2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The FFmpeg Project
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * Matroska file demuxer
25 * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28 * @see specs available on the Matroska project page: http://www.matroska.org/
36 #include "libavutil/avstring.h"
37 #include "libavutil/base64.h"
38 #include "libavutil/dict.h"
39 #include "libavutil/intfloat.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/lzo.h"
42 #include "libavutil/mastering_display_metadata.h"
43 #include "libavutil/mathematics.h"
44 #include "libavutil/opt.h"
45 #include "libavutil/time_internal.h"
46 #include "libavutil/spherical.h"
48 #include "libavcodec/bytestream.h"
49 #include "libavcodec/flac.h"
50 #include "libavcodec/mpeg4audio.h"
53 #include "avio_internal.h"
58 /* For ff_codec_get_id(). */
69 #include "qtpalette.h"
71 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
72 #define NEEDS_CHECKING 2 /* Indicates that some error checks
73 * still need to be performed */
74 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
75 * syntax level used for parsing ended. */
76 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
77 * of unkown, potentially damaged data is encountered,
78 * it is considered an error. */
79 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
80 * to this many bytes of unknown data for the
81 * SKIP_THRESHOLD check. */
97 typedef const struct EbmlSyntax {
100 size_t list_elem_size;
107 const struct EbmlSyntax *n;
111 typedef struct EbmlList {
113 unsigned int alloc_elem_size;
117 typedef struct EbmlBin {
124 typedef struct Ebml {
129 uint64_t doctype_version;
132 typedef struct MatroskaTrackCompression {
135 } MatroskaTrackCompression;
137 typedef struct MatroskaTrackEncryption {
140 } MatroskaTrackEncryption;
142 typedef struct MatroskaTrackEncoding {
145 MatroskaTrackCompression compression;
146 MatroskaTrackEncryption encryption;
147 } MatroskaTrackEncoding;
149 typedef struct MatroskaMasteringMeta {
158 double max_luminance;
159 double min_luminance;
160 } MatroskaMasteringMeta;
162 typedef struct MatroskaTrackVideoColor {
163 uint64_t matrix_coefficients;
164 uint64_t bits_per_channel;
165 uint64_t chroma_sub_horz;
166 uint64_t chroma_sub_vert;
167 uint64_t cb_sub_horz;
168 uint64_t cb_sub_vert;
169 uint64_t chroma_siting_horz;
170 uint64_t chroma_siting_vert;
172 uint64_t transfer_characteristics;
176 MatroskaMasteringMeta mastering_meta;
177 } MatroskaTrackVideoColor;
179 typedef struct MatroskaTrackVideoProjection {
185 } MatroskaTrackVideoProjection;
187 typedef struct MatroskaTrackVideo {
189 uint64_t display_width;
190 uint64_t display_height;
191 uint64_t pixel_width;
192 uint64_t pixel_height;
194 uint64_t display_unit;
196 uint64_t field_order;
197 uint64_t stereo_mode;
200 MatroskaTrackVideoProjection projection;
201 } MatroskaTrackVideo;
203 typedef struct MatroskaTrackAudio {
205 double out_samplerate;
209 /* real audio header (extracted from extradata) */
216 uint64_t buf_timecode;
218 } MatroskaTrackAudio;
220 typedef struct MatroskaTrackPlane {
223 } MatroskaTrackPlane;
225 typedef struct MatroskaTrackOperation {
226 EbmlList combine_planes;
227 } MatroskaTrackOperation;
229 typedef struct MatroskaTrack {
238 uint64_t default_duration;
239 uint64_t flag_default;
240 uint64_t flag_forced;
241 uint64_t seek_preroll;
242 MatroskaTrackVideo video;
243 MatroskaTrackAudio audio;
244 MatroskaTrackOperation operation;
246 uint64_t codec_delay;
247 uint64_t codec_delay_in_track_tb;
250 int64_t end_timecode;
252 uint64_t max_block_additional_id;
254 uint32_t palette[AVPALETTE_COUNT];
258 typedef struct MatroskaAttachment {
265 } MatroskaAttachment;
267 typedef struct MatroskaChapter {
276 typedef struct MatroskaIndexPos {
281 typedef struct MatroskaIndex {
286 typedef struct MatroskaTag {
294 typedef struct MatroskaTagTarget {
302 typedef struct MatroskaTags {
303 MatroskaTagTarget target;
307 typedef struct MatroskaSeekhead {
312 typedef struct MatroskaLevel {
317 typedef struct MatroskaBlock {
322 uint64_t additional_id;
324 int64_t discard_padding;
327 typedef struct MatroskaCluster {
333 typedef struct MatroskaLevel1Element {
337 } MatroskaLevel1Element;
339 typedef struct MatroskaDemuxContext {
340 const AVClass *class;
341 AVFormatContext *ctx;
344 MatroskaLevel levels[EBML_MAX_DEPTH];
356 EbmlList attachments;
362 /* byte position of the segment inside the stream */
363 int64_t segment_start;
365 /* the packet queue */
367 AVPacketList *queue_end;
371 /* What to skip before effectively reading a packet. */
372 int skip_to_keyframe;
373 uint64_t skip_to_timecode;
375 /* File has a CUES element, but we defer parsing until it is needed. */
376 int cues_parsing_deferred;
378 /* Level1 elements and whether they were read yet */
379 MatroskaLevel1Element level1_elems[64];
380 int num_level1_elems;
382 MatroskaCluster current_cluster;
384 /* WebM DASH Manifest live flag */
387 /* Bandwidth value for WebM DASH Manifest */
389 } MatroskaDemuxContext;
391 #define CHILD_OF(parent) { .def = { .n = parent } }
393 // The following forward declarations need their size because
394 // a tentative definition with internal linkage must not be an
395 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
396 // Removing the sizes breaks MSVC.
397 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
398 matroska_track[27], matroska_track_encoding[6], matroska_track_encodings[2],
399 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_tracks[2],
400 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
401 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
402 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
404 static EbmlSyntax ebml_header[] = {
405 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
406 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
407 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
408 { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
409 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
410 { EBML_ID_EBMLVERSION, EBML_NONE },
411 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
412 CHILD_OF(ebml_syntax)
415 static EbmlSyntax ebml_syntax[] = {
416 { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
417 { MATROSKA_ID_SEGMENT, EBML_STOP },
421 static EbmlSyntax matroska_info[] = {
422 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
423 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, offsetof(MatroskaDemuxContext, duration) },
424 { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
425 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
426 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
427 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
428 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
429 CHILD_OF(matroska_segment)
432 static EbmlSyntax matroska_mastering_meta[] = {
433 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
434 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
435 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
436 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
437 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
438 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
439 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
440 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
441 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
442 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
443 CHILD_OF(matroska_track_video_color)
446 static EbmlSyntax matroska_track_video_color[] = {
447 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
448 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=0 } },
449 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
450 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
451 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
452 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
453 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
454 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
455 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
456 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
457 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
458 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
459 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
460 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
461 CHILD_OF(matroska_track_video)
464 static EbmlSyntax matroska_track_video_projection[] = {
465 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
466 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrackVideoProjection, private) },
467 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f=0.0 } },
468 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f=0.0 } },
469 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f=0.0 } },
470 CHILD_OF(matroska_track_video)
473 static EbmlSyntax matroska_track_video[] = {
474 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
475 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
476 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
477 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
478 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
479 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
480 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
481 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
482 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
483 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
484 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
485 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
486 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
487 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
488 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
489 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
490 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
491 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
492 CHILD_OF(matroska_track)
495 static EbmlSyntax matroska_track_audio[] = {
496 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
497 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
498 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
499 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
500 CHILD_OF(matroska_track)
503 static EbmlSyntax matroska_track_encoding_compression[] = {
504 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, offsetof(MatroskaTrackCompression, algo), { .u = 0 } },
505 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, offsetof(MatroskaTrackCompression, settings) },
506 CHILD_OF(matroska_track_encoding)
509 static EbmlSyntax matroska_track_encoding_encryption[] = {
510 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
511 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, offsetof(MatroskaTrackEncryption,key_id) },
512 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
513 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
514 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
515 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
516 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
517 CHILD_OF(matroska_track_encoding)
519 static EbmlSyntax matroska_track_encoding[] = {
520 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
521 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
522 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
523 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
524 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
525 CHILD_OF(matroska_track_encodings)
528 static EbmlSyntax matroska_track_encodings[] = {
529 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
530 CHILD_OF(matroska_track)
533 static EbmlSyntax matroska_track_plane[] = {
534 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, offsetof(MatroskaTrackPlane,uid) },
535 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, offsetof(MatroskaTrackPlane,type) },
536 CHILD_OF(matroska_track_combine_planes)
539 static EbmlSyntax matroska_track_combine_planes[] = {
540 { MATROSKA_ID_TRACKPLANE, EBML_NEST, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
541 CHILD_OF(matroska_track_operation)
544 static EbmlSyntax matroska_track_operation[] = {
545 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, {.n = matroska_track_combine_planes} },
546 CHILD_OF(matroska_track)
549 static EbmlSyntax matroska_track[] = {
550 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
551 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, offsetof(MatroskaTrack, name) },
552 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
553 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, offsetof(MatroskaTrack, type) },
554 { MATROSKA_ID_CODECID, EBML_STR, 0, offsetof(MatroskaTrack, codec_id) },
555 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
556 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
557 { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
558 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
559 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
560 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
561 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
562 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
563 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
564 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
565 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, { .n = matroska_track_encodings } },
566 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
567 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
568 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
569 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
570 { MATROSKA_ID_CODECNAME, EBML_NONE },
571 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
572 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
573 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
574 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
575 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
576 CHILD_OF(matroska_tracks)
579 static EbmlSyntax matroska_tracks[] = {
580 { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
581 CHILD_OF(matroska_segment)
584 static EbmlSyntax matroska_attachment[] = {
585 { MATROSKA_ID_FILEUID, EBML_UINT, 0, offsetof(MatroskaAttachment, uid) },
586 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
587 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
588 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
589 { MATROSKA_ID_FILEDESC, EBML_NONE },
590 CHILD_OF(matroska_attachments)
593 static EbmlSyntax matroska_attachments[] = {
594 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
595 CHILD_OF(matroska_segment)
598 static EbmlSyntax matroska_chapter_display[] = {
599 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
600 { MATROSKA_ID_CHAPLANG, EBML_NONE },
601 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
602 CHILD_OF(matroska_chapter_entry)
605 static EbmlSyntax matroska_chapter_entry[] = {
606 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
607 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
608 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaChapter, uid) },
609 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, { .n = matroska_chapter_display } },
610 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
611 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
612 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
613 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
614 CHILD_OF(matroska_chapter)
617 static EbmlSyntax matroska_chapter[] = {
618 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
619 { MATROSKA_ID_EDITIONUID, EBML_NONE },
620 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
621 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
622 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
623 CHILD_OF(matroska_chapters)
626 static EbmlSyntax matroska_chapters[] = {
627 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, { .n = matroska_chapter } },
628 CHILD_OF(matroska_segment)
631 static EbmlSyntax matroska_index_pos[] = {
632 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
633 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, offsetof(MatroskaIndexPos, pos) },
634 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
635 { MATROSKA_ID_CUEDURATION, EBML_NONE },
636 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
637 CHILD_OF(matroska_index_entry)
640 static EbmlSyntax matroska_index_entry[] = {
641 { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
642 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
643 CHILD_OF(matroska_index)
646 static EbmlSyntax matroska_index[] = {
647 { MATROSKA_ID_POINTENTRY, EBML_NEST, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
648 CHILD_OF(matroska_segment)
651 static EbmlSyntax matroska_simpletag[] = {
652 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
653 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
654 { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
655 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTag, def) },
656 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
657 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
658 CHILD_OF(matroska_tag)
661 static EbmlSyntax matroska_tagtargets[] = {
662 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, offsetof(MatroskaTagTarget, type) },
663 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
664 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
665 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
666 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
667 CHILD_OF(matroska_tag)
670 static EbmlSyntax matroska_tag[] = {
671 { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
672 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
673 CHILD_OF(matroska_tags)
676 static EbmlSyntax matroska_tags[] = {
677 { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
678 CHILD_OF(matroska_segment)
681 static EbmlSyntax matroska_seekhead_entry[] = {
682 { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
683 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
684 CHILD_OF(matroska_seekhead)
687 static EbmlSyntax matroska_seekhead[] = {
688 { MATROSKA_ID_SEEKENTRY, EBML_NEST, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
689 CHILD_OF(matroska_segment)
692 static EbmlSyntax matroska_segment[] = {
693 { MATROSKA_ID_CLUSTER, EBML_STOP },
694 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
695 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
696 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, { .n = matroska_attachments } },
697 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, { .n = matroska_chapters } },
698 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, { .n = matroska_index } },
699 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
700 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, { .n = matroska_seekhead } },
701 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
704 static EbmlSyntax matroska_segments[] = {
705 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
709 static EbmlSyntax matroska_blockmore[] = {
710 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
711 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
712 CHILD_OF(matroska_blockadditions)
715 static EbmlSyntax matroska_blockadditions[] = {
716 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, {.n = matroska_blockmore} },
717 CHILD_OF(matroska_blockgroup)
720 static EbmlSyntax matroska_blockgroup[] = {
721 { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
722 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, { .n = matroska_blockadditions} },
723 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, offsetof(MatroskaBlock, duration) },
724 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
725 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference), { .i = INT64_MIN } },
726 { MATROSKA_ID_CODECSTATE, EBML_NONE },
727 { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
728 CHILD_OF(matroska_cluster_parsing)
731 // The following array contains SimpleBlock and BlockGroup twice
732 // in order to reuse the other values for matroska_cluster_enter.
733 static EbmlSyntax matroska_cluster_parsing[] = {
734 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
735 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, { .n = matroska_blockgroup } },
736 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
737 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
738 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
739 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
740 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
741 CHILD_OF(matroska_segment)
744 static EbmlSyntax matroska_cluster_enter[] = {
745 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = &matroska_cluster_parsing[2] } },
750 static const char *const matroska_doctypes[] = { "matroska", "webm" };
752 static int matroska_read_close(AVFormatContext *s);
755 * This function prepares the status for parsing of level 1 elements.
757 static int matroska_reset_status(MatroskaDemuxContext *matroska,
758 uint32_t id, int64_t position)
761 int64_t err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
766 matroska->current_id = id;
767 matroska->num_levels = 1;
768 matroska->unknown_count = 0;
769 matroska->resync_pos = avio_tell(matroska->ctx->pb);
771 matroska->resync_pos -= (av_log2(id) + 7) / 8;
776 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
778 AVIOContext *pb = matroska->ctx->pb;
781 /* Try to seek to the last position to resync from. If this doesn't work,
782 * we resync from the earliest position available: The start of the buffer. */
783 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
784 av_log(matroska->ctx, AV_LOG_WARNING,
785 "Seek to desired resync point failed. Seeking to "
786 "earliest point available instead.\n");
787 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
788 last_pos + 1), SEEK_SET);
793 // try to find a toplevel element
794 while (!avio_feof(pb)) {
795 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
796 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
797 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
798 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
799 /* Prepare the context for parsing of a level 1 element. */
800 matroska_reset_status(matroska, id, -1);
801 /* Given that we are here means that an error has occurred,
802 * so treat the segment as unknown length in order not to
803 * discard valid data that happens to be beyond the designated
804 * end of the segment. */
805 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
808 id = (id << 8) | avio_r8(pb);
812 return pb->error ? pb->error : AVERROR_EOF;
816 * Read: an "EBML number", which is defined as a variable-length
817 * array of bytes. The first byte indicates the length by giving a
818 * number of 0-bits followed by a one. The position of the first
819 * "one" bit inside the first byte indicates the length of this
821 * Returns: number of bytes read, < 0 on error
823 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
824 int max_size, uint64_t *number, int eof_forbidden)
830 /* The first byte tells us the length in bytes - except when it is zero. */
835 /* get the length of the EBML number */
836 read = 8 - ff_log2_tab[total];
838 if (!total || read > max_size) {
839 pos = avio_tell(pb) - 1;
841 av_log(matroska->ctx, AV_LOG_ERROR,
842 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
843 "of an EBML number\n", pos, pos);
845 av_log(matroska->ctx, AV_LOG_ERROR,
846 "Length %d indicated by an EBML number's first byte 0x%02x "
847 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
848 read, (uint8_t) total, pos, pos, max_size);
850 return AVERROR_INVALIDDATA;
853 /* read out length */
854 total ^= 1 << ff_log2_tab[total];
856 total = (total << 8) | avio_r8(pb);
858 if (pb->eof_reached) {
870 av_log(matroska->ctx, AV_LOG_ERROR,
871 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
876 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
877 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
884 * Read a EBML length value.
885 * This needs special handling for the "unknown length" case which has multiple
888 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
891 int res = ebml_read_num(matroska, pb, 8, number, 1);
892 if (res > 0 && *number + 1 == 1ULL << (7 * res))
893 *number = EBML_UNKNOWN_LENGTH;
898 * Read the next element as an unsigned int.
899 * Returns NEEDS_CHECKING.
901 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
905 /* big-endian ordering; build up number */
908 *num = (*num << 8) | avio_r8(pb);
910 return NEEDS_CHECKING;
914 * Read the next element as a signed int.
915 * Returns NEEDS_CHECKING.
917 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
924 *num = sign_extend(avio_r8(pb), 8);
926 /* big-endian ordering; build up number */
928 *num = ((uint64_t)*num << 8) | avio_r8(pb);
931 return NEEDS_CHECKING;
935 * Read the next element as a float.
936 * Returns NEEDS_CHECKING or < 0 on obvious failure.
938 static int ebml_read_float(AVIOContext *pb, int size, double *num)
943 *num = av_int2float(avio_rb32(pb));
945 *num = av_int2double(avio_rb64(pb));
947 return AVERROR_INVALIDDATA;
949 return NEEDS_CHECKING;
953 * Read the next element as an ASCII string.
954 * 0 is success, < 0 or NEEDS_CHECKING is failure.
956 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
961 /* EBML strings are usually not 0-terminated, so we allocate one
962 * byte more, read the string and NULL-terminate it ourselves. */
963 if (!(res = av_malloc(size + 1)))
964 return AVERROR(ENOMEM);
965 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
967 return ret < 0 ? ret : NEEDS_CHECKING;
977 * Read the next element as binary data.
978 * 0 is success, < 0 or NEEDS_CHECKING is failure.
980 static int ebml_read_binary(AVIOContext *pb, int length,
981 int64_t pos, EbmlBin *bin)
985 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
988 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
990 bin->data = bin->buf->data;
993 if ((ret = avio_read(pb, bin->data, length)) != length) {
994 av_buffer_unref(&bin->buf);
997 return ret < 0 ? ret : NEEDS_CHECKING;
1004 * Read the next element, but only the header. The contents
1005 * are supposed to be sub-elements which can be read separately.
1006 * 0 is success, < 0 is failure.
1008 static int ebml_read_master(MatroskaDemuxContext *matroska,
1009 uint64_t length, int64_t pos)
1011 MatroskaLevel *level;
1013 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1014 av_log(matroska->ctx, AV_LOG_ERROR,
1015 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1016 return AVERROR(ENOSYS);
1019 level = &matroska->levels[matroska->num_levels++];
1021 level->length = length;
1027 * Read signed/unsigned "EBML" numbers.
1028 * Return: number of bytes processed, < 0 on error
1030 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
1031 uint8_t *data, uint32_t size, uint64_t *num)
1034 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
1035 return ebml_read_num(matroska, &pb, FFMIN(size, 8), num, 1);
1039 * Same as above, but signed.
1041 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1042 uint8_t *data, uint32_t size, int64_t *num)
1047 /* read as unsigned number first */
1048 if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
1051 /* make signed (weird way) */
1052 *num = unum - ((1LL << (7 * res - 1)) - 1);
1057 static int ebml_parse(MatroskaDemuxContext *matroska,
1058 EbmlSyntax *syntax, void *data);
1060 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1064 // Whoever touches this should be aware of the duplication
1065 // existing in matroska_cluster_parsing.
1066 for (i = 0; syntax[i].id; i++)
1067 if (id == syntax[i].id)
1073 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1079 for (int i = 0; syntax[i].id; i++)
1080 switch (syntax[i].type) {
1082 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1085 *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1088 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1092 // the default may be NULL
1093 if (syntax[i].def.s) {
1094 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1095 *dst = av_strdup(syntax[i].def.s);
1097 return AVERROR(ENOMEM);
1102 if (!matroska->levels[matroska->num_levels - 1].length) {
1103 matroska->num_levels--;
1109 res = ebml_parse(matroska, syntax, data);
1112 return res == LEVEL_ENDED ? 0 : res;
1115 static int is_ebml_id_valid(uint32_t id)
1117 // Due to endian nonsense in Matroska, the highest byte with any bits set
1118 // will contain the leading length bit. This bit in turn identifies the
1119 // total byte length of the element by its position within the byte.
1120 unsigned int bits = av_log2(id);
1121 return id && (bits + 7) / 8 == (8 - bits % 8);
1125 * Allocate and return the entry for the level1 element with the given ID. If
1126 * an entry already exists, return the existing entry.
1128 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1132 MatroskaLevel1Element *elem;
1134 if (!is_ebml_id_valid(id))
1137 // Some files link to all clusters; useless.
1138 if (id == MATROSKA_ID_CLUSTER)
1141 // There can be multiple seekheads.
1142 if (id != MATROSKA_ID_SEEKHEAD) {
1143 for (i = 0; i < matroska->num_level1_elems; i++) {
1144 if (matroska->level1_elems[i].id == id)
1145 return &matroska->level1_elems[i];
1149 // Only a completely broken file would have more elements.
1150 // It also provides a low-effort way to escape from circular seekheads
1151 // (every iteration will add a level1 entry).
1152 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1153 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1157 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1158 *elem = (MatroskaLevel1Element){.id = id};
1163 static int ebml_parse(MatroskaDemuxContext *matroska,
1164 EbmlSyntax *syntax, void *data)
1166 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1167 // Forbid unknown-length EBML_NONE elements.
1168 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1172 // max. 16 MB for strings
1173 [EBML_STR] = 0x1000000,
1174 [EBML_UTF8] = 0x1000000,
1175 // max. 256 MB for binary data
1176 [EBML_BIN] = 0x10000000,
1177 // no limits for anything else
1179 AVIOContext *pb = matroska->ctx->pb;
1182 int64_t pos = avio_tell(pb), pos_alt;
1183 int res, update_pos = 1, level_check;
1184 MatroskaLevel1Element *level1_elem;
1185 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1187 if (!matroska->current_id) {
1189 res = ebml_read_num(matroska, pb, 4, &id, 0);
1191 if (pb->eof_reached && res == AVERROR_EOF) {
1192 if (matroska->is_live)
1193 // in live mode, finish parsing if EOF is reached.
1195 if (level && pos == avio_tell(pb)) {
1196 if (level->length == EBML_UNKNOWN_LENGTH) {
1197 // Unknown-length levels automatically end at EOF.
1198 matroska->num_levels--;
1201 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1202 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1208 matroska->current_id = id | 1 << 7 * res;
1209 pos_alt = pos + res;
1212 pos -= (av_log2(matroska->current_id) + 7) / 8;
1215 id = matroska->current_id;
1217 syntax = ebml_parse_id(syntax, id);
1218 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1219 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1220 // Unknown-length levels end when an element from an upper level
1221 // in the hierarchy is encountered.
1222 while (syntax->def.n) {
1223 syntax = ebml_parse_id(syntax->def.n, id);
1225 matroska->num_levels--;
1231 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1232 "%"PRId64"\n", id, pos);
1233 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1237 data = (char *) data + syntax->data_offset;
1238 if (syntax->list_elem_size) {
1239 EbmlList *list = data;
1242 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1243 return AVERROR(ENOMEM);
1244 newelem = av_fast_realloc(list->elem,
1245 &list->alloc_elem_size,
1246 (list->nb_elem + 1) * syntax->list_elem_size);
1248 return AVERROR(ENOMEM);
1249 list->elem = newelem;
1250 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1251 memset(data, 0, syntax->list_elem_size);
1256 if (syntax->type != EBML_STOP) {
1257 matroska->current_id = 0;
1258 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1263 if (matroska->num_levels > 0) {
1264 if (length != EBML_UNKNOWN_LENGTH &&
1265 level->length != EBML_UNKNOWN_LENGTH) {
1266 uint64_t elem_end = pos_alt + length,
1267 level_end = level->start + level->length;
1269 if (elem_end < level_end) {
1271 } else if (elem_end == level_end) {
1272 level_check = LEVEL_ENDED;
1274 av_log(matroska->ctx, AV_LOG_ERROR,
1275 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1276 "containing master element ending at 0x%"PRIx64"\n",
1277 pos, elem_end, level_end);
1278 return AVERROR_INVALIDDATA;
1280 } else if (length != EBML_UNKNOWN_LENGTH) {
1282 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1283 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1284 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1285 return AVERROR_INVALIDDATA;
1288 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1289 || syntax->type == EBML_NEST)) {
1290 // According to the current specifications only clusters and
1291 // segments are allowed to be unknown-length. We also accept
1292 // other unknown-length master elements.
1293 av_log(matroska->ctx, AV_LOG_WARNING,
1294 "Found unknown-length element 0x%"PRIX32" other than "
1295 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1296 "parsing will nevertheless be attempted.\n", id, pos);
1303 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1304 if (length != EBML_UNKNOWN_LENGTH) {
1305 av_log(matroska->ctx, AV_LOG_ERROR,
1306 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1307 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1308 length, max_lengths[syntax->type], id, pos);
1309 } else if (syntax->type != EBML_NONE) {
1310 av_log(matroska->ctx, AV_LOG_ERROR,
1311 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1312 "unknown length, yet the length of an element of its "
1313 "type must be known.\n", id, pos);
1315 av_log(matroska->ctx, AV_LOG_ERROR,
1316 "Found unknown-length element with ID 0x%"PRIX32" at "
1317 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1318 "available.\n", id, pos);
1320 return AVERROR_INVALIDDATA;
1323 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1324 // Loosing sync will likely manifest itself as encountering unknown
1325 // elements which are not reliably distinguishable from elements
1326 // belonging to future extensions of the format.
1327 // We use a heuristic to detect such situations: If the current
1328 // element is not expected at the current syntax level and there
1329 // were only a few unknown elements in a row, then the element is
1330 // skipped or considered defective based upon the length of the
1331 // current element (i.e. how much would be skipped); if there were
1332 // more than a few skipped elements in a row and skipping the current
1333 // element would lead us more than SKIP_THRESHOLD away from the last
1334 // known good position, then it is inferred that an error occurred.
1335 // The dependency on the number of unknown elements in a row exists
1336 // because the distance to the last known good position is
1337 // automatically big if the last parsed element was big.
1338 // In both cases, each unknown element is considered equivalent to
1339 // UNKNOWN_EQUIV of skipped bytes for the check.
1340 // The whole check is only done for non-seekable output, because
1341 // in this situation skipped data can't simply be rechecked later.
1342 // This is especially important when using unkown length elements
1343 // as the check for whether a child exceeds its containing master
1344 // element is not effective in this situation.
1346 matroska->unknown_count = 0;
1348 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1350 if (matroska->unknown_count > 3)
1351 dist += pos_alt - matroska->resync_pos;
1353 if (dist > SKIP_THRESHOLD) {
1354 av_log(matroska->ctx, AV_LOG_ERROR,
1355 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1356 "length 0x%"PRIx64" considered as invalid data. Last "
1357 "known good position 0x%"PRIx64", %d unknown elements"
1358 " in a row\n", id, pos, length, matroska->resync_pos,
1359 matroska->unknown_count);
1360 return AVERROR_INVALIDDATA;
1365 if (update_pos > 0) {
1366 // We have found an element that is allowed at this place
1367 // in the hierarchy and it passed all checks, so treat the beginning
1368 // of the element as the "last known good" position.
1369 matroska->resync_pos = pos;
1372 if (!data && length != EBML_UNKNOWN_LENGTH)
1376 switch (syntax->type) {
1378 res = ebml_read_uint(pb, length, data);
1381 res = ebml_read_sint(pb, length, data);
1384 res = ebml_read_float(pb, length, data);
1388 res = ebml_read_ascii(pb, length, data);
1391 res = ebml_read_binary(pb, length, pos_alt, data);
1395 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1397 if (id == MATROSKA_ID_SEGMENT)
1398 matroska->segment_start = pos_alt;
1399 if (id == MATROSKA_ID_CUES)
1400 matroska->cues_parsing_deferred = 0;
1401 if (syntax->type == EBML_LEVEL1 &&
1402 (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1403 if (!level1_elem->pos) {
1404 // Zero is not a valid position for a level 1 element.
1405 level1_elem->pos = pos;
1406 } else if (level1_elem->pos != pos)
1407 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1408 level1_elem->parsed = 1;
1410 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1419 if (ffio_limit(pb, length) != length) {
1420 // ffio_limit emits its own error message,
1421 // so we don't have to.
1422 return AVERROR(EIO);
1424 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1425 // avio_skip might take us past EOF. We check for this
1426 // by skipping only length - 1 bytes, reading a byte and
1427 // checking the error flags. This is done in order to check
1428 // that the element has been properly skipped even when
1429 // no filesize (that ffio_limit relies on) is available.
1431 res = NEEDS_CHECKING;
1438 if (res == NEEDS_CHECKING) {
1439 if (pb->eof_reached) {
1448 if (res == AVERROR_INVALIDDATA)
1449 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1450 else if (res == AVERROR(EIO))
1451 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1452 else if (res == AVERROR_EOF) {
1453 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1461 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1462 level = &matroska->levels[matroska->num_levels - 1];
1463 pos = avio_tell(pb);
1465 // Given that pos >= level->start no check for
1466 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1467 while (matroska->num_levels && pos == level->start + level->length) {
1468 matroska->num_levels--;
1476 static void ebml_free(EbmlSyntax *syntax, void *data)
1479 for (i = 0; syntax[i].id; i++) {
1480 void *data_off = (char *) data + syntax[i].data_offset;
1481 switch (syntax[i].type) {
1487 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1491 if (syntax[i].list_elem_size) {
1492 EbmlList *list = data_off;
1493 char *ptr = list->elem;
1494 for (j = 0; j < list->nb_elem;
1495 j++, ptr += syntax[i].list_elem_size)
1496 ebml_free(syntax[i].def.n, ptr);
1497 av_freep(&list->elem);
1499 list->alloc_elem_size = 0;
1501 ebml_free(syntax[i].def.n, data_off);
1511 static int matroska_probe(const AVProbeData *p)
1514 int len_mask = 0x80, size = 1, n = 1, i;
1517 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1520 /* length of header */
1522 while (size <= 8 && !(total & len_mask)) {
1528 total &= (len_mask - 1);
1530 total = (total << 8) | p->buf[4 + n++];
1532 if (total + 1 == 1ULL << (7 * size)){
1533 /* Unknown-length header - simply parse the whole buffer. */
1534 total = p->buf_size - 4 - size;
1536 /* Does the probe data contain the whole header? */
1537 if (p->buf_size < 4 + size + total)
1541 /* The header should contain a known document type. For now,
1542 * we don't parse the whole header but simply check for the
1543 * availability of that array of characters inside the header.
1544 * Not fully fool-proof, but good enough. */
1545 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1546 size_t probelen = strlen(matroska_doctypes[i]);
1547 if (total < probelen)
1549 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1550 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1551 return AVPROBE_SCORE_MAX;
1554 // probably valid EBML header but no recognized doctype
1555 return AVPROBE_SCORE_EXTENSION;
1558 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1561 MatroskaTrack *tracks = matroska->tracks.elem;
1564 for (i = 0; i < matroska->tracks.nb_elem; i++)
1565 if (tracks[i].num == num)
1568 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %d\n", num);
1572 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1573 MatroskaTrack *track)
1575 MatroskaTrackEncoding *encodings = track->encodings.elem;
1576 uint8_t *data = *buf;
1577 int isize = *buf_size;
1578 uint8_t *pkt_data = NULL;
1579 uint8_t av_unused *newpktdata;
1580 int pkt_size = isize;
1584 if (pkt_size >= 10000000U)
1585 return AVERROR_INVALIDDATA;
1587 switch (encodings[0].compression.algo) {
1588 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1590 int header_size = encodings[0].compression.settings.size;
1591 uint8_t *header = encodings[0].compression.settings.data;
1593 if (header_size && !header) {
1594 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1601 pkt_size = isize + header_size;
1602 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1604 return AVERROR(ENOMEM);
1606 memcpy(pkt_data, header, header_size);
1607 memcpy(pkt_data + header_size, data, isize);
1611 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1613 olen = pkt_size *= 3;
1614 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1615 + AV_INPUT_BUFFER_PADDING_SIZE);
1617 result = AVERROR(ENOMEM);
1620 pkt_data = newpktdata;
1621 result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1622 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1624 result = AVERROR_INVALIDDATA;
1631 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1633 z_stream zstream = { 0 };
1634 if (inflateInit(&zstream) != Z_OK)
1636 zstream.next_in = data;
1637 zstream.avail_in = isize;
1640 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1642 inflateEnd(&zstream);
1643 result = AVERROR(ENOMEM);
1646 pkt_data = newpktdata;
1647 zstream.avail_out = pkt_size - zstream.total_out;
1648 zstream.next_out = pkt_data + zstream.total_out;
1649 result = inflate(&zstream, Z_NO_FLUSH);
1650 } while (result == Z_OK && pkt_size < 10000000);
1651 pkt_size = zstream.total_out;
1652 inflateEnd(&zstream);
1653 if (result != Z_STREAM_END) {
1654 if (result == Z_MEM_ERROR)
1655 result = AVERROR(ENOMEM);
1657 result = AVERROR_INVALIDDATA;
1664 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1666 bz_stream bzstream = { 0 };
1667 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1669 bzstream.next_in = data;
1670 bzstream.avail_in = isize;
1673 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1675 BZ2_bzDecompressEnd(&bzstream);
1676 result = AVERROR(ENOMEM);
1679 pkt_data = newpktdata;
1680 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1681 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1682 result = BZ2_bzDecompress(&bzstream);
1683 } while (result == BZ_OK && pkt_size < 10000000);
1684 pkt_size = bzstream.total_out_lo32;
1685 BZ2_bzDecompressEnd(&bzstream);
1686 if (result != BZ_STREAM_END) {
1687 if (result == BZ_MEM_ERROR)
1688 result = AVERROR(ENOMEM);
1690 result = AVERROR_INVALIDDATA;
1697 return AVERROR_INVALIDDATA;
1700 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1703 *buf_size = pkt_size;
1711 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1712 AVDictionary **metadata, char *prefix)
1714 MatroskaTag *tags = list->elem;
1718 for (i = 0; i < list->nb_elem; i++) {
1719 const char *lang = tags[i].lang &&
1720 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1722 if (!tags[i].name) {
1723 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1727 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1729 av_strlcpy(key, tags[i].name, sizeof(key));
1730 if (tags[i].def || !lang) {
1731 av_dict_set(metadata, key, tags[i].string, 0);
1732 if (tags[i].sub.nb_elem)
1733 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1736 av_strlcat(key, "-", sizeof(key));
1737 av_strlcat(key, lang, sizeof(key));
1738 av_dict_set(metadata, key, tags[i].string, 0);
1739 if (tags[i].sub.nb_elem)
1740 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1743 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1746 static void matroska_convert_tags(AVFormatContext *s)
1748 MatroskaDemuxContext *matroska = s->priv_data;
1749 MatroskaTags *tags = matroska->tags.elem;
1752 for (i = 0; i < matroska->tags.nb_elem; i++) {
1753 if (tags[i].target.attachuid) {
1754 MatroskaAttachment *attachment = matroska->attachments.elem;
1756 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1757 if (attachment[j].uid == tags[i].target.attachuid &&
1758 attachment[j].stream) {
1759 matroska_convert_tag(s, &tags[i].tag,
1760 &attachment[j].stream->metadata, NULL);
1765 av_log(NULL, AV_LOG_WARNING,
1766 "The tags at index %d refer to a "
1767 "non-existent attachment %"PRId64".\n",
1768 i, tags[i].target.attachuid);
1770 } else if (tags[i].target.chapteruid) {
1771 MatroskaChapter *chapter = matroska->chapters.elem;
1773 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1774 if (chapter[j].uid == tags[i].target.chapteruid &&
1775 chapter[j].chapter) {
1776 matroska_convert_tag(s, &tags[i].tag,
1777 &chapter[j].chapter->metadata, NULL);
1782 av_log(NULL, AV_LOG_WARNING,
1783 "The tags at index %d refer to a non-existent chapter "
1785 i, tags[i].target.chapteruid);
1787 } else if (tags[i].target.trackuid) {
1788 MatroskaTrack *track = matroska->tracks.elem;
1790 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1791 if (track[j].uid == tags[i].target.trackuid &&
1793 matroska_convert_tag(s, &tags[i].tag,
1794 &track[j].stream->metadata, NULL);
1799 av_log(NULL, AV_LOG_WARNING,
1800 "The tags at index %d refer to a non-existent track "
1802 i, tags[i].target.trackuid);
1805 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1806 tags[i].target.type);
1811 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1814 uint32_t saved_id = matroska->current_id;
1815 int64_t before_pos = avio_tell(matroska->ctx->pb);
1819 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1820 /* We don't want to lose our seekhead level, so we add
1821 * a dummy. This is a crude hack. */
1822 if (matroska->num_levels == EBML_MAX_DEPTH) {
1823 av_log(matroska->ctx, AV_LOG_INFO,
1824 "Max EBML element depth (%d) reached, "
1825 "cannot parse further.\n", EBML_MAX_DEPTH);
1826 ret = AVERROR_INVALIDDATA;
1828 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1829 matroska->num_levels++;
1830 matroska->current_id = 0;
1832 ret = ebml_parse(matroska, matroska_segment, matroska);
1833 if (ret == LEVEL_ENDED) {
1834 /* This can only happen if the seek brought us beyond EOF. */
1839 /* Seek back - notice that in all instances where this is used
1840 * it is safe to set the level to 1. */
1841 matroska_reset_status(matroska, saved_id, before_pos);
1846 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1848 EbmlList *seekhead_list = &matroska->seekhead;
1851 // we should not do any seeking in the streaming case
1852 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1855 for (i = 0; i < seekhead_list->nb_elem; i++) {
1856 MatroskaSeekhead *seekheads = seekhead_list->elem;
1857 uint32_t id = seekheads[i].id;
1858 int64_t pos = seekheads[i].pos + matroska->segment_start;
1860 MatroskaLevel1Element *elem = matroska_find_level1_elem(matroska, id);
1861 if (!elem || elem->parsed)
1866 // defer cues parsing until we actually need cue data.
1867 if (id == MATROSKA_ID_CUES)
1870 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1871 // mark index as broken
1872 matroska->cues_parsing_deferred = -1;
1880 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1882 EbmlList *index_list;
1883 MatroskaIndex *index;
1884 uint64_t index_scale = 1;
1887 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1890 index_list = &matroska->index;
1891 index = index_list->elem;
1892 if (index_list->nb_elem < 2)
1894 if (index[1].time > 1E14 / matroska->time_scale) {
1895 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1898 for (i = 0; i < index_list->nb_elem; i++) {
1899 EbmlList *pos_list = &index[i].pos;
1900 MatroskaIndexPos *pos = pos_list->elem;
1901 for (j = 0; j < pos_list->nb_elem; j++) {
1902 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1904 if (track && track->stream)
1905 av_add_index_entry(track->stream,
1906 pos[j].pos + matroska->segment_start,
1907 index[i].time / index_scale, 0, 0,
1913 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1916 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1919 for (i = 0; i < matroska->num_level1_elems; i++) {
1920 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1921 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1922 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1923 matroska->cues_parsing_deferred = -1;
1929 matroska_add_index_entries(matroska);
1932 static int matroska_aac_profile(char *codec_id)
1934 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1937 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1938 if (strstr(codec_id, aac_profiles[profile]))
1943 static int matroska_aac_sri(int samplerate)
1947 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1948 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1953 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1955 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1956 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1959 static int matroska_parse_flac(AVFormatContext *s,
1960 MatroskaTrack *track,
1963 AVStream *st = track->stream;
1964 uint8_t *p = track->codec_priv.data;
1965 int size = track->codec_priv.size;
1967 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1968 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1969 track->codec_priv.size = 0;
1973 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1975 p += track->codec_priv.size;
1976 size -= track->codec_priv.size;
1978 /* parse the remaining metadata blocks if present */
1980 int block_last, block_type, block_size;
1982 flac_parse_block_header(p, &block_last, &block_type, &block_size);
1986 if (block_size > size)
1989 /* check for the channel mask */
1990 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
1991 AVDictionary *dict = NULL;
1992 AVDictionaryEntry *chmask;
1994 ff_vorbis_comment(s, &dict, p, block_size, 0);
1995 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
1997 uint64_t mask = strtol(chmask->value, NULL, 0);
1998 if (!mask || mask & ~0x3ffffULL) {
1999 av_log(s, AV_LOG_WARNING,
2000 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2002 st->codecpar->channel_layout = mask;
2004 av_dict_free(&dict);
2014 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2016 int major, minor, micro, bttb = 0;
2018 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2019 * this function, and fixed in 57.52 */
2020 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, µ) == 3)
2021 bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
2023 switch (field_order) {
2024 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2025 return AV_FIELD_PROGRESSIVE;
2026 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2027 return AV_FIELD_UNKNOWN;
2028 case MATROSKA_VIDEO_FIELDORDER_TT:
2030 case MATROSKA_VIDEO_FIELDORDER_BB:
2032 case MATROSKA_VIDEO_FIELDORDER_BT:
2033 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2034 case MATROSKA_VIDEO_FIELDORDER_TB:
2035 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2037 return AV_FIELD_UNKNOWN;
2041 static void mkv_stereo_mode_display_mul(int stereo_mode,
2042 int *h_width, int *h_height)
2044 switch (stereo_mode) {
2045 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2046 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2047 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2048 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2049 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2051 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2052 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2053 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2054 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2057 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2058 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2059 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2060 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2066 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2067 const MatroskaTrackVideoColor *color = track->video.color.elem;
2068 const MatroskaMasteringMeta *mastering_meta;
2069 int has_mastering_primaries, has_mastering_luminance;
2071 if (!track->video.color.nb_elem)
2074 mastering_meta = &color->mastering_meta;
2075 // Mastering primaries are CIE 1931 coords, and must be > 0.
2076 has_mastering_primaries =
2077 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2078 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2079 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2080 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2081 has_mastering_luminance = mastering_meta->max_luminance > 0;
2083 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2084 st->codecpar->color_space = color->matrix_coefficients;
2085 if (color->primaries != AVCOL_PRI_RESERVED &&
2086 color->primaries != AVCOL_PRI_RESERVED0)
2087 st->codecpar->color_primaries = color->primaries;
2088 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2089 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2090 st->codecpar->color_trc = color->transfer_characteristics;
2091 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2092 color->range <= AVCOL_RANGE_JPEG)
2093 st->codecpar->color_range = color->range;
2094 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2095 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2096 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2097 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2098 st->codecpar->chroma_location =
2099 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2100 (color->chroma_siting_vert - 1) << 7);
2102 if (color->max_cll && color->max_fall) {
2105 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2107 return AVERROR(ENOMEM);
2108 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2109 (uint8_t *)metadata, size);
2111 av_freep(&metadata);
2114 metadata->MaxCLL = color->max_cll;
2115 metadata->MaxFALL = color->max_fall;
2118 if (has_mastering_primaries || has_mastering_luminance) {
2119 AVMasteringDisplayMetadata *metadata =
2120 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2121 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2122 sizeof(AVMasteringDisplayMetadata));
2124 return AVERROR(ENOMEM);
2126 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2127 if (has_mastering_primaries) {
2128 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2129 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2130 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2131 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2132 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2133 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2134 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2135 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2136 metadata->has_primaries = 1;
2138 if (has_mastering_luminance) {
2139 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2140 metadata->min_luminance = av_d2q(mastering_meta->min_luminance, INT_MAX);
2141 metadata->has_luminance = 1;
2147 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track) {
2148 AVSphericalMapping *spherical;
2149 enum AVSphericalProjection projection;
2150 size_t spherical_size;
2151 uint32_t l = 0, t = 0, r = 0, b = 0;
2152 uint32_t padding = 0;
2156 bytestream2_init(&gb, track->video.projection.private.data,
2157 track->video.projection.private.size);
2159 if (bytestream2_get_byte(&gb) != 0) {
2160 av_log(NULL, AV_LOG_WARNING, "Unknown spherical metadata\n");
2164 bytestream2_skip(&gb, 3); // flags
2166 switch (track->video.projection.type) {
2167 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2168 if (track->video.projection.private.size == 20) {
2169 t = bytestream2_get_be32(&gb);
2170 b = bytestream2_get_be32(&gb);
2171 l = bytestream2_get_be32(&gb);
2172 r = bytestream2_get_be32(&gb);
2174 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2175 av_log(NULL, AV_LOG_ERROR,
2176 "Invalid bounding rectangle coordinates "
2177 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2179 return AVERROR_INVALIDDATA;
2181 } else if (track->video.projection.private.size != 0) {
2182 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2183 return AVERROR_INVALIDDATA;
2186 if (l || t || r || b)
2187 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2189 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2191 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2192 if (track->video.projection.private.size < 4) {
2193 av_log(NULL, AV_LOG_ERROR, "Missing projection private properties\n");
2194 return AVERROR_INVALIDDATA;
2195 } else if (track->video.projection.private.size == 12) {
2196 uint32_t layout = bytestream2_get_be32(&gb);
2198 av_log(NULL, AV_LOG_WARNING,
2199 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2202 projection = AV_SPHERICAL_CUBEMAP;
2203 padding = bytestream2_get_be32(&gb);
2205 av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2206 return AVERROR_INVALIDDATA;
2209 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2210 /* No Spherical metadata */
2213 av_log(NULL, AV_LOG_WARNING,
2214 "Unknown spherical metadata type %"PRIu64"\n",
2215 track->video.projection.type);
2219 spherical = av_spherical_alloc(&spherical_size);
2221 return AVERROR(ENOMEM);
2223 spherical->projection = projection;
2225 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2226 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2227 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2229 spherical->padding = padding;
2231 spherical->bound_left = l;
2232 spherical->bound_top = t;
2233 spherical->bound_right = r;
2234 spherical->bound_bottom = b;
2236 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2239 av_freep(&spherical);
2246 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2248 const AVCodecTag *codec_tags;
2250 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2251 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2253 /* Normalize noncompliant private data that starts with the fourcc
2254 * by expanding/shifting the data by 4 bytes and storing the data
2255 * size at the start. */
2256 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2257 int ret = av_buffer_realloc(&track->codec_priv.buf,
2258 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2262 track->codec_priv.data = track->codec_priv.buf->data;
2263 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2264 track->codec_priv.size += 4;
2265 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2268 *fourcc = AV_RL32(track->codec_priv.data + 4);
2269 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2274 static int matroska_parse_tracks(AVFormatContext *s)
2276 MatroskaDemuxContext *matroska = s->priv_data;
2277 MatroskaTrack *tracks = matroska->tracks.elem;
2282 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2283 MatroskaTrack *track = &tracks[i];
2284 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2285 EbmlList *encodings_list = &track->encodings;
2286 MatroskaTrackEncoding *encodings = encodings_list->elem;
2287 uint8_t *extradata = NULL;
2288 int extradata_size = 0;
2289 int extradata_offset = 0;
2290 uint32_t fourcc = 0;
2292 char* key_id_base64 = NULL;
2295 /* Apply some sanity checks. */
2296 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2297 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2298 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2299 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2300 av_log(matroska->ctx, AV_LOG_INFO,
2301 "Unknown or unsupported track type %"PRIu64"\n",
2305 if (!track->codec_id)
2308 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2309 isnan(track->audio.samplerate)) {
2310 av_log(matroska->ctx, AV_LOG_WARNING,
2311 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2312 track->audio.samplerate);
2313 track->audio.samplerate = 8000;
2316 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2317 if (!track->default_duration && track->video.frame_rate > 0) {
2318 double default_duration = 1000000000 / track->video.frame_rate;
2319 if (default_duration > UINT64_MAX || default_duration < 0) {
2320 av_log(matroska->ctx, AV_LOG_WARNING,
2321 "Invalid frame rate %e. Cannot calculate default duration.\n",
2322 track->video.frame_rate);
2324 track->default_duration = default_duration;
2327 if (track->video.display_width == -1)
2328 track->video.display_width = track->video.pixel_width;
2329 if (track->video.display_height == -1)
2330 track->video.display_height = track->video.pixel_height;
2331 if (track->video.color_space.size == 4)
2332 fourcc = AV_RL32(track->video.color_space.data);
2333 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2334 if (!track->audio.out_samplerate)
2335 track->audio.out_samplerate = track->audio.samplerate;
2337 if (encodings_list->nb_elem > 1) {
2338 av_log(matroska->ctx, AV_LOG_ERROR,
2339 "Multiple combined encodings not supported");
2340 } else if (encodings_list->nb_elem == 1) {
2341 if (encodings[0].type) {
2342 if (encodings[0].encryption.key_id.size > 0) {
2343 /* Save the encryption key id to be stored later as a
2345 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2346 key_id_base64 = av_malloc(b64_size);
2347 if (key_id_base64 == NULL)
2348 return AVERROR(ENOMEM);
2350 av_base64_encode(key_id_base64, b64_size,
2351 encodings[0].encryption.key_id.data,
2352 encodings[0].encryption.key_id.size);
2354 encodings[0].scope = 0;
2355 av_log(matroska->ctx, AV_LOG_ERROR,
2356 "Unsupported encoding type");
2360 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2363 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2366 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2368 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2369 encodings[0].scope = 0;
2370 av_log(matroska->ctx, AV_LOG_ERROR,
2371 "Unsupported encoding type");
2372 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2373 uint8_t *codec_priv = track->codec_priv.data;
2374 int ret = matroska_decode_buffer(&track->codec_priv.data,
2375 &track->codec_priv.size,
2378 track->codec_priv.data = NULL;
2379 track->codec_priv.size = 0;
2380 av_log(matroska->ctx, AV_LOG_ERROR,
2381 "Failed to decode codec private data\n");
2384 if (codec_priv != track->codec_priv.data) {
2385 av_buffer_unref(&track->codec_priv.buf);
2386 if (track->codec_priv.data) {
2387 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2388 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2390 if (!track->codec_priv.buf) {
2391 av_freep(&track->codec_priv.data);
2392 track->codec_priv.size = 0;
2393 return AVERROR(ENOMEM);
2400 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2401 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2402 strlen(ff_mkv_codec_tags[j].str))) {
2403 codec_id = ff_mkv_codec_tags[j].id;
2408 st = track->stream = avformat_new_stream(s, NULL);
2410 av_free(key_id_base64);
2411 return AVERROR(ENOMEM);
2414 if (key_id_base64) {
2415 /* export encryption key id as base64 metadata tag */
2416 av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
2417 av_freep(&key_id_base64);
2420 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2421 track->codec_priv.size >= 40 &&
2422 track->codec_priv.data) {
2423 track->ms_compat = 1;
2424 bit_depth = AV_RL16(track->codec_priv.data + 14);
2425 fourcc = AV_RL32(track->codec_priv.data + 16);
2426 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2429 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2431 extradata_offset = 40;
2432 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2433 track->codec_priv.size >= 14 &&
2434 track->codec_priv.data) {
2436 ffio_init_context(&b, track->codec_priv.data,
2437 track->codec_priv.size,
2438 0, NULL, NULL, NULL, NULL);
2439 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2442 codec_id = st->codecpar->codec_id;
2443 fourcc = st->codecpar->codec_tag;
2444 extradata_offset = FFMIN(track->codec_priv.size, 18);
2445 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2446 /* Normally 36, but allow noncompliant private data */
2447 && (track->codec_priv.size >= 32)
2448 && (track->codec_priv.data)) {
2449 uint16_t sample_size;
2450 int ret = get_qt_codec(track, &fourcc, &codec_id);
2453 sample_size = AV_RB16(track->codec_priv.data + 26);
2455 if (sample_size == 8) {
2456 fourcc = MKTAG('r','a','w',' ');
2457 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2458 } else if (sample_size == 16) {
2459 fourcc = MKTAG('t','w','o','s');
2460 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2463 if ((fourcc == MKTAG('t','w','o','s') ||
2464 fourcc == MKTAG('s','o','w','t')) &&
2466 codec_id = AV_CODEC_ID_PCM_S8;
2467 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2468 (track->codec_priv.size >= 21) &&
2469 (track->codec_priv.data)) {
2470 int ret = get_qt_codec(track, &fourcc, &codec_id);
2473 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2474 fourcc = MKTAG('S','V','Q','3');
2475 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2477 if (codec_id == AV_CODEC_ID_NONE)
2478 av_log(matroska->ctx, AV_LOG_ERROR,
2479 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2480 if (track->codec_priv.size >= 86) {
2481 bit_depth = AV_RB16(track->codec_priv.data + 82);
2482 ffio_init_context(&b, track->codec_priv.data,
2483 track->codec_priv.size,
2484 0, NULL, NULL, NULL, NULL);
2485 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2487 track->has_palette = 1;
2490 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2491 switch (track->audio.bitdepth) {
2493 codec_id = AV_CODEC_ID_PCM_U8;
2496 codec_id = AV_CODEC_ID_PCM_S24BE;
2499 codec_id = AV_CODEC_ID_PCM_S32BE;
2502 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2503 switch (track->audio.bitdepth) {
2505 codec_id = AV_CODEC_ID_PCM_U8;
2508 codec_id = AV_CODEC_ID_PCM_S24LE;
2511 codec_id = AV_CODEC_ID_PCM_S32LE;
2514 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2515 track->audio.bitdepth == 64) {
2516 codec_id = AV_CODEC_ID_PCM_F64LE;
2517 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2518 int profile = matroska_aac_profile(track->codec_id);
2519 int sri = matroska_aac_sri(track->audio.samplerate);
2520 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2522 return AVERROR(ENOMEM);
2523 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2524 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2525 if (strstr(track->codec_id, "SBR")) {
2526 sri = matroska_aac_sri(track->audio.out_samplerate);
2527 extradata[2] = 0x56;
2528 extradata[3] = 0xE5;
2529 extradata[4] = 0x80 | (sri << 3);
2533 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2534 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2535 * Create the "atom size", "tag", and "tag version" fields the
2536 * decoder expects manually. */
2537 extradata_size = 12 + track->codec_priv.size;
2538 extradata = av_mallocz(extradata_size +
2539 AV_INPUT_BUFFER_PADDING_SIZE);
2541 return AVERROR(ENOMEM);
2542 AV_WB32(extradata, extradata_size);
2543 memcpy(&extradata[4], "alac", 4);
2544 AV_WB32(&extradata[8], 0);
2545 memcpy(&extradata[12], track->codec_priv.data,
2546 track->codec_priv.size);
2547 } else if (codec_id == AV_CODEC_ID_TTA) {
2549 if (track->audio.channels > UINT16_MAX ||
2550 track->audio.bitdepth > UINT16_MAX) {
2551 av_log(matroska->ctx, AV_LOG_WARNING,
2552 "Too large audio channel number %"PRIu64
2553 " or bitdepth %"PRIu64". Skipping track.\n",
2554 track->audio.channels, track->audio.bitdepth);
2555 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2556 return AVERROR_INVALIDDATA;
2560 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2561 return AVERROR_INVALIDDATA;
2562 extradata_size = 22;
2563 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2565 return AVERROR(ENOMEM);
2567 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2568 bytestream_put_le16(&ptr, 1);
2569 bytestream_put_le16(&ptr, track->audio.channels);
2570 bytestream_put_le16(&ptr, track->audio.bitdepth);
2571 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2572 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2573 track->audio.out_samplerate,
2574 AV_TIME_BASE * 1000));
2575 } else if (codec_id == AV_CODEC_ID_RV10 ||
2576 codec_id == AV_CODEC_ID_RV20 ||
2577 codec_id == AV_CODEC_ID_RV30 ||
2578 codec_id == AV_CODEC_ID_RV40) {
2579 extradata_offset = 26;
2580 } else if (codec_id == AV_CODEC_ID_RA_144) {
2581 track->audio.out_samplerate = 8000;
2582 track->audio.channels = 1;
2583 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2584 codec_id == AV_CODEC_ID_COOK ||
2585 codec_id == AV_CODEC_ID_ATRAC3 ||
2586 codec_id == AV_CODEC_ID_SIPR)
2587 && track->codec_priv.data) {
2590 ffio_init_context(&b, track->codec_priv.data,
2591 track->codec_priv.size,
2592 0, NULL, NULL, NULL, NULL);
2594 flavor = avio_rb16(&b);
2595 track->audio.coded_framesize = avio_rb32(&b);
2597 track->audio.sub_packet_h = avio_rb16(&b);
2598 track->audio.frame_size = avio_rb16(&b);
2599 track->audio.sub_packet_size = avio_rb16(&b);
2601 track->audio.coded_framesize <= 0 ||
2602 track->audio.sub_packet_h <= 0 ||
2603 track->audio.frame_size <= 0 ||
2604 track->audio.sub_packet_size <= 0 && codec_id != AV_CODEC_ID_SIPR)
2605 return AVERROR_INVALIDDATA;
2606 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2607 track->audio.frame_size);
2608 if (!track->audio.buf)
2609 return AVERROR(ENOMEM);
2610 if (codec_id == AV_CODEC_ID_RA_288) {
2611 st->codecpar->block_align = track->audio.coded_framesize;
2612 track->codec_priv.size = 0;
2614 if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2615 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2616 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2617 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2619 st->codecpar->block_align = track->audio.sub_packet_size;
2620 extradata_offset = 78;
2622 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2623 ret = matroska_parse_flac(s, track, &extradata_offset);
2626 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2627 fourcc = AV_RL32(track->codec_priv.data);
2628 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2629 /* we don't need any value stored in CodecPrivate.
2630 make sure that it's not exported as extradata. */
2631 track->codec_priv.size = 0;
2632 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2633 /* For now, propagate only the OBUs, if any. Once libavcodec is
2634 updated to handle isobmff style extradata this can be removed. */
2635 extradata_offset = 4;
2637 track->codec_priv.size -= extradata_offset;
2639 if (codec_id == AV_CODEC_ID_NONE)
2640 av_log(matroska->ctx, AV_LOG_INFO,
2641 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2643 if (track->time_scale < 0.01)
2644 track->time_scale = 1.0;
2645 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2646 1000 * 1000 * 1000); /* 64 bit pts in ns */
2648 /* convert the delay from ns to the track timebase */
2649 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2650 (AVRational){ 1, 1000000000 },
2653 st->codecpar->codec_id = codec_id;
2655 if (strcmp(track->language, "und"))
2656 av_dict_set(&st->metadata, "language", track->language, 0);
2657 av_dict_set(&st->metadata, "title", track->name, 0);
2659 if (track->flag_default)
2660 st->disposition |= AV_DISPOSITION_DEFAULT;
2661 if (track->flag_forced)
2662 st->disposition |= AV_DISPOSITION_FORCED;
2664 if (!st->codecpar->extradata) {
2666 st->codecpar->extradata = extradata;
2667 st->codecpar->extradata_size = extradata_size;
2668 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2669 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2670 return AVERROR(ENOMEM);
2671 memcpy(st->codecpar->extradata,
2672 track->codec_priv.data + extradata_offset,
2673 track->codec_priv.size);
2677 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2678 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2679 int display_width_mul = 1;
2680 int display_height_mul = 1;
2682 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2683 st->codecpar->codec_tag = fourcc;
2685 st->codecpar->bits_per_coded_sample = bit_depth;
2686 st->codecpar->width = track->video.pixel_width;
2687 st->codecpar->height = track->video.pixel_height;
2689 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2690 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2691 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2692 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2694 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2695 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2697 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2698 av_reduce(&st->sample_aspect_ratio.num,
2699 &st->sample_aspect_ratio.den,
2700 st->codecpar->height * track->video.display_width * display_width_mul,
2701 st->codecpar->width * track->video.display_height * display_height_mul,
2704 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2705 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2707 if (track->default_duration) {
2708 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2709 1000000000, track->default_duration, 30000);
2710 #if FF_API_R_FRAME_RATE
2711 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2712 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2713 st->r_frame_rate = st->avg_frame_rate;
2717 /* export stereo mode flag as metadata tag */
2718 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2719 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2721 /* export alpha mode flag as metadata tag */
2722 if (track->video.alpha_mode)
2723 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2725 /* if we have virtual track, mark the real tracks */
2726 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2728 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2730 snprintf(buf, sizeof(buf), "%s_%d",
2731 ff_matroska_video_stereo_plane[planes[j].type], i);
2732 for (k=0; k < matroska->tracks.nb_elem; k++)
2733 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2734 av_dict_set(&tracks[k].stream->metadata,
2735 "stereo_mode", buf, 0);
2739 // add stream level stereo3d side data if it is a supported format
2740 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2741 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2742 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2747 ret = mkv_parse_video_color(st, track);
2750 ret = mkv_parse_video_projection(st, track);
2753 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2754 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2755 st->codecpar->codec_tag = fourcc;
2756 st->codecpar->sample_rate = track->audio.out_samplerate;
2757 st->codecpar->channels = track->audio.channels;
2758 if (!st->codecpar->bits_per_coded_sample)
2759 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2760 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2761 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2762 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2763 st->need_parsing = AVSTREAM_PARSE_FULL;
2764 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2765 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2766 if (track->codec_delay > 0) {
2767 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2768 (AVRational){1, 1000000000},
2769 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2770 48000 : st->codecpar->sample_rate});
2772 if (track->seek_preroll > 0) {
2773 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2774 (AVRational){1, 1000000000},
2775 (AVRational){1, st->codecpar->sample_rate});
2777 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2778 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2780 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2781 st->disposition |= AV_DISPOSITION_CAPTIONS;
2782 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2783 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2784 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2785 st->disposition |= AV_DISPOSITION_METADATA;
2787 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2788 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2795 static int matroska_read_header(AVFormatContext *s)
2797 MatroskaDemuxContext *matroska = s->priv_data;
2798 EbmlList *attachments_list = &matroska->attachments;
2799 EbmlList *chapters_list = &matroska->chapters;
2800 MatroskaAttachment *attachments;
2801 MatroskaChapter *chapters;
2802 uint64_t max_start = 0;
2808 matroska->cues_parsing_deferred = 1;
2810 /* First read the EBML header. */
2811 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2812 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2813 ebml_free(ebml_syntax, &ebml);
2814 return AVERROR_INVALIDDATA;
2816 if (ebml.version > EBML_VERSION ||
2817 ebml.max_size > sizeof(uint64_t) ||
2818 ebml.id_length > sizeof(uint32_t) ||
2819 ebml.doctype_version > 3) {
2820 avpriv_report_missing_feature(matroska->ctx,
2821 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2822 ebml.version, ebml.doctype, ebml.doctype_version);
2823 ebml_free(ebml_syntax, &ebml);
2824 return AVERROR_PATCHWELCOME;
2825 } else if (ebml.doctype_version == 3) {
2826 av_log(matroska->ctx, AV_LOG_WARNING,
2827 "EBML header using unsupported features\n"
2828 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2829 ebml.version, ebml.doctype, ebml.doctype_version);
2831 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2832 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2834 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2835 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2836 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2837 ebml_free(ebml_syntax, &ebml);
2838 return AVERROR_INVALIDDATA;
2841 ebml_free(ebml_syntax, &ebml);
2843 /* The next thing is a segment. */
2844 pos = avio_tell(matroska->ctx->pb);
2845 res = ebml_parse(matroska, matroska_segments, matroska);
2846 // Try resyncing until we find an EBML_STOP type element.
2848 res = matroska_resync(matroska, pos);
2851 pos = avio_tell(matroska->ctx->pb);
2852 res = ebml_parse(matroska, matroska_segment, matroska);
2854 /* Set data_offset as it might be needed later by seek_frame_generic. */
2855 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2856 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2857 matroska_execute_seekhead(matroska);
2859 if (!matroska->time_scale)
2860 matroska->time_scale = 1000000;
2861 if (matroska->duration)
2862 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2863 1000 / AV_TIME_BASE;
2864 av_dict_set(&s->metadata, "title", matroska->title, 0);
2865 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2867 if (matroska->date_utc.size == 8)
2868 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2870 res = matroska_parse_tracks(s);
2874 attachments = attachments_list->elem;
2875 for (j = 0; j < attachments_list->nb_elem; j++) {
2876 if (!(attachments[j].filename && attachments[j].mime &&
2877 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2878 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2880 AVStream *st = avformat_new_stream(s, NULL);
2883 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2884 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2885 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2887 for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2888 if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2889 strlen(ff_mkv_image_mime_tags[i].str))) {
2890 st->codecpar->codec_id = ff_mkv_image_mime_tags[i].id;
2895 attachments[j].stream = st;
2897 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2898 AVPacket *pkt = &st->attached_pic;
2900 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2901 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2903 av_init_packet(pkt);
2904 pkt->buf = av_buffer_ref(attachments[j].bin.buf);
2906 return AVERROR(ENOMEM);
2907 pkt->data = attachments[j].bin.data;
2908 pkt->size = attachments[j].bin.size;
2909 pkt->stream_index = st->index;
2910 pkt->flags |= AV_PKT_FLAG_KEY;
2912 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2913 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2915 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2916 attachments[j].bin.size);
2918 for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2919 if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2920 strlen(ff_mkv_mime_tags[i].str))) {
2921 st->codecpar->codec_id = ff_mkv_mime_tags[i].id;
2929 chapters = chapters_list->elem;
2930 for (i = 0; i < chapters_list->nb_elem; i++)
2931 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2932 (max_start == 0 || chapters[i].start > max_start)) {
2933 chapters[i].chapter =
2934 avpriv_new_chapter(s, chapters[i].uid,
2935 (AVRational) { 1, 1000000000 },
2936 chapters[i].start, chapters[i].end,
2938 if (chapters[i].chapter) {
2939 av_dict_set(&chapters[i].chapter->metadata,
2940 "title", chapters[i].title, 0);
2942 max_start = chapters[i].start;
2945 matroska_add_index_entries(matroska);
2947 matroska_convert_tags(s);
2951 matroska_read_close(s);
2956 * Put one packet in an application-supplied AVPacket struct.
2957 * Returns 0 on success or -1 on failure.
2959 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2962 if (matroska->queue) {
2963 MatroskaTrack *tracks = matroska->tracks.elem;
2964 MatroskaTrack *track;
2966 ff_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
2967 track = &tracks[pkt->stream_index];
2968 if (track->has_palette) {
2969 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2971 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
2973 memcpy(pal, track->palette, AVPALETTE_SIZE);
2975 track->has_palette = 0;
2984 * Free all packets in our internal queue.
2986 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
2988 ff_packet_list_free(&matroska->queue, &matroska->queue_end);
2991 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
2993 uint32_t lace_size[256], int *laces)
2996 uint8_t *data = *buf;
3000 lace_size[0] = size;
3004 av_assert0(size > 0);
3010 case 0x1: /* Xiph lacing */
3014 for (n = 0; n < *laces - 1; n++) {
3018 if (size <= total) {
3019 return AVERROR_INVALIDDATA;
3023 lace_size[n] += temp;
3030 if (size <= total) {
3031 return AVERROR_INVALIDDATA;
3034 lace_size[n] = size - total;
3038 case 0x2: /* fixed-size lacing */
3039 if (size % (*laces)) {
3040 return AVERROR_INVALIDDATA;
3042 for (n = 0; n < *laces; n++)
3043 lace_size[n] = size / *laces;
3046 case 0x3: /* EBML lacing */
3050 n = matroska_ebmlnum_uint(matroska, data, size, &num);
3051 if (n < 0 || num > INT_MAX) {
3052 av_log(matroska->ctx, AV_LOG_INFO,
3053 "EBML block data error\n");
3054 return n < 0 ? n : AVERROR_INVALIDDATA;
3058 total = lace_size[0] = num;
3059 for (n = 1; n < *laces - 1; n++) {
3062 r = matroska_ebmlnum_sint(matroska, data, size, &snum);
3063 if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
3064 av_log(matroska->ctx, AV_LOG_INFO,
3065 "EBML block data error\n");
3066 return r < 0 ? r : AVERROR_INVALIDDATA;
3070 lace_size[n] = lace_size[n - 1] + snum;
3071 total += lace_size[n];
3073 if (size <= total) {
3074 return AVERROR_INVALIDDATA;
3076 lace_size[*laces - 1] = size - total;
3086 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3087 MatroskaTrack *track, AVStream *st,
3088 uint8_t *data, int size, uint64_t timecode,
3091 int a = st->codecpar->block_align;
3092 int sps = track->audio.sub_packet_size;
3093 int cfs = track->audio.coded_framesize;
3094 int h = track->audio.sub_packet_h;
3095 int y = track->audio.sub_packet_cnt;
3096 int w = track->audio.frame_size;
3099 if (!track->audio.pkt_cnt) {
3100 if (track->audio.sub_packet_cnt == 0)
3101 track->audio.buf_timecode = timecode;
3102 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3103 if (size < cfs * h / 2) {
3104 av_log(matroska->ctx, AV_LOG_ERROR,
3105 "Corrupt int4 RM-style audio packet size\n");
3106 return AVERROR_INVALIDDATA;
3108 for (x = 0; x < h / 2; x++)
3109 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3110 data + x * cfs, cfs);
3111 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3113 av_log(matroska->ctx, AV_LOG_ERROR,
3114 "Corrupt sipr RM-style audio packet size\n");
3115 return AVERROR_INVALIDDATA;
3117 memcpy(track->audio.buf + y * w, data, w);
3119 if (size < sps * w / sps || h<=0 || w%sps) {
3120 av_log(matroska->ctx, AV_LOG_ERROR,
3121 "Corrupt generic RM-style audio packet size\n");
3122 return AVERROR_INVALIDDATA;
3124 for (x = 0; x < w / sps; x++)
3125 memcpy(track->audio.buf +
3126 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3127 data + x * sps, sps);
3130 if (++track->audio.sub_packet_cnt >= h) {
3131 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3132 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3133 track->audio.sub_packet_cnt = 0;
3134 track->audio.pkt_cnt = h * w / a;
3138 while (track->audio.pkt_cnt) {
3140 AVPacket pktl, *pkt = &pktl;
3142 ret = av_new_packet(pkt, a);
3147 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3149 pkt->pts = track->audio.buf_timecode;
3150 track->audio.buf_timecode = AV_NOPTS_VALUE;
3152 pkt->stream_index = st->index;
3153 ret = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3155 av_packet_unref(pkt);
3156 return AVERROR(ENOMEM);
3163 /* reconstruct full wavpack blocks from mangled matroska ones */
3164 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
3165 uint8_t **pdst, int *size)
3167 uint8_t *dst = NULL;
3172 int ret, offset = 0;
3174 if (srclen < 12 || track->stream->codecpar->extradata_size < 2)
3175 return AVERROR_INVALIDDATA;
3177 ver = AV_RL16(track->stream->codecpar->extradata);
3179 samples = AV_RL32(src);
3183 while (srclen >= 8) {
3188 uint32_t flags = AV_RL32(src);
3189 uint32_t crc = AV_RL32(src + 4);
3193 multiblock = (flags & 0x1800) != 0x1800;
3196 ret = AVERROR_INVALIDDATA;
3199 blocksize = AV_RL32(src);
3205 if (blocksize > srclen) {
3206 ret = AVERROR_INVALIDDATA;
3210 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3212 ret = AVERROR(ENOMEM);
3216 dstlen += blocksize + 32;
3218 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3219 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3220 AV_WL16(dst + offset + 8, ver); // version
3221 AV_WL16(dst + offset + 10, 0); // track/index_no
3222 AV_WL32(dst + offset + 12, 0); // total samples
3223 AV_WL32(dst + offset + 16, 0); // block index
3224 AV_WL32(dst + offset + 20, samples); // number of samples
3225 AV_WL32(dst + offset + 24, flags); // flags
3226 AV_WL32(dst + offset + 28, crc); // crc
3227 memcpy(dst + offset + 32, src, blocksize); // block data
3230 srclen -= blocksize;
3231 offset += blocksize + 32;
3234 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3246 static int matroska_parse_prores(MatroskaTrack *track, uint8_t *src,
3247 uint8_t **pdst, int *size)
3252 if (AV_RB32(&src[4]) != MKBETAG('i', 'c', 'p', 'f')) {
3255 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3257 return AVERROR(ENOMEM);
3259 AV_WB32(dst, dstlen);
3260 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3261 memcpy(dst + 8, src, dstlen - 8);
3262 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3271 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3272 MatroskaTrack *track,
3274 uint8_t *data, int data_len,
3279 AVPacket pktl, *pkt = &pktl;
3280 uint8_t *id, *settings, *text, *buf;
3281 int id_len, settings_len, text_len;
3286 return AVERROR_INVALIDDATA;
3289 q = data + data_len;
3294 if (*p == '\r' || *p == '\n') {
3303 if (p >= q || *p != '\n')
3304 return AVERROR_INVALIDDATA;
3310 if (*p == '\r' || *p == '\n') {
3311 settings_len = p - settings;
3319 if (p >= q || *p != '\n')
3320 return AVERROR_INVALIDDATA;
3325 while (text_len > 0) {
3326 const int len = text_len - 1;
3327 const uint8_t c = p[len];
3328 if (c != '\r' && c != '\n')
3334 return AVERROR_INVALIDDATA;
3336 err = av_new_packet(pkt, text_len);
3341 memcpy(pkt->data, text, text_len);
3344 buf = av_packet_new_side_data(pkt,
3345 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3348 av_packet_unref(pkt);
3349 return AVERROR(ENOMEM);
3351 memcpy(buf, id, id_len);
3354 if (settings_len > 0) {
3355 buf = av_packet_new_side_data(pkt,
3356 AV_PKT_DATA_WEBVTT_SETTINGS,
3359 av_packet_unref(pkt);
3360 return AVERROR(ENOMEM);
3362 memcpy(buf, settings, settings_len);
3365 // Do we need this for subtitles?
3366 // pkt->flags = AV_PKT_FLAG_KEY;
3368 pkt->stream_index = st->index;
3369 pkt->pts = timecode;
3371 // Do we need this for subtitles?
3372 // pkt->dts = timecode;
3374 pkt->duration = duration;
3377 err = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3379 av_packet_unref(pkt);
3380 return AVERROR(ENOMEM);
3386 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3387 MatroskaTrack *track, AVStream *st,
3388 AVBufferRef *buf, uint8_t *data, int pkt_size,
3389 uint64_t timecode, uint64_t lace_duration,
3390 int64_t pos, int is_keyframe,
3391 uint8_t *additional, uint64_t additional_id, int additional_size,
3392 int64_t discard_padding)
3394 MatroskaTrackEncoding *encodings = track->encodings.elem;
3395 uint8_t *pkt_data = data;
3397 AVPacket pktl, *pkt = &pktl;
3399 if (encodings && !encodings->type && encodings->scope & 1) {
3400 res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
3405 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3407 res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
3409 av_log(matroska->ctx, AV_LOG_ERROR,
3410 "Error parsing a wavpack block.\n");
3413 if (pkt_data != data)
3414 av_freep(&pkt_data);
3418 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES) {
3420 res = matroska_parse_prores(track, pkt_data, &pr_data, &pkt_size);
3422 av_log(matroska->ctx, AV_LOG_ERROR,
3423 "Error parsing a prores block.\n");
3426 if (pkt_data != data)
3427 av_freep(&pkt_data);
3431 av_init_packet(pkt);
3432 if (pkt_data != data)
3433 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3436 pkt->buf = av_buffer_ref(buf);
3439 res = AVERROR(ENOMEM);
3443 pkt->data = pkt_data;
3444 pkt->size = pkt_size;
3445 pkt->flags = is_keyframe;
3446 pkt->stream_index = st->index;
3448 if (additional_size > 0) {
3449 uint8_t *side_data = av_packet_new_side_data(pkt,
3450 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3451 additional_size + 8);
3453 av_packet_unref(pkt);
3454 return AVERROR(ENOMEM);
3456 AV_WB64(side_data, additional_id);
3457 memcpy(side_data + 8, additional, additional_size);
3460 if (discard_padding) {
3461 uint8_t *side_data = av_packet_new_side_data(pkt,
3462 AV_PKT_DATA_SKIP_SAMPLES,
3465 av_packet_unref(pkt);
3466 return AVERROR(ENOMEM);
3468 discard_padding = av_rescale_q(discard_padding,
3469 (AVRational){1, 1000000000},
3470 (AVRational){1, st->codecpar->sample_rate});
3471 if (discard_padding > 0) {
3472 AV_WL32(side_data + 4, discard_padding);
3474 AV_WL32(side_data, -discard_padding);
3478 if (track->ms_compat)
3479 pkt->dts = timecode;
3481 pkt->pts = timecode;
3483 pkt->duration = lace_duration;
3485 #if FF_API_CONVERGENCE_DURATION
3486 FF_DISABLE_DEPRECATION_WARNINGS
3487 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3488 pkt->convergence_duration = lace_duration;
3490 FF_ENABLE_DEPRECATION_WARNINGS
3493 res = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3495 av_packet_unref(pkt);
3496 return AVERROR(ENOMEM);
3502 if (pkt_data != data)
3503 av_freep(&pkt_data);
3507 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3508 int size, int64_t pos, uint64_t cluster_time,
3509 uint64_t block_duration, int is_keyframe,
3510 uint8_t *additional, uint64_t additional_id, int additional_size,
3511 int64_t cluster_pos, int64_t discard_padding)
3513 uint64_t timecode = AV_NOPTS_VALUE;
3514 MatroskaTrack *track;
3518 uint32_t lace_size[256];
3519 int n, flags, laces = 0;
3521 int trust_default_duration = 1;
3523 if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
3529 track = matroska_find_track_by_num(matroska, num);
3530 if (!track || !track->stream) {
3531 av_log(matroska->ctx, AV_LOG_INFO,
3532 "Invalid stream %"PRIu64"\n", num);
3533 return AVERROR_INVALIDDATA;
3534 } else if (size <= 3)
3537 if (st->discard >= AVDISCARD_ALL)
3539 av_assert1(block_duration != AV_NOPTS_VALUE);
3541 block_time = sign_extend(AV_RB16(data), 16);
3545 if (is_keyframe == -1)
3546 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3548 if (cluster_time != (uint64_t) -1 &&
3549 (block_time >= 0 || cluster_time >= -block_time)) {
3550 timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3551 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3552 timecode < track->end_timecode)
3553 is_keyframe = 0; /* overlapping subtitles are not key frame */
3555 ff_reduce_index(matroska->ctx, st->index);
3556 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3561 if (matroska->skip_to_keyframe &&
3562 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3563 // Compare signed timecodes. Timecode may be negative due to codec delay
3564 // offset. We don't support timestamps greater than int64_t anyway - see
3566 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3569 matroska->skip_to_keyframe = 0;
3570 else if (!st->skip_to_keyframe) {
3571 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3572 matroska->skip_to_keyframe = 0;
3576 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3581 if (track->audio.samplerate == 8000) {
3582 // If this is needed for more codecs, then add them here
3583 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3584 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3585 trust_default_duration = 0;
3589 if (!block_duration && trust_default_duration)
3590 block_duration = track->default_duration * laces / matroska->time_scale;
3592 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3593 track->end_timecode =
3594 FFMAX(track->end_timecode, timecode + block_duration);
3596 for (n = 0; n < laces; n++) {
3597 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3599 if ((st->codecpar->codec_id == AV_CODEC_ID_RA_288 ||
3600 st->codecpar->codec_id == AV_CODEC_ID_COOK ||
3601 st->codecpar->codec_id == AV_CODEC_ID_SIPR ||
3602 st->codecpar->codec_id == AV_CODEC_ID_ATRAC3) &&
3603 st->codecpar->block_align && track->audio.sub_packet_size) {
3604 res = matroska_parse_rm_audio(matroska, track, st, data,
3610 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3611 res = matroska_parse_webvtt(matroska, track, st,
3613 timecode, lace_duration,
3618 res = matroska_parse_frame(matroska, track, st, buf, data, lace_size[n],
3619 timecode, lace_duration, pos,
3620 !n ? is_keyframe : 0,
3621 additional, additional_id, additional_size,
3627 if (timecode != AV_NOPTS_VALUE)
3628 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3629 data += lace_size[n];
3635 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3637 MatroskaCluster *cluster = &matroska->current_cluster;
3638 MatroskaBlock *block = &cluster->block;
3641 av_assert0(matroska->num_levels <= 2);
3643 if (matroska->num_levels == 1) {
3644 res = ebml_parse(matroska, matroska_segment, NULL);
3647 /* Found a cluster: subtract the size of the ID already read. */
3648 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3650 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3656 if (matroska->num_levels == 2) {
3657 /* We are inside a cluster. */
3658 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3660 if (res >= 0 && block->bin.size > 0) {
3661 int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3662 uint8_t* additional = block->additional.size > 0 ?
3663 block->additional.data : NULL;
3665 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3666 block->bin.size, block->bin.pos,
3667 cluster->timecode, block->duration,
3668 is_keyframe, additional, block->additional_id,
3669 block->additional.size, cluster->pos,
3670 block->discard_padding);
3673 ebml_free(matroska_blockgroup, block);
3674 memset(block, 0, sizeof(*block));
3675 } else if (!matroska->num_levels) {
3676 if (!avio_feof(matroska->ctx->pb)) {
3677 avio_r8(matroska->ctx->pb);
3678 if (!avio_feof(matroska->ctx->pb)) {
3679 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3680 "end of segment.\n");
3681 return AVERROR_INVALIDDATA;
3691 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3693 MatroskaDemuxContext *matroska = s->priv_data;
3696 if (matroska->resync_pos == -1) {
3697 // This can only happen if generic seeking has been used.
3698 matroska->resync_pos = avio_tell(s->pb);
3701 while (matroska_deliver_packet(matroska, pkt)) {
3703 return (ret < 0) ? ret : AVERROR_EOF;
3704 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3705 ret = matroska_resync(matroska, matroska->resync_pos);
3711 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3712 int64_t timestamp, int flags)
3714 MatroskaDemuxContext *matroska = s->priv_data;
3715 MatroskaTrack *tracks = NULL;
3716 AVStream *st = s->streams[stream_index];
3719 /* Parse the CUES now since we need the index data to seek. */
3720 if (matroska->cues_parsing_deferred > 0) {
3721 matroska->cues_parsing_deferred = 0;
3722 matroska_parse_cues(matroska);
3725 if (!st->nb_index_entries)
3727 timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3729 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3730 matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3731 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3732 matroska_clear_queue(matroska);
3733 if (matroska_parse_cluster(matroska) < 0)
3738 matroska_clear_queue(matroska);
3739 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3742 tracks = matroska->tracks.elem;
3743 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3744 tracks[i].audio.pkt_cnt = 0;
3745 tracks[i].audio.sub_packet_cnt = 0;
3746 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3747 tracks[i].end_timecode = 0;
3750 /* We seek to a level 1 element, so set the appropriate status. */
3751 matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3752 if (flags & AVSEEK_FLAG_ANY) {
3753 st->skip_to_keyframe = 0;
3754 matroska->skip_to_timecode = timestamp;
3756 st->skip_to_keyframe = 1;
3757 matroska->skip_to_timecode = st->index_entries[index].timestamp;
3759 matroska->skip_to_keyframe = 1;
3761 ff_update_cur_dts(s, st, st->index_entries[index].timestamp);
3764 // slightly hackish but allows proper fallback to
3765 // the generic seeking code.
3766 matroska_reset_status(matroska, 0, -1);
3767 matroska->resync_pos = -1;
3768 matroska_clear_queue(matroska);
3769 st->skip_to_keyframe =
3770 matroska->skip_to_keyframe = 0;
3775 static int matroska_read_close(AVFormatContext *s)
3777 MatroskaDemuxContext *matroska = s->priv_data;
3778 MatroskaTrack *tracks = matroska->tracks.elem;
3781 matroska_clear_queue(matroska);
3783 for (n = 0; n < matroska->tracks.nb_elem; n++)
3784 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3785 av_freep(&tracks[n].audio.buf);
3786 ebml_free(matroska_segment, matroska);
3792 int64_t start_time_ns;
3793 int64_t end_time_ns;
3794 int64_t start_offset;
3798 /* This function searches all the Cues and returns the CueDesc corresponding to
3799 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3800 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3802 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3803 MatroskaDemuxContext *matroska = s->priv_data;
3806 int nb_index_entries = s->streams[0]->nb_index_entries;
3807 AVIndexEntry *index_entries = s->streams[0]->index_entries;
3808 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3809 for (i = 1; i < nb_index_entries; i++) {
3810 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3811 index_entries[i].timestamp * matroska->time_scale > ts) {
3816 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3817 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3818 if (i != nb_index_entries - 1) {
3819 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3820 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3822 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3823 // FIXME: this needs special handling for files where Cues appear
3824 // before Clusters. the current logic assumes Cues appear after
3826 cue_desc.end_offset = cues_start - matroska->segment_start;
3831 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3833 MatroskaDemuxContext *matroska = s->priv_data;
3834 uint32_t id = matroska->current_id;
3835 int64_t cluster_pos, before_pos;
3837 if (s->streams[0]->nb_index_entries <= 0) return 0;
3838 // seek to the first cluster using cues.
3839 index = av_index_search_timestamp(s->streams[0], 0, 0);
3840 if (index < 0) return 0;
3841 cluster_pos = s->streams[0]->index_entries[index].pos;
3842 before_pos = avio_tell(s->pb);
3844 uint64_t cluster_id, cluster_length;
3847 avio_seek(s->pb, cluster_pos, SEEK_SET);
3848 // read cluster id and length
3849 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3850 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3852 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3856 matroska_reset_status(matroska, 0, cluster_pos);
3857 matroska_clear_queue(matroska);
3858 if (matroska_parse_cluster(matroska) < 0 ||
3862 pkt = &matroska->queue->pkt;
3863 // 4 + read is the length of the cluster id and the cluster length field.
3864 cluster_pos += 4 + read + cluster_length;
3865 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3871 /* Restore the status after matroska_read_header: */
3872 matroska_reset_status(matroska, id, before_pos);
3877 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3878 double min_buffer, double* buffer,
3879 double* sec_to_download, AVFormatContext *s,
3882 double nano_seconds_per_second = 1000000000.0;
3883 double time_sec = time_ns / nano_seconds_per_second;
3885 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3886 int64_t end_time_ns = time_ns + time_to_search_ns;
3887 double sec_downloaded = 0.0;
3888 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3889 if (desc_curr.start_time_ns == -1)
3891 *sec_to_download = 0.0;
3893 // Check for non cue start time.
3894 if (time_ns > desc_curr.start_time_ns) {
3895 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3896 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3897 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3898 double timeToDownload = (cueBytes * 8.0) / bps;
3900 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3901 *sec_to_download += timeToDownload;
3903 // Check if the search ends within the first cue.
3904 if (desc_curr.end_time_ns >= end_time_ns) {
3905 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3906 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3907 sec_downloaded = percent_to_sub * sec_downloaded;
3908 *sec_to_download = percent_to_sub * *sec_to_download;
3911 if ((sec_downloaded + *buffer) <= min_buffer) {
3915 // Get the next Cue.
3916 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3919 while (desc_curr.start_time_ns != -1) {
3920 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3921 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3922 double desc_sec = desc_ns / nano_seconds_per_second;
3923 double bits = (desc_bytes * 8.0);
3924 double time_to_download = bits / bps;
3926 sec_downloaded += desc_sec - time_to_download;
3927 *sec_to_download += time_to_download;
3929 if (desc_curr.end_time_ns >= end_time_ns) {
3930 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3931 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3932 sec_downloaded = percent_to_sub * sec_downloaded;
3933 *sec_to_download = percent_to_sub * *sec_to_download;
3935 if ((sec_downloaded + *buffer) <= min_buffer)
3940 if ((sec_downloaded + *buffer) <= min_buffer) {
3945 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3947 *buffer = *buffer + sec_downloaded;
3951 /* This function computes the bandwidth of the WebM file with the help of
3952 * buffer_size_after_time_downloaded() function. Both of these functions are
3953 * adapted from WebM Tools project and are adapted to work with FFmpeg's
3954 * Matroska parsing mechanism.
3956 * Returns the bandwidth of the file on success; -1 on error.
3958 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
3960 MatroskaDemuxContext *matroska = s->priv_data;
3961 AVStream *st = s->streams[0];
3962 double bandwidth = 0.0;
3965 for (i = 0; i < st->nb_index_entries; i++) {
3966 int64_t prebuffer_ns = 1000000000;
3967 int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
3968 double nano_seconds_per_second = 1000000000.0;
3969 int64_t prebuffered_ns = time_ns + prebuffer_ns;
3970 double prebuffer_bytes = 0.0;
3971 int64_t temp_prebuffer_ns = prebuffer_ns;
3972 int64_t pre_bytes, pre_ns;
3973 double pre_sec, prebuffer, bits_per_second;
3974 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
3976 // Start with the first Cue.
3977 CueDesc desc_end = desc_beg;
3979 // Figure out how much data we have downloaded for the prebuffer. This will
3980 // be used later to adjust the bits per sample to try.
3981 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
3982 // Prebuffered the entire Cue.
3983 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
3984 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
3985 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
3987 if (desc_end.start_time_ns == -1) {
3988 // The prebuffer is larger than the duration.
3989 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
3991 bits_per_second = 0.0;
3993 // The prebuffer ends in the last Cue. Estimate how much data was
3995 pre_bytes = desc_end.end_offset - desc_end.start_offset;
3996 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
3997 pre_sec = pre_ns / nano_seconds_per_second;
3999 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4001 prebuffer = prebuffer_ns / nano_seconds_per_second;
4003 // Set this to 0.0 in case our prebuffer buffers the entire video.
4004 bits_per_second = 0.0;
4006 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4007 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4008 double desc_sec = desc_ns / nano_seconds_per_second;
4009 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4011 // Drop the bps by the percentage of bytes buffered.
4012 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4013 double mod_bits_per_second = calc_bits_per_second * percent;
4015 if (prebuffer < desc_sec) {
4017 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4019 // Add 1 so the bits per second should be a little bit greater than file
4021 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4022 const double min_buffer = 0.0;
4023 double buffer = prebuffer;
4024 double sec_to_download = 0.0;
4026 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4027 min_buffer, &buffer, &sec_to_download,
4031 } else if (rv == 0) {
4032 bits_per_second = (double)(bps);
4037 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4038 } while (desc_end.start_time_ns != -1);
4040 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4042 return (int64_t)bandwidth;
4045 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4047 MatroskaDemuxContext *matroska = s->priv_data;
4048 EbmlList *seekhead_list = &matroska->seekhead;
4049 MatroskaSeekhead *seekhead = seekhead_list->elem;
4051 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4055 // determine cues start and end positions
4056 for (i = 0; i < seekhead_list->nb_elem; i++)
4057 if (seekhead[i].id == MATROSKA_ID_CUES)
4060 if (i >= seekhead_list->nb_elem) return -1;
4062 before_pos = avio_tell(matroska->ctx->pb);
4063 cues_start = seekhead[i].pos + matroska->segment_start;
4064 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4065 // cues_end is computed as cues_start + cues_length + length of the
4066 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4067 // cues_end is inclusive and the above sum is reduced by 1.
4068 uint64_t cues_length, cues_id;
4070 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4071 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4072 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4073 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4076 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4078 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4079 if (cues_start == -1 || cues_end == -1) return -1;
4082 matroska_parse_cues(matroska);
4085 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4088 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4090 // if the file has cues at the start, fix up the init range so that
4091 // it does not include it
4092 if (cues_start <= init_range)
4093 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4096 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4097 if (bandwidth < 0) return -1;
4098 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4100 // check if all clusters start with key frames
4101 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4103 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4104 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4105 buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4106 if (!buf) return -1;
4108 for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4109 int ret = snprintf(buf + end, 20,
4110 "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4111 i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4112 if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4113 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4115 return AVERROR_INVALIDDATA;
4119 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
4125 static int webm_dash_manifest_read_header(AVFormatContext *s)
4128 int ret = matroska_read_header(s);
4130 MatroskaTrack *tracks;
4131 MatroskaDemuxContext *matroska = s->priv_data;
4133 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4136 if (!s->nb_streams) {
4137 matroska_read_close(s);
4138 av_log(s, AV_LOG_ERROR, "No streams found\n");
4139 return AVERROR_INVALIDDATA;
4142 if (!matroska->is_live) {
4143 buf = av_asprintf("%g", matroska->duration);
4144 if (!buf) return AVERROR(ENOMEM);
4145 av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
4148 // initialization range
4149 // 5 is the offset of Cluster ID.
4150 init_range = avio_tell(s->pb) - 5;
4151 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4154 // basename of the file
4155 buf = strrchr(s->url, '/');
4156 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4159 tracks = matroska->tracks.elem;
4160 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4162 // parse the cues and populate Cue related fields
4163 if (!matroska->is_live) {
4164 ret = webm_dash_manifest_cues(s, init_range);
4166 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4171 // use the bandwidth from the command line if it was provided
4172 if (matroska->bandwidth > 0) {
4173 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4174 matroska->bandwidth, 0);
4179 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4184 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4185 static const AVOption options[] = {
4186 { "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 },
4187 { "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 },
4191 static const AVClass webm_dash_class = {
4192 .class_name = "WebM DASH Manifest demuxer",
4193 .item_name = av_default_item_name,
4195 .version = LIBAVUTIL_VERSION_INT,
4198 AVInputFormat ff_matroska_demuxer = {
4199 .name = "matroska,webm",
4200 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4201 .extensions = "mkv,mk3d,mka,mks",
4202 .priv_data_size = sizeof(MatroskaDemuxContext),
4203 .read_probe = matroska_probe,
4204 .read_header = matroska_read_header,
4205 .read_packet = matroska_read_packet,
4206 .read_close = matroska_read_close,
4207 .read_seek = matroska_read_seek,
4208 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4211 AVInputFormat ff_webm_dash_manifest_demuxer = {
4212 .name = "webm_dash_manifest",
4213 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4214 .priv_data_size = sizeof(MatroskaDemuxContext),
4215 .read_header = webm_dash_manifest_read_header,
4216 .read_packet = webm_dash_manifest_read_packet,
4217 .read_close = matroska_read_close,
4218 .priv_class = &webm_dash_class,