2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The FFmpeg Project
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * Matroska file demuxer
25 * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28 * @see specs available on the Matroska project page: http://www.matroska.org/
36 #include "libavutil/avstring.h"
37 #include "libavutil/base64.h"
38 #include "libavutil/dict.h"
39 #include "libavutil/intfloat.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/lzo.h"
42 #include "libavutil/mastering_display_metadata.h"
43 #include "libavutil/mathematics.h"
44 #include "libavutil/opt.h"
45 #include "libavutil/time_internal.h"
46 #include "libavutil/spherical.h"
48 #include "libavcodec/bytestream.h"
49 #include "libavcodec/flac.h"
50 #include "libavcodec/mpeg4audio.h"
53 #include "avio_internal.h"
58 /* For ff_codec_get_id(). */
69 #include "qtpalette.h"
71 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
72 #define NEEDS_CHECKING 2 /* Indicates that some error checks
73 * still need to be performed */
74 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
75 * syntax level used for parsing ended. */
76 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
77 * of unkown, potentially damaged data is encountered,
78 * it is considered an error. */
79 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
80 * to this many bytes of unknown data for the
81 * SKIP_THRESHOLD check. */
97 typedef const struct EbmlSyntax {
100 size_t list_elem_size;
107 const struct EbmlSyntax *n;
111 typedef struct EbmlList {
113 unsigned int alloc_elem_size;
117 typedef struct EbmlBin {
124 typedef struct Ebml {
129 uint64_t doctype_version;
132 typedef struct MatroskaTrackCompression {
135 } MatroskaTrackCompression;
137 typedef struct MatroskaTrackEncryption {
140 } MatroskaTrackEncryption;
142 typedef struct MatroskaTrackEncoding {
145 MatroskaTrackCompression compression;
146 MatroskaTrackEncryption encryption;
147 } MatroskaTrackEncoding;
149 typedef struct MatroskaMasteringMeta {
158 double max_luminance;
159 double min_luminance;
160 } MatroskaMasteringMeta;
162 typedef struct MatroskaTrackVideoColor {
163 uint64_t matrix_coefficients;
164 uint64_t bits_per_channel;
165 uint64_t chroma_sub_horz;
166 uint64_t chroma_sub_vert;
167 uint64_t cb_sub_horz;
168 uint64_t cb_sub_vert;
169 uint64_t chroma_siting_horz;
170 uint64_t chroma_siting_vert;
172 uint64_t transfer_characteristics;
176 MatroskaMasteringMeta mastering_meta;
177 } MatroskaTrackVideoColor;
179 typedef struct MatroskaTrackVideoProjection {
185 } MatroskaTrackVideoProjection;
187 typedef struct MatroskaTrackVideo {
189 uint64_t display_width;
190 uint64_t display_height;
191 uint64_t pixel_width;
192 uint64_t pixel_height;
194 uint64_t display_unit;
196 uint64_t field_order;
197 uint64_t stereo_mode;
200 MatroskaTrackVideoProjection projection;
201 } MatroskaTrackVideo;
203 typedef struct MatroskaTrackAudio {
205 double out_samplerate;
209 /* real audio header (extracted from extradata) */
216 uint64_t buf_timecode;
218 } MatroskaTrackAudio;
220 typedef struct MatroskaTrackPlane {
223 } MatroskaTrackPlane;
225 typedef struct MatroskaTrackOperation {
226 EbmlList combine_planes;
227 } MatroskaTrackOperation;
229 typedef struct MatroskaTrack {
238 uint64_t default_duration;
239 uint64_t flag_default;
240 uint64_t flag_forced;
241 uint64_t seek_preroll;
242 MatroskaTrackVideo video;
243 MatroskaTrackAudio audio;
244 MatroskaTrackOperation operation;
246 uint64_t codec_delay;
247 uint64_t codec_delay_in_track_tb;
250 int64_t end_timecode;
252 uint64_t max_block_additional_id;
254 uint32_t palette[AVPALETTE_COUNT];
258 typedef struct MatroskaAttachment {
265 } MatroskaAttachment;
267 typedef struct MatroskaChapter {
276 typedef struct MatroskaIndexPos {
281 typedef struct MatroskaIndex {
286 typedef struct MatroskaTag {
294 typedef struct MatroskaTagTarget {
302 typedef struct MatroskaTags {
303 MatroskaTagTarget target;
307 typedef struct MatroskaSeekhead {
312 typedef struct MatroskaLevel {
317 typedef struct MatroskaBlock {
322 uint64_t additional_id;
324 int64_t discard_padding;
327 typedef struct MatroskaCluster {
333 typedef struct MatroskaLevel1Element {
337 } MatroskaLevel1Element;
339 typedef struct MatroskaDemuxContext {
340 const AVClass *class;
341 AVFormatContext *ctx;
344 MatroskaLevel levels[EBML_MAX_DEPTH];
356 EbmlList attachments;
362 /* byte position of the segment inside the stream */
363 int64_t segment_start;
365 /* the packet queue */
367 AVPacketList *queue_end;
371 /* What to skip before effectively reading a packet. */
372 int skip_to_keyframe;
373 uint64_t skip_to_timecode;
375 /* File has a CUES element, but we defer parsing until it is needed. */
376 int cues_parsing_deferred;
378 /* Level1 elements and whether they were read yet */
379 MatroskaLevel1Element level1_elems[64];
380 int num_level1_elems;
382 MatroskaCluster current_cluster;
384 /* WebM DASH Manifest live flag */
387 /* Bandwidth value for WebM DASH Manifest */
389 } MatroskaDemuxContext;
391 #define CHILD_OF(parent) { .def = { .n = parent } }
393 // The following forward declarations need their size because
394 // a tentative definition with internal linkage must not be an
395 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
396 // Removing the sizes breaks MSVC.
397 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
398 matroska_track[27], matroska_track_encoding[6], matroska_track_encodings[2],
399 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_tracks[2],
400 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
401 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
402 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
404 static EbmlSyntax ebml_header[] = {
405 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
406 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
407 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
408 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
409 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
410 { EBML_ID_EBMLVERSION, EBML_NONE },
411 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
412 CHILD_OF(ebml_syntax)
415 static EbmlSyntax ebml_syntax[] = {
416 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
417 { MATROSKA_ID_SEGMENT, EBML_STOP },
421 static EbmlSyntax matroska_info[] = {
422 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
423 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
424 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
425 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
426 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
427 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
428 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
429 CHILD_OF(matroska_segment)
432 static EbmlSyntax matroska_mastering_meta[] = {
433 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
434 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
435 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
436 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
437 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
438 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
439 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
440 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
441 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
442 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
443 CHILD_OF(matroska_track_video_color)
446 static EbmlSyntax matroska_track_video_color[] = {
447 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
448 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=0 } },
449 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
450 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
451 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
452 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
453 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
454 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
455 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
456 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
457 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
458 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
459 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
460 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
461 CHILD_OF(matroska_track_video)
464 static EbmlSyntax matroska_track_video_projection[] = {
465 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
466 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrackVideoProjection, private) },
467 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f=0.0 } },
468 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f=0.0 } },
469 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f=0.0 } },
470 CHILD_OF(matroska_track_video)
473 static EbmlSyntax matroska_track_video[] = {
474 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
475 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
476 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
477 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
478 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
479 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
480 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
481 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
482 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
483 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
484 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
485 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
486 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
487 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
488 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
489 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
490 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
491 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
492 CHILD_OF(matroska_track)
495 static EbmlSyntax matroska_track_audio[] = {
496 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
497 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
498 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
499 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
500 CHILD_OF(matroska_track)
503 static EbmlSyntax matroska_track_encoding_compression[] = {
504 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
505 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
506 CHILD_OF(matroska_track_encoding)
509 static EbmlSyntax matroska_track_encoding_encryption[] = {
510 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
511 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
512 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
513 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
514 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
515 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
516 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
517 CHILD_OF(matroska_track_encoding)
519 static EbmlSyntax matroska_track_encoding[] = {
520 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
521 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
522 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
523 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
524 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
525 CHILD_OF(matroska_track_encodings)
528 static EbmlSyntax matroska_track_encodings[] = {
529 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
530 CHILD_OF(matroska_track)
533 static EbmlSyntax matroska_track_plane[] = {
534 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
535 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
536 CHILD_OF(matroska_track_combine_planes)
539 static EbmlSyntax matroska_track_combine_planes[] = {
540 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
541 CHILD_OF(matroska_track_operation)
544 static EbmlSyntax matroska_track_operation[] = {
545 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
546 CHILD_OF(matroska_track)
549 static EbmlSyntax matroska_track[] = {
550 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
551 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
552 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
553 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
554 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
555 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
556 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
557 { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
558 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
559 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
560 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
561 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
562 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
563 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
564 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
565 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
566 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
567 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
568 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
569 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
570 { MATROSKA_ID_CODECNAME, EBML_NONE },
571 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
572 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
573 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
574 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
575 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
576 CHILD_OF(matroska_tracks)
579 static EbmlSyntax matroska_tracks[] = {
580 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
581 CHILD_OF(matroska_segment)
584 static EbmlSyntax matroska_attachment[] = {
585 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
586 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
587 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
588 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
589 { MATROSKA_ID_FILEDESC, EBML_NONE },
590 CHILD_OF(matroska_attachments)
593 static EbmlSyntax matroska_attachments[] = {
594 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
595 CHILD_OF(matroska_segment)
598 static EbmlSyntax matroska_chapter_display[] = {
599 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
600 { MATROSKA_ID_CHAPLANG, EBML_NONE },
601 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
602 CHILD_OF(matroska_chapter_entry)
605 static EbmlSyntax matroska_chapter_entry[] = {
606 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
607 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
608 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
609 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
610 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
611 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
612 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
613 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
614 CHILD_OF(matroska_chapter)
617 static EbmlSyntax matroska_chapter[] = {
618 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
619 { MATROSKA_ID_EDITIONUID, EBML_NONE },
620 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
621 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
622 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
623 CHILD_OF(matroska_chapters)
626 static EbmlSyntax matroska_chapters[] = {
627 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
628 CHILD_OF(matroska_segment)
631 static EbmlSyntax matroska_index_pos[] = {
632 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
633 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
634 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
635 { MATROSKA_ID_CUEDURATION, EBML_NONE },
636 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
637 CHILD_OF(matroska_index_entry)
640 static EbmlSyntax matroska_index_entry[] = {
641 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
642 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
643 CHILD_OF(matroska_index)
646 static EbmlSyntax matroska_index[] = {
647 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
648 CHILD_OF(matroska_segment)
651 static EbmlSyntax matroska_simpletag[] = {
652 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
653 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
654 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
655 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
656 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
657 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
658 CHILD_OF(matroska_tag)
661 static EbmlSyntax matroska_tagtargets[] = {
662 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
663 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
664 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
665 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
666 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
667 CHILD_OF(matroska_tag)
670 static EbmlSyntax matroska_tag[] = {
671 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
672 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
673 CHILD_OF(matroska_tags)
676 static EbmlSyntax matroska_tags[] = {
677 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
678 CHILD_OF(matroska_segment)
681 static EbmlSyntax matroska_seekhead_entry[] = {
682 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
683 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
684 CHILD_OF(matroska_seekhead)
687 static EbmlSyntax matroska_seekhead[] = {
688 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
689 CHILD_OF(matroska_segment)
692 static EbmlSyntax matroska_segment[] = {
693 { MATROSKA_ID_CLUSTER, EBML_STOP },
694 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
695 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
696 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
697 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
698 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
699 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
700 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
701 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
704 static EbmlSyntax matroska_segments[] = {
705 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
709 static EbmlSyntax matroska_blockmore[] = {
710 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
711 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
712 CHILD_OF(matroska_blockadditions)
715 static EbmlSyntax matroska_blockadditions[] = {
716 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
717 CHILD_OF(matroska_blockgroup)
720 static EbmlSyntax matroska_blockgroup[] = {
721 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
722 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
723 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
724 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
725 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference), { .i = INT64_MIN } },
726 { MATROSKA_ID_CODECSTATE, EBML_NONE },
727 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
728 CHILD_OF(matroska_cluster_parsing)
731 // The following array contains SimpleBlock and BlockGroup twice
732 // in order to reuse the other values for matroska_cluster_enter.
733 static EbmlSyntax matroska_cluster_parsing[] = {
734 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
735 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, { .n = matroska_blockgroup } },
736 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
737 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
738 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
739 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
740 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
741 CHILD_OF(matroska_segment)
744 static EbmlSyntax matroska_cluster_enter[] = {
745 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = &matroska_cluster_parsing[2] } },
750 static const CodecMime mkv_image_mime_tags[] = {
751 {"image/gif" , AV_CODEC_ID_GIF},
752 {"image/jpeg" , AV_CODEC_ID_MJPEG},
753 {"image/png" , AV_CODEC_ID_PNG},
754 {"image/tiff" , AV_CODEC_ID_TIFF},
756 {"" , AV_CODEC_ID_NONE}
759 static const CodecMime mkv_mime_tags[] = {
760 {"text/plain" , AV_CODEC_ID_TEXT},
761 {"application/x-truetype-font", AV_CODEC_ID_TTF},
762 {"application/x-font" , AV_CODEC_ID_TTF},
763 {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
764 {"binary" , AV_CODEC_ID_BIN_DATA},
766 {"" , AV_CODEC_ID_NONE}
769 static const char *const matroska_doctypes[] = { "matroska", "webm" };
771 static int matroska_read_close(AVFormatContext *s);
774 * This function prepares the status for parsing of level 1 elements.
776 static int matroska_reset_status(MatroskaDemuxContext *matroska,
777 uint32_t id, int64_t position)
780 int64_t err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
785 matroska->current_id = id;
786 matroska->num_levels = 1;
787 matroska->unknown_count = 0;
788 matroska->resync_pos = avio_tell(matroska->ctx->pb);
790 matroska->resync_pos -= (av_log2(id) + 7) / 8;
795 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
797 AVIOContext *pb = matroska->ctx->pb;
800 /* Try to seek to the last position to resync from. If this doesn't work,
801 * we resync from the earliest position available: The start of the buffer. */
802 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
803 av_log(matroska->ctx, AV_LOG_WARNING,
804 "Seek to desired resync point failed. Seeking to "
805 "earliest point available instead.\n");
806 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
807 last_pos + 1), SEEK_SET);
812 // try to find a toplevel element
813 while (!avio_feof(pb)) {
814 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
815 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
816 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
817 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
818 /* Prepare the context for parsing of a level 1 element. */
819 matroska_reset_status(matroska, id, -1);
820 /* Given that we are here means that an error has occurred,
821 * so treat the segment as unknown length in order not to
822 * discard valid data that happens to be beyond the designated
823 * end of the segment. */
824 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
827 id = (id << 8) | avio_r8(pb);
831 return pb->error ? pb->error : AVERROR_EOF;
835 * Read: an "EBML number", which is defined as a variable-length
836 * array of bytes. The first byte indicates the length by giving a
837 * number of 0-bits followed by a one. The position of the first
838 * "one" bit inside the first byte indicates the length of this
840 * Returns: number of bytes read, < 0 on error
842 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
843 int max_size, uint64_t *number, int eof_forbidden)
849 /* The first byte tells us the length in bytes - except when it is zero. */
854 /* get the length of the EBML number */
855 read = 8 - ff_log2_tab[total];
857 if (!total || read > max_size) {
858 pos = avio_tell(pb) - 1;
860 av_log(matroska->ctx, AV_LOG_ERROR,
861 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
862 "of an EBML number\n", pos, pos);
864 av_log(matroska->ctx, AV_LOG_ERROR,
865 "Length %d indicated by an EBML number's first byte 0x%02x "
866 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
867 read, (uint8_t) total, pos, pos, max_size);
869 return AVERROR_INVALIDDATA;
872 /* read out length */
873 total ^= 1 << ff_log2_tab[total];
875 total = (total << 8) | avio_r8(pb);
877 if (pb->eof_reached) {
889 av_log(matroska->ctx, AV_LOG_ERROR,
890 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
895 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
896 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
903 * Read a EBML length value.
904 * This needs special handling for the "unknown length" case which has multiple
907 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
910 int res = ebml_read_num(matroska, pb, 8, number, 1);
911 if (res > 0 && *number + 1 == 1ULL << (7 * res))
912 *number = EBML_UNKNOWN_LENGTH;
917 * Read the next element as an unsigned int.
918 * Returns NEEDS_CHECKING.
920 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
924 /* big-endian ordering; build up number */
927 *num = (*num << 8) | avio_r8(pb);
929 return NEEDS_CHECKING;
933 * Read the next element as a signed int.
934 * Returns NEEDS_CHECKING.
936 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
943 *num = sign_extend(avio_r8(pb), 8);
945 /* big-endian ordering; build up number */
947 *num = ((uint64_t)*num << 8) | avio_r8(pb);
950 return NEEDS_CHECKING;
954 * Read the next element as a float.
955 * Returns NEEDS_CHECKING or < 0 on obvious failure.
957 static int ebml_read_float(AVIOContext *pb, int size, double *num)
962 *num = av_int2float(avio_rb32(pb));
964 *num = av_int2double(avio_rb64(pb));
966 return AVERROR_INVALIDDATA;
968 return NEEDS_CHECKING;
972 * Read the next element as an ASCII string.
973 * 0 is success, < 0 or NEEDS_CHECKING is failure.
975 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
980 /* EBML strings are usually not 0-terminated, so we allocate one
981 * byte more, read the string and NULL-terminate it ourselves. */
982 if (!(res = av_malloc(size + 1)))
983 return AVERROR(ENOMEM);
984 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
986 return ret < 0 ? ret : NEEDS_CHECKING;
996 * Read the next element as binary data.
997 * 0 is success, < 0 or NEEDS_CHECKING is failure.
999 static int ebml_read_binary(AVIOContext *pb, int length,
1000 int64_t pos, EbmlBin *bin)
1004 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1007 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1009 bin->data = bin->buf->data;
1012 if ((ret = avio_read(pb, bin->data, length)) != length) {
1013 av_buffer_unref(&bin->buf);
1016 return ret < 0 ? ret : NEEDS_CHECKING;
1023 * Read the next element, but only the header. The contents
1024 * are supposed to be sub-elements which can be read separately.
1025 * 0 is success, < 0 is failure.
1027 static int ebml_read_master(MatroskaDemuxContext *matroska,
1028 uint64_t length, int64_t pos)
1030 MatroskaLevel *level;
1032 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1033 av_log(matroska->ctx, AV_LOG_ERROR,
1034 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1035 return AVERROR(ENOSYS);
1038 level = &matroska->levels[matroska->num_levels++];
1040 level->length = length;
1046 * Read a signed "EBML number"
1047 * Return: number of bytes processed, < 0 on error
1049 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1050 AVIOContext *pb, int64_t *num)
1055 /* read as unsigned number first */
1056 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1059 /* make signed (weird way) */
1060 *num = unum - ((1LL << (7 * res - 1)) - 1);
1065 static int ebml_parse(MatroskaDemuxContext *matroska,
1066 EbmlSyntax *syntax, void *data);
1068 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1072 // Whoever touches this should be aware of the duplication
1073 // existing in matroska_cluster_parsing.
1074 for (i = 0; syntax[i].id; i++)
1075 if (id == syntax[i].id)
1081 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1087 for (int i = 0; syntax[i].id; i++)
1088 switch (syntax[i].type) {
1090 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1093 *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1096 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1100 // the default may be NULL
1101 if (syntax[i].def.s) {
1102 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1103 *dst = av_strdup(syntax[i].def.s);
1105 return AVERROR(ENOMEM);
1110 if (!matroska->levels[matroska->num_levels - 1].length) {
1111 matroska->num_levels--;
1117 res = ebml_parse(matroska, syntax, data);
1120 return res == LEVEL_ENDED ? 0 : res;
1123 static int is_ebml_id_valid(uint32_t id)
1125 // Due to endian nonsense in Matroska, the highest byte with any bits set
1126 // will contain the leading length bit. This bit in turn identifies the
1127 // total byte length of the element by its position within the byte.
1128 unsigned int bits = av_log2(id);
1129 return id && (bits + 7) / 8 == (8 - bits % 8);
1133 * Allocate and return the entry for the level1 element with the given ID. If
1134 * an entry already exists, return the existing entry.
1136 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1140 MatroskaLevel1Element *elem;
1142 if (!is_ebml_id_valid(id))
1145 // Some files link to all clusters; useless.
1146 if (id == MATROSKA_ID_CLUSTER)
1149 // There can be multiple seekheads.
1150 if (id != MATROSKA_ID_SEEKHEAD) {
1151 for (i = 0; i < matroska->num_level1_elems; i++) {
1152 if (matroska->level1_elems[i].id == id)
1153 return &matroska->level1_elems[i];
1157 // Only a completely broken file would have more elements.
1158 // It also provides a low-effort way to escape from circular seekheads
1159 // (every iteration will add a level1 entry).
1160 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1161 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1165 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1166 *elem = (MatroskaLevel1Element){.id = id};
1171 static int ebml_parse(MatroskaDemuxContext *matroska,
1172 EbmlSyntax *syntax, void *data)
1174 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1175 // Forbid unknown-length EBML_NONE elements.
1176 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1180 // max. 16 MB for strings
1181 [EBML_STR] = 0x1000000,
1182 [EBML_UTF8] = 0x1000000,
1183 // max. 256 MB for binary data
1184 [EBML_BIN] = 0x10000000,
1185 // no limits for anything else
1187 AVIOContext *pb = matroska->ctx->pb;
1190 int64_t pos = avio_tell(pb), pos_alt;
1191 int res, update_pos = 1, level_check;
1192 MatroskaLevel1Element *level1_elem;
1193 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1195 if (!matroska->current_id) {
1197 res = ebml_read_num(matroska, pb, 4, &id, 0);
1199 if (pb->eof_reached && res == AVERROR_EOF) {
1200 if (matroska->is_live)
1201 // in live mode, finish parsing if EOF is reached.
1203 if (level && pos == avio_tell(pb)) {
1204 if (level->length == EBML_UNKNOWN_LENGTH) {
1205 // Unknown-length levels automatically end at EOF.
1206 matroska->num_levels--;
1209 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1210 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1216 matroska->current_id = id | 1 << 7 * res;
1217 pos_alt = pos + res;
1220 pos -= (av_log2(matroska->current_id) + 7) / 8;
1223 id = matroska->current_id;
1225 syntax = ebml_parse_id(syntax, id);
1226 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1227 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1228 // Unknown-length levels end when an element from an upper level
1229 // in the hierarchy is encountered.
1230 while (syntax->def.n) {
1231 syntax = ebml_parse_id(syntax->def.n, id);
1233 matroska->num_levels--;
1239 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1240 "%"PRId64"\n", id, pos);
1241 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1245 data = (char *) data + syntax->data_offset;
1246 if (syntax->list_elem_size) {
1247 EbmlList *list = data;
1250 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1251 return AVERROR(ENOMEM);
1252 newelem = av_fast_realloc(list->elem,
1253 &list->alloc_elem_size,
1254 (list->nb_elem + 1) * syntax->list_elem_size);
1256 return AVERROR(ENOMEM);
1257 list->elem = newelem;
1258 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1259 memset(data, 0, syntax->list_elem_size);
1264 if (syntax->type != EBML_STOP) {
1265 matroska->current_id = 0;
1266 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1271 if (matroska->num_levels > 0) {
1272 if (length != EBML_UNKNOWN_LENGTH &&
1273 level->length != EBML_UNKNOWN_LENGTH) {
1274 uint64_t elem_end = pos_alt + length,
1275 level_end = level->start + level->length;
1277 if (elem_end < level_end) {
1279 } else if (elem_end == level_end) {
1280 level_check = LEVEL_ENDED;
1282 av_log(matroska->ctx, AV_LOG_ERROR,
1283 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1284 "containing master element ending at 0x%"PRIx64"\n",
1285 pos, elem_end, level_end);
1286 return AVERROR_INVALIDDATA;
1288 } else if (length != EBML_UNKNOWN_LENGTH) {
1290 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1291 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1292 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1293 return AVERROR_INVALIDDATA;
1296 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1297 || syntax->type == EBML_NEST)) {
1298 // According to the current specifications only clusters and
1299 // segments are allowed to be unknown-length. We also accept
1300 // other unknown-length master elements.
1301 av_log(matroska->ctx, AV_LOG_WARNING,
1302 "Found unknown-length element 0x%"PRIX32" other than "
1303 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1304 "parsing will nevertheless be attempted.\n", id, pos);
1311 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1312 if (length != EBML_UNKNOWN_LENGTH) {
1313 av_log(matroska->ctx, AV_LOG_ERROR,
1314 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1315 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1316 length, max_lengths[syntax->type], id, pos);
1317 } else if (syntax->type != EBML_NONE) {
1318 av_log(matroska->ctx, AV_LOG_ERROR,
1319 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1320 "unknown length, yet the length of an element of its "
1321 "type must be known.\n", id, pos);
1323 av_log(matroska->ctx, AV_LOG_ERROR,
1324 "Found unknown-length element with ID 0x%"PRIX32" at "
1325 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1326 "available.\n", id, pos);
1328 return AVERROR_INVALIDDATA;
1331 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1332 // Loosing sync will likely manifest itself as encountering unknown
1333 // elements which are not reliably distinguishable from elements
1334 // belonging to future extensions of the format.
1335 // We use a heuristic to detect such situations: If the current
1336 // element is not expected at the current syntax level and there
1337 // were only a few unknown elements in a row, then the element is
1338 // skipped or considered defective based upon the length of the
1339 // current element (i.e. how much would be skipped); if there were
1340 // more than a few skipped elements in a row and skipping the current
1341 // element would lead us more than SKIP_THRESHOLD away from the last
1342 // known good position, then it is inferred that an error occurred.
1343 // The dependency on the number of unknown elements in a row exists
1344 // because the distance to the last known good position is
1345 // automatically big if the last parsed element was big.
1346 // In both cases, each unknown element is considered equivalent to
1347 // UNKNOWN_EQUIV of skipped bytes for the check.
1348 // The whole check is only done for non-seekable output, because
1349 // in this situation skipped data can't simply be rechecked later.
1350 // This is especially important when using unkown length elements
1351 // as the check for whether a child exceeds its containing master
1352 // element is not effective in this situation.
1354 matroska->unknown_count = 0;
1356 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1358 if (matroska->unknown_count > 3)
1359 dist += pos_alt - matroska->resync_pos;
1361 if (dist > SKIP_THRESHOLD) {
1362 av_log(matroska->ctx, AV_LOG_ERROR,
1363 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1364 "length 0x%"PRIx64" considered as invalid data. Last "
1365 "known good position 0x%"PRIx64", %d unknown elements"
1366 " in a row\n", id, pos, length, matroska->resync_pos,
1367 matroska->unknown_count);
1368 return AVERROR_INVALIDDATA;
1373 if (update_pos > 0) {
1374 // We have found an element that is allowed at this place
1375 // in the hierarchy and it passed all checks, so treat the beginning
1376 // of the element as the "last known good" position.
1377 matroska->resync_pos = pos;
1380 if (!data && length != EBML_UNKNOWN_LENGTH)
1384 switch (syntax->type) {
1386 res = ebml_read_uint(pb, length, data);
1389 res = ebml_read_sint(pb, length, data);
1392 res = ebml_read_float(pb, length, data);
1396 res = ebml_read_ascii(pb, length, data);
1399 res = ebml_read_binary(pb, length, pos_alt, data);
1403 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1405 if (id == MATROSKA_ID_SEGMENT)
1406 matroska->segment_start = pos_alt;
1407 if (id == MATROSKA_ID_CUES)
1408 matroska->cues_parsing_deferred = 0;
1409 if (syntax->type == EBML_LEVEL1 &&
1410 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1411 if (!level1_elem->pos) {
1412 // Zero is not a valid position for a level 1 element.
1413 level1_elem->pos = pos;
1414 } else if (level1_elem->pos != pos)
1415 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1416 level1_elem->parsed = 1;
1418 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1427 if (ffio_limit(pb, length) != length) {
1428 // ffio_limit emits its own error message,
1429 // so we don't have to.
1430 return AVERROR(EIO);
1432 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1433 // avio_skip might take us past EOF. We check for this
1434 // by skipping only length - 1 bytes, reading a byte and
1435 // checking the error flags. This is done in order to check
1436 // that the element has been properly skipped even when
1437 // no filesize (that ffio_limit relies on) is available.
1439 res = NEEDS_CHECKING;
1446 if (res == NEEDS_CHECKING) {
1447 if (pb->eof_reached) {
1456 if (res == AVERROR_INVALIDDATA)
1457 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1458 else if (res == AVERROR(EIO))
1459 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1460 else if (res == AVERROR_EOF) {
1461 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1469 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1470 level = &matroska->levels[matroska->num_levels - 1];
1471 pos = avio_tell(pb);
1473 // Given that pos >= level->start no check for
1474 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1475 while (matroska->num_levels && pos == level->start + level->length) {
1476 matroska->num_levels--;
1484 static void ebml_free(EbmlSyntax *syntax, void *data)
1487 for (i = 0; syntax[i].id; i++) {
1488 void *data_off = (char *) data + syntax[i].data_offset;
1489 switch (syntax[i].type) {
1495 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1499 if (syntax[i].list_elem_size) {
1500 EbmlList *list = data_off;
1501 char *ptr = list->elem;
1502 for (j = 0; j < list->nb_elem;
1503 j++, ptr += syntax[i].list_elem_size)
1504 ebml_free(syntax[i].def.n, ptr);
1505 av_freep(&list->elem);
1507 list->alloc_elem_size = 0;
1509 ebml_free(syntax[i].def.n, data_off);
1519 static int matroska_probe(const AVProbeData *p)
1522 int len_mask = 0x80, size = 1, n = 1, i;
1525 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1528 /* length of header */
1530 while (size <= 8 && !(total & len_mask)) {
1536 total &= (len_mask - 1);
1538 total = (total << 8) | p->buf[4 + n++];
1540 if (total + 1 == 1ULL << (7 * size)){
1541 /* Unknown-length header - simply parse the whole buffer. */
1542 total = p->buf_size - 4 - size;
1544 /* Does the probe data contain the whole header? */
1545 if (p->buf_size < 4 + size + total)
1549 /* The header should contain a known document type. For now,
1550 * we don't parse the whole header but simply check for the
1551 * availability of that array of characters inside the header.
1552 * Not fully fool-proof, but good enough. */
1553 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1554 size_t probelen = strlen(matroska_doctypes[i]);
1555 if (total < probelen)
1557 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1558 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1559 return AVPROBE_SCORE_MAX;
1562 // probably valid EBML header but no recognized doctype
1563 return AVPROBE_SCORE_EXTENSION;
1566 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1569 MatroskaTrack *tracks = matroska->tracks.elem;
1572 for (i = 0; i < matroska->tracks.nb_elem; i++)
1573 if (tracks[i].num == num)
1576 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1580 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1581 MatroskaTrack *track)
1583 MatroskaTrackEncoding *encodings = track->encodings.elem;
1584 uint8_t *data = *buf;
1585 int isize = *buf_size;
1586 uint8_t *pkt_data = NULL;
1587 uint8_t av_unused *newpktdata;
1588 int pkt_size = isize;
1592 if (pkt_size >= 10000000U)
1593 return AVERROR_INVALIDDATA;
1595 switch (encodings[0].compression.algo) {
1596 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1598 int header_size = encodings[0].compression.settings.size;
1599 uint8_t *header = encodings[0].compression.settings.data;
1601 if (header_size && !header) {
1602 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1609 pkt_size = isize + header_size;
1610 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1612 return AVERROR(ENOMEM);
1614 memcpy(pkt_data, header, header_size);
1615 memcpy(pkt_data + header_size, data, isize);
1619 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1622 olen = pkt_size *= 3;
1623 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1624 + AV_INPUT_BUFFER_PADDING_SIZE);
1626 result = AVERROR(ENOMEM);
1629 pkt_data = newpktdata;
1630 result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1631 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1633 result = AVERROR_INVALIDDATA;
1640 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1642 z_stream zstream = { 0 };
1643 if (inflateInit(&zstream) != Z_OK)
1645 zstream.next_in = data;
1646 zstream.avail_in = isize;
1649 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1651 inflateEnd(&zstream);
1652 result = AVERROR(ENOMEM);
1655 pkt_data = newpktdata;
1656 zstream.avail_out = pkt_size - zstream.total_out;
1657 zstream.next_out = pkt_data + zstream.total_out;
1658 result = inflate(&zstream, Z_NO_FLUSH);
1659 } while (result == Z_OK && pkt_size < 10000000);
1660 pkt_size = zstream.total_out;
1661 inflateEnd(&zstream);
1662 if (result != Z_STREAM_END) {
1663 if (result == Z_MEM_ERROR)
1664 result = AVERROR(ENOMEM);
1666 result = AVERROR_INVALIDDATA;
1673 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1675 bz_stream bzstream = { 0 };
1676 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1678 bzstream.next_in = data;
1679 bzstream.avail_in = isize;
1682 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1684 BZ2_bzDecompressEnd(&bzstream);
1685 result = AVERROR(ENOMEM);
1688 pkt_data = newpktdata;
1689 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1690 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1691 result = BZ2_bzDecompress(&bzstream);
1692 } while (result == BZ_OK && pkt_size < 10000000);
1693 pkt_size = bzstream.total_out_lo32;
1694 BZ2_bzDecompressEnd(&bzstream);
1695 if (result != BZ_STREAM_END) {
1696 if (result == BZ_MEM_ERROR)
1697 result = AVERROR(ENOMEM);
1699 result = AVERROR_INVALIDDATA;
1706 return AVERROR_INVALIDDATA;
1709 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1712 *buf_size = pkt_size;
1720 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1721 AVDictionary **metadata, char *prefix)
1723 MatroskaTag *tags = list->elem;
1727 for (i = 0; i < list->nb_elem; i++) {
1728 const char *lang = tags[i].lang &&
1729 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1731 if (!tags[i].name) {
1732 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1736 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1738 av_strlcpy(key, tags[i].name, sizeof(key));
1739 if (tags[i].def || !lang) {
1740 av_dict_set(metadata, key, tags[i].string, 0);
1741 if (tags[i].sub.nb_elem)
1742 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1745 av_strlcat(key, "-", sizeof(key));
1746 av_strlcat(key, lang, sizeof(key));
1747 av_dict_set(metadata, key, tags[i].string, 0);
1748 if (tags[i].sub.nb_elem)
1749 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1752 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1755 static void matroska_convert_tags(AVFormatContext *s)
1757 MatroskaDemuxContext *matroska = s->priv_data;
1758 MatroskaTags *tags = matroska->tags.elem;
1761 for (i = 0; i < matroska->tags.nb_elem; i++) {
1762 if (tags[i].target.attachuid) {
1763 MatroskaAttachment *attachment = matroska->attachments.elem;
1765 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1766 if (attachment[j].uid == tags[i].target.attachuid &&
1767 attachment[j].stream) {
1768 matroska_convert_tag(s, &tags[i].tag,
1769 &attachment[j].stream->metadata, NULL);
1774 av_log(NULL, AV_LOG_WARNING,
1775 "The tags at index %d refer to a "
1776 "non-existent attachment %"PRId64".\n",
1777 i, tags[i].target.attachuid);
1779 } else if (tags[i].target.chapteruid) {
1780 MatroskaChapter *chapter = matroska->chapters.elem;
1782 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1783 if (chapter[j].uid == tags[i].target.chapteruid &&
1784 chapter[j].chapter) {
1785 matroska_convert_tag(s, &tags[i].tag,
1786 &chapter[j].chapter->metadata, NULL);
1791 av_log(NULL, AV_LOG_WARNING,
1792 "The tags at index %d refer to a non-existent chapter "
1794 i, tags[i].target.chapteruid);
1796 } else if (tags[i].target.trackuid) {
1797 MatroskaTrack *track = matroska->tracks.elem;
1799 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1800 if (track[j].uid == tags[i].target.trackuid &&
1802 matroska_convert_tag(s, &tags[i].tag,
1803 &track[j].stream->metadata, NULL);
1808 av_log(NULL, AV_LOG_WARNING,
1809 "The tags at index %d refer to a non-existent track "
1811 i, tags[i].target.trackuid);
1814 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1815 tags[i].target.type);
1820 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1823 uint32_t saved_id = matroska->current_id;
1824 int64_t before_pos = avio_tell(matroska->ctx->pb);
1828 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1829 /* We don't want to lose our seekhead level, so we add
1830 * a dummy. This is a crude hack. */
1831 if (matroska->num_levels == EBML_MAX_DEPTH) {
1832 av_log(matroska->ctx, AV_LOG_INFO,
1833 "Max EBML element depth (%d) reached, "
1834 "cannot parse further.\n", EBML_MAX_DEPTH);
1835 ret = AVERROR_INVALIDDATA;
1837 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1838 matroska->num_levels++;
1839 matroska->current_id = 0;
1841 ret = ebml_parse(matroska, matroska_segment, matroska);
1842 if (ret == LEVEL_ENDED) {
1843 /* This can only happen if the seek brought us beyond EOF. */
1848 /* Seek back - notice that in all instances where this is used
1849 * it is safe to set the level to 1. */
1850 matroska_reset_status(matroska, saved_id, before_pos);
1855 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1857 EbmlList *seekhead_list = &matroska->seekhead;
1860 // we should not do any seeking in the streaming case
1861 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1864 for (i = 0; i < seekhead_list->nb_elem; i++) {
1865 MatroskaSeekhead *seekheads = seekhead_list->elem;
1866 uint32_t id = seekheads[i].id;
1867 int64_t pos = seekheads[i].pos + matroska->segment_start;
1869 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1870 if (!elem || elem->parsed)
1875 // defer cues parsing until we actually need cue data.
1876 if (id == MATROSKA_ID_CUES)
1879 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1880 // mark index as broken
1881 matroska->cues_parsing_deferred = -1;
1889 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1891 EbmlList *index_list;
1892 MatroskaIndex *index;
1893 uint64_t index_scale = 1;
1896 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1899 index_list = &matroska->index;
1900 index = index_list->elem;
1901 if (index_list->nb_elem < 2)
1903 if (index[1].time > 1E14 / matroska->time_scale) {
1904 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1907 for (i = 0; i < index_list->nb_elem; i++) {
1908 EbmlList *pos_list = &index[i].pos;
1909 MatroskaIndexPos *pos = pos_list->elem;
1910 for (j = 0; j < pos_list->nb_elem; j++) {
1911 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1913 if (track && track->stream)
1914 av_add_index_entry(track->stream,
1915 pos[j].pos + matroska->segment_start,
1916 index[i].time / index_scale, 0, 0,
1922 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1925 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1928 for (i = 0; i < matroska->num_level1_elems; i++) {
1929 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1930 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1931 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1932 matroska->cues_parsing_deferred = -1;
1938 matroska_add_index_entries(matroska);
1941 static int matroska_aac_profile(char *codec_id)
1943 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1946 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1947 if (strstr(codec_id, aac_profiles[profile]))
1952 static int matroska_aac_sri(int samplerate)
1956 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1957 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1962 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1964 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1965 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1968 static int matroska_parse_flac(AVFormatContext *s,
1969 MatroskaTrack *track,
1972 AVStream *st = track->stream;
1973 uint8_t *p = track->codec_priv.data;
1974 int size = track->codec_priv.size;
1976 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1977 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1978 track->codec_priv.size = 0;
1982 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1984 p += track->codec_priv.size;
1985 size -= track->codec_priv.size;
1987 /* parse the remaining metadata blocks if present */
1989 int block_last, block_type, block_size;
1991 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1995 if (block_size > size)
1998 /* check for the channel mask */
1999 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2000 AVDictionary *dict = NULL;
2001 AVDictionaryEntry *chmask;
2003 ff_vorbis_comment(s, &dict, p, block_size, 0);
2004 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2006 uint64_t mask = strtol(chmask->value, NULL, 0);
2007 if (!mask || mask & ~0x3ffffULL) {
2008 av_log(s, AV_LOG_WARNING,
2009 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2011 st->codecpar->channel_layout = mask;
2013 av_dict_free(&dict);
2023 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2025 int major, minor, micro, bttb = 0;
2027 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2028 * this function, and fixed in 57.52 */
2029 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, µ) == 3)
2030 bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
2032 switch (field_order) {
2033 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2034 return AV_FIELD_PROGRESSIVE;
2035 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2036 return AV_FIELD_UNKNOWN;
2037 case MATROSKA_VIDEO_FIELDORDER_TT:
2039 case MATROSKA_VIDEO_FIELDORDER_BB:
2041 case MATROSKA_VIDEO_FIELDORDER_BT:
2042 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2043 case MATROSKA_VIDEO_FIELDORDER_TB:
2044 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2046 return AV_FIELD_UNKNOWN;
2050 static void mkv_stereo_mode_display_mul(int stereo_mode,
2051 int *h_width, int *h_height)
2053 switch (stereo_mode) {
2054 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2055 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2056 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2057 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2058 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2060 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2061 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2062 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2063 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2066 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2067 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2068 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2069 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2075 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2076 const MatroskaTrackVideoColor *color = track->video.color.elem;
2077 const MatroskaMasteringMeta *mastering_meta;
2078 int has_mastering_primaries, has_mastering_luminance;
2080 if (!track->video.color.nb_elem)
2083 mastering_meta = &color->mastering_meta;
2084 // Mastering primaries are CIE 1931 coords, and must be > 0.
2085 has_mastering_primaries =
2086 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2087 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2088 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2089 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2090 has_mastering_luminance = mastering_meta->max_luminance > 0;
2092 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2093 st->codecpar->color_space = color->matrix_coefficients;
2094 if (color->primaries != AVCOL_PRI_RESERVED &&
2095 color->primaries != AVCOL_PRI_RESERVED0)
2096 st->codecpar->color_primaries = color->primaries;
2097 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2098 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2099 st->codecpar->color_trc = color->transfer_characteristics;
2100 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2101 color->range <= AVCOL_RANGE_JPEG)
2102 st->codecpar->color_range = color->range;
2103 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2104 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2105 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2106 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2107 st->codecpar->chroma_location =
2108 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2109 (color->chroma_siting_vert - 1) << 7);
2111 if (color->max_cll && color->max_fall) {
2114 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2116 return AVERROR(ENOMEM);
2117 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2118 (uint8_t *)metadata, size);
2120 av_freep(&metadata);
2123 metadata->MaxCLL = color->max_cll;
2124 metadata->MaxFALL = color->max_fall;
2127 if (has_mastering_primaries || has_mastering_luminance) {
2128 AVMasteringDisplayMetadata *metadata =
2129 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2130 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2131 sizeof(AVMasteringDisplayMetadata));
2133 return AVERROR(ENOMEM);
2135 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2136 if (has_mastering_primaries) {
2137 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2138 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2139 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2140 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2141 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2142 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2143 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2144 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2145 metadata->has_primaries = 1;
2147 if (has_mastering_luminance) {
2148 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2149 metadata->min_luminance = av_d2q(mastering_meta->min_luminance, INT_MAX);
2150 metadata->has_luminance = 1;
2156 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track) {
2157 AVSphericalMapping *spherical;
2158 enum AVSphericalProjection projection;
2159 size_t spherical_size;
2160 uint32_t l = 0, t = 0, r = 0, b = 0;
2161 uint32_t padding = 0;
2165 bytestream2_init(&gb, track->video.projection.private.data,
2166 track->video.projection.private.size);
2168 if (bytestream2_get_byte(&gb) != 0) {
2169 av_log(NULL, AV_LOG_WARNING, "Unknown spherical metadata\n");
2173 bytestream2_skip(&gb, 3); // flags
2175 switch (track->video.projection.type) {
2176 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2177 if (track->video.projection.private.size == 20) {
2178 t = bytestream2_get_be32(&gb);
2179 b = bytestream2_get_be32(&gb);
2180 l = bytestream2_get_be32(&gb);
2181 r = bytestream2_get_be32(&gb);
2183 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2184 av_log(NULL, AV_LOG_ERROR,
2185 "Invalid bounding rectangle coordinates "
2186 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2188 return AVERROR_INVALIDDATA;
2190 } else if (track->video.projection.private.size != 0) {
2191 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2192 return AVERROR_INVALIDDATA;
2195 if (l || t || r || b)
2196 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2198 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2200 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2201 if (track->video.projection.private.size < 4) {
2202 av_log(NULL, AV_LOG_ERROR, "Missing projection private properties\n");
2203 return AVERROR_INVALIDDATA;
2204 } else if (track->video.projection.private.size == 12) {
2205 uint32_t layout = bytestream2_get_be32(&gb);
2207 av_log(NULL, AV_LOG_WARNING,
2208 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2211 projection = AV_SPHERICAL_CUBEMAP;
2212 padding = bytestream2_get_be32(&gb);
2214 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2215 return AVERROR_INVALIDDATA;
2218 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2219 /* No Spherical metadata */
2222 av_log(NULL, AV_LOG_WARNING,
2223 "Unknown spherical metadata type %"PRIu64"\n",
2224 track->video.projection.type);
2228 spherical = av_spherical_alloc(&spherical_size);
2230 return AVERROR(ENOMEM);
2232 spherical->projection = projection;
2234 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2235 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2236 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2238 spherical->padding = padding;
2240 spherical->bound_left = l;
2241 spherical->bound_top = t;
2242 spherical->bound_right = r;
2243 spherical->bound_bottom = b;
2245 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2248 av_freep(&spherical);
2255 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2257 const AVCodecTag *codec_tags;
2259 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2260 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2262 /* Normalize noncompliant private data that starts with the fourcc
2263 * by expanding/shifting the data by 4 bytes and storing the data
2264 * size at the start. */
2265 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2266 int ret = av_buffer_realloc(&track->codec_priv.buf,
2267 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2271 track->codec_priv.data = track->codec_priv.buf->data;
2272 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2273 track->codec_priv.size += 4;
2274 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2277 *fourcc = AV_RL32(track->codec_priv.data + 4);
2278 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2283 static int matroska_parse_tracks(AVFormatContext *s)
2285 MatroskaDemuxContext *matroska = s->priv_data;
2286 MatroskaTrack *tracks = matroska->tracks.elem;
2291 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2292 MatroskaTrack *track = &tracks[i];
2293 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2294 EbmlList *encodings_list = &track->encodings;
2295 MatroskaTrackEncoding *encodings = encodings_list->elem;
2296 uint8_t *extradata = NULL;
2297 int extradata_size = 0;
2298 int extradata_offset = 0;
2299 uint32_t fourcc = 0;
2301 char* key_id_base64 = NULL;
2304 /* Apply some sanity checks. */
2305 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2306 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2307 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2308 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2309 av_log(matroska->ctx, AV_LOG_INFO,
2310 "Unknown or unsupported track type %"PRIu64"\n",
2314 if (!track->codec_id)
2317 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2318 isnan(track->audio.samplerate)) {
2319 av_log(matroska->ctx, AV_LOG_WARNING,
2320 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2321 track->audio.samplerate);
2322 track->audio.samplerate = 8000;
2325 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2326 if (!track->default_duration && track->video.frame_rate > 0) {
2327 double default_duration = 1000000000 / track->video.frame_rate;
2328 if (default_duration > UINT64_MAX || default_duration < 0) {
2329 av_log(matroska->ctx, AV_LOG_WARNING,
2330 "Invalid frame rate %e. Cannot calculate default duration.\n",
2331 track->video.frame_rate);
2333 track->default_duration = default_duration;
2336 if (track->video.display_width == -1)
2337 track->video.display_width = track->video.pixel_width;
2338 if (track->video.display_height == -1)
2339 track->video.display_height = track->video.pixel_height;
2340 if (track->video.color_space.size == 4)
2341 fourcc = AV_RL32(track->video.color_space.data);
2342 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2343 if (!track->audio.out_samplerate)
2344 track->audio.out_samplerate = track->audio.samplerate;
2346 if (encodings_list->nb_elem > 1) {
2347 av_log(matroska->ctx, AV_LOG_ERROR,
2348 "Multiple combined encodings not supported");
2349 } else if (encodings_list->nb_elem == 1) {
2350 if (encodings[0].type) {
2351 if (encodings[0].encryption.key_id.size > 0) {
2352 /* Save the encryption key id to be stored later as a
2354 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2355 key_id_base64 = av_malloc(b64_size);
2356 if (key_id_base64 == NULL)
2357 return AVERROR(ENOMEM);
2359 av_base64_encode(key_id_base64, b64_size,
2360 encodings[0].encryption.key_id.data,
2361 encodings[0].encryption.key_id.size);
2363 encodings[0].scope = 0;
2364 av_log(matroska->ctx, AV_LOG_ERROR,
2365 "Unsupported encoding type");
2369 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2372 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2375 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2377 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2378 encodings[0].scope = 0;
2379 av_log(matroska->ctx, AV_LOG_ERROR,
2380 "Unsupported encoding type");
2381 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2382 uint8_t *codec_priv = track->codec_priv.data;
2383 int ret = matroska_decode_buffer(&track->codec_priv.data,
2384 &track->codec_priv.size,
2387 track->codec_priv.data = NULL;
2388 track->codec_priv.size = 0;
2389 av_log(matroska->ctx, AV_LOG_ERROR,
2390 "Failed to decode codec private data\n");
2393 if (codec_priv != track->codec_priv.data) {
2394 av_buffer_unref(&track->codec_priv.buf);
2395 if (track->codec_priv.data) {
2396 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2397 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2399 if (!track->codec_priv.buf) {
2400 av_freep(&track->codec_priv.data);
2401 track->codec_priv.size = 0;
2402 return AVERROR(ENOMEM);
2409 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2410 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2411 strlen(ff_mkv_codec_tags[j].str))) {
2412 codec_id = ff_mkv_codec_tags[j].id;
2417 st = track->stream = avformat_new_stream(s, NULL);
2419 av_free(key_id_base64);
2420 return AVERROR(ENOMEM);
2423 if (key_id_base64) {
2424 /* export encryption key id as base64 metadata tag */
2425 av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2426 AV_DICT_DONT_STRDUP_VAL);
2429 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2430 track->codec_priv.size >= 40 &&
2431 track->codec_priv.data) {
2432 track->ms_compat = 1;
2433 bit_depth = AV_RL16(track->codec_priv.data + 14);
2434 fourcc = AV_RL32(track->codec_priv.data + 16);
2435 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2438 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2440 extradata_offset = 40;
2441 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2442 track->codec_priv.size >= 14 &&
2443 track->codec_priv.data) {
2445 ffio_init_context(&b, track->codec_priv.data,
2446 track->codec_priv.size,
2447 0, NULL, NULL, NULL, NULL);
2448 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2451 codec_id = st->codecpar->codec_id;
2452 fourcc = st->codecpar->codec_tag;
2453 extradata_offset = FFMIN(track->codec_priv.size, 18);
2454 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2455 /* Normally 36, but allow noncompliant private data */
2456 && (track->codec_priv.size >= 32)
2457 && (track->codec_priv.data)) {
2458 uint16_t sample_size;
2459 int ret = get_qt_codec(track, &fourcc, &codec_id);
2462 sample_size = AV_RB16(track->codec_priv.data + 26);
2464 if (sample_size == 8) {
2465 fourcc = MKTAG('r','a','w',' ');
2466 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2467 } else if (sample_size == 16) {
2468 fourcc = MKTAG('t','w','o','s');
2469 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2472 if ((fourcc == MKTAG('t','w','o','s') ||
2473 fourcc == MKTAG('s','o','w','t')) &&
2475 codec_id = AV_CODEC_ID_PCM_S8;
2476 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2477 (track->codec_priv.size >= 21) &&
2478 (track->codec_priv.data)) {
2479 int ret = get_qt_codec(track, &fourcc, &codec_id);
2482 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2483 fourcc = MKTAG('S','V','Q','3');
2484 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2486 if (codec_id == AV_CODEC_ID_NONE)
2487 av_log(matroska->ctx, AV_LOG_ERROR,
2488 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2489 if (track->codec_priv.size >= 86) {
2490 bit_depth = AV_RB16(track->codec_priv.data + 82);
2491 ffio_init_context(&b, track->codec_priv.data,
2492 track->codec_priv.size,
2493 0, NULL, NULL, NULL, NULL);
2494 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2496 track->has_palette = 1;
2499 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2500 switch (track->audio.bitdepth) {
2502 codec_id = AV_CODEC_ID_PCM_U8;
2505 codec_id = AV_CODEC_ID_PCM_S24BE;
2508 codec_id = AV_CODEC_ID_PCM_S32BE;
2511 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2512 switch (track->audio.bitdepth) {
2514 codec_id = AV_CODEC_ID_PCM_U8;
2517 codec_id = AV_CODEC_ID_PCM_S24LE;
2520 codec_id = AV_CODEC_ID_PCM_S32LE;
2523 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2524 track->audio.bitdepth == 64) {
2525 codec_id = AV_CODEC_ID_PCM_F64LE;
2526 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2527 int profile = matroska_aac_profile(track->codec_id);
2528 int sri = matroska_aac_sri(track->audio.samplerate);
2529 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2531 return AVERROR(ENOMEM);
2532 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2533 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2534 if (strstr(track->codec_id, "SBR")) {
2535 sri = matroska_aac_sri(track->audio.out_samplerate);
2536 extradata[2] = 0x56;
2537 extradata[3] = 0xE5;
2538 extradata[4] = 0x80 | (sri << 3);
2542 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2543 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2544 * Create the "atom size", "tag", and "tag version" fields the
2545 * decoder expects manually. */
2546 extradata_size = 12 + track->codec_priv.size;
2547 extradata = av_mallocz(extradata_size +
2548 AV_INPUT_BUFFER_PADDING_SIZE);
2550 return AVERROR(ENOMEM);
2551 AV_WB32(extradata, extradata_size);
2552 memcpy(&extradata[4], "alac", 4);
2553 AV_WB32(&extradata[8], 0);
2554 memcpy(&extradata[12], track->codec_priv.data,
2555 track->codec_priv.size);
2556 } else if (codec_id == AV_CODEC_ID_TTA) {
2558 if (track->audio.channels > UINT16_MAX ||
2559 track->audio.bitdepth > UINT16_MAX) {
2560 av_log(matroska->ctx, AV_LOG_WARNING,
2561 "Too large audio channel number %"PRIu64
2562 " or bitdepth %"PRIu64". Skipping track.\n",
2563 track->audio.channels, track->audio.bitdepth);
2564 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2565 return AVERROR_INVALIDDATA;
2569 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2570 return AVERROR_INVALIDDATA;
2571 extradata_size = 22;
2572 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2574 return AVERROR(ENOMEM);
2576 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2577 bytestream_put_le16(&ptr, 1);
2578 bytestream_put_le16(&ptr, track->audio.channels);
2579 bytestream_put_le16(&ptr, track->audio.bitdepth);
2580 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2581 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2582 track->audio.out_samplerate,
2583 AV_TIME_BASE * 1000));
2584 } else if (codec_id == AV_CODEC_ID_RV10 ||
2585 codec_id == AV_CODEC_ID_RV20 ||
2586 codec_id == AV_CODEC_ID_RV30 ||
2587 codec_id == AV_CODEC_ID_RV40) {
2588 extradata_offset = 26;
2589 } else if (codec_id == AV_CODEC_ID_RA_144) {
2590 track->audio.out_samplerate = 8000;
2591 track->audio.channels = 1;
2592 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2593 codec_id == AV_CODEC_ID_COOK ||
2594 codec_id == AV_CODEC_ID_ATRAC3 ||
2595 codec_id == AV_CODEC_ID_SIPR)
2596 && track->codec_priv.data) {
2599 ffio_init_context(&b, track->codec_priv.data,
2600 track->codec_priv.size,
2601 0, NULL, NULL, NULL, NULL);
2603 flavor = avio_rb16(&b);
2604 track->audio.coded_framesize = avio_rb32(&b);
2606 track->audio.sub_packet_h = avio_rb16(&b);
2607 track->audio.frame_size = avio_rb16(&b);
2608 track->audio.sub_packet_size = avio_rb16(&b);
2609 if (track->audio.coded_framesize <= 0 ||
2610 track->audio.sub_packet_h <= 0 ||
2611 track->audio.frame_size <= 0)
2612 return AVERROR_INVALIDDATA;
2614 if (codec_id == AV_CODEC_ID_RA_288) {
2615 if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2616 != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2617 return AVERROR_INVALIDDATA;
2618 st->codecpar->block_align = track->audio.coded_framesize;
2619 track->codec_priv.size = 0;
2621 if (codec_id == AV_CODEC_ID_SIPR) {
2622 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2624 return AVERROR_INVALIDDATA;
2625 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2626 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2627 } else if (track->audio.sub_packet_size <= 0 ||
2628 track->audio.frame_size % track->audio.sub_packet_size)
2629 return AVERROR_INVALIDDATA;
2630 st->codecpar->block_align = track->audio.sub_packet_size;
2631 extradata_offset = 78;
2633 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2634 track->audio.frame_size);
2635 if (!track->audio.buf)
2636 return AVERROR(ENOMEM);
2637 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2638 ret = matroska_parse_flac(s, track, &extradata_offset);
2641 } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2642 av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2643 "in absence of valid CodecPrivate.\n");
2645 extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2647 return AVERROR(ENOMEM);
2648 AV_WL16(extradata, 0x410);
2649 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2650 fourcc = AV_RL32(track->codec_priv.data);
2651 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2652 /* we don't need any value stored in CodecPrivate.
2653 make sure that it's not exported as extradata. */
2654 track->codec_priv.size = 0;
2655 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2656 /* For now, propagate only the OBUs, if any. Once libavcodec is
2657 updated to handle isobmff style extradata this can be removed. */
2658 extradata_offset = 4;
2660 track->codec_priv.size -= extradata_offset;
2662 if (codec_id == AV_CODEC_ID_NONE)
2663 av_log(matroska->ctx, AV_LOG_INFO,
2664 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2666 if (track->time_scale < 0.01)
2667 track->time_scale = 1.0;
2668 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2669 1000 * 1000 * 1000); /* 64 bit pts in ns */
2671 /* convert the delay from ns to the track timebase */
2672 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2673 (AVRational){ 1, 1000000000 },
2676 st->codecpar->codec_id = codec_id;
2678 if (strcmp(track->language, "und"))
2679 av_dict_set(&st->metadata, "language", track->language, 0);
2680 av_dict_set(&st->metadata, "title", track->name, 0);
2682 if (track->flag_default)
2683 st->disposition |= AV_DISPOSITION_DEFAULT;
2684 if (track->flag_forced)
2685 st->disposition |= AV_DISPOSITION_FORCED;
2687 if (!st->codecpar->extradata) {
2689 st->codecpar->extradata = extradata;
2690 st->codecpar->extradata_size = extradata_size;
2691 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2692 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2693 return AVERROR(ENOMEM);
2694 memcpy(st->codecpar->extradata,
2695 track->codec_priv.data + extradata_offset,
2696 track->codec_priv.size);
2700 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2701 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2702 int display_width_mul = 1;
2703 int display_height_mul = 1;
2705 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2706 st->codecpar->codec_tag = fourcc;
2708 st->codecpar->bits_per_coded_sample = bit_depth;
2709 st->codecpar->width = track->video.pixel_width;
2710 st->codecpar->height = track->video.pixel_height;
2712 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2713 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2714 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2715 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2717 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2718 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2720 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2721 av_reduce(&st->sample_aspect_ratio.num,
2722 &st->sample_aspect_ratio.den,
2723 st->codecpar->height * track->video.display_width * display_width_mul,
2724 st->codecpar->width * track->video.display_height * display_height_mul,
2727 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2728 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2730 if (track->default_duration) {
2731 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2732 1000000000, track->default_duration, 30000);
2733 #if FF_API_R_FRAME_RATE
2734 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2735 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2736 st->r_frame_rate = st->avg_frame_rate;
2740 /* export stereo mode flag as metadata tag */
2741 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2742 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2744 /* export alpha mode flag as metadata tag */
2745 if (track->video.alpha_mode)
2746 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2748 /* if we have virtual track, mark the real tracks */
2749 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2751 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2753 snprintf(buf, sizeof(buf), "%s_%d",
2754 ff_matroska_video_stereo_plane[planes[j].type], i);
2755 for (k=0; k < matroska->tracks.nb_elem; k++)
2756 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2757 av_dict_set(&tracks[k].stream->metadata,
2758 "stereo_mode", buf, 0);
2762 // add stream level stereo3d side data if it is a supported format
2763 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2764 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2765 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2770 ret = mkv_parse_video_color(st, track);
2773 ret = mkv_parse_video_projection(st, track);
2776 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2777 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2778 st->codecpar->codec_tag = fourcc;
2779 st->codecpar->sample_rate = track->audio.out_samplerate;
2780 st->codecpar->channels = track->audio.channels;
2781 if (!st->codecpar->bits_per_coded_sample)
2782 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2783 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2784 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2785 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2786 st->need_parsing = AVSTREAM_PARSE_FULL;
2787 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2788 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2789 if (track->codec_delay > 0) {
2790 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2791 (AVRational){1, 1000000000},
2792 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2793 48000 : st->codecpar->sample_rate});
2795 if (track->seek_preroll > 0) {
2796 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2797 (AVRational){1, 1000000000},
2798 (AVRational){1, st->codecpar->sample_rate});
2800 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2801 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2803 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2804 st->disposition |= AV_DISPOSITION_CAPTIONS;
2805 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2806 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2807 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2808 st->disposition |= AV_DISPOSITION_METADATA;
2810 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2811 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2818 static int matroska_read_header(AVFormatContext *s)
2820 MatroskaDemuxContext *matroska = s->priv_data;
2821 EbmlList *attachments_list = &matroska->attachments;
2822 EbmlList *chapters_list = &matroska->chapters;
2823 MatroskaAttachment *attachments;
2824 MatroskaChapter *chapters;
2825 uint64_t max_start = 0;
2831 matroska->cues_parsing_deferred = 1;
2833 /* First read the EBML header. */
2834 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2835 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2836 ebml_free(ebml_syntax, &ebml);
2837 return AVERROR_INVALIDDATA;
2839 if (ebml.version > EBML_VERSION ||
2840 ebml.max_size > sizeof(uint64_t) ||
2841 ebml.id_length > sizeof(uint32_t) ||
2842 ebml.doctype_version > 3) {
2843 avpriv_report_missing_feature(matroska->ctx,
2844 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2845 ebml.version, ebml.doctype, ebml.doctype_version);
2846 ebml_free(ebml_syntax, &ebml);
2847 return AVERROR_PATCHWELCOME;
2848 } else if (ebml.doctype_version == 3) {
2849 av_log(matroska->ctx, AV_LOG_WARNING,
2850 "EBML header using unsupported features\n"
2851 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2852 ebml.version, ebml.doctype, ebml.doctype_version);
2854 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2855 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2857 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2858 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2859 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2860 ebml_free(ebml_syntax, &ebml);
2861 return AVERROR_INVALIDDATA;
2864 ebml_free(ebml_syntax, &ebml);
2866 /* The next thing is a segment. */
2867 pos = avio_tell(matroska->ctx->pb);
2868 res = ebml_parse(matroska, matroska_segments, matroska);
2869 // Try resyncing until we find an EBML_STOP type element.
2871 res = matroska_resync(matroska, pos);
2874 pos = avio_tell(matroska->ctx->pb);
2875 res = ebml_parse(matroska, matroska_segment, matroska);
2877 /* Set data_offset as it might be needed later by seek_frame_generic. */
2878 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2879 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2880 matroska_execute_seekhead(matroska);
2882 if (!matroska->time_scale)
2883 matroska->time_scale = 1000000;
2884 if (matroska->duration)
2885 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2886 1000 / AV_TIME_BASE;
2887 av_dict_set(&s->metadata, "title", matroska->title, 0);
2888 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2890 if (matroska->date_utc.size == 8)
2891 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2893 res = matroska_parse_tracks(s);
2897 attachments = attachments_list->elem;
2898 for (j = 0; j < attachments_list->nb_elem; j++) {
2899 if (!(attachments[j].filename && attachments[j].mime &&
2900 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2901 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2903 AVStream *st = avformat_new_stream(s, NULL);
2906 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2907 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2908 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2910 for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2911 if (!strncmp(mkv_image_mime_tags[i].str, attachments[j].mime,
2912 strlen(mkv_image_mime_tags[i].str))) {
2913 st->codecpar->codec_id = mkv_image_mime_tags[i].id;
2918 attachments[j].stream = st;
2920 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2921 AVPacket *pkt = &st->attached_pic;
2923 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2924 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2926 av_init_packet(pkt);
2927 pkt->buf = av_buffer_ref(attachments[j].bin.buf);
2929 return AVERROR(ENOMEM);
2930 pkt->data = attachments[j].bin.data;
2931 pkt->size = attachments[j].bin.size;
2932 pkt->stream_index = st->index;
2933 pkt->flags |= AV_PKT_FLAG_KEY;
2935 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2936 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2938 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2939 attachments[j].bin.size);
2941 for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2942 if (!strncmp(mkv_mime_tags[i].str, attachments[j].mime,
2943 strlen(mkv_mime_tags[i].str))) {
2944 st->codecpar->codec_id = mkv_mime_tags[i].id;
2952 chapters = chapters_list->elem;
2953 for (i = 0; i < chapters_list->nb_elem; i++)
2954 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2955 (max_start == 0 || chapters[i].start > max_start)) {
2956 chapters[i].chapter =
2957 avpriv_new_chapter(s, chapters[i].uid,
2958 (AVRational) { 1, 1000000000 },
2959 chapters[i].start, chapters[i].end,
2961 max_start = chapters[i].start;
2964 matroska_add_index_entries(matroska);
2966 matroska_convert_tags(s);
2970 matroska_read_close(s);
2975 * Put one packet in an application-supplied AVPacket struct.
2976 * Returns 0 on success or -1 on failure.
2978 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2981 if (matroska->queue) {
2982 MatroskaTrack *tracks = matroska->tracks.elem;
2983 MatroskaTrack *track;
2985 ff_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
2986 track = &tracks[pkt->stream_index];
2987 if (track->has_palette) {
2988 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2990 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
2992 memcpy(pal, track->palette, AVPALETTE_SIZE);
2994 track->has_palette = 0;
3003 * Free all packets in our internal queue.
3005 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3007 ff_packet_list_free(&matroska->queue, &matroska->queue_end);
3010 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3011 int size, int type, AVIOContext *pb,
3012 uint32_t lace_size[256], int *laces)
3015 uint8_t *data = *buf;
3019 lace_size[0] = size;
3024 return AVERROR_INVALIDDATA;
3031 case 0x1: /* Xiph lacing */
3035 for (n = 0; n < *laces - 1; n++) {
3039 if (size <= total) {
3040 return AVERROR_INVALIDDATA;
3044 lace_size[n] += temp;
3052 return AVERROR_INVALIDDATA;
3055 lace_size[n] = size - total;
3059 case 0x2: /* fixed-size lacing */
3060 if (size % (*laces)) {
3061 return AVERROR_INVALIDDATA;
3063 for (n = 0; n < *laces; n++)
3064 lace_size[n] = size / *laces;
3067 case 0x3: /* EBML lacing */
3075 n = ebml_read_num(matroska, pb, 8, &num, 1);
3079 return AVERROR_INVALIDDATA;
3081 total = lace_size[0] = num;
3083 for (n = 1; n < *laces - 1; n++) {
3086 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3089 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3090 return AVERROR_INVALIDDATA;
3092 lace_size[n] = lace_size[n - 1] + snum;
3093 total += lace_size[n];
3099 return AVERROR_INVALIDDATA;
3101 lace_size[*laces - 1] = size - total;
3111 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3112 MatroskaTrack *track, AVStream *st,
3113 uint8_t *data, int size, uint64_t timecode,
3116 int a = st->codecpar->block_align;
3117 int sps = track->audio.sub_packet_size;
3118 int cfs = track->audio.coded_framesize;
3119 int h = track->audio.sub_packet_h;
3120 int y = track->audio.sub_packet_cnt;
3121 int w = track->audio.frame_size;
3124 if (!track->audio.pkt_cnt) {
3125 if (track->audio.sub_packet_cnt == 0)
3126 track->audio.buf_timecode = timecode;
3127 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3128 if (size < cfs * h / 2) {
3129 av_log(matroska->ctx, AV_LOG_ERROR,
3130 "Corrupt int4 RM-style audio packet size\n");
3131 return AVERROR_INVALIDDATA;
3133 for (x = 0; x < h / 2; x++)
3134 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3135 data + x * cfs, cfs);
3136 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3138 av_log(matroska->ctx, AV_LOG_ERROR,
3139 "Corrupt sipr RM-style audio packet size\n");
3140 return AVERROR_INVALIDDATA;
3142 memcpy(track->audio.buf + y * w, data, w);
3145 av_log(matroska->ctx, AV_LOG_ERROR,
3146 "Corrupt generic RM-style audio packet size\n");
3147 return AVERROR_INVALIDDATA;
3149 for (x = 0; x < w / sps; x++)
3150 memcpy(track->audio.buf +
3151 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3152 data + x * sps, sps);
3155 if (++track->audio.sub_packet_cnt >= h) {
3156 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3157 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3158 track->audio.sub_packet_cnt = 0;
3159 track->audio.pkt_cnt = h * w / a;
3163 while (track->audio.pkt_cnt) {
3165 AVPacket pktl, *pkt = &pktl;
3167 ret = av_new_packet(pkt, a);
3172 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3174 pkt->pts = track->audio.buf_timecode;
3175 track->audio.buf_timecode = AV_NOPTS_VALUE;
3177 pkt->stream_index = st->index;
3178 ret = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3180 av_packet_unref(pkt);
3181 return AVERROR(ENOMEM);
3188 /* reconstruct full wavpack blocks from mangled matroska ones */
3189 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
3190 uint8_t **pdst, int *size)
3192 uint8_t *dst = NULL;
3197 int ret, offset = 0;
3200 return AVERROR_INVALIDDATA;
3202 av_assert1(track->stream->codecpar->extradata_size >= 2);
3203 ver = AV_RL16(track->stream->codecpar->extradata);
3205 samples = AV_RL32(src);
3209 while (srclen >= 8) {
3214 uint32_t flags = AV_RL32(src);
3215 uint32_t crc = AV_RL32(src + 4);
3219 multiblock = (flags & 0x1800) != 0x1800;
3222 ret = AVERROR_INVALIDDATA;
3225 blocksize = AV_RL32(src);
3231 if (blocksize > srclen) {
3232 ret = AVERROR_INVALIDDATA;
3236 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3238 ret = AVERROR(ENOMEM);
3242 dstlen += blocksize + 32;
3244 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3245 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3246 AV_WL16(dst + offset + 8, ver); // version
3247 AV_WL16(dst + offset + 10, 0); // track/index_no
3248 AV_WL32(dst + offset + 12, 0); // total samples
3249 AV_WL32(dst + offset + 16, 0); // block index
3250 AV_WL32(dst + offset + 20, samples); // number of samples
3251 AV_WL32(dst + offset + 24, flags); // flags
3252 AV_WL32(dst + offset + 28, crc); // crc
3253 memcpy(dst + offset + 32, src, blocksize); // block data
3256 srclen -= blocksize;
3257 offset += blocksize + 32;
3260 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3272 static int matroska_parse_prores(MatroskaTrack *track, uint8_t *src,
3273 uint8_t **pdst, int *size)
3276 int dstlen = *size + 8;
3278 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3280 return AVERROR(ENOMEM);
3282 AV_WB32(dst, dstlen);
3283 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3284 memcpy(dst + 8, src, dstlen - 8);
3285 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3293 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3294 MatroskaTrack *track,
3296 uint8_t *data, int data_len,
3301 AVPacket pktl, *pkt = &pktl;
3302 uint8_t *id, *settings, *text, *buf;
3303 int id_len, settings_len, text_len;
3308 return AVERROR_INVALIDDATA;
3311 q = data + data_len;
3316 if (*p == '\r' || *p == '\n') {
3325 if (p >= q || *p != '\n')
3326 return AVERROR_INVALIDDATA;
3332 if (*p == '\r' || *p == '\n') {
3333 settings_len = p - settings;
3341 if (p >= q || *p != '\n')
3342 return AVERROR_INVALIDDATA;
3347 while (text_len > 0) {
3348 const int len = text_len - 1;
3349 const uint8_t c = p[len];
3350 if (c != '\r' && c != '\n')
3356 return AVERROR_INVALIDDATA;
3358 err = av_new_packet(pkt, text_len);
3363 memcpy(pkt->data, text, text_len);
3366 buf = av_packet_new_side_data(pkt,
3367 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3370 av_packet_unref(pkt);
3371 return AVERROR(ENOMEM);
3373 memcpy(buf, id, id_len);
3376 if (settings_len > 0) {
3377 buf = av_packet_new_side_data(pkt,
3378 AV_PKT_DATA_WEBVTT_SETTINGS,
3381 av_packet_unref(pkt);
3382 return AVERROR(ENOMEM);
3384 memcpy(buf, settings, settings_len);
3387 // Do we need this for subtitles?
3388 // pkt->flags = AV_PKT_FLAG_KEY;
3390 pkt->stream_index = st->index;
3391 pkt->pts = timecode;
3393 // Do we need this for subtitles?
3394 // pkt->dts = timecode;
3396 pkt->duration = duration;
3399 err = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3401 av_packet_unref(pkt);
3402 return AVERROR(ENOMEM);
3408 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3409 MatroskaTrack *track, AVStream *st,
3410 AVBufferRef *buf, uint8_t *data, int pkt_size,
3411 uint64_t timecode, uint64_t lace_duration,
3412 int64_t pos, int is_keyframe,
3413 uint8_t *additional, uint64_t additional_id, int additional_size,
3414 int64_t discard_padding)
3416 MatroskaTrackEncoding *encodings = track->encodings.elem;
3417 uint8_t *pkt_data = data;
3419 AVPacket pktl, *pkt = &pktl;
3421 if (encodings && !encodings->type && encodings->scope & 1) {
3422 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
3427 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3429 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
3431 av_log(matroska->ctx, AV_LOG_ERROR,
3432 "Error parsing a wavpack block.\n");
3435 if (pkt_data != data)
3436 av_freep(&pkt_data);
3440 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3441 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3443 res = matroska_parse_prores(track, pkt_data, &pr_data, &pkt_size);
3445 av_log(matroska->ctx, AV_LOG_ERROR,
3446 "Error parsing a prores block.\n");
3449 if (pkt_data != data)
3450 av_freep(&pkt_data);
3454 if (!pkt_size && !additional_size)
3457 av_init_packet(pkt);
3458 if (pkt_data != data)
3459 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3462 pkt->buf = av_buffer_ref(buf);
3465 res = AVERROR(ENOMEM);
3469 pkt->data = pkt_data;
3470 pkt->size = pkt_size;
3471 pkt->flags = is_keyframe;
3472 pkt->stream_index = st->index;
3474 if (additional_size > 0) {
3475 uint8_t *side_data = av_packet_new_side_data(pkt,
3476 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3477 additional_size + 8);
3479 av_packet_unref(pkt);
3480 return AVERROR(ENOMEM);
3482 AV_WB64(side_data, additional_id);
3483 memcpy(side_data + 8, additional, additional_size);
3486 if (discard_padding) {
3487 uint8_t *side_data = av_packet_new_side_data(pkt,
3488 AV_PKT_DATA_SKIP_SAMPLES,
3491 av_packet_unref(pkt);
3492 return AVERROR(ENOMEM);
3494 discard_padding = av_rescale_q(discard_padding,
3495 (AVRational){1, 1000000000},
3496 (AVRational){1, st->codecpar->sample_rate});
3497 if (discard_padding > 0) {
3498 AV_WL32(side_data + 4, discard_padding);
3500 AV_WL32(side_data, -discard_padding);
3504 if (track->ms_compat)
3505 pkt->dts = timecode;
3507 pkt->pts = timecode;
3509 pkt->duration = lace_duration;
3511 #if FF_API_CONVERGENCE_DURATION
3512 FF_DISABLE_DEPRECATION_WARNINGS
3513 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3514 pkt->convergence_duration = lace_duration;
3516 FF_ENABLE_DEPRECATION_WARNINGS
3519 res = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3521 av_packet_unref(pkt);
3522 return AVERROR(ENOMEM);
3529 if (pkt_data != data)
3530 av_freep(&pkt_data);
3534 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3535 int size, int64_t pos, uint64_t cluster_time,
3536 uint64_t block_duration, int is_keyframe,
3537 uint8_t *additional, uint64_t additional_id, int additional_size,
3538 int64_t cluster_pos, int64_t discard_padding)
3540 uint64_t timecode = AV_NOPTS_VALUE;
3541 MatroskaTrack *track;
3546 uint32_t lace_size[256];
3547 int n, flags, laces = 0;
3549 int trust_default_duration = 1;
3551 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3553 if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3558 track = matroska_find_track_by_num(matroska, num);
3559 if (!track || !track->stream) {
3560 av_log(matroska->ctx, AV_LOG_INFO,
3561 "Invalid stream %"PRIu64"\n", num);
3562 return AVERROR_INVALIDDATA;
3563 } else if (size < 3)
3564 return AVERROR_INVALIDDATA;
3566 if (st->discard >= AVDISCARD_ALL)
3568 av_assert1(block_duration != AV_NOPTS_VALUE);
3570 block_time = sign_extend(AV_RB16(data), 16);
3574 if (is_keyframe == -1)
3575 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3577 if (cluster_time != (uint64_t) -1 &&
3578 (block_time >= 0 || cluster_time >= -block_time)) {
3579 timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3580 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3581 timecode < track->end_timecode)
3582 is_keyframe = 0; /* overlapping subtitles are not key frame */
3584 ff_reduce_index(matroska->ctx, st->index);
3585 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3590 if (matroska->skip_to_keyframe &&
3591 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3592 // Compare signed timecodes. Timecode may be negative due to codec delay
3593 // offset. We don't support timestamps greater than int64_t anyway - see
3595 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3598 matroska->skip_to_keyframe = 0;
3599 else if (!st->skip_to_keyframe) {
3600 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3601 matroska->skip_to_keyframe = 0;
3605 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3606 &pb, lace_size, &laces);
3608 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3612 if (track->audio.samplerate == 8000) {
3613 // If this is needed for more codecs, then add them here
3614 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3615 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3616 trust_default_duration = 0;
3620 if (!block_duration && trust_default_duration)
3621 block_duration = track->default_duration * laces / matroska->time_scale;
3623 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3624 track->end_timecode =
3625 FFMAX(track->end_timecode, timecode + block_duration);
3627 for (n = 0; n < laces; n++) {
3628 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3630 if (track->audio.buf) {
3631 res = matroska_parse_rm_audio(matroska, track, st, data,
3637 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3638 res = matroska_parse_webvtt(matroska, track, st,
3640 timecode, lace_duration,
3645 res = matroska_parse_frame(matroska, track, st, buf, data, lace_size[n],
3646 timecode, lace_duration, pos,
3647 !n ? is_keyframe : 0,
3648 additional, additional_id, additional_size,
3654 if (timecode != AV_NOPTS_VALUE)
3655 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3656 data += lace_size[n];
3662 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3664 MatroskaCluster *cluster = &matroska->current_cluster;
3665 MatroskaBlock *block = &cluster->block;
3668 av_assert0(matroska->num_levels <= 2);
3670 if (matroska->num_levels == 1) {
3671 res = ebml_parse(matroska, matroska_segment, NULL);
3674 /* Found a cluster: subtract the size of the ID already read. */
3675 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3677 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3683 if (matroska->num_levels == 2) {
3684 /* We are inside a cluster. */
3685 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3687 if (res >= 0 && block->bin.size > 0) {
3688 int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3689 uint8_t* additional = block->additional.size > 0 ?
3690 block->additional.data : NULL;
3692 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3693 block->bin.size, block->bin.pos,
3694 cluster->timecode, block->duration,
3695 is_keyframe, additional, block->additional_id,
3696 block->additional.size, cluster->pos,
3697 block->discard_padding);
3700 ebml_free(matroska_blockgroup, block);
3701 memset(block, 0, sizeof(*block));
3702 } else if (!matroska->num_levels) {
3703 if (!avio_feof(matroska->ctx->pb)) {
3704 avio_r8(matroska->ctx->pb);
3705 if (!avio_feof(matroska->ctx->pb)) {
3706 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3707 "end of segment.\n");
3708 return AVERROR_INVALIDDATA;
3718 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3720 MatroskaDemuxContext *matroska = s->priv_data;
3723 if (matroska->resync_pos == -1) {
3724 // This can only happen if generic seeking has been used.
3725 matroska->resync_pos = avio_tell(s->pb);
3728 while (matroska_deliver_packet(matroska, pkt)) {
3730 return (ret < 0) ? ret : AVERROR_EOF;
3731 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3732 ret = matroska_resync(matroska, matroska->resync_pos);
3738 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3739 int64_t timestamp, int flags)
3741 MatroskaDemuxContext *matroska = s->priv_data;
3742 MatroskaTrack *tracks = NULL;
3743 AVStream *st = s->streams[stream_index];
3746 /* Parse the CUES now since we need the index data to seek. */
3747 if (matroska->cues_parsing_deferred > 0) {
3748 matroska->cues_parsing_deferred = 0;
3749 matroska_parse_cues(matroska);
3752 if (!st->nb_index_entries)
3754 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3756 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3757 matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3758 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3759 matroska_clear_queue(matroska);
3760 if (matroska_parse_cluster(matroska) < 0)
3765 matroska_clear_queue(matroska);
3766 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3769 tracks = matroska->tracks.elem;
3770 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3771 tracks[i].audio.pkt_cnt = 0;
3772 tracks[i].audio.sub_packet_cnt = 0;
3773 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3774 tracks[i].end_timecode = 0;
3777 /* We seek to a level 1 element, so set the appropriate status. */
3778 matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3779 if (flags & AVSEEK_FLAG_ANY) {
3780 st->skip_to_keyframe = 0;
3781 matroska->skip_to_timecode = timestamp;
3783 st->skip_to_keyframe = 1;
3784 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3786 matroska->skip_to_keyframe = 1;
3788 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3791 // slightly hackish but allows proper fallback to
3792 // the generic seeking code.
3793 matroska_reset_status(matroska, 0, -1);
3794 matroska->resync_pos = -1;
3795 matroska_clear_queue(matroska);
3796 st->skip_to_keyframe =
3797 matroska->skip_to_keyframe = 0;
3802 static int matroska_read_close(AVFormatContext *s)
3804 MatroskaDemuxContext *matroska = s->priv_data;
3805 MatroskaTrack *tracks = matroska->tracks.elem;
3808 matroska_clear_queue(matroska);
3810 for (n = 0; n < matroska->tracks.nb_elem; n++)
3811 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3812 av_freep(&tracks[n].audio.buf);
3813 ebml_free(matroska_segment, matroska);
3819 int64_t start_time_ns;
3820 int64_t end_time_ns;
3821 int64_t start_offset;
3825 /* This function searches all the Cues and returns the CueDesc corresponding to
3826 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3827 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3829 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3830 MatroskaDemuxContext *matroska = s->priv_data;
3833 int nb_index_entries = s->streams[0]->nb_index_entries;
3834 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3835 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3836 for (i = 1; i < nb_index_entries; i++) {
3837 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3838 index_entries[i].timestamp * matroska->time_scale > ts) {
3843 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3844 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3845 if (i != nb_index_entries - 1) {
3846 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3847 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3849 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3850 // FIXME: this needs special handling for files where Cues appear
3851 // before Clusters. the current logic assumes Cues appear after
3853 cue_desc.end_offset = cues_start - matroska->segment_start;
3858 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3860 MatroskaDemuxContext *matroska = s->priv_data;
3861 uint32_t id = matroska->current_id;
3862 int64_t cluster_pos, before_pos;
3864 if (s->streams[0]->nb_index_entries <= 0) return 0;
3865 // seek to the first cluster using cues.
3866 index = av_index_search_timestamp(s->streams[0], 0, 0);
3867 if (index < 0) return 0;
3868 cluster_pos = s->streams[0]->index_entries[index].pos;
3869 before_pos = avio_tell(s->pb);
3871 uint64_t cluster_id, cluster_length;
3874 avio_seek(s->pb, cluster_pos, SEEK_SET);
3875 // read cluster id and length
3876 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3877 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3879 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3883 matroska_reset_status(matroska, 0, cluster_pos);
3884 matroska_clear_queue(matroska);
3885 if (matroska_parse_cluster(matroska) < 0 ||
3889 pkt = &matroska->queue->pkt;
3890 // 4 + read is the length of the cluster id and the cluster length field.
3891 cluster_pos += 4 + read + cluster_length;
3892 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3898 /* Restore the status after matroska_read_header: */
3899 matroska_reset_status(matroska, id, before_pos);
3904 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3905 double min_buffer, double* buffer,
3906 double* sec_to_download, AVFormatContext *s,
3909 double nano_seconds_per_second = 1000000000.0;
3910 double time_sec = time_ns / nano_seconds_per_second;
3912 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3913 int64_t end_time_ns = time_ns + time_to_search_ns;
3914 double sec_downloaded = 0.0;
3915 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3916 if (desc_curr.start_time_ns == -1)
3918 *sec_to_download = 0.0;
3920 // Check for non cue start time.
3921 if (time_ns > desc_curr.start_time_ns) {
3922 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3923 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3924 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3925 double timeToDownload = (cueBytes * 8.0) / bps;
3927 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3928 *sec_to_download += timeToDownload;
3930 // Check if the search ends within the first cue.
3931 if (desc_curr.end_time_ns >= end_time_ns) {
3932 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3933 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3934 sec_downloaded = percent_to_sub * sec_downloaded;
3935 *sec_to_download = percent_to_sub * *sec_to_download;
3938 if ((sec_downloaded + *buffer) <= min_buffer) {
3942 // Get the next Cue.
3943 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3946 while (desc_curr.start_time_ns != -1) {
3947 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3948 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3949 double desc_sec = desc_ns / nano_seconds_per_second;
3950 double bits = (desc_bytes * 8.0);
3951 double time_to_download = bits / bps;
3953 sec_downloaded += desc_sec - time_to_download;
3954 *sec_to_download += time_to_download;
3956 if (desc_curr.end_time_ns >= end_time_ns) {
3957 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3958 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3959 sec_downloaded = percent_to_sub * sec_downloaded;
3960 *sec_to_download = percent_to_sub * *sec_to_download;
3962 if ((sec_downloaded + *buffer) <= min_buffer)
3967 if ((sec_downloaded + *buffer) <= min_buffer) {
3972 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3974 *buffer = *buffer + sec_downloaded;
3978 /* This function computes the bandwidth of the WebM file with the help of
3979 * buffer_size_after_time_downloaded() function. Both of these functions are
3980 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3981 * Matroska parsing mechanism.
3983 * Returns the bandwidth of the file on success; -1 on error.
3985 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3987 MatroskaDemuxContext *matroska = s->priv_data;
3988 AVStream *st = s->streams[0];
3989 double bandwidth = 0.0;
3992 for (i = 0; i < st->nb_index_entries; i++) {
3993 int64_t prebuffer_ns = 1000000000;
3994 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3995 double nano_seconds_per_second = 1000000000.0;
3996 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3997 double prebuffer_bytes = 0.0;
3998 int64_t temp_prebuffer_ns = prebuffer_ns;
3999 int64_t pre_bytes, pre_ns;
4000 double pre_sec, prebuffer, bits_per_second;
4001 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4003 // Start with the first Cue.
4004 CueDesc desc_end = desc_beg;
4006 // Figure out how much data we have downloaded for the prebuffer. This will
4007 // be used later to adjust the bits per sample to try.
4008 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4009 // Prebuffered the entire Cue.
4010 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4011 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4012 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4014 if (desc_end.start_time_ns == -1) {
4015 // The prebuffer is larger than the duration.
4016 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4018 bits_per_second = 0.0;
4020 // The prebuffer ends in the last Cue. Estimate how much data was
4022 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4023 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4024 pre_sec = pre_ns / nano_seconds_per_second;
4026 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4028 prebuffer = prebuffer_ns / nano_seconds_per_second;
4030 // Set this to 0.0 in case our prebuffer buffers the entire video.
4031 bits_per_second = 0.0;
4033 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4034 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4035 double desc_sec = desc_ns / nano_seconds_per_second;
4036 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4038 // Drop the bps by the percentage of bytes buffered.
4039 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4040 double mod_bits_per_second = calc_bits_per_second * percent;
4042 if (prebuffer < desc_sec) {
4044 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4046 // Add 1 so the bits per second should be a little bit greater than file
4048 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4049 const double min_buffer = 0.0;
4050 double buffer = prebuffer;
4051 double sec_to_download = 0.0;
4053 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4054 min_buffer, &buffer, &sec_to_download,
4058 } else if (rv == 0) {
4059 bits_per_second = (double)(bps);
4064 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4065 } while (desc_end.start_time_ns != -1);
4067 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4069 return (int64_t)bandwidth;
4072 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4074 MatroskaDemuxContext *matroska = s->priv_data;
4075 EbmlList *seekhead_list = &matroska->seekhead;
4076 MatroskaSeekhead *seekhead = seekhead_list->elem;
4078 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4082 // determine cues start and end positions
4083 for (i = 0; i < seekhead_list->nb_elem; i++)
4084 if (seekhead[i].id == MATROSKA_ID_CUES)
4087 if (i >= seekhead_list->nb_elem) return -1;
4089 before_pos = avio_tell(matroska->ctx->pb);
4090 cues_start = seekhead[i].pos + matroska->segment_start;
4091 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4092 // cues_end is computed as cues_start + cues_length + length of the
4093 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4094 // cues_end is inclusive and the above sum is reduced by 1.
4095 uint64_t cues_length, cues_id;
4097 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4098 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4099 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4100 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4103 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4105 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4106 if (cues_start == -1 || cues_end == -1) return -1;
4109 matroska_parse_cues(matroska);
4112 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4115 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4117 // if the file has cues at the start, fix up the init range so that
4118 // it does not include it
4119 if (cues_start <= init_range)
4120 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4123 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4124 if (bandwidth < 0) return -1;
4125 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4127 // check if all clusters start with key frames
4128 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4130 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4131 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4132 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4133 if (!buf) return -1;
4135 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4136 int ret = snprintf(buf + end, 20,
4137 "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4138 i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4139 if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4140 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4142 return AVERROR_INVALIDDATA;
4146 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4147 buf, AV_DICT_DONT_STRDUP_VAL);
4152 static int webm_dash_manifest_read_header(AVFormatContext *s)
4155 int ret = matroska_read_header(s);
4157 MatroskaTrack *tracks;
4158 MatroskaDemuxContext *matroska = s->priv_data;
4160 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4163 if (!s->nb_streams) {
4164 matroska_read_close(s);
4165 av_log(s, AV_LOG_ERROR, "No streams found\n");
4166 return AVERROR_INVALIDDATA;
4169 if (!matroska->is_live) {
4170 buf = av_asprintf("%g", matroska->duration);
4171 if (!buf) return AVERROR(ENOMEM);
4172 av_dict_set(&s->streams[0]->metadata, DURATION,
4173 buf, AV_DICT_DONT_STRDUP_VAL);
4175 // initialization range
4176 // 5 is the offset of Cluster ID.
4177 init_range = avio_tell(s->pb) - 5;
4178 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4181 // basename of the file
4182 buf = strrchr(s->url, '/');
4183 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4186 tracks = matroska->tracks.elem;
4187 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4189 // parse the cues and populate Cue related fields
4190 if (!matroska->is_live) {
4191 ret = webm_dash_manifest_cues(s, init_range);
4193 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4198 // use the bandwidth from the command line if it was provided
4199 if (matroska->bandwidth > 0) {
4200 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4201 matroska->bandwidth, 0);
4206 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4211 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4212 static const AVOption options[] = {
4213 { "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 },
4214 { "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 },
4218 static const AVClass webm_dash_class = {
4219 .class_name = "WebM DASH Manifest demuxer",
4220 .item_name = av_default_item_name,
4222 .version = LIBAVUTIL_VERSION_INT,
4225 AVInputFormat ff_matroska_demuxer = {
4226 .name = "matroska,webm",
4227 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4228 .extensions = "mkv,mk3d,mka,mks",
4229 .priv_data_size = sizeof(MatroskaDemuxContext),
4230 .read_probe = matroska_probe,
4231 .read_header = matroska_read_header,
4232 .read_packet = matroska_read_packet,
4233 .read_close = matroska_read_close,
4234 .read_seek = matroska_read_seek,
4235 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4238 AVInputFormat ff_webm_dash_manifest_demuxer = {
4239 .name = "webm_dash_manifest",
4240 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4241 .priv_data_size = sizeof(MatroskaDemuxContext),
4242 .read_header = webm_dash_manifest_read_header,
4243 .read_packet = webm_dash_manifest_read_packet,
4244 .read_close = matroska_read_close,
4245 .priv_class = &webm_dash_class,