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. */
76 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
77 * of unkown, potentially damaged data is encountered,
78 * it is considered an error. */
79 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
80 * to this many bytes of unknown data for the
81 * SKIP_THRESHOLD check. */
97 typedef const struct EbmlSyntax {
107 const struct EbmlSyntax *n;
111 typedef struct EbmlList {
116 typedef struct EbmlBin {
123 typedef struct Ebml {
128 uint64_t doctype_version;
131 typedef struct MatroskaTrackCompression {
134 } MatroskaTrackCompression;
136 typedef struct MatroskaTrackEncryption {
139 } MatroskaTrackEncryption;
141 typedef struct MatroskaTrackEncoding {
144 MatroskaTrackCompression compression;
145 MatroskaTrackEncryption encryption;
146 } MatroskaTrackEncoding;
148 typedef struct MatroskaMasteringMeta {
157 double max_luminance;
158 double min_luminance;
159 } MatroskaMasteringMeta;
161 typedef struct MatroskaTrackVideoColor {
162 uint64_t matrix_coefficients;
163 uint64_t bits_per_channel;
164 uint64_t chroma_sub_horz;
165 uint64_t chroma_sub_vert;
166 uint64_t cb_sub_horz;
167 uint64_t cb_sub_vert;
168 uint64_t chroma_siting_horz;
169 uint64_t chroma_siting_vert;
171 uint64_t transfer_characteristics;
175 MatroskaMasteringMeta mastering_meta;
176 } MatroskaTrackVideoColor;
178 typedef struct MatroskaTrackVideoProjection {
184 } MatroskaTrackVideoProjection;
186 typedef struct MatroskaTrackVideo {
188 uint64_t display_width;
189 uint64_t display_height;
190 uint64_t pixel_width;
191 uint64_t pixel_height;
193 uint64_t display_unit;
195 uint64_t field_order;
196 uint64_t stereo_mode;
199 MatroskaTrackVideoProjection projection;
200 } MatroskaTrackVideo;
202 typedef struct MatroskaTrackAudio {
204 double out_samplerate;
208 /* real audio header (extracted from extradata) */
215 uint64_t buf_timecode;
217 } MatroskaTrackAudio;
219 typedef struct MatroskaTrackPlane {
222 } MatroskaTrackPlane;
224 typedef struct MatroskaTrackOperation {
225 EbmlList combine_planes;
226 } MatroskaTrackOperation;
228 typedef struct MatroskaTrack {
237 uint64_t default_duration;
238 uint64_t flag_default;
239 uint64_t flag_forced;
240 uint64_t seek_preroll;
241 MatroskaTrackVideo video;
242 MatroskaTrackAudio audio;
243 MatroskaTrackOperation operation;
245 uint64_t codec_delay;
246 uint64_t codec_delay_in_track_tb;
249 int64_t end_timecode;
251 uint64_t max_block_additional_id;
253 uint32_t palette[AVPALETTE_COUNT];
257 typedef struct MatroskaAttachment {
264 } MatroskaAttachment;
266 typedef struct MatroskaChapter {
275 typedef struct MatroskaIndexPos {
280 typedef struct MatroskaIndex {
285 typedef struct MatroskaTag {
293 typedef struct MatroskaTagTarget {
301 typedef struct MatroskaTags {
302 MatroskaTagTarget target;
306 typedef struct MatroskaSeekhead {
311 typedef struct MatroskaLevel {
316 typedef struct MatroskaBlock {
321 uint64_t additional_id;
323 int64_t discard_padding;
326 typedef struct MatroskaCluster {
332 typedef struct MatroskaLevel1Element {
336 } MatroskaLevel1Element;
338 typedef struct MatroskaDemuxContext {
339 const AVClass *class;
340 AVFormatContext *ctx;
343 MatroskaLevel levels[EBML_MAX_DEPTH];
355 EbmlList attachments;
361 /* byte position of the segment inside the stream */
362 int64_t segment_start;
364 /* the packet queue */
366 AVPacketList *queue_end;
370 /* What to skip before effectively reading a packet. */
371 int skip_to_keyframe;
372 uint64_t skip_to_timecode;
374 /* File has a CUES element, but we defer parsing until it is needed. */
375 int cues_parsing_deferred;
377 /* Level1 elements and whether they were read yet */
378 MatroskaLevel1Element level1_elems[64];
379 int num_level1_elems;
381 MatroskaCluster current_cluster;
383 /* WebM DASH Manifest live flag */
386 /* Bandwidth value for WebM DASH Manifest */
388 } MatroskaDemuxContext;
390 #define CHILD_OF(parent) { .def = { .n = parent } }
392 static const EbmlSyntax ebml_syntax[], matroska_segment[], matroska_track_video_color[], matroska_track_video[],
393 matroska_track[], matroska_track_encoding[], matroska_track_encodings[],
394 matroska_track_combine_planes[], matroska_track_operation[], matroska_tracks[],
395 matroska_attachments[], matroska_chapter_entry[], matroska_chapter[], matroska_chapters[],
396 matroska_index_entry[], matroska_index[], matroska_tag[], matroska_tags[], matroska_seekhead[],
397 matroska_blockadditions[], matroska_blockgroup[], matroska_cluster_parsing[];
399 static const EbmlSyntax ebml_header[] = {
400 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
401 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
402 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
403 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
404 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
405 { EBML_ID_EBMLVERSION, EBML_NONE },
406 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
407 CHILD_OF(ebml_syntax)
410 static const EbmlSyntax ebml_syntax[] = {
411 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
412 { MATROSKA_ID_SEGMENT, EBML_STOP },
416 static const EbmlSyntax matroska_info[] = {
417 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
418 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
419 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
420 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
421 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
422 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
423 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
424 CHILD_OF(matroska_segment)
427 static const EbmlSyntax matroska_mastering_meta[] = {
428 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
429 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
430 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
431 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
432 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
433 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
434 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
435 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
436 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
437 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
438 CHILD_OF(matroska_track_video_color)
441 static const EbmlSyntax matroska_track_video_color[] = {
442 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
443 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=0 } },
444 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
445 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
446 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
447 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
448 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
449 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
450 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
451 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
452 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
453 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
454 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
455 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
456 CHILD_OF(matroska_track_video)
459 static const EbmlSyntax matroska_track_video_projection[] = {
460 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
461 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrackVideoProjection, private) },
462 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f=0.0 } },
463 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f=0.0 } },
464 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f=0.0 } },
465 CHILD_OF(matroska_track_video)
468 static const EbmlSyntax matroska_track_video[] = {
469 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
470 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
471 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
472 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
473 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
474 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
475 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
476 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
477 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
478 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
479 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
480 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
481 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
482 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
483 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
484 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
485 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
486 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
487 CHILD_OF(matroska_track)
490 static const EbmlSyntax matroska_track_audio[] = {
491 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
492 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
493 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
494 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
495 CHILD_OF(matroska_track)
498 static const EbmlSyntax matroska_track_encoding_compression[] = {
499 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
500 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
501 CHILD_OF(matroska_track_encoding)
504 static const EbmlSyntax matroska_track_encoding_encryption[] = {
505 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
506 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
507 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
508 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
509 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
510 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
511 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
512 CHILD_OF(matroska_track_encoding)
514 static const EbmlSyntax matroska_track_encoding[] = {
515 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
516 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
517 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
518 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
519 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
520 CHILD_OF(matroska_track_encodings)
523 static const EbmlSyntax matroska_track_encodings[] = {
524 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
525 CHILD_OF(matroska_track)
528 static const EbmlSyntax matroska_track_plane[] = {
529 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
530 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
531 CHILD_OF(matroska_track_combine_planes)
534 static const EbmlSyntax matroska_track_combine_planes[] = {
535 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
536 CHILD_OF(matroska_track_operation)
539 static const EbmlSyntax matroska_track_operation[] = {
540 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
541 CHILD_OF(matroska_track)
544 static const EbmlSyntax matroska_track[] = {
545 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
546 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
547 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
548 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
549 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
550 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
551 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
552 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
553 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
554 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
555 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
556 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
557 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
558 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
559 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
560 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
561 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
562 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
563 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
564 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
565 { MATROSKA_ID_CODECNAME, EBML_NONE },
566 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
567 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
568 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
569 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
570 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
571 CHILD_OF(matroska_tracks)
574 static const EbmlSyntax matroska_tracks[] = {
575 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
576 CHILD_OF(matroska_segment)
579 static const EbmlSyntax matroska_attachment[] = {
580 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
581 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
582 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
583 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
584 { MATROSKA_ID_FILEDESC, EBML_NONE },
585 CHILD_OF(matroska_attachments)
588 static const EbmlSyntax matroska_attachments[] = {
589 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
590 CHILD_OF(matroska_segment)
593 static const EbmlSyntax matroska_chapter_display[] = {
594 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
595 { MATROSKA_ID_CHAPLANG, EBML_NONE },
596 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
597 CHILD_OF(matroska_chapter_entry)
600 static const EbmlSyntax matroska_chapter_entry[] = {
601 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
602 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
603 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
604 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
605 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
606 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
607 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
608 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
609 CHILD_OF(matroska_chapter)
612 static const EbmlSyntax matroska_chapter[] = {
613 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
614 { MATROSKA_ID_EDITIONUID, EBML_NONE },
615 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
616 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
617 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
618 CHILD_OF(matroska_chapters)
621 static const EbmlSyntax matroska_chapters[] = {
622 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
623 CHILD_OF(matroska_segment)
626 static const EbmlSyntax matroska_index_pos[] = {
627 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
628 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
629 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
630 { MATROSKA_ID_CUEDURATION, EBML_NONE },
631 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
632 CHILD_OF(matroska_index_entry)
635 static const EbmlSyntax matroska_index_entry[] = {
636 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
637 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
638 CHILD_OF(matroska_index)
641 static const EbmlSyntax matroska_index[] = {
642 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
643 CHILD_OF(matroska_segment)
646 static const EbmlSyntax matroska_simpletag[] = {
647 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
648 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
649 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
650 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
651 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
652 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
653 CHILD_OF(matroska_tag)
656 static const EbmlSyntax matroska_tagtargets[] = {
657 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
658 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
659 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
660 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
661 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
662 CHILD_OF(matroska_tag)
665 static const EbmlSyntax matroska_tag[] = {
666 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
667 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
668 CHILD_OF(matroska_tags)
671 static const EbmlSyntax matroska_tags[] = {
672 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
673 CHILD_OF(matroska_segment)
676 static const EbmlSyntax matroska_seekhead_entry[] = {
677 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
678 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
679 CHILD_OF(matroska_seekhead)
682 static const EbmlSyntax matroska_seekhead[] = {
683 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
684 CHILD_OF(matroska_segment)
687 static const EbmlSyntax matroska_segment[] = {
688 { MATROSKA_ID_CLUSTER, EBML_STOP },
689 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
690 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
691 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
692 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
693 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
694 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
695 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
696 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
699 static const EbmlSyntax matroska_segments[] = {
700 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
704 static const EbmlSyntax matroska_blockmore[] = {
705 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id) },
706 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
707 CHILD_OF(matroska_blockadditions)
710 static const EbmlSyntax matroska_blockadditions[] = {
711 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
712 CHILD_OF(matroska_blockgroup)
715 static const EbmlSyntax matroska_blockgroup[] = {
716 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
717 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
718 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
719 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
720 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference), { .i = INT64_MIN } },
721 { MATROSKA_ID_CODECSTATE, EBML_NONE },
722 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
723 CHILD_OF(matroska_cluster_parsing)
726 // The following array contains SimpleBlock and BlockGroup twice
727 // in order to reuse the other values for matroska_cluster_enter.
728 static const EbmlSyntax matroska_cluster_parsing[] = {
729 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
730 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, { .n = matroska_blockgroup } },
731 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
732 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
733 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
734 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
735 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
736 CHILD_OF(matroska_segment)
739 static const EbmlSyntax matroska_cluster_enter[] = {
740 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = &matroska_cluster_parsing[2] } },
745 static const char *const matroska_doctypes[] = { "matroska", "webm" };
747 static int matroska_read_close(AVFormatContext *s);
750 * This function prepares the status for parsing of level 1 elements.
752 static int matroska_reset_status(MatroskaDemuxContext *matroska,
753 uint32_t id, int64_t position)
756 int err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
761 matroska->current_id = id;
762 matroska->num_levels = 1;
763 matroska->unknown_count = 0;
764 matroska->resync_pos = avio_tell(matroska->ctx->pb);
766 matroska->resync_pos -= (av_log2(id) + 7) / 8;
771 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
773 AVIOContext *pb = matroska->ctx->pb;
776 /* Try to seek to the last position to resync from. If this doesn't work,
777 * we resync from the earliest position available: The start of the buffer. */
778 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
779 av_log(matroska->ctx, AV_LOG_WARNING,
780 "Seek to desired resync point failed. Seeking to "
781 "earliest point available instead.\n");
782 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
783 last_pos + 1), SEEK_SET);
788 // try to find a toplevel element
789 while (!avio_feof(pb)) {
790 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
791 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
792 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
793 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
794 /* Prepare the context for parsing of a level 1 element. */
795 matroska_reset_status(matroska, id, -1);
796 /* Given that we are here means that an error has occured,
797 * so treat the segment as unknown length in order not to
798 * discard valid data that happens to be beyond the designated
799 * end of the segment. */
800 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
803 id = (id << 8) | avio_r8(pb);
807 return pb->error ? pb->error : AVERROR_EOF;
811 * Read: an "EBML number", which is defined as a variable-length
812 * array of bytes. The first byte indicates the length by giving a
813 * number of 0-bits followed by a one. The position of the first
814 * "one" bit inside the first byte indicates the length of this
816 * Returns: number of bytes read, < 0 on error
818 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
819 int max_size, uint64_t *number, int eof_forbidden)
825 /* The first byte tells us the length in bytes - except when it is zero. */
830 /* get the length of the EBML number */
831 read = 8 - ff_log2_tab[total];
833 if (!total || read > max_size) {
834 pos = avio_tell(pb) - 1;
836 av_log(matroska->ctx, AV_LOG_ERROR,
837 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
838 "of an EBML number\n", pos, pos);
840 av_log(matroska->ctx, AV_LOG_ERROR,
841 "Length %d indicated by an EBML number's first byte 0x%02x "
842 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
843 read, (uint8_t) total, pos, pos, max_size);
845 return AVERROR_INVALIDDATA;
848 /* read out length */
849 total ^= 1 << ff_log2_tab[total];
851 total = (total << 8) | avio_r8(pb);
853 if (pb->eof_reached) {
865 av_log(matroska->ctx, AV_LOG_ERROR,
866 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
871 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
872 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
879 * Read a EBML length value.
880 * This needs special handling for the "unknown length" case which has multiple
883 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
886 int res = ebml_read_num(matroska, pb, 8, number, 1);
887 if (res > 0 && *number + 1 == 1ULL << (7 * res))
888 *number = EBML_UNKNOWN_LENGTH;
893 * Read the next element as an unsigned int.
894 * Returns NEEDS_CHECKING.
896 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
900 /* big-endian ordering; build up number */
903 *num = (*num << 8) | avio_r8(pb);
905 return NEEDS_CHECKING;
909 * Read the next element as a signed int.
910 * Returns NEEDS_CHECKING.
912 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
919 *num = sign_extend(avio_r8(pb), 8);
921 /* big-endian ordering; build up number */
923 *num = ((uint64_t)*num << 8) | avio_r8(pb);
926 return NEEDS_CHECKING;
930 * Read the next element as a float.
931 * Returns NEEDS_CHECKING or < 0 on obvious failure.
933 static int ebml_read_float(AVIOContext *pb, int size, double *num)
938 *num = av_int2float(avio_rb32(pb));
940 *num = av_int2double(avio_rb64(pb));
942 return AVERROR_INVALIDDATA;
944 return NEEDS_CHECKING;
948 * Read the next element as an ASCII string.
949 * 0 is success, < 0 or NEEDS_CHECKING is failure.
951 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
956 /* EBML strings are usually not 0-terminated, so we allocate one
957 * byte more, read the string and NULL-terminate it ourselves. */
958 if (!(res = av_malloc(size + 1)))
959 return AVERROR(ENOMEM);
960 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
962 return ret < 0 ? ret : NEEDS_CHECKING;
972 * Read the next element as binary data.
973 * 0 is success, < 0 or NEEDS_CHECKING is failure.
975 static int ebml_read_binary(AVIOContext *pb, int length,
976 int64_t pos, EbmlBin *bin)
980 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
983 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
985 bin->data = bin->buf->data;
988 if ((ret = avio_read(pb, bin->data, length)) != length) {
989 av_buffer_unref(&bin->buf);
992 return ret < 0 ? ret : NEEDS_CHECKING;
999 * Read the next element, but only the header. The contents
1000 * are supposed to be sub-elements which can be read separately.
1001 * 0 is success, < 0 is failure.
1003 static int ebml_read_master(MatroskaDemuxContext *matroska,
1004 uint64_t length, int64_t pos)
1006 MatroskaLevel *level;
1008 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1009 av_log(matroska->ctx, AV_LOG_ERROR,
1010 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1011 return AVERROR(ENOSYS);
1014 level = &matroska->levels[matroska->num_levels++];
1016 level->length = length;
1022 * Read signed/unsigned "EBML" numbers.
1023 * Return: number of bytes processed, < 0 on error
1025 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
1026 uint8_t *data, uint32_t size, uint64_t *num)
1029 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
1030 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num, 1);
1034 * Same as above, but signed.
1036 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1037 uint8_t *data, uint32_t size, int64_t *num)
1042 /* read as unsigned number first */
1043 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
1046 /* make signed (weird way) */
1047 *num = unum - ((1LL << (7 * res - 1)) - 1);
1052 static int ebml_parse(MatroskaDemuxContext *matroska,
1053 EbmlSyntax *syntax, void *data);
1055 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1059 // Whoever touches this should be aware of the duplication
1060 // existing in matroska_cluster_parsing.
1061 for (i = 0; syntax[i].id; i++)
1062 if (id == syntax[i].id)
1068 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1074 for (int i = 0; syntax[i].id; i++)
1075 switch (syntax[i].type) {
1077 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1080 *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1083 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1087 // the default may be NULL
1088 if (syntax[i].def.s) {
1089 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1090 *dst = av_strdup(syntax[i].def.s);
1092 return AVERROR(ENOMEM);
1097 if (!matroska->levels[matroska->num_levels - 1].length) {
1098 matroska->num_levels--;
1104 res = ebml_parse(matroska, syntax, data);
1107 return res == LEVEL_ENDED ? 0 : res;
1110 static int is_ebml_id_valid(uint32_t id)
1112 // Due to endian nonsense in Matroska, the highest byte with any bits set
1113 // will contain the leading length bit. This bit in turn identifies the
1114 // total byte length of the element by its position within the byte.
1115 unsigned int bits = av_log2(id);
1116 return id && (bits + 7) / 8 == (8 - bits % 8);
1120 * Allocate and return the entry for the level1 element with the given ID. If
1121 * an entry already exists, return the existing entry.
1123 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1127 MatroskaLevel1Element *elem;
1129 if (!is_ebml_id_valid(id))
1132 // Some files link to all clusters; useless.
1133 if (id == MATROSKA_ID_CLUSTER)
1136 // There can be multiple seekheads.
1137 if (id != MATROSKA_ID_SEEKHEAD) {
1138 for (i = 0; i < matroska->num_level1_elems; i++) {
1139 if (matroska->level1_elems[i].id == id)
1140 return &matroska->level1_elems[i];
1144 // Only a completely broken file would have more elements.
1145 // It also provides a low-effort way to escape from circular seekheads
1146 // (every iteration will add a level1 entry).
1147 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1148 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1152 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1153 *elem = (MatroskaLevel1Element){.id = id};
1158 static int ebml_parse(MatroskaDemuxContext *matroska,
1159 EbmlSyntax *syntax, void *data)
1161 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1162 // Forbid unknown-length EBML_NONE elements.
1163 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1167 // max. 16 MB for strings
1168 [EBML_STR] = 0x1000000,
1169 [EBML_UTF8] = 0x1000000,
1170 // max. 256 MB for binary data
1171 [EBML_BIN] = 0x10000000,
1172 // no limits for anything else
1174 AVIOContext *pb = matroska->ctx->pb;
1177 int64_t pos = avio_tell(pb), pos_alt;
1178 int res, update_pos = 1, level_check;
1179 MatroskaLevel1Element *level1_elem;
1180 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1182 if (!matroska->current_id) {
1184 res = ebml_read_num(matroska, pb, 4, &id, 0);
1186 if (pb->eof_reached && res == AVERROR_EOF) {
1187 if (matroska->is_live)
1188 // in live mode, finish parsing if EOF is reached.
1190 if (level && pos == avio_tell(pb)) {
1191 if (level->length == EBML_UNKNOWN_LENGTH) {
1192 // Unknown-length levels automatically end at EOF.
1193 matroska->num_levels--;
1196 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1197 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1203 matroska->current_id = id | 1 << 7 * res;
1204 pos_alt = pos + res;
1207 pos -= (av_log2(matroska->current_id) + 7) / 8;
1210 id = matroska->current_id;
1212 syntax = ebml_parse_id(syntax, id);
1213 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1214 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1215 // Unknown-length levels end when an element from an upper level
1216 // in the hierarchy is encountered.
1217 while (syntax->def.n) {
1218 syntax = ebml_parse_id(syntax->def.n, id);
1220 matroska->num_levels--;
1226 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1227 "%"PRId64"\n", id, pos);
1228 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1232 data = (char *) data + syntax->data_offset;
1233 if (syntax->list_elem_size) {
1234 EbmlList *list = data;
1235 void *newelem = av_realloc_array(list->elem, list->nb_elem + 1,
1236 syntax->list_elem_size);
1238 return AVERROR(ENOMEM);
1239 list->elem = newelem;
1240 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1241 memset(data, 0, syntax->list_elem_size);
1246 if (syntax->type != EBML_STOP) {
1247 matroska->current_id = 0;
1248 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1253 if (matroska->num_levels > 0) {
1254 if (length != EBML_UNKNOWN_LENGTH &&
1255 level->length != EBML_UNKNOWN_LENGTH) {
1256 uint64_t elem_end = pos_alt + length,
1257 level_end = level->start + level->length;
1259 if (elem_end < level_end) {
1261 } else if (elem_end == level_end) {
1262 level_check = LEVEL_ENDED;
1264 av_log(matroska->ctx, AV_LOG_ERROR,
1265 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1266 "containing master element ending at 0x%"PRIx64"\n",
1267 pos, elem_end, level_end);
1268 return AVERROR_INVALIDDATA;
1270 } else if (length != EBML_UNKNOWN_LENGTH) {
1272 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1273 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1274 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1275 return AVERROR_INVALIDDATA;
1278 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1279 || syntax->type == EBML_NEST)) {
1280 // According to the current specifications only clusters and
1281 // segments are allowed to be unknown-length. We also accept
1282 // other unknown-length master elements.
1283 av_log(matroska->ctx, AV_LOG_WARNING,
1284 "Found unknown-length element 0x%"PRIX32" other than "
1285 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1286 "parsing will nevertheless be attempted.\n", id, pos);
1293 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1294 if (length != EBML_UNKNOWN_LENGTH) {
1295 av_log(matroska->ctx, AV_LOG_ERROR,
1296 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1297 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1298 length, max_lengths[syntax->type], id, pos);
1299 } else if (syntax->type != EBML_NONE) {
1300 av_log(matroska->ctx, AV_LOG_ERROR,
1301 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1302 "unknown length, yet the length of an element of its "
1303 "type must be known.\n", id, pos);
1305 av_log(matroska->ctx, AV_LOG_ERROR,
1306 "Found unknown-length element with ID 0x%"PRIX32" at "
1307 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1308 "available.\n", id, pos);
1310 return AVERROR_INVALIDDATA;
1313 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1314 // Loosing sync will likely manifest itself as encountering unknown
1315 // elements which are not reliably distinguishable from elements
1316 // belonging to future extensions of the format.
1317 // We use a heuristic to detect such situations: If the current
1318 // element is not expected at the current syntax level and there
1319 // were only a few unknown elements in a row, then the element is
1320 // skipped or considered defective based upon the length of the
1321 // current element (i.e. how much would be skipped); if there were
1322 // more than a few skipped elements in a row and skipping the current
1323 // element would lead us more than SKIP_THRESHOLD away from the last
1324 // known good position, then it is inferred that an error occured.
1325 // The dependency on the number of unknown elements in a row exists
1326 // because the distance to the last known good position is
1327 // automatically big if the last parsed element was big.
1328 // In both cases, each unknown element is considered equivalent to
1329 // UNKNOWN_EQUIV of skipped bytes for the check.
1330 // The whole check is only done for non-seekable output, because
1331 // in this situation skipped data can't simply be rechecked later.
1332 // This is especially important when using unkown length elements
1333 // as the check for whether a child exceeds its containing master
1334 // element is not effective in this situation.
1336 matroska->unknown_count = 0;
1338 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1340 if (matroska->unknown_count > 3)
1341 dist += pos_alt - matroska->resync_pos;
1343 if (dist > SKIP_THRESHOLD) {
1344 av_log(matroska->ctx, AV_LOG_ERROR,
1345 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1346 "length 0x%"PRIx64" considered as invalid data. Last "
1347 "known good position 0x%"PRIx64", %d unknown elements"
1348 " in a row\n", id, pos, length, matroska->resync_pos,
1349 matroska->unknown_count);
1350 return AVERROR_INVALIDDATA;
1355 if (update_pos > 0) {
1356 // We have found an element that is allowed at this place
1357 // in the hierarchy and it passed all checks, so treat the beginning
1358 // of the element as the "last known good" position.
1359 matroska->resync_pos = pos;
1362 if (!data && length != EBML_UNKNOWN_LENGTH)
1366 switch (syntax->type) {
1368 res = ebml_read_uint(pb, length, data);
1371 res = ebml_read_sint(pb, length, data);
1374 res = ebml_read_float(pb, length, data);
1378 res = ebml_read_ascii(pb, length, data);
1381 res = ebml_read_binary(pb, length, pos_alt, data);
1385 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1387 if (id == MATROSKA_ID_SEGMENT)
1388 matroska->segment_start = pos_alt;
1389 if (id == MATROSKA_ID_CUES)
1390 matroska->cues_parsing_deferred = 0;
1391 if (syntax->type == EBML_LEVEL1 &&
1392 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1393 if (level1_elem->parsed)
1394 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1395 level1_elem->parsed = 1;
1397 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1406 if (ffio_limit(pb, length) != length) {
1407 // ffio_limit emits its own error message,
1408 // so we don't have to.
1409 return AVERROR(EIO);
1411 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1412 // avio_skip might take us past EOF. We check for this
1413 // by skipping only length - 1 bytes, reading a byte and
1414 // checking the error flags. This is done in order to check
1415 // that the element has been properly skipped even when
1416 // no filesize (that ffio_limit relies on) is available.
1418 res = NEEDS_CHECKING;
1425 if (res == NEEDS_CHECKING) {
1426 if (pb->eof_reached) {
1435 if (res == AVERROR_INVALIDDATA)
1436 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1437 else if (res == AVERROR(EIO))
1438 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1439 else if (res == AVERROR_EOF) {
1440 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1448 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1449 level = &matroska->levels[matroska->num_levels - 1];
1450 pos = avio_tell(pb);
1452 // Given that pos >= level->start no check for
1453 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1454 while (matroska->num_levels && pos == level->start + level->length) {
1455 matroska->num_levels--;
1463 static void ebml_free(EbmlSyntax *syntax, void *data)
1466 for (i = 0; syntax[i].id; i++) {
1467 void *data_off = (char *) data + syntax[i].data_offset;
1468 switch (syntax[i].type) {
1474 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1478 if (syntax[i].list_elem_size) {
1479 EbmlList *list = data_off;
1480 char *ptr = list->elem;
1481 for (j = 0; j < list->nb_elem;
1482 j++, ptr += syntax[i].list_elem_size)
1483 ebml_free(syntax[i].def.n, ptr);
1484 av_freep(&list->elem);
1487 ebml_free(syntax[i].def.n, data_off);
1497 static int matroska_probe(const AVProbeData *p)
1500 int len_mask = 0x80, size = 1, n = 1, i;
1503 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1506 /* length of header */
1508 while (size <= 8 && !(total & len_mask)) {
1514 total &= (len_mask - 1);
1516 total = (total << 8) | p->buf[4 + n++];
1518 if (total + 1 == 1ULL << (7 * size)){
1519 /* Unknown-length header - simply parse the whole buffer. */
1520 total = p->buf_size - 4 - size;
1522 /* Does the probe data contain the whole header? */
1523 if (p->buf_size < 4 + size + total)
1527 /* The header should contain a known document type. For now,
1528 * we don't parse the whole header but simply check for the
1529 * availability of that array of characters inside the header.
1530 * Not fully fool-proof, but good enough. */
1531 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1532 size_t probelen = strlen(matroska_doctypes[i]);
1533 if (total < probelen)
1535 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1536 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1537 return AVPROBE_SCORE_MAX;
1540 // probably valid EBML header but no recognized doctype
1541 return AVPROBE_SCORE_EXTENSION;
1544 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1547 MatroskaTrack *tracks = matroska->tracks.elem;
1550 for (i = 0; i < matroska->tracks.nb_elem; i++)
1551 if (tracks[i].num == num)
1554 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1558 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1559 MatroskaTrack *track)
1561 MatroskaTrackEncoding *encodings = track->encodings.elem;
1562 uint8_t *data = *buf;
1563 int isize = *buf_size;
1564 uint8_t *pkt_data = NULL;
1565 uint8_t av_unused *newpktdata;
1566 int pkt_size = isize;
1570 if (pkt_size >= 10000000U)
1571 return AVERROR_INVALIDDATA;
1573 switch (encodings[0].compression.algo) {
1574 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1576 int header_size = encodings[0].compression.settings.size;
1577 uint8_t *header = encodings[0].compression.settings.data;
1579 if (header_size && !header) {
1580 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1587 pkt_size = isize + header_size;
1588 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1590 return AVERROR(ENOMEM);
1592 memcpy(pkt_data, header, header_size);
1593 memcpy(pkt_data + header_size, data, isize);
1597 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1599 olen = pkt_size *= 3;
1600 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1601 + AV_INPUT_BUFFER_PADDING_SIZE);
1603 result = AVERROR(ENOMEM);
1606 pkt_data = newpktdata;
1607 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1608 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1610 result = AVERROR_INVALIDDATA;
1617 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1619 z_stream zstream = { 0 };
1620 if (inflateInit(&zstream) != Z_OK)
1622 zstream.next_in = data;
1623 zstream.avail_in = isize;
1626 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1628 inflateEnd(&zstream);
1629 result = AVERROR(ENOMEM);
1632 pkt_data = newpktdata;
1633 zstream.avail_out = pkt_size - zstream.total_out;
1634 zstream.next_out = pkt_data + zstream.total_out;
1635 result = inflate(&zstream, Z_NO_FLUSH);
1636 } while (result == Z_OK && pkt_size < 10000000);
1637 pkt_size = zstream.total_out;
1638 inflateEnd(&zstream);
1639 if (result != Z_STREAM_END) {
1640 if (result == Z_MEM_ERROR)
1641 result = AVERROR(ENOMEM);
1643 result = AVERROR_INVALIDDATA;
1650 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1652 bz_stream bzstream = { 0 };
1653 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1655 bzstream.next_in = data;
1656 bzstream.avail_in = isize;
1659 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1661 BZ2_bzDecompressEnd(&bzstream);
1662 result = AVERROR(ENOMEM);
1665 pkt_data = newpktdata;
1666 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1667 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1668 result = BZ2_bzDecompress(&bzstream);
1669 } while (result == BZ_OK && pkt_size < 10000000);
1670 pkt_size = bzstream.total_out_lo32;
1671 BZ2_bzDecompressEnd(&bzstream);
1672 if (result != BZ_STREAM_END) {
1673 if (result == BZ_MEM_ERROR)
1674 result = AVERROR(ENOMEM);
1676 result = AVERROR_INVALIDDATA;
1683 return AVERROR_INVALIDDATA;
1686 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1689 *buf_size = pkt_size;
1697 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1698 AVDictionary **metadata, char *prefix)
1700 MatroskaTag *tags = list->elem;
1704 for (i = 0; i < list->nb_elem; i++) {
1705 const char *lang = tags[i].lang &&
1706 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1708 if (!tags[i].name) {
1709 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1713 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1715 av_strlcpy(key, tags[i].name, sizeof(key));
1716 if (tags[i].def || !lang) {
1717 av_dict_set(metadata, key, tags[i].string, 0);
1718 if (tags[i].sub.nb_elem)
1719 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1722 av_strlcat(key, "-", sizeof(key));
1723 av_strlcat(key, lang, sizeof(key));
1724 av_dict_set(metadata, key, tags[i].string, 0);
1725 if (tags[i].sub.nb_elem)
1726 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1729 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1732 static void matroska_convert_tags(AVFormatContext *s)
1734 MatroskaDemuxContext *matroska = s->priv_data;
1735 MatroskaTags *tags = matroska->tags.elem;
1738 for (i = 0; i < matroska->tags.nb_elem; i++) {
1739 if (tags[i].target.attachuid) {
1740 MatroskaAttachment *attachment = matroska->attachments.elem;
1742 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1743 if (attachment[j].uid == tags[i].target.attachuid &&
1744 attachment[j].stream) {
1745 matroska_convert_tag(s, &tags[i].tag,
1746 &attachment[j].stream->metadata, NULL);
1751 av_log(NULL, AV_LOG_WARNING,
1752 "The tags at index %d refer to a "
1753 "non-existent attachment %"PRId64".\n",
1754 i, tags[i].target.attachuid);
1756 } else if (tags[i].target.chapteruid) {
1757 MatroskaChapter *chapter = matroska->chapters.elem;
1759 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1760 if (chapter[j].uid == tags[i].target.chapteruid &&
1761 chapter[j].chapter) {
1762 matroska_convert_tag(s, &tags[i].tag,
1763 &chapter[j].chapter->metadata, NULL);
1768 av_log(NULL, AV_LOG_WARNING,
1769 "The tags at index %d refer to a non-existent chapter "
1771 i, tags[i].target.chapteruid);
1773 } else if (tags[i].target.trackuid) {
1774 MatroskaTrack *track = matroska->tracks.elem;
1776 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1777 if (track[j].uid == tags[i].target.trackuid &&
1779 matroska_convert_tag(s, &tags[i].tag,
1780 &track[j].stream->metadata, NULL);
1785 av_log(NULL, AV_LOG_WARNING,
1786 "The tags at index %d refer to a non-existent track "
1788 i, tags[i].target.trackuid);
1791 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1792 tags[i].target.type);
1797 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1800 uint32_t saved_id = matroska->current_id;
1801 int64_t before_pos = avio_tell(matroska->ctx->pb);
1806 offset = pos + matroska->segment_start;
1807 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1808 /* We don't want to lose our seekhead level, so we add
1809 * a dummy. This is a crude hack. */
1810 if (matroska->num_levels == EBML_MAX_DEPTH) {
1811 av_log(matroska->ctx, AV_LOG_INFO,
1812 "Max EBML element depth (%d) reached, "
1813 "cannot parse further.\n", EBML_MAX_DEPTH);
1814 ret = AVERROR_INVALIDDATA;
1816 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1817 matroska->num_levels++;
1818 matroska->current_id = 0;
1820 ret = ebml_parse(matroska, matroska_segment, matroska);
1821 if (ret == LEVEL_ENDED) {
1822 /* This can only happen if the seek brought us beyond EOF. */
1827 /* Seek back - notice that in all instances where this is used
1828 * it is safe to set the level to 1. */
1829 matroska_reset_status(matroska, saved_id, before_pos);
1834 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1836 EbmlList *seekhead_list = &matroska->seekhead;
1839 // we should not do any seeking in the streaming case
1840 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1843 for (i = 0; i < seekhead_list->nb_elem; i++) {
1844 MatroskaSeekhead *seekheads = seekhead_list->elem;
1845 uint32_t id = seekheads[i].id;
1846 uint64_t pos = seekheads[i].pos;
1848 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1849 if (!elem || elem->parsed)
1854 // defer cues parsing until we actually need cue data.
1855 if (id == MATROSKA_ID_CUES)
1858 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1859 // mark index as broken
1860 matroska->cues_parsing_deferred = -1;
1868 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1870 EbmlList *index_list;
1871 MatroskaIndex *index;
1872 uint64_t index_scale = 1;
1875 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1878 index_list = &matroska->index;
1879 index = index_list->elem;
1880 if (index_list->nb_elem < 2)
1882 if (index[1].time > 1E14 / matroska->time_scale) {
1883 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1886 for (i = 0; i < index_list->nb_elem; i++) {
1887 EbmlList *pos_list = &index[i].pos;
1888 MatroskaIndexPos *pos = pos_list->elem;
1889 for (j = 0; j < pos_list->nb_elem; j++) {
1890 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1892 if (track && track->stream)
1893 av_add_index_entry(track->stream,
1894 pos[j].pos + matroska->segment_start,
1895 index[i].time / index_scale, 0, 0,
1901 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1904 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1907 for (i = 0; i < matroska->num_level1_elems; i++) {
1908 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1909 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1910 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1911 matroska->cues_parsing_deferred = -1;
1917 matroska_add_index_entries(matroska);
1920 static int matroska_aac_profile(char *codec_id)
1922 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1925 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1926 if (strstr(codec_id, aac_profiles[profile]))
1931 static int matroska_aac_sri(int samplerate)
1935 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1936 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1941 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1943 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1944 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1947 static int matroska_parse_flac(AVFormatContext *s,
1948 MatroskaTrack *track,
1951 AVStream *st = track->stream;
1952 uint8_t *p = track->codec_priv.data;
1953 int size = track->codec_priv.size;
1955 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1956 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1957 track->codec_priv.size = 0;
1961 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1963 p += track->codec_priv.size;
1964 size -= track->codec_priv.size;
1966 /* parse the remaining metadata blocks if present */
1968 int block_last, block_type, block_size;
1970 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1974 if (block_size > size)
1977 /* check for the channel mask */
1978 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1979 AVDictionary *dict = NULL;
1980 AVDictionaryEntry *chmask;
1982 ff_vorbis_comment(s, &dict, p, block_size, 0);
1983 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1985 uint64_t mask = strtol(chmask->value, NULL, 0);
1986 if (!mask || mask & ~0x3ffffULL) {
1987 av_log(s, AV_LOG_WARNING,
1988 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1990 st->codecpar->channel_layout = mask;
1992 av_dict_free(&dict);
2002 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2004 int major, minor, micro, bttb = 0;
2006 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2007 * this function, and fixed in 57.52 */
2008 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, µ) == 3)
2009 bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
2011 switch (field_order) {
2012 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2013 return AV_FIELD_PROGRESSIVE;
2014 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2015 return AV_FIELD_UNKNOWN;
2016 case MATROSKA_VIDEO_FIELDORDER_TT:
2018 case MATROSKA_VIDEO_FIELDORDER_BB:
2020 case MATROSKA_VIDEO_FIELDORDER_BT:
2021 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2022 case MATROSKA_VIDEO_FIELDORDER_TB:
2023 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2025 return AV_FIELD_UNKNOWN;
2029 static void mkv_stereo_mode_display_mul(int stereo_mode,
2030 int *h_width, int *h_height)
2032 switch (stereo_mode) {
2033 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2034 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2035 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2036 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2037 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2039 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2040 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2041 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2042 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2045 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2046 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2047 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2048 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2054 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2055 const MatroskaTrackVideoColor *color = track->video.color.elem;
2056 const MatroskaMasteringMeta *mastering_meta;
2057 int has_mastering_primaries, has_mastering_luminance;
2059 if (!track->video.color.nb_elem)
2062 mastering_meta = &color->mastering_meta;
2063 // Mastering primaries are CIE 1931 coords, and must be > 0.
2064 has_mastering_primaries =
2065 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2066 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2067 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2068 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2069 has_mastering_luminance = mastering_meta->max_luminance > 0;
2071 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2072 st->codecpar->color_space = color->matrix_coefficients;
2073 if (color->primaries != AVCOL_PRI_RESERVED &&
2074 color->primaries != AVCOL_PRI_RESERVED0)
2075 st->codecpar->color_primaries = color->primaries;
2076 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2077 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2078 st->codecpar->color_trc = color->transfer_characteristics;
2079 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2080 color->range <= AVCOL_RANGE_JPEG)
2081 st->codecpar->color_range = color->range;
2082 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2083 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2084 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2085 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2086 st->codecpar->chroma_location =
2087 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2088 (color->chroma_siting_vert - 1) << 7);
2090 if (color->max_cll && color->max_fall) {
2093 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2095 return AVERROR(ENOMEM);
2096 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2097 (uint8_t *)metadata, size);
2099 av_freep(&metadata);
2102 metadata->MaxCLL = color->max_cll;
2103 metadata->MaxFALL = color->max_fall;
2106 if (has_mastering_primaries || has_mastering_luminance) {
2107 // Use similar rationals as other standards.
2108 const int chroma_den = 50000;
2109 const int luma_den = 10000;
2110 AVMasteringDisplayMetadata *metadata =
2111 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2112 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2113 sizeof(AVMasteringDisplayMetadata));
2115 return AVERROR(ENOMEM);
2117 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2118 if (has_mastering_primaries) {
2119 metadata->display_primaries[0][0] = av_make_q(
2120 round(mastering_meta->r_x * chroma_den), chroma_den);
2121 metadata->display_primaries[0][1] = av_make_q(
2122 round(mastering_meta->r_y * chroma_den), chroma_den);
2123 metadata->display_primaries[1][0] = av_make_q(
2124 round(mastering_meta->g_x * chroma_den), chroma_den);
2125 metadata->display_primaries[1][1] = av_make_q(
2126 round(mastering_meta->g_y * chroma_den), chroma_den);
2127 metadata->display_primaries[2][0] = av_make_q(
2128 round(mastering_meta->b_x * chroma_den), chroma_den);
2129 metadata->display_primaries[2][1] = av_make_q(
2130 round(mastering_meta->b_y * chroma_den), chroma_den);
2131 metadata->white_point[0] = av_make_q(
2132 round(mastering_meta->white_x * chroma_den), chroma_den);
2133 metadata->white_point[1] = av_make_q(
2134 round(mastering_meta->white_y * chroma_den), chroma_den);
2135 metadata->has_primaries = 1;
2137 if (has_mastering_luminance) {
2138 metadata->max_luminance = av_make_q(
2139 round(mastering_meta->max_luminance * luma_den), luma_den);
2140 metadata->min_luminance = av_make_q(
2141 round(mastering_meta->min_luminance * luma_den), luma_den);
2142 metadata->has_luminance = 1;
2148 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track) {
2149 AVSphericalMapping *spherical;
2150 enum AVSphericalProjection projection;
2151 size_t spherical_size;
2152 uint32_t l = 0, t = 0, r = 0, b = 0;
2153 uint32_t padding = 0;
2157 bytestream2_init(&gb, track->video.projection.private.data,
2158 track->video.projection.private.size);
2160 if (bytestream2_get_byte(&gb) != 0) {
2161 av_log(NULL, AV_LOG_WARNING, "Unknown spherical metadata\n");
2165 bytestream2_skip(&gb, 3); // flags
2167 switch (track->video.projection.type) {
2168 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2169 if (track->video.projection.private.size == 20) {
2170 t = bytestream2_get_be32(&gb);
2171 b = bytestream2_get_be32(&gb);
2172 l = bytestream2_get_be32(&gb);
2173 r = bytestream2_get_be32(&gb);
2175 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2176 av_log(NULL, AV_LOG_ERROR,
2177 "Invalid bounding rectangle coordinates "
2178 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2180 return AVERROR_INVALIDDATA;
2182 } else if (track->video.projection.private.size != 0) {
2183 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2184 return AVERROR_INVALIDDATA;
2187 if (l || t || r || b)
2188 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2190 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2192 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2193 if (track->video.projection.private.size < 4) {
2194 av_log(NULL, AV_LOG_ERROR, "Missing projection private properties\n");
2195 return AVERROR_INVALIDDATA;
2196 } else if (track->video.projection.private.size == 12) {
2197 uint32_t layout = bytestream2_get_be32(&gb);
2199 av_log(NULL, AV_LOG_WARNING,
2200 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2203 projection = AV_SPHERICAL_CUBEMAP;
2204 padding = bytestream2_get_be32(&gb);
2206 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2207 return AVERROR_INVALIDDATA;
2210 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2211 /* No Spherical metadata */
2214 av_log(NULL, AV_LOG_WARNING,
2215 "Unknown spherical metadata type %"PRIu64"\n",
2216 track->video.projection.type);
2220 spherical = av_spherical_alloc(&spherical_size);
2222 return AVERROR(ENOMEM);
2224 spherical->projection = projection;
2226 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2227 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2228 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2230 spherical->padding = padding;
2232 spherical->bound_left = l;
2233 spherical->bound_top = t;
2234 spherical->bound_right = r;
2235 spherical->bound_bottom = b;
2237 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2240 av_freep(&spherical);
2247 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2249 const AVCodecTag *codec_tags;
2251 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2252 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2254 /* Normalize noncompliant private data that starts with the fourcc
2255 * by expanding/shifting the data by 4 bytes and storing the data
2256 * size at the start. */
2257 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2258 int ret = av_buffer_realloc(&track->codec_priv.buf,
2259 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2263 track->codec_priv.data = track->codec_priv.buf->data;
2264 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2265 track->codec_priv.size += 4;
2266 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2269 *fourcc = AV_RL32(track->codec_priv.data + 4);
2270 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2275 static int matroska_parse_tracks(AVFormatContext *s)
2277 MatroskaDemuxContext *matroska = s->priv_data;
2278 MatroskaTrack *tracks = matroska->tracks.elem;
2283 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2284 MatroskaTrack *track = &tracks[i];
2285 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2286 EbmlList *encodings_list = &track->encodings;
2287 MatroskaTrackEncoding *encodings = encodings_list->elem;
2288 uint8_t *extradata = NULL;
2289 int extradata_size = 0;
2290 int extradata_offset = 0;
2291 uint32_t fourcc = 0;
2293 char* key_id_base64 = NULL;
2296 /* Apply some sanity checks. */
2297 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2298 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2299 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2300 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2301 av_log(matroska->ctx, AV_LOG_INFO,
2302 "Unknown or unsupported track type %"PRIu64"\n",
2306 if (!track->codec_id)
2309 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2310 isnan(track->audio.samplerate)) {
2311 av_log(matroska->ctx, AV_LOG_WARNING,
2312 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2313 track->audio.samplerate);
2314 track->audio.samplerate = 8000;
2317 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2318 if (!track->default_duration && track->video.frame_rate > 0) {
2319 double default_duration = 1000000000 / track->video.frame_rate;
2320 if (default_duration > UINT64_MAX || default_duration < 0) {
2321 av_log(matroska->ctx, AV_LOG_WARNING,
2322 "Invalid frame rate %e. Cannot calculate default duration.\n",
2323 track->video.frame_rate);
2325 track->default_duration = default_duration;
2328 if (track->video.display_width == -1)
2329 track->video.display_width = track->video.pixel_width;
2330 if (track->video.display_height == -1)
2331 track->video.display_height = track->video.pixel_height;
2332 if (track->video.color_space.size == 4)
2333 fourcc = AV_RL32(track->video.color_space.data);
2334 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2335 if (!track->audio.out_samplerate)
2336 track->audio.out_samplerate = track->audio.samplerate;
2338 if (encodings_list->nb_elem > 1) {
2339 av_log(matroska->ctx, AV_LOG_ERROR,
2340 "Multiple combined encodings not supported");
2341 } else if (encodings_list->nb_elem == 1) {
2342 if (encodings[0].type) {
2343 if (encodings[0].encryption.key_id.size > 0) {
2344 /* Save the encryption key id to be stored later as a
2346 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2347 key_id_base64 = av_malloc(b64_size);
2348 if (key_id_base64 == NULL)
2349 return AVERROR(ENOMEM);
2351 av_base64_encode(key_id_base64, b64_size,
2352 encodings[0].encryption.key_id.data,
2353 encodings[0].encryption.key_id.size);
2355 encodings[0].scope = 0;
2356 av_log(matroska->ctx, AV_LOG_ERROR,
2357 "Unsupported encoding type");
2361 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2364 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2367 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2369 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2370 encodings[0].scope = 0;
2371 av_log(matroska->ctx, AV_LOG_ERROR,
2372 "Unsupported encoding type");
2373 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2374 uint8_t *codec_priv = track->codec_priv.data;
2375 int ret = matroska_decode_buffer(&track->codec_priv.data,
2376 &track->codec_priv.size,
2379 track->codec_priv.data = NULL;
2380 track->codec_priv.size = 0;
2381 av_log(matroska->ctx, AV_LOG_ERROR,
2382 "Failed to decode codec private data\n");
2385 if (codec_priv != track->codec_priv.data) {
2386 av_buffer_unref(&track->codec_priv.buf);
2387 if (track->codec_priv.data) {
2388 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2389 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2391 if (!track->codec_priv.buf) {
2392 av_freep(&track->codec_priv.data);
2393 track->codec_priv.size = 0;
2394 return AVERROR(ENOMEM);
2401 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2402 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2403 strlen(ff_mkv_codec_tags[j].str))) {
2404 codec_id = ff_mkv_codec_tags[j].id;
2409 st = track->stream = avformat_new_stream(s, NULL);
2411 av_free(key_id_base64);
2412 return AVERROR(ENOMEM);
2415 if (key_id_base64) {
2416 /* export encryption key id as base64 metadata tag */
2417 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
2418 av_freep(&key_id_base64);
2421 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2422 track->codec_priv.size >= 40 &&
2423 track->codec_priv.data) {
2424 track->ms_compat = 1;
2425 bit_depth = AV_RL16(track->codec_priv.data + 14);
2426 fourcc = AV_RL32(track->codec_priv.data + 16);
2427 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2430 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2432 extradata_offset = 40;
2433 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2434 track->codec_priv.size >= 14 &&
2435 track->codec_priv.data) {
2437 ffio_init_context(&b, track->codec_priv.data,
2438 track->codec_priv.size,
2439 0, NULL, NULL, NULL, NULL);
2440 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2443 codec_id = st->codecpar->codec_id;
2444 fourcc = st->codecpar->codec_tag;
2445 extradata_offset = FFMIN(track->codec_priv.size, 18);
2446 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2447 /* Normally 36, but allow noncompliant private data */
2448 && (track->codec_priv.size >= 32)
2449 && (track->codec_priv.data)) {
2450 uint16_t sample_size;
2451 int ret = get_qt_codec(track, &fourcc, &codec_id);
2454 sample_size = AV_RB16(track->codec_priv.data + 26);
2456 if (sample_size == 8) {
2457 fourcc = MKTAG('r','a','w',' ');
2458 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2459 } else if (sample_size == 16) {
2460 fourcc = MKTAG('t','w','o','s');
2461 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2464 if ((fourcc == MKTAG('t','w','o','s') ||
2465 fourcc == MKTAG('s','o','w','t')) &&
2467 codec_id = AV_CODEC_ID_PCM_S8;
2468 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2469 (track->codec_priv.size >= 21) &&
2470 (track->codec_priv.data)) {
2471 int ret = get_qt_codec(track, &fourcc, &codec_id);
2474 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2475 fourcc = MKTAG('S','V','Q','3');
2476 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2478 if (codec_id == AV_CODEC_ID_NONE)
2479 av_log(matroska->ctx, AV_LOG_ERROR,
2480 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2481 if (track->codec_priv.size >= 86) {
2482 bit_depth = AV_RB16(track->codec_priv.data + 82);
2483 ffio_init_context(&b, track->codec_priv.data,
2484 track->codec_priv.size,
2485 0, NULL, NULL, NULL, NULL);
2486 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2488 track->has_palette = 1;
2491 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2492 switch (track->audio.bitdepth) {
2494 codec_id = AV_CODEC_ID_PCM_U8;
2497 codec_id = AV_CODEC_ID_PCM_S24BE;
2500 codec_id = AV_CODEC_ID_PCM_S32BE;
2503 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2504 switch (track->audio.bitdepth) {
2506 codec_id = AV_CODEC_ID_PCM_U8;
2509 codec_id = AV_CODEC_ID_PCM_S24LE;
2512 codec_id = AV_CODEC_ID_PCM_S32LE;
2515 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2516 track->audio.bitdepth == 64) {
2517 codec_id = AV_CODEC_ID_PCM_F64LE;
2518 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2519 int profile = matroska_aac_profile(track->codec_id);
2520 int sri = matroska_aac_sri(track->audio.samplerate);
2521 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2523 return AVERROR(ENOMEM);
2524 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2525 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2526 if (strstr(track->codec_id, "SBR")) {
2527 sri = matroska_aac_sri(track->audio.out_samplerate);
2528 extradata[2] = 0x56;
2529 extradata[3] = 0xE5;
2530 extradata[4] = 0x80 | (sri << 3);
2534 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2535 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2536 * Create the "atom size", "tag", and "tag version" fields the
2537 * decoder expects manually. */
2538 extradata_size = 12 + track->codec_priv.size;
2539 extradata = av_mallocz(extradata_size +
2540 AV_INPUT_BUFFER_PADDING_SIZE);
2542 return AVERROR(ENOMEM);
2543 AV_WB32(extradata, extradata_size);
2544 memcpy(&extradata[4], "alac", 4);
2545 AV_WB32(&extradata[8], 0);
2546 memcpy(&extradata[12], track->codec_priv.data,
2547 track->codec_priv.size);
2548 } else if (codec_id == AV_CODEC_ID_TTA) {
2549 extradata_size = 30;
2550 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2552 return AVERROR(ENOMEM);
2553 ffio_init_context(&b, extradata, extradata_size, 1,
2554 NULL, NULL, NULL, NULL);
2555 avio_write(&b, "TTA1", 4);
2557 if (track->audio.channels > UINT16_MAX ||
2558 track->audio.bitdepth > UINT16_MAX) {
2559 av_log(matroska->ctx, AV_LOG_WARNING,
2560 "Too large audio channel number %"PRIu64
2561 " or bitdepth %"PRIu64". Skipping track.\n",
2562 track->audio.channels, track->audio.bitdepth);
2563 av_freep(&extradata);
2564 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2565 return AVERROR_INVALIDDATA;
2569 avio_wl16(&b, track->audio.channels);
2570 avio_wl16(&b, track->audio.bitdepth);
2571 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2572 return AVERROR_INVALIDDATA;
2573 avio_wl32(&b, track->audio.out_samplerate);
2574 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
2575 track->audio.out_samplerate,
2576 AV_TIME_BASE * 1000));
2577 } else if (codec_id == AV_CODEC_ID_RV10 ||
2578 codec_id == AV_CODEC_ID_RV20 ||
2579 codec_id == AV_CODEC_ID_RV30 ||
2580 codec_id == AV_CODEC_ID_RV40) {
2581 extradata_offset = 26;
2582 } else if (codec_id == AV_CODEC_ID_RA_144) {
2583 track->audio.out_samplerate = 8000;
2584 track->audio.channels = 1;
2585 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2586 codec_id == AV_CODEC_ID_COOK ||
2587 codec_id == AV_CODEC_ID_ATRAC3 ||
2588 codec_id == AV_CODEC_ID_SIPR)
2589 && track->codec_priv.data) {
2592 ffio_init_context(&b, track->codec_priv.data,
2593 track->codec_priv.size,
2594 0, NULL, NULL, NULL, NULL);
2596 flavor = avio_rb16(&b);
2597 track->audio.coded_framesize = avio_rb32(&b);
2599 track->audio.sub_packet_h = avio_rb16(&b);
2600 track->audio.frame_size = avio_rb16(&b);
2601 track->audio.sub_packet_size = avio_rb16(&b);
2603 track->audio.coded_framesize <= 0 ||
2604 track->audio.sub_packet_h <= 0 ||
2605 track->audio.frame_size <= 0 ||
2606 track->audio.sub_packet_size <= 0 && codec_id != AV_CODEC_ID_SIPR)
2607 return AVERROR_INVALIDDATA;
2608 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2609 track->audio.frame_size);
2610 if (!track->audio.buf)
2611 return AVERROR(ENOMEM);
2612 if (codec_id == AV_CODEC_ID_RA_288) {
2613 st->codecpar->block_align = track->audio.coded_framesize;
2614 track->codec_priv.size = 0;
2616 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2617 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2618 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2619 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2621 st->codecpar->block_align = track->audio.sub_packet_size;
2622 extradata_offset = 78;
2624 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2625 ret = matroska_parse_flac(s, track, &extradata_offset);
2628 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2629 fourcc = AV_RL32(track->codec_priv.data);
2630 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2631 /* we don't need any value stored in CodecPrivate.
2632 make sure that it's not exported as extradata. */
2633 track->codec_priv.size = 0;
2634 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2635 /* For now, propagate only the OBUs, if any. Once libavcodec is
2636 updated to handle isobmff style extradata this can be removed. */
2637 extradata_offset = 4;
2639 track->codec_priv.size -= extradata_offset;
2641 if (codec_id == AV_CODEC_ID_NONE)
2642 av_log(matroska->ctx, AV_LOG_INFO,
2643 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2645 if (track->time_scale < 0.01)
2646 track->time_scale = 1.0;
2647 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2648 1000 * 1000 * 1000); /* 64 bit pts in ns */
2650 /* convert the delay from ns to the track timebase */
2651 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2652 (AVRational){ 1, 1000000000 },
2655 st->codecpar->codec_id = codec_id;
2657 if (strcmp(track->language, "und"))
2658 av_dict_set(&st->metadata, "language", track->language, 0);
2659 av_dict_set(&st->metadata, "title", track->name, 0);
2661 if (track->flag_default)
2662 st->disposition |= AV_DISPOSITION_DEFAULT;
2663 if (track->flag_forced)
2664 st->disposition |= AV_DISPOSITION_FORCED;
2666 if (!st->codecpar->extradata) {
2668 st->codecpar->extradata = extradata;
2669 st->codecpar->extradata_size = extradata_size;
2670 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2671 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2672 return AVERROR(ENOMEM);
2673 memcpy(st->codecpar->extradata,
2674 track->codec_priv.data + extradata_offset,
2675 track->codec_priv.size);
2679 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2680 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2681 int display_width_mul = 1;
2682 int display_height_mul = 1;
2684 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2685 st->codecpar->codec_tag = fourcc;
2687 st->codecpar->bits_per_coded_sample = bit_depth;
2688 st->codecpar->width = track->video.pixel_width;
2689 st->codecpar->height = track->video.pixel_height;
2691 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2692 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2693 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2694 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2696 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2697 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2699 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2700 av_reduce(&st->sample_aspect_ratio.num,
2701 &st->sample_aspect_ratio.den,
2702 st->codecpar->height * track->video.display_width * display_width_mul,
2703 st->codecpar->width * track->video.display_height * display_height_mul,
2706 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2707 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2709 if (track->default_duration) {
2710 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2711 1000000000, track->default_duration, 30000);
2712 #if FF_API_R_FRAME_RATE
2713 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2714 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2715 st->r_frame_rate = st->avg_frame_rate;
2719 /* export stereo mode flag as metadata tag */
2720 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2721 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2723 /* export alpha mode flag as metadata tag */
2724 if (track->video.alpha_mode)
2725 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2727 /* if we have virtual track, mark the real tracks */
2728 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2730 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2732 snprintf(buf, sizeof(buf), "%s_%d",
2733 ff_matroska_video_stereo_plane[planes[j].type], i);
2734 for (k=0; k < matroska->tracks.nb_elem; k++)
2735 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2736 av_dict_set(&tracks[k].stream->metadata,
2737 "stereo_mode", buf, 0);
2741 // add stream level stereo3d side data if it is a supported format
2742 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2743 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2744 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2749 ret = mkv_parse_video_color(st, track);
2752 ret = mkv_parse_video_projection(st, track);
2755 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2756 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2757 st->codecpar->codec_tag = fourcc;
2758 st->codecpar->sample_rate = track->audio.out_samplerate;
2759 st->codecpar->channels = track->audio.channels;
2760 if (!st->codecpar->bits_per_coded_sample)
2761 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2762 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2763 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2764 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2765 st->need_parsing = AVSTREAM_PARSE_FULL;
2766 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2767 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2768 if (track->codec_delay > 0) {
2769 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2770 (AVRational){1, 1000000000},
2771 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2772 48000 : st->codecpar->sample_rate});
2774 if (track->seek_preroll > 0) {
2775 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2776 (AVRational){1, 1000000000},
2777 (AVRational){1, st->codecpar->sample_rate});
2779 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2780 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2782 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2783 st->disposition |= AV_DISPOSITION_CAPTIONS;
2784 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2785 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2786 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2787 st->disposition |= AV_DISPOSITION_METADATA;
2789 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2790 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2797 static int matroska_read_header(AVFormatContext *s)
2799 MatroskaDemuxContext *matroska = s->priv_data;
2800 EbmlList *attachments_list = &matroska->attachments;
2801 EbmlList *chapters_list = &matroska->chapters;
2802 MatroskaAttachment *attachments;
2803 MatroskaChapter *chapters;
2804 uint64_t max_start = 0;
2810 matroska->cues_parsing_deferred = 1;
2812 /* First read the EBML header. */
2813 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2814 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2815 ebml_free(ebml_syntax, &ebml);
2816 return AVERROR_INVALIDDATA;
2818 if (ebml.version > EBML_VERSION ||
2819 ebml.max_size > sizeof(uint64_t) ||
2820 ebml.id_length > sizeof(uint32_t) ||
2821 ebml.doctype_version > 3) {
2822 avpriv_report_missing_feature(matroska->ctx,
2823 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2824 ebml.version, ebml.doctype, ebml.doctype_version);
2825 ebml_free(ebml_syntax, &ebml);
2826 return AVERROR_PATCHWELCOME;
2827 } else if (ebml.doctype_version == 3) {
2828 av_log(matroska->ctx, AV_LOG_WARNING,
2829 "EBML header using unsupported features\n"
2830 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2831 ebml.version, ebml.doctype, ebml.doctype_version);
2833 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2834 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2836 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2837 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2838 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2839 ebml_free(ebml_syntax, &ebml);
2840 return AVERROR_INVALIDDATA;
2843 ebml_free(ebml_syntax, &ebml);
2845 /* The next thing is a segment. */
2846 pos = avio_tell(matroska->ctx->pb);
2847 res = ebml_parse(matroska, matroska_segments, matroska);
2848 // Try resyncing until we find an EBML_STOP type element.
2850 res = matroska_resync(matroska, pos);
2853 pos = avio_tell(matroska->ctx->pb);
2854 res = ebml_parse(matroska, matroska_segment, matroska);
2856 /* Set data_offset as it might be needed later by seek_frame_generic. */
2857 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2858 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2859 matroska_execute_seekhead(matroska);
2861 if (!matroska->time_scale)
2862 matroska->time_scale = 1000000;
2863 if (matroska->duration)
2864 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2865 1000 / AV_TIME_BASE;
2866 av_dict_set(&s->metadata, "title", matroska->title, 0);
2867 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2869 if (matroska->date_utc.size == 8)
2870 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2872 res = matroska_parse_tracks(s);
2876 attachments = attachments_list->elem;
2877 for (j = 0; j < attachments_list->nb_elem; j++) {
2878 if (!(attachments[j].filename && attachments[j].mime &&
2879 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2880 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2882 AVStream *st = avformat_new_stream(s, NULL);
2885 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2886 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2887 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2889 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2890 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2891 strlen(ff_mkv_image_mime_tags[i].str))) {
2892 st->codecpar->codec_id = ff_mkv_image_mime_tags[i].id;
2897 attachments[j].stream = st;
2899 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2900 AVPacket *pkt = &st->attached_pic;
2902 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2903 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2905 av_init_packet(pkt);
2906 pkt->buf = av_buffer_ref(attachments[j].bin.buf);
2908 return AVERROR(ENOMEM);
2909 pkt->data = attachments[j].bin.data;
2910 pkt->size = attachments[j].bin.size;
2911 pkt->stream_index = st->index;
2912 pkt->flags |= AV_PKT_FLAG_KEY;
2914 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2915 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2917 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2918 attachments[j].bin.size);
2920 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2921 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2922 strlen(ff_mkv_mime_tags[i].str))) {
2923 st->codecpar->codec_id = ff_mkv_mime_tags[i].id;
2931 chapters = chapters_list->elem;
2932 for (i = 0; i < chapters_list->nb_elem; i++)
2933 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2934 (max_start == 0 || chapters[i].start > max_start)) {
2935 chapters[i].chapter =
2936 avpriv_new_chapter(s, chapters[i].uid,
2937 (AVRational) { 1, 1000000000 },
2938 chapters[i].start, chapters[i].end,
2940 if (chapters[i].chapter) {
2941 av_dict_set(&chapters[i].chapter->metadata,
2942 "title", chapters[i].title, 0);
2944 max_start = chapters[i].start;
2947 matroska_add_index_entries(matroska);
2949 matroska_convert_tags(s);
2953 matroska_read_close(s);
2958 * Put one packet in an application-supplied AVPacket struct.
2959 * Returns 0 on success or -1 on failure.
2961 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2964 if (matroska->queue) {
2965 MatroskaTrack *tracks = matroska->tracks.elem;
2966 MatroskaTrack *track;
2968 ff_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
2969 track = &tracks[pkt->stream_index];
2970 if (track->has_palette) {
2971 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2973 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
2975 memcpy(pal, track->palette, AVPALETTE_SIZE);
2977 track->has_palette = 0;
2986 * Free all packets in our internal queue.
2988 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2990 ff_packet_list_free(&matroska->queue, &matroska->queue_end);
2993 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2994 int *buf_size, int type,
2995 uint32_t **lace_buf, int *laces)
2997 int res = 0, n, size = *buf_size;
2998 uint8_t *data = *buf;
2999 uint32_t *lace_size;
3003 *lace_buf = av_malloc(sizeof(**lace_buf));
3005 return AVERROR(ENOMEM);
3007 *lace_buf[0] = size;
3011 av_assert0(size > 0);
3015 lace_size = av_malloc_array(*laces, sizeof(*lace_size));
3017 return AVERROR(ENOMEM);
3020 case 0x1: /* Xiph lacing */
3024 for (n = 0; res == 0 && n < *laces - 1; n++) {
3028 if (size <= total) {
3029 res = AVERROR_INVALIDDATA;
3034 lace_size[n] += temp;
3041 if (size <= total) {
3042 res = AVERROR_INVALIDDATA;
3046 lace_size[n] = size - total;
3050 case 0x2: /* fixed-size lacing */
3051 if (size % (*laces)) {
3052 res = AVERROR_INVALIDDATA;
3055 for (n = 0; n < *laces; n++)
3056 lace_size[n] = size / *laces;
3059 case 0x3: /* EBML lacing */
3063 n = matroska_ebmlnum_uint(matroska, data, size, &num);
3064 if (n < 0 || num > INT_MAX) {
3065 av_log(matroska->ctx, AV_LOG_INFO,
3066 "EBML block data error\n");
3067 res = n<0 ? n : AVERROR_INVALIDDATA;
3072 total = lace_size[0] = num;
3073 for (n = 1; res == 0 && n < *laces - 1; n++) {
3076 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
3077 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
3078 av_log(matroska->ctx, AV_LOG_INFO,
3079 "EBML block data error\n");
3080 res = r<0 ? r : AVERROR_INVALIDDATA;
3085 lace_size[n] = lace_size[n - 1] + snum;
3086 total += lace_size[n];
3088 if (size <= total) {
3089 res = AVERROR_INVALIDDATA;
3092 lace_size[*laces - 1] = size - total;
3098 *lace_buf = lace_size;
3104 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3105 MatroskaTrack *track, AVStream *st,
3106 uint8_t *data, int size, uint64_t timecode,
3109 int a = st->codecpar->block_align;
3110 int sps = track->audio.sub_packet_size;
3111 int cfs = track->audio.coded_framesize;
3112 int h = track->audio.sub_packet_h;
3113 int y = track->audio.sub_packet_cnt;
3114 int w = track->audio.frame_size;
3117 if (!track->audio.pkt_cnt) {
3118 if (track->audio.sub_packet_cnt == 0)
3119 track->audio.buf_timecode = timecode;
3120 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3121 if (size < cfs * h / 2) {
3122 av_log(matroska->ctx, AV_LOG_ERROR,
3123 "Corrupt int4 RM-style audio packet size\n");
3124 return AVERROR_INVALIDDATA;
3126 for (x = 0; x < h / 2; x++)
3127 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3128 data + x * cfs, cfs);
3129 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3131 av_log(matroska->ctx, AV_LOG_ERROR,
3132 "Corrupt sipr RM-style audio packet size\n");
3133 return AVERROR_INVALIDDATA;
3135 memcpy(track->audio.buf + y * w, data, w);
3137 if (size < sps * w / sps || h<=0 || w%sps) {
3138 av_log(matroska->ctx, AV_LOG_ERROR,
3139 "Corrupt generic RM-style audio packet size\n");
3140 return AVERROR_INVALIDDATA;
3142 for (x = 0; x < w / sps; x++)
3143 memcpy(track->audio.buf +
3144 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3145 data + x * sps, sps);
3148 if (++track->audio.sub_packet_cnt >= h) {
3149 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3150 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3151 track->audio.sub_packet_cnt = 0;
3152 track->audio.pkt_cnt = h * w / a;
3156 while (track->audio.pkt_cnt) {
3158 AVPacket pktl, *pkt = &pktl;
3160 ret = av_new_packet(pkt, a);
3165 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3167 pkt->pts = track->audio.buf_timecode;
3168 track->audio.buf_timecode = AV_NOPTS_VALUE;
3170 pkt->stream_index = st->index;
3171 ret = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3173 av_packet_unref(pkt);
3174 return AVERROR(ENOMEM);
3181 /* reconstruct full wavpack blocks from mangled matroska ones */
3182 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
3183 uint8_t **pdst, int *size)
3185 uint8_t *dst = NULL;
3190 int ret, offset = 0;
3192 if (srclen < 12 || track->stream->codecpar->extradata_size < 2)
3193 return AVERROR_INVALIDDATA;
3195 ver = AV_RL16(track->stream->codecpar->extradata);
3197 samples = AV_RL32(src);
3201 while (srclen >= 8) {
3206 uint32_t flags = AV_RL32(src);
3207 uint32_t crc = AV_RL32(src + 4);
3211 multiblock = (flags & 0x1800) != 0x1800;
3214 ret = AVERROR_INVALIDDATA;
3217 blocksize = AV_RL32(src);
3223 if (blocksize > srclen) {
3224 ret = AVERROR_INVALIDDATA;
3228 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3230 ret = AVERROR(ENOMEM);
3234 dstlen += blocksize + 32;
3236 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3237 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3238 AV_WL16(dst + offset + 8, ver); // version
3239 AV_WL16(dst + offset + 10, 0); // track/index_no
3240 AV_WL32(dst + offset + 12, 0); // total samples
3241 AV_WL32(dst + offset + 16, 0); // block index
3242 AV_WL32(dst + offset + 20, samples); // number of samples
3243 AV_WL32(dst + offset + 24, flags); // flags
3244 AV_WL32(dst + offset + 28, crc); // crc
3245 memcpy(dst + offset + 32, src, blocksize); // block data
3248 srclen -= blocksize;
3249 offset += blocksize + 32;
3252 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3264 static int matroska_parse_prores(MatroskaTrack *track, uint8_t *src,
3265 uint8_t **pdst, int *size)
3270 if (AV_RB32(&src[4]) != MKBETAG('i', 'c', 'p', 'f')) {
3271 dst = av_malloc(dstlen + 8 + AV_INPUT_BUFFER_PADDING_SIZE);
3273 return AVERROR(ENOMEM);
3275 AV_WB32(dst, dstlen);
3276 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3277 memcpy(dst + 8, src, dstlen);
3278 memset(dst + 8 + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3288 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3289 MatroskaTrack *track,
3291 uint8_t *data, int data_len,
3296 AVPacket pktl, *pkt = &pktl;
3297 uint8_t *id, *settings, *text, *buf;
3298 int id_len, settings_len, text_len;
3303 return AVERROR_INVALIDDATA;
3306 q = data + data_len;
3311 if (*p == '\r' || *p == '\n') {
3320 if (p >= q || *p != '\n')
3321 return AVERROR_INVALIDDATA;
3327 if (*p == '\r' || *p == '\n') {
3328 settings_len = p - settings;
3336 if (p >= q || *p != '\n')
3337 return AVERROR_INVALIDDATA;
3342 while (text_len > 0) {
3343 const int len = text_len - 1;
3344 const uint8_t c = p[len];
3345 if (c != '\r' && c != '\n')
3351 return AVERROR_INVALIDDATA;
3353 err = av_new_packet(pkt, text_len);
3358 memcpy(pkt->data, text, text_len);
3361 buf = av_packet_new_side_data(pkt,
3362 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3365 av_packet_unref(pkt);
3366 return AVERROR(ENOMEM);
3368 memcpy(buf, id, id_len);
3371 if (settings_len > 0) {
3372 buf = av_packet_new_side_data(pkt,
3373 AV_PKT_DATA_WEBVTT_SETTINGS,
3376 av_packet_unref(pkt);
3377 return AVERROR(ENOMEM);
3379 memcpy(buf, settings, settings_len);
3382 // Do we need this for subtitles?
3383 // pkt->flags = AV_PKT_FLAG_KEY;
3385 pkt->stream_index = st->index;
3386 pkt->pts = timecode;
3388 // Do we need this for subtitles?
3389 // pkt->dts = timecode;
3391 pkt->duration = duration;
3394 err = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3396 av_packet_unref(pkt);
3397 return AVERROR(ENOMEM);
3403 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3404 MatroskaTrack *track, AVStream *st,
3405 AVBufferRef *buf, uint8_t *data, int pkt_size,
3406 uint64_t timecode, uint64_t lace_duration,
3407 int64_t pos, int is_keyframe,
3408 uint8_t *additional, uint64_t additional_id, int additional_size,
3409 int64_t discard_padding)
3411 MatroskaTrackEncoding *encodings = track->encodings.elem;
3412 uint8_t *pkt_data = data;
3414 AVPacket pktl, *pkt = &pktl;
3416 if (encodings && !encodings->type && encodings->scope & 1) {
3417 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
3422 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3424 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
3426 av_log(matroska->ctx, AV_LOG_ERROR,
3427 "Error parsing a wavpack block.\n");
3430 if (pkt_data != data)
3431 av_freep(&pkt_data);
3435 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES) {
3437 res = matroska_parse_prores(track, pkt_data, &pr_data, &pkt_size);
3439 av_log(matroska->ctx, AV_LOG_ERROR,
3440 "Error parsing a prores block.\n");
3443 if (pkt_data != data)
3444 av_freep(&pkt_data);
3448 av_init_packet(pkt);
3449 if (pkt_data != data)
3450 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3453 pkt->buf = av_buffer_ref(buf);
3456 res = AVERROR(ENOMEM);
3460 pkt->data = pkt_data;
3461 pkt->size = pkt_size;
3462 pkt->flags = is_keyframe;
3463 pkt->stream_index = st->index;
3465 if (additional_size > 0) {
3466 uint8_t *side_data = av_packet_new_side_data(pkt,
3467 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3468 additional_size + 8);
3470 av_packet_unref(pkt);
3471 return AVERROR(ENOMEM);
3473 AV_WB64(side_data, additional_id);
3474 memcpy(side_data + 8, additional, additional_size);
3477 if (discard_padding) {
3478 uint8_t *side_data = av_packet_new_side_data(pkt,
3479 AV_PKT_DATA_SKIP_SAMPLES,
3482 av_packet_unref(pkt);
3483 return AVERROR(ENOMEM);
3485 discard_padding = av_rescale_q(discard_padding,
3486 (AVRational){1, 1000000000},
3487 (AVRational){1, st->codecpar->sample_rate});
3488 if (discard_padding > 0) {
3489 AV_WL32(side_data + 4, discard_padding);
3491 AV_WL32(side_data, -discard_padding);
3495 if (track->ms_compat)
3496 pkt->dts = timecode;
3498 pkt->pts = timecode;
3500 pkt->duration = lace_duration;
3502 #if FF_API_CONVERGENCE_DURATION
3503 FF_DISABLE_DEPRECATION_WARNINGS
3504 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3505 pkt->convergence_duration = lace_duration;
3507 FF_ENABLE_DEPRECATION_WARNINGS
3510 res = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3512 av_packet_unref(pkt);
3513 return AVERROR(ENOMEM);
3519 if (pkt_data != data)
3520 av_freep(&pkt_data);
3524 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3525 int size, int64_t pos, uint64_t cluster_time,
3526 uint64_t block_duration, int is_keyframe,
3527 uint8_t *additional, uint64_t additional_id, int additional_size,
3528 int64_t cluster_pos, int64_t discard_padding)
3530 uint64_t timecode = AV_NOPTS_VALUE;
3531 MatroskaTrack *track;
3535 uint32_t *lace_size = NULL;
3536 int n, flags, laces = 0;
3538 int trust_default_duration = 1;
3540 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
3546 track = matroska_find_track_by_num(matroska, num);
3547 if (!track || !track->stream) {
3548 av_log(matroska->ctx, AV_LOG_INFO,
3549 "Invalid stream %"PRIu64"\n", num);
3550 return AVERROR_INVALIDDATA;
3551 } else if (size <= 3)
3554 if (st->discard >= AVDISCARD_ALL)
3556 av_assert1(block_duration != AV_NOPTS_VALUE);
3558 block_time = sign_extend(AV_RB16(data), 16);
3562 if (is_keyframe == -1)
3563 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3565 if (cluster_time != (uint64_t) -1 &&
3566 (block_time >= 0 || cluster_time >= -block_time)) {
3567 timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3568 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3569 timecode < track->end_timecode)
3570 is_keyframe = 0; /* overlapping subtitles are not key frame */
3572 ff_reduce_index(matroska->ctx, st->index);
3573 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3578 if (matroska->skip_to_keyframe &&
3579 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3580 // Compare signed timecodes. Timecode may be negative due to codec delay
3581 // offset. We don't support timestamps greater than int64_t anyway - see
3583 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3586 matroska->skip_to_keyframe = 0;
3587 else if (!st->skip_to_keyframe) {
3588 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3589 matroska->skip_to_keyframe = 0;
3593 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
3594 &lace_size, &laces);
3599 if (track->audio.samplerate == 8000) {
3600 // If this is needed for more codecs, then add them here
3601 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3602 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3603 trust_default_duration = 0;
3607 if (!block_duration && trust_default_duration)
3608 block_duration = track->default_duration * laces / matroska->time_scale;
3610 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3611 track->end_timecode =
3612 FFMAX(track->end_timecode, timecode + block_duration);
3614 for (n = 0; n < laces; n++) {
3615 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3617 if (lace_size[n] > size) {
3618 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
3622 if ((st->codecpar->codec_id == AV_CODEC_ID_RA_288 ||
3623 st->codecpar->codec_id == AV_CODEC_ID_COOK ||
3624 st->codecpar->codec_id == AV_CODEC_ID_SIPR ||
3625 st->codecpar->codec_id == AV_CODEC_ID_ATRAC3) &&
3626 st->codecpar->block_align && track->audio.sub_packet_size) {
3627 res = matroska_parse_rm_audio(matroska, track, st, data,
3633 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3634 res = matroska_parse_webvtt(matroska, track, st,
3636 timecode, lace_duration,
3641 res = matroska_parse_frame(matroska, track, st, buf, data, lace_size[n],
3642 timecode, lace_duration, pos,
3643 !n ? is_keyframe : 0,
3644 additional, additional_id, additional_size,
3650 if (timecode != AV_NOPTS_VALUE)
3651 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3652 data += lace_size[n];
3653 size -= lace_size[n];
3661 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3663 MatroskaCluster *cluster = &matroska->current_cluster;
3664 MatroskaBlock *block = &cluster->block;
3667 av_assert0(matroska->num_levels <= 2);
3669 if (matroska->num_levels == 1) {
3670 res = ebml_parse(matroska, matroska_segment, NULL);
3673 /* Found a cluster: subtract the size of the ID already read. */
3674 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3676 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3682 if (matroska->num_levels == 2) {
3683 /* We are inside a cluster. */
3684 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3686 if (res >= 0 && block->bin.size > 0) {
3687 int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3688 uint8_t* additional = block->additional.size > 0 ?
3689 block->additional.data : NULL;
3691 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3692 block->bin.size, block->bin.pos,
3693 cluster->timecode, block->duration,
3694 is_keyframe, additional, block->additional_id,
3695 block->additional.size, cluster->pos,
3696 block->discard_padding);
3699 ebml_free(matroska_blockgroup, block);
3700 memset(block, 0, sizeof(*block));
3701 } else if (!matroska->num_levels) {
3702 if (!avio_feof(matroska->ctx->pb)) {
3703 avio_r8(matroska->ctx->pb);
3704 if (!avio_feof(matroska->ctx->pb)) {
3705 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3706 "end of segment.\n");
3707 return AVERROR_INVALIDDATA;
3717 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3719 MatroskaDemuxContext *matroska = s->priv_data;
3722 if (matroska->resync_pos == -1) {
3723 // This can only happen if generic seeking has been used.
3724 matroska->resync_pos = avio_tell(s->pb);
3727 while (matroska_deliver_packet(matroska, pkt)) {
3729 return (ret < 0) ? ret : AVERROR_EOF;
3730 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3731 ret = matroska_resync(matroska, matroska->resync_pos);
3737 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3738 int64_t timestamp, int flags)
3740 MatroskaDemuxContext *matroska = s->priv_data;
3741 MatroskaTrack *tracks = NULL;
3742 AVStream *st = s->streams[stream_index];
3745 /* Parse the CUES now since we need the index data to seek. */
3746 if (matroska->cues_parsing_deferred > 0) {
3747 matroska->cues_parsing_deferred = 0;
3748 matroska_parse_cues(matroska);
3751 if (!st->nb_index_entries)
3753 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3755 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3756 matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3757 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3758 matroska_clear_queue(matroska);
3759 if (matroska_parse_cluster(matroska) < 0)
3764 matroska_clear_queue(matroska);
3765 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3768 tracks = matroska->tracks.elem;
3769 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3770 tracks[i].audio.pkt_cnt = 0;
3771 tracks[i].audio.sub_packet_cnt = 0;
3772 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3773 tracks[i].end_timecode = 0;
3776 /* We seek to a level 1 element, so set the appropriate status. */
3777 matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3778 if (flags & AVSEEK_FLAG_ANY) {
3779 st->skip_to_keyframe = 0;
3780 matroska->skip_to_timecode = timestamp;
3782 st->skip_to_keyframe = 1;
3783 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3785 matroska->skip_to_keyframe = 1;
3787 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3790 // slightly hackish but allows proper fallback to
3791 // the generic seeking code.
3792 matroska_reset_status(matroska, 0, -1);
3793 matroska->resync_pos = -1;
3794 matroska_clear_queue(matroska);
3795 st->skip_to_keyframe =
3796 matroska->skip_to_keyframe = 0;
3801 static int matroska_read_close(AVFormatContext *s)
3803 MatroskaDemuxContext *matroska = s->priv_data;
3804 MatroskaTrack *tracks = matroska->tracks.elem;
3807 matroska_clear_queue(matroska);
3809 for (n = 0; n < matroska->tracks.nb_elem; n++)
3810 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3811 av_freep(&tracks[n].audio.buf);
3812 ebml_free(matroska_segment, matroska);
3818 int64_t start_time_ns;
3819 int64_t end_time_ns;
3820 int64_t start_offset;
3824 /* This function searches all the Cues and returns the CueDesc corresponding to
3825 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3826 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3828 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3829 MatroskaDemuxContext *matroska = s->priv_data;
3832 int nb_index_entries = s->streams[0]->nb_index_entries;
3833 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3834 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3835 for (i = 1; i < nb_index_entries; i++) {
3836 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3837 index_entries[i].timestamp * matroska->time_scale > ts) {
3842 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3843 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3844 if (i != nb_index_entries - 1) {
3845 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3846 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3848 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3849 // FIXME: this needs special handling for files where Cues appear
3850 // before Clusters. the current logic assumes Cues appear after
3852 cue_desc.end_offset = cues_start - matroska->segment_start;
3857 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3859 MatroskaDemuxContext *matroska = s->priv_data;
3860 uint32_t id = matroska->current_id;
3861 int64_t cluster_pos, before_pos;
3863 if (s->streams[0]->nb_index_entries <= 0) return 0;
3864 // seek to the first cluster using cues.
3865 index = av_index_search_timestamp(s->streams[0], 0, 0);
3866 if (index < 0) return 0;
3867 cluster_pos = s->streams[0]->index_entries[index].pos;
3868 before_pos = avio_tell(s->pb);
3870 uint64_t cluster_id, cluster_length;
3873 avio_seek(s->pb, cluster_pos, SEEK_SET);
3874 // read cluster id and length
3875 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3876 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3878 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3882 matroska_reset_status(matroska, 0, cluster_pos);
3883 matroska_clear_queue(matroska);
3884 if (matroska_parse_cluster(matroska) < 0 ||
3888 pkt = &matroska->queue->pkt;
3889 // 4 + read is the length of the cluster id and the cluster length field.
3890 cluster_pos += 4 + read + cluster_length;
3891 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3897 /* Restore the status after matroska_read_header: */
3898 matroska_reset_status(matroska, id, before_pos);
3903 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3904 double min_buffer, double* buffer,
3905 double* sec_to_download, AVFormatContext *s,
3908 double nano_seconds_per_second = 1000000000.0;
3909 double time_sec = time_ns / nano_seconds_per_second;
3911 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3912 int64_t end_time_ns = time_ns + time_to_search_ns;
3913 double sec_downloaded = 0.0;
3914 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3915 if (desc_curr.start_time_ns == -1)
3917 *sec_to_download = 0.0;
3919 // Check for non cue start time.
3920 if (time_ns > desc_curr.start_time_ns) {
3921 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3922 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3923 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3924 double timeToDownload = (cueBytes * 8.0) / bps;
3926 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3927 *sec_to_download += timeToDownload;
3929 // Check if the search ends within the first cue.
3930 if (desc_curr.end_time_ns >= end_time_ns) {
3931 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3932 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3933 sec_downloaded = percent_to_sub * sec_downloaded;
3934 *sec_to_download = percent_to_sub * *sec_to_download;
3937 if ((sec_downloaded + *buffer) <= min_buffer) {
3941 // Get the next Cue.
3942 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3945 while (desc_curr.start_time_ns != -1) {
3946 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3947 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3948 double desc_sec = desc_ns / nano_seconds_per_second;
3949 double bits = (desc_bytes * 8.0);
3950 double time_to_download = bits / bps;
3952 sec_downloaded += desc_sec - time_to_download;
3953 *sec_to_download += time_to_download;
3955 if (desc_curr.end_time_ns >= end_time_ns) {
3956 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3957 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3958 sec_downloaded = percent_to_sub * sec_downloaded;
3959 *sec_to_download = percent_to_sub * *sec_to_download;
3961 if ((sec_downloaded + *buffer) <= min_buffer)
3966 if ((sec_downloaded + *buffer) <= min_buffer) {
3971 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3973 *buffer = *buffer + sec_downloaded;
3977 /* This function computes the bandwidth of the WebM file with the help of
3978 * buffer_size_after_time_downloaded() function. Both of these functions are
3979 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3980 * Matroska parsing mechanism.
3982 * Returns the bandwidth of the file on success; -1 on error.
3984 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3986 MatroskaDemuxContext *matroska = s->priv_data;
3987 AVStream *st = s->streams[0];
3988 double bandwidth = 0.0;
3991 for (i = 0; i < st->nb_index_entries; i++) {
3992 int64_t prebuffer_ns = 1000000000;
3993 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3994 double nano_seconds_per_second = 1000000000.0;
3995 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3996 double prebuffer_bytes = 0.0;
3997 int64_t temp_prebuffer_ns = prebuffer_ns;
3998 int64_t pre_bytes, pre_ns;
3999 double pre_sec, prebuffer, bits_per_second;
4000 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4002 // Start with the first Cue.
4003 CueDesc desc_end = desc_beg;
4005 // Figure out how much data we have downloaded for the prebuffer. This will
4006 // be used later to adjust the bits per sample to try.
4007 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4008 // Prebuffered the entire Cue.
4009 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4010 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4011 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4013 if (desc_end.start_time_ns == -1) {
4014 // The prebuffer is larger than the duration.
4015 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4017 bits_per_second = 0.0;
4019 // The prebuffer ends in the last Cue. Estimate how much data was
4021 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4022 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4023 pre_sec = pre_ns / nano_seconds_per_second;
4025 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4027 prebuffer = prebuffer_ns / nano_seconds_per_second;
4029 // Set this to 0.0 in case our prebuffer buffers the entire video.
4030 bits_per_second = 0.0;
4032 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4033 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4034 double desc_sec = desc_ns / nano_seconds_per_second;
4035 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4037 // Drop the bps by the percentage of bytes buffered.
4038 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4039 double mod_bits_per_second = calc_bits_per_second * percent;
4041 if (prebuffer < desc_sec) {
4043 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4045 // Add 1 so the bits per second should be a little bit greater than file
4047 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4048 const double min_buffer = 0.0;
4049 double buffer = prebuffer;
4050 double sec_to_download = 0.0;
4052 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4053 min_buffer, &buffer, &sec_to_download,
4057 } else if (rv == 0) {
4058 bits_per_second = (double)(bps);
4063 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4064 } while (desc_end.start_time_ns != -1);
4066 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4068 return (int64_t)bandwidth;
4071 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4073 MatroskaDemuxContext *matroska = s->priv_data;
4074 EbmlList *seekhead_list = &matroska->seekhead;
4075 MatroskaSeekhead *seekhead = seekhead_list->elem;
4077 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4081 // determine cues start and end positions
4082 for (i = 0; i < seekhead_list->nb_elem; i++)
4083 if (seekhead[i].id == MATROSKA_ID_CUES)
4086 if (i >= seekhead_list->nb_elem) return -1;
4088 before_pos = avio_tell(matroska->ctx->pb);
4089 cues_start = seekhead[i].pos + matroska->segment_start;
4090 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4091 // cues_end is computed as cues_start + cues_length + length of the
4092 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4093 // cues_end is inclusive and the above sum is reduced by 1.
4094 uint64_t cues_length, cues_id;
4096 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4097 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4098 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4099 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4102 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4104 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4105 if (cues_start == -1 || cues_end == -1) return -1;
4108 matroska_parse_cues(matroska);
4111 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4114 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4116 // if the file has cues at the start, fix up the init range so that
4117 // it does not include it
4118 if (cues_start <= init_range)
4119 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4122 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4123 if (bandwidth < 0) return -1;
4124 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4126 // check if all clusters start with key frames
4127 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4129 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4130 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4131 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4132 if (!buf) return -1;
4134 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4135 int ret = snprintf(buf + end, 20,
4136 "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4137 i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4138 if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4139 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4141 return AVERROR_INVALIDDATA;
4145 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
4151 static int webm_dash_manifest_read_header(AVFormatContext *s)
4154 int ret = matroska_read_header(s);
4156 MatroskaTrack *tracks;
4157 MatroskaDemuxContext *matroska = s->priv_data;
4159 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4162 if (!s->nb_streams) {
4163 matroska_read_close(s);
4164 av_log(s, AV_LOG_ERROR, "No streams found\n");
4165 return AVERROR_INVALIDDATA;
4168 if (!matroska->is_live) {
4169 buf = av_asprintf("%g", matroska->duration);
4170 if (!buf) return AVERROR(ENOMEM);
4171 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
4174 // initialization range
4175 // 5 is the offset of Cluster ID.
4176 init_range = avio_tell(s->pb) - 5;
4177 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4180 // basename of the file
4181 buf = strrchr(s->url, '/');
4182 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4185 tracks = matroska->tracks.elem;
4186 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4188 // parse the cues and populate Cue related fields
4189 if (!matroska->is_live) {
4190 ret = webm_dash_manifest_cues(s, init_range);
4192 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4197 // use the bandwidth from the command line if it was provided
4198 if (matroska->bandwidth > 0) {
4199 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4200 matroska->bandwidth, 0);
4205 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4210 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4211 static const AVOption options[] = {
4212 { "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 },
4213 { "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 },
4217 static const AVClass webm_dash_class = {
4218 .class_name = "WebM DASH Manifest demuxer",
4219 .item_name = av_default_item_name,
4221 .version = LIBAVUTIL_VERSION_INT,
4224 AVInputFormat ff_matroska_demuxer = {
4225 .name = "matroska,webm",
4226 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4227 .extensions = "mkv,mk3d,mka,mks",
4228 .priv_data_size = sizeof(MatroskaDemuxContext),
4229 .read_probe = matroska_probe,
4230 .read_header = matroska_read_header,
4231 .read_packet = matroska_read_packet,
4232 .read_close = matroska_read_close,
4233 .read_seek = matroska_read_seek,
4234 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4237 AVInputFormat ff_webm_dash_manifest_demuxer = {
4238 .name = "webm_dash_manifest",
4239 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4240 .priv_data_size = sizeof(MatroskaDemuxContext),
4241 .read_header = webm_dash_manifest_read_header,
4242 .read_packet = webm_dash_manifest_read_packet,
4243 .read_close = matroska_read_close,
4244 .priv_class = &webm_dash_class,