return AVERROR(ENOMEM);
2057 memcpy(st->codec->extradata,
2058 track->codec_priv.data + extradata_offset,
2059 track->codec_priv.size);
2063 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2064 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2065 int display_width_mul = 1;
2066 int display_height_mul = 1;
2068 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
2069 st->codec->codec_tag = fourcc;
2071 st->codec->bits_per_coded_sample = bit_depth;
2072 st->codec->width = track->video.pixel_width;
2073 st->codec->height = track->video.pixel_height;
2075 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2076 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2078 av_reduce(&st->sample_aspect_ratio.num,
2079 &st->sample_aspect_ratio.den,
2080 st->codec->height * track->video.display_width * display_width_mul,
2081 st->codec->width * track->video.display_height * display_height_mul,
2083 if (st->codec->codec_id != AV_CODEC_ID_HEVC)
2084 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2086 if (track->default_duration) {
2087 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2088 1000000000, track->default_duration, 30000);
2089 #if FF_API_R_FRAME_RATE
2090 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2091 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2092 st->r_frame_rate = st->avg_frame_rate;
2096 /* export stereo mode flag as metadata tag */
2097 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2098 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2100 /* export alpha mode flag as metadata tag */
2101 if (track->video.alpha_mode)
2102 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2104 /* if we have virtual track, mark the real tracks */
2105 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2107 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2109 snprintf(buf, sizeof(buf), "%s_%d",
2110 ff_matroska_video_stereo_plane[planes[j].type], i);
2111 for (k=0; k < matroska->tracks.nb_elem; k++)
2112 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2113 av_dict_set(&tracks[k].stream->metadata,
2114 "stereo_mode", buf, 0);
2118 // add stream level stereo3d side data if it is a supported format
2119 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2120 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2121 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2125 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2126 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
2127 st->codec->sample_rate = track->audio.out_samplerate;
2128 st->codec->channels = track->audio.channels;
2129 if (!st->codec->bits_per_coded_sample)
2130 st->codec->bits_per_coded_sample = track->audio.bitdepth;
2131 if (st->codec->codec_id == AV_CODEC_ID_MP3)
2132 st->need_parsing = AVSTREAM_PARSE_FULL;
2133 else if (st->codec->codec_id != AV_CODEC_ID_AAC)
2134 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2135 if (track->codec_delay > 0) {
2136 st->codec->delay = av_rescale_q(track->codec_delay,
2138 (AVRational){1, st->codec->sample_rate});
2140 if (track->seek_preroll > 0) {
2141 av_codec_set_seek_preroll(st->codec,
2142 av_rescale_q(track->seek_preroll,
2143 (AVRational){1, 1000000000},
2144 (AVRational){1, st->codec->sample_rate}));
2146 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2147 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
2149 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2150 st->disposition |= AV_DISPOSITION_CAPTIONS;
2151 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2152 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2153 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2154 st->disposition |= AV_DISPOSITION_METADATA;
2156 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2157 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
2158 if (st->codec->codec_id == AV_CODEC_ID_ASS)
2159 matroska->contains_ssa = 1;
2166 static int matroska_read_header(AVFormatContext *s)
2168 MatroskaDemuxContext *matroska = s->priv_data;
2169 EbmlList *attachments_list = &matroska->attachments;
2170 EbmlList *chapters_list = &matroska->chapters;
2171 MatroskaAttachment *attachments;
2172 MatroskaChapter *chapters;
2173 uint64_t max_start = 0;
2179 matroska->cues_parsing_deferred = 1;
2181 /* First read the EBML header. */
2182 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2183 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2184 ebml_free(ebml_syntax, &ebml);
2185 return AVERROR_INVALIDDATA;
2187 if (ebml.version > EBML_VERSION ||
2188 ebml.max_size > sizeof(uint64_t) ||
2189 ebml.id_length > sizeof(uint32_t) ||
2190 ebml.doctype_version > 3) {
2191 av_log(matroska->ctx, AV_LOG_ERROR,
2192 "EBML header using unsupported features\n"
2193 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2194 ebml.version, ebml.doctype, ebml.doctype_version);
2195 ebml_free(ebml_syntax, &ebml);
2196 return AVERROR_PATCHWELCOME;
2197 } else if (ebml.doctype_version == 3) {
2198 av_log(matroska->ctx, AV_LOG_WARNING,
2199 "EBML header using unsupported features\n"
2200 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2201 ebml.version, ebml.doctype, ebml.doctype_version);
2203 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2204 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2206 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2207 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2208 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2209 ebml_free(ebml_syntax, &ebml);
2210 return AVERROR_INVALIDDATA;
2213 ebml_free(ebml_syntax, &ebml);
2215 /* The next thing is a segment. */
2216 pos = avio_tell(matroska->ctx->pb);
2217 res = ebml_parse(matroska, matroska_segments, matroska);
2218 // try resyncing until we find a EBML_STOP type element.
2220 res = matroska_resync(matroska, pos);
2223 pos = avio_tell(matroska->ctx->pb);
2224 res = ebml_parse(matroska, matroska_segment, matroska);
2226 matroska_execute_seekhead(matroska);
2228 if (!matroska->time_scale)
2229 matroska->time_scale = 1000000;
2230 if (matroska->duration)
2231 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2232 1000 / AV_TIME_BASE;
2233 av_dict_set(&s->metadata, "title", matroska->title, 0);
2234 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2236 if (matroska->date_utc.size == 8)
2237 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2239 res = matroska_parse_tracks(s);
2243 attachments = attachments_list->elem;
2244 for (j = 0; j < attachments_list->nb_elem; j++) {
2245 if (!(attachments[j].filename && attachments[j].mime &&
2246 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2247 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2249 AVStream *st = avformat_new_stream(s, NULL);
2252 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2253 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2254 st->codec->codec_id = AV_CODEC_ID_NONE;
2256 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2257 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2258 strlen(ff_mkv_image_mime_tags[i].str))) {
2259 st->codec->codec_id = ff_mkv_image_mime_tags[i].id;
2264 attachments[j].stream = st;
2266 if (st->codec->codec_id != AV_CODEC_ID_NONE) {
2267 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2268 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
2270 av_init_packet(&st->attached_pic);
2271 if ((res = av_new_packet(&st->attached_pic, attachments[j].bin.size)) < 0)
2273 memcpy(st->attached_pic.data, attachments[j].bin.data, attachments[j].bin.size);
2274 st->attached_pic.stream_index = st->index;
2275 st->attached_pic.flags |= AV_PKT_FLAG_KEY;
2277 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2278 if (ff_alloc_extradata(st->codec, attachments[j].bin.size))
2280 memcpy(st->codec->extradata, attachments[j].bin.data,
2281 attachments[j].bin.size);
2283 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2284 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2285 strlen(ff_mkv_mime_tags[i].str))) {
2286 st->codec->codec_id = ff_mkv_mime_tags[i].id;
2294 chapters = chapters_list->elem;
2295 for (i = 0; i < chapters_list->nb_elem; i++)
2296 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2297 (max_start == 0 || chapters[i].start > max_start)) {
2298 chapters[i].chapter =
2299 avpriv_new_chapter(s, chapters[i].uid,
2300 (AVRational) { 1, 1000000000 },
2301 chapters[i].start, chapters[i].end,
2303 if (chapters[i].chapter) {
2304 av_dict_set(&chapters[i].chapter->metadata,
2305 "title", chapters[i].title, 0);
2307 max_start = chapters[i].start;
2310 matroska_add_index_entries(matroska);
2312 matroska_convert_tags(s);
2318 * Put one packet in an application-supplied AVPacket struct.
2319 * Returns 0 on success or -1 on failure.
2321 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2324 if (matroska->num_packets > 0) {
2325 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2326 av_freep(&matroska->packets[0]);
2327 if (matroska->num_packets > 1) {
2329 memmove(&matroska->packets[0], &matroska->packets[1],
2330 (matroska->num_packets - 1) * sizeof(AVPacket *));
2331 newpackets = av_realloc(matroska->packets,
2332 (matroska->num_packets - 1) *
2333 sizeof(AVPacket *));
2335 matroska->packets = newpackets;
2337 av_freep(&matroska->packets);
2338 matroska->prev_pkt = NULL;
2340 matroska->num_packets--;
2348 * Free all packets in our internal queue.
2350 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2352 matroska->prev_pkt = NULL;
2353 if (matroska->packets) {
2355 for (n = 0; n < matroska->num_packets; n++) {
2356 av_packet_unref(matroska->packets[n]);
2357 av_freep(&matroska->packets[n]);
2359 av_freep(&matroska->packets);
2360 matroska->num_packets = 0;
2364 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2365 int *buf_size, int type,
2366 uint32_t **lace_buf, int *laces)
2368 int res = 0, n, size = *buf_size;
2369 uint8_t *data = *buf;
2370 uint32_t *lace_size;
2374 *lace_buf = av_mallocz(sizeof(int));
2376 return AVERROR(ENOMEM);
2378 *lace_buf[0] = size;
2382 av_assert0(size > 0);
2386 lace_size = av_mallocz(*laces * sizeof(int));
2388 return AVERROR(ENOMEM);
2391 case 0x1: /* Xiph lacing */
2395 for (n = 0; res == 0 && n < *laces - 1; n++) {
2397 if (size <= total) {
2398 res = AVERROR_INVALIDDATA;
2403 lace_size[n] += temp;
2410 if (size <= total) {
2411 res = AVERROR_INVALIDDATA;
2415 lace_size[n] = size - total;
2419 case 0x2: /* fixed-size lacing */
2420 if (size % (*laces)) {
2421 res = AVERROR_INVALIDDATA;
2424 for (n = 0; n < *laces; n++)
2425 lace_size[n] = size / *laces;
2428 case 0x3: /* EBML lacing */
2432 n = matroska_ebmlnum_uint(matroska, data, size, &num);
2433 if (n < 0 || num > INT_MAX) {
2434 av_log(matroska->ctx, AV_LOG_INFO,
2435 "EBML block data error\n");
2436 res = n<0 ? n : AVERROR_INVALIDDATA;
2441 total = lace_size[0] = num;
2442 for (n = 1; res == 0 && n < *laces - 1; n++) {
2445 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2446 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2447 av_log(matroska->ctx, AV_LOG_INFO,
2448 "EBML block data error\n");
2449 res = r<0 ? r : AVERROR_INVALIDDATA;
2454 lace_size[n] = lace_size[n - 1] + snum;
2455 total += lace_size[n];
2457 if (size <= total) {
2458 res = AVERROR_INVALIDDATA;
2461 lace_size[*laces - 1] = size - total;
2467 *lace_buf = lace_size;
2473 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2474 MatroskaTrack *track, AVStream *st,
2475 uint8_t *data, int size, uint64_t timecode,
2478 int a = st->codec->block_align;
2479 int sps = track->audio.sub_packet_size;
2480 int cfs = track->audio.coded_framesize;
2481 int h = track->audio.sub_packet_h;
2482 int y = track->audio.sub_packet_cnt;
2483 int w = track->audio.frame_size;
2486 if (!track->audio.pkt_cnt) {
2487 if (track->audio.sub_packet_cnt == 0)
2488 track->audio.buf_timecode = timecode;
2489 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2490 if (size < cfs * h / 2) {
2491 av_log(matroska->ctx, AV_LOG_ERROR,
2492 "Corrupt int4 RM-style audio packet size\n");
2493 return AVERROR_INVALIDDATA;
2495 for (x = 0; x < h / 2; x++)
2496 memcpy(track->audio.buf + x * 2 * w + y * cfs,
2497 data + x * cfs, cfs);
2498 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2500 av_log(matroska->ctx, AV_LOG_ERROR,
2501 "Corrupt sipr RM-style audio packet size\n");
2502 return AVERROR_INVALIDDATA;
2504 memcpy(track->audio.buf + y * w, data, w);
2506 if (size < sps * w / sps || h<=0 || w%sps) {
2507 av_log(matroska->ctx, AV_LOG_ERROR,
2508 "Corrupt generic RM-style audio packet size\n");
2509 return AVERROR_INVALIDDATA;
2511 for (x = 0; x < w / sps; x++)
2512 memcpy(track->audio.buf +
2513 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2514 data + x * sps, sps);
2517 if (++track->audio.sub_packet_cnt >= h) {
2518 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2519 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2520 track->audio.sub_packet_cnt = 0;
2521 track->audio.pkt_cnt = h * w / a;
2525 while (track->audio.pkt_cnt) {
2527 AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2529 return AVERROR(ENOMEM);
2531 ret = av_new_packet(pkt, a);
2537 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2539 pkt->pts = track->audio.buf_timecode;
2540 track->audio.buf_timecode = AV_NOPTS_VALUE;
2542 pkt->stream_index = st->index;
2543 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2549 /* reconstruct full wavpack blocks from mangled matroska ones */
2550 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
2551 uint8_t **pdst, int *size)
2553 uint8_t *dst = NULL;
2558 int ret, offset = 0;
2560 if (srclen < 12 || track->stream->codec->extradata_size < 2)
2561 return AVERROR_INVALIDDATA;
2563 ver = AV_RL16(track->stream->codec->extradata);
2565 samples = AV_RL32(src);
2569 while (srclen >= 8) {
2574 uint32_t flags = AV_RL32(src);
2575 uint32_t crc = AV_RL32(src + 4);
2579 multiblock = (flags & 0x1800) != 0x1800;
2582 ret = AVERROR_INVALIDDATA;
2585 blocksize = AV_RL32(src);
2591 if (blocksize > srclen) {
2592 ret = AVERROR_INVALIDDATA;
2596 tmp = av_realloc(dst, dstlen + blocksize + 32);
2598 ret = AVERROR(ENOMEM);
2602 dstlen += blocksize + 32;
2604 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2605 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2606 AV_WL16(dst + offset + 8, ver); // version
2607 AV_WL16(dst + offset + 10, 0); // track/index_no
2608 AV_WL32(dst + offset + 12, 0); // total samples
2609 AV_WL32(dst + offset + 16, 0); // block index
2610 AV_WL32(dst + offset + 20, samples); // number of samples
2611 AV_WL32(dst + offset + 24, flags); // flags
2612 AV_WL32(dst + offset + 28, crc); // crc
2613 memcpy(dst + offset + 32, src, blocksize); // block data
2616 srclen -= blocksize;
2617 offset += blocksize + 32;
2630 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
2631 MatroskaTrack *track,
2633 uint8_t *data, int data_len,
2639 uint8_t *id, *settings, *text, *buf;
2640 int id_len, settings_len, text_len;
2645 return AVERROR_INVALIDDATA;
2648 q = data + data_len;
2653 if (*p == '\r' || *p == '\n') {
2662 if (p >= q || *p != '\n')
2663 return AVERROR_INVALIDDATA;
2669 if (*p == '\r' || *p == '\n') {
2670 settings_len = p - settings;
2678 if (p >= q || *p != '\n')
2679 return AVERROR_INVALIDDATA;
2684 while (text_len > 0) {
2685 const int len = text_len - 1;
2686 const uint8_t c = p[len];
2687 if (c != '\r' && c != '\n')
2693 return AVERROR_INVALIDDATA;
2695 pkt = av_mallocz(sizeof(*pkt));
2697 return AVERROR(ENOMEM);
2698 err = av_new_packet(pkt, text_len);
2701 return AVERROR(err);
2704 memcpy(pkt->data, text, text_len);
2707 buf = av_packet_new_side_data(pkt,
2708 AV_PKT_DATA_WEBVTT_IDENTIFIER,
2712 return AVERROR(ENOMEM);
2714 memcpy(buf, id, id_len);
2717 if (settings_len > 0) {
2718 buf = av_packet_new_side_data(pkt,
2719 AV_PKT_DATA_WEBVTT_SETTINGS,
2723 return AVERROR(ENOMEM);
2725 memcpy(buf, settings, settings_len);
2728 // Do we need this for subtitles?
2729 // pkt->flags = AV_PKT_FLAG_KEY;
2731 pkt->stream_index = st->index;
2732 pkt->pts = timecode;
2734 // Do we need this for subtitles?
2735 // pkt->dts = timecode;
2737 pkt->duration = duration;
2740 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2741 matroska->prev_pkt = pkt;
2746 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2747 MatroskaTrack *track, AVStream *st,
2748 uint8_t *data, int pkt_size,
2749 uint64_t timecode, uint64_t lace_duration,
2750 int64_t pos, int is_keyframe,
2751 uint8_t *additional, uint64_t additional_id, int additional_size,
2752 int64_t discard_padding)
2754 MatroskaTrackEncoding *encodings = track->encodings.elem;
2755 uint8_t *pkt_data = data;
2756 int offset = 0, res;
2759 if (encodings && !encodings->type && encodings->scope & 1) {
2760 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2765 if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
2767 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2769 av_log(matroska->ctx, AV_LOG_ERROR,
2770 "Error parsing a wavpack block.\n");
2773 if (pkt_data != data)
2774 av_freep(&pkt_data);
2778 if (st->codec->codec_id == AV_CODEC_ID_PRORES &&
2779 AV_RB32(&data[4]) != MKBETAG('i', 'c', 'p', 'f'))
2782 pkt = av_mallocz(sizeof(AVPacket));
2784 if (pkt_data != data)
2785 av_freep(&pkt_data);
2786 return AVERROR(ENOMEM);
2788 /* XXX: prevent data copy... */
2789 if (av_new_packet(pkt, pkt_size + offset) < 0) {
2791 res = AVERROR(ENOMEM);
2795 if (st->codec->codec_id == AV_CODEC_ID_PRORES && offset == 8) {
2796 uint8_t *buf = pkt->data;
2797 bytestream_put_be32(&buf, pkt_size);
2798 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2801 memcpy(pkt->data + offset, pkt_data, pkt_size);
2803 if (pkt_data != data)
2804 av_freep(&pkt_data);
2806 pkt->flags = is_keyframe;
2807 pkt->stream_index = st->index;
2809 if (additional_size > 0) {
2810 uint8_t *side_data = av_packet_new_side_data(pkt,
2811 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
2812 additional_size + 8);
2814 av_packet_unref(pkt);
2816 return AVERROR(ENOMEM);
2818 AV_WB64(side_data, additional_id);
2819 memcpy(side_data + 8, additional, additional_size);
2822 if (discard_padding) {
2823 uint8_t *side_data = av_packet_new_side_data(pkt,
2824 AV_PKT_DATA_SKIP_SAMPLES,
2827 av_packet_unref(pkt);
2829 return AVERROR(ENOMEM);
2831 AV_WL32(side_data, 0);
2832 AV_WL32(side_data + 4, av_rescale_q(discard_padding,
2833 (AVRational){1, 1000000000},
2834 (AVRational){1, st->codec->sample_rate}));
2837 if (track->ms_compat)
2838 pkt->dts = timecode;
2840 pkt->pts = timecode;
2842 pkt->duration = lace_duration;
2844 #if FF_API_CONVERGENCE_DURATION
2845 FF_DISABLE_DEPRECATION_WARNINGS
2846 if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2847 pkt->convergence_duration = lace_duration;
2849 FF_ENABLE_DEPRECATION_WARNINGS
2852 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2853 matroska->prev_pkt = pkt;
2858 if (pkt_data != data)
2859 av_freep(&pkt_data);
2863 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2864 int size, int64_t pos, uint64_t cluster_time,
2865 uint64_t block_duration, int is_keyframe,
2866 uint8_t *additional, uint64_t additional_id, int additional_size,
2867 int64_t cluster_pos, int64_t discard_padding)
2869 uint64_t timecode = AV_NOPTS_VALUE;
2870 MatroskaTrack *track;
2874 uint32_t *lace_size = NULL;
2875 int n, flags, laces = 0;
2877 int trust_default_duration = 1;
2879 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2880 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2886 track = matroska_find_track_by_num(matroska, num);
2887 if (!track || !track->stream) {
2888 av_log(matroska->ctx, AV_LOG_INFO,
2889 "Invalid stream %"PRIu64" or size %u\n", num, size);
2890 return AVERROR_INVALIDDATA;
2891 } else if (size <= 3)
2894 if (st->discard >= AVDISCARD_ALL)
2896 av_assert1(block_duration != AV_NOPTS_VALUE);
2898 block_time = sign_extend(AV_RB16(data), 16);
2902 if (is_keyframe == -1)
2903 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2905 if (cluster_time != (uint64_t) -1 &&
2906 (block_time >= 0 || cluster_time >= -block_time)) {
2907 timecode = cluster_time + block_time - track->codec_delay;
2908 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2909 timecode < track->end_timecode)
2910 is_keyframe = 0; /* overlapping subtitles are not key frame */
2912 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
2916 if (matroska->skip_to_keyframe &&
2917 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2918 if (timecode < matroska->skip_to_timecode)
2921 matroska->skip_to_keyframe = 0;
2922 else if (!st->skip_to_keyframe) {
2923 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2924 matroska->skip_to_keyframe = 0;
2928 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2929 &lace_size, &laces);
2934 if (track->audio.samplerate == 8000) {
2935 // If this is needed for more codecs, then add them here
2936 if (st->codec->codec_id == AV_CODEC_ID_AC3) {
2937 if (track->audio.samplerate != st->codec->sample_rate || !st->codec->frame_size)
2938 trust_default_duration = 0;
2942 if (!block_duration && trust_default_duration)
2943 block_duration = track->default_duration * laces / matroska->time_scale;
2945 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
2946 track->end_timecode =
2947 FFMAX(track->end_timecode, timecode + block_duration);
2949 for (n = 0; n < laces; n++) {
2950 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
2952 if (lace_size[n] > size) {
2953 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
2957 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
2958 st->codec->codec_id == AV_CODEC_ID_COOK ||
2959 st->codec->codec_id == AV_CODEC_ID_SIPR ||
2960 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
2961 st->codec->block_align && track->audio.sub_packet_size) {
2962 res = matroska_parse_rm_audio(matroska, track, st, data,
2968 } else if (st->codec->codec_id == AV_CODEC_ID_WEBVTT) {
2969 res = matroska_parse_webvtt(matroska, track, st,
2971 timecode, lace_duration,
2976 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
2977 timecode, lace_duration, pos,
2978 !n ? is_keyframe : 0,
2979 additional, additional_id, additional_size,
2985 if (timecode != AV_NOPTS_VALUE)
2986 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
2987 data += lace_size[n];
2988 size -= lace_size[n];
2996 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
2998 EbmlList *blocks_list;
2999 MatroskaBlock *blocks;
3001 res = ebml_parse(matroska,
3002 matroska_cluster_incremental_parsing,
3003 &matroska->current_cluster);
3006 if (matroska->current_cluster_pos)
3007 ebml_level_end(matroska);
3008 ebml_free(matroska_cluster, &matroska->current_cluster);
3009 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
3010 matroska->current_cluster_num_blocks = 0;
3011 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
3012 matroska->prev_pkt = NULL;
3013 /* sizeof the ID which was already read */
3014 if (matroska->current_id)
3015 matroska->current_cluster_pos -= 4;
3016 res = ebml_parse(matroska,
3017 matroska_clusters_incremental,
3018 &matroska->current_cluster);
3019 /* Try parsing the block again. */
3021 res = ebml_parse(matroska,
3022 matroska_cluster_incremental_parsing,
3023 &matroska->current_cluster);
3027 matroska->current_cluster_num_blocks <
3028 matroska->current_cluster.blocks.nb_elem) {
3029 blocks_list = &matroska->current_cluster.blocks;
3030 blocks = blocks_list->elem;
3032 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
3033 i = blocks_list->nb_elem - 1;
3034 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
3035 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
3036 uint8_t* additional = blocks[i].additional.size > 0 ?
3037 blocks[i].additional.data : NULL;
3038 if (!blocks[i].non_simple)
3039 blocks[i].duration = 0;
3040 res = matroska_parse_block(matroska, blocks[i].bin.data,
3041 blocks[i].bin.size, blocks[i].bin.pos,
3042 matroska->current_cluster.timecode,
3043 blocks[i].duration, is_keyframe,
3044 additional, blocks[i].additional_id,
3045 blocks[i].additional.size,
3046 matroska->current_cluster_pos,
3047 blocks[i].discard_padding);
3054 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3056 MatroskaCluster cluster = { 0 };
3057 EbmlList *blocks_list;
3058 MatroskaBlock *blocks;
3062 if (!matroska->contains_ssa)
3063 return matroska_parse_cluster_incremental(matroska);
3064 pos = avio_tell(matroska->ctx->pb);
3065 matroska->prev_pkt = NULL;
3066 if (matroska->current_id)
3067 pos -= 4; /* sizeof the ID which was already read */
3068 res = ebml_parse(matroska, matroska_clusters, &cluster);
3069 blocks_list = &cluster.blocks;
3070 blocks = blocks_list->elem;
3071 for (i = 0; i < blocks_list->nb_elem; i++)
3072 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
3073 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
3074 res = matroska_parse_block(matroska, blocks[i].bin.data,
3075 blocks[i].bin.size, blocks[i].bin.pos,
3076 cluster.timecode, blocks[i].duration,
3077 is_keyframe, NULL, 0, 0, pos,
3078 blocks[i].discard_padding);
3080 ebml_free(matroska_cluster, &cluster);
3084 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3086 MatroskaDemuxContext *matroska = s->priv_data;
3088 while (matroska_deliver_packet(matroska, pkt)) {
3089 int64_t pos = avio_tell(matroska->ctx->pb);
3092 if (matroska_parse_cluster(matroska) < 0)
3093 matroska_resync(matroska, pos);
3099 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3100 int64_t timestamp, int flags)
3102 MatroskaDemuxContext *matroska = s->priv_data;
3103 MatroskaTrack *tracks = NULL;
3104 AVStream *st = s->streams[stream_index];
3105 int i, index, index_sub, index_min;
3107 /* Parse the CUES now since we need the index data to seek. */
3108 if (matroska->cues_parsing_deferred > 0) {
3109 matroska->cues_parsing_deferred = 0;
3110 matroska_parse_cues(matroska);
3113 if (!st->nb_index_entries)
3115 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3117 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3118 avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
3120 matroska->current_id = 0;
3121 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3122 matroska_clear_queue(matroska);
3123 if (matroska_parse_cluster(matroska) < 0)
3128 matroska_clear_queue(matroska);
3129 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3133 tracks = matroska->tracks.elem;
3134 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3135 tracks[i].audio.pkt_cnt = 0;
3136 tracks[i].audio.sub_packet_cnt = 0;
3137 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3138 tracks[i].end_timecode = 0;
3139 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3141 tracks[i].stream->discard != AVDISCARD_ALL) {
3142 index_sub = av_index_search_timestamp(
3143 tracks[i].stream, st->index_entries[index].timestamp,
3144 AVSEEK_FLAG_BACKWARD);
3145 while (index_sub >= 0 &&
3147 tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
3148 st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
3153 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
3154 matroska->current_id = 0;
3155 if (flags & AVSEEK_FLAG_ANY) {
3156 st->skip_to_keyframe = 0;
3157 matroska->skip_to_timecode = timestamp;
3159 st->skip_to_keyframe = 1;
3160 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3162 matroska->skip_to_keyframe = 1;
3164 matroska->num_levels = 0;
3165 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3168 // slightly hackish but allows proper fallback to
3169 // the generic seeking code.
3170 matroska_clear_queue(matroska);
3171 matroska->current_id = 0;
3172 st->skip_to_keyframe =
3173 matroska->skip_to_keyframe = 0;
3175 matroska->num_levels = 0;
3179 static int matroska_read_close(AVFormatContext *s)
3181 MatroskaDemuxContext *matroska = s->priv_data;
3182 MatroskaTrack *tracks = matroska->tracks.elem;
3185 matroska_clear_queue(matroska);
3187 for (n = 0; n < matroska->tracks.nb_elem; n++)
3188 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3189 av_freep(&tracks[n].audio.buf);
3190 ebml_free(matroska_cluster, &matroska->current_cluster);
3191 ebml_free(matroska_segment, matroska);
3197 int64_t start_time_ns;
3198 int64_t end_time_ns;
3199 int64_t start_offset;
3203 /* This function searches all the Cues and returns the CueDesc corresponding the
3204 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3205 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3207 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3208 MatroskaDemuxContext *matroska = s->priv_data;
3211 int nb_index_entries = s->streams[0]->nb_index_entries;
3212 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3213 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3214 for (i = 1; i < nb_index_entries; i++) {
3215 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3216 index_entries[i].timestamp * matroska->time_scale > ts) {
3221 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3222 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3223 if (i != nb_index_entries - 1) {
3224 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3225 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3227 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3228 // FIXME: this needs special handling for files where Cues appear
3229 // before Clusters. the current logic assumes Cues appear after
3231 cue_desc.end_offset = cues_start - matroska->segment_start;
3236 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3238 MatroskaDemuxContext *matroska = s->priv_data;
3239 int64_t cluster_pos, before_pos;
3241 if (s->streams[0]->nb_index_entries <= 0) return 0;
3242 // seek to the first cluster using cues.
3243 index = av_index_search_timestamp(s->streams[0], 0, 0);
3244 if (index < 0) return 0;
3245 cluster_pos = s->streams[0]->index_entries[index].pos;
3246 before_pos = avio_tell(s->pb);
3248 int64_t cluster_id = 0, cluster_length = 0;
3250 avio_seek(s->pb, cluster_pos, SEEK_SET);
3251 // read cluster id and length
3252 ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id);
3253 ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3254 if (cluster_id != 0xF43B675) { // done with all clusters
3257 avio_seek(s->pb, cluster_pos, SEEK_SET);
3258 matroska->current_id = 0;
3259 matroska_clear_queue(matroska);
3260 if (matroska_parse_cluster(matroska) < 0 ||
3261 matroska->num_packets <= 0) {
3264 pkt = matroska->packets[0];
3265 cluster_pos += cluster_length + 12; // 12 is the offset of the cluster id and length.
3266 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3271 avio_seek(s->pb, before_pos, SEEK_SET);
3275 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3276 double min_buffer, double* buffer,
3277 double* sec_to_download, AVFormatContext *s,
3280 double nano_seconds_per_second = 1000000000.0;
3281 double time_sec = time_ns / nano_seconds_per_second;
3283 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3284 int64_t end_time_ns = time_ns + time_to_search_ns;
3285 double sec_downloaded = 0.0;
3286 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3287 if (desc_curr.start_time_ns == -1)
3289 *sec_to_download = 0.0;
3291 // Check for non cue start time.
3292 if (time_ns > desc_curr.start_time_ns) {
3293 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3294 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3295 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3296 double timeToDownload = (cueBytes * 8.0) / bps;
3298 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3299 *sec_to_download += timeToDownload;
3301 // Check if the search ends within the first cue.
3302 if (desc_curr.end_time_ns >= end_time_ns) {
3303 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3304 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3305 sec_downloaded = percent_to_sub * sec_downloaded;
3306 *sec_to_download = percent_to_sub * *sec_to_download;
3309 if ((sec_downloaded + *buffer) <= min_buffer) {
3313 // Get the next Cue.
3314 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3317 while (desc_curr.start_time_ns != -1) {
3318 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3319 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3320 double desc_sec = desc_ns / nano_seconds_per_second;
3321 double bits = (desc_bytes * 8.0);
3322 double time_to_download = bits / bps;
3324 sec_downloaded += desc_sec - time_to_download;
3325 *sec_to_download += time_to_download;
3327 if (desc_curr.end_time_ns >= end_time_ns) {
3328 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3329 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3330 sec_downloaded = percent_to_sub * sec_downloaded;
3331 *sec_to_download = percent_to_sub * *sec_to_download;
3333 if ((sec_downloaded + *buffer) <= min_buffer)
3338 if ((sec_downloaded + *buffer) <= min_buffer) {
3343 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3345 *buffer = *buffer + sec_downloaded;
3349 /* This function computes the bandwidth of the WebM file with the help of
3350 * buffer_size_after_time_downloaded() function. Both of these functions are
3351 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3352 * Matroska parsing mechanism.
3354 * Returns the bandwidth of the file on success; -1 on error.
3356 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3358 MatroskaDemuxContext *matroska = s->priv_data;
3359 AVStream *st = s->streams[0];
3360 double bandwidth = 0.0;
3363 for (i = 0; i < st->nb_index_entries; i++) {
3364 int64_t prebuffer_ns = 1000000000;
3365 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3366 double nano_seconds_per_second = 1000000000.0;
3367 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3368 double prebuffer_bytes = 0.0;
3369 int64_t temp_prebuffer_ns = prebuffer_ns;
3370 int64_t pre_bytes, pre_ns;
3371 double pre_sec, prebuffer, bits_per_second;
3372 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3374 // Start with the first Cue.
3375 CueDesc desc_end = desc_beg;
3377 // Figure out how much data we have downloaded for the prebuffer. This will
3378 // be used later to adjust the bits per sample to try.
3379 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3380 // Prebuffered the entire Cue.
3381 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3382 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3383 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3385 if (desc_end.start_time_ns == -1) {
3386 // The prebuffer is larger than the duration.
3387 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3389 bits_per_second = 0.0;
3391 // The prebuffer ends in the last Cue. Estimate how much data was
3393 pre_bytes = desc_end.end_offset - desc_end.start_offset;
3394 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3395 pre_sec = pre_ns / nano_seconds_per_second;
3397 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3399 prebuffer = prebuffer_ns / nano_seconds_per_second;
3401 // Set this to 0.0 in case our prebuffer buffers the entire video.
3402 bits_per_second = 0.0;
3404 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3405 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3406 double desc_sec = desc_ns / nano_seconds_per_second;
3407 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3409 // Drop the bps by the percentage of bytes buffered.
3410 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3411 double mod_bits_per_second = calc_bits_per_second * percent;
3413 if (prebuffer < desc_sec) {
3415 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3417 // Add 1 so the bits per second should be a little bit greater than file
3419 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3420 const double min_buffer = 0.0;
3421 double buffer = prebuffer;
3422 double sec_to_download = 0.0;
3424 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3425 min_buffer, &buffer, &sec_to_download,
3429 } else if (rv == 0) {
3430 bits_per_second = (double)(bps);
3435 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3436 } while (desc_end.start_time_ns != -1);
3438 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
3440 return (int64_t)bandwidth;
3443 static int webm_dash_manifest_cues(AVFormatContext *s)
3445 MatroskaDemuxContext *matroska = s->priv_data;
3446 EbmlList *seekhead_list = &matroska->seekhead;
3447 MatroskaSeekhead *seekhead = seekhead_list->elem;
3449 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
3452 // determine cues start and end positions
3453 for (i = 0; i < seekhead_list->nb_elem; i++)
3454 if (seekhead[i].id == MATROSKA_ID_CUES)
3457 if (i >= seekhead_list->nb_elem) return -1;
3459 before_pos = avio_tell(matroska->ctx->pb);
3460 cues_start = seekhead[i].pos + matroska->segment_start;
3461 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
3462 // cues_end is computed as cues_start + cues_length + length of the
3463 // Cues element ID + EBML length of the Cues element. cues_end is
3464 // inclusive and the above sum is reduced by 1.
3465 uint64_t cues_length = 0, cues_id = 0, bytes_read = 0;
3466 bytes_read += ebml_read_num(matroska, matroska->ctx->pb, 4, &cues_id);
3467 bytes_read += ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
3468 cues_end = cues_start + cues_length + bytes_read - 1;
3470 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
3471 if (cues_start == -1 || cues_end == -1) return -1;
3474 matroska_parse_cues(matroska);
3477 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
3480 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
3483 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
3484 if (bandwidth < 0) return -1;
3485 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
3487 // check if all clusters start with key frames
3488 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
3490 // store cue point timestamps as a comma separated list for checking subsegment alignment in
3491 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
3492 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20 * sizeof(char));
3493 if (!buf) return -1;
3495 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
3496 snprintf(buf, (i + 1) * 20 * sizeof(char),
3497 "%s%" PRId64, buf, s->streams[0]->index_entries[i].timestamp);
3498 if (i != s->streams[0]->nb_index_entries - 1)
3499 strncat(buf, ",", sizeof(char));
3501 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
3507 static int webm_dash_manifest_read_header(AVFormatContext *s)
3510 int ret = matroska_read_header(s);
3511 MatroskaTrack *tracks;
3512 MatroskaDemuxContext *matroska = s->priv_data;
3514 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
3518 if (!matroska->is_live) {
3519 buf = av_asprintf("%g", matroska->duration);
3520 if (!buf) return AVERROR(ENOMEM);
3521 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
3524 // initialization range
3525 // 5 is the offset of Cluster ID.
3526 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, avio_tell(s->pb) - 5, 0);
3529 // basename of the file
3530 buf = strrchr(s->filename, '/');
3531 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->filename, 0);
3534 tracks = matroska->tracks.elem;
3535 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
3537 // parse the cues and populate Cue related fields
3538 return matroska->is_live ? 0 : webm_dash_manifest_cues(s);
3541 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
3546 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
3547 static const AVOption options[] = {
3548 { "live", "flag indicating that the input is a live file that only has the headers.", OFFSET(is_live), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_DECODING_PARAM },
3552 static const AVClass webm_dash_class = {
3553 .class_name = "WebM DASH Manifest demuxer",
3554 .item_name = av_default_item_name,
3556 .version = LIBAVUTIL_VERSION_INT,
3559 AVInputFormat ff_matroska_demuxer = {
3560 .name = "matroska,webm",
3561 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
3562 .extensions = "mkv,mk3d,mka,mks",
3563 .priv_data_size = sizeof(MatroskaDemuxContext),
3564 .read_probe = matroska_probe,
3565 .read_header = matroska_read_header,
3566 .read_packet = matroska_read_packet,
3567 .read_close = matroska_read_close,
3568 .read_seek = matroska_read_seek,
3569 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
3572 AVInputFormat ff_webm_dash_manifest_demuxer = {
3573 .name = "webm_dash_manifest",
3574 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
3575 .priv_data_size = sizeof(MatroskaDemuxContext),
3576 .read_header = webm_dash_manifest_read_header,
3577 .read_packet = webm_dash_manifest_read_packet,
3578 .read_close = matroska_read_close,
3579 .priv_class = &webm_dash_class,