2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The FFmpeg Project
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * Matroska file demuxer
25 * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28 * @see specs available on the Matroska project page: http://www.matroska.org/
36 #include "libavutil/avstring.h"
37 #include "libavutil/base64.h"
38 #include "libavutil/dict.h"
39 #include "libavutil/intfloat.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/lzo.h"
42 #include "libavutil/mastering_display_metadata.h"
43 #include "libavutil/mathematics.h"
44 #include "libavutil/opt.h"
45 #include "libavutil/time_internal.h"
46 #include "libavutil/spherical.h"
48 #include "libavcodec/bytestream.h"
49 #include "libavcodec/flac.h"
50 #include "libavcodec/mpeg4audio.h"
51 #include "libavcodec/packet_internal.h"
54 #include "avio_internal.h"
59 /* For ff_codec_get_id(). */
70 #include "qtpalette.h"
72 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
73 #define NEEDS_CHECKING 2 /* Indicates that some error checks
74 * still need to be performed */
75 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
76 * syntax level used for parsing ended. */
77 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
78 * of unkown, potentially damaged data is encountered,
79 * it is considered an error. */
80 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
81 * to this many bytes of unknown data for the
82 * SKIP_THRESHOLD check. */
98 typedef struct CountedElement {
108 typedef const struct EbmlSyntax {
112 size_t list_elem_size;
119 const struct EbmlSyntax *n;
123 typedef struct EbmlList {
125 unsigned int alloc_elem_size;
129 typedef struct EbmlBin {
136 typedef struct Ebml {
141 uint64_t doctype_version;
144 typedef struct MatroskaTrackCompression {
147 } MatroskaTrackCompression;
149 typedef struct MatroskaTrackEncryption {
152 } MatroskaTrackEncryption;
154 typedef struct MatroskaTrackEncoding {
157 MatroskaTrackCompression compression;
158 MatroskaTrackEncryption encryption;
159 } MatroskaTrackEncoding;
161 typedef struct MatroskaMasteringMeta {
170 double max_luminance;
171 CountedElement min_luminance;
172 } MatroskaMasteringMeta;
174 typedef struct MatroskaTrackVideoColor {
175 uint64_t matrix_coefficients;
176 uint64_t bits_per_channel;
177 uint64_t chroma_sub_horz;
178 uint64_t chroma_sub_vert;
179 uint64_t cb_sub_horz;
180 uint64_t cb_sub_vert;
181 uint64_t chroma_siting_horz;
182 uint64_t chroma_siting_vert;
184 uint64_t transfer_characteristics;
188 MatroskaMasteringMeta mastering_meta;
189 } MatroskaTrackVideoColor;
191 typedef struct MatroskaTrackVideoProjection {
197 } MatroskaTrackVideoProjection;
199 typedef struct MatroskaTrackVideo {
201 uint64_t display_width;
202 uint64_t display_height;
203 uint64_t pixel_width;
204 uint64_t pixel_height;
206 uint64_t display_unit;
208 uint64_t field_order;
209 uint64_t stereo_mode;
212 MatroskaTrackVideoProjection projection;
213 } MatroskaTrackVideo;
215 typedef struct MatroskaTrackAudio {
217 double out_samplerate;
221 /* real audio header (extracted from extradata) */
228 uint64_t buf_timecode;
230 } MatroskaTrackAudio;
232 typedef struct MatroskaTrackPlane {
235 } MatroskaTrackPlane;
237 typedef struct MatroskaTrackOperation {
238 EbmlList combine_planes;
239 } MatroskaTrackOperation;
241 typedef struct MatroskaTrack {
250 uint64_t default_duration;
251 uint64_t flag_default;
252 uint64_t flag_forced;
253 uint64_t flag_comment;
254 uint64_t flag_hearingimpaired;
255 uint64_t flag_visualimpaired;
256 uint64_t flag_textdescriptions;
257 uint64_t seek_preroll;
258 MatroskaTrackVideo video;
259 MatroskaTrackAudio audio;
260 MatroskaTrackOperation operation;
262 uint64_t codec_delay;
263 uint64_t codec_delay_in_track_tb;
266 int64_t end_timecode;
269 uint64_t max_block_additional_id;
271 uint32_t palette[AVPALETTE_COUNT];
275 typedef struct MatroskaAttachment {
283 } MatroskaAttachment;
285 typedef struct MatroskaChapter {
294 typedef struct MatroskaIndexPos {
299 typedef struct MatroskaIndex {
304 typedef struct MatroskaTag {
312 typedef struct MatroskaTagTarget {
320 typedef struct MatroskaTags {
321 MatroskaTagTarget target;
325 typedef struct MatroskaSeekhead {
330 typedef struct MatroskaLevel {
335 typedef struct MatroskaBlock {
337 CountedElement reference;
340 uint64_t additional_id;
342 int64_t discard_padding;
345 typedef struct MatroskaCluster {
351 typedef struct MatroskaLevel1Element {
355 } MatroskaLevel1Element;
357 typedef struct MatroskaDemuxContext {
358 const AVClass *class;
359 AVFormatContext *ctx;
362 MatroskaLevel levels[EBML_MAX_DEPTH];
374 EbmlList attachments;
380 /* byte position of the segment inside the stream */
381 int64_t segment_start;
383 /* the packet queue */
385 AVPacketList *queue_end;
389 /* What to skip before effectively reading a packet. */
390 int skip_to_keyframe;
391 uint64_t skip_to_timecode;
393 /* File has a CUES element, but we defer parsing until it is needed. */
394 int cues_parsing_deferred;
396 /* Level1 elements and whether they were read yet */
397 MatroskaLevel1Element level1_elems[64];
398 int num_level1_elems;
400 MatroskaCluster current_cluster;
402 /* WebM DASH Manifest live flag */
405 /* Bandwidth value for WebM DASH Manifest */
407 } MatroskaDemuxContext;
409 #define CHILD_OF(parent) { .def = { .n = parent } }
411 // The following forward declarations need their size because
412 // a tentative definition with internal linkage must not be an
413 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
414 // Removing the sizes breaks MSVC.
415 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
416 matroska_track[31], matroska_track_encoding[6], matroska_track_encodings[2],
417 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_tracks[2],
418 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
419 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
420 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
422 static EbmlSyntax ebml_header[] = {
423 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
424 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, 0, offsetof(Ebml, max_size), { .u = 8 } },
425 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, 0, offsetof(Ebml, id_length), { .u = 4 } },
426 { EBML_ID_DOCTYPE, EBML_STR, 0, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
427 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
428 { EBML_ID_EBMLVERSION, EBML_NONE },
429 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
430 CHILD_OF(ebml_syntax)
433 static EbmlSyntax ebml_syntax[] = {
434 { EBML_ID_HEADER, EBML_NEST, 0, 0, 0, { .n = ebml_header } },
435 { MATROSKA_ID_SEGMENT, EBML_STOP },
439 static EbmlSyntax matroska_info[] = {
440 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
441 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, 0, offsetof(MatroskaDemuxContext, duration) },
442 { MATROSKA_ID_TITLE, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, title) },
443 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
444 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, muxingapp) },
445 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, 0, offsetof(MatroskaDemuxContext, date_utc) },
446 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
447 CHILD_OF(matroska_segment)
450 static EbmlSyntax matroska_mastering_meta[] = {
451 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_x) },
452 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_y) },
453 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_x) },
454 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_y) },
455 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_x) },
456 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_y) },
457 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_x) },
458 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_y) },
459 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 1, 0, offsetof(MatroskaMasteringMeta, min_luminance) },
460 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, max_luminance) },
461 CHILD_OF(matroska_track_video_color)
464 static EbmlSyntax matroska_track_video_color[] = {
465 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
466 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u = 0 } },
467 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz) },
468 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert) },
469 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz) },
470 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert) },
471 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
472 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
473 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
474 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
475 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
476 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_cll) },
477 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_fall) },
478 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
479 CHILD_OF(matroska_track_video)
482 static EbmlSyntax matroska_track_video_projection[] = {
483 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
484 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideoProjection, private) },
485 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f = 0.0 } },
486 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f = 0.0 } },
487 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f = 0.0 } },
488 CHILD_OF(matroska_track_video)
491 static EbmlSyntax matroska_track_video[] = {
492 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideo, frame_rate) },
493 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
494 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
495 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_width) },
496 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_height) },
497 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideo, color_space) },
498 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, alpha_mode), { .u = 0 } },
499 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, 0, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
500 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
501 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
502 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
503 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
504 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
505 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
506 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
507 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
508 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
509 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
510 CHILD_OF(matroska_track)
513 static EbmlSyntax matroska_track_audio[] = {
514 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
515 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
516 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, bitdepth) },
517 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
518 CHILD_OF(matroska_track)
521 static EbmlSyntax matroska_track_encoding_compression[] = {
522 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackCompression, algo), { .u = MATROSKA_TRACK_ENCODING_COMP_ZLIB } },
523 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, 0, offsetof(MatroskaTrackCompression, settings) },
524 CHILD_OF(matroska_track_encoding)
527 static EbmlSyntax matroska_track_encoding_encryption[] = {
528 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
529 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, 0, offsetof(MatroskaTrackEncryption,key_id) },
530 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
531 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
532 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
533 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
534 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
535 CHILD_OF(matroska_track_encoding)
537 static EbmlSyntax matroska_track_encoding[] = {
538 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
539 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
540 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
541 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
542 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
543 CHILD_OF(matroska_track_encodings)
546 static EbmlSyntax matroska_track_encodings[] = {
547 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, 0, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
548 CHILD_OF(matroska_track)
551 static EbmlSyntax matroska_track_plane[] = {
552 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,uid) },
553 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,type) },
554 CHILD_OF(matroska_track_combine_planes)
557 static EbmlSyntax matroska_track_combine_planes[] = {
558 { MATROSKA_ID_TRACKPLANE, EBML_NEST, 0, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
559 CHILD_OF(matroska_track_operation)
562 static EbmlSyntax matroska_track_operation[] = {
563 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, 0, {.n = matroska_track_combine_planes} },
564 CHILD_OF(matroska_track)
567 static EbmlSyntax matroska_track[] = {
568 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, 0, offsetof(MatroskaTrack, num) },
569 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTrack, name) },
570 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, uid) },
571 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrack, type) },
572 { MATROSKA_ID_CODECID, EBML_STR, 0, 0, offsetof(MatroskaTrack, codec_id) },
573 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrack, codec_priv) },
574 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, codec_delay), { .u = 0 } },
575 { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
576 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, 0, offsetof(MatroskaTrack, default_duration) },
577 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
578 { MATROSKA_ID_TRACKFLAGCOMMENTARY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_comment), { .u = 0 } },
579 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
580 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
581 { MATROSKA_ID_TRACKFLAGHEARINGIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_hearingimpaired), { .u = 0 } },
582 { MATROSKA_ID_TRACKFLAGVISUALIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_visualimpaired), { .u = 0 } },
583 { MATROSKA_ID_TRACKFLAGTEXTDESCRIPTIONS, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_textdescriptions), { .u = 0 } },
584 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
585 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
586 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
587 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, 0, { .n = matroska_track_encodings } },
588 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, max_block_additional_id), { .u = 0 } },
589 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, 0, offsetof(MatroskaTrack, seek_preroll), { .u = 0 } },
590 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
591 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
592 { MATROSKA_ID_CODECNAME, EBML_NONE },
593 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
594 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
595 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
596 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
597 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
598 CHILD_OF(matroska_tracks)
601 static EbmlSyntax matroska_tracks[] = {
602 { MATROSKA_ID_TRACKENTRY, EBML_NEST, 0, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
603 CHILD_OF(matroska_segment)
606 static EbmlSyntax matroska_attachment[] = {
607 { MATROSKA_ID_FILEUID, EBML_UINT, 0, 0, offsetof(MatroskaAttachment, uid) },
608 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, filename) },
609 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, 0, offsetof(MatroskaAttachment, mime) },
610 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, 0, offsetof(MatroskaAttachment, bin) },
611 { MATROSKA_ID_FILEDESC, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, description) },
612 CHILD_OF(matroska_attachments)
615 static EbmlSyntax matroska_attachments[] = {
616 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, 0, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
617 CHILD_OF(matroska_segment)
620 static EbmlSyntax matroska_chapter_display[] = {
621 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaChapter, title) },
622 { MATROSKA_ID_CHAPLANG, EBML_NONE },
623 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
624 CHILD_OF(matroska_chapter_entry)
627 static EbmlSyntax matroska_chapter_entry[] = {
628 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
629 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
630 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaChapter, uid) },
631 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter_display } },
632 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
633 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
634 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
635 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
636 CHILD_OF(matroska_chapter)
639 static EbmlSyntax matroska_chapter[] = {
640 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, 0, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
641 { MATROSKA_ID_EDITIONUID, EBML_NONE },
642 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
643 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
644 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
645 CHILD_OF(matroska_chapters)
648 static EbmlSyntax matroska_chapters[] = {
649 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter } },
650 CHILD_OF(matroska_segment)
653 static EbmlSyntax matroska_index_pos[] = {
654 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, track) },
655 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, pos) },
656 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
657 { MATROSKA_ID_CUEDURATION, EBML_NONE },
658 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
659 CHILD_OF(matroska_index_entry)
662 static EbmlSyntax matroska_index_entry[] = {
663 { MATROSKA_ID_CUETIME, EBML_UINT, 0, 0, offsetof(MatroskaIndex, time) },
664 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, 0, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
665 CHILD_OF(matroska_index)
668 static EbmlSyntax matroska_index[] = {
669 { MATROSKA_ID_POINTENTRY, EBML_NEST, 0, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
670 CHILD_OF(matroska_segment)
673 static EbmlSyntax matroska_simpletag[] = {
674 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTag, name) },
675 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaTag, string) },
676 { MATROSKA_ID_TAGLANG, EBML_STR, 0, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
677 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
678 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
679 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
680 CHILD_OF(matroska_tag)
683 static EbmlSyntax matroska_tagtargets[] = {
684 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, 0, offsetof(MatroskaTagTarget, type) },
685 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
686 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, trackuid), { .u = 0 } },
687 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, chapteruid), { .u = 0 } },
688 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, attachuid), { .u = 0 } },
689 CHILD_OF(matroska_tag)
692 static EbmlSyntax matroska_tag[] = {
693 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
694 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
695 CHILD_OF(matroska_tags)
698 static EbmlSyntax matroska_tags[] = {
699 { MATROSKA_ID_TAG, EBML_NEST, 0, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
700 CHILD_OF(matroska_segment)
703 static EbmlSyntax matroska_seekhead_entry[] = {
704 { MATROSKA_ID_SEEKID, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, id) },
705 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
706 CHILD_OF(matroska_seekhead)
709 static EbmlSyntax matroska_seekhead[] = {
710 { MATROSKA_ID_SEEKENTRY, EBML_NEST, 0, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
711 CHILD_OF(matroska_segment)
714 static EbmlSyntax matroska_segment[] = {
715 { MATROSKA_ID_CLUSTER, EBML_STOP },
716 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, 0, { .n = matroska_info } },
717 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tracks } },
718 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_attachments } },
719 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_chapters } },
720 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, 0, { .n = matroska_index } },
721 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tags } },
722 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, 0, { .n = matroska_seekhead } },
723 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
726 static EbmlSyntax matroska_segments[] = {
727 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, 0, { .n = matroska_segment } },
731 static EbmlSyntax matroska_blockmore[] = {
732 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
733 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, 0, offsetof(MatroskaBlock,additional) },
734 CHILD_OF(matroska_blockadditions)
737 static EbmlSyntax matroska_blockadditions[] = {
738 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, 0, {.n = matroska_blockmore} },
739 CHILD_OF(matroska_blockgroup)
742 static EbmlSyntax matroska_blockgroup[] = {
743 { MATROSKA_ID_BLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
744 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, 0, { .n = matroska_blockadditions} },
745 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, 0, offsetof(MatroskaBlock, duration) },
746 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, 0, offsetof(MatroskaBlock, discard_padding) },
747 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 1, 0, offsetof(MatroskaBlock, reference) },
748 { MATROSKA_ID_CODECSTATE, EBML_NONE },
749 { 1, EBML_UINT, 0, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
750 CHILD_OF(matroska_cluster_parsing)
753 // The following array contains SimpleBlock and BlockGroup twice
754 // in order to reuse the other values for matroska_cluster_enter.
755 static EbmlSyntax matroska_cluster_parsing[] = {
756 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
757 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, 0, { .n = matroska_blockgroup } },
758 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, 0, offsetof(MatroskaCluster, timecode) },
759 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
760 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
761 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
762 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
763 CHILD_OF(matroska_segment)
766 static EbmlSyntax matroska_cluster_enter[] = {
767 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, 0, { .n = &matroska_cluster_parsing[2] } },
772 static const CodecMime mkv_image_mime_tags[] = {
773 {"image/gif" , AV_CODEC_ID_GIF},
774 {"image/jpeg" , AV_CODEC_ID_MJPEG},
775 {"image/png" , AV_CODEC_ID_PNG},
776 {"image/tiff" , AV_CODEC_ID_TIFF},
778 {"" , AV_CODEC_ID_NONE}
781 static const CodecMime mkv_mime_tags[] = {
782 {"text/plain" , AV_CODEC_ID_TEXT},
783 {"application/x-truetype-font", AV_CODEC_ID_TTF},
784 {"application/x-font" , AV_CODEC_ID_TTF},
785 {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
786 {"binary" , AV_CODEC_ID_BIN_DATA},
788 {"" , AV_CODEC_ID_NONE}
791 static const char *const matroska_doctypes[] = { "matroska", "webm" };
793 static int matroska_read_close(AVFormatContext *s);
796 * This function prepares the status for parsing of level 1 elements.
798 static int matroska_reset_status(MatroskaDemuxContext *matroska,
799 uint32_t id, int64_t position)
802 int64_t err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
807 matroska->current_id = id;
808 matroska->num_levels = 1;
809 matroska->unknown_count = 0;
810 matroska->resync_pos = avio_tell(matroska->ctx->pb);
812 matroska->resync_pos -= (av_log2(id) + 7) / 8;
817 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
819 AVIOContext *pb = matroska->ctx->pb;
822 /* Try to seek to the last position to resync from. If this doesn't work,
823 * we resync from the earliest position available: The start of the buffer. */
824 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
825 av_log(matroska->ctx, AV_LOG_WARNING,
826 "Seek to desired resync point failed. Seeking to "
827 "earliest point available instead.\n");
828 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
829 last_pos + 1), SEEK_SET);
834 // try to find a toplevel element
835 while (!avio_feof(pb)) {
836 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
837 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
838 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
839 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
840 /* Prepare the context for parsing of a level 1 element. */
841 matroska_reset_status(matroska, id, -1);
842 /* Given that we are here means that an error has occurred,
843 * so treat the segment as unknown length in order not to
844 * discard valid data that happens to be beyond the designated
845 * end of the segment. */
846 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
849 id = (id << 8) | avio_r8(pb);
853 return pb->error ? pb->error : AVERROR_EOF;
857 * Read: an "EBML number", which is defined as a variable-length
858 * array of bytes. The first byte indicates the length by giving a
859 * number of 0-bits followed by a one. The position of the first
860 * "one" bit inside the first byte indicates the length of this
862 * Returns: number of bytes read, < 0 on error
864 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
865 int max_size, uint64_t *number, int eof_forbidden)
871 /* The first byte tells us the length in bytes - except when it is zero. */
876 /* get the length of the EBML number */
877 read = 8 - ff_log2_tab[total];
879 if (!total || read > max_size) {
880 pos = avio_tell(pb) - 1;
882 av_log(matroska->ctx, AV_LOG_ERROR,
883 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
884 "of an EBML number\n", pos, pos);
886 av_log(matroska->ctx, AV_LOG_ERROR,
887 "Length %d indicated by an EBML number's first byte 0x%02x "
888 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
889 read, (uint8_t) total, pos, pos, max_size);
891 return AVERROR_INVALIDDATA;
894 /* read out length */
895 total ^= 1 << ff_log2_tab[total];
897 total = (total << 8) | avio_r8(pb);
899 if (pb->eof_reached) {
911 av_log(matroska->ctx, AV_LOG_ERROR,
912 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
917 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
918 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
925 * Read a EBML length value.
926 * This needs special handling for the "unknown length" case which has multiple
929 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
932 int res = ebml_read_num(matroska, pb, 8, number, 1);
933 if (res > 0 && *number + 1 == 1ULL << (7 * res))
934 *number = EBML_UNKNOWN_LENGTH;
939 * Read the next element as an unsigned int.
940 * Returns NEEDS_CHECKING unless size == 0.
942 static int ebml_read_uint(AVIOContext *pb, int size,
943 uint64_t default_value, uint64_t *num)
948 *num = default_value;
951 /* big-endian ordering; build up number */
954 *num = (*num << 8) | avio_r8(pb);
956 return NEEDS_CHECKING;
960 * Read the next element as a signed int.
961 * Returns NEEDS_CHECKING unless size == 0.
963 static int ebml_read_sint(AVIOContext *pb, int size,
964 int64_t default_value, int64_t *num)
969 *num = default_value;
972 *num = sign_extend(avio_r8(pb), 8);
974 /* big-endian ordering; build up number */
976 *num = ((uint64_t)*num << 8) | avio_r8(pb);
979 return NEEDS_CHECKING;
983 * Read the next element as a float.
984 * Returns 0 if size == 0, NEEDS_CHECKING or < 0 on obvious failure.
986 static int ebml_read_float(AVIOContext *pb, int size,
987 double default_value, double *num)
990 *num = default_value;
992 } else if (size == 4) {
993 *num = av_int2float(avio_rb32(pb));
994 } else if (size == 8) {
995 *num = av_int2double(avio_rb64(pb));
997 return AVERROR_INVALIDDATA;
999 return NEEDS_CHECKING;
1003 * Read the next element as an ASCII string.
1004 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1006 static int ebml_read_ascii(AVIOContext *pb, int size,
1007 const char *default_value, char **str)
1012 if (size == 0 && default_value) {
1013 res = av_strdup(default_value);
1015 return AVERROR(ENOMEM);
1017 /* EBML strings are usually not 0-terminated, so we allocate one
1018 * byte more, read the string and NUL-terminate it ourselves. */
1019 if (!(res = av_malloc(size + 1)))
1020 return AVERROR(ENOMEM);
1021 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
1023 return ret < 0 ? ret : NEEDS_CHECKING;
1034 * Read the next element as binary data.
1035 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1037 static int ebml_read_binary(AVIOContext *pb, int length,
1038 int64_t pos, EbmlBin *bin)
1042 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1045 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1047 bin->data = bin->buf->data;
1050 if ((ret = avio_read(pb, bin->data, length)) != length) {
1051 av_buffer_unref(&bin->buf);
1054 return ret < 0 ? ret : NEEDS_CHECKING;
1061 * Read the next element, but only the header. The contents
1062 * are supposed to be sub-elements which can be read separately.
1063 * 0 is success, < 0 is failure.
1065 static int ebml_read_master(MatroskaDemuxContext *matroska,
1066 uint64_t length, int64_t pos)
1068 MatroskaLevel *level;
1070 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1071 av_log(matroska->ctx, AV_LOG_ERROR,
1072 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1073 return AVERROR(ENOSYS);
1076 level = &matroska->levels[matroska->num_levels++];
1078 level->length = length;
1084 * Read a signed "EBML number"
1085 * Return: number of bytes processed, < 0 on error
1087 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1088 AVIOContext *pb, int64_t *num)
1093 /* read as unsigned number first */
1094 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1097 /* make signed (weird way) */
1098 *num = unum - ((1LL << (7 * res - 1)) - 1);
1103 static int ebml_parse(MatroskaDemuxContext *matroska,
1104 EbmlSyntax *syntax, void *data);
1106 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1110 // Whoever touches this should be aware of the duplication
1111 // existing in matroska_cluster_parsing.
1112 for (i = 0; syntax[i].id; i++)
1113 if (id == syntax[i].id)
1119 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1125 for (int i = 0; syntax[i].id; i++) {
1126 void *dst = (char *)data + syntax[i].data_offset;
1127 switch (syntax[i].type) {
1129 *(uint64_t *)dst = syntax[i].def.u;
1132 *(int64_t *) dst = syntax[i].def.i;
1135 *(double *) dst = syntax[i].def.f;
1139 // the default may be NULL
1140 if (syntax[i].def.s) {
1141 *(char**)dst = av_strdup(syntax[i].def.s);
1143 return AVERROR(ENOMEM);
1149 if (!matroska->levels[matroska->num_levels - 1].length) {
1150 matroska->num_levels--;
1156 res = ebml_parse(matroska, syntax, data);
1159 return res == LEVEL_ENDED ? 0 : res;
1162 static int is_ebml_id_valid(uint32_t id)
1164 // Due to endian nonsense in Matroska, the highest byte with any bits set
1165 // will contain the leading length bit. This bit in turn identifies the
1166 // total byte length of the element by its position within the byte.
1167 unsigned int bits = av_log2(id);
1168 return id && (bits + 7) / 8 == (8 - bits % 8);
1172 * Allocate and return the entry for the level1 element with the given ID. If
1173 * an entry already exists, return the existing entry.
1175 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1176 uint32_t id, int64_t pos)
1179 MatroskaLevel1Element *elem;
1181 if (!is_ebml_id_valid(id))
1184 // Some files link to all clusters; useless.
1185 if (id == MATROSKA_ID_CLUSTER)
1188 // There can be multiple SeekHeads and Tags.
1189 for (i = 0; i < matroska->num_level1_elems; i++) {
1190 if (matroska->level1_elems[i].id == id) {
1191 if (matroska->level1_elems[i].pos == pos ||
1192 id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1193 return &matroska->level1_elems[i];
1197 // Only a completely broken file would have more elements.
1198 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1199 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1203 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1204 *elem = (MatroskaLevel1Element){.id = id};
1209 static int ebml_parse(MatroskaDemuxContext *matroska,
1210 EbmlSyntax *syntax, void *data)
1212 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1213 // Forbid unknown-length EBML_NONE elements.
1214 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1218 // max. 16 MB for strings
1219 [EBML_STR] = 0x1000000,
1220 [EBML_UTF8] = 0x1000000,
1221 // max. 256 MB for binary data
1222 [EBML_BIN] = 0x10000000,
1223 // no limits for anything else
1225 AVIOContext *pb = matroska->ctx->pb;
1228 int64_t pos = avio_tell(pb), pos_alt;
1229 int res, update_pos = 1, level_check;
1230 MatroskaLevel1Element *level1_elem;
1231 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1233 if (!matroska->current_id) {
1235 res = ebml_read_num(matroska, pb, 4, &id, 0);
1237 if (pb->eof_reached && res == AVERROR_EOF) {
1238 if (matroska->is_live)
1239 // in live mode, finish parsing if EOF is reached.
1241 if (level && pos == avio_tell(pb)) {
1242 if (level->length == EBML_UNKNOWN_LENGTH) {
1243 // Unknown-length levels automatically end at EOF.
1244 matroska->num_levels--;
1247 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1248 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1254 matroska->current_id = id | 1 << 7 * res;
1255 pos_alt = pos + res;
1258 pos -= (av_log2(matroska->current_id) + 7) / 8;
1261 id = matroska->current_id;
1263 syntax = ebml_parse_id(syntax, id);
1264 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1265 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1266 // Unknown-length levels end when an element from an upper level
1267 // in the hierarchy is encountered.
1268 while (syntax->def.n) {
1269 syntax = ebml_parse_id(syntax->def.n, id);
1271 matroska->num_levels--;
1277 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1278 "%"PRId64"\n", id, pos);
1279 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1283 data = (char *) data + syntax->data_offset;
1284 if (syntax->list_elem_size) {
1285 EbmlList *list = data;
1288 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1289 return AVERROR(ENOMEM);
1290 newelem = av_fast_realloc(list->elem,
1291 &list->alloc_elem_size,
1292 (list->nb_elem + 1) * syntax->list_elem_size);
1294 return AVERROR(ENOMEM);
1295 list->elem = newelem;
1296 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1297 memset(data, 0, syntax->list_elem_size);
1302 if (syntax->type != EBML_STOP) {
1303 matroska->current_id = 0;
1304 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1309 if (matroska->num_levels > 0) {
1310 if (length != EBML_UNKNOWN_LENGTH &&
1311 level->length != EBML_UNKNOWN_LENGTH) {
1312 uint64_t elem_end = pos_alt + length,
1313 level_end = level->start + level->length;
1315 if (elem_end < level_end) {
1317 } else if (elem_end == level_end) {
1318 level_check = LEVEL_ENDED;
1320 av_log(matroska->ctx, AV_LOG_ERROR,
1321 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1322 "containing master element ending at 0x%"PRIx64"\n",
1323 pos, elem_end, level_end);
1324 return AVERROR_INVALIDDATA;
1326 } else if (length != EBML_UNKNOWN_LENGTH) {
1328 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1329 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1330 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1331 return AVERROR_INVALIDDATA;
1334 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1335 || syntax->type == EBML_NEST)) {
1336 // According to the current specifications only clusters and
1337 // segments are allowed to be unknown-length. We also accept
1338 // other unknown-length master elements.
1339 av_log(matroska->ctx, AV_LOG_WARNING,
1340 "Found unknown-length element 0x%"PRIX32" other than "
1341 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1342 "parsing will nevertheless be attempted.\n", id, pos);
1349 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1350 if (length != EBML_UNKNOWN_LENGTH) {
1351 av_log(matroska->ctx, AV_LOG_ERROR,
1352 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1353 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1354 length, max_lengths[syntax->type], id, pos);
1355 } else if (syntax->type != EBML_NONE) {
1356 av_log(matroska->ctx, AV_LOG_ERROR,
1357 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1358 "unknown length, yet the length of an element of its "
1359 "type must be known.\n", id, pos);
1361 av_log(matroska->ctx, AV_LOG_ERROR,
1362 "Found unknown-length element with ID 0x%"PRIX32" at "
1363 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1364 "available.\n", id, pos);
1366 return AVERROR_INVALIDDATA;
1369 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1370 // Loosing sync will likely manifest itself as encountering unknown
1371 // elements which are not reliably distinguishable from elements
1372 // belonging to future extensions of the format.
1373 // We use a heuristic to detect such situations: If the current
1374 // element is not expected at the current syntax level and there
1375 // were only a few unknown elements in a row, then the element is
1376 // skipped or considered defective based upon the length of the
1377 // current element (i.e. how much would be skipped); if there were
1378 // more than a few skipped elements in a row and skipping the current
1379 // element would lead us more than SKIP_THRESHOLD away from the last
1380 // known good position, then it is inferred that an error occurred.
1381 // The dependency on the number of unknown elements in a row exists
1382 // because the distance to the last known good position is
1383 // automatically big if the last parsed element was big.
1384 // In both cases, each unknown element is considered equivalent to
1385 // UNKNOWN_EQUIV of skipped bytes for the check.
1386 // The whole check is only done for non-seekable output, because
1387 // in this situation skipped data can't simply be rechecked later.
1388 // This is especially important when using unkown length elements
1389 // as the check for whether a child exceeds its containing master
1390 // element is not effective in this situation.
1392 matroska->unknown_count = 0;
1394 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1396 if (matroska->unknown_count > 3)
1397 dist += pos_alt - matroska->resync_pos;
1399 if (dist > SKIP_THRESHOLD) {
1400 av_log(matroska->ctx, AV_LOG_ERROR,
1401 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1402 "length 0x%"PRIx64" considered as invalid data. Last "
1403 "known good position 0x%"PRIx64", %d unknown elements"
1404 " in a row\n", id, pos, length, matroska->resync_pos,
1405 matroska->unknown_count);
1406 return AVERROR_INVALIDDATA;
1411 if (update_pos > 0) {
1412 // We have found an element that is allowed at this place
1413 // in the hierarchy and it passed all checks, so treat the beginning
1414 // of the element as the "last known good" position.
1415 matroska->resync_pos = pos;
1418 if (!data && length != EBML_UNKNOWN_LENGTH)
1422 switch (syntax->type) {
1424 res = ebml_read_uint(pb, length, syntax->def.u, data);
1427 res = ebml_read_sint(pb, length, syntax->def.i, data);
1430 res = ebml_read_float(pb, length, syntax->def.f, data);
1434 res = ebml_read_ascii(pb, length, syntax->def.s, data);
1437 res = ebml_read_binary(pb, length, pos_alt, data);
1441 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1443 if (id == MATROSKA_ID_SEGMENT)
1444 matroska->segment_start = pos_alt;
1445 if (id == MATROSKA_ID_CUES)
1446 matroska->cues_parsing_deferred = 0;
1447 if (syntax->type == EBML_LEVEL1 &&
1448 (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1449 if (!level1_elem->pos) {
1450 // Zero is not a valid position for a level 1 element.
1451 level1_elem->pos = pos;
1452 } else if (level1_elem->pos != pos)
1453 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1454 level1_elem->parsed = 1;
1456 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1465 if (ffio_limit(pb, length) != length) {
1466 // ffio_limit emits its own error message,
1467 // so we don't have to.
1468 return AVERROR(EIO);
1470 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1471 // avio_skip might take us past EOF. We check for this
1472 // by skipping only length - 1 bytes, reading a byte and
1473 // checking the error flags. This is done in order to check
1474 // that the element has been properly skipped even when
1475 // no filesize (that ffio_limit relies on) is available.
1477 res = NEEDS_CHECKING;
1484 if (res == NEEDS_CHECKING) {
1485 if (pb->eof_reached) {
1494 if (res == AVERROR_INVALIDDATA)
1495 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1496 else if (res == AVERROR(EIO))
1497 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1498 else if (res == AVERROR_EOF) {
1499 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1507 if (syntax->is_counted && data) {
1508 CountedElement *elem = data;
1509 if (elem->count != UINT_MAX)
1513 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1514 level = &matroska->levels[matroska->num_levels - 1];
1515 pos = avio_tell(pb);
1517 // Given that pos >= level->start no check for
1518 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1519 while (matroska->num_levels && pos == level->start + level->length) {
1520 matroska->num_levels--;
1528 static void ebml_free(EbmlSyntax *syntax, void *data)
1531 for (i = 0; syntax[i].id; i++) {
1532 void *data_off = (char *) data + syntax[i].data_offset;
1533 switch (syntax[i].type) {
1539 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1543 if (syntax[i].list_elem_size) {
1544 EbmlList *list = data_off;
1545 char *ptr = list->elem;
1546 for (j = 0; j < list->nb_elem;
1547 j++, ptr += syntax[i].list_elem_size)
1548 ebml_free(syntax[i].def.n, ptr);
1549 av_freep(&list->elem);
1551 list->alloc_elem_size = 0;
1553 ebml_free(syntax[i].def.n, data_off);
1563 static int matroska_probe(const AVProbeData *p)
1566 int len_mask = 0x80, size = 1, n = 1, i;
1569 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1572 /* length of header */
1574 while (size <= 8 && !(total & len_mask)) {
1580 total &= (len_mask - 1);
1582 total = (total << 8) | p->buf[4 + n++];
1584 if (total + 1 == 1ULL << (7 * size)){
1585 /* Unknown-length header - simply parse the whole buffer. */
1586 total = p->buf_size - 4 - size;
1588 /* Does the probe data contain the whole header? */
1589 if (p->buf_size < 4 + size + total)
1593 /* The header should contain a known document type. For now,
1594 * we don't parse the whole header but simply check for the
1595 * availability of that array of characters inside the header.
1596 * Not fully fool-proof, but good enough. */
1597 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1598 size_t probelen = strlen(matroska_doctypes[i]);
1599 if (total < probelen)
1601 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1602 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1603 return AVPROBE_SCORE_MAX;
1606 // probably valid EBML header but no recognized doctype
1607 return AVPROBE_SCORE_EXTENSION;
1610 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1613 MatroskaTrack *tracks = matroska->tracks.elem;
1616 for (i = 0; i < matroska->tracks.nb_elem; i++)
1617 if (tracks[i].num == num)
1620 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1624 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1625 MatroskaTrack *track)
1627 MatroskaTrackEncoding *encodings = track->encodings.elem;
1628 uint8_t *data = *buf;
1629 int isize = *buf_size;
1630 uint8_t *pkt_data = NULL;
1631 uint8_t av_unused *newpktdata;
1632 int pkt_size = isize;
1636 if (pkt_size >= 10000000U)
1637 return AVERROR_INVALIDDATA;
1639 switch (encodings[0].compression.algo) {
1640 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1642 int header_size = encodings[0].compression.settings.size;
1643 uint8_t *header = encodings[0].compression.settings.data;
1645 if (header_size && !header) {
1646 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1653 pkt_size = isize + header_size;
1654 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1656 return AVERROR(ENOMEM);
1658 memcpy(pkt_data, header, header_size);
1659 memcpy(pkt_data + header_size, data, isize);
1663 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1666 olen = pkt_size *= 3;
1667 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1668 + AV_INPUT_BUFFER_PADDING_SIZE);
1670 result = AVERROR(ENOMEM);
1673 pkt_data = newpktdata;
1674 result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1675 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1677 result = AVERROR_INVALIDDATA;
1684 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1686 z_stream zstream = { 0 };
1687 if (inflateInit(&zstream) != Z_OK)
1689 zstream.next_in = data;
1690 zstream.avail_in = isize;
1693 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1695 inflateEnd(&zstream);
1696 result = AVERROR(ENOMEM);
1699 pkt_data = newpktdata;
1700 zstream.avail_out = pkt_size - zstream.total_out;
1701 zstream.next_out = pkt_data + zstream.total_out;
1702 result = inflate(&zstream, Z_NO_FLUSH);
1703 } while (result == Z_OK && pkt_size < 10000000);
1704 pkt_size = zstream.total_out;
1705 inflateEnd(&zstream);
1706 if (result != Z_STREAM_END) {
1707 if (result == Z_MEM_ERROR)
1708 result = AVERROR(ENOMEM);
1710 result = AVERROR_INVALIDDATA;
1717 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1719 bz_stream bzstream = { 0 };
1720 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1722 bzstream.next_in = data;
1723 bzstream.avail_in = isize;
1726 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1728 BZ2_bzDecompressEnd(&bzstream);
1729 result = AVERROR(ENOMEM);
1732 pkt_data = newpktdata;
1733 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1734 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1735 result = BZ2_bzDecompress(&bzstream);
1736 } while (result == BZ_OK && pkt_size < 10000000);
1737 pkt_size = bzstream.total_out_lo32;
1738 BZ2_bzDecompressEnd(&bzstream);
1739 if (result != BZ_STREAM_END) {
1740 if (result == BZ_MEM_ERROR)
1741 result = AVERROR(ENOMEM);
1743 result = AVERROR_INVALIDDATA;
1750 return AVERROR_INVALIDDATA;
1753 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1756 *buf_size = pkt_size;
1764 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1765 AVDictionary **metadata, char *prefix)
1767 MatroskaTag *tags = list->elem;
1771 for (i = 0; i < list->nb_elem; i++) {
1772 const char *lang = tags[i].lang &&
1773 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1775 if (!tags[i].name) {
1776 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1780 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1782 av_strlcpy(key, tags[i].name, sizeof(key));
1783 if (tags[i].def || !lang) {
1784 av_dict_set(metadata, key, tags[i].string, 0);
1785 if (tags[i].sub.nb_elem)
1786 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1789 av_strlcat(key, "-", sizeof(key));
1790 av_strlcat(key, lang, sizeof(key));
1791 av_dict_set(metadata, key, tags[i].string, 0);
1792 if (tags[i].sub.nb_elem)
1793 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1796 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1799 static void matroska_convert_tags(AVFormatContext *s)
1801 MatroskaDemuxContext *matroska = s->priv_data;
1802 MatroskaTags *tags = matroska->tags.elem;
1805 for (i = 0; i < matroska->tags.nb_elem; i++) {
1806 if (tags[i].target.attachuid) {
1807 MatroskaAttachment *attachment = matroska->attachments.elem;
1809 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1810 if (attachment[j].uid == tags[i].target.attachuid &&
1811 attachment[j].stream) {
1812 matroska_convert_tag(s, &tags[i].tag,
1813 &attachment[j].stream->metadata, NULL);
1818 av_log(s, AV_LOG_WARNING,
1819 "The tags at index %d refer to a "
1820 "non-existent attachment %"PRId64".\n",
1821 i, tags[i].target.attachuid);
1823 } else if (tags[i].target.chapteruid) {
1824 MatroskaChapter *chapter = matroska->chapters.elem;
1826 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1827 if (chapter[j].uid == tags[i].target.chapteruid &&
1828 chapter[j].chapter) {
1829 matroska_convert_tag(s, &tags[i].tag,
1830 &chapter[j].chapter->metadata, NULL);
1835 av_log(s, AV_LOG_WARNING,
1836 "The tags at index %d refer to a non-existent chapter "
1838 i, tags[i].target.chapteruid);
1840 } else if (tags[i].target.trackuid) {
1841 MatroskaTrack *track = matroska->tracks.elem;
1843 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1844 if (track[j].uid == tags[i].target.trackuid &&
1846 matroska_convert_tag(s, &tags[i].tag,
1847 &track[j].stream->metadata, NULL);
1852 av_log(s, AV_LOG_WARNING,
1853 "The tags at index %d refer to a non-existent track "
1855 i, tags[i].target.trackuid);
1858 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1859 tags[i].target.type);
1864 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1867 uint32_t saved_id = matroska->current_id;
1868 int64_t before_pos = avio_tell(matroska->ctx->pb);
1872 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1873 /* We don't want to lose our seekhead level, so we add
1874 * a dummy. This is a crude hack. */
1875 if (matroska->num_levels == EBML_MAX_DEPTH) {
1876 av_log(matroska->ctx, AV_LOG_INFO,
1877 "Max EBML element depth (%d) reached, "
1878 "cannot parse further.\n", EBML_MAX_DEPTH);
1879 ret = AVERROR_INVALIDDATA;
1881 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1882 matroska->num_levels++;
1883 matroska->current_id = 0;
1885 ret = ebml_parse(matroska, matroska_segment, matroska);
1886 if (ret == LEVEL_ENDED) {
1887 /* This can only happen if the seek brought us beyond EOF. */
1892 /* Seek back - notice that in all instances where this is used
1893 * it is safe to set the level to 1. */
1894 matroska_reset_status(matroska, saved_id, before_pos);
1899 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1901 EbmlList *seekhead_list = &matroska->seekhead;
1904 // we should not do any seeking in the streaming case
1905 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1908 for (i = 0; i < seekhead_list->nb_elem; i++) {
1909 MatroskaSeekhead *seekheads = seekhead_list->elem;
1910 uint32_t id = seekheads[i].id;
1911 int64_t pos = seekheads[i].pos + matroska->segment_start;
1912 MatroskaLevel1Element *elem;
1914 if (id != seekheads[i].id || pos < matroska->segment_start)
1917 elem = matroska_find_level1_elem(matroska, id, pos);
1918 if (!elem || elem->parsed)
1923 // defer cues parsing until we actually need cue data.
1924 if (id == MATROSKA_ID_CUES)
1927 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1928 // mark index as broken
1929 matroska->cues_parsing_deferred = -1;
1937 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1939 EbmlList *index_list;
1940 MatroskaIndex *index;
1941 uint64_t index_scale = 1;
1944 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1947 index_list = &matroska->index;
1948 index = index_list->elem;
1949 if (index_list->nb_elem < 2)
1951 if (index[1].time > 1E14 / matroska->time_scale) {
1952 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1955 for (i = 0; i < index_list->nb_elem; i++) {
1956 EbmlList *pos_list = &index[i].pos;
1957 MatroskaIndexPos *pos = pos_list->elem;
1958 for (j = 0; j < pos_list->nb_elem; j++) {
1959 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1961 if (track && track->stream)
1962 av_add_index_entry(track->stream,
1963 pos[j].pos + matroska->segment_start,
1964 index[i].time / index_scale, 0, 0,
1970 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1973 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1976 for (i = 0; i < matroska->num_level1_elems; i++) {
1977 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1978 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1979 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1980 matroska->cues_parsing_deferred = -1;
1986 matroska_add_index_entries(matroska);
1989 static int matroska_aac_profile(char *codec_id)
1991 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1994 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1995 if (strstr(codec_id, aac_profiles[profile]))
2000 static int matroska_aac_sri(int samplerate)
2004 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
2005 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
2010 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
2012 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
2013 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
2016 static int matroska_parse_flac(AVFormatContext *s,
2017 MatroskaTrack *track,
2020 AVStream *st = track->stream;
2021 uint8_t *p = track->codec_priv.data;
2022 int size = track->codec_priv.size;
2024 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
2025 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
2026 track->codec_priv.size = 0;
2030 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
2032 p += track->codec_priv.size;
2033 size -= track->codec_priv.size;
2035 /* parse the remaining metadata blocks if present */
2037 int block_last, block_type, block_size;
2039 flac_parse_block_header(p, &block_last, &block_type, &block_size);
2043 if (block_size > size)
2046 /* check for the channel mask */
2047 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2048 AVDictionary *dict = NULL;
2049 AVDictionaryEntry *chmask;
2051 ff_vorbis_comment(s, &dict, p, block_size, 0);
2052 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2054 uint64_t mask = strtol(chmask->value, NULL, 0);
2055 if (!mask || mask & ~0x3ffffULL) {
2056 av_log(s, AV_LOG_WARNING,
2057 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2059 st->codecpar->channel_layout = mask;
2061 av_dict_free(&dict);
2071 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2073 int minor, micro, bttb = 0;
2075 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2076 * this function, and fixed in 57.52 */
2077 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, µ) == 2)
2078 bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2080 switch (field_order) {
2081 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2082 return AV_FIELD_PROGRESSIVE;
2083 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2084 return AV_FIELD_UNKNOWN;
2085 case MATROSKA_VIDEO_FIELDORDER_TT:
2087 case MATROSKA_VIDEO_FIELDORDER_BB:
2089 case MATROSKA_VIDEO_FIELDORDER_BT:
2090 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2091 case MATROSKA_VIDEO_FIELDORDER_TB:
2092 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2094 return AV_FIELD_UNKNOWN;
2098 static void mkv_stereo_mode_display_mul(int stereo_mode,
2099 int *h_width, int *h_height)
2101 switch (stereo_mode) {
2102 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2103 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2104 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2105 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2106 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2108 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2109 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2110 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2111 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2114 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2115 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2116 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2117 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2123 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2124 const MatroskaTrackVideoColor *color = track->video.color.elem;
2125 const MatroskaMasteringMeta *mastering_meta;
2126 int has_mastering_primaries, has_mastering_luminance;
2128 if (!track->video.color.nb_elem)
2131 mastering_meta = &color->mastering_meta;
2132 // Mastering primaries are CIE 1931 coords, and must be > 0.
2133 has_mastering_primaries =
2134 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2135 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2136 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2137 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2138 has_mastering_luminance = mastering_meta->max_luminance >
2139 mastering_meta->min_luminance.el.f &&
2140 mastering_meta->min_luminance.el.f >= 0 &&
2141 mastering_meta->min_luminance.count;
2143 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2144 st->codecpar->color_space = color->matrix_coefficients;
2145 if (color->primaries != AVCOL_PRI_RESERVED &&
2146 color->primaries != AVCOL_PRI_RESERVED0)
2147 st->codecpar->color_primaries = color->primaries;
2148 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2149 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2150 st->codecpar->color_trc = color->transfer_characteristics;
2151 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2152 color->range <= AVCOL_RANGE_JPEG)
2153 st->codecpar->color_range = color->range;
2154 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2155 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2156 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2157 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2158 st->codecpar->chroma_location =
2159 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2160 (color->chroma_siting_vert - 1) << 7);
2162 if (color->max_cll && color->max_fall) {
2165 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2167 return AVERROR(ENOMEM);
2168 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2169 (uint8_t *)metadata, size);
2171 av_freep(&metadata);
2174 metadata->MaxCLL = color->max_cll;
2175 metadata->MaxFALL = color->max_fall;
2178 if (has_mastering_primaries || has_mastering_luminance) {
2179 AVMasteringDisplayMetadata *metadata =
2180 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2181 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2182 sizeof(AVMasteringDisplayMetadata));
2184 return AVERROR(ENOMEM);
2186 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2187 if (has_mastering_primaries) {
2188 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2189 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2190 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2191 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2192 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2193 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2194 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2195 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2196 metadata->has_primaries = 1;
2198 if (has_mastering_luminance) {
2199 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2200 metadata->min_luminance = av_d2q(mastering_meta->min_luminance.el.f, INT_MAX);
2201 metadata->has_luminance = 1;
2207 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2210 AVSphericalMapping *spherical;
2211 const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2212 const uint8_t *priv_data = mkv_projection->private.data;
2213 enum AVSphericalProjection projection;
2214 size_t spherical_size;
2215 uint32_t l = 0, t = 0, r = 0, b = 0;
2216 uint32_t padding = 0;
2219 if (mkv_projection->private.size && priv_data[0] != 0) {
2220 av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2224 switch (track->video.projection.type) {
2225 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2226 if (track->video.projection.private.size == 20) {
2227 t = AV_RB32(priv_data + 4);
2228 b = AV_RB32(priv_data + 8);
2229 l = AV_RB32(priv_data + 12);
2230 r = AV_RB32(priv_data + 16);
2232 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2233 av_log(logctx, AV_LOG_ERROR,
2234 "Invalid bounding rectangle coordinates "
2235 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2237 return AVERROR_INVALIDDATA;
2239 } else if (track->video.projection.private.size != 0) {
2240 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2241 return AVERROR_INVALIDDATA;
2244 if (l || t || r || b)
2245 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2247 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2249 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2250 if (track->video.projection.private.size < 4) {
2251 av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2252 return AVERROR_INVALIDDATA;
2253 } else if (track->video.projection.private.size == 12) {
2254 uint32_t layout = AV_RB32(priv_data + 4);
2256 av_log(logctx, AV_LOG_WARNING,
2257 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2260 projection = AV_SPHERICAL_CUBEMAP;
2261 padding = AV_RB32(priv_data + 8);
2263 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2264 return AVERROR_INVALIDDATA;
2267 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2268 /* No Spherical metadata */
2271 av_log(logctx, AV_LOG_WARNING,
2272 "Unknown spherical metadata type %"PRIu64"\n",
2273 track->video.projection.type);
2277 spherical = av_spherical_alloc(&spherical_size);
2279 return AVERROR(ENOMEM);
2281 spherical->projection = projection;
2283 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2284 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2285 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2287 spherical->padding = padding;
2289 spherical->bound_left = l;
2290 spherical->bound_top = t;
2291 spherical->bound_right = r;
2292 spherical->bound_bottom = b;
2294 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2297 av_freep(&spherical);
2304 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2306 const AVCodecTag *codec_tags;
2308 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2309 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2311 /* Normalize noncompliant private data that starts with the fourcc
2312 * by expanding/shifting the data by 4 bytes and storing the data
2313 * size at the start. */
2314 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2315 int ret = av_buffer_realloc(&track->codec_priv.buf,
2316 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2320 track->codec_priv.data = track->codec_priv.buf->data;
2321 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2322 track->codec_priv.size += 4;
2323 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2326 *fourcc = AV_RL32(track->codec_priv.data + 4);
2327 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2332 static int matroska_parse_tracks(AVFormatContext *s)
2334 MatroskaDemuxContext *matroska = s->priv_data;
2335 MatroskaTrack *tracks = matroska->tracks.elem;
2340 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2341 MatroskaTrack *track = &tracks[i];
2342 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2343 EbmlList *encodings_list = &track->encodings;
2344 MatroskaTrackEncoding *encodings = encodings_list->elem;
2345 uint8_t *extradata = NULL;
2346 int extradata_size = 0;
2347 int extradata_offset = 0;
2348 uint32_t fourcc = 0;
2350 char* key_id_base64 = NULL;
2353 /* Apply some sanity checks. */
2354 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2355 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2356 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2357 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2358 av_log(matroska->ctx, AV_LOG_INFO,
2359 "Unknown or unsupported track type %"PRIu64"\n",
2363 if (!track->codec_id)
2366 if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
2367 || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
2368 || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2369 || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2371 av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
2375 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2376 isnan(track->audio.samplerate)) {
2377 av_log(matroska->ctx, AV_LOG_WARNING,
2378 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2379 track->audio.samplerate);
2380 track->audio.samplerate = 8000;
2383 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2384 if (!track->default_duration && track->video.frame_rate > 0) {
2385 double default_duration = 1000000000 / track->video.frame_rate;
2386 if (default_duration > UINT64_MAX || default_duration < 0) {
2387 av_log(matroska->ctx, AV_LOG_WARNING,
2388 "Invalid frame rate %e. Cannot calculate default duration.\n",
2389 track->video.frame_rate);
2391 track->default_duration = default_duration;
2394 if (track->video.display_width == -1)
2395 track->video.display_width = track->video.pixel_width;
2396 if (track->video.display_height == -1)
2397 track->video.display_height = track->video.pixel_height;
2398 if (track->video.color_space.size == 4)
2399 fourcc = AV_RL32(track->video.color_space.data);
2400 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2401 if (!track->audio.out_samplerate)
2402 track->audio.out_samplerate = track->audio.samplerate;
2404 if (encodings_list->nb_elem > 1) {
2405 av_log(matroska->ctx, AV_LOG_ERROR,
2406 "Multiple combined encodings not supported");
2407 } else if (encodings_list->nb_elem == 1) {
2408 if (encodings[0].type) {
2409 if (encodings[0].encryption.key_id.size > 0) {
2410 /* Save the encryption key id to be stored later as a
2412 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2413 key_id_base64 = av_malloc(b64_size);
2414 if (key_id_base64 == NULL)
2415 return AVERROR(ENOMEM);
2417 av_base64_encode(key_id_base64, b64_size,
2418 encodings[0].encryption.key_id.data,
2419 encodings[0].encryption.key_id.size);
2421 encodings[0].scope = 0;
2422 av_log(matroska->ctx, AV_LOG_ERROR,
2423 "Unsupported encoding type");
2427 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2430 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2433 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2435 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2436 encodings[0].scope = 0;
2437 av_log(matroska->ctx, AV_LOG_ERROR,
2438 "Unsupported encoding type");
2439 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2440 uint8_t *codec_priv = track->codec_priv.data;
2441 int ret = matroska_decode_buffer(&track->codec_priv.data,
2442 &track->codec_priv.size,
2445 track->codec_priv.data = NULL;
2446 track->codec_priv.size = 0;
2447 av_log(matroska->ctx, AV_LOG_ERROR,
2448 "Failed to decode codec private data\n");
2451 if (codec_priv != track->codec_priv.data) {
2452 av_buffer_unref(&track->codec_priv.buf);
2453 if (track->codec_priv.data) {
2454 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2455 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2457 if (!track->codec_priv.buf) {
2458 av_freep(&track->codec_priv.data);
2459 track->codec_priv.size = 0;
2460 return AVERROR(ENOMEM);
2466 track->needs_decoding = encodings && !encodings[0].type &&
2467 encodings[0].scope & 1 &&
2468 (encodings[0].compression.algo !=
2469 MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP ||
2470 encodings[0].compression.settings.size);
2472 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2473 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2474 strlen(ff_mkv_codec_tags[j].str))) {
2475 codec_id = ff_mkv_codec_tags[j].id;
2480 st = track->stream = avformat_new_stream(s, NULL);
2482 av_free(key_id_base64);
2483 return AVERROR(ENOMEM);
2486 if (key_id_base64) {
2487 /* export encryption key id as base64 metadata tag */
2488 av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2489 AV_DICT_DONT_STRDUP_VAL);
2492 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2493 track->codec_priv.size >= 40 &&
2494 track->codec_priv.data) {
2495 track->ms_compat = 1;
2496 bit_depth = AV_RL16(track->codec_priv.data + 14);
2497 fourcc = AV_RL32(track->codec_priv.data + 16);
2498 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2501 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2503 extradata_offset = 40;
2504 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2505 track->codec_priv.size >= 14 &&
2506 track->codec_priv.data) {
2508 ffio_init_context(&b, track->codec_priv.data,
2509 track->codec_priv.size,
2510 0, NULL, NULL, NULL, NULL);
2511 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2514 codec_id = st->codecpar->codec_id;
2515 fourcc = st->codecpar->codec_tag;
2516 extradata_offset = FFMIN(track->codec_priv.size, 18);
2517 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2518 /* Normally 36, but allow noncompliant private data */
2519 && (track->codec_priv.size >= 32)
2520 && (track->codec_priv.data)) {
2521 uint16_t sample_size;
2522 int ret = get_qt_codec(track, &fourcc, &codec_id);
2525 sample_size = AV_RB16(track->codec_priv.data + 26);
2527 if (sample_size == 8) {
2528 fourcc = MKTAG('r','a','w',' ');
2529 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2530 } else if (sample_size == 16) {
2531 fourcc = MKTAG('t','w','o','s');
2532 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2535 if ((fourcc == MKTAG('t','w','o','s') ||
2536 fourcc == MKTAG('s','o','w','t')) &&
2538 codec_id = AV_CODEC_ID_PCM_S8;
2539 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2540 (track->codec_priv.size >= 21) &&
2541 (track->codec_priv.data)) {
2542 int ret = get_qt_codec(track, &fourcc, &codec_id);
2545 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2546 fourcc = MKTAG('S','V','Q','3');
2547 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2549 if (codec_id == AV_CODEC_ID_NONE)
2550 av_log(matroska->ctx, AV_LOG_ERROR,
2551 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2552 if (track->codec_priv.size >= 86) {
2553 bit_depth = AV_RB16(track->codec_priv.data + 82);
2554 ffio_init_context(&b, track->codec_priv.data,
2555 track->codec_priv.size,
2556 0, NULL, NULL, NULL, NULL);
2557 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2559 track->has_palette = 1;
2562 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2563 switch (track->audio.bitdepth) {
2565 codec_id = AV_CODEC_ID_PCM_U8;
2568 codec_id = AV_CODEC_ID_PCM_S24BE;
2571 codec_id = AV_CODEC_ID_PCM_S32BE;
2574 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2575 switch (track->audio.bitdepth) {
2577 codec_id = AV_CODEC_ID_PCM_U8;
2580 codec_id = AV_CODEC_ID_PCM_S24LE;
2583 codec_id = AV_CODEC_ID_PCM_S32LE;
2586 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2587 track->audio.bitdepth == 64) {
2588 codec_id = AV_CODEC_ID_PCM_F64LE;
2589 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2590 int profile = matroska_aac_profile(track->codec_id);
2591 int sri = matroska_aac_sri(track->audio.samplerate);
2592 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2594 return AVERROR(ENOMEM);
2595 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2596 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2597 if (strstr(track->codec_id, "SBR")) {
2598 sri = matroska_aac_sri(track->audio.out_samplerate);
2599 extradata[2] = 0x56;
2600 extradata[3] = 0xE5;
2601 extradata[4] = 0x80 | (sri << 3);
2605 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2606 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2607 * Create the "atom size", "tag", and "tag version" fields the
2608 * decoder expects manually. */
2609 extradata_size = 12 + track->codec_priv.size;
2610 extradata = av_mallocz(extradata_size +
2611 AV_INPUT_BUFFER_PADDING_SIZE);
2613 return AVERROR(ENOMEM);
2614 AV_WB32(extradata, extradata_size);
2615 memcpy(&extradata[4], "alac", 4);
2616 AV_WB32(&extradata[8], 0);
2617 memcpy(&extradata[12], track->codec_priv.data,
2618 track->codec_priv.size);
2619 } else if (codec_id == AV_CODEC_ID_TTA) {
2621 if (track->audio.channels > UINT16_MAX ||
2622 track->audio.bitdepth > UINT16_MAX) {
2623 av_log(matroska->ctx, AV_LOG_WARNING,
2624 "Too large audio channel number %"PRIu64
2625 " or bitdepth %"PRIu64". Skipping track.\n",
2626 track->audio.channels, track->audio.bitdepth);
2627 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2628 return AVERROR_INVALIDDATA;
2632 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2633 return AVERROR_INVALIDDATA;
2634 extradata_size = 22;
2635 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2637 return AVERROR(ENOMEM);
2639 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2640 bytestream_put_le16(&ptr, 1);
2641 bytestream_put_le16(&ptr, track->audio.channels);
2642 bytestream_put_le16(&ptr, track->audio.bitdepth);
2643 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2644 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2645 track->audio.out_samplerate,
2646 AV_TIME_BASE * 1000));
2647 } else if (codec_id == AV_CODEC_ID_RV10 ||
2648 codec_id == AV_CODEC_ID_RV20 ||
2649 codec_id == AV_CODEC_ID_RV30 ||
2650 codec_id == AV_CODEC_ID_RV40) {
2651 extradata_offset = 26;
2652 } else if (codec_id == AV_CODEC_ID_RA_144) {
2653 track->audio.out_samplerate = 8000;
2654 track->audio.channels = 1;
2655 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2656 codec_id == AV_CODEC_ID_COOK ||
2657 codec_id == AV_CODEC_ID_ATRAC3 ||
2658 codec_id == AV_CODEC_ID_SIPR)
2659 && track->codec_priv.data) {
2662 ffio_init_context(&b, track->codec_priv.data,
2663 track->codec_priv.size,
2664 0, NULL, NULL, NULL, NULL);
2666 flavor = avio_rb16(&b);
2667 track->audio.coded_framesize = avio_rb32(&b);
2669 track->audio.sub_packet_h = avio_rb16(&b);
2670 track->audio.frame_size = avio_rb16(&b);
2671 track->audio.sub_packet_size = avio_rb16(&b);
2672 if (track->audio.coded_framesize <= 0 ||
2673 track->audio.sub_packet_h <= 0 ||
2674 track->audio.frame_size <= 0)
2675 return AVERROR_INVALIDDATA;
2677 if (codec_id == AV_CODEC_ID_RA_288) {
2678 if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2679 != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2680 return AVERROR_INVALIDDATA;
2681 st->codecpar->block_align = track->audio.coded_framesize;
2682 track->codec_priv.size = 0;
2684 if (codec_id == AV_CODEC_ID_SIPR) {
2685 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2687 return AVERROR_INVALIDDATA;
2688 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2689 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2690 } else if (track->audio.sub_packet_size <= 0 ||
2691 track->audio.frame_size % track->audio.sub_packet_size)
2692 return AVERROR_INVALIDDATA;
2693 st->codecpar->block_align = track->audio.sub_packet_size;
2694 extradata_offset = 78;
2696 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2697 track->audio.frame_size);
2698 if (!track->audio.buf)
2699 return AVERROR(ENOMEM);
2700 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2701 ret = matroska_parse_flac(s, track, &extradata_offset);
2704 } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2705 av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2706 "in absence of valid CodecPrivate.\n");
2708 extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2710 return AVERROR(ENOMEM);
2711 AV_WL16(extradata, 0x410);
2712 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2713 fourcc = AV_RL32(track->codec_priv.data);
2714 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2715 /* we don't need any value stored in CodecPrivate.
2716 make sure that it's not exported as extradata. */
2717 track->codec_priv.size = 0;
2718 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2719 /* For now, propagate only the OBUs, if any. Once libavcodec is
2720 updated to handle isobmff style extradata this can be removed. */
2721 extradata_offset = 4;
2723 track->codec_priv.size -= extradata_offset;
2725 if (codec_id == AV_CODEC_ID_NONE)
2726 av_log(matroska->ctx, AV_LOG_INFO,
2727 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2729 if (track->time_scale < 0.01) {
2730 av_log(matroska->ctx, AV_LOG_WARNING,
2731 "Track TimestampScale too small %f, assuming 1.0.\n",
2733 track->time_scale = 1.0;
2735 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2736 1000 * 1000 * 1000); /* 64 bit pts in ns */
2738 /* convert the delay from ns to the track timebase */
2739 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2740 (AVRational){ 1, 1000000000 },
2743 st->codecpar->codec_id = codec_id;
2745 if (strcmp(track->language, "und"))
2746 av_dict_set(&st->metadata, "language", track->language, 0);
2747 av_dict_set(&st->metadata, "title", track->name, 0);
2749 if (track->flag_default)
2750 st->disposition |= AV_DISPOSITION_DEFAULT;
2751 if (track->flag_forced)
2752 st->disposition |= AV_DISPOSITION_FORCED;
2753 if (track->flag_comment)
2754 st->disposition |= AV_DISPOSITION_COMMENT;
2755 if (track->flag_hearingimpaired)
2756 st->disposition |= AV_DISPOSITION_HEARING_IMPAIRED;
2757 if (track->flag_visualimpaired)
2758 st->disposition |= AV_DISPOSITION_VISUAL_IMPAIRED;
2760 if (!st->codecpar->extradata) {
2762 st->codecpar->extradata = extradata;
2763 st->codecpar->extradata_size = extradata_size;
2764 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2765 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2766 return AVERROR(ENOMEM);
2767 memcpy(st->codecpar->extradata,
2768 track->codec_priv.data + extradata_offset,
2769 track->codec_priv.size);
2773 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2774 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2775 int display_width_mul = 1;
2776 int display_height_mul = 1;
2778 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2779 st->codecpar->codec_tag = fourcc;
2781 st->codecpar->bits_per_coded_sample = bit_depth;
2782 st->codecpar->width = track->video.pixel_width;
2783 st->codecpar->height = track->video.pixel_height;
2785 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2786 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2787 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2788 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2790 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2791 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2793 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2794 av_reduce(&st->sample_aspect_ratio.num,
2795 &st->sample_aspect_ratio.den,
2796 st->codecpar->height * track->video.display_width * display_width_mul,
2797 st->codecpar->width * track->video.display_height * display_height_mul,
2800 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2801 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2803 if (track->default_duration) {
2804 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2805 1000000000, track->default_duration, 30000);
2806 #if FF_API_R_FRAME_RATE
2807 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2808 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2809 st->r_frame_rate = st->avg_frame_rate;
2813 /* export stereo mode flag as metadata tag */
2814 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2815 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2817 /* export alpha mode flag as metadata tag */
2818 if (track->video.alpha_mode)
2819 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2821 /* if we have virtual track, mark the real tracks */
2822 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2824 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2826 snprintf(buf, sizeof(buf), "%s_%d",
2827 ff_matroska_video_stereo_plane[planes[j].type], i);
2828 for (k=0; k < matroska->tracks.nb_elem; k++)
2829 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2830 av_dict_set(&tracks[k].stream->metadata,
2831 "stereo_mode", buf, 0);
2835 // add stream level stereo3d side data if it is a supported format
2836 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2837 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2838 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2843 ret = mkv_parse_video_color(st, track);
2846 ret = mkv_parse_video_projection(st, track, matroska->ctx);
2849 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2850 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2851 st->codecpar->codec_tag = fourcc;
2852 st->codecpar->sample_rate = track->audio.out_samplerate;
2853 st->codecpar->channels = track->audio.channels;
2854 if (!st->codecpar->bits_per_coded_sample)
2855 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2856 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2857 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2858 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2859 st->need_parsing = AVSTREAM_PARSE_FULL;
2860 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2861 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2862 if (track->codec_delay > 0) {
2863 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2864 (AVRational){1, 1000000000},
2865 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2866 48000 : st->codecpar->sample_rate});
2868 if (track->seek_preroll > 0) {
2869 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2870 (AVRational){1, 1000000000},
2871 (AVRational){1, st->codecpar->sample_rate});
2873 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2874 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2876 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2877 st->disposition |= AV_DISPOSITION_CAPTIONS;
2878 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2879 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2880 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2881 st->disposition |= AV_DISPOSITION_METADATA;
2883 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2884 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2886 if (track->flag_textdescriptions)
2887 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2894 static int matroska_read_header(AVFormatContext *s)
2896 MatroskaDemuxContext *matroska = s->priv_data;
2897 EbmlList *attachments_list = &matroska->attachments;
2898 EbmlList *chapters_list = &matroska->chapters;
2899 MatroskaAttachment *attachments;
2900 MatroskaChapter *chapters;
2901 uint64_t max_start = 0;
2907 matroska->cues_parsing_deferred = 1;
2909 /* First read the EBML header. */
2910 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2911 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2912 ebml_free(ebml_syntax, &ebml);
2913 return AVERROR_INVALIDDATA;
2915 if (ebml.version > EBML_VERSION ||
2916 ebml.max_size > sizeof(uint64_t) ||
2917 ebml.id_length > sizeof(uint32_t) ||
2918 ebml.doctype_version > 3) {
2919 avpriv_report_missing_feature(matroska->ctx,
2920 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2921 ebml.version, ebml.doctype, ebml.doctype_version);
2922 ebml_free(ebml_syntax, &ebml);
2923 return AVERROR_PATCHWELCOME;
2924 } else if (ebml.doctype_version == 3) {
2925 av_log(matroska->ctx, AV_LOG_WARNING,
2926 "EBML header using unsupported features\n"
2927 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2928 ebml.version, ebml.doctype, ebml.doctype_version);
2930 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2931 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2933 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2934 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2935 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2936 ebml_free(ebml_syntax, &ebml);
2937 return AVERROR_INVALIDDATA;
2940 ebml_free(ebml_syntax, &ebml);
2942 /* The next thing is a segment. */
2943 pos = avio_tell(matroska->ctx->pb);
2944 res = ebml_parse(matroska, matroska_segments, matroska);
2945 // Try resyncing until we find an EBML_STOP type element.
2947 res = matroska_resync(matroska, pos);
2950 pos = avio_tell(matroska->ctx->pb);
2951 res = ebml_parse(matroska, matroska_segment, matroska);
2953 /* Set data_offset as it might be needed later by seek_frame_generic. */
2954 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2955 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2956 matroska_execute_seekhead(matroska);
2958 if (!matroska->time_scale)
2959 matroska->time_scale = 1000000;
2960 if (matroska->duration)
2961 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2962 1000 / AV_TIME_BASE;
2963 av_dict_set(&s->metadata, "title", matroska->title, 0);
2964 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2966 if (matroska->date_utc.size == 8)
2967 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2969 res = matroska_parse_tracks(s);
2973 attachments = attachments_list->elem;
2974 for (j = 0; j < attachments_list->nb_elem; j++) {
2975 if (!(attachments[j].filename && attachments[j].mime &&
2976 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2977 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2979 AVStream *st = avformat_new_stream(s, NULL);
2982 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2983 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2984 if (attachments[j].description)
2985 av_dict_set(&st->metadata, "title", attachments[j].description, 0);
2986 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2988 for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2989 if (!strncmp(mkv_image_mime_tags[i].str, attachments[j].mime,
2990 strlen(mkv_image_mime_tags[i].str))) {
2991 st->codecpar->codec_id = mkv_image_mime_tags[i].id;
2996 attachments[j].stream = st;
2998 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2999 AVPacket *pkt = &st->attached_pic;
3001 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
3002 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
3004 av_init_packet(pkt);
3005 pkt->buf = attachments[j].bin.buf;
3006 attachments[j].bin.buf = NULL;
3007 pkt->data = attachments[j].bin.data;
3008 pkt->size = attachments[j].bin.size;
3009 pkt->stream_index = st->index;
3010 pkt->flags |= AV_PKT_FLAG_KEY;
3012 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
3013 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
3015 memcpy(st->codecpar->extradata, attachments[j].bin.data,
3016 attachments[j].bin.size);
3018 for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3019 if (!strncmp(mkv_mime_tags[i].str, attachments[j].mime,
3020 strlen(mkv_mime_tags[i].str))) {
3021 st->codecpar->codec_id = mkv_mime_tags[i].id;
3029 chapters = chapters_list->elem;
3030 for (i = 0; i < chapters_list->nb_elem; i++)
3031 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
3032 (max_start == 0 || chapters[i].start > max_start)) {
3033 chapters[i].chapter =
3034 avpriv_new_chapter(s, chapters[i].uid,
3035 (AVRational) { 1, 1000000000 },
3036 chapters[i].start, chapters[i].end,
3038 max_start = chapters[i].start;
3041 matroska_add_index_entries(matroska);
3043 matroska_convert_tags(s);
3047 matroska_read_close(s);
3052 * Put one packet in an application-supplied AVPacket struct.
3053 * Returns 0 on success or -1 on failure.
3055 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3058 if (matroska->queue) {
3059 MatroskaTrack *tracks = matroska->tracks.elem;
3060 MatroskaTrack *track;
3062 avpriv_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
3063 track = &tracks[pkt->stream_index];
3064 if (track->has_palette) {
3065 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
3067 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3069 memcpy(pal, track->palette, AVPALETTE_SIZE);
3071 track->has_palette = 0;
3080 * Free all packets in our internal queue.
3082 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3084 avpriv_packet_list_free(&matroska->queue, &matroska->queue_end);
3087 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3088 int size, int type, AVIOContext *pb,
3089 uint32_t lace_size[256], int *laces)
3092 uint8_t *data = *buf;
3096 lace_size[0] = size;
3101 return AVERROR_INVALIDDATA;
3108 case 0x1: /* Xiph lacing */
3112 for (n = 0; n < *laces - 1; n++) {
3117 return AVERROR_INVALIDDATA;
3120 lace_size[n] += temp;
3123 } while (temp == 0xff);
3126 return AVERROR_INVALIDDATA;
3128 lace_size[n] = size - total;
3132 case 0x2: /* fixed-size lacing */
3133 if (size % (*laces))
3134 return AVERROR_INVALIDDATA;
3135 for (n = 0; n < *laces; n++)
3136 lace_size[n] = size / *laces;
3139 case 0x3: /* EBML lacing */
3147 n = ebml_read_num(matroska, pb, 8, &num, 1);
3151 return AVERROR_INVALIDDATA;
3153 total = lace_size[0] = num;
3155 for (n = 1; n < *laces - 1; n++) {
3158 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3161 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3162 return AVERROR_INVALIDDATA;
3164 lace_size[n] = lace_size[n - 1] + snum;
3165 total += lace_size[n];
3171 return AVERROR_INVALIDDATA;
3173 lace_size[*laces - 1] = size - total;
3183 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3184 MatroskaTrack *track, AVStream *st,
3185 uint8_t *data, int size, uint64_t timecode,
3188 const int a = st->codecpar->block_align;
3189 const int sps = track->audio.sub_packet_size;
3190 const int cfs = track->audio.coded_framesize;
3191 const int h = track->audio.sub_packet_h;
3192 const int w = track->audio.frame_size;
3193 int y = track->audio.sub_packet_cnt;
3196 if (!track->audio.pkt_cnt) {
3197 if (track->audio.sub_packet_cnt == 0)
3198 track->audio.buf_timecode = timecode;
3199 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3200 if (size < cfs * h / 2) {
3201 av_log(matroska->ctx, AV_LOG_ERROR,
3202 "Corrupt int4 RM-style audio packet size\n");
3203 return AVERROR_INVALIDDATA;
3205 for (x = 0; x < h / 2; x++)
3206 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3207 data + x * cfs, cfs);
3208 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3210 av_log(matroska->ctx, AV_LOG_ERROR,
3211 "Corrupt sipr RM-style audio packet size\n");
3212 return AVERROR_INVALIDDATA;
3214 memcpy(track->audio.buf + y * w, data, w);
3217 av_log(matroska->ctx, AV_LOG_ERROR,
3218 "Corrupt generic RM-style audio packet size\n");
3219 return AVERROR_INVALIDDATA;
3221 for (x = 0; x < w / sps; x++)
3222 memcpy(track->audio.buf +
3223 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3224 data + x * sps, sps);
3227 if (++track->audio.sub_packet_cnt >= h) {
3228 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3229 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3230 track->audio.sub_packet_cnt = 0;
3231 track->audio.pkt_cnt = h * w / a;
3235 while (track->audio.pkt_cnt) {
3237 AVPacket pktl, *pkt = &pktl;
3239 ret = av_new_packet(pkt, a);
3244 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3246 pkt->pts = track->audio.buf_timecode;
3247 track->audio.buf_timecode = AV_NOPTS_VALUE;
3249 pkt->stream_index = st->index;
3250 ret = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3252 av_packet_unref(pkt);
3253 return AVERROR(ENOMEM);
3260 /* reconstruct full wavpack blocks from mangled matroska ones */
3261 static int matroska_parse_wavpack(MatroskaTrack *track,
3262 uint8_t **data, int *size)
3264 uint8_t *dst = NULL;
3265 uint8_t *src = *data;
3270 int ret, offset = 0;
3273 return AVERROR_INVALIDDATA;
3275 av_assert1(track->stream->codecpar->extradata_size >= 2);
3276 ver = AV_RL16(track->stream->codecpar->extradata);
3278 samples = AV_RL32(src);
3282 while (srclen >= 8) {
3287 uint32_t flags = AV_RL32(src);
3288 uint32_t crc = AV_RL32(src + 4);
3292 multiblock = (flags & 0x1800) != 0x1800;
3295 ret = AVERROR_INVALIDDATA;
3298 blocksize = AV_RL32(src);
3304 if (blocksize > srclen) {
3305 ret = AVERROR_INVALIDDATA;
3309 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3311 ret = AVERROR(ENOMEM);
3315 dstlen += blocksize + 32;
3317 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3318 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3319 AV_WL16(dst + offset + 8, ver); // version
3320 AV_WL16(dst + offset + 10, 0); // track/index_no
3321 AV_WL32(dst + offset + 12, 0); // total samples
3322 AV_WL32(dst + offset + 16, 0); // block index
3323 AV_WL32(dst + offset + 20, samples); // number of samples
3324 AV_WL32(dst + offset + 24, flags); // flags
3325 AV_WL32(dst + offset + 28, crc); // crc
3326 memcpy(dst + offset + 32, src, blocksize); // block data
3329 srclen -= blocksize;
3330 offset += blocksize + 32;
3333 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3345 static int matroska_parse_prores(MatroskaTrack *track,
3346 uint8_t **data, int *size)
3349 int dstlen = *size + 8;
3351 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3353 return AVERROR(ENOMEM);
3355 AV_WB32(dst, dstlen);
3356 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3357 memcpy(dst + 8, *data, dstlen - 8);
3358 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3366 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3367 MatroskaTrack *track,
3369 uint8_t *data, int data_len,
3374 AVPacket pktl, *pkt = &pktl;
3375 uint8_t *id, *settings, *text, *buf;
3376 int id_len, settings_len, text_len;
3381 return AVERROR_INVALIDDATA;
3384 q = data + data_len;
3389 if (*p == '\r' || *p == '\n') {
3398 if (p >= q || *p != '\n')
3399 return AVERROR_INVALIDDATA;
3405 if (*p == '\r' || *p == '\n') {
3406 settings_len = p - settings;
3414 if (p >= q || *p != '\n')
3415 return AVERROR_INVALIDDATA;
3420 while (text_len > 0) {
3421 const int len = text_len - 1;
3422 const uint8_t c = p[len];
3423 if (c != '\r' && c != '\n')
3429 return AVERROR_INVALIDDATA;
3431 err = av_new_packet(pkt, text_len);
3436 memcpy(pkt->data, text, text_len);
3439 buf = av_packet_new_side_data(pkt,
3440 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3443 av_packet_unref(pkt);
3444 return AVERROR(ENOMEM);
3446 memcpy(buf, id, id_len);
3449 if (settings_len > 0) {
3450 buf = av_packet_new_side_data(pkt,
3451 AV_PKT_DATA_WEBVTT_SETTINGS,
3454 av_packet_unref(pkt);
3455 return AVERROR(ENOMEM);
3457 memcpy(buf, settings, settings_len);
3460 // Do we need this for subtitles?
3461 // pkt->flags = AV_PKT_FLAG_KEY;
3463 pkt->stream_index = st->index;
3464 pkt->pts = timecode;
3466 // Do we need this for subtitles?
3467 // pkt->dts = timecode;
3469 pkt->duration = duration;
3472 err = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3474 av_packet_unref(pkt);
3475 return AVERROR(ENOMEM);
3481 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3482 MatroskaTrack *track, AVStream *st,
3483 AVBufferRef *buf, uint8_t *data, int pkt_size,
3484 uint64_t timecode, uint64_t lace_duration,
3485 int64_t pos, int is_keyframe,
3486 uint8_t *additional, uint64_t additional_id, int additional_size,
3487 int64_t discard_padding)
3489 uint8_t *pkt_data = data;
3491 AVPacket pktl, *pkt = &pktl;
3493 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3494 res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3496 av_log(matroska->ctx, AV_LOG_ERROR,
3497 "Error parsing a wavpack block.\n");
3505 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3506 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3507 res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3509 av_log(matroska->ctx, AV_LOG_ERROR,
3510 "Error parsing a prores block.\n");
3518 if (!pkt_size && !additional_size)
3521 av_init_packet(pkt);
3523 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3526 pkt->buf = av_buffer_ref(buf);
3529 res = AVERROR(ENOMEM);
3533 pkt->data = pkt_data;
3534 pkt->size = pkt_size;
3535 pkt->flags = is_keyframe;
3536 pkt->stream_index = st->index;
3538 if (additional_size > 0) {
3539 uint8_t *side_data = av_packet_new_side_data(pkt,
3540 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3541 additional_size + 8);
3543 av_packet_unref(pkt);
3544 return AVERROR(ENOMEM);
3546 AV_WB64(side_data, additional_id);
3547 memcpy(side_data + 8, additional, additional_size);
3550 if (discard_padding) {
3551 uint8_t *side_data = av_packet_new_side_data(pkt,
3552 AV_PKT_DATA_SKIP_SAMPLES,
3555 av_packet_unref(pkt);
3556 return AVERROR(ENOMEM);
3558 discard_padding = av_rescale_q(discard_padding,
3559 (AVRational){1, 1000000000},
3560 (AVRational){1, st->codecpar->sample_rate});
3561 if (discard_padding > 0) {
3562 AV_WL32(side_data + 4, discard_padding);
3564 AV_WL32(side_data, -discard_padding);
3568 if (track->ms_compat)
3569 pkt->dts = timecode;
3571 pkt->pts = timecode;
3573 pkt->duration = lace_duration;
3575 #if FF_API_CONVERGENCE_DURATION
3576 FF_DISABLE_DEPRECATION_WARNINGS
3577 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3578 pkt->convergence_duration = lace_duration;
3580 FF_ENABLE_DEPRECATION_WARNINGS
3583 res = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3585 av_packet_unref(pkt);
3586 return AVERROR(ENOMEM);
3598 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3599 int size, int64_t pos, uint64_t cluster_time,
3600 uint64_t block_duration, int is_keyframe,
3601 uint8_t *additional, uint64_t additional_id, int additional_size,
3602 int64_t cluster_pos, int64_t discard_padding)
3604 uint64_t timecode = AV_NOPTS_VALUE;
3605 MatroskaTrack *track;
3610 uint32_t lace_size[256];
3611 int n, flags, laces = 0;
3613 int trust_default_duration;
3615 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3617 if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3622 track = matroska_find_track_by_num(matroska, num);
3623 if (!track || size < 3)
3624 return AVERROR_INVALIDDATA;
3626 if (!(st = track->stream)) {
3627 av_log(matroska->ctx, AV_LOG_VERBOSE,
3628 "No stream associated to TrackNumber %"PRIu64". "
3629 "Ignoring Block with this TrackNumber.\n", num);
3633 if (st->discard >= AVDISCARD_ALL)
3635 if (block_duration > INT64_MAX)
3636 block_duration = INT64_MAX;
3638 block_time = sign_extend(AV_RB16(data), 16);
3642 if (is_keyframe == -1)
3643 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3645 if (cluster_time != (uint64_t) -1 &&
3646 (block_time >= 0 || cluster_time >= -block_time)) {
3647 uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
3648 timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
3649 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3650 timecode < track->end_timecode)
3651 is_keyframe = 0; /* overlapping subtitles are not key frame */
3653 ff_reduce_index(matroska->ctx, st->index);
3654 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3659 if (matroska->skip_to_keyframe &&
3660 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3661 // Compare signed timecodes. Timecode may be negative due to codec delay
3662 // offset. We don't support timestamps greater than int64_t anyway - see
3664 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3667 matroska->skip_to_keyframe = 0;
3668 else if (!st->internal->skip_to_keyframe) {
3669 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3670 matroska->skip_to_keyframe = 0;
3674 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3675 &pb, lace_size, &laces);
3677 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3681 trust_default_duration = track->default_duration != 0;
3682 if (track->audio.samplerate == 8000 && trust_default_duration) {
3683 // If this is needed for more codecs, then add them here
3684 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3685 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3686 trust_default_duration = 0;
3690 if (!block_duration && trust_default_duration)
3691 block_duration = track->default_duration * laces / matroska->time_scale;
3693 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3694 track->end_timecode =
3695 FFMAX(track->end_timecode, timecode + block_duration);
3697 for (n = 0; n < laces; n++) {
3698 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3699 uint8_t *out_data = data;
3700 int out_size = lace_size[n];
3702 if (track->needs_decoding) {
3703 res = matroska_decode_buffer(&out_data, &out_size, track);
3706 /* Given that we are here means that out_data is no longer
3707 * owned by buf, so set it to NULL. This depends upon
3708 * zero-length header removal compression being ignored. */
3709 av_assert1(out_data != data);
3713 if (track->audio.buf) {
3714 res = matroska_parse_rm_audio(matroska, track, st,
3721 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3722 res = matroska_parse_webvtt(matroska, track, st,
3724 timecode, lace_duration,
3731 res = matroska_parse_frame(matroska, track, st, buf, out_data,
3732 out_size, timecode, lace_duration,
3733 pos, !n ? is_keyframe : 0,
3734 additional, additional_id, additional_size,
3740 if (timecode != AV_NOPTS_VALUE)
3741 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3742 data += lace_size[n];
3748 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3750 MatroskaCluster *cluster = &matroska->current_cluster;
3751 MatroskaBlock *block = &cluster->block;
3754 av_assert0(matroska->num_levels <= 2);
3756 if (matroska->num_levels == 1) {
3757 res = ebml_parse(matroska, matroska_segment, NULL);
3760 /* Found a cluster: subtract the size of the ID already read. */
3761 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3763 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3769 if (matroska->num_levels == 2) {
3770 /* We are inside a cluster. */
3771 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3773 if (res >= 0 && block->bin.size > 0) {
3774 int is_keyframe = block->non_simple ? block->reference.count == 0 : -1;
3775 uint8_t* additional = block->additional.size > 0 ?
3776 block->additional.data : NULL;
3778 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3779 block->bin.size, block->bin.pos,
3780 cluster->timecode, block->duration,
3781 is_keyframe, additional, block->additional_id,
3782 block->additional.size, cluster->pos,
3783 block->discard_padding);
3786 ebml_free(matroska_blockgroup, block);
3787 memset(block, 0, sizeof(*block));
3788 } else if (!matroska->num_levels) {
3789 if (!avio_feof(matroska->ctx->pb)) {
3790 avio_r8(matroska->ctx->pb);
3791 if (!avio_feof(matroska->ctx->pb)) {
3792 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3793 "end of segment.\n");
3794 return AVERROR_INVALIDDATA;
3804 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3806 MatroskaDemuxContext *matroska = s->priv_data;
3809 if (matroska->resync_pos == -1) {
3810 // This can only happen if generic seeking has been used.
3811 matroska->resync_pos = avio_tell(s->pb);
3814 while (matroska_deliver_packet(matroska, pkt)) {
3816 return (ret < 0) ? ret : AVERROR_EOF;
3817 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3818 ret = matroska_resync(matroska, matroska->resync_pos);
3824 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3825 int64_t timestamp, int flags)
3827 MatroskaDemuxContext *matroska = s->priv_data;
3828 MatroskaTrack *tracks = NULL;
3829 AVStream *st = s->streams[stream_index];
3832 /* Parse the CUES now since we need the index data to seek. */
3833 if (matroska->cues_parsing_deferred > 0) {
3834 matroska->cues_parsing_deferred = 0;
3835 matroska_parse_cues(matroska);
3838 if (!st->internal->nb_index_entries)
3840 timestamp = FFMAX(timestamp, st->internal->index_entries[0].timestamp);
3842 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3843 matroska_reset_status(matroska, 0, st->internal->index_entries[st->internal->nb_index_entries - 1].pos);
3844 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3845 matroska_clear_queue(matroska);
3846 if (matroska_parse_cluster(matroska) < 0)
3851 matroska_clear_queue(matroska);
3852 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->internal->nb_index_entries - 1))
3855 tracks = matroska->tracks.elem;
3856 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3857 tracks[i].audio.pkt_cnt = 0;
3858 tracks[i].audio.sub_packet_cnt = 0;
3859 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3860 tracks[i].end_timecode = 0;
3863 /* We seek to a level 1 element, so set the appropriate status. */
3864 matroska_reset_status(matroska, 0, st->internal->index_entries[index].pos);
3865 if (flags & AVSEEK_FLAG_ANY) {
3866 st->internal->skip_to_keyframe = 0;
3867 matroska->skip_to_timecode = timestamp;
3869 st->internal->skip_to_keyframe = 1;
3870 matroska->skip_to_timecode = st->internal->index_entries[index].timestamp;
3872 matroska->skip_to_keyframe = 1;
3874 ff_update_cur_dts(s, st, st->internal->index_entries[index].timestamp);
3877 // slightly hackish but allows proper fallback to
3878 // the generic seeking code.
3879 matroska_reset_status(matroska, 0, -1);
3880 matroska->resync_pos = -1;
3881 matroska_clear_queue(matroska);
3882 st->internal->skip_to_keyframe =
3883 matroska->skip_to_keyframe = 0;
3888 static int matroska_read_close(AVFormatContext *s)
3890 MatroskaDemuxContext *matroska = s->priv_data;
3891 MatroskaTrack *tracks = matroska->tracks.elem;
3894 matroska_clear_queue(matroska);
3896 for (n = 0; n < matroska->tracks.nb_elem; n++)
3897 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3898 av_freep(&tracks[n].audio.buf);
3899 ebml_free(matroska_segment, matroska);
3905 int64_t start_time_ns;
3906 int64_t end_time_ns;
3907 int64_t start_offset;
3911 /* This function searches all the Cues and returns the CueDesc corresponding to
3912 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3913 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3915 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3916 MatroskaDemuxContext *matroska = s->priv_data;
3919 int nb_index_entries = s->streams[0]->internal->nb_index_entries;
3920 AVIndexEntry *index_entries = s->streams[0]->internal->index_entries;
3921 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3922 for (i = 1; i < nb_index_entries; i++) {
3923 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3924 index_entries[i].timestamp * matroska->time_scale > ts) {
3929 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3930 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3931 if (i != nb_index_entries - 1) {
3932 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3933 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3935 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3936 // FIXME: this needs special handling for files where Cues appear
3937 // before Clusters. the current logic assumes Cues appear after
3939 cue_desc.end_offset = cues_start - matroska->segment_start;
3944 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3946 MatroskaDemuxContext *matroska = s->priv_data;
3947 uint32_t id = matroska->current_id;
3948 int64_t cluster_pos, before_pos;
3950 if (s->streams[0]->internal->nb_index_entries <= 0) return 0;
3951 // seek to the first cluster using cues.
3952 index = av_index_search_timestamp(s->streams[0], 0, 0);
3953 if (index < 0) return 0;
3954 cluster_pos = s->streams[0]->internal->index_entries[index].pos;
3955 before_pos = avio_tell(s->pb);
3957 uint64_t cluster_id, cluster_length;
3960 avio_seek(s->pb, cluster_pos, SEEK_SET);
3961 // read cluster id and length
3962 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3963 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3965 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3969 matroska_reset_status(matroska, 0, cluster_pos);
3970 matroska_clear_queue(matroska);
3971 if (matroska_parse_cluster(matroska) < 0 ||
3975 pkt = &matroska->queue->pkt;
3976 // 4 + read is the length of the cluster id and the cluster length field.
3977 cluster_pos += 4 + read + cluster_length;
3978 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3984 /* Restore the status after matroska_read_header: */
3985 matroska_reset_status(matroska, id, before_pos);
3990 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3991 double min_buffer, double* buffer,
3992 double* sec_to_download, AVFormatContext *s,
3995 double nano_seconds_per_second = 1000000000.0;
3996 double time_sec = time_ns / nano_seconds_per_second;
3998 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3999 int64_t end_time_ns = time_ns + time_to_search_ns;
4000 double sec_downloaded = 0.0;
4001 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
4002 if (desc_curr.start_time_ns == -1)
4004 *sec_to_download = 0.0;
4006 // Check for non cue start time.
4007 if (time_ns > desc_curr.start_time_ns) {
4008 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
4009 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
4010 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
4011 double timeToDownload = (cueBytes * 8.0) / bps;
4013 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
4014 *sec_to_download += timeToDownload;
4016 // Check if the search ends within the first cue.
4017 if (desc_curr.end_time_ns >= end_time_ns) {
4018 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4019 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4020 sec_downloaded = percent_to_sub * sec_downloaded;
4021 *sec_to_download = percent_to_sub * *sec_to_download;
4024 if ((sec_downloaded + *buffer) <= min_buffer) {
4028 // Get the next Cue.
4029 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4032 while (desc_curr.start_time_ns != -1) {
4033 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
4034 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
4035 double desc_sec = desc_ns / nano_seconds_per_second;
4036 double bits = (desc_bytes * 8.0);
4037 double time_to_download = bits / bps;
4039 sec_downloaded += desc_sec - time_to_download;
4040 *sec_to_download += time_to_download;
4042 if (desc_curr.end_time_ns >= end_time_ns) {
4043 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4044 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4045 sec_downloaded = percent_to_sub * sec_downloaded;
4046 *sec_to_download = percent_to_sub * *sec_to_download;
4048 if ((sec_downloaded + *buffer) <= min_buffer)
4053 if ((sec_downloaded + *buffer) <= min_buffer) {
4058 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4060 *buffer = *buffer + sec_downloaded;
4064 /* This function computes the bandwidth of the WebM file with the help of
4065 * buffer_size_after_time_downloaded() function. Both of these functions are
4066 * adapted from WebM Tools project and are adapted to work with FFmpeg's
4067 * Matroska parsing mechanism.
4069 * Returns the bandwidth of the file on success; -1 on error.
4071 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4073 MatroskaDemuxContext *matroska = s->priv_data;
4074 AVStream *st = s->streams[0];
4075 double bandwidth = 0.0;
4078 for (i = 0; i < st->internal->nb_index_entries; i++) {
4079 int64_t prebuffer_ns = 1000000000;
4080 int64_t time_ns = st->internal->index_entries[i].timestamp * matroska->time_scale;
4081 double nano_seconds_per_second = 1000000000.0;
4082 int64_t prebuffered_ns = time_ns + prebuffer_ns;
4083 double prebuffer_bytes = 0.0;
4084 int64_t temp_prebuffer_ns = prebuffer_ns;
4085 int64_t pre_bytes, pre_ns;
4086 double pre_sec, prebuffer, bits_per_second;
4087 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4089 // Start with the first Cue.
4090 CueDesc desc_end = desc_beg;
4092 // Figure out how much data we have downloaded for the prebuffer. This will
4093 // be used later to adjust the bits per sample to try.
4094 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4095 // Prebuffered the entire Cue.
4096 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4097 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4098 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4100 if (desc_end.start_time_ns == -1) {
4101 // The prebuffer is larger than the duration.
4102 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4104 bits_per_second = 0.0;
4106 // The prebuffer ends in the last Cue. Estimate how much data was
4108 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4109 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4110 pre_sec = pre_ns / nano_seconds_per_second;
4112 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4114 prebuffer = prebuffer_ns / nano_seconds_per_second;
4116 // Set this to 0.0 in case our prebuffer buffers the entire video.
4117 bits_per_second = 0.0;
4119 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4120 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4121 double desc_sec = desc_ns / nano_seconds_per_second;
4122 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4124 // Drop the bps by the percentage of bytes buffered.
4125 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4126 double mod_bits_per_second = calc_bits_per_second * percent;
4128 if (prebuffer < desc_sec) {
4130 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4132 // Add 1 so the bits per second should be a little bit greater than file
4134 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4135 const double min_buffer = 0.0;
4136 double buffer = prebuffer;
4137 double sec_to_download = 0.0;
4139 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4140 min_buffer, &buffer, &sec_to_download,
4144 } else if (rv == 0) {
4145 bits_per_second = (double)(bps);
4150 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4151 } while (desc_end.start_time_ns != -1);
4153 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4155 return (int64_t)bandwidth;
4158 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4160 MatroskaDemuxContext *matroska = s->priv_data;
4161 EbmlList *seekhead_list = &matroska->seekhead;
4162 MatroskaSeekhead *seekhead = seekhead_list->elem;
4164 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4168 // determine cues start and end positions
4169 for (i = 0; i < seekhead_list->nb_elem; i++)
4170 if (seekhead[i].id == MATROSKA_ID_CUES)
4173 if (i >= seekhead_list->nb_elem) return -1;
4175 before_pos = avio_tell(matroska->ctx->pb);
4176 cues_start = seekhead[i].pos + matroska->segment_start;
4177 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4178 // cues_end is computed as cues_start + cues_length + length of the
4179 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4180 // cues_end is inclusive and the above sum is reduced by 1.
4181 uint64_t cues_length, cues_id;
4183 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4184 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4185 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4186 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4189 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4191 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4192 if (cues_start == -1 || cues_end == -1) return -1;
4195 matroska_parse_cues(matroska);
4198 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4201 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4203 // if the file has cues at the start, fix up the init range so that
4204 // it does not include it
4205 if (cues_start <= init_range)
4206 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4209 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4210 if (bandwidth < 0) return -1;
4211 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4213 // check if all clusters start with key frames
4214 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4216 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4217 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4218 buf = av_malloc_array(s->streams[0]->internal->nb_index_entries, 20);
4219 if (!buf) return -1;
4221 for (i = 0; i < s->streams[0]->internal->nb_index_entries; i++) {
4222 int ret = snprintf(buf + end, 20,
4223 "%" PRId64"%s", s->streams[0]->internal->index_entries[i].timestamp,
4224 i != s->streams[0]->internal->nb_index_entries - 1 ? "," : "");
4225 if (ret <= 0 || (ret == 20 && i == s->streams[0]->internal->nb_index_entries - 1)) {
4226 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4228 return AVERROR_INVALIDDATA;
4232 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4233 buf, AV_DICT_DONT_STRDUP_VAL);
4238 static int webm_dash_manifest_read_header(AVFormatContext *s)
4241 int ret = matroska_read_header(s);
4243 MatroskaTrack *tracks;
4244 MatroskaDemuxContext *matroska = s->priv_data;
4246 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4249 if (!matroska->tracks.nb_elem || !s->nb_streams) {
4250 av_log(s, AV_LOG_ERROR, "No track found\n");
4251 ret = AVERROR_INVALIDDATA;
4255 if (!matroska->is_live) {
4256 buf = av_asprintf("%g", matroska->duration);
4258 ret = AVERROR(ENOMEM);
4261 av_dict_set(&s->streams[0]->metadata, DURATION,
4262 buf, AV_DICT_DONT_STRDUP_VAL);
4264 // initialization range
4265 // 5 is the offset of Cluster ID.
4266 init_range = avio_tell(s->pb) - 5;
4267 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4270 // basename of the file
4271 buf = strrchr(s->url, '/');
4272 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4275 tracks = matroska->tracks.elem;
4276 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4278 // parse the cues and populate Cue related fields
4279 if (!matroska->is_live) {
4280 ret = webm_dash_manifest_cues(s, init_range);
4282 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4287 // use the bandwidth from the command line if it was provided
4288 if (matroska->bandwidth > 0) {
4289 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4290 matroska->bandwidth, 0);
4294 matroska_read_close(s);
4298 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4303 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4304 static const AVOption options[] = {
4305 { "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 },
4306 { "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 },
4310 static const AVClass webm_dash_class = {
4311 .class_name = "WebM DASH Manifest demuxer",
4312 .item_name = av_default_item_name,
4314 .version = LIBAVUTIL_VERSION_INT,
4317 AVInputFormat ff_matroska_demuxer = {
4318 .name = "matroska,webm",
4319 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4320 .extensions = "mkv,mk3d,mka,mks",
4321 .priv_data_size = sizeof(MatroskaDemuxContext),
4322 .read_probe = matroska_probe,
4323 .read_header = matroska_read_header,
4324 .read_packet = matroska_read_packet,
4325 .read_close = matroska_read_close,
4326 .read_seek = matroska_read_seek,
4327 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4330 AVInputFormat ff_webm_dash_manifest_demuxer = {
4331 .name = "webm_dash_manifest",
4332 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4333 .priv_data_size = sizeof(MatroskaDemuxContext),
4334 .read_header = webm_dash_manifest_read_header,
4335 .read_packet = webm_dash_manifest_read_packet,
4336 .read_close = matroska_read_close,
4337 .priv_class = &webm_dash_class,