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/mastering_display_metadata.h"
43 #include "libavutil/mathematics.h"
44 #include "libavutil/opt.h"
45 #include "libavutil/time_internal.h"
47 #include "libavcodec/bytestream.h"
48 #include "libavcodec/flac.h"
49 #include "libavcodec/mpeg4audio.h"
52 #include "avio_internal.h"
57 /* For ff_codec_get_id(). */
68 #include "qtpalette.h"
85 typedef const struct EbmlSyntax {
94 const struct EbmlSyntax *n;
98 typedef struct EbmlList {
103 typedef struct EbmlBin {
109 typedef struct Ebml {
114 uint64_t doctype_version;
117 typedef struct MatroskaTrackCompression {
120 } MatroskaTrackCompression;
122 typedef struct MatroskaTrackEncryption {
125 } MatroskaTrackEncryption;
127 typedef struct MatroskaTrackEncoding {
130 MatroskaTrackCompression compression;
131 MatroskaTrackEncryption encryption;
132 } MatroskaTrackEncoding;
134 typedef struct MatroskaMasteringMeta {
143 double max_luminance;
144 double min_luminance;
145 } MatroskaMasteringMeta;
147 typedef struct MatroskaTrackVideoColor {
148 uint64_t matrix_coefficients;
149 uint64_t bits_per_channel;
150 uint64_t chroma_sub_horz;
151 uint64_t chroma_sub_vert;
152 uint64_t cb_sub_horz;
153 uint64_t cb_sub_vert;
154 uint64_t chroma_siting_horz;
155 uint64_t chroma_siting_vert;
157 uint64_t transfer_characteristics;
161 MatroskaMasteringMeta mastering_meta;
162 } MatroskaTrackVideoColor;
164 typedef struct MatroskaTrackVideo {
166 uint64_t display_width;
167 uint64_t display_height;
168 uint64_t pixel_width;
169 uint64_t pixel_height;
172 uint64_t field_order;
173 uint64_t stereo_mode;
175 MatroskaTrackVideoColor color;
176 } MatroskaTrackVideo;
178 typedef struct MatroskaTrackAudio {
180 double out_samplerate;
184 /* real audio header (extracted from extradata) */
191 uint64_t buf_timecode;
193 } MatroskaTrackAudio;
195 typedef struct MatroskaTrackPlane {
198 } MatroskaTrackPlane;
200 typedef struct MatroskaTrackOperation {
201 EbmlList combine_planes;
202 } MatroskaTrackOperation;
204 typedef struct MatroskaTrack {
213 uint64_t default_duration;
214 uint64_t flag_default;
215 uint64_t flag_forced;
216 uint64_t seek_preroll;
217 MatroskaTrackVideo video;
218 MatroskaTrackAudio audio;
219 MatroskaTrackOperation operation;
221 uint64_t codec_delay;
222 uint64_t codec_delay_in_track_tb;
225 int64_t end_timecode;
227 uint64_t max_block_additional_id;
229 uint32_t palette[AVPALETTE_COUNT];
233 typedef struct MatroskaAttachment {
240 } MatroskaAttachment;
242 typedef struct MatroskaChapter {
251 typedef struct MatroskaIndexPos {
256 typedef struct MatroskaIndex {
261 typedef struct MatroskaTag {
269 typedef struct MatroskaTagTarget {
277 typedef struct MatroskaTags {
278 MatroskaTagTarget target;
282 typedef struct MatroskaSeekhead {
287 typedef struct MatroskaLevel {
292 typedef struct MatroskaCluster {
297 typedef struct MatroskaLevel1Element {
301 } MatroskaLevel1Element;
303 typedef struct MatroskaDemuxContext {
304 const AVClass *class;
305 AVFormatContext *ctx;
309 MatroskaLevel levels[EBML_MAX_DEPTH];
319 EbmlList attachments;
325 /* byte position of the segment inside the stream */
326 int64_t segment_start;
328 /* the packet queue */
335 /* What to skip before effectively reading a packet. */
336 int skip_to_keyframe;
337 uint64_t skip_to_timecode;
339 /* File has a CUES element, but we defer parsing until it is needed. */
340 int cues_parsing_deferred;
342 /* Level1 elements and whether they were read yet */
343 MatroskaLevel1Element level1_elems[64];
344 int num_level1_elems;
346 int current_cluster_num_blocks;
347 int64_t current_cluster_pos;
348 MatroskaCluster current_cluster;
350 /* File has SSA subtitles which prevent incremental cluster parsing. */
353 /* WebM DASH Manifest live flag/ */
355 } MatroskaDemuxContext;
357 typedef struct MatroskaBlock {
362 uint64_t additional_id;
364 int64_t discard_padding;
367 static const EbmlSyntax ebml_header[] = {
368 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
369 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
370 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
371 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
372 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
373 { EBML_ID_EBMLVERSION, EBML_NONE },
374 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
378 static const EbmlSyntax ebml_syntax[] = {
379 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
383 static const EbmlSyntax matroska_info[] = {
384 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
385 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
386 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
387 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
388 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
389 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
390 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
394 static const EbmlSyntax matroska_mastering_meta[] = {
395 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
396 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
397 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
398 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
399 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
400 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
401 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
402 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
403 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
404 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
408 static const EbmlSyntax matroska_track_video_color[] = {
409 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u=2 } },
410 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=8 } },
411 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
412 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
413 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
414 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
415 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u=0 } },
416 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u=0 } },
417 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, range), { .u=0 } },
418 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u=2 } },
419 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u=2 } },
420 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
421 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
422 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
426 static const EbmlSyntax matroska_track_video[] = {
427 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
428 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
429 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
430 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
431 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
432 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
433 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
434 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, 0, offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
435 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
436 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
437 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
438 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
439 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_NONE },
440 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
441 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
442 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
443 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
447 static const EbmlSyntax matroska_track_audio[] = {
448 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
449 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
450 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
451 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
455 static const EbmlSyntax matroska_track_encoding_compression[] = {
456 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
457 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
461 static const EbmlSyntax matroska_track_encoding_encryption[] = {
462 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
463 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
464 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
465 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
466 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
467 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
468 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
471 static const EbmlSyntax matroska_track_encoding[] = {
472 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
473 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
474 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
475 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
476 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
480 static const EbmlSyntax matroska_track_encodings[] = {
481 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
485 static const EbmlSyntax matroska_track_plane[] = {
486 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
487 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
491 static const EbmlSyntax matroska_track_combine_planes[] = {
492 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
496 static const EbmlSyntax matroska_track_operation[] = {
497 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
501 static const EbmlSyntax matroska_track[] = {
502 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
503 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
504 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
505 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
506 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
507 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
508 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
509 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
510 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
511 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
512 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
513 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
514 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
515 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
516 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
517 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
518 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
519 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
520 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
521 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
522 { MATROSKA_ID_CODECNAME, EBML_NONE },
523 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
524 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
525 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
526 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
527 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
531 static const EbmlSyntax matroska_tracks[] = {
532 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
536 static const EbmlSyntax matroska_attachment[] = {
537 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
538 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
539 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
540 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
541 { MATROSKA_ID_FILEDESC, EBML_NONE },
545 static const EbmlSyntax matroska_attachments[] = {
546 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
550 static const EbmlSyntax matroska_chapter_display[] = {
551 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
552 { MATROSKA_ID_CHAPLANG, EBML_NONE },
553 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
557 static const EbmlSyntax matroska_chapter_entry[] = {
558 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
559 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
560 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
561 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
562 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
563 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
564 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
565 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
569 static const EbmlSyntax matroska_chapter[] = {
570 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
571 { MATROSKA_ID_EDITIONUID, EBML_NONE },
572 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
573 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
574 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
578 static const EbmlSyntax matroska_chapters[] = {
579 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
583 static const EbmlSyntax matroska_index_pos[] = {
584 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
585 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
586 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
587 { MATROSKA_ID_CUEDURATION, EBML_NONE },
588 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
592 static const EbmlSyntax matroska_index_entry[] = {
593 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
594 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
598 static const EbmlSyntax matroska_index[] = {
599 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
603 static const EbmlSyntax matroska_simpletag[] = {
604 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
605 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
606 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
607 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
608 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
609 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
613 static const EbmlSyntax matroska_tagtargets[] = {
614 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
615 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
616 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
617 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
618 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
622 static const EbmlSyntax matroska_tag[] = {
623 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
624 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
628 static const EbmlSyntax matroska_tags[] = {
629 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
633 static const EbmlSyntax matroska_seekhead_entry[] = {
634 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
635 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
639 static const EbmlSyntax matroska_seekhead[] = {
640 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
644 static const EbmlSyntax matroska_segment[] = {
645 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
646 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
647 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
648 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
649 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
650 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
651 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
652 { MATROSKA_ID_CLUSTER, EBML_STOP },
656 static const EbmlSyntax matroska_segments[] = {
657 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
661 static const EbmlSyntax matroska_blockmore[] = {
662 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
663 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
667 static const EbmlSyntax matroska_blockadditions[] = {
668 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
672 static const EbmlSyntax matroska_blockgroup[] = {
673 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
674 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
675 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
676 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
677 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
678 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference) },
679 { MATROSKA_ID_CODECSTATE, EBML_NONE },
680 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
684 static const EbmlSyntax matroska_cluster[] = {
685 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
686 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
687 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
688 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
689 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
693 static const EbmlSyntax matroska_clusters[] = {
694 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
695 { MATROSKA_ID_INFO, EBML_NONE },
696 { MATROSKA_ID_CUES, EBML_NONE },
697 { MATROSKA_ID_TAGS, EBML_NONE },
698 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
702 static const EbmlSyntax matroska_cluster_incremental_parsing[] = {
703 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
704 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
705 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
706 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
707 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
708 { MATROSKA_ID_INFO, EBML_NONE },
709 { MATROSKA_ID_CUES, EBML_NONE },
710 { MATROSKA_ID_TAGS, EBML_NONE },
711 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
712 { MATROSKA_ID_CLUSTER, EBML_STOP },
716 static const EbmlSyntax matroska_cluster_incremental[] = {
717 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
718 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
719 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
720 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
721 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
725 static const EbmlSyntax matroska_clusters_incremental[] = {
726 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
727 { MATROSKA_ID_INFO, EBML_NONE },
728 { MATROSKA_ID_CUES, EBML_NONE },
729 { MATROSKA_ID_TAGS, EBML_NONE },
730 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
734 static const char *const matroska_doctypes[] = { "matroska", "webm" };
736 static int matroska_read_close(AVFormatContext *s);
738 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
740 AVIOContext *pb = matroska->ctx->pb;
742 matroska->current_id = 0;
743 matroska->num_levels = 0;
745 /* seek to next position to resync from */
746 if (avio_seek(pb, last_pos + 1, SEEK_SET) < 0)
751 // try to find a toplevel element
752 while (!avio_feof(pb)) {
753 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
754 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
755 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
756 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
757 matroska->current_id = id;
760 id = (id << 8) | avio_r8(pb);
769 * Return: Whether we reached the end of a level in the hierarchy or not.
771 static int ebml_level_end(MatroskaDemuxContext *matroska)
773 AVIOContext *pb = matroska->ctx->pb;
774 int64_t pos = avio_tell(pb);
776 if (matroska->num_levels > 0) {
777 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
778 if (pos - level->start >= level->length || matroska->current_id) {
779 matroska->num_levels--;
783 return (matroska->is_live && matroska->ctx->pb->eof_reached) ? 1 : 0;
787 * Read: an "EBML number", which is defined as a variable-length
788 * array of bytes. The first byte indicates the length by giving a
789 * number of 0-bits followed by a one. The position of the first
790 * "one" bit inside the first byte indicates the length of this
792 * Returns: number of bytes read, < 0 on error
794 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
795 int max_size, uint64_t *number)
800 /* The first byte tells us the length in bytes - avio_r8() can normally
801 * return 0, but since that's not a valid first ebmlID byte, we can
802 * use it safely here to catch EOS. */
803 if (!(total = avio_r8(pb))) {
804 /* we might encounter EOS here */
805 if (!avio_feof(pb)) {
806 int64_t pos = avio_tell(pb);
807 av_log(matroska->ctx, AV_LOG_ERROR,
808 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
810 return pb->error ? pb->error : AVERROR(EIO);
815 /* get the length of the EBML number */
816 read = 8 - ff_log2_tab[total];
817 if (read > max_size) {
818 int64_t pos = avio_tell(pb) - 1;
819 av_log(matroska->ctx, AV_LOG_ERROR,
820 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
821 (uint8_t) total, pos, pos);
822 return AVERROR_INVALIDDATA;
825 /* read out length */
826 total ^= 1 << ff_log2_tab[total];
828 total = (total << 8) | avio_r8(pb);
836 * Read a EBML length value.
837 * This needs special handling for the "unknown length" case which has multiple
840 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
843 int res = ebml_read_num(matroska, pb, 8, number);
844 if (res > 0 && *number + 1 == 1ULL << (7 * res))
845 *number = 0xffffffffffffffULL;
850 * Read the next element as an unsigned int.
851 * 0 is success, < 0 is failure.
853 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
858 return AVERROR_INVALIDDATA;
860 /* big-endian ordering; build up number */
863 *num = (*num << 8) | avio_r8(pb);
869 * Read the next element as a signed int.
870 * 0 is success, < 0 is failure.
872 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
877 return AVERROR_INVALIDDATA;
882 *num = sign_extend(avio_r8(pb), 8);
884 /* big-endian ordering; build up number */
886 *num = ((uint64_t)*num << 8) | avio_r8(pb);
893 * Read the next element as a float.
894 * 0 is success, < 0 is failure.
896 static int ebml_read_float(AVIOContext *pb, int size, double *num)
901 *num = av_int2float(avio_rb32(pb));
903 *num = av_int2double(avio_rb64(pb));
905 return AVERROR_INVALIDDATA;
911 * Read the next element as an ASCII string.
912 * 0 is success, < 0 is failure.
914 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
918 /* EBML strings are usually not 0-terminated, so we allocate one
919 * byte more, read the string and NULL-terminate it ourselves. */
920 if (!(res = av_malloc(size + 1)))
921 return AVERROR(ENOMEM);
922 if (avio_read(pb, (uint8_t *) res, size) != size) {
934 * Read the next element as binary data.
935 * 0 is success, < 0 is failure.
937 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
939 av_fast_padded_malloc(&bin->data, &bin->size, length);
941 return AVERROR(ENOMEM);
944 bin->pos = avio_tell(pb);
945 if (avio_read(pb, bin->data, length) != length) {
946 av_freep(&bin->data);
955 * Read the next element, but only the header. The contents
956 * are supposed to be sub-elements which can be read separately.
957 * 0 is success, < 0 is failure.
959 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
961 AVIOContext *pb = matroska->ctx->pb;
962 MatroskaLevel *level;
964 if (matroska->num_levels >= EBML_MAX_DEPTH) {
965 av_log(matroska->ctx, AV_LOG_ERROR,
966 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
967 return AVERROR(ENOSYS);
970 level = &matroska->levels[matroska->num_levels++];
971 level->start = avio_tell(pb);
972 level->length = length;
978 * Read signed/unsigned "EBML" numbers.
979 * Return: number of bytes processed, < 0 on error
981 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
982 uint8_t *data, uint32_t size, uint64_t *num)
985 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
986 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
990 * Same as above, but signed.
992 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
993 uint8_t *data, uint32_t size, int64_t *num)
998 /* read as unsigned number first */
999 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
1002 /* make signed (weird way) */
1003 *num = unum - ((1LL << (7 * res - 1)) - 1);
1008 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
1009 EbmlSyntax *syntax, void *data);
1011 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1012 uint32_t id, void *data)
1015 for (i = 0; syntax[i].id; i++)
1016 if (id == syntax[i].id)
1018 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
1019 matroska->num_levels > 0 &&
1020 matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
1021 return 0; // we reached the end of an unknown size cluster
1022 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1023 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32"\n", id);
1025 return ebml_parse_elem(matroska, &syntax[i], data);
1028 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1031 if (!matroska->current_id) {
1033 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
1035 // in live mode, finish parsing if EOF is reached.
1036 return (matroska->is_live && matroska->ctx->pb->eof_reached &&
1037 res == AVERROR_EOF) ? 1 : res;
1039 matroska->current_id = id | 1 << 7 * res;
1041 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
1044 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1049 for (i = 0; syntax[i].id; i++)
1050 switch (syntax[i].type) {
1052 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1055 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1059 // the default may be NULL
1060 if (syntax[i].def.s) {
1061 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1062 *dst = av_strdup(syntax[i].def.s);
1064 return AVERROR(ENOMEM);
1069 while (!res && !ebml_level_end(matroska))
1070 res = ebml_parse(matroska, syntax, data);
1075 static int is_ebml_id_valid(uint32_t id)
1077 // Due to endian nonsense in Matroska, the highest byte with any bits set
1078 // will contain the leading length bit. This bit in turn identifies the
1079 // total byte length of the element by its position within the byte.
1080 unsigned int bits = av_log2(id);
1081 return id && (bits + 7) / 8 == (8 - bits % 8);
1085 * Allocate and return the entry for the level1 element with the given ID. If
1086 * an entry already exists, return the existing entry.
1088 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1092 MatroskaLevel1Element *elem;
1094 if (!is_ebml_id_valid(id))
1097 // Some files link to all clusters; useless.
1098 if (id == MATROSKA_ID_CLUSTER)
1101 // There can be multiple seekheads.
1102 if (id != MATROSKA_ID_SEEKHEAD) {
1103 for (i = 0; i < matroska->num_level1_elems; i++) {
1104 if (matroska->level1_elems[i].id == id)
1105 return &matroska->level1_elems[i];
1109 // Only a completely broken file would have more elements.
1110 // It also provides a low-effort way to escape from circular seekheads
1111 // (every iteration will add a level1 entry).
1112 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1113 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1117 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1118 *elem = (MatroskaLevel1Element){.id = id};
1123 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
1124 EbmlSyntax *syntax, void *data)
1126 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1129 // max. 16 MB for strings
1130 [EBML_STR] = 0x1000000,
1131 [EBML_UTF8] = 0x1000000,
1132 // max. 256 MB for binary data
1133 [EBML_BIN] = 0x10000000,
1134 // no limits for anything else
1136 AVIOContext *pb = matroska->ctx->pb;
1137 uint32_t id = syntax->id;
1141 MatroskaLevel1Element *level1_elem;
1143 data = (char *) data + syntax->data_offset;
1144 if (syntax->list_elem_size) {
1145 EbmlList *list = data;
1146 newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1148 return AVERROR(ENOMEM);
1149 list->elem = newelem;
1150 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1151 memset(data, 0, syntax->list_elem_size);
1155 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1156 matroska->current_id = 0;
1157 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1159 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1160 av_log(matroska->ctx, AV_LOG_ERROR,
1161 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1162 length, max_lengths[syntax->type], syntax->type);
1163 return AVERROR_INVALIDDATA;
1167 switch (syntax->type) {
1169 res = ebml_read_uint(pb, length, data);
1172 res = ebml_read_sint(pb, length, data);
1175 res = ebml_read_float(pb, length, data);
1179 res = ebml_read_ascii(pb, length, data);
1182 res = ebml_read_binary(pb, length, data);
1186 if ((res = ebml_read_master(matroska, length)) < 0)
1188 if (id == MATROSKA_ID_SEGMENT)
1189 matroska->segment_start = avio_tell(matroska->ctx->pb);
1190 if (id == MATROSKA_ID_CUES)
1191 matroska->cues_parsing_deferred = 0;
1192 if (syntax->type == EBML_LEVEL1 &&
1193 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1194 if (level1_elem->parsed)
1195 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1196 level1_elem->parsed = 1;
1198 return ebml_parse_nest(matroska, syntax->def.n, data);
1200 return ebml_parse_id(matroska, syntax->def.n, id, data);
1204 if (ffio_limit(pb, length) != length)
1205 return AVERROR(EIO);
1206 return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
1208 if (res == AVERROR_INVALIDDATA)
1209 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1210 else if (res == AVERROR(EIO))
1211 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1215 static void ebml_free(EbmlSyntax *syntax, void *data)
1218 for (i = 0; syntax[i].id; i++) {
1219 void *data_off = (char *) data + syntax[i].data_offset;
1220 switch (syntax[i].type) {
1226 av_freep(&((EbmlBin *) data_off)->data);
1230 if (syntax[i].list_elem_size) {
1231 EbmlList *list = data_off;
1232 char *ptr = list->elem;
1233 for (j = 0; j < list->nb_elem;
1234 j++, ptr += syntax[i].list_elem_size)
1235 ebml_free(syntax[i].def.n, ptr);
1236 av_freep(&list->elem);
1238 ebml_free(syntax[i].def.n, data_off);
1248 static int matroska_probe(AVProbeData *p)
1251 int len_mask = 0x80, size = 1, n = 1, i;
1254 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1257 /* length of header */
1259 while (size <= 8 && !(total & len_mask)) {
1265 total &= (len_mask - 1);
1267 total = (total << 8) | p->buf[4 + n++];
1269 /* Does the probe data contain the whole header? */
1270 if (p->buf_size < 4 + size + total)
1273 /* The header should contain a known document type. For now,
1274 * we don't parse the whole header but simply check for the
1275 * availability of that array of characters inside the header.
1276 * Not fully fool-proof, but good enough. */
1277 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1278 size_t probelen = strlen(matroska_doctypes[i]);
1279 if (total < probelen)
1281 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1282 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1283 return AVPROBE_SCORE_MAX;
1286 // probably valid EBML header but no recognized doctype
1287 return AVPROBE_SCORE_EXTENSION;
1290 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1293 MatroskaTrack *tracks = matroska->tracks.elem;
1296 for (i = 0; i < matroska->tracks.nb_elem; i++)
1297 if (tracks[i].num == num)
1300 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1304 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1305 MatroskaTrack *track)
1307 MatroskaTrackEncoding *encodings = track->encodings.elem;
1308 uint8_t *data = *buf;
1309 int isize = *buf_size;
1310 uint8_t *pkt_data = NULL;
1311 uint8_t av_unused *newpktdata;
1312 int pkt_size = isize;
1316 if (pkt_size >= 10000000U)
1317 return AVERROR_INVALIDDATA;
1319 switch (encodings[0].compression.algo) {
1320 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1322 int header_size = encodings[0].compression.settings.size;
1323 uint8_t *header = encodings[0].compression.settings.data;
1325 if (header_size && !header) {
1326 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1333 pkt_size = isize + header_size;
1334 pkt_data = av_malloc(pkt_size);
1336 return AVERROR(ENOMEM);
1338 memcpy(pkt_data, header, header_size);
1339 memcpy(pkt_data + header_size, data, isize);
1343 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1345 olen = pkt_size *= 3;
1346 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1348 result = AVERROR(ENOMEM);
1351 pkt_data = newpktdata;
1352 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1353 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1355 result = AVERROR_INVALIDDATA;
1362 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1364 z_stream zstream = { 0 };
1365 if (inflateInit(&zstream) != Z_OK)
1367 zstream.next_in = data;
1368 zstream.avail_in = isize;
1371 newpktdata = av_realloc(pkt_data, pkt_size);
1373 inflateEnd(&zstream);
1374 result = AVERROR(ENOMEM);
1377 pkt_data = newpktdata;
1378 zstream.avail_out = pkt_size - zstream.total_out;
1379 zstream.next_out = pkt_data + zstream.total_out;
1380 result = inflate(&zstream, Z_NO_FLUSH);
1381 } while (result == Z_OK && pkt_size < 10000000);
1382 pkt_size = zstream.total_out;
1383 inflateEnd(&zstream);
1384 if (result != Z_STREAM_END) {
1385 if (result == Z_MEM_ERROR)
1386 result = AVERROR(ENOMEM);
1388 result = AVERROR_INVALIDDATA;
1395 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1397 bz_stream bzstream = { 0 };
1398 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1400 bzstream.next_in = data;
1401 bzstream.avail_in = isize;
1404 newpktdata = av_realloc(pkt_data, pkt_size);
1406 BZ2_bzDecompressEnd(&bzstream);
1407 result = AVERROR(ENOMEM);
1410 pkt_data = newpktdata;
1411 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1412 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1413 result = BZ2_bzDecompress(&bzstream);
1414 } while (result == BZ_OK && pkt_size < 10000000);
1415 pkt_size = bzstream.total_out_lo32;
1416 BZ2_bzDecompressEnd(&bzstream);
1417 if (result != BZ_STREAM_END) {
1418 if (result == BZ_MEM_ERROR)
1419 result = AVERROR(ENOMEM);
1421 result = AVERROR_INVALIDDATA;
1428 return AVERROR_INVALIDDATA;
1432 *buf_size = pkt_size;
1440 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1441 AVDictionary **metadata, char *prefix)
1443 MatroskaTag *tags = list->elem;
1447 for (i = 0; i < list->nb_elem; i++) {
1448 const char *lang = tags[i].lang &&
1449 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1451 if (!tags[i].name) {
1452 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1456 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1458 av_strlcpy(key, tags[i].name, sizeof(key));
1459 if (tags[i].def || !lang) {
1460 av_dict_set(metadata, key, tags[i].string, 0);
1461 if (tags[i].sub.nb_elem)
1462 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1465 av_strlcat(key, "-", sizeof(key));
1466 av_strlcat(key, lang, sizeof(key));
1467 av_dict_set(metadata, key, tags[i].string, 0);
1468 if (tags[i].sub.nb_elem)
1469 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1472 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1475 static void matroska_convert_tags(AVFormatContext *s)
1477 MatroskaDemuxContext *matroska = s->priv_data;
1478 MatroskaTags *tags = matroska->tags.elem;
1481 for (i = 0; i < matroska->tags.nb_elem; i++) {
1482 if (tags[i].target.attachuid) {
1483 MatroskaAttachment *attachment = matroska->attachments.elem;
1485 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1486 if (attachment[j].uid == tags[i].target.attachuid &&
1487 attachment[j].stream) {
1488 matroska_convert_tag(s, &tags[i].tag,
1489 &attachment[j].stream->metadata, NULL);
1494 av_log(NULL, AV_LOG_WARNING,
1495 "The tags at index %d refer to a "
1496 "non-existent attachment %"PRId64".\n",
1497 i, tags[i].target.attachuid);
1499 } else if (tags[i].target.chapteruid) {
1500 MatroskaChapter *chapter = matroska->chapters.elem;
1502 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1503 if (chapter[j].uid == tags[i].target.chapteruid &&
1504 chapter[j].chapter) {
1505 matroska_convert_tag(s, &tags[i].tag,
1506 &chapter[j].chapter->metadata, NULL);
1511 av_log(NULL, AV_LOG_WARNING,
1512 "The tags at index %d refer to a non-existent chapter "
1514 i, tags[i].target.chapteruid);
1516 } else if (tags[i].target.trackuid) {
1517 MatroskaTrack *track = matroska->tracks.elem;
1519 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1520 if (track[j].uid == tags[i].target.trackuid &&
1522 matroska_convert_tag(s, &tags[i].tag,
1523 &track[j].stream->metadata, NULL);
1528 av_log(NULL, AV_LOG_WARNING,
1529 "The tags at index %d refer to a non-existent track "
1531 i, tags[i].target.trackuid);
1534 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1535 tags[i].target.type);
1540 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1543 uint32_t level_up = matroska->level_up;
1544 uint32_t saved_id = matroska->current_id;
1545 int64_t before_pos = avio_tell(matroska->ctx->pb);
1546 MatroskaLevel level;
1551 offset = pos + matroska->segment_start;
1552 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1553 /* We don't want to lose our seekhead level, so we add
1554 * a dummy. This is a crude hack. */
1555 if (matroska->num_levels == EBML_MAX_DEPTH) {
1556 av_log(matroska->ctx, AV_LOG_INFO,
1557 "Max EBML element depth (%d) reached, "
1558 "cannot parse further.\n", EBML_MAX_DEPTH);
1559 ret = AVERROR_INVALIDDATA;
1562 level.length = (uint64_t) -1;
1563 matroska->levels[matroska->num_levels] = level;
1564 matroska->num_levels++;
1565 matroska->current_id = 0;
1567 ret = ebml_parse(matroska, matroska_segment, matroska);
1569 /* remove dummy level */
1570 while (matroska->num_levels) {
1571 uint64_t length = matroska->levels[--matroska->num_levels].length;
1572 if (length == (uint64_t) -1)
1578 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1579 matroska->level_up = level_up;
1580 matroska->current_id = saved_id;
1585 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1587 EbmlList *seekhead_list = &matroska->seekhead;
1590 // we should not do any seeking in the streaming case
1591 if (!matroska->ctx->pb->seekable)
1594 for (i = 0; i < seekhead_list->nb_elem; i++) {
1595 MatroskaSeekhead *seekheads = seekhead_list->elem;
1596 uint32_t id = seekheads[i].id;
1597 uint64_t pos = seekheads[i].pos;
1599 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1600 if (!elem || elem->parsed)
1605 // defer cues parsing until we actually need cue data.
1606 if (id == MATROSKA_ID_CUES)
1609 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1610 // mark index as broken
1611 matroska->cues_parsing_deferred = -1;
1619 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1621 EbmlList *index_list;
1622 MatroskaIndex *index;
1623 uint64_t index_scale = 1;
1626 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1629 index_list = &matroska->index;
1630 index = index_list->elem;
1631 if (index_list->nb_elem < 2)
1633 if (index[1].time > 1E14 / matroska->time_scale) {
1634 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1637 for (i = 0; i < index_list->nb_elem; i++) {
1638 EbmlList *pos_list = &index[i].pos;
1639 MatroskaIndexPos *pos = pos_list->elem;
1640 for (j = 0; j < pos_list->nb_elem; j++) {
1641 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1643 if (track && track->stream)
1644 av_add_index_entry(track->stream,
1645 pos[j].pos + matroska->segment_start,
1646 index[i].time / index_scale, 0, 0,
1652 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1655 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1658 for (i = 0; i < matroska->num_level1_elems; i++) {
1659 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1660 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1661 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1662 matroska->cues_parsing_deferred = -1;
1668 matroska_add_index_entries(matroska);
1671 static int matroska_aac_profile(char *codec_id)
1673 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1676 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1677 if (strstr(codec_id, aac_profiles[profile]))
1682 static int matroska_aac_sri(int samplerate)
1686 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1687 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1692 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1695 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1696 time_t creation_time = date_utc / 1000000000 + 978307200;
1697 struct tm tmpbuf, *ptm = gmtime_r(&creation_time, &tmpbuf);
1699 if (strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm))
1700 av_dict_set(metadata, "creation_time", buffer, 0);
1703 static int matroska_parse_flac(AVFormatContext *s,
1704 MatroskaTrack *track,
1707 AVStream *st = track->stream;
1708 uint8_t *p = track->codec_priv.data;
1709 int size = track->codec_priv.size;
1711 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1712 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1713 track->codec_priv.size = 0;
1717 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1719 p += track->codec_priv.size;
1720 size -= track->codec_priv.size;
1722 /* parse the remaining metadata blocks if present */
1724 int block_last, block_type, block_size;
1726 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1730 if (block_size > size)
1733 /* check for the channel mask */
1734 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1735 AVDictionary *dict = NULL;
1736 AVDictionaryEntry *chmask;
1738 ff_vorbis_comment(s, &dict, p, block_size, 0);
1739 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1741 uint64_t mask = strtol(chmask->value, NULL, 0);
1742 if (!mask || mask & ~0x3ffffULL) {
1743 av_log(s, AV_LOG_WARNING,
1744 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1746 st->codecpar->channel_layout = mask;
1748 av_dict_free(&dict);
1758 static int mkv_field_order(int64_t field_order)
1760 switch (field_order) {
1761 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
1762 return AV_FIELD_PROGRESSIVE;
1763 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
1764 return AV_FIELD_UNKNOWN;
1765 case MATROSKA_VIDEO_FIELDORDER_TT:
1767 case MATROSKA_VIDEO_FIELDORDER_BB:
1769 case MATROSKA_VIDEO_FIELDORDER_BT:
1771 case MATROSKA_VIDEO_FIELDORDER_TB:
1774 return AV_FIELD_UNKNOWN;
1778 static void mkv_stereo_mode_display_mul(int stereo_mode,
1779 int *h_width, int *h_height)
1781 switch (stereo_mode) {
1782 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
1783 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
1784 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
1785 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
1786 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
1788 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
1789 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
1790 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
1791 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
1794 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
1795 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
1796 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
1797 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
1803 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
1804 const MatroskaMasteringMeta* mastering_meta =
1805 &track->video.color.mastering_meta;
1806 // Mastering primaries are CIE 1931 coords, and must be > 0.
1807 const int has_mastering_primaries =
1808 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
1809 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
1810 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
1811 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
1812 const int has_mastering_luminance = mastering_meta->max_luminance > 0;
1814 if (track->video.color.matrix_coefficients != AVCOL_SPC_RESERVED)
1815 st->codecpar->color_space = track->video.color.matrix_coefficients;
1816 if (track->video.color.primaries != AVCOL_PRI_RESERVED)
1817 st->codecpar->color_primaries = track->video.color.primaries;
1818 if (track->video.color.transfer_characteristics != AVCOL_TRC_RESERVED)
1819 st->codecpar->color_trc = track->video.color.transfer_characteristics;
1820 if (track->video.color.range != AVCOL_RANGE_UNSPECIFIED &&
1821 track->video.color.range <= AVCOL_RANGE_JPEG)
1822 st->codecpar->color_range = track->video.color.range;
1824 if (has_mastering_primaries || has_mastering_luminance) {
1825 // Use similar rationals as other standards.
1826 const int chroma_den = 50000;
1827 const int luma_den = 10000;
1828 AVMasteringDisplayMetadata *metadata =
1829 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
1830 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
1831 sizeof(AVMasteringDisplayMetadata));
1833 return AVERROR(ENOMEM);
1835 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
1836 if (has_mastering_primaries) {
1837 metadata->display_primaries[0][0] = av_make_q(
1838 round(mastering_meta->r_x * chroma_den), chroma_den);
1839 metadata->display_primaries[0][1] = av_make_q(
1840 round(mastering_meta->r_y * chroma_den), chroma_den);
1841 metadata->display_primaries[1][0] = av_make_q(
1842 round(mastering_meta->g_x * chroma_den), chroma_den);
1843 metadata->display_primaries[1][1] = av_make_q(
1844 round(mastering_meta->g_y * chroma_den), chroma_den);
1845 metadata->display_primaries[2][0] = av_make_q(
1846 round(mastering_meta->b_x * chroma_den), chroma_den);
1847 metadata->display_primaries[2][1] = av_make_q(
1848 round(mastering_meta->b_y * chroma_den), chroma_den);
1849 metadata->white_point[0] = av_make_q(
1850 round(mastering_meta->white_x * chroma_den), chroma_den);
1851 metadata->white_point[1] = av_make_q(
1852 round(mastering_meta->white_y * chroma_den), chroma_den);
1853 metadata->has_primaries = 1;
1855 if (has_mastering_luminance) {
1856 metadata->max_luminance = av_make_q(
1857 round(mastering_meta->max_luminance * luma_den), luma_den);
1858 metadata->min_luminance = av_make_q(
1859 round(mastering_meta->min_luminance * luma_den), luma_den);
1860 metadata->has_luminance = 1;
1866 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
1868 const AVCodecTag *codec_tags;
1870 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
1871 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
1873 /* Normalize noncompliant private data that starts with the fourcc
1874 * by expanding/shifting the data by 4 bytes and storing the data
1875 * size at the start. */
1876 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
1877 uint8_t *p = av_realloc(track->codec_priv.data,
1878 track->codec_priv.size + 4);
1880 return AVERROR(ENOMEM);
1881 memmove(p + 4, p, track->codec_priv.size);
1882 track->codec_priv.data = p;
1883 track->codec_priv.size += 4;
1884 AV_WB32(track->codec_priv.data, track->codec_priv.size);
1887 *fourcc = AV_RL32(track->codec_priv.data + 4);
1888 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
1893 static int matroska_parse_tracks(AVFormatContext *s)
1895 MatroskaDemuxContext *matroska = s->priv_data;
1896 MatroskaTrack *tracks = matroska->tracks.elem;
1901 for (i = 0; i < matroska->tracks.nb_elem; i++) {
1902 MatroskaTrack *track = &tracks[i];
1903 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1904 EbmlList *encodings_list = &track->encodings;
1905 MatroskaTrackEncoding *encodings = encodings_list->elem;
1906 uint8_t *extradata = NULL;
1907 int extradata_size = 0;
1908 int extradata_offset = 0;
1909 uint32_t fourcc = 0;
1911 char* key_id_base64 = NULL;
1914 /* Apply some sanity checks. */
1915 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1916 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1917 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1918 track->type != MATROSKA_TRACK_TYPE_METADATA) {
1919 av_log(matroska->ctx, AV_LOG_INFO,
1920 "Unknown or unsupported track type %"PRIu64"\n",
1924 if (!track->codec_id)
1927 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
1928 isnan(track->audio.samplerate)) {
1929 av_log(matroska->ctx, AV_LOG_WARNING,
1930 "Invalid sample rate %f, defaulting to 8000 instead.\n",
1931 track->audio.samplerate);
1932 track->audio.samplerate = 8000;
1935 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1936 if (!track->default_duration && track->video.frame_rate > 0)
1937 track->default_duration = 1000000000 / track->video.frame_rate;
1938 if (track->video.display_width == -1)
1939 track->video.display_width = track->video.pixel_width;
1940 if (track->video.display_height == -1)
1941 track->video.display_height = track->video.pixel_height;
1942 if (track->video.color_space.size == 4)
1943 fourcc = AV_RL32(track->video.color_space.data);
1944 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1945 if (!track->audio.out_samplerate)
1946 track->audio.out_samplerate = track->audio.samplerate;
1948 if (encodings_list->nb_elem > 1) {
1949 av_log(matroska->ctx, AV_LOG_ERROR,
1950 "Multiple combined encodings not supported");
1951 } else if (encodings_list->nb_elem == 1) {
1952 if (encodings[0].type) {
1953 if (encodings[0].encryption.key_id.size > 0) {
1954 /* Save the encryption key id to be stored later as a
1956 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1957 key_id_base64 = av_malloc(b64_size);
1958 if (key_id_base64 == NULL)
1959 return AVERROR(ENOMEM);
1961 av_base64_encode(key_id_base64, b64_size,
1962 encodings[0].encryption.key_id.data,
1963 encodings[0].encryption.key_id.size);
1965 encodings[0].scope = 0;
1966 av_log(matroska->ctx, AV_LOG_ERROR,
1967 "Unsupported encoding type");
1971 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1974 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1977 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1979 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
1980 encodings[0].scope = 0;
1981 av_log(matroska->ctx, AV_LOG_ERROR,
1982 "Unsupported encoding type");
1983 } else if (track->codec_priv.size && encodings[0].scope & 2) {
1984 uint8_t *codec_priv = track->codec_priv.data;
1985 int ret = matroska_decode_buffer(&track->codec_priv.data,
1986 &track->codec_priv.size,
1989 track->codec_priv.data = NULL;
1990 track->codec_priv.size = 0;
1991 av_log(matroska->ctx, AV_LOG_ERROR,
1992 "Failed to decode codec private data\n");
1995 if (codec_priv != track->codec_priv.data)
1996 av_free(codec_priv);
2000 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2001 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2002 strlen(ff_mkv_codec_tags[j].str))) {
2003 codec_id = ff_mkv_codec_tags[j].id;
2008 st = track->stream = avformat_new_stream(s, NULL);
2010 av_free(key_id_base64);
2011 return AVERROR(ENOMEM);
2014 if (key_id_base64) {
2015 /* export encryption key id as base64 metadata tag */
2016 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
2017 av_freep(&key_id_base64);
2020 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2021 track->codec_priv.size >= 40 &&
2022 track->codec_priv.data) {
2023 track->ms_compat = 1;
2024 bit_depth = AV_RL16(track->codec_priv.data + 14);
2025 fourcc = AV_RL32(track->codec_priv.data + 16);
2026 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2029 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2031 extradata_offset = 40;
2032 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2033 track->codec_priv.size >= 14 &&
2034 track->codec_priv.data) {
2036 ffio_init_context(&b, track->codec_priv.data,
2037 track->codec_priv.size,
2038 0, NULL, NULL, NULL, NULL);
2039 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2042 codec_id = st->codecpar->codec_id;
2043 fourcc = st->codecpar->codec_tag;
2044 extradata_offset = FFMIN(track->codec_priv.size, 18);
2045 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2046 /* Normally 36, but allow noncompliant private data */
2047 && (track->codec_priv.size >= 32)
2048 && (track->codec_priv.data)) {
2049 uint16_t sample_size;
2050 int ret = get_qt_codec(track, &fourcc, &codec_id);
2053 sample_size = AV_RB16(track->codec_priv.data + 26);
2055 if (sample_size == 8) {
2056 fourcc = MKTAG('r','a','w',' ');
2057 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2058 } else if (sample_size == 16) {
2059 fourcc = MKTAG('t','w','o','s');
2060 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2063 if ((fourcc == MKTAG('t','w','o','s') ||
2064 fourcc == MKTAG('s','o','w','t')) &&
2066 codec_id = AV_CODEC_ID_PCM_S8;
2067 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2068 (track->codec_priv.size >= 21) &&
2069 (track->codec_priv.data)) {
2070 int ret = get_qt_codec(track, &fourcc, &codec_id);
2073 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2074 fourcc = MKTAG('S','V','Q','3');
2075 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2077 if (codec_id == AV_CODEC_ID_NONE) {
2079 av_get_codec_tag_string(buf, sizeof(buf), fourcc);
2080 av_log(matroska->ctx, AV_LOG_ERROR,
2081 "mov FourCC not found %s.\n", buf);
2083 if (track->codec_priv.size >= 86) {
2084 bit_depth = AV_RB16(track->codec_priv.data + 82);
2085 ffio_init_context(&b, track->codec_priv.data,
2086 track->codec_priv.size,
2087 0, NULL, NULL, NULL, NULL);
2088 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2090 track->has_palette = 1;
2093 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2094 switch (track->audio.bitdepth) {
2096 codec_id = AV_CODEC_ID_PCM_U8;
2099 codec_id = AV_CODEC_ID_PCM_S24BE;
2102 codec_id = AV_CODEC_ID_PCM_S32BE;
2105 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2106 switch (track->audio.bitdepth) {
2108 codec_id = AV_CODEC_ID_PCM_U8;
2111 codec_id = AV_CODEC_ID_PCM_S24LE;
2114 codec_id = AV_CODEC_ID_PCM_S32LE;
2117 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2118 track->audio.bitdepth == 64) {
2119 codec_id = AV_CODEC_ID_PCM_F64LE;
2120 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2121 int profile = matroska_aac_profile(track->codec_id);
2122 int sri = matroska_aac_sri(track->audio.samplerate);
2123 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2125 return AVERROR(ENOMEM);
2126 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2127 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2128 if (strstr(track->codec_id, "SBR")) {
2129 sri = matroska_aac_sri(track->audio.out_samplerate);
2130 extradata[2] = 0x56;
2131 extradata[3] = 0xE5;
2132 extradata[4] = 0x80 | (sri << 3);
2136 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2137 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2138 * Create the "atom size", "tag", and "tag version" fields the
2139 * decoder expects manually. */
2140 extradata_size = 12 + track->codec_priv.size;
2141 extradata = av_mallocz(extradata_size +
2142 AV_INPUT_BUFFER_PADDING_SIZE);
2144 return AVERROR(ENOMEM);
2145 AV_WB32(extradata, extradata_size);
2146 memcpy(&extradata[4], "alac", 4);
2147 AV_WB32(&extradata[8], 0);
2148 memcpy(&extradata[12], track->codec_priv.data,
2149 track->codec_priv.size);
2150 } else if (codec_id == AV_CODEC_ID_TTA) {
2151 extradata_size = 30;
2152 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2154 return AVERROR(ENOMEM);
2155 ffio_init_context(&b, extradata, extradata_size, 1,
2156 NULL, NULL, NULL, NULL);
2157 avio_write(&b, "TTA1", 4);
2159 if (track->audio.channels > UINT16_MAX ||
2160 track->audio.bitdepth > UINT16_MAX) {
2161 av_log(matroska->ctx, AV_LOG_WARNING,
2162 "Too large audio channel number %"PRIu64
2163 " or bitdepth %"PRIu64". Skipping track.\n",
2164 track->audio.channels, track->audio.bitdepth);
2165 av_freep(&extradata);
2166 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2167 return AVERROR_INVALIDDATA;
2171 avio_wl16(&b, track->audio.channels);
2172 avio_wl16(&b, track->audio.bitdepth);
2173 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2174 return AVERROR_INVALIDDATA;
2175 avio_wl32(&b, track->audio.out_samplerate);
2176 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
2177 track->audio.out_samplerate,
2178 AV_TIME_BASE * 1000));
2179 } else if (codec_id == AV_CODEC_ID_RV10 ||
2180 codec_id == AV_CODEC_ID_RV20 ||
2181 codec_id == AV_CODEC_ID_RV30 ||
2182 codec_id == AV_CODEC_ID_RV40) {
2183 extradata_offset = 26;
2184 } else if (codec_id == AV_CODEC_ID_RA_144) {
2185 track->audio.out_samplerate = 8000;
2186 track->audio.channels = 1;
2187 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2188 codec_id == AV_CODEC_ID_COOK ||
2189 codec_id == AV_CODEC_ID_ATRAC3 ||
2190 codec_id == AV_CODEC_ID_SIPR)
2191 && track->codec_priv.data) {
2194 ffio_init_context(&b, track->codec_priv.data,
2195 track->codec_priv.size,
2196 0, NULL, NULL, NULL, NULL);
2198 flavor = avio_rb16(&b);
2199 track->audio.coded_framesize = avio_rb32(&b);
2201 track->audio.sub_packet_h = avio_rb16(&b);
2202 track->audio.frame_size = avio_rb16(&b);
2203 track->audio.sub_packet_size = avio_rb16(&b);
2205 track->audio.coded_framesize <= 0 ||
2206 track->audio.sub_packet_h <= 0 ||
2207 track->audio.frame_size <= 0 ||
2208 track->audio.sub_packet_size <= 0)
2209 return AVERROR_INVALIDDATA;
2210 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2211 track->audio.frame_size);
2212 if (!track->audio.buf)
2213 return AVERROR(ENOMEM);
2214 if (codec_id == AV_CODEC_ID_RA_288) {
2215 st->codecpar->block_align = track->audio.coded_framesize;
2216 track->codec_priv.size = 0;
2218 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2219 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2220 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2221 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2223 st->codecpar->block_align = track->audio.sub_packet_size;
2224 extradata_offset = 78;
2226 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2227 ret = matroska_parse_flac(s, track, &extradata_offset);
2230 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2231 fourcc = AV_RL32(track->codec_priv.data);
2233 track->codec_priv.size -= extradata_offset;
2235 if (codec_id == AV_CODEC_ID_NONE)
2236 av_log(matroska->ctx, AV_LOG_INFO,
2237 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2239 if (track->time_scale < 0.01)
2240 track->time_scale = 1.0;
2241 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2242 1000 * 1000 * 1000); /* 64 bit pts in ns */
2244 /* convert the delay from ns to the track timebase */
2245 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2246 (AVRational){ 1, 1000000000 },
2249 st->codecpar->codec_id = codec_id;
2251 if (strcmp(track->language, "und"))
2252 av_dict_set(&st->metadata, "language", track->language, 0);
2253 av_dict_set(&st->metadata, "title", track->name, 0);
2255 if (track->flag_default)
2256 st->disposition |= AV_DISPOSITION_DEFAULT;
2257 if (track->flag_forced)
2258 st->disposition |= AV_DISPOSITION_FORCED;
2260 if (!st->codecpar->extradata) {
2262 st->codecpar->extradata = extradata;
2263 st->codecpar->extradata_size = extradata_size;
2264 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2265 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2266 return AVERROR(ENOMEM);
2267 memcpy(st->codecpar->extradata,
2268 track->codec_priv.data + extradata_offset,
2269 track->codec_priv.size);
2273 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2274 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2275 int display_width_mul = 1;
2276 int display_height_mul = 1;
2278 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2279 st->codecpar->codec_tag = fourcc;
2281 st->codecpar->bits_per_coded_sample = bit_depth;
2282 st->codecpar->width = track->video.pixel_width;
2283 st->codecpar->height = track->video.pixel_height;
2285 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2286 st->codecpar->field_order = mkv_field_order(track->video.field_order);
2288 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2289 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2291 av_reduce(&st->sample_aspect_ratio.num,
2292 &st->sample_aspect_ratio.den,
2293 st->codecpar->height * track->video.display_width * display_width_mul,
2294 st->codecpar->width * track->video.display_height * display_height_mul,
2296 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2297 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2299 if (track->default_duration) {
2300 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2301 1000000000, track->default_duration, 30000);
2302 #if FF_API_R_FRAME_RATE
2303 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2304 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2305 st->r_frame_rate = st->avg_frame_rate;
2309 /* export stereo mode flag as metadata tag */
2310 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2311 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2313 /* export alpha mode flag as metadata tag */
2314 if (track->video.alpha_mode)
2315 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2317 /* if we have virtual track, mark the real tracks */
2318 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2320 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2322 snprintf(buf, sizeof(buf), "%s_%d",
2323 ff_matroska_video_stereo_plane[planes[j].type], i);
2324 for (k=0; k < matroska->tracks.nb_elem; k++)
2325 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2326 av_dict_set(&tracks[k].stream->metadata,
2327 "stereo_mode", buf, 0);
2331 // add stream level stereo3d side data if it is a supported format
2332 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2333 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2334 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2339 if (s->strict_std_compliance <= FF_COMPLIANCE_UNOFFICIAL) {
2340 int ret = mkv_parse_video_color(st, track);
2344 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2345 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2346 st->codecpar->codec_tag = fourcc;
2347 st->codecpar->sample_rate = track->audio.out_samplerate;
2348 st->codecpar->channels = track->audio.channels;
2349 if (!st->codecpar->bits_per_coded_sample)
2350 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2351 if (st->codecpar->codec_id == AV_CODEC_ID_MP3)
2352 st->need_parsing = AVSTREAM_PARSE_FULL;
2353 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2354 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2355 if (track->codec_delay > 0) {
2356 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2357 (AVRational){1, 1000000000},
2358 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2359 48000 : st->codecpar->sample_rate});
2361 if (track->seek_preroll > 0) {
2362 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2363 (AVRational){1, 1000000000},
2364 (AVRational){1, st->codecpar->sample_rate});
2366 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2367 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2369 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2370 st->disposition |= AV_DISPOSITION_CAPTIONS;
2371 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2372 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2373 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2374 st->disposition |= AV_DISPOSITION_METADATA;
2376 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2377 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2378 if (st->codecpar->codec_id == AV_CODEC_ID_ASS)
2379 matroska->contains_ssa = 1;
2386 static int matroska_read_header(AVFormatContext *s)
2388 MatroskaDemuxContext *matroska = s->priv_data;
2389 EbmlList *attachments_list = &matroska->attachments;
2390 EbmlList *chapters_list = &matroska->chapters;
2391 MatroskaAttachment *attachments;
2392 MatroskaChapter *chapters;
2393 uint64_t max_start = 0;
2399 matroska->cues_parsing_deferred = 1;
2401 /* First read the EBML header. */
2402 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2403 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2404 ebml_free(ebml_syntax, &ebml);
2405 return AVERROR_INVALIDDATA;
2407 if (ebml.version > EBML_VERSION ||
2408 ebml.max_size > sizeof(uint64_t) ||
2409 ebml.id_length > sizeof(uint32_t) ||
2410 ebml.doctype_version > 3) {
2411 av_log(matroska->ctx, AV_LOG_ERROR,
2412 "EBML header using unsupported features\n"
2413 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2414 ebml.version, ebml.doctype, ebml.doctype_version);
2415 ebml_free(ebml_syntax, &ebml);
2416 return AVERROR_PATCHWELCOME;
2417 } else if (ebml.doctype_version == 3) {
2418 av_log(matroska->ctx, AV_LOG_WARNING,
2419 "EBML header using unsupported features\n"
2420 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2421 ebml.version, ebml.doctype, ebml.doctype_version);
2423 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2424 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2426 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2427 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2428 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2429 ebml_free(ebml_syntax, &ebml);
2430 return AVERROR_INVALIDDATA;
2433 ebml_free(ebml_syntax, &ebml);
2435 /* The next thing is a segment. */
2436 pos = avio_tell(matroska->ctx->pb);
2437 res = ebml_parse(matroska, matroska_segments, matroska);
2438 // try resyncing until we find a EBML_STOP type element.
2440 res = matroska_resync(matroska, pos);
2443 pos = avio_tell(matroska->ctx->pb);
2444 res = ebml_parse(matroska, matroska_segment, matroska);
2446 matroska_execute_seekhead(matroska);
2448 if (!matroska->time_scale)
2449 matroska->time_scale = 1000000;
2450 if (matroska->duration)
2451 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2452 1000 / AV_TIME_BASE;
2453 av_dict_set(&s->metadata, "title", matroska->title, 0);
2454 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2456 if (matroska->date_utc.size == 8)
2457 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2459 res = matroska_parse_tracks(s);
2463 attachments = attachments_list->elem;
2464 for (j = 0; j < attachments_list->nb_elem; j++) {
2465 if (!(attachments[j].filename && attachments[j].mime &&
2466 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2467 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2469 AVStream *st = avformat_new_stream(s, NULL);
2472 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2473 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2474 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2476 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2477 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2478 strlen(ff_mkv_image_mime_tags[i].str))) {
2479 st->codecpar->codec_id = ff_mkv_image_mime_tags[i].id;
2484 attachments[j].stream = st;
2486 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2487 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2488 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2490 av_init_packet(&st->attached_pic);
2491 if ((res = av_new_packet(&st->attached_pic, attachments[j].bin.size)) < 0)
2493 memcpy(st->attached_pic.data, attachments[j].bin.data, attachments[j].bin.size);
2494 st->attached_pic.stream_index = st->index;
2495 st->attached_pic.flags |= AV_PKT_FLAG_KEY;
2497 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2498 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2500 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2501 attachments[j].bin.size);
2503 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2504 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2505 strlen(ff_mkv_mime_tags[i].str))) {
2506 st->codecpar->codec_id = ff_mkv_mime_tags[i].id;
2514 chapters = chapters_list->elem;
2515 for (i = 0; i < chapters_list->nb_elem; i++)
2516 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2517 (max_start == 0 || chapters[i].start > max_start)) {
2518 chapters[i].chapter =
2519 avpriv_new_chapter(s, chapters[i].uid,
2520 (AVRational) { 1, 1000000000 },
2521 chapters[i].start, chapters[i].end,
2523 if (chapters[i].chapter) {
2524 av_dict_set(&chapters[i].chapter->metadata,
2525 "title", chapters[i].title, 0);
2527 max_start = chapters[i].start;
2530 matroska_add_index_entries(matroska);
2532 matroska_convert_tags(s);
2536 matroska_read_close(s);
2541 * Put one packet in an application-supplied AVPacket struct.
2542 * Returns 0 on success or -1 on failure.
2544 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2547 if (matroska->num_packets > 0) {
2548 MatroskaTrack *tracks = matroska->tracks.elem;
2549 MatroskaTrack *track;
2550 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2551 av_freep(&matroska->packets[0]);
2552 track = &tracks[pkt->stream_index];
2553 if (track->has_palette) {
2554 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2556 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
2558 memcpy(pal, track->palette, AVPALETTE_SIZE);
2560 track->has_palette = 0;
2562 if (matroska->num_packets > 1) {
2564 memmove(&matroska->packets[0], &matroska->packets[1],
2565 (matroska->num_packets - 1) * sizeof(AVPacket *));
2566 newpackets = av_realloc(matroska->packets,
2567 (matroska->num_packets - 1) *
2568 sizeof(AVPacket *));
2570 matroska->packets = newpackets;
2572 av_freep(&matroska->packets);
2573 matroska->prev_pkt = NULL;
2575 matroska->num_packets--;
2583 * Free all packets in our internal queue.
2585 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2587 matroska->prev_pkt = NULL;
2588 if (matroska->packets) {
2590 for (n = 0; n < matroska->num_packets; n++) {
2591 av_packet_unref(matroska->packets[n]);
2592 av_freep(&matroska->packets[n]);
2594 av_freep(&matroska->packets);
2595 matroska->num_packets = 0;
2599 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2600 int *buf_size, int type,
2601 uint32_t **lace_buf, int *laces)
2603 int res = 0, n, size = *buf_size;
2604 uint8_t *data = *buf;
2605 uint32_t *lace_size;
2609 *lace_buf = av_mallocz(sizeof(int));
2611 return AVERROR(ENOMEM);
2613 *lace_buf[0] = size;
2617 av_assert0(size > 0);
2621 lace_size = av_mallocz(*laces * sizeof(int));
2623 return AVERROR(ENOMEM);
2626 case 0x1: /* Xiph lacing */
2630 for (n = 0; res == 0 && n < *laces - 1; n++) {
2632 if (size <= total) {
2633 res = AVERROR_INVALIDDATA;
2638 lace_size[n] += temp;
2645 if (size <= total) {
2646 res = AVERROR_INVALIDDATA;
2650 lace_size[n] = size - total;
2654 case 0x2: /* fixed-size lacing */
2655 if (size % (*laces)) {
2656 res = AVERROR_INVALIDDATA;
2659 for (n = 0; n < *laces; n++)
2660 lace_size[n] = size / *laces;
2663 case 0x3: /* EBML lacing */
2667 n = matroska_ebmlnum_uint(matroska, data, size, &num);
2668 if (n < 0 || num > INT_MAX) {
2669 av_log(matroska->ctx, AV_LOG_INFO,
2670 "EBML block data error\n");
2671 res = n<0 ? n : AVERROR_INVALIDDATA;
2676 total = lace_size[0] = num;
2677 for (n = 1; res == 0 && n < *laces - 1; n++) {
2680 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2681 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2682 av_log(matroska->ctx, AV_LOG_INFO,
2683 "EBML block data error\n");
2684 res = r<0 ? r : AVERROR_INVALIDDATA;
2689 lace_size[n] = lace_size[n - 1] + snum;
2690 total += lace_size[n];
2692 if (size <= total) {
2693 res = AVERROR_INVALIDDATA;
2696 lace_size[*laces - 1] = size - total;
2702 *lace_buf = lace_size;
2708 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2709 MatroskaTrack *track, AVStream *st,
2710 uint8_t *data, int size, uint64_t timecode,
2713 int a = st->codecpar->block_align;
2714 int sps = track->audio.sub_packet_size;
2715 int cfs = track->audio.coded_framesize;
2716 int h = track->audio.sub_packet_h;
2717 int y = track->audio.sub_packet_cnt;
2718 int w = track->audio.frame_size;
2721 if (!track->audio.pkt_cnt) {
2722 if (track->audio.sub_packet_cnt == 0)
2723 track->audio.buf_timecode = timecode;
2724 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
2725 if (size < cfs * h / 2) {
2726 av_log(matroska->ctx, AV_LOG_ERROR,
2727 "Corrupt int4 RM-style audio packet size\n");
2728 return AVERROR_INVALIDDATA;
2730 for (x = 0; x < h / 2; x++)
2731 memcpy(track->audio.buf + x * 2 * w + y * cfs,
2732 data + x * cfs, cfs);
2733 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
2735 av_log(matroska->ctx, AV_LOG_ERROR,
2736 "Corrupt sipr RM-style audio packet size\n");
2737 return AVERROR_INVALIDDATA;
2739 memcpy(track->audio.buf + y * w, data, w);
2741 if (size < sps * w / sps || h<=0 || w%sps) {
2742 av_log(matroska->ctx, AV_LOG_ERROR,
2743 "Corrupt generic RM-style audio packet size\n");
2744 return AVERROR_INVALIDDATA;
2746 for (x = 0; x < w / sps; x++)
2747 memcpy(track->audio.buf +
2748 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2749 data + x * sps, sps);
2752 if (++track->audio.sub_packet_cnt >= h) {
2753 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
2754 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2755 track->audio.sub_packet_cnt = 0;
2756 track->audio.pkt_cnt = h * w / a;
2760 while (track->audio.pkt_cnt) {
2762 AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2764 return AVERROR(ENOMEM);
2766 ret = av_new_packet(pkt, a);
2772 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2774 pkt->pts = track->audio.buf_timecode;
2775 track->audio.buf_timecode = AV_NOPTS_VALUE;
2777 pkt->stream_index = st->index;
2778 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2784 /* reconstruct full wavpack blocks from mangled matroska ones */
2785 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
2786 uint8_t **pdst, int *size)
2788 uint8_t *dst = NULL;
2793 int ret, offset = 0;
2795 if (srclen < 12 || track->stream->codecpar->extradata_size < 2)
2796 return AVERROR_INVALIDDATA;
2798 ver = AV_RL16(track->stream->codecpar->extradata);
2800 samples = AV_RL32(src);
2804 while (srclen >= 8) {
2809 uint32_t flags = AV_RL32(src);
2810 uint32_t crc = AV_RL32(src + 4);
2814 multiblock = (flags & 0x1800) != 0x1800;
2817 ret = AVERROR_INVALIDDATA;
2820 blocksize = AV_RL32(src);
2826 if (blocksize > srclen) {
2827 ret = AVERROR_INVALIDDATA;
2831 tmp = av_realloc(dst, dstlen + blocksize + 32);
2833 ret = AVERROR(ENOMEM);
2837 dstlen += blocksize + 32;
2839 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2840 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2841 AV_WL16(dst + offset + 8, ver); // version
2842 AV_WL16(dst + offset + 10, 0); // track/index_no
2843 AV_WL32(dst + offset + 12, 0); // total samples
2844 AV_WL32(dst + offset + 16, 0); // block index
2845 AV_WL32(dst + offset + 20, samples); // number of samples
2846 AV_WL32(dst + offset + 24, flags); // flags
2847 AV_WL32(dst + offset + 28, crc); // crc
2848 memcpy(dst + offset + 32, src, blocksize); // block data
2851 srclen -= blocksize;
2852 offset += blocksize + 32;
2865 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
2866 MatroskaTrack *track,
2868 uint8_t *data, int data_len,
2874 uint8_t *id, *settings, *text, *buf;
2875 int id_len, settings_len, text_len;
2880 return AVERROR_INVALIDDATA;
2883 q = data + data_len;
2888 if (*p == '\r' || *p == '\n') {
2897 if (p >= q || *p != '\n')
2898 return AVERROR_INVALIDDATA;
2904 if (*p == '\r' || *p == '\n') {
2905 settings_len = p - settings;
2913 if (p >= q || *p != '\n')
2914 return AVERROR_INVALIDDATA;
2919 while (text_len > 0) {
2920 const int len = text_len - 1;
2921 const uint8_t c = p[len];
2922 if (c != '\r' && c != '\n')
2928 return AVERROR_INVALIDDATA;
2930 pkt = av_mallocz(sizeof(*pkt));
2932 return AVERROR(ENOMEM);
2933 err = av_new_packet(pkt, text_len);
2936 return AVERROR(err);
2939 memcpy(pkt->data, text, text_len);
2942 buf = av_packet_new_side_data(pkt,
2943 AV_PKT_DATA_WEBVTT_IDENTIFIER,
2947 return AVERROR(ENOMEM);
2949 memcpy(buf, id, id_len);
2952 if (settings_len > 0) {
2953 buf = av_packet_new_side_data(pkt,
2954 AV_PKT_DATA_WEBVTT_SETTINGS,
2958 return AVERROR(ENOMEM);
2960 memcpy(buf, settings, settings_len);
2963 // Do we need this for subtitles?
2964 // pkt->flags = AV_PKT_FLAG_KEY;
2966 pkt->stream_index = st->index;
2967 pkt->pts = timecode;
2969 // Do we need this for subtitles?
2970 // pkt->dts = timecode;
2972 pkt->duration = duration;
2975 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2976 matroska->prev_pkt = pkt;
2981 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
2982 MatroskaTrack *track, AVStream *st,
2983 uint8_t *data, int pkt_size,
2984 uint64_t timecode, uint64_t lace_duration,
2985 int64_t pos, int is_keyframe,
2986 uint8_t *additional, uint64_t additional_id, int additional_size,
2987 int64_t discard_padding)
2989 MatroskaTrackEncoding *encodings = track->encodings.elem;
2990 uint8_t *pkt_data = data;
2991 int offset = 0, res;
2994 if (encodings && !encodings->type && encodings->scope & 1) {
2995 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
3000 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3002 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
3004 av_log(matroska->ctx, AV_LOG_ERROR,
3005 "Error parsing a wavpack block.\n");
3008 if (pkt_data != data)
3009 av_freep(&pkt_data);
3013 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3014 AV_RB32(&data[4]) != MKBETAG('i', 'c', 'p', 'f'))
3017 pkt = av_mallocz(sizeof(AVPacket));
3019 if (pkt_data != data)
3020 av_freep(&pkt_data);
3021 return AVERROR(ENOMEM);
3023 /* XXX: prevent data copy... */
3024 if (av_new_packet(pkt, pkt_size + offset) < 0) {
3026 res = AVERROR(ENOMEM);
3030 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES && offset == 8) {
3031 uint8_t *buf = pkt->data;
3032 bytestream_put_be32(&buf, pkt_size);
3033 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
3036 memcpy(pkt->data + offset, pkt_data, pkt_size);
3038 if (pkt_data != data)
3039 av_freep(&pkt_data);
3041 pkt->flags = is_keyframe;
3042 pkt->stream_index = st->index;
3044 if (additional_size > 0) {
3045 uint8_t *side_data = av_packet_new_side_data(pkt,
3046 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3047 additional_size + 8);
3049 av_packet_unref(pkt);
3051 return AVERROR(ENOMEM);
3053 AV_WB64(side_data, additional_id);
3054 memcpy(side_data + 8, additional, additional_size);
3057 if (discard_padding) {
3058 uint8_t *side_data = av_packet_new_side_data(pkt,
3059 AV_PKT_DATA_SKIP_SAMPLES,
3062 av_packet_unref(pkt);
3064 return AVERROR(ENOMEM);
3066 AV_WL32(side_data, 0);
3067 AV_WL32(side_data + 4, av_rescale_q(discard_padding,
3068 (AVRational){1, 1000000000},
3069 (AVRational){1, st->codecpar->sample_rate}));
3072 if (track->ms_compat)
3073 pkt->dts = timecode;
3075 pkt->pts = timecode;
3077 pkt->duration = lace_duration;
3079 #if FF_API_CONVERGENCE_DURATION
3080 FF_DISABLE_DEPRECATION_WARNINGS
3081 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3082 pkt->convergence_duration = lace_duration;
3084 FF_ENABLE_DEPRECATION_WARNINGS
3087 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
3088 matroska->prev_pkt = pkt;
3093 if (pkt_data != data)
3094 av_freep(&pkt_data);
3098 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
3099 int size, int64_t pos, uint64_t cluster_time,
3100 uint64_t block_duration, int is_keyframe,
3101 uint8_t *additional, uint64_t additional_id, int additional_size,
3102 int64_t cluster_pos, int64_t discard_padding)
3104 uint64_t timecode = AV_NOPTS_VALUE;
3105 MatroskaTrack *track;
3109 uint32_t *lace_size = NULL;
3110 int n, flags, laces = 0;
3112 int trust_default_duration = 1;
3114 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
3115 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
3121 track = matroska_find_track_by_num(matroska, num);
3122 if (!track || !track->stream) {
3123 av_log(matroska->ctx, AV_LOG_INFO,
3124 "Invalid stream %"PRIu64" or size %u\n", num, size);
3125 return AVERROR_INVALIDDATA;
3126 } else if (size <= 3)
3129 if (st->discard >= AVDISCARD_ALL)
3131 av_assert1(block_duration != AV_NOPTS_VALUE);
3133 block_time = sign_extend(AV_RB16(data), 16);
3137 if (is_keyframe == -1)
3138 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3140 if (cluster_time != (uint64_t) -1 &&
3141 (block_time >= 0 || cluster_time >= -block_time)) {
3142 timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3143 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3144 timecode < track->end_timecode)
3145 is_keyframe = 0; /* overlapping subtitles are not key frame */
3147 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3151 if (matroska->skip_to_keyframe &&
3152 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3153 if (timecode < matroska->skip_to_timecode)
3156 matroska->skip_to_keyframe = 0;
3157 else if (!st->skip_to_keyframe) {
3158 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3159 matroska->skip_to_keyframe = 0;
3163 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
3164 &lace_size, &laces);
3169 if (track->audio.samplerate == 8000) {
3170 // If this is needed for more codecs, then add them here
3171 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3172 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3173 trust_default_duration = 0;
3177 if (!block_duration && trust_default_duration)
3178 block_duration = track->default_duration * laces / matroska->time_scale;
3180 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3181 track->end_timecode =
3182 FFMAX(track->end_timecode, timecode + block_duration);
3184 for (n = 0; n < laces; n++) {
3185 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3187 if (lace_size[n] > size) {
3188 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
3192 if ((st->codecpar->codec_id == AV_CODEC_ID_RA_288 ||
3193 st->codecpar->codec_id == AV_CODEC_ID_COOK ||
3194 st->codecpar->codec_id == AV_CODEC_ID_SIPR ||
3195 st->codecpar->codec_id == AV_CODEC_ID_ATRAC3) &&
3196 st->codecpar->block_align && track->audio.sub_packet_size) {
3197 res = matroska_parse_rm_audio(matroska, track, st, data,
3203 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3204 res = matroska_parse_webvtt(matroska, track, st,
3206 timecode, lace_duration,
3211 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
3212 timecode, lace_duration, pos,
3213 !n ? is_keyframe : 0,
3214 additional, additional_id, additional_size,
3220 if (timecode != AV_NOPTS_VALUE)
3221 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3222 data += lace_size[n];
3223 size -= lace_size[n];
3231 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
3233 EbmlList *blocks_list;
3234 MatroskaBlock *blocks;
3236 res = ebml_parse(matroska,
3237 matroska_cluster_incremental_parsing,
3238 &matroska->current_cluster);
3241 if (matroska->current_cluster_pos)
3242 ebml_level_end(matroska);
3243 ebml_free(matroska_cluster, &matroska->current_cluster);
3244 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
3245 matroska->current_cluster_num_blocks = 0;
3246 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
3247 matroska->prev_pkt = NULL;
3248 /* sizeof the ID which was already read */
3249 if (matroska->current_id)
3250 matroska->current_cluster_pos -= 4;
3251 res = ebml_parse(matroska,
3252 matroska_clusters_incremental,
3253 &matroska->current_cluster);
3254 /* Try parsing the block again. */
3256 res = ebml_parse(matroska,
3257 matroska_cluster_incremental_parsing,
3258 &matroska->current_cluster);
3262 matroska->current_cluster_num_blocks <
3263 matroska->current_cluster.blocks.nb_elem) {
3264 blocks_list = &matroska->current_cluster.blocks;
3265 blocks = blocks_list->elem;
3267 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
3268 i = blocks_list->nb_elem - 1;
3269 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
3270 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
3271 uint8_t* additional = blocks[i].additional.size > 0 ?
3272 blocks[i].additional.data : NULL;
3273 if (!blocks[i].non_simple)
3274 blocks[i].duration = 0;
3275 res = matroska_parse_block(matroska, blocks[i].bin.data,
3276 blocks[i].bin.size, blocks[i].bin.pos,
3277 matroska->current_cluster.timecode,
3278 blocks[i].duration, is_keyframe,
3279 additional, blocks[i].additional_id,
3280 blocks[i].additional.size,
3281 matroska->current_cluster_pos,
3282 blocks[i].discard_padding);
3289 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3291 MatroskaCluster cluster = { 0 };
3292 EbmlList *blocks_list;
3293 MatroskaBlock *blocks;
3297 if (!matroska->contains_ssa)
3298 return matroska_parse_cluster_incremental(matroska);
3299 pos = avio_tell(matroska->ctx->pb);
3300 matroska->prev_pkt = NULL;
3301 if (matroska->current_id)
3302 pos -= 4; /* sizeof the ID which was already read */
3303 res = ebml_parse(matroska, matroska_clusters, &cluster);
3304 blocks_list = &cluster.blocks;
3305 blocks = blocks_list->elem;
3306 for (i = 0; i < blocks_list->nb_elem; i++)
3307 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
3308 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
3309 res = matroska_parse_block(matroska, blocks[i].bin.data,
3310 blocks[i].bin.size, blocks[i].bin.pos,
3311 cluster.timecode, blocks[i].duration,
3312 is_keyframe, NULL, 0, 0, pos,
3313 blocks[i].discard_padding);
3315 ebml_free(matroska_cluster, &cluster);
3319 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3321 MatroskaDemuxContext *matroska = s->priv_data;
3323 while (matroska_deliver_packet(matroska, pkt)) {
3324 int64_t pos = avio_tell(matroska->ctx->pb);
3327 if (matroska_parse_cluster(matroska) < 0)
3328 matroska_resync(matroska, pos);
3334 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3335 int64_t timestamp, int flags)
3337 MatroskaDemuxContext *matroska = s->priv_data;
3338 MatroskaTrack *tracks = NULL;
3339 AVStream *st = s->streams[stream_index];
3340 int i, index, index_sub, index_min;
3342 /* Parse the CUES now since we need the index data to seek. */
3343 if (matroska->cues_parsing_deferred > 0) {
3344 matroska->cues_parsing_deferred = 0;
3345 matroska_parse_cues(matroska);
3348 if (!st->nb_index_entries)
3350 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3352 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3353 avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
3355 matroska->current_id = 0;
3356 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3357 matroska_clear_queue(matroska);
3358 if (matroska_parse_cluster(matroska) < 0)
3363 matroska_clear_queue(matroska);
3364 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3368 tracks = matroska->tracks.elem;
3369 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3370 tracks[i].audio.pkt_cnt = 0;
3371 tracks[i].audio.sub_packet_cnt = 0;
3372 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3373 tracks[i].end_timecode = 0;
3374 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3376 tracks[i].stream->discard != AVDISCARD_ALL) {
3377 index_sub = av_index_search_timestamp(
3378 tracks[i].stream, st->index_entries[index].timestamp,
3379 AVSEEK_FLAG_BACKWARD);
3380 while (index_sub >= 0 &&
3382 tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
3383 st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
3388 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
3389 matroska->current_id = 0;
3390 if (flags & AVSEEK_FLAG_ANY) {
3391 st->skip_to_keyframe = 0;
3392 matroska->skip_to_timecode = timestamp;
3394 st->skip_to_keyframe = 1;
3395 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3397 matroska->skip_to_keyframe = 1;
3399 matroska->num_levels = 0;
3400 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3403 // slightly hackish but allows proper fallback to
3404 // the generic seeking code.
3405 matroska_clear_queue(matroska);
3406 matroska->current_id = 0;
3407 st->skip_to_keyframe =
3408 matroska->skip_to_keyframe = 0;
3410 matroska->num_levels = 0;
3414 static int matroska_read_close(AVFormatContext *s)
3416 MatroskaDemuxContext *matroska = s->priv_data;
3417 MatroskaTrack *tracks = matroska->tracks.elem;
3420 matroska_clear_queue(matroska);
3422 for (n = 0; n < matroska->tracks.nb_elem; n++)
3423 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3424 av_freep(&tracks[n].audio.buf);
3425 ebml_free(matroska_cluster, &matroska->current_cluster);
3426 ebml_free(matroska_segment, matroska);
3432 int64_t start_time_ns;
3433 int64_t end_time_ns;
3434 int64_t start_offset;
3438 /* This function searches all the Cues and returns the CueDesc corresponding the
3439 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3440 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3442 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3443 MatroskaDemuxContext *matroska = s->priv_data;
3446 int nb_index_entries = s->streams[0]->nb_index_entries;
3447 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3448 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3449 for (i = 1; i < nb_index_entries; i++) {
3450 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3451 index_entries[i].timestamp * matroska->time_scale > ts) {
3456 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3457 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3458 if (i != nb_index_entries - 1) {
3459 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3460 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3462 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3463 // FIXME: this needs special handling for files where Cues appear
3464 // before Clusters. the current logic assumes Cues appear after
3466 cue_desc.end_offset = cues_start - matroska->segment_start;
3471 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3473 MatroskaDemuxContext *matroska = s->priv_data;
3474 int64_t cluster_pos, before_pos;
3476 if (s->streams[0]->nb_index_entries <= 0) return 0;
3477 // seek to the first cluster using cues.
3478 index = av_index_search_timestamp(s->streams[0], 0, 0);
3479 if (index < 0) return 0;
3480 cluster_pos = s->streams[0]->index_entries[index].pos;
3481 before_pos = avio_tell(s->pb);
3483 int64_t cluster_id = 0, cluster_length = 0;
3485 avio_seek(s->pb, cluster_pos, SEEK_SET);
3486 // read cluster id and length
3487 ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id);
3488 ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3489 if (cluster_id != 0xF43B675) { // done with all clusters
3492 avio_seek(s->pb, cluster_pos, SEEK_SET);
3493 matroska->current_id = 0;
3494 matroska_clear_queue(matroska);
3495 if (matroska_parse_cluster(matroska) < 0 ||
3496 matroska->num_packets <= 0) {
3499 pkt = matroska->packets[0];
3500 cluster_pos += cluster_length + 12; // 12 is the offset of the cluster id and length.
3501 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3506 avio_seek(s->pb, before_pos, SEEK_SET);
3510 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3511 double min_buffer, double* buffer,
3512 double* sec_to_download, AVFormatContext *s,
3515 double nano_seconds_per_second = 1000000000.0;
3516 double time_sec = time_ns / nano_seconds_per_second;
3518 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3519 int64_t end_time_ns = time_ns + time_to_search_ns;
3520 double sec_downloaded = 0.0;
3521 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3522 if (desc_curr.start_time_ns == -1)
3524 *sec_to_download = 0.0;
3526 // Check for non cue start time.
3527 if (time_ns > desc_curr.start_time_ns) {
3528 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3529 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3530 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3531 double timeToDownload = (cueBytes * 8.0) / bps;
3533 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3534 *sec_to_download += timeToDownload;
3536 // Check if the search ends within the first cue.
3537 if (desc_curr.end_time_ns >= end_time_ns) {
3538 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3539 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3540 sec_downloaded = percent_to_sub * sec_downloaded;
3541 *sec_to_download = percent_to_sub * *sec_to_download;
3544 if ((sec_downloaded + *buffer) <= min_buffer) {
3548 // Get the next Cue.
3549 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3552 while (desc_curr.start_time_ns != -1) {
3553 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3554 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3555 double desc_sec = desc_ns / nano_seconds_per_second;
3556 double bits = (desc_bytes * 8.0);
3557 double time_to_download = bits / bps;
3559 sec_downloaded += desc_sec - time_to_download;
3560 *sec_to_download += time_to_download;
3562 if (desc_curr.end_time_ns >= end_time_ns) {
3563 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3564 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3565 sec_downloaded = percent_to_sub * sec_downloaded;
3566 *sec_to_download = percent_to_sub * *sec_to_download;
3568 if ((sec_downloaded + *buffer) <= min_buffer)
3573 if ((sec_downloaded + *buffer) <= min_buffer) {
3578 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3580 *buffer = *buffer + sec_downloaded;
3584 /* This function computes the bandwidth of the WebM file with the help of
3585 * buffer_size_after_time_downloaded() function. Both of these functions are
3586 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3587 * Matroska parsing mechanism.
3589 * Returns the bandwidth of the file on success; -1 on error.
3591 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3593 MatroskaDemuxContext *matroska = s->priv_data;
3594 AVStream *st = s->streams[0];
3595 double bandwidth = 0.0;
3598 for (i = 0; i < st->nb_index_entries; i++) {
3599 int64_t prebuffer_ns = 1000000000;
3600 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3601 double nano_seconds_per_second = 1000000000.0;
3602 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3603 double prebuffer_bytes = 0.0;
3604 int64_t temp_prebuffer_ns = prebuffer_ns;
3605 int64_t pre_bytes, pre_ns;
3606 double pre_sec, prebuffer, bits_per_second;
3607 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3609 // Start with the first Cue.
3610 CueDesc desc_end = desc_beg;
3612 // Figure out how much data we have downloaded for the prebuffer. This will
3613 // be used later to adjust the bits per sample to try.
3614 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3615 // Prebuffered the entire Cue.
3616 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3617 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3618 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3620 if (desc_end.start_time_ns == -1) {
3621 // The prebuffer is larger than the duration.
3622 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3624 bits_per_second = 0.0;
3626 // The prebuffer ends in the last Cue. Estimate how much data was
3628 pre_bytes = desc_end.end_offset - desc_end.start_offset;
3629 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3630 pre_sec = pre_ns / nano_seconds_per_second;
3632 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3634 prebuffer = prebuffer_ns / nano_seconds_per_second;
3636 // Set this to 0.0 in case our prebuffer buffers the entire video.
3637 bits_per_second = 0.0;
3639 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3640 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3641 double desc_sec = desc_ns / nano_seconds_per_second;
3642 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3644 // Drop the bps by the percentage of bytes buffered.
3645 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3646 double mod_bits_per_second = calc_bits_per_second * percent;
3648 if (prebuffer < desc_sec) {
3650 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3652 // Add 1 so the bits per second should be a little bit greater than file
3654 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3655 const double min_buffer = 0.0;
3656 double buffer = prebuffer;
3657 double sec_to_download = 0.0;
3659 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3660 min_buffer, &buffer, &sec_to_download,
3664 } else if (rv == 0) {
3665 bits_per_second = (double)(bps);
3670 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3671 } while (desc_end.start_time_ns != -1);
3673 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
3675 return (int64_t)bandwidth;
3678 static int webm_dash_manifest_cues(AVFormatContext *s)
3680 MatroskaDemuxContext *matroska = s->priv_data;
3681 EbmlList *seekhead_list = &matroska->seekhead;
3682 MatroskaSeekhead *seekhead = seekhead_list->elem;
3684 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
3687 // determine cues start and end positions
3688 for (i = 0; i < seekhead_list->nb_elem; i++)
3689 if (seekhead[i].id == MATROSKA_ID_CUES)
3692 if (i >= seekhead_list->nb_elem) return -1;
3694 before_pos = avio_tell(matroska->ctx->pb);
3695 cues_start = seekhead[i].pos + matroska->segment_start;
3696 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
3697 // cues_end is computed as cues_start + cues_length + length of the
3698 // Cues element ID + EBML length of the Cues element. cues_end is
3699 // inclusive and the above sum is reduced by 1.
3700 uint64_t cues_length = 0, cues_id = 0, bytes_read = 0;
3701 bytes_read += ebml_read_num(matroska, matroska->ctx->pb, 4, &cues_id);
3702 bytes_read += ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
3703 cues_end = cues_start + cues_length + bytes_read - 1;
3705 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
3706 if (cues_start == -1 || cues_end == -1) return -1;
3709 matroska_parse_cues(matroska);
3712 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
3715 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
3718 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
3719 if (bandwidth < 0) return -1;
3720 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
3722 // check if all clusters start with key frames
3723 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
3725 // store cue point timestamps as a comma separated list for checking subsegment alignment in
3726 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
3727 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20 * sizeof(char));
3728 if (!buf) return -1;
3730 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
3731 snprintf(buf, (i + 1) * 20 * sizeof(char),
3732 "%s%" PRId64, buf, s->streams[0]->index_entries[i].timestamp);
3733 if (i != s->streams[0]->nb_index_entries - 1)
3734 strncat(buf, ",", sizeof(char));
3736 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
3742 static int webm_dash_manifest_read_header(AVFormatContext *s)
3745 int ret = matroska_read_header(s);
3746 MatroskaTrack *tracks;
3747 MatroskaDemuxContext *matroska = s->priv_data;
3749 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
3753 if (!matroska->is_live) {
3754 buf = av_asprintf("%g", matroska->duration);
3755 if (!buf) return AVERROR(ENOMEM);
3756 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
3759 // initialization range
3760 // 5 is the offset of Cluster ID.
3761 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, avio_tell(s->pb) - 5, 0);
3764 // basename of the file
3765 buf = strrchr(s->filename, '/');
3766 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->filename, 0);
3769 tracks = matroska->tracks.elem;
3770 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
3772 // parse the cues and populate Cue related fields
3773 return matroska->is_live ? 0 : webm_dash_manifest_cues(s);
3776 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
3781 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
3782 static const AVOption options[] = {
3783 { "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 },
3787 static const AVClass webm_dash_class = {
3788 .class_name = "WebM DASH Manifest demuxer",
3789 .item_name = av_default_item_name,
3791 .version = LIBAVUTIL_VERSION_INT,
3794 AVInputFormat ff_matroska_demuxer = {
3795 .name = "matroska,webm",
3796 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
3797 .extensions = "mkv,mk3d,mka,mks",
3798 .priv_data_size = sizeof(MatroskaDemuxContext),
3799 .read_probe = matroska_probe,
3800 .read_header = matroska_read_header,
3801 .read_packet = matroska_read_packet,
3802 .read_close = matroska_read_close,
3803 .read_seek = matroska_read_seek,
3804 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
3807 AVInputFormat ff_webm_dash_manifest_demuxer = {
3808 .name = "webm_dash_manifest",
3809 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
3810 .priv_data_size = sizeof(MatroskaDemuxContext),
3811 .read_header = webm_dash_manifest_read_header,
3812 .read_packet = webm_dash_manifest_read_packet,
3813 .read_close = matroska_read_close,
3814 .priv_class = &webm_dash_class,