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(). */
67 #include "qtpalette.h"
84 typedef const struct EbmlSyntax {
93 const struct EbmlSyntax *n;
97 typedef struct EbmlList {
102 typedef struct EbmlBin {
108 typedef struct Ebml {
113 uint64_t doctype_version;
116 typedef struct MatroskaTrackCompression {
119 } MatroskaTrackCompression;
121 typedef struct MatroskaTrackEncryption {
124 } MatroskaTrackEncryption;
126 typedef struct MatroskaTrackEncoding {
129 MatroskaTrackCompression compression;
130 MatroskaTrackEncryption encryption;
131 } MatroskaTrackEncoding;
133 typedef struct MatroskaTrackVideo {
135 uint64_t display_width;
136 uint64_t display_height;
137 uint64_t pixel_width;
138 uint64_t pixel_height;
140 uint64_t stereo_mode;
142 } MatroskaTrackVideo;
144 typedef struct MatroskaTrackAudio {
146 double out_samplerate;
150 /* real audio header (extracted from extradata) */
157 uint64_t buf_timecode;
159 } MatroskaTrackAudio;
161 typedef struct MatroskaTrackPlane {
164 } MatroskaTrackPlane;
166 typedef struct MatroskaTrackOperation {
167 EbmlList combine_planes;
168 } MatroskaTrackOperation;
170 typedef struct MatroskaTrack {
179 uint64_t default_duration;
180 uint64_t flag_default;
181 uint64_t flag_forced;
182 uint64_t seek_preroll;
183 MatroskaTrackVideo video;
184 MatroskaTrackAudio audio;
185 MatroskaTrackOperation operation;
187 uint64_t codec_delay;
190 int64_t end_timecode;
192 uint64_t max_block_additional_id;
195 typedef struct MatroskaAttachment {
202 } MatroskaAttachment;
204 typedef struct MatroskaChapter {
213 typedef struct MatroskaIndexPos {
218 typedef struct MatroskaIndex {
223 typedef struct MatroskaTag {
231 typedef struct MatroskaTagTarget {
239 typedef struct MatroskaTags {
240 MatroskaTagTarget target;
244 typedef struct MatroskaSeekhead {
249 typedef struct MatroskaLevel {
254 typedef struct MatroskaCluster {
259 typedef struct MatroskaLevel1Element {
263 } MatroskaLevel1Element;
265 typedef struct MatroskaDemuxContext {
266 const AVClass *class;
267 AVFormatContext *ctx;
271 MatroskaLevel levels[EBML_MAX_DEPTH];
281 EbmlList attachments;
287 /* byte position of the segment inside the stream */
288 int64_t segment_start;
290 /* the packet queue */
297 /* What to skip before effectively reading a packet. */
298 int skip_to_keyframe;
299 uint64_t skip_to_timecode;
301 /* File has a CUES element, but we defer parsing until it is needed. */
302 int cues_parsing_deferred;
304 /* Level1 elements and whether they were read yet */
305 MatroskaLevel1Element level1_elems[64];
306 int num_level1_elems;
308 int current_cluster_num_blocks;
309 int64_t current_cluster_pos;
310 MatroskaCluster current_cluster;
312 /* File has SSA subtitles which prevent incremental cluster parsing. */
315 /* WebM DASH Manifest live flag/ */
318 uint32_t palette[AVPALETTE_COUNT];
320 } MatroskaDemuxContext;
322 typedef struct MatroskaBlock {
327 uint64_t additional_id;
329 int64_t discard_padding;
332 static const EbmlSyntax ebml_header[] = {
333 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
334 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
335 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
336 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
337 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
338 { EBML_ID_EBMLVERSION, EBML_NONE },
339 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
343 static const EbmlSyntax ebml_syntax[] = {
344 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
348 static const EbmlSyntax matroska_info[] = {
349 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
350 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
351 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
352 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
353 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
354 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
355 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
359 static const EbmlSyntax matroska_track_video[] = {
360 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
361 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
362 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
363 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
364 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
365 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
366 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
367 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
368 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
369 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
370 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
371 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
372 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_NONE },
373 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
374 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
378 static const EbmlSyntax matroska_track_audio[] = {
379 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
380 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
381 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
382 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
386 static const EbmlSyntax matroska_track_encoding_compression[] = {
387 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
388 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
392 static const EbmlSyntax matroska_track_encoding_encryption[] = {
393 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
394 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
395 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
396 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
397 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
398 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
399 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
402 static const EbmlSyntax matroska_track_encoding[] = {
403 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
404 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
405 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
406 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
407 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
411 static const EbmlSyntax matroska_track_encodings[] = {
412 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
416 static const EbmlSyntax matroska_track_plane[] = {
417 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
418 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
422 static const EbmlSyntax matroska_track_combine_planes[] = {
423 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
427 static const EbmlSyntax matroska_track_operation[] = {
428 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
432 static const EbmlSyntax matroska_track[] = {
433 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
434 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
435 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
436 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
437 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
438 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
439 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
440 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
441 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
442 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
443 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
444 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
445 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
446 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
447 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
448 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
449 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
450 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
451 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
452 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
453 { MATROSKA_ID_CODECNAME, EBML_NONE },
454 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
455 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
456 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
457 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
458 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
462 static const EbmlSyntax matroska_tracks[] = {
463 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
467 static const EbmlSyntax matroska_attachment[] = {
468 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
469 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
470 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
471 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
472 { MATROSKA_ID_FILEDESC, EBML_NONE },
476 static const EbmlSyntax matroska_attachments[] = {
477 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
481 static const EbmlSyntax matroska_chapter_display[] = {
482 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
483 { MATROSKA_ID_CHAPLANG, EBML_NONE },
484 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
488 static const EbmlSyntax matroska_chapter_entry[] = {
489 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
490 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
491 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
492 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
493 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
494 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
495 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
496 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
500 static const EbmlSyntax matroska_chapter[] = {
501 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
502 { MATROSKA_ID_EDITIONUID, EBML_NONE },
503 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
504 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
505 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
509 static const EbmlSyntax matroska_chapters[] = {
510 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
514 static const EbmlSyntax matroska_index_pos[] = {
515 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
516 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
517 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
518 { MATROSKA_ID_CUEDURATION, EBML_NONE },
519 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
523 static const EbmlSyntax matroska_index_entry[] = {
524 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
525 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
529 static const EbmlSyntax matroska_index[] = {
530 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
534 static const EbmlSyntax matroska_simpletag[] = {
535 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
536 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
537 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
538 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
539 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
540 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
544 static const EbmlSyntax matroska_tagtargets[] = {
545 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
546 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
547 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
548 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
549 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
553 static const EbmlSyntax matroska_tag[] = {
554 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
555 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
559 static const EbmlSyntax matroska_tags[] = {
560 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
564 static const EbmlSyntax matroska_seekhead_entry[] = {
565 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
566 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
570 static const EbmlSyntax matroska_seekhead[] = {
571 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
575 static const EbmlSyntax matroska_segment[] = {
576 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
577 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
578 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
579 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
580 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
581 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
582 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
583 { MATROSKA_ID_CLUSTER, EBML_STOP },
587 static const EbmlSyntax matroska_segments[] = {
588 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
592 static const EbmlSyntax matroska_blockmore[] = {
593 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
594 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
598 static const EbmlSyntax matroska_blockadditions[] = {
599 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
603 static const EbmlSyntax matroska_blockgroup[] = {
604 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
605 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
606 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
607 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
608 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
609 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference) },
610 { MATROSKA_ID_CODECSTATE, EBML_NONE },
611 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
615 static const EbmlSyntax matroska_cluster[] = {
616 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
617 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
618 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
619 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
620 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
624 static const EbmlSyntax matroska_clusters[] = {
625 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
626 { MATROSKA_ID_INFO, EBML_NONE },
627 { MATROSKA_ID_CUES, EBML_NONE },
628 { MATROSKA_ID_TAGS, EBML_NONE },
629 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
633 static const EbmlSyntax matroska_cluster_incremental_parsing[] = {
634 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
635 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
636 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
637 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
638 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
639 { MATROSKA_ID_INFO, EBML_NONE },
640 { MATROSKA_ID_CUES, EBML_NONE },
641 { MATROSKA_ID_TAGS, EBML_NONE },
642 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
643 { MATROSKA_ID_CLUSTER, EBML_STOP },
647 static const EbmlSyntax matroska_cluster_incremental[] = {
648 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
649 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
650 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
651 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
652 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
656 static const EbmlSyntax matroska_clusters_incremental[] = {
657 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
658 { MATROSKA_ID_INFO, EBML_NONE },
659 { MATROSKA_ID_CUES, EBML_NONE },
660 { MATROSKA_ID_TAGS, EBML_NONE },
661 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
665 static const char *const matroska_doctypes[] = { "matroska", "webm" };
667 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
669 AVIOContext *pb = matroska->ctx->pb;
671 matroska->current_id = 0;
672 matroska->num_levels = 0;
674 /* seek to next position to resync from */
675 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
680 // try to find a toplevel element
681 while (!avio_feof(pb)) {
682 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
683 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
684 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
685 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
686 matroska->current_id = id;
689 id = (id << 8) | avio_r8(pb);
698 * Return: Whether we reached the end of a level in the hierarchy or not.
700 static int ebml_level_end(MatroskaDemuxContext *matroska)
702 AVIOContext *pb = matroska->ctx->pb;
703 int64_t pos = avio_tell(pb);
705 if (matroska->num_levels > 0) {
706 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
707 if (pos - level->start >= level->length || matroska->current_id) {
708 matroska->num_levels--;
712 return (matroska->is_live && matroska->ctx->pb->eof_reached) ? 1 : 0;
716 * Read: an "EBML number", which is defined as a variable-length
717 * array of bytes. The first byte indicates the length by giving a
718 * number of 0-bits followed by a one. The position of the first
719 * "one" bit inside the first byte indicates the length of this
721 * Returns: number of bytes read, < 0 on error
723 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
724 int max_size, uint64_t *number)
729 /* The first byte tells us the length in bytes - avio_r8() can normally
730 * return 0, but since that's not a valid first ebmlID byte, we can
731 * use it safely here to catch EOS. */
732 if (!(total = avio_r8(pb))) {
733 /* we might encounter EOS here */
734 if (!avio_feof(pb)) {
735 int64_t pos = avio_tell(pb);
736 av_log(matroska->ctx, AV_LOG_ERROR,
737 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
739 return pb->error ? pb->error : AVERROR(EIO);
744 /* get the length of the EBML number */
745 read = 8 - ff_log2_tab[total];
746 if (read > max_size) {
747 int64_t pos = avio_tell(pb) - 1;
748 av_log(matroska->ctx, AV_LOG_ERROR,
749 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
750 (uint8_t) total, pos, pos);
751 return AVERROR_INVALIDDATA;
754 /* read out length */
755 total ^= 1 << ff_log2_tab[total];
757 total = (total << 8) | avio_r8(pb);
765 * Read a EBML length value.
766 * This needs special handling for the "unknown length" case which has multiple
769 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
772 int res = ebml_read_num(matroska, pb, 8, number);
773 if (res > 0 && *number + 1 == 1ULL << (7 * res))
774 *number = 0xffffffffffffffULL;
779 * Read the next element as an unsigned int.
780 * 0 is success, < 0 is failure.
782 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
787 return AVERROR_INVALIDDATA;
789 /* big-endian ordering; build up number */
792 *num = (*num << 8) | avio_r8(pb);
798 * Read the next element as a signed int.
799 * 0 is success, < 0 is failure.
801 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
806 return AVERROR_INVALIDDATA;
811 *num = sign_extend(avio_r8(pb), 8);
813 /* big-endian ordering; build up number */
815 *num = ((uint64_t)*num << 8) | avio_r8(pb);
822 * Read the next element as a float.
823 * 0 is success, < 0 is failure.
825 static int ebml_read_float(AVIOContext *pb, int size, double *num)
830 *num = av_int2float(avio_rb32(pb));
832 *num = av_int2double(avio_rb64(pb));
834 return AVERROR_INVALIDDATA;
840 * Read the next element as an ASCII string.
841 * 0 is success, < 0 is failure.
843 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
847 /* EBML strings are usually not 0-terminated, so we allocate one
848 * byte more, read the string and NULL-terminate it ourselves. */
849 if (!(res = av_malloc(size + 1)))
850 return AVERROR(ENOMEM);
851 if (avio_read(pb, (uint8_t *) res, size) != size) {
863 * Read the next element as binary data.
864 * 0 is success, < 0 is failure.
866 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
868 av_fast_padded_malloc(&bin->data, &bin->size, length);
870 return AVERROR(ENOMEM);
873 bin->pos = avio_tell(pb);
874 if (avio_read(pb, bin->data, length) != length) {
875 av_freep(&bin->data);
884 * Read the next element, but only the header. The contents
885 * are supposed to be sub-elements which can be read separately.
886 * 0 is success, < 0 is failure.
888 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
890 AVIOContext *pb = matroska->ctx->pb;
891 MatroskaLevel *level;
893 if (matroska->num_levels >= EBML_MAX_DEPTH) {
894 av_log(matroska->ctx, AV_LOG_ERROR,
895 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
896 return AVERROR(ENOSYS);
899 level = &matroska->levels[matroska->num_levels++];
900 level->start = avio_tell(pb);
901 level->length = length;
907 * Read signed/unsigned "EBML" numbers.
908 * Return: number of bytes processed, < 0 on error
910 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
911 uint8_t *data, uint32_t size, uint64_t *num)
914 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
915 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
919 * Same as above, but signed.
921 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
922 uint8_t *data, uint32_t size, int64_t *num)
927 /* read as unsigned number first */
928 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
931 /* make signed (weird way) */
932 *num = unum - ((1LL << (7 * res - 1)) - 1);
937 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
938 EbmlSyntax *syntax, void *data);
940 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
941 uint32_t id, void *data)
944 for (i = 0; syntax[i].id; i++)
945 if (id == syntax[i].id)
947 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
948 matroska->num_levels > 0 &&
949 matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
950 return 0; // we reached the end of an unknown size cluster
951 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
952 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32"\n", id);
954 return ebml_parse_elem(matroska, &syntax[i], data);
957 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
960 if (!matroska->current_id) {
962 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
964 // in live mode, finish parsing if EOF is reached.
965 return (matroska->is_live && matroska->ctx->pb->eof_reached &&
966 res == AVERROR_EOF) ? 1 : res;
968 matroska->current_id = id | 1 << 7 * res;
970 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
973 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
978 for (i = 0; syntax[i].id; i++)
979 switch (syntax[i].type) {
981 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
984 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
988 // the default may be NULL
989 if (syntax[i].def.s) {
990 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
991 *dst = av_strdup(syntax[i].def.s);
993 return AVERROR(ENOMEM);
998 while (!res && !ebml_level_end(matroska))
999 res = ebml_parse(matroska, syntax, data);
1004 static int is_ebml_id_valid(uint32_t id)
1006 // Due to endian nonsense in Matroska, the highest byte with any bits set
1007 // will contain the leading length bit. This bit in turn identifies the
1008 // total byte length of the element by its position within the byte.
1009 unsigned int bits = av_log2(id);
1010 return id && (bits + 7) / 8 == (8 - bits % 8);
1014 * Allocate and return the entry for the level1 element with the given ID. If
1015 * an entry already exists, return the existing entry.
1017 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1021 MatroskaLevel1Element *elem;
1023 if (!is_ebml_id_valid(id))
1026 // Some files link to all clusters; useless.
1027 if (id == MATROSKA_ID_CLUSTER)
1030 // There can be multiple seekheads.
1031 if (id != MATROSKA_ID_SEEKHEAD) {
1032 for (i = 0; i < matroska->num_level1_elems; i++) {
1033 if (matroska->level1_elems[i].id == id)
1034 return &matroska->level1_elems[i];
1038 // Only a completely broken file would have more elements.
1039 // It also provides a low-effort way to escape from circular seekheads
1040 // (every iteration will add a level1 entry).
1041 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1042 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1046 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1047 *elem = (MatroskaLevel1Element){.id = id};
1052 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
1053 EbmlSyntax *syntax, void *data)
1055 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1058 // max. 16 MB for strings
1059 [EBML_STR] = 0x1000000,
1060 [EBML_UTF8] = 0x1000000,
1061 // max. 256 MB for binary data
1062 [EBML_BIN] = 0x10000000,
1063 // no limits for anything else
1065 AVIOContext *pb = matroska->ctx->pb;
1066 uint32_t id = syntax->id;
1070 MatroskaLevel1Element *level1_elem;
1072 data = (char *) data + syntax->data_offset;
1073 if (syntax->list_elem_size) {
1074 EbmlList *list = data;
1075 newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1077 return AVERROR(ENOMEM);
1078 list->elem = newelem;
1079 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1080 memset(data, 0, syntax->list_elem_size);
1084 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1085 matroska->current_id = 0;
1086 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1088 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1089 av_log(matroska->ctx, AV_LOG_ERROR,
1090 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1091 length, max_lengths[syntax->type], syntax->type);
1092 return AVERROR_INVALIDDATA;
1096 switch (syntax->type) {
1098 res = ebml_read_uint(pb, length, data);
1101 res = ebml_read_sint(pb, length, data);
1104 res = ebml_read_float(pb, length, data);
1108 res = ebml_read_ascii(pb, length, data);
1111 res = ebml_read_binary(pb, length, data);
1115 if ((res = ebml_read_master(matroska, length)) < 0)
1117 if (id == MATROSKA_ID_SEGMENT)
1118 matroska->segment_start = avio_tell(matroska->ctx->pb);
1119 if (id == MATROSKA_ID_CUES)
1120 matroska->cues_parsing_deferred = 0;
1121 if (syntax->type == EBML_LEVEL1 &&
1122 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1123 if (level1_elem->parsed)
1124 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1125 level1_elem->parsed = 1;
1127 return ebml_parse_nest(matroska, syntax->def.n, data);
1129 return ebml_parse_id(matroska, syntax->def.n, id, data);
1133 if (ffio_limit(pb, length) != length)
1134 return AVERROR(EIO);
1135 return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
1137 if (res == AVERROR_INVALIDDATA)
1138 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1139 else if (res == AVERROR(EIO))
1140 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1144 static void ebml_free(EbmlSyntax *syntax, void *data)
1147 for (i = 0; syntax[i].id; i++) {
1148 void *data_off = (char *) data + syntax[i].data_offset;
1149 switch (syntax[i].type) {
1155 av_freep(&((EbmlBin *) data_off)->data);
1159 if (syntax[i].list_elem_size) {
1160 EbmlList *list = data_off;
1161 char *ptr = list->elem;
1162 for (j = 0; j < list->nb_elem;
1163 j++, ptr += syntax[i].list_elem_size)
1164 ebml_free(syntax[i].def.n, ptr);
1165 av_freep(&list->elem);
1167 ebml_free(syntax[i].def.n, data_off);
1177 static int matroska_probe(AVProbeData *p)
1180 int len_mask = 0x80, size = 1, n = 1, i;
1183 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1186 /* length of header */
1188 while (size <= 8 && !(total & len_mask)) {
1194 total &= (len_mask - 1);
1196 total = (total << 8) | p->buf[4 + n++];
1198 /* Does the probe data contain the whole header? */
1199 if (p->buf_size < 4 + size + total)
1202 /* The header should contain a known document type. For now,
1203 * we don't parse the whole header but simply check for the
1204 * availability of that array of characters inside the header.
1205 * Not fully fool-proof, but good enough. */
1206 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1207 size_t probelen = strlen(matroska_doctypes[i]);
1208 if (total < probelen)
1210 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1211 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1212 return AVPROBE_SCORE_MAX;
1215 // probably valid EBML header but no recognized doctype
1216 return AVPROBE_SCORE_EXTENSION;
1219 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1222 MatroskaTrack *tracks = matroska->tracks.elem;
1225 for (i = 0; i < matroska->tracks.nb_elem; i++)
1226 if (tracks[i].num == num)
1229 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1233 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1234 MatroskaTrack *track)
1236 MatroskaTrackEncoding *encodings = track->encodings.elem;
1237 uint8_t *data = *buf;
1238 int isize = *buf_size;
1239 uint8_t *pkt_data = NULL;
1240 uint8_t av_unused *newpktdata;
1241 int pkt_size = isize;
1245 if (pkt_size >= 10000000U)
1246 return AVERROR_INVALIDDATA;
1248 switch (encodings[0].compression.algo) {
1249 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1251 int header_size = encodings[0].compression.settings.size;
1252 uint8_t *header = encodings[0].compression.settings.data;
1254 if (header_size && !header) {
1255 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1262 pkt_size = isize + header_size;
1263 pkt_data = av_malloc(pkt_size);
1265 return AVERROR(ENOMEM);
1267 memcpy(pkt_data, header, header_size);
1268 memcpy(pkt_data + header_size, data, isize);
1272 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1274 olen = pkt_size *= 3;
1275 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1277 result = AVERROR(ENOMEM);
1280 pkt_data = newpktdata;
1281 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1282 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1284 result = AVERROR_INVALIDDATA;
1291 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1293 z_stream zstream = { 0 };
1294 if (inflateInit(&zstream) != Z_OK)
1296 zstream.next_in = data;
1297 zstream.avail_in = isize;
1300 newpktdata = av_realloc(pkt_data, pkt_size);
1302 inflateEnd(&zstream);
1303 result = AVERROR(ENOMEM);
1306 pkt_data = newpktdata;
1307 zstream.avail_out = pkt_size - zstream.total_out;
1308 zstream.next_out = pkt_data + zstream.total_out;
1309 result = inflate(&zstream, Z_NO_FLUSH);
1310 } while (result == Z_OK && pkt_size < 10000000);
1311 pkt_size = zstream.total_out;
1312 inflateEnd(&zstream);
1313 if (result != Z_STREAM_END) {
1314 if (result == Z_MEM_ERROR)
1315 result = AVERROR(ENOMEM);
1317 result = AVERROR_INVALIDDATA;
1324 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1326 bz_stream bzstream = { 0 };
1327 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1329 bzstream.next_in = data;
1330 bzstream.avail_in = isize;
1333 newpktdata = av_realloc(pkt_data, pkt_size);
1335 BZ2_bzDecompressEnd(&bzstream);
1336 result = AVERROR(ENOMEM);
1339 pkt_data = newpktdata;
1340 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1341 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1342 result = BZ2_bzDecompress(&bzstream);
1343 } while (result == BZ_OK && pkt_size < 10000000);
1344 pkt_size = bzstream.total_out_lo32;
1345 BZ2_bzDecompressEnd(&bzstream);
1346 if (result != BZ_STREAM_END) {
1347 if (result == BZ_MEM_ERROR)
1348 result = AVERROR(ENOMEM);
1350 result = AVERROR_INVALIDDATA;
1357 return AVERROR_INVALIDDATA;
1361 *buf_size = pkt_size;
1369 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1370 AVDictionary **metadata, char *prefix)
1372 MatroskaTag *tags = list->elem;
1376 for (i = 0; i < list->nb_elem; i++) {
1377 const char *lang = tags[i].lang &&
1378 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1380 if (!tags[i].name) {
1381 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1385 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1387 av_strlcpy(key, tags[i].name, sizeof(key));
1388 if (tags[i].def || !lang) {
1389 av_dict_set(metadata, key, tags[i].string, 0);
1390 if (tags[i].sub.nb_elem)
1391 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1394 av_strlcat(key, "-", sizeof(key));
1395 av_strlcat(key, lang, sizeof(key));
1396 av_dict_set(metadata, key, tags[i].string, 0);
1397 if (tags[i].sub.nb_elem)
1398 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1401 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1404 static void matroska_convert_tags(AVFormatContext *s)
1406 MatroskaDemuxContext *matroska = s->priv_data;
1407 MatroskaTags *tags = matroska->tags.elem;
1410 for (i = 0; i < matroska->tags.nb_elem; i++) {
1411 if (tags[i].target.attachuid) {
1412 MatroskaAttachment *attachment = matroska->attachments.elem;
1414 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1415 if (attachment[j].uid == tags[i].target.attachuid &&
1416 attachment[j].stream) {
1417 matroska_convert_tag(s, &tags[i].tag,
1418 &attachment[j].stream->metadata, NULL);
1423 av_log(NULL, AV_LOG_WARNING,
1424 "The tags at index %d refer to a "
1425 "non-existent attachment %"PRId64".\n",
1426 i, tags[i].target.attachuid);
1428 } else if (tags[i].target.chapteruid) {
1429 MatroskaChapter *chapter = matroska->chapters.elem;
1431 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1432 if (chapter[j].uid == tags[i].target.chapteruid &&
1433 chapter[j].chapter) {
1434 matroska_convert_tag(s, &tags[i].tag,
1435 &chapter[j].chapter->metadata, NULL);
1440 av_log(NULL, AV_LOG_WARNING,
1441 "The tags at index %d refer to a non-existent chapter "
1443 i, tags[i].target.chapteruid);
1445 } else if (tags[i].target.trackuid) {
1446 MatroskaTrack *track = matroska->tracks.elem;
1448 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1449 if (track[j].uid == tags[i].target.trackuid &&
1451 matroska_convert_tag(s, &tags[i].tag,
1452 &track[j].stream->metadata, NULL);
1457 av_log(NULL, AV_LOG_WARNING,
1458 "The tags at index %d refer to a non-existent track "
1460 i, tags[i].target.trackuid);
1463 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1464 tags[i].target.type);
1469 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1472 uint32_t level_up = matroska->level_up;
1473 uint32_t saved_id = matroska->current_id;
1474 int64_t before_pos = avio_tell(matroska->ctx->pb);
1475 MatroskaLevel level;
1480 offset = pos + matroska->segment_start;
1481 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1482 /* We don't want to lose our seekhead level, so we add
1483 * a dummy. This is a crude hack. */
1484 if (matroska->num_levels == EBML_MAX_DEPTH) {
1485 av_log(matroska->ctx, AV_LOG_INFO,
1486 "Max EBML element depth (%d) reached, "
1487 "cannot parse further.\n", EBML_MAX_DEPTH);
1488 ret = AVERROR_INVALIDDATA;
1491 level.length = (uint64_t) -1;
1492 matroska->levels[matroska->num_levels] = level;
1493 matroska->num_levels++;
1494 matroska->current_id = 0;
1496 ret = ebml_parse(matroska, matroska_segment, matroska);
1498 /* remove dummy level */
1499 while (matroska->num_levels) {
1500 uint64_t length = matroska->levels[--matroska->num_levels].length;
1501 if (length == (uint64_t) -1)
1507 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1508 matroska->level_up = level_up;
1509 matroska->current_id = saved_id;
1514 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1516 EbmlList *seekhead_list = &matroska->seekhead;
1519 // we should not do any seeking in the streaming case
1520 if (!matroska->ctx->pb->seekable)
1523 for (i = 0; i < seekhead_list->nb_elem; i++) {
1524 MatroskaSeekhead *seekheads = seekhead_list->elem;
1525 uint32_t id = seekheads[i].id;
1526 uint64_t pos = seekheads[i].pos;
1528 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1529 if (!elem || elem->parsed)
1534 // defer cues parsing until we actually need cue data.
1535 if (id == MATROSKA_ID_CUES)
1538 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1539 // mark index as broken
1540 matroska->cues_parsing_deferred = -1;
1548 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1550 EbmlList *index_list;
1551 MatroskaIndex *index;
1552 uint64_t index_scale = 1;
1555 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1558 index_list = &matroska->index;
1559 index = index_list->elem;
1560 if (index_list->nb_elem < 2)
1562 if (index[1].time > 1E14 / matroska->time_scale) {
1563 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1566 for (i = 0; i < index_list->nb_elem; i++) {
1567 EbmlList *pos_list = &index[i].pos;
1568 MatroskaIndexPos *pos = pos_list->elem;
1569 for (j = 0; j < pos_list->nb_elem; j++) {
1570 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1572 if (track && track->stream)
1573 av_add_index_entry(track->stream,
1574 pos[j].pos + matroska->segment_start,
1575 index[i].time / index_scale, 0, 0,
1581 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1584 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1587 for (i = 0; i < matroska->num_level1_elems; i++) {
1588 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1589 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1590 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1591 matroska->cues_parsing_deferred = -1;
1597 matroska_add_index_entries(matroska);
1600 static int matroska_aac_profile(char *codec_id)
1602 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1605 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1606 if (strstr(codec_id, aac_profiles[profile]))
1611 static int matroska_aac_sri(int samplerate)
1615 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1616 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1621 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1624 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1625 time_t creation_time = date_utc / 1000000000 + 978307200;
1626 struct tm tmpbuf, *ptm = gmtime_r(&creation_time, &tmpbuf);
1628 if (strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm))
1629 av_dict_set(metadata, "creation_time", buffer, 0);
1632 static int matroska_parse_flac(AVFormatContext *s,
1633 MatroskaTrack *track,
1636 AVStream *st = track->stream;
1637 uint8_t *p = track->codec_priv.data;
1638 int size = track->codec_priv.size;
1640 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1641 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1642 track->codec_priv.size = 0;
1646 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1648 p += track->codec_priv.size;
1649 size -= track->codec_priv.size;
1651 /* parse the remaining metadata blocks if present */
1653 int block_last, block_type, block_size;
1655 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1659 if (block_size > size)
1662 /* check for the channel mask */
1663 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1664 AVDictionary *dict = NULL;
1665 AVDictionaryEntry *chmask;
1667 ff_vorbis_comment(s, &dict, p, block_size, 0);
1668 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1670 uint64_t mask = strtol(chmask->value, NULL, 0);
1671 if (!mask || mask & ~0x3ffffULL) {
1672 av_log(s, AV_LOG_WARNING,
1673 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1675 st->codec->channel_layout = mask;
1677 av_dict_free(&dict);
1687 static void mkv_stereo_mode_display_mul(int stereo_mode, int *h_width, int *h_height)
1689 switch (stereo_mode) {
1690 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
1691 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
1692 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
1693 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
1694 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
1696 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
1697 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
1698 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
1699 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
1702 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
1703 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
1704 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
1705 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
1711 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
1713 const AVCodecTag *codec_tags;
1715 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
1716 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
1718 /* Normalize noncompliant private data that starts with the fourcc
1719 * by expanding/shifting the data by 4 bytes and storing the data
1720 * size at the start. */
1721 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
1722 uint8_t *p = av_realloc(track->codec_priv.data,
1723 track->codec_priv.size + 4);
1725 return AVERROR(ENOMEM);
1726 memmove(p + 4, p, track->codec_priv.size);
1727 track->codec_priv.data = p;
1728 track->codec_priv.size += 4;
1729 AV_WB32(track->codec_priv.data, track->codec_priv.size);
1732 *fourcc = AV_RL32(track->codec_priv.data + 4);
1733 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
1738 static int matroska_parse_tracks(AVFormatContext *s)
1740 MatroskaDemuxContext *matroska = s->priv_data;
1741 MatroskaTrack *tracks = matroska->tracks.elem;
1746 for (i = 0; i < matroska->tracks.nb_elem; i++) {
1747 MatroskaTrack *track = &tracks[i];
1748 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1749 EbmlList *encodings_list = &track->encodings;
1750 MatroskaTrackEncoding *encodings = encodings_list->elem;
1751 uint8_t *extradata = NULL;
1752 int extradata_size = 0;
1753 int extradata_offset = 0;
1754 uint32_t fourcc = 0;
1756 char* key_id_base64 = NULL;
1759 /* Apply some sanity checks. */
1760 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1761 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1762 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1763 track->type != MATROSKA_TRACK_TYPE_METADATA) {
1764 av_log(matroska->ctx, AV_LOG_INFO,
1765 "Unknown or unsupported track type %"PRIu64"\n",
1769 if (!track->codec_id)
1772 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
1773 isnan(track->audio.samplerate)) {
1774 av_log(matroska->ctx, AV_LOG_WARNING,
1775 "Invalid sample rate %f, defaulting to 8000 instead.\n",
1776 track->audio.samplerate);
1777 track->audio.samplerate = 8000;
1780 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1781 if (!track->default_duration && track->video.frame_rate > 0)
1782 track->default_duration = 1000000000 / track->video.frame_rate;
1783 if (track->video.display_width == -1)
1784 track->video.display_width = track->video.pixel_width;
1785 if (track->video.display_height == -1)
1786 track->video.display_height = track->video.pixel_height;
1787 if (track->video.color_space.size == 4)
1788 fourcc = AV_RL32(track->video.color_space.data);
1789 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1790 if (!track->audio.out_samplerate)
1791 track->audio.out_samplerate = track->audio.samplerate;
1793 if (encodings_list->nb_elem > 1) {
1794 av_log(matroska->ctx, AV_LOG_ERROR,
1795 "Multiple combined encodings not supported");
1796 } else if (encodings_list->nb_elem == 1) {
1797 if (encodings[0].type) {
1798 if (encodings[0].encryption.key_id.size > 0) {
1799 /* Save the encryption key id to be stored later as a
1801 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1802 key_id_base64 = av_malloc(b64_size);
1803 if (key_id_base64 == NULL)
1804 return AVERROR(ENOMEM);
1806 av_base64_encode(key_id_base64, b64_size,
1807 encodings[0].encryption.key_id.data,
1808 encodings[0].encryption.key_id.size);
1810 encodings[0].scope = 0;
1811 av_log(matroska->ctx, AV_LOG_ERROR,
1812 "Unsupported encoding type");
1816 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1819 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1822 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1824 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
1825 encodings[0].scope = 0;
1826 av_log(matroska->ctx, AV_LOG_ERROR,
1827 "Unsupported encoding type");
1828 } else if (track->codec_priv.size && encodings[0].scope & 2) {
1829 uint8_t *codec_priv = track->codec_priv.data;
1830 int ret = matroska_decode_buffer(&track->codec_priv.data,
1831 &track->codec_priv.size,
1834 track->codec_priv.data = NULL;
1835 track->codec_priv.size = 0;
1836 av_log(matroska->ctx, AV_LOG_ERROR,
1837 "Failed to decode codec private data\n");
1840 if (codec_priv != track->codec_priv.data)
1841 av_free(codec_priv);
1845 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
1846 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
1847 strlen(ff_mkv_codec_tags[j].str))) {
1848 codec_id = ff_mkv_codec_tags[j].id;
1853 st = track->stream = avformat_new_stream(s, NULL);
1855 av_free(key_id_base64);
1856 return AVERROR(ENOMEM);
1859 if (key_id_base64) {
1860 /* export encryption key id as base64 metadata tag */
1861 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
1862 av_freep(&key_id_base64);
1865 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
1866 track->codec_priv.size >= 40 &&
1867 track->codec_priv.data) {
1868 track->ms_compat = 1;
1869 bit_depth = AV_RL16(track->codec_priv.data + 14);
1870 fourcc = AV_RL32(track->codec_priv.data + 16);
1871 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
1874 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
1876 extradata_offset = 40;
1877 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
1878 track->codec_priv.size >= 14 &&
1879 track->codec_priv.data) {
1881 ffio_init_context(&b, track->codec_priv.data,
1882 track->codec_priv.size,
1883 0, NULL, NULL, NULL, NULL);
1884 ret = ff_get_wav_header(s, &b, st->codec, track->codec_priv.size, 0);
1887 codec_id = st->codec->codec_id;
1888 fourcc = st->codec->codec_tag;
1889 extradata_offset = FFMIN(track->codec_priv.size, 18);
1890 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
1891 /* Normally 36, but allow noncompliant private data */
1892 && (track->codec_priv.size >= 32)
1893 && (track->codec_priv.data)) {
1894 uint16_t sample_size;
1895 int ret = get_qt_codec(track, &fourcc, &codec_id);
1898 sample_size = AV_RB16(track->codec_priv.data + 26);
1900 if (sample_size == 8) {
1901 fourcc = MKTAG('r','a','w',' ');
1902 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1903 } else if (sample_size == 16) {
1904 fourcc = MKTAG('t','w','o','s');
1905 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
1908 if ((fourcc == MKTAG('t','w','o','s') ||
1909 fourcc == MKTAG('s','o','w','t')) &&
1911 codec_id = AV_CODEC_ID_PCM_S8;
1912 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
1913 (track->codec_priv.size >= 21) &&
1914 (track->codec_priv.data)) {
1915 int ret = get_qt_codec(track, &fourcc, &codec_id);
1918 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
1919 fourcc = MKTAG('S','V','Q','3');
1920 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
1922 if (codec_id == AV_CODEC_ID_NONE) {
1924 av_get_codec_tag_string(buf, sizeof(buf), fourcc);
1925 av_log(matroska->ctx, AV_LOG_ERROR,
1926 "mov FourCC not found %s.\n", buf);
1928 if (track->codec_priv.size >= 86) {
1929 bit_depth = AV_RB16(track->codec_priv.data + 82);
1930 ffio_init_context(&b, track->codec_priv.data,
1931 track->codec_priv.size,
1932 0, NULL, NULL, NULL, NULL);
1933 if (ff_get_qtpalette(codec_id, &b, matroska->palette)) {
1935 matroska->has_palette = 1;
1938 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
1939 switch (track->audio.bitdepth) {
1941 codec_id = AV_CODEC_ID_PCM_U8;
1944 codec_id = AV_CODEC_ID_PCM_S24BE;
1947 codec_id = AV_CODEC_ID_PCM_S32BE;
1950 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
1951 switch (track->audio.bitdepth) {
1953 codec_id = AV_CODEC_ID_PCM_U8;
1956 codec_id = AV_CODEC_ID_PCM_S24LE;
1959 codec_id = AV_CODEC_ID_PCM_S32LE;
1962 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
1963 track->audio.bitdepth == 64) {
1964 codec_id = AV_CODEC_ID_PCM_F64LE;
1965 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
1966 int profile = matroska_aac_profile(track->codec_id);
1967 int sri = matroska_aac_sri(track->audio.samplerate);
1968 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
1970 return AVERROR(ENOMEM);
1971 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
1972 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
1973 if (strstr(track->codec_id, "SBR")) {
1974 sri = matroska_aac_sri(track->audio.out_samplerate);
1975 extradata[2] = 0x56;
1976 extradata[3] = 0xE5;
1977 extradata[4] = 0x80 | (sri << 3);
1981 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
1982 /* Only ALAC's magic cookie is stored in Matroska's track headers.
1983 * Create the "atom size", "tag", and "tag version" fields the
1984 * decoder expects manually. */
1985 extradata_size = 12 + track->codec_priv.size;
1986 extradata = av_mallocz(extradata_size +
1987 AV_INPUT_BUFFER_PADDING_SIZE);
1989 return AVERROR(ENOMEM);
1990 AV_WB32(extradata, extradata_size);
1991 memcpy(&extradata[4], "alac", 4);
1992 AV_WB32(&extradata[8], 0);
1993 memcpy(&extradata[12], track->codec_priv.data,
1994 track->codec_priv.size);
1995 } else if (codec_id == AV_CODEC_ID_TTA) {
1996 extradata_size = 30;
1997 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
1999 return AVERROR(ENOMEM);
2000 ffio_init_context(&b, extradata, extradata_size, 1,
2001 NULL, NULL, NULL, NULL);
2002 avio_write(&b, "TTA1", 4);
2004 if (track->audio.channels > UINT16_MAX ||
2005 track->audio.bitdepth > UINT16_MAX) {
2006 av_log(matroska->ctx, AV_LOG_WARNING,
2007 "Too large audio channel number %"PRIu64
2008 " or bitdepth %"PRIu64". Skipping track.\n",
2009 track->audio.channels, track->audio.bitdepth);
2010 av_freep(&extradata);
2011 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2012 return AVERROR_INVALIDDATA;
2016 avio_wl16(&b, track->audio.channels);
2017 avio_wl16(&b, track->audio.bitdepth);
2018 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2019 return AVERROR_INVALIDDATA;
2020 avio_wl32(&b, track->audio.out_samplerate);
2021 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
2022 track->audio.out_samplerate,
2023 AV_TIME_BASE * 1000));
2024 } else if (codec_id == AV_CODEC_ID_RV10 ||
2025 codec_id == AV_CODEC_ID_RV20 ||
2026 codec_id == AV_CODEC_ID_RV30 ||
2027 codec_id == AV_CODEC_ID_RV40) {
2028 extradata_offset = 26;
2029 } else if (codec_id == AV_CODEC_ID_RA_144) {
2030 track->audio.out_samplerate = 8000;
2031 track->audio.channels = 1;
2032 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2033 codec_id == AV_CODEC_ID_COOK ||
2034 codec_id == AV_CODEC_ID_ATRAC3 ||
2035 codec_id == AV_CODEC_ID_SIPR)
2036 && track->codec_priv.data) {
2039 ffio_init_context(&b, track->codec_priv.data,
2040 track->codec_priv.size,
2041 0, NULL, NULL, NULL, NULL);
2043 flavor = avio_rb16(&b);
2044 track->audio.coded_framesize = avio_rb32(&b);
2046 track->audio.sub_packet_h = avio_rb16(&b);
2047 track->audio.frame_size = avio_rb16(&b);
2048 track->audio.sub_packet_size = avio_rb16(&b);
2050 track->audio.coded_framesize <= 0 ||
2051 track->audio.sub_packet_h <= 0 ||
2052 track->audio.frame_size <= 0 ||
2053 track->audio.sub_packet_size <= 0)
2054 return AVERROR_INVALIDDATA;
2055 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2056 track->audio.frame_size);
2057 if (!track->audio.buf)
2058 return AVERROR(ENOMEM);
2059 if (codec_id == AV_CODEC_ID_RA_288) {
2060 st->codec->block_align = track->audio.coded_framesize;
2061 track->codec_priv.size = 0;
2063 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2064 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2065 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2066 st->codec->bit_rate = sipr_bit_rate[flavor];
2068 st->codec->block_align = track->audio.sub_packet_size;
2069 extradata_offset = 78;
2071 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2072 ret = matroska_parse_flac(s, track, &extradata_offset);
2075 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2076 fourcc = AV_RL32(track->codec_priv.data);
2078 track->codec_priv.size -= extradata_offset;
2080 if (codec_id == AV_CODEC_ID_NONE)
2081 av_log(matroska->ctx, AV_LOG_INFO,
2082 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2084 if (track->time_scale < 0.01)
2085 track->time_scale = 1.0;
2086 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2087 1000 * 1000 * 1000); /* 64 bit pts in ns */
2089 /* convert the delay from ns to the track timebase */
2090 track->codec_delay = av_rescale_q(track->codec_delay,
2091 (AVRational){ 1, 1000000000 },
2094 st->codec->codec_id = codec_id;
2096 if (strcmp(track->language, "und"))
2097 av_dict_set(&st->metadata, "language", track->language, 0);
2098 av_dict_set(&st->metadata, "title", track->name, 0);
2100 if (track->flag_default)
2101 st->disposition |= AV_DISPOSITION_DEFAULT;
2102 if (track->flag_forced)
2103 st->disposition |= AV_DISPOSITION_FORCED;
2105 if (!st->codec->extradata) {
2107 st->codec->extradata = extradata;
2108 st->codec->extradata_size = extradata_size;
2109 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2110 if (ff_alloc_extradata(st->codec, track->codec_priv.size))
2111 return AVERROR(ENOMEM);
2112 memcpy(st->codec->extradata,
2113 track->codec_priv.data + extradata_offset,
2114 track->codec_priv.size);
2118 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2119 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2120 int display_width_mul = 1;
2121 int display_height_mul = 1;
2123 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
2124 st->codec->codec_tag = fourcc;
2126 st->codec->bits_per_coded_sample = bit_depth;
2127 st->codec->width = track->video.pixel_width;
2128 st->codec->height = track->video.pixel_height;
2130 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2131 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2133 av_reduce(&st->sample_aspect_ratio.num,
2134 &st->sample_aspect_ratio.den,
2135 st->codec->height * track->video.display_width * display_width_mul,
2136 st->codec->width * track->video.display_height * display_height_mul,
2138 if (st->codec->codec_id != AV_CODEC_ID_HEVC)
2139 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2141 if (track->default_duration) {
2142 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2143 1000000000, track->default_duration, 30000);
2144 #if FF_API_R_FRAME_RATE
2145 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2146 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2147 st->r_frame_rate = st->avg_frame_rate;
2151 /* export stereo mode flag as metadata tag */
2152 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2153 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2155 /* export alpha mode flag as metadata tag */
2156 if (track->video.alpha_mode)
2157 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2159 /* if we have virtual track, mark the real tracks */
2160 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2162 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2164 snprintf(buf, sizeof(buf), "%s_%d",
2165 ff_matroska_video_stereo_plane[planes[j].type], i);
2166 for (k=0; k < matroska->tracks.nb_elem; k++)
2167 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2168 av_dict_set(&tracks[k].stream->metadata,
2169 "stereo_mode", buf, 0);
2173 // add stream level stereo3d side data if it is a supported format
2174 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2175 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2176 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2180 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2181 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
2182 st->codec->codec_tag = fourcc;
2183 st->codec->sample_rate = track->audio.out_samplerate;
2184 st->codec->channels = track->audio.channels;
2185 if (!st->codec->bits_per_coded_sample)
2186 st->codec->bits_per_coded_sample = track->audio.bitdepth;
2187 if (st->codec->codec_id == AV_CODEC_ID_MP3)
2188 st->need_parsing = AVSTREAM_PARSE_FULL;
2189 else if (st->codec->codec_id != AV_CODEC_ID_AAC)
2190 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2191 if (track->codec_delay > 0) {
2192 st->codec->delay = av_rescale_q(track->codec_delay,
2194 (AVRational){1, st->codec->sample_rate});
2196 if (track->seek_preroll > 0) {
2197 av_codec_set_seek_preroll(st->codec,
2198 av_rescale_q(track->seek_preroll,
2199 (AVRational){1, 1000000000},
2200 (AVRational){1, st->codec->sample_rate}));
2202 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2203 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
2205 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2206 st->disposition |= AV_DISPOSITION_CAPTIONS;
2207 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2208 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2209 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2210 st->disposition |= AV_DISPOSITION_METADATA;
2212 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2213 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
2214 if (st->codec->codec_id == AV_CODEC_ID_ASS)
2215 matroska->contains_ssa = 1;
2222 static int matroska_read_header(AVFormatContext *s)
2224 MatroskaDemuxContext *matroska = s->priv_data;
2225 EbmlList *attachments_list = &matroska->attachments;
2226 EbmlList *chapters_list = &matroska->chapters;
2227 MatroskaAttachment *attachments;
2228 MatroskaChapter *chapters;
2229 uint64_t max_start = 0;
2235 matroska->cues_parsing_deferred = 1;
2237 /* First read the EBML header. */
2238 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2239 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2240 ebml_free(ebml_syntax, &ebml);
2241 return AVERROR_INVALIDDATA;
2243 if (ebml.version > EBML_VERSION ||
2244 ebml.max_size > sizeof(uint64_t) ||
2245 ebml.id_length > sizeof(uint32_t) ||
2246 ebml.doctype_version > 3) {
2247 av_log(matroska->ctx, AV_LOG_ERROR,
2248 "EBML header using unsupported features\n"
2249 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2250 ebml.version, ebml.doctype, ebml.doctype_version);
2251 ebml_free(ebml_syntax, &ebml);
2252 return AVERROR_PATCHWELCOME;
2253 } else if (ebml.doctype_version == 3) {
2254 av_log(matroska->ctx, AV_LOG_WARNING,
2255 "EBML header using unsupported features\n"
2256 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2257 ebml.version, ebml.doctype, ebml.doctype_version);
2259 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2260 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2262 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2263 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2264 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2265 ebml_free(ebml_syntax, &ebml);
2266 return AVERROR_INVALIDDATA;
2269 ebml_free(ebml_syntax, &ebml);
2271 /* The next thing is a segment. */
2272 pos = avio_tell(matroska->ctx->pb);
2273 res = ebml_parse(matroska, matroska_segments, matroska);
2274 // try resyncing until we find a EBML_STOP type element.
2276 res = matroska_resync(matroska, pos);
2279 pos = avio_tell(matroska->ctx->pb);
2280 res = ebml_parse(matroska, matroska_segment, matroska);
2282 matroska_execute_seekhead(matroska);
2284 if (!matroska->time_scale)
2285 matroska->time_scale = 1000000;
2286 if (matroska->duration)
2287 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2288 1000 / AV_TIME_BASE;
2289 av_dict_set(&s->metadata, "title", matroska->title, 0);
2290 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2292 if (matroska->date_utc.size == 8)
2293 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2295 res = matroska_parse_tracks(s);
2299 attachments = attachments_list->elem;
2300 for (j = 0; j < attachments_list->nb_elem; j++) {
2301 if (!(attachments[j].filename && attachments[j].mime &&
2302 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2303 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2305 AVStream *st = avformat_new_stream(s, NULL);
2308 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2309 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2310 st->codec->codec_id = AV_CODEC_ID_NONE;
2312 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2313 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2314 strlen(ff_mkv_image_mime_tags[i].str))) {
2315 st->codec->codec_id = ff_mkv_image_mime_tags[i].id;
2320 attachments[j].stream = st;
2322 if (st->codec->codec_id != AV_CODEC_ID_NONE) {
2323 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2324 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
2326 av_init_packet(&st->attached_pic);
2327 if ((res = av_new_packet(&st->attached_pic, attachments[j].bin.size)) < 0)
2329 memcpy(st->attached_pic.data, attachments[j].bin.data, attachments[j].bin.size);
2330 st->attached_pic.stream_index = st->index;
2331 st->attached_pic.flags |= AV_PKT_FLAG_KEY;
2333 st->codec->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2334 if (ff_alloc_extradata(st->codec, attachments[j].bin.size))
2336 memcpy(st->codec->extradata, attachments[j].bin.data,
2337 attachments[j].bin.size);
2339 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2340 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2341 strlen(ff_mkv_mime_tags[i].str))) {
2342 st->codec->codec_id = ff_mkv_mime_tags[i].id;
2350 chapters = chapters_list->elem;
2351 for (i = 0; i < chapters_list->nb_elem; i++)
2352 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2353 (max_start == 0 || chapters[i].start > max_start)) {
2354 chapters[i].chapter =
2355 avpriv_new_chapter(s, chapters[i].uid,
2356 (AVRational) { 1, 1000000000 },
2357 chapters[i].start, chapters[i].end,
2359 if (chapters[i].chapter) {
2360 av_dict_set(&chapters[i].chapter->metadata,
2361 "title", chapters[i].title, 0);
2363 max_start = chapters[i].start;
2366 matroska_add_index_entries(matroska);
2368 matroska_convert_tags(s);
2374 * Put one packet in an application-supplied AVPacket struct.
2375 * Returns 0 on success or -1 on failure.
2377 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2380 if (matroska->num_packets > 0) {
2381 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2382 av_freep(&matroska->packets[0]);
2383 if (matroska->has_palette) {
2384 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2386 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
2388 memcpy(pal, matroska->palette, AVPALETTE_SIZE);
2390 matroska->has_palette = 0;
2392 if (matroska->num_packets > 1) {
2394 memmove(&matroska->packets[0], &matroska->packets[1],
2395 (matroska->num_packets - 1) * sizeof(AVPacket *));
2396 newpackets = av_realloc(matroska->packets,
2397 (matroska->num_packets - 1) *
2398 sizeof(AVPacket *));
2400 matroska->packets = newpackets;
2402 av_freep(&matroska->packets);
2403 matroska->prev_pkt = NULL;
2405 matroska->num_packets--;
2413 * Free all packets in our internal queue.
2415 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2417 matroska->prev_pkt = NULL;
2418 if (matroska->packets) {
2420 for (n = 0; n < matroska->num_packets; n++) {
2421 av_packet_unref(matroska->packets[n]);
2422 av_freep(&matroska->packets[n]);
2424 av_freep(&matroska->packets);
2425 matroska->num_packets = 0;
2429 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2430 int *buf_size, int type,
2431 uint32_t **lace_buf, int *laces)
2433 int res = 0, n, size = *buf_size;
2434 uint8_t *data = *buf;
2435 uint32_t *lace_size;
2439 *lace_buf = av_mallocz(sizeof(int));
2441 return AVERROR(ENOMEM);
2443 *lace_buf[0] = size;
2447 av_assert0(size > 0);
2451 lace_size = av_mallocz(*laces * sizeof(int));
2453 return AVERROR(ENOMEM);
2456 case 0x1: /* Xiph lacing */
2460 for (n = 0; res == 0 && n < *laces - 1; n++) {
2462 if (size <= total) {
2463 res = AVERROR_INVALIDDATA;
2468 lace_size[n] += temp;
2475 if (size <= total) {
2476 res = AVERROR_INVALIDDATA;
2480 lace_size[n] = size - total;
2484 case 0x2: /* fixed-size lacing */
2485 if (size % (*laces)) {
2486 res = AVERROR_INVALIDDATA;
2489 for (n = 0; n < *laces; n++)
2490 lace_size[n] = size / *laces;
2493 case 0x3: /* EBML lacing */
2497 n = matroska_ebmlnum_uint(matroska, data, size, &num);
2498 if (n < 0 || num > INT_MAX) {
2499 av_log(matroska->ctx, AV_LOG_INFO,
2500 "EBML block data error\n");
2501 res = n<0 ? n : AVERROR_INVALIDDATA;
2506 total = lace_size[0] = num;
2507 for (n = 1; res == 0 && n < *laces - 1; n++) {
2510 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2511 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2512 av_log(matroska->ctx, AV_LOG_INFO,
2513 "EBML block data error\n");
2514 res = r<0 ? r : AVERROR_INVALIDDATA;
2519 lace_size[n] = lace_size[n - 1] + snum;
2520 total += lace_size[n];
2522 if (size <= total) {
2523 res = AVERROR_INVALIDDATA;
2526 lace_size[*laces - 1] = size - total;
2532 *lace_buf = lace_size;
2538 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2539 MatroskaTrack *track, AVStream *st,
2540 uint8_t *data, int size, uint64_t timecode,
2543 int a = st->codec->block_align;
2544 int sps = track->audio.sub_packet_size;
2545 int cfs = track->audio.coded_framesize;
2546 int h = track->audio.sub_packet_h;
2547 int y = track->audio.sub_packet_cnt;
2548 int w = track->audio.frame_size;
2551 if (!track->audio.pkt_cnt) {
2552 if (track->audio.sub_packet_cnt == 0)
2553 track->audio.buf_timecode = timecode;
2554 if (st->codec->codec_id == AV_CODEC_ID_RA_288) {
2555 if (size < cfs * h / 2) {
2556 av_log(matroska->ctx, AV_LOG_ERROR,
2557 "Corrupt int4 RM-style audio packet size\n");
2558 return AVERROR_INVALIDDATA;
2560 for (x = 0; x < h / 2; x++)
2561 memcpy(track->audio.buf + x * 2 * w + y * cfs,
2562 data + x * cfs, cfs);
2563 } else if (st->codec->codec_id == AV_CODEC_ID_SIPR) {
2565 av_log(matroska->ctx, AV_LOG_ERROR,
2566 "Corrupt sipr RM-style audio packet size\n");
2567 return AVERROR_INVALIDDATA;
2569 memcpy(track->audio.buf + y * w, data, w);
2571 if (size < sps * w / sps || h<=0 || w%sps) {
2572 av_log(matroska->ctx, AV_LOG_ERROR,
2573 "Corrupt generic RM-style audio packet size\n");
2574 return AVERROR_INVALIDDATA;
2576 for (x = 0; x < w / sps; x++)
2577 memcpy(track->audio.buf +
2578 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2579 data + x * sps, sps);
2582 if (++track->audio.sub_packet_cnt >= h) {
2583 if (st->codec->codec_id == AV_CODEC_ID_SIPR)
2584 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2585 track->audio.sub_packet_cnt = 0;
2586 track->audio.pkt_cnt = h * w / a;
2590 while (track->audio.pkt_cnt) {
2592 AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2594 return AVERROR(ENOMEM);
2596 ret = av_new_packet(pkt, a);
2602 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2604 pkt->pts = track->audio.buf_timecode;
2605 track->audio.buf_timecode = AV_NOPTS_VALUE;
2607 pkt->stream_index = st->index;
2608 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2614 /* reconstruct full wavpack blocks from mangled matroska ones */
2615 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
2616 uint8_t **pdst, int *size)
2618 uint8_t *dst = NULL;
2623 int ret, offset = 0;
2625 if (srclen < 12 || track->stream->codec->extradata_size < 2)
2626 return AVERROR_INVALIDDATA;
2628 ver = AV_RL16(track->stream->codec->extradata);
2630 samples = AV_RL32(src);
2634 while (srclen >= 8) {
2639 uint32_t flags = AV_RL32(src);
2640 uint32_t crc = AV_RL32(src + 4);
2644 multiblock = (flags & 0x1800) != 0x1800;
2647 ret = AVERROR_INVALIDDATA;
2650 blocksize = AV_RL32(src);
2656 if (blocksize > srclen) {
2657 ret = AVERROR_INVALIDDATA;
2661 tmp = av_realloc(dst, dstlen + blocksize + 32);
2663 ret = AVERROR(ENOMEM);
2667 dstlen += blocksize + 32;
2669 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2670 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2671 AV_WL16(dst + offset + 8, ver); // version
2672 AV_WL16(dst + offset + 10, 0); // track/index_no
2673 AV_WL32(dst + offset + 12, 0); // total samples
2674 AV_WL32(dst + offset + 16, 0); // block index
2675 AV_WL32(dst + offset + 20, samples); // number of samples
2676 AV_WL32(dst + offset + 24, flags); // flags
2677 AV_WL32(dst + offset + 28, crc); // crc
2678 memcpy(dst + offset + 32, src, blocksize); // block data
2681 srclen -= blocksize;
2682 offset += blocksize + 32;
2695 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
2696 MatroskaTrack *track,
2698 uint8_t *data, int data_len,
2704 uint8_t *id, *settings, *text, *buf;
2705 int id_len, settings_len, text_len;
2710 return AVERROR_INVALIDDATA;
2713 q = data + data_len;
2718 if (*p == '\r' || *p == '\n') {
2727 if (p >= q || *p != '\n')
2728 return AVERROR_INVALIDDATA;
2734 if (*p == '\r' || *p == '\n') {
2735 settings_len = p - settings;
2743 if (p >= q || *p != '\n')
2744 return AVERROR_INVALIDDATA;
2749 while (text_len > 0) {
2750 const int len = text_len - 1;
2751 const uint8_t c = p[len];
2752 if (c != '\r' && c != '\n')
2758 return AVERROR_INVALIDDATA;
2760 pkt = av_mallocz(sizeof(*pkt));
2762 return AVERROR(ENOMEM);
2763 err = av_new_packet(pkt, text_len);
2766 return AVERROR(err);
2769 memcpy(pkt->data, text, text_len);
2772 buf = av_packet_new_side_data(pkt,
2773 AV_PKT_DATA_WEBVTT_IDENTIFIER,
2777 return AVERROR(ENOMEM);
2779 memcpy(buf, id, id_len);
2782 if (settings_len > 0) {
2783 buf = av_packet_new_side_data(pkt,
2784 AV_PKT_DATA_WEBVTT_SETTINGS,
2788 return AVERROR(ENOMEM);
2790 memcpy(buf, settings, settings_len);
2793 // Do we need this for subtitles?
2794 // pkt->flags = AV_PKT_FLAG_KEY;
2796 pkt->stream_index = st->index;
2797 pkt->pts = timecode;
2799 // Do we need this for subtitles?
2800 // pkt->dts = timecode;
2802 pkt->duration = duration;
2805 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2806 matroska->prev_pkt = pkt;
2811 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2812 MatroskaTrack *track, AVStream *st,
2813 uint8_t *data, int pkt_size,
2814 uint64_t timecode, uint64_t lace_duration,
2815 int64_t pos, int is_keyframe,
2816 uint8_t *additional, uint64_t additional_id, int additional_size,
2817 int64_t discard_padding)
2819 MatroskaTrackEncoding *encodings = track->encodings.elem;
2820 uint8_t *pkt_data = data;
2821 int offset = 0, res;
2824 if (encodings && !encodings->type && encodings->scope & 1) {
2825 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
2830 if (st->codec->codec_id == AV_CODEC_ID_WAVPACK) {
2832 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
2834 av_log(matroska->ctx, AV_LOG_ERROR,
2835 "Error parsing a wavpack block.\n");
2838 if (pkt_data != data)
2839 av_freep(&pkt_data);
2843 if (st->codec->codec_id == AV_CODEC_ID_PRORES &&
2844 AV_RB32(&data[4]) != MKBETAG('i', 'c', 'p', 'f'))
2847 pkt = av_mallocz(sizeof(AVPacket));
2849 if (pkt_data != data)
2850 av_freep(&pkt_data);
2851 return AVERROR(ENOMEM);
2853 /* XXX: prevent data copy... */
2854 if (av_new_packet(pkt, pkt_size + offset) < 0) {
2856 res = AVERROR(ENOMEM);
2860 if (st->codec->codec_id == AV_CODEC_ID_PRORES && offset == 8) {
2861 uint8_t *buf = pkt->data;
2862 bytestream_put_be32(&buf, pkt_size);
2863 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
2866 memcpy(pkt->data + offset, pkt_data, pkt_size);
2868 if (pkt_data != data)
2869 av_freep(&pkt_data);
2871 pkt->flags = is_keyframe;
2872 pkt->stream_index = st->index;
2874 if (additional_size > 0) {
2875 uint8_t *side_data = av_packet_new_side_data(pkt,
2876 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
2877 additional_size + 8);
2879 av_packet_unref(pkt);
2881 return AVERROR(ENOMEM);
2883 AV_WB64(side_data, additional_id);
2884 memcpy(side_data + 8, additional, additional_size);
2887 if (discard_padding) {
2888 uint8_t *side_data = av_packet_new_side_data(pkt,
2889 AV_PKT_DATA_SKIP_SAMPLES,
2892 av_packet_unref(pkt);
2894 return AVERROR(ENOMEM);
2896 AV_WL32(side_data, 0);
2897 AV_WL32(side_data + 4, av_rescale_q(discard_padding,
2898 (AVRational){1, 1000000000},
2899 (AVRational){1, st->codec->sample_rate}));
2902 if (track->ms_compat)
2903 pkt->dts = timecode;
2905 pkt->pts = timecode;
2907 pkt->duration = lace_duration;
2909 #if FF_API_CONVERGENCE_DURATION
2910 FF_DISABLE_DEPRECATION_WARNINGS
2911 if (st->codec->codec_id == AV_CODEC_ID_SUBRIP) {
2912 pkt->convergence_duration = lace_duration;
2914 FF_ENABLE_DEPRECATION_WARNINGS
2917 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2918 matroska->prev_pkt = pkt;
2923 if (pkt_data != data)
2924 av_freep(&pkt_data);
2928 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
2929 int size, int64_t pos, uint64_t cluster_time,
2930 uint64_t block_duration, int is_keyframe,
2931 uint8_t *additional, uint64_t additional_id, int additional_size,
2932 int64_t cluster_pos, int64_t discard_padding)
2934 uint64_t timecode = AV_NOPTS_VALUE;
2935 MatroskaTrack *track;
2939 uint32_t *lace_size = NULL;
2940 int n, flags, laces = 0;
2942 int trust_default_duration = 1;
2944 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
2945 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
2951 track = matroska_find_track_by_num(matroska, num);
2952 if (!track || !track->stream) {
2953 av_log(matroska->ctx, AV_LOG_INFO,
2954 "Invalid stream %"PRIu64" or size %u\n", num, size);
2955 return AVERROR_INVALIDDATA;
2956 } else if (size <= 3)
2959 if (st->discard >= AVDISCARD_ALL)
2961 av_assert1(block_duration != AV_NOPTS_VALUE);
2963 block_time = sign_extend(AV_RB16(data), 16);
2967 if (is_keyframe == -1)
2968 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
2970 if (cluster_time != (uint64_t) -1 &&
2971 (block_time >= 0 || cluster_time >= -block_time)) {
2972 timecode = cluster_time + block_time - track->codec_delay;
2973 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
2974 timecode < track->end_timecode)
2975 is_keyframe = 0; /* overlapping subtitles are not key frame */
2977 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
2981 if (matroska->skip_to_keyframe &&
2982 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
2983 if (timecode < matroska->skip_to_timecode)
2986 matroska->skip_to_keyframe = 0;
2987 else if (!st->skip_to_keyframe) {
2988 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
2989 matroska->skip_to_keyframe = 0;
2993 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
2994 &lace_size, &laces);
2999 if (track->audio.samplerate == 8000) {
3000 // If this is needed for more codecs, then add them here
3001 if (st->codec->codec_id == AV_CODEC_ID_AC3) {
3002 if (track->audio.samplerate != st->codec->sample_rate || !st->codec->frame_size)
3003 trust_default_duration = 0;
3007 if (!block_duration && trust_default_duration)
3008 block_duration = track->default_duration * laces / matroska->time_scale;
3010 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3011 track->end_timecode =
3012 FFMAX(track->end_timecode, timecode + block_duration);
3014 for (n = 0; n < laces; n++) {
3015 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3017 if (lace_size[n] > size) {
3018 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
3022 if ((st->codec->codec_id == AV_CODEC_ID_RA_288 ||
3023 st->codec->codec_id == AV_CODEC_ID_COOK ||
3024 st->codec->codec_id == AV_CODEC_ID_SIPR ||
3025 st->codec->codec_id == AV_CODEC_ID_ATRAC3) &&
3026 st->codec->block_align && track->audio.sub_packet_size) {
3027 res = matroska_parse_rm_audio(matroska, track, st, data,
3033 } else if (st->codec->codec_id == AV_CODEC_ID_WEBVTT) {
3034 res = matroska_parse_webvtt(matroska, track, st,
3036 timecode, lace_duration,
3041 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
3042 timecode, lace_duration, pos,
3043 !n ? is_keyframe : 0,
3044 additional, additional_id, additional_size,
3050 if (timecode != AV_NOPTS_VALUE)
3051 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3052 data += lace_size[n];
3053 size -= lace_size[n];
3061 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
3063 EbmlList *blocks_list;
3064 MatroskaBlock *blocks;
3066 res = ebml_parse(matroska,
3067 matroska_cluster_incremental_parsing,
3068 &matroska->current_cluster);
3071 if (matroska->current_cluster_pos)
3072 ebml_level_end(matroska);
3073 ebml_free(matroska_cluster, &matroska->current_cluster);
3074 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
3075 matroska->current_cluster_num_blocks = 0;
3076 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
3077 matroska->prev_pkt = NULL;
3078 /* sizeof the ID which was already read */
3079 if (matroska->current_id)
3080 matroska->current_cluster_pos -= 4;
3081 res = ebml_parse(matroska,
3082 matroska_clusters_incremental,
3083 &matroska->current_cluster);
3084 /* Try parsing the block again. */
3086 res = ebml_parse(matroska,
3087 matroska_cluster_incremental_parsing,
3088 &matroska->current_cluster);
3092 matroska->current_cluster_num_blocks <
3093 matroska->current_cluster.blocks.nb_elem) {
3094 blocks_list = &matroska->current_cluster.blocks;
3095 blocks = blocks_list->elem;
3097 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
3098 i = blocks_list->nb_elem - 1;
3099 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
3100 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
3101 uint8_t* additional = blocks[i].additional.size > 0 ?
3102 blocks[i].additional.data : NULL;
3103 if (!blocks[i].non_simple)
3104 blocks[i].duration = 0;
3105 res = matroska_parse_block(matroska, blocks[i].bin.data,
3106 blocks[i].bin.size, blocks[i].bin.pos,
3107 matroska->current_cluster.timecode,
3108 blocks[i].duration, is_keyframe,
3109 additional, blocks[i].additional_id,
3110 blocks[i].additional.size,
3111 matroska->current_cluster_pos,
3112 blocks[i].discard_padding);
3119 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3121 MatroskaCluster cluster = { 0 };
3122 EbmlList *blocks_list;
3123 MatroskaBlock *blocks;
3127 if (!matroska->contains_ssa)
3128 return matroska_parse_cluster_incremental(matroska);
3129 pos = avio_tell(matroska->ctx->pb);
3130 matroska->prev_pkt = NULL;
3131 if (matroska->current_id)
3132 pos -= 4; /* sizeof the ID which was already read */
3133 res = ebml_parse(matroska, matroska_clusters, &cluster);
3134 blocks_list = &cluster.blocks;
3135 blocks = blocks_list->elem;
3136 for (i = 0; i < blocks_list->nb_elem; i++)
3137 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
3138 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
3139 res = matroska_parse_block(matroska, blocks[i].bin.data,
3140 blocks[i].bin.size, blocks[i].bin.pos,
3141 cluster.timecode, blocks[i].duration,
3142 is_keyframe, NULL, 0, 0, pos,
3143 blocks[i].discard_padding);
3145 ebml_free(matroska_cluster, &cluster);
3149 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3151 MatroskaDemuxContext *matroska = s->priv_data;
3153 while (matroska_deliver_packet(matroska, pkt)) {
3154 int64_t pos = avio_tell(matroska->ctx->pb);
3157 if (matroska_parse_cluster(matroska) < 0)
3158 matroska_resync(matroska, pos);
3164 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3165 int64_t timestamp, int flags)
3167 MatroskaDemuxContext *matroska = s->priv_data;
3168 MatroskaTrack *tracks = NULL;
3169 AVStream *st = s->streams[stream_index];
3170 int i, index, index_sub, index_min;
3172 /* Parse the CUES now since we need the index data to seek. */
3173 if (matroska->cues_parsing_deferred > 0) {
3174 matroska->cues_parsing_deferred = 0;
3175 matroska_parse_cues(matroska);
3178 if (!st->nb_index_entries)
3180 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3182 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3183 avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
3185 matroska->current_id = 0;
3186 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3187 matroska_clear_queue(matroska);
3188 if (matroska_parse_cluster(matroska) < 0)
3193 matroska_clear_queue(matroska);
3194 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3198 tracks = matroska->tracks.elem;
3199 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3200 tracks[i].audio.pkt_cnt = 0;
3201 tracks[i].audio.sub_packet_cnt = 0;
3202 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3203 tracks[i].end_timecode = 0;
3204 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3206 tracks[i].stream->discard != AVDISCARD_ALL) {
3207 index_sub = av_index_search_timestamp(
3208 tracks[i].stream, st->index_entries[index].timestamp,
3209 AVSEEK_FLAG_BACKWARD);
3210 while (index_sub >= 0 &&
3212 tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
3213 st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
3218 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
3219 matroska->current_id = 0;
3220 if (flags & AVSEEK_FLAG_ANY) {
3221 st->skip_to_keyframe = 0;
3222 matroska->skip_to_timecode = timestamp;
3224 st->skip_to_keyframe = 1;
3225 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3227 matroska->skip_to_keyframe = 1;
3229 matroska->num_levels = 0;
3230 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3233 // slightly hackish but allows proper fallback to
3234 // the generic seeking code.
3235 matroska_clear_queue(matroska);
3236 matroska->current_id = 0;
3237 st->skip_to_keyframe =
3238 matroska->skip_to_keyframe = 0;
3240 matroska->num_levels = 0;
3244 static int matroska_read_close(AVFormatContext *s)
3246 MatroskaDemuxContext *matroska = s->priv_data;
3247 MatroskaTrack *tracks = matroska->tracks.elem;
3250 matroska_clear_queue(matroska);
3252 for (n = 0; n < matroska->tracks.nb_elem; n++)
3253 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3254 av_freep(&tracks[n].audio.buf);
3255 ebml_free(matroska_cluster, &matroska->current_cluster);
3256 ebml_free(matroska_segment, matroska);
3262 int64_t start_time_ns;
3263 int64_t end_time_ns;
3264 int64_t start_offset;
3268 /* This function searches all the Cues and returns the CueDesc corresponding the
3269 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3270 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3272 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3273 MatroskaDemuxContext *matroska = s->priv_data;
3276 int nb_index_entries = s->streams[0]->nb_index_entries;
3277 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3278 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3279 for (i = 1; i < nb_index_entries; i++) {
3280 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3281 index_entries[i].timestamp * matroska->time_scale > ts) {
3286 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3287 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3288 if (i != nb_index_entries - 1) {
3289 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3290 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3292 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3293 // FIXME: this needs special handling for files where Cues appear
3294 // before Clusters. the current logic assumes Cues appear after
3296 cue_desc.end_offset = cues_start - matroska->segment_start;
3301 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3303 MatroskaDemuxContext *matroska = s->priv_data;
3304 int64_t cluster_pos, before_pos;
3306 if (s->streams[0]->nb_index_entries <= 0) return 0;
3307 // seek to the first cluster using cues.
3308 index = av_index_search_timestamp(s->streams[0], 0, 0);
3309 if (index < 0) return 0;
3310 cluster_pos = s->streams[0]->index_entries[index].pos;
3311 before_pos = avio_tell(s->pb);
3313 int64_t cluster_id = 0, cluster_length = 0;
3315 avio_seek(s->pb, cluster_pos, SEEK_SET);
3316 // read cluster id and length
3317 ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id);
3318 ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3319 if (cluster_id != 0xF43B675) { // done with all clusters
3322 avio_seek(s->pb, cluster_pos, SEEK_SET);
3323 matroska->current_id = 0;
3324 matroska_clear_queue(matroska);
3325 if (matroska_parse_cluster(matroska) < 0 ||
3326 matroska->num_packets <= 0) {
3329 pkt = matroska->packets[0];
3330 cluster_pos += cluster_length + 12; // 12 is the offset of the cluster id and length.
3331 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3336 avio_seek(s->pb, before_pos, SEEK_SET);
3340 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3341 double min_buffer, double* buffer,
3342 double* sec_to_download, AVFormatContext *s,
3345 double nano_seconds_per_second = 1000000000.0;
3346 double time_sec = time_ns / nano_seconds_per_second;
3348 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3349 int64_t end_time_ns = time_ns + time_to_search_ns;
3350 double sec_downloaded = 0.0;
3351 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3352 if (desc_curr.start_time_ns == -1)
3354 *sec_to_download = 0.0;
3356 // Check for non cue start time.
3357 if (time_ns > desc_curr.start_time_ns) {
3358 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3359 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3360 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3361 double timeToDownload = (cueBytes * 8.0) / bps;
3363 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3364 *sec_to_download += timeToDownload;
3366 // Check if the search ends within the first cue.
3367 if (desc_curr.end_time_ns >= end_time_ns) {
3368 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3369 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3370 sec_downloaded = percent_to_sub * sec_downloaded;
3371 *sec_to_download = percent_to_sub * *sec_to_download;
3374 if ((sec_downloaded + *buffer) <= min_buffer) {
3378 // Get the next Cue.
3379 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3382 while (desc_curr.start_time_ns != -1) {
3383 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3384 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3385 double desc_sec = desc_ns / nano_seconds_per_second;
3386 double bits = (desc_bytes * 8.0);
3387 double time_to_download = bits / bps;
3389 sec_downloaded += desc_sec - time_to_download;
3390 *sec_to_download += time_to_download;
3392 if (desc_curr.end_time_ns >= end_time_ns) {
3393 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3394 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3395 sec_downloaded = percent_to_sub * sec_downloaded;
3396 *sec_to_download = percent_to_sub * *sec_to_download;
3398 if ((sec_downloaded + *buffer) <= min_buffer)
3403 if ((sec_downloaded + *buffer) <= min_buffer) {
3408 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3410 *buffer = *buffer + sec_downloaded;
3414 /* This function computes the bandwidth of the WebM file with the help of
3415 * buffer_size_after_time_downloaded() function. Both of these functions are
3416 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3417 * Matroska parsing mechanism.
3419 * Returns the bandwidth of the file on success; -1 on error.
3421 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3423 MatroskaDemuxContext *matroska = s->priv_data;
3424 AVStream *st = s->streams[0];
3425 double bandwidth = 0.0;
3428 for (i = 0; i < st->nb_index_entries; i++) {
3429 int64_t prebuffer_ns = 1000000000;
3430 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3431 double nano_seconds_per_second = 1000000000.0;
3432 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3433 double prebuffer_bytes = 0.0;
3434 int64_t temp_prebuffer_ns = prebuffer_ns;
3435 int64_t pre_bytes, pre_ns;
3436 double pre_sec, prebuffer, bits_per_second;
3437 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3439 // Start with the first Cue.
3440 CueDesc desc_end = desc_beg;
3442 // Figure out how much data we have downloaded for the prebuffer. This will
3443 // be used later to adjust the bits per sample to try.
3444 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3445 // Prebuffered the entire Cue.
3446 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3447 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3448 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3450 if (desc_end.start_time_ns == -1) {
3451 // The prebuffer is larger than the duration.
3452 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3454 bits_per_second = 0.0;
3456 // The prebuffer ends in the last Cue. Estimate how much data was
3458 pre_bytes = desc_end.end_offset - desc_end.start_offset;
3459 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3460 pre_sec = pre_ns / nano_seconds_per_second;
3462 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3464 prebuffer = prebuffer_ns / nano_seconds_per_second;
3466 // Set this to 0.0 in case our prebuffer buffers the entire video.
3467 bits_per_second = 0.0;
3469 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3470 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3471 double desc_sec = desc_ns / nano_seconds_per_second;
3472 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3474 // Drop the bps by the percentage of bytes buffered.
3475 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3476 double mod_bits_per_second = calc_bits_per_second * percent;
3478 if (prebuffer < desc_sec) {
3480 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3482 // Add 1 so the bits per second should be a little bit greater than file
3484 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3485 const double min_buffer = 0.0;
3486 double buffer = prebuffer;
3487 double sec_to_download = 0.0;
3489 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3490 min_buffer, &buffer, &sec_to_download,
3494 } else if (rv == 0) {
3495 bits_per_second = (double)(bps);
3500 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3501 } while (desc_end.start_time_ns != -1);
3503 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
3505 return (int64_t)bandwidth;
3508 static int webm_dash_manifest_cues(AVFormatContext *s)
3510 MatroskaDemuxContext *matroska = s->priv_data;
3511 EbmlList *seekhead_list = &matroska->seekhead;
3512 MatroskaSeekhead *seekhead = seekhead_list->elem;
3514 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
3517 // determine cues start and end positions
3518 for (i = 0; i < seekhead_list->nb_elem; i++)
3519 if (seekhead[i].id == MATROSKA_ID_CUES)
3522 if (i >= seekhead_list->nb_elem) return -1;
3524 before_pos = avio_tell(matroska->ctx->pb);
3525 cues_start = seekhead[i].pos + matroska->segment_start;
3526 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
3527 // cues_end is computed as cues_start + cues_length + length of the
3528 // Cues element ID + EBML length of the Cues element. cues_end is
3529 // inclusive and the above sum is reduced by 1.
3530 uint64_t cues_length = 0, cues_id = 0, bytes_read = 0;
3531 bytes_read += ebml_read_num(matroska, matroska->ctx->pb, 4, &cues_id);
3532 bytes_read += ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
3533 cues_end = cues_start + cues_length + bytes_read - 1;
3535 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
3536 if (cues_start == -1 || cues_end == -1) return -1;
3539 matroska_parse_cues(matroska);
3542 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
3545 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
3548 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
3549 if (bandwidth < 0) return -1;
3550 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
3552 // check if all clusters start with key frames
3553 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
3555 // store cue point timestamps as a comma separated list for checking subsegment alignment in
3556 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
3557 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20 * sizeof(char));
3558 if (!buf) return -1;
3560 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
3561 snprintf(buf, (i + 1) * 20 * sizeof(char),
3562 "%s%" PRId64, buf, s->streams[0]->index_entries[i].timestamp);
3563 if (i != s->streams[0]->nb_index_entries - 1)
3564 strncat(buf, ",", sizeof(char));
3566 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
3572 static int webm_dash_manifest_read_header(AVFormatContext *s)
3575 int ret = matroska_read_header(s);
3576 MatroskaTrack *tracks;
3577 MatroskaDemuxContext *matroska = s->priv_data;
3579 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
3583 if (!matroska->is_live) {
3584 buf = av_asprintf("%g", matroska->duration);
3585 if (!buf) return AVERROR(ENOMEM);
3586 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
3589 // initialization range
3590 // 5 is the offset of Cluster ID.
3591 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, avio_tell(s->pb) - 5, 0);
3594 // basename of the file
3595 buf = strrchr(s->filename, '/');
3596 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->filename, 0);
3599 tracks = matroska->tracks.elem;
3600 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
3602 // parse the cues and populate Cue related fields
3603 return matroska->is_live ? 0 : webm_dash_manifest_cues(s);
3606 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
3611 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
3612 static const AVOption options[] = {
3613 { "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 },
3617 static const AVClass webm_dash_class = {
3618 .class_name = "WebM DASH Manifest demuxer",
3619 .item_name = av_default_item_name,
3621 .version = LIBAVUTIL_VERSION_INT,
3624 AVInputFormat ff_matroska_demuxer = {
3625 .name = "matroska,webm",
3626 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
3627 .extensions = "mkv,mk3d,mka,mks",
3628 .priv_data_size = sizeof(MatroskaDemuxContext),
3629 .read_probe = matroska_probe,
3630 .read_header = matroska_read_header,
3631 .read_packet = matroska_read_packet,
3632 .read_close = matroska_read_close,
3633 .read_seek = matroska_read_seek,
3634 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
3637 AVInputFormat ff_webm_dash_manifest_demuxer = {
3638 .name = "webm_dash_manifest",
3639 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
3640 .priv_data_size = sizeof(MatroskaDemuxContext),
3641 .read_header = webm_dash_manifest_read_header,
3642 .read_packet = webm_dash_manifest_read_packet,
3643 .read_close = matroska_read_close,
3644 .priv_class = &webm_dash_class,