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"
51 #include "libavcodec/packet_internal.h"
54 #include "avio_internal.h"
59 /* For ff_codec_get_id(). */
70 #include "qtpalette.h"
72 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
73 #define NEEDS_CHECKING 2 /* Indicates that some error checks
74 * still need to be performed */
75 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
76 * syntax level used for parsing ended. */
77 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
78 * of unkown, potentially damaged data is encountered,
79 * it is considered an error. */
80 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
81 * to this many bytes of unknown data for the
82 * SKIP_THRESHOLD check. */
98 typedef const struct EbmlSyntax {
101 size_t list_elem_size;
108 const struct EbmlSyntax *n;
112 typedef struct EbmlList {
114 unsigned int alloc_elem_size;
118 typedef struct EbmlBin {
125 typedef struct Ebml {
130 uint64_t doctype_version;
133 typedef struct MatroskaTrackCompression {
136 } MatroskaTrackCompression;
138 typedef struct MatroskaTrackEncryption {
141 } MatroskaTrackEncryption;
143 typedef struct MatroskaTrackEncoding {
146 MatroskaTrackCompression compression;
147 MatroskaTrackEncryption encryption;
148 } MatroskaTrackEncoding;
150 typedef struct MatroskaMasteringMeta {
159 double max_luminance;
160 double min_luminance;
161 } MatroskaMasteringMeta;
163 typedef struct MatroskaTrackVideoColor {
164 uint64_t matrix_coefficients;
165 uint64_t bits_per_channel;
166 uint64_t chroma_sub_horz;
167 uint64_t chroma_sub_vert;
168 uint64_t cb_sub_horz;
169 uint64_t cb_sub_vert;
170 uint64_t chroma_siting_horz;
171 uint64_t chroma_siting_vert;
173 uint64_t transfer_characteristics;
177 MatroskaMasteringMeta mastering_meta;
178 } MatroskaTrackVideoColor;
180 typedef struct MatroskaTrackVideoProjection {
186 } MatroskaTrackVideoProjection;
188 typedef struct MatroskaTrackVideo {
190 uint64_t display_width;
191 uint64_t display_height;
192 uint64_t pixel_width;
193 uint64_t pixel_height;
195 uint64_t display_unit;
197 uint64_t field_order;
198 uint64_t stereo_mode;
201 MatroskaTrackVideoProjection projection;
202 } MatroskaTrackVideo;
204 typedef struct MatroskaTrackAudio {
206 double out_samplerate;
210 /* real audio header (extracted from extradata) */
217 uint64_t buf_timecode;
219 } MatroskaTrackAudio;
221 typedef struct MatroskaTrackPlane {
224 } MatroskaTrackPlane;
226 typedef struct MatroskaTrackOperation {
227 EbmlList combine_planes;
228 } MatroskaTrackOperation;
230 typedef struct MatroskaTrack {
239 uint64_t default_duration;
240 uint64_t flag_default;
241 uint64_t flag_forced;
242 uint64_t seek_preroll;
243 MatroskaTrackVideo video;
244 MatroskaTrackAudio audio;
245 MatroskaTrackOperation operation;
247 uint64_t codec_delay;
248 uint64_t codec_delay_in_track_tb;
251 int64_t end_timecode;
254 uint64_t max_block_additional_id;
256 uint32_t palette[AVPALETTE_COUNT];
260 typedef struct MatroskaAttachment {
268 } MatroskaAttachment;
270 typedef struct MatroskaChapter {
279 typedef struct MatroskaIndexPos {
284 typedef struct MatroskaIndex {
289 typedef struct MatroskaTag {
297 typedef struct MatroskaTagTarget {
305 typedef struct MatroskaTags {
306 MatroskaTagTarget target;
310 typedef struct MatroskaSeekhead {
315 typedef struct MatroskaLevel {
320 typedef struct MatroskaBlock {
325 uint64_t additional_id;
327 int64_t discard_padding;
330 typedef struct MatroskaCluster {
336 typedef struct MatroskaLevel1Element {
340 } MatroskaLevel1Element;
342 typedef struct MatroskaDemuxContext {
343 const AVClass *class;
344 AVFormatContext *ctx;
347 MatroskaLevel levels[EBML_MAX_DEPTH];
359 EbmlList attachments;
365 /* byte position of the segment inside the stream */
366 int64_t segment_start;
368 /* the packet queue */
370 AVPacketList *queue_end;
374 /* What to skip before effectively reading a packet. */
375 int skip_to_keyframe;
376 uint64_t skip_to_timecode;
378 /* File has a CUES element, but we defer parsing until it is needed. */
379 int cues_parsing_deferred;
381 /* Level1 elements and whether they were read yet */
382 MatroskaLevel1Element level1_elems[64];
383 int num_level1_elems;
385 MatroskaCluster current_cluster;
387 /* WebM DASH Manifest live flag */
390 /* Bandwidth value for WebM DASH Manifest */
392 } MatroskaDemuxContext;
394 #define CHILD_OF(parent) { .def = { .n = parent } }
396 // The following forward declarations need their size because
397 // a tentative definition with internal linkage must not be an
398 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
399 // Removing the sizes breaks MSVC.
400 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
401 matroska_track[27], matroska_track_encoding[6], matroska_track_encodings[2],
402 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_tracks[2],
403 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
404 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
405 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
407 static EbmlSyntax ebml_header[] = {
408 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
409 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
410 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
411 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
412 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
413 { EBML_ID_EBMLVERSION, EBML_NONE },
414 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
415 CHILD_OF(ebml_syntax)
418 static EbmlSyntax ebml_syntax[] = {
419 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
420 { MATROSKA_ID_SEGMENT, EBML_STOP },
424 static EbmlSyntax matroska_info[] = {
425 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
426 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
427 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
428 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
429 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
430 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
431 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
432 CHILD_OF(matroska_segment)
435 static EbmlSyntax matroska_mastering_meta[] = {
436 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
437 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
438 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
439 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
440 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
441 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
442 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
443 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
444 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
445 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
446 CHILD_OF(matroska_track_video_color)
449 static EbmlSyntax matroska_track_video_color[] = {
450 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
451 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=0 } },
452 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
453 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
454 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
455 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
456 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
457 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
458 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
459 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
460 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
461 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
462 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
463 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
464 CHILD_OF(matroska_track_video)
467 static EbmlSyntax matroska_track_video_projection[] = {
468 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
469 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrackVideoProjection, private) },
470 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f=0.0 } },
471 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f=0.0 } },
472 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f=0.0 } },
473 CHILD_OF(matroska_track_video)
476 static EbmlSyntax matroska_track_video[] = {
477 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
478 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
479 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
480 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
481 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
482 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
483 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
484 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
485 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
486 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
487 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
488 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
489 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
490 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
491 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
492 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
493 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
494 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
495 CHILD_OF(matroska_track)
498 static EbmlSyntax matroska_track_audio[] = {
499 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
500 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
501 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
502 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
503 CHILD_OF(matroska_track)
506 static EbmlSyntax matroska_track_encoding_compression[] = {
507 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
508 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
509 CHILD_OF(matroska_track_encoding)
512 static EbmlSyntax matroska_track_encoding_encryption[] = {
513 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
514 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
515 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
516 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
517 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
518 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
519 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
520 CHILD_OF(matroska_track_encoding)
522 static EbmlSyntax matroska_track_encoding[] = {
523 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
524 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
525 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
526 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
527 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
528 CHILD_OF(matroska_track_encodings)
531 static EbmlSyntax matroska_track_encodings[] = {
532 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
533 CHILD_OF(matroska_track)
536 static EbmlSyntax matroska_track_plane[] = {
537 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
538 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
539 CHILD_OF(matroska_track_combine_planes)
542 static EbmlSyntax matroska_track_combine_planes[] = {
543 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
544 CHILD_OF(matroska_track_operation)
547 static EbmlSyntax matroska_track_operation[] = {
548 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
549 CHILD_OF(matroska_track)
552 static EbmlSyntax matroska_track[] = {
553 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
554 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
555 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
556 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
557 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
558 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
559 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
560 { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
561 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
562 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
563 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
564 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
565 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
566 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
567 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
568 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
569 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
570 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
571 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
572 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
573 { MATROSKA_ID_CODECNAME, EBML_NONE },
574 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
575 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
576 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
577 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
578 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
579 CHILD_OF(matroska_tracks)
582 static EbmlSyntax matroska_tracks[] = {
583 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
584 CHILD_OF(matroska_segment)
587 static EbmlSyntax matroska_attachment[] = {
588 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
589 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
590 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
591 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
592 { MATROSKA_ID_FILEDESC, EBML_UTF8, 0, offsetof(MatroskaAttachment, description) },
593 CHILD_OF(matroska_attachments)
596 static EbmlSyntax matroska_attachments[] = {
597 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
598 CHILD_OF(matroska_segment)
601 static EbmlSyntax matroska_chapter_display[] = {
602 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
603 { MATROSKA_ID_CHAPLANG, EBML_NONE },
604 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
605 CHILD_OF(matroska_chapter_entry)
608 static EbmlSyntax matroska_chapter_entry[] = {
609 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
610 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
611 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
612 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
613 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
614 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
615 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
616 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
617 CHILD_OF(matroska_chapter)
620 static EbmlSyntax matroska_chapter[] = {
621 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
622 { MATROSKA_ID_EDITIONUID, EBML_NONE },
623 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
624 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
625 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
626 CHILD_OF(matroska_chapters)
629 static EbmlSyntax matroska_chapters[] = {
630 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
631 CHILD_OF(matroska_segment)
634 static EbmlSyntax matroska_index_pos[] = {
635 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
636 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
637 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
638 { MATROSKA_ID_CUEDURATION, EBML_NONE },
639 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
640 CHILD_OF(matroska_index_entry)
643 static EbmlSyntax matroska_index_entry[] = {
644 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
645 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
646 CHILD_OF(matroska_index)
649 static EbmlSyntax matroska_index[] = {
650 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
651 CHILD_OF(matroska_segment)
654 static EbmlSyntax matroska_simpletag[] = {
655 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
656 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
657 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
658 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
659 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
660 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
661 CHILD_OF(matroska_tag)
664 static EbmlSyntax matroska_tagtargets[] = {
665 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
666 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
667 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
668 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
669 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
670 CHILD_OF(matroska_tag)
673 static EbmlSyntax matroska_tag[] = {
674 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
675 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
676 CHILD_OF(matroska_tags)
679 static EbmlSyntax matroska_tags[] = {
680 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
681 CHILD_OF(matroska_segment)
684 static EbmlSyntax matroska_seekhead_entry[] = {
685 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
686 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
687 CHILD_OF(matroska_seekhead)
690 static EbmlSyntax matroska_seekhead[] = {
691 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
692 CHILD_OF(matroska_segment)
695 static EbmlSyntax matroska_segment[] = {
696 { MATROSKA_ID_CLUSTER, EBML_STOP },
697 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
698 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
699 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
700 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
701 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
702 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
703 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
704 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
707 static EbmlSyntax matroska_segments[] = {
708 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
712 static EbmlSyntax matroska_blockmore[] = {
713 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
714 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
715 CHILD_OF(matroska_blockadditions)
718 static EbmlSyntax matroska_blockadditions[] = {
719 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
720 CHILD_OF(matroska_blockgroup)
723 static EbmlSyntax matroska_blockgroup[] = {
724 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
725 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
726 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
727 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
728 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference), { .i = INT64_MIN } },
729 { MATROSKA_ID_CODECSTATE, EBML_NONE },
730 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
731 CHILD_OF(matroska_cluster_parsing)
734 // The following array contains SimpleBlock and BlockGroup twice
735 // in order to reuse the other values for matroska_cluster_enter.
736 static EbmlSyntax matroska_cluster_parsing[] = {
737 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
738 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, { .n = matroska_blockgroup } },
739 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
740 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
741 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
742 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
743 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
744 CHILD_OF(matroska_segment)
747 static EbmlSyntax matroska_cluster_enter[] = {
748 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = &matroska_cluster_parsing[2] } },
753 static const CodecMime mkv_image_mime_tags[] = {
754 {"image/gif" , AV_CODEC_ID_GIF},
755 {"image/jpeg" , AV_CODEC_ID_MJPEG},
756 {"image/png" , AV_CODEC_ID_PNG},
757 {"image/tiff" , AV_CODEC_ID_TIFF},
759 {"" , AV_CODEC_ID_NONE}
762 static const CodecMime mkv_mime_tags[] = {
763 {"text/plain" , AV_CODEC_ID_TEXT},
764 {"application/x-truetype-font", AV_CODEC_ID_TTF},
765 {"application/x-font" , AV_CODEC_ID_TTF},
766 {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
767 {"binary" , AV_CODEC_ID_BIN_DATA},
769 {"" , AV_CODEC_ID_NONE}
772 static const char *const matroska_doctypes[] = { "matroska", "webm" };
774 static int matroska_read_close(AVFormatContext *s);
777 * This function prepares the status for parsing of level 1 elements.
779 static int matroska_reset_status(MatroskaDemuxContext *matroska,
780 uint32_t id, int64_t position)
783 int64_t err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
788 matroska->current_id = id;
789 matroska->num_levels = 1;
790 matroska->unknown_count = 0;
791 matroska->resync_pos = avio_tell(matroska->ctx->pb);
793 matroska->resync_pos -= (av_log2(id) + 7) / 8;
798 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
800 AVIOContext *pb = matroska->ctx->pb;
803 /* Try to seek to the last position to resync from. If this doesn't work,
804 * we resync from the earliest position available: The start of the buffer. */
805 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
806 av_log(matroska->ctx, AV_LOG_WARNING,
807 "Seek to desired resync point failed. Seeking to "
808 "earliest point available instead.\n");
809 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
810 last_pos + 1), SEEK_SET);
815 // try to find a toplevel element
816 while (!avio_feof(pb)) {
817 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
818 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
819 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
820 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
821 /* Prepare the context for parsing of a level 1 element. */
822 matroska_reset_status(matroska, id, -1);
823 /* Given that we are here means that an error has occurred,
824 * so treat the segment as unknown length in order not to
825 * discard valid data that happens to be beyond the designated
826 * end of the segment. */
827 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
830 id = (id << 8) | avio_r8(pb);
834 return pb->error ? pb->error : AVERROR_EOF;
838 * Read: an "EBML number", which is defined as a variable-length
839 * array of bytes. The first byte indicates the length by giving a
840 * number of 0-bits followed by a one. The position of the first
841 * "one" bit inside the first byte indicates the length of this
843 * Returns: number of bytes read, < 0 on error
845 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
846 int max_size, uint64_t *number, int eof_forbidden)
852 /* The first byte tells us the length in bytes - except when it is zero. */
857 /* get the length of the EBML number */
858 read = 8 - ff_log2_tab[total];
860 if (!total || read > max_size) {
861 pos = avio_tell(pb) - 1;
863 av_log(matroska->ctx, AV_LOG_ERROR,
864 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
865 "of an EBML number\n", pos, pos);
867 av_log(matroska->ctx, AV_LOG_ERROR,
868 "Length %d indicated by an EBML number's first byte 0x%02x "
869 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
870 read, (uint8_t) total, pos, pos, max_size);
872 return AVERROR_INVALIDDATA;
875 /* read out length */
876 total ^= 1 << ff_log2_tab[total];
878 total = (total << 8) | avio_r8(pb);
880 if (pb->eof_reached) {
892 av_log(matroska->ctx, AV_LOG_ERROR,
893 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
898 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
899 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
906 * Read a EBML length value.
907 * This needs special handling for the "unknown length" case which has multiple
910 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
913 int res = ebml_read_num(matroska, pb, 8, number, 1);
914 if (res > 0 && *number + 1 == 1ULL << (7 * res))
915 *number = EBML_UNKNOWN_LENGTH;
920 * Read the next element as an unsigned int.
921 * Returns NEEDS_CHECKING.
923 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
927 /* big-endian ordering; build up number */
930 *num = (*num << 8) | avio_r8(pb);
932 return NEEDS_CHECKING;
936 * Read the next element as a signed int.
937 * Returns NEEDS_CHECKING.
939 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
946 *num = sign_extend(avio_r8(pb), 8);
948 /* big-endian ordering; build up number */
950 *num = ((uint64_t)*num << 8) | avio_r8(pb);
953 return NEEDS_CHECKING;
957 * Read the next element as a float.
958 * Returns NEEDS_CHECKING or < 0 on obvious failure.
960 static int ebml_read_float(AVIOContext *pb, int size, double *num)
965 *num = av_int2float(avio_rb32(pb));
967 *num = av_int2double(avio_rb64(pb));
969 return AVERROR_INVALIDDATA;
971 return NEEDS_CHECKING;
975 * Read the next element as an ASCII string.
976 * 0 is success, < 0 or NEEDS_CHECKING is failure.
978 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
983 /* EBML strings are usually not 0-terminated, so we allocate one
984 * byte more, read the string and NULL-terminate it ourselves. */
985 if (!(res = av_malloc(size + 1)))
986 return AVERROR(ENOMEM);
987 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
989 return ret < 0 ? ret : NEEDS_CHECKING;
999 * Read the next element as binary data.
1000 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1002 static int ebml_read_binary(AVIOContext *pb, int length,
1003 int64_t pos, EbmlBin *bin)
1007 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1010 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1012 bin->data = bin->buf->data;
1015 if ((ret = avio_read(pb, bin->data, length)) != length) {
1016 av_buffer_unref(&bin->buf);
1019 return ret < 0 ? ret : NEEDS_CHECKING;
1026 * Read the next element, but only the header. The contents
1027 * are supposed to be sub-elements which can be read separately.
1028 * 0 is success, < 0 is failure.
1030 static int ebml_read_master(MatroskaDemuxContext *matroska,
1031 uint64_t length, int64_t pos)
1033 MatroskaLevel *level;
1035 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1036 av_log(matroska->ctx, AV_LOG_ERROR,
1037 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1038 return AVERROR(ENOSYS);
1041 level = &matroska->levels[matroska->num_levels++];
1043 level->length = length;
1049 * Read a signed "EBML number"
1050 * Return: number of bytes processed, < 0 on error
1052 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1053 AVIOContext *pb, int64_t *num)
1058 /* read as unsigned number first */
1059 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1062 /* make signed (weird way) */
1063 *num = unum - ((1LL << (7 * res - 1)) - 1);
1068 static int ebml_parse(MatroskaDemuxContext *matroska,
1069 EbmlSyntax *syntax, void *data);
1071 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1075 // Whoever touches this should be aware of the duplication
1076 // existing in matroska_cluster_parsing.
1077 for (i = 0; syntax[i].id; i++)
1078 if (id == syntax[i].id)
1084 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1090 for (int i = 0; syntax[i].id; i++)
1091 switch (syntax[i].type) {
1093 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1096 *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1099 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1103 // the default may be NULL
1104 if (syntax[i].def.s) {
1105 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1106 *dst = av_strdup(syntax[i].def.s);
1108 return AVERROR(ENOMEM);
1113 if (!matroska->levels[matroska->num_levels - 1].length) {
1114 matroska->num_levels--;
1120 res = ebml_parse(matroska, syntax, data);
1123 return res == LEVEL_ENDED ? 0 : res;
1126 static int is_ebml_id_valid(uint32_t id)
1128 // Due to endian nonsense in Matroska, the highest byte with any bits set
1129 // will contain the leading length bit. This bit in turn identifies the
1130 // total byte length of the element by its position within the byte.
1131 unsigned int bits = av_log2(id);
1132 return id && (bits + 7) / 8 == (8 - bits % 8);
1136 * Allocate and return the entry for the level1 element with the given ID. If
1137 * an entry already exists, return the existing entry.
1139 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1140 uint32_t id, int64_t pos)
1143 MatroskaLevel1Element *elem;
1145 if (!is_ebml_id_valid(id))
1148 // Some files link to all clusters; useless.
1149 if (id == MATROSKA_ID_CLUSTER)
1152 // There can be multiple SeekHeads and Tags.
1153 for (i = 0; i < matroska->num_level1_elems; i++) {
1154 if (matroska->level1_elems[i].id == id) {
1155 if (matroska->level1_elems[i].pos == pos ||
1156 id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1157 return &matroska->level1_elems[i];
1161 // Only a completely broken file would have more elements.
1162 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1163 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1167 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1168 *elem = (MatroskaLevel1Element){.id = id};
1173 static int ebml_parse(MatroskaDemuxContext *matroska,
1174 EbmlSyntax *syntax, void *data)
1176 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1177 // Forbid unknown-length EBML_NONE elements.
1178 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1182 // max. 16 MB for strings
1183 [EBML_STR] = 0x1000000,
1184 [EBML_UTF8] = 0x1000000,
1185 // max. 256 MB for binary data
1186 [EBML_BIN] = 0x10000000,
1187 // no limits for anything else
1189 AVIOContext *pb = matroska->ctx->pb;
1192 int64_t pos = avio_tell(pb), pos_alt;
1193 int res, update_pos = 1, level_check;
1194 MatroskaLevel1Element *level1_elem;
1195 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1197 if (!matroska->current_id) {
1199 res = ebml_read_num(matroska, pb, 4, &id, 0);
1201 if (pb->eof_reached && res == AVERROR_EOF) {
1202 if (matroska->is_live)
1203 // in live mode, finish parsing if EOF is reached.
1205 if (level && pos == avio_tell(pb)) {
1206 if (level->length == EBML_UNKNOWN_LENGTH) {
1207 // Unknown-length levels automatically end at EOF.
1208 matroska->num_levels--;
1211 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1212 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1218 matroska->current_id = id | 1 << 7 * res;
1219 pos_alt = pos + res;
1222 pos -= (av_log2(matroska->current_id) + 7) / 8;
1225 id = matroska->current_id;
1227 syntax = ebml_parse_id(syntax, id);
1228 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1229 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1230 // Unknown-length levels end when an element from an upper level
1231 // in the hierarchy is encountered.
1232 while (syntax->def.n) {
1233 syntax = ebml_parse_id(syntax->def.n, id);
1235 matroska->num_levels--;
1241 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1242 "%"PRId64"\n", id, pos);
1243 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1247 data = (char *) data + syntax->data_offset;
1248 if (syntax->list_elem_size) {
1249 EbmlList *list = data;
1252 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1253 return AVERROR(ENOMEM);
1254 newelem = av_fast_realloc(list->elem,
1255 &list->alloc_elem_size,
1256 (list->nb_elem + 1) * syntax->list_elem_size);
1258 return AVERROR(ENOMEM);
1259 list->elem = newelem;
1260 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1261 memset(data, 0, syntax->list_elem_size);
1266 if (syntax->type != EBML_STOP) {
1267 matroska->current_id = 0;
1268 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1273 if (matroska->num_levels > 0) {
1274 if (length != EBML_UNKNOWN_LENGTH &&
1275 level->length != EBML_UNKNOWN_LENGTH) {
1276 uint64_t elem_end = pos_alt + length,
1277 level_end = level->start + level->length;
1279 if (elem_end < level_end) {
1281 } else if (elem_end == level_end) {
1282 level_check = LEVEL_ENDED;
1284 av_log(matroska->ctx, AV_LOG_ERROR,
1285 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1286 "containing master element ending at 0x%"PRIx64"\n",
1287 pos, elem_end, level_end);
1288 return AVERROR_INVALIDDATA;
1290 } else if (length != EBML_UNKNOWN_LENGTH) {
1292 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1293 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1294 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1295 return AVERROR_INVALIDDATA;
1298 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1299 || syntax->type == EBML_NEST)) {
1300 // According to the current specifications only clusters and
1301 // segments are allowed to be unknown-length. We also accept
1302 // other unknown-length master elements.
1303 av_log(matroska->ctx, AV_LOG_WARNING,
1304 "Found unknown-length element 0x%"PRIX32" other than "
1305 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1306 "parsing will nevertheless be attempted.\n", id, pos);
1313 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1314 if (length != EBML_UNKNOWN_LENGTH) {
1315 av_log(matroska->ctx, AV_LOG_ERROR,
1316 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1317 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1318 length, max_lengths[syntax->type], id, pos);
1319 } else if (syntax->type != EBML_NONE) {
1320 av_log(matroska->ctx, AV_LOG_ERROR,
1321 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1322 "unknown length, yet the length of an element of its "
1323 "type must be known.\n", id, pos);
1325 av_log(matroska->ctx, AV_LOG_ERROR,
1326 "Found unknown-length element with ID 0x%"PRIX32" at "
1327 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1328 "available.\n", id, pos);
1330 return AVERROR_INVALIDDATA;
1333 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1334 // Loosing sync will likely manifest itself as encountering unknown
1335 // elements which are not reliably distinguishable from elements
1336 // belonging to future extensions of the format.
1337 // We use a heuristic to detect such situations: If the current
1338 // element is not expected at the current syntax level and there
1339 // were only a few unknown elements in a row, then the element is
1340 // skipped or considered defective based upon the length of the
1341 // current element (i.e. how much would be skipped); if there were
1342 // more than a few skipped elements in a row and skipping the current
1343 // element would lead us more than SKIP_THRESHOLD away from the last
1344 // known good position, then it is inferred that an error occurred.
1345 // The dependency on the number of unknown elements in a row exists
1346 // because the distance to the last known good position is
1347 // automatically big if the last parsed element was big.
1348 // In both cases, each unknown element is considered equivalent to
1349 // UNKNOWN_EQUIV of skipped bytes for the check.
1350 // The whole check is only done for non-seekable output, because
1351 // in this situation skipped data can't simply be rechecked later.
1352 // This is especially important when using unkown length elements
1353 // as the check for whether a child exceeds its containing master
1354 // element is not effective in this situation.
1356 matroska->unknown_count = 0;
1358 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1360 if (matroska->unknown_count > 3)
1361 dist += pos_alt - matroska->resync_pos;
1363 if (dist > SKIP_THRESHOLD) {
1364 av_log(matroska->ctx, AV_LOG_ERROR,
1365 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1366 "length 0x%"PRIx64" considered as invalid data. Last "
1367 "known good position 0x%"PRIx64", %d unknown elements"
1368 " in a row\n", id, pos, length, matroska->resync_pos,
1369 matroska->unknown_count);
1370 return AVERROR_INVALIDDATA;
1375 if (update_pos > 0) {
1376 // We have found an element that is allowed at this place
1377 // in the hierarchy and it passed all checks, so treat the beginning
1378 // of the element as the "last known good" position.
1379 matroska->resync_pos = pos;
1382 if (!data && length != EBML_UNKNOWN_LENGTH)
1386 switch (syntax->type) {
1388 res = ebml_read_uint(pb, length, data);
1391 res = ebml_read_sint(pb, length, data);
1394 res = ebml_read_float(pb, length, data);
1398 res = ebml_read_ascii(pb, length, data);
1401 res = ebml_read_binary(pb, length, pos_alt, data);
1405 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1407 if (id == MATROSKA_ID_SEGMENT)
1408 matroska->segment_start = pos_alt;
1409 if (id == MATROSKA_ID_CUES)
1410 matroska->cues_parsing_deferred = 0;
1411 if (syntax->type == EBML_LEVEL1 &&
1412 (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1413 if (!level1_elem->pos) {
1414 // Zero is not a valid position for a level 1 element.
1415 level1_elem->pos = pos;
1416 } else if (level1_elem->pos != pos)
1417 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1418 level1_elem->parsed = 1;
1420 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1429 if (ffio_limit(pb, length) != length) {
1430 // ffio_limit emits its own error message,
1431 // so we don't have to.
1432 return AVERROR(EIO);
1434 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1435 // avio_skip might take us past EOF. We check for this
1436 // by skipping only length - 1 bytes, reading a byte and
1437 // checking the error flags. This is done in order to check
1438 // that the element has been properly skipped even when
1439 // no filesize (that ffio_limit relies on) is available.
1441 res = NEEDS_CHECKING;
1448 if (res == NEEDS_CHECKING) {
1449 if (pb->eof_reached) {
1458 if (res == AVERROR_INVALIDDATA)
1459 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1460 else if (res == AVERROR(EIO))
1461 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1462 else if (res == AVERROR_EOF) {
1463 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1471 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1472 level = &matroska->levels[matroska->num_levels - 1];
1473 pos = avio_tell(pb);
1475 // Given that pos >= level->start no check for
1476 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1477 while (matroska->num_levels && pos == level->start + level->length) {
1478 matroska->num_levels--;
1486 static void ebml_free(EbmlSyntax *syntax, void *data)
1489 for (i = 0; syntax[i].id; i++) {
1490 void *data_off = (char *) data + syntax[i].data_offset;
1491 switch (syntax[i].type) {
1497 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1501 if (syntax[i].list_elem_size) {
1502 EbmlList *list = data_off;
1503 char *ptr = list->elem;
1504 for (j = 0; j < list->nb_elem;
1505 j++, ptr += syntax[i].list_elem_size)
1506 ebml_free(syntax[i].def.n, ptr);
1507 av_freep(&list->elem);
1509 list->alloc_elem_size = 0;
1511 ebml_free(syntax[i].def.n, data_off);
1521 static int matroska_probe(const AVProbeData *p)
1524 int len_mask = 0x80, size = 1, n = 1, i;
1527 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1530 /* length of header */
1532 while (size <= 8 && !(total & len_mask)) {
1538 total &= (len_mask - 1);
1540 total = (total << 8) | p->buf[4 + n++];
1542 if (total + 1 == 1ULL << (7 * size)){
1543 /* Unknown-length header - simply parse the whole buffer. */
1544 total = p->buf_size - 4 - size;
1546 /* Does the probe data contain the whole header? */
1547 if (p->buf_size < 4 + size + total)
1551 /* The header should contain a known document type. For now,
1552 * we don't parse the whole header but simply check for the
1553 * availability of that array of characters inside the header.
1554 * Not fully fool-proof, but good enough. */
1555 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1556 size_t probelen = strlen(matroska_doctypes[i]);
1557 if (total < probelen)
1559 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1560 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1561 return AVPROBE_SCORE_MAX;
1564 // probably valid EBML header but no recognized doctype
1565 return AVPROBE_SCORE_EXTENSION;
1568 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1571 MatroskaTrack *tracks = matroska->tracks.elem;
1574 for (i = 0; i < matroska->tracks.nb_elem; i++)
1575 if (tracks[i].num == num)
1578 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1582 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1583 MatroskaTrack *track)
1585 MatroskaTrackEncoding *encodings = track->encodings.elem;
1586 uint8_t *data = *buf;
1587 int isize = *buf_size;
1588 uint8_t *pkt_data = NULL;
1589 uint8_t av_unused *newpktdata;
1590 int pkt_size = isize;
1594 if (pkt_size >= 10000000U)
1595 return AVERROR_INVALIDDATA;
1597 switch (encodings[0].compression.algo) {
1598 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1600 int header_size = encodings[0].compression.settings.size;
1601 uint8_t *header = encodings[0].compression.settings.data;
1603 if (header_size && !header) {
1604 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1611 pkt_size = isize + header_size;
1612 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1614 return AVERROR(ENOMEM);
1616 memcpy(pkt_data, header, header_size);
1617 memcpy(pkt_data + header_size, data, isize);
1621 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1624 olen = pkt_size *= 3;
1625 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1626 + AV_INPUT_BUFFER_PADDING_SIZE);
1628 result = AVERROR(ENOMEM);
1631 pkt_data = newpktdata;
1632 result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1633 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1635 result = AVERROR_INVALIDDATA;
1642 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1644 z_stream zstream = { 0 };
1645 if (inflateInit(&zstream) != Z_OK)
1647 zstream.next_in = data;
1648 zstream.avail_in = isize;
1651 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1653 inflateEnd(&zstream);
1654 result = AVERROR(ENOMEM);
1657 pkt_data = newpktdata;
1658 zstream.avail_out = pkt_size - zstream.total_out;
1659 zstream.next_out = pkt_data + zstream.total_out;
1660 result = inflate(&zstream, Z_NO_FLUSH);
1661 } while (result == Z_OK && pkt_size < 10000000);
1662 pkt_size = zstream.total_out;
1663 inflateEnd(&zstream);
1664 if (result != Z_STREAM_END) {
1665 if (result == Z_MEM_ERROR)
1666 result = AVERROR(ENOMEM);
1668 result = AVERROR_INVALIDDATA;
1675 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1677 bz_stream bzstream = { 0 };
1678 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1680 bzstream.next_in = data;
1681 bzstream.avail_in = isize;
1684 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1686 BZ2_bzDecompressEnd(&bzstream);
1687 result = AVERROR(ENOMEM);
1690 pkt_data = newpktdata;
1691 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1692 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1693 result = BZ2_bzDecompress(&bzstream);
1694 } while (result == BZ_OK && pkt_size < 10000000);
1695 pkt_size = bzstream.total_out_lo32;
1696 BZ2_bzDecompressEnd(&bzstream);
1697 if (result != BZ_STREAM_END) {
1698 if (result == BZ_MEM_ERROR)
1699 result = AVERROR(ENOMEM);
1701 result = AVERROR_INVALIDDATA;
1708 return AVERROR_INVALIDDATA;
1711 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1714 *buf_size = pkt_size;
1722 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1723 AVDictionary **metadata, char *prefix)
1725 MatroskaTag *tags = list->elem;
1729 for (i = 0; i < list->nb_elem; i++) {
1730 const char *lang = tags[i].lang &&
1731 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1733 if (!tags[i].name) {
1734 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1738 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1740 av_strlcpy(key, tags[i].name, sizeof(key));
1741 if (tags[i].def || !lang) {
1742 av_dict_set(metadata, key, tags[i].string, 0);
1743 if (tags[i].sub.nb_elem)
1744 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1747 av_strlcat(key, "-", sizeof(key));
1748 av_strlcat(key, lang, sizeof(key));
1749 av_dict_set(metadata, key, tags[i].string, 0);
1750 if (tags[i].sub.nb_elem)
1751 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1754 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1757 static void matroska_convert_tags(AVFormatContext *s)
1759 MatroskaDemuxContext *matroska = s->priv_data;
1760 MatroskaTags *tags = matroska->tags.elem;
1763 for (i = 0; i < matroska->tags.nb_elem; i++) {
1764 if (tags[i].target.attachuid) {
1765 MatroskaAttachment *attachment = matroska->attachments.elem;
1767 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1768 if (attachment[j].uid == tags[i].target.attachuid &&
1769 attachment[j].stream) {
1770 matroska_convert_tag(s, &tags[i].tag,
1771 &attachment[j].stream->metadata, NULL);
1776 av_log(s, AV_LOG_WARNING,
1777 "The tags at index %d refer to a "
1778 "non-existent attachment %"PRId64".\n",
1779 i, tags[i].target.attachuid);
1781 } else if (tags[i].target.chapteruid) {
1782 MatroskaChapter *chapter = matroska->chapters.elem;
1784 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1785 if (chapter[j].uid == tags[i].target.chapteruid &&
1786 chapter[j].chapter) {
1787 matroska_convert_tag(s, &tags[i].tag,
1788 &chapter[j].chapter->metadata, NULL);
1793 av_log(s, AV_LOG_WARNING,
1794 "The tags at index %d refer to a non-existent chapter "
1796 i, tags[i].target.chapteruid);
1798 } else if (tags[i].target.trackuid) {
1799 MatroskaTrack *track = matroska->tracks.elem;
1801 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1802 if (track[j].uid == tags[i].target.trackuid &&
1804 matroska_convert_tag(s, &tags[i].tag,
1805 &track[j].stream->metadata, NULL);
1810 av_log(s, AV_LOG_WARNING,
1811 "The tags at index %d refer to a non-existent track "
1813 i, tags[i].target.trackuid);
1816 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1817 tags[i].target.type);
1822 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1825 uint32_t saved_id = matroska->current_id;
1826 int64_t before_pos = avio_tell(matroska->ctx->pb);
1830 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1831 /* We don't want to lose our seekhead level, so we add
1832 * a dummy. This is a crude hack. */
1833 if (matroska->num_levels == EBML_MAX_DEPTH) {
1834 av_log(matroska->ctx, AV_LOG_INFO,
1835 "Max EBML element depth (%d) reached, "
1836 "cannot parse further.\n", EBML_MAX_DEPTH);
1837 ret = AVERROR_INVALIDDATA;
1839 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1840 matroska->num_levels++;
1841 matroska->current_id = 0;
1843 ret = ebml_parse(matroska, matroska_segment, matroska);
1844 if (ret == LEVEL_ENDED) {
1845 /* This can only happen if the seek brought us beyond EOF. */
1850 /* Seek back - notice that in all instances where this is used
1851 * it is safe to set the level to 1. */
1852 matroska_reset_status(matroska, saved_id, before_pos);
1857 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1859 EbmlList *seekhead_list = &matroska->seekhead;
1862 // we should not do any seeking in the streaming case
1863 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1866 for (i = 0; i < seekhead_list->nb_elem; i++) {
1867 MatroskaSeekhead *seekheads = seekhead_list->elem;
1868 uint32_t id = seekheads[i].id;
1869 int64_t pos = seekheads[i].pos + matroska->segment_start;
1870 MatroskaLevel1Element *elem;
1872 if (id != seekheads[i].id || pos < matroska->segment_start)
1875 elem = matroska_find_level1_elem(matroska, id, pos);
1876 if (!elem || elem->parsed)
1881 // defer cues parsing until we actually need cue data.
1882 if (id == MATROSKA_ID_CUES)
1885 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1886 // mark index as broken
1887 matroska->cues_parsing_deferred = -1;
1895 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1897 EbmlList *index_list;
1898 MatroskaIndex *index;
1899 uint64_t index_scale = 1;
1902 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1905 index_list = &matroska->index;
1906 index = index_list->elem;
1907 if (index_list->nb_elem < 2)
1909 if (index[1].time > 1E14 / matroska->time_scale) {
1910 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1913 for (i = 0; i < index_list->nb_elem; i++) {
1914 EbmlList *pos_list = &index[i].pos;
1915 MatroskaIndexPos *pos = pos_list->elem;
1916 for (j = 0; j < pos_list->nb_elem; j++) {
1917 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1919 if (track && track->stream)
1920 av_add_index_entry(track->stream,
1921 pos[j].pos + matroska->segment_start,
1922 index[i].time / index_scale, 0, 0,
1928 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1931 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1934 for (i = 0; i < matroska->num_level1_elems; i++) {
1935 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1936 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1937 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1938 matroska->cues_parsing_deferred = -1;
1944 matroska_add_index_entries(matroska);
1947 static int matroska_aac_profile(char *codec_id)
1949 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1952 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1953 if (strstr(codec_id, aac_profiles[profile]))
1958 static int matroska_aac_sri(int samplerate)
1962 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1963 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1968 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1970 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1971 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1974 static int matroska_parse_flac(AVFormatContext *s,
1975 MatroskaTrack *track,
1978 AVStream *st = track->stream;
1979 uint8_t *p = track->codec_priv.data;
1980 int size = track->codec_priv.size;
1982 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1983 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1984 track->codec_priv.size = 0;
1988 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1990 p += track->codec_priv.size;
1991 size -= track->codec_priv.size;
1993 /* parse the remaining metadata blocks if present */
1995 int block_last, block_type, block_size;
1997 flac_parse_block_header(p, &block_last, &block_type, &block_size);
2001 if (block_size > size)
2004 /* check for the channel mask */
2005 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2006 AVDictionary *dict = NULL;
2007 AVDictionaryEntry *chmask;
2009 ff_vorbis_comment(s, &dict, p, block_size, 0);
2010 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2012 uint64_t mask = strtol(chmask->value, NULL, 0);
2013 if (!mask || mask & ~0x3ffffULL) {
2014 av_log(s, AV_LOG_WARNING,
2015 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2017 st->codecpar->channel_layout = mask;
2019 av_dict_free(&dict);
2029 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2031 int minor, micro, bttb = 0;
2033 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2034 * this function, and fixed in 57.52 */
2035 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, µ) == 2)
2036 bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2038 switch (field_order) {
2039 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2040 return AV_FIELD_PROGRESSIVE;
2041 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2042 return AV_FIELD_UNKNOWN;
2043 case MATROSKA_VIDEO_FIELDORDER_TT:
2045 case MATROSKA_VIDEO_FIELDORDER_BB:
2047 case MATROSKA_VIDEO_FIELDORDER_BT:
2048 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2049 case MATROSKA_VIDEO_FIELDORDER_TB:
2050 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2052 return AV_FIELD_UNKNOWN;
2056 static void mkv_stereo_mode_display_mul(int stereo_mode,
2057 int *h_width, int *h_height)
2059 switch (stereo_mode) {
2060 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2061 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2062 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2063 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2064 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2066 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2067 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2068 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2069 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2072 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2073 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2074 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2075 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2081 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2082 const MatroskaTrackVideoColor *color = track->video.color.elem;
2083 const MatroskaMasteringMeta *mastering_meta;
2084 int has_mastering_primaries, has_mastering_luminance;
2086 if (!track->video.color.nb_elem)
2089 mastering_meta = &color->mastering_meta;
2090 // Mastering primaries are CIE 1931 coords, and must be > 0.
2091 has_mastering_primaries =
2092 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2093 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2094 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2095 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2096 has_mastering_luminance = mastering_meta->max_luminance > 0;
2098 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2099 st->codecpar->color_space = color->matrix_coefficients;
2100 if (color->primaries != AVCOL_PRI_RESERVED &&
2101 color->primaries != AVCOL_PRI_RESERVED0)
2102 st->codecpar->color_primaries = color->primaries;
2103 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2104 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2105 st->codecpar->color_trc = color->transfer_characteristics;
2106 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2107 color->range <= AVCOL_RANGE_JPEG)
2108 st->codecpar->color_range = color->range;
2109 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2110 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2111 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2112 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2113 st->codecpar->chroma_location =
2114 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2115 (color->chroma_siting_vert - 1) << 7);
2117 if (color->max_cll && color->max_fall) {
2120 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2122 return AVERROR(ENOMEM);
2123 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2124 (uint8_t *)metadata, size);
2126 av_freep(&metadata);
2129 metadata->MaxCLL = color->max_cll;
2130 metadata->MaxFALL = color->max_fall;
2133 if (has_mastering_primaries || has_mastering_luminance) {
2134 AVMasteringDisplayMetadata *metadata =
2135 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2136 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2137 sizeof(AVMasteringDisplayMetadata));
2139 return AVERROR(ENOMEM);
2141 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2142 if (has_mastering_primaries) {
2143 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2144 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2145 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2146 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2147 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2148 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2149 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2150 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2151 metadata->has_primaries = 1;
2153 if (has_mastering_luminance) {
2154 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2155 metadata->min_luminance = av_d2q(mastering_meta->min_luminance, INT_MAX);
2156 metadata->has_luminance = 1;
2162 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2165 AVSphericalMapping *spherical;
2166 const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2167 const uint8_t *priv_data = mkv_projection->private.data;
2168 enum AVSphericalProjection projection;
2169 size_t spherical_size;
2170 uint32_t l = 0, t = 0, r = 0, b = 0;
2171 uint32_t padding = 0;
2174 if (mkv_projection->private.size && priv_data[0] != 0) {
2175 av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2179 switch (track->video.projection.type) {
2180 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2181 if (track->video.projection.private.size == 20) {
2182 t = AV_RB32(priv_data + 4);
2183 b = AV_RB32(priv_data + 8);
2184 l = AV_RB32(priv_data + 12);
2185 r = AV_RB32(priv_data + 16);
2187 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2188 av_log(logctx, AV_LOG_ERROR,
2189 "Invalid bounding rectangle coordinates "
2190 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2192 return AVERROR_INVALIDDATA;
2194 } else if (track->video.projection.private.size != 0) {
2195 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2196 return AVERROR_INVALIDDATA;
2199 if (l || t || r || b)
2200 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2202 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2204 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2205 if (track->video.projection.private.size < 4) {
2206 av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2207 return AVERROR_INVALIDDATA;
2208 } else if (track->video.projection.private.size == 12) {
2209 uint32_t layout = AV_RB32(priv_data + 4);
2211 av_log(logctx, AV_LOG_WARNING,
2212 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2215 projection = AV_SPHERICAL_CUBEMAP;
2216 padding = AV_RB32(priv_data + 8);
2218 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2219 return AVERROR_INVALIDDATA;
2222 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2223 /* No Spherical metadata */
2226 av_log(logctx, AV_LOG_WARNING,
2227 "Unknown spherical metadata type %"PRIu64"\n",
2228 track->video.projection.type);
2232 spherical = av_spherical_alloc(&spherical_size);
2234 return AVERROR(ENOMEM);
2236 spherical->projection = projection;
2238 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2239 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2240 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2242 spherical->padding = padding;
2244 spherical->bound_left = l;
2245 spherical->bound_top = t;
2246 spherical->bound_right = r;
2247 spherical->bound_bottom = b;
2249 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2252 av_freep(&spherical);
2259 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2261 const AVCodecTag *codec_tags;
2263 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2264 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2266 /* Normalize noncompliant private data that starts with the fourcc
2267 * by expanding/shifting the data by 4 bytes and storing the data
2268 * size at the start. */
2269 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2270 int ret = av_buffer_realloc(&track->codec_priv.buf,
2271 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2275 track->codec_priv.data = track->codec_priv.buf->data;
2276 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2277 track->codec_priv.size += 4;
2278 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2281 *fourcc = AV_RL32(track->codec_priv.data + 4);
2282 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2287 static int matroska_parse_tracks(AVFormatContext *s)
2289 MatroskaDemuxContext *matroska = s->priv_data;
2290 MatroskaTrack *tracks = matroska->tracks.elem;
2295 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2296 MatroskaTrack *track = &tracks[i];
2297 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2298 EbmlList *encodings_list = &track->encodings;
2299 MatroskaTrackEncoding *encodings = encodings_list->elem;
2300 uint8_t *extradata = NULL;
2301 int extradata_size = 0;
2302 int extradata_offset = 0;
2303 uint32_t fourcc = 0;
2305 char* key_id_base64 = NULL;
2308 /* Apply some sanity checks. */
2309 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2310 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2311 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2312 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2313 av_log(matroska->ctx, AV_LOG_INFO,
2314 "Unknown or unsupported track type %"PRIu64"\n",
2318 if (!track->codec_id)
2321 if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
2322 || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
2323 || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2324 || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2326 av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
2330 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2331 isnan(track->audio.samplerate)) {
2332 av_log(matroska->ctx, AV_LOG_WARNING,
2333 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2334 track->audio.samplerate);
2335 track->audio.samplerate = 8000;
2338 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2339 if (!track->default_duration && track->video.frame_rate > 0) {
2340 double default_duration = 1000000000 / track->video.frame_rate;
2341 if (default_duration > UINT64_MAX || default_duration < 0) {
2342 av_log(matroska->ctx, AV_LOG_WARNING,
2343 "Invalid frame rate %e. Cannot calculate default duration.\n",
2344 track->video.frame_rate);
2346 track->default_duration = default_duration;
2349 if (track->video.display_width == -1)
2350 track->video.display_width = track->video.pixel_width;
2351 if (track->video.display_height == -1)
2352 track->video.display_height = track->video.pixel_height;
2353 if (track->video.color_space.size == 4)
2354 fourcc = AV_RL32(track->video.color_space.data);
2355 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2356 if (!track->audio.out_samplerate)
2357 track->audio.out_samplerate = track->audio.samplerate;
2359 if (encodings_list->nb_elem > 1) {
2360 av_log(matroska->ctx, AV_LOG_ERROR,
2361 "Multiple combined encodings not supported");
2362 } else if (encodings_list->nb_elem == 1) {
2363 if (encodings[0].type) {
2364 if (encodings[0].encryption.key_id.size > 0) {
2365 /* Save the encryption key id to be stored later as a
2367 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2368 key_id_base64 = av_malloc(b64_size);
2369 if (key_id_base64 == NULL)
2370 return AVERROR(ENOMEM);
2372 av_base64_encode(key_id_base64, b64_size,
2373 encodings[0].encryption.key_id.data,
2374 encodings[0].encryption.key_id.size);
2376 encodings[0].scope = 0;
2377 av_log(matroska->ctx, AV_LOG_ERROR,
2378 "Unsupported encoding type");
2382 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2385 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2388 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2390 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2391 encodings[0].scope = 0;
2392 av_log(matroska->ctx, AV_LOG_ERROR,
2393 "Unsupported encoding type");
2394 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2395 uint8_t *codec_priv = track->codec_priv.data;
2396 int ret = matroska_decode_buffer(&track->codec_priv.data,
2397 &track->codec_priv.size,
2400 track->codec_priv.data = NULL;
2401 track->codec_priv.size = 0;
2402 av_log(matroska->ctx, AV_LOG_ERROR,
2403 "Failed to decode codec private data\n");
2406 if (codec_priv != track->codec_priv.data) {
2407 av_buffer_unref(&track->codec_priv.buf);
2408 if (track->codec_priv.data) {
2409 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2410 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2412 if (!track->codec_priv.buf) {
2413 av_freep(&track->codec_priv.data);
2414 track->codec_priv.size = 0;
2415 return AVERROR(ENOMEM);
2421 track->needs_decoding = encodings && !encodings[0].type &&
2422 encodings[0].scope & 1 &&
2423 (encodings[0].compression.algo !=
2424 MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP ||
2425 encodings[0].compression.settings.size);
2427 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2428 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2429 strlen(ff_mkv_codec_tags[j].str))) {
2430 codec_id = ff_mkv_codec_tags[j].id;
2435 st = track->stream = avformat_new_stream(s, NULL);
2437 av_free(key_id_base64);
2438 return AVERROR(ENOMEM);
2441 if (key_id_base64) {
2442 /* export encryption key id as base64 metadata tag */
2443 av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2444 AV_DICT_DONT_STRDUP_VAL);
2447 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2448 track->codec_priv.size >= 40 &&
2449 track->codec_priv.data) {
2450 track->ms_compat = 1;
2451 bit_depth = AV_RL16(track->codec_priv.data + 14);
2452 fourcc = AV_RL32(track->codec_priv.data + 16);
2453 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2456 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2458 extradata_offset = 40;
2459 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2460 track->codec_priv.size >= 14 &&
2461 track->codec_priv.data) {
2463 ffio_init_context(&b, track->codec_priv.data,
2464 track->codec_priv.size,
2465 0, NULL, NULL, NULL, NULL);
2466 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2469 codec_id = st->codecpar->codec_id;
2470 fourcc = st->codecpar->codec_tag;
2471 extradata_offset = FFMIN(track->codec_priv.size, 18);
2472 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2473 /* Normally 36, but allow noncompliant private data */
2474 && (track->codec_priv.size >= 32)
2475 && (track->codec_priv.data)) {
2476 uint16_t sample_size;
2477 int ret = get_qt_codec(track, &fourcc, &codec_id);
2480 sample_size = AV_RB16(track->codec_priv.data + 26);
2482 if (sample_size == 8) {
2483 fourcc = MKTAG('r','a','w',' ');
2484 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2485 } else if (sample_size == 16) {
2486 fourcc = MKTAG('t','w','o','s');
2487 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2490 if ((fourcc == MKTAG('t','w','o','s') ||
2491 fourcc == MKTAG('s','o','w','t')) &&
2493 codec_id = AV_CODEC_ID_PCM_S8;
2494 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2495 (track->codec_priv.size >= 21) &&
2496 (track->codec_priv.data)) {
2497 int ret = get_qt_codec(track, &fourcc, &codec_id);
2500 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2501 fourcc = MKTAG('S','V','Q','3');
2502 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2504 if (codec_id == AV_CODEC_ID_NONE)
2505 av_log(matroska->ctx, AV_LOG_ERROR,
2506 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2507 if (track->codec_priv.size >= 86) {
2508 bit_depth = AV_RB16(track->codec_priv.data + 82);
2509 ffio_init_context(&b, track->codec_priv.data,
2510 track->codec_priv.size,
2511 0, NULL, NULL, NULL, NULL);
2512 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2514 track->has_palette = 1;
2517 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2518 switch (track->audio.bitdepth) {
2520 codec_id = AV_CODEC_ID_PCM_U8;
2523 codec_id = AV_CODEC_ID_PCM_S24BE;
2526 codec_id = AV_CODEC_ID_PCM_S32BE;
2529 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2530 switch (track->audio.bitdepth) {
2532 codec_id = AV_CODEC_ID_PCM_U8;
2535 codec_id = AV_CODEC_ID_PCM_S24LE;
2538 codec_id = AV_CODEC_ID_PCM_S32LE;
2541 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2542 track->audio.bitdepth == 64) {
2543 codec_id = AV_CODEC_ID_PCM_F64LE;
2544 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2545 int profile = matroska_aac_profile(track->codec_id);
2546 int sri = matroska_aac_sri(track->audio.samplerate);
2547 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2549 return AVERROR(ENOMEM);
2550 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2551 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2552 if (strstr(track->codec_id, "SBR")) {
2553 sri = matroska_aac_sri(track->audio.out_samplerate);
2554 extradata[2] = 0x56;
2555 extradata[3] = 0xE5;
2556 extradata[4] = 0x80 | (sri << 3);
2560 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2561 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2562 * Create the "atom size", "tag", and "tag version" fields the
2563 * decoder expects manually. */
2564 extradata_size = 12 + track->codec_priv.size;
2565 extradata = av_mallocz(extradata_size +
2566 AV_INPUT_BUFFER_PADDING_SIZE);
2568 return AVERROR(ENOMEM);
2569 AV_WB32(extradata, extradata_size);
2570 memcpy(&extradata[4], "alac", 4);
2571 AV_WB32(&extradata[8], 0);
2572 memcpy(&extradata[12], track->codec_priv.data,
2573 track->codec_priv.size);
2574 } else if (codec_id == AV_CODEC_ID_TTA) {
2576 if (track->audio.channels > UINT16_MAX ||
2577 track->audio.bitdepth > UINT16_MAX) {
2578 av_log(matroska->ctx, AV_LOG_WARNING,
2579 "Too large audio channel number %"PRIu64
2580 " or bitdepth %"PRIu64". Skipping track.\n",
2581 track->audio.channels, track->audio.bitdepth);
2582 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2583 return AVERROR_INVALIDDATA;
2587 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2588 return AVERROR_INVALIDDATA;
2589 extradata_size = 22;
2590 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2592 return AVERROR(ENOMEM);
2594 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2595 bytestream_put_le16(&ptr, 1);
2596 bytestream_put_le16(&ptr, track->audio.channels);
2597 bytestream_put_le16(&ptr, track->audio.bitdepth);
2598 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2599 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2600 track->audio.out_samplerate,
2601 AV_TIME_BASE * 1000));
2602 } else if (codec_id == AV_CODEC_ID_RV10 ||
2603 codec_id == AV_CODEC_ID_RV20 ||
2604 codec_id == AV_CODEC_ID_RV30 ||
2605 codec_id == AV_CODEC_ID_RV40) {
2606 extradata_offset = 26;
2607 } else if (codec_id == AV_CODEC_ID_RA_144) {
2608 track->audio.out_samplerate = 8000;
2609 track->audio.channels = 1;
2610 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2611 codec_id == AV_CODEC_ID_COOK ||
2612 codec_id == AV_CODEC_ID_ATRAC3 ||
2613 codec_id == AV_CODEC_ID_SIPR)
2614 && track->codec_priv.data) {
2617 ffio_init_context(&b, track->codec_priv.data,
2618 track->codec_priv.size,
2619 0, NULL, NULL, NULL, NULL);
2621 flavor = avio_rb16(&b);
2622 track->audio.coded_framesize = avio_rb32(&b);
2624 track->audio.sub_packet_h = avio_rb16(&b);
2625 track->audio.frame_size = avio_rb16(&b);
2626 track->audio.sub_packet_size = avio_rb16(&b);
2627 if (track->audio.coded_framesize <= 0 ||
2628 track->audio.sub_packet_h <= 0 ||
2629 track->audio.frame_size <= 0)
2630 return AVERROR_INVALIDDATA;
2632 if (codec_id == AV_CODEC_ID_RA_288) {
2633 if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2634 != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2635 return AVERROR_INVALIDDATA;
2636 st->codecpar->block_align = track->audio.coded_framesize;
2637 track->codec_priv.size = 0;
2639 if (codec_id == AV_CODEC_ID_SIPR) {
2640 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2642 return AVERROR_INVALIDDATA;
2643 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2644 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2645 } else if (track->audio.sub_packet_size <= 0 ||
2646 track->audio.frame_size % track->audio.sub_packet_size)
2647 return AVERROR_INVALIDDATA;
2648 st->codecpar->block_align = track->audio.sub_packet_size;
2649 extradata_offset = 78;
2651 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2652 track->audio.frame_size);
2653 if (!track->audio.buf)
2654 return AVERROR(ENOMEM);
2655 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2656 ret = matroska_parse_flac(s, track, &extradata_offset);
2659 } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2660 av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2661 "in absence of valid CodecPrivate.\n");
2663 extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2665 return AVERROR(ENOMEM);
2666 AV_WL16(extradata, 0x410);
2667 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2668 fourcc = AV_RL32(track->codec_priv.data);
2669 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2670 /* we don't need any value stored in CodecPrivate.
2671 make sure that it's not exported as extradata. */
2672 track->codec_priv.size = 0;
2673 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2674 /* For now, propagate only the OBUs, if any. Once libavcodec is
2675 updated to handle isobmff style extradata this can be removed. */
2676 extradata_offset = 4;
2678 track->codec_priv.size -= extradata_offset;
2680 if (codec_id == AV_CODEC_ID_NONE)
2681 av_log(matroska->ctx, AV_LOG_INFO,
2682 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2684 if (track->time_scale < 0.01) {
2685 av_log(matroska->ctx, AV_LOG_WARNING,
2686 "Track TimestampScale too small %f, assuming 1.0.\n",
2688 track->time_scale = 1.0;
2690 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2691 1000 * 1000 * 1000); /* 64 bit pts in ns */
2693 /* convert the delay from ns to the track timebase */
2694 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2695 (AVRational){ 1, 1000000000 },
2698 st->codecpar->codec_id = codec_id;
2700 if (strcmp(track->language, "und"))
2701 av_dict_set(&st->metadata, "language", track->language, 0);
2702 av_dict_set(&st->metadata, "title", track->name, 0);
2704 if (track->flag_default)
2705 st->disposition |= AV_DISPOSITION_DEFAULT;
2706 if (track->flag_forced)
2707 st->disposition |= AV_DISPOSITION_FORCED;
2709 if (!st->codecpar->extradata) {
2711 st->codecpar->extradata = extradata;
2712 st->codecpar->extradata_size = extradata_size;
2713 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2714 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2715 return AVERROR(ENOMEM);
2716 memcpy(st->codecpar->extradata,
2717 track->codec_priv.data + extradata_offset,
2718 track->codec_priv.size);
2722 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2723 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2724 int display_width_mul = 1;
2725 int display_height_mul = 1;
2727 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2728 st->codecpar->codec_tag = fourcc;
2730 st->codecpar->bits_per_coded_sample = bit_depth;
2731 st->codecpar->width = track->video.pixel_width;
2732 st->codecpar->height = track->video.pixel_height;
2734 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2735 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2736 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2737 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2739 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2740 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2742 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2743 av_reduce(&st->sample_aspect_ratio.num,
2744 &st->sample_aspect_ratio.den,
2745 st->codecpar->height * track->video.display_width * display_width_mul,
2746 st->codecpar->width * track->video.display_height * display_height_mul,
2749 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2750 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2752 if (track->default_duration) {
2753 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2754 1000000000, track->default_duration, 30000);
2755 #if FF_API_R_FRAME_RATE
2756 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2757 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2758 st->r_frame_rate = st->avg_frame_rate;
2762 /* export stereo mode flag as metadata tag */
2763 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2764 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2766 /* export alpha mode flag as metadata tag */
2767 if (track->video.alpha_mode)
2768 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2770 /* if we have virtual track, mark the real tracks */
2771 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2773 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2775 snprintf(buf, sizeof(buf), "%s_%d",
2776 ff_matroska_video_stereo_plane[planes[j].type], i);
2777 for (k=0; k < matroska->tracks.nb_elem; k++)
2778 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2779 av_dict_set(&tracks[k].stream->metadata,
2780 "stereo_mode", buf, 0);
2784 // add stream level stereo3d side data if it is a supported format
2785 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2786 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2787 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2792 ret = mkv_parse_video_color(st, track);
2795 ret = mkv_parse_video_projection(st, track, matroska->ctx);
2798 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2799 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2800 st->codecpar->codec_tag = fourcc;
2801 st->codecpar->sample_rate = track->audio.out_samplerate;
2802 st->codecpar->channels = track->audio.channels;
2803 if (!st->codecpar->bits_per_coded_sample)
2804 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2805 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2806 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2807 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2808 st->need_parsing = AVSTREAM_PARSE_FULL;
2809 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2810 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2811 if (track->codec_delay > 0) {
2812 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2813 (AVRational){1, 1000000000},
2814 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2815 48000 : st->codecpar->sample_rate});
2817 if (track->seek_preroll > 0) {
2818 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2819 (AVRational){1, 1000000000},
2820 (AVRational){1, st->codecpar->sample_rate});
2822 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2823 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2825 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2826 st->disposition |= AV_DISPOSITION_CAPTIONS;
2827 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2828 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2829 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2830 st->disposition |= AV_DISPOSITION_METADATA;
2832 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2833 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2840 static int matroska_read_header(AVFormatContext *s)
2842 MatroskaDemuxContext *matroska = s->priv_data;
2843 EbmlList *attachments_list = &matroska->attachments;
2844 EbmlList *chapters_list = &matroska->chapters;
2845 MatroskaAttachment *attachments;
2846 MatroskaChapter *chapters;
2847 uint64_t max_start = 0;
2853 matroska->cues_parsing_deferred = 1;
2855 /* First read the EBML header. */
2856 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2857 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2858 ebml_free(ebml_syntax, &ebml);
2859 return AVERROR_INVALIDDATA;
2861 if (ebml.version > EBML_VERSION ||
2862 ebml.max_size > sizeof(uint64_t) ||
2863 ebml.id_length > sizeof(uint32_t) ||
2864 ebml.doctype_version > 3) {
2865 avpriv_report_missing_feature(matroska->ctx,
2866 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2867 ebml.version, ebml.doctype, ebml.doctype_version);
2868 ebml_free(ebml_syntax, &ebml);
2869 return AVERROR_PATCHWELCOME;
2870 } else if (ebml.doctype_version == 3) {
2871 av_log(matroska->ctx, AV_LOG_WARNING,
2872 "EBML header using unsupported features\n"
2873 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2874 ebml.version, ebml.doctype, ebml.doctype_version);
2876 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2877 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2879 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2880 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2881 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2882 ebml_free(ebml_syntax, &ebml);
2883 return AVERROR_INVALIDDATA;
2886 ebml_free(ebml_syntax, &ebml);
2888 /* The next thing is a segment. */
2889 pos = avio_tell(matroska->ctx->pb);
2890 res = ebml_parse(matroska, matroska_segments, matroska);
2891 // Try resyncing until we find an EBML_STOP type element.
2893 res = matroska_resync(matroska, pos);
2896 pos = avio_tell(matroska->ctx->pb);
2897 res = ebml_parse(matroska, matroska_segment, matroska);
2899 /* Set data_offset as it might be needed later by seek_frame_generic. */
2900 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2901 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2902 matroska_execute_seekhead(matroska);
2904 if (!matroska->time_scale)
2905 matroska->time_scale = 1000000;
2906 if (matroska->duration)
2907 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2908 1000 / AV_TIME_BASE;
2909 av_dict_set(&s->metadata, "title", matroska->title, 0);
2910 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2912 if (matroska->date_utc.size == 8)
2913 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2915 res = matroska_parse_tracks(s);
2919 attachments = attachments_list->elem;
2920 for (j = 0; j < attachments_list->nb_elem; j++) {
2921 if (!(attachments[j].filename && attachments[j].mime &&
2922 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2923 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2925 AVStream *st = avformat_new_stream(s, NULL);
2928 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2929 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2930 if (attachments[j].description)
2931 av_dict_set(&st->metadata, "title", attachments[j].description, 0);
2932 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2934 for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2935 if (!strncmp(mkv_image_mime_tags[i].str, attachments[j].mime,
2936 strlen(mkv_image_mime_tags[i].str))) {
2937 st->codecpar->codec_id = mkv_image_mime_tags[i].id;
2942 attachments[j].stream = st;
2944 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2945 AVPacket *pkt = &st->attached_pic;
2947 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2948 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2950 av_init_packet(pkt);
2951 pkt->buf = attachments[j].bin.buf;
2952 attachments[j].bin.buf = NULL;
2953 pkt->data = attachments[j].bin.data;
2954 pkt->size = attachments[j].bin.size;
2955 pkt->stream_index = st->index;
2956 pkt->flags |= AV_PKT_FLAG_KEY;
2958 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2959 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2961 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2962 attachments[j].bin.size);
2964 for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2965 if (!strncmp(mkv_mime_tags[i].str, attachments[j].mime,
2966 strlen(mkv_mime_tags[i].str))) {
2967 st->codecpar->codec_id = mkv_mime_tags[i].id;
2975 chapters = chapters_list->elem;
2976 for (i = 0; i < chapters_list->nb_elem; i++)
2977 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2978 (max_start == 0 || chapters[i].start > max_start)) {
2979 chapters[i].chapter =
2980 avpriv_new_chapter(s, chapters[i].uid,
2981 (AVRational) { 1, 1000000000 },
2982 chapters[i].start, chapters[i].end,
2984 max_start = chapters[i].start;
2987 matroska_add_index_entries(matroska);
2989 matroska_convert_tags(s);
2993 matroska_read_close(s);
2998 * Put one packet in an application-supplied AVPacket struct.
2999 * Returns 0 on success or -1 on failure.
3001 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3004 if (matroska->queue) {
3005 MatroskaTrack *tracks = matroska->tracks.elem;
3006 MatroskaTrack *track;
3008 avpriv_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
3009 track = &tracks[pkt->stream_index];
3010 if (track->has_palette) {
3011 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
3013 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3015 memcpy(pal, track->palette, AVPALETTE_SIZE);
3017 track->has_palette = 0;
3026 * Free all packets in our internal queue.
3028 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3030 avpriv_packet_list_free(&matroska->queue, &matroska->queue_end);
3033 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3034 int size, int type, AVIOContext *pb,
3035 uint32_t lace_size[256], int *laces)
3038 uint8_t *data = *buf;
3042 lace_size[0] = size;
3047 return AVERROR_INVALIDDATA;
3054 case 0x1: /* Xiph lacing */
3058 for (n = 0; n < *laces - 1; n++) {
3063 return AVERROR_INVALIDDATA;
3066 lace_size[n] += temp;
3069 } while (temp == 0xff);
3072 return AVERROR_INVALIDDATA;
3074 lace_size[n] = size - total;
3078 case 0x2: /* fixed-size lacing */
3079 if (size % (*laces))
3080 return AVERROR_INVALIDDATA;
3081 for (n = 0; n < *laces; n++)
3082 lace_size[n] = size / *laces;
3085 case 0x3: /* EBML lacing */
3093 n = ebml_read_num(matroska, pb, 8, &num, 1);
3097 return AVERROR_INVALIDDATA;
3099 total = lace_size[0] = num;
3101 for (n = 1; n < *laces - 1; n++) {
3104 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3107 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3108 return AVERROR_INVALIDDATA;
3110 lace_size[n] = lace_size[n - 1] + snum;
3111 total += lace_size[n];
3117 return AVERROR_INVALIDDATA;
3119 lace_size[*laces - 1] = size - total;
3129 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3130 MatroskaTrack *track, AVStream *st,
3131 uint8_t *data, int size, uint64_t timecode,
3134 const int a = st->codecpar->block_align;
3135 const int sps = track->audio.sub_packet_size;
3136 const int cfs = track->audio.coded_framesize;
3137 const int h = track->audio.sub_packet_h;
3138 const int w = track->audio.frame_size;
3139 int y = track->audio.sub_packet_cnt;
3142 if (!track->audio.pkt_cnt) {
3143 if (track->audio.sub_packet_cnt == 0)
3144 track->audio.buf_timecode = timecode;
3145 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3146 if (size < cfs * h / 2) {
3147 av_log(matroska->ctx, AV_LOG_ERROR,
3148 "Corrupt int4 RM-style audio packet size\n");
3149 return AVERROR_INVALIDDATA;
3151 for (x = 0; x < h / 2; x++)
3152 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3153 data + x * cfs, cfs);
3154 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3156 av_log(matroska->ctx, AV_LOG_ERROR,
3157 "Corrupt sipr RM-style audio packet size\n");
3158 return AVERROR_INVALIDDATA;
3160 memcpy(track->audio.buf + y * w, data, w);
3163 av_log(matroska->ctx, AV_LOG_ERROR,
3164 "Corrupt generic RM-style audio packet size\n");
3165 return AVERROR_INVALIDDATA;
3167 for (x = 0; x < w / sps; x++)
3168 memcpy(track->audio.buf +
3169 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3170 data + x * sps, sps);
3173 if (++track->audio.sub_packet_cnt >= h) {
3174 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3175 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3176 track->audio.sub_packet_cnt = 0;
3177 track->audio.pkt_cnt = h * w / a;
3181 while (track->audio.pkt_cnt) {
3183 AVPacket pktl, *pkt = &pktl;
3185 ret = av_new_packet(pkt, a);
3190 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3192 pkt->pts = track->audio.buf_timecode;
3193 track->audio.buf_timecode = AV_NOPTS_VALUE;
3195 pkt->stream_index = st->index;
3196 ret = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3198 av_packet_unref(pkt);
3199 return AVERROR(ENOMEM);
3206 /* reconstruct full wavpack blocks from mangled matroska ones */
3207 static int matroska_parse_wavpack(MatroskaTrack *track,
3208 uint8_t **data, int *size)
3210 uint8_t *dst = NULL;
3211 uint8_t *src = *data;
3216 int ret, offset = 0;
3219 return AVERROR_INVALIDDATA;
3221 av_assert1(track->stream->codecpar->extradata_size >= 2);
3222 ver = AV_RL16(track->stream->codecpar->extradata);
3224 samples = AV_RL32(src);
3228 while (srclen >= 8) {
3233 uint32_t flags = AV_RL32(src);
3234 uint32_t crc = AV_RL32(src + 4);
3238 multiblock = (flags & 0x1800) != 0x1800;
3241 ret = AVERROR_INVALIDDATA;
3244 blocksize = AV_RL32(src);
3250 if (blocksize > srclen) {
3251 ret = AVERROR_INVALIDDATA;
3255 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3257 ret = AVERROR(ENOMEM);
3261 dstlen += blocksize + 32;
3263 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3264 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3265 AV_WL16(dst + offset + 8, ver); // version
3266 AV_WL16(dst + offset + 10, 0); // track/index_no
3267 AV_WL32(dst + offset + 12, 0); // total samples
3268 AV_WL32(dst + offset + 16, 0); // block index
3269 AV_WL32(dst + offset + 20, samples); // number of samples
3270 AV_WL32(dst + offset + 24, flags); // flags
3271 AV_WL32(dst + offset + 28, crc); // crc
3272 memcpy(dst + offset + 32, src, blocksize); // block data
3275 srclen -= blocksize;
3276 offset += blocksize + 32;
3279 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3291 static int matroska_parse_prores(MatroskaTrack *track,
3292 uint8_t **data, int *size)
3295 int dstlen = *size + 8;
3297 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3299 return AVERROR(ENOMEM);
3301 AV_WB32(dst, dstlen);
3302 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3303 memcpy(dst + 8, *data, dstlen - 8);
3304 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3312 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3313 MatroskaTrack *track,
3315 uint8_t *data, int data_len,
3320 AVPacket pktl, *pkt = &pktl;
3321 uint8_t *id, *settings, *text, *buf;
3322 int id_len, settings_len, text_len;
3327 return AVERROR_INVALIDDATA;
3330 q = data + data_len;
3335 if (*p == '\r' || *p == '\n') {
3344 if (p >= q || *p != '\n')
3345 return AVERROR_INVALIDDATA;
3351 if (*p == '\r' || *p == '\n') {
3352 settings_len = p - settings;
3360 if (p >= q || *p != '\n')
3361 return AVERROR_INVALIDDATA;
3366 while (text_len > 0) {
3367 const int len = text_len - 1;
3368 const uint8_t c = p[len];
3369 if (c != '\r' && c != '\n')
3375 return AVERROR_INVALIDDATA;
3377 err = av_new_packet(pkt, text_len);
3382 memcpy(pkt->data, text, text_len);
3385 buf = av_packet_new_side_data(pkt,
3386 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3389 av_packet_unref(pkt);
3390 return AVERROR(ENOMEM);
3392 memcpy(buf, id, id_len);
3395 if (settings_len > 0) {
3396 buf = av_packet_new_side_data(pkt,
3397 AV_PKT_DATA_WEBVTT_SETTINGS,
3400 av_packet_unref(pkt);
3401 return AVERROR(ENOMEM);
3403 memcpy(buf, settings, settings_len);
3406 // Do we need this for subtitles?
3407 // pkt->flags = AV_PKT_FLAG_KEY;
3409 pkt->stream_index = st->index;
3410 pkt->pts = timecode;
3412 // Do we need this for subtitles?
3413 // pkt->dts = timecode;
3415 pkt->duration = duration;
3418 err = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3420 av_packet_unref(pkt);
3421 return AVERROR(ENOMEM);
3427 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3428 MatroskaTrack *track, AVStream *st,
3429 AVBufferRef *buf, uint8_t *data, int pkt_size,
3430 uint64_t timecode, uint64_t lace_duration,
3431 int64_t pos, int is_keyframe,
3432 uint8_t *additional, uint64_t additional_id, int additional_size,
3433 int64_t discard_padding)
3435 uint8_t *pkt_data = data;
3437 AVPacket pktl, *pkt = &pktl;
3439 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3440 res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3442 av_log(matroska->ctx, AV_LOG_ERROR,
3443 "Error parsing a wavpack block.\n");
3451 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3452 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3453 res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3455 av_log(matroska->ctx, AV_LOG_ERROR,
3456 "Error parsing a prores block.\n");
3464 if (!pkt_size && !additional_size)
3467 av_init_packet(pkt);
3469 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3472 pkt->buf = av_buffer_ref(buf);
3475 res = AVERROR(ENOMEM);
3479 pkt->data = pkt_data;
3480 pkt->size = pkt_size;
3481 pkt->flags = is_keyframe;
3482 pkt->stream_index = st->index;
3484 if (additional_size > 0) {
3485 uint8_t *side_data = av_packet_new_side_data(pkt,
3486 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3487 additional_size + 8);
3489 av_packet_unref(pkt);
3490 return AVERROR(ENOMEM);
3492 AV_WB64(side_data, additional_id);
3493 memcpy(side_data + 8, additional, additional_size);
3496 if (discard_padding) {
3497 uint8_t *side_data = av_packet_new_side_data(pkt,
3498 AV_PKT_DATA_SKIP_SAMPLES,
3501 av_packet_unref(pkt);
3502 return AVERROR(ENOMEM);
3504 discard_padding = av_rescale_q(discard_padding,
3505 (AVRational){1, 1000000000},
3506 (AVRational){1, st->codecpar->sample_rate});
3507 if (discard_padding > 0) {
3508 AV_WL32(side_data + 4, discard_padding);
3510 AV_WL32(side_data, -discard_padding);
3514 if (track->ms_compat)
3515 pkt->dts = timecode;
3517 pkt->pts = timecode;
3519 pkt->duration = lace_duration;
3521 #if FF_API_CONVERGENCE_DURATION
3522 FF_DISABLE_DEPRECATION_WARNINGS
3523 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3524 pkt->convergence_duration = lace_duration;
3526 FF_ENABLE_DEPRECATION_WARNINGS
3529 res = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3531 av_packet_unref(pkt);
3532 return AVERROR(ENOMEM);
3544 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3545 int size, int64_t pos, uint64_t cluster_time,
3546 uint64_t block_duration, int is_keyframe,
3547 uint8_t *additional, uint64_t additional_id, int additional_size,
3548 int64_t cluster_pos, int64_t discard_padding)
3550 uint64_t timecode = AV_NOPTS_VALUE;
3551 MatroskaTrack *track;
3556 uint32_t lace_size[256];
3557 int n, flags, laces = 0;
3559 int trust_default_duration;
3561 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3563 if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3568 track = matroska_find_track_by_num(matroska, num);
3569 if (!track || size < 3)
3570 return AVERROR_INVALIDDATA;
3572 if (!(st = track->stream)) {
3573 av_log(matroska->ctx, AV_LOG_VERBOSE,
3574 "No stream associated to TrackNumber %"PRIu64". "
3575 "Ignoring Block with this TrackNumber.\n", num);
3579 if (st->discard >= AVDISCARD_ALL)
3581 if (block_duration > INT64_MAX)
3582 block_duration = INT64_MAX;
3584 block_time = sign_extend(AV_RB16(data), 16);
3588 if (is_keyframe == -1)
3589 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3591 if (cluster_time != (uint64_t) -1 &&
3592 (block_time >= 0 || cluster_time >= -block_time)) {
3593 uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
3594 timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
3595 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3596 timecode < track->end_timecode)
3597 is_keyframe = 0; /* overlapping subtitles are not key frame */
3599 ff_reduce_index(matroska->ctx, st->index);
3600 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3605 if (matroska->skip_to_keyframe &&
3606 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3607 // Compare signed timecodes. Timecode may be negative due to codec delay
3608 // offset. We don't support timestamps greater than int64_t anyway - see
3610 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3613 matroska->skip_to_keyframe = 0;
3614 else if (!st->internal->skip_to_keyframe) {
3615 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3616 matroska->skip_to_keyframe = 0;
3620 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3621 &pb, lace_size, &laces);
3623 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3627 trust_default_duration = track->default_duration != 0;
3628 if (track->audio.samplerate == 8000 && trust_default_duration) {
3629 // If this is needed for more codecs, then add them here
3630 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3631 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3632 trust_default_duration = 0;
3636 if (!block_duration && trust_default_duration)
3637 block_duration = track->default_duration * laces / matroska->time_scale;
3639 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3640 track->end_timecode =
3641 FFMAX(track->end_timecode, timecode + block_duration);
3643 for (n = 0; n < laces; n++) {
3644 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3645 uint8_t *out_data = data;
3646 int out_size = lace_size[n];
3648 if (track->needs_decoding) {
3649 res = matroska_decode_buffer(&out_data, &out_size, track);
3652 /* Given that we are here means that out_data is no longer
3653 * owned by buf, so set it to NULL. This depends upon
3654 * zero-length header removal compression being ignored. */
3655 av_assert1(out_data != data);
3659 if (track->audio.buf) {
3660 res = matroska_parse_rm_audio(matroska, track, st,
3667 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3668 res = matroska_parse_webvtt(matroska, track, st,
3670 timecode, lace_duration,
3677 res = matroska_parse_frame(matroska, track, st, buf, out_data,
3678 out_size, timecode, lace_duration,
3679 pos, !n ? is_keyframe : 0,
3680 additional, additional_id, additional_size,
3686 if (timecode != AV_NOPTS_VALUE)
3687 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3688 data += lace_size[n];
3694 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3696 MatroskaCluster *cluster = &matroska->current_cluster;
3697 MatroskaBlock *block = &cluster->block;
3700 av_assert0(matroska->num_levels <= 2);
3702 if (matroska->num_levels == 1) {
3703 res = ebml_parse(matroska, matroska_segment, NULL);
3706 /* Found a cluster: subtract the size of the ID already read. */
3707 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3709 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3715 if (matroska->num_levels == 2) {
3716 /* We are inside a cluster. */
3717 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3719 if (res >= 0 && block->bin.size > 0) {
3720 int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3721 uint8_t* additional = block->additional.size > 0 ?
3722 block->additional.data : NULL;
3724 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3725 block->bin.size, block->bin.pos,
3726 cluster->timecode, block->duration,
3727 is_keyframe, additional, block->additional_id,
3728 block->additional.size, cluster->pos,
3729 block->discard_padding);
3732 ebml_free(matroska_blockgroup, block);
3733 memset(block, 0, sizeof(*block));
3734 } else if (!matroska->num_levels) {
3735 if (!avio_feof(matroska->ctx->pb)) {
3736 avio_r8(matroska->ctx->pb);
3737 if (!avio_feof(matroska->ctx->pb)) {
3738 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3739 "end of segment.\n");
3740 return AVERROR_INVALIDDATA;
3750 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3752 MatroskaDemuxContext *matroska = s->priv_data;
3755 if (matroska->resync_pos == -1) {
3756 // This can only happen if generic seeking has been used.
3757 matroska->resync_pos = avio_tell(s->pb);
3760 while (matroska_deliver_packet(matroska, pkt)) {
3762 return (ret < 0) ? ret : AVERROR_EOF;
3763 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3764 ret = matroska_resync(matroska, matroska->resync_pos);
3770 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3771 int64_t timestamp, int flags)
3773 MatroskaDemuxContext *matroska = s->priv_data;
3774 MatroskaTrack *tracks = NULL;
3775 AVStream *st = s->streams[stream_index];
3778 /* Parse the CUES now since we need the index data to seek. */
3779 if (matroska->cues_parsing_deferred > 0) {
3780 matroska->cues_parsing_deferred = 0;
3781 matroska_parse_cues(matroska);
3784 if (!st->internal->nb_index_entries)
3786 timestamp = FFMAX(timestamp, st->internal->index_entries[0].timestamp);
3788 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3789 matroska_reset_status(matroska, 0, st->internal->index_entries[st->internal->nb_index_entries - 1].pos);
3790 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3791 matroska_clear_queue(matroska);
3792 if (matroska_parse_cluster(matroska) < 0)
3797 matroska_clear_queue(matroska);
3798 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->internal->nb_index_entries - 1))
3801 tracks = matroska->tracks.elem;
3802 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3803 tracks[i].audio.pkt_cnt = 0;
3804 tracks[i].audio.sub_packet_cnt = 0;
3805 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3806 tracks[i].end_timecode = 0;
3809 /* We seek to a level 1 element, so set the appropriate status. */
3810 matroska_reset_status(matroska, 0, st->internal->index_entries[index].pos);
3811 if (flags & AVSEEK_FLAG_ANY) {
3812 st->internal->skip_to_keyframe = 0;
3813 matroska->skip_to_timecode = timestamp;
3815 st->internal->skip_to_keyframe = 1;
3816 matroska->skip_to_timecode = st->internal->index_entries[index].timestamp;
3818 matroska->skip_to_keyframe = 1;
3820 ff_update_cur_dts(s, st, st->internal->index_entries[index].timestamp);
3823 // slightly hackish but allows proper fallback to
3824 // the generic seeking code.
3825 matroska_reset_status(matroska, 0, -1);
3826 matroska->resync_pos = -1;
3827 matroska_clear_queue(matroska);
3828 st->internal->skip_to_keyframe =
3829 matroska->skip_to_keyframe = 0;
3834 static int matroska_read_close(AVFormatContext *s)
3836 MatroskaDemuxContext *matroska = s->priv_data;
3837 MatroskaTrack *tracks = matroska->tracks.elem;
3840 matroska_clear_queue(matroska);
3842 for (n = 0; n < matroska->tracks.nb_elem; n++)
3843 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3844 av_freep(&tracks[n].audio.buf);
3845 ebml_free(matroska_segment, matroska);
3851 int64_t start_time_ns;
3852 int64_t end_time_ns;
3853 int64_t start_offset;
3857 /* This function searches all the Cues and returns the CueDesc corresponding to
3858 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3859 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3861 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3862 MatroskaDemuxContext *matroska = s->priv_data;
3865 int nb_index_entries = s->streams[0]->internal->nb_index_entries;
3866 AVIndexEntry *index_entries = s->streams[0]->internal->index_entries;
3867 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3868 for (i = 1; i < nb_index_entries; i++) {
3869 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3870 index_entries[i].timestamp * matroska->time_scale > ts) {
3875 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3876 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3877 if (i != nb_index_entries - 1) {
3878 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3879 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3881 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3882 // FIXME: this needs special handling for files where Cues appear
3883 // before Clusters. the current logic assumes Cues appear after
3885 cue_desc.end_offset = cues_start - matroska->segment_start;
3890 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3892 MatroskaDemuxContext *matroska = s->priv_data;
3893 uint32_t id = matroska->current_id;
3894 int64_t cluster_pos, before_pos;
3896 if (s->streams[0]->internal->nb_index_entries <= 0) return 0;
3897 // seek to the first cluster using cues.
3898 index = av_index_search_timestamp(s->streams[0], 0, 0);
3899 if (index < 0) return 0;
3900 cluster_pos = s->streams[0]->internal->index_entries[index].pos;
3901 before_pos = avio_tell(s->pb);
3903 uint64_t cluster_id, cluster_length;
3906 avio_seek(s->pb, cluster_pos, SEEK_SET);
3907 // read cluster id and length
3908 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3909 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3911 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3915 matroska_reset_status(matroska, 0, cluster_pos);
3916 matroska_clear_queue(matroska);
3917 if (matroska_parse_cluster(matroska) < 0 ||
3921 pkt = &matroska->queue->pkt;
3922 // 4 + read is the length of the cluster id and the cluster length field.
3923 cluster_pos += 4 + read + cluster_length;
3924 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3930 /* Restore the status after matroska_read_header: */
3931 matroska_reset_status(matroska, id, before_pos);
3936 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3937 double min_buffer, double* buffer,
3938 double* sec_to_download, AVFormatContext *s,
3941 double nano_seconds_per_second = 1000000000.0;
3942 double time_sec = time_ns / nano_seconds_per_second;
3944 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3945 int64_t end_time_ns = time_ns + time_to_search_ns;
3946 double sec_downloaded = 0.0;
3947 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3948 if (desc_curr.start_time_ns == -1)
3950 *sec_to_download = 0.0;
3952 // Check for non cue start time.
3953 if (time_ns > desc_curr.start_time_ns) {
3954 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3955 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3956 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3957 double timeToDownload = (cueBytes * 8.0) / bps;
3959 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3960 *sec_to_download += timeToDownload;
3962 // Check if the search ends within the first cue.
3963 if (desc_curr.end_time_ns >= end_time_ns) {
3964 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3965 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3966 sec_downloaded = percent_to_sub * sec_downloaded;
3967 *sec_to_download = percent_to_sub * *sec_to_download;
3970 if ((sec_downloaded + *buffer) <= min_buffer) {
3974 // Get the next Cue.
3975 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3978 while (desc_curr.start_time_ns != -1) {
3979 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3980 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3981 double desc_sec = desc_ns / nano_seconds_per_second;
3982 double bits = (desc_bytes * 8.0);
3983 double time_to_download = bits / bps;
3985 sec_downloaded += desc_sec - time_to_download;
3986 *sec_to_download += time_to_download;
3988 if (desc_curr.end_time_ns >= end_time_ns) {
3989 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3990 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3991 sec_downloaded = percent_to_sub * sec_downloaded;
3992 *sec_to_download = percent_to_sub * *sec_to_download;
3994 if ((sec_downloaded + *buffer) <= min_buffer)
3999 if ((sec_downloaded + *buffer) <= min_buffer) {
4004 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4006 *buffer = *buffer + sec_downloaded;
4010 /* This function computes the bandwidth of the WebM file with the help of
4011 * buffer_size_after_time_downloaded() function. Both of these functions are
4012 * adapted from WebM Tools project and are adapted to work with FFmpeg's
4013 * Matroska parsing mechanism.
4015 * Returns the bandwidth of the file on success; -1 on error.
4017 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4019 MatroskaDemuxContext *matroska = s->priv_data;
4020 AVStream *st = s->streams[0];
4021 double bandwidth = 0.0;
4024 for (i = 0; i < st->internal->nb_index_entries; i++) {
4025 int64_t prebuffer_ns = 1000000000;
4026 int64_t time_ns = st->internal->index_entries[i].timestamp * matroska->time_scale;
4027 double nano_seconds_per_second = 1000000000.0;
4028 int64_t prebuffered_ns = time_ns + prebuffer_ns;
4029 double prebuffer_bytes = 0.0;
4030 int64_t temp_prebuffer_ns = prebuffer_ns;
4031 int64_t pre_bytes, pre_ns;
4032 double pre_sec, prebuffer, bits_per_second;
4033 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4035 // Start with the first Cue.
4036 CueDesc desc_end = desc_beg;
4038 // Figure out how much data we have downloaded for the prebuffer. This will
4039 // be used later to adjust the bits per sample to try.
4040 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4041 // Prebuffered the entire Cue.
4042 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4043 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4044 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4046 if (desc_end.start_time_ns == -1) {
4047 // The prebuffer is larger than the duration.
4048 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4050 bits_per_second = 0.0;
4052 // The prebuffer ends in the last Cue. Estimate how much data was
4054 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4055 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4056 pre_sec = pre_ns / nano_seconds_per_second;
4058 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4060 prebuffer = prebuffer_ns / nano_seconds_per_second;
4062 // Set this to 0.0 in case our prebuffer buffers the entire video.
4063 bits_per_second = 0.0;
4065 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4066 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4067 double desc_sec = desc_ns / nano_seconds_per_second;
4068 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4070 // Drop the bps by the percentage of bytes buffered.
4071 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4072 double mod_bits_per_second = calc_bits_per_second * percent;
4074 if (prebuffer < desc_sec) {
4076 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4078 // Add 1 so the bits per second should be a little bit greater than file
4080 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4081 const double min_buffer = 0.0;
4082 double buffer = prebuffer;
4083 double sec_to_download = 0.0;
4085 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4086 min_buffer, &buffer, &sec_to_download,
4090 } else if (rv == 0) {
4091 bits_per_second = (double)(bps);
4096 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4097 } while (desc_end.start_time_ns != -1);
4099 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4101 return (int64_t)bandwidth;
4104 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4106 MatroskaDemuxContext *matroska = s->priv_data;
4107 EbmlList *seekhead_list = &matroska->seekhead;
4108 MatroskaSeekhead *seekhead = seekhead_list->elem;
4110 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4114 // determine cues start and end positions
4115 for (i = 0; i < seekhead_list->nb_elem; i++)
4116 if (seekhead[i].id == MATROSKA_ID_CUES)
4119 if (i >= seekhead_list->nb_elem) return -1;
4121 before_pos = avio_tell(matroska->ctx->pb);
4122 cues_start = seekhead[i].pos + matroska->segment_start;
4123 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4124 // cues_end is computed as cues_start + cues_length + length of the
4125 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4126 // cues_end is inclusive and the above sum is reduced by 1.
4127 uint64_t cues_length, cues_id;
4129 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4130 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4131 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4132 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4135 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4137 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4138 if (cues_start == -1 || cues_end == -1) return -1;
4141 matroska_parse_cues(matroska);
4144 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4147 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4149 // if the file has cues at the start, fix up the init range so that
4150 // it does not include it
4151 if (cues_start <= init_range)
4152 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4155 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4156 if (bandwidth < 0) return -1;
4157 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4159 // check if all clusters start with key frames
4160 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4162 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4163 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4164 buf = av_malloc_array(s->streams[0]->internal->nb_index_entries, 20);
4165 if (!buf) return -1;
4167 for (i = 0; i < s->streams[0]->internal->nb_index_entries; i++) {
4168 int ret = snprintf(buf + end, 20,
4169 "%" PRId64"%s", s->streams[0]->internal->index_entries[i].timestamp,
4170 i != s->streams[0]->internal->nb_index_entries - 1 ? "," : "");
4171 if (ret <= 0 || (ret == 20 && i == s->streams[0]->internal->nb_index_entries - 1)) {
4172 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4174 return AVERROR_INVALIDDATA;
4178 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4179 buf, AV_DICT_DONT_STRDUP_VAL);
4184 static int webm_dash_manifest_read_header(AVFormatContext *s)
4187 int ret = matroska_read_header(s);
4189 MatroskaTrack *tracks;
4190 MatroskaDemuxContext *matroska = s->priv_data;
4192 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4195 if (!matroska->tracks.nb_elem || !s->nb_streams) {
4196 av_log(s, AV_LOG_ERROR, "No track found\n");
4197 ret = AVERROR_INVALIDDATA;
4201 if (!matroska->is_live) {
4202 buf = av_asprintf("%g", matroska->duration);
4204 ret = AVERROR(ENOMEM);
4207 av_dict_set(&s->streams[0]->metadata, DURATION,
4208 buf, AV_DICT_DONT_STRDUP_VAL);
4210 // initialization range
4211 // 5 is the offset of Cluster ID.
4212 init_range = avio_tell(s->pb) - 5;
4213 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4216 // basename of the file
4217 buf = strrchr(s->url, '/');
4218 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4221 tracks = matroska->tracks.elem;
4222 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4224 // parse the cues and populate Cue related fields
4225 if (!matroska->is_live) {
4226 ret = webm_dash_manifest_cues(s, init_range);
4228 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4233 // use the bandwidth from the command line if it was provided
4234 if (matroska->bandwidth > 0) {
4235 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4236 matroska->bandwidth, 0);
4240 matroska_read_close(s);
4244 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4249 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4250 static const AVOption options[] = {
4251 { "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 },
4252 { "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 },
4256 static const AVClass webm_dash_class = {
4257 .class_name = "WebM DASH Manifest demuxer",
4258 .item_name = av_default_item_name,
4260 .version = LIBAVUTIL_VERSION_INT,
4263 AVInputFormat ff_matroska_demuxer = {
4264 .name = "matroska,webm",
4265 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4266 .extensions = "mkv,mk3d,mka,mks",
4267 .priv_data_size = sizeof(MatroskaDemuxContext),
4268 .read_probe = matroska_probe,
4269 .read_header = matroska_read_header,
4270 .read_packet = matroska_read_packet,
4271 .read_close = matroska_read_close,
4272 .read_seek = matroska_read_seek,
4273 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4276 AVInputFormat ff_webm_dash_manifest_demuxer = {
4277 .name = "webm_dash_manifest",
4278 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4279 .priv_data_size = sizeof(MatroskaDemuxContext),
4280 .read_header = webm_dash_manifest_read_header,
4281 .read_packet = webm_dash_manifest_read_packet,
4282 .read_close = matroska_read_close,
4283 .priv_class = &webm_dash_class,