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;
253 uint64_t max_block_additional_id;
255 uint32_t palette[AVPALETTE_COUNT];
259 typedef struct MatroskaAttachment {
266 } MatroskaAttachment;
268 typedef struct MatroskaChapter {
277 typedef struct MatroskaIndexPos {
282 typedef struct MatroskaIndex {
287 typedef struct MatroskaTag {
295 typedef struct MatroskaTagTarget {
303 typedef struct MatroskaTags {
304 MatroskaTagTarget target;
308 typedef struct MatroskaSeekhead {
313 typedef struct MatroskaLevel {
318 typedef struct MatroskaBlock {
323 uint64_t additional_id;
325 int64_t discard_padding;
328 typedef struct MatroskaCluster {
334 typedef struct MatroskaLevel1Element {
338 } MatroskaLevel1Element;
340 typedef struct MatroskaDemuxContext {
341 const AVClass *class;
342 AVFormatContext *ctx;
345 MatroskaLevel levels[EBML_MAX_DEPTH];
357 EbmlList attachments;
363 /* byte position of the segment inside the stream */
364 int64_t segment_start;
366 /* the packet queue */
368 AVPacketList *queue_end;
372 /* What to skip before effectively reading a packet. */
373 int skip_to_keyframe;
374 uint64_t skip_to_timecode;
376 /* File has a CUES element, but we defer parsing until it is needed. */
377 int cues_parsing_deferred;
379 /* Level1 elements and whether they were read yet */
380 MatroskaLevel1Element level1_elems[64];
381 int num_level1_elems;
383 MatroskaCluster current_cluster;
385 /* WebM DASH Manifest live flag */
388 /* Bandwidth value for WebM DASH Manifest */
390 } MatroskaDemuxContext;
392 #define CHILD_OF(parent) { .def = { .n = parent } }
394 // The following forward declarations need their size because
395 // a tentative definition with internal linkage must not be an
396 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
397 // Removing the sizes breaks MSVC.
398 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
399 matroska_track[27], matroska_track_encoding[6], matroska_track_encodings[2],
400 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_tracks[2],
401 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
402 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
403 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
405 static EbmlSyntax ebml_header[] = {
406 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
407 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
408 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
409 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
410 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
411 { EBML_ID_EBMLVERSION, EBML_NONE },
412 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
413 CHILD_OF(ebml_syntax)
416 static EbmlSyntax ebml_syntax[] = {
417 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
418 { MATROSKA_ID_SEGMENT, EBML_STOP },
422 static EbmlSyntax matroska_info[] = {
423 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
424 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
425 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
426 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
427 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
428 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
429 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
430 CHILD_OF(matroska_segment)
433 static EbmlSyntax matroska_mastering_meta[] = {
434 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
435 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
436 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
437 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
438 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
439 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
440 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
441 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
442 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
443 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
444 CHILD_OF(matroska_track_video_color)
447 static EbmlSyntax matroska_track_video_color[] = {
448 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
449 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=0 } },
450 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
451 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
452 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
453 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
454 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
455 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
456 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
457 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
458 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
459 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
460 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
461 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
462 CHILD_OF(matroska_track_video)
465 static EbmlSyntax matroska_track_video_projection[] = {
466 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
467 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrackVideoProjection, private) },
468 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f=0.0 } },
469 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f=0.0 } },
470 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f=0.0 } },
471 CHILD_OF(matroska_track_video)
474 static EbmlSyntax matroska_track_video[] = {
475 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
476 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
477 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
478 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
479 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
480 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
481 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
482 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
483 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
484 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
485 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
486 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
487 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
488 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
489 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
490 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
491 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
492 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
493 CHILD_OF(matroska_track)
496 static EbmlSyntax matroska_track_audio[] = {
497 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
498 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
499 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
500 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
501 CHILD_OF(matroska_track)
504 static EbmlSyntax matroska_track_encoding_compression[] = {
505 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
506 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
507 CHILD_OF(matroska_track_encoding)
510 static EbmlSyntax matroska_track_encoding_encryption[] = {
511 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
512 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
513 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
514 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
515 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
516 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
517 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
518 CHILD_OF(matroska_track_encoding)
520 static EbmlSyntax matroska_track_encoding[] = {
521 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
522 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
523 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
524 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
525 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
526 CHILD_OF(matroska_track_encodings)
529 static EbmlSyntax matroska_track_encodings[] = {
530 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
531 CHILD_OF(matroska_track)
534 static EbmlSyntax matroska_track_plane[] = {
535 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
536 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
537 CHILD_OF(matroska_track_combine_planes)
540 static EbmlSyntax matroska_track_combine_planes[] = {
541 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
542 CHILD_OF(matroska_track_operation)
545 static EbmlSyntax matroska_track_operation[] = {
546 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
547 CHILD_OF(matroska_track)
550 static EbmlSyntax matroska_track[] = {
551 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
552 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
553 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
554 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
555 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
556 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
557 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
558 { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
559 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
560 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
561 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
562 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
563 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
564 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
565 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
566 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
567 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
568 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
569 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
570 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
571 { MATROSKA_ID_CODECNAME, EBML_NONE },
572 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
573 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
574 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
575 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
576 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
577 CHILD_OF(matroska_tracks)
580 static EbmlSyntax matroska_tracks[] = {
581 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
582 CHILD_OF(matroska_segment)
585 static EbmlSyntax matroska_attachment[] = {
586 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
587 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
588 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
589 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
590 { MATROSKA_ID_FILEDESC, EBML_NONE },
591 CHILD_OF(matroska_attachments)
594 static EbmlSyntax matroska_attachments[] = {
595 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
596 CHILD_OF(matroska_segment)
599 static EbmlSyntax matroska_chapter_display[] = {
600 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
601 { MATROSKA_ID_CHAPLANG, EBML_NONE },
602 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
603 CHILD_OF(matroska_chapter_entry)
606 static EbmlSyntax matroska_chapter_entry[] = {
607 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
608 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
609 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
610 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
611 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
612 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
613 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
614 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
615 CHILD_OF(matroska_chapter)
618 static EbmlSyntax matroska_chapter[] = {
619 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
620 { MATROSKA_ID_EDITIONUID, EBML_NONE },
621 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
622 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
623 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
624 CHILD_OF(matroska_chapters)
627 static EbmlSyntax matroska_chapters[] = {
628 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
629 CHILD_OF(matroska_segment)
632 static EbmlSyntax matroska_index_pos[] = {
633 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
634 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
635 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
636 { MATROSKA_ID_CUEDURATION, EBML_NONE },
637 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
638 CHILD_OF(matroska_index_entry)
641 static EbmlSyntax matroska_index_entry[] = {
642 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
643 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
644 CHILD_OF(matroska_index)
647 static EbmlSyntax matroska_index[] = {
648 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
649 CHILD_OF(matroska_segment)
652 static EbmlSyntax matroska_simpletag[] = {
653 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
654 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
655 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
656 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
657 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
658 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
659 CHILD_OF(matroska_tag)
662 static EbmlSyntax matroska_tagtargets[] = {
663 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
664 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
665 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
666 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
667 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
668 CHILD_OF(matroska_tag)
671 static EbmlSyntax matroska_tag[] = {
672 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
673 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
674 CHILD_OF(matroska_tags)
677 static EbmlSyntax matroska_tags[] = {
678 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
679 CHILD_OF(matroska_segment)
682 static EbmlSyntax matroska_seekhead_entry[] = {
683 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
684 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
685 CHILD_OF(matroska_seekhead)
688 static EbmlSyntax matroska_seekhead[] = {
689 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
690 CHILD_OF(matroska_segment)
693 static EbmlSyntax matroska_segment[] = {
694 { MATROSKA_ID_CLUSTER, EBML_STOP },
695 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
696 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
697 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
698 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
699 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
700 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
701 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
702 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
705 static EbmlSyntax matroska_segments[] = {
706 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
710 static EbmlSyntax matroska_blockmore[] = {
711 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
712 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
713 CHILD_OF(matroska_blockadditions)
716 static EbmlSyntax matroska_blockadditions[] = {
717 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
718 CHILD_OF(matroska_blockgroup)
721 static EbmlSyntax matroska_blockgroup[] = {
722 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
723 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
724 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
725 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
726 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference), { .i = INT64_MIN } },
727 { MATROSKA_ID_CODECSTATE, EBML_NONE },
728 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
729 CHILD_OF(matroska_cluster_parsing)
732 // The following array contains SimpleBlock and BlockGroup twice
733 // in order to reuse the other values for matroska_cluster_enter.
734 static EbmlSyntax matroska_cluster_parsing[] = {
735 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
736 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, { .n = matroska_blockgroup } },
737 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
738 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
739 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
740 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
741 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
742 CHILD_OF(matroska_segment)
745 static EbmlSyntax matroska_cluster_enter[] = {
746 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = &matroska_cluster_parsing[2] } },
751 static const CodecMime mkv_image_mime_tags[] = {
752 {"image/gif" , AV_CODEC_ID_GIF},
753 {"image/jpeg" , AV_CODEC_ID_MJPEG},
754 {"image/png" , AV_CODEC_ID_PNG},
755 {"image/tiff" , AV_CODEC_ID_TIFF},
757 {"" , AV_CODEC_ID_NONE}
760 static const CodecMime mkv_mime_tags[] = {
761 {"text/plain" , AV_CODEC_ID_TEXT},
762 {"application/x-truetype-font", AV_CODEC_ID_TTF},
763 {"application/x-font" , AV_CODEC_ID_TTF},
764 {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
765 {"binary" , AV_CODEC_ID_BIN_DATA},
767 {"" , AV_CODEC_ID_NONE}
770 static const char *const matroska_doctypes[] = { "matroska", "webm" };
772 static int matroska_read_close(AVFormatContext *s);
775 * This function prepares the status for parsing of level 1 elements.
777 static int matroska_reset_status(MatroskaDemuxContext *matroska,
778 uint32_t id, int64_t position)
781 int64_t err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
786 matroska->current_id = id;
787 matroska->num_levels = 1;
788 matroska->unknown_count = 0;
789 matroska->resync_pos = avio_tell(matroska->ctx->pb);
791 matroska->resync_pos -= (av_log2(id) + 7) / 8;
796 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
798 AVIOContext *pb = matroska->ctx->pb;
801 /* Try to seek to the last position to resync from. If this doesn't work,
802 * we resync from the earliest position available: The start of the buffer. */
803 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
804 av_log(matroska->ctx, AV_LOG_WARNING,
805 "Seek to desired resync point failed. Seeking to "
806 "earliest point available instead.\n");
807 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
808 last_pos + 1), SEEK_SET);
813 // try to find a toplevel element
814 while (!avio_feof(pb)) {
815 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
816 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
817 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
818 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
819 /* Prepare the context for parsing of a level 1 element. */
820 matroska_reset_status(matroska, id, -1);
821 /* Given that we are here means that an error has occurred,
822 * so treat the segment as unknown length in order not to
823 * discard valid data that happens to be beyond the designated
824 * end of the segment. */
825 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
828 id = (id << 8) | avio_r8(pb);
832 return pb->error ? pb->error : AVERROR_EOF;
836 * Read: an "EBML number", which is defined as a variable-length
837 * array of bytes. The first byte indicates the length by giving a
838 * number of 0-bits followed by a one. The position of the first
839 * "one" bit inside the first byte indicates the length of this
841 * Returns: number of bytes read, < 0 on error
843 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
844 int max_size, uint64_t *number, int eof_forbidden)
850 /* The first byte tells us the length in bytes - except when it is zero. */
855 /* get the length of the EBML number */
856 read = 8 - ff_log2_tab[total];
858 if (!total || read > max_size) {
859 pos = avio_tell(pb) - 1;
861 av_log(matroska->ctx, AV_LOG_ERROR,
862 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
863 "of an EBML number\n", pos, pos);
865 av_log(matroska->ctx, AV_LOG_ERROR,
866 "Length %d indicated by an EBML number's first byte 0x%02x "
867 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
868 read, (uint8_t) total, pos, pos, max_size);
870 return AVERROR_INVALIDDATA;
873 /* read out length */
874 total ^= 1 << ff_log2_tab[total];
876 total = (total << 8) | avio_r8(pb);
878 if (pb->eof_reached) {
890 av_log(matroska->ctx, AV_LOG_ERROR,
891 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
896 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
897 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
904 * Read a EBML length value.
905 * This needs special handling for the "unknown length" case which has multiple
908 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
911 int res = ebml_read_num(matroska, pb, 8, number, 1);
912 if (res > 0 && *number + 1 == 1ULL << (7 * res))
913 *number = EBML_UNKNOWN_LENGTH;
918 * Read the next element as an unsigned int.
919 * Returns NEEDS_CHECKING.
921 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
925 /* big-endian ordering; build up number */
928 *num = (*num << 8) | avio_r8(pb);
930 return NEEDS_CHECKING;
934 * Read the next element as a signed int.
935 * Returns NEEDS_CHECKING.
937 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
944 *num = sign_extend(avio_r8(pb), 8);
946 /* big-endian ordering; build up number */
948 *num = ((uint64_t)*num << 8) | avio_r8(pb);
951 return NEEDS_CHECKING;
955 * Read the next element as a float.
956 * Returns NEEDS_CHECKING or < 0 on obvious failure.
958 static int ebml_read_float(AVIOContext *pb, int size, double *num)
963 *num = av_int2float(avio_rb32(pb));
965 *num = av_int2double(avio_rb64(pb));
967 return AVERROR_INVALIDDATA;
969 return NEEDS_CHECKING;
973 * Read the next element as an ASCII string.
974 * 0 is success, < 0 or NEEDS_CHECKING is failure.
976 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
981 /* EBML strings are usually not 0-terminated, so we allocate one
982 * byte more, read the string and NULL-terminate it ourselves. */
983 if (!(res = av_malloc(size + 1)))
984 return AVERROR(ENOMEM);
985 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
987 return ret < 0 ? ret : NEEDS_CHECKING;
997 * Read the next element as binary data.
998 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1000 static int ebml_read_binary(AVIOContext *pb, int length,
1001 int64_t pos, EbmlBin *bin)
1005 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1008 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1010 bin->data = bin->buf->data;
1013 if ((ret = avio_read(pb, bin->data, length)) != length) {
1014 av_buffer_unref(&bin->buf);
1017 return ret < 0 ? ret : NEEDS_CHECKING;
1024 * Read the next element, but only the header. The contents
1025 * are supposed to be sub-elements which can be read separately.
1026 * 0 is success, < 0 is failure.
1028 static int ebml_read_master(MatroskaDemuxContext *matroska,
1029 uint64_t length, int64_t pos)
1031 MatroskaLevel *level;
1033 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1034 av_log(matroska->ctx, AV_LOG_ERROR,
1035 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1036 return AVERROR(ENOSYS);
1039 level = &matroska->levels[matroska->num_levels++];
1041 level->length = length;
1047 * Read a signed "EBML number"
1048 * Return: number of bytes processed, < 0 on error
1050 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1051 AVIOContext *pb, int64_t *num)
1056 /* read as unsigned number first */
1057 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1060 /* make signed (weird way) */
1061 *num = unum - ((1LL << (7 * res - 1)) - 1);
1066 static int ebml_parse(MatroskaDemuxContext *matroska,
1067 EbmlSyntax *syntax, void *data);
1069 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1073 // Whoever touches this should be aware of the duplication
1074 // existing in matroska_cluster_parsing.
1075 for (i = 0; syntax[i].id; i++)
1076 if (id == syntax[i].id)
1082 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1088 for (int i = 0; syntax[i].id; i++)
1089 switch (syntax[i].type) {
1091 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1094 *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1097 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1101 // the default may be NULL
1102 if (syntax[i].def.s) {
1103 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1104 *dst = av_strdup(syntax[i].def.s);
1106 return AVERROR(ENOMEM);
1111 if (!matroska->levels[matroska->num_levels - 1].length) {
1112 matroska->num_levels--;
1118 res = ebml_parse(matroska, syntax, data);
1121 return res == LEVEL_ENDED ? 0 : res;
1124 static int is_ebml_id_valid(uint32_t id)
1126 // Due to endian nonsense in Matroska, the highest byte with any bits set
1127 // will contain the leading length bit. This bit in turn identifies the
1128 // total byte length of the element by its position within the byte.
1129 unsigned int bits = av_log2(id);
1130 return id && (bits + 7) / 8 == (8 - bits % 8);
1134 * Allocate and return the entry for the level1 element with the given ID. If
1135 * an entry already exists, return the existing entry.
1137 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1141 MatroskaLevel1Element *elem;
1143 if (!is_ebml_id_valid(id))
1146 // Some files link to all clusters; useless.
1147 if (id == MATROSKA_ID_CLUSTER)
1150 // There can be multiple seekheads.
1151 if (id != MATROSKA_ID_SEEKHEAD) {
1152 for (i = 0; i < matroska->num_level1_elems; i++) {
1153 if (matroska->level1_elems[i].id == id)
1154 return &matroska->level1_elems[i];
1158 // Only a completely broken file would have more elements.
1159 // It also provides a low-effort way to escape from circular seekheads
1160 // (every iteration will add a level1 entry).
1161 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1162 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1166 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1167 *elem = (MatroskaLevel1Element){.id = id};
1172 static int ebml_parse(MatroskaDemuxContext *matroska,
1173 EbmlSyntax *syntax, void *data)
1175 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1176 // Forbid unknown-length EBML_NONE elements.
1177 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1181 // max. 16 MB for strings
1182 [EBML_STR] = 0x1000000,
1183 [EBML_UTF8] = 0x1000000,
1184 // max. 256 MB for binary data
1185 [EBML_BIN] = 0x10000000,
1186 // no limits for anything else
1188 AVIOContext *pb = matroska->ctx->pb;
1191 int64_t pos = avio_tell(pb), pos_alt;
1192 int res, update_pos = 1, level_check;
1193 MatroskaLevel1Element *level1_elem;
1194 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1196 if (!matroska->current_id) {
1198 res = ebml_read_num(matroska, pb, 4, &id, 0);
1200 if (pb->eof_reached && res == AVERROR_EOF) {
1201 if (matroska->is_live)
1202 // in live mode, finish parsing if EOF is reached.
1204 if (level && pos == avio_tell(pb)) {
1205 if (level->length == EBML_UNKNOWN_LENGTH) {
1206 // Unknown-length levels automatically end at EOF.
1207 matroska->num_levels--;
1210 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1211 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1217 matroska->current_id = id | 1 << 7 * res;
1218 pos_alt = pos + res;
1221 pos -= (av_log2(matroska->current_id) + 7) / 8;
1224 id = matroska->current_id;
1226 syntax = ebml_parse_id(syntax, id);
1227 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1228 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1229 // Unknown-length levels end when an element from an upper level
1230 // in the hierarchy is encountered.
1231 while (syntax->def.n) {
1232 syntax = ebml_parse_id(syntax->def.n, id);
1234 matroska->num_levels--;
1240 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1241 "%"PRId64"\n", id, pos);
1242 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1246 data = (char *) data + syntax->data_offset;
1247 if (syntax->list_elem_size) {
1248 EbmlList *list = data;
1251 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1252 return AVERROR(ENOMEM);
1253 newelem = av_fast_realloc(list->elem,
1254 &list->alloc_elem_size,
1255 (list->nb_elem + 1) * syntax->list_elem_size);
1257 return AVERROR(ENOMEM);
1258 list->elem = newelem;
1259 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1260 memset(data, 0, syntax->list_elem_size);
1265 if (syntax->type != EBML_STOP) {
1266 matroska->current_id = 0;
1267 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1272 if (matroska->num_levels > 0) {
1273 if (length != EBML_UNKNOWN_LENGTH &&
1274 level->length != EBML_UNKNOWN_LENGTH) {
1275 uint64_t elem_end = pos_alt + length,
1276 level_end = level->start + level->length;
1278 if (elem_end < level_end) {
1280 } else if (elem_end == level_end) {
1281 level_check = LEVEL_ENDED;
1283 av_log(matroska->ctx, AV_LOG_ERROR,
1284 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1285 "containing master element ending at 0x%"PRIx64"\n",
1286 pos, elem_end, level_end);
1287 return AVERROR_INVALIDDATA;
1289 } else if (length != EBML_UNKNOWN_LENGTH) {
1291 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1292 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1293 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1294 return AVERROR_INVALIDDATA;
1297 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1298 || syntax->type == EBML_NEST)) {
1299 // According to the current specifications only clusters and
1300 // segments are allowed to be unknown-length. We also accept
1301 // other unknown-length master elements.
1302 av_log(matroska->ctx, AV_LOG_WARNING,
1303 "Found unknown-length element 0x%"PRIX32" other than "
1304 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1305 "parsing will nevertheless be attempted.\n", id, pos);
1312 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1313 if (length != EBML_UNKNOWN_LENGTH) {
1314 av_log(matroska->ctx, AV_LOG_ERROR,
1315 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1316 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1317 length, max_lengths[syntax->type], id, pos);
1318 } else if (syntax->type != EBML_NONE) {
1319 av_log(matroska->ctx, AV_LOG_ERROR,
1320 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1321 "unknown length, yet the length of an element of its "
1322 "type must be known.\n", id, pos);
1324 av_log(matroska->ctx, AV_LOG_ERROR,
1325 "Found unknown-length element with ID 0x%"PRIX32" at "
1326 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1327 "available.\n", id, pos);
1329 return AVERROR_INVALIDDATA;
1332 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1333 // Loosing sync will likely manifest itself as encountering unknown
1334 // elements which are not reliably distinguishable from elements
1335 // belonging to future extensions of the format.
1336 // We use a heuristic to detect such situations: If the current
1337 // element is not expected at the current syntax level and there
1338 // were only a few unknown elements in a row, then the element is
1339 // skipped or considered defective based upon the length of the
1340 // current element (i.e. how much would be skipped); if there were
1341 // more than a few skipped elements in a row and skipping the current
1342 // element would lead us more than SKIP_THRESHOLD away from the last
1343 // known good position, then it is inferred that an error occurred.
1344 // The dependency on the number of unknown elements in a row exists
1345 // because the distance to the last known good position is
1346 // automatically big if the last parsed element was big.
1347 // In both cases, each unknown element is considered equivalent to
1348 // UNKNOWN_EQUIV of skipped bytes for the check.
1349 // The whole check is only done for non-seekable output, because
1350 // in this situation skipped data can't simply be rechecked later.
1351 // This is especially important when using unkown length elements
1352 // as the check for whether a child exceeds its containing master
1353 // element is not effective in this situation.
1355 matroska->unknown_count = 0;
1357 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1359 if (matroska->unknown_count > 3)
1360 dist += pos_alt - matroska->resync_pos;
1362 if (dist > SKIP_THRESHOLD) {
1363 av_log(matroska->ctx, AV_LOG_ERROR,
1364 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1365 "length 0x%"PRIx64" considered as invalid data. Last "
1366 "known good position 0x%"PRIx64", %d unknown elements"
1367 " in a row\n", id, pos, length, matroska->resync_pos,
1368 matroska->unknown_count);
1369 return AVERROR_INVALIDDATA;
1374 if (update_pos > 0) {
1375 // We have found an element that is allowed at this place
1376 // in the hierarchy and it passed all checks, so treat the beginning
1377 // of the element as the "last known good" position.
1378 matroska->resync_pos = pos;
1381 if (!data && length != EBML_UNKNOWN_LENGTH)
1385 switch (syntax->type) {
1387 res = ebml_read_uint(pb, length, data);
1390 res = ebml_read_sint(pb, length, data);
1393 res = ebml_read_float(pb, length, data);
1397 res = ebml_read_ascii(pb, length, data);
1400 res = ebml_read_binary(pb, length, pos_alt, data);
1404 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1406 if (id == MATROSKA_ID_SEGMENT)
1407 matroska->segment_start = pos_alt;
1408 if (id == MATROSKA_ID_CUES)
1409 matroska->cues_parsing_deferred = 0;
1410 if (syntax->type == EBML_LEVEL1 &&
1411 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1412 if (!level1_elem->pos) {
1413 // Zero is not a valid position for a level 1 element.
1414 level1_elem->pos = pos;
1415 } else if (level1_elem->pos != pos)
1416 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1417 level1_elem->parsed = 1;
1419 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1428 if (ffio_limit(pb, length) != length) {
1429 // ffio_limit emits its own error message,
1430 // so we don't have to.
1431 return AVERROR(EIO);
1433 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1434 // avio_skip might take us past EOF. We check for this
1435 // by skipping only length - 1 bytes, reading a byte and
1436 // checking the error flags. This is done in order to check
1437 // that the element has been properly skipped even when
1438 // no filesize (that ffio_limit relies on) is available.
1440 res = NEEDS_CHECKING;
1447 if (res == NEEDS_CHECKING) {
1448 if (pb->eof_reached) {
1457 if (res == AVERROR_INVALIDDATA)
1458 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1459 else if (res == AVERROR(EIO))
1460 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1461 else if (res == AVERROR_EOF) {
1462 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1470 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1471 level = &matroska->levels[matroska->num_levels - 1];
1472 pos = avio_tell(pb);
1474 // Given that pos >= level->start no check for
1475 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1476 while (matroska->num_levels && pos == level->start + level->length) {
1477 matroska->num_levels--;
1485 static void ebml_free(EbmlSyntax *syntax, void *data)
1488 for (i = 0; syntax[i].id; i++) {
1489 void *data_off = (char *) data + syntax[i].data_offset;
1490 switch (syntax[i].type) {
1496 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1500 if (syntax[i].list_elem_size) {
1501 EbmlList *list = data_off;
1502 char *ptr = list->elem;
1503 for (j = 0; j < list->nb_elem;
1504 j++, ptr += syntax[i].list_elem_size)
1505 ebml_free(syntax[i].def.n, ptr);
1506 av_freep(&list->elem);
1508 list->alloc_elem_size = 0;
1510 ebml_free(syntax[i].def.n, data_off);
1520 static int matroska_probe(const AVProbeData *p)
1523 int len_mask = 0x80, size = 1, n = 1, i;
1526 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1529 /* length of header */
1531 while (size <= 8 && !(total & len_mask)) {
1537 total &= (len_mask - 1);
1539 total = (total << 8) | p->buf[4 + n++];
1541 if (total + 1 == 1ULL << (7 * size)){
1542 /* Unknown-length header - simply parse the whole buffer. */
1543 total = p->buf_size - 4 - size;
1545 /* Does the probe data contain the whole header? */
1546 if (p->buf_size < 4 + size + total)
1550 /* The header should contain a known document type. For now,
1551 * we don't parse the whole header but simply check for the
1552 * availability of that array of characters inside the header.
1553 * Not fully fool-proof, but good enough. */
1554 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1555 size_t probelen = strlen(matroska_doctypes[i]);
1556 if (total < probelen)
1558 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1559 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1560 return AVPROBE_SCORE_MAX;
1563 // probably valid EBML header but no recognized doctype
1564 return AVPROBE_SCORE_EXTENSION;
1567 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1570 MatroskaTrack *tracks = matroska->tracks.elem;
1573 for (i = 0; i < matroska->tracks.nb_elem; i++)
1574 if (tracks[i].num == num)
1577 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1581 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1582 MatroskaTrack *track)
1584 MatroskaTrackEncoding *encodings = track->encodings.elem;
1585 uint8_t *data = *buf;
1586 int isize = *buf_size;
1587 uint8_t *pkt_data = NULL;
1588 uint8_t av_unused *newpktdata;
1589 int pkt_size = isize;
1593 if (pkt_size >= 10000000U)
1594 return AVERROR_INVALIDDATA;
1596 switch (encodings[0].compression.algo) {
1597 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1599 int header_size = encodings[0].compression.settings.size;
1600 uint8_t *header = encodings[0].compression.settings.data;
1602 if (header_size && !header) {
1603 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1610 pkt_size = isize + header_size;
1611 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1613 return AVERROR(ENOMEM);
1615 memcpy(pkt_data, header, header_size);
1616 memcpy(pkt_data + header_size, data, isize);
1620 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1623 olen = pkt_size *= 3;
1624 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1625 + AV_INPUT_BUFFER_PADDING_SIZE);
1627 result = AVERROR(ENOMEM);
1630 pkt_data = newpktdata;
1631 result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1632 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1634 result = AVERROR_INVALIDDATA;
1641 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1643 z_stream zstream = { 0 };
1644 if (inflateInit(&zstream) != Z_OK)
1646 zstream.next_in = data;
1647 zstream.avail_in = isize;
1650 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1652 inflateEnd(&zstream);
1653 result = AVERROR(ENOMEM);
1656 pkt_data = newpktdata;
1657 zstream.avail_out = pkt_size - zstream.total_out;
1658 zstream.next_out = pkt_data + zstream.total_out;
1659 result = inflate(&zstream, Z_NO_FLUSH);
1660 } while (result == Z_OK && pkt_size < 10000000);
1661 pkt_size = zstream.total_out;
1662 inflateEnd(&zstream);
1663 if (result != Z_STREAM_END) {
1664 if (result == Z_MEM_ERROR)
1665 result = AVERROR(ENOMEM);
1667 result = AVERROR_INVALIDDATA;
1674 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1676 bz_stream bzstream = { 0 };
1677 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1679 bzstream.next_in = data;
1680 bzstream.avail_in = isize;
1683 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1685 BZ2_bzDecompressEnd(&bzstream);
1686 result = AVERROR(ENOMEM);
1689 pkt_data = newpktdata;
1690 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1691 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1692 result = BZ2_bzDecompress(&bzstream);
1693 } while (result == BZ_OK && pkt_size < 10000000);
1694 pkt_size = bzstream.total_out_lo32;
1695 BZ2_bzDecompressEnd(&bzstream);
1696 if (result != BZ_STREAM_END) {
1697 if (result == BZ_MEM_ERROR)
1698 result = AVERROR(ENOMEM);
1700 result = AVERROR_INVALIDDATA;
1707 return AVERROR_INVALIDDATA;
1710 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1713 *buf_size = pkt_size;
1721 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1722 AVDictionary **metadata, char *prefix)
1724 MatroskaTag *tags = list->elem;
1728 for (i = 0; i < list->nb_elem; i++) {
1729 const char *lang = tags[i].lang &&
1730 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1732 if (!tags[i].name) {
1733 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1737 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1739 av_strlcpy(key, tags[i].name, sizeof(key));
1740 if (tags[i].def || !lang) {
1741 av_dict_set(metadata, key, tags[i].string, 0);
1742 if (tags[i].sub.nb_elem)
1743 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1746 av_strlcat(key, "-", sizeof(key));
1747 av_strlcat(key, lang, sizeof(key));
1748 av_dict_set(metadata, key, tags[i].string, 0);
1749 if (tags[i].sub.nb_elem)
1750 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1753 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1756 static void matroska_convert_tags(AVFormatContext *s)
1758 MatroskaDemuxContext *matroska = s->priv_data;
1759 MatroskaTags *tags = matroska->tags.elem;
1762 for (i = 0; i < matroska->tags.nb_elem; i++) {
1763 if (tags[i].target.attachuid) {
1764 MatroskaAttachment *attachment = matroska->attachments.elem;
1766 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1767 if (attachment[j].uid == tags[i].target.attachuid &&
1768 attachment[j].stream) {
1769 matroska_convert_tag(s, &tags[i].tag,
1770 &attachment[j].stream->metadata, NULL);
1775 av_log(NULL, AV_LOG_WARNING,
1776 "The tags at index %d refer to a "
1777 "non-existent attachment %"PRId64".\n",
1778 i, tags[i].target.attachuid);
1780 } else if (tags[i].target.chapteruid) {
1781 MatroskaChapter *chapter = matroska->chapters.elem;
1783 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1784 if (chapter[j].uid == tags[i].target.chapteruid &&
1785 chapter[j].chapter) {
1786 matroska_convert_tag(s, &tags[i].tag,
1787 &chapter[j].chapter->metadata, NULL);
1792 av_log(NULL, AV_LOG_WARNING,
1793 "The tags at index %d refer to a non-existent chapter "
1795 i, tags[i].target.chapteruid);
1797 } else if (tags[i].target.trackuid) {
1798 MatroskaTrack *track = matroska->tracks.elem;
1800 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1801 if (track[j].uid == tags[i].target.trackuid &&
1803 matroska_convert_tag(s, &tags[i].tag,
1804 &track[j].stream->metadata, NULL);
1809 av_log(NULL, AV_LOG_WARNING,
1810 "The tags at index %d refer to a non-existent track "
1812 i, tags[i].target.trackuid);
1815 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1816 tags[i].target.type);
1821 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1824 uint32_t saved_id = matroska->current_id;
1825 int64_t before_pos = avio_tell(matroska->ctx->pb);
1829 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1830 /* We don't want to lose our seekhead level, so we add
1831 * a dummy. This is a crude hack. */
1832 if (matroska->num_levels == EBML_MAX_DEPTH) {
1833 av_log(matroska->ctx, AV_LOG_INFO,
1834 "Max EBML element depth (%d) reached, "
1835 "cannot parse further.\n", EBML_MAX_DEPTH);
1836 ret = AVERROR_INVALIDDATA;
1838 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1839 matroska->num_levels++;
1840 matroska->current_id = 0;
1842 ret = ebml_parse(matroska, matroska_segment, matroska);
1843 if (ret == LEVEL_ENDED) {
1844 /* This can only happen if the seek brought us beyond EOF. */
1849 /* Seek back - notice that in all instances where this is used
1850 * it is safe to set the level to 1. */
1851 matroska_reset_status(matroska, saved_id, before_pos);
1856 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1858 EbmlList *seekhead_list = &matroska->seekhead;
1861 // we should not do any seeking in the streaming case
1862 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1865 for (i = 0; i < seekhead_list->nb_elem; i++) {
1866 MatroskaSeekhead *seekheads = seekhead_list->elem;
1867 uint32_t id = seekheads[i].id;
1868 int64_t pos = seekheads[i].pos + matroska->segment_start;
1869 MatroskaLevel1Element *elem;
1871 if (id != seekheads[i].id || pos < matroska->segment_start)
1874 elem = matroska_find_level1_elem(matroska, id);
1875 if (!elem || elem->parsed)
1880 // defer cues parsing until we actually need cue data.
1881 if (id == MATROSKA_ID_CUES)
1884 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1885 // mark index as broken
1886 matroska->cues_parsing_deferred = -1;
1894 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1896 EbmlList *index_list;
1897 MatroskaIndex *index;
1898 uint64_t index_scale = 1;
1901 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1904 index_list = &matroska->index;
1905 index = index_list->elem;
1906 if (index_list->nb_elem < 2)
1908 if (index[1].time > 1E14 / matroska->time_scale) {
1909 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1912 for (i = 0; i < index_list->nb_elem; i++) {
1913 EbmlList *pos_list = &index[i].pos;
1914 MatroskaIndexPos *pos = pos_list->elem;
1915 for (j = 0; j < pos_list->nb_elem; j++) {
1916 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1918 if (track && track->stream)
1919 av_add_index_entry(track->stream,
1920 pos[j].pos + matroska->segment_start,
1921 index[i].time / index_scale, 0, 0,
1927 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1930 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1933 for (i = 0; i < matroska->num_level1_elems; i++) {
1934 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1935 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1936 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1937 matroska->cues_parsing_deferred = -1;
1943 matroska_add_index_entries(matroska);
1946 static int matroska_aac_profile(char *codec_id)
1948 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1951 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1952 if (strstr(codec_id, aac_profiles[profile]))
1957 static int matroska_aac_sri(int samplerate)
1961 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1962 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1967 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1969 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1970 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1973 static int matroska_parse_flac(AVFormatContext *s,
1974 MatroskaTrack *track,
1977 AVStream *st = track->stream;
1978 uint8_t *p = track->codec_priv.data;
1979 int size = track->codec_priv.size;
1981 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1982 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1983 track->codec_priv.size = 0;
1987 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1989 p += track->codec_priv.size;
1990 size -= track->codec_priv.size;
1992 /* parse the remaining metadata blocks if present */
1994 int block_last, block_type, block_size;
1996 flac_parse_block_header(p, &block_last, &block_type, &block_size);
2000 if (block_size > size)
2003 /* check for the channel mask */
2004 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2005 AVDictionary *dict = NULL;
2006 AVDictionaryEntry *chmask;
2008 ff_vorbis_comment(s, &dict, p, block_size, 0);
2009 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2011 uint64_t mask = strtol(chmask->value, NULL, 0);
2012 if (!mask || mask & ~0x3ffffULL) {
2013 av_log(s, AV_LOG_WARNING,
2014 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2016 st->codecpar->channel_layout = mask;
2018 av_dict_free(&dict);
2028 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2030 int major, minor, micro, bttb = 0;
2032 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2033 * this function, and fixed in 57.52 */
2034 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, µ) == 3)
2035 bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
2037 switch (field_order) {
2038 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2039 return AV_FIELD_PROGRESSIVE;
2040 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2041 return AV_FIELD_UNKNOWN;
2042 case MATROSKA_VIDEO_FIELDORDER_TT:
2044 case MATROSKA_VIDEO_FIELDORDER_BB:
2046 case MATROSKA_VIDEO_FIELDORDER_BT:
2047 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2048 case MATROSKA_VIDEO_FIELDORDER_TB:
2049 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2051 return AV_FIELD_UNKNOWN;
2055 static void mkv_stereo_mode_display_mul(int stereo_mode,
2056 int *h_width, int *h_height)
2058 switch (stereo_mode) {
2059 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2060 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2061 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2062 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2063 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2065 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2066 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2067 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2068 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2071 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2072 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2073 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2074 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2080 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2081 const MatroskaTrackVideoColor *color = track->video.color.elem;
2082 const MatroskaMasteringMeta *mastering_meta;
2083 int has_mastering_primaries, has_mastering_luminance;
2085 if (!track->video.color.nb_elem)
2088 mastering_meta = &color->mastering_meta;
2089 // Mastering primaries are CIE 1931 coords, and must be > 0.
2090 has_mastering_primaries =
2091 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2092 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2093 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2094 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2095 has_mastering_luminance = mastering_meta->max_luminance > 0;
2097 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2098 st->codecpar->color_space = color->matrix_coefficients;
2099 if (color->primaries != AVCOL_PRI_RESERVED &&
2100 color->primaries != AVCOL_PRI_RESERVED0)
2101 st->codecpar->color_primaries = color->primaries;
2102 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2103 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2104 st->codecpar->color_trc = color->transfer_characteristics;
2105 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2106 color->range <= AVCOL_RANGE_JPEG)
2107 st->codecpar->color_range = color->range;
2108 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2109 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2110 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2111 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2112 st->codecpar->chroma_location =
2113 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2114 (color->chroma_siting_vert - 1) << 7);
2116 if (color->max_cll && color->max_fall) {
2119 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2121 return AVERROR(ENOMEM);
2122 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2123 (uint8_t *)metadata, size);
2125 av_freep(&metadata);
2128 metadata->MaxCLL = color->max_cll;
2129 metadata->MaxFALL = color->max_fall;
2132 if (has_mastering_primaries || has_mastering_luminance) {
2133 AVMasteringDisplayMetadata *metadata =
2134 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2135 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2136 sizeof(AVMasteringDisplayMetadata));
2138 return AVERROR(ENOMEM);
2140 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2141 if (has_mastering_primaries) {
2142 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2143 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2144 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2145 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2146 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2147 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2148 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2149 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2150 metadata->has_primaries = 1;
2152 if (has_mastering_luminance) {
2153 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2154 metadata->min_luminance = av_d2q(mastering_meta->min_luminance, INT_MAX);
2155 metadata->has_luminance = 1;
2161 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track) {
2162 AVSphericalMapping *spherical;
2163 enum AVSphericalProjection projection;
2164 size_t spherical_size;
2165 uint32_t l = 0, t = 0, r = 0, b = 0;
2166 uint32_t padding = 0;
2170 bytestream2_init(&gb, track->video.projection.private.data,
2171 track->video.projection.private.size);
2173 if (bytestream2_get_byte(&gb) != 0) {
2174 av_log(NULL, AV_LOG_WARNING, "Unknown spherical metadata\n");
2178 bytestream2_skip(&gb, 3); // flags
2180 switch (track->video.projection.type) {
2181 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2182 if (track->video.projection.private.size == 20) {
2183 t = bytestream2_get_be32(&gb);
2184 b = bytestream2_get_be32(&gb);
2185 l = bytestream2_get_be32(&gb);
2186 r = bytestream2_get_be32(&gb);
2188 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2189 av_log(NULL, AV_LOG_ERROR,
2190 "Invalid bounding rectangle coordinates "
2191 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2193 return AVERROR_INVALIDDATA;
2195 } else if (track->video.projection.private.size != 0) {
2196 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2197 return AVERROR_INVALIDDATA;
2200 if (l || t || r || b)
2201 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2203 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2205 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2206 if (track->video.projection.private.size < 4) {
2207 av_log(NULL, AV_LOG_ERROR, "Missing projection private properties\n");
2208 return AVERROR_INVALIDDATA;
2209 } else if (track->video.projection.private.size == 12) {
2210 uint32_t layout = bytestream2_get_be32(&gb);
2212 av_log(NULL, AV_LOG_WARNING,
2213 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2216 projection = AV_SPHERICAL_CUBEMAP;
2217 padding = bytestream2_get_be32(&gb);
2219 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2220 return AVERROR_INVALIDDATA;
2223 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2224 /* No Spherical metadata */
2227 av_log(NULL, AV_LOG_WARNING,
2228 "Unknown spherical metadata type %"PRIu64"\n",
2229 track->video.projection.type);
2233 spherical = av_spherical_alloc(&spherical_size);
2235 return AVERROR(ENOMEM);
2237 spherical->projection = projection;
2239 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2240 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2241 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2243 spherical->padding = padding;
2245 spherical->bound_left = l;
2246 spherical->bound_top = t;
2247 spherical->bound_right = r;
2248 spherical->bound_bottom = b;
2250 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2253 av_freep(&spherical);
2260 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2262 const AVCodecTag *codec_tags;
2264 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2265 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2267 /* Normalize noncompliant private data that starts with the fourcc
2268 * by expanding/shifting the data by 4 bytes and storing the data
2269 * size at the start. */
2270 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2271 int ret = av_buffer_realloc(&track->codec_priv.buf,
2272 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2276 track->codec_priv.data = track->codec_priv.buf->data;
2277 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2278 track->codec_priv.size += 4;
2279 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2282 *fourcc = AV_RL32(track->codec_priv.data + 4);
2283 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2288 static int matroska_parse_tracks(AVFormatContext *s)
2290 MatroskaDemuxContext *matroska = s->priv_data;
2291 MatroskaTrack *tracks = matroska->tracks.elem;
2296 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2297 MatroskaTrack *track = &tracks[i];
2298 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2299 EbmlList *encodings_list = &track->encodings;
2300 MatroskaTrackEncoding *encodings = encodings_list->elem;
2301 uint8_t *extradata = NULL;
2302 int extradata_size = 0;
2303 int extradata_offset = 0;
2304 uint32_t fourcc = 0;
2306 char* key_id_base64 = NULL;
2309 /* Apply some sanity checks. */
2310 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2311 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2312 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2313 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2314 av_log(matroska->ctx, AV_LOG_INFO,
2315 "Unknown or unsupported track type %"PRIu64"\n",
2319 if (!track->codec_id)
2322 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2323 isnan(track->audio.samplerate)) {
2324 av_log(matroska->ctx, AV_LOG_WARNING,
2325 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2326 track->audio.samplerate);
2327 track->audio.samplerate = 8000;
2330 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2331 if (!track->default_duration && track->video.frame_rate > 0) {
2332 double default_duration = 1000000000 / track->video.frame_rate;
2333 if (default_duration > UINT64_MAX || default_duration < 0) {
2334 av_log(matroska->ctx, AV_LOG_WARNING,
2335 "Invalid frame rate %e. Cannot calculate default duration.\n",
2336 track->video.frame_rate);
2338 track->default_duration = default_duration;
2341 if (track->video.display_width == -1)
2342 track->video.display_width = track->video.pixel_width;
2343 if (track->video.display_height == -1)
2344 track->video.display_height = track->video.pixel_height;
2345 if (track->video.color_space.size == 4)
2346 fourcc = AV_RL32(track->video.color_space.data);
2347 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2348 if (!track->audio.out_samplerate)
2349 track->audio.out_samplerate = track->audio.samplerate;
2351 if (encodings_list->nb_elem > 1) {
2352 av_log(matroska->ctx, AV_LOG_ERROR,
2353 "Multiple combined encodings not supported");
2354 } else if (encodings_list->nb_elem == 1) {
2355 if (encodings[0].type) {
2356 if (encodings[0].encryption.key_id.size > 0) {
2357 /* Save the encryption key id to be stored later as a
2359 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2360 key_id_base64 = av_malloc(b64_size);
2361 if (key_id_base64 == NULL)
2362 return AVERROR(ENOMEM);
2364 av_base64_encode(key_id_base64, b64_size,
2365 encodings[0].encryption.key_id.data,
2366 encodings[0].encryption.key_id.size);
2368 encodings[0].scope = 0;
2369 av_log(matroska->ctx, AV_LOG_ERROR,
2370 "Unsupported encoding type");
2374 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2377 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2380 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2382 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2383 encodings[0].scope = 0;
2384 av_log(matroska->ctx, AV_LOG_ERROR,
2385 "Unsupported encoding type");
2386 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2387 uint8_t *codec_priv = track->codec_priv.data;
2388 int ret = matroska_decode_buffer(&track->codec_priv.data,
2389 &track->codec_priv.size,
2392 track->codec_priv.data = NULL;
2393 track->codec_priv.size = 0;
2394 av_log(matroska->ctx, AV_LOG_ERROR,
2395 "Failed to decode codec private data\n");
2398 if (codec_priv != track->codec_priv.data) {
2399 av_buffer_unref(&track->codec_priv.buf);
2400 if (track->codec_priv.data) {
2401 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2402 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2404 if (!track->codec_priv.buf) {
2405 av_freep(&track->codec_priv.data);
2406 track->codec_priv.size = 0;
2407 return AVERROR(ENOMEM);
2413 track->needs_decoding = encodings && !encodings[0].type &&
2414 encodings[0].scope & 1 &&
2415 (encodings[0].compression.algo !=
2416 MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP ||
2417 encodings[0].compression.settings.size);
2419 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2420 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2421 strlen(ff_mkv_codec_tags[j].str))) {
2422 codec_id = ff_mkv_codec_tags[j].id;
2427 st = track->stream = avformat_new_stream(s, NULL);
2429 av_free(key_id_base64);
2430 return AVERROR(ENOMEM);
2433 if (key_id_base64) {
2434 /* export encryption key id as base64 metadata tag */
2435 av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2436 AV_DICT_DONT_STRDUP_VAL);
2439 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2440 track->codec_priv.size >= 40 &&
2441 track->codec_priv.data) {
2442 track->ms_compat = 1;
2443 bit_depth = AV_RL16(track->codec_priv.data + 14);
2444 fourcc = AV_RL32(track->codec_priv.data + 16);
2445 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2448 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2450 extradata_offset = 40;
2451 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2452 track->codec_priv.size >= 14 &&
2453 track->codec_priv.data) {
2455 ffio_init_context(&b, track->codec_priv.data,
2456 track->codec_priv.size,
2457 0, NULL, NULL, NULL, NULL);
2458 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2461 codec_id = st->codecpar->codec_id;
2462 fourcc = st->codecpar->codec_tag;
2463 extradata_offset = FFMIN(track->codec_priv.size, 18);
2464 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2465 /* Normally 36, but allow noncompliant private data */
2466 && (track->codec_priv.size >= 32)
2467 && (track->codec_priv.data)) {
2468 uint16_t sample_size;
2469 int ret = get_qt_codec(track, &fourcc, &codec_id);
2472 sample_size = AV_RB16(track->codec_priv.data + 26);
2474 if (sample_size == 8) {
2475 fourcc = MKTAG('r','a','w',' ');
2476 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2477 } else if (sample_size == 16) {
2478 fourcc = MKTAG('t','w','o','s');
2479 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2482 if ((fourcc == MKTAG('t','w','o','s') ||
2483 fourcc == MKTAG('s','o','w','t')) &&
2485 codec_id = AV_CODEC_ID_PCM_S8;
2486 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2487 (track->codec_priv.size >= 21) &&
2488 (track->codec_priv.data)) {
2489 int ret = get_qt_codec(track, &fourcc, &codec_id);
2492 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2493 fourcc = MKTAG('S','V','Q','3');
2494 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2496 if (codec_id == AV_CODEC_ID_NONE)
2497 av_log(matroska->ctx, AV_LOG_ERROR,
2498 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2499 if (track->codec_priv.size >= 86) {
2500 bit_depth = AV_RB16(track->codec_priv.data + 82);
2501 ffio_init_context(&b, track->codec_priv.data,
2502 track->codec_priv.size,
2503 0, NULL, NULL, NULL, NULL);
2504 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2506 track->has_palette = 1;
2509 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2510 switch (track->audio.bitdepth) {
2512 codec_id = AV_CODEC_ID_PCM_U8;
2515 codec_id = AV_CODEC_ID_PCM_S24BE;
2518 codec_id = AV_CODEC_ID_PCM_S32BE;
2521 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2522 switch (track->audio.bitdepth) {
2524 codec_id = AV_CODEC_ID_PCM_U8;
2527 codec_id = AV_CODEC_ID_PCM_S24LE;
2530 codec_id = AV_CODEC_ID_PCM_S32LE;
2533 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2534 track->audio.bitdepth == 64) {
2535 codec_id = AV_CODEC_ID_PCM_F64LE;
2536 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2537 int profile = matroska_aac_profile(track->codec_id);
2538 int sri = matroska_aac_sri(track->audio.samplerate);
2539 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2541 return AVERROR(ENOMEM);
2542 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2543 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2544 if (strstr(track->codec_id, "SBR")) {
2545 sri = matroska_aac_sri(track->audio.out_samplerate);
2546 extradata[2] = 0x56;
2547 extradata[3] = 0xE5;
2548 extradata[4] = 0x80 | (sri << 3);
2552 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2553 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2554 * Create the "atom size", "tag", and "tag version" fields the
2555 * decoder expects manually. */
2556 extradata_size = 12 + track->codec_priv.size;
2557 extradata = av_mallocz(extradata_size +
2558 AV_INPUT_BUFFER_PADDING_SIZE);
2560 return AVERROR(ENOMEM);
2561 AV_WB32(extradata, extradata_size);
2562 memcpy(&extradata[4], "alac", 4);
2563 AV_WB32(&extradata[8], 0);
2564 memcpy(&extradata[12], track->codec_priv.data,
2565 track->codec_priv.size);
2566 } else if (codec_id == AV_CODEC_ID_TTA) {
2568 if (track->audio.channels > UINT16_MAX ||
2569 track->audio.bitdepth > UINT16_MAX) {
2570 av_log(matroska->ctx, AV_LOG_WARNING,
2571 "Too large audio channel number %"PRIu64
2572 " or bitdepth %"PRIu64". Skipping track.\n",
2573 track->audio.channels, track->audio.bitdepth);
2574 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2575 return AVERROR_INVALIDDATA;
2579 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2580 return AVERROR_INVALIDDATA;
2581 extradata_size = 22;
2582 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2584 return AVERROR(ENOMEM);
2586 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2587 bytestream_put_le16(&ptr, 1);
2588 bytestream_put_le16(&ptr, track->audio.channels);
2589 bytestream_put_le16(&ptr, track->audio.bitdepth);
2590 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2591 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2592 track->audio.out_samplerate,
2593 AV_TIME_BASE * 1000));
2594 } else if (codec_id == AV_CODEC_ID_RV10 ||
2595 codec_id == AV_CODEC_ID_RV20 ||
2596 codec_id == AV_CODEC_ID_RV30 ||
2597 codec_id == AV_CODEC_ID_RV40) {
2598 extradata_offset = 26;
2599 } else if (codec_id == AV_CODEC_ID_RA_144) {
2600 track->audio.out_samplerate = 8000;
2601 track->audio.channels = 1;
2602 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2603 codec_id == AV_CODEC_ID_COOK ||
2604 codec_id == AV_CODEC_ID_ATRAC3 ||
2605 codec_id == AV_CODEC_ID_SIPR)
2606 && track->codec_priv.data) {
2609 ffio_init_context(&b, track->codec_priv.data,
2610 track->codec_priv.size,
2611 0, NULL, NULL, NULL, NULL);
2613 flavor = avio_rb16(&b);
2614 track->audio.coded_framesize = avio_rb32(&b);
2616 track->audio.sub_packet_h = avio_rb16(&b);
2617 track->audio.frame_size = avio_rb16(&b);
2618 track->audio.sub_packet_size = avio_rb16(&b);
2619 if (track->audio.coded_framesize <= 0 ||
2620 track->audio.sub_packet_h <= 0 ||
2621 track->audio.frame_size <= 0)
2622 return AVERROR_INVALIDDATA;
2624 if (codec_id == AV_CODEC_ID_RA_288) {
2625 if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2626 != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2627 return AVERROR_INVALIDDATA;
2628 st->codecpar->block_align = track->audio.coded_framesize;
2629 track->codec_priv.size = 0;
2631 if (codec_id == AV_CODEC_ID_SIPR) {
2632 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2634 return AVERROR_INVALIDDATA;
2635 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2636 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2637 } else if (track->audio.sub_packet_size <= 0 ||
2638 track->audio.frame_size % track->audio.sub_packet_size)
2639 return AVERROR_INVALIDDATA;
2640 st->codecpar->block_align = track->audio.sub_packet_size;
2641 extradata_offset = 78;
2643 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2644 track->audio.frame_size);
2645 if (!track->audio.buf)
2646 return AVERROR(ENOMEM);
2647 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2648 ret = matroska_parse_flac(s, track, &extradata_offset);
2651 } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2652 av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2653 "in absence of valid CodecPrivate.\n");
2655 extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2657 return AVERROR(ENOMEM);
2658 AV_WL16(extradata, 0x410);
2659 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2660 fourcc = AV_RL32(track->codec_priv.data);
2661 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2662 /* we don't need any value stored in CodecPrivate.
2663 make sure that it's not exported as extradata. */
2664 track->codec_priv.size = 0;
2665 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2666 /* For now, propagate only the OBUs, if any. Once libavcodec is
2667 updated to handle isobmff style extradata this can be removed. */
2668 extradata_offset = 4;
2670 track->codec_priv.size -= extradata_offset;
2672 if (codec_id == AV_CODEC_ID_NONE)
2673 av_log(matroska->ctx, AV_LOG_INFO,
2674 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2676 if (track->time_scale < 0.01)
2677 track->time_scale = 1.0;
2678 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2679 1000 * 1000 * 1000); /* 64 bit pts in ns */
2681 /* convert the delay from ns to the track timebase */
2682 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2683 (AVRational){ 1, 1000000000 },
2686 st->codecpar->codec_id = codec_id;
2688 if (strcmp(track->language, "und"))
2689 av_dict_set(&st->metadata, "language", track->language, 0);
2690 av_dict_set(&st->metadata, "title", track->name, 0);
2692 if (track->flag_default)
2693 st->disposition |= AV_DISPOSITION_DEFAULT;
2694 if (track->flag_forced)
2695 st->disposition |= AV_DISPOSITION_FORCED;
2697 if (!st->codecpar->extradata) {
2699 st->codecpar->extradata = extradata;
2700 st->codecpar->extradata_size = extradata_size;
2701 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2702 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2703 return AVERROR(ENOMEM);
2704 memcpy(st->codecpar->extradata,
2705 track->codec_priv.data + extradata_offset,
2706 track->codec_priv.size);
2710 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2711 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2712 int display_width_mul = 1;
2713 int display_height_mul = 1;
2715 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2716 st->codecpar->codec_tag = fourcc;
2718 st->codecpar->bits_per_coded_sample = bit_depth;
2719 st->codecpar->width = track->video.pixel_width;
2720 st->codecpar->height = track->video.pixel_height;
2722 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2723 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2724 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2725 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2727 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2728 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2730 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2731 av_reduce(&st->sample_aspect_ratio.num,
2732 &st->sample_aspect_ratio.den,
2733 st->codecpar->height * track->video.display_width * display_width_mul,
2734 st->codecpar->width * track->video.display_height * display_height_mul,
2737 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2738 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2740 if (track->default_duration) {
2741 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2742 1000000000, track->default_duration, 30000);
2743 #if FF_API_R_FRAME_RATE
2744 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2745 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2746 st->r_frame_rate = st->avg_frame_rate;
2750 /* export stereo mode flag as metadata tag */
2751 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2752 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2754 /* export alpha mode flag as metadata tag */
2755 if (track->video.alpha_mode)
2756 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2758 /* if we have virtual track, mark the real tracks */
2759 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2761 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2763 snprintf(buf, sizeof(buf), "%s_%d",
2764 ff_matroska_video_stereo_plane[planes[j].type], i);
2765 for (k=0; k < matroska->tracks.nb_elem; k++)
2766 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2767 av_dict_set(&tracks[k].stream->metadata,
2768 "stereo_mode", buf, 0);
2772 // add stream level stereo3d side data if it is a supported format
2773 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2774 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2775 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2780 ret = mkv_parse_video_color(st, track);
2783 ret = mkv_parse_video_projection(st, track);
2786 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2787 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2788 st->codecpar->codec_tag = fourcc;
2789 st->codecpar->sample_rate = track->audio.out_samplerate;
2790 st->codecpar->channels = track->audio.channels;
2791 if (!st->codecpar->bits_per_coded_sample)
2792 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2793 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2794 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2795 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2796 st->need_parsing = AVSTREAM_PARSE_FULL;
2797 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2798 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2799 if (track->codec_delay > 0) {
2800 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2801 (AVRational){1, 1000000000},
2802 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2803 48000 : st->codecpar->sample_rate});
2805 if (track->seek_preroll > 0) {
2806 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2807 (AVRational){1, 1000000000},
2808 (AVRational){1, st->codecpar->sample_rate});
2810 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2811 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2813 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2814 st->disposition |= AV_DISPOSITION_CAPTIONS;
2815 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2816 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2817 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2818 st->disposition |= AV_DISPOSITION_METADATA;
2820 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2821 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2828 static int matroska_read_header(AVFormatContext *s)
2830 MatroskaDemuxContext *matroska = s->priv_data;
2831 EbmlList *attachments_list = &matroska->attachments;
2832 EbmlList *chapters_list = &matroska->chapters;
2833 MatroskaAttachment *attachments;
2834 MatroskaChapter *chapters;
2835 uint64_t max_start = 0;
2841 matroska->cues_parsing_deferred = 1;
2843 /* First read the EBML header. */
2844 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2845 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2846 ebml_free(ebml_syntax, &ebml);
2847 return AVERROR_INVALIDDATA;
2849 if (ebml.version > EBML_VERSION ||
2850 ebml.max_size > sizeof(uint64_t) ||
2851 ebml.id_length > sizeof(uint32_t) ||
2852 ebml.doctype_version > 3) {
2853 avpriv_report_missing_feature(matroska->ctx,
2854 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2855 ebml.version, ebml.doctype, ebml.doctype_version);
2856 ebml_free(ebml_syntax, &ebml);
2857 return AVERROR_PATCHWELCOME;
2858 } else if (ebml.doctype_version == 3) {
2859 av_log(matroska->ctx, AV_LOG_WARNING,
2860 "EBML header using unsupported features\n"
2861 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2862 ebml.version, ebml.doctype, ebml.doctype_version);
2864 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2865 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2867 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2868 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2869 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2870 ebml_free(ebml_syntax, &ebml);
2871 return AVERROR_INVALIDDATA;
2874 ebml_free(ebml_syntax, &ebml);
2876 /* The next thing is a segment. */
2877 pos = avio_tell(matroska->ctx->pb);
2878 res = ebml_parse(matroska, matroska_segments, matroska);
2879 // Try resyncing until we find an EBML_STOP type element.
2881 res = matroska_resync(matroska, pos);
2884 pos = avio_tell(matroska->ctx->pb);
2885 res = ebml_parse(matroska, matroska_segment, matroska);
2887 /* Set data_offset as it might be needed later by seek_frame_generic. */
2888 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2889 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2890 matroska_execute_seekhead(matroska);
2892 if (!matroska->time_scale)
2893 matroska->time_scale = 1000000;
2894 if (matroska->duration)
2895 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2896 1000 / AV_TIME_BASE;
2897 av_dict_set(&s->metadata, "title", matroska->title, 0);
2898 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2900 if (matroska->date_utc.size == 8)
2901 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2903 res = matroska_parse_tracks(s);
2907 attachments = attachments_list->elem;
2908 for (j = 0; j < attachments_list->nb_elem; j++) {
2909 if (!(attachments[j].filename && attachments[j].mime &&
2910 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2911 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2913 AVStream *st = avformat_new_stream(s, NULL);
2916 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2917 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2918 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2920 for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2921 if (!strncmp(mkv_image_mime_tags[i].str, attachments[j].mime,
2922 strlen(mkv_image_mime_tags[i].str))) {
2923 st->codecpar->codec_id = mkv_image_mime_tags[i].id;
2928 attachments[j].stream = st;
2930 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2931 AVPacket *pkt = &st->attached_pic;
2933 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2934 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2936 av_init_packet(pkt);
2937 pkt->buf = av_buffer_ref(attachments[j].bin.buf);
2939 return AVERROR(ENOMEM);
2940 pkt->data = attachments[j].bin.data;
2941 pkt->size = attachments[j].bin.size;
2942 pkt->stream_index = st->index;
2943 pkt->flags |= AV_PKT_FLAG_KEY;
2945 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2946 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2948 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2949 attachments[j].bin.size);
2951 for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2952 if (!strncmp(mkv_mime_tags[i].str, attachments[j].mime,
2953 strlen(mkv_mime_tags[i].str))) {
2954 st->codecpar->codec_id = mkv_mime_tags[i].id;
2962 chapters = chapters_list->elem;
2963 for (i = 0; i < chapters_list->nb_elem; i++)
2964 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2965 (max_start == 0 || chapters[i].start > max_start)) {
2966 chapters[i].chapter =
2967 avpriv_new_chapter(s, chapters[i].uid,
2968 (AVRational) { 1, 1000000000 },
2969 chapters[i].start, chapters[i].end,
2971 max_start = chapters[i].start;
2974 matroska_add_index_entries(matroska);
2976 matroska_convert_tags(s);
2980 matroska_read_close(s);
2985 * Put one packet in an application-supplied AVPacket struct.
2986 * Returns 0 on success or -1 on failure.
2988 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2991 if (matroska->queue) {
2992 MatroskaTrack *tracks = matroska->tracks.elem;
2993 MatroskaTrack *track;
2995 ff_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
2996 track = &tracks[pkt->stream_index];
2997 if (track->has_palette) {
2998 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
3000 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3002 memcpy(pal, track->palette, AVPALETTE_SIZE);
3004 track->has_palette = 0;
3013 * Free all packets in our internal queue.
3015 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3017 ff_packet_list_free(&matroska->queue, &matroska->queue_end);
3020 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3021 int size, int type, AVIOContext *pb,
3022 uint32_t lace_size[256], int *laces)
3025 uint8_t *data = *buf;
3029 lace_size[0] = size;
3034 return AVERROR_INVALIDDATA;
3041 case 0x1: /* Xiph lacing */
3045 for (n = 0; n < *laces - 1; n++) {
3049 if (size <= total) {
3050 return AVERROR_INVALIDDATA;
3054 lace_size[n] += temp;
3062 return AVERROR_INVALIDDATA;
3065 lace_size[n] = size - total;
3069 case 0x2: /* fixed-size lacing */
3070 if (size % (*laces)) {
3071 return AVERROR_INVALIDDATA;
3073 for (n = 0; n < *laces; n++)
3074 lace_size[n] = size / *laces;
3077 case 0x3: /* EBML lacing */
3085 n = ebml_read_num(matroska, pb, 8, &num, 1);
3089 return AVERROR_INVALIDDATA;
3091 total = lace_size[0] = num;
3093 for (n = 1; n < *laces - 1; n++) {
3096 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3099 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3100 return AVERROR_INVALIDDATA;
3102 lace_size[n] = lace_size[n - 1] + snum;
3103 total += lace_size[n];
3109 return AVERROR_INVALIDDATA;
3111 lace_size[*laces - 1] = size - total;
3121 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3122 MatroskaTrack *track, AVStream *st,
3123 uint8_t *data, int size, uint64_t timecode,
3126 const int a = st->codecpar->block_align;
3127 const int sps = track->audio.sub_packet_size;
3128 const int cfs = track->audio.coded_framesize;
3129 const int h = track->audio.sub_packet_h;
3130 const int w = track->audio.frame_size;
3131 int y = track->audio.sub_packet_cnt;
3134 if (!track->audio.pkt_cnt) {
3135 if (track->audio.sub_packet_cnt == 0)
3136 track->audio.buf_timecode = timecode;
3137 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3138 if (size < cfs * h / 2) {
3139 av_log(matroska->ctx, AV_LOG_ERROR,
3140 "Corrupt int4 RM-style audio packet size\n");
3141 return AVERROR_INVALIDDATA;
3143 for (x = 0; x < h / 2; x++)
3144 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3145 data + x * cfs, cfs);
3146 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3148 av_log(matroska->ctx, AV_LOG_ERROR,
3149 "Corrupt sipr RM-style audio packet size\n");
3150 return AVERROR_INVALIDDATA;
3152 memcpy(track->audio.buf + y * w, data, w);
3155 av_log(matroska->ctx, AV_LOG_ERROR,
3156 "Corrupt generic RM-style audio packet size\n");
3157 return AVERROR_INVALIDDATA;
3159 for (x = 0; x < w / sps; x++)
3160 memcpy(track->audio.buf +
3161 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3162 data + x * sps, sps);
3165 if (++track->audio.sub_packet_cnt >= h) {
3166 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3167 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3168 track->audio.sub_packet_cnt = 0;
3169 track->audio.pkt_cnt = h * w / a;
3173 while (track->audio.pkt_cnt) {
3175 AVPacket pktl, *pkt = &pktl;
3177 ret = av_new_packet(pkt, a);
3182 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3184 pkt->pts = track->audio.buf_timecode;
3185 track->audio.buf_timecode = AV_NOPTS_VALUE;
3187 pkt->stream_index = st->index;
3188 ret = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3190 av_packet_unref(pkt);
3191 return AVERROR(ENOMEM);
3198 /* reconstruct full wavpack blocks from mangled matroska ones */
3199 static int matroska_parse_wavpack(MatroskaTrack *track,
3200 uint8_t **data, int *size)
3202 uint8_t *dst = NULL;
3203 uint8_t *src = *data;
3208 int ret, offset = 0;
3211 return AVERROR_INVALIDDATA;
3213 av_assert1(track->stream->codecpar->extradata_size >= 2);
3214 ver = AV_RL16(track->stream->codecpar->extradata);
3216 samples = AV_RL32(src);
3220 while (srclen >= 8) {
3225 uint32_t flags = AV_RL32(src);
3226 uint32_t crc = AV_RL32(src + 4);
3230 multiblock = (flags & 0x1800) != 0x1800;
3233 ret = AVERROR_INVALIDDATA;
3236 blocksize = AV_RL32(src);
3242 if (blocksize > srclen) {
3243 ret = AVERROR_INVALIDDATA;
3247 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3249 ret = AVERROR(ENOMEM);
3253 dstlen += blocksize + 32;
3255 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3256 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3257 AV_WL16(dst + offset + 8, ver); // version
3258 AV_WL16(dst + offset + 10, 0); // track/index_no
3259 AV_WL32(dst + offset + 12, 0); // total samples
3260 AV_WL32(dst + offset + 16, 0); // block index
3261 AV_WL32(dst + offset + 20, samples); // number of samples
3262 AV_WL32(dst + offset + 24, flags); // flags
3263 AV_WL32(dst + offset + 28, crc); // crc
3264 memcpy(dst + offset + 32, src, blocksize); // block data
3267 srclen -= blocksize;
3268 offset += blocksize + 32;
3271 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3283 static int matroska_parse_prores(MatroskaTrack *track,
3284 uint8_t **data, int *size)
3287 int dstlen = *size + 8;
3289 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3291 return AVERROR(ENOMEM);
3293 AV_WB32(dst, dstlen);
3294 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3295 memcpy(dst + 8, *data, dstlen - 8);
3296 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3304 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3305 MatroskaTrack *track,
3307 uint8_t *data, int data_len,
3312 AVPacket pktl, *pkt = &pktl;
3313 uint8_t *id, *settings, *text, *buf;
3314 int id_len, settings_len, text_len;
3319 return AVERROR_INVALIDDATA;
3322 q = data + data_len;
3327 if (*p == '\r' || *p == '\n') {
3336 if (p >= q || *p != '\n')
3337 return AVERROR_INVALIDDATA;
3343 if (*p == '\r' || *p == '\n') {
3344 settings_len = p - settings;
3352 if (p >= q || *p != '\n')
3353 return AVERROR_INVALIDDATA;
3358 while (text_len > 0) {
3359 const int len = text_len - 1;
3360 const uint8_t c = p[len];
3361 if (c != '\r' && c != '\n')
3367 return AVERROR_INVALIDDATA;
3369 err = av_new_packet(pkt, text_len);
3374 memcpy(pkt->data, text, text_len);
3377 buf = av_packet_new_side_data(pkt,
3378 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3381 av_packet_unref(pkt);
3382 return AVERROR(ENOMEM);
3384 memcpy(buf, id, id_len);
3387 if (settings_len > 0) {
3388 buf = av_packet_new_side_data(pkt,
3389 AV_PKT_DATA_WEBVTT_SETTINGS,
3392 av_packet_unref(pkt);
3393 return AVERROR(ENOMEM);
3395 memcpy(buf, settings, settings_len);
3398 // Do we need this for subtitles?
3399 // pkt->flags = AV_PKT_FLAG_KEY;
3401 pkt->stream_index = st->index;
3402 pkt->pts = timecode;
3404 // Do we need this for subtitles?
3405 // pkt->dts = timecode;
3407 pkt->duration = duration;
3410 err = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3412 av_packet_unref(pkt);
3413 return AVERROR(ENOMEM);
3419 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3420 MatroskaTrack *track, AVStream *st,
3421 AVBufferRef *buf, uint8_t *data, int pkt_size,
3422 uint64_t timecode, uint64_t lace_duration,
3423 int64_t pos, int is_keyframe,
3424 uint8_t *additional, uint64_t additional_id, int additional_size,
3425 int64_t discard_padding)
3427 uint8_t *pkt_data = data;
3429 AVPacket pktl, *pkt = &pktl;
3431 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3432 res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3434 av_log(matroska->ctx, AV_LOG_ERROR,
3435 "Error parsing a wavpack block.\n");
3443 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3444 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3445 res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3447 av_log(matroska->ctx, AV_LOG_ERROR,
3448 "Error parsing a prores block.\n");
3456 if (!pkt_size && !additional_size)
3459 av_init_packet(pkt);
3461 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3464 pkt->buf = av_buffer_ref(buf);
3467 res = AVERROR(ENOMEM);
3471 pkt->data = pkt_data;
3472 pkt->size = pkt_size;
3473 pkt->flags = is_keyframe;
3474 pkt->stream_index = st->index;
3476 if (additional_size > 0) {
3477 uint8_t *side_data = av_packet_new_side_data(pkt,
3478 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3479 additional_size + 8);
3481 av_packet_unref(pkt);
3482 return AVERROR(ENOMEM);
3484 AV_WB64(side_data, additional_id);
3485 memcpy(side_data + 8, additional, additional_size);
3488 if (discard_padding) {
3489 uint8_t *side_data = av_packet_new_side_data(pkt,
3490 AV_PKT_DATA_SKIP_SAMPLES,
3493 av_packet_unref(pkt);
3494 return AVERROR(ENOMEM);
3496 discard_padding = av_rescale_q(discard_padding,
3497 (AVRational){1, 1000000000},
3498 (AVRational){1, st->codecpar->sample_rate});
3499 if (discard_padding > 0) {
3500 AV_WL32(side_data + 4, discard_padding);
3502 AV_WL32(side_data, -discard_padding);
3506 if (track->ms_compat)
3507 pkt->dts = timecode;
3509 pkt->pts = timecode;
3511 pkt->duration = lace_duration;
3513 #if FF_API_CONVERGENCE_DURATION
3514 FF_DISABLE_DEPRECATION_WARNINGS
3515 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3516 pkt->convergence_duration = lace_duration;
3518 FF_ENABLE_DEPRECATION_WARNINGS
3521 res = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3523 av_packet_unref(pkt);
3524 return AVERROR(ENOMEM);
3536 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3537 int size, int64_t pos, uint64_t cluster_time,
3538 uint64_t block_duration, int is_keyframe,
3539 uint8_t *additional, uint64_t additional_id, int additional_size,
3540 int64_t cluster_pos, int64_t discard_padding)
3542 uint64_t timecode = AV_NOPTS_VALUE;
3543 MatroskaTrack *track;
3548 uint32_t lace_size[256];
3549 int n, flags, laces = 0;
3551 int trust_default_duration = 1;
3553 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3555 if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3560 track = matroska_find_track_by_num(matroska, num);
3561 if (!track || size < 3)
3562 return AVERROR_INVALIDDATA;
3564 if (!(st = track->stream)) {
3565 av_log(matroska->ctx, AV_LOG_VERBOSE,
3566 "No stream associated to TrackNumber %"PRIu64". "
3567 "Ignoring Block with this TrackNumber.\n", num);
3571 if (st->discard >= AVDISCARD_ALL)
3573 av_assert1(block_duration != AV_NOPTS_VALUE);
3575 block_time = sign_extend(AV_RB16(data), 16);
3579 if (is_keyframe == -1)
3580 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3582 if (cluster_time != (uint64_t) -1 &&
3583 (block_time >= 0 || cluster_time >= -block_time)) {
3584 timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3585 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3586 timecode < track->end_timecode)
3587 is_keyframe = 0; /* overlapping subtitles are not key frame */
3589 ff_reduce_index(matroska->ctx, st->index);
3590 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3595 if (matroska->skip_to_keyframe &&
3596 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3597 // Compare signed timecodes. Timecode may be negative due to codec delay
3598 // offset. We don't support timestamps greater than int64_t anyway - see
3600 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3603 matroska->skip_to_keyframe = 0;
3604 else if (!st->skip_to_keyframe) {
3605 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3606 matroska->skip_to_keyframe = 0;
3610 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3611 &pb, lace_size, &laces);
3613 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3617 if (track->audio.samplerate == 8000) {
3618 // If this is needed for more codecs, then add them here
3619 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3620 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3621 trust_default_duration = 0;
3625 if (!block_duration && trust_default_duration)
3626 block_duration = track->default_duration * laces / matroska->time_scale;
3628 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3629 track->end_timecode =
3630 FFMAX(track->end_timecode, timecode + block_duration);
3632 for (n = 0; n < laces; n++) {
3633 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3634 uint8_t *out_data = data;
3635 int out_size = lace_size[n];
3637 if (track->needs_decoding) {
3638 res = matroska_decode_buffer(&out_data, &out_size, track);
3641 /* Given that we are here means that out_data is no longer
3642 * owned by buf, so set it to NULL. This depends upon
3643 * zero-length header removal compression being ignored. */
3644 av_assert1(out_data != data);
3648 if (track->audio.buf) {
3649 res = matroska_parse_rm_audio(matroska, track, st,
3656 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3657 res = matroska_parse_webvtt(matroska, track, st,
3659 timecode, lace_duration,
3666 res = matroska_parse_frame(matroska, track, st, buf, out_data,
3667 out_size, timecode, lace_duration,
3668 pos, !n ? is_keyframe : 0,
3669 additional, additional_id, additional_size,
3675 if (timecode != AV_NOPTS_VALUE)
3676 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3677 data += lace_size[n];
3683 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3685 MatroskaCluster *cluster = &matroska->current_cluster;
3686 MatroskaBlock *block = &cluster->block;
3689 av_assert0(matroska->num_levels <= 2);
3691 if (matroska->num_levels == 1) {
3692 res = ebml_parse(matroska, matroska_segment, NULL);
3695 /* Found a cluster: subtract the size of the ID already read. */
3696 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3698 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3704 if (matroska->num_levels == 2) {
3705 /* We are inside a cluster. */
3706 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3708 if (res >= 0 && block->bin.size > 0) {
3709 int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3710 uint8_t* additional = block->additional.size > 0 ?
3711 block->additional.data : NULL;
3713 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3714 block->bin.size, block->bin.pos,
3715 cluster->timecode, block->duration,
3716 is_keyframe, additional, block->additional_id,
3717 block->additional.size, cluster->pos,
3718 block->discard_padding);
3721 ebml_free(matroska_blockgroup, block);
3722 memset(block, 0, sizeof(*block));
3723 } else if (!matroska->num_levels) {
3724 if (!avio_feof(matroska->ctx->pb)) {
3725 avio_r8(matroska->ctx->pb);
3726 if (!avio_feof(matroska->ctx->pb)) {
3727 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3728 "end of segment.\n");
3729 return AVERROR_INVALIDDATA;
3739 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3741 MatroskaDemuxContext *matroska = s->priv_data;
3744 if (matroska->resync_pos == -1) {
3745 // This can only happen if generic seeking has been used.
3746 matroska->resync_pos = avio_tell(s->pb);
3749 while (matroska_deliver_packet(matroska, pkt)) {
3751 return (ret < 0) ? ret : AVERROR_EOF;
3752 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3753 ret = matroska_resync(matroska, matroska->resync_pos);
3759 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3760 int64_t timestamp, int flags)
3762 MatroskaDemuxContext *matroska = s->priv_data;
3763 MatroskaTrack *tracks = NULL;
3764 AVStream *st = s->streams[stream_index];
3767 /* Parse the CUES now since we need the index data to seek. */
3768 if (matroska->cues_parsing_deferred > 0) {
3769 matroska->cues_parsing_deferred = 0;
3770 matroska_parse_cues(matroska);
3773 if (!st->nb_index_entries)
3775 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3777 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3778 matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3779 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3780 matroska_clear_queue(matroska);
3781 if (matroska_parse_cluster(matroska) < 0)
3786 matroska_clear_queue(matroska);
3787 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3790 tracks = matroska->tracks.elem;
3791 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3792 tracks[i].audio.pkt_cnt = 0;
3793 tracks[i].audio.sub_packet_cnt = 0;
3794 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3795 tracks[i].end_timecode = 0;
3798 /* We seek to a level 1 element, so set the appropriate status. */
3799 matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3800 if (flags & AVSEEK_FLAG_ANY) {
3801 st->skip_to_keyframe = 0;
3802 matroska->skip_to_timecode = timestamp;
3804 st->skip_to_keyframe = 1;
3805 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3807 matroska->skip_to_keyframe = 1;
3809 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3812 // slightly hackish but allows proper fallback to
3813 // the generic seeking code.
3814 matroska_reset_status(matroska, 0, -1);
3815 matroska->resync_pos = -1;
3816 matroska_clear_queue(matroska);
3817 st->skip_to_keyframe =
3818 matroska->skip_to_keyframe = 0;
3823 static int matroska_read_close(AVFormatContext *s)
3825 MatroskaDemuxContext *matroska = s->priv_data;
3826 MatroskaTrack *tracks = matroska->tracks.elem;
3829 matroska_clear_queue(matroska);
3831 for (n = 0; n < matroska->tracks.nb_elem; n++)
3832 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3833 av_freep(&tracks[n].audio.buf);
3834 ebml_free(matroska_segment, matroska);
3840 int64_t start_time_ns;
3841 int64_t end_time_ns;
3842 int64_t start_offset;
3846 /* This function searches all the Cues and returns the CueDesc corresponding to
3847 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3848 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3850 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3851 MatroskaDemuxContext *matroska = s->priv_data;
3854 int nb_index_entries = s->streams[0]->nb_index_entries;
3855 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3856 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3857 for (i = 1; i < nb_index_entries; i++) {
3858 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3859 index_entries[i].timestamp * matroska->time_scale > ts) {
3864 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3865 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3866 if (i != nb_index_entries - 1) {
3867 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3868 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3870 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3871 // FIXME: this needs special handling for files where Cues appear
3872 // before Clusters. the current logic assumes Cues appear after
3874 cue_desc.end_offset = cues_start - matroska->segment_start;
3879 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3881 MatroskaDemuxContext *matroska = s->priv_data;
3882 uint32_t id = matroska->current_id;
3883 int64_t cluster_pos, before_pos;
3885 if (s->streams[0]->nb_index_entries <= 0) return 0;
3886 // seek to the first cluster using cues.
3887 index = av_index_search_timestamp(s->streams[0], 0, 0);
3888 if (index < 0) return 0;
3889 cluster_pos = s->streams[0]->index_entries[index].pos;
3890 before_pos = avio_tell(s->pb);
3892 uint64_t cluster_id, cluster_length;
3895 avio_seek(s->pb, cluster_pos, SEEK_SET);
3896 // read cluster id and length
3897 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3898 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3900 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3904 matroska_reset_status(matroska, 0, cluster_pos);
3905 matroska_clear_queue(matroska);
3906 if (matroska_parse_cluster(matroska) < 0 ||
3910 pkt = &matroska->queue->pkt;
3911 // 4 + read is the length of the cluster id and the cluster length field.
3912 cluster_pos += 4 + read + cluster_length;
3913 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3919 /* Restore the status after matroska_read_header: */
3920 matroska_reset_status(matroska, id, before_pos);
3925 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3926 double min_buffer, double* buffer,
3927 double* sec_to_download, AVFormatContext *s,
3930 double nano_seconds_per_second = 1000000000.0;
3931 double time_sec = time_ns / nano_seconds_per_second;
3933 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3934 int64_t end_time_ns = time_ns + time_to_search_ns;
3935 double sec_downloaded = 0.0;
3936 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3937 if (desc_curr.start_time_ns == -1)
3939 *sec_to_download = 0.0;
3941 // Check for non cue start time.
3942 if (time_ns > desc_curr.start_time_ns) {
3943 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3944 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3945 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3946 double timeToDownload = (cueBytes * 8.0) / bps;
3948 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3949 *sec_to_download += timeToDownload;
3951 // Check if the search ends within the first cue.
3952 if (desc_curr.end_time_ns >= end_time_ns) {
3953 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3954 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3955 sec_downloaded = percent_to_sub * sec_downloaded;
3956 *sec_to_download = percent_to_sub * *sec_to_download;
3959 if ((sec_downloaded + *buffer) <= min_buffer) {
3963 // Get the next Cue.
3964 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3967 while (desc_curr.start_time_ns != -1) {
3968 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3969 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3970 double desc_sec = desc_ns / nano_seconds_per_second;
3971 double bits = (desc_bytes * 8.0);
3972 double time_to_download = bits / bps;
3974 sec_downloaded += desc_sec - time_to_download;
3975 *sec_to_download += time_to_download;
3977 if (desc_curr.end_time_ns >= end_time_ns) {
3978 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3979 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3980 sec_downloaded = percent_to_sub * sec_downloaded;
3981 *sec_to_download = percent_to_sub * *sec_to_download;
3983 if ((sec_downloaded + *buffer) <= min_buffer)
3988 if ((sec_downloaded + *buffer) <= min_buffer) {
3993 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3995 *buffer = *buffer + sec_downloaded;
3999 /* This function computes the bandwidth of the WebM file with the help of
4000 * buffer_size_after_time_downloaded() function. Both of these functions are
4001 * adapted from WebM Tools project and are adapted to work with FFmpeg's
4002 * Matroska parsing mechanism.
4004 * Returns the bandwidth of the file on success; -1 on error.
4006 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4008 MatroskaDemuxContext *matroska = s->priv_data;
4009 AVStream *st = s->streams[0];
4010 double bandwidth = 0.0;
4013 for (i = 0; i < st->nb_index_entries; i++) {
4014 int64_t prebuffer_ns = 1000000000;
4015 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
4016 double nano_seconds_per_second = 1000000000.0;
4017 int64_t prebuffered_ns = time_ns + prebuffer_ns;
4018 double prebuffer_bytes = 0.0;
4019 int64_t temp_prebuffer_ns = prebuffer_ns;
4020 int64_t pre_bytes, pre_ns;
4021 double pre_sec, prebuffer, bits_per_second;
4022 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4024 // Start with the first Cue.
4025 CueDesc desc_end = desc_beg;
4027 // Figure out how much data we have downloaded for the prebuffer. This will
4028 // be used later to adjust the bits per sample to try.
4029 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4030 // Prebuffered the entire Cue.
4031 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4032 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4033 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4035 if (desc_end.start_time_ns == -1) {
4036 // The prebuffer is larger than the duration.
4037 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4039 bits_per_second = 0.0;
4041 // The prebuffer ends in the last Cue. Estimate how much data was
4043 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4044 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4045 pre_sec = pre_ns / nano_seconds_per_second;
4047 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4049 prebuffer = prebuffer_ns / nano_seconds_per_second;
4051 // Set this to 0.0 in case our prebuffer buffers the entire video.
4052 bits_per_second = 0.0;
4054 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4055 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4056 double desc_sec = desc_ns / nano_seconds_per_second;
4057 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4059 // Drop the bps by the percentage of bytes buffered.
4060 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4061 double mod_bits_per_second = calc_bits_per_second * percent;
4063 if (prebuffer < desc_sec) {
4065 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4067 // Add 1 so the bits per second should be a little bit greater than file
4069 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4070 const double min_buffer = 0.0;
4071 double buffer = prebuffer;
4072 double sec_to_download = 0.0;
4074 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4075 min_buffer, &buffer, &sec_to_download,
4079 } else if (rv == 0) {
4080 bits_per_second = (double)(bps);
4085 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4086 } while (desc_end.start_time_ns != -1);
4088 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4090 return (int64_t)bandwidth;
4093 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4095 MatroskaDemuxContext *matroska = s->priv_data;
4096 EbmlList *seekhead_list = &matroska->seekhead;
4097 MatroskaSeekhead *seekhead = seekhead_list->elem;
4099 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4103 // determine cues start and end positions
4104 for (i = 0; i < seekhead_list->nb_elem; i++)
4105 if (seekhead[i].id == MATROSKA_ID_CUES)
4108 if (i >= seekhead_list->nb_elem) return -1;
4110 before_pos = avio_tell(matroska->ctx->pb);
4111 cues_start = seekhead[i].pos + matroska->segment_start;
4112 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4113 // cues_end is computed as cues_start + cues_length + length of the
4114 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4115 // cues_end is inclusive and the above sum is reduced by 1.
4116 uint64_t cues_length, cues_id;
4118 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4119 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4120 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4121 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4124 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4126 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4127 if (cues_start == -1 || cues_end == -1) return -1;
4130 matroska_parse_cues(matroska);
4133 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4136 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4138 // if the file has cues at the start, fix up the init range so that
4139 // it does not include it
4140 if (cues_start <= init_range)
4141 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4144 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4145 if (bandwidth < 0) return -1;
4146 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4148 // check if all clusters start with key frames
4149 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4151 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4152 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4153 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4154 if (!buf) return -1;
4156 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4157 int ret = snprintf(buf + end, 20,
4158 "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4159 i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4160 if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4161 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4163 return AVERROR_INVALIDDATA;
4167 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4168 buf, AV_DICT_DONT_STRDUP_VAL);
4173 static int webm_dash_manifest_read_header(AVFormatContext *s)
4176 int ret = matroska_read_header(s);
4178 MatroskaTrack *tracks;
4179 MatroskaDemuxContext *matroska = s->priv_data;
4181 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4184 if (!s->nb_streams) {
4185 matroska_read_close(s);
4186 av_log(s, AV_LOG_ERROR, "No streams found\n");
4187 return AVERROR_INVALIDDATA;
4190 if (!matroska->is_live) {
4191 buf = av_asprintf("%g", matroska->duration);
4192 if (!buf) return AVERROR(ENOMEM);
4193 av_dict_set(&s->streams[0]->metadata, DURATION,
4194 buf, AV_DICT_DONT_STRDUP_VAL);
4196 // initialization range
4197 // 5 is the offset of Cluster ID.
4198 init_range = avio_tell(s->pb) - 5;
4199 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4202 // basename of the file
4203 buf = strrchr(s->url, '/');
4204 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4207 tracks = matroska->tracks.elem;
4208 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4210 // parse the cues and populate Cue related fields
4211 if (!matroska->is_live) {
4212 ret = webm_dash_manifest_cues(s, init_range);
4214 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4219 // use the bandwidth from the command line if it was provided
4220 if (matroska->bandwidth > 0) {
4221 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4222 matroska->bandwidth, 0);
4227 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4232 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4233 static const AVOption options[] = {
4234 { "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 },
4235 { "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 },
4239 static const AVClass webm_dash_class = {
4240 .class_name = "WebM DASH Manifest demuxer",
4241 .item_name = av_default_item_name,
4243 .version = LIBAVUTIL_VERSION_INT,
4246 AVInputFormat ff_matroska_demuxer = {
4247 .name = "matroska,webm",
4248 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4249 .extensions = "mkv,mk3d,mka,mks",
4250 .priv_data_size = sizeof(MatroskaDemuxContext),
4251 .read_probe = matroska_probe,
4252 .read_header = matroska_read_header,
4253 .read_packet = matroska_read_packet,
4254 .read_close = matroska_read_close,
4255 .read_seek = matroska_read_seek,
4256 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4259 AVInputFormat ff_webm_dash_manifest_demuxer = {
4260 .name = "webm_dash_manifest",
4261 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4262 .priv_data_size = sizeof(MatroskaDemuxContext),
4263 .read_header = webm_dash_manifest_read_header,
4264 .read_packet = webm_dash_manifest_read_packet,
4265 .read_close = matroska_read_close,
4266 .priv_class = &webm_dash_class,