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"
46 #include "libavutil/spherical.h"
48 #include "libavcodec/bytestream.h"
49 #include "libavcodec/flac.h"
50 #include "libavcodec/mpeg4audio.h"
53 #include "avio_internal.h"
58 /* For ff_codec_get_id(). */
69 #include "qtpalette.h"
71 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
72 #define NEEDS_CHECKING 2 /* Indicates that some error checks
73 * still need to be performed */
74 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
75 * syntax level used for parsing ended. */
91 typedef const struct EbmlSyntax {
101 const struct EbmlSyntax *n;
105 typedef struct EbmlList {
110 typedef struct EbmlBin {
117 typedef struct Ebml {
122 uint64_t doctype_version;
125 typedef struct MatroskaTrackCompression {
128 } MatroskaTrackCompression;
130 typedef struct MatroskaTrackEncryption {
133 } MatroskaTrackEncryption;
135 typedef struct MatroskaTrackEncoding {
138 MatroskaTrackCompression compression;
139 MatroskaTrackEncryption encryption;
140 } MatroskaTrackEncoding;
142 typedef struct MatroskaMasteringMeta {
151 double max_luminance;
152 double min_luminance;
153 } MatroskaMasteringMeta;
155 typedef struct MatroskaTrackVideoColor {
156 uint64_t matrix_coefficients;
157 uint64_t bits_per_channel;
158 uint64_t chroma_sub_horz;
159 uint64_t chroma_sub_vert;
160 uint64_t cb_sub_horz;
161 uint64_t cb_sub_vert;
162 uint64_t chroma_siting_horz;
163 uint64_t chroma_siting_vert;
165 uint64_t transfer_characteristics;
169 MatroskaMasteringMeta mastering_meta;
170 } MatroskaTrackVideoColor;
172 typedef struct MatroskaTrackVideoProjection {
178 } MatroskaTrackVideoProjection;
180 typedef struct MatroskaTrackVideo {
182 uint64_t display_width;
183 uint64_t display_height;
184 uint64_t pixel_width;
185 uint64_t pixel_height;
187 uint64_t display_unit;
189 uint64_t field_order;
190 uint64_t stereo_mode;
193 MatroskaTrackVideoProjection projection;
194 } MatroskaTrackVideo;
196 typedef struct MatroskaTrackAudio {
198 double out_samplerate;
202 /* real audio header (extracted from extradata) */
209 uint64_t buf_timecode;
211 } MatroskaTrackAudio;
213 typedef struct MatroskaTrackPlane {
216 } MatroskaTrackPlane;
218 typedef struct MatroskaTrackOperation {
219 EbmlList combine_planes;
220 } MatroskaTrackOperation;
222 typedef struct MatroskaTrack {
231 uint64_t default_duration;
232 uint64_t flag_default;
233 uint64_t flag_forced;
234 uint64_t seek_preroll;
235 MatroskaTrackVideo video;
236 MatroskaTrackAudio audio;
237 MatroskaTrackOperation operation;
239 uint64_t codec_delay;
240 uint64_t codec_delay_in_track_tb;
243 int64_t end_timecode;
245 uint64_t max_block_additional_id;
247 uint32_t palette[AVPALETTE_COUNT];
251 typedef struct MatroskaAttachment {
258 } MatroskaAttachment;
260 typedef struct MatroskaChapter {
269 typedef struct MatroskaIndexPos {
274 typedef struct MatroskaIndex {
279 typedef struct MatroskaTag {
287 typedef struct MatroskaTagTarget {
295 typedef struct MatroskaTags {
296 MatroskaTagTarget target;
300 typedef struct MatroskaSeekhead {
305 typedef struct MatroskaLevel {
310 typedef struct MatroskaBlock {
315 uint64_t additional_id;
317 int64_t discard_padding;
320 typedef struct MatroskaCluster {
326 typedef struct MatroskaLevel1Element {
330 } MatroskaLevel1Element;
332 typedef struct MatroskaDemuxContext {
333 const AVClass *class;
334 AVFormatContext *ctx;
338 MatroskaLevel levels[EBML_MAX_DEPTH];
348 EbmlList attachments;
354 /* byte position of the segment inside the stream */
355 int64_t segment_start;
357 /* the packet queue */
359 AVPacketList *queue_end;
363 /* What to skip before effectively reading a packet. */
364 int skip_to_keyframe;
365 uint64_t skip_to_timecode;
367 /* File has a CUES element, but we defer parsing until it is needed. */
368 int cues_parsing_deferred;
370 /* Level1 elements and whether they were read yet */
371 MatroskaLevel1Element level1_elems[64];
372 int num_level1_elems;
374 MatroskaCluster current_cluster;
376 /* WebM DASH Manifest live flag */
379 /* Bandwidth value for WebM DASH Manifest */
381 } MatroskaDemuxContext;
383 #define CHILD_OF(parent) { .def = { .n = parent } }
385 static const EbmlSyntax ebml_syntax[], matroska_segment[], matroska_track_video_color[], matroska_track_video[],
386 matroska_track[], matroska_track_encoding[], matroska_track_encodings[],
387 matroska_track_combine_planes[], matroska_track_operation[], matroska_tracks[],
388 matroska_attachments[], matroska_chapter_entry[], matroska_chapter[], matroska_chapters[],
389 matroska_index_entry[], matroska_index[], matroska_tag[], matroska_tags[], matroska_seekhead[],
390 matroska_blockadditions[], matroska_blockgroup[], matroska_cluster_parsing[];
392 static const EbmlSyntax ebml_header[] = {
393 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
394 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
395 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
396 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
397 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
398 { EBML_ID_EBMLVERSION, EBML_NONE },
399 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
400 CHILD_OF(ebml_syntax)
403 static const EbmlSyntax ebml_syntax[] = {
404 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
405 { MATROSKA_ID_SEGMENT, EBML_STOP },
409 static const EbmlSyntax matroska_info[] = {
410 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
411 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
412 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
413 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
414 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
415 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
416 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
417 CHILD_OF(matroska_segment)
420 static const EbmlSyntax matroska_mastering_meta[] = {
421 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
422 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
423 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
424 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
425 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
426 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
427 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
428 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
429 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
430 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
431 CHILD_OF(matroska_track_video_color)
434 static const EbmlSyntax matroska_track_video_color[] = {
435 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
436 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=0 } },
437 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
438 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
439 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
440 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
441 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
442 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
443 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
444 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
445 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
446 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
447 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
448 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
449 CHILD_OF(matroska_track_video)
452 static const EbmlSyntax matroska_track_video_projection[] = {
453 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
454 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrackVideoProjection, private) },
455 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f=0.0 } },
456 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f=0.0 } },
457 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f=0.0 } },
458 CHILD_OF(matroska_track_video)
461 static const EbmlSyntax matroska_track_video[] = {
462 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
463 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
464 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
465 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
466 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
467 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
468 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
469 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
470 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
471 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
472 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
473 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
474 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
475 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
476 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
477 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
478 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
479 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
480 CHILD_OF(matroska_track)
483 static const EbmlSyntax matroska_track_audio[] = {
484 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
485 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
486 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
487 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
488 CHILD_OF(matroska_track)
491 static const EbmlSyntax matroska_track_encoding_compression[] = {
492 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
493 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
494 CHILD_OF(matroska_track_encoding)
497 static const EbmlSyntax matroska_track_encoding_encryption[] = {
498 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
499 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
500 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
501 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
502 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
503 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
504 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
505 CHILD_OF(matroska_track_encoding)
507 static const EbmlSyntax matroska_track_encoding[] = {
508 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
509 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
510 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
511 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
512 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
513 CHILD_OF(matroska_track_encodings)
516 static const EbmlSyntax matroska_track_encodings[] = {
517 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
518 CHILD_OF(matroska_track)
521 static const EbmlSyntax matroska_track_plane[] = {
522 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
523 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
524 CHILD_OF(matroska_track_combine_planes)
527 static const EbmlSyntax matroska_track_combine_planes[] = {
528 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
529 CHILD_OF(matroska_track_operation)
532 static const EbmlSyntax matroska_track_operation[] = {
533 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
534 CHILD_OF(matroska_track)
537 static const EbmlSyntax matroska_track[] = {
538 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
539 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
540 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
541 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
542 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
543 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
544 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
545 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
546 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
547 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
548 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
549 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
550 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
551 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
552 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
553 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
554 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
555 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
556 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
557 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
558 { MATROSKA_ID_CODECNAME, EBML_NONE },
559 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
560 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
561 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
562 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
563 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
564 CHILD_OF(matroska_tracks)
567 static const EbmlSyntax matroska_tracks[] = {
568 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
569 CHILD_OF(matroska_segment)
572 static const EbmlSyntax matroska_attachment[] = {
573 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
574 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
575 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
576 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
577 { MATROSKA_ID_FILEDESC, EBML_NONE },
578 CHILD_OF(matroska_attachments)
581 static const EbmlSyntax matroska_attachments[] = {
582 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
583 CHILD_OF(matroska_segment)
586 static const EbmlSyntax matroska_chapter_display[] = {
587 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
588 { MATROSKA_ID_CHAPLANG, EBML_NONE },
589 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
590 CHILD_OF(matroska_chapter_entry)
593 static const EbmlSyntax matroska_chapter_entry[] = {
594 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
595 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
596 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
597 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
598 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
599 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
600 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
601 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
602 CHILD_OF(matroska_chapter)
605 static const EbmlSyntax matroska_chapter[] = {
606 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
607 { MATROSKA_ID_EDITIONUID, EBML_NONE },
608 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
609 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
610 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
611 CHILD_OF(matroska_chapters)
614 static const EbmlSyntax matroska_chapters[] = {
615 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
616 CHILD_OF(matroska_segment)
619 static const EbmlSyntax matroska_index_pos[] = {
620 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
621 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
622 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
623 { MATROSKA_ID_CUEDURATION, EBML_NONE },
624 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
625 CHILD_OF(matroska_index_entry)
628 static const EbmlSyntax matroska_index_entry[] = {
629 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
630 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
631 CHILD_OF(matroska_index)
634 static const EbmlSyntax matroska_index[] = {
635 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
636 CHILD_OF(matroska_segment)
639 static const EbmlSyntax matroska_simpletag[] = {
640 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
641 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
642 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
643 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
644 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
645 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
646 CHILD_OF(matroska_tag)
649 static const EbmlSyntax matroska_tagtargets[] = {
650 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
651 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
652 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
653 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
654 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
655 CHILD_OF(matroska_tag)
658 static const EbmlSyntax matroska_tag[] = {
659 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
660 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
661 CHILD_OF(matroska_tags)
664 static const EbmlSyntax matroska_tags[] = {
665 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
666 CHILD_OF(matroska_segment)
669 static const EbmlSyntax matroska_seekhead_entry[] = {
670 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
671 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
672 CHILD_OF(matroska_seekhead)
675 static const EbmlSyntax matroska_seekhead[] = {
676 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
677 CHILD_OF(matroska_segment)
680 static const EbmlSyntax matroska_segment[] = {
681 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
682 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
683 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
684 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
685 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
686 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
687 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
688 { MATROSKA_ID_CLUSTER, EBML_STOP },
689 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
692 static const EbmlSyntax matroska_segments[] = {
693 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
697 static const EbmlSyntax matroska_blockmore[] = {
698 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
699 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
700 CHILD_OF(matroska_blockadditions)
703 static const EbmlSyntax matroska_blockadditions[] = {
704 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
705 CHILD_OF(matroska_blockgroup)
708 static const EbmlSyntax matroska_blockgroup[] = {
709 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
710 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
711 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
712 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
713 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference), { .i = INT64_MIN } },
714 { MATROSKA_ID_CODECSTATE, EBML_NONE },
715 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
716 CHILD_OF(matroska_cluster_parsing)
719 // The following array contains SimpleBlock and BlockGroup twice
720 // in order to reuse the other values for matroska_cluster_enter.
721 static const EbmlSyntax matroska_cluster_parsing[] = {
722 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
723 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, { .n = matroska_blockgroup } },
724 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
725 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
726 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
727 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
728 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
729 CHILD_OF(matroska_segment)
732 static const EbmlSyntax matroska_cluster_enter[] = {
733 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = &matroska_cluster_parsing[2] } },
737 static const EbmlSyntax matroska_clusters[] = {
738 { MATROSKA_ID_CLUSTER, EBML_STOP },
739 { MATROSKA_ID_CUES, EBML_NONE },
740 { MATROSKA_ID_TAGS, EBML_NONE },
741 { MATROSKA_ID_INFO, EBML_NONE },
742 { MATROSKA_ID_TRACKS, EBML_NONE },
743 { MATROSKA_ID_ATTACHMENTS, EBML_NONE },
744 { MATROSKA_ID_CHAPTERS, EBML_NONE },
745 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
746 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
750 static const char *const matroska_doctypes[] = { "matroska", "webm" };
752 static int matroska_read_close(AVFormatContext *s);
755 * This function prepares the status for parsing of level 1 elements.
757 static int matroska_reset_status(MatroskaDemuxContext *matroska,
758 uint32_t id, int64_t position)
761 int err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
766 matroska->current_id = id;
767 matroska->num_levels = 1;
768 matroska->resync_pos = avio_tell(matroska->ctx->pb);
770 matroska->resync_pos -= (av_log2(id) + 7) / 8;
775 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
777 AVIOContext *pb = matroska->ctx->pb;
780 /* Try to seek to the last position to resync from. If this doesn't work,
781 * we resync from the earliest position available: The start of the buffer. */
782 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
783 av_log(matroska->ctx, AV_LOG_WARNING,
784 "Seek to desired resync point failed. Seeking to "
785 "earliest point available instead.\n");
786 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
787 last_pos + 1), SEEK_SET);
792 // try to find a toplevel element
793 while (!avio_feof(pb)) {
794 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
795 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
796 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
797 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
798 /* Prepare the context for parsing of a level 1 element. */
799 matroska_reset_status(matroska, id, -1);
800 /* Given that we are here means that an error has occured,
801 * so treat the segment as unknown length in order not to
802 * discard valid data that happens to be beyond the designated
803 * end of the segment. */
804 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
807 id = (id << 8) | avio_r8(pb);
811 return pb->error ? pb->error : AVERROR_EOF;
815 * Read: an "EBML number", which is defined as a variable-length
816 * array of bytes. The first byte indicates the length by giving a
817 * number of 0-bits followed by a one. The position of the first
818 * "one" bit inside the first byte indicates the length of this
820 * Returns: number of bytes read, < 0 on error
822 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
823 int max_size, uint64_t *number, int eof_forbidden)
829 /* The first byte tells us the length in bytes - except when it is zero. */
834 /* get the length of the EBML number */
835 read = 8 - ff_log2_tab[total];
837 if (!total || read > max_size) {
838 pos = avio_tell(pb) - 1;
840 av_log(matroska->ctx, AV_LOG_ERROR,
841 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
842 "of an EBML number\n", pos, pos);
844 av_log(matroska->ctx, AV_LOG_ERROR,
845 "Length %d indicated by an EBML number's first byte 0x%02x "
846 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
847 read, (uint8_t) total, pos, pos, max_size);
849 return AVERROR_INVALIDDATA;
852 /* read out length */
853 total ^= 1 << ff_log2_tab[total];
855 total = (total << 8) | avio_r8(pb);
857 if (pb->eof_reached) {
869 av_log(matroska->ctx, AV_LOG_ERROR,
870 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
875 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
876 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
883 * Read a EBML length value.
884 * This needs special handling for the "unknown length" case which has multiple
887 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
890 int res = ebml_read_num(matroska, pb, 8, number, 1);
891 if (res > 0 && *number + 1 == 1ULL << (7 * res))
892 *number = EBML_UNKNOWN_LENGTH;
897 * Read the next element as an unsigned int.
898 * Returns NEEDS_CHECKING.
900 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
904 /* big-endian ordering; build up number */
907 *num = (*num << 8) | avio_r8(pb);
909 return NEEDS_CHECKING;
913 * Read the next element as a signed int.
914 * Returns NEEDS_CHECKING.
916 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
923 *num = sign_extend(avio_r8(pb), 8);
925 /* big-endian ordering; build up number */
927 *num = ((uint64_t)*num << 8) | avio_r8(pb);
930 return NEEDS_CHECKING;
934 * Read the next element as a float.
935 * Returns NEEDS_CHECKING or < 0 on obvious failure.
937 static int ebml_read_float(AVIOContext *pb, int size, double *num)
942 *num = av_int2float(avio_rb32(pb));
944 *num = av_int2double(avio_rb64(pb));
946 return AVERROR_INVALIDDATA;
948 return NEEDS_CHECKING;
952 * Read the next element as an ASCII string.
953 * 0 is success, < 0 or NEEDS_CHECKING is failure.
955 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
960 /* EBML strings are usually not 0-terminated, so we allocate one
961 * byte more, read the string and NULL-terminate it ourselves. */
962 if (!(res = av_malloc(size + 1)))
963 return AVERROR(ENOMEM);
964 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
966 return ret < 0 ? ret : NEEDS_CHECKING;
976 * Read the next element as binary data.
977 * 0 is success, < 0 or NEEDS_CHECKING is failure.
979 static int ebml_read_binary(AVIOContext *pb, int length,
980 int64_t pos, EbmlBin *bin)
984 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
987 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
989 bin->data = bin->buf->data;
992 if ((ret = avio_read(pb, bin->data, length)) != length) {
993 av_buffer_unref(&bin->buf);
996 return ret < 0 ? ret : NEEDS_CHECKING;
1003 * Read the next element, but only the header. The contents
1004 * are supposed to be sub-elements which can be read separately.
1005 * 0 is success, < 0 is failure.
1007 static int ebml_read_master(MatroskaDemuxContext *matroska,
1008 uint64_t length, int64_t pos)
1010 MatroskaLevel *level;
1012 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1013 av_log(matroska->ctx, AV_LOG_ERROR,
1014 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1015 return AVERROR(ENOSYS);
1018 level = &matroska->levels[matroska->num_levels++];
1020 level->length = length;
1026 * Read signed/unsigned "EBML" numbers.
1027 * Return: number of bytes processed, < 0 on error
1029 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
1030 uint8_t *data, uint32_t size, uint64_t *num)
1033 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
1034 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num, 1);
1038 * Same as above, but signed.
1040 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1041 uint8_t *data, uint32_t size, int64_t *num)
1046 /* read as unsigned number first */
1047 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
1050 /* make signed (weird way) */
1051 *num = unum - ((1LL << (7 * res - 1)) - 1);
1056 static int ebml_parse(MatroskaDemuxContext *matroska,
1057 EbmlSyntax *syntax, void *data);
1059 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1063 // Whoever touches this should be aware of the duplication
1064 // existing in matroska_cluster_parsing.
1065 for (i = 0; syntax[i].id; i++)
1066 if (id == syntax[i].id)
1072 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1077 for (i = 0; syntax[i].id; i++)
1078 switch (syntax[i].type) {
1080 *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1083 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1086 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1090 // the default may be NULL
1091 if (syntax[i].def.s) {
1092 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1093 *dst = av_strdup(syntax[i].def.s);
1095 return AVERROR(ENOMEM);
1100 if (!matroska->levels[matroska->num_levels - 1].length) {
1101 matroska->num_levels--;
1106 res = ebml_parse(matroska, syntax, data);
1109 return res == LEVEL_ENDED ? 0 : res;
1112 static int is_ebml_id_valid(uint32_t id)
1114 // Due to endian nonsense in Matroska, the highest byte with any bits set
1115 // will contain the leading length bit. This bit in turn identifies the
1116 // total byte length of the element by its position within the byte.
1117 unsigned int bits = av_log2(id);
1118 return id && (bits + 7) / 8 == (8 - bits % 8);
1122 * Allocate and return the entry for the level1 element with the given ID. If
1123 * an entry already exists, return the existing entry.
1125 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1129 MatroskaLevel1Element *elem;
1131 if (!is_ebml_id_valid(id))
1134 // Some files link to all clusters; useless.
1135 if (id == MATROSKA_ID_CLUSTER)
1138 // There can be multiple seekheads.
1139 if (id != MATROSKA_ID_SEEKHEAD) {
1140 for (i = 0; i < matroska->num_level1_elems; i++) {
1141 if (matroska->level1_elems[i].id == id)
1142 return &matroska->level1_elems[i];
1146 // Only a completely broken file would have more elements.
1147 // It also provides a low-effort way to escape from circular seekheads
1148 // (every iteration will add a level1 entry).
1149 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1150 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1154 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1155 *elem = (MatroskaLevel1Element){.id = id};
1160 static int ebml_parse(MatroskaDemuxContext *matroska,
1161 EbmlSyntax *syntax, void *data)
1163 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1167 // max. 16 MB for strings
1168 [EBML_STR] = 0x1000000,
1169 [EBML_UTF8] = 0x1000000,
1170 // max. 256 MB for binary data
1171 [EBML_BIN] = 0x10000000,
1172 // no limits for anything else
1174 AVIOContext *pb = matroska->ctx->pb;
1177 int64_t pos = avio_tell(pb), pos_alt;
1178 int res, update_pos = 1, level_check;
1180 MatroskaLevel1Element *level1_elem;
1181 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1183 if (!matroska->current_id) {
1185 res = ebml_read_num(matroska, pb, 4, &id, 0);
1187 if (pb->eof_reached && res == AVERROR_EOF) {
1188 if (matroska->is_live)
1189 // in live mode, finish parsing if EOF is reached.
1191 if (level && pos == avio_tell(pb)) {
1192 if (level->length == EBML_UNKNOWN_LENGTH) {
1193 // Unknown-length levels automatically end at EOF.
1194 matroska->num_levels--;
1197 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1198 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1204 matroska->current_id = id | 1 << 7 * res;
1205 pos_alt = pos + res;
1208 pos -= (av_log2(matroska->current_id) + 7) / 8;
1211 id = matroska->current_id;
1213 syntax = ebml_parse_id(syntax, id);
1214 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1215 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1216 // Unknown-length levels end when an element from an upper level
1217 // in the hierarchy is encountered.
1218 while (syntax->def.n) {
1219 syntax = ebml_parse_id(syntax->def.n, id);
1221 matroska->num_levels--;
1227 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1228 "%"PRId64"\n", id, pos);
1229 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1232 data = (char *) data + syntax->data_offset;
1233 if (syntax->list_elem_size) {
1234 EbmlList *list = data;
1235 newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1237 return AVERROR(ENOMEM);
1238 list->elem = newelem;
1239 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1240 memset(data, 0, syntax->list_elem_size);
1244 if (syntax->type != EBML_STOP) {
1245 matroska->current_id = 0;
1246 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1248 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1249 av_log(matroska->ctx, AV_LOG_ERROR,
1250 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1251 length, max_lengths[syntax->type], syntax->type);
1252 return AVERROR_INVALIDDATA;
1257 if (matroska->num_levels > 0) {
1258 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
1260 if (length != EBML_UNKNOWN_LENGTH &&
1261 level->length != EBML_UNKNOWN_LENGTH) {
1262 uint64_t elem_end = pos_alt + length,
1263 level_end = level->start + level->length;
1265 if (elem_end < level_end) {
1267 } else if (elem_end == level_end) {
1268 level_check = LEVEL_ENDED;
1270 av_log(matroska->ctx, AV_LOG_ERROR,
1271 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1272 "containing master element ending at 0x%"PRIx64"\n",
1273 pos, elem_end, level_end);
1274 return AVERROR_INVALIDDATA;
1276 } else if (length != EBML_UNKNOWN_LENGTH) {
1278 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1279 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1280 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1281 return AVERROR_INVALIDDATA;
1282 } else if (id != MATROSKA_ID_CLUSTER) {
1283 // According to the specifications only clusters and segments
1284 // are allowed to be unknown-sized.
1285 av_log(matroska->ctx, AV_LOG_ERROR,
1286 "Found unknown-sized element other than a cluster at "
1287 "0x%"PRIx64". Dropping the invalid element.\n", pos);
1288 return AVERROR_INVALIDDATA;
1295 // We have found an element that is allowed at this place
1296 // in the hierarchy and it passed all checks, so treat the beginning
1297 // of the element as the "last known good" position.
1298 matroska->resync_pos = pos;
1302 switch (syntax->type) {
1304 res = ebml_read_uint(pb, length, data);
1307 res = ebml_read_sint(pb, length, data);
1310 res = ebml_read_float(pb, length, data);
1314 res = ebml_read_ascii(pb, length, data);
1317 res = ebml_read_binary(pb, length, pos_alt, data);
1321 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1323 if (id == MATROSKA_ID_SEGMENT)
1324 matroska->segment_start = pos_alt;
1325 if (id == MATROSKA_ID_CUES)
1326 matroska->cues_parsing_deferred = 0;
1327 if (syntax->type == EBML_LEVEL1 &&
1328 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1329 if (level1_elem->parsed)
1330 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1331 level1_elem->parsed = 1;
1333 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1341 if (ffio_limit(pb, length) != length) {
1342 // ffio_limit emits its own error message,
1343 // so we don't have to.
1344 return AVERROR(EIO);
1346 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1347 // avio_skip might take us past EOF. We check for this
1348 // by skipping only length - 1 bytes, reading a byte and
1349 // checking the error flags. This is done in order to check
1350 // that the element has been properly skipped even when
1351 // no filesize (that ffio_limit relies on) is available.
1353 res = NEEDS_CHECKING;
1360 if (res == NEEDS_CHECKING) {
1361 if (pb->eof_reached) {
1370 if (res == AVERROR_INVALIDDATA)
1371 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1372 else if (res == AVERROR(EIO))
1373 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1374 else if (res == AVERROR_EOF) {
1375 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1383 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1384 level = &matroska->levels[matroska->num_levels - 1];
1385 pos = avio_tell(pb);
1387 // Given that pos >= level->start no check for
1388 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1389 while (matroska->num_levels && pos == level->start + level->length) {
1390 matroska->num_levels--;
1398 static void ebml_free(EbmlSyntax *syntax, void *data)
1401 for (i = 0; syntax[i].id; i++) {
1402 void *data_off = (char *) data + syntax[i].data_offset;
1403 switch (syntax[i].type) {
1409 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1413 if (syntax[i].list_elem_size) {
1414 EbmlList *list = data_off;
1415 char *ptr = list->elem;
1416 for (j = 0; j < list->nb_elem;
1417 j++, ptr += syntax[i].list_elem_size)
1418 ebml_free(syntax[i].def.n, ptr);
1419 av_freep(&list->elem);
1422 ebml_free(syntax[i].def.n, data_off);
1432 static int matroska_probe(const AVProbeData *p)
1435 int len_mask = 0x80, size = 1, n = 1, i;
1438 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1441 /* length of header */
1443 while (size <= 8 && !(total & len_mask)) {
1449 total &= (len_mask - 1);
1451 total = (total << 8) | p->buf[4 + n++];
1453 /* Does the probe data contain the whole header? */
1454 if (p->buf_size < 4 + size + total)
1457 /* The header should contain a known document type. For now,
1458 * we don't parse the whole header but simply check for the
1459 * availability of that array of characters inside the header.
1460 * Not fully fool-proof, but good enough. */
1461 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1462 size_t probelen = strlen(matroska_doctypes[i]);
1463 if (total < probelen)
1465 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1466 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1467 return AVPROBE_SCORE_MAX;
1470 // probably valid EBML header but no recognized doctype
1471 return AVPROBE_SCORE_EXTENSION;
1474 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1477 MatroskaTrack *tracks = matroska->tracks.elem;
1480 for (i = 0; i < matroska->tracks.nb_elem; i++)
1481 if (tracks[i].num == num)
1484 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1488 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1489 MatroskaTrack *track)
1491 MatroskaTrackEncoding *encodings = track->encodings.elem;
1492 uint8_t *data = *buf;
1493 int isize = *buf_size;
1494 uint8_t *pkt_data = NULL;
1495 uint8_t av_unused *newpktdata;
1496 int pkt_size = isize;
1500 if (pkt_size >= 10000000U)
1501 return AVERROR_INVALIDDATA;
1503 switch (encodings[0].compression.algo) {
1504 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1506 int header_size = encodings[0].compression.settings.size;
1507 uint8_t *header = encodings[0].compression.settings.data;
1509 if (header_size && !header) {
1510 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1517 pkt_size = isize + header_size;
1518 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1520 return AVERROR(ENOMEM);
1522 memcpy(pkt_data, header, header_size);
1523 memcpy(pkt_data + header_size, data, isize);
1527 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1529 olen = pkt_size *= 3;
1530 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1531 + AV_INPUT_BUFFER_PADDING_SIZE);
1533 result = AVERROR(ENOMEM);
1536 pkt_data = newpktdata;
1537 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1538 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1540 result = AVERROR_INVALIDDATA;
1547 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1549 z_stream zstream = { 0 };
1550 if (inflateInit(&zstream) != Z_OK)
1552 zstream.next_in = data;
1553 zstream.avail_in = isize;
1556 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1558 inflateEnd(&zstream);
1559 result = AVERROR(ENOMEM);
1562 pkt_data = newpktdata;
1563 zstream.avail_out = pkt_size - zstream.total_out;
1564 zstream.next_out = pkt_data + zstream.total_out;
1565 result = inflate(&zstream, Z_NO_FLUSH);
1566 } while (result == Z_OK && pkt_size < 10000000);
1567 pkt_size = zstream.total_out;
1568 inflateEnd(&zstream);
1569 if (result != Z_STREAM_END) {
1570 if (result == Z_MEM_ERROR)
1571 result = AVERROR(ENOMEM);
1573 result = AVERROR_INVALIDDATA;
1580 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1582 bz_stream bzstream = { 0 };
1583 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1585 bzstream.next_in = data;
1586 bzstream.avail_in = isize;
1589 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1591 BZ2_bzDecompressEnd(&bzstream);
1592 result = AVERROR(ENOMEM);
1595 pkt_data = newpktdata;
1596 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1597 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1598 result = BZ2_bzDecompress(&bzstream);
1599 } while (result == BZ_OK && pkt_size < 10000000);
1600 pkt_size = bzstream.total_out_lo32;
1601 BZ2_bzDecompressEnd(&bzstream);
1602 if (result != BZ_STREAM_END) {
1603 if (result == BZ_MEM_ERROR)
1604 result = AVERROR(ENOMEM);
1606 result = AVERROR_INVALIDDATA;
1613 return AVERROR_INVALIDDATA;
1616 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1619 *buf_size = pkt_size;
1627 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1628 AVDictionary **metadata, char *prefix)
1630 MatroskaTag *tags = list->elem;
1634 for (i = 0; i < list->nb_elem; i++) {
1635 const char *lang = tags[i].lang &&
1636 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1638 if (!tags[i].name) {
1639 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1643 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1645 av_strlcpy(key, tags[i].name, sizeof(key));
1646 if (tags[i].def || !lang) {
1647 av_dict_set(metadata, key, tags[i].string, 0);
1648 if (tags[i].sub.nb_elem)
1649 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1652 av_strlcat(key, "-", sizeof(key));
1653 av_strlcat(key, lang, sizeof(key));
1654 av_dict_set(metadata, key, tags[i].string, 0);
1655 if (tags[i].sub.nb_elem)
1656 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1659 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1662 static void matroska_convert_tags(AVFormatContext *s)
1664 MatroskaDemuxContext *matroska = s->priv_data;
1665 MatroskaTags *tags = matroska->tags.elem;
1668 for (i = 0; i < matroska->tags.nb_elem; i++) {
1669 if (tags[i].target.attachuid) {
1670 MatroskaAttachment *attachment = matroska->attachments.elem;
1672 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1673 if (attachment[j].uid == tags[i].target.attachuid &&
1674 attachment[j].stream) {
1675 matroska_convert_tag(s, &tags[i].tag,
1676 &attachment[j].stream->metadata, NULL);
1681 av_log(NULL, AV_LOG_WARNING,
1682 "The tags at index %d refer to a "
1683 "non-existent attachment %"PRId64".\n",
1684 i, tags[i].target.attachuid);
1686 } else if (tags[i].target.chapteruid) {
1687 MatroskaChapter *chapter = matroska->chapters.elem;
1689 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1690 if (chapter[j].uid == tags[i].target.chapteruid &&
1691 chapter[j].chapter) {
1692 matroska_convert_tag(s, &tags[i].tag,
1693 &chapter[j].chapter->metadata, NULL);
1698 av_log(NULL, AV_LOG_WARNING,
1699 "The tags at index %d refer to a non-existent chapter "
1701 i, tags[i].target.chapteruid);
1703 } else if (tags[i].target.trackuid) {
1704 MatroskaTrack *track = matroska->tracks.elem;
1706 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1707 if (track[j].uid == tags[i].target.trackuid &&
1709 matroska_convert_tag(s, &tags[i].tag,
1710 &track[j].stream->metadata, NULL);
1715 av_log(NULL, AV_LOG_WARNING,
1716 "The tags at index %d refer to a non-existent track "
1718 i, tags[i].target.trackuid);
1721 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1722 tags[i].target.type);
1727 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1730 uint32_t saved_id = matroska->current_id;
1731 int64_t before_pos = avio_tell(matroska->ctx->pb);
1732 MatroskaLevel level;
1737 offset = pos + matroska->segment_start;
1738 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1739 /* We don't want to lose our seekhead level, so we add
1740 * a dummy. This is a crude hack. */
1741 if (matroska->num_levels == EBML_MAX_DEPTH) {
1742 av_log(matroska->ctx, AV_LOG_INFO,
1743 "Max EBML element depth (%d) reached, "
1744 "cannot parse further.\n", EBML_MAX_DEPTH);
1745 ret = AVERROR_INVALIDDATA;
1748 level.length = EBML_UNKNOWN_LENGTH;
1749 matroska->levels[matroska->num_levels] = level;
1750 matroska->num_levels++;
1751 matroska->current_id = 0;
1753 ret = ebml_parse(matroska, matroska_segment, matroska);
1754 if (ret == LEVEL_ENDED) {
1755 /* This can only happen if the seek brought us beyond EOF. */
1760 /* Seek back - notice that in all instances where this is used
1761 * it is safe to set the level to 1. */
1762 matroska_reset_status(matroska, saved_id, before_pos);
1767 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1769 EbmlList *seekhead_list = &matroska->seekhead;
1772 // we should not do any seeking in the streaming case
1773 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1776 for (i = 0; i < seekhead_list->nb_elem; i++) {
1777 MatroskaSeekhead *seekheads = seekhead_list->elem;
1778 uint32_t id = seekheads[i].id;
1779 uint64_t pos = seekheads[i].pos;
1781 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1782 if (!elem || elem->parsed)
1787 // defer cues parsing until we actually need cue data.
1788 if (id == MATROSKA_ID_CUES)
1791 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1792 // mark index as broken
1793 matroska->cues_parsing_deferred = -1;
1801 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1803 EbmlList *index_list;
1804 MatroskaIndex *index;
1805 uint64_t index_scale = 1;
1808 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1811 index_list = &matroska->index;
1812 index = index_list->elem;
1813 if (index_list->nb_elem < 2)
1815 if (index[1].time > 1E14 / matroska->time_scale) {
1816 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1819 for (i = 0; i < index_list->nb_elem; i++) {
1820 EbmlList *pos_list = &index[i].pos;
1821 MatroskaIndexPos *pos = pos_list->elem;
1822 for (j = 0; j < pos_list->nb_elem; j++) {
1823 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1825 if (track && track->stream)
1826 av_add_index_entry(track->stream,
1827 pos[j].pos + matroska->segment_start,
1828 index[i].time / index_scale, 0, 0,
1834 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1837 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1840 for (i = 0; i < matroska->num_level1_elems; i++) {
1841 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1842 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1843 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1844 matroska->cues_parsing_deferred = -1;
1850 matroska_add_index_entries(matroska);
1853 static int matroska_aac_profile(char *codec_id)
1855 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1858 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1859 if (strstr(codec_id, aac_profiles[profile]))
1864 static int matroska_aac_sri(int samplerate)
1868 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1869 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1874 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1876 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1877 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1880 static int matroska_parse_flac(AVFormatContext *s,
1881 MatroskaTrack *track,
1884 AVStream *st = track->stream;
1885 uint8_t *p = track->codec_priv.data;
1886 int size = track->codec_priv.size;
1888 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1889 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1890 track->codec_priv.size = 0;
1894 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1896 p += track->codec_priv.size;
1897 size -= track->codec_priv.size;
1899 /* parse the remaining metadata blocks if present */
1901 int block_last, block_type, block_size;
1903 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1907 if (block_size > size)
1910 /* check for the channel mask */
1911 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1912 AVDictionary *dict = NULL;
1913 AVDictionaryEntry *chmask;
1915 ff_vorbis_comment(s, &dict, p, block_size, 0);
1916 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1918 uint64_t mask = strtol(chmask->value, NULL, 0);
1919 if (!mask || mask & ~0x3ffffULL) {
1920 av_log(s, AV_LOG_WARNING,
1921 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1923 st->codecpar->channel_layout = mask;
1925 av_dict_free(&dict);
1935 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
1937 int major, minor, micro, bttb = 0;
1939 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
1940 * this function, and fixed in 57.52 */
1941 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, µ) == 3)
1942 bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
1944 switch (field_order) {
1945 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
1946 return AV_FIELD_PROGRESSIVE;
1947 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
1948 return AV_FIELD_UNKNOWN;
1949 case MATROSKA_VIDEO_FIELDORDER_TT:
1951 case MATROSKA_VIDEO_FIELDORDER_BB:
1953 case MATROSKA_VIDEO_FIELDORDER_BT:
1954 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
1955 case MATROSKA_VIDEO_FIELDORDER_TB:
1956 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
1958 return AV_FIELD_UNKNOWN;
1962 static void mkv_stereo_mode_display_mul(int stereo_mode,
1963 int *h_width, int *h_height)
1965 switch (stereo_mode) {
1966 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
1967 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
1968 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
1969 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
1970 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
1972 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
1973 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
1974 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
1975 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
1978 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
1979 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
1980 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
1981 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
1987 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
1988 const MatroskaTrackVideoColor *color = track->video.color.elem;
1989 const MatroskaMasteringMeta *mastering_meta;
1990 int has_mastering_primaries, has_mastering_luminance;
1992 if (!track->video.color.nb_elem)
1995 mastering_meta = &color->mastering_meta;
1996 // Mastering primaries are CIE 1931 coords, and must be > 0.
1997 has_mastering_primaries =
1998 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
1999 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2000 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2001 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2002 has_mastering_luminance = mastering_meta->max_luminance > 0;
2004 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2005 st->codecpar->color_space = color->matrix_coefficients;
2006 if (color->primaries != AVCOL_PRI_RESERVED &&
2007 color->primaries != AVCOL_PRI_RESERVED0)
2008 st->codecpar->color_primaries = color->primaries;
2009 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2010 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2011 st->codecpar->color_trc = color->transfer_characteristics;
2012 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2013 color->range <= AVCOL_RANGE_JPEG)
2014 st->codecpar->color_range = color->range;
2015 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2016 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2017 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2018 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2019 st->codecpar->chroma_location =
2020 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2021 (color->chroma_siting_vert - 1) << 7);
2023 if (color->max_cll && color->max_fall) {
2026 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2028 return AVERROR(ENOMEM);
2029 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2030 (uint8_t *)metadata, size);
2032 av_freep(&metadata);
2035 metadata->MaxCLL = color->max_cll;
2036 metadata->MaxFALL = color->max_fall;
2039 if (has_mastering_primaries || has_mastering_luminance) {
2040 // Use similar rationals as other standards.
2041 const int chroma_den = 50000;
2042 const int luma_den = 10000;
2043 AVMasteringDisplayMetadata *metadata =
2044 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2045 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2046 sizeof(AVMasteringDisplayMetadata));
2048 return AVERROR(ENOMEM);
2050 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2051 if (has_mastering_primaries) {
2052 metadata->display_primaries[0][0] = av_make_q(
2053 round(mastering_meta->r_x * chroma_den), chroma_den);
2054 metadata->display_primaries[0][1] = av_make_q(
2055 round(mastering_meta->r_y * chroma_den), chroma_den);
2056 metadata->display_primaries[1][0] = av_make_q(
2057 round(mastering_meta->g_x * chroma_den), chroma_den);
2058 metadata->display_primaries[1][1] = av_make_q(
2059 round(mastering_meta->g_y * chroma_den), chroma_den);
2060 metadata->display_primaries[2][0] = av_make_q(
2061 round(mastering_meta->b_x * chroma_den), chroma_den);
2062 metadata->display_primaries[2][1] = av_make_q(
2063 round(mastering_meta->b_y * chroma_den), chroma_den);
2064 metadata->white_point[0] = av_make_q(
2065 round(mastering_meta->white_x * chroma_den), chroma_den);
2066 metadata->white_point[1] = av_make_q(
2067 round(mastering_meta->white_y * chroma_den), chroma_den);
2068 metadata->has_primaries = 1;
2070 if (has_mastering_luminance) {
2071 metadata->max_luminance = av_make_q(
2072 round(mastering_meta->max_luminance * luma_den), luma_den);
2073 metadata->min_luminance = av_make_q(
2074 round(mastering_meta->min_luminance * luma_den), luma_den);
2075 metadata->has_luminance = 1;
2081 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track) {
2082 AVSphericalMapping *spherical;
2083 enum AVSphericalProjection projection;
2084 size_t spherical_size;
2085 uint32_t l = 0, t = 0, r = 0, b = 0;
2086 uint32_t padding = 0;
2090 bytestream2_init(&gb, track->video.projection.private.data,
2091 track->video.projection.private.size);
2093 if (bytestream2_get_byte(&gb) != 0) {
2094 av_log(NULL, AV_LOG_WARNING, "Unknown spherical metadata\n");
2098 bytestream2_skip(&gb, 3); // flags
2100 switch (track->video.projection.type) {
2101 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2102 if (track->video.projection.private.size == 20) {
2103 t = bytestream2_get_be32(&gb);
2104 b = bytestream2_get_be32(&gb);
2105 l = bytestream2_get_be32(&gb);
2106 r = bytestream2_get_be32(&gb);
2108 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2109 av_log(NULL, AV_LOG_ERROR,
2110 "Invalid bounding rectangle coordinates "
2111 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2113 return AVERROR_INVALIDDATA;
2115 } else if (track->video.projection.private.size != 0) {
2116 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2117 return AVERROR_INVALIDDATA;
2120 if (l || t || r || b)
2121 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2123 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2125 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2126 if (track->video.projection.private.size < 4) {
2127 av_log(NULL, AV_LOG_ERROR, "Missing projection private properties\n");
2128 return AVERROR_INVALIDDATA;
2129 } else if (track->video.projection.private.size == 12) {
2130 uint32_t layout = bytestream2_get_be32(&gb);
2132 av_log(NULL, AV_LOG_WARNING,
2133 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2136 projection = AV_SPHERICAL_CUBEMAP;
2137 padding = bytestream2_get_be32(&gb);
2139 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2140 return AVERROR_INVALIDDATA;
2143 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2144 /* No Spherical metadata */
2147 av_log(NULL, AV_LOG_WARNING,
2148 "Unknown spherical metadata type %"PRIu64"\n",
2149 track->video.projection.type);
2153 spherical = av_spherical_alloc(&spherical_size);
2155 return AVERROR(ENOMEM);
2157 spherical->projection = projection;
2159 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2160 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2161 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2163 spherical->padding = padding;
2165 spherical->bound_left = l;
2166 spherical->bound_top = t;
2167 spherical->bound_right = r;
2168 spherical->bound_bottom = b;
2170 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2173 av_freep(&spherical);
2180 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2182 const AVCodecTag *codec_tags;
2184 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2185 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2187 /* Normalize noncompliant private data that starts with the fourcc
2188 * by expanding/shifting the data by 4 bytes and storing the data
2189 * size at the start. */
2190 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2191 int ret = av_buffer_realloc(&track->codec_priv.buf,
2192 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2196 track->codec_priv.data = track->codec_priv.buf->data;
2197 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2198 track->codec_priv.size += 4;
2199 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2202 *fourcc = AV_RL32(track->codec_priv.data + 4);
2203 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2208 static int matroska_parse_tracks(AVFormatContext *s)
2210 MatroskaDemuxContext *matroska = s->priv_data;
2211 MatroskaTrack *tracks = matroska->tracks.elem;
2216 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2217 MatroskaTrack *track = &tracks[i];
2218 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2219 EbmlList *encodings_list = &track->encodings;
2220 MatroskaTrackEncoding *encodings = encodings_list->elem;
2221 uint8_t *extradata = NULL;
2222 int extradata_size = 0;
2223 int extradata_offset = 0;
2224 uint32_t fourcc = 0;
2226 char* key_id_base64 = NULL;
2229 /* Apply some sanity checks. */
2230 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2231 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2232 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2233 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2234 av_log(matroska->ctx, AV_LOG_INFO,
2235 "Unknown or unsupported track type %"PRIu64"\n",
2239 if (!track->codec_id)
2242 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2243 isnan(track->audio.samplerate)) {
2244 av_log(matroska->ctx, AV_LOG_WARNING,
2245 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2246 track->audio.samplerate);
2247 track->audio.samplerate = 8000;
2250 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2251 if (!track->default_duration && track->video.frame_rate > 0) {
2252 double default_duration = 1000000000 / track->video.frame_rate;
2253 if (default_duration > UINT64_MAX || default_duration < 0) {
2254 av_log(matroska->ctx, AV_LOG_WARNING,
2255 "Invalid frame rate %e. Cannot calculate default duration.\n",
2256 track->video.frame_rate);
2258 track->default_duration = default_duration;
2261 if (track->video.display_width == -1)
2262 track->video.display_width = track->video.pixel_width;
2263 if (track->video.display_height == -1)
2264 track->video.display_height = track->video.pixel_height;
2265 if (track->video.color_space.size == 4)
2266 fourcc = AV_RL32(track->video.color_space.data);
2267 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2268 if (!track->audio.out_samplerate)
2269 track->audio.out_samplerate = track->audio.samplerate;
2271 if (encodings_list->nb_elem > 1) {
2272 av_log(matroska->ctx, AV_LOG_ERROR,
2273 "Multiple combined encodings not supported");
2274 } else if (encodings_list->nb_elem == 1) {
2275 if (encodings[0].type) {
2276 if (encodings[0].encryption.key_id.size > 0) {
2277 /* Save the encryption key id to be stored later as a
2279 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2280 key_id_base64 = av_malloc(b64_size);
2281 if (key_id_base64 == NULL)
2282 return AVERROR(ENOMEM);
2284 av_base64_encode(key_id_base64, b64_size,
2285 encodings[0].encryption.key_id.data,
2286 encodings[0].encryption.key_id.size);
2288 encodings[0].scope = 0;
2289 av_log(matroska->ctx, AV_LOG_ERROR,
2290 "Unsupported encoding type");
2294 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2297 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2300 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2302 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2303 encodings[0].scope = 0;
2304 av_log(matroska->ctx, AV_LOG_ERROR,
2305 "Unsupported encoding type");
2306 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2307 uint8_t *codec_priv = track->codec_priv.data;
2308 int ret = matroska_decode_buffer(&track->codec_priv.data,
2309 &track->codec_priv.size,
2312 track->codec_priv.data = NULL;
2313 track->codec_priv.size = 0;
2314 av_log(matroska->ctx, AV_LOG_ERROR,
2315 "Failed to decode codec private data\n");
2318 if (codec_priv != track->codec_priv.data) {
2319 av_buffer_unref(&track->codec_priv.buf);
2320 if (track->codec_priv.data) {
2321 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2322 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2324 if (!track->codec_priv.buf) {
2325 av_freep(&track->codec_priv.data);
2326 track->codec_priv.size = 0;
2327 return AVERROR(ENOMEM);
2334 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2335 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2336 strlen(ff_mkv_codec_tags[j].str))) {
2337 codec_id = ff_mkv_codec_tags[j].id;
2342 st = track->stream = avformat_new_stream(s, NULL);
2344 av_free(key_id_base64);
2345 return AVERROR(ENOMEM);
2348 if (key_id_base64) {
2349 /* export encryption key id as base64 metadata tag */
2350 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
2351 av_freep(&key_id_base64);
2354 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2355 track->codec_priv.size >= 40 &&
2356 track->codec_priv.data) {
2357 track->ms_compat = 1;
2358 bit_depth = AV_RL16(track->codec_priv.data + 14);
2359 fourcc = AV_RL32(track->codec_priv.data + 16);
2360 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2363 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2365 extradata_offset = 40;
2366 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2367 track->codec_priv.size >= 14 &&
2368 track->codec_priv.data) {
2370 ffio_init_context(&b, track->codec_priv.data,
2371 track->codec_priv.size,
2372 0, NULL, NULL, NULL, NULL);
2373 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2376 codec_id = st->codecpar->codec_id;
2377 fourcc = st->codecpar->codec_tag;
2378 extradata_offset = FFMIN(track->codec_priv.size, 18);
2379 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2380 /* Normally 36, but allow noncompliant private data */
2381 && (track->codec_priv.size >= 32)
2382 && (track->codec_priv.data)) {
2383 uint16_t sample_size;
2384 int ret = get_qt_codec(track, &fourcc, &codec_id);
2387 sample_size = AV_RB16(track->codec_priv.data + 26);
2389 if (sample_size == 8) {
2390 fourcc = MKTAG('r','a','w',' ');
2391 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2392 } else if (sample_size == 16) {
2393 fourcc = MKTAG('t','w','o','s');
2394 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2397 if ((fourcc == MKTAG('t','w','o','s') ||
2398 fourcc == MKTAG('s','o','w','t')) &&
2400 codec_id = AV_CODEC_ID_PCM_S8;
2401 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2402 (track->codec_priv.size >= 21) &&
2403 (track->codec_priv.data)) {
2404 int ret = get_qt_codec(track, &fourcc, &codec_id);
2407 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2408 fourcc = MKTAG('S','V','Q','3');
2409 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2411 if (codec_id == AV_CODEC_ID_NONE)
2412 av_log(matroska->ctx, AV_LOG_ERROR,
2413 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2414 if (track->codec_priv.size >= 86) {
2415 bit_depth = AV_RB16(track->codec_priv.data + 82);
2416 ffio_init_context(&b, track->codec_priv.data,
2417 track->codec_priv.size,
2418 0, NULL, NULL, NULL, NULL);
2419 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2421 track->has_palette = 1;
2424 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2425 switch (track->audio.bitdepth) {
2427 codec_id = AV_CODEC_ID_PCM_U8;
2430 codec_id = AV_CODEC_ID_PCM_S24BE;
2433 codec_id = AV_CODEC_ID_PCM_S32BE;
2436 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2437 switch (track->audio.bitdepth) {
2439 codec_id = AV_CODEC_ID_PCM_U8;
2442 codec_id = AV_CODEC_ID_PCM_S24LE;
2445 codec_id = AV_CODEC_ID_PCM_S32LE;
2448 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2449 track->audio.bitdepth == 64) {
2450 codec_id = AV_CODEC_ID_PCM_F64LE;
2451 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2452 int profile = matroska_aac_profile(track->codec_id);
2453 int sri = matroska_aac_sri(track->audio.samplerate);
2454 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2456 return AVERROR(ENOMEM);
2457 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2458 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2459 if (strstr(track->codec_id, "SBR")) {
2460 sri = matroska_aac_sri(track->audio.out_samplerate);
2461 extradata[2] = 0x56;
2462 extradata[3] = 0xE5;
2463 extradata[4] = 0x80 | (sri << 3);
2467 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2468 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2469 * Create the "atom size", "tag", and "tag version" fields the
2470 * decoder expects manually. */
2471 extradata_size = 12 + track->codec_priv.size;
2472 extradata = av_mallocz(extradata_size +
2473 AV_INPUT_BUFFER_PADDING_SIZE);
2475 return AVERROR(ENOMEM);
2476 AV_WB32(extradata, extradata_size);
2477 memcpy(&extradata[4], "alac", 4);
2478 AV_WB32(&extradata[8], 0);
2479 memcpy(&extradata[12], track->codec_priv.data,
2480 track->codec_priv.size);
2481 } else if (codec_id == AV_CODEC_ID_TTA) {
2482 extradata_size = 30;
2483 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2485 return AVERROR(ENOMEM);
2486 ffio_init_context(&b, extradata, extradata_size, 1,
2487 NULL, NULL, NULL, NULL);
2488 avio_write(&b, "TTA1", 4);
2490 if (track->audio.channels > UINT16_MAX ||
2491 track->audio.bitdepth > UINT16_MAX) {
2492 av_log(matroska->ctx, AV_LOG_WARNING,
2493 "Too large audio channel number %"PRIu64
2494 " or bitdepth %"PRIu64". Skipping track.\n",
2495 track->audio.channels, track->audio.bitdepth);
2496 av_freep(&extradata);
2497 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2498 return AVERROR_INVALIDDATA;
2502 avio_wl16(&b, track->audio.channels);
2503 avio_wl16(&b, track->audio.bitdepth);
2504 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2505 return AVERROR_INVALIDDATA;
2506 avio_wl32(&b, track->audio.out_samplerate);
2507 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
2508 track->audio.out_samplerate,
2509 AV_TIME_BASE * 1000));
2510 } else if (codec_id == AV_CODEC_ID_RV10 ||
2511 codec_id == AV_CODEC_ID_RV20 ||
2512 codec_id == AV_CODEC_ID_RV30 ||
2513 codec_id == AV_CODEC_ID_RV40) {
2514 extradata_offset = 26;
2515 } else if (codec_id == AV_CODEC_ID_RA_144) {
2516 track->audio.out_samplerate = 8000;
2517 track->audio.channels = 1;
2518 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2519 codec_id == AV_CODEC_ID_COOK ||
2520 codec_id == AV_CODEC_ID_ATRAC3 ||
2521 codec_id == AV_CODEC_ID_SIPR)
2522 && track->codec_priv.data) {
2525 ffio_init_context(&b, track->codec_priv.data,
2526 track->codec_priv.size,
2527 0, NULL, NULL, NULL, NULL);
2529 flavor = avio_rb16(&b);
2530 track->audio.coded_framesize = avio_rb32(&b);
2532 track->audio.sub_packet_h = avio_rb16(&b);
2533 track->audio.frame_size = avio_rb16(&b);
2534 track->audio.sub_packet_size = avio_rb16(&b);
2536 track->audio.coded_framesize <= 0 ||
2537 track->audio.sub_packet_h <= 0 ||
2538 track->audio.frame_size <= 0 ||
2539 track->audio.sub_packet_size <= 0 && codec_id != AV_CODEC_ID_SIPR)
2540 return AVERROR_INVALIDDATA;
2541 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2542 track->audio.frame_size);
2543 if (!track->audio.buf)
2544 return AVERROR(ENOMEM);
2545 if (codec_id == AV_CODEC_ID_RA_288) {
2546 st->codecpar->block_align = track->audio.coded_framesize;
2547 track->codec_priv.size = 0;
2549 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2550 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2551 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2552 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2554 st->codecpar->block_align = track->audio.sub_packet_size;
2555 extradata_offset = 78;
2557 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2558 ret = matroska_parse_flac(s, track, &extradata_offset);
2561 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2562 fourcc = AV_RL32(track->codec_priv.data);
2563 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2564 /* we don't need any value stored in CodecPrivate.
2565 make sure that it's not exported as extradata. */
2566 track->codec_priv.size = 0;
2567 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2568 /* For now, propagate only the OBUs, if any. Once libavcodec is
2569 updated to handle isobmff style extradata this can be removed. */
2570 extradata_offset = 4;
2572 track->codec_priv.size -= extradata_offset;
2574 if (codec_id == AV_CODEC_ID_NONE)
2575 av_log(matroska->ctx, AV_LOG_INFO,
2576 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2578 if (track->time_scale < 0.01)
2579 track->time_scale = 1.0;
2580 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2581 1000 * 1000 * 1000); /* 64 bit pts in ns */
2583 /* convert the delay from ns to the track timebase */
2584 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2585 (AVRational){ 1, 1000000000 },
2588 st->codecpar->codec_id = codec_id;
2590 if (strcmp(track->language, "und"))
2591 av_dict_set(&st->metadata, "language", track->language, 0);
2592 av_dict_set(&st->metadata, "title", track->name, 0);
2594 if (track->flag_default)
2595 st->disposition |= AV_DISPOSITION_DEFAULT;
2596 if (track->flag_forced)
2597 st->disposition |= AV_DISPOSITION_FORCED;
2599 if (!st->codecpar->extradata) {
2601 st->codecpar->extradata = extradata;
2602 st->codecpar->extradata_size = extradata_size;
2603 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2604 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2605 return AVERROR(ENOMEM);
2606 memcpy(st->codecpar->extradata,
2607 track->codec_priv.data + extradata_offset,
2608 track->codec_priv.size);
2612 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2613 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2614 int display_width_mul = 1;
2615 int display_height_mul = 1;
2617 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2618 st->codecpar->codec_tag = fourcc;
2620 st->codecpar->bits_per_coded_sample = bit_depth;
2621 st->codecpar->width = track->video.pixel_width;
2622 st->codecpar->height = track->video.pixel_height;
2624 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2625 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2626 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2627 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2629 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2630 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2632 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2633 av_reduce(&st->sample_aspect_ratio.num,
2634 &st->sample_aspect_ratio.den,
2635 st->codecpar->height * track->video.display_width * display_width_mul,
2636 st->codecpar->width * track->video.display_height * display_height_mul,
2639 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2640 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2642 if (track->default_duration) {
2643 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2644 1000000000, track->default_duration, 30000);
2645 #if FF_API_R_FRAME_RATE
2646 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2647 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2648 st->r_frame_rate = st->avg_frame_rate;
2652 /* export stereo mode flag as metadata tag */
2653 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2654 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2656 /* export alpha mode flag as metadata tag */
2657 if (track->video.alpha_mode)
2658 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2660 /* if we have virtual track, mark the real tracks */
2661 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2663 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2665 snprintf(buf, sizeof(buf), "%s_%d",
2666 ff_matroska_video_stereo_plane[planes[j].type], i);
2667 for (k=0; k < matroska->tracks.nb_elem; k++)
2668 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2669 av_dict_set(&tracks[k].stream->metadata,
2670 "stereo_mode", buf, 0);
2674 // add stream level stereo3d side data if it is a supported format
2675 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2676 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2677 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2682 ret = mkv_parse_video_color(st, track);
2685 ret = mkv_parse_video_projection(st, track);
2688 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2689 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2690 st->codecpar->codec_tag = fourcc;
2691 st->codecpar->sample_rate = track->audio.out_samplerate;
2692 st->codecpar->channels = track->audio.channels;
2693 if (!st->codecpar->bits_per_coded_sample)
2694 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2695 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2696 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2697 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2698 st->need_parsing = AVSTREAM_PARSE_FULL;
2699 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2700 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2701 if (track->codec_delay > 0) {
2702 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2703 (AVRational){1, 1000000000},
2704 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2705 48000 : st->codecpar->sample_rate});
2707 if (track->seek_preroll > 0) {
2708 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2709 (AVRational){1, 1000000000},
2710 (AVRational){1, st->codecpar->sample_rate});
2712 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2713 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2715 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2716 st->disposition |= AV_DISPOSITION_CAPTIONS;
2717 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2718 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2719 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2720 st->disposition |= AV_DISPOSITION_METADATA;
2722 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2723 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2730 static int matroska_read_header(AVFormatContext *s)
2732 MatroskaDemuxContext *matroska = s->priv_data;
2733 EbmlList *attachments_list = &matroska->attachments;
2734 EbmlList *chapters_list = &matroska->chapters;
2735 MatroskaAttachment *attachments;
2736 MatroskaChapter *chapters;
2737 uint64_t max_start = 0;
2743 matroska->cues_parsing_deferred = 1;
2745 /* First read the EBML header. */
2746 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2747 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2748 ebml_free(ebml_syntax, &ebml);
2749 return AVERROR_INVALIDDATA;
2751 if (ebml.version > EBML_VERSION ||
2752 ebml.max_size > sizeof(uint64_t) ||
2753 ebml.id_length > sizeof(uint32_t) ||
2754 ebml.doctype_version > 3) {
2755 avpriv_report_missing_feature(matroska->ctx,
2756 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2757 ebml.version, ebml.doctype, ebml.doctype_version);
2758 ebml_free(ebml_syntax, &ebml);
2759 return AVERROR_PATCHWELCOME;
2760 } else if (ebml.doctype_version == 3) {
2761 av_log(matroska->ctx, AV_LOG_WARNING,
2762 "EBML header using unsupported features\n"
2763 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2764 ebml.version, ebml.doctype, ebml.doctype_version);
2766 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2767 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2769 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2770 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2771 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2772 ebml_free(ebml_syntax, &ebml);
2773 return AVERROR_INVALIDDATA;
2776 ebml_free(ebml_syntax, &ebml);
2778 /* The next thing is a segment. */
2779 pos = avio_tell(matroska->ctx->pb);
2780 res = ebml_parse(matroska, matroska_segments, matroska);
2781 // try resyncing until we find a EBML_STOP type element.
2783 res = matroska_resync(matroska, pos);
2786 pos = avio_tell(matroska->ctx->pb);
2787 res = ebml_parse(matroska, matroska_segment, matroska);
2789 /* Set data_offset as it might be needed later by seek_frame_generic. */
2790 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2791 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2792 matroska_execute_seekhead(matroska);
2794 if (!matroska->time_scale)
2795 matroska->time_scale = 1000000;
2796 if (matroska->duration)
2797 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2798 1000 / AV_TIME_BASE;
2799 av_dict_set(&s->metadata, "title", matroska->title, 0);
2800 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2802 if (matroska->date_utc.size == 8)
2803 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2805 res = matroska_parse_tracks(s);
2809 attachments = attachments_list->elem;
2810 for (j = 0; j < attachments_list->nb_elem; j++) {
2811 if (!(attachments[j].filename && attachments[j].mime &&
2812 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2813 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2815 AVStream *st = avformat_new_stream(s, NULL);
2818 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2819 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2820 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2822 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2823 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2824 strlen(ff_mkv_image_mime_tags[i].str))) {
2825 st->codecpar->codec_id = ff_mkv_image_mime_tags[i].id;
2830 attachments[j].stream = st;
2832 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2833 AVPacket *pkt = &st->attached_pic;
2835 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2836 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2838 av_init_packet(pkt);
2839 pkt->buf = av_buffer_ref(attachments[j].bin.buf);
2841 return AVERROR(ENOMEM);
2842 pkt->data = attachments[j].bin.data;
2843 pkt->size = attachments[j].bin.size;
2844 pkt->stream_index = st->index;
2845 pkt->flags |= AV_PKT_FLAG_KEY;
2847 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2848 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2850 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2851 attachments[j].bin.size);
2853 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2854 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2855 strlen(ff_mkv_mime_tags[i].str))) {
2856 st->codecpar->codec_id = ff_mkv_mime_tags[i].id;
2864 chapters = chapters_list->elem;
2865 for (i = 0; i < chapters_list->nb_elem; i++)
2866 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2867 (max_start == 0 || chapters[i].start > max_start)) {
2868 chapters[i].chapter =
2869 avpriv_new_chapter(s, chapters[i].uid,
2870 (AVRational) { 1, 1000000000 },
2871 chapters[i].start, chapters[i].end,
2873 if (chapters[i].chapter) {
2874 av_dict_set(&chapters[i].chapter->metadata,
2875 "title", chapters[i].title, 0);
2877 max_start = chapters[i].start;
2880 matroska_add_index_entries(matroska);
2882 matroska_convert_tags(s);
2886 matroska_read_close(s);
2891 * Put one packet in an application-supplied AVPacket struct.
2892 * Returns 0 on success or -1 on failure.
2894 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2897 if (matroska->queue) {
2898 MatroskaTrack *tracks = matroska->tracks.elem;
2899 MatroskaTrack *track;
2901 ff_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
2902 track = &tracks[pkt->stream_index];
2903 if (track->has_palette) {
2904 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2906 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
2908 memcpy(pal, track->palette, AVPALETTE_SIZE);
2910 track->has_palette = 0;
2919 * Free all packets in our internal queue.
2921 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2923 ff_packet_list_free(&matroska->queue, &matroska->queue_end);
2926 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2927 int *buf_size, int type,
2928 uint32_t **lace_buf, int *laces)
2930 int res = 0, n, size = *buf_size;
2931 uint8_t *data = *buf;
2932 uint32_t *lace_size;
2936 *lace_buf = av_malloc(sizeof(**lace_buf));
2938 return AVERROR(ENOMEM);
2940 *lace_buf[0] = size;
2944 av_assert0(size > 0);
2948 lace_size = av_malloc_array(*laces, sizeof(*lace_size));
2950 return AVERROR(ENOMEM);
2953 case 0x1: /* Xiph lacing */
2957 for (n = 0; res == 0 && n < *laces - 1; n++) {
2961 if (size <= total) {
2962 res = AVERROR_INVALIDDATA;
2967 lace_size[n] += temp;
2974 if (size <= total) {
2975 res = AVERROR_INVALIDDATA;
2979 lace_size[n] = size - total;
2983 case 0x2: /* fixed-size lacing */
2984 if (size % (*laces)) {
2985 res = AVERROR_INVALIDDATA;
2988 for (n = 0; n < *laces; n++)
2989 lace_size[n] = size / *laces;
2992 case 0x3: /* EBML lacing */
2996 n = matroska_ebmlnum_uint(matroska, data, size, &num);
2997 if (n < 0 || num > INT_MAX) {
2998 av_log(matroska->ctx, AV_LOG_INFO,
2999 "EBML block data error\n");
3000 res = n<0 ? n : AVERROR_INVALIDDATA;
3005 total = lace_size[0] = num;
3006 for (n = 1; res == 0 && n < *laces - 1; n++) {
3009 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
3010 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
3011 av_log(matroska->ctx, AV_LOG_INFO,
3012 "EBML block data error\n");
3013 res = r<0 ? r : AVERROR_INVALIDDATA;
3018 lace_size[n] = lace_size[n - 1] + snum;
3019 total += lace_size[n];
3021 if (size <= total) {
3022 res = AVERROR_INVALIDDATA;
3025 lace_size[*laces - 1] = size - total;
3031 *lace_buf = lace_size;
3037 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3038 MatroskaTrack *track, AVStream *st,
3039 uint8_t *data, int size, uint64_t timecode,
3042 int a = st->codecpar->block_align;
3043 int sps = track->audio.sub_packet_size;
3044 int cfs = track->audio.coded_framesize;
3045 int h = track->audio.sub_packet_h;
3046 int y = track->audio.sub_packet_cnt;
3047 int w = track->audio.frame_size;
3050 if (!track->audio.pkt_cnt) {
3051 if (track->audio.sub_packet_cnt == 0)
3052 track->audio.buf_timecode = timecode;
3053 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3054 if (size < cfs * h / 2) {
3055 av_log(matroska->ctx, AV_LOG_ERROR,
3056 "Corrupt int4 RM-style audio packet size\n");
3057 return AVERROR_INVALIDDATA;
3059 for (x = 0; x < h / 2; x++)
3060 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3061 data + x * cfs, cfs);
3062 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3064 av_log(matroska->ctx, AV_LOG_ERROR,
3065 "Corrupt sipr RM-style audio packet size\n");
3066 return AVERROR_INVALIDDATA;
3068 memcpy(track->audio.buf + y * w, data, w);
3070 if (size < sps * w / sps || h<=0 || w%sps) {
3071 av_log(matroska->ctx, AV_LOG_ERROR,
3072 "Corrupt generic RM-style audio packet size\n");
3073 return AVERROR_INVALIDDATA;
3075 for (x = 0; x < w / sps; x++)
3076 memcpy(track->audio.buf +
3077 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3078 data + x * sps, sps);
3081 if (++track->audio.sub_packet_cnt >= h) {
3082 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3083 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3084 track->audio.sub_packet_cnt = 0;
3085 track->audio.pkt_cnt = h * w / a;
3089 while (track->audio.pkt_cnt) {
3091 AVPacket pktl, *pkt = &pktl;
3093 ret = av_new_packet(pkt, a);
3098 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3100 pkt->pts = track->audio.buf_timecode;
3101 track->audio.buf_timecode = AV_NOPTS_VALUE;
3103 pkt->stream_index = st->index;
3104 ret = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3106 av_packet_unref(pkt);
3107 return AVERROR(ENOMEM);
3114 /* reconstruct full wavpack blocks from mangled matroska ones */
3115 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
3116 uint8_t **pdst, int *size)
3118 uint8_t *dst = NULL;
3123 int ret, offset = 0;
3125 if (srclen < 12 || track->stream->codecpar->extradata_size < 2)
3126 return AVERROR_INVALIDDATA;
3128 ver = AV_RL16(track->stream->codecpar->extradata);
3130 samples = AV_RL32(src);
3134 while (srclen >= 8) {
3139 uint32_t flags = AV_RL32(src);
3140 uint32_t crc = AV_RL32(src + 4);
3144 multiblock = (flags & 0x1800) != 0x1800;
3147 ret = AVERROR_INVALIDDATA;
3150 blocksize = AV_RL32(src);
3156 if (blocksize > srclen) {
3157 ret = AVERROR_INVALIDDATA;
3161 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3163 ret = AVERROR(ENOMEM);
3167 dstlen += blocksize + 32;
3169 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3170 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3171 AV_WL16(dst + offset + 8, ver); // version
3172 AV_WL16(dst + offset + 10, 0); // track/index_no
3173 AV_WL32(dst + offset + 12, 0); // total samples
3174 AV_WL32(dst + offset + 16, 0); // block index
3175 AV_WL32(dst + offset + 20, samples); // number of samples
3176 AV_WL32(dst + offset + 24, flags); // flags
3177 AV_WL32(dst + offset + 28, crc); // crc
3178 memcpy(dst + offset + 32, src, blocksize); // block data
3181 srclen -= blocksize;
3182 offset += blocksize + 32;
3185 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3197 static int matroska_parse_prores(MatroskaTrack *track, uint8_t *src,
3198 uint8_t **pdst, int *size)
3203 if (AV_RB32(&src[4]) != MKBETAG('i', 'c', 'p', 'f')) {
3204 dst = av_malloc(dstlen + 8 + AV_INPUT_BUFFER_PADDING_SIZE);
3206 return AVERROR(ENOMEM);
3208 AV_WB32(dst, dstlen);
3209 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3210 memcpy(dst + 8, src, dstlen);
3211 memset(dst + 8 + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3221 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3222 MatroskaTrack *track,
3224 uint8_t *data, int data_len,
3229 AVPacket pktl, *pkt = &pktl;
3230 uint8_t *id, *settings, *text, *buf;
3231 int id_len, settings_len, text_len;
3236 return AVERROR_INVALIDDATA;
3239 q = data + data_len;
3244 if (*p == '\r' || *p == '\n') {
3253 if (p >= q || *p != '\n')
3254 return AVERROR_INVALIDDATA;
3260 if (*p == '\r' || *p == '\n') {
3261 settings_len = p - settings;
3269 if (p >= q || *p != '\n')
3270 return AVERROR_INVALIDDATA;
3275 while (text_len > 0) {
3276 const int len = text_len - 1;
3277 const uint8_t c = p[len];
3278 if (c != '\r' && c != '\n')
3284 return AVERROR_INVALIDDATA;
3286 err = av_new_packet(pkt, text_len);
3291 memcpy(pkt->data, text, text_len);
3294 buf = av_packet_new_side_data(pkt,
3295 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3298 av_packet_unref(pkt);
3299 return AVERROR(ENOMEM);
3301 memcpy(buf, id, id_len);
3304 if (settings_len > 0) {
3305 buf = av_packet_new_side_data(pkt,
3306 AV_PKT_DATA_WEBVTT_SETTINGS,
3309 av_packet_unref(pkt);
3310 return AVERROR(ENOMEM);
3312 memcpy(buf, settings, settings_len);
3315 // Do we need this for subtitles?
3316 // pkt->flags = AV_PKT_FLAG_KEY;
3318 pkt->stream_index = st->index;
3319 pkt->pts = timecode;
3321 // Do we need this for subtitles?
3322 // pkt->dts = timecode;
3324 pkt->duration = duration;
3327 err = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3329 av_packet_unref(pkt);
3330 return AVERROR(ENOMEM);
3336 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3337 MatroskaTrack *track, AVStream *st,
3338 AVBufferRef *buf, uint8_t *data, int pkt_size,
3339 uint64_t timecode, uint64_t lace_duration,
3340 int64_t pos, int is_keyframe,
3341 uint8_t *additional, uint64_t additional_id, int additional_size,
3342 int64_t discard_padding)
3344 MatroskaTrackEncoding *encodings = track->encodings.elem;
3345 uint8_t *pkt_data = data;
3347 AVPacket pktl, *pkt = &pktl;
3349 if (encodings && !encodings->type && encodings->scope & 1) {
3350 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
3355 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3357 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
3359 av_log(matroska->ctx, AV_LOG_ERROR,
3360 "Error parsing a wavpack block.\n");
3363 if (pkt_data != data)
3364 av_freep(&pkt_data);
3368 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES) {
3370 res = matroska_parse_prores(track, pkt_data, &pr_data, &pkt_size);
3372 av_log(matroska->ctx, AV_LOG_ERROR,
3373 "Error parsing a prores block.\n");
3376 if (pkt_data != data)
3377 av_freep(&pkt_data);
3381 av_init_packet(pkt);
3382 if (pkt_data != data)
3383 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3386 pkt->buf = av_buffer_ref(buf);
3389 res = AVERROR(ENOMEM);
3393 pkt->data = pkt_data;
3394 pkt->size = pkt_size;
3395 pkt->flags = is_keyframe;
3396 pkt->stream_index = st->index;
3398 if (additional_size > 0) {
3399 uint8_t *side_data = av_packet_new_side_data(pkt,
3400 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3401 additional_size + 8);
3403 av_packet_unref(pkt);
3404 return AVERROR(ENOMEM);
3406 AV_WB64(side_data, additional_id);
3407 memcpy(side_data + 8, additional, additional_size);
3410 if (discard_padding) {
3411 uint8_t *side_data = av_packet_new_side_data(pkt,
3412 AV_PKT_DATA_SKIP_SAMPLES,
3415 av_packet_unref(pkt);
3416 return AVERROR(ENOMEM);
3418 discard_padding = av_rescale_q(discard_padding,
3419 (AVRational){1, 1000000000},
3420 (AVRational){1, st->codecpar->sample_rate});
3421 if (discard_padding > 0) {
3422 AV_WL32(side_data + 4, discard_padding);
3424 AV_WL32(side_data, -discard_padding);
3428 if (track->ms_compat)
3429 pkt->dts = timecode;
3431 pkt->pts = timecode;
3433 pkt->duration = lace_duration;
3435 #if FF_API_CONVERGENCE_DURATION
3436 FF_DISABLE_DEPRECATION_WARNINGS
3437 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3438 pkt->convergence_duration = lace_duration;
3440 FF_ENABLE_DEPRECATION_WARNINGS
3443 res = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3445 av_packet_unref(pkt);
3446 return AVERROR(ENOMEM);
3452 if (pkt_data != data)
3453 av_freep(&pkt_data);
3457 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3458 int size, int64_t pos, uint64_t cluster_time,
3459 uint64_t block_duration, int is_keyframe,
3460 uint8_t *additional, uint64_t additional_id, int additional_size,
3461 int64_t cluster_pos, int64_t discard_padding)
3463 uint64_t timecode = AV_NOPTS_VALUE;
3464 MatroskaTrack *track;
3468 uint32_t *lace_size = NULL;
3469 int n, flags, laces = 0;
3471 int trust_default_duration = 1;
3473 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
3479 track = matroska_find_track_by_num(matroska, num);
3480 if (!track || !track->stream) {
3481 av_log(matroska->ctx, AV_LOG_INFO,
3482 "Invalid stream %"PRIu64"\n", num);
3483 return AVERROR_INVALIDDATA;
3484 } else if (size <= 3)
3487 if (st->discard >= AVDISCARD_ALL)
3489 av_assert1(block_duration != AV_NOPTS_VALUE);
3491 block_time = sign_extend(AV_RB16(data), 16);
3495 if (is_keyframe == -1)
3496 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3498 if (cluster_time != (uint64_t) -1 &&
3499 (block_time >= 0 || cluster_time >= -block_time)) {
3500 timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3501 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3502 timecode < track->end_timecode)
3503 is_keyframe = 0; /* overlapping subtitles are not key frame */
3505 ff_reduce_index(matroska->ctx, st->index);
3506 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3511 if (matroska->skip_to_keyframe &&
3512 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3513 // Compare signed timecodes. Timecode may be negative due to codec delay
3514 // offset. We don't support timestamps greater than int64_t anyway - see
3516 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3519 matroska->skip_to_keyframe = 0;
3520 else if (!st->skip_to_keyframe) {
3521 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3522 matroska->skip_to_keyframe = 0;
3526 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
3527 &lace_size, &laces);
3532 if (track->audio.samplerate == 8000) {
3533 // If this is needed for more codecs, then add them here
3534 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3535 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3536 trust_default_duration = 0;
3540 if (!block_duration && trust_default_duration)
3541 block_duration = track->default_duration * laces / matroska->time_scale;
3543 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3544 track->end_timecode =
3545 FFMAX(track->end_timecode, timecode + block_duration);
3547 for (n = 0; n < laces; n++) {
3548 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3550 if (lace_size[n] > size) {
3551 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
3555 if ((st->codecpar->codec_id == AV_CODEC_ID_RA_288 ||
3556 st->codecpar->codec_id == AV_CODEC_ID_COOK ||
3557 st->codecpar->codec_id == AV_CODEC_ID_SIPR ||
3558 st->codecpar->codec_id == AV_CODEC_ID_ATRAC3) &&
3559 st->codecpar->block_align && track->audio.sub_packet_size) {
3560 res = matroska_parse_rm_audio(matroska, track, st, data,
3566 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3567 res = matroska_parse_webvtt(matroska, track, st,
3569 timecode, lace_duration,
3574 res = matroska_parse_frame(matroska, track, st, buf, data, lace_size[n],
3575 timecode, lace_duration, pos,
3576 !n ? is_keyframe : 0,
3577 additional, additional_id, additional_size,
3583 if (timecode != AV_NOPTS_VALUE)
3584 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3585 data += lace_size[n];
3586 size -= lace_size[n];
3594 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3596 MatroskaCluster *cluster = &matroska->current_cluster;
3597 MatroskaBlock *block = &cluster->block;
3600 av_assert0(matroska->num_levels <= 2);
3602 if (matroska->num_levels == 1) {
3603 res = ebml_parse(matroska, matroska_clusters, NULL);
3606 /* Found a cluster: subtract the size of the ID already read. */
3607 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3609 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3615 if (matroska->num_levels == 2) {
3616 /* We are inside a cluster. */
3617 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3619 if (res >= 0 && block->bin.size > 0) {
3620 int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3621 uint8_t* additional = block->additional.size > 0 ?
3622 block->additional.data : NULL;
3624 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3625 block->bin.size, block->bin.pos,
3626 cluster->timecode, block->duration,
3627 is_keyframe, additional, block->additional_id,
3628 block->additional.size, cluster->pos,
3629 block->discard_padding);
3632 ebml_free(matroska_blockgroup, block);
3633 memset(block, 0, sizeof(*block));
3634 } else if (!matroska->num_levels) {
3635 if (!avio_feof(matroska->ctx->pb)) {
3636 avio_r8(matroska->ctx->pb);
3637 if (!avio_feof(matroska->ctx->pb)) {
3638 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3639 "end of segment.\n");
3640 return AVERROR_INVALIDDATA;
3650 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3652 MatroskaDemuxContext *matroska = s->priv_data;
3655 if (matroska->resync_pos == -1) {
3656 // This can only happen if generic seeking has been used.
3657 matroska->resync_pos = avio_tell(s->pb);
3660 while (matroska_deliver_packet(matroska, pkt)) {
3662 return (ret < 0) ? ret : AVERROR_EOF;
3663 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3664 ret = matroska_resync(matroska, matroska->resync_pos);
3670 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3671 int64_t timestamp, int flags)
3673 MatroskaDemuxContext *matroska = s->priv_data;
3674 MatroskaTrack *tracks = NULL;
3675 AVStream *st = s->streams[stream_index];
3678 /* Parse the CUES now since we need the index data to seek. */
3679 if (matroska->cues_parsing_deferred > 0) {
3680 matroska->cues_parsing_deferred = 0;
3681 matroska_parse_cues(matroska);
3684 if (!st->nb_index_entries)
3686 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3688 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3689 matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3690 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3691 matroska_clear_queue(matroska);
3692 if (matroska_parse_cluster(matroska) < 0)
3697 matroska_clear_queue(matroska);
3698 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3701 tracks = matroska->tracks.elem;
3702 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3703 tracks[i].audio.pkt_cnt = 0;
3704 tracks[i].audio.sub_packet_cnt = 0;
3705 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3706 tracks[i].end_timecode = 0;
3709 /* We seek to a level 1 element, so set the appropriate status. */
3710 matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3711 if (flags & AVSEEK_FLAG_ANY) {
3712 st->skip_to_keyframe = 0;
3713 matroska->skip_to_timecode = timestamp;
3715 st->skip_to_keyframe = 1;
3716 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3718 matroska->skip_to_keyframe = 1;
3720 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3723 // slightly hackish but allows proper fallback to
3724 // the generic seeking code.
3725 matroska_reset_status(matroska, 0, -1);
3726 matroska->resync_pos = -1;
3727 matroska_clear_queue(matroska);
3728 st->skip_to_keyframe =
3729 matroska->skip_to_keyframe = 0;
3734 static int matroska_read_close(AVFormatContext *s)
3736 MatroskaDemuxContext *matroska = s->priv_data;
3737 MatroskaTrack *tracks = matroska->tracks.elem;
3740 matroska_clear_queue(matroska);
3742 for (n = 0; n < matroska->tracks.nb_elem; n++)
3743 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3744 av_freep(&tracks[n].audio.buf);
3745 ebml_free(matroska_segment, matroska);
3751 int64_t start_time_ns;
3752 int64_t end_time_ns;
3753 int64_t start_offset;
3757 /* This function searches all the Cues and returns the CueDesc corresponding to
3758 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3759 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3761 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3762 MatroskaDemuxContext *matroska = s->priv_data;
3765 int nb_index_entries = s->streams[0]->nb_index_entries;
3766 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3767 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3768 for (i = 1; i < nb_index_entries; i++) {
3769 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3770 index_entries[i].timestamp * matroska->time_scale > ts) {
3775 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3776 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3777 if (i != nb_index_entries - 1) {
3778 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3779 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3781 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3782 // FIXME: this needs special handling for files where Cues appear
3783 // before Clusters. the current logic assumes Cues appear after
3785 cue_desc.end_offset = cues_start - matroska->segment_start;
3790 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3792 MatroskaDemuxContext *matroska = s->priv_data;
3793 uint32_t id = matroska->current_id;
3794 int64_t cluster_pos, before_pos;
3796 if (s->streams[0]->nb_index_entries <= 0) return 0;
3797 // seek to the first cluster using cues.
3798 index = av_index_search_timestamp(s->streams[0], 0, 0);
3799 if (index < 0) return 0;
3800 cluster_pos = s->streams[0]->index_entries[index].pos;
3801 before_pos = avio_tell(s->pb);
3803 uint64_t cluster_id, cluster_length;
3806 avio_seek(s->pb, cluster_pos, SEEK_SET);
3807 // read cluster id and length
3808 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3809 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3811 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3815 matroska_reset_status(matroska, 0, cluster_pos);
3816 matroska_clear_queue(matroska);
3817 if (matroska_parse_cluster(matroska) < 0 ||
3821 pkt = &matroska->queue->pkt;
3822 // 4 + read is the length of the cluster id and the cluster length field.
3823 cluster_pos += 4 + read + cluster_length;
3824 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3830 /* Restore the status after matroska_read_header: */
3831 matroska_reset_status(matroska, id, before_pos);
3836 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3837 double min_buffer, double* buffer,
3838 double* sec_to_download, AVFormatContext *s,
3841 double nano_seconds_per_second = 1000000000.0;
3842 double time_sec = time_ns / nano_seconds_per_second;
3844 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3845 int64_t end_time_ns = time_ns + time_to_search_ns;
3846 double sec_downloaded = 0.0;
3847 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3848 if (desc_curr.start_time_ns == -1)
3850 *sec_to_download = 0.0;
3852 // Check for non cue start time.
3853 if (time_ns > desc_curr.start_time_ns) {
3854 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3855 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3856 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3857 double timeToDownload = (cueBytes * 8.0) / bps;
3859 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3860 *sec_to_download += timeToDownload;
3862 // Check if the search ends within the first cue.
3863 if (desc_curr.end_time_ns >= end_time_ns) {
3864 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3865 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3866 sec_downloaded = percent_to_sub * sec_downloaded;
3867 *sec_to_download = percent_to_sub * *sec_to_download;
3870 if ((sec_downloaded + *buffer) <= min_buffer) {
3874 // Get the next Cue.
3875 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3878 while (desc_curr.start_time_ns != -1) {
3879 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3880 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3881 double desc_sec = desc_ns / nano_seconds_per_second;
3882 double bits = (desc_bytes * 8.0);
3883 double time_to_download = bits / bps;
3885 sec_downloaded += desc_sec - time_to_download;
3886 *sec_to_download += time_to_download;
3888 if (desc_curr.end_time_ns >= end_time_ns) {
3889 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3890 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3891 sec_downloaded = percent_to_sub * sec_downloaded;
3892 *sec_to_download = percent_to_sub * *sec_to_download;
3894 if ((sec_downloaded + *buffer) <= min_buffer)
3899 if ((sec_downloaded + *buffer) <= min_buffer) {
3904 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3906 *buffer = *buffer + sec_downloaded;
3910 /* This function computes the bandwidth of the WebM file with the help of
3911 * buffer_size_after_time_downloaded() function. Both of these functions are
3912 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3913 * Matroska parsing mechanism.
3915 * Returns the bandwidth of the file on success; -1 on error.
3917 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3919 MatroskaDemuxContext *matroska = s->priv_data;
3920 AVStream *st = s->streams[0];
3921 double bandwidth = 0.0;
3924 for (i = 0; i < st->nb_index_entries; i++) {
3925 int64_t prebuffer_ns = 1000000000;
3926 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3927 double nano_seconds_per_second = 1000000000.0;
3928 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3929 double prebuffer_bytes = 0.0;
3930 int64_t temp_prebuffer_ns = prebuffer_ns;
3931 int64_t pre_bytes, pre_ns;
3932 double pre_sec, prebuffer, bits_per_second;
3933 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3935 // Start with the first Cue.
3936 CueDesc desc_end = desc_beg;
3938 // Figure out how much data we have downloaded for the prebuffer. This will
3939 // be used later to adjust the bits per sample to try.
3940 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3941 // Prebuffered the entire Cue.
3942 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3943 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3944 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3946 if (desc_end.start_time_ns == -1) {
3947 // The prebuffer is larger than the duration.
3948 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3950 bits_per_second = 0.0;
3952 // The prebuffer ends in the last Cue. Estimate how much data was
3954 pre_bytes = desc_end.end_offset - desc_end.start_offset;
3955 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3956 pre_sec = pre_ns / nano_seconds_per_second;
3958 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3960 prebuffer = prebuffer_ns / nano_seconds_per_second;
3962 // Set this to 0.0 in case our prebuffer buffers the entire video.
3963 bits_per_second = 0.0;
3965 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3966 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3967 double desc_sec = desc_ns / nano_seconds_per_second;
3968 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3970 // Drop the bps by the percentage of bytes buffered.
3971 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3972 double mod_bits_per_second = calc_bits_per_second * percent;
3974 if (prebuffer < desc_sec) {
3976 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3978 // Add 1 so the bits per second should be a little bit greater than file
3980 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3981 const double min_buffer = 0.0;
3982 double buffer = prebuffer;
3983 double sec_to_download = 0.0;
3985 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3986 min_buffer, &buffer, &sec_to_download,
3990 } else if (rv == 0) {
3991 bits_per_second = (double)(bps);
3996 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3997 } while (desc_end.start_time_ns != -1);
3999 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4001 return (int64_t)bandwidth;
4004 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4006 MatroskaDemuxContext *matroska = s->priv_data;
4007 EbmlList *seekhead_list = &matroska->seekhead;
4008 MatroskaSeekhead *seekhead = seekhead_list->elem;
4010 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4014 // determine cues start and end positions
4015 for (i = 0; i < seekhead_list->nb_elem; i++)
4016 if (seekhead[i].id == MATROSKA_ID_CUES)
4019 if (i >= seekhead_list->nb_elem) return -1;
4021 before_pos = avio_tell(matroska->ctx->pb);
4022 cues_start = seekhead[i].pos + matroska->segment_start;
4023 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4024 // cues_end is computed as cues_start + cues_length + length of the
4025 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4026 // cues_end is inclusive and the above sum is reduced by 1.
4027 uint64_t cues_length, cues_id;
4029 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4030 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4031 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4032 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4035 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4037 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4038 if (cues_start == -1 || cues_end == -1) return -1;
4041 matroska_parse_cues(matroska);
4044 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4047 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4049 // if the file has cues at the start, fix up the init range so tht
4050 // it does not include it
4051 if (cues_start <= init_range)
4052 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4055 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4056 if (bandwidth < 0) return -1;
4057 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4059 // check if all clusters start with key frames
4060 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4062 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4063 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4064 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4065 if (!buf) return -1;
4067 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4068 int ret = snprintf(buf + end, 20,
4069 "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4070 i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4071 if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4072 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4074 return AVERROR_INVALIDDATA;
4078 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
4084 static int webm_dash_manifest_read_header(AVFormatContext *s)
4087 int ret = matroska_read_header(s);
4089 MatroskaTrack *tracks;
4090 MatroskaDemuxContext *matroska = s->priv_data;
4092 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4095 if (!s->nb_streams) {
4096 matroska_read_close(s);
4097 av_log(s, AV_LOG_ERROR, "No streams found\n");
4098 return AVERROR_INVALIDDATA;
4101 if (!matroska->is_live) {
4102 buf = av_asprintf("%g", matroska->duration);
4103 if (!buf) return AVERROR(ENOMEM);
4104 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
4107 // initialization range
4108 // 5 is the offset of Cluster ID.
4109 init_range = avio_tell(s->pb) - 5;
4110 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4113 // basename of the file
4114 buf = strrchr(s->url, '/');
4115 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4118 tracks = matroska->tracks.elem;
4119 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4121 // parse the cues and populate Cue related fields
4122 if (!matroska->is_live) {
4123 ret = webm_dash_manifest_cues(s, init_range);
4125 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4130 // use the bandwidth from the command line if it was provided
4131 if (matroska->bandwidth > 0) {
4132 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4133 matroska->bandwidth, 0);
4138 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4143 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4144 static const AVOption options[] = {
4145 { "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 },
4146 { "bandwidth", "bandwidth of this stream to be specified in the DASH manifest.", OFFSET(bandwidth), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, AV_OPT_FLAG_DECODING_PARAM },
4150 static const AVClass webm_dash_class = {
4151 .class_name = "WebM DASH Manifest demuxer",
4152 .item_name = av_default_item_name,
4154 .version = LIBAVUTIL_VERSION_INT,
4157 AVInputFormat ff_matroska_demuxer = {
4158 .name = "matroska,webm",
4159 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4160 .extensions = "mkv,mk3d,mka,mks",
4161 .priv_data_size = sizeof(MatroskaDemuxContext),
4162 .read_probe = matroska_probe,
4163 .read_header = matroska_read_header,
4164 .read_packet = matroska_read_packet,
4165 .read_close = matroska_read_close,
4166 .read_seek = matroska_read_seek,
4167 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4170 AVInputFormat ff_webm_dash_manifest_demuxer = {
4171 .name = "webm_dash_manifest",
4172 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4173 .priv_data_size = sizeof(MatroskaDemuxContext),
4174 .read_header = webm_dash_manifest_read_header,
4175 .read_packet = webm_dash_manifest_read_packet,
4176 .read_close = matroska_read_close,
4177 .priv_class = &webm_dash_class,