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_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
689 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
690 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
691 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
692 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
693 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
694 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
695 { MATROSKA_ID_CLUSTER, EBML_STOP },
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] } },
744 static const EbmlSyntax matroska_clusters[] = {
745 { MATROSKA_ID_CLUSTER, EBML_STOP },
746 { MATROSKA_ID_CUES, EBML_NONE },
747 { MATROSKA_ID_TAGS, EBML_NONE },
748 { MATROSKA_ID_INFO, EBML_NONE },
749 { MATROSKA_ID_TRACKS, EBML_NONE },
750 { MATROSKA_ID_ATTACHMENTS, EBML_NONE },
751 { MATROSKA_ID_CHAPTERS, EBML_NONE },
752 { MATROSKA_ID_SEEKHEAD, EBML_NONE },
753 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
757 static const char *const matroska_doctypes[] = { "matroska", "webm" };
759 static int matroska_read_close(AVFormatContext *s);
762 * This function prepares the status for parsing of level 1 elements.
764 static int matroska_reset_status(MatroskaDemuxContext *matroska,
765 uint32_t id, int64_t position)
768 int err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
773 matroska->current_id = id;
774 matroska->num_levels = 1;
775 matroska->unknown_count = 0;
776 matroska->resync_pos = avio_tell(matroska->ctx->pb);
778 matroska->resync_pos -= (av_log2(id) + 7) / 8;
783 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
785 AVIOContext *pb = matroska->ctx->pb;
788 /* Try to seek to the last position to resync from. If this doesn't work,
789 * we resync from the earliest position available: The start of the buffer. */
790 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
791 av_log(matroska->ctx, AV_LOG_WARNING,
792 "Seek to desired resync point failed. Seeking to "
793 "earliest point available instead.\n");
794 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
795 last_pos + 1), SEEK_SET);
800 // try to find a toplevel element
801 while (!avio_feof(pb)) {
802 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
803 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
804 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
805 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
806 /* Prepare the context for parsing of a level 1 element. */
807 matroska_reset_status(matroska, id, -1);
808 /* Given that we are here means that an error has occured,
809 * so treat the segment as unknown length in order not to
810 * discard valid data that happens to be beyond the designated
811 * end of the segment. */
812 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
815 id = (id << 8) | avio_r8(pb);
819 return pb->error ? pb->error : AVERROR_EOF;
823 * Read: an "EBML number", which is defined as a variable-length
824 * array of bytes. The first byte indicates the length by giving a
825 * number of 0-bits followed by a one. The position of the first
826 * "one" bit inside the first byte indicates the length of this
828 * Returns: number of bytes read, < 0 on error
830 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
831 int max_size, uint64_t *number, int eof_forbidden)
837 /* The first byte tells us the length in bytes - except when it is zero. */
842 /* get the length of the EBML number */
843 read = 8 - ff_log2_tab[total];
845 if (!total || read > max_size) {
846 pos = avio_tell(pb) - 1;
848 av_log(matroska->ctx, AV_LOG_ERROR,
849 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
850 "of an EBML number\n", pos, pos);
852 av_log(matroska->ctx, AV_LOG_ERROR,
853 "Length %d indicated by an EBML number's first byte 0x%02x "
854 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
855 read, (uint8_t) total, pos, pos, max_size);
857 return AVERROR_INVALIDDATA;
860 /* read out length */
861 total ^= 1 << ff_log2_tab[total];
863 total = (total << 8) | avio_r8(pb);
865 if (pb->eof_reached) {
877 av_log(matroska->ctx, AV_LOG_ERROR,
878 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
883 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
884 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
891 * Read a EBML length value.
892 * This needs special handling for the "unknown length" case which has multiple
895 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
898 int res = ebml_read_num(matroska, pb, 8, number, 1);
899 if (res > 0 && *number + 1 == 1ULL << (7 * res))
900 *number = EBML_UNKNOWN_LENGTH;
905 * Read the next element as an unsigned int.
906 * Returns NEEDS_CHECKING.
908 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
912 /* big-endian ordering; build up number */
915 *num = (*num << 8) | avio_r8(pb);
917 return NEEDS_CHECKING;
921 * Read the next element as a signed int.
922 * Returns NEEDS_CHECKING.
924 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
931 *num = sign_extend(avio_r8(pb), 8);
933 /* big-endian ordering; build up number */
935 *num = ((uint64_t)*num << 8) | avio_r8(pb);
938 return NEEDS_CHECKING;
942 * Read the next element as a float.
943 * Returns NEEDS_CHECKING or < 0 on obvious failure.
945 static int ebml_read_float(AVIOContext *pb, int size, double *num)
950 *num = av_int2float(avio_rb32(pb));
952 *num = av_int2double(avio_rb64(pb));
954 return AVERROR_INVALIDDATA;
956 return NEEDS_CHECKING;
960 * Read the next element as an ASCII string.
961 * 0 is success, < 0 or NEEDS_CHECKING is failure.
963 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
968 /* EBML strings are usually not 0-terminated, so we allocate one
969 * byte more, read the string and NULL-terminate it ourselves. */
970 if (!(res = av_malloc(size + 1)))
971 return AVERROR(ENOMEM);
972 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
974 return ret < 0 ? ret : NEEDS_CHECKING;
984 * Read the next element as binary data.
985 * 0 is success, < 0 or NEEDS_CHECKING is failure.
987 static int ebml_read_binary(AVIOContext *pb, int length,
988 int64_t pos, EbmlBin *bin)
992 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
995 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
997 bin->data = bin->buf->data;
1000 if ((ret = avio_read(pb, bin->data, length)) != length) {
1001 av_buffer_unref(&bin->buf);
1004 return ret < 0 ? ret : NEEDS_CHECKING;
1011 * Read the next element, but only the header. The contents
1012 * are supposed to be sub-elements which can be read separately.
1013 * 0 is success, < 0 is failure.
1015 static int ebml_read_master(MatroskaDemuxContext *matroska,
1016 uint64_t length, int64_t pos)
1018 MatroskaLevel *level;
1020 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1021 av_log(matroska->ctx, AV_LOG_ERROR,
1022 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1023 return AVERROR(ENOSYS);
1026 level = &matroska->levels[matroska->num_levels++];
1028 level->length = length;
1034 * Read signed/unsigned "EBML" numbers.
1035 * Return: number of bytes processed, < 0 on error
1037 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
1038 uint8_t *data, uint32_t size, uint64_t *num)
1041 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
1042 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num, 1);
1046 * Same as above, but signed.
1048 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1049 uint8_t *data, uint32_t size, int64_t *num)
1054 /* read as unsigned number first */
1055 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
1058 /* make signed (weird way) */
1059 *num = unum - ((1LL << (7 * res - 1)) - 1);
1064 static int ebml_parse(MatroskaDemuxContext *matroska,
1065 EbmlSyntax *syntax, void *data);
1067 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1071 // Whoever touches this should be aware of the duplication
1072 // existing in matroska_cluster_parsing.
1073 for (i = 0; syntax[i].id; i++)
1074 if (id == syntax[i].id)
1080 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1085 for (i = 0; syntax[i].id; i++)
1086 switch (syntax[i].type) {
1088 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1091 *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1094 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1098 // the default may be NULL
1099 if (syntax[i].def.s) {
1100 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1101 *dst = av_strdup(syntax[i].def.s);
1103 return AVERROR(ENOMEM);
1108 if (!matroska->levels[matroska->num_levels - 1].length) {
1109 matroska->num_levels--;
1114 res = ebml_parse(matroska, syntax, data);
1117 return res == LEVEL_ENDED ? 0 : res;
1120 static int is_ebml_id_valid(uint32_t id)
1122 // Due to endian nonsense in Matroska, the highest byte with any bits set
1123 // will contain the leading length bit. This bit in turn identifies the
1124 // total byte length of the element by its position within the byte.
1125 unsigned int bits = av_log2(id);
1126 return id && (bits + 7) / 8 == (8 - bits % 8);
1130 * Allocate and return the entry for the level1 element with the given ID. If
1131 * an entry already exists, return the existing entry.
1133 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1137 MatroskaLevel1Element *elem;
1139 if (!is_ebml_id_valid(id))
1142 // Some files link to all clusters; useless.
1143 if (id == MATROSKA_ID_CLUSTER)
1146 // There can be multiple seekheads.
1147 if (id != MATROSKA_ID_SEEKHEAD) {
1148 for (i = 0; i < matroska->num_level1_elems; i++) {
1149 if (matroska->level1_elems[i].id == id)
1150 return &matroska->level1_elems[i];
1154 // Only a completely broken file would have more elements.
1155 // It also provides a low-effort way to escape from circular seekheads
1156 // (every iteration will add a level1 entry).
1157 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1158 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1162 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1163 *elem = (MatroskaLevel1Element){.id = id};
1168 static int ebml_parse(MatroskaDemuxContext *matroska,
1169 EbmlSyntax *syntax, void *data)
1171 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1172 // Forbid unknown-length EBML_NONE elements.
1173 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1177 // max. 16 MB for strings
1178 [EBML_STR] = 0x1000000,
1179 [EBML_UTF8] = 0x1000000,
1180 // max. 256 MB for binary data
1181 [EBML_BIN] = 0x10000000,
1182 // no limits for anything else
1184 AVIOContext *pb = matroska->ctx->pb;
1187 int64_t pos = avio_tell(pb), pos_alt;
1188 int res, update_pos = 1, level_check;
1189 MatroskaLevel1Element *level1_elem;
1190 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1192 if (!matroska->current_id) {
1194 res = ebml_read_num(matroska, pb, 4, &id, 0);
1196 if (pb->eof_reached && res == AVERROR_EOF) {
1197 if (matroska->is_live)
1198 // in live mode, finish parsing if EOF is reached.
1200 if (level && pos == avio_tell(pb)) {
1201 if (level->length == EBML_UNKNOWN_LENGTH) {
1202 // Unknown-length levels automatically end at EOF.
1203 matroska->num_levels--;
1206 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1207 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1213 matroska->current_id = id | 1 << 7 * res;
1214 pos_alt = pos + res;
1217 pos -= (av_log2(matroska->current_id) + 7) / 8;
1220 id = matroska->current_id;
1222 syntax = ebml_parse_id(syntax, id);
1223 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1224 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1225 // Unknown-length levels end when an element from an upper level
1226 // in the hierarchy is encountered.
1227 while (syntax->def.n) {
1228 syntax = ebml_parse_id(syntax->def.n, id);
1230 matroska->num_levels--;
1236 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1237 "%"PRId64"\n", id, pos);
1238 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1241 data = (char *) data + syntax->data_offset;
1242 if (syntax->list_elem_size) {
1243 EbmlList *list = data;
1244 void *newelem = av_realloc_array(list->elem, list->nb_elem + 1,
1245 syntax->list_elem_size);
1247 return AVERROR(ENOMEM);
1248 list->elem = newelem;
1249 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1250 memset(data, 0, syntax->list_elem_size);
1254 if (syntax->type != EBML_STOP) {
1255 matroska->current_id = 0;
1256 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1261 if (matroska->num_levels > 0) {
1262 if (length != EBML_UNKNOWN_LENGTH &&
1263 level->length != EBML_UNKNOWN_LENGTH) {
1264 uint64_t elem_end = pos_alt + length,
1265 level_end = level->start + level->length;
1267 if (elem_end < level_end) {
1269 } else if (elem_end == level_end) {
1270 level_check = LEVEL_ENDED;
1272 av_log(matroska->ctx, AV_LOG_ERROR,
1273 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1274 "containing master element ending at 0x%"PRIx64"\n",
1275 pos, elem_end, level_end);
1276 return AVERROR_INVALIDDATA;
1278 } else if (length != EBML_UNKNOWN_LENGTH) {
1280 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1281 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1282 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1283 return AVERROR_INVALIDDATA;
1284 } else if (id != MATROSKA_ID_CLUSTER) {
1285 // According to the specifications only clusters and segments
1286 // are allowed to be unknown-sized.
1287 av_log(matroska->ctx, AV_LOG_ERROR,
1288 "Found unknown-sized element other than a cluster at "
1289 "0x%"PRIx64". Dropping the invalid element.\n", pos);
1290 return AVERROR_INVALIDDATA;
1296 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1297 if (length != EBML_UNKNOWN_LENGTH) {
1298 av_log(matroska->ctx, AV_LOG_ERROR,
1299 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1300 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1301 length, max_lengths[syntax->type], id, pos);
1302 } else if (syntax->type != EBML_NONE) {
1303 av_log(matroska->ctx, AV_LOG_ERROR,
1304 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1305 "unknown length, yet the length of an element of its "
1306 "type must be known.\n", id, pos);
1308 av_log(matroska->ctx, AV_LOG_ERROR,
1309 "Found unknown-length element with ID 0x%"PRIX32" at "
1310 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1311 "available.\n", id, pos);
1313 return AVERROR_INVALIDDATA;
1316 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1317 // Loosing sync will likely manifest itself as encountering unknown
1318 // elements which are not reliably distinguishable from elements
1319 // belonging to future extensions of the format.
1320 // We use a heuristic to detect such situations: If the current
1321 // element is not expected at the current syntax level and there
1322 // were only a few unknown elements in a row, then the element is
1323 // skipped or considered defective based upon the length of the
1324 // current element (i.e. how much would be skipped); if there were
1325 // more than a few skipped elements in a row and skipping the current
1326 // element would lead us more than SKIP_THRESHOLD away from the last
1327 // known good position, then it is inferred that an error occured.
1328 // The dependency on the number of unknown elements in a row exists
1329 // because the distance to the last known good position is
1330 // automatically big if the last parsed element was big.
1331 // In both cases, each unknown element is considered equivalent to
1332 // UNKNOWN_EQUIV of skipped bytes for the check.
1333 // The whole check is only done for non-seekable output, because
1334 // in this situation skipped data can't simply be rechecked later.
1335 // This is especially important when using unkown length elements
1336 // as the check for whether a child exceeds its containing master
1337 // element is not effective in this situation.
1339 matroska->unknown_count = 0;
1341 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1343 if (matroska->unknown_count > 3)
1344 dist += pos_alt - matroska->resync_pos;
1346 if (dist > SKIP_THRESHOLD) {
1347 av_log(matroska->ctx, AV_LOG_ERROR,
1348 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1349 "length 0x%"PRIx64" considered as invalid data. Last "
1350 "known good position 0x%"PRIx64", %d unknown elements"
1351 " in a row\n", id, pos, length, matroska->resync_pos,
1352 matroska->unknown_count);
1353 return AVERROR_INVALIDDATA;
1359 // We have found an element that is allowed at this place
1360 // in the hierarchy and it passed all checks, so treat the beginning
1361 // of the element as the "last known good" position.
1362 matroska->resync_pos = pos;
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))
1405 if (ffio_limit(pb, length) != length) {
1406 // ffio_limit emits its own error message,
1407 // so we don't have to.
1408 return AVERROR(EIO);
1410 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1411 // avio_skip might take us past EOF. We check for this
1412 // by skipping only length - 1 bytes, reading a byte and
1413 // checking the error flags. This is done in order to check
1414 // that the element has been properly skipped even when
1415 // no filesize (that ffio_limit relies on) is available.
1417 res = NEEDS_CHECKING;
1424 if (res == NEEDS_CHECKING) {
1425 if (pb->eof_reached) {
1434 if (res == AVERROR_INVALIDDATA)
1435 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1436 else if (res == AVERROR(EIO))
1437 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1438 else if (res == AVERROR_EOF) {
1439 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1447 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1448 level = &matroska->levels[matroska->num_levels - 1];
1449 pos = avio_tell(pb);
1451 // Given that pos >= level->start no check for
1452 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1453 while (matroska->num_levels && pos == level->start + level->length) {
1454 matroska->num_levels--;
1462 static void ebml_free(EbmlSyntax *syntax, void *data)
1465 for (i = 0; syntax[i].id; i++) {
1466 void *data_off = (char *) data + syntax[i].data_offset;
1467 switch (syntax[i].type) {
1473 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1477 if (syntax[i].list_elem_size) {
1478 EbmlList *list = data_off;
1479 char *ptr = list->elem;
1480 for (j = 0; j < list->nb_elem;
1481 j++, ptr += syntax[i].list_elem_size)
1482 ebml_free(syntax[i].def.n, ptr);
1483 av_freep(&list->elem);
1486 ebml_free(syntax[i].def.n, data_off);
1496 static int matroska_probe(const AVProbeData *p)
1499 int len_mask = 0x80, size = 1, n = 1, i;
1502 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1505 /* length of header */
1507 while (size <= 8 && !(total & len_mask)) {
1513 total &= (len_mask - 1);
1515 total = (total << 8) | p->buf[4 + n++];
1517 /* Does the probe data contain the whole header? */
1518 if (p->buf_size < 4 + size + total)
1521 /* The header should contain a known document type. For now,
1522 * we don't parse the whole header but simply check for the
1523 * availability of that array of characters inside the header.
1524 * Not fully fool-proof, but good enough. */
1525 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1526 size_t probelen = strlen(matroska_doctypes[i]);
1527 if (total < probelen)
1529 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1530 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1531 return AVPROBE_SCORE_MAX;
1534 // probably valid EBML header but no recognized doctype
1535 return AVPROBE_SCORE_EXTENSION;
1538 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1541 MatroskaTrack *tracks = matroska->tracks.elem;
1544 for (i = 0; i < matroska->tracks.nb_elem; i++)
1545 if (tracks[i].num == num)
1548 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1552 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1553 MatroskaTrack *track)
1555 MatroskaTrackEncoding *encodings = track->encodings.elem;
1556 uint8_t *data = *buf;
1557 int isize = *buf_size;
1558 uint8_t *pkt_data = NULL;
1559 uint8_t av_unused *newpktdata;
1560 int pkt_size = isize;
1564 if (pkt_size >= 10000000U)
1565 return AVERROR_INVALIDDATA;
1567 switch (encodings[0].compression.algo) {
1568 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1570 int header_size = encodings[0].compression.settings.size;
1571 uint8_t *header = encodings[0].compression.settings.data;
1573 if (header_size && !header) {
1574 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1581 pkt_size = isize + header_size;
1582 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1584 return AVERROR(ENOMEM);
1586 memcpy(pkt_data, header, header_size);
1587 memcpy(pkt_data + header_size, data, isize);
1591 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1593 olen = pkt_size *= 3;
1594 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1595 + AV_INPUT_BUFFER_PADDING_SIZE);
1597 result = AVERROR(ENOMEM);
1600 pkt_data = newpktdata;
1601 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1602 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1604 result = AVERROR_INVALIDDATA;
1611 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1613 z_stream zstream = { 0 };
1614 if (inflateInit(&zstream) != Z_OK)
1616 zstream.next_in = data;
1617 zstream.avail_in = isize;
1620 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1622 inflateEnd(&zstream);
1623 result = AVERROR(ENOMEM);
1626 pkt_data = newpktdata;
1627 zstream.avail_out = pkt_size - zstream.total_out;
1628 zstream.next_out = pkt_data + zstream.total_out;
1629 result = inflate(&zstream, Z_NO_FLUSH);
1630 } while (result == Z_OK && pkt_size < 10000000);
1631 pkt_size = zstream.total_out;
1632 inflateEnd(&zstream);
1633 if (result != Z_STREAM_END) {
1634 if (result == Z_MEM_ERROR)
1635 result = AVERROR(ENOMEM);
1637 result = AVERROR_INVALIDDATA;
1644 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1646 bz_stream bzstream = { 0 };
1647 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1649 bzstream.next_in = data;
1650 bzstream.avail_in = isize;
1653 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1655 BZ2_bzDecompressEnd(&bzstream);
1656 result = AVERROR(ENOMEM);
1659 pkt_data = newpktdata;
1660 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1661 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1662 result = BZ2_bzDecompress(&bzstream);
1663 } while (result == BZ_OK && pkt_size < 10000000);
1664 pkt_size = bzstream.total_out_lo32;
1665 BZ2_bzDecompressEnd(&bzstream);
1666 if (result != BZ_STREAM_END) {
1667 if (result == BZ_MEM_ERROR)
1668 result = AVERROR(ENOMEM);
1670 result = AVERROR_INVALIDDATA;
1677 return AVERROR_INVALIDDATA;
1680 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1683 *buf_size = pkt_size;
1691 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1692 AVDictionary **metadata, char *prefix)
1694 MatroskaTag *tags = list->elem;
1698 for (i = 0; i < list->nb_elem; i++) {
1699 const char *lang = tags[i].lang &&
1700 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1702 if (!tags[i].name) {
1703 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1707 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1709 av_strlcpy(key, tags[i].name, sizeof(key));
1710 if (tags[i].def || !lang) {
1711 av_dict_set(metadata, key, tags[i].string, 0);
1712 if (tags[i].sub.nb_elem)
1713 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1716 av_strlcat(key, "-", sizeof(key));
1717 av_strlcat(key, lang, sizeof(key));
1718 av_dict_set(metadata, key, tags[i].string, 0);
1719 if (tags[i].sub.nb_elem)
1720 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1723 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1726 static void matroska_convert_tags(AVFormatContext *s)
1728 MatroskaDemuxContext *matroska = s->priv_data;
1729 MatroskaTags *tags = matroska->tags.elem;
1732 for (i = 0; i < matroska->tags.nb_elem; i++) {
1733 if (tags[i].target.attachuid) {
1734 MatroskaAttachment *attachment = matroska->attachments.elem;
1736 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1737 if (attachment[j].uid == tags[i].target.attachuid &&
1738 attachment[j].stream) {
1739 matroska_convert_tag(s, &tags[i].tag,
1740 &attachment[j].stream->metadata, NULL);
1745 av_log(NULL, AV_LOG_WARNING,
1746 "The tags at index %d refer to a "
1747 "non-existent attachment %"PRId64".\n",
1748 i, tags[i].target.attachuid);
1750 } else if (tags[i].target.chapteruid) {
1751 MatroskaChapter *chapter = matroska->chapters.elem;
1753 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1754 if (chapter[j].uid == tags[i].target.chapteruid &&
1755 chapter[j].chapter) {
1756 matroska_convert_tag(s, &tags[i].tag,
1757 &chapter[j].chapter->metadata, NULL);
1762 av_log(NULL, AV_LOG_WARNING,
1763 "The tags at index %d refer to a non-existent chapter "
1765 i, tags[i].target.chapteruid);
1767 } else if (tags[i].target.trackuid) {
1768 MatroskaTrack *track = matroska->tracks.elem;
1770 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1771 if (track[j].uid == tags[i].target.trackuid &&
1773 matroska_convert_tag(s, &tags[i].tag,
1774 &track[j].stream->metadata, NULL);
1779 av_log(NULL, AV_LOG_WARNING,
1780 "The tags at index %d refer to a non-existent track "
1782 i, tags[i].target.trackuid);
1785 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1786 tags[i].target.type);
1791 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1794 uint32_t saved_id = matroska->current_id;
1795 int64_t before_pos = avio_tell(matroska->ctx->pb);
1800 offset = pos + matroska->segment_start;
1801 if (avio_seek(matroska->ctx->pb, offset, SEEK_SET) == offset) {
1802 /* We don't want to lose our seekhead level, so we add
1803 * a dummy. This is a crude hack. */
1804 if (matroska->num_levels == EBML_MAX_DEPTH) {
1805 av_log(matroska->ctx, AV_LOG_INFO,
1806 "Max EBML element depth (%d) reached, "
1807 "cannot parse further.\n", EBML_MAX_DEPTH);
1808 ret = AVERROR_INVALIDDATA;
1810 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1811 matroska->num_levels++;
1812 matroska->current_id = 0;
1814 ret = ebml_parse(matroska, matroska_segment, matroska);
1815 if (ret == LEVEL_ENDED) {
1816 /* This can only happen if the seek brought us beyond EOF. */
1821 /* Seek back - notice that in all instances where this is used
1822 * it is safe to set the level to 1. */
1823 matroska_reset_status(matroska, saved_id, before_pos);
1828 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1830 EbmlList *seekhead_list = &matroska->seekhead;
1833 // we should not do any seeking in the streaming case
1834 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1837 for (i = 0; i < seekhead_list->nb_elem; i++) {
1838 MatroskaSeekhead *seekheads = seekhead_list->elem;
1839 uint32_t id = seekheads[i].id;
1840 uint64_t pos = seekheads[i].pos;
1842 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1843 if (!elem || elem->parsed)
1848 // defer cues parsing until we actually need cue data.
1849 if (id == MATROSKA_ID_CUES)
1852 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1853 // mark index as broken
1854 matroska->cues_parsing_deferred = -1;
1862 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1864 EbmlList *index_list;
1865 MatroskaIndex *index;
1866 uint64_t index_scale = 1;
1869 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1872 index_list = &matroska->index;
1873 index = index_list->elem;
1874 if (index_list->nb_elem < 2)
1876 if (index[1].time > 1E14 / matroska->time_scale) {
1877 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1880 for (i = 0; i < index_list->nb_elem; i++) {
1881 EbmlList *pos_list = &index[i].pos;
1882 MatroskaIndexPos *pos = pos_list->elem;
1883 for (j = 0; j < pos_list->nb_elem; j++) {
1884 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1886 if (track && track->stream)
1887 av_add_index_entry(track->stream,
1888 pos[j].pos + matroska->segment_start,
1889 index[i].time / index_scale, 0, 0,
1895 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1898 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1901 for (i = 0; i < matroska->num_level1_elems; i++) {
1902 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1903 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1904 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1905 matroska->cues_parsing_deferred = -1;
1911 matroska_add_index_entries(matroska);
1914 static int matroska_aac_profile(char *codec_id)
1916 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1919 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1920 if (strstr(codec_id, aac_profiles[profile]))
1925 static int matroska_aac_sri(int samplerate)
1929 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1930 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1935 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1937 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1938 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1941 static int matroska_parse_flac(AVFormatContext *s,
1942 MatroskaTrack *track,
1945 AVStream *st = track->stream;
1946 uint8_t *p = track->codec_priv.data;
1947 int size = track->codec_priv.size;
1949 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1950 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1951 track->codec_priv.size = 0;
1955 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1957 p += track->codec_priv.size;
1958 size -= track->codec_priv.size;
1960 /* parse the remaining metadata blocks if present */
1962 int block_last, block_type, block_size;
1964 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1968 if (block_size > size)
1971 /* check for the channel mask */
1972 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1973 AVDictionary *dict = NULL;
1974 AVDictionaryEntry *chmask;
1976 ff_vorbis_comment(s, &dict, p, block_size, 0);
1977 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1979 uint64_t mask = strtol(chmask->value, NULL, 0);
1980 if (!mask || mask & ~0x3ffffULL) {
1981 av_log(s, AV_LOG_WARNING,
1982 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1984 st->codecpar->channel_layout = mask;
1986 av_dict_free(&dict);
1996 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
1998 int major, minor, micro, bttb = 0;
2000 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2001 * this function, and fixed in 57.52 */
2002 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, µ) == 3)
2003 bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
2005 switch (field_order) {
2006 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2007 return AV_FIELD_PROGRESSIVE;
2008 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2009 return AV_FIELD_UNKNOWN;
2010 case MATROSKA_VIDEO_FIELDORDER_TT:
2012 case MATROSKA_VIDEO_FIELDORDER_BB:
2014 case MATROSKA_VIDEO_FIELDORDER_BT:
2015 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2016 case MATROSKA_VIDEO_FIELDORDER_TB:
2017 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2019 return AV_FIELD_UNKNOWN;
2023 static void mkv_stereo_mode_display_mul(int stereo_mode,
2024 int *h_width, int *h_height)
2026 switch (stereo_mode) {
2027 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2028 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2029 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2030 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2031 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2033 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2034 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2035 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2036 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2039 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2040 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2041 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2042 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2048 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2049 const MatroskaTrackVideoColor *color = track->video.color.elem;
2050 const MatroskaMasteringMeta *mastering_meta;
2051 int has_mastering_primaries, has_mastering_luminance;
2053 if (!track->video.color.nb_elem)
2056 mastering_meta = &color->mastering_meta;
2057 // Mastering primaries are CIE 1931 coords, and must be > 0.
2058 has_mastering_primaries =
2059 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2060 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2061 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2062 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2063 has_mastering_luminance = mastering_meta->max_luminance > 0;
2065 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2066 st->codecpar->color_space = color->matrix_coefficients;
2067 if (color->primaries != AVCOL_PRI_RESERVED &&
2068 color->primaries != AVCOL_PRI_RESERVED0)
2069 st->codecpar->color_primaries = color->primaries;
2070 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2071 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2072 st->codecpar->color_trc = color->transfer_characteristics;
2073 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2074 color->range <= AVCOL_RANGE_JPEG)
2075 st->codecpar->color_range = color->range;
2076 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2077 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2078 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2079 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2080 st->codecpar->chroma_location =
2081 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2082 (color->chroma_siting_vert - 1) << 7);
2084 if (color->max_cll && color->max_fall) {
2087 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2089 return AVERROR(ENOMEM);
2090 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2091 (uint8_t *)metadata, size);
2093 av_freep(&metadata);
2096 metadata->MaxCLL = color->max_cll;
2097 metadata->MaxFALL = color->max_fall;
2100 if (has_mastering_primaries || has_mastering_luminance) {
2101 // Use similar rationals as other standards.
2102 const int chroma_den = 50000;
2103 const int luma_den = 10000;
2104 AVMasteringDisplayMetadata *metadata =
2105 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2106 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2107 sizeof(AVMasteringDisplayMetadata));
2109 return AVERROR(ENOMEM);
2111 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2112 if (has_mastering_primaries) {
2113 metadata->display_primaries[0][0] = av_make_q(
2114 round(mastering_meta->r_x * chroma_den), chroma_den);
2115 metadata->display_primaries[0][1] = av_make_q(
2116 round(mastering_meta->r_y * chroma_den), chroma_den);
2117 metadata->display_primaries[1][0] = av_make_q(
2118 round(mastering_meta->g_x * chroma_den), chroma_den);
2119 metadata->display_primaries[1][1] = av_make_q(
2120 round(mastering_meta->g_y * chroma_den), chroma_den);
2121 metadata->display_primaries[2][0] = av_make_q(
2122 round(mastering_meta->b_x * chroma_den), chroma_den);
2123 metadata->display_primaries[2][1] = av_make_q(
2124 round(mastering_meta->b_y * chroma_den), chroma_den);
2125 metadata->white_point[0] = av_make_q(
2126 round(mastering_meta->white_x * chroma_den), chroma_den);
2127 metadata->white_point[1] = av_make_q(
2128 round(mastering_meta->white_y * chroma_den), chroma_den);
2129 metadata->has_primaries = 1;
2131 if (has_mastering_luminance) {
2132 metadata->max_luminance = av_make_q(
2133 round(mastering_meta->max_luminance * luma_den), luma_den);
2134 metadata->min_luminance = av_make_q(
2135 round(mastering_meta->min_luminance * luma_den), luma_den);
2136 metadata->has_luminance = 1;
2142 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track) {
2143 AVSphericalMapping *spherical;
2144 enum AVSphericalProjection projection;
2145 size_t spherical_size;
2146 uint32_t l = 0, t = 0, r = 0, b = 0;
2147 uint32_t padding = 0;
2151 bytestream2_init(&gb, track->video.projection.private.data,
2152 track->video.projection.private.size);
2154 if (bytestream2_get_byte(&gb) != 0) {
2155 av_log(NULL, AV_LOG_WARNING, "Unknown spherical metadata\n");
2159 bytestream2_skip(&gb, 3); // flags
2161 switch (track->video.projection.type) {
2162 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2163 if (track->video.projection.private.size == 20) {
2164 t = bytestream2_get_be32(&gb);
2165 b = bytestream2_get_be32(&gb);
2166 l = bytestream2_get_be32(&gb);
2167 r = bytestream2_get_be32(&gb);
2169 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2170 av_log(NULL, AV_LOG_ERROR,
2171 "Invalid bounding rectangle coordinates "
2172 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2174 return AVERROR_INVALIDDATA;
2176 } else if (track->video.projection.private.size != 0) {
2177 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2178 return AVERROR_INVALIDDATA;
2181 if (l || t || r || b)
2182 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2184 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2186 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2187 if (track->video.projection.private.size < 4) {
2188 av_log(NULL, AV_LOG_ERROR, "Missing projection private properties\n");
2189 return AVERROR_INVALIDDATA;
2190 } else if (track->video.projection.private.size == 12) {
2191 uint32_t layout = bytestream2_get_be32(&gb);
2193 av_log(NULL, AV_LOG_WARNING,
2194 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2197 projection = AV_SPHERICAL_CUBEMAP;
2198 padding = bytestream2_get_be32(&gb);
2200 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2201 return AVERROR_INVALIDDATA;
2204 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2205 /* No Spherical metadata */
2208 av_log(NULL, AV_LOG_WARNING,
2209 "Unknown spherical metadata type %"PRIu64"\n",
2210 track->video.projection.type);
2214 spherical = av_spherical_alloc(&spherical_size);
2216 return AVERROR(ENOMEM);
2218 spherical->projection = projection;
2220 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2221 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2222 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2224 spherical->padding = padding;
2226 spherical->bound_left = l;
2227 spherical->bound_top = t;
2228 spherical->bound_right = r;
2229 spherical->bound_bottom = b;
2231 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2234 av_freep(&spherical);
2241 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2243 const AVCodecTag *codec_tags;
2245 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2246 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2248 /* Normalize noncompliant private data that starts with the fourcc
2249 * by expanding/shifting the data by 4 bytes and storing the data
2250 * size at the start. */
2251 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2252 int ret = av_buffer_realloc(&track->codec_priv.buf,
2253 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2257 track->codec_priv.data = track->codec_priv.buf->data;
2258 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2259 track->codec_priv.size += 4;
2260 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2263 *fourcc = AV_RL32(track->codec_priv.data + 4);
2264 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2269 static int matroska_parse_tracks(AVFormatContext *s)
2271 MatroskaDemuxContext *matroska = s->priv_data;
2272 MatroskaTrack *tracks = matroska->tracks.elem;
2277 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2278 MatroskaTrack *track = &tracks[i];
2279 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2280 EbmlList *encodings_list = &track->encodings;
2281 MatroskaTrackEncoding *encodings = encodings_list->elem;
2282 uint8_t *extradata = NULL;
2283 int extradata_size = 0;
2284 int extradata_offset = 0;
2285 uint32_t fourcc = 0;
2287 char* key_id_base64 = NULL;
2290 /* Apply some sanity checks. */
2291 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2292 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2293 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2294 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2295 av_log(matroska->ctx, AV_LOG_INFO,
2296 "Unknown or unsupported track type %"PRIu64"\n",
2300 if (!track->codec_id)
2303 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2304 isnan(track->audio.samplerate)) {
2305 av_log(matroska->ctx, AV_LOG_WARNING,
2306 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2307 track->audio.samplerate);
2308 track->audio.samplerate = 8000;
2311 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2312 if (!track->default_duration && track->video.frame_rate > 0) {
2313 double default_duration = 1000000000 / track->video.frame_rate;
2314 if (default_duration > UINT64_MAX || default_duration < 0) {
2315 av_log(matroska->ctx, AV_LOG_WARNING,
2316 "Invalid frame rate %e. Cannot calculate default duration.\n",
2317 track->video.frame_rate);
2319 track->default_duration = default_duration;
2322 if (track->video.display_width == -1)
2323 track->video.display_width = track->video.pixel_width;
2324 if (track->video.display_height == -1)
2325 track->video.display_height = track->video.pixel_height;
2326 if (track->video.color_space.size == 4)
2327 fourcc = AV_RL32(track->video.color_space.data);
2328 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2329 if (!track->audio.out_samplerate)
2330 track->audio.out_samplerate = track->audio.samplerate;
2332 if (encodings_list->nb_elem > 1) {
2333 av_log(matroska->ctx, AV_LOG_ERROR,
2334 "Multiple combined encodings not supported");
2335 } else if (encodings_list->nb_elem == 1) {
2336 if (encodings[0].type) {
2337 if (encodings[0].encryption.key_id.size > 0) {
2338 /* Save the encryption key id to be stored later as a
2340 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2341 key_id_base64 = av_malloc(b64_size);
2342 if (key_id_base64 == NULL)
2343 return AVERROR(ENOMEM);
2345 av_base64_encode(key_id_base64, b64_size,
2346 encodings[0].encryption.key_id.data,
2347 encodings[0].encryption.key_id.size);
2349 encodings[0].scope = 0;
2350 av_log(matroska->ctx, AV_LOG_ERROR,
2351 "Unsupported encoding type");
2355 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2358 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2361 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2363 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2364 encodings[0].scope = 0;
2365 av_log(matroska->ctx, AV_LOG_ERROR,
2366 "Unsupported encoding type");
2367 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2368 uint8_t *codec_priv = track->codec_priv.data;
2369 int ret = matroska_decode_buffer(&track->codec_priv.data,
2370 &track->codec_priv.size,
2373 track->codec_priv.data = NULL;
2374 track->codec_priv.size = 0;
2375 av_log(matroska->ctx, AV_LOG_ERROR,
2376 "Failed to decode codec private data\n");
2379 if (codec_priv != track->codec_priv.data) {
2380 av_buffer_unref(&track->codec_priv.buf);
2381 if (track->codec_priv.data) {
2382 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2383 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2385 if (!track->codec_priv.buf) {
2386 av_freep(&track->codec_priv.data);
2387 track->codec_priv.size = 0;
2388 return AVERROR(ENOMEM);
2395 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2396 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2397 strlen(ff_mkv_codec_tags[j].str))) {
2398 codec_id = ff_mkv_codec_tags[j].id;
2403 st = track->stream = avformat_new_stream(s, NULL);
2405 av_free(key_id_base64);
2406 return AVERROR(ENOMEM);
2409 if (key_id_base64) {
2410 /* export encryption key id as base64 metadata tag */
2411 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
2412 av_freep(&key_id_base64);
2415 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2416 track->codec_priv.size >= 40 &&
2417 track->codec_priv.data) {
2418 track->ms_compat = 1;
2419 bit_depth = AV_RL16(track->codec_priv.data + 14);
2420 fourcc = AV_RL32(track->codec_priv.data + 16);
2421 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2424 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2426 extradata_offset = 40;
2427 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2428 track->codec_priv.size >= 14 &&
2429 track->codec_priv.data) {
2431 ffio_init_context(&b, track->codec_priv.data,
2432 track->codec_priv.size,
2433 0, NULL, NULL, NULL, NULL);
2434 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2437 codec_id = st->codecpar->codec_id;
2438 fourcc = st->codecpar->codec_tag;
2439 extradata_offset = FFMIN(track->codec_priv.size, 18);
2440 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2441 /* Normally 36, but allow noncompliant private data */
2442 && (track->codec_priv.size >= 32)
2443 && (track->codec_priv.data)) {
2444 uint16_t sample_size;
2445 int ret = get_qt_codec(track, &fourcc, &codec_id);
2448 sample_size = AV_RB16(track->codec_priv.data + 26);
2450 if (sample_size == 8) {
2451 fourcc = MKTAG('r','a','w',' ');
2452 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2453 } else if (sample_size == 16) {
2454 fourcc = MKTAG('t','w','o','s');
2455 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2458 if ((fourcc == MKTAG('t','w','o','s') ||
2459 fourcc == MKTAG('s','o','w','t')) &&
2461 codec_id = AV_CODEC_ID_PCM_S8;
2462 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2463 (track->codec_priv.size >= 21) &&
2464 (track->codec_priv.data)) {
2465 int ret = get_qt_codec(track, &fourcc, &codec_id);
2468 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2469 fourcc = MKTAG('S','V','Q','3');
2470 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2472 if (codec_id == AV_CODEC_ID_NONE)
2473 av_log(matroska->ctx, AV_LOG_ERROR,
2474 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2475 if (track->codec_priv.size >= 86) {
2476 bit_depth = AV_RB16(track->codec_priv.data + 82);
2477 ffio_init_context(&b, track->codec_priv.data,
2478 track->codec_priv.size,
2479 0, NULL, NULL, NULL, NULL);
2480 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2482 track->has_palette = 1;
2485 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2486 switch (track->audio.bitdepth) {
2488 codec_id = AV_CODEC_ID_PCM_U8;
2491 codec_id = AV_CODEC_ID_PCM_S24BE;
2494 codec_id = AV_CODEC_ID_PCM_S32BE;
2497 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2498 switch (track->audio.bitdepth) {
2500 codec_id = AV_CODEC_ID_PCM_U8;
2503 codec_id = AV_CODEC_ID_PCM_S24LE;
2506 codec_id = AV_CODEC_ID_PCM_S32LE;
2509 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2510 track->audio.bitdepth == 64) {
2511 codec_id = AV_CODEC_ID_PCM_F64LE;
2512 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2513 int profile = matroska_aac_profile(track->codec_id);
2514 int sri = matroska_aac_sri(track->audio.samplerate);
2515 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2517 return AVERROR(ENOMEM);
2518 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2519 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2520 if (strstr(track->codec_id, "SBR")) {
2521 sri = matroska_aac_sri(track->audio.out_samplerate);
2522 extradata[2] = 0x56;
2523 extradata[3] = 0xE5;
2524 extradata[4] = 0x80 | (sri << 3);
2528 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2529 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2530 * Create the "atom size", "tag", and "tag version" fields the
2531 * decoder expects manually. */
2532 extradata_size = 12 + track->codec_priv.size;
2533 extradata = av_mallocz(extradata_size +
2534 AV_INPUT_BUFFER_PADDING_SIZE);
2536 return AVERROR(ENOMEM);
2537 AV_WB32(extradata, extradata_size);
2538 memcpy(&extradata[4], "alac", 4);
2539 AV_WB32(&extradata[8], 0);
2540 memcpy(&extradata[12], track->codec_priv.data,
2541 track->codec_priv.size);
2542 } else if (codec_id == AV_CODEC_ID_TTA) {
2543 extradata_size = 30;
2544 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2546 return AVERROR(ENOMEM);
2547 ffio_init_context(&b, extradata, extradata_size, 1,
2548 NULL, NULL, NULL, NULL);
2549 avio_write(&b, "TTA1", 4);
2551 if (track->audio.channels > UINT16_MAX ||
2552 track->audio.bitdepth > UINT16_MAX) {
2553 av_log(matroska->ctx, AV_LOG_WARNING,
2554 "Too large audio channel number %"PRIu64
2555 " or bitdepth %"PRIu64". Skipping track.\n",
2556 track->audio.channels, track->audio.bitdepth);
2557 av_freep(&extradata);
2558 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2559 return AVERROR_INVALIDDATA;
2563 avio_wl16(&b, track->audio.channels);
2564 avio_wl16(&b, track->audio.bitdepth);
2565 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2566 return AVERROR_INVALIDDATA;
2567 avio_wl32(&b, track->audio.out_samplerate);
2568 avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
2569 track->audio.out_samplerate,
2570 AV_TIME_BASE * 1000));
2571 } else if (codec_id == AV_CODEC_ID_RV10 ||
2572 codec_id == AV_CODEC_ID_RV20 ||
2573 codec_id == AV_CODEC_ID_RV30 ||
2574 codec_id == AV_CODEC_ID_RV40) {
2575 extradata_offset = 26;
2576 } else if (codec_id == AV_CODEC_ID_RA_144) {
2577 track->audio.out_samplerate = 8000;
2578 track->audio.channels = 1;
2579 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2580 codec_id == AV_CODEC_ID_COOK ||
2581 codec_id == AV_CODEC_ID_ATRAC3 ||
2582 codec_id == AV_CODEC_ID_SIPR)
2583 && track->codec_priv.data) {
2586 ffio_init_context(&b, track->codec_priv.data,
2587 track->codec_priv.size,
2588 0, NULL, NULL, NULL, NULL);
2590 flavor = avio_rb16(&b);
2591 track->audio.coded_framesize = avio_rb32(&b);
2593 track->audio.sub_packet_h = avio_rb16(&b);
2594 track->audio.frame_size = avio_rb16(&b);
2595 track->audio.sub_packet_size = avio_rb16(&b);
2597 track->audio.coded_framesize <= 0 ||
2598 track->audio.sub_packet_h <= 0 ||
2599 track->audio.frame_size <= 0 ||
2600 track->audio.sub_packet_size <= 0 && codec_id != AV_CODEC_ID_SIPR)
2601 return AVERROR_INVALIDDATA;
2602 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2603 track->audio.frame_size);
2604 if (!track->audio.buf)
2605 return AVERROR(ENOMEM);
2606 if (codec_id == AV_CODEC_ID_RA_288) {
2607 st->codecpar->block_align = track->audio.coded_framesize;
2608 track->codec_priv.size = 0;
2610 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2611 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2612 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2613 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2615 st->codecpar->block_align = track->audio.sub_packet_size;
2616 extradata_offset = 78;
2618 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2619 ret = matroska_parse_flac(s, track, &extradata_offset);
2622 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2623 fourcc = AV_RL32(track->codec_priv.data);
2624 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2625 /* we don't need any value stored in CodecPrivate.
2626 make sure that it's not exported as extradata. */
2627 track->codec_priv.size = 0;
2628 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2629 /* For now, propagate only the OBUs, if any. Once libavcodec is
2630 updated to handle isobmff style extradata this can be removed. */
2631 extradata_offset = 4;
2633 track->codec_priv.size -= extradata_offset;
2635 if (codec_id == AV_CODEC_ID_NONE)
2636 av_log(matroska->ctx, AV_LOG_INFO,
2637 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2639 if (track->time_scale < 0.01)
2640 track->time_scale = 1.0;
2641 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2642 1000 * 1000 * 1000); /* 64 bit pts in ns */
2644 /* convert the delay from ns to the track timebase */
2645 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2646 (AVRational){ 1, 1000000000 },
2649 st->codecpar->codec_id = codec_id;
2651 if (strcmp(track->language, "und"))
2652 av_dict_set(&st->metadata, "language", track->language, 0);
2653 av_dict_set(&st->metadata, "title", track->name, 0);
2655 if (track->flag_default)
2656 st->disposition |= AV_DISPOSITION_DEFAULT;
2657 if (track->flag_forced)
2658 st->disposition |= AV_DISPOSITION_FORCED;
2660 if (!st->codecpar->extradata) {
2662 st->codecpar->extradata = extradata;
2663 st->codecpar->extradata_size = extradata_size;
2664 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2665 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2666 return AVERROR(ENOMEM);
2667 memcpy(st->codecpar->extradata,
2668 track->codec_priv.data + extradata_offset,
2669 track->codec_priv.size);
2673 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2674 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2675 int display_width_mul = 1;
2676 int display_height_mul = 1;
2678 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2679 st->codecpar->codec_tag = fourcc;
2681 st->codecpar->bits_per_coded_sample = bit_depth;
2682 st->codecpar->width = track->video.pixel_width;
2683 st->codecpar->height = track->video.pixel_height;
2685 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2686 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2687 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2688 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2690 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2691 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2693 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2694 av_reduce(&st->sample_aspect_ratio.num,
2695 &st->sample_aspect_ratio.den,
2696 st->codecpar->height * track->video.display_width * display_width_mul,
2697 st->codecpar->width * track->video.display_height * display_height_mul,
2700 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2701 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2703 if (track->default_duration) {
2704 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2705 1000000000, track->default_duration, 30000);
2706 #if FF_API_R_FRAME_RATE
2707 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2708 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2709 st->r_frame_rate = st->avg_frame_rate;
2713 /* export stereo mode flag as metadata tag */
2714 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2715 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2717 /* export alpha mode flag as metadata tag */
2718 if (track->video.alpha_mode)
2719 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2721 /* if we have virtual track, mark the real tracks */
2722 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2724 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2726 snprintf(buf, sizeof(buf), "%s_%d",
2727 ff_matroska_video_stereo_plane[planes[j].type], i);
2728 for (k=0; k < matroska->tracks.nb_elem; k++)
2729 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2730 av_dict_set(&tracks[k].stream->metadata,
2731 "stereo_mode", buf, 0);
2735 // add stream level stereo3d side data if it is a supported format
2736 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2737 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2738 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2743 ret = mkv_parse_video_color(st, track);
2746 ret = mkv_parse_video_projection(st, track);
2749 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2750 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2751 st->codecpar->codec_tag = fourcc;
2752 st->codecpar->sample_rate = track->audio.out_samplerate;
2753 st->codecpar->channels = track->audio.channels;
2754 if (!st->codecpar->bits_per_coded_sample)
2755 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2756 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2757 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2758 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2759 st->need_parsing = AVSTREAM_PARSE_FULL;
2760 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2761 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2762 if (track->codec_delay > 0) {
2763 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2764 (AVRational){1, 1000000000},
2765 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2766 48000 : st->codecpar->sample_rate});
2768 if (track->seek_preroll > 0) {
2769 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2770 (AVRational){1, 1000000000},
2771 (AVRational){1, st->codecpar->sample_rate});
2773 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2774 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2776 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2777 st->disposition |= AV_DISPOSITION_CAPTIONS;
2778 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2779 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2780 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2781 st->disposition |= AV_DISPOSITION_METADATA;
2783 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2784 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2791 static int matroska_read_header(AVFormatContext *s)
2793 MatroskaDemuxContext *matroska = s->priv_data;
2794 EbmlList *attachments_list = &matroska->attachments;
2795 EbmlList *chapters_list = &matroska->chapters;
2796 MatroskaAttachment *attachments;
2797 MatroskaChapter *chapters;
2798 uint64_t max_start = 0;
2804 matroska->cues_parsing_deferred = 1;
2806 /* First read the EBML header. */
2807 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2808 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2809 ebml_free(ebml_syntax, &ebml);
2810 return AVERROR_INVALIDDATA;
2812 if (ebml.version > EBML_VERSION ||
2813 ebml.max_size > sizeof(uint64_t) ||
2814 ebml.id_length > sizeof(uint32_t) ||
2815 ebml.doctype_version > 3) {
2816 avpriv_report_missing_feature(matroska->ctx,
2817 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2818 ebml.version, ebml.doctype, ebml.doctype_version);
2819 ebml_free(ebml_syntax, &ebml);
2820 return AVERROR_PATCHWELCOME;
2821 } else if (ebml.doctype_version == 3) {
2822 av_log(matroska->ctx, AV_LOG_WARNING,
2823 "EBML header using unsupported features\n"
2824 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2825 ebml.version, ebml.doctype, ebml.doctype_version);
2827 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2828 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2830 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2831 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2832 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2833 ebml_free(ebml_syntax, &ebml);
2834 return AVERROR_INVALIDDATA;
2837 ebml_free(ebml_syntax, &ebml);
2839 /* The next thing is a segment. */
2840 pos = avio_tell(matroska->ctx->pb);
2841 res = ebml_parse(matroska, matroska_segments, matroska);
2842 // Try resyncing until we find an EBML_STOP type element.
2844 res = matroska_resync(matroska, pos);
2847 pos = avio_tell(matroska->ctx->pb);
2848 res = ebml_parse(matroska, matroska_segment, matroska);
2850 /* Set data_offset as it might be needed later by seek_frame_generic. */
2851 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2852 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2853 matroska_execute_seekhead(matroska);
2855 if (!matroska->time_scale)
2856 matroska->time_scale = 1000000;
2857 if (matroska->duration)
2858 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2859 1000 / AV_TIME_BASE;
2860 av_dict_set(&s->metadata, "title", matroska->title, 0);
2861 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2863 if (matroska->date_utc.size == 8)
2864 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2866 res = matroska_parse_tracks(s);
2870 attachments = attachments_list->elem;
2871 for (j = 0; j < attachments_list->nb_elem; j++) {
2872 if (!(attachments[j].filename && attachments[j].mime &&
2873 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2874 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2876 AVStream *st = avformat_new_stream(s, NULL);
2879 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2880 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2881 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2883 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2884 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2885 strlen(ff_mkv_image_mime_tags[i].str))) {
2886 st->codecpar->codec_id = ff_mkv_image_mime_tags[i].id;
2891 attachments[j].stream = st;
2893 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2894 AVPacket *pkt = &st->attached_pic;
2896 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2897 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2899 av_init_packet(pkt);
2900 pkt->buf = av_buffer_ref(attachments[j].bin.buf);
2902 return AVERROR(ENOMEM);
2903 pkt->data = attachments[j].bin.data;
2904 pkt->size = attachments[j].bin.size;
2905 pkt->stream_index = st->index;
2906 pkt->flags |= AV_PKT_FLAG_KEY;
2908 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2909 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2911 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2912 attachments[j].bin.size);
2914 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2915 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2916 strlen(ff_mkv_mime_tags[i].str))) {
2917 st->codecpar->codec_id = ff_mkv_mime_tags[i].id;
2925 chapters = chapters_list->elem;
2926 for (i = 0; i < chapters_list->nb_elem; i++)
2927 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2928 (max_start == 0 || chapters[i].start > max_start)) {
2929 chapters[i].chapter =
2930 avpriv_new_chapter(s, chapters[i].uid,
2931 (AVRational) { 1, 1000000000 },
2932 chapters[i].start, chapters[i].end,
2934 if (chapters[i].chapter) {
2935 av_dict_set(&chapters[i].chapter->metadata,
2936 "title", chapters[i].title, 0);
2938 max_start = chapters[i].start;
2941 matroska_add_index_entries(matroska);
2943 matroska_convert_tags(s);
2947 matroska_read_close(s);
2952 * Put one packet in an application-supplied AVPacket struct.
2953 * Returns 0 on success or -1 on failure.
2955 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2958 if (matroska->queue) {
2959 MatroskaTrack *tracks = matroska->tracks.elem;
2960 MatroskaTrack *track;
2962 ff_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
2963 track = &tracks[pkt->stream_index];
2964 if (track->has_palette) {
2965 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2967 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
2969 memcpy(pal, track->palette, AVPALETTE_SIZE);
2971 track->has_palette = 0;
2980 * Free all packets in our internal queue.
2982 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2984 ff_packet_list_free(&matroska->queue, &matroska->queue_end);
2987 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2988 int *buf_size, int type,
2989 uint32_t **lace_buf, int *laces)
2991 int res = 0, n, size = *buf_size;
2992 uint8_t *data = *buf;
2993 uint32_t *lace_size;
2997 *lace_buf = av_malloc(sizeof(**lace_buf));
2999 return AVERROR(ENOMEM);
3001 *lace_buf[0] = size;
3005 av_assert0(size > 0);
3009 lace_size = av_malloc_array(*laces, sizeof(*lace_size));
3011 return AVERROR(ENOMEM);
3014 case 0x1: /* Xiph lacing */
3018 for (n = 0; res == 0 && n < *laces - 1; n++) {
3022 if (size <= total) {
3023 res = AVERROR_INVALIDDATA;
3028 lace_size[n] += temp;
3035 if (size <= total) {
3036 res = AVERROR_INVALIDDATA;
3040 lace_size[n] = size - total;
3044 case 0x2: /* fixed-size lacing */
3045 if (size % (*laces)) {
3046 res = AVERROR_INVALIDDATA;
3049 for (n = 0; n < *laces; n++)
3050 lace_size[n] = size / *laces;
3053 case 0x3: /* EBML lacing */
3057 n = matroska_ebmlnum_uint(matroska, data, size, &num);
3058 if (n < 0 || num > INT_MAX) {
3059 av_log(matroska->ctx, AV_LOG_INFO,
3060 "EBML block data error\n");
3061 res = n<0 ? n : AVERROR_INVALIDDATA;
3066 total = lace_size[0] = num;
3067 for (n = 1; res == 0 && n < *laces - 1; n++) {
3070 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
3071 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
3072 av_log(matroska->ctx, AV_LOG_INFO,
3073 "EBML block data error\n");
3074 res = r<0 ? r : AVERROR_INVALIDDATA;
3079 lace_size[n] = lace_size[n - 1] + snum;
3080 total += lace_size[n];
3082 if (size <= total) {
3083 res = AVERROR_INVALIDDATA;
3086 lace_size[*laces - 1] = size - total;
3092 *lace_buf = lace_size;
3098 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3099 MatroskaTrack *track, AVStream *st,
3100 uint8_t *data, int size, uint64_t timecode,
3103 int a = st->codecpar->block_align;
3104 int sps = track->audio.sub_packet_size;
3105 int cfs = track->audio.coded_framesize;
3106 int h = track->audio.sub_packet_h;
3107 int y = track->audio.sub_packet_cnt;
3108 int w = track->audio.frame_size;
3111 if (!track->audio.pkt_cnt) {
3112 if (track->audio.sub_packet_cnt == 0)
3113 track->audio.buf_timecode = timecode;
3114 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3115 if (size < cfs * h / 2) {
3116 av_log(matroska->ctx, AV_LOG_ERROR,
3117 "Corrupt int4 RM-style audio packet size\n");
3118 return AVERROR_INVALIDDATA;
3120 for (x = 0; x < h / 2; x++)
3121 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3122 data + x * cfs, cfs);
3123 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3125 av_log(matroska->ctx, AV_LOG_ERROR,
3126 "Corrupt sipr RM-style audio packet size\n");
3127 return AVERROR_INVALIDDATA;
3129 memcpy(track->audio.buf + y * w, data, w);
3131 if (size < sps * w / sps || h<=0 || w%sps) {
3132 av_log(matroska->ctx, AV_LOG_ERROR,
3133 "Corrupt generic RM-style audio packet size\n");
3134 return AVERROR_INVALIDDATA;
3136 for (x = 0; x < w / sps; x++)
3137 memcpy(track->audio.buf +
3138 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3139 data + x * sps, sps);
3142 if (++track->audio.sub_packet_cnt >= h) {
3143 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3144 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3145 track->audio.sub_packet_cnt = 0;
3146 track->audio.pkt_cnt = h * w / a;
3150 while (track->audio.pkt_cnt) {
3152 AVPacket pktl, *pkt = &pktl;
3154 ret = av_new_packet(pkt, a);
3159 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3161 pkt->pts = track->audio.buf_timecode;
3162 track->audio.buf_timecode = AV_NOPTS_VALUE;
3164 pkt->stream_index = st->index;
3165 ret = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3167 av_packet_unref(pkt);
3168 return AVERROR(ENOMEM);
3175 /* reconstruct full wavpack blocks from mangled matroska ones */
3176 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
3177 uint8_t **pdst, int *size)
3179 uint8_t *dst = NULL;
3184 int ret, offset = 0;
3186 if (srclen < 12 || track->stream->codecpar->extradata_size < 2)
3187 return AVERROR_INVALIDDATA;
3189 ver = AV_RL16(track->stream->codecpar->extradata);
3191 samples = AV_RL32(src);
3195 while (srclen >= 8) {
3200 uint32_t flags = AV_RL32(src);
3201 uint32_t crc = AV_RL32(src + 4);
3205 multiblock = (flags & 0x1800) != 0x1800;
3208 ret = AVERROR_INVALIDDATA;
3211 blocksize = AV_RL32(src);
3217 if (blocksize > srclen) {
3218 ret = AVERROR_INVALIDDATA;
3222 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3224 ret = AVERROR(ENOMEM);
3228 dstlen += blocksize + 32;
3230 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3231 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3232 AV_WL16(dst + offset + 8, ver); // version
3233 AV_WL16(dst + offset + 10, 0); // track/index_no
3234 AV_WL32(dst + offset + 12, 0); // total samples
3235 AV_WL32(dst + offset + 16, 0); // block index
3236 AV_WL32(dst + offset + 20, samples); // number of samples
3237 AV_WL32(dst + offset + 24, flags); // flags
3238 AV_WL32(dst + offset + 28, crc); // crc
3239 memcpy(dst + offset + 32, src, blocksize); // block data
3242 srclen -= blocksize;
3243 offset += blocksize + 32;
3246 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3258 static int matroska_parse_prores(MatroskaTrack *track, uint8_t *src,
3259 uint8_t **pdst, int *size)
3264 if (AV_RB32(&src[4]) != MKBETAG('i', 'c', 'p', 'f')) {
3265 dst = av_malloc(dstlen + 8 + AV_INPUT_BUFFER_PADDING_SIZE);
3267 return AVERROR(ENOMEM);
3269 AV_WB32(dst, dstlen);
3270 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3271 memcpy(dst + 8, src, dstlen);
3272 memset(dst + 8 + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3282 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3283 MatroskaTrack *track,
3285 uint8_t *data, int data_len,
3290 AVPacket pktl, *pkt = &pktl;
3291 uint8_t *id, *settings, *text, *buf;
3292 int id_len, settings_len, text_len;
3297 return AVERROR_INVALIDDATA;
3300 q = data + data_len;
3305 if (*p == '\r' || *p == '\n') {
3314 if (p >= q || *p != '\n')
3315 return AVERROR_INVALIDDATA;
3321 if (*p == '\r' || *p == '\n') {
3322 settings_len = p - settings;
3330 if (p >= q || *p != '\n')
3331 return AVERROR_INVALIDDATA;
3336 while (text_len > 0) {
3337 const int len = text_len - 1;
3338 const uint8_t c = p[len];
3339 if (c != '\r' && c != '\n')
3345 return AVERROR_INVALIDDATA;
3347 err = av_new_packet(pkt, text_len);
3352 memcpy(pkt->data, text, text_len);
3355 buf = av_packet_new_side_data(pkt,
3356 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3359 av_packet_unref(pkt);
3360 return AVERROR(ENOMEM);
3362 memcpy(buf, id, id_len);
3365 if (settings_len > 0) {
3366 buf = av_packet_new_side_data(pkt,
3367 AV_PKT_DATA_WEBVTT_SETTINGS,
3370 av_packet_unref(pkt);
3371 return AVERROR(ENOMEM);
3373 memcpy(buf, settings, settings_len);
3376 // Do we need this for subtitles?
3377 // pkt->flags = AV_PKT_FLAG_KEY;
3379 pkt->stream_index = st->index;
3380 pkt->pts = timecode;
3382 // Do we need this for subtitles?
3383 // pkt->dts = timecode;
3385 pkt->duration = duration;
3388 err = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3390 av_packet_unref(pkt);
3391 return AVERROR(ENOMEM);
3397 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3398 MatroskaTrack *track, AVStream *st,
3399 AVBufferRef *buf, uint8_t *data, int pkt_size,
3400 uint64_t timecode, uint64_t lace_duration,
3401 int64_t pos, int is_keyframe,
3402 uint8_t *additional, uint64_t additional_id, int additional_size,
3403 int64_t discard_padding)
3405 MatroskaTrackEncoding *encodings = track->encodings.elem;
3406 uint8_t *pkt_data = data;
3408 AVPacket pktl, *pkt = &pktl;
3410 if (encodings && !encodings->type && encodings->scope & 1) {
3411 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
3416 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3418 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
3420 av_log(matroska->ctx, AV_LOG_ERROR,
3421 "Error parsing a wavpack block.\n");
3424 if (pkt_data != data)
3425 av_freep(&pkt_data);
3429 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES) {
3431 res = matroska_parse_prores(track, pkt_data, &pr_data, &pkt_size);
3433 av_log(matroska->ctx, AV_LOG_ERROR,
3434 "Error parsing a prores block.\n");
3437 if (pkt_data != data)
3438 av_freep(&pkt_data);
3442 av_init_packet(pkt);
3443 if (pkt_data != data)
3444 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3447 pkt->buf = av_buffer_ref(buf);
3450 res = AVERROR(ENOMEM);
3454 pkt->data = pkt_data;
3455 pkt->size = pkt_size;
3456 pkt->flags = is_keyframe;
3457 pkt->stream_index = st->index;
3459 if (additional_size > 0) {
3460 uint8_t *side_data = av_packet_new_side_data(pkt,
3461 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3462 additional_size + 8);
3464 av_packet_unref(pkt);
3465 return AVERROR(ENOMEM);
3467 AV_WB64(side_data, additional_id);
3468 memcpy(side_data + 8, additional, additional_size);
3471 if (discard_padding) {
3472 uint8_t *side_data = av_packet_new_side_data(pkt,
3473 AV_PKT_DATA_SKIP_SAMPLES,
3476 av_packet_unref(pkt);
3477 return AVERROR(ENOMEM);
3479 discard_padding = av_rescale_q(discard_padding,
3480 (AVRational){1, 1000000000},
3481 (AVRational){1, st->codecpar->sample_rate});
3482 if (discard_padding > 0) {
3483 AV_WL32(side_data + 4, discard_padding);
3485 AV_WL32(side_data, -discard_padding);
3489 if (track->ms_compat)
3490 pkt->dts = timecode;
3492 pkt->pts = timecode;
3494 pkt->duration = lace_duration;
3496 #if FF_API_CONVERGENCE_DURATION
3497 FF_DISABLE_DEPRECATION_WARNINGS
3498 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3499 pkt->convergence_duration = lace_duration;
3501 FF_ENABLE_DEPRECATION_WARNINGS
3504 res = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3506 av_packet_unref(pkt);
3507 return AVERROR(ENOMEM);
3513 if (pkt_data != data)
3514 av_freep(&pkt_data);
3518 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3519 int size, int64_t pos, uint64_t cluster_time,
3520 uint64_t block_duration, int is_keyframe,
3521 uint8_t *additional, uint64_t additional_id, int additional_size,
3522 int64_t cluster_pos, int64_t discard_padding)
3524 uint64_t timecode = AV_NOPTS_VALUE;
3525 MatroskaTrack *track;
3529 uint32_t *lace_size = NULL;
3530 int n, flags, laces = 0;
3532 int trust_default_duration = 1;
3534 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
3540 track = matroska_find_track_by_num(matroska, num);
3541 if (!track || !track->stream) {
3542 av_log(matroska->ctx, AV_LOG_INFO,
3543 "Invalid stream %"PRIu64"\n", num);
3544 return AVERROR_INVALIDDATA;
3545 } else if (size <= 3)
3548 if (st->discard >= AVDISCARD_ALL)
3550 av_assert1(block_duration != AV_NOPTS_VALUE);
3552 block_time = sign_extend(AV_RB16(data), 16);
3556 if (is_keyframe == -1)
3557 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3559 if (cluster_time != (uint64_t) -1 &&
3560 (block_time >= 0 || cluster_time >= -block_time)) {
3561 timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3562 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3563 timecode < track->end_timecode)
3564 is_keyframe = 0; /* overlapping subtitles are not key frame */
3566 ff_reduce_index(matroska->ctx, st->index);
3567 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3572 if (matroska->skip_to_keyframe &&
3573 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3574 // Compare signed timecodes. Timecode may be negative due to codec delay
3575 // offset. We don't support timestamps greater than int64_t anyway - see
3577 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3580 matroska->skip_to_keyframe = 0;
3581 else if (!st->skip_to_keyframe) {
3582 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3583 matroska->skip_to_keyframe = 0;
3587 res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
3588 &lace_size, &laces);
3593 if (track->audio.samplerate == 8000) {
3594 // If this is needed for more codecs, then add them here
3595 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3596 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3597 trust_default_duration = 0;
3601 if (!block_duration && trust_default_duration)
3602 block_duration = track->default_duration * laces / matroska->time_scale;
3604 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3605 track->end_timecode =
3606 FFMAX(track->end_timecode, timecode + block_duration);
3608 for (n = 0; n < laces; n++) {
3609 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3611 if (lace_size[n] > size) {
3612 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
3616 if ((st->codecpar->codec_id == AV_CODEC_ID_RA_288 ||
3617 st->codecpar->codec_id == AV_CODEC_ID_COOK ||
3618 st->codecpar->codec_id == AV_CODEC_ID_SIPR ||
3619 st->codecpar->codec_id == AV_CODEC_ID_ATRAC3) &&
3620 st->codecpar->block_align && track->audio.sub_packet_size) {
3621 res = matroska_parse_rm_audio(matroska, track, st, data,
3627 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3628 res = matroska_parse_webvtt(matroska, track, st,
3630 timecode, lace_duration,
3635 res = matroska_parse_frame(matroska, track, st, buf, data, lace_size[n],
3636 timecode, lace_duration, pos,
3637 !n ? is_keyframe : 0,
3638 additional, additional_id, additional_size,
3644 if (timecode != AV_NOPTS_VALUE)
3645 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3646 data += lace_size[n];
3647 size -= lace_size[n];
3655 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3657 MatroskaCluster *cluster = &matroska->current_cluster;
3658 MatroskaBlock *block = &cluster->block;
3661 av_assert0(matroska->num_levels <= 2);
3663 if (matroska->num_levels == 1) {
3664 res = ebml_parse(matroska, matroska_clusters, NULL);
3667 /* Found a cluster: subtract the size of the ID already read. */
3668 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3670 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3676 if (matroska->num_levels == 2) {
3677 /* We are inside a cluster. */
3678 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3680 if (res >= 0 && block->bin.size > 0) {
3681 int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3682 uint8_t* additional = block->additional.size > 0 ?
3683 block->additional.data : NULL;
3685 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3686 block->bin.size, block->bin.pos,
3687 cluster->timecode, block->duration,
3688 is_keyframe, additional, block->additional_id,
3689 block->additional.size, cluster->pos,
3690 block->discard_padding);
3693 ebml_free(matroska_blockgroup, block);
3694 memset(block, 0, sizeof(*block));
3695 } else if (!matroska->num_levels) {
3696 if (!avio_feof(matroska->ctx->pb)) {
3697 avio_r8(matroska->ctx->pb);
3698 if (!avio_feof(matroska->ctx->pb)) {
3699 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3700 "end of segment.\n");
3701 return AVERROR_INVALIDDATA;
3711 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3713 MatroskaDemuxContext *matroska = s->priv_data;
3716 if (matroska->resync_pos == -1) {
3717 // This can only happen if generic seeking has been used.
3718 matroska->resync_pos = avio_tell(s->pb);
3721 while (matroska_deliver_packet(matroska, pkt)) {
3723 return (ret < 0) ? ret : AVERROR_EOF;
3724 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3725 ret = matroska_resync(matroska, matroska->resync_pos);
3731 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3732 int64_t timestamp, int flags)
3734 MatroskaDemuxContext *matroska = s->priv_data;
3735 MatroskaTrack *tracks = NULL;
3736 AVStream *st = s->streams[stream_index];
3739 /* Parse the CUES now since we need the index data to seek. */
3740 if (matroska->cues_parsing_deferred > 0) {
3741 matroska->cues_parsing_deferred = 0;
3742 matroska_parse_cues(matroska);
3745 if (!st->nb_index_entries)
3747 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3749 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3750 matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3751 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3752 matroska_clear_queue(matroska);
3753 if (matroska_parse_cluster(matroska) < 0)
3758 matroska_clear_queue(matroska);
3759 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3762 tracks = matroska->tracks.elem;
3763 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3764 tracks[i].audio.pkt_cnt = 0;
3765 tracks[i].audio.sub_packet_cnt = 0;
3766 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3767 tracks[i].end_timecode = 0;
3770 /* We seek to a level 1 element, so set the appropriate status. */
3771 matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3772 if (flags & AVSEEK_FLAG_ANY) {
3773 st->skip_to_keyframe = 0;
3774 matroska->skip_to_timecode = timestamp;
3776 st->skip_to_keyframe = 1;
3777 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3779 matroska->skip_to_keyframe = 1;
3781 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3784 // slightly hackish but allows proper fallback to
3785 // the generic seeking code.
3786 matroska_reset_status(matroska, 0, -1);
3787 matroska->resync_pos = -1;
3788 matroska_clear_queue(matroska);
3789 st->skip_to_keyframe =
3790 matroska->skip_to_keyframe = 0;
3795 static int matroska_read_close(AVFormatContext *s)
3797 MatroskaDemuxContext *matroska = s->priv_data;
3798 MatroskaTrack *tracks = matroska->tracks.elem;
3801 matroska_clear_queue(matroska);
3803 for (n = 0; n < matroska->tracks.nb_elem; n++)
3804 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3805 av_freep(&tracks[n].audio.buf);
3806 ebml_free(matroska_segment, matroska);
3812 int64_t start_time_ns;
3813 int64_t end_time_ns;
3814 int64_t start_offset;
3818 /* This function searches all the Cues and returns the CueDesc corresponding to
3819 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3820 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3822 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3823 MatroskaDemuxContext *matroska = s->priv_data;
3826 int nb_index_entries = s->streams[0]->nb_index_entries;
3827 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3828 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3829 for (i = 1; i < nb_index_entries; i++) {
3830 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3831 index_entries[i].timestamp * matroska->time_scale > ts) {
3836 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3837 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3838 if (i != nb_index_entries - 1) {
3839 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3840 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3842 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3843 // FIXME: this needs special handling for files where Cues appear
3844 // before Clusters. the current logic assumes Cues appear after
3846 cue_desc.end_offset = cues_start - matroska->segment_start;
3851 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3853 MatroskaDemuxContext *matroska = s->priv_data;
3854 uint32_t id = matroska->current_id;
3855 int64_t cluster_pos, before_pos;
3857 if (s->streams[0]->nb_index_entries <= 0) return 0;
3858 // seek to the first cluster using cues.
3859 index = av_index_search_timestamp(s->streams[0], 0, 0);
3860 if (index < 0) return 0;
3861 cluster_pos = s->streams[0]->index_entries[index].pos;
3862 before_pos = avio_tell(s->pb);
3864 uint64_t cluster_id, cluster_length;
3867 avio_seek(s->pb, cluster_pos, SEEK_SET);
3868 // read cluster id and length
3869 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3870 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3872 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3876 matroska_reset_status(matroska, 0, cluster_pos);
3877 matroska_clear_queue(matroska);
3878 if (matroska_parse_cluster(matroska) < 0 ||
3882 pkt = &matroska->queue->pkt;
3883 // 4 + read is the length of the cluster id and the cluster length field.
3884 cluster_pos += 4 + read + cluster_length;
3885 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3891 /* Restore the status after matroska_read_header: */
3892 matroska_reset_status(matroska, id, before_pos);
3897 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3898 double min_buffer, double* buffer,
3899 double* sec_to_download, AVFormatContext *s,
3902 double nano_seconds_per_second = 1000000000.0;
3903 double time_sec = time_ns / nano_seconds_per_second;
3905 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3906 int64_t end_time_ns = time_ns + time_to_search_ns;
3907 double sec_downloaded = 0.0;
3908 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3909 if (desc_curr.start_time_ns == -1)
3911 *sec_to_download = 0.0;
3913 // Check for non cue start time.
3914 if (time_ns > desc_curr.start_time_ns) {
3915 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3916 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3917 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3918 double timeToDownload = (cueBytes * 8.0) / bps;
3920 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3921 *sec_to_download += timeToDownload;
3923 // Check if the search ends within the first cue.
3924 if (desc_curr.end_time_ns >= end_time_ns) {
3925 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3926 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3927 sec_downloaded = percent_to_sub * sec_downloaded;
3928 *sec_to_download = percent_to_sub * *sec_to_download;
3931 if ((sec_downloaded + *buffer) <= min_buffer) {
3935 // Get the next Cue.
3936 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3939 while (desc_curr.start_time_ns != -1) {
3940 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3941 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3942 double desc_sec = desc_ns / nano_seconds_per_second;
3943 double bits = (desc_bytes * 8.0);
3944 double time_to_download = bits / bps;
3946 sec_downloaded += desc_sec - time_to_download;
3947 *sec_to_download += time_to_download;
3949 if (desc_curr.end_time_ns >= end_time_ns) {
3950 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3951 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3952 sec_downloaded = percent_to_sub * sec_downloaded;
3953 *sec_to_download = percent_to_sub * *sec_to_download;
3955 if ((sec_downloaded + *buffer) <= min_buffer)
3960 if ((sec_downloaded + *buffer) <= min_buffer) {
3965 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3967 *buffer = *buffer + sec_downloaded;
3971 /* This function computes the bandwidth of the WebM file with the help of
3972 * buffer_size_after_time_downloaded() function. Both of these functions are
3973 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3974 * Matroska parsing mechanism.
3976 * Returns the bandwidth of the file on success; -1 on error.
3978 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3980 MatroskaDemuxContext *matroska = s->priv_data;
3981 AVStream *st = s->streams[0];
3982 double bandwidth = 0.0;
3985 for (i = 0; i < st->nb_index_entries; i++) {
3986 int64_t prebuffer_ns = 1000000000;
3987 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3988 double nano_seconds_per_second = 1000000000.0;
3989 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3990 double prebuffer_bytes = 0.0;
3991 int64_t temp_prebuffer_ns = prebuffer_ns;
3992 int64_t pre_bytes, pre_ns;
3993 double pre_sec, prebuffer, bits_per_second;
3994 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3996 // Start with the first Cue.
3997 CueDesc desc_end = desc_beg;
3999 // Figure out how much data we have downloaded for the prebuffer. This will
4000 // be used later to adjust the bits per sample to try.
4001 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4002 // Prebuffered the entire Cue.
4003 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4004 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4005 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4007 if (desc_end.start_time_ns == -1) {
4008 // The prebuffer is larger than the duration.
4009 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4011 bits_per_second = 0.0;
4013 // The prebuffer ends in the last Cue. Estimate how much data was
4015 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4016 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4017 pre_sec = pre_ns / nano_seconds_per_second;
4019 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4021 prebuffer = prebuffer_ns / nano_seconds_per_second;
4023 // Set this to 0.0 in case our prebuffer buffers the entire video.
4024 bits_per_second = 0.0;
4026 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4027 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4028 double desc_sec = desc_ns / nano_seconds_per_second;
4029 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4031 // Drop the bps by the percentage of bytes buffered.
4032 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4033 double mod_bits_per_second = calc_bits_per_second * percent;
4035 if (prebuffer < desc_sec) {
4037 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4039 // Add 1 so the bits per second should be a little bit greater than file
4041 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4042 const double min_buffer = 0.0;
4043 double buffer = prebuffer;
4044 double sec_to_download = 0.0;
4046 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4047 min_buffer, &buffer, &sec_to_download,
4051 } else if (rv == 0) {
4052 bits_per_second = (double)(bps);
4057 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4058 } while (desc_end.start_time_ns != -1);
4060 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4062 return (int64_t)bandwidth;
4065 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4067 MatroskaDemuxContext *matroska = s->priv_data;
4068 EbmlList *seekhead_list = &matroska->seekhead;
4069 MatroskaSeekhead *seekhead = seekhead_list->elem;
4071 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4075 // determine cues start and end positions
4076 for (i = 0; i < seekhead_list->nb_elem; i++)
4077 if (seekhead[i].id == MATROSKA_ID_CUES)
4080 if (i >= seekhead_list->nb_elem) return -1;
4082 before_pos = avio_tell(matroska->ctx->pb);
4083 cues_start = seekhead[i].pos + matroska->segment_start;
4084 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4085 // cues_end is computed as cues_start + cues_length + length of the
4086 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4087 // cues_end is inclusive and the above sum is reduced by 1.
4088 uint64_t cues_length, cues_id;
4090 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4091 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4092 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4093 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4096 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4098 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4099 if (cues_start == -1 || cues_end == -1) return -1;
4102 matroska_parse_cues(matroska);
4105 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4108 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4110 // if the file has cues at the start, fix up the init range so that
4111 // it does not include it
4112 if (cues_start <= init_range)
4113 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4116 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4117 if (bandwidth < 0) return -1;
4118 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4120 // check if all clusters start with key frames
4121 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4123 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4124 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4125 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4126 if (!buf) return -1;
4128 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4129 int ret = snprintf(buf + end, 20,
4130 "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4131 i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4132 if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4133 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4135 return AVERROR_INVALIDDATA;
4139 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
4145 static int webm_dash_manifest_read_header(AVFormatContext *s)
4148 int ret = matroska_read_header(s);
4150 MatroskaTrack *tracks;
4151 MatroskaDemuxContext *matroska = s->priv_data;
4153 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4156 if (!s->nb_streams) {
4157 matroska_read_close(s);
4158 av_log(s, AV_LOG_ERROR, "No streams found\n");
4159 return AVERROR_INVALIDDATA;
4162 if (!matroska->is_live) {
4163 buf = av_asprintf("%g", matroska->duration);
4164 if (!buf) return AVERROR(ENOMEM);
4165 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
4168 // initialization range
4169 // 5 is the offset of Cluster ID.
4170 init_range = avio_tell(s->pb) - 5;
4171 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4174 // basename of the file
4175 buf = strrchr(s->url, '/');
4176 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4179 tracks = matroska->tracks.elem;
4180 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4182 // parse the cues and populate Cue related fields
4183 if (!matroska->is_live) {
4184 ret = webm_dash_manifest_cues(s, init_range);
4186 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4191 // use the bandwidth from the command line if it was provided
4192 if (matroska->bandwidth > 0) {
4193 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4194 matroska->bandwidth, 0);
4199 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4204 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4205 static const AVOption options[] = {
4206 { "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 },
4207 { "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 },
4211 static const AVClass webm_dash_class = {
4212 .class_name = "WebM DASH Manifest demuxer",
4213 .item_name = av_default_item_name,
4215 .version = LIBAVUTIL_VERSION_INT,
4218 AVInputFormat ff_matroska_demuxer = {
4219 .name = "matroska,webm",
4220 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4221 .extensions = "mkv,mk3d,mka,mks",
4222 .priv_data_size = sizeof(MatroskaDemuxContext),
4223 .read_probe = matroska_probe,
4224 .read_header = matroska_read_header,
4225 .read_packet = matroska_read_packet,
4226 .read_close = matroska_read_close,
4227 .read_seek = matroska_read_seek,
4228 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4231 AVInputFormat ff_webm_dash_manifest_demuxer = {
4232 .name = "webm_dash_manifest",
4233 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4234 .priv_data_size = sizeof(MatroskaDemuxContext),
4235 .read_header = webm_dash_manifest_read_header,
4236 .read_packet = webm_dash_manifest_read_packet,
4237 .read_close = matroska_read_close,
4238 .priv_class = &webm_dash_class,