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 {
100 size_t list_elem_size;
107 const struct EbmlSyntax *n;
111 typedef struct EbmlList {
113 unsigned int alloc_elem_size;
117 typedef struct EbmlBin {
124 typedef struct Ebml {
129 uint64_t doctype_version;
132 typedef struct MatroskaTrackCompression {
135 } MatroskaTrackCompression;
137 typedef struct MatroskaTrackEncryption {
140 } MatroskaTrackEncryption;
142 typedef struct MatroskaTrackEncoding {
145 MatroskaTrackCompression compression;
146 MatroskaTrackEncryption encryption;
147 } MatroskaTrackEncoding;
149 typedef struct MatroskaMasteringMeta {
158 double max_luminance;
159 double min_luminance;
160 } MatroskaMasteringMeta;
162 typedef struct MatroskaTrackVideoColor {
163 uint64_t matrix_coefficients;
164 uint64_t bits_per_channel;
165 uint64_t chroma_sub_horz;
166 uint64_t chroma_sub_vert;
167 uint64_t cb_sub_horz;
168 uint64_t cb_sub_vert;
169 uint64_t chroma_siting_horz;
170 uint64_t chroma_siting_vert;
172 uint64_t transfer_characteristics;
176 MatroskaMasteringMeta mastering_meta;
177 } MatroskaTrackVideoColor;
179 typedef struct MatroskaTrackVideoProjection {
185 } MatroskaTrackVideoProjection;
187 typedef struct MatroskaTrackVideo {
189 uint64_t display_width;
190 uint64_t display_height;
191 uint64_t pixel_width;
192 uint64_t pixel_height;
194 uint64_t display_unit;
196 uint64_t field_order;
197 uint64_t stereo_mode;
200 MatroskaTrackVideoProjection projection;
201 } MatroskaTrackVideo;
203 typedef struct MatroskaTrackAudio {
205 double out_samplerate;
209 /* real audio header (extracted from extradata) */
216 uint64_t buf_timecode;
218 } MatroskaTrackAudio;
220 typedef struct MatroskaTrackPlane {
223 } MatroskaTrackPlane;
225 typedef struct MatroskaTrackOperation {
226 EbmlList combine_planes;
227 } MatroskaTrackOperation;
229 typedef struct MatroskaTrack {
238 uint64_t default_duration;
239 uint64_t flag_default;
240 uint64_t flag_forced;
241 uint64_t seek_preroll;
242 MatroskaTrackVideo video;
243 MatroskaTrackAudio audio;
244 MatroskaTrackOperation operation;
246 uint64_t codec_delay;
247 uint64_t codec_delay_in_track_tb;
250 int64_t end_timecode;
252 uint64_t max_block_additional_id;
254 uint32_t palette[AVPALETTE_COUNT];
258 typedef struct MatroskaAttachment {
265 } MatroskaAttachment;
267 typedef struct MatroskaChapter {
276 typedef struct MatroskaIndexPos {
281 typedef struct MatroskaIndex {
286 typedef struct MatroskaTag {
294 typedef struct MatroskaTagTarget {
302 typedef struct MatroskaTags {
303 MatroskaTagTarget target;
307 typedef struct MatroskaSeekhead {
312 typedef struct MatroskaLevel {
317 typedef struct MatroskaBlock {
322 uint64_t additional_id;
324 int64_t discard_padding;
327 typedef struct MatroskaCluster {
333 typedef struct MatroskaLevel1Element {
337 } MatroskaLevel1Element;
339 typedef struct MatroskaDemuxContext {
340 const AVClass *class;
341 AVFormatContext *ctx;
344 MatroskaLevel levels[EBML_MAX_DEPTH];
356 EbmlList attachments;
362 /* byte position of the segment inside the stream */
363 int64_t segment_start;
365 /* the packet queue */
367 AVPacketList *queue_end;
371 /* What to skip before effectively reading a packet. */
372 int skip_to_keyframe;
373 uint64_t skip_to_timecode;
375 /* File has a CUES element, but we defer parsing until it is needed. */
376 int cues_parsing_deferred;
378 /* Level1 elements and whether they were read yet */
379 MatroskaLevel1Element level1_elems[64];
380 int num_level1_elems;
382 MatroskaCluster current_cluster;
384 /* WebM DASH Manifest live flag */
387 /* Bandwidth value for WebM DASH Manifest */
389 } MatroskaDemuxContext;
391 #define CHILD_OF(parent) { .def = { .n = parent } }
393 // The following forward declarations need their size because
394 // a tentative definition with internal linkage must not be an
395 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
396 // Removing the sizes breaks MSVC.
397 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
398 matroska_track[27], matroska_track_encoding[6], matroska_track_encodings[2],
399 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_tracks[2],
400 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
401 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
402 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
404 static EbmlSyntax ebml_header[] = {
405 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
406 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
407 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
408 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
409 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
410 { EBML_ID_EBMLVERSION, EBML_NONE },
411 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
412 CHILD_OF(ebml_syntax)
415 static EbmlSyntax ebml_syntax[] = {
416 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
417 { MATROSKA_ID_SEGMENT, EBML_STOP },
421 static EbmlSyntax matroska_info[] = {
422 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
423 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
424 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
425 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
426 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
427 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
428 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
429 CHILD_OF(matroska_segment)
432 static EbmlSyntax matroska_mastering_meta[] = {
433 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
434 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
435 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
436 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
437 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
438 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
439 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
440 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
441 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
442 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
443 CHILD_OF(matroska_track_video_color)
446 static EbmlSyntax matroska_track_video_color[] = {
447 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
448 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=0 } },
449 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
450 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
451 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
452 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
453 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
454 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
455 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
456 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
457 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
458 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
459 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
460 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
461 CHILD_OF(matroska_track_video)
464 static EbmlSyntax matroska_track_video_projection[] = {
465 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
466 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrackVideoProjection, private) },
467 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f=0.0 } },
468 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f=0.0 } },
469 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f=0.0 } },
470 CHILD_OF(matroska_track_video)
473 static EbmlSyntax matroska_track_video[] = {
474 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
475 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
476 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
477 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
478 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
479 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
480 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
481 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
482 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
483 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
484 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
485 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
486 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
487 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
488 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
489 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
490 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
491 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
492 CHILD_OF(matroska_track)
495 static EbmlSyntax matroska_track_audio[] = {
496 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
497 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
498 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
499 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
500 CHILD_OF(matroska_track)
503 static EbmlSyntax matroska_track_encoding_compression[] = {
504 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
505 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
506 CHILD_OF(matroska_track_encoding)
509 static EbmlSyntax matroska_track_encoding_encryption[] = {
510 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
511 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
512 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
513 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
514 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
515 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
516 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
517 CHILD_OF(matroska_track_encoding)
519 static EbmlSyntax matroska_track_encoding[] = {
520 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
521 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
522 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
523 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
524 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
525 CHILD_OF(matroska_track_encodings)
528 static EbmlSyntax matroska_track_encodings[] = {
529 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
530 CHILD_OF(matroska_track)
533 static EbmlSyntax matroska_track_plane[] = {
534 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
535 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
536 CHILD_OF(matroska_track_combine_planes)
539 static EbmlSyntax matroska_track_combine_planes[] = {
540 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
541 CHILD_OF(matroska_track_operation)
544 static EbmlSyntax matroska_track_operation[] = {
545 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
546 CHILD_OF(matroska_track)
549 static EbmlSyntax matroska_track[] = {
550 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
551 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
552 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
553 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
554 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
555 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
556 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
557 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
558 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
559 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
560 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
561 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
562 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
563 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
564 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
565 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
566 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
567 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
568 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
569 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
570 { MATROSKA_ID_CODECNAME, EBML_NONE },
571 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
572 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
573 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
574 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
575 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
576 CHILD_OF(matroska_tracks)
579 static EbmlSyntax matroska_tracks[] = {
580 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
581 CHILD_OF(matroska_segment)
584 static EbmlSyntax matroska_attachment[] = {
585 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
586 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
587 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
588 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
589 { MATROSKA_ID_FILEDESC, EBML_NONE },
590 CHILD_OF(matroska_attachments)
593 static EbmlSyntax matroska_attachments[] = {
594 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
595 CHILD_OF(matroska_segment)
598 static EbmlSyntax matroska_chapter_display[] = {
599 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
600 { MATROSKA_ID_CHAPLANG, EBML_NONE },
601 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
602 CHILD_OF(matroska_chapter_entry)
605 static EbmlSyntax matroska_chapter_entry[] = {
606 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
607 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
608 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
609 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
610 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
611 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
612 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
613 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
614 CHILD_OF(matroska_chapter)
617 static EbmlSyntax matroska_chapter[] = {
618 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
619 { MATROSKA_ID_EDITIONUID, EBML_NONE },
620 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
621 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
622 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
623 CHILD_OF(matroska_chapters)
626 static EbmlSyntax matroska_chapters[] = {
627 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
628 CHILD_OF(matroska_segment)
631 static EbmlSyntax matroska_index_pos[] = {
632 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
633 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
634 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
635 { MATROSKA_ID_CUEDURATION, EBML_NONE },
636 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
637 CHILD_OF(matroska_index_entry)
640 static EbmlSyntax matroska_index_entry[] = {
641 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
642 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
643 CHILD_OF(matroska_index)
646 static EbmlSyntax matroska_index[] = {
647 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
648 CHILD_OF(matroska_segment)
651 static EbmlSyntax matroska_simpletag[] = {
652 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
653 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
654 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
655 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
656 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
657 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
658 CHILD_OF(matroska_tag)
661 static EbmlSyntax matroska_tagtargets[] = {
662 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
663 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
664 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
665 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
666 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
667 CHILD_OF(matroska_tag)
670 static EbmlSyntax matroska_tag[] = {
671 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
672 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
673 CHILD_OF(matroska_tags)
676 static EbmlSyntax matroska_tags[] = {
677 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
678 CHILD_OF(matroska_segment)
681 static EbmlSyntax matroska_seekhead_entry[] = {
682 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
683 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
684 CHILD_OF(matroska_seekhead)
687 static EbmlSyntax matroska_seekhead[] = {
688 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
689 CHILD_OF(matroska_segment)
692 static EbmlSyntax matroska_segment[] = {
693 { MATROSKA_ID_CLUSTER, EBML_STOP },
694 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
695 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
696 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
697 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
698 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
699 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
700 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
701 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
704 static EbmlSyntax matroska_segments[] = {
705 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
709 static EbmlSyntax matroska_blockmore[] = {
710 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
711 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
712 CHILD_OF(matroska_blockadditions)
715 static EbmlSyntax matroska_blockadditions[] = {
716 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
717 CHILD_OF(matroska_blockgroup)
720 static EbmlSyntax matroska_blockgroup[] = {
721 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
722 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
723 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
724 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
725 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference), { .i = INT64_MIN } },
726 { MATROSKA_ID_CODECSTATE, EBML_NONE },
727 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
728 CHILD_OF(matroska_cluster_parsing)
731 // The following array contains SimpleBlock and BlockGroup twice
732 // in order to reuse the other values for matroska_cluster_enter.
733 static EbmlSyntax matroska_cluster_parsing[] = {
734 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
735 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, { .n = matroska_blockgroup } },
736 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
737 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
738 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
739 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
740 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
741 CHILD_OF(matroska_segment)
744 static EbmlSyntax matroska_cluster_enter[] = {
745 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = &matroska_cluster_parsing[2] } },
750 static const char *const matroska_doctypes[] = { "matroska", "webm" };
752 static int matroska_read_close(AVFormatContext *s);
755 * This function prepares the status for parsing of level 1 elements.
757 static int matroska_reset_status(MatroskaDemuxContext *matroska,
758 uint32_t id, int64_t position)
761 int64_t err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
766 matroska->current_id = id;
767 matroska->num_levels = 1;
768 matroska->unknown_count = 0;
769 matroska->resync_pos = avio_tell(matroska->ctx->pb);
771 matroska->resync_pos -= (av_log2(id) + 7) / 8;
776 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
778 AVIOContext *pb = matroska->ctx->pb;
781 /* Try to seek to the last position to resync from. If this doesn't work,
782 * we resync from the earliest position available: The start of the buffer. */
783 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
784 av_log(matroska->ctx, AV_LOG_WARNING,
785 "Seek to desired resync point failed. Seeking to "
786 "earliest point available instead.\n");
787 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
788 last_pos + 1), SEEK_SET);
793 // try to find a toplevel element
794 while (!avio_feof(pb)) {
795 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
796 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
797 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
798 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
799 /* Prepare the context for parsing of a level 1 element. */
800 matroska_reset_status(matroska, id, -1);
801 /* Given that we are here means that an error has occurred,
802 * so treat the segment as unknown length in order not to
803 * discard valid data that happens to be beyond the designated
804 * end of the segment. */
805 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
808 id = (id << 8) | avio_r8(pb);
812 return pb->error ? pb->error : AVERROR_EOF;
816 * Read: an "EBML number", which is defined as a variable-length
817 * array of bytes. The first byte indicates the length by giving a
818 * number of 0-bits followed by a one. The position of the first
819 * "one" bit inside the first byte indicates the length of this
821 * Returns: number of bytes read, < 0 on error
823 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
824 int max_size, uint64_t *number, int eof_forbidden)
830 /* The first byte tells us the length in bytes - except when it is zero. */
835 /* get the length of the EBML number */
836 read = 8 - ff_log2_tab[total];
838 if (!total || read > max_size) {
839 pos = avio_tell(pb) - 1;
841 av_log(matroska->ctx, AV_LOG_ERROR,
842 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
843 "of an EBML number\n", pos, pos);
845 av_log(matroska->ctx, AV_LOG_ERROR,
846 "Length %d indicated by an EBML number's first byte 0x%02x "
847 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
848 read, (uint8_t) total, pos, pos, max_size);
850 return AVERROR_INVALIDDATA;
853 /* read out length */
854 total ^= 1 << ff_log2_tab[total];
856 total = (total << 8) | avio_r8(pb);
858 if (pb->eof_reached) {
870 av_log(matroska->ctx, AV_LOG_ERROR,
871 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
876 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
877 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
884 * Read a EBML length value.
885 * This needs special handling for the "unknown length" case which has multiple
888 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
891 int res = ebml_read_num(matroska, pb, 8, number, 1);
892 if (res > 0 && *number + 1 == 1ULL << (7 * res))
893 *number = EBML_UNKNOWN_LENGTH;
898 * Read the next element as an unsigned int.
899 * Returns NEEDS_CHECKING.
901 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
905 /* big-endian ordering; build up number */
908 *num = (*num << 8) | avio_r8(pb);
910 return NEEDS_CHECKING;
914 * Read the next element as a signed int.
915 * Returns NEEDS_CHECKING.
917 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
924 *num = sign_extend(avio_r8(pb), 8);
926 /* big-endian ordering; build up number */
928 *num = ((uint64_t)*num << 8) | avio_r8(pb);
931 return NEEDS_CHECKING;
935 * Read the next element as a float.
936 * Returns NEEDS_CHECKING or < 0 on obvious failure.
938 static int ebml_read_float(AVIOContext *pb, int size, double *num)
943 *num = av_int2float(avio_rb32(pb));
945 *num = av_int2double(avio_rb64(pb));
947 return AVERROR_INVALIDDATA;
949 return NEEDS_CHECKING;
953 * Read the next element as an ASCII string.
954 * 0 is success, < 0 or NEEDS_CHECKING is failure.
956 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
961 /* EBML strings are usually not 0-terminated, so we allocate one
962 * byte more, read the string and NULL-terminate it ourselves. */
963 if (!(res = av_malloc(size + 1)))
964 return AVERROR(ENOMEM);
965 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
967 return ret < 0 ? ret : NEEDS_CHECKING;
977 * Read the next element as binary data.
978 * 0 is success, < 0 or NEEDS_CHECKING is failure.
980 static int ebml_read_binary(AVIOContext *pb, int length,
981 int64_t pos, EbmlBin *bin)
985 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
988 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
990 bin->data = bin->buf->data;
993 if ((ret = avio_read(pb, bin->data, length)) != length) {
994 av_buffer_unref(&bin->buf);
997 return ret < 0 ? ret : NEEDS_CHECKING;
1004 * Read the next element, but only the header. The contents
1005 * are supposed to be sub-elements which can be read separately.
1006 * 0 is success, < 0 is failure.
1008 static int ebml_read_master(MatroskaDemuxContext *matroska,
1009 uint64_t length, int64_t pos)
1011 MatroskaLevel *level;
1013 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1014 av_log(matroska->ctx, AV_LOG_ERROR,
1015 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1016 return AVERROR(ENOSYS);
1019 level = &matroska->levels[matroska->num_levels++];
1021 level->length = length;
1027 * Read a signed "EBML number"
1028 * Return: number of bytes processed, < 0 on error
1030 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1031 AVIOContext *pb, int64_t *num)
1036 /* read as unsigned number first */
1037 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1040 /* make signed (weird way) */
1041 *num = unum - ((1LL << (7 * res - 1)) - 1);
1046 static int ebml_parse(MatroskaDemuxContext *matroska,
1047 EbmlSyntax *syntax, void *data);
1049 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1053 // Whoever touches this should be aware of the duplication
1054 // existing in matroska_cluster_parsing.
1055 for (i = 0; syntax[i].id; i++)
1056 if (id == syntax[i].id)
1062 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1068 for (int i = 0; syntax[i].id; i++)
1069 switch (syntax[i].type) {
1071 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1074 *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1077 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1081 // the default may be NULL
1082 if (syntax[i].def.s) {
1083 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1084 *dst = av_strdup(syntax[i].def.s);
1086 return AVERROR(ENOMEM);
1091 if (!matroska->levels[matroska->num_levels - 1].length) {
1092 matroska->num_levels--;
1098 res = ebml_parse(matroska, syntax, data);
1101 return res == LEVEL_ENDED ? 0 : res;
1104 static int is_ebml_id_valid(uint32_t id)
1106 // Due to endian nonsense in Matroska, the highest byte with any bits set
1107 // will contain the leading length bit. This bit in turn identifies the
1108 // total byte length of the element by its position within the byte.
1109 unsigned int bits = av_log2(id);
1110 return id && (bits + 7) / 8 == (8 - bits % 8);
1114 * Allocate and return the entry for the level1 element with the given ID. If
1115 * an entry already exists, return the existing entry.
1117 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1121 MatroskaLevel1Element *elem;
1123 if (!is_ebml_id_valid(id))
1126 // Some files link to all clusters; useless.
1127 if (id == MATROSKA_ID_CLUSTER)
1130 // There can be multiple seekheads.
1131 if (id != MATROSKA_ID_SEEKHEAD) {
1132 for (i = 0; i < matroska->num_level1_elems; i++) {
1133 if (matroska->level1_elems[i].id == id)
1134 return &matroska->level1_elems[i];
1138 // Only a completely broken file would have more elements.
1139 // It also provides a low-effort way to escape from circular seekheads
1140 // (every iteration will add a level1 entry).
1141 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1142 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1146 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1147 *elem = (MatroskaLevel1Element){.id = id};
1152 static int ebml_parse(MatroskaDemuxContext *matroska,
1153 EbmlSyntax *syntax, void *data)
1155 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1156 // Forbid unknown-length EBML_NONE elements.
1157 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1161 // max. 16 MB for strings
1162 [EBML_STR] = 0x1000000,
1163 [EBML_UTF8] = 0x1000000,
1164 // max. 256 MB for binary data
1165 [EBML_BIN] = 0x10000000,
1166 // no limits for anything else
1168 AVIOContext *pb = matroska->ctx->pb;
1171 int64_t pos = avio_tell(pb), pos_alt;
1172 int res, update_pos = 1, level_check;
1173 MatroskaLevel1Element *level1_elem;
1174 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1176 if (!matroska->current_id) {
1178 res = ebml_read_num(matroska, pb, 4, &id, 0);
1180 if (pb->eof_reached && res == AVERROR_EOF) {
1181 if (matroska->is_live)
1182 // in live mode, finish parsing if EOF is reached.
1184 if (level && pos == avio_tell(pb)) {
1185 if (level->length == EBML_UNKNOWN_LENGTH) {
1186 // Unknown-length levels automatically end at EOF.
1187 matroska->num_levels--;
1190 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1191 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1197 matroska->current_id = id | 1 << 7 * res;
1198 pos_alt = pos + res;
1201 pos -= (av_log2(matroska->current_id) + 7) / 8;
1204 id = matroska->current_id;
1206 syntax = ebml_parse_id(syntax, id);
1207 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1208 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1209 // Unknown-length levels end when an element from an upper level
1210 // in the hierarchy is encountered.
1211 while (syntax->def.n) {
1212 syntax = ebml_parse_id(syntax->def.n, id);
1214 matroska->num_levels--;
1220 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1221 "%"PRId64"\n", id, pos);
1222 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1226 data = (char *) data + syntax->data_offset;
1227 if (syntax->list_elem_size) {
1228 EbmlList *list = data;
1231 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1232 return AVERROR(ENOMEM);
1233 newelem = av_fast_realloc(list->elem,
1234 &list->alloc_elem_size,
1235 (list->nb_elem + 1) * syntax->list_elem_size);
1237 return AVERROR(ENOMEM);
1238 list->elem = newelem;
1239 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1240 memset(data, 0, syntax->list_elem_size);
1245 if (syntax->type != EBML_STOP) {
1246 matroska->current_id = 0;
1247 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1252 if (matroska->num_levels > 0) {
1253 if (length != EBML_UNKNOWN_LENGTH &&
1254 level->length != EBML_UNKNOWN_LENGTH) {
1255 uint64_t elem_end = pos_alt + length,
1256 level_end = level->start + level->length;
1258 if (elem_end < level_end) {
1260 } else if (elem_end == level_end) {
1261 level_check = LEVEL_ENDED;
1263 av_log(matroska->ctx, AV_LOG_ERROR,
1264 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1265 "containing master element ending at 0x%"PRIx64"\n",
1266 pos, elem_end, level_end);
1267 return AVERROR_INVALIDDATA;
1269 } else if (length != EBML_UNKNOWN_LENGTH) {
1271 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1272 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1273 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1274 return AVERROR_INVALIDDATA;
1277 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1278 || syntax->type == EBML_NEST)) {
1279 // According to the current specifications only clusters and
1280 // segments are allowed to be unknown-length. We also accept
1281 // other unknown-length master elements.
1282 av_log(matroska->ctx, AV_LOG_WARNING,
1283 "Found unknown-length element 0x%"PRIX32" other than "
1284 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1285 "parsing will nevertheless be attempted.\n", id, pos);
1292 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1293 if (length != EBML_UNKNOWN_LENGTH) {
1294 av_log(matroska->ctx, AV_LOG_ERROR,
1295 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1296 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1297 length, max_lengths[syntax->type], id, pos);
1298 } else if (syntax->type != EBML_NONE) {
1299 av_log(matroska->ctx, AV_LOG_ERROR,
1300 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1301 "unknown length, yet the length of an element of its "
1302 "type must be known.\n", id, pos);
1304 av_log(matroska->ctx, AV_LOG_ERROR,
1305 "Found unknown-length element with ID 0x%"PRIX32" at "
1306 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1307 "available.\n", id, pos);
1309 return AVERROR_INVALIDDATA;
1312 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1313 // Loosing sync will likely manifest itself as encountering unknown
1314 // elements which are not reliably distinguishable from elements
1315 // belonging to future extensions of the format.
1316 // We use a heuristic to detect such situations: If the current
1317 // element is not expected at the current syntax level and there
1318 // were only a few unknown elements in a row, then the element is
1319 // skipped or considered defective based upon the length of the
1320 // current element (i.e. how much would be skipped); if there were
1321 // more than a few skipped elements in a row and skipping the current
1322 // element would lead us more than SKIP_THRESHOLD away from the last
1323 // known good position, then it is inferred that an error occurred.
1324 // The dependency on the number of unknown elements in a row exists
1325 // because the distance to the last known good position is
1326 // automatically big if the last parsed element was big.
1327 // In both cases, each unknown element is considered equivalent to
1328 // UNKNOWN_EQUIV of skipped bytes for the check.
1329 // The whole check is only done for non-seekable output, because
1330 // in this situation skipped data can't simply be rechecked later.
1331 // This is especially important when using unkown length elements
1332 // as the check for whether a child exceeds its containing master
1333 // element is not effective in this situation.
1335 matroska->unknown_count = 0;
1337 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1339 if (matroska->unknown_count > 3)
1340 dist += pos_alt - matroska->resync_pos;
1342 if (dist > SKIP_THRESHOLD) {
1343 av_log(matroska->ctx, AV_LOG_ERROR,
1344 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1345 "length 0x%"PRIx64" considered as invalid data. Last "
1346 "known good position 0x%"PRIx64", %d unknown elements"
1347 " in a row\n", id, pos, length, matroska->resync_pos,
1348 matroska->unknown_count);
1349 return AVERROR_INVALIDDATA;
1354 if (update_pos > 0) {
1355 // We have found an element that is allowed at this place
1356 // in the hierarchy and it passed all checks, so treat the beginning
1357 // of the element as the "last known good" position.
1358 matroska->resync_pos = pos;
1361 if (!data && length != EBML_UNKNOWN_LENGTH)
1365 switch (syntax->type) {
1367 res = ebml_read_uint(pb, length, data);
1370 res = ebml_read_sint(pb, length, data);
1373 res = ebml_read_float(pb, length, data);
1377 res = ebml_read_ascii(pb, length, data);
1380 res = ebml_read_binary(pb, length, pos_alt, data);
1384 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1386 if (id == MATROSKA_ID_SEGMENT)
1387 matroska->segment_start = pos_alt;
1388 if (id == MATROSKA_ID_CUES)
1389 matroska->cues_parsing_deferred = 0;
1390 if (syntax->type == EBML_LEVEL1 &&
1391 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1392 if (!level1_elem->pos) {
1393 // Zero is not a valid position for a level 1 element.
1394 level1_elem->pos = pos;
1395 } else if (level1_elem->pos != pos)
1396 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1397 level1_elem->parsed = 1;
1399 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1408 if (ffio_limit(pb, length) != length) {
1409 // ffio_limit emits its own error message,
1410 // so we don't have to.
1411 return AVERROR(EIO);
1413 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1414 // avio_skip might take us past EOF. We check for this
1415 // by skipping only length - 1 bytes, reading a byte and
1416 // checking the error flags. This is done in order to check
1417 // that the element has been properly skipped even when
1418 // no filesize (that ffio_limit relies on) is available.
1420 res = NEEDS_CHECKING;
1427 if (res == NEEDS_CHECKING) {
1428 if (pb->eof_reached) {
1437 if (res == AVERROR_INVALIDDATA)
1438 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1439 else if (res == AVERROR(EIO))
1440 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1441 else if (res == AVERROR_EOF) {
1442 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1450 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1451 level = &matroska->levels[matroska->num_levels - 1];
1452 pos = avio_tell(pb);
1454 // Given that pos >= level->start no check for
1455 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1456 while (matroska->num_levels && pos == level->start + level->length) {
1457 matroska->num_levels--;
1465 static void ebml_free(EbmlSyntax *syntax, void *data)
1468 for (i = 0; syntax[i].id; i++) {
1469 void *data_off = (char *) data + syntax[i].data_offset;
1470 switch (syntax[i].type) {
1476 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1480 if (syntax[i].list_elem_size) {
1481 EbmlList *list = data_off;
1482 char *ptr = list->elem;
1483 for (j = 0; j < list->nb_elem;
1484 j++, ptr += syntax[i].list_elem_size)
1485 ebml_free(syntax[i].def.n, ptr);
1486 av_freep(&list->elem);
1488 list->alloc_elem_size = 0;
1490 ebml_free(syntax[i].def.n, data_off);
1500 static int matroska_probe(const AVProbeData *p)
1503 int len_mask = 0x80, size = 1, n = 1, i;
1506 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1509 /* length of header */
1511 while (size <= 8 && !(total & len_mask)) {
1517 total &= (len_mask - 1);
1519 total = (total << 8) | p->buf[4 + n++];
1521 if (total + 1 == 1ULL << (7 * size)){
1522 /* Unknown-length header - simply parse the whole buffer. */
1523 total = p->buf_size - 4 - size;
1525 /* Does the probe data contain the whole header? */
1526 if (p->buf_size < 4 + size + total)
1530 /* The header should contain a known document type. For now,
1531 * we don't parse the whole header but simply check for the
1532 * availability of that array of characters inside the header.
1533 * Not fully fool-proof, but good enough. */
1534 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1535 size_t probelen = strlen(matroska_doctypes[i]);
1536 if (total < probelen)
1538 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1539 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1540 return AVPROBE_SCORE_MAX;
1543 // probably valid EBML header but no recognized doctype
1544 return AVPROBE_SCORE_EXTENSION;
1547 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1550 MatroskaTrack *tracks = matroska->tracks.elem;
1553 for (i = 0; i < matroska->tracks.nb_elem; i++)
1554 if (tracks[i].num == num)
1557 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1561 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1562 MatroskaTrack *track)
1564 MatroskaTrackEncoding *encodings = track->encodings.elem;
1565 uint8_t *data = *buf;
1566 int isize = *buf_size;
1567 uint8_t *pkt_data = NULL;
1568 uint8_t av_unused *newpktdata;
1569 int pkt_size = isize;
1573 if (pkt_size >= 10000000U)
1574 return AVERROR_INVALIDDATA;
1576 switch (encodings[0].compression.algo) {
1577 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1579 int header_size = encodings[0].compression.settings.size;
1580 uint8_t *header = encodings[0].compression.settings.data;
1582 if (header_size && !header) {
1583 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1590 pkt_size = isize + header_size;
1591 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1593 return AVERROR(ENOMEM);
1595 memcpy(pkt_data, header, header_size);
1596 memcpy(pkt_data + header_size, data, isize);
1600 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1602 olen = pkt_size *= 3;
1603 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1604 + AV_INPUT_BUFFER_PADDING_SIZE);
1606 result = AVERROR(ENOMEM);
1609 pkt_data = newpktdata;
1610 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1611 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1613 result = AVERROR_INVALIDDATA;
1620 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1622 z_stream zstream = { 0 };
1623 if (inflateInit(&zstream) != Z_OK)
1625 zstream.next_in = data;
1626 zstream.avail_in = isize;
1629 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1631 inflateEnd(&zstream);
1632 result = AVERROR(ENOMEM);
1635 pkt_data = newpktdata;
1636 zstream.avail_out = pkt_size - zstream.total_out;
1637 zstream.next_out = pkt_data + zstream.total_out;
1638 result = inflate(&zstream, Z_NO_FLUSH);
1639 } while (result == Z_OK && pkt_size < 10000000);
1640 pkt_size = zstream.total_out;
1641 inflateEnd(&zstream);
1642 if (result != Z_STREAM_END) {
1643 if (result == Z_MEM_ERROR)
1644 result = AVERROR(ENOMEM);
1646 result = AVERROR_INVALIDDATA;
1653 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1655 bz_stream bzstream = { 0 };
1656 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1658 bzstream.next_in = data;
1659 bzstream.avail_in = isize;
1662 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1664 BZ2_bzDecompressEnd(&bzstream);
1665 result = AVERROR(ENOMEM);
1668 pkt_data = newpktdata;
1669 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1670 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1671 result = BZ2_bzDecompress(&bzstream);
1672 } while (result == BZ_OK && pkt_size < 10000000);
1673 pkt_size = bzstream.total_out_lo32;
1674 BZ2_bzDecompressEnd(&bzstream);
1675 if (result != BZ_STREAM_END) {
1676 if (result == BZ_MEM_ERROR)
1677 result = AVERROR(ENOMEM);
1679 result = AVERROR_INVALIDDATA;
1686 return AVERROR_INVALIDDATA;
1689 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1692 *buf_size = pkt_size;
1700 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1701 AVDictionary **metadata, char *prefix)
1703 MatroskaTag *tags = list->elem;
1707 for (i = 0; i < list->nb_elem; i++) {
1708 const char *lang = tags[i].lang &&
1709 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1711 if (!tags[i].name) {
1712 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1716 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1718 av_strlcpy(key, tags[i].name, sizeof(key));
1719 if (tags[i].def || !lang) {
1720 av_dict_set(metadata, key, tags[i].string, 0);
1721 if (tags[i].sub.nb_elem)
1722 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1725 av_strlcat(key, "-", sizeof(key));
1726 av_strlcat(key, lang, sizeof(key));
1727 av_dict_set(metadata, key, tags[i].string, 0);
1728 if (tags[i].sub.nb_elem)
1729 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1732 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1735 static void matroska_convert_tags(AVFormatContext *s)
1737 MatroskaDemuxContext *matroska = s->priv_data;
1738 MatroskaTags *tags = matroska->tags.elem;
1741 for (i = 0; i < matroska->tags.nb_elem; i++) {
1742 if (tags[i].target.attachuid) {
1743 MatroskaAttachment *attachment = matroska->attachments.elem;
1745 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1746 if (attachment[j].uid == tags[i].target.attachuid &&
1747 attachment[j].stream) {
1748 matroska_convert_tag(s, &tags[i].tag,
1749 &attachment[j].stream->metadata, NULL);
1754 av_log(NULL, AV_LOG_WARNING,
1755 "The tags at index %d refer to a "
1756 "non-existent attachment %"PRId64".\n",
1757 i, tags[i].target.attachuid);
1759 } else if (tags[i].target.chapteruid) {
1760 MatroskaChapter *chapter = matroska->chapters.elem;
1762 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1763 if (chapter[j].uid == tags[i].target.chapteruid &&
1764 chapter[j].chapter) {
1765 matroska_convert_tag(s, &tags[i].tag,
1766 &chapter[j].chapter->metadata, NULL);
1771 av_log(NULL, AV_LOG_WARNING,
1772 "The tags at index %d refer to a non-existent chapter "
1774 i, tags[i].target.chapteruid);
1776 } else if (tags[i].target.trackuid) {
1777 MatroskaTrack *track = matroska->tracks.elem;
1779 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1780 if (track[j].uid == tags[i].target.trackuid &&
1782 matroska_convert_tag(s, &tags[i].tag,
1783 &track[j].stream->metadata, NULL);
1788 av_log(NULL, AV_LOG_WARNING,
1789 "The tags at index %d refer to a non-existent track "
1791 i, tags[i].target.trackuid);
1794 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1795 tags[i].target.type);
1800 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1803 uint32_t saved_id = matroska->current_id;
1804 int64_t before_pos = avio_tell(matroska->ctx->pb);
1808 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1809 /* We don't want to lose our seekhead level, so we add
1810 * a dummy. This is a crude hack. */
1811 if (matroska->num_levels == EBML_MAX_DEPTH) {
1812 av_log(matroska->ctx, AV_LOG_INFO,
1813 "Max EBML element depth (%d) reached, "
1814 "cannot parse further.\n", EBML_MAX_DEPTH);
1815 ret = AVERROR_INVALIDDATA;
1817 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1818 matroska->num_levels++;
1819 matroska->current_id = 0;
1821 ret = ebml_parse(matroska, matroska_segment, matroska);
1822 if (ret == LEVEL_ENDED) {
1823 /* This can only happen if the seek brought us beyond EOF. */
1828 /* Seek back - notice that in all instances where this is used
1829 * it is safe to set the level to 1. */
1830 matroska_reset_status(matroska, saved_id, before_pos);
1835 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1837 EbmlList *seekhead_list = &matroska->seekhead;
1840 // we should not do any seeking in the streaming case
1841 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1844 for (i = 0; i < seekhead_list->nb_elem; i++) {
1845 MatroskaSeekhead *seekheads = seekhead_list->elem;
1846 uint32_t id = seekheads[i].id;
1847 int64_t pos = seekheads[i].pos + matroska->segment_start;
1849 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1850 if (!elem || elem->parsed)
1855 // defer cues parsing until we actually need cue data.
1856 if (id == MATROSKA_ID_CUES)
1859 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1860 // mark index as broken
1861 matroska->cues_parsing_deferred = -1;
1869 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1871 EbmlList *index_list;
1872 MatroskaIndex *index;
1873 uint64_t index_scale = 1;
1876 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1879 index_list = &matroska->index;
1880 index = index_list->elem;
1881 if (index_list->nb_elem < 2)
1883 if (index[1].time > 1E14 / matroska->time_scale) {
1884 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1887 for (i = 0; i < index_list->nb_elem; i++) {
1888 EbmlList *pos_list = &index[i].pos;
1889 MatroskaIndexPos *pos = pos_list->elem;
1890 for (j = 0; j < pos_list->nb_elem; j++) {
1891 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1893 if (track && track->stream)
1894 av_add_index_entry(track->stream,
1895 pos[j].pos + matroska->segment_start,
1896 index[i].time / index_scale, 0, 0,
1902 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1905 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1908 for (i = 0; i < matroska->num_level1_elems; i++) {
1909 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1910 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1911 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1912 matroska->cues_parsing_deferred = -1;
1918 matroska_add_index_entries(matroska);
1921 static int matroska_aac_profile(char *codec_id)
1923 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1926 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1927 if (strstr(codec_id, aac_profiles[profile]))
1932 static int matroska_aac_sri(int samplerate)
1936 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1937 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1942 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1944 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1945 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1948 static int matroska_parse_flac(AVFormatContext *s,
1949 MatroskaTrack *track,
1952 AVStream *st = track->stream;
1953 uint8_t *p = track->codec_priv.data;
1954 int size = track->codec_priv.size;
1956 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1957 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1958 track->codec_priv.size = 0;
1962 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1964 p += track->codec_priv.size;
1965 size -= track->codec_priv.size;
1967 /* parse the remaining metadata blocks if present */
1969 int block_last, block_type, block_size;
1971 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1975 if (block_size > size)
1978 /* check for the channel mask */
1979 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1980 AVDictionary *dict = NULL;
1981 AVDictionaryEntry *chmask;
1983 ff_vorbis_comment(s, &dict, p, block_size, 0);
1984 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1986 uint64_t mask = strtol(chmask->value, NULL, 0);
1987 if (!mask || mask & ~0x3ffffULL) {
1988 av_log(s, AV_LOG_WARNING,
1989 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
1991 st->codecpar->channel_layout = mask;
1993 av_dict_free(&dict);
2003 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2005 int major, minor, micro, bttb = 0;
2007 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2008 * this function, and fixed in 57.52 */
2009 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, µ) == 3)
2010 bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
2012 switch (field_order) {
2013 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2014 return AV_FIELD_PROGRESSIVE;
2015 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2016 return AV_FIELD_UNKNOWN;
2017 case MATROSKA_VIDEO_FIELDORDER_TT:
2019 case MATROSKA_VIDEO_FIELDORDER_BB:
2021 case MATROSKA_VIDEO_FIELDORDER_BT:
2022 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2023 case MATROSKA_VIDEO_FIELDORDER_TB:
2024 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2026 return AV_FIELD_UNKNOWN;
2030 static void mkv_stereo_mode_display_mul(int stereo_mode,
2031 int *h_width, int *h_height)
2033 switch (stereo_mode) {
2034 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2035 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2036 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2037 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2038 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2040 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2041 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2042 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2043 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2046 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2047 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2048 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2049 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2055 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2056 const MatroskaTrackVideoColor *color = track->video.color.elem;
2057 const MatroskaMasteringMeta *mastering_meta;
2058 int has_mastering_primaries, has_mastering_luminance;
2060 if (!track->video.color.nb_elem)
2063 mastering_meta = &color->mastering_meta;
2064 // Mastering primaries are CIE 1931 coords, and must be > 0.
2065 has_mastering_primaries =
2066 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2067 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2068 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2069 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2070 has_mastering_luminance = mastering_meta->max_luminance > 0;
2072 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2073 st->codecpar->color_space = color->matrix_coefficients;
2074 if (color->primaries != AVCOL_PRI_RESERVED &&
2075 color->primaries != AVCOL_PRI_RESERVED0)
2076 st->codecpar->color_primaries = color->primaries;
2077 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2078 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2079 st->codecpar->color_trc = color->transfer_characteristics;
2080 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2081 color->range <= AVCOL_RANGE_JPEG)
2082 st->codecpar->color_range = color->range;
2083 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2084 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2085 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2086 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2087 st->codecpar->chroma_location =
2088 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2089 (color->chroma_siting_vert - 1) << 7);
2091 if (color->max_cll && color->max_fall) {
2094 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2096 return AVERROR(ENOMEM);
2097 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2098 (uint8_t *)metadata, size);
2100 av_freep(&metadata);
2103 metadata->MaxCLL = color->max_cll;
2104 metadata->MaxFALL = color->max_fall;
2107 if (has_mastering_primaries || has_mastering_luminance) {
2108 AVMasteringDisplayMetadata *metadata =
2109 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2110 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2111 sizeof(AVMasteringDisplayMetadata));
2113 return AVERROR(ENOMEM);
2115 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2116 if (has_mastering_primaries) {
2117 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2118 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2119 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2120 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2121 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2122 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2123 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2124 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2125 metadata->has_primaries = 1;
2127 if (has_mastering_luminance) {
2128 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2129 metadata->min_luminance = av_d2q(mastering_meta->min_luminance, INT_MAX);
2130 metadata->has_luminance = 1;
2136 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track) {
2137 AVSphericalMapping *spherical;
2138 enum AVSphericalProjection projection;
2139 size_t spherical_size;
2140 uint32_t l = 0, t = 0, r = 0, b = 0;
2141 uint32_t padding = 0;
2145 bytestream2_init(&gb, track->video.projection.private.data,
2146 track->video.projection.private.size);
2148 if (bytestream2_get_byte(&gb) != 0) {
2149 av_log(NULL, AV_LOG_WARNING, "Unknown spherical metadata\n");
2153 bytestream2_skip(&gb, 3); // flags
2155 switch (track->video.projection.type) {
2156 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2157 if (track->video.projection.private.size == 20) {
2158 t = bytestream2_get_be32(&gb);
2159 b = bytestream2_get_be32(&gb);
2160 l = bytestream2_get_be32(&gb);
2161 r = bytestream2_get_be32(&gb);
2163 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2164 av_log(NULL, AV_LOG_ERROR,
2165 "Invalid bounding rectangle coordinates "
2166 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2168 return AVERROR_INVALIDDATA;
2170 } else if (track->video.projection.private.size != 0) {
2171 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2172 return AVERROR_INVALIDDATA;
2175 if (l || t || r || b)
2176 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2178 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2180 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2181 if (track->video.projection.private.size < 4) {
2182 av_log(NULL, AV_LOG_ERROR, "Missing projection private properties\n");
2183 return AVERROR_INVALIDDATA;
2184 } else if (track->video.projection.private.size == 12) {
2185 uint32_t layout = bytestream2_get_be32(&gb);
2187 av_log(NULL, AV_LOG_WARNING,
2188 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2191 projection = AV_SPHERICAL_CUBEMAP;
2192 padding = bytestream2_get_be32(&gb);
2194 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2195 return AVERROR_INVALIDDATA;
2198 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2199 /* No Spherical metadata */
2202 av_log(NULL, AV_LOG_WARNING,
2203 "Unknown spherical metadata type %"PRIu64"\n",
2204 track->video.projection.type);
2208 spherical = av_spherical_alloc(&spherical_size);
2210 return AVERROR(ENOMEM);
2212 spherical->projection = projection;
2214 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2215 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2216 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2218 spherical->padding = padding;
2220 spherical->bound_left = l;
2221 spherical->bound_top = t;
2222 spherical->bound_right = r;
2223 spherical->bound_bottom = b;
2225 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2228 av_freep(&spherical);
2235 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2237 const AVCodecTag *codec_tags;
2239 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2240 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2242 /* Normalize noncompliant private data that starts with the fourcc
2243 * by expanding/shifting the data by 4 bytes and storing the data
2244 * size at the start. */
2245 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2246 int ret = av_buffer_realloc(&track->codec_priv.buf,
2247 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2251 track->codec_priv.data = track->codec_priv.buf->data;
2252 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2253 track->codec_priv.size += 4;
2254 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2257 *fourcc = AV_RL32(track->codec_priv.data + 4);
2258 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2263 static int matroska_parse_tracks(AVFormatContext *s)
2265 MatroskaDemuxContext *matroska = s->priv_data;
2266 MatroskaTrack *tracks = matroska->tracks.elem;
2271 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2272 MatroskaTrack *track = &tracks[i];
2273 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2274 EbmlList *encodings_list = &track->encodings;
2275 MatroskaTrackEncoding *encodings = encodings_list->elem;
2276 uint8_t *extradata = NULL;
2277 int extradata_size = 0;
2278 int extradata_offset = 0;
2279 uint32_t fourcc = 0;
2281 char* key_id_base64 = NULL;
2284 /* Apply some sanity checks. */
2285 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2286 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2287 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2288 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2289 av_log(matroska->ctx, AV_LOG_INFO,
2290 "Unknown or unsupported track type %"PRIu64"\n",
2294 if (!track->codec_id)
2297 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2298 isnan(track->audio.samplerate)) {
2299 av_log(matroska->ctx, AV_LOG_WARNING,
2300 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2301 track->audio.samplerate);
2302 track->audio.samplerate = 8000;
2305 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2306 if (!track->default_duration && track->video.frame_rate > 0) {
2307 double default_duration = 1000000000 / track->video.frame_rate;
2308 if (default_duration > UINT64_MAX || default_duration < 0) {
2309 av_log(matroska->ctx, AV_LOG_WARNING,
2310 "Invalid frame rate %e. Cannot calculate default duration.\n",
2311 track->video.frame_rate);
2313 track->default_duration = default_duration;
2316 if (track->video.display_width == -1)
2317 track->video.display_width = track->video.pixel_width;
2318 if (track->video.display_height == -1)
2319 track->video.display_height = track->video.pixel_height;
2320 if (track->video.color_space.size == 4)
2321 fourcc = AV_RL32(track->video.color_space.data);
2322 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2323 if (!track->audio.out_samplerate)
2324 track->audio.out_samplerate = track->audio.samplerate;
2326 if (encodings_list->nb_elem > 1) {
2327 av_log(matroska->ctx, AV_LOG_ERROR,
2328 "Multiple combined encodings not supported");
2329 } else if (encodings_list->nb_elem == 1) {
2330 if (encodings[0].type) {
2331 if (encodings[0].encryption.key_id.size > 0) {
2332 /* Save the encryption key id to be stored later as a
2334 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2335 key_id_base64 = av_malloc(b64_size);
2336 if (key_id_base64 == NULL)
2337 return AVERROR(ENOMEM);
2339 av_base64_encode(key_id_base64, b64_size,
2340 encodings[0].encryption.key_id.data,
2341 encodings[0].encryption.key_id.size);
2343 encodings[0].scope = 0;
2344 av_log(matroska->ctx, AV_LOG_ERROR,
2345 "Unsupported encoding type");
2349 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2352 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2355 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2357 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2358 encodings[0].scope = 0;
2359 av_log(matroska->ctx, AV_LOG_ERROR,
2360 "Unsupported encoding type");
2361 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2362 uint8_t *codec_priv = track->codec_priv.data;
2363 int ret = matroska_decode_buffer(&track->codec_priv.data,
2364 &track->codec_priv.size,
2367 track->codec_priv.data = NULL;
2368 track->codec_priv.size = 0;
2369 av_log(matroska->ctx, AV_LOG_ERROR,
2370 "Failed to decode codec private data\n");
2373 if (codec_priv != track->codec_priv.data) {
2374 av_buffer_unref(&track->codec_priv.buf);
2375 if (track->codec_priv.data) {
2376 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2377 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2379 if (!track->codec_priv.buf) {
2380 av_freep(&track->codec_priv.data);
2381 track->codec_priv.size = 0;
2382 return AVERROR(ENOMEM);
2389 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2390 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2391 strlen(ff_mkv_codec_tags[j].str))) {
2392 codec_id = ff_mkv_codec_tags[j].id;
2397 st = track->stream = avformat_new_stream(s, NULL);
2399 av_free(key_id_base64);
2400 return AVERROR(ENOMEM);
2403 if (key_id_base64) {
2404 /* export encryption key id as base64 metadata tag */
2405 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
2406 av_freep(&key_id_base64);
2409 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2410 track->codec_priv.size >= 40 &&
2411 track->codec_priv.data) {
2412 track->ms_compat = 1;
2413 bit_depth = AV_RL16(track->codec_priv.data + 14);
2414 fourcc = AV_RL32(track->codec_priv.data + 16);
2415 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2418 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2420 extradata_offset = 40;
2421 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2422 track->codec_priv.size >= 14 &&
2423 track->codec_priv.data) {
2425 ffio_init_context(&b, track->codec_priv.data,
2426 track->codec_priv.size,
2427 0, NULL, NULL, NULL, NULL);
2428 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2431 codec_id = st->codecpar->codec_id;
2432 fourcc = st->codecpar->codec_tag;
2433 extradata_offset = FFMIN(track->codec_priv.size, 18);
2434 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2435 /* Normally 36, but allow noncompliant private data */
2436 && (track->codec_priv.size >= 32)
2437 && (track->codec_priv.data)) {
2438 uint16_t sample_size;
2439 int ret = get_qt_codec(track, &fourcc, &codec_id);
2442 sample_size = AV_RB16(track->codec_priv.data + 26);
2444 if (sample_size == 8) {
2445 fourcc = MKTAG('r','a','w',' ');
2446 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2447 } else if (sample_size == 16) {
2448 fourcc = MKTAG('t','w','o','s');
2449 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2452 if ((fourcc == MKTAG('t','w','o','s') ||
2453 fourcc == MKTAG('s','o','w','t')) &&
2455 codec_id = AV_CODEC_ID_PCM_S8;
2456 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2457 (track->codec_priv.size >= 21) &&
2458 (track->codec_priv.data)) {
2459 int ret = get_qt_codec(track, &fourcc, &codec_id);
2462 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2463 fourcc = MKTAG('S','V','Q','3');
2464 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2466 if (codec_id == AV_CODEC_ID_NONE)
2467 av_log(matroska->ctx, AV_LOG_ERROR,
2468 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2469 if (track->codec_priv.size >= 86) {
2470 bit_depth = AV_RB16(track->codec_priv.data + 82);
2471 ffio_init_context(&b, track->codec_priv.data,
2472 track->codec_priv.size,
2473 0, NULL, NULL, NULL, NULL);
2474 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2476 track->has_palette = 1;
2479 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2480 switch (track->audio.bitdepth) {
2482 codec_id = AV_CODEC_ID_PCM_U8;
2485 codec_id = AV_CODEC_ID_PCM_S24BE;
2488 codec_id = AV_CODEC_ID_PCM_S32BE;
2491 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2492 switch (track->audio.bitdepth) {
2494 codec_id = AV_CODEC_ID_PCM_U8;
2497 codec_id = AV_CODEC_ID_PCM_S24LE;
2500 codec_id = AV_CODEC_ID_PCM_S32LE;
2503 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2504 track->audio.bitdepth == 64) {
2505 codec_id = AV_CODEC_ID_PCM_F64LE;
2506 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2507 int profile = matroska_aac_profile(track->codec_id);
2508 int sri = matroska_aac_sri(track->audio.samplerate);
2509 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2511 return AVERROR(ENOMEM);
2512 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2513 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2514 if (strstr(track->codec_id, "SBR")) {
2515 sri = matroska_aac_sri(track->audio.out_samplerate);
2516 extradata[2] = 0x56;
2517 extradata[3] = 0xE5;
2518 extradata[4] = 0x80 | (sri << 3);
2522 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2523 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2524 * Create the "atom size", "tag", and "tag version" fields the
2525 * decoder expects manually. */
2526 extradata_size = 12 + track->codec_priv.size;
2527 extradata = av_mallocz(extradata_size +
2528 AV_INPUT_BUFFER_PADDING_SIZE);
2530 return AVERROR(ENOMEM);
2531 AV_WB32(extradata, extradata_size);
2532 memcpy(&extradata[4], "alac", 4);
2533 AV_WB32(&extradata[8], 0);
2534 memcpy(&extradata[12], track->codec_priv.data,
2535 track->codec_priv.size);
2536 } else if (codec_id == AV_CODEC_ID_TTA) {
2538 if (track->audio.channels > UINT16_MAX ||
2539 track->audio.bitdepth > UINT16_MAX) {
2540 av_log(matroska->ctx, AV_LOG_WARNING,
2541 "Too large audio channel number %"PRIu64
2542 " or bitdepth %"PRIu64". Skipping track.\n",
2543 track->audio.channels, track->audio.bitdepth);
2544 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2545 return AVERROR_INVALIDDATA;
2549 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2550 return AVERROR_INVALIDDATA;
2551 extradata_size = 22;
2552 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2554 return AVERROR(ENOMEM);
2556 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2557 bytestream_put_le16(&ptr, 1);
2558 bytestream_put_le16(&ptr, track->audio.channels);
2559 bytestream_put_le16(&ptr, track->audio.bitdepth);
2560 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2561 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2562 track->audio.out_samplerate,
2563 AV_TIME_BASE * 1000));
2564 } else if (codec_id == AV_CODEC_ID_RV10 ||
2565 codec_id == AV_CODEC_ID_RV20 ||
2566 codec_id == AV_CODEC_ID_RV30 ||
2567 codec_id == AV_CODEC_ID_RV40) {
2568 extradata_offset = 26;
2569 } else if (codec_id == AV_CODEC_ID_RA_144) {
2570 track->audio.out_samplerate = 8000;
2571 track->audio.channels = 1;
2572 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2573 codec_id == AV_CODEC_ID_COOK ||
2574 codec_id == AV_CODEC_ID_ATRAC3 ||
2575 codec_id == AV_CODEC_ID_SIPR)
2576 && track->codec_priv.data) {
2579 ffio_init_context(&b, track->codec_priv.data,
2580 track->codec_priv.size,
2581 0, NULL, NULL, NULL, NULL);
2583 flavor = avio_rb16(&b);
2584 track->audio.coded_framesize = avio_rb32(&b);
2586 track->audio.sub_packet_h = avio_rb16(&b);
2587 track->audio.frame_size = avio_rb16(&b);
2588 track->audio.sub_packet_size = avio_rb16(&b);
2590 track->audio.coded_framesize <= 0 ||
2591 track->audio.sub_packet_h <= 0 ||
2592 track->audio.frame_size <= 0 ||
2593 track->audio.sub_packet_size <= 0 && codec_id != AV_CODEC_ID_SIPR)
2594 return AVERROR_INVALIDDATA;
2595 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2596 track->audio.frame_size);
2597 if (!track->audio.buf)
2598 return AVERROR(ENOMEM);
2599 if (codec_id == AV_CODEC_ID_RA_288) {
2600 st->codecpar->block_align = track->audio.coded_framesize;
2601 track->codec_priv.size = 0;
2603 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2604 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2605 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2606 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2608 st->codecpar->block_align = track->audio.sub_packet_size;
2609 extradata_offset = 78;
2611 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2612 ret = matroska_parse_flac(s, track, &extradata_offset);
2615 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2616 fourcc = AV_RL32(track->codec_priv.data);
2617 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2618 /* we don't need any value stored in CodecPrivate.
2619 make sure that it's not exported as extradata. */
2620 track->codec_priv.size = 0;
2621 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2622 /* For now, propagate only the OBUs, if any. Once libavcodec is
2623 updated to handle isobmff style extradata this can be removed. */
2624 extradata_offset = 4;
2626 track->codec_priv.size -= extradata_offset;
2628 if (codec_id == AV_CODEC_ID_NONE)
2629 av_log(matroska->ctx, AV_LOG_INFO,
2630 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2632 if (track->time_scale < 0.01)
2633 track->time_scale = 1.0;
2634 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2635 1000 * 1000 * 1000); /* 64 bit pts in ns */
2637 /* convert the delay from ns to the track timebase */
2638 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2639 (AVRational){ 1, 1000000000 },
2642 st->codecpar->codec_id = codec_id;
2644 if (strcmp(track->language, "und"))
2645 av_dict_set(&st->metadata, "language", track->language, 0);
2646 av_dict_set(&st->metadata, "title", track->name, 0);
2648 if (track->flag_default)
2649 st->disposition |= AV_DISPOSITION_DEFAULT;
2650 if (track->flag_forced)
2651 st->disposition |= AV_DISPOSITION_FORCED;
2653 if (!st->codecpar->extradata) {
2655 st->codecpar->extradata = extradata;
2656 st->codecpar->extradata_size = extradata_size;
2657 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2658 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2659 return AVERROR(ENOMEM);
2660 memcpy(st->codecpar->extradata,
2661 track->codec_priv.data + extradata_offset,
2662 track->codec_priv.size);
2666 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2667 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2668 int display_width_mul = 1;
2669 int display_height_mul = 1;
2671 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2672 st->codecpar->codec_tag = fourcc;
2674 st->codecpar->bits_per_coded_sample = bit_depth;
2675 st->codecpar->width = track->video.pixel_width;
2676 st->codecpar->height = track->video.pixel_height;
2678 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2679 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2680 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2681 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2683 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2684 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2686 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2687 av_reduce(&st->sample_aspect_ratio.num,
2688 &st->sample_aspect_ratio.den,
2689 st->codecpar->height * track->video.display_width * display_width_mul,
2690 st->codecpar->width * track->video.display_height * display_height_mul,
2693 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2694 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2696 if (track->default_duration) {
2697 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2698 1000000000, track->default_duration, 30000);
2699 #if FF_API_R_FRAME_RATE
2700 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2701 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2702 st->r_frame_rate = st->avg_frame_rate;
2706 /* export stereo mode flag as metadata tag */
2707 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2708 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2710 /* export alpha mode flag as metadata tag */
2711 if (track->video.alpha_mode)
2712 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2714 /* if we have virtual track, mark the real tracks */
2715 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2717 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2719 snprintf(buf, sizeof(buf), "%s_%d",
2720 ff_matroska_video_stereo_plane[planes[j].type], i);
2721 for (k=0; k < matroska->tracks.nb_elem; k++)
2722 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2723 av_dict_set(&tracks[k].stream->metadata,
2724 "stereo_mode", buf, 0);
2728 // add stream level stereo3d side data if it is a supported format
2729 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2730 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2731 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2736 ret = mkv_parse_video_color(st, track);
2739 ret = mkv_parse_video_projection(st, track);
2742 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2743 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2744 st->codecpar->codec_tag = fourcc;
2745 st->codecpar->sample_rate = track->audio.out_samplerate;
2746 st->codecpar->channels = track->audio.channels;
2747 if (!st->codecpar->bits_per_coded_sample)
2748 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2749 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2750 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2751 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2752 st->need_parsing = AVSTREAM_PARSE_FULL;
2753 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2754 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2755 if (track->codec_delay > 0) {
2756 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2757 (AVRational){1, 1000000000},
2758 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2759 48000 : st->codecpar->sample_rate});
2761 if (track->seek_preroll > 0) {
2762 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2763 (AVRational){1, 1000000000},
2764 (AVRational){1, st->codecpar->sample_rate});
2766 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2767 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2769 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2770 st->disposition |= AV_DISPOSITION_CAPTIONS;
2771 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2772 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2773 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2774 st->disposition |= AV_DISPOSITION_METADATA;
2776 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2777 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2784 static int matroska_read_header(AVFormatContext *s)
2786 MatroskaDemuxContext *matroska = s->priv_data;
2787 EbmlList *attachments_list = &matroska->attachments;
2788 EbmlList *chapters_list = &matroska->chapters;
2789 MatroskaAttachment *attachments;
2790 MatroskaChapter *chapters;
2791 uint64_t max_start = 0;
2797 matroska->cues_parsing_deferred = 1;
2799 /* First read the EBML header. */
2800 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2801 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2802 ebml_free(ebml_syntax, &ebml);
2803 return AVERROR_INVALIDDATA;
2805 if (ebml.version > EBML_VERSION ||
2806 ebml.max_size > sizeof(uint64_t) ||
2807 ebml.id_length > sizeof(uint32_t) ||
2808 ebml.doctype_version > 3) {
2809 avpriv_report_missing_feature(matroska->ctx,
2810 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2811 ebml.version, ebml.doctype, ebml.doctype_version);
2812 ebml_free(ebml_syntax, &ebml);
2813 return AVERROR_PATCHWELCOME;
2814 } else if (ebml.doctype_version == 3) {
2815 av_log(matroska->ctx, AV_LOG_WARNING,
2816 "EBML header using unsupported features\n"
2817 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2818 ebml.version, ebml.doctype, ebml.doctype_version);
2820 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2821 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2823 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2824 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2825 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2826 ebml_free(ebml_syntax, &ebml);
2827 return AVERROR_INVALIDDATA;
2830 ebml_free(ebml_syntax, &ebml);
2832 /* The next thing is a segment. */
2833 pos = avio_tell(matroska->ctx->pb);
2834 res = ebml_parse(matroska, matroska_segments, matroska);
2835 // Try resyncing until we find an EBML_STOP type element.
2837 res = matroska_resync(matroska, pos);
2840 pos = avio_tell(matroska->ctx->pb);
2841 res = ebml_parse(matroska, matroska_segment, matroska);
2843 /* Set data_offset as it might be needed later by seek_frame_generic. */
2844 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2845 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2846 matroska_execute_seekhead(matroska);
2848 if (!matroska->time_scale)
2849 matroska->time_scale = 1000000;
2850 if (matroska->duration)
2851 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2852 1000 / AV_TIME_BASE;
2853 av_dict_set(&s->metadata, "title", matroska->title, 0);
2854 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2856 if (matroska->date_utc.size == 8)
2857 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2859 res = matroska_parse_tracks(s);
2863 attachments = attachments_list->elem;
2864 for (j = 0; j < attachments_list->nb_elem; j++) {
2865 if (!(attachments[j].filename && attachments[j].mime &&
2866 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2867 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2869 AVStream *st = avformat_new_stream(s, NULL);
2872 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2873 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2874 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2876 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2877 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2878 strlen(ff_mkv_image_mime_tags[i].str))) {
2879 st->codecpar->codec_id = ff_mkv_image_mime_tags[i].id;
2884 attachments[j].stream = st;
2886 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2887 AVPacket *pkt = &st->attached_pic;
2889 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2890 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2892 av_init_packet(pkt);
2893 pkt->buf = av_buffer_ref(attachments[j].bin.buf);
2895 return AVERROR(ENOMEM);
2896 pkt->data = attachments[j].bin.data;
2897 pkt->size = attachments[j].bin.size;
2898 pkt->stream_index = st->index;
2899 pkt->flags |= AV_PKT_FLAG_KEY;
2901 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2902 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2904 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2905 attachments[j].bin.size);
2907 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2908 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2909 strlen(ff_mkv_mime_tags[i].str))) {
2910 st->codecpar->codec_id = ff_mkv_mime_tags[i].id;
2918 chapters = chapters_list->elem;
2919 for (i = 0; i < chapters_list->nb_elem; i++)
2920 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2921 (max_start == 0 || chapters[i].start > max_start)) {
2922 chapters[i].chapter =
2923 avpriv_new_chapter(s, chapters[i].uid,
2924 (AVRational) { 1, 1000000000 },
2925 chapters[i].start, chapters[i].end,
2927 if (chapters[i].chapter) {
2928 av_dict_set(&chapters[i].chapter->metadata,
2929 "title", chapters[i].title, 0);
2931 max_start = chapters[i].start;
2934 matroska_add_index_entries(matroska);
2936 matroska_convert_tags(s);
2940 matroska_read_close(s);
2945 * Put one packet in an application-supplied AVPacket struct.
2946 * Returns 0 on success or -1 on failure.
2948 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2951 if (matroska->queue) {
2952 MatroskaTrack *tracks = matroska->tracks.elem;
2953 MatroskaTrack *track;
2955 ff_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
2956 track = &tracks[pkt->stream_index];
2957 if (track->has_palette) {
2958 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2960 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
2962 memcpy(pal, track->palette, AVPALETTE_SIZE);
2964 track->has_palette = 0;
2973 * Free all packets in our internal queue.
2975 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2977 ff_packet_list_free(&matroska->queue, &matroska->queue_end);
2980 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2981 int size, int type, AVIOContext *pb,
2982 uint32_t lace_size[256], int *laces)
2985 uint8_t *data = *buf;
2989 lace_size[0] = size;
2993 av_assert0(size > 0);
2999 case 0x1: /* Xiph lacing */
3003 for (n = 0; n < *laces - 1; n++) {
3007 if (size <= total) {
3008 return AVERROR_INVALIDDATA;
3012 lace_size[n] += temp;
3019 if (size <= total) {
3020 return AVERROR_INVALIDDATA;
3023 lace_size[n] = size - total;
3027 case 0x2: /* fixed-size lacing */
3028 if (size % (*laces)) {
3029 return AVERROR_INVALIDDATA;
3031 for (n = 0; n < *laces; n++)
3032 lace_size[n] = size / *laces;
3035 case 0x3: /* EBML lacing */
3043 n = ebml_read_num(matroska, pb, 8, &num, 1);
3047 return AVERROR_INVALIDDATA;
3049 total = lace_size[0] = num;
3051 for (n = 1; n < *laces - 1; n++) {
3054 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3057 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3058 return AVERROR_INVALIDDATA;
3060 lace_size[n] = lace_size[n - 1] + snum;
3061 total += lace_size[n];
3066 if (size <= total) {
3067 return AVERROR_INVALIDDATA;
3069 lace_size[*laces - 1] = size - total;
3079 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3080 MatroskaTrack *track, AVStream *st,
3081 uint8_t *data, int size, uint64_t timecode,
3084 int a = st->codecpar->block_align;
3085 int sps = track->audio.sub_packet_size;
3086 int cfs = track->audio.coded_framesize;
3087 int h = track->audio.sub_packet_h;
3088 int y = track->audio.sub_packet_cnt;
3089 int w = track->audio.frame_size;
3092 if (!track->audio.pkt_cnt) {
3093 if (track->audio.sub_packet_cnt == 0)
3094 track->audio.buf_timecode = timecode;
3095 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3096 if (size < cfs * h / 2) {
3097 av_log(matroska->ctx, AV_LOG_ERROR,
3098 "Corrupt int4 RM-style audio packet size\n");
3099 return AVERROR_INVALIDDATA;
3101 for (x = 0; x < h / 2; x++)
3102 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3103 data + x * cfs, cfs);
3104 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3106 av_log(matroska->ctx, AV_LOG_ERROR,
3107 "Corrupt sipr RM-style audio packet size\n");
3108 return AVERROR_INVALIDDATA;
3110 memcpy(track->audio.buf + y * w, data, w);
3112 if (size < sps * w / sps || h<=0 || w%sps) {
3113 av_log(matroska->ctx, AV_LOG_ERROR,
3114 "Corrupt generic RM-style audio packet size\n");
3115 return AVERROR_INVALIDDATA;
3117 for (x = 0; x < w / sps; x++)
3118 memcpy(track->audio.buf +
3119 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3120 data + x * sps, sps);
3123 if (++track->audio.sub_packet_cnt >= h) {
3124 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3125 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3126 track->audio.sub_packet_cnt = 0;
3127 track->audio.pkt_cnt = h * w / a;
3131 while (track->audio.pkt_cnt) {
3133 AVPacket pktl, *pkt = &pktl;
3135 ret = av_new_packet(pkt, a);
3140 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3142 pkt->pts = track->audio.buf_timecode;
3143 track->audio.buf_timecode = AV_NOPTS_VALUE;
3145 pkt->stream_index = st->index;
3146 ret = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3148 av_packet_unref(pkt);
3149 return AVERROR(ENOMEM);
3156 /* reconstruct full wavpack blocks from mangled matroska ones */
3157 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
3158 uint8_t **pdst, int *size)
3160 uint8_t *dst = NULL;
3165 int ret, offset = 0;
3167 if (srclen < 12 || track->stream->codecpar->extradata_size < 2)
3168 return AVERROR_INVALIDDATA;
3170 ver = AV_RL16(track->stream->codecpar->extradata);
3172 samples = AV_RL32(src);
3176 while (srclen >= 8) {
3181 uint32_t flags = AV_RL32(src);
3182 uint32_t crc = AV_RL32(src + 4);
3186 multiblock = (flags & 0x1800) != 0x1800;
3189 ret = AVERROR_INVALIDDATA;
3192 blocksize = AV_RL32(src);
3198 if (blocksize > srclen) {
3199 ret = AVERROR_INVALIDDATA;
3203 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3205 ret = AVERROR(ENOMEM);
3209 dstlen += blocksize + 32;
3211 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3212 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3213 AV_WL16(dst + offset + 8, ver); // version
3214 AV_WL16(dst + offset + 10, 0); // track/index_no
3215 AV_WL32(dst + offset + 12, 0); // total samples
3216 AV_WL32(dst + offset + 16, 0); // block index
3217 AV_WL32(dst + offset + 20, samples); // number of samples
3218 AV_WL32(dst + offset + 24, flags); // flags
3219 AV_WL32(dst + offset + 28, crc); // crc
3220 memcpy(dst + offset + 32, src, blocksize); // block data
3223 srclen -= blocksize;
3224 offset += blocksize + 32;
3227 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3239 static int matroska_parse_prores(MatroskaTrack *track, uint8_t *src,
3240 uint8_t **pdst, int *size)
3243 int dstlen = *size + 8;
3245 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3247 return AVERROR(ENOMEM);
3249 AV_WB32(dst, dstlen);
3250 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3251 memcpy(dst + 8, src, dstlen - 8);
3252 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3260 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3261 MatroskaTrack *track,
3263 uint8_t *data, int data_len,
3268 AVPacket pktl, *pkt = &pktl;
3269 uint8_t *id, *settings, *text, *buf;
3270 int id_len, settings_len, text_len;
3275 return AVERROR_INVALIDDATA;
3278 q = data + data_len;
3283 if (*p == '\r' || *p == '\n') {
3292 if (p >= q || *p != '\n')
3293 return AVERROR_INVALIDDATA;
3299 if (*p == '\r' || *p == '\n') {
3300 settings_len = p - settings;
3308 if (p >= q || *p != '\n')
3309 return AVERROR_INVALIDDATA;
3314 while (text_len > 0) {
3315 const int len = text_len - 1;
3316 const uint8_t c = p[len];
3317 if (c != '\r' && c != '\n')
3323 return AVERROR_INVALIDDATA;
3325 err = av_new_packet(pkt, text_len);
3330 memcpy(pkt->data, text, text_len);
3333 buf = av_packet_new_side_data(pkt,
3334 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3337 av_packet_unref(pkt);
3338 return AVERROR(ENOMEM);
3340 memcpy(buf, id, id_len);
3343 if (settings_len > 0) {
3344 buf = av_packet_new_side_data(pkt,
3345 AV_PKT_DATA_WEBVTT_SETTINGS,
3348 av_packet_unref(pkt);
3349 return AVERROR(ENOMEM);
3351 memcpy(buf, settings, settings_len);
3354 // Do we need this for subtitles?
3355 // pkt->flags = AV_PKT_FLAG_KEY;
3357 pkt->stream_index = st->index;
3358 pkt->pts = timecode;
3360 // Do we need this for subtitles?
3361 // pkt->dts = timecode;
3363 pkt->duration = duration;
3366 err = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3368 av_packet_unref(pkt);
3369 return AVERROR(ENOMEM);
3375 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3376 MatroskaTrack *track, AVStream *st,
3377 AVBufferRef *buf, uint8_t *data, int pkt_size,
3378 uint64_t timecode, uint64_t lace_duration,
3379 int64_t pos, int is_keyframe,
3380 uint8_t *additional, uint64_t additional_id, int additional_size,
3381 int64_t discard_padding)
3383 MatroskaTrackEncoding *encodings = track->encodings.elem;
3384 uint8_t *pkt_data = data;
3386 AVPacket pktl, *pkt = &pktl;
3388 if (encodings && !encodings->type && encodings->scope & 1) {
3389 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
3394 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3396 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
3398 av_log(matroska->ctx, AV_LOG_ERROR,
3399 "Error parsing a wavpack block.\n");
3402 if (pkt_data != data)
3403 av_freep(&pkt_data);
3407 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3408 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3410 res = matroska_parse_prores(track, pkt_data, &pr_data, &pkt_size);
3412 av_log(matroska->ctx, AV_LOG_ERROR,
3413 "Error parsing a prores block.\n");
3416 if (pkt_data != data)
3417 av_freep(&pkt_data);
3421 av_init_packet(pkt);
3422 if (pkt_data != data)
3423 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3426 pkt->buf = av_buffer_ref(buf);
3429 res = AVERROR(ENOMEM);
3433 pkt->data = pkt_data;
3434 pkt->size = pkt_size;
3435 pkt->flags = is_keyframe;
3436 pkt->stream_index = st->index;
3438 if (additional_size > 0) {
3439 uint8_t *side_data = av_packet_new_side_data(pkt,
3440 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3441 additional_size + 8);
3443 av_packet_unref(pkt);
3444 return AVERROR(ENOMEM);
3446 AV_WB64(side_data, additional_id);
3447 memcpy(side_data + 8, additional, additional_size);
3450 if (discard_padding) {
3451 uint8_t *side_data = av_packet_new_side_data(pkt,
3452 AV_PKT_DATA_SKIP_SAMPLES,
3455 av_packet_unref(pkt);
3456 return AVERROR(ENOMEM);
3458 discard_padding = av_rescale_q(discard_padding,
3459 (AVRational){1, 1000000000},
3460 (AVRational){1, st->codecpar->sample_rate});
3461 if (discard_padding > 0) {
3462 AV_WL32(side_data + 4, discard_padding);
3464 AV_WL32(side_data, -discard_padding);
3468 if (track->ms_compat)
3469 pkt->dts = timecode;
3471 pkt->pts = timecode;
3473 pkt->duration = lace_duration;
3475 #if FF_API_CONVERGENCE_DURATION
3476 FF_DISABLE_DEPRECATION_WARNINGS
3477 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3478 pkt->convergence_duration = lace_duration;
3480 FF_ENABLE_DEPRECATION_WARNINGS
3483 res = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3485 av_packet_unref(pkt);
3486 return AVERROR(ENOMEM);
3492 if (pkt_data != data)
3493 av_freep(&pkt_data);
3497 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3498 int size, int64_t pos, uint64_t cluster_time,
3499 uint64_t block_duration, int is_keyframe,
3500 uint8_t *additional, uint64_t additional_id, int additional_size,
3501 int64_t cluster_pos, int64_t discard_padding)
3503 uint64_t timecode = AV_NOPTS_VALUE;
3504 MatroskaTrack *track;
3509 uint32_t lace_size[256];
3510 int n, flags, laces = 0;
3512 int trust_default_duration = 1;
3514 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3516 if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3521 track = matroska_find_track_by_num(matroska, num);
3522 if (!track || !track->stream) {
3523 av_log(matroska->ctx, AV_LOG_INFO,
3524 "Invalid stream %"PRIu64"\n", num);
3525 return AVERROR_INVALIDDATA;
3526 } else if (size <= 3)
3529 if (st->discard >= AVDISCARD_ALL)
3531 av_assert1(block_duration != AV_NOPTS_VALUE);
3533 block_time = sign_extend(AV_RB16(data), 16);
3537 if (is_keyframe == -1)
3538 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3540 if (cluster_time != (uint64_t) -1 &&
3541 (block_time >= 0 || cluster_time >= -block_time)) {
3542 timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3543 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3544 timecode < track->end_timecode)
3545 is_keyframe = 0; /* overlapping subtitles are not key frame */
3547 ff_reduce_index(matroska->ctx, st->index);
3548 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3553 if (matroska->skip_to_keyframe &&
3554 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3555 // Compare signed timecodes. Timecode may be negative due to codec delay
3556 // offset. We don't support timestamps greater than int64_t anyway - see
3558 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3561 matroska->skip_to_keyframe = 0;
3562 else if (!st->skip_to_keyframe) {
3563 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3564 matroska->skip_to_keyframe = 0;
3568 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3569 &pb, lace_size, &laces);
3571 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3575 if (track->audio.samplerate == 8000) {
3576 // If this is needed for more codecs, then add them here
3577 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3578 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3579 trust_default_duration = 0;
3583 if (!block_duration && trust_default_duration)
3584 block_duration = track->default_duration * laces / matroska->time_scale;
3586 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3587 track->end_timecode =
3588 FFMAX(track->end_timecode, timecode + block_duration);
3590 for (n = 0; n < laces; n++) {
3591 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3593 if ((st->codecpar->codec_id == AV_CODEC_ID_RA_288 ||
3594 st->codecpar->codec_id == AV_CODEC_ID_COOK ||
3595 st->codecpar->codec_id == AV_CODEC_ID_SIPR ||
3596 st->codecpar->codec_id == AV_CODEC_ID_ATRAC3) &&
3597 st->codecpar->block_align && track->audio.sub_packet_size) {
3598 res = matroska_parse_rm_audio(matroska, track, st, data,
3604 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3605 res = matroska_parse_webvtt(matroska, track, st,
3607 timecode, lace_duration,
3612 res = matroska_parse_frame(matroska, track, st, buf, data, lace_size[n],
3613 timecode, lace_duration, pos,
3614 !n ? is_keyframe : 0,
3615 additional, additional_id, additional_size,
3621 if (timecode != AV_NOPTS_VALUE)
3622 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3623 data += lace_size[n];
3629 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3631 MatroskaCluster *cluster = &matroska->current_cluster;
3632 MatroskaBlock *block = &cluster->block;
3635 av_assert0(matroska->num_levels <= 2);
3637 if (matroska->num_levels == 1) {
3638 res = ebml_parse(matroska, matroska_segment, NULL);
3641 /* Found a cluster: subtract the size of the ID already read. */
3642 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3644 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3650 if (matroska->num_levels == 2) {
3651 /* We are inside a cluster. */
3652 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3654 if (res >= 0 && block->bin.size > 0) {
3655 int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3656 uint8_t* additional = block->additional.size > 0 ?
3657 block->additional.data : NULL;
3659 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3660 block->bin.size, block->bin.pos,
3661 cluster->timecode, block->duration,
3662 is_keyframe, additional, block->additional_id,
3663 block->additional.size, cluster->pos,
3664 block->discard_padding);
3667 ebml_free(matroska_blockgroup, block);
3668 memset(block, 0, sizeof(*block));
3669 } else if (!matroska->num_levels) {
3670 if (!avio_feof(matroska->ctx->pb)) {
3671 avio_r8(matroska->ctx->pb);
3672 if (!avio_feof(matroska->ctx->pb)) {
3673 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3674 "end of segment.\n");
3675 return AVERROR_INVALIDDATA;
3685 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3687 MatroskaDemuxContext *matroska = s->priv_data;
3690 if (matroska->resync_pos == -1) {
3691 // This can only happen if generic seeking has been used.
3692 matroska->resync_pos = avio_tell(s->pb);
3695 while (matroska_deliver_packet(matroska, pkt)) {
3697 return (ret < 0) ? ret : AVERROR_EOF;
3698 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3699 ret = matroska_resync(matroska, matroska->resync_pos);
3705 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3706 int64_t timestamp, int flags)
3708 MatroskaDemuxContext *matroska = s->priv_data;
3709 MatroskaTrack *tracks = NULL;
3710 AVStream *st = s->streams[stream_index];
3713 /* Parse the CUES now since we need the index data to seek. */
3714 if (matroska->cues_parsing_deferred > 0) {
3715 matroska->cues_parsing_deferred = 0;
3716 matroska_parse_cues(matroska);
3719 if (!st->nb_index_entries)
3721 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3723 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3724 matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3725 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3726 matroska_clear_queue(matroska);
3727 if (matroska_parse_cluster(matroska) < 0)
3732 matroska_clear_queue(matroska);
3733 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3736 tracks = matroska->tracks.elem;
3737 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3738 tracks[i].audio.pkt_cnt = 0;
3739 tracks[i].audio.sub_packet_cnt = 0;
3740 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3741 tracks[i].end_timecode = 0;
3744 /* We seek to a level 1 element, so set the appropriate status. */
3745 matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3746 if (flags & AVSEEK_FLAG_ANY) {
3747 st->skip_to_keyframe = 0;
3748 matroska->skip_to_timecode = timestamp;
3750 st->skip_to_keyframe = 1;
3751 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3753 matroska->skip_to_keyframe = 1;
3755 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3758 // slightly hackish but allows proper fallback to
3759 // the generic seeking code.
3760 matroska_reset_status(matroska, 0, -1);
3761 matroska->resync_pos = -1;
3762 matroska_clear_queue(matroska);
3763 st->skip_to_keyframe =
3764 matroska->skip_to_keyframe = 0;
3769 static int matroska_read_close(AVFormatContext *s)
3771 MatroskaDemuxContext *matroska = s->priv_data;
3772 MatroskaTrack *tracks = matroska->tracks.elem;
3775 matroska_clear_queue(matroska);
3777 for (n = 0; n < matroska->tracks.nb_elem; n++)
3778 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3779 av_freep(&tracks[n].audio.buf);
3780 ebml_free(matroska_segment, matroska);
3786 int64_t start_time_ns;
3787 int64_t end_time_ns;
3788 int64_t start_offset;
3792 /* This function searches all the Cues and returns the CueDesc corresponding to
3793 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3794 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3796 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3797 MatroskaDemuxContext *matroska = s->priv_data;
3800 int nb_index_entries = s->streams[0]->nb_index_entries;
3801 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3802 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3803 for (i = 1; i < nb_index_entries; i++) {
3804 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3805 index_entries[i].timestamp * matroska->time_scale > ts) {
3810 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3811 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3812 if (i != nb_index_entries - 1) {
3813 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3814 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3816 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3817 // FIXME: this needs special handling for files where Cues appear
3818 // before Clusters. the current logic assumes Cues appear after
3820 cue_desc.end_offset = cues_start - matroska->segment_start;
3825 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3827 MatroskaDemuxContext *matroska = s->priv_data;
3828 uint32_t id = matroska->current_id;
3829 int64_t cluster_pos, before_pos;
3831 if (s->streams[0]->nb_index_entries <= 0) return 0;
3832 // seek to the first cluster using cues.
3833 index = av_index_search_timestamp(s->streams[0], 0, 0);
3834 if (index < 0) return 0;
3835 cluster_pos = s->streams[0]->index_entries[index].pos;
3836 before_pos = avio_tell(s->pb);
3838 uint64_t cluster_id, cluster_length;
3841 avio_seek(s->pb, cluster_pos, SEEK_SET);
3842 // read cluster id and length
3843 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3844 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3846 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3850 matroska_reset_status(matroska, 0, cluster_pos);
3851 matroska_clear_queue(matroska);
3852 if (matroska_parse_cluster(matroska) < 0 ||
3856 pkt = &matroska->queue->pkt;
3857 // 4 + read is the length of the cluster id and the cluster length field.
3858 cluster_pos += 4 + read + cluster_length;
3859 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3865 /* Restore the status after matroska_read_header: */
3866 matroska_reset_status(matroska, id, before_pos);
3871 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3872 double min_buffer, double* buffer,
3873 double* sec_to_download, AVFormatContext *s,
3876 double nano_seconds_per_second = 1000000000.0;
3877 double time_sec = time_ns / nano_seconds_per_second;
3879 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3880 int64_t end_time_ns = time_ns + time_to_search_ns;
3881 double sec_downloaded = 0.0;
3882 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3883 if (desc_curr.start_time_ns == -1)
3885 *sec_to_download = 0.0;
3887 // Check for non cue start time.
3888 if (time_ns > desc_curr.start_time_ns) {
3889 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3890 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3891 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3892 double timeToDownload = (cueBytes * 8.0) / bps;
3894 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3895 *sec_to_download += timeToDownload;
3897 // Check if the search ends within the first cue.
3898 if (desc_curr.end_time_ns >= end_time_ns) {
3899 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3900 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3901 sec_downloaded = percent_to_sub * sec_downloaded;
3902 *sec_to_download = percent_to_sub * *sec_to_download;
3905 if ((sec_downloaded + *buffer) <= min_buffer) {
3909 // Get the next Cue.
3910 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3913 while (desc_curr.start_time_ns != -1) {
3914 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3915 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3916 double desc_sec = desc_ns / nano_seconds_per_second;
3917 double bits = (desc_bytes * 8.0);
3918 double time_to_download = bits / bps;
3920 sec_downloaded += desc_sec - time_to_download;
3921 *sec_to_download += time_to_download;
3923 if (desc_curr.end_time_ns >= end_time_ns) {
3924 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3925 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3926 sec_downloaded = percent_to_sub * sec_downloaded;
3927 *sec_to_download = percent_to_sub * *sec_to_download;
3929 if ((sec_downloaded + *buffer) <= min_buffer)
3934 if ((sec_downloaded + *buffer) <= min_buffer) {
3939 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3941 *buffer = *buffer + sec_downloaded;
3945 /* This function computes the bandwidth of the WebM file with the help of
3946 * buffer_size_after_time_downloaded() function. Both of these functions are
3947 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3948 * Matroska parsing mechanism.
3950 * Returns the bandwidth of the file on success; -1 on error.
3952 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3954 MatroskaDemuxContext *matroska = s->priv_data;
3955 AVStream *st = s->streams[0];
3956 double bandwidth = 0.0;
3959 for (i = 0; i < st->nb_index_entries; i++) {
3960 int64_t prebuffer_ns = 1000000000;
3961 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3962 double nano_seconds_per_second = 1000000000.0;
3963 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3964 double prebuffer_bytes = 0.0;
3965 int64_t temp_prebuffer_ns = prebuffer_ns;
3966 int64_t pre_bytes, pre_ns;
3967 double pre_sec, prebuffer, bits_per_second;
3968 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3970 // Start with the first Cue.
3971 CueDesc desc_end = desc_beg;
3973 // Figure out how much data we have downloaded for the prebuffer. This will
3974 // be used later to adjust the bits per sample to try.
3975 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3976 // Prebuffered the entire Cue.
3977 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3978 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3979 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3981 if (desc_end.start_time_ns == -1) {
3982 // The prebuffer is larger than the duration.
3983 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3985 bits_per_second = 0.0;
3987 // The prebuffer ends in the last Cue. Estimate how much data was
3989 pre_bytes = desc_end.end_offset - desc_end.start_offset;
3990 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3991 pre_sec = pre_ns / nano_seconds_per_second;
3993 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
3995 prebuffer = prebuffer_ns / nano_seconds_per_second;
3997 // Set this to 0.0 in case our prebuffer buffers the entire video.
3998 bits_per_second = 0.0;
4000 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4001 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4002 double desc_sec = desc_ns / nano_seconds_per_second;
4003 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4005 // Drop the bps by the percentage of bytes buffered.
4006 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4007 double mod_bits_per_second = calc_bits_per_second * percent;
4009 if (prebuffer < desc_sec) {
4011 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4013 // Add 1 so the bits per second should be a little bit greater than file
4015 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4016 const double min_buffer = 0.0;
4017 double buffer = prebuffer;
4018 double sec_to_download = 0.0;
4020 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4021 min_buffer, &buffer, &sec_to_download,
4025 } else if (rv == 0) {
4026 bits_per_second = (double)(bps);
4031 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4032 } while (desc_end.start_time_ns != -1);
4034 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4036 return (int64_t)bandwidth;
4039 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4041 MatroskaDemuxContext *matroska = s->priv_data;
4042 EbmlList *seekhead_list = &matroska->seekhead;
4043 MatroskaSeekhead *seekhead = seekhead_list->elem;
4045 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4049 // determine cues start and end positions
4050 for (i = 0; i < seekhead_list->nb_elem; i++)
4051 if (seekhead[i].id == MATROSKA_ID_CUES)
4054 if (i >= seekhead_list->nb_elem) return -1;
4056 before_pos = avio_tell(matroska->ctx->pb);
4057 cues_start = seekhead[i].pos + matroska->segment_start;
4058 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4059 // cues_end is computed as cues_start + cues_length + length of the
4060 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4061 // cues_end is inclusive and the above sum is reduced by 1.
4062 uint64_t cues_length, cues_id;
4064 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4065 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4066 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4067 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4070 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4072 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4073 if (cues_start == -1 || cues_end == -1) return -1;
4076 matroska_parse_cues(matroska);
4079 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4082 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4084 // if the file has cues at the start, fix up the init range so that
4085 // it does not include it
4086 if (cues_start <= init_range)
4087 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4090 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4091 if (bandwidth < 0) return -1;
4092 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4094 // check if all clusters start with key frames
4095 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4097 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4098 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4099 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4100 if (!buf) return -1;
4102 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4103 int ret = snprintf(buf + end, 20,
4104 "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4105 i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4106 if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4107 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4109 return AVERROR_INVALIDDATA;
4113 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
4119 static int webm_dash_manifest_read_header(AVFormatContext *s)
4122 int ret = matroska_read_header(s);
4124 MatroskaTrack *tracks;
4125 MatroskaDemuxContext *matroska = s->priv_data;
4127 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4130 if (!s->nb_streams) {
4131 matroska_read_close(s);
4132 av_log(s, AV_LOG_ERROR, "No streams found\n");
4133 return AVERROR_INVALIDDATA;
4136 if (!matroska->is_live) {
4137 buf = av_asprintf("%g", matroska->duration);
4138 if (!buf) return AVERROR(ENOMEM);
4139 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
4142 // initialization range
4143 // 5 is the offset of Cluster ID.
4144 init_range = avio_tell(s->pb) - 5;
4145 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4148 // basename of the file
4149 buf = strrchr(s->url, '/');
4150 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4153 tracks = matroska->tracks.elem;
4154 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4156 // parse the cues and populate Cue related fields
4157 if (!matroska->is_live) {
4158 ret = webm_dash_manifest_cues(s, init_range);
4160 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4165 // use the bandwidth from the command line if it was provided
4166 if (matroska->bandwidth > 0) {
4167 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4168 matroska->bandwidth, 0);
4173 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4178 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4179 static const AVOption options[] = {
4180 { "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 },
4181 { "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 },
4185 static const AVClass webm_dash_class = {
4186 .class_name = "WebM DASH Manifest demuxer",
4187 .item_name = av_default_item_name,
4189 .version = LIBAVUTIL_VERSION_INT,
4192 AVInputFormat ff_matroska_demuxer = {
4193 .name = "matroska,webm",
4194 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4195 .extensions = "mkv,mk3d,mka,mks",
4196 .priv_data_size = sizeof(MatroskaDemuxContext),
4197 .read_probe = matroska_probe,
4198 .read_header = matroska_read_header,
4199 .read_packet = matroska_read_packet,
4200 .read_close = matroska_read_close,
4201 .read_seek = matroska_read_seek,
4202 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4205 AVInputFormat ff_webm_dash_manifest_demuxer = {
4206 .name = "webm_dash_manifest",
4207 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4208 .priv_data_size = sizeof(MatroskaDemuxContext),
4209 .read_header = webm_dash_manifest_read_header,
4210 .read_packet = webm_dash_manifest_read_packet,
4211 .read_close = matroska_read_close,
4212 .priv_class = &webm_dash_class,