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;
171 uint64_t display_unit;
173 uint64_t field_order;
174 uint64_t stereo_mode;
176 MatroskaTrackVideoColor color;
177 } MatroskaTrackVideo;
179 typedef struct MatroskaTrackAudio {
181 double out_samplerate;
185 /* real audio header (extracted from extradata) */
192 uint64_t buf_timecode;
194 } MatroskaTrackAudio;
196 typedef struct MatroskaTrackPlane {
199 } MatroskaTrackPlane;
201 typedef struct MatroskaTrackOperation {
202 EbmlList combine_planes;
203 } MatroskaTrackOperation;
205 typedef struct MatroskaTrack {
214 uint64_t default_duration;
215 uint64_t flag_default;
216 uint64_t flag_forced;
217 uint64_t seek_preroll;
218 MatroskaTrackVideo video;
219 MatroskaTrackAudio audio;
220 MatroskaTrackOperation operation;
222 uint64_t codec_delay;
223 uint64_t codec_delay_in_track_tb;
226 int64_t end_timecode;
228 uint64_t max_block_additional_id;
230 uint32_t palette[AVPALETTE_COUNT];
234 typedef struct MatroskaAttachment {
241 } MatroskaAttachment;
243 typedef struct MatroskaChapter {
252 typedef struct MatroskaIndexPos {
257 typedef struct MatroskaIndex {
262 typedef struct MatroskaTag {
270 typedef struct MatroskaTagTarget {
278 typedef struct MatroskaTags {
279 MatroskaTagTarget target;
283 typedef struct MatroskaSeekhead {
288 typedef struct MatroskaLevel {
293 typedef struct MatroskaCluster {
298 typedef struct MatroskaLevel1Element {
302 } MatroskaLevel1Element;
304 typedef struct MatroskaDemuxContext {
305 const AVClass *class;
306 AVFormatContext *ctx;
310 MatroskaLevel levels[EBML_MAX_DEPTH];
320 EbmlList attachments;
326 /* byte position of the segment inside the stream */
327 int64_t segment_start;
329 /* the packet queue */
336 /* What to skip before effectively reading a packet. */
337 int skip_to_keyframe;
338 uint64_t skip_to_timecode;
340 /* File has a CUES element, but we defer parsing until it is needed. */
341 int cues_parsing_deferred;
343 /* Level1 elements and whether they were read yet */
344 MatroskaLevel1Element level1_elems[64];
345 int num_level1_elems;
347 int current_cluster_num_blocks;
348 int64_t current_cluster_pos;
349 MatroskaCluster current_cluster;
351 /* File has SSA subtitles which prevent incremental cluster parsing. */
354 /* WebM DASH Manifest live flag/ */
356 } MatroskaDemuxContext;
358 typedef struct MatroskaBlock {
363 uint64_t additional_id;
365 int64_t discard_padding;
368 static const EbmlSyntax ebml_header[] = {
369 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
370 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
371 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
372 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
373 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
374 { EBML_ID_EBMLVERSION, EBML_NONE },
375 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
379 static const EbmlSyntax ebml_syntax[] = {
380 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
384 static const EbmlSyntax matroska_info[] = {
385 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
386 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
387 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
388 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
389 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
390 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
391 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
395 static const EbmlSyntax matroska_mastering_meta[] = {
396 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
397 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
398 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
399 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
400 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
401 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
402 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
403 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
404 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
405 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
409 static const EbmlSyntax matroska_track_video_color[] = {
410 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
411 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=0 } },
412 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
413 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
414 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
415 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
416 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
417 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
418 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
419 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
420 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
421 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
422 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
423 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
427 static const EbmlSyntax matroska_track_video[] = {
428 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
429 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
430 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
431 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
432 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
433 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
434 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
435 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, 0, offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
436 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
437 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
438 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
439 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
440 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
441 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
442 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
443 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
444 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
448 static const EbmlSyntax matroska_track_audio[] = {
449 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
450 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
451 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
452 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
456 static const EbmlSyntax matroska_track_encoding_compression[] = {
457 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
458 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
462 static const EbmlSyntax matroska_track_encoding_encryption[] = {
463 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
464 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
465 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
466 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
467 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
468 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
469 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
472 static const EbmlSyntax matroska_track_encoding[] = {
473 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
474 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
475 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
476 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
477 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
481 static const EbmlSyntax matroska_track_encodings[] = {
482 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
486 static const EbmlSyntax matroska_track_plane[] = {
487 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
488 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
492 static const EbmlSyntax matroska_track_combine_planes[] = {
493 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
497 static const EbmlSyntax matroska_track_operation[] = {
498 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
502 static const EbmlSyntax matroska_track[] = {
503 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
504 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
505 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
506 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
507 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
508 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
509 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
510 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
511 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
512 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
513 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
514 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
515 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
516 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
517 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
518 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
519 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
520 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
521 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
522 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
523 { MATROSKA_ID_CODECNAME, EBML_NONE },
524 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
525 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
526 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
527 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
528 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
532 static const EbmlSyntax matroska_tracks[] = {
533 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
537 static const EbmlSyntax matroska_attachment[] = {
538 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
539 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
540 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
541 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
542 { MATROSKA_ID_FILEDESC, EBML_NONE },
546 static const EbmlSyntax matroska_attachments[] = {
547 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
551 static const EbmlSyntax matroska_chapter_display[] = {
552 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
553 { MATROSKA_ID_CHAPLANG, EBML_NONE },
554 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
558 static const EbmlSyntax matroska_chapter_entry[] = {
559 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
560 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
561 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
562 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
563 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
564 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
565 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
566 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
570 static const EbmlSyntax matroska_chapter[] = {
571 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
572 { MATROSKA_ID_EDITIONUID, EBML_NONE },
573 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
574 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
575 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
579 static const EbmlSyntax matroska_chapters[] = {
580 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
584 static const EbmlSyntax matroska_index_pos[] = {
585 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
586 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
587 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
588 { MATROSKA_ID_CUEDURATION, EBML_NONE },
589 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
593 static const EbmlSyntax matroska_index_entry[] = {
594 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
595 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
599 static const EbmlSyntax matroska_index[] = {
600 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
604 static const EbmlSyntax matroska_simpletag[] = {
605 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
606 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
607 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
608 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
609 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
610 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
614 static const EbmlSyntax matroska_tagtargets[] = {
615 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
616 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
617 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
618 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
619 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
623 static const EbmlSyntax matroska_tag[] = {
624 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
625 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
629 static const EbmlSyntax matroska_tags[] = {
630 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
634 static const EbmlSyntax matroska_seekhead_entry[] = {
635 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
636 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
640 static const EbmlSyntax matroska_seekhead[] = {
641 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
645 static const EbmlSyntax matroska_segment[] = {
646 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
647 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
648 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
649 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
650 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
651 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
652 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
653 { MATROSKA_ID_CLUSTER, EBML_STOP },
657 static const EbmlSyntax matroska_segments[] = {
658 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
662 static const EbmlSyntax matroska_blockmore[] = {
663 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
664 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
668 static const EbmlSyntax matroska_blockadditions[] = {
669 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
673 static const EbmlSyntax matroska_blockgroup[] = {
674 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
675 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
676 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
677 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
678 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
679 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference) },
680 { MATROSKA_ID_CODECSTATE, EBML_NONE },
681 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
685 static const EbmlSyntax matroska_cluster[] = {
686 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
687 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
688 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
689 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
690 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
694 static const EbmlSyntax matroska_clusters[] = {
695 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
696 { MATROSKA_ID_INFO, EBML_NONE },
697 { MATROSKA_ID_CUES, EBML_NONE },
698 { MATROSKA_ID_TAGS, EBML_NONE },
699 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
703 static const EbmlSyntax matroska_cluster_incremental_parsing[] = {
704 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
705 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
706 { MATROSKA_ID_SIMPLEBLOCK, EBML_PASS, sizeof(MatroskaBlock), offsetof(MatroskaCluster, blocks), { .n = matroska_blockgroup } },
707 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
708 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
709 { MATROSKA_ID_INFO, EBML_NONE },
710 { MATROSKA_ID_CUES, EBML_NONE },
711 { MATROSKA_ID_TAGS, EBML_NONE },
712 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
713 { MATROSKA_ID_CLUSTER, EBML_STOP },
717 static const EbmlSyntax matroska_cluster_incremental[] = {
718 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
719 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
720 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
721 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
722 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
726 static const EbmlSyntax matroska_clusters_incremental[] = {
727 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster_incremental } },
728 { MATROSKA_ID_INFO, EBML_NONE },
729 { MATROSKA_ID_CUES, EBML_NONE },
730 { MATROSKA_ID_TAGS, EBML_NONE },
731 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
735 static const char *const matroska_doctypes[] = { "matroska", "webm" };
737 static int matroska_read_close(AVFormatContext *s);
739 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
741 AVIOContext *pb = matroska->ctx->pb;
744 matroska->current_id = 0;
745 matroska->num_levels = 0;
747 /* seek to next position to resync from */
748 if ((ret = avio_seek(pb, last_pos + 1, SEEK_SET)) < 0) {
755 // try to find a toplevel element
756 while (!avio_feof(pb)) {
757 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
758 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
759 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
760 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
761 matroska->current_id = id;
764 id = (id << 8) | avio_r8(pb);
772 * Return: Whether we reached the end of a level in the hierarchy or not.
774 static int ebml_level_end(MatroskaDemuxContext *matroska)
776 AVIOContext *pb = matroska->ctx->pb;
777 int64_t pos = avio_tell(pb);
779 if (matroska->num_levels > 0) {
780 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
781 if (pos - level->start >= level->length || matroska->current_id) {
782 matroska->num_levels--;
786 return (matroska->is_live && matroska->ctx->pb->eof_reached) ? 1 : 0;
790 * Read: an "EBML number", which is defined as a variable-length
791 * array of bytes. The first byte indicates the length by giving a
792 * number of 0-bits followed by a one. The position of the first
793 * "one" bit inside the first byte indicates the length of this
795 * Returns: number of bytes read, < 0 on error
797 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
798 int max_size, uint64_t *number)
803 /* The first byte tells us the length in bytes - avio_r8() can normally
804 * return 0, but since that's not a valid first ebmlID byte, we can
805 * use it safely here to catch EOS. */
806 if (!(total = avio_r8(pb))) {
807 /* we might encounter EOS here */
808 if (!avio_feof(pb)) {
809 int64_t pos = avio_tell(pb);
810 av_log(matroska->ctx, AV_LOG_ERROR,
811 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
813 return pb->error ? pb->error : AVERROR(EIO);
818 /* get the length of the EBML number */
819 read = 8 - ff_log2_tab[total];
820 if (read > max_size) {
821 int64_t pos = avio_tell(pb) - 1;
822 av_log(matroska->ctx, AV_LOG_ERROR,
823 "Invalid EBML number size tag 0x%02x at pos %"PRIu64" (0x%"PRIx64")\n",
824 (uint8_t) total, pos, pos);
825 return AVERROR_INVALIDDATA;
828 /* read out length */
829 total ^= 1 << ff_log2_tab[total];
831 total = (total << 8) | avio_r8(pb);
839 * Read a EBML length value.
840 * This needs special handling for the "unknown length" case which has multiple
843 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
846 int res = ebml_read_num(matroska, pb, 8, number);
847 if (res > 0 && *number + 1 == 1ULL << (7 * res))
848 *number = 0xffffffffffffffULL;
853 * Read the next element as an unsigned int.
854 * 0 is success, < 0 is failure.
856 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
861 return AVERROR_INVALIDDATA;
863 /* big-endian ordering; build up number */
866 *num = (*num << 8) | avio_r8(pb);
872 * Read the next element as a signed int.
873 * 0 is success, < 0 is failure.
875 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
880 return AVERROR_INVALIDDATA;
885 *num = sign_extend(avio_r8(pb), 8);
887 /* big-endian ordering; build up number */
889 *num = ((uint64_t)*num << 8) | avio_r8(pb);
896 * Read the next element as a float.
897 * 0 is success, < 0 is failure.
899 static int ebml_read_float(AVIOContext *pb, int size, double *num)
904 *num = av_int2float(avio_rb32(pb));
906 *num = av_int2double(avio_rb64(pb));
908 return AVERROR_INVALIDDATA;
914 * Read the next element as an ASCII string.
915 * 0 is success, < 0 is failure.
917 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
921 /* EBML strings are usually not 0-terminated, so we allocate one
922 * byte more, read the string and NULL-terminate it ourselves. */
923 if (!(res = av_malloc(size + 1)))
924 return AVERROR(ENOMEM);
925 if (avio_read(pb, (uint8_t *) res, size) != size) {
937 * Read the next element as binary data.
938 * 0 is success, < 0 is failure.
940 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
942 av_fast_padded_malloc(&bin->data, &bin->size, length);
944 return AVERROR(ENOMEM);
947 bin->pos = avio_tell(pb);
948 if (avio_read(pb, bin->data, length) != length) {
949 av_freep(&bin->data);
958 * Read the next element, but only the header. The contents
959 * are supposed to be sub-elements which can be read separately.
960 * 0 is success, < 0 is failure.
962 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
964 AVIOContext *pb = matroska->ctx->pb;
965 MatroskaLevel *level;
967 if (matroska->num_levels >= EBML_MAX_DEPTH) {
968 av_log(matroska->ctx, AV_LOG_ERROR,
969 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
970 return AVERROR(ENOSYS);
973 level = &matroska->levels[matroska->num_levels++];
974 level->start = avio_tell(pb);
975 level->length = length;
981 * Read signed/unsigned "EBML" numbers.
982 * Return: number of bytes processed, < 0 on error
984 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
985 uint8_t *data, uint32_t size, uint64_t *num)
988 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
989 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num);
993 * Same as above, but signed.
995 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
996 uint8_t *data, uint32_t size, int64_t *num)
1001 /* read as unsigned number first */
1002 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
1005 /* make signed (weird way) */
1006 *num = unum - ((1LL << (7 * res - 1)) - 1);
1011 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
1012 EbmlSyntax *syntax, void *data);
1014 static int ebml_parse_id(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1015 uint32_t id, void *data)
1018 for (i = 0; syntax[i].id; i++)
1019 if (id == syntax[i].id)
1021 if (!syntax[i].id && id == MATROSKA_ID_CLUSTER &&
1022 matroska->num_levels > 0 &&
1023 matroska->levels[matroska->num_levels - 1].length == 0xffffffffffffff)
1024 return 0; // we reached the end of an unknown size cluster
1025 if (!syntax[i].id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1026 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32"\n", id);
1028 return ebml_parse_elem(matroska, &syntax[i], data);
1031 static int ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1034 if (!matroska->current_id) {
1036 int res = ebml_read_num(matroska, matroska->ctx->pb, 4, &id);
1038 // in live mode, finish parsing if EOF is reached.
1039 return (matroska->is_live && matroska->ctx->pb->eof_reached &&
1040 res == AVERROR_EOF) ? 1 : res;
1042 matroska->current_id = id | 1 << 7 * res;
1044 return ebml_parse_id(matroska, syntax, matroska->current_id, data);
1047 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1052 for (i = 0; syntax[i].id; i++)
1053 switch (syntax[i].type) {
1055 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1058 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1062 // the default may be NULL
1063 if (syntax[i].def.s) {
1064 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1065 *dst = av_strdup(syntax[i].def.s);
1067 return AVERROR(ENOMEM);
1072 while (!res && !ebml_level_end(matroska))
1073 res = ebml_parse(matroska, syntax, data);
1078 static int is_ebml_id_valid(uint32_t id)
1080 // Due to endian nonsense in Matroska, the highest byte with any bits set
1081 // will contain the leading length bit. This bit in turn identifies the
1082 // total byte length of the element by its position within the byte.
1083 unsigned int bits = av_log2(id);
1084 return id && (bits + 7) / 8 == (8 - bits % 8);
1088 * Allocate and return the entry for the level1 element with the given ID. If
1089 * an entry already exists, return the existing entry.
1091 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1095 MatroskaLevel1Element *elem;
1097 if (!is_ebml_id_valid(id))
1100 // Some files link to all clusters; useless.
1101 if (id == MATROSKA_ID_CLUSTER)
1104 // There can be multiple seekheads.
1105 if (id != MATROSKA_ID_SEEKHEAD) {
1106 for (i = 0; i < matroska->num_level1_elems; i++) {
1107 if (matroska->level1_elems[i].id == id)
1108 return &matroska->level1_elems[i];
1112 // Only a completely broken file would have more elements.
1113 // It also provides a low-effort way to escape from circular seekheads
1114 // (every iteration will add a level1 entry).
1115 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1116 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1120 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1121 *elem = (MatroskaLevel1Element){.id = id};
1126 static int ebml_parse_elem(MatroskaDemuxContext *matroska,
1127 EbmlSyntax *syntax, void *data)
1129 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1132 // max. 16 MB for strings
1133 [EBML_STR] = 0x1000000,
1134 [EBML_UTF8] = 0x1000000,
1135 // max. 256 MB for binary data
1136 [EBML_BIN] = 0x10000000,
1137 // no limits for anything else
1139 AVIOContext *pb = matroska->ctx->pb;
1140 uint32_t id = syntax->id;
1144 MatroskaLevel1Element *level1_elem;
1146 data = (char *) data + syntax->data_offset;
1147 if (syntax->list_elem_size) {
1148 EbmlList *list = data;
1149 newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1151 return AVERROR(ENOMEM);
1152 list->elem = newelem;
1153 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1154 memset(data, 0, syntax->list_elem_size);
1158 if (syntax->type != EBML_PASS && syntax->type != EBML_STOP) {
1159 matroska->current_id = 0;
1160 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1162 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1163 av_log(matroska->ctx, AV_LOG_ERROR,
1164 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1165 length, max_lengths[syntax->type], syntax->type);
1166 return AVERROR_INVALIDDATA;
1170 switch (syntax->type) {
1172 res = ebml_read_uint(pb, length, data);
1175 res = ebml_read_sint(pb, length, data);
1178 res = ebml_read_float(pb, length, data);
1182 res = ebml_read_ascii(pb, length, data);
1185 res = ebml_read_binary(pb, length, data);
1189 if ((res = ebml_read_master(matroska, length)) < 0)
1191 if (id == MATROSKA_ID_SEGMENT)
1192 matroska->segment_start = avio_tell(matroska->ctx->pb);
1193 if (id == MATROSKA_ID_CUES)
1194 matroska->cues_parsing_deferred = 0;
1195 if (syntax->type == EBML_LEVEL1 &&
1196 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1197 if (level1_elem->parsed)
1198 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1199 level1_elem->parsed = 1;
1201 return ebml_parse_nest(matroska, syntax->def.n, data);
1203 return ebml_parse_id(matroska, syntax->def.n, id, data);
1207 if (ffio_limit(pb, length) != length)
1208 return AVERROR(EIO);
1209 return avio_skip(pb, length) < 0 ? AVERROR(EIO) : 0;
1211 if (res == AVERROR_INVALIDDATA)
1212 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1213 else if (res == AVERROR(EIO))
1214 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1218 static void ebml_free(EbmlSyntax *syntax, void *data)
1221 for (i = 0; syntax[i].id; i++) {
1222 void *data_off = (char *) data + syntax[i].data_offset;
1223 switch (syntax[i].type) {
1229 av_freep(&((EbmlBin *) data_off)->data);
1233 if (syntax[i].list_elem_size) {
1234 EbmlList *list = data_off;
1235 char *ptr = list->elem;
1236 for (j = 0; j < list->nb_elem;
1237 j++, ptr += syntax[i].list_elem_size)
1238 ebml_free(syntax[i].def.n, ptr);
1239 av_freep(&list->elem);
1241 ebml_free(syntax[i].def.n, data_off);
1251 static int matroska_probe(AVProbeData *p)
1254 int len_mask = 0x80, size = 1, n = 1, i;
1257 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1260 /* length of header */
1262 while (size <= 8 && !(total & len_mask)) {
1268 total &= (len_mask - 1);
1270 total = (total << 8) | p->buf[4 + n++];
1272 /* Does the probe data contain the whole header? */
1273 if (p->buf_size < 4 + size + total)
1276 /* The header should contain a known document type. For now,
1277 * we don't parse the whole header but simply check for the
1278 * availability of that array of characters inside the header.
1279 * Not fully fool-proof, but good enough. */
1280 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1281 size_t probelen = strlen(matroska_doctypes[i]);
1282 if (total < probelen)
1284 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1285 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1286 return AVPROBE_SCORE_MAX;
1289 // probably valid EBML header but no recognized doctype
1290 return AVPROBE_SCORE_EXTENSION;
1293 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1296 MatroskaTrack *tracks = matroska->tracks.elem;
1299 for (i = 0; i < matroska->tracks.nb_elem; i++)
1300 if (tracks[i].num == num)
1303 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1307 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1308 MatroskaTrack *track)
1310 MatroskaTrackEncoding *encodings = track->encodings.elem;
1311 uint8_t *data = *buf;
1312 int isize = *buf_size;
1313 uint8_t *pkt_data = NULL;
1314 uint8_t av_unused *newpktdata;
1315 int pkt_size = isize;
1319 if (pkt_size >= 10000000U)
1320 return AVERROR_INVALIDDATA;
1322 switch (encodings[0].compression.algo) {
1323 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1325 int header_size = encodings[0].compression.settings.size;
1326 uint8_t *header = encodings[0].compression.settings.data;
1328 if (header_size && !header) {
1329 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1336 pkt_size = isize + header_size;
1337 pkt_data = av_malloc(pkt_size);
1339 return AVERROR(ENOMEM);
1341 memcpy(pkt_data, header, header_size);
1342 memcpy(pkt_data + header_size, data, isize);
1346 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1348 olen = pkt_size *= 3;
1349 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING);
1351 result = AVERROR(ENOMEM);
1354 pkt_data = newpktdata;
1355 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1356 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1358 result = AVERROR_INVALIDDATA;
1365 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1367 z_stream zstream = { 0 };
1368 if (inflateInit(&zstream) != Z_OK)
1370 zstream.next_in = data;
1371 zstream.avail_in = isize;
1374 newpktdata = av_realloc(pkt_data, pkt_size);
1376 inflateEnd(&zstream);
1377 result = AVERROR(ENOMEM);
1380 pkt_data = newpktdata;
1381 zstream.avail_out = pkt_size - zstream.total_out;
1382 zstream.next_out = pkt_data + zstream.total_out;
1383 result = inflate(&zstream, Z_NO_FLUSH);
1384 } while (result == Z_OK && pkt_size < 10000000);
1385 pkt_size = zstream.total_out;
1386 inflateEnd(&zstream);
1387 if (result != Z_STREAM_END) {
1388 if (result == Z_MEM_ERROR)
1389 result = AVERROR(ENOMEM);
1391 result = AVERROR_INVALIDDATA;
1398 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1400 bz_stream bzstream = { 0 };
1401 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1403 bzstream.next_in = data;
1404 bzstream.avail_in = isize;
1407 newpktdata = av_realloc(pkt_data, pkt_size);
1409 BZ2_bzDecompressEnd(&bzstream);
1410 result = AVERROR(ENOMEM);
1413 pkt_data = newpktdata;
1414 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1415 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1416 result = BZ2_bzDecompress(&bzstream);
1417 } while (result == BZ_OK && pkt_size < 10000000);
1418 pkt_size = bzstream.total_out_lo32;
1419 BZ2_bzDecompressEnd(&bzstream);
1420 if (result != BZ_STREAM_END) {
1421 if (result == BZ_MEM_ERROR)
1422 result = AVERROR(ENOMEM);
1424 result = AVERROR_INVALIDDATA;
1431 return AVERROR_INVALIDDATA;
1435 *buf_size = pkt_size;
1443 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1444 AVDictionary **metadata, char *prefix)
1446 MatroskaTag *tags = list->elem;
1450 for (i = 0; i < list->nb_elem; i++) {
1451 const char *lang = tags[i].lang &&
1452 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1454 if (!tags[i].name) {
1455 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1459 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1461 av_strlcpy(key, tags[i].name, sizeof(key));
1462 if (tags[i].def || !lang) {
1463 av_dict_set(metadata, key, tags[i].string, 0);
1464 if (tags[i].sub.nb_elem)
1465 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1468 av_strlcat(key, "-", sizeof(key));
1469 av_strlcat(key, lang, sizeof(key));
1470 av_dict_set(metadata, key, tags[i].string, 0);
1471 if (tags[i].sub.nb_elem)
1472 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1475 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1478 static void matroska_convert_tags(AVFormatContext *s)
1480 MatroskaDemuxContext *matroska = s->priv_data;
1481 MatroskaTags *tags = matroska->tags.elem;
1484 for (i = 0; i < matroska->tags.nb_elem; i++) {
1485 if (tags[i].target.attachuid) {
1486 MatroskaAttachment *attachment = matroska->attachments.elem;
1488 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1489 if (attachment[j].uid == tags[i].target.attachuid &&
1490 attachment[j].stream) {
1491 matroska_convert_tag(s, &tags[i].tag,
1492 &attachment[j].stream->metadata, NULL);
1497 av_log(NULL, AV_LOG_WARNING,
1498 "The tags at index %d refer to a "
1499 "non-existent attachment %"PRId64".\n",
1500 i, tags[i].target.attachuid);
1502 } else if (tags[i].target.chapteruid) {
1503 MatroskaChapter *chapter = matroska->chapters.elem;
1505 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1506 if (chapter[j].uid == tags[i].target.chapteruid &&
1507 chapter[j].chapter) {
1508 matroska_convert_tag(s, &tags[i].tag,
1509 &chapter[j].chapter->metadata, NULL);
1514 av_log(NULL, AV_LOG_WARNING,
1515 "The tags at index %d refer to a non-existent chapter "
1517 i, tags[i].target.chapteruid);
1519 } else if (tags[i].target.trackuid) {
1520 MatroskaTrack *track = matroska->tracks.elem;
1522 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1523 if (track[j].uid == tags[i].target.trackuid &&
1525 matroska_convert_tag(s, &tags[i].tag,
1526 &track[j].stream->metadata, NULL);
1531 av_log(NULL, AV_LOG_WARNING,
1532 "The tags at index %d refer to a non-existent track "
1534 i, tags[i].target.trackuid);
1537 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1538 tags[i].target.type);
1543 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1546 uint32_t level_up = matroska->level_up;
1547 uint32_t saved_id = matroska->current_id;
1548 int64_t before_pos = avio_tell(matroska->ctx->pb);
1549 MatroskaLevel level;
1554 offset = pos + matroska->segment_start;
1555 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1556 /* We don't want to lose our seekhead level, so we add
1557 * a dummy. This is a crude hack. */
1558 if (matroska->num_levels == EBML_MAX_DEPTH) {
1559 av_log(matroska->ctx, AV_LOG_INFO,
1560 "Max EBML element depth (%d) reached, "
1561 "cannot parse further.\n", EBML_MAX_DEPTH);
1562 ret = AVERROR_INVALIDDATA;
1565 level.length = (uint64_t) -1;
1566 matroska->levels[matroska->num_levels] = level;
1567 matroska->num_levels++;
1568 matroska->current_id = 0;
1570 ret = ebml_parse(matroska, matroska_segment, matroska);
1572 /* remove dummy level */
1573 while (matroska->num_levels) {
1574 uint64_t length = matroska->levels[--matroska->num_levels].length;
1575 if (length == (uint64_t) -1)
1581 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
1582 matroska->level_up = level_up;
1583 matroska->current_id = saved_id;
1588 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1590 EbmlList *seekhead_list = &matroska->seekhead;
1593 // we should not do any seeking in the streaming case
1594 if (!matroska->ctx->pb->seekable)
1597 for (i = 0; i < seekhead_list->nb_elem; i++) {
1598 MatroskaSeekhead *seekheads = seekhead_list->elem;
1599 uint32_t id = seekheads[i].id;
1600 uint64_t pos = seekheads[i].pos;
1602 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1603 if (!elem || elem->parsed)
1608 // defer cues parsing until we actually need cue data.
1609 if (id == MATROSKA_ID_CUES)
1612 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1613 // mark index as broken
1614 matroska->cues_parsing_deferred = -1;
1622 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1624 EbmlList *index_list;
1625 MatroskaIndex *index;
1626 uint64_t index_scale = 1;
1629 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1632 index_list = &matroska->index;
1633 index = index_list->elem;
1634 if (index_list->nb_elem < 2)
1636 if (index[1].time > 1E14 / matroska->time_scale) {
1637 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1640 for (i = 0; i < index_list->nb_elem; i++) {
1641 EbmlList *pos_list = &index[i].pos;
1642 MatroskaIndexPos *pos = pos_list->elem;
1643 for (j = 0; j < pos_list->nb_elem; j++) {
1644 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1646 if (track && track->stream)
1647 av_add_index_entry(track->stream,
1648 pos[j].pos + matroska->segment_start,
1649 index[i].time / index_scale, 0, 0,
1655 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1658 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1661 for (i = 0; i < matroska->num_level1_elems; i++) {
1662 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1663 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1664 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1665 matroska->cues_parsing_deferred = -1;
1671 matroska_add_index_entries(matroska);
1674 static int matroska_aac_profile(char *codec_id)
1676 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1679 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1680 if (strstr(codec_id, aac_profiles[profile]))
1685 static int matroska_aac_sri(int samplerate)
1689 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1690 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1695 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1697 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1698 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1701 static int matroska_parse_flac(AVFormatContext *s,
1702 MatroskaTrack *track,
1705 AVStream *st = track->stream;
1706 uint8_t *p = track->codec_priv.data;
1707 int size = track->codec_priv.size;
1709 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1710 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1711 track->codec_priv.size = 0;
1715 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1717 p += track->codec_priv.size;
1718 size -= track->codec_priv.size;
1720 /* parse the remaining metadata blocks if present */
1722 int block_last, block_type, block_size;
1724 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1728 if (block_size > size)
1731 /* check for the channel mask */
1732 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1733 AVDictionary *dict = NULL;
1734 AVDictionaryEntry *chmask;
1736 ff_vorbis_comment(s, &dict, p, block_size, 0);
1737 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1739 uint64_t mask = strtol(chmask->value, NULL, 0);
1740 if (!mask || mask & ~0x3ffffULL) {
1741 av_log(s, AV_LOG_WARNING,
1742 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1744 st->codecpar->channel_layout = mask;
1746 av_dict_free(&dict);
1756 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
1758 int major, minor, micro, bttb = 0;
1760 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
1761 * this function, and fixed in 57.52 */
1762 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, µ) == 3)
1763 bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
1765 switch (field_order) {
1766 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
1767 return AV_FIELD_PROGRESSIVE;
1768 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
1769 return AV_FIELD_UNKNOWN;
1770 case MATROSKA_VIDEO_FIELDORDER_TT:
1772 case MATROSKA_VIDEO_FIELDORDER_BB:
1774 case MATROSKA_VIDEO_FIELDORDER_BT:
1775 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
1776 case MATROSKA_VIDEO_FIELDORDER_TB:
1777 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
1779 return AV_FIELD_UNKNOWN;
1783 static void mkv_stereo_mode_display_mul(int stereo_mode,
1784 int *h_width, int *h_height)
1786 switch (stereo_mode) {
1787 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
1788 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
1789 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
1790 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
1791 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
1793 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
1794 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
1795 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
1796 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
1799 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
1800 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
1801 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
1802 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
1808 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
1809 const MatroskaMasteringMeta* mastering_meta =
1810 &track->video.color.mastering_meta;
1811 // Mastering primaries are CIE 1931 coords, and must be > 0.
1812 const int has_mastering_primaries =
1813 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
1814 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
1815 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
1816 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
1817 const int has_mastering_luminance = mastering_meta->max_luminance > 0;
1819 if (track->video.color.matrix_coefficients != AVCOL_SPC_RESERVED)
1820 st->codecpar->color_space = track->video.color.matrix_coefficients;
1821 if (track->video.color.primaries != AVCOL_PRI_RESERVED &&
1822 track->video.color.primaries != AVCOL_PRI_RESERVED0)
1823 st->codecpar->color_primaries = track->video.color.primaries;
1824 if (track->video.color.transfer_characteristics != AVCOL_TRC_RESERVED &&
1825 track->video.color.transfer_characteristics != AVCOL_TRC_RESERVED0)
1826 st->codecpar->color_trc = track->video.color.transfer_characteristics;
1827 if (track->video.color.range != AVCOL_RANGE_UNSPECIFIED &&
1828 track->video.color.range <= AVCOL_RANGE_JPEG)
1829 st->codecpar->color_range = track->video.color.range;
1830 if (track->video.color.chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
1831 track->video.color.chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
1832 track->video.color.chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
1833 track->video.color.chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
1834 st->codecpar->chroma_location =
1835 avcodec_chroma_pos_to_enum((track->video.color.chroma_siting_horz - 1) << 7,
1836 (track->video.color.chroma_siting_vert - 1) << 7);
1839 if (has_mastering_primaries || has_mastering_luminance) {
1840 // Use similar rationals as other standards.
1841 const int chroma_den = 50000;
1842 const int luma_den = 10000;
1843 AVMasteringDisplayMetadata *metadata =
1844 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
1845 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
1846 sizeof(AVMasteringDisplayMetadata));
1848 return AVERROR(ENOMEM);
1850 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
1851 if (has_mastering_primaries) {
1852 metadata->display_primaries[0][0] = av_make_q(
1853 round(mastering_meta->r_x * chroma_den), chroma_den);
1854 metadata->display_primaries[0][1] = av_make_q(
1855 round(mastering_meta->r_y * chroma_den), chroma_den);
1856 metadata->display_primaries[1][0] = av_make_q(
1857 round(mastering_meta->g_x * chroma_den), chroma_den);
1858 metadata->display_primaries[1][1] = av_make_q(
1859 round(mastering_meta->g_y * chroma_den), chroma_den);
1860 metadata->display_primaries[2][0] = av_make_q(
1861 round(mastering_meta->b_x * chroma_den), chroma_den);
1862 metadata->display_primaries[2][1] = av_make_q(
1863 round(mastering_meta->b_y * chroma_den), chroma_den);
1864 metadata->white_point[0] = av_make_q(
1865 round(mastering_meta->white_x * chroma_den), chroma_den);
1866 metadata->white_point[1] = av_make_q(
1867 round(mastering_meta->white_y * chroma_den), chroma_den);
1868 metadata->has_primaries = 1;
1870 if (has_mastering_luminance) {
1871 metadata->max_luminance = av_make_q(
1872 round(mastering_meta->max_luminance * luma_den), luma_den);
1873 metadata->min_luminance = av_make_q(
1874 round(mastering_meta->min_luminance * luma_den), luma_den);
1875 metadata->has_luminance = 1;
1881 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
1883 const AVCodecTag *codec_tags;
1885 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
1886 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
1888 /* Normalize noncompliant private data that starts with the fourcc
1889 * by expanding/shifting the data by 4 bytes and storing the data
1890 * size at the start. */
1891 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
1892 uint8_t *p = av_realloc(track->codec_priv.data,
1893 track->codec_priv.size + 4);
1895 return AVERROR(ENOMEM);
1896 memmove(p + 4, p, track->codec_priv.size);
1897 track->codec_priv.data = p;
1898 track->codec_priv.size += 4;
1899 AV_WB32(track->codec_priv.data, track->codec_priv.size);
1902 *fourcc = AV_RL32(track->codec_priv.data + 4);
1903 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
1908 static int matroska_parse_tracks(AVFormatContext *s)
1910 MatroskaDemuxContext *matroska = s->priv_data;
1911 MatroskaTrack *tracks = matroska->tracks.elem;
1916 for (i = 0; i < matroska->tracks.nb_elem; i++) {
1917 MatroskaTrack *track = &tracks[i];
1918 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
1919 EbmlList *encodings_list = &track->encodings;
1920 MatroskaTrackEncoding *encodings = encodings_list->elem;
1921 uint8_t *extradata = NULL;
1922 int extradata_size = 0;
1923 int extradata_offset = 0;
1924 uint32_t fourcc = 0;
1926 char* key_id_base64 = NULL;
1929 /* Apply some sanity checks. */
1930 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
1931 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
1932 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
1933 track->type != MATROSKA_TRACK_TYPE_METADATA) {
1934 av_log(matroska->ctx, AV_LOG_INFO,
1935 "Unknown or unsupported track type %"PRIu64"\n",
1939 if (!track->codec_id)
1942 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
1943 isnan(track->audio.samplerate)) {
1944 av_log(matroska->ctx, AV_LOG_WARNING,
1945 "Invalid sample rate %f, defaulting to 8000 instead.\n",
1946 track->audio.samplerate);
1947 track->audio.samplerate = 8000;
1950 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
1951 if (!track->default_duration && track->video.frame_rate > 0)
1952 track->default_duration = 1000000000 / track->video.frame_rate;
1953 if (track->video.display_width == -1)
1954 track->video.display_width = track->video.pixel_width;
1955 if (track->video.display_height == -1)
1956 track->video.display_height = track->video.pixel_height;
1957 if (track->video.color_space.size == 4)
1958 fourcc = AV_RL32(track->video.color_space.data);
1959 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
1960 if (!track->audio.out_samplerate)
1961 track->audio.out_samplerate = track->audio.samplerate;
1963 if (encodings_list->nb_elem > 1) {
1964 av_log(matroska->ctx, AV_LOG_ERROR,
1965 "Multiple combined encodings not supported");
1966 } else if (encodings_list->nb_elem == 1) {
1967 if (encodings[0].type) {
1968 if (encodings[0].encryption.key_id.size > 0) {
1969 /* Save the encryption key id to be stored later as a
1971 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
1972 key_id_base64 = av_malloc(b64_size);
1973 if (key_id_base64 == NULL)
1974 return AVERROR(ENOMEM);
1976 av_base64_encode(key_id_base64, b64_size,
1977 encodings[0].encryption.key_id.data,
1978 encodings[0].encryption.key_id.size);
1980 encodings[0].scope = 0;
1981 av_log(matroska->ctx, AV_LOG_ERROR,
1982 "Unsupported encoding type");
1986 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
1989 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
1992 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
1994 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
1995 encodings[0].scope = 0;
1996 av_log(matroska->ctx, AV_LOG_ERROR,
1997 "Unsupported encoding type");
1998 } else if (track->codec_priv.size && encodings[0].scope & 2) {
1999 uint8_t *codec_priv = track->codec_priv.data;
2000 int ret = matroska_decode_buffer(&track->codec_priv.data,
2001 &track->codec_priv.size,
2004 track->codec_priv.data = NULL;
2005 track->codec_priv.size = 0;
2006 av_log(matroska->ctx, AV_LOG_ERROR,
2007 "Failed to decode codec private data\n");
2010 if (codec_priv != track->codec_priv.data)
2011 av_free(codec_priv);
2015 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2016 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2017 strlen(ff_mkv_codec_tags[j].str))) {
2018 codec_id = ff_mkv_codec_tags[j].id;
2023 st = track->stream = avformat_new_stream(s, NULL);
2025 av_free(key_id_base64);
2026 return AVERROR(ENOMEM);
2029 if (key_id_base64) {
2030 /* export encryption key id as base64 metadata tag */
2031 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
2032 av_freep(&key_id_base64);
2035 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2036 track->codec_priv.size >= 40 &&
2037 track->codec_priv.data) {
2038 track->ms_compat = 1;
2039 bit_depth = AV_RL16(track->codec_priv.data + 14);
2040 fourcc = AV_RL32(track->codec_priv.data + 16);
2041 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2044 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2046 extradata_offset = 40;
2047 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2048 track->codec_priv.size >= 14 &&
2049 track->codec_priv.data) {
2051 ffio_init_context(&b, track->codec_priv.data,
2052 track->codec_priv.size,
2053 0, NULL, NULL, NULL, NULL);
2054 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2057 codec_id = st->codecpar->codec_id;
2058 fourcc = st->codecpar->codec_tag;
2059 extradata_offset = FFMIN(track->codec_priv.size, 18);
2060 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2061 /* Normally 36, but allow noncompliant private data */
2062 && (track->codec_priv.size >= 32)
2063 && (track->codec_priv.data)) {
2064 uint16_t sample_size;
2065 int ret = get_qt_codec(track, &fourcc, &codec_id);
2068 sample_size = AV_RB16(track->codec_priv.data + 26);
2070 if (sample_size == 8) {
2071 fourcc = MKTAG('r','a','w',' ');
2072 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2073 } else if (sample_size == 16) {
2074 fourcc = MKTAG('t','w','o','s');
2075 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2078 if ((fourcc == MKTAG('t','w','o','s') ||
2079 fourcc == MKTAG('s','o','w','t')) &&
2081 codec_id = AV_CODEC_ID_PCM_S8;
2082 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2083 (track->codec_priv.size >= 21) &&
2084 (track->codec_priv.data)) {
2085 int ret = get_qt_codec(track, &fourcc, &codec_id);
2088 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2089 fourcc = MKTAG('S','V','Q','3');
2090 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2092 if (codec_id == AV_CODEC_ID_NONE) {
2094 av_get_codec_tag_string(buf, sizeof(buf), fourcc);
2095 av_log(matroska->ctx, AV_LOG_ERROR,
2096 "mov FourCC not found %s.\n", buf);
2098 if (track->codec_priv.size >= 86) {
2099 bit_depth = AV_RB16(track->codec_priv.data + 82);
2100 ffio_init_context(&b, track->codec_priv.data,
2101 track->codec_priv.size,
2102 0, NULL, NULL, NULL, NULL);
2103 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2105 track->has_palette = 1;
2108 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2109 switch (track->audio.bitdepth) {
2111 codec_id = AV_CODEC_ID_PCM_U8;
2114 codec_id = AV_CODEC_ID_PCM_S24BE;
2117 codec_id = AV_CODEC_ID_PCM_S32BE;
2120 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2121 switch (track->audio.bitdepth) {
2123 codec_id = AV_CODEC_ID_PCM_U8;
2126 codec_id = AV_CODEC_ID_PCM_S24LE;
2129 codec_id = AV_CODEC_ID_PCM_S32LE;
2132 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2133 track->audio.bitdepth == 64) {
2134 codec_id = AV_CODEC_ID_PCM_F64LE;
2135 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2136 int profile = matroska_aac_profile(track->codec_id);
2137 int sri = matroska_aac_sri(track->audio.samplerate);
2138 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2140 return AVERROR(ENOMEM);
2141 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2142 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2143 if (strstr(track->codec_id, "SBR")) {
2144 sri = matroska_aac_sri(track->audio.out_samplerate);
2145 extradata[2] = 0x56;
2146 extradata[3] = 0xE5;
2147 extradata[4] = 0x80 | (sri << 3);
2151 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2152 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2153 * Create the "atom size", "tag", and "tag version" fields the
2154 * decoder expects manually. */
2155 extradata_size = 12 + track->codec_priv.size;
2156 extradata = av_mallocz(extradata_size +
2157 AV_INPUT_BUFFER_PADDING_SIZE);
2159 return AVERROR(ENOMEM);
2160 AV_WB32(extradata, extradata_size);
2161 memcpy(&extradata[4], "alac", 4);
2162 AV_WB32(&extradata[8], 0);
2163 memcpy(&extradata[12], track->codec_priv.data,
2164 track->codec_priv.size);
2165 } else if (codec_id == AV_CODEC_ID_TTA) {
2166 extradata_size = 30;
2167 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2169 return AVERROR(ENOMEM);
2170 ffio_init_context(&b, extradata, extradata_size, 1,
2171 NULL, NULL, NULL, NULL);
2172 avio_write(&b, "TTA1", 4);
2174 if (track->audio.channels > UINT16_MAX ||
2175 track->audio.bitdepth > UINT16_MAX) {
2176 av_log(matroska->ctx, AV_LOG_WARNING,
2177 "Too large audio channel number %"PRIu64
2178 " or bitdepth %"PRIu64". Skipping track.\n",
2179 track->audio.channels, track->audio.bitdepth);
2180 av_freep(&extradata);
2181 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2182 return AVERROR_INVALIDDATA;
2186 avio_wl16(&b, track->audio.channels);
2187 avio_wl16(&b, track->audio.bitdepth);
2188 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2189 return AVERROR_INVALIDDATA;
2190 avio_wl32(&b, track->audio.out_samplerate);
2191 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
2192 track->audio.out_samplerate,
2193 AV_TIME_BASE * 1000));
2194 } else if (codec_id == AV_CODEC_ID_RV10 ||
2195 codec_id == AV_CODEC_ID_RV20 ||
2196 codec_id == AV_CODEC_ID_RV30 ||
2197 codec_id == AV_CODEC_ID_RV40) {
2198 extradata_offset = 26;
2199 } else if (codec_id == AV_CODEC_ID_RA_144) {
2200 track->audio.out_samplerate = 8000;
2201 track->audio.channels = 1;
2202 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2203 codec_id == AV_CODEC_ID_COOK ||
2204 codec_id == AV_CODEC_ID_ATRAC3 ||
2205 codec_id == AV_CODEC_ID_SIPR)
2206 && track->codec_priv.data) {
2209 ffio_init_context(&b, track->codec_priv.data,
2210 track->codec_priv.size,
2211 0, NULL, NULL, NULL, NULL);
2213 flavor = avio_rb16(&b);
2214 track->audio.coded_framesize = avio_rb32(&b);
2216 track->audio.sub_packet_h = avio_rb16(&b);
2217 track->audio.frame_size = avio_rb16(&b);
2218 track->audio.sub_packet_size = avio_rb16(&b);
2220 track->audio.coded_framesize <= 0 ||
2221 track->audio.sub_packet_h <= 0 ||
2222 track->audio.frame_size <= 0 ||
2223 track->audio.sub_packet_size <= 0)
2224 return AVERROR_INVALIDDATA;
2225 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2226 track->audio.frame_size);
2227 if (!track->audio.buf)
2228 return AVERROR(ENOMEM);
2229 if (codec_id == AV_CODEC_ID_RA_288) {
2230 st->codecpar->block_align = track->audio.coded_framesize;
2231 track->codec_priv.size = 0;
2233 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2234 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2235 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2236 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2238 st->codecpar->block_align = track->audio.sub_packet_size;
2239 extradata_offset = 78;
2241 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2242 ret = matroska_parse_flac(s, track, &extradata_offset);
2245 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2246 fourcc = AV_RL32(track->codec_priv.data);
2248 track->codec_priv.size -= extradata_offset;
2250 if (codec_id == AV_CODEC_ID_NONE)
2251 av_log(matroska->ctx, AV_LOG_INFO,
2252 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2254 if (track->time_scale < 0.01)
2255 track->time_scale = 1.0;
2256 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2257 1000 * 1000 * 1000); /* 64 bit pts in ns */
2259 /* convert the delay from ns to the track timebase */
2260 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2261 (AVRational){ 1, 1000000000 },
2264 st->codecpar->codec_id = codec_id;
2266 if (strcmp(track->language, "und"))
2267 av_dict_set(&st->metadata, "language", track->language, 0);
2268 av_dict_set(&st->metadata, "title", track->name, 0);
2270 if (track->flag_default)
2271 st->disposition |= AV_DISPOSITION_DEFAULT;
2272 if (track->flag_forced)
2273 st->disposition |= AV_DISPOSITION_FORCED;
2275 if (!st->codecpar->extradata) {
2277 st->codecpar->extradata = extradata;
2278 st->codecpar->extradata_size = extradata_size;
2279 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2280 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2281 return AVERROR(ENOMEM);
2282 memcpy(st->codecpar->extradata,
2283 track->codec_priv.data + extradata_offset,
2284 track->codec_priv.size);
2288 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2289 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2290 int display_width_mul = 1;
2291 int display_height_mul = 1;
2293 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2294 st->codecpar->codec_tag = fourcc;
2296 st->codecpar->bits_per_coded_sample = bit_depth;
2297 st->codecpar->width = track->video.pixel_width;
2298 st->codecpar->height = track->video.pixel_height;
2300 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2301 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2302 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2303 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2305 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2306 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2308 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2309 av_reduce(&st->sample_aspect_ratio.num,
2310 &st->sample_aspect_ratio.den,
2311 st->codecpar->height * track->video.display_width * display_width_mul,
2312 st->codecpar->width * track->video.display_height * display_height_mul,
2315 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2316 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2318 if (track->default_duration) {
2319 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2320 1000000000, track->default_duration, 30000);
2321 #if FF_API_R_FRAME_RATE
2322 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2323 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2324 st->r_frame_rate = st->avg_frame_rate;
2328 /* export stereo mode flag as metadata tag */
2329 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2330 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2332 /* export alpha mode flag as metadata tag */
2333 if (track->video.alpha_mode)
2334 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2336 /* if we have virtual track, mark the real tracks */
2337 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2339 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2341 snprintf(buf, sizeof(buf), "%s_%d",
2342 ff_matroska_video_stereo_plane[planes[j].type], i);
2343 for (k=0; k < matroska->tracks.nb_elem; k++)
2344 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2345 av_dict_set(&tracks[k].stream->metadata,
2346 "stereo_mode", buf, 0);
2350 // add stream level stereo3d side data if it is a supported format
2351 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2352 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2353 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2358 if (s->strict_std_compliance <= FF_COMPLIANCE_UNOFFICIAL) {
2359 int ret = mkv_parse_video_color(st, track);
2363 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2364 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2365 st->codecpar->codec_tag = fourcc;
2366 st->codecpar->sample_rate = track->audio.out_samplerate;
2367 st->codecpar->channels = track->audio.channels;
2368 if (!st->codecpar->bits_per_coded_sample)
2369 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2370 if (st->codecpar->codec_id == AV_CODEC_ID_MP3)
2371 st->need_parsing = AVSTREAM_PARSE_FULL;
2372 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2373 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2374 if (track->codec_delay > 0) {
2375 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2376 (AVRational){1, 1000000000},
2377 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2378 48000 : st->codecpar->sample_rate});
2380 if (track->seek_preroll > 0) {
2381 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2382 (AVRational){1, 1000000000},
2383 (AVRational){1, st->codecpar->sample_rate});
2385 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2386 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2388 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2389 st->disposition |= AV_DISPOSITION_CAPTIONS;
2390 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2391 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2392 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2393 st->disposition |= AV_DISPOSITION_METADATA;
2395 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2396 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2397 if (st->codecpar->codec_id == AV_CODEC_ID_ASS)
2398 matroska->contains_ssa = 1;
2405 static int matroska_read_header(AVFormatContext *s)
2407 MatroskaDemuxContext *matroska = s->priv_data;
2408 EbmlList *attachments_list = &matroska->attachments;
2409 EbmlList *chapters_list = &matroska->chapters;
2410 MatroskaAttachment *attachments;
2411 MatroskaChapter *chapters;
2412 uint64_t max_start = 0;
2418 matroska->cues_parsing_deferred = 1;
2420 /* First read the EBML header. */
2421 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2422 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2423 ebml_free(ebml_syntax, &ebml);
2424 return AVERROR_INVALIDDATA;
2426 if (ebml.version > EBML_VERSION ||
2427 ebml.max_size > sizeof(uint64_t) ||
2428 ebml.id_length > sizeof(uint32_t) ||
2429 ebml.doctype_version > 3) {
2430 av_log(matroska->ctx, AV_LOG_ERROR,
2431 "EBML header using unsupported features\n"
2432 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2433 ebml.version, ebml.doctype, ebml.doctype_version);
2434 ebml_free(ebml_syntax, &ebml);
2435 return AVERROR_PATCHWELCOME;
2436 } else if (ebml.doctype_version == 3) {
2437 av_log(matroska->ctx, AV_LOG_WARNING,
2438 "EBML header using unsupported features\n"
2439 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2440 ebml.version, ebml.doctype, ebml.doctype_version);
2442 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2443 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2445 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2446 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2447 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2448 ebml_free(ebml_syntax, &ebml);
2449 return AVERROR_INVALIDDATA;
2452 ebml_free(ebml_syntax, &ebml);
2454 /* The next thing is a segment. */
2455 pos = avio_tell(matroska->ctx->pb);
2456 res = ebml_parse(matroska, matroska_segments, matroska);
2457 // try resyncing until we find a EBML_STOP type element.
2459 res = matroska_resync(matroska, pos);
2462 pos = avio_tell(matroska->ctx->pb);
2463 res = ebml_parse(matroska, matroska_segment, matroska);
2465 matroska_execute_seekhead(matroska);
2467 if (!matroska->time_scale)
2468 matroska->time_scale = 1000000;
2469 if (matroska->duration)
2470 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2471 1000 / AV_TIME_BASE;
2472 av_dict_set(&s->metadata, "title", matroska->title, 0);
2473 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2475 if (matroska->date_utc.size == 8)
2476 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2478 res = matroska_parse_tracks(s);
2482 attachments = attachments_list->elem;
2483 for (j = 0; j < attachments_list->nb_elem; j++) {
2484 if (!(attachments[j].filename && attachments[j].mime &&
2485 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2486 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2488 AVStream *st = avformat_new_stream(s, NULL);
2491 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2492 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2493 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2495 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2496 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2497 strlen(ff_mkv_image_mime_tags[i].str))) {
2498 st->codecpar->codec_id = ff_mkv_image_mime_tags[i].id;
2503 attachments[j].stream = st;
2505 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2506 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2507 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2509 av_init_packet(&st->attached_pic);
2510 if ((res = av_new_packet(&st->attached_pic, attachments[j].bin.size)) < 0)
2512 memcpy(st->attached_pic.data, attachments[j].bin.data, attachments[j].bin.size);
2513 st->attached_pic.stream_index = st->index;
2514 st->attached_pic.flags |= AV_PKT_FLAG_KEY;
2516 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2517 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2519 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2520 attachments[j].bin.size);
2522 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2523 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2524 strlen(ff_mkv_mime_tags[i].str))) {
2525 st->codecpar->codec_id = ff_mkv_mime_tags[i].id;
2533 chapters = chapters_list->elem;
2534 for (i = 0; i < chapters_list->nb_elem; i++)
2535 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2536 (max_start == 0 || chapters[i].start > max_start)) {
2537 chapters[i].chapter =
2538 avpriv_new_chapter(s, chapters[i].uid,
2539 (AVRational) { 1, 1000000000 },
2540 chapters[i].start, chapters[i].end,
2542 if (chapters[i].chapter) {
2543 av_dict_set(&chapters[i].chapter->metadata,
2544 "title", chapters[i].title, 0);
2546 max_start = chapters[i].start;
2549 matroska_add_index_entries(matroska);
2551 matroska_convert_tags(s);
2555 matroska_read_close(s);
2560 * Put one packet in an application-supplied AVPacket struct.
2561 * Returns 0 on success or -1 on failure.
2563 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2566 if (matroska->num_packets > 0) {
2567 MatroskaTrack *tracks = matroska->tracks.elem;
2568 MatroskaTrack *track;
2569 memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
2570 av_freep(&matroska->packets[0]);
2571 track = &tracks[pkt->stream_index];
2572 if (track->has_palette) {
2573 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2575 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
2577 memcpy(pal, track->palette, AVPALETTE_SIZE);
2579 track->has_palette = 0;
2581 if (matroska->num_packets > 1) {
2583 memmove(&matroska->packets[0], &matroska->packets[1],
2584 (matroska->num_packets - 1) * sizeof(AVPacket *));
2585 newpackets = av_realloc(matroska->packets,
2586 (matroska->num_packets - 1) *
2587 sizeof(AVPacket *));
2589 matroska->packets = newpackets;
2591 av_freep(&matroska->packets);
2592 matroska->prev_pkt = NULL;
2594 matroska->num_packets--;
2602 * Free all packets in our internal queue.
2604 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2606 matroska->prev_pkt = NULL;
2607 if (matroska->packets) {
2609 for (n = 0; n < matroska->num_packets; n++) {
2610 av_packet_unref(matroska->packets[n]);
2611 av_freep(&matroska->packets[n]);
2613 av_freep(&matroska->packets);
2614 matroska->num_packets = 0;
2618 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2619 int *buf_size, int type,
2620 uint32_t **lace_buf, int *laces)
2622 int res = 0, n, size = *buf_size;
2623 uint8_t *data = *buf;
2624 uint32_t *lace_size;
2628 *lace_buf = av_mallocz(sizeof(int));
2630 return AVERROR(ENOMEM);
2632 *lace_buf[0] = size;
2636 av_assert0(size > 0);
2640 lace_size = av_mallocz(*laces * sizeof(int));
2642 return AVERROR(ENOMEM);
2645 case 0x1: /* Xiph lacing */
2649 for (n = 0; res == 0 && n < *laces - 1; n++) {
2651 if (size <= total) {
2652 res = AVERROR_INVALIDDATA;
2657 lace_size[n] += temp;
2664 if (size <= total) {
2665 res = AVERROR_INVALIDDATA;
2669 lace_size[n] = size - total;
2673 case 0x2: /* fixed-size lacing */
2674 if (size % (*laces)) {
2675 res = AVERROR_INVALIDDATA;
2678 for (n = 0; n < *laces; n++)
2679 lace_size[n] = size / *laces;
2682 case 0x3: /* EBML lacing */
2686 n = matroska_ebmlnum_uint(matroska, data, size, &num);
2687 if (n < 0 || num > INT_MAX) {
2688 av_log(matroska->ctx, AV_LOG_INFO,
2689 "EBML block data error\n");
2690 res = n<0 ? n : AVERROR_INVALIDDATA;
2695 total = lace_size[0] = num;
2696 for (n = 1; res == 0 && n < *laces - 1; n++) {
2699 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
2700 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
2701 av_log(matroska->ctx, AV_LOG_INFO,
2702 "EBML block data error\n");
2703 res = r<0 ? r : AVERROR_INVALIDDATA;
2708 lace_size[n] = lace_size[n - 1] + snum;
2709 total += lace_size[n];
2711 if (size <= total) {
2712 res = AVERROR_INVALIDDATA;
2715 lace_size[*laces - 1] = size - total;
2721 *lace_buf = lace_size;
2727 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
2728 MatroskaTrack *track, AVStream *st,
2729 uint8_t *data, int size, uint64_t timecode,
2732 int a = st->codecpar->block_align;
2733 int sps = track->audio.sub_packet_size;
2734 int cfs = track->audio.coded_framesize;
2735 int h = track->audio.sub_packet_h;
2736 int y = track->audio.sub_packet_cnt;
2737 int w = track->audio.frame_size;
2740 if (!track->audio.pkt_cnt) {
2741 if (track->audio.sub_packet_cnt == 0)
2742 track->audio.buf_timecode = timecode;
2743 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
2744 if (size < cfs * h / 2) {
2745 av_log(matroska->ctx, AV_LOG_ERROR,
2746 "Corrupt int4 RM-style audio packet size\n");
2747 return AVERROR_INVALIDDATA;
2749 for (x = 0; x < h / 2; x++)
2750 memcpy(track->audio.buf + x * 2 * w + y * cfs,
2751 data + x * cfs, cfs);
2752 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
2754 av_log(matroska->ctx, AV_LOG_ERROR,
2755 "Corrupt sipr RM-style audio packet size\n");
2756 return AVERROR_INVALIDDATA;
2758 memcpy(track->audio.buf + y * w, data, w);
2760 if (size < sps * w / sps || h<=0 || w%sps) {
2761 av_log(matroska->ctx, AV_LOG_ERROR,
2762 "Corrupt generic RM-style audio packet size\n");
2763 return AVERROR_INVALIDDATA;
2765 for (x = 0; x < w / sps; x++)
2766 memcpy(track->audio.buf +
2767 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
2768 data + x * sps, sps);
2771 if (++track->audio.sub_packet_cnt >= h) {
2772 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
2773 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
2774 track->audio.sub_packet_cnt = 0;
2775 track->audio.pkt_cnt = h * w / a;
2779 while (track->audio.pkt_cnt) {
2781 AVPacket *pkt = av_mallocz(sizeof(AVPacket));
2783 return AVERROR(ENOMEM);
2785 ret = av_new_packet(pkt, a);
2791 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
2793 pkt->pts = track->audio.buf_timecode;
2794 track->audio.buf_timecode = AV_NOPTS_VALUE;
2796 pkt->stream_index = st->index;
2797 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2803 /* reconstruct full wavpack blocks from mangled matroska ones */
2804 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
2805 uint8_t **pdst, int *size)
2807 uint8_t *dst = NULL;
2812 int ret, offset = 0;
2814 if (srclen < 12 || track->stream->codecpar->extradata_size < 2)
2815 return AVERROR_INVALIDDATA;
2817 ver = AV_RL16(track->stream->codecpar->extradata);
2819 samples = AV_RL32(src);
2823 while (srclen >= 8) {
2828 uint32_t flags = AV_RL32(src);
2829 uint32_t crc = AV_RL32(src + 4);
2833 multiblock = (flags & 0x1800) != 0x1800;
2836 ret = AVERROR_INVALIDDATA;
2839 blocksize = AV_RL32(src);
2845 if (blocksize > srclen) {
2846 ret = AVERROR_INVALIDDATA;
2850 tmp = av_realloc(dst, dstlen + blocksize + 32);
2852 ret = AVERROR(ENOMEM);
2856 dstlen += blocksize + 32;
2858 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
2859 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
2860 AV_WL16(dst + offset + 8, ver); // version
2861 AV_WL16(dst + offset + 10, 0); // track/index_no
2862 AV_WL32(dst + offset + 12, 0); // total samples
2863 AV_WL32(dst + offset + 16, 0); // block index
2864 AV_WL32(dst + offset + 20, samples); // number of samples
2865 AV_WL32(dst + offset + 24, flags); // flags
2866 AV_WL32(dst + offset + 28, crc); // crc
2867 memcpy(dst + offset + 32, src, blocksize); // block data
2870 srclen -= blocksize;
2871 offset += blocksize + 32;
2884 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
2885 MatroskaTrack *track,
2887 uint8_t *data, int data_len,
2893 uint8_t *id, *settings, *text, *buf;
2894 int id_len, settings_len, text_len;
2899 return AVERROR_INVALIDDATA;
2902 q = data + data_len;
2907 if (*p == '\r' || *p == '\n') {
2916 if (p >= q || *p != '\n')
2917 return AVERROR_INVALIDDATA;
2923 if (*p == '\r' || *p == '\n') {
2924 settings_len = p - settings;
2932 if (p >= q || *p != '\n')
2933 return AVERROR_INVALIDDATA;
2938 while (text_len > 0) {
2939 const int len = text_len - 1;
2940 const uint8_t c = p[len];
2941 if (c != '\r' && c != '\n')
2947 return AVERROR_INVALIDDATA;
2949 pkt = av_mallocz(sizeof(*pkt));
2951 return AVERROR(ENOMEM);
2952 err = av_new_packet(pkt, text_len);
2955 return AVERROR(err);
2958 memcpy(pkt->data, text, text_len);
2961 buf = av_packet_new_side_data(pkt,
2962 AV_PKT_DATA_WEBVTT_IDENTIFIER,
2966 return AVERROR(ENOMEM);
2968 memcpy(buf, id, id_len);
2971 if (settings_len > 0) {
2972 buf = av_packet_new_side_data(pkt,
2973 AV_PKT_DATA_WEBVTT_SETTINGS,
2977 return AVERROR(ENOMEM);
2979 memcpy(buf, settings, settings_len);
2982 // Do we need this for subtitles?
2983 // pkt->flags = AV_PKT_FLAG_KEY;
2985 pkt->stream_index = st->index;
2986 pkt->pts = timecode;
2988 // Do we need this for subtitles?
2989 // pkt->dts = timecode;
2991 pkt->duration = duration;
2994 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
2995 matroska->prev_pkt = pkt;
3000 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3001 MatroskaTrack *track, AVStream *st,
3002 uint8_t *data, int pkt_size,
3003 uint64_t timecode, uint64_t lace_duration,
3004 int64_t pos, int is_keyframe,
3005 uint8_t *additional, uint64_t additional_id, int additional_size,
3006 int64_t discard_padding)
3008 MatroskaTrackEncoding *encodings = track->encodings.elem;
3009 uint8_t *pkt_data = data;
3010 int offset = 0, res;
3013 if (encodings && !encodings->type && encodings->scope & 1) {
3014 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
3019 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3021 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
3023 av_log(matroska->ctx, AV_LOG_ERROR,
3024 "Error parsing a wavpack block.\n");
3027 if (pkt_data != data)
3028 av_freep(&pkt_data);
3032 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3033 AV_RB32(&data[4]) != MKBETAG('i', 'c', 'p', 'f'))
3036 pkt = av_mallocz(sizeof(AVPacket));
3038 if (pkt_data != data)
3039 av_freep(&pkt_data);
3040 return AVERROR(ENOMEM);
3042 /* XXX: prevent data copy... */
3043 if (av_new_packet(pkt, pkt_size + offset) < 0) {
3045 res = AVERROR(ENOMEM);
3049 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES && offset == 8) {
3050 uint8_t *buf = pkt->data;
3051 bytestream_put_be32(&buf, pkt_size);
3052 bytestream_put_be32(&buf, MKBETAG('i', 'c', 'p', 'f'));
3055 memcpy(pkt->data + offset, pkt_data, pkt_size);
3057 if (pkt_data != data)
3058 av_freep(&pkt_data);
3060 pkt->flags = is_keyframe;
3061 pkt->stream_index = st->index;
3063 if (additional_size > 0) {
3064 uint8_t *side_data = av_packet_new_side_data(pkt,
3065 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3066 additional_size + 8);
3068 av_packet_unref(pkt);
3070 return AVERROR(ENOMEM);
3072 AV_WB64(side_data, additional_id);
3073 memcpy(side_data + 8, additional, additional_size);
3076 if (discard_padding) {
3077 uint8_t *side_data = av_packet_new_side_data(pkt,
3078 AV_PKT_DATA_SKIP_SAMPLES,
3081 av_packet_unref(pkt);
3083 return AVERROR(ENOMEM);
3085 discard_padding = av_rescale_q(discard_padding,
3086 (AVRational){1, 1000000000},
3087 (AVRational){1, st->codecpar->sample_rate});
3088 if (discard_padding > 0) {
3089 AV_WL32(side_data + 4, discard_padding);
3091 AV_WL32(side_data, -discard_padding);
3095 if (track->ms_compat)
3096 pkt->dts = timecode;
3098 pkt->pts = timecode;
3100 pkt->duration = lace_duration;
3102 #if FF_API_CONVERGENCE_DURATION
3103 FF_DISABLE_DEPRECATION_WARNINGS
3104 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3105 pkt->convergence_duration = lace_duration;
3107 FF_ENABLE_DEPRECATION_WARNINGS
3110 dynarray_add(&matroska->packets, &matroska->num_packets, pkt);
3111 matroska->prev_pkt = pkt;
3116 if (pkt_data != data)
3117 av_freep(&pkt_data);
3121 static int matroska_parse_block(MatroskaDemuxContext *matroska, uint8_t *data,
3122 int size, int64_t pos, uint64_t cluster_time,
3123 uint64_t block_duration, int is_keyframe,
3124 uint8_t *additional, uint64_t additional_id, int additional_size,
3125 int64_t cluster_pos, int64_t discard_padding)
3127 uint64_t timecode = AV_NOPTS_VALUE;
3128 MatroskaTrack *track;
3132 uint32_t *lace_size = NULL;
3133 int n, flags, laces = 0;
3135 int trust_default_duration = 1;
3137 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
3138 av_log(matroska->ctx, AV_LOG_ERROR, "EBML block data error\n");
3144 track = matroska_find_track_by_num(matroska, num);
3145 if (!track || !track->stream) {
3146 av_log(matroska->ctx, AV_LOG_INFO,
3147 "Invalid stream %"PRIu64" or size %u\n", num, size);
3148 return AVERROR_INVALIDDATA;
3149 } else if (size <= 3)
3152 if (st->discard >= AVDISCARD_ALL)
3154 av_assert1(block_duration != AV_NOPTS_VALUE);
3156 block_time = sign_extend(AV_RB16(data), 16);
3160 if (is_keyframe == -1)
3161 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3163 if (cluster_time != (uint64_t) -1 &&
3164 (block_time >= 0 || cluster_time >= -block_time)) {
3165 timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3166 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3167 timecode < track->end_timecode)
3168 is_keyframe = 0; /* overlapping subtitles are not key frame */
3170 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3174 if (matroska->skip_to_keyframe &&
3175 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3176 // Compare signed timecodes. Timecode may be negative due to codec delay
3177 // offset. We don't support timestamps greater than int64_t anyway - see
3179 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3182 matroska->skip_to_keyframe = 0;
3183 else if (!st->skip_to_keyframe) {
3184 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3185 matroska->skip_to_keyframe = 0;
3189 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
3190 &lace_size, &laces);
3195 if (track->audio.samplerate == 8000) {
3196 // If this is needed for more codecs, then add them here
3197 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3198 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3199 trust_default_duration = 0;
3203 if (!block_duration && trust_default_duration)
3204 block_duration = track->default_duration * laces / matroska->time_scale;
3206 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3207 track->end_timecode =
3208 FFMAX(track->end_timecode, timecode + block_duration);
3210 for (n = 0; n < laces; n++) {
3211 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3213 if (lace_size[n] > size) {
3214 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
3218 if ((st->codecpar->codec_id == AV_CODEC_ID_RA_288 ||
3219 st->codecpar->codec_id == AV_CODEC_ID_COOK ||
3220 st->codecpar->codec_id == AV_CODEC_ID_SIPR ||
3221 st->codecpar->codec_id == AV_CODEC_ID_ATRAC3) &&
3222 st->codecpar->block_align && track->audio.sub_packet_size) {
3223 res = matroska_parse_rm_audio(matroska, track, st, data,
3229 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3230 res = matroska_parse_webvtt(matroska, track, st,
3232 timecode, lace_duration,
3237 res = matroska_parse_frame(matroska, track, st, data, lace_size[n],
3238 timecode, lace_duration, pos,
3239 !n ? is_keyframe : 0,
3240 additional, additional_id, additional_size,
3246 if (timecode != AV_NOPTS_VALUE)
3247 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3248 data += lace_size[n];
3249 size -= lace_size[n];
3257 static int matroska_parse_cluster_incremental(MatroskaDemuxContext *matroska)
3259 EbmlList *blocks_list;
3260 MatroskaBlock *blocks;
3262 res = ebml_parse(matroska,
3263 matroska_cluster_incremental_parsing,
3264 &matroska->current_cluster);
3267 if (matroska->current_cluster_pos)
3268 ebml_level_end(matroska);
3269 ebml_free(matroska_cluster, &matroska->current_cluster);
3270 memset(&matroska->current_cluster, 0, sizeof(MatroskaCluster));
3271 matroska->current_cluster_num_blocks = 0;
3272 matroska->current_cluster_pos = avio_tell(matroska->ctx->pb);
3273 matroska->prev_pkt = NULL;
3274 /* sizeof the ID which was already read */
3275 if (matroska->current_id)
3276 matroska->current_cluster_pos -= 4;
3277 res = ebml_parse(matroska,
3278 matroska_clusters_incremental,
3279 &matroska->current_cluster);
3280 /* Try parsing the block again. */
3282 res = ebml_parse(matroska,
3283 matroska_cluster_incremental_parsing,
3284 &matroska->current_cluster);
3288 matroska->current_cluster_num_blocks <
3289 matroska->current_cluster.blocks.nb_elem) {
3290 blocks_list = &matroska->current_cluster.blocks;
3291 blocks = blocks_list->elem;
3293 matroska->current_cluster_num_blocks = blocks_list->nb_elem;
3294 i = blocks_list->nb_elem - 1;
3295 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
3296 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
3297 uint8_t* additional = blocks[i].additional.size > 0 ?
3298 blocks[i].additional.data : NULL;
3299 if (!blocks[i].non_simple)
3300 blocks[i].duration = 0;
3301 res = matroska_parse_block(matroska, blocks[i].bin.data,
3302 blocks[i].bin.size, blocks[i].bin.pos,
3303 matroska->current_cluster.timecode,
3304 blocks[i].duration, is_keyframe,
3305 additional, blocks[i].additional_id,
3306 blocks[i].additional.size,
3307 matroska->current_cluster_pos,
3308 blocks[i].discard_padding);
3315 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3317 MatroskaCluster cluster = { 0 };
3318 EbmlList *blocks_list;
3319 MatroskaBlock *blocks;
3323 if (!matroska->contains_ssa)
3324 return matroska_parse_cluster_incremental(matroska);
3325 pos = avio_tell(matroska->ctx->pb);
3326 matroska->prev_pkt = NULL;
3327 if (matroska->current_id)
3328 pos -= 4; /* sizeof the ID which was already read */
3329 res = ebml_parse(matroska, matroska_clusters, &cluster);
3330 blocks_list = &cluster.blocks;
3331 blocks = blocks_list->elem;
3332 for (i = 0; i < blocks_list->nb_elem; i++)
3333 if (blocks[i].bin.size > 0 && blocks[i].bin.data) {
3334 int is_keyframe = blocks[i].non_simple ? !blocks[i].reference : -1;
3335 res = matroska_parse_block(matroska, blocks[i].bin.data,
3336 blocks[i].bin.size, blocks[i].bin.pos,
3337 cluster.timecode, blocks[i].duration,
3338 is_keyframe, NULL, 0, 0, pos,
3339 blocks[i].discard_padding);
3341 ebml_free(matroska_cluster, &cluster);
3345 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3347 MatroskaDemuxContext *matroska = s->priv_data;
3350 while (matroska_deliver_packet(matroska, pkt)) {
3351 int64_t pos = avio_tell(matroska->ctx->pb);
3353 return (ret < 0) ? ret : AVERROR_EOF;
3354 if (matroska_parse_cluster(matroska) < 0)
3355 ret = matroska_resync(matroska, pos);
3361 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3362 int64_t timestamp, int flags)
3364 MatroskaDemuxContext *matroska = s->priv_data;
3365 MatroskaTrack *tracks = NULL;
3366 AVStream *st = s->streams[stream_index];
3367 int i, index, index_sub, index_min;
3369 /* Parse the CUES now since we need the index data to seek. */
3370 if (matroska->cues_parsing_deferred > 0) {
3371 matroska->cues_parsing_deferred = 0;
3372 matroska_parse_cues(matroska);
3375 if (!st->nb_index_entries)
3377 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3379 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3380 avio_seek(s->pb, st->index_entries[st->nb_index_entries - 1].pos,
3382 matroska->current_id = 0;
3383 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3384 matroska_clear_queue(matroska);
3385 if (matroska_parse_cluster(matroska) < 0)
3390 matroska_clear_queue(matroska);
3391 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3395 tracks = matroska->tracks.elem;
3396 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3397 tracks[i].audio.pkt_cnt = 0;
3398 tracks[i].audio.sub_packet_cnt = 0;
3399 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3400 tracks[i].end_timecode = 0;
3401 if (tracks[i].type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3403 tracks[i].stream->discard != AVDISCARD_ALL) {
3404 index_sub = av_index_search_timestamp(
3405 tracks[i].stream, st->index_entries[index].timestamp,
3406 AVSEEK_FLAG_BACKWARD);
3407 while (index_sub >= 0 &&
3409 tracks[i].stream->index_entries[index_sub].pos < st->index_entries[index_min].pos &&
3410 st->index_entries[index].timestamp - tracks[i].stream->index_entries[index_sub].timestamp < 30000000000 / matroska->time_scale)
3415 avio_seek(s->pb, st->index_entries[index_min].pos, SEEK_SET);
3416 matroska->current_id = 0;
3417 if (flags & AVSEEK_FLAG_ANY) {
3418 st->skip_to_keyframe = 0;
3419 matroska->skip_to_timecode = timestamp;
3421 st->skip_to_keyframe = 1;
3422 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3424 matroska->skip_to_keyframe = 1;
3426 matroska->num_levels = 0;
3427 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3430 // slightly hackish but allows proper fallback to
3431 // the generic seeking code.
3432 matroska_clear_queue(matroska);
3433 matroska->current_id = 0;
3434 st->skip_to_keyframe =
3435 matroska->skip_to_keyframe = 0;
3437 matroska->num_levels = 0;
3441 static int matroska_read_close(AVFormatContext *s)
3443 MatroskaDemuxContext *matroska = s->priv_data;
3444 MatroskaTrack *tracks = matroska->tracks.elem;
3447 matroska_clear_queue(matroska);
3449 for (n = 0; n < matroska->tracks.nb_elem; n++)
3450 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3451 av_freep(&tracks[n].audio.buf);
3452 ebml_free(matroska_cluster, &matroska->current_cluster);
3453 ebml_free(matroska_segment, matroska);
3459 int64_t start_time_ns;
3460 int64_t end_time_ns;
3461 int64_t start_offset;
3465 /* This function searches all the Cues and returns the CueDesc corresponding the
3466 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3467 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3469 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3470 MatroskaDemuxContext *matroska = s->priv_data;
3473 int nb_index_entries = s->streams[0]->nb_index_entries;
3474 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3475 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3476 for (i = 1; i < nb_index_entries; i++) {
3477 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3478 index_entries[i].timestamp * matroska->time_scale > ts) {
3483 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3484 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3485 if (i != nb_index_entries - 1) {
3486 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3487 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3489 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3490 // FIXME: this needs special handling for files where Cues appear
3491 // before Clusters. the current logic assumes Cues appear after
3493 cue_desc.end_offset = cues_start - matroska->segment_start;
3498 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3500 MatroskaDemuxContext *matroska = s->priv_data;
3501 int64_t cluster_pos, before_pos;
3503 if (s->streams[0]->nb_index_entries <= 0) return 0;
3504 // seek to the first cluster using cues.
3505 index = av_index_search_timestamp(s->streams[0], 0, 0);
3506 if (index < 0) return 0;
3507 cluster_pos = s->streams[0]->index_entries[index].pos;
3508 before_pos = avio_tell(s->pb);
3510 int64_t cluster_id = 0, cluster_length = 0;
3512 avio_seek(s->pb, cluster_pos, SEEK_SET);
3513 // read cluster id and length
3514 ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id);
3515 ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3516 if (cluster_id != 0xF43B675) { // done with all clusters
3519 avio_seek(s->pb, cluster_pos, SEEK_SET);
3520 matroska->current_id = 0;
3521 matroska_clear_queue(matroska);
3522 if (matroska_parse_cluster(matroska) < 0 ||
3523 matroska->num_packets <= 0) {
3526 pkt = matroska->packets[0];
3527 cluster_pos += cluster_length + 12; // 12 is the offset of the cluster id and length.
3528 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3533 avio_seek(s->pb, before_pos, SEEK_SET);
3537 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3538 double min_buffer, double* buffer,
3539 double* sec_to_download, AVFormatContext *s,
3542 double nano_seconds_per_second = 1000000000.0;
3543 double time_sec = time_ns / nano_seconds_per_second;
3545 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3546 int64_t end_time_ns = time_ns + time_to_search_ns;
3547 double sec_downloaded = 0.0;
3548 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3549 if (desc_curr.start_time_ns == -1)
3551 *sec_to_download = 0.0;
3553 // Check for non cue start time.
3554 if (time_ns > desc_curr.start_time_ns) {
3555 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3556 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3557 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3558 double timeToDownload = (cueBytes * 8.0) / bps;
3560 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3561 *sec_to_download += timeToDownload;
3563 // Check if the search ends within the first cue.
3564 if (desc_curr.end_time_ns >= end_time_ns) {
3565 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3566 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3567 sec_downloaded = percent_to_sub * sec_downloaded;
3568 *sec_to_download = percent_to_sub * *sec_to_download;
3571 if ((sec_downloaded + *buffer) <= min_buffer) {
3575 // Get the next Cue.
3576 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3579 while (desc_curr.start_time_ns != -1) {
3580 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3581 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3582 double desc_sec = desc_ns / nano_seconds_per_second;
3583 double bits = (desc_bytes * 8.0);
3584 double time_to_download = bits / bps;
3586 sec_downloaded += desc_sec - time_to_download;
3587 *sec_to_download += time_to_download;
3589 if (desc_curr.end_time_ns >= end_time_ns) {
3590 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3591 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3592 sec_downloaded = percent_to_sub * sec_downloaded;
3593 *sec_to_download = percent_to_sub * *sec_to_download;
3595 if ((sec_downloaded + *buffer) <= min_buffer)
3600 if ((sec_downloaded + *buffer) <= min_buffer) {
3605 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3607 *buffer = *buffer + sec_downloaded;
3611 /* This function computes the bandwidth of the WebM file with the help of
3612 * buffer_size_after_time_downloaded() function. Both of these functions are
3613 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3614 * Matroska parsing mechanism.
3616 * Returns the bandwidth of the file on success; -1 on error.
3618 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3620 MatroskaDemuxContext *matroska = s->priv_data;
3621 AVStream *st = s->streams[0];
3622 double bandwidth = 0.0;
3625 for (i = 0; i < st->nb_index_entries; i++) {
3626 int64_t prebuffer_ns = 1000000000;
3627 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3628 double nano_seconds_per_second = 1000000000.0;
3629 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3630 double prebuffer_bytes = 0.0;
3631 int64_t temp_prebuffer_ns = prebuffer_ns;
3632 int64_t pre_bytes, pre_ns;
3633 double pre_sec, prebuffer, bits_per_second;
3634 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3636 // Start with the first Cue.
3637 CueDesc desc_end = desc_beg;
3639 // Figure out how much data we have downloaded for the prebuffer. This will
3640 // be used later to adjust the bits per sample to try.
3641 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3642 // Prebuffered the entire Cue.
3643 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3644 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3645 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3647 if (desc_end.start_time_ns == -1) {
3648 // The prebuffer is larger than the duration.
3649 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3651 bits_per_second = 0.0;
3653 // The prebuffer ends in the last Cue. Estimate how much data was
3655 pre_bytes = desc_end.end_offset - desc_end.start_offset;
3656 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3657 pre_sec = pre_ns / nano_seconds_per_second;
3659 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3661 prebuffer = prebuffer_ns / nano_seconds_per_second;
3663 // Set this to 0.0 in case our prebuffer buffers the entire video.
3664 bits_per_second = 0.0;
3666 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3667 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3668 double desc_sec = desc_ns / nano_seconds_per_second;
3669 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3671 // Drop the bps by the percentage of bytes buffered.
3672 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3673 double mod_bits_per_second = calc_bits_per_second * percent;
3675 if (prebuffer < desc_sec) {
3677 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3679 // Add 1 so the bits per second should be a little bit greater than file
3681 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3682 const double min_buffer = 0.0;
3683 double buffer = prebuffer;
3684 double sec_to_download = 0.0;
3686 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3687 min_buffer, &buffer, &sec_to_download,
3691 } else if (rv == 0) {
3692 bits_per_second = (double)(bps);
3697 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3698 } while (desc_end.start_time_ns != -1);
3700 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
3702 return (int64_t)bandwidth;
3705 static int webm_dash_manifest_cues(AVFormatContext *s)
3707 MatroskaDemuxContext *matroska = s->priv_data;
3708 EbmlList *seekhead_list = &matroska->seekhead;
3709 MatroskaSeekhead *seekhead = seekhead_list->elem;
3711 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
3714 // determine cues start and end positions
3715 for (i = 0; i < seekhead_list->nb_elem; i++)
3716 if (seekhead[i].id == MATROSKA_ID_CUES)
3719 if (i >= seekhead_list->nb_elem) return -1;
3721 before_pos = avio_tell(matroska->ctx->pb);
3722 cues_start = seekhead[i].pos + matroska->segment_start;
3723 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
3724 // cues_end is computed as cues_start + cues_length + length of the
3725 // Cues element ID + EBML length of the Cues element. cues_end is
3726 // inclusive and the above sum is reduced by 1.
3727 uint64_t cues_length = 0, cues_id = 0, bytes_read = 0;
3728 bytes_read += ebml_read_num(matroska, matroska->ctx->pb, 4, &cues_id);
3729 bytes_read += ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
3730 cues_end = cues_start + cues_length + bytes_read - 1;
3732 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
3733 if (cues_start == -1 || cues_end == -1) return -1;
3736 matroska_parse_cues(matroska);
3739 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
3742 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
3745 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
3746 if (bandwidth < 0) return -1;
3747 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
3749 // check if all clusters start with key frames
3750 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
3752 // store cue point timestamps as a comma separated list for checking subsegment alignment in
3753 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
3754 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20 * sizeof(char));
3755 if (!buf) return -1;
3757 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
3758 snprintf(buf, (i + 1) * 20 * sizeof(char),
3759 "%s%" PRId64, buf, s->streams[0]->index_entries[i].timestamp);
3760 if (i != s->streams[0]->nb_index_entries - 1)
3761 strncat(buf, ",", sizeof(char));
3763 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
3769 static int webm_dash_manifest_read_header(AVFormatContext *s)
3772 int ret = matroska_read_header(s);
3773 MatroskaTrack *tracks;
3774 MatroskaDemuxContext *matroska = s->priv_data;
3776 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
3779 if (!s->nb_streams) {
3780 matroska_read_close(s);
3781 av_log(s, AV_LOG_ERROR, "No streams found\n");
3782 return AVERROR_INVALIDDATA;
3785 if (!matroska->is_live) {
3786 buf = av_asprintf("%g", matroska->duration);
3787 if (!buf) return AVERROR(ENOMEM);
3788 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
3791 // initialization range
3792 // 5 is the offset of Cluster ID.
3793 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, avio_tell(s->pb) - 5, 0);
3796 // basename of the file
3797 buf = strrchr(s->filename, '/');
3798 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->filename, 0);
3801 tracks = matroska->tracks.elem;
3802 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
3804 // parse the cues and populate Cue related fields
3805 return matroska->is_live ? 0 : webm_dash_manifest_cues(s);
3808 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
3813 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
3814 static const AVOption options[] = {
3815 { "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 },
3819 static const AVClass webm_dash_class = {
3820 .class_name = "WebM DASH Manifest demuxer",
3821 .item_name = av_default_item_name,
3823 .version = LIBAVUTIL_VERSION_INT,
3826 AVInputFormat ff_matroska_demuxer = {
3827 .name = "matroska,webm",
3828 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
3829 .extensions = "mkv,mk3d,mka,mks",
3830 .priv_data_size = sizeof(MatroskaDemuxContext),
3831 .read_probe = matroska_probe,
3832 .read_header = matroska_read_header,
3833 .read_packet = matroska_read_packet,
3834 .read_close = matroska_read_close,
3835 .read_seek = matroska_read_seek,
3836 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
3839 AVInputFormat ff_webm_dash_manifest_demuxer = {
3840 .name = "webm_dash_manifest",
3841 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
3842 .priv_data_size = sizeof(MatroskaDemuxContext),
3843 .read_header = webm_dash_manifest_read_header,
3844 .read_packet = webm_dash_manifest_read_packet,
3845 .read_close = matroska_read_close,
3846 .priv_class = &webm_dash_class,