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 static const EbmlSyntax matroska_cluster_parsing[] = {
720 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
721 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, { .n = matroska_blockgroup } },
722 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
723 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
724 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
725 { MATROSKA_ID_INFO, EBML_NONE },
726 { MATROSKA_ID_CUES, EBML_NONE },
727 { MATROSKA_ID_TAGS, EBML_NONE },
728 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
729 { MATROSKA_ID_CLUSTER, EBML_STOP },
730 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
733 static const EbmlSyntax matroska_cluster[] = {
734 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
735 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
736 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
737 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
738 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
739 CHILD_OF(matroska_segment)
742 static const EbmlSyntax matroska_clusters[] = {
743 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = matroska_cluster } },
744 { MATROSKA_ID_INFO, EBML_NONE },
745 { MATROSKA_ID_CUES, EBML_NONE },
746 { MATROSKA_ID_TAGS, EBML_NONE },
747 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
748 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
752 static const char *const matroska_doctypes[] = { "matroska", "webm" };
754 static int matroska_read_close(AVFormatContext *s);
757 * This function prepares the status for parsing of level 1 elements.
759 static int matroska_reset_status(MatroskaDemuxContext *matroska,
760 uint32_t id, int64_t position)
763 int err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
768 matroska->current_id = id;
769 matroska->num_levels = 1;
770 matroska->current_cluster.pos = 0;
771 matroska->resync_pos = avio_tell(matroska->ctx->pb);
773 matroska->resync_pos -= (av_log2(id) + 7) / 8;
778 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
780 AVIOContext *pb = matroska->ctx->pb;
783 /* Try to seek to the last position to resync from. If this doesn't work,
784 * we resync from the earliest position available: The start of the buffer. */
785 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
786 av_log(matroska->ctx, AV_LOG_WARNING,
787 "Seek to desired resync point failed. Seeking to "
788 "earliest point available instead.\n");
789 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
790 last_pos + 1), SEEK_SET);
795 // try to find a toplevel element
796 while (!avio_feof(pb)) {
797 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
798 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
799 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
800 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
801 /* Prepare the context for parsing of a level 1 element. */
802 matroska_reset_status(matroska, id, -1);
803 /* Given that we are here means that an error has occured,
804 * so treat the segment as unknown length in order not to
805 * discard valid data that happens to be beyond the designated
806 * end of the segment. */
807 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
810 id = (id << 8) | avio_r8(pb);
814 return pb->error ? pb->error : AVERROR_EOF;
818 * Return: Whether we reached the end of a level in the hierarchy or not.
820 static int ebml_level_end(MatroskaDemuxContext *matroska)
822 AVIOContext *pb = matroska->ctx->pb;
823 int64_t pos = avio_tell(pb);
825 if (matroska->num_levels > 0) {
826 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
827 if (pos - level->start >= level->length || matroska->current_id) {
828 matroska->num_levels--;
832 return (matroska->is_live && matroska->ctx->pb->eof_reached) ? 1 : 0;
836 * Read: an "EBML number", which is defined as a variable-length
837 * array of bytes. The first byte indicates the length by giving a
838 * number of 0-bits followed by a one. The position of the first
839 * "one" bit inside the first byte indicates the length of this
841 * Returns: number of bytes read, < 0 on error
843 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
844 int max_size, uint64_t *number, int eof_forbidden)
850 /* The first byte tells us the length in bytes - except when it is zero. */
855 /* get the length of the EBML number */
856 read = 8 - ff_log2_tab[total];
858 if (!total || read > max_size) {
859 pos = avio_tell(pb) - 1;
861 av_log(matroska->ctx, AV_LOG_ERROR,
862 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
863 "of an EBML number\n", pos, pos);
865 av_log(matroska->ctx, AV_LOG_ERROR,
866 "Length %d indicated by an EBML number's first byte 0x%02x "
867 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
868 read, (uint8_t) total, pos, pos, max_size);
870 return AVERROR_INVALIDDATA;
873 /* read out length */
874 total ^= 1 << ff_log2_tab[total];
876 total = (total << 8) | avio_r8(pb);
878 if (pb->eof_reached) {
890 av_log(matroska->ctx, AV_LOG_ERROR,
891 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
896 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
897 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
904 * Read a EBML length value.
905 * This needs special handling for the "unknown length" case which has multiple
908 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
911 int res = ebml_read_num(matroska, pb, 8, number, 1);
912 if (res > 0 && *number + 1 == 1ULL << (7 * res))
913 *number = EBML_UNKNOWN_LENGTH;
918 * Read the next element as an unsigned int.
919 * Returns NEEDS_CHECKING.
921 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
925 /* big-endian ordering; build up number */
928 *num = (*num << 8) | avio_r8(pb);
930 return NEEDS_CHECKING;
934 * Read the next element as a signed int.
935 * Returns NEEDS_CHECKING.
937 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
944 *num = sign_extend(avio_r8(pb), 8);
946 /* big-endian ordering; build up number */
948 *num = ((uint64_t)*num << 8) | avio_r8(pb);
951 return NEEDS_CHECKING;
955 * Read the next element as a float.
956 * Returns NEEDS_CHECKING or < 0 on obvious failure.
958 static int ebml_read_float(AVIOContext *pb, int size, double *num)
963 *num = av_int2float(avio_rb32(pb));
965 *num = av_int2double(avio_rb64(pb));
967 return AVERROR_INVALIDDATA;
969 return NEEDS_CHECKING;
973 * Read the next element as an ASCII string.
974 * 0 is success, < 0 or NEEDS_CHECKING is failure.
976 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
981 /* EBML strings are usually not 0-terminated, so we allocate one
982 * byte more, read the string and NULL-terminate it ourselves. */
983 if (!(res = av_malloc(size + 1)))
984 return AVERROR(ENOMEM);
985 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
987 return ret < 0 ? ret : NEEDS_CHECKING;
997 * Read the next element as binary data.
998 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1000 static int ebml_read_binary(AVIOContext *pb, int length, EbmlBin *bin)
1004 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1007 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1009 bin->data = bin->buf->data;
1011 bin->pos = avio_tell(pb);
1012 if ((ret = avio_read(pb, bin->data, length)) != length) {
1013 av_buffer_unref(&bin->buf);
1016 return ret < 0 ? ret : NEEDS_CHECKING;
1023 * Read the next element, but only the header. The contents
1024 * are supposed to be sub-elements which can be read separately.
1025 * 0 is success, < 0 is failure.
1027 static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length)
1029 AVIOContext *pb = matroska->ctx->pb;
1030 MatroskaLevel *level;
1032 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1033 av_log(matroska->ctx, AV_LOG_ERROR,
1034 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1035 return AVERROR(ENOSYS);
1038 level = &matroska->levels[matroska->num_levels++];
1039 level->start = avio_tell(pb);
1040 level->length = length;
1046 * Read signed/unsigned "EBML" numbers.
1047 * Return: number of bytes processed, < 0 on error
1049 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
1050 uint8_t *data, uint32_t size, uint64_t *num)
1053 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
1054 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num, 1);
1058 * Same as above, but signed.
1060 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1061 uint8_t *data, uint32_t size, int64_t *num)
1066 /* read as unsigned number first */
1067 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
1070 /* make signed (weird way) */
1071 *num = unum - ((1LL << (7 * res - 1)) - 1);
1076 static int ebml_parse(MatroskaDemuxContext *matroska,
1077 EbmlSyntax *syntax, void *data);
1079 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1082 for (i = 0; syntax[i].id; i++)
1083 if (id == syntax[i].id)
1089 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1094 for (i = 0; syntax[i].id; i++)
1095 switch (syntax[i].type) {
1097 *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1100 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1103 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1107 // the default may be NULL
1108 if (syntax[i].def.s) {
1109 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1110 *dst = av_strdup(syntax[i].def.s);
1112 return AVERROR(ENOMEM);
1117 if (!matroska->levels[matroska->num_levels - 1].length) {
1118 matroska->num_levels--;
1123 res = ebml_parse(matroska, syntax, data);
1126 return res == LEVEL_ENDED ? 0 : res;
1129 static int is_ebml_id_valid(uint32_t id)
1131 // Due to endian nonsense in Matroska, the highest byte with any bits set
1132 // will contain the leading length bit. This bit in turn identifies the
1133 // total byte length of the element by its position within the byte.
1134 unsigned int bits = av_log2(id);
1135 return id && (bits + 7) / 8 == (8 - bits % 8);
1139 * Allocate and return the entry for the level1 element with the given ID. If
1140 * an entry already exists, return the existing entry.
1142 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1146 MatroskaLevel1Element *elem;
1148 if (!is_ebml_id_valid(id))
1151 // Some files link to all clusters; useless.
1152 if (id == MATROSKA_ID_CLUSTER)
1155 // There can be multiple seekheads.
1156 if (id != MATROSKA_ID_SEEKHEAD) {
1157 for (i = 0; i < matroska->num_level1_elems; i++) {
1158 if (matroska->level1_elems[i].id == id)
1159 return &matroska->level1_elems[i];
1163 // Only a completely broken file would have more elements.
1164 // It also provides a low-effort way to escape from circular seekheads
1165 // (every iteration will add a level1 entry).
1166 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1167 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1171 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1172 *elem = (MatroskaLevel1Element){.id = id};
1177 static int ebml_parse(MatroskaDemuxContext *matroska,
1178 EbmlSyntax *syntax, void *data)
1180 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1184 // max. 16 MB for strings
1185 [EBML_STR] = 0x1000000,
1186 [EBML_UTF8] = 0x1000000,
1187 // max. 256 MB for binary data
1188 [EBML_BIN] = 0x10000000,
1189 // no limits for anything else
1191 AVIOContext *pb = matroska->ctx->pb;
1194 int64_t pos = avio_tell(pb);
1195 int res, update_pos = 1, level_check;
1197 MatroskaLevel1Element *level1_elem;
1198 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1200 if (!matroska->current_id) {
1202 res = ebml_read_num(matroska, pb, 4, &id, 0);
1204 if (pb->eof_reached && res == AVERROR_EOF) {
1205 if (matroska->is_live)
1206 // in live mode, finish parsing if EOF is reached.
1208 if (level && level->length == EBML_UNKNOWN_LENGTH && pos == avio_tell(pb)) {
1209 // Unknown-length levels automatically end at EOF.
1210 matroska->num_levels--;
1216 matroska->current_id = id | 1 << 7 * res;
1218 pos -= (av_log2(matroska->current_id) + 7) / 8;
1220 id = matroska->current_id;
1222 syntax = ebml_parse_id(syntax, id);
1223 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1224 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1225 // Unknown-length levels end when an element from an upper level
1226 // in the hierarchy is encountered.
1227 while (syntax->def.n) {
1228 syntax = ebml_parse_id(syntax->def.n, id);
1230 matroska->num_levels--;
1236 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1237 "%"PRId64"\n", id, pos);
1238 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1241 data = (char *) data + syntax->data_offset;
1242 if (syntax->list_elem_size) {
1243 EbmlList *list = data;
1244 newelem = av_realloc_array(list->elem, list->nb_elem + 1, syntax->list_elem_size);
1246 return AVERROR(ENOMEM);
1247 list->elem = newelem;
1248 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1249 memset(data, 0, syntax->list_elem_size);
1253 if (syntax->type != EBML_STOP) {
1254 matroska->current_id = 0;
1255 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1257 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1258 av_log(matroska->ctx, AV_LOG_ERROR,
1259 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for syntax element %i\n",
1260 length, max_lengths[syntax->type], syntax->type);
1261 return AVERROR_INVALIDDATA;
1263 if (matroska->num_levels > 0) {
1264 MatroskaLevel *level = &matroska->levels[matroska->num_levels - 1];
1265 AVIOContext *pb = matroska->ctx->pb;
1266 int64_t pos = avio_tell(pb);
1268 if (length != EBML_UNKNOWN_LENGTH &&
1269 level->length != EBML_UNKNOWN_LENGTH) {
1270 uint64_t elem_end = pos + length,
1271 level_end = level->start + level->length;
1273 if (elem_end < level_end) {
1275 } else if (elem_end == level_end) {
1276 level_check = LEVEL_ENDED;
1278 av_log(matroska->ctx, AV_LOG_ERROR,
1279 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1280 "containing master element ending at 0x%"PRIx64"\n",
1281 pos, elem_end, level_end);
1282 return AVERROR_INVALIDDATA;
1284 } else if (length != EBML_UNKNOWN_LENGTH) {
1286 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1287 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1288 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1289 return AVERROR_INVALIDDATA;
1290 } else if (id != MATROSKA_ID_CLUSTER) {
1291 // According to the specifications only clusters and segments
1292 // are allowed to be unknown-sized.
1293 av_log(matroska->ctx, AV_LOG_ERROR,
1294 "Found unknown-sized element other than a cluster at "
1295 "0x%"PRIx64". Dropping the invalid element.\n", pos);
1296 return AVERROR_INVALIDDATA;
1303 // We have found an element that is allowed at this place
1304 // in the hierarchy and it passed all checks, so treat the beginning
1305 // of the element as the "last known good" position.
1306 matroska->resync_pos = pos;
1310 switch (syntax->type) {
1312 res = ebml_read_uint(pb, length, data);
1315 res = ebml_read_sint(pb, length, data);
1318 res = ebml_read_float(pb, length, data);
1322 res = ebml_read_ascii(pb, length, data);
1325 res = ebml_read_binary(pb, length, data);
1329 if ((res = ebml_read_master(matroska, length)) < 0)
1331 if (id == MATROSKA_ID_SEGMENT)
1332 matroska->segment_start = avio_tell(matroska->ctx->pb);
1333 if (id == MATROSKA_ID_CUES)
1334 matroska->cues_parsing_deferred = 0;
1335 if (syntax->type == EBML_LEVEL1 &&
1336 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1337 if (level1_elem->parsed)
1338 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1339 level1_elem->parsed = 1;
1341 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1349 if (ffio_limit(pb, length) != length) {
1350 // ffio_limit emits its own error message,
1351 // so we don't have to.
1352 return AVERROR(EIO);
1354 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1355 // avio_skip might take us past EOF. We check for this
1356 // by skipping only length - 1 bytes, reading a byte and
1357 // checking the error flags. This is done in order to check
1358 // that the element has been properly skipped even when
1359 // no filesize (that ffio_limit relies on) is available.
1361 res = NEEDS_CHECKING;
1368 if (res == NEEDS_CHECKING) {
1369 if (pb->eof_reached) {
1378 if (res == AVERROR_INVALIDDATA)
1379 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1380 else if (res == AVERROR(EIO))
1381 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1382 else if (res == AVERROR_EOF) {
1383 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1391 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1392 level = &matroska->levels[matroska->num_levels - 1];
1393 pos = avio_tell(pb);
1395 // Given that pos >= level->start no check for
1396 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1397 while (matroska->num_levels && pos == level->start + level->length) {
1398 matroska->num_levels--;
1406 static void ebml_free(EbmlSyntax *syntax, void *data)
1409 for (i = 0; syntax[i].id; i++) {
1410 void *data_off = (char *) data + syntax[i].data_offset;
1411 switch (syntax[i].type) {
1417 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1421 if (syntax[i].list_elem_size) {
1422 EbmlList *list = data_off;
1423 char *ptr = list->elem;
1424 for (j = 0; j < list->nb_elem;
1425 j++, ptr += syntax[i].list_elem_size)
1426 ebml_free(syntax[i].def.n, ptr);
1427 av_freep(&list->elem);
1430 ebml_free(syntax[i].def.n, data_off);
1440 static int matroska_probe(const AVProbeData *p)
1443 int len_mask = 0x80, size = 1, n = 1, i;
1446 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1449 /* length of header */
1451 while (size <= 8 && !(total & len_mask)) {
1457 total &= (len_mask - 1);
1459 total = (total << 8) | p->buf[4 + n++];
1461 /* Does the probe data contain the whole header? */
1462 if (p->buf_size < 4 + size + total)
1465 /* The header should contain a known document type. For now,
1466 * we don't parse the whole header but simply check for the
1467 * availability of that array of characters inside the header.
1468 * Not fully fool-proof, but good enough. */
1469 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1470 size_t probelen = strlen(matroska_doctypes[i]);
1471 if (total < probelen)
1473 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1474 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1475 return AVPROBE_SCORE_MAX;
1478 // probably valid EBML header but no recognized doctype
1479 return AVPROBE_SCORE_EXTENSION;
1482 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1485 MatroskaTrack *tracks = matroska->tracks.elem;
1488 for (i = 0; i < matroska->tracks.nb_elem; i++)
1489 if (tracks[i].num == num)
1492 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1496 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1497 MatroskaTrack *track)
1499 MatroskaTrackEncoding *encodings = track->encodings.elem;
1500 uint8_t *data = *buf;
1501 int isize = *buf_size;
1502 uint8_t *pkt_data = NULL;
1503 uint8_t av_unused *newpktdata;
1504 int pkt_size = isize;
1508 if (pkt_size >= 10000000U)
1509 return AVERROR_INVALIDDATA;
1511 switch (encodings[0].compression.algo) {
1512 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1514 int header_size = encodings[0].compression.settings.size;
1515 uint8_t *header = encodings[0].compression.settings.data;
1517 if (header_size && !header) {
1518 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1525 pkt_size = isize + header_size;
1526 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1528 return AVERROR(ENOMEM);
1530 memcpy(pkt_data, header, header_size);
1531 memcpy(pkt_data + header_size, data, isize);
1535 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1537 olen = pkt_size *= 3;
1538 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1539 + AV_INPUT_BUFFER_PADDING_SIZE);
1541 result = AVERROR(ENOMEM);
1544 pkt_data = newpktdata;
1545 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1546 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1548 result = AVERROR_INVALIDDATA;
1555 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1557 z_stream zstream = { 0 };
1558 if (inflateInit(&zstream) != Z_OK)
1560 zstream.next_in = data;
1561 zstream.avail_in = isize;
1564 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1566 inflateEnd(&zstream);
1567 result = AVERROR(ENOMEM);
1570 pkt_data = newpktdata;
1571 zstream.avail_out = pkt_size - zstream.total_out;
1572 zstream.next_out = pkt_data + zstream.total_out;
1573 result = inflate(&zstream, Z_NO_FLUSH);
1574 } while (result == Z_OK && pkt_size < 10000000);
1575 pkt_size = zstream.total_out;
1576 inflateEnd(&zstream);
1577 if (result != Z_STREAM_END) {
1578 if (result == Z_MEM_ERROR)
1579 result = AVERROR(ENOMEM);
1581 result = AVERROR_INVALIDDATA;
1588 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1590 bz_stream bzstream = { 0 };
1591 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1593 bzstream.next_in = data;
1594 bzstream.avail_in = isize;
1597 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1599 BZ2_bzDecompressEnd(&bzstream);
1600 result = AVERROR(ENOMEM);
1603 pkt_data = newpktdata;
1604 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1605 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1606 result = BZ2_bzDecompress(&bzstream);
1607 } while (result == BZ_OK && pkt_size < 10000000);
1608 pkt_size = bzstream.total_out_lo32;
1609 BZ2_bzDecompressEnd(&bzstream);
1610 if (result != BZ_STREAM_END) {
1611 if (result == BZ_MEM_ERROR)
1612 result = AVERROR(ENOMEM);
1614 result = AVERROR_INVALIDDATA;
1621 return AVERROR_INVALIDDATA;
1624 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1627 *buf_size = pkt_size;
1635 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1636 AVDictionary **metadata, char *prefix)
1638 MatroskaTag *tags = list->elem;
1642 for (i = 0; i < list->nb_elem; i++) {
1643 const char *lang = tags[i].lang &&
1644 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1646 if (!tags[i].name) {
1647 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1651 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1653 av_strlcpy(key, tags[i].name, sizeof(key));
1654 if (tags[i].def || !lang) {
1655 av_dict_set(metadata, key, tags[i].string, 0);
1656 if (tags[i].sub.nb_elem)
1657 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1660 av_strlcat(key, "-", sizeof(key));
1661 av_strlcat(key, lang, sizeof(key));
1662 av_dict_set(metadata, key, tags[i].string, 0);
1663 if (tags[i].sub.nb_elem)
1664 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1667 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1670 static void matroska_convert_tags(AVFormatContext *s)
1672 MatroskaDemuxContext *matroska = s->priv_data;
1673 MatroskaTags *tags = matroska->tags.elem;
1676 for (i = 0; i < matroska->tags.nb_elem; i++) {
1677 if (tags[i].target.attachuid) {
1678 MatroskaAttachment *attachment = matroska->attachments.elem;
1680 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1681 if (attachment[j].uid == tags[i].target.attachuid &&
1682 attachment[j].stream) {
1683 matroska_convert_tag(s, &tags[i].tag,
1684 &attachment[j].stream->metadata, NULL);
1689 av_log(NULL, AV_LOG_WARNING,
1690 "The tags at index %d refer to a "
1691 "non-existent attachment %"PRId64".\n",
1692 i, tags[i].target.attachuid);
1694 } else if (tags[i].target.chapteruid) {
1695 MatroskaChapter *chapter = matroska->chapters.elem;
1697 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1698 if (chapter[j].uid == tags[i].target.chapteruid &&
1699 chapter[j].chapter) {
1700 matroska_convert_tag(s, &tags[i].tag,
1701 &chapter[j].chapter->metadata, NULL);
1706 av_log(NULL, AV_LOG_WARNING,
1707 "The tags at index %d refer to a non-existent chapter "
1709 i, tags[i].target.chapteruid);
1711 } else if (tags[i].target.trackuid) {
1712 MatroskaTrack *track = matroska->tracks.elem;
1714 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1715 if (track[j].uid == tags[i].target.trackuid &&
1717 matroska_convert_tag(s, &tags[i].tag,
1718 &track[j].stream->metadata, NULL);
1723 av_log(NULL, AV_LOG_WARNING,
1724 "The tags at index %d refer to a non-existent track "
1726 i, tags[i].target.trackuid);
1729 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1730 tags[i].target.type);
1735 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1738 uint32_t saved_id = matroska->current_id;
1739 int64_t before_pos = avio_tell(matroska->ctx->pb);
1740 MatroskaLevel level;
1745 offset = pos + matroska->segment_start;
1746 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1747 /* We don't want to lose our seekhead level, so we add
1748 * a dummy. This is a crude hack. */
1749 if (matroska->num_levels == EBML_MAX_DEPTH) {
1750 av_log(matroska->ctx, AV_LOG_INFO,
1751 "Max EBML element depth (%d) reached, "
1752 "cannot parse further.\n", EBML_MAX_DEPTH);
1753 ret = AVERROR_INVALIDDATA;
1756 level.length = EBML_UNKNOWN_LENGTH;
1757 matroska->levels[matroska->num_levels] = level;
1758 matroska->num_levels++;
1759 matroska->current_id = 0;
1761 ret = ebml_parse(matroska, matroska_segment, matroska);
1762 if (ret == LEVEL_ENDED) {
1763 /* This can only happen if the seek brought us beyond EOF. */
1768 /* Seek back - notice that in all instances where this is used it is safe
1769 * to set the level to 1 and unset the position of the current cluster. */
1770 matroska_reset_status(matroska, saved_id, before_pos);
1775 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1777 EbmlList *seekhead_list = &matroska->seekhead;
1780 // we should not do any seeking in the streaming case
1781 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1784 for (i = 0; i < seekhead_list->nb_elem; i++) {
1785 MatroskaSeekhead *seekheads = seekhead_list->elem;
1786 uint32_t id = seekheads[i].id;
1787 uint64_t pos = seekheads[i].pos;
1789 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1790 if (!elem || elem->parsed)
1795 // defer cues parsing until we actually need cue data.
1796 if (id == MATROSKA_ID_CUES)
1799 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1800 // mark index as broken
1801 matroska->cues_parsing_deferred = -1;
1809 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1811 EbmlList *index_list;
1812 MatroskaIndex *index;
1813 uint64_t index_scale = 1;
1816 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1819 index_list = &matroska->index;
1820 index = index_list->elem;
1821 if (index_list->nb_elem < 2)
1823 if (index[1].time > 1E14 / matroska->time_scale) {
1824 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1827 for (i = 0; i < index_list->nb_elem; i++) {
1828 EbmlList *pos_list = &index[i].pos;
1829 MatroskaIndexPos *pos = pos_list->elem;
1830 for (j = 0; j < pos_list->nb_elem; j++) {
1831 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1833 if (track && track->stream)
1834 av_add_index_entry(track->stream,
1835 pos[j].pos + matroska->segment_start,
1836 index[i].time / index_scale, 0, 0,
1842 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1845 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1848 for (i = 0; i < matroska->num_level1_elems; i++) {
1849 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1850 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1851 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1852 matroska->cues_parsing_deferred = -1;
1858 matroska_add_index_entries(matroska);
1861 static int matroska_aac_profile(char *codec_id)
1863 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1866 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1867 if (strstr(codec_id, aac_profiles[profile]))
1872 static int matroska_aac_sri(int samplerate)
1876 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1877 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1882 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1884 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1885 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1888 static int matroska_parse_flac(AVFormatContext *s,
1889 MatroskaTrack *track,
1892 AVStream *st = track->stream;
1893 uint8_t *p = track->codec_priv.data;
1894 int size = track->codec_priv.size;
1896 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1897 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1898 track->codec_priv.size = 0;
1902 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1904 p += track->codec_priv.size;
1905 size -= track->codec_priv.size;
1907 /* parse the remaining metadata blocks if present */
1909 int block_last, block_type, block_size;
1911 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1915 if (block_size > size)
1918 /* check for the channel mask */
1919 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1920 AVDictionary *dict = NULL;
1921 AVDictionaryEntry *chmask;
1923 ff_vorbis_comment(s, &dict, p, block_size, 0);
1924 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1926 uint64_t mask = strtol(chmask->value, NULL, 0);
1927 if (!mask || mask & ~0x3ffffULL) {
1928 av_log(s, AV_LOG_WARNING,
1929 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1931 st->codecpar->channel_layout = mask;
1933 av_dict_free(&dict);
1943 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
1945 int major, minor, micro, bttb = 0;
1947 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
1948 * this function, and fixed in 57.52 */
1949 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, µ) == 3)
1950 bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
1952 switch (field_order) {
1953 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
1954 return AV_FIELD_PROGRESSIVE;
1955 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
1956 return AV_FIELD_UNKNOWN;
1957 case MATROSKA_VIDEO_FIELDORDER_TT:
1959 case MATROSKA_VIDEO_FIELDORDER_BB:
1961 case MATROSKA_VIDEO_FIELDORDER_BT:
1962 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
1963 case MATROSKA_VIDEO_FIELDORDER_TB:
1964 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
1966 return AV_FIELD_UNKNOWN;
1970 static void mkv_stereo_mode_display_mul(int stereo_mode,
1971 int *h_width, int *h_height)
1973 switch (stereo_mode) {
1974 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
1975 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
1976 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
1977 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
1978 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
1980 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
1981 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
1982 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
1983 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
1986 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
1987 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
1988 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
1989 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
1995 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
1996 const MatroskaTrackVideoColor *color = track->video.color.elem;
1997 const MatroskaMasteringMeta *mastering_meta;
1998 int has_mastering_primaries, has_mastering_luminance;
2000 if (!track->video.color.nb_elem)
2003 mastering_meta = &color->mastering_meta;
2004 // Mastering primaries are CIE 1931 coords, and must be > 0.
2005 has_mastering_primaries =
2006 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2007 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2008 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2009 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2010 has_mastering_luminance = mastering_meta->max_luminance > 0;
2012 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2013 st->codecpar->color_space = color->matrix_coefficients;
2014 if (color->primaries != AVCOL_PRI_RESERVED &&
2015 color->primaries != AVCOL_PRI_RESERVED0)
2016 st->codecpar->color_primaries = color->primaries;
2017 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2018 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2019 st->codecpar->color_trc = color->transfer_characteristics;
2020 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2021 color->range <= AVCOL_RANGE_JPEG)
2022 st->codecpar->color_range = color->range;
2023 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2024 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2025 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2026 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2027 st->codecpar->chroma_location =
2028 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2029 (color->chroma_siting_vert - 1) << 7);
2031 if (color->max_cll && color->max_fall) {
2034 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2036 return AVERROR(ENOMEM);
2037 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2038 (uint8_t *)metadata, size);
2040 av_freep(&metadata);
2043 metadata->MaxCLL = color->max_cll;
2044 metadata->MaxFALL = color->max_fall;
2047 if (has_mastering_primaries || has_mastering_luminance) {
2048 // Use similar rationals as other standards.
2049 const int chroma_den = 50000;
2050 const int luma_den = 10000;
2051 AVMasteringDisplayMetadata *metadata =
2052 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2053 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2054 sizeof(AVMasteringDisplayMetadata));
2056 return AVERROR(ENOMEM);
2058 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2059 if (has_mastering_primaries) {
2060 metadata->display_primaries[0][0] = av_make_q(
2061 round(mastering_meta->r_x * chroma_den), chroma_den);
2062 metadata->display_primaries[0][1] = av_make_q(
2063 round(mastering_meta->r_y * chroma_den), chroma_den);
2064 metadata->display_primaries[1][0] = av_make_q(
2065 round(mastering_meta->g_x * chroma_den), chroma_den);
2066 metadata->display_primaries[1][1] = av_make_q(
2067 round(mastering_meta->g_y * chroma_den), chroma_den);
2068 metadata->display_primaries[2][0] = av_make_q(
2069 round(mastering_meta->b_x * chroma_den), chroma_den);
2070 metadata->display_primaries[2][1] = av_make_q(
2071 round(mastering_meta->b_y * chroma_den), chroma_den);
2072 metadata->white_point[0] = av_make_q(
2073 round(mastering_meta->white_x * chroma_den), chroma_den);
2074 metadata->white_point[1] = av_make_q(
2075 round(mastering_meta->white_y * chroma_den), chroma_den);
2076 metadata->has_primaries = 1;
2078 if (has_mastering_luminance) {
2079 metadata->max_luminance = av_make_q(
2080 round(mastering_meta->max_luminance * luma_den), luma_den);
2081 metadata->min_luminance = av_make_q(
2082 round(mastering_meta->min_luminance * luma_den), luma_den);
2083 metadata->has_luminance = 1;
2089 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track) {
2090 AVSphericalMapping *spherical;
2091 enum AVSphericalProjection projection;
2092 size_t spherical_size;
2093 uint32_t l = 0, t = 0, r = 0, b = 0;
2094 uint32_t padding = 0;
2098 bytestream2_init(&gb, track->video.projection.private.data,
2099 track->video.projection.private.size);
2101 if (bytestream2_get_byte(&gb) != 0) {
2102 av_log(NULL, AV_LOG_WARNING, "Unknown spherical metadata\n");
2106 bytestream2_skip(&gb, 3); // flags
2108 switch (track->video.projection.type) {
2109 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2110 if (track->video.projection.private.size == 20) {
2111 t = bytestream2_get_be32(&gb);
2112 b = bytestream2_get_be32(&gb);
2113 l = bytestream2_get_be32(&gb);
2114 r = bytestream2_get_be32(&gb);
2116 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2117 av_log(NULL, AV_LOG_ERROR,
2118 "Invalid bounding rectangle coordinates "
2119 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2121 return AVERROR_INVALIDDATA;
2123 } else if (track->video.projection.private.size != 0) {
2124 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2125 return AVERROR_INVALIDDATA;
2128 if (l || t || r || b)
2129 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2131 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2133 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2134 if (track->video.projection.private.size < 4) {
2135 av_log(NULL, AV_LOG_ERROR, "Missing projection private properties\n");
2136 return AVERROR_INVALIDDATA;
2137 } else if (track->video.projection.private.size == 12) {
2138 uint32_t layout = bytestream2_get_be32(&gb);
2140 av_log(NULL, AV_LOG_WARNING,
2141 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2144 projection = AV_SPHERICAL_CUBEMAP;
2145 padding = bytestream2_get_be32(&gb);
2147 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2148 return AVERROR_INVALIDDATA;
2151 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2152 /* No Spherical metadata */
2155 av_log(NULL, AV_LOG_WARNING,
2156 "Unknown spherical metadata type %"PRIu64"\n",
2157 track->video.projection.type);
2161 spherical = av_spherical_alloc(&spherical_size);
2163 return AVERROR(ENOMEM);
2165 spherical->projection = projection;
2167 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2168 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2169 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2171 spherical->padding = padding;
2173 spherical->bound_left = l;
2174 spherical->bound_top = t;
2175 spherical->bound_right = r;
2176 spherical->bound_bottom = b;
2178 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2181 av_freep(&spherical);
2188 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2190 const AVCodecTag *codec_tags;
2192 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2193 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2195 /* Normalize noncompliant private data that starts with the fourcc
2196 * by expanding/shifting the data by 4 bytes and storing the data
2197 * size at the start. */
2198 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2199 int ret = av_buffer_realloc(&track->codec_priv.buf,
2200 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2204 track->codec_priv.data = track->codec_priv.buf->data;
2205 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2206 track->codec_priv.size += 4;
2207 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2210 *fourcc = AV_RL32(track->codec_priv.data + 4);
2211 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2216 static int matroska_parse_tracks(AVFormatContext *s)
2218 MatroskaDemuxContext *matroska = s->priv_data;
2219 MatroskaTrack *tracks = matroska->tracks.elem;
2224 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2225 MatroskaTrack *track = &tracks[i];
2226 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2227 EbmlList *encodings_list = &track->encodings;
2228 MatroskaTrackEncoding *encodings = encodings_list->elem;
2229 uint8_t *extradata = NULL;
2230 int extradata_size = 0;
2231 int extradata_offset = 0;
2232 uint32_t fourcc = 0;
2234 char* key_id_base64 = NULL;
2237 /* Apply some sanity checks. */
2238 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2239 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2240 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2241 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2242 av_log(matroska->ctx, AV_LOG_INFO,
2243 "Unknown or unsupported track type %"PRIu64"\n",
2247 if (!track->codec_id)
2250 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2251 isnan(track->audio.samplerate)) {
2252 av_log(matroska->ctx, AV_LOG_WARNING,
2253 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2254 track->audio.samplerate);
2255 track->audio.samplerate = 8000;
2258 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2259 if (!track->default_duration && track->video.frame_rate > 0) {
2260 double default_duration = 1000000000 / track->video.frame_rate;
2261 if (default_duration > UINT64_MAX || default_duration < 0) {
2262 av_log(matroska->ctx, AV_LOG_WARNING,
2263 "Invalid frame rate %e. Cannot calculate default duration.\n",
2264 track->video.frame_rate);
2266 track->default_duration = default_duration;
2269 if (track->video.display_width == -1)
2270 track->video.display_width = track->video.pixel_width;
2271 if (track->video.display_height == -1)
2272 track->video.display_height = track->video.pixel_height;
2273 if (track->video.color_space.size == 4)
2274 fourcc = AV_RL32(track->video.color_space.data);
2275 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2276 if (!track->audio.out_samplerate)
2277 track->audio.out_samplerate = track->audio.samplerate;
2279 if (encodings_list->nb_elem > 1) {
2280 av_log(matroska->ctx, AV_LOG_ERROR,
2281 "Multiple combined encodings not supported");
2282 } else if (encodings_list->nb_elem == 1) {
2283 if (encodings[0].type) {
2284 if (encodings[0].encryption.key_id.size > 0) {
2285 /* Save the encryption key id to be stored later as a
2287 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2288 key_id_base64 = av_malloc(b64_size);
2289 if (key_id_base64 == NULL)
2290 return AVERROR(ENOMEM);
2292 av_base64_encode(key_id_base64, b64_size,
2293 encodings[0].encryption.key_id.data,
2294 encodings[0].encryption.key_id.size);
2296 encodings[0].scope = 0;
2297 av_log(matroska->ctx, AV_LOG_ERROR,
2298 "Unsupported encoding type");
2302 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2305 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2308 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2310 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2311 encodings[0].scope = 0;
2312 av_log(matroska->ctx, AV_LOG_ERROR,
2313 "Unsupported encoding type");
2314 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2315 uint8_t *codec_priv = track->codec_priv.data;
2316 int ret = matroska_decode_buffer(&track->codec_priv.data,
2317 &track->codec_priv.size,
2320 track->codec_priv.data = NULL;
2321 track->codec_priv.size = 0;
2322 av_log(matroska->ctx, AV_LOG_ERROR,
2323 "Failed to decode codec private data\n");
2326 if (codec_priv != track->codec_priv.data) {
2327 av_buffer_unref(&track->codec_priv.buf);
2328 if (track->codec_priv.data) {
2329 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2330 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2332 if (!track->codec_priv.buf) {
2333 av_freep(&track->codec_priv.data);
2334 track->codec_priv.size = 0;
2335 return AVERROR(ENOMEM);
2342 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2343 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2344 strlen(ff_mkv_codec_tags[j].str))) {
2345 codec_id = ff_mkv_codec_tags[j].id;
2350 st = track->stream = avformat_new_stream(s, NULL);
2352 av_free(key_id_base64);
2353 return AVERROR(ENOMEM);
2356 if (key_id_base64) {
2357 /* export encryption key id as base64 metadata tag */
2358 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
2359 av_freep(&key_id_base64);
2362 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2363 track->codec_priv.size >= 40 &&
2364 track->codec_priv.data) {
2365 track->ms_compat = 1;
2366 bit_depth = AV_RL16(track->codec_priv.data + 14);
2367 fourcc = AV_RL32(track->codec_priv.data + 16);
2368 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2371 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2373 extradata_offset = 40;
2374 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2375 track->codec_priv.size >= 14 &&
2376 track->codec_priv.data) {
2378 ffio_init_context(&b, track->codec_priv.data,
2379 track->codec_priv.size,
2380 0, NULL, NULL, NULL, NULL);
2381 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2384 codec_id = st->codecpar->codec_id;
2385 fourcc = st->codecpar->codec_tag;
2386 extradata_offset = FFMIN(track->codec_priv.size, 18);
2387 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2388 /* Normally 36, but allow noncompliant private data */
2389 && (track->codec_priv.size >= 32)
2390 && (track->codec_priv.data)) {
2391 uint16_t sample_size;
2392 int ret = get_qt_codec(track, &fourcc, &codec_id);
2395 sample_size = AV_RB16(track->codec_priv.data + 26);
2397 if (sample_size == 8) {
2398 fourcc = MKTAG('r','a','w',' ');
2399 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2400 } else if (sample_size == 16) {
2401 fourcc = MKTAG('t','w','o','s');
2402 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2405 if ((fourcc == MKTAG('t','w','o','s') ||
2406 fourcc == MKTAG('s','o','w','t')) &&
2408 codec_id = AV_CODEC_ID_PCM_S8;
2409 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2410 (track->codec_priv.size >= 21) &&
2411 (track->codec_priv.data)) {
2412 int ret = get_qt_codec(track, &fourcc, &codec_id);
2415 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2416 fourcc = MKTAG('S','V','Q','3');
2417 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2419 if (codec_id == AV_CODEC_ID_NONE)
2420 av_log(matroska->ctx, AV_LOG_ERROR,
2421 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2422 if (track->codec_priv.size >= 86) {
2423 bit_depth = AV_RB16(track->codec_priv.data + 82);
2424 ffio_init_context(&b, track->codec_priv.data,
2425 track->codec_priv.size,
2426 0, NULL, NULL, NULL, NULL);
2427 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2429 track->has_palette = 1;
2432 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2433 switch (track->audio.bitdepth) {
2435 codec_id = AV_CODEC_ID_PCM_U8;
2438 codec_id = AV_CODEC_ID_PCM_S24BE;
2441 codec_id = AV_CODEC_ID_PCM_S32BE;
2444 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2445 switch (track->audio.bitdepth) {
2447 codec_id = AV_CODEC_ID_PCM_U8;
2450 codec_id = AV_CODEC_ID_PCM_S24LE;
2453 codec_id = AV_CODEC_ID_PCM_S32LE;
2456 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2457 track->audio.bitdepth == 64) {
2458 codec_id = AV_CODEC_ID_PCM_F64LE;
2459 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2460 int profile = matroska_aac_profile(track->codec_id);
2461 int sri = matroska_aac_sri(track->audio.samplerate);
2462 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2464 return AVERROR(ENOMEM);
2465 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2466 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2467 if (strstr(track->codec_id, "SBR")) {
2468 sri = matroska_aac_sri(track->audio.out_samplerate);
2469 extradata[2] = 0x56;
2470 extradata[3] = 0xE5;
2471 extradata[4] = 0x80 | (sri << 3);
2475 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2476 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2477 * Create the "atom size", "tag", and "tag version" fields the
2478 * decoder expects manually. */
2479 extradata_size = 12 + track->codec_priv.size;
2480 extradata = av_mallocz(extradata_size +
2481 AV_INPUT_BUFFER_PADDING_SIZE);
2483 return AVERROR(ENOMEM);
2484 AV_WB32(extradata, extradata_size);
2485 memcpy(&extradata[4], "alac", 4);
2486 AV_WB32(&extradata[8], 0);
2487 memcpy(&extradata[12], track->codec_priv.data,
2488 track->codec_priv.size);
2489 } else if (codec_id == AV_CODEC_ID_TTA) {
2490 extradata_size = 30;
2491 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2493 return AVERROR(ENOMEM);
2494 ffio_init_context(&b, extradata, extradata_size, 1,
2495 NULL, NULL, NULL, NULL);
2496 avio_write(&b, "TTA1", 4);
2498 if (track->audio.channels > UINT16_MAX ||
2499 track->audio.bitdepth > UINT16_MAX) {
2500 av_log(matroska->ctx, AV_LOG_WARNING,
2501 "Too large audio channel number %"PRIu64
2502 " or bitdepth %"PRIu64". Skipping track.\n",
2503 track->audio.channels, track->audio.bitdepth);
2504 av_freep(&extradata);
2505 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2506 return AVERROR_INVALIDDATA;
2510 avio_wl16(&b, track->audio.channels);
2511 avio_wl16(&b, track->audio.bitdepth);
2512 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2513 return AVERROR_INVALIDDATA;
2514 avio_wl32(&b, track->audio.out_samplerate);
2515 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
2516 track->audio.out_samplerate,
2517 AV_TIME_BASE * 1000));
2518 } else if (codec_id == AV_CODEC_ID_RV10 ||
2519 codec_id == AV_CODEC_ID_RV20 ||
2520 codec_id == AV_CODEC_ID_RV30 ||
2521 codec_id == AV_CODEC_ID_RV40) {
2522 extradata_offset = 26;
2523 } else if (codec_id == AV_CODEC_ID_RA_144) {
2524 track->audio.out_samplerate = 8000;
2525 track->audio.channels = 1;
2526 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2527 codec_id == AV_CODEC_ID_COOK ||
2528 codec_id == AV_CODEC_ID_ATRAC3 ||
2529 codec_id == AV_CODEC_ID_SIPR)
2530 && track->codec_priv.data) {
2533 ffio_init_context(&b, track->codec_priv.data,
2534 track->codec_priv.size,
2535 0, NULL, NULL, NULL, NULL);
2537 flavor = avio_rb16(&b);
2538 track->audio.coded_framesize = avio_rb32(&b);
2540 track->audio.sub_packet_h = avio_rb16(&b);
2541 track->audio.frame_size = avio_rb16(&b);
2542 track->audio.sub_packet_size = avio_rb16(&b);
2544 track->audio.coded_framesize <= 0 ||
2545 track->audio.sub_packet_h <= 0 ||
2546 track->audio.frame_size <= 0 ||
2547 track->audio.sub_packet_size <= 0 && codec_id != AV_CODEC_ID_SIPR)
2548 return AVERROR_INVALIDDATA;
2549 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2550 track->audio.frame_size);
2551 if (!track->audio.buf)
2552 return AVERROR(ENOMEM);
2553 if (codec_id == AV_CODEC_ID_RA_288) {
2554 st->codecpar->block_align = track->audio.coded_framesize;
2555 track->codec_priv.size = 0;
2557 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2558 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2559 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2560 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2562 st->codecpar->block_align = track->audio.sub_packet_size;
2563 extradata_offset = 78;
2565 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2566 ret = matroska_parse_flac(s, track, &extradata_offset);
2569 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2570 fourcc = AV_RL32(track->codec_priv.data);
2571 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2572 /* we don't need any value stored in CodecPrivate.
2573 make sure that it's not exported as extradata. */
2574 track->codec_priv.size = 0;
2575 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2576 /* For now, propagate only the OBUs, if any. Once libavcodec is
2577 updated to handle isobmff style extradata this can be removed. */
2578 extradata_offset = 4;
2580 track->codec_priv.size -= extradata_offset;
2582 if (codec_id == AV_CODEC_ID_NONE)
2583 av_log(matroska->ctx, AV_LOG_INFO,
2584 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2586 if (track->time_scale < 0.01)
2587 track->time_scale = 1.0;
2588 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2589 1000 * 1000 * 1000); /* 64 bit pts in ns */
2591 /* convert the delay from ns to the track timebase */
2592 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2593 (AVRational){ 1, 1000000000 },
2596 st->codecpar->codec_id = codec_id;
2598 if (strcmp(track->language, "und"))
2599 av_dict_set(&st->metadata, "language", track->language, 0);
2600 av_dict_set(&st->metadata, "title", track->name, 0);
2602 if (track->flag_default)
2603 st->disposition |= AV_DISPOSITION_DEFAULT;
2604 if (track->flag_forced)
2605 st->disposition |= AV_DISPOSITION_FORCED;
2607 if (!st->codecpar->extradata) {
2609 st->codecpar->extradata = extradata;
2610 st->codecpar->extradata_size = extradata_size;
2611 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2612 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2613 return AVERROR(ENOMEM);
2614 memcpy(st->codecpar->extradata,
2615 track->codec_priv.data + extradata_offset,
2616 track->codec_priv.size);
2620 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2621 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2622 int display_width_mul = 1;
2623 int display_height_mul = 1;
2625 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2626 st->codecpar->codec_tag = fourcc;
2628 st->codecpar->bits_per_coded_sample = bit_depth;
2629 st->codecpar->width = track->video.pixel_width;
2630 st->codecpar->height = track->video.pixel_height;
2632 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2633 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2634 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2635 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2637 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2638 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2640 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2641 av_reduce(&st->sample_aspect_ratio.num,
2642 &st->sample_aspect_ratio.den,
2643 st->codecpar->height * track->video.display_width * display_width_mul,
2644 st->codecpar->width * track->video.display_height * display_height_mul,
2647 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2648 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2650 if (track->default_duration) {
2651 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2652 1000000000, track->default_duration, 30000);
2653 #if FF_API_R_FRAME_RATE
2654 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2655 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2656 st->r_frame_rate = st->avg_frame_rate;
2660 /* export stereo mode flag as metadata tag */
2661 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2662 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2664 /* export alpha mode flag as metadata tag */
2665 if (track->video.alpha_mode)
2666 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2668 /* if we have virtual track, mark the real tracks */
2669 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2671 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2673 snprintf(buf, sizeof(buf), "%s_%d",
2674 ff_matroska_video_stereo_plane[planes[j].type], i);
2675 for (k=0; k < matroska->tracks.nb_elem; k++)
2676 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2677 av_dict_set(&tracks[k].stream->metadata,
2678 "stereo_mode", buf, 0);
2682 // add stream level stereo3d side data if it is a supported format
2683 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2684 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2685 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2690 ret = mkv_parse_video_color(st, track);
2693 ret = mkv_parse_video_projection(st, track);
2696 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2697 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2698 st->codecpar->codec_tag = fourcc;
2699 st->codecpar->sample_rate = track->audio.out_samplerate;
2700 st->codecpar->channels = track->audio.channels;
2701 if (!st->codecpar->bits_per_coded_sample)
2702 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2703 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2704 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2705 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2706 st->need_parsing = AVSTREAM_PARSE_FULL;
2707 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2708 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2709 if (track->codec_delay > 0) {
2710 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2711 (AVRational){1, 1000000000},
2712 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2713 48000 : st->codecpar->sample_rate});
2715 if (track->seek_preroll > 0) {
2716 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2717 (AVRational){1, 1000000000},
2718 (AVRational){1, st->codecpar->sample_rate});
2720 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2721 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2723 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2724 st->disposition |= AV_DISPOSITION_CAPTIONS;
2725 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2726 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2727 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2728 st->disposition |= AV_DISPOSITION_METADATA;
2730 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2731 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2738 static int matroska_read_header(AVFormatContext *s)
2740 MatroskaDemuxContext *matroska = s->priv_data;
2741 EbmlList *attachments_list = &matroska->attachments;
2742 EbmlList *chapters_list = &matroska->chapters;
2743 MatroskaAttachment *attachments;
2744 MatroskaChapter *chapters;
2745 uint64_t max_start = 0;
2751 matroska->cues_parsing_deferred = 1;
2753 /* First read the EBML header. */
2754 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2755 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2756 ebml_free(ebml_syntax, &ebml);
2757 return AVERROR_INVALIDDATA;
2759 if (ebml.version > EBML_VERSION ||
2760 ebml.max_size > sizeof(uint64_t) ||
2761 ebml.id_length > sizeof(uint32_t) ||
2762 ebml.doctype_version > 3) {
2763 avpriv_report_missing_feature(matroska->ctx,
2764 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2765 ebml.version, ebml.doctype, ebml.doctype_version);
2766 ebml_free(ebml_syntax, &ebml);
2767 return AVERROR_PATCHWELCOME;
2768 } else if (ebml.doctype_version == 3) {
2769 av_log(matroska->ctx, AV_LOG_WARNING,
2770 "EBML header using unsupported features\n"
2771 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2772 ebml.version, ebml.doctype, ebml.doctype_version);
2774 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2775 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2777 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2778 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2779 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2780 ebml_free(ebml_syntax, &ebml);
2781 return AVERROR_INVALIDDATA;
2784 ebml_free(ebml_syntax, &ebml);
2786 /* The next thing is a segment. */
2787 pos = avio_tell(matroska->ctx->pb);
2788 res = ebml_parse(matroska, matroska_segments, matroska);
2789 // try resyncing until we find a EBML_STOP type element.
2791 res = matroska_resync(matroska, pos);
2794 pos = avio_tell(matroska->ctx->pb);
2795 res = ebml_parse(matroska, matroska_segment, matroska);
2797 /* Set data_offset as it might be needed later by seek_frame_generic. */
2798 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2799 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2800 matroska_execute_seekhead(matroska);
2802 if (!matroska->time_scale)
2803 matroska->time_scale = 1000000;
2804 if (matroska->duration)
2805 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2806 1000 / AV_TIME_BASE;
2807 av_dict_set(&s->metadata, "title", matroska->title, 0);
2808 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2810 if (matroska->date_utc.size == 8)
2811 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2813 res = matroska_parse_tracks(s);
2817 attachments = attachments_list->elem;
2818 for (j = 0; j < attachments_list->nb_elem; j++) {
2819 if (!(attachments[j].filename && attachments[j].mime &&
2820 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2821 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2823 AVStream *st = avformat_new_stream(s, NULL);
2826 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2827 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2828 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2830 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2831 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2832 strlen(ff_mkv_image_mime_tags[i].str))) {
2833 st->codecpar->codec_id = ff_mkv_image_mime_tags[i].id;
2838 attachments[j].stream = st;
2840 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2841 AVPacket *pkt = &st->attached_pic;
2843 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2844 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2846 av_init_packet(pkt);
2847 pkt->buf = av_buffer_ref(attachments[j].bin.buf);
2849 return AVERROR(ENOMEM);
2850 pkt->data = attachments[j].bin.data;
2851 pkt->size = attachments[j].bin.size;
2852 pkt->stream_index = st->index;
2853 pkt->flags |= AV_PKT_FLAG_KEY;
2855 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2856 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2858 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2859 attachments[j].bin.size);
2861 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2862 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2863 strlen(ff_mkv_mime_tags[i].str))) {
2864 st->codecpar->codec_id = ff_mkv_mime_tags[i].id;
2872 chapters = chapters_list->elem;
2873 for (i = 0; i < chapters_list->nb_elem; i++)
2874 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2875 (max_start == 0 || chapters[i].start > max_start)) {
2876 chapters[i].chapter =
2877 avpriv_new_chapter(s, chapters[i].uid,
2878 (AVRational) { 1, 1000000000 },
2879 chapters[i].start, chapters[i].end,
2881 if (chapters[i].chapter) {
2882 av_dict_set(&chapters[i].chapter->metadata,
2883 "title", chapters[i].title, 0);
2885 max_start = chapters[i].start;
2888 matroska_add_index_entries(matroska);
2890 matroska_convert_tags(s);
2894 matroska_read_close(s);
2899 * Put one packet in an application-supplied AVPacket struct.
2900 * Returns 0 on success or -1 on failure.
2902 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2905 if (matroska->queue) {
2906 MatroskaTrack *tracks = matroska->tracks.elem;
2907 MatroskaTrack *track;
2909 ff_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
2910 track = &tracks[pkt->stream_index];
2911 if (track->has_palette) {
2912 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2914 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
2916 memcpy(pal, track->palette, AVPALETTE_SIZE);
2918 track->has_palette = 0;
2927 * Free all packets in our internal queue.
2929 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2931 ff_packet_list_free(&matroska->queue, &matroska->queue_end);
2934 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2935 int *buf_size, int type,
2936 uint32_t **lace_buf, int *laces)
2938 int res = 0, n, size = *buf_size;
2939 uint8_t *data = *buf;
2940 uint32_t *lace_size;
2944 *lace_buf = av_malloc(sizeof(**lace_buf));
2946 return AVERROR(ENOMEM);
2948 *lace_buf[0] = size;
2952 av_assert0(size > 0);
2956 lace_size = av_malloc_array(*laces, sizeof(*lace_size));
2958 return AVERROR(ENOMEM);
2961 case 0x1: /* Xiph lacing */
2965 for (n = 0; res == 0 && n < *laces - 1; n++) {
2969 if (size <= total) {
2970 res = AVERROR_INVALIDDATA;
2975 lace_size[n] += temp;
2982 if (size <= total) {
2983 res = AVERROR_INVALIDDATA;
2987 lace_size[n] = size - total;
2991 case 0x2: /* fixed-size lacing */
2992 if (size % (*laces)) {
2993 res = AVERROR_INVALIDDATA;
2996 for (n = 0; n < *laces; n++)
2997 lace_size[n] = size / *laces;
3000 case 0x3: /* EBML lacing */
3004 n = matroska_ebmlnum_uint(matroska, data, size, &num);
3005 if (n < 0 || num > INT_MAX) {
3006 av_log(matroska->ctx, AV_LOG_INFO,
3007 "EBML block data error\n");
3008 res = n<0 ? n : AVERROR_INVALIDDATA;
3013 total = lace_size[0] = num;
3014 for (n = 1; res == 0 && n < *laces - 1; n++) {
3017 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
3018 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
3019 av_log(matroska->ctx, AV_LOG_INFO,
3020 "EBML block data error\n");
3021 res = r<0 ? r : AVERROR_INVALIDDATA;
3026 lace_size[n] = lace_size[n - 1] + snum;
3027 total += lace_size[n];
3029 if (size <= total) {
3030 res = AVERROR_INVALIDDATA;
3033 lace_size[*laces - 1] = size - total;
3039 *lace_buf = lace_size;
3045 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3046 MatroskaTrack *track, AVStream *st,
3047 uint8_t *data, int size, uint64_t timecode,
3050 int a = st->codecpar->block_align;
3051 int sps = track->audio.sub_packet_size;
3052 int cfs = track->audio.coded_framesize;
3053 int h = track->audio.sub_packet_h;
3054 int y = track->audio.sub_packet_cnt;
3055 int w = track->audio.frame_size;
3058 if (!track->audio.pkt_cnt) {
3059 if (track->audio.sub_packet_cnt == 0)
3060 track->audio.buf_timecode = timecode;
3061 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3062 if (size < cfs * h / 2) {
3063 av_log(matroska->ctx, AV_LOG_ERROR,
3064 "Corrupt int4 RM-style audio packet size\n");
3065 return AVERROR_INVALIDDATA;
3067 for (x = 0; x < h / 2; x++)
3068 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3069 data + x * cfs, cfs);
3070 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3072 av_log(matroska->ctx, AV_LOG_ERROR,
3073 "Corrupt sipr RM-style audio packet size\n");
3074 return AVERROR_INVALIDDATA;
3076 memcpy(track->audio.buf + y * w, data, w);
3078 if (size < sps * w / sps || h<=0 || w%sps) {
3079 av_log(matroska->ctx, AV_LOG_ERROR,
3080 "Corrupt generic RM-style audio packet size\n");
3081 return AVERROR_INVALIDDATA;
3083 for (x = 0; x < w / sps; x++)
3084 memcpy(track->audio.buf +
3085 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3086 data + x * sps, sps);
3089 if (++track->audio.sub_packet_cnt >= h) {
3090 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3091 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3092 track->audio.sub_packet_cnt = 0;
3093 track->audio.pkt_cnt = h * w / a;
3097 while (track->audio.pkt_cnt) {
3099 AVPacket pktl, *pkt = &pktl;
3101 ret = av_new_packet(pkt, a);
3106 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3108 pkt->pts = track->audio.buf_timecode;
3109 track->audio.buf_timecode = AV_NOPTS_VALUE;
3111 pkt->stream_index = st->index;
3112 ret = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3114 av_packet_unref(pkt);
3115 return AVERROR(ENOMEM);
3122 /* reconstruct full wavpack blocks from mangled matroska ones */
3123 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
3124 uint8_t **pdst, int *size)
3126 uint8_t *dst = NULL;
3131 int ret, offset = 0;
3133 if (srclen < 12 || track->stream->codecpar->extradata_size < 2)
3134 return AVERROR_INVALIDDATA;
3136 ver = AV_RL16(track->stream->codecpar->extradata);
3138 samples = AV_RL32(src);
3142 while (srclen >= 8) {
3147 uint32_t flags = AV_RL32(src);
3148 uint32_t crc = AV_RL32(src + 4);
3152 multiblock = (flags & 0x1800) != 0x1800;
3155 ret = AVERROR_INVALIDDATA;
3158 blocksize = AV_RL32(src);
3164 if (blocksize > srclen) {
3165 ret = AVERROR_INVALIDDATA;
3169 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3171 ret = AVERROR(ENOMEM);
3175 dstlen += blocksize + 32;
3177 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3178 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3179 AV_WL16(dst + offset + 8, ver); // version
3180 AV_WL16(dst + offset + 10, 0); // track/index_no
3181 AV_WL32(dst + offset + 12, 0); // total samples
3182 AV_WL32(dst + offset + 16, 0); // block index
3183 AV_WL32(dst + offset + 20, samples); // number of samples
3184 AV_WL32(dst + offset + 24, flags); // flags
3185 AV_WL32(dst + offset + 28, crc); // crc
3186 memcpy(dst + offset + 32, src, blocksize); // block data
3189 srclen -= blocksize;
3190 offset += blocksize + 32;
3193 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3205 static int matroska_parse_prores(MatroskaTrack *track, uint8_t *src,
3206 uint8_t **pdst, int *size)
3211 if (AV_RB32(&src[4]) != MKBETAG('i', 'c', 'p', 'f')) {
3212 dst = av_malloc(dstlen + 8 + AV_INPUT_BUFFER_PADDING_SIZE);
3214 return AVERROR(ENOMEM);
3216 AV_WB32(dst, dstlen);
3217 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3218 memcpy(dst + 8, src, dstlen);
3219 memset(dst + 8 + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3229 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3230 MatroskaTrack *track,
3232 uint8_t *data, int data_len,
3237 AVPacket pktl, *pkt = &pktl;
3238 uint8_t *id, *settings, *text, *buf;
3239 int id_len, settings_len, text_len;
3244 return AVERROR_INVALIDDATA;
3247 q = data + data_len;
3252 if (*p == '\r' || *p == '\n') {
3261 if (p >= q || *p != '\n')
3262 return AVERROR_INVALIDDATA;
3268 if (*p == '\r' || *p == '\n') {
3269 settings_len = p - settings;
3277 if (p >= q || *p != '\n')
3278 return AVERROR_INVALIDDATA;
3283 while (text_len > 0) {
3284 const int len = text_len - 1;
3285 const uint8_t c = p[len];
3286 if (c != '\r' && c != '\n')
3292 return AVERROR_INVALIDDATA;
3294 err = av_new_packet(pkt, text_len);
3299 memcpy(pkt->data, text, text_len);
3302 buf = av_packet_new_side_data(pkt,
3303 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3306 av_packet_unref(pkt);
3307 return AVERROR(ENOMEM);
3309 memcpy(buf, id, id_len);
3312 if (settings_len > 0) {
3313 buf = av_packet_new_side_data(pkt,
3314 AV_PKT_DATA_WEBVTT_SETTINGS,
3317 av_packet_unref(pkt);
3318 return AVERROR(ENOMEM);
3320 memcpy(buf, settings, settings_len);
3323 // Do we need this for subtitles?
3324 // pkt->flags = AV_PKT_FLAG_KEY;
3326 pkt->stream_index = st->index;
3327 pkt->pts = timecode;
3329 // Do we need this for subtitles?
3330 // pkt->dts = timecode;
3332 pkt->duration = duration;
3335 err = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3337 av_packet_unref(pkt);
3338 return AVERROR(ENOMEM);
3344 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3345 MatroskaTrack *track, AVStream *st,
3346 AVBufferRef *buf, uint8_t *data, int pkt_size,
3347 uint64_t timecode, uint64_t lace_duration,
3348 int64_t pos, int is_keyframe,
3349 uint8_t *additional, uint64_t additional_id, int additional_size,
3350 int64_t discard_padding)
3352 MatroskaTrackEncoding *encodings = track->encodings.elem;
3353 uint8_t *pkt_data = data;
3355 AVPacket pktl, *pkt = &pktl;
3357 if (encodings && !encodings->type && encodings->scope & 1) {
3358 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
3363 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3365 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
3367 av_log(matroska->ctx, AV_LOG_ERROR,
3368 "Error parsing a wavpack block.\n");
3371 if (pkt_data != data)
3372 av_freep(&pkt_data);
3376 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES) {
3378 res = matroska_parse_prores(track, pkt_data, &pr_data, &pkt_size);
3380 av_log(matroska->ctx, AV_LOG_ERROR,
3381 "Error parsing a prores block.\n");
3384 if (pkt_data != data)
3385 av_freep(&pkt_data);
3389 av_init_packet(pkt);
3390 if (pkt_data != data)
3391 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3394 pkt->buf = av_buffer_ref(buf);
3397 res = AVERROR(ENOMEM);
3401 pkt->data = pkt_data;
3402 pkt->size = pkt_size;
3403 pkt->flags = is_keyframe;
3404 pkt->stream_index = st->index;
3406 if (additional_size > 0) {
3407 uint8_t *side_data = av_packet_new_side_data(pkt,
3408 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3409 additional_size + 8);
3411 av_packet_unref(pkt);
3412 return AVERROR(ENOMEM);
3414 AV_WB64(side_data, additional_id);
3415 memcpy(side_data + 8, additional, additional_size);
3418 if (discard_padding) {
3419 uint8_t *side_data = av_packet_new_side_data(pkt,
3420 AV_PKT_DATA_SKIP_SAMPLES,
3423 av_packet_unref(pkt);
3424 return AVERROR(ENOMEM);
3426 discard_padding = av_rescale_q(discard_padding,
3427 (AVRational){1, 1000000000},
3428 (AVRational){1, st->codecpar->sample_rate});
3429 if (discard_padding > 0) {
3430 AV_WL32(side_data + 4, discard_padding);
3432 AV_WL32(side_data, -discard_padding);
3436 if (track->ms_compat)
3437 pkt->dts = timecode;
3439 pkt->pts = timecode;
3441 pkt->duration = lace_duration;
3443 #if FF_API_CONVERGENCE_DURATION
3444 FF_DISABLE_DEPRECATION_WARNINGS
3445 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3446 pkt->convergence_duration = lace_duration;
3448 FF_ENABLE_DEPRECATION_WARNINGS
3451 res = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3453 av_packet_unref(pkt);
3454 return AVERROR(ENOMEM);
3460 if (pkt_data != data)
3461 av_freep(&pkt_data);
3465 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3466 int size, int64_t pos, uint64_t cluster_time,
3467 uint64_t block_duration, int is_keyframe,
3468 uint8_t *additional, uint64_t additional_id, int additional_size,
3469 int64_t cluster_pos, int64_t discard_padding)
3471 uint64_t timecode = AV_NOPTS_VALUE;
3472 MatroskaTrack *track;
3476 uint32_t *lace_size = NULL;
3477 int n, flags, laces = 0;
3479 int trust_default_duration = 1;
3481 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
3487 track = matroska_find_track_by_num(matroska, num);
3488 if (!track || !track->stream) {
3489 av_log(matroska->ctx, AV_LOG_INFO,
3490 "Invalid stream %"PRIu64"\n", num);
3491 return AVERROR_INVALIDDATA;
3492 } else if (size <= 3)
3495 if (st->discard >= AVDISCARD_ALL)
3497 av_assert1(block_duration != AV_NOPTS_VALUE);
3499 block_time = sign_extend(AV_RB16(data), 16);
3503 if (is_keyframe == -1)
3504 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3506 if (cluster_time != (uint64_t) -1 &&
3507 (block_time >= 0 || cluster_time >= -block_time)) {
3508 timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3509 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3510 timecode < track->end_timecode)
3511 is_keyframe = 0; /* overlapping subtitles are not key frame */
3513 ff_reduce_index(matroska->ctx, st->index);
3514 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3519 if (matroska->skip_to_keyframe &&
3520 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3521 // Compare signed timecodes. Timecode may be negative due to codec delay
3522 // offset. We don't support timestamps greater than int64_t anyway - see
3524 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3527 matroska->skip_to_keyframe = 0;
3528 else if (!st->skip_to_keyframe) {
3529 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3530 matroska->skip_to_keyframe = 0;
3534 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
3535 &lace_size, &laces);
3540 if (track->audio.samplerate == 8000) {
3541 // If this is needed for more codecs, then add them here
3542 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3543 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3544 trust_default_duration = 0;
3548 if (!block_duration && trust_default_duration)
3549 block_duration = track->default_duration * laces / matroska->time_scale;
3551 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3552 track->end_timecode =
3553 FFMAX(track->end_timecode, timecode + block_duration);
3555 for (n = 0; n < laces; n++) {
3556 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3558 if (lace_size[n] > size) {
3559 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
3563 if ((st->codecpar->codec_id == AV_CODEC_ID_RA_288 ||
3564 st->codecpar->codec_id == AV_CODEC_ID_COOK ||
3565 st->codecpar->codec_id == AV_CODEC_ID_SIPR ||
3566 st->codecpar->codec_id == AV_CODEC_ID_ATRAC3) &&
3567 st->codecpar->block_align && track->audio.sub_packet_size) {
3568 res = matroska_parse_rm_audio(matroska, track, st, data,
3574 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3575 res = matroska_parse_webvtt(matroska, track, st,
3577 timecode, lace_duration,
3582 res = matroska_parse_frame(matroska, track, st, buf, data, lace_size[n],
3583 timecode, lace_duration, pos,
3584 !n ? is_keyframe : 0,
3585 additional, additional_id, additional_size,
3591 if (timecode != AV_NOPTS_VALUE)
3592 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3593 data += lace_size[n];
3594 size -= lace_size[n];
3602 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3604 MatroskaCluster *cluster = &matroska->current_cluster;
3605 MatroskaBlock *block = &cluster->block;
3607 res = ebml_parse(matroska,
3608 matroska_cluster_parsing,
3613 ebml_level_end(matroska);
3614 cluster->pos = avio_tell(matroska->ctx->pb);
3615 /* sizeof the ID which was already read */
3616 if (matroska->current_id)
3618 res = ebml_parse(matroska,
3621 /* Try parsing the block again. */
3623 res = ebml_parse(matroska,
3624 matroska_cluster_parsing,
3630 if (res >= 0 && block->bin.size > 0) {
3631 int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3632 uint8_t* additional = block->additional.size > 0 ?
3633 block->additional.data : NULL;
3635 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3636 block->bin.size, block->bin.pos,
3637 matroska->current_cluster.timecode,
3638 block->duration, is_keyframe,
3639 additional, block->additional_id,
3640 block->additional.size,
3642 block->discard_padding);
3645 if (res == LEVEL_ENDED)
3648 ebml_free(matroska_blockgroup, block);
3649 memset(block, 0, sizeof(*block));
3654 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3656 MatroskaDemuxContext *matroska = s->priv_data;
3659 if (matroska->resync_pos == -1) {
3660 // This can only happen if generic seeking has been used.
3661 matroska->resync_pos = avio_tell(s->pb);
3664 while (matroska_deliver_packet(matroska, pkt)) {
3666 return (ret < 0) ? ret : AVERROR_EOF;
3667 if (matroska_parse_cluster(matroska) < 0)
3668 ret = matroska_resync(matroska, matroska->resync_pos);
3674 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3675 int64_t timestamp, int flags)
3677 MatroskaDemuxContext *matroska = s->priv_data;
3678 MatroskaTrack *tracks = NULL;
3679 AVStream *st = s->streams[stream_index];
3682 /* Parse the CUES now since we need the index data to seek. */
3683 if (matroska->cues_parsing_deferred > 0) {
3684 matroska->cues_parsing_deferred = 0;
3685 matroska_parse_cues(matroska);
3688 if (!st->nb_index_entries)
3690 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3692 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3693 matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3694 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3695 matroska_clear_queue(matroska);
3696 if (matroska_parse_cluster(matroska) < 0)
3701 matroska_clear_queue(matroska);
3702 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3705 tracks = matroska->tracks.elem;
3706 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3707 tracks[i].audio.pkt_cnt = 0;
3708 tracks[i].audio.sub_packet_cnt = 0;
3709 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3710 tracks[i].end_timecode = 0;
3713 /* We seek to a level 1 element, so set the appropriate status. */
3714 matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3715 if (flags & AVSEEK_FLAG_ANY) {
3716 st->skip_to_keyframe = 0;
3717 matroska->skip_to_timecode = timestamp;
3719 st->skip_to_keyframe = 1;
3720 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3722 matroska->skip_to_keyframe = 1;
3724 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3727 // slightly hackish but allows proper fallback to
3728 // the generic seeking code.
3729 matroska_reset_status(matroska, 0, -1);
3730 matroska->resync_pos = -1;
3731 matroska_clear_queue(matroska);
3732 st->skip_to_keyframe =
3733 matroska->skip_to_keyframe = 0;
3738 static int matroska_read_close(AVFormatContext *s)
3740 MatroskaDemuxContext *matroska = s->priv_data;
3741 MatroskaTrack *tracks = matroska->tracks.elem;
3744 matroska_clear_queue(matroska);
3746 for (n = 0; n < matroska->tracks.nb_elem; n++)
3747 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3748 av_freep(&tracks[n].audio.buf);
3749 ebml_free(matroska_segment, matroska);
3755 int64_t start_time_ns;
3756 int64_t end_time_ns;
3757 int64_t start_offset;
3761 /* This function searches all the Cues and returns the CueDesc corresponding to
3762 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3763 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3765 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3766 MatroskaDemuxContext *matroska = s->priv_data;
3769 int nb_index_entries = s->streams[0]->nb_index_entries;
3770 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3771 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3772 for (i = 1; i < nb_index_entries; i++) {
3773 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3774 index_entries[i].timestamp * matroska->time_scale > ts) {
3779 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3780 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3781 if (i != nb_index_entries - 1) {
3782 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3783 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3785 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3786 // FIXME: this needs special handling for files where Cues appear
3787 // before Clusters. the current logic assumes Cues appear after
3789 cue_desc.end_offset = cues_start - matroska->segment_start;
3794 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3796 MatroskaDemuxContext *matroska = s->priv_data;
3797 uint32_t id = matroska->current_id;
3798 int64_t cluster_pos, before_pos;
3800 if (s->streams[0]->nb_index_entries <= 0) return 0;
3801 // seek to the first cluster using cues.
3802 index = av_index_search_timestamp(s->streams[0], 0, 0);
3803 if (index < 0) return 0;
3804 cluster_pos = s->streams[0]->index_entries[index].pos;
3805 before_pos = avio_tell(s->pb);
3807 uint64_t cluster_id, cluster_length;
3810 avio_seek(s->pb, cluster_pos, SEEK_SET);
3811 // read cluster id and length
3812 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3813 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3815 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3819 matroska_reset_status(matroska, 0, cluster_pos);
3820 matroska_clear_queue(matroska);
3821 if (matroska_parse_cluster(matroska) < 0 ||
3825 pkt = &matroska->queue->pkt;
3826 // 4 + read is the length of the cluster id and the cluster length field.
3827 cluster_pos += 4 + read + cluster_length;
3828 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3834 /* Restore the status after matroska_read_header: */
3835 matroska_reset_status(matroska, id, before_pos);
3840 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3841 double min_buffer, double* buffer,
3842 double* sec_to_download, AVFormatContext *s,
3845 double nano_seconds_per_second = 1000000000.0;
3846 double time_sec = time_ns / nano_seconds_per_second;
3848 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3849 int64_t end_time_ns = time_ns + time_to_search_ns;
3850 double sec_downloaded = 0.0;
3851 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3852 if (desc_curr.start_time_ns == -1)
3854 *sec_to_download = 0.0;
3856 // Check for non cue start time.
3857 if (time_ns > desc_curr.start_time_ns) {
3858 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3859 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3860 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3861 double timeToDownload = (cueBytes * 8.0) / bps;
3863 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3864 *sec_to_download += timeToDownload;
3866 // Check if the search ends within the first cue.
3867 if (desc_curr.end_time_ns >= end_time_ns) {
3868 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3869 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3870 sec_downloaded = percent_to_sub * sec_downloaded;
3871 *sec_to_download = percent_to_sub * *sec_to_download;
3874 if ((sec_downloaded + *buffer) <= min_buffer) {
3878 // Get the next Cue.
3879 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3882 while (desc_curr.start_time_ns != -1) {
3883 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3884 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3885 double desc_sec = desc_ns / nano_seconds_per_second;
3886 double bits = (desc_bytes * 8.0);
3887 double time_to_download = bits / bps;
3889 sec_downloaded += desc_sec - time_to_download;
3890 *sec_to_download += time_to_download;
3892 if (desc_curr.end_time_ns >= end_time_ns) {
3893 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3894 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3895 sec_downloaded = percent_to_sub * sec_downloaded;
3896 *sec_to_download = percent_to_sub * *sec_to_download;
3898 if ((sec_downloaded + *buffer) <= min_buffer)
3903 if ((sec_downloaded + *buffer) <= min_buffer) {
3908 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3910 *buffer = *buffer + sec_downloaded;
3914 /* This function computes the bandwidth of the WebM file with the help of
3915 * buffer_size_after_time_downloaded() function. Both of these functions are
3916 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3917 * Matroska parsing mechanism.
3919 * Returns the bandwidth of the file on success; -1 on error.
3921 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3923 MatroskaDemuxContext *matroska = s->priv_data;
3924 AVStream *st = s->streams[0];
3925 double bandwidth = 0.0;
3928 for (i = 0; i < st->nb_index_entries; i++) {
3929 int64_t prebuffer_ns = 1000000000;
3930 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3931 double nano_seconds_per_second = 1000000000.0;
3932 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3933 double prebuffer_bytes = 0.0;
3934 int64_t temp_prebuffer_ns = prebuffer_ns;
3935 int64_t pre_bytes, pre_ns;
3936 double pre_sec, prebuffer, bits_per_second;
3937 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3939 // Start with the first Cue.
3940 CueDesc desc_end = desc_beg;
3942 // Figure out how much data we have downloaded for the prebuffer. This will
3943 // be used later to adjust the bits per sample to try.
3944 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3945 // Prebuffered the entire Cue.
3946 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3947 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3948 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3950 if (desc_end.start_time_ns == -1) {
3951 // The prebuffer is larger than the duration.
3952 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3954 bits_per_second = 0.0;
3956 // The prebuffer ends in the last Cue. Estimate how much data was
3958 pre_bytes = desc_end.end_offset - desc_end.start_offset;
3959 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3960 pre_sec = pre_ns / nano_seconds_per_second;
3962 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3964 prebuffer = prebuffer_ns / nano_seconds_per_second;
3966 // Set this to 0.0 in case our prebuffer buffers the entire video.
3967 bits_per_second = 0.0;
3969 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
3970 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
3971 double desc_sec = desc_ns / nano_seconds_per_second;
3972 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
3974 // Drop the bps by the percentage of bytes buffered.
3975 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
3976 double mod_bits_per_second = calc_bits_per_second * percent;
3978 if (prebuffer < desc_sec) {
3980 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
3982 // Add 1 so the bits per second should be a little bit greater than file
3984 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
3985 const double min_buffer = 0.0;
3986 double buffer = prebuffer;
3987 double sec_to_download = 0.0;
3989 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
3990 min_buffer, &buffer, &sec_to_download,
3994 } else if (rv == 0) {
3995 bits_per_second = (double)(bps);
4000 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4001 } while (desc_end.start_time_ns != -1);
4003 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4005 return (int64_t)bandwidth;
4008 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4010 MatroskaDemuxContext *matroska = s->priv_data;
4011 EbmlList *seekhead_list = &matroska->seekhead;
4012 MatroskaSeekhead *seekhead = seekhead_list->elem;
4014 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4018 // determine cues start and end positions
4019 for (i = 0; i < seekhead_list->nb_elem; i++)
4020 if (seekhead[i].id == MATROSKA_ID_CUES)
4023 if (i >= seekhead_list->nb_elem) return -1;
4025 before_pos = avio_tell(matroska->ctx->pb);
4026 cues_start = seekhead[i].pos + matroska->segment_start;
4027 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4028 // cues_end is computed as cues_start + cues_length + length of the
4029 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4030 // cues_end is inclusive and the above sum is reduced by 1.
4031 uint64_t cues_length, cues_id;
4033 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4034 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4035 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4036 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4039 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4041 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4042 if (cues_start == -1 || cues_end == -1) return -1;
4045 matroska_parse_cues(matroska);
4048 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4051 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4053 // if the file has cues at the start, fix up the init range so tht
4054 // it does not include it
4055 if (cues_start <= init_range)
4056 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4059 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4060 if (bandwidth < 0) return -1;
4061 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4063 // check if all clusters start with key frames
4064 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4066 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4067 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4068 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4069 if (!buf) return -1;
4071 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4072 int ret = snprintf(buf + end, 20,
4073 "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4074 i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4075 if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4076 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4078 return AVERROR_INVALIDDATA;
4082 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
4088 static int webm_dash_manifest_read_header(AVFormatContext *s)
4091 int ret = matroska_read_header(s);
4093 MatroskaTrack *tracks;
4094 MatroskaDemuxContext *matroska = s->priv_data;
4096 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4099 if (!s->nb_streams) {
4100 matroska_read_close(s);
4101 av_log(s, AV_LOG_ERROR, "No streams found\n");
4102 return AVERROR_INVALIDDATA;
4105 if (!matroska->is_live) {
4106 buf = av_asprintf("%g", matroska->duration);
4107 if (!buf) return AVERROR(ENOMEM);
4108 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
4111 // initialization range
4112 // 5 is the offset of Cluster ID.
4113 init_range = avio_tell(s->pb) - 5;
4114 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4117 // basename of the file
4118 buf = strrchr(s->url, '/');
4119 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4122 tracks = matroska->tracks.elem;
4123 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4125 // parse the cues and populate Cue related fields
4126 if (!matroska->is_live) {
4127 ret = webm_dash_manifest_cues(s, init_range);
4129 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4134 // use the bandwidth from the command line if it was provided
4135 if (matroska->bandwidth > 0) {
4136 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4137 matroska->bandwidth, 0);
4142 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4147 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4148 static const AVOption options[] = {
4149 { "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 },
4150 { "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 },
4154 static const AVClass webm_dash_class = {
4155 .class_name = "WebM DASH Manifest demuxer",
4156 .item_name = av_default_item_name,
4158 .version = LIBAVUTIL_VERSION_INT,
4161 AVInputFormat ff_matroska_demuxer = {
4162 .name = "matroska,webm",
4163 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4164 .extensions = "mkv,mk3d,mka,mks",
4165 .priv_data_size = sizeof(MatroskaDemuxContext),
4166 .read_probe = matroska_probe,
4167 .read_header = matroska_read_header,
4168 .read_packet = matroska_read_packet,
4169 .read_close = matroska_read_close,
4170 .read_seek = matroska_read_seek,
4171 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4174 AVInputFormat ff_webm_dash_manifest_demuxer = {
4175 .name = "webm_dash_manifest",
4176 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4177 .priv_data_size = sizeof(MatroskaDemuxContext),
4178 .read_header = webm_dash_manifest_read_header,
4179 .read_packet = webm_dash_manifest_read_packet,
4180 .read_close = matroska_read_close,
4181 .priv_class = &webm_dash_class,