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 seek_preroll;
254 MatroskaTrackVideo video;
255 MatroskaTrackAudio audio;
256 MatroskaTrackOperation operation;
258 uint64_t codec_delay;
259 uint64_t codec_delay_in_track_tb;
262 int64_t end_timecode;
265 uint64_t max_block_additional_id;
267 uint32_t palette[AVPALETTE_COUNT];
271 typedef struct MatroskaAttachment {
279 } MatroskaAttachment;
281 typedef struct MatroskaChapter {
290 typedef struct MatroskaIndexPos {
295 typedef struct MatroskaIndex {
300 typedef struct MatroskaTag {
308 typedef struct MatroskaTagTarget {
316 typedef struct MatroskaTags {
317 MatroskaTagTarget target;
321 typedef struct MatroskaSeekhead {
326 typedef struct MatroskaLevel {
331 typedef struct MatroskaBlock {
333 CountedElement reference;
336 uint64_t additional_id;
338 int64_t discard_padding;
341 typedef struct MatroskaCluster {
347 typedef struct MatroskaLevel1Element {
351 } MatroskaLevel1Element;
353 typedef struct MatroskaDemuxContext {
354 const AVClass *class;
355 AVFormatContext *ctx;
358 MatroskaLevel levels[EBML_MAX_DEPTH];
370 EbmlList attachments;
376 /* byte position of the segment inside the stream */
377 int64_t segment_start;
379 /* the packet queue */
381 AVPacketList *queue_end;
385 /* What to skip before effectively reading a packet. */
386 int skip_to_keyframe;
387 uint64_t skip_to_timecode;
389 /* File has a CUES element, but we defer parsing until it is needed. */
390 int cues_parsing_deferred;
392 /* Level1 elements and whether they were read yet */
393 MatroskaLevel1Element level1_elems[64];
394 int num_level1_elems;
396 MatroskaCluster current_cluster;
398 /* WebM DASH Manifest live flag */
401 /* Bandwidth value for WebM DASH Manifest */
403 } MatroskaDemuxContext;
405 #define CHILD_OF(parent) { .def = { .n = parent } }
407 // The following forward declarations need their size because
408 // a tentative definition with internal linkage must not be an
409 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
410 // Removing the sizes breaks MSVC.
411 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
412 matroska_track[27], matroska_track_encoding[6], matroska_track_encodings[2],
413 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_tracks[2],
414 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
415 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
416 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
418 static EbmlSyntax ebml_header[] = {
419 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
420 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, 0, offsetof(Ebml, max_size), { .u = 8 } },
421 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, 0, offsetof(Ebml, id_length), { .u = 4 } },
422 { EBML_ID_DOCTYPE, EBML_STR, 0, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
423 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
424 { EBML_ID_EBMLVERSION, EBML_NONE },
425 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
426 CHILD_OF(ebml_syntax)
429 static EbmlSyntax ebml_syntax[] = {
430 { EBML_ID_HEADER, EBML_NEST, 0, 0, 0, { .n = ebml_header } },
431 { MATROSKA_ID_SEGMENT, EBML_STOP },
435 static EbmlSyntax matroska_info[] = {
436 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
437 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, 0, offsetof(MatroskaDemuxContext, duration) },
438 { MATROSKA_ID_TITLE, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, title) },
439 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
440 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, muxingapp) },
441 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, 0, offsetof(MatroskaDemuxContext, date_utc) },
442 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
443 CHILD_OF(matroska_segment)
446 static EbmlSyntax matroska_mastering_meta[] = {
447 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_x) },
448 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_y) },
449 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_x) },
450 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_y) },
451 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_x) },
452 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_y) },
453 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_x) },
454 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_y) },
455 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 1, 0, offsetof(MatroskaMasteringMeta, min_luminance) },
456 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, max_luminance) },
457 CHILD_OF(matroska_track_video_color)
460 static EbmlSyntax matroska_track_video_color[] = {
461 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
462 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u = 0 } },
463 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz) },
464 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert) },
465 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz) },
466 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert) },
467 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
468 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
469 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
470 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
471 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
472 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_cll) },
473 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_fall) },
474 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
475 CHILD_OF(matroska_track_video)
478 static EbmlSyntax matroska_track_video_projection[] = {
479 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
480 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideoProjection, private) },
481 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f = 0.0 } },
482 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f = 0.0 } },
483 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f = 0.0 } },
484 CHILD_OF(matroska_track_video)
487 static EbmlSyntax matroska_track_video[] = {
488 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideo, frame_rate) },
489 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
490 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
491 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_width) },
492 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_height) },
493 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideo, color_space) },
494 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, alpha_mode), { .u = 0 } },
495 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, 0, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
496 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
497 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
498 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
499 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
500 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
501 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
502 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
503 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
504 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
505 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
506 CHILD_OF(matroska_track)
509 static EbmlSyntax matroska_track_audio[] = {
510 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
511 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
512 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, bitdepth) },
513 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
514 CHILD_OF(matroska_track)
517 static EbmlSyntax matroska_track_encoding_compression[] = {
518 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackCompression, algo), { .u = MATROSKA_TRACK_ENCODING_COMP_ZLIB } },
519 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, 0, offsetof(MatroskaTrackCompression, settings) },
520 CHILD_OF(matroska_track_encoding)
523 static EbmlSyntax matroska_track_encoding_encryption[] = {
524 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
525 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, 0, offsetof(MatroskaTrackEncryption,key_id) },
526 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
527 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
528 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
529 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
530 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
531 CHILD_OF(matroska_track_encoding)
533 static EbmlSyntax matroska_track_encoding[] = {
534 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
535 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
536 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
537 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
538 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
539 CHILD_OF(matroska_track_encodings)
542 static EbmlSyntax matroska_track_encodings[] = {
543 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, 0, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
544 CHILD_OF(matroska_track)
547 static EbmlSyntax matroska_track_plane[] = {
548 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,uid) },
549 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,type) },
550 CHILD_OF(matroska_track_combine_planes)
553 static EbmlSyntax matroska_track_combine_planes[] = {
554 { MATROSKA_ID_TRACKPLANE, EBML_NEST, 0, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
555 CHILD_OF(matroska_track_operation)
558 static EbmlSyntax matroska_track_operation[] = {
559 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, 0, {.n = matroska_track_combine_planes} },
560 CHILD_OF(matroska_track)
563 static EbmlSyntax matroska_track[] = {
564 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, 0, offsetof(MatroskaTrack, num) },
565 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTrack, name) },
566 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, uid) },
567 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrack, type) },
568 { MATROSKA_ID_CODECID, EBML_STR, 0, 0, offsetof(MatroskaTrack, codec_id) },
569 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrack, codec_priv) },
570 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, codec_delay), { .u = 0 } },
571 { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
572 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, 0, offsetof(MatroskaTrack, default_duration) },
573 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
574 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
575 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
576 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
577 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
578 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
579 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, 0, { .n = matroska_track_encodings } },
580 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, max_block_additional_id), { .u = 0 } },
581 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, 0, offsetof(MatroskaTrack, seek_preroll), { .u = 0 } },
582 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
583 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
584 { MATROSKA_ID_CODECNAME, EBML_NONE },
585 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
586 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
587 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
588 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
589 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
590 CHILD_OF(matroska_tracks)
593 static EbmlSyntax matroska_tracks[] = {
594 { MATROSKA_ID_TRACKENTRY, EBML_NEST, 0, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
595 CHILD_OF(matroska_segment)
598 static EbmlSyntax matroska_attachment[] = {
599 { MATROSKA_ID_FILEUID, EBML_UINT, 0, 0, offsetof(MatroskaAttachment, uid) },
600 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, filename) },
601 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, 0, offsetof(MatroskaAttachment, mime) },
602 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, 0, offsetof(MatroskaAttachment, bin) },
603 { MATROSKA_ID_FILEDESC, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, description) },
604 CHILD_OF(matroska_attachments)
607 static EbmlSyntax matroska_attachments[] = {
608 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, 0, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
609 CHILD_OF(matroska_segment)
612 static EbmlSyntax matroska_chapter_display[] = {
613 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaChapter, title) },
614 { MATROSKA_ID_CHAPLANG, EBML_NONE },
615 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
616 CHILD_OF(matroska_chapter_entry)
619 static EbmlSyntax matroska_chapter_entry[] = {
620 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
621 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
622 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaChapter, uid) },
623 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter_display } },
624 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
625 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
626 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
627 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
628 CHILD_OF(matroska_chapter)
631 static EbmlSyntax matroska_chapter[] = {
632 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, 0, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
633 { MATROSKA_ID_EDITIONUID, EBML_NONE },
634 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
635 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
636 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
637 CHILD_OF(matroska_chapters)
640 static EbmlSyntax matroska_chapters[] = {
641 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter } },
642 CHILD_OF(matroska_segment)
645 static EbmlSyntax matroska_index_pos[] = {
646 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, track) },
647 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, pos) },
648 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
649 { MATROSKA_ID_CUEDURATION, EBML_NONE },
650 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
651 CHILD_OF(matroska_index_entry)
654 static EbmlSyntax matroska_index_entry[] = {
655 { MATROSKA_ID_CUETIME, EBML_UINT, 0, 0, offsetof(MatroskaIndex, time) },
656 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, 0, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
657 CHILD_OF(matroska_index)
660 static EbmlSyntax matroska_index[] = {
661 { MATROSKA_ID_POINTENTRY, EBML_NEST, 0, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
662 CHILD_OF(matroska_segment)
665 static EbmlSyntax matroska_simpletag[] = {
666 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTag, name) },
667 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaTag, string) },
668 { MATROSKA_ID_TAGLANG, EBML_STR, 0, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
669 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
670 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
671 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
672 CHILD_OF(matroska_tag)
675 static EbmlSyntax matroska_tagtargets[] = {
676 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, 0, offsetof(MatroskaTagTarget, type) },
677 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
678 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, trackuid), { .u = 0 } },
679 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, chapteruid), { .u = 0 } },
680 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, attachuid), { .u = 0 } },
681 CHILD_OF(matroska_tag)
684 static EbmlSyntax matroska_tag[] = {
685 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
686 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
687 CHILD_OF(matroska_tags)
690 static EbmlSyntax matroska_tags[] = {
691 { MATROSKA_ID_TAG, EBML_NEST, 0, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
692 CHILD_OF(matroska_segment)
695 static EbmlSyntax matroska_seekhead_entry[] = {
696 { MATROSKA_ID_SEEKID, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, id) },
697 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
698 CHILD_OF(matroska_seekhead)
701 static EbmlSyntax matroska_seekhead[] = {
702 { MATROSKA_ID_SEEKENTRY, EBML_NEST, 0, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
703 CHILD_OF(matroska_segment)
706 static EbmlSyntax matroska_segment[] = {
707 { MATROSKA_ID_CLUSTER, EBML_STOP },
708 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, 0, { .n = matroska_info } },
709 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tracks } },
710 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_attachments } },
711 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_chapters } },
712 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, 0, { .n = matroska_index } },
713 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tags } },
714 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, 0, { .n = matroska_seekhead } },
715 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
718 static EbmlSyntax matroska_segments[] = {
719 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, 0, { .n = matroska_segment } },
723 static EbmlSyntax matroska_blockmore[] = {
724 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
725 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, 0, offsetof(MatroskaBlock,additional) },
726 CHILD_OF(matroska_blockadditions)
729 static EbmlSyntax matroska_blockadditions[] = {
730 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, 0, {.n = matroska_blockmore} },
731 CHILD_OF(matroska_blockgroup)
734 static EbmlSyntax matroska_blockgroup[] = {
735 { MATROSKA_ID_BLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
736 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, 0, { .n = matroska_blockadditions} },
737 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, 0, offsetof(MatroskaBlock, duration) },
738 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, 0, offsetof(MatroskaBlock, discard_padding) },
739 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 1, 0, offsetof(MatroskaBlock, reference) },
740 { MATROSKA_ID_CODECSTATE, EBML_NONE },
741 { 1, EBML_UINT, 0, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
742 CHILD_OF(matroska_cluster_parsing)
745 // The following array contains SimpleBlock and BlockGroup twice
746 // in order to reuse the other values for matroska_cluster_enter.
747 static EbmlSyntax matroska_cluster_parsing[] = {
748 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
749 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, 0, { .n = matroska_blockgroup } },
750 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, 0, offsetof(MatroskaCluster, timecode) },
751 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
752 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
753 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
754 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
755 CHILD_OF(matroska_segment)
758 static EbmlSyntax matroska_cluster_enter[] = {
759 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, 0, { .n = &matroska_cluster_parsing[2] } },
764 static const CodecMime mkv_image_mime_tags[] = {
765 {"image/gif" , AV_CODEC_ID_GIF},
766 {"image/jpeg" , AV_CODEC_ID_MJPEG},
767 {"image/png" , AV_CODEC_ID_PNG},
768 {"image/tiff" , AV_CODEC_ID_TIFF},
770 {"" , AV_CODEC_ID_NONE}
773 static const CodecMime mkv_mime_tags[] = {
774 {"text/plain" , AV_CODEC_ID_TEXT},
775 {"application/x-truetype-font", AV_CODEC_ID_TTF},
776 {"application/x-font" , AV_CODEC_ID_TTF},
777 {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
778 {"binary" , AV_CODEC_ID_BIN_DATA},
780 {"" , AV_CODEC_ID_NONE}
783 static const char *const matroska_doctypes[] = { "matroska", "webm" };
785 static int matroska_read_close(AVFormatContext *s);
788 * This function prepares the status for parsing of level 1 elements.
790 static int matroska_reset_status(MatroskaDemuxContext *matroska,
791 uint32_t id, int64_t position)
794 int64_t err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
799 matroska->current_id = id;
800 matroska->num_levels = 1;
801 matroska->unknown_count = 0;
802 matroska->resync_pos = avio_tell(matroska->ctx->pb);
804 matroska->resync_pos -= (av_log2(id) + 7) / 8;
809 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
811 AVIOContext *pb = matroska->ctx->pb;
814 /* Try to seek to the last position to resync from. If this doesn't work,
815 * we resync from the earliest position available: The start of the buffer. */
816 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
817 av_log(matroska->ctx, AV_LOG_WARNING,
818 "Seek to desired resync point failed. Seeking to "
819 "earliest point available instead.\n");
820 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
821 last_pos + 1), SEEK_SET);
826 // try to find a toplevel element
827 while (!avio_feof(pb)) {
828 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
829 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
830 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
831 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
832 /* Prepare the context for parsing of a level 1 element. */
833 matroska_reset_status(matroska, id, -1);
834 /* Given that we are here means that an error has occurred,
835 * so treat the segment as unknown length in order not to
836 * discard valid data that happens to be beyond the designated
837 * end of the segment. */
838 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
841 id = (id << 8) | avio_r8(pb);
845 return pb->error ? pb->error : AVERROR_EOF;
849 * Read: an "EBML number", which is defined as a variable-length
850 * array of bytes. The first byte indicates the length by giving a
851 * number of 0-bits followed by a one. The position of the first
852 * "one" bit inside the first byte indicates the length of this
854 * Returns: number of bytes read, < 0 on error
856 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
857 int max_size, uint64_t *number, int eof_forbidden)
863 /* The first byte tells us the length in bytes - except when it is zero. */
868 /* get the length of the EBML number */
869 read = 8 - ff_log2_tab[total];
871 if (!total || read > max_size) {
872 pos = avio_tell(pb) - 1;
874 av_log(matroska->ctx, AV_LOG_ERROR,
875 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
876 "of an EBML number\n", pos, pos);
878 av_log(matroska->ctx, AV_LOG_ERROR,
879 "Length %d indicated by an EBML number's first byte 0x%02x "
880 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
881 read, (uint8_t) total, pos, pos, max_size);
883 return AVERROR_INVALIDDATA;
886 /* read out length */
887 total ^= 1 << ff_log2_tab[total];
889 total = (total << 8) | avio_r8(pb);
891 if (pb->eof_reached) {
903 av_log(matroska->ctx, AV_LOG_ERROR,
904 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
909 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
910 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
917 * Read a EBML length value.
918 * This needs special handling for the "unknown length" case which has multiple
921 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
924 int res = ebml_read_num(matroska, pb, 8, number, 1);
925 if (res > 0 && *number + 1 == 1ULL << (7 * res))
926 *number = EBML_UNKNOWN_LENGTH;
931 * Read the next element as an unsigned int.
932 * Returns NEEDS_CHECKING.
934 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
938 /* big-endian ordering; build up number */
941 *num = (*num << 8) | avio_r8(pb);
943 return NEEDS_CHECKING;
947 * Read the next element as a signed int.
948 * Returns NEEDS_CHECKING.
950 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
957 *num = sign_extend(avio_r8(pb), 8);
959 /* big-endian ordering; build up number */
961 *num = ((uint64_t)*num << 8) | avio_r8(pb);
964 return NEEDS_CHECKING;
968 * Read the next element as a float.
969 * Returns NEEDS_CHECKING or < 0 on obvious failure.
971 static int ebml_read_float(AVIOContext *pb, int size, double *num)
976 *num = av_int2float(avio_rb32(pb));
978 *num = av_int2double(avio_rb64(pb));
980 return AVERROR_INVALIDDATA;
982 return NEEDS_CHECKING;
986 * Read the next element as an ASCII string.
987 * 0 is success, < 0 or NEEDS_CHECKING is failure.
989 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
994 /* EBML strings are usually not 0-terminated, so we allocate one
995 * byte more, read the string and NULL-terminate it ourselves. */
996 if (!(res = av_malloc(size + 1)))
997 return AVERROR(ENOMEM);
998 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
1000 return ret < 0 ? ret : NEEDS_CHECKING;
1010 * Read the next element as binary data.
1011 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1013 static int ebml_read_binary(AVIOContext *pb, int length,
1014 int64_t pos, EbmlBin *bin)
1018 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1021 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1023 bin->data = bin->buf->data;
1026 if ((ret = avio_read(pb, bin->data, length)) != length) {
1027 av_buffer_unref(&bin->buf);
1030 return ret < 0 ? ret : NEEDS_CHECKING;
1037 * Read the next element, but only the header. The contents
1038 * are supposed to be sub-elements which can be read separately.
1039 * 0 is success, < 0 is failure.
1041 static int ebml_read_master(MatroskaDemuxContext *matroska,
1042 uint64_t length, int64_t pos)
1044 MatroskaLevel *level;
1046 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1047 av_log(matroska->ctx, AV_LOG_ERROR,
1048 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1049 return AVERROR(ENOSYS);
1052 level = &matroska->levels[matroska->num_levels++];
1054 level->length = length;
1060 * Read a signed "EBML number"
1061 * Return: number of bytes processed, < 0 on error
1063 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1064 AVIOContext *pb, int64_t *num)
1069 /* read as unsigned number first */
1070 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1073 /* make signed (weird way) */
1074 *num = unum - ((1LL << (7 * res - 1)) - 1);
1079 static int ebml_parse(MatroskaDemuxContext *matroska,
1080 EbmlSyntax *syntax, void *data);
1082 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1086 // Whoever touches this should be aware of the duplication
1087 // existing in matroska_cluster_parsing.
1088 for (i = 0; syntax[i].id; i++)
1089 if (id == syntax[i].id)
1095 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1101 for (int i = 0; syntax[i].id; i++)
1102 switch (syntax[i].type) {
1104 *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1107 *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1110 *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1114 // the default may be NULL
1115 if (syntax[i].def.s) {
1116 uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1117 *dst = av_strdup(syntax[i].def.s);
1119 return AVERROR(ENOMEM);
1124 if (!matroska->levels[matroska->num_levels - 1].length) {
1125 matroska->num_levels--;
1131 res = ebml_parse(matroska, syntax, data);
1134 return res == LEVEL_ENDED ? 0 : res;
1137 static int is_ebml_id_valid(uint32_t id)
1139 // Due to endian nonsense in Matroska, the highest byte with any bits set
1140 // will contain the leading length bit. This bit in turn identifies the
1141 // total byte length of the element by its position within the byte.
1142 unsigned int bits = av_log2(id);
1143 return id && (bits + 7) / 8 == (8 - bits % 8);
1147 * Allocate and return the entry for the level1 element with the given ID. If
1148 * an entry already exists, return the existing entry.
1150 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1151 uint32_t id, int64_t pos)
1154 MatroskaLevel1Element *elem;
1156 if (!is_ebml_id_valid(id))
1159 // Some files link to all clusters; useless.
1160 if (id == MATROSKA_ID_CLUSTER)
1163 // There can be multiple SeekHeads and Tags.
1164 for (i = 0; i < matroska->num_level1_elems; i++) {
1165 if (matroska->level1_elems[i].id == id) {
1166 if (matroska->level1_elems[i].pos == pos ||
1167 id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1168 return &matroska->level1_elems[i];
1172 // Only a completely broken file would have more elements.
1173 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1174 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1178 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1179 *elem = (MatroskaLevel1Element){.id = id};
1184 static int ebml_parse(MatroskaDemuxContext *matroska,
1185 EbmlSyntax *syntax, void *data)
1187 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1188 // Forbid unknown-length EBML_NONE elements.
1189 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1193 // max. 16 MB for strings
1194 [EBML_STR] = 0x1000000,
1195 [EBML_UTF8] = 0x1000000,
1196 // max. 256 MB for binary data
1197 [EBML_BIN] = 0x10000000,
1198 // no limits for anything else
1200 AVIOContext *pb = matroska->ctx->pb;
1203 int64_t pos = avio_tell(pb), pos_alt;
1204 int res, update_pos = 1, level_check;
1205 MatroskaLevel1Element *level1_elem;
1206 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1208 if (!matroska->current_id) {
1210 res = ebml_read_num(matroska, pb, 4, &id, 0);
1212 if (pb->eof_reached && res == AVERROR_EOF) {
1213 if (matroska->is_live)
1214 // in live mode, finish parsing if EOF is reached.
1216 if (level && pos == avio_tell(pb)) {
1217 if (level->length == EBML_UNKNOWN_LENGTH) {
1218 // Unknown-length levels automatically end at EOF.
1219 matroska->num_levels--;
1222 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1223 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1229 matroska->current_id = id | 1 << 7 * res;
1230 pos_alt = pos + res;
1233 pos -= (av_log2(matroska->current_id) + 7) / 8;
1236 id = matroska->current_id;
1238 syntax = ebml_parse_id(syntax, id);
1239 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1240 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1241 // Unknown-length levels end when an element from an upper level
1242 // in the hierarchy is encountered.
1243 while (syntax->def.n) {
1244 syntax = ebml_parse_id(syntax->def.n, id);
1246 matroska->num_levels--;
1252 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1253 "%"PRId64"\n", id, pos);
1254 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1258 data = (char *) data + syntax->data_offset;
1259 if (syntax->list_elem_size) {
1260 EbmlList *list = data;
1263 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1264 return AVERROR(ENOMEM);
1265 newelem = av_fast_realloc(list->elem,
1266 &list->alloc_elem_size,
1267 (list->nb_elem + 1) * syntax->list_elem_size);
1269 return AVERROR(ENOMEM);
1270 list->elem = newelem;
1271 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1272 memset(data, 0, syntax->list_elem_size);
1277 if (syntax->type != EBML_STOP) {
1278 matroska->current_id = 0;
1279 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1284 if (matroska->num_levels > 0) {
1285 if (length != EBML_UNKNOWN_LENGTH &&
1286 level->length != EBML_UNKNOWN_LENGTH) {
1287 uint64_t elem_end = pos_alt + length,
1288 level_end = level->start + level->length;
1290 if (elem_end < level_end) {
1292 } else if (elem_end == level_end) {
1293 level_check = LEVEL_ENDED;
1295 av_log(matroska->ctx, AV_LOG_ERROR,
1296 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1297 "containing master element ending at 0x%"PRIx64"\n",
1298 pos, elem_end, level_end);
1299 return AVERROR_INVALIDDATA;
1301 } else if (length != EBML_UNKNOWN_LENGTH) {
1303 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1304 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1305 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1306 return AVERROR_INVALIDDATA;
1309 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1310 || syntax->type == EBML_NEST)) {
1311 // According to the current specifications only clusters and
1312 // segments are allowed to be unknown-length. We also accept
1313 // other unknown-length master elements.
1314 av_log(matroska->ctx, AV_LOG_WARNING,
1315 "Found unknown-length element 0x%"PRIX32" other than "
1316 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1317 "parsing will nevertheless be attempted.\n", id, pos);
1324 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1325 if (length != EBML_UNKNOWN_LENGTH) {
1326 av_log(matroska->ctx, AV_LOG_ERROR,
1327 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1328 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1329 length, max_lengths[syntax->type], id, pos);
1330 } else if (syntax->type != EBML_NONE) {
1331 av_log(matroska->ctx, AV_LOG_ERROR,
1332 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1333 "unknown length, yet the length of an element of its "
1334 "type must be known.\n", id, pos);
1336 av_log(matroska->ctx, AV_LOG_ERROR,
1337 "Found unknown-length element with ID 0x%"PRIX32" at "
1338 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1339 "available.\n", id, pos);
1341 return AVERROR_INVALIDDATA;
1344 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1345 // Loosing sync will likely manifest itself as encountering unknown
1346 // elements which are not reliably distinguishable from elements
1347 // belonging to future extensions of the format.
1348 // We use a heuristic to detect such situations: If the current
1349 // element is not expected at the current syntax level and there
1350 // were only a few unknown elements in a row, then the element is
1351 // skipped or considered defective based upon the length of the
1352 // current element (i.e. how much would be skipped); if there were
1353 // more than a few skipped elements in a row and skipping the current
1354 // element would lead us more than SKIP_THRESHOLD away from the last
1355 // known good position, then it is inferred that an error occurred.
1356 // The dependency on the number of unknown elements in a row exists
1357 // because the distance to the last known good position is
1358 // automatically big if the last parsed element was big.
1359 // In both cases, each unknown element is considered equivalent to
1360 // UNKNOWN_EQUIV of skipped bytes for the check.
1361 // The whole check is only done for non-seekable output, because
1362 // in this situation skipped data can't simply be rechecked later.
1363 // This is especially important when using unkown length elements
1364 // as the check for whether a child exceeds its containing master
1365 // element is not effective in this situation.
1367 matroska->unknown_count = 0;
1369 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1371 if (matroska->unknown_count > 3)
1372 dist += pos_alt - matroska->resync_pos;
1374 if (dist > SKIP_THRESHOLD) {
1375 av_log(matroska->ctx, AV_LOG_ERROR,
1376 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1377 "length 0x%"PRIx64" considered as invalid data. Last "
1378 "known good position 0x%"PRIx64", %d unknown elements"
1379 " in a row\n", id, pos, length, matroska->resync_pos,
1380 matroska->unknown_count);
1381 return AVERROR_INVALIDDATA;
1386 if (update_pos > 0) {
1387 // We have found an element that is allowed at this place
1388 // in the hierarchy and it passed all checks, so treat the beginning
1389 // of the element as the "last known good" position.
1390 matroska->resync_pos = pos;
1393 if (!data && length != EBML_UNKNOWN_LENGTH)
1397 switch (syntax->type) {
1399 res = ebml_read_uint(pb, length, data);
1402 res = ebml_read_sint(pb, length, data);
1405 res = ebml_read_float(pb, length, data);
1409 res = ebml_read_ascii(pb, length, data);
1412 res = ebml_read_binary(pb, length, pos_alt, data);
1416 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1418 if (id == MATROSKA_ID_SEGMENT)
1419 matroska->segment_start = pos_alt;
1420 if (id == MATROSKA_ID_CUES)
1421 matroska->cues_parsing_deferred = 0;
1422 if (syntax->type == EBML_LEVEL1 &&
1423 (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1424 if (!level1_elem->pos) {
1425 // Zero is not a valid position for a level 1 element.
1426 level1_elem->pos = pos;
1427 } else if (level1_elem->pos != pos)
1428 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1429 level1_elem->parsed = 1;
1431 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1440 if (ffio_limit(pb, length) != length) {
1441 // ffio_limit emits its own error message,
1442 // so we don't have to.
1443 return AVERROR(EIO);
1445 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1446 // avio_skip might take us past EOF. We check for this
1447 // by skipping only length - 1 bytes, reading a byte and
1448 // checking the error flags. This is done in order to check
1449 // that the element has been properly skipped even when
1450 // no filesize (that ffio_limit relies on) is available.
1452 res = NEEDS_CHECKING;
1459 if (res == NEEDS_CHECKING) {
1460 if (pb->eof_reached) {
1469 if (res == AVERROR_INVALIDDATA)
1470 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1471 else if (res == AVERROR(EIO))
1472 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1473 else if (res == AVERROR_EOF) {
1474 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1482 if (syntax->is_counted && data) {
1483 CountedElement *elem = data;
1484 if (elem->count != UINT_MAX)
1488 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1489 level = &matroska->levels[matroska->num_levels - 1];
1490 pos = avio_tell(pb);
1492 // Given that pos >= level->start no check for
1493 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1494 while (matroska->num_levels && pos == level->start + level->length) {
1495 matroska->num_levels--;
1503 static void ebml_free(EbmlSyntax *syntax, void *data)
1506 for (i = 0; syntax[i].id; i++) {
1507 void *data_off = (char *) data + syntax[i].data_offset;
1508 switch (syntax[i].type) {
1514 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1518 if (syntax[i].list_elem_size) {
1519 EbmlList *list = data_off;
1520 char *ptr = list->elem;
1521 for (j = 0; j < list->nb_elem;
1522 j++, ptr += syntax[i].list_elem_size)
1523 ebml_free(syntax[i].def.n, ptr);
1524 av_freep(&list->elem);
1526 list->alloc_elem_size = 0;
1528 ebml_free(syntax[i].def.n, data_off);
1538 static int matroska_probe(const AVProbeData *p)
1541 int len_mask = 0x80, size = 1, n = 1, i;
1544 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1547 /* length of header */
1549 while (size <= 8 && !(total & len_mask)) {
1555 total &= (len_mask - 1);
1557 total = (total << 8) | p->buf[4 + n++];
1559 if (total + 1 == 1ULL << (7 * size)){
1560 /* Unknown-length header - simply parse the whole buffer. */
1561 total = p->buf_size - 4 - size;
1563 /* Does the probe data contain the whole header? */
1564 if (p->buf_size < 4 + size + total)
1568 /* The header should contain a known document type. For now,
1569 * we don't parse the whole header but simply check for the
1570 * availability of that array of characters inside the header.
1571 * Not fully fool-proof, but good enough. */
1572 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1573 size_t probelen = strlen(matroska_doctypes[i]);
1574 if (total < probelen)
1576 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1577 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1578 return AVPROBE_SCORE_MAX;
1581 // probably valid EBML header but no recognized doctype
1582 return AVPROBE_SCORE_EXTENSION;
1585 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1588 MatroskaTrack *tracks = matroska->tracks.elem;
1591 for (i = 0; i < matroska->tracks.nb_elem; i++)
1592 if (tracks[i].num == num)
1595 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1599 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1600 MatroskaTrack *track)
1602 MatroskaTrackEncoding *encodings = track->encodings.elem;
1603 uint8_t *data = *buf;
1604 int isize = *buf_size;
1605 uint8_t *pkt_data = NULL;
1606 uint8_t av_unused *newpktdata;
1607 int pkt_size = isize;
1611 if (pkt_size >= 10000000U)
1612 return AVERROR_INVALIDDATA;
1614 switch (encodings[0].compression.algo) {
1615 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1617 int header_size = encodings[0].compression.settings.size;
1618 uint8_t *header = encodings[0].compression.settings.data;
1620 if (header_size && !header) {
1621 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1628 pkt_size = isize + header_size;
1629 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1631 return AVERROR(ENOMEM);
1633 memcpy(pkt_data, header, header_size);
1634 memcpy(pkt_data + header_size, data, isize);
1638 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1641 olen = pkt_size *= 3;
1642 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1643 + AV_INPUT_BUFFER_PADDING_SIZE);
1645 result = AVERROR(ENOMEM);
1648 pkt_data = newpktdata;
1649 result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1650 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1652 result = AVERROR_INVALIDDATA;
1659 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1661 z_stream zstream = { 0 };
1662 if (inflateInit(&zstream) != Z_OK)
1664 zstream.next_in = data;
1665 zstream.avail_in = isize;
1668 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1670 inflateEnd(&zstream);
1671 result = AVERROR(ENOMEM);
1674 pkt_data = newpktdata;
1675 zstream.avail_out = pkt_size - zstream.total_out;
1676 zstream.next_out = pkt_data + zstream.total_out;
1677 result = inflate(&zstream, Z_NO_FLUSH);
1678 } while (result == Z_OK && pkt_size < 10000000);
1679 pkt_size = zstream.total_out;
1680 inflateEnd(&zstream);
1681 if (result != Z_STREAM_END) {
1682 if (result == Z_MEM_ERROR)
1683 result = AVERROR(ENOMEM);
1685 result = AVERROR_INVALIDDATA;
1692 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1694 bz_stream bzstream = { 0 };
1695 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1697 bzstream.next_in = data;
1698 bzstream.avail_in = isize;
1701 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1703 BZ2_bzDecompressEnd(&bzstream);
1704 result = AVERROR(ENOMEM);
1707 pkt_data = newpktdata;
1708 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1709 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1710 result = BZ2_bzDecompress(&bzstream);
1711 } while (result == BZ_OK && pkt_size < 10000000);
1712 pkt_size = bzstream.total_out_lo32;
1713 BZ2_bzDecompressEnd(&bzstream);
1714 if (result != BZ_STREAM_END) {
1715 if (result == BZ_MEM_ERROR)
1716 result = AVERROR(ENOMEM);
1718 result = AVERROR_INVALIDDATA;
1725 return AVERROR_INVALIDDATA;
1728 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1731 *buf_size = pkt_size;
1739 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1740 AVDictionary **metadata, char *prefix)
1742 MatroskaTag *tags = list->elem;
1746 for (i = 0; i < list->nb_elem; i++) {
1747 const char *lang = tags[i].lang &&
1748 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1750 if (!tags[i].name) {
1751 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1755 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1757 av_strlcpy(key, tags[i].name, sizeof(key));
1758 if (tags[i].def || !lang) {
1759 av_dict_set(metadata, key, tags[i].string, 0);
1760 if (tags[i].sub.nb_elem)
1761 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1764 av_strlcat(key, "-", sizeof(key));
1765 av_strlcat(key, lang, sizeof(key));
1766 av_dict_set(metadata, key, tags[i].string, 0);
1767 if (tags[i].sub.nb_elem)
1768 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1771 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1774 static void matroska_convert_tags(AVFormatContext *s)
1776 MatroskaDemuxContext *matroska = s->priv_data;
1777 MatroskaTags *tags = matroska->tags.elem;
1780 for (i = 0; i < matroska->tags.nb_elem; i++) {
1781 if (tags[i].target.attachuid) {
1782 MatroskaAttachment *attachment = matroska->attachments.elem;
1784 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1785 if (attachment[j].uid == tags[i].target.attachuid &&
1786 attachment[j].stream) {
1787 matroska_convert_tag(s, &tags[i].tag,
1788 &attachment[j].stream->metadata, NULL);
1793 av_log(s, AV_LOG_WARNING,
1794 "The tags at index %d refer to a "
1795 "non-existent attachment %"PRId64".\n",
1796 i, tags[i].target.attachuid);
1798 } else if (tags[i].target.chapteruid) {
1799 MatroskaChapter *chapter = matroska->chapters.elem;
1801 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1802 if (chapter[j].uid == tags[i].target.chapteruid &&
1803 chapter[j].chapter) {
1804 matroska_convert_tag(s, &tags[i].tag,
1805 &chapter[j].chapter->metadata, NULL);
1810 av_log(s, AV_LOG_WARNING,
1811 "The tags at index %d refer to a non-existent chapter "
1813 i, tags[i].target.chapteruid);
1815 } else if (tags[i].target.trackuid) {
1816 MatroskaTrack *track = matroska->tracks.elem;
1818 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1819 if (track[j].uid == tags[i].target.trackuid &&
1821 matroska_convert_tag(s, &tags[i].tag,
1822 &track[j].stream->metadata, NULL);
1827 av_log(s, AV_LOG_WARNING,
1828 "The tags at index %d refer to a non-existent track "
1830 i, tags[i].target.trackuid);
1833 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1834 tags[i].target.type);
1839 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1842 uint32_t saved_id = matroska->current_id;
1843 int64_t before_pos = avio_tell(matroska->ctx->pb);
1847 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1848 /* We don't want to lose our seekhead level, so we add
1849 * a dummy. This is a crude hack. */
1850 if (matroska->num_levels == EBML_MAX_DEPTH) {
1851 av_log(matroska->ctx, AV_LOG_INFO,
1852 "Max EBML element depth (%d) reached, "
1853 "cannot parse further.\n", EBML_MAX_DEPTH);
1854 ret = AVERROR_INVALIDDATA;
1856 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1857 matroska->num_levels++;
1858 matroska->current_id = 0;
1860 ret = ebml_parse(matroska, matroska_segment, matroska);
1861 if (ret == LEVEL_ENDED) {
1862 /* This can only happen if the seek brought us beyond EOF. */
1867 /* Seek back - notice that in all instances where this is used
1868 * it is safe to set the level to 1. */
1869 matroska_reset_status(matroska, saved_id, before_pos);
1874 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1876 EbmlList *seekhead_list = &matroska->seekhead;
1879 // we should not do any seeking in the streaming case
1880 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1883 for (i = 0; i < seekhead_list->nb_elem; i++) {
1884 MatroskaSeekhead *seekheads = seekhead_list->elem;
1885 uint32_t id = seekheads[i].id;
1886 int64_t pos = seekheads[i].pos + matroska->segment_start;
1887 MatroskaLevel1Element *elem;
1889 if (id != seekheads[i].id || pos < matroska->segment_start)
1892 elem = matroska_find_level1_elem(matroska, id, pos);
1893 if (!elem || elem->parsed)
1898 // defer cues parsing until we actually need cue data.
1899 if (id == MATROSKA_ID_CUES)
1902 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1903 // mark index as broken
1904 matroska->cues_parsing_deferred = -1;
1912 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1914 EbmlList *index_list;
1915 MatroskaIndex *index;
1916 uint64_t index_scale = 1;
1919 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1922 index_list = &matroska->index;
1923 index = index_list->elem;
1924 if (index_list->nb_elem < 2)
1926 if (index[1].time > 1E14 / matroska->time_scale) {
1927 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1930 for (i = 0; i < index_list->nb_elem; i++) {
1931 EbmlList *pos_list = &index[i].pos;
1932 MatroskaIndexPos *pos = pos_list->elem;
1933 for (j = 0; j < pos_list->nb_elem; j++) {
1934 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1936 if (track && track->stream)
1937 av_add_index_entry(track->stream,
1938 pos[j].pos + matroska->segment_start,
1939 index[i].time / index_scale, 0, 0,
1945 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1948 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1951 for (i = 0; i < matroska->num_level1_elems; i++) {
1952 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1953 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1954 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1955 matroska->cues_parsing_deferred = -1;
1961 matroska_add_index_entries(matroska);
1964 static int matroska_aac_profile(char *codec_id)
1966 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1969 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1970 if (strstr(codec_id, aac_profiles[profile]))
1975 static int matroska_aac_sri(int samplerate)
1979 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1980 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1985 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1987 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1988 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1991 static int matroska_parse_flac(AVFormatContext *s,
1992 MatroskaTrack *track,
1995 AVStream *st = track->stream;
1996 uint8_t *p = track->codec_priv.data;
1997 int size = track->codec_priv.size;
1999 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
2000 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
2001 track->codec_priv.size = 0;
2005 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
2007 p += track->codec_priv.size;
2008 size -= track->codec_priv.size;
2010 /* parse the remaining metadata blocks if present */
2012 int block_last, block_type, block_size;
2014 flac_parse_block_header(p, &block_last, &block_type, &block_size);
2018 if (block_size > size)
2021 /* check for the channel mask */
2022 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2023 AVDictionary *dict = NULL;
2024 AVDictionaryEntry *chmask;
2026 ff_vorbis_comment(s, &dict, p, block_size, 0);
2027 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2029 uint64_t mask = strtol(chmask->value, NULL, 0);
2030 if (!mask || mask & ~0x3ffffULL) {
2031 av_log(s, AV_LOG_WARNING,
2032 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2034 st->codecpar->channel_layout = mask;
2036 av_dict_free(&dict);
2046 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2048 int minor, micro, bttb = 0;
2050 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2051 * this function, and fixed in 57.52 */
2052 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, µ) == 2)
2053 bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2055 switch (field_order) {
2056 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2057 return AV_FIELD_PROGRESSIVE;
2058 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2059 return AV_FIELD_UNKNOWN;
2060 case MATROSKA_VIDEO_FIELDORDER_TT:
2062 case MATROSKA_VIDEO_FIELDORDER_BB:
2064 case MATROSKA_VIDEO_FIELDORDER_BT:
2065 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2066 case MATROSKA_VIDEO_FIELDORDER_TB:
2067 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2069 return AV_FIELD_UNKNOWN;
2073 static void mkv_stereo_mode_display_mul(int stereo_mode,
2074 int *h_width, int *h_height)
2076 switch (stereo_mode) {
2077 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2078 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2079 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2080 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2081 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2083 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2084 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2085 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2086 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2089 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2090 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2091 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2092 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2098 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2099 const MatroskaTrackVideoColor *color = track->video.color.elem;
2100 const MatroskaMasteringMeta *mastering_meta;
2101 int has_mastering_primaries, has_mastering_luminance;
2103 if (!track->video.color.nb_elem)
2106 mastering_meta = &color->mastering_meta;
2107 // Mastering primaries are CIE 1931 coords, and must be > 0.
2108 has_mastering_primaries =
2109 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2110 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2111 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2112 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2113 has_mastering_luminance = mastering_meta->max_luminance >
2114 mastering_meta->min_luminance.el.f &&
2115 mastering_meta->min_luminance.el.f >= 0 &&
2116 mastering_meta->min_luminance.count;
2118 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2119 st->codecpar->color_space = color->matrix_coefficients;
2120 if (color->primaries != AVCOL_PRI_RESERVED &&
2121 color->primaries != AVCOL_PRI_RESERVED0)
2122 st->codecpar->color_primaries = color->primaries;
2123 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2124 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2125 st->codecpar->color_trc = color->transfer_characteristics;
2126 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2127 color->range <= AVCOL_RANGE_JPEG)
2128 st->codecpar->color_range = color->range;
2129 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2130 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2131 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2132 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2133 st->codecpar->chroma_location =
2134 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2135 (color->chroma_siting_vert - 1) << 7);
2137 if (color->max_cll && color->max_fall) {
2140 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2142 return AVERROR(ENOMEM);
2143 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2144 (uint8_t *)metadata, size);
2146 av_freep(&metadata);
2149 metadata->MaxCLL = color->max_cll;
2150 metadata->MaxFALL = color->max_fall;
2153 if (has_mastering_primaries || has_mastering_luminance) {
2154 AVMasteringDisplayMetadata *metadata =
2155 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2156 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2157 sizeof(AVMasteringDisplayMetadata));
2159 return AVERROR(ENOMEM);
2161 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2162 if (has_mastering_primaries) {
2163 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2164 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2165 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2166 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2167 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2168 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2169 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2170 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2171 metadata->has_primaries = 1;
2173 if (has_mastering_luminance) {
2174 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2175 metadata->min_luminance = av_d2q(mastering_meta->min_luminance.el.f, INT_MAX);
2176 metadata->has_luminance = 1;
2182 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2185 AVSphericalMapping *spherical;
2186 const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2187 const uint8_t *priv_data = mkv_projection->private.data;
2188 enum AVSphericalProjection projection;
2189 size_t spherical_size;
2190 uint32_t l = 0, t = 0, r = 0, b = 0;
2191 uint32_t padding = 0;
2194 if (mkv_projection->private.size && priv_data[0] != 0) {
2195 av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2199 switch (track->video.projection.type) {
2200 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2201 if (track->video.projection.private.size == 20) {
2202 t = AV_RB32(priv_data + 4);
2203 b = AV_RB32(priv_data + 8);
2204 l = AV_RB32(priv_data + 12);
2205 r = AV_RB32(priv_data + 16);
2207 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2208 av_log(logctx, AV_LOG_ERROR,
2209 "Invalid bounding rectangle coordinates "
2210 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2212 return AVERROR_INVALIDDATA;
2214 } else if (track->video.projection.private.size != 0) {
2215 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2216 return AVERROR_INVALIDDATA;
2219 if (l || t || r || b)
2220 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2222 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2224 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2225 if (track->video.projection.private.size < 4) {
2226 av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2227 return AVERROR_INVALIDDATA;
2228 } else if (track->video.projection.private.size == 12) {
2229 uint32_t layout = AV_RB32(priv_data + 4);
2231 av_log(logctx, AV_LOG_WARNING,
2232 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2235 projection = AV_SPHERICAL_CUBEMAP;
2236 padding = AV_RB32(priv_data + 8);
2238 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2239 return AVERROR_INVALIDDATA;
2242 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2243 /* No Spherical metadata */
2246 av_log(logctx, AV_LOG_WARNING,
2247 "Unknown spherical metadata type %"PRIu64"\n",
2248 track->video.projection.type);
2252 spherical = av_spherical_alloc(&spherical_size);
2254 return AVERROR(ENOMEM);
2256 spherical->projection = projection;
2258 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2259 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2260 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2262 spherical->padding = padding;
2264 spherical->bound_left = l;
2265 spherical->bound_top = t;
2266 spherical->bound_right = r;
2267 spherical->bound_bottom = b;
2269 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2272 av_freep(&spherical);
2279 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2281 const AVCodecTag *codec_tags;
2283 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2284 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2286 /* Normalize noncompliant private data that starts with the fourcc
2287 * by expanding/shifting the data by 4 bytes and storing the data
2288 * size at the start. */
2289 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2290 int ret = av_buffer_realloc(&track->codec_priv.buf,
2291 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2295 track->codec_priv.data = track->codec_priv.buf->data;
2296 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2297 track->codec_priv.size += 4;
2298 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2301 *fourcc = AV_RL32(track->codec_priv.data + 4);
2302 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2307 static int matroska_parse_tracks(AVFormatContext *s)
2309 MatroskaDemuxContext *matroska = s->priv_data;
2310 MatroskaTrack *tracks = matroska->tracks.elem;
2315 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2316 MatroskaTrack *track = &tracks[i];
2317 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2318 EbmlList *encodings_list = &track->encodings;
2319 MatroskaTrackEncoding *encodings = encodings_list->elem;
2320 uint8_t *extradata = NULL;
2321 int extradata_size = 0;
2322 int extradata_offset = 0;
2323 uint32_t fourcc = 0;
2325 char* key_id_base64 = NULL;
2328 /* Apply some sanity checks. */
2329 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2330 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2331 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2332 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2333 av_log(matroska->ctx, AV_LOG_INFO,
2334 "Unknown or unsupported track type %"PRIu64"\n",
2338 if (!track->codec_id)
2341 if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
2342 || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
2343 || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2344 || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2346 av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
2350 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2351 isnan(track->audio.samplerate)) {
2352 av_log(matroska->ctx, AV_LOG_WARNING,
2353 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2354 track->audio.samplerate);
2355 track->audio.samplerate = 8000;
2358 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2359 if (!track->default_duration && track->video.frame_rate > 0) {
2360 double default_duration = 1000000000 / track->video.frame_rate;
2361 if (default_duration > UINT64_MAX || default_duration < 0) {
2362 av_log(matroska->ctx, AV_LOG_WARNING,
2363 "Invalid frame rate %e. Cannot calculate default duration.\n",
2364 track->video.frame_rate);
2366 track->default_duration = default_duration;
2369 if (track->video.display_width == -1)
2370 track->video.display_width = track->video.pixel_width;
2371 if (track->video.display_height == -1)
2372 track->video.display_height = track->video.pixel_height;
2373 if (track->video.color_space.size == 4)
2374 fourcc = AV_RL32(track->video.color_space.data);
2375 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2376 if (!track->audio.out_samplerate)
2377 track->audio.out_samplerate = track->audio.samplerate;
2379 if (encodings_list->nb_elem > 1) {
2380 av_log(matroska->ctx, AV_LOG_ERROR,
2381 "Multiple combined encodings not supported");
2382 } else if (encodings_list->nb_elem == 1) {
2383 if (encodings[0].type) {
2384 if (encodings[0].encryption.key_id.size > 0) {
2385 /* Save the encryption key id to be stored later as a
2387 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2388 key_id_base64 = av_malloc(b64_size);
2389 if (key_id_base64 == NULL)
2390 return AVERROR(ENOMEM);
2392 av_base64_encode(key_id_base64, b64_size,
2393 encodings[0].encryption.key_id.data,
2394 encodings[0].encryption.key_id.size);
2396 encodings[0].scope = 0;
2397 av_log(matroska->ctx, AV_LOG_ERROR,
2398 "Unsupported encoding type");
2402 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2405 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2408 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2410 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2411 encodings[0].scope = 0;
2412 av_log(matroska->ctx, AV_LOG_ERROR,
2413 "Unsupported encoding type");
2414 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2415 uint8_t *codec_priv = track->codec_priv.data;
2416 int ret = matroska_decode_buffer(&track->codec_priv.data,
2417 &track->codec_priv.size,
2420 track->codec_priv.data = NULL;
2421 track->codec_priv.size = 0;
2422 av_log(matroska->ctx, AV_LOG_ERROR,
2423 "Failed to decode codec private data\n");
2426 if (codec_priv != track->codec_priv.data) {
2427 av_buffer_unref(&track->codec_priv.buf);
2428 if (track->codec_priv.data) {
2429 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2430 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2432 if (!track->codec_priv.buf) {
2433 av_freep(&track->codec_priv.data);
2434 track->codec_priv.size = 0;
2435 return AVERROR(ENOMEM);
2441 track->needs_decoding = encodings && !encodings[0].type &&
2442 encodings[0].scope & 1 &&
2443 (encodings[0].compression.algo !=
2444 MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP ||
2445 encodings[0].compression.settings.size);
2447 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2448 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2449 strlen(ff_mkv_codec_tags[j].str))) {
2450 codec_id = ff_mkv_codec_tags[j].id;
2455 st = track->stream = avformat_new_stream(s, NULL);
2457 av_free(key_id_base64);
2458 return AVERROR(ENOMEM);
2461 if (key_id_base64) {
2462 /* export encryption key id as base64 metadata tag */
2463 av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2464 AV_DICT_DONT_STRDUP_VAL);
2467 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2468 track->codec_priv.size >= 40 &&
2469 track->codec_priv.data) {
2470 track->ms_compat = 1;
2471 bit_depth = AV_RL16(track->codec_priv.data + 14);
2472 fourcc = AV_RL32(track->codec_priv.data + 16);
2473 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2476 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2478 extradata_offset = 40;
2479 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2480 track->codec_priv.size >= 14 &&
2481 track->codec_priv.data) {
2483 ffio_init_context(&b, track->codec_priv.data,
2484 track->codec_priv.size,
2485 0, NULL, NULL, NULL, NULL);
2486 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2489 codec_id = st->codecpar->codec_id;
2490 fourcc = st->codecpar->codec_tag;
2491 extradata_offset = FFMIN(track->codec_priv.size, 18);
2492 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2493 /* Normally 36, but allow noncompliant private data */
2494 && (track->codec_priv.size >= 32)
2495 && (track->codec_priv.data)) {
2496 uint16_t sample_size;
2497 int ret = get_qt_codec(track, &fourcc, &codec_id);
2500 sample_size = AV_RB16(track->codec_priv.data + 26);
2502 if (sample_size == 8) {
2503 fourcc = MKTAG('r','a','w',' ');
2504 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2505 } else if (sample_size == 16) {
2506 fourcc = MKTAG('t','w','o','s');
2507 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2510 if ((fourcc == MKTAG('t','w','o','s') ||
2511 fourcc == MKTAG('s','o','w','t')) &&
2513 codec_id = AV_CODEC_ID_PCM_S8;
2514 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2515 (track->codec_priv.size >= 21) &&
2516 (track->codec_priv.data)) {
2517 int ret = get_qt_codec(track, &fourcc, &codec_id);
2520 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2521 fourcc = MKTAG('S','V','Q','3');
2522 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2524 if (codec_id == AV_CODEC_ID_NONE)
2525 av_log(matroska->ctx, AV_LOG_ERROR,
2526 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2527 if (track->codec_priv.size >= 86) {
2528 bit_depth = AV_RB16(track->codec_priv.data + 82);
2529 ffio_init_context(&b, track->codec_priv.data,
2530 track->codec_priv.size,
2531 0, NULL, NULL, NULL, NULL);
2532 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2534 track->has_palette = 1;
2537 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2538 switch (track->audio.bitdepth) {
2540 codec_id = AV_CODEC_ID_PCM_U8;
2543 codec_id = AV_CODEC_ID_PCM_S24BE;
2546 codec_id = AV_CODEC_ID_PCM_S32BE;
2549 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2550 switch (track->audio.bitdepth) {
2552 codec_id = AV_CODEC_ID_PCM_U8;
2555 codec_id = AV_CODEC_ID_PCM_S24LE;
2558 codec_id = AV_CODEC_ID_PCM_S32LE;
2561 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2562 track->audio.bitdepth == 64) {
2563 codec_id = AV_CODEC_ID_PCM_F64LE;
2564 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2565 int profile = matroska_aac_profile(track->codec_id);
2566 int sri = matroska_aac_sri(track->audio.samplerate);
2567 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2569 return AVERROR(ENOMEM);
2570 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2571 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2572 if (strstr(track->codec_id, "SBR")) {
2573 sri = matroska_aac_sri(track->audio.out_samplerate);
2574 extradata[2] = 0x56;
2575 extradata[3] = 0xE5;
2576 extradata[4] = 0x80 | (sri << 3);
2580 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2581 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2582 * Create the "atom size", "tag", and "tag version" fields the
2583 * decoder expects manually. */
2584 extradata_size = 12 + track->codec_priv.size;
2585 extradata = av_mallocz(extradata_size +
2586 AV_INPUT_BUFFER_PADDING_SIZE);
2588 return AVERROR(ENOMEM);
2589 AV_WB32(extradata, extradata_size);
2590 memcpy(&extradata[4], "alac", 4);
2591 AV_WB32(&extradata[8], 0);
2592 memcpy(&extradata[12], track->codec_priv.data,
2593 track->codec_priv.size);
2594 } else if (codec_id == AV_CODEC_ID_TTA) {
2596 if (track->audio.channels > UINT16_MAX ||
2597 track->audio.bitdepth > UINT16_MAX) {
2598 av_log(matroska->ctx, AV_LOG_WARNING,
2599 "Too large audio channel number %"PRIu64
2600 " or bitdepth %"PRIu64". Skipping track.\n",
2601 track->audio.channels, track->audio.bitdepth);
2602 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2603 return AVERROR_INVALIDDATA;
2607 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2608 return AVERROR_INVALIDDATA;
2609 extradata_size = 22;
2610 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2612 return AVERROR(ENOMEM);
2614 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2615 bytestream_put_le16(&ptr, 1);
2616 bytestream_put_le16(&ptr, track->audio.channels);
2617 bytestream_put_le16(&ptr, track->audio.bitdepth);
2618 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2619 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2620 track->audio.out_samplerate,
2621 AV_TIME_BASE * 1000));
2622 } else if (codec_id == AV_CODEC_ID_RV10 ||
2623 codec_id == AV_CODEC_ID_RV20 ||
2624 codec_id == AV_CODEC_ID_RV30 ||
2625 codec_id == AV_CODEC_ID_RV40) {
2626 extradata_offset = 26;
2627 } else if (codec_id == AV_CODEC_ID_RA_144) {
2628 track->audio.out_samplerate = 8000;
2629 track->audio.channels = 1;
2630 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2631 codec_id == AV_CODEC_ID_COOK ||
2632 codec_id == AV_CODEC_ID_ATRAC3 ||
2633 codec_id == AV_CODEC_ID_SIPR)
2634 && track->codec_priv.data) {
2637 ffio_init_context(&b, track->codec_priv.data,
2638 track->codec_priv.size,
2639 0, NULL, NULL, NULL, NULL);
2641 flavor = avio_rb16(&b);
2642 track->audio.coded_framesize = avio_rb32(&b);
2644 track->audio.sub_packet_h = avio_rb16(&b);
2645 track->audio.frame_size = avio_rb16(&b);
2646 track->audio.sub_packet_size = avio_rb16(&b);
2647 if (track->audio.coded_framesize <= 0 ||
2648 track->audio.sub_packet_h <= 0 ||
2649 track->audio.frame_size <= 0)
2650 return AVERROR_INVALIDDATA;
2652 if (codec_id == AV_CODEC_ID_RA_288) {
2653 if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2654 != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2655 return AVERROR_INVALIDDATA;
2656 st->codecpar->block_align = track->audio.coded_framesize;
2657 track->codec_priv.size = 0;
2659 if (codec_id == AV_CODEC_ID_SIPR) {
2660 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2662 return AVERROR_INVALIDDATA;
2663 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2664 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2665 } else if (track->audio.sub_packet_size <= 0 ||
2666 track->audio.frame_size % track->audio.sub_packet_size)
2667 return AVERROR_INVALIDDATA;
2668 st->codecpar->block_align = track->audio.sub_packet_size;
2669 extradata_offset = 78;
2671 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2672 track->audio.frame_size);
2673 if (!track->audio.buf)
2674 return AVERROR(ENOMEM);
2675 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2676 ret = matroska_parse_flac(s, track, &extradata_offset);
2679 } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2680 av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2681 "in absence of valid CodecPrivate.\n");
2683 extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2685 return AVERROR(ENOMEM);
2686 AV_WL16(extradata, 0x410);
2687 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2688 fourcc = AV_RL32(track->codec_priv.data);
2689 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2690 /* we don't need any value stored in CodecPrivate.
2691 make sure that it's not exported as extradata. */
2692 track->codec_priv.size = 0;
2693 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2694 /* For now, propagate only the OBUs, if any. Once libavcodec is
2695 updated to handle isobmff style extradata this can be removed. */
2696 extradata_offset = 4;
2698 track->codec_priv.size -= extradata_offset;
2700 if (codec_id == AV_CODEC_ID_NONE)
2701 av_log(matroska->ctx, AV_LOG_INFO,
2702 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2704 if (track->time_scale < 0.01) {
2705 av_log(matroska->ctx, AV_LOG_WARNING,
2706 "Track TimestampScale too small %f, assuming 1.0.\n",
2708 track->time_scale = 1.0;
2710 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2711 1000 * 1000 * 1000); /* 64 bit pts in ns */
2713 /* convert the delay from ns to the track timebase */
2714 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2715 (AVRational){ 1, 1000000000 },
2718 st->codecpar->codec_id = codec_id;
2720 if (strcmp(track->language, "und"))
2721 av_dict_set(&st->metadata, "language", track->language, 0);
2722 av_dict_set(&st->metadata, "title", track->name, 0);
2724 if (track->flag_default)
2725 st->disposition |= AV_DISPOSITION_DEFAULT;
2726 if (track->flag_forced)
2727 st->disposition |= AV_DISPOSITION_FORCED;
2729 if (!st->codecpar->extradata) {
2731 st->codecpar->extradata = extradata;
2732 st->codecpar->extradata_size = extradata_size;
2733 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2734 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2735 return AVERROR(ENOMEM);
2736 memcpy(st->codecpar->extradata,
2737 track->codec_priv.data + extradata_offset,
2738 track->codec_priv.size);
2742 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2743 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2744 int display_width_mul = 1;
2745 int display_height_mul = 1;
2747 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2748 st->codecpar->codec_tag = fourcc;
2750 st->codecpar->bits_per_coded_sample = bit_depth;
2751 st->codecpar->width = track->video.pixel_width;
2752 st->codecpar->height = track->video.pixel_height;
2754 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2755 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2756 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2757 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2759 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2760 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2762 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2763 av_reduce(&st->sample_aspect_ratio.num,
2764 &st->sample_aspect_ratio.den,
2765 st->codecpar->height * track->video.display_width * display_width_mul,
2766 st->codecpar->width * track->video.display_height * display_height_mul,
2769 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2770 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2772 if (track->default_duration) {
2773 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2774 1000000000, track->default_duration, 30000);
2775 #if FF_API_R_FRAME_RATE
2776 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2777 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2778 st->r_frame_rate = st->avg_frame_rate;
2782 /* export stereo mode flag as metadata tag */
2783 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2784 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2786 /* export alpha mode flag as metadata tag */
2787 if (track->video.alpha_mode)
2788 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2790 /* if we have virtual track, mark the real tracks */
2791 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2793 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2795 snprintf(buf, sizeof(buf), "%s_%d",
2796 ff_matroska_video_stereo_plane[planes[j].type], i);
2797 for (k=0; k < matroska->tracks.nb_elem; k++)
2798 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2799 av_dict_set(&tracks[k].stream->metadata,
2800 "stereo_mode", buf, 0);
2804 // add stream level stereo3d side data if it is a supported format
2805 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2806 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2807 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2812 ret = mkv_parse_video_color(st, track);
2815 ret = mkv_parse_video_projection(st, track, matroska->ctx);
2818 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2819 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2820 st->codecpar->codec_tag = fourcc;
2821 st->codecpar->sample_rate = track->audio.out_samplerate;
2822 st->codecpar->channels = track->audio.channels;
2823 if (!st->codecpar->bits_per_coded_sample)
2824 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2825 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2826 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2827 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2828 st->need_parsing = AVSTREAM_PARSE_FULL;
2829 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2830 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2831 if (track->codec_delay > 0) {
2832 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2833 (AVRational){1, 1000000000},
2834 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2835 48000 : st->codecpar->sample_rate});
2837 if (track->seek_preroll > 0) {
2838 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2839 (AVRational){1, 1000000000},
2840 (AVRational){1, st->codecpar->sample_rate});
2842 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2843 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2845 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2846 st->disposition |= AV_DISPOSITION_CAPTIONS;
2847 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2848 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2849 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2850 st->disposition |= AV_DISPOSITION_METADATA;
2852 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2853 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2860 static int matroska_read_header(AVFormatContext *s)
2862 MatroskaDemuxContext *matroska = s->priv_data;
2863 EbmlList *attachments_list = &matroska->attachments;
2864 EbmlList *chapters_list = &matroska->chapters;
2865 MatroskaAttachment *attachments;
2866 MatroskaChapter *chapters;
2867 uint64_t max_start = 0;
2873 matroska->cues_parsing_deferred = 1;
2875 /* First read the EBML header. */
2876 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2877 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2878 ebml_free(ebml_syntax, &ebml);
2879 return AVERROR_INVALIDDATA;
2881 if (ebml.version > EBML_VERSION ||
2882 ebml.max_size > sizeof(uint64_t) ||
2883 ebml.id_length > sizeof(uint32_t) ||
2884 ebml.doctype_version > 3) {
2885 avpriv_report_missing_feature(matroska->ctx,
2886 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2887 ebml.version, ebml.doctype, ebml.doctype_version);
2888 ebml_free(ebml_syntax, &ebml);
2889 return AVERROR_PATCHWELCOME;
2890 } else if (ebml.doctype_version == 3) {
2891 av_log(matroska->ctx, AV_LOG_WARNING,
2892 "EBML header using unsupported features\n"
2893 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2894 ebml.version, ebml.doctype, ebml.doctype_version);
2896 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2897 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2899 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2900 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2901 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2902 ebml_free(ebml_syntax, &ebml);
2903 return AVERROR_INVALIDDATA;
2906 ebml_free(ebml_syntax, &ebml);
2908 /* The next thing is a segment. */
2909 pos = avio_tell(matroska->ctx->pb);
2910 res = ebml_parse(matroska, matroska_segments, matroska);
2911 // Try resyncing until we find an EBML_STOP type element.
2913 res = matroska_resync(matroska, pos);
2916 pos = avio_tell(matroska->ctx->pb);
2917 res = ebml_parse(matroska, matroska_segment, matroska);
2919 /* Set data_offset as it might be needed later by seek_frame_generic. */
2920 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2921 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2922 matroska_execute_seekhead(matroska);
2924 if (!matroska->time_scale)
2925 matroska->time_scale = 1000000;
2926 if (matroska->duration)
2927 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2928 1000 / AV_TIME_BASE;
2929 av_dict_set(&s->metadata, "title", matroska->title, 0);
2930 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2932 if (matroska->date_utc.size == 8)
2933 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2935 res = matroska_parse_tracks(s);
2939 attachments = attachments_list->elem;
2940 for (j = 0; j < attachments_list->nb_elem; j++) {
2941 if (!(attachments[j].filename && attachments[j].mime &&
2942 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2943 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2945 AVStream *st = avformat_new_stream(s, NULL);
2948 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2949 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2950 if (attachments[j].description)
2951 av_dict_set(&st->metadata, "title", attachments[j].description, 0);
2952 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2954 for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2955 if (!strncmp(mkv_image_mime_tags[i].str, attachments[j].mime,
2956 strlen(mkv_image_mime_tags[i].str))) {
2957 st->codecpar->codec_id = mkv_image_mime_tags[i].id;
2962 attachments[j].stream = st;
2964 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2965 AVPacket *pkt = &st->attached_pic;
2967 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2968 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2970 av_init_packet(pkt);
2971 pkt->buf = attachments[j].bin.buf;
2972 attachments[j].bin.buf = NULL;
2973 pkt->data = attachments[j].bin.data;
2974 pkt->size = attachments[j].bin.size;
2975 pkt->stream_index = st->index;
2976 pkt->flags |= AV_PKT_FLAG_KEY;
2978 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2979 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2981 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2982 attachments[j].bin.size);
2984 for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2985 if (!strncmp(mkv_mime_tags[i].str, attachments[j].mime,
2986 strlen(mkv_mime_tags[i].str))) {
2987 st->codecpar->codec_id = mkv_mime_tags[i].id;
2995 chapters = chapters_list->elem;
2996 for (i = 0; i < chapters_list->nb_elem; i++)
2997 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2998 (max_start == 0 || chapters[i].start > max_start)) {
2999 chapters[i].chapter =
3000 avpriv_new_chapter(s, chapters[i].uid,
3001 (AVRational) { 1, 1000000000 },
3002 chapters[i].start, chapters[i].end,
3004 max_start = chapters[i].start;
3007 matroska_add_index_entries(matroska);
3009 matroska_convert_tags(s);
3013 matroska_read_close(s);
3018 * Put one packet in an application-supplied AVPacket struct.
3019 * Returns 0 on success or -1 on failure.
3021 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3024 if (matroska->queue) {
3025 MatroskaTrack *tracks = matroska->tracks.elem;
3026 MatroskaTrack *track;
3028 avpriv_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
3029 track = &tracks[pkt->stream_index];
3030 if (track->has_palette) {
3031 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
3033 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3035 memcpy(pal, track->palette, AVPALETTE_SIZE);
3037 track->has_palette = 0;
3046 * Free all packets in our internal queue.
3048 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3050 avpriv_packet_list_free(&matroska->queue, &matroska->queue_end);
3053 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3054 int size, int type, AVIOContext *pb,
3055 uint32_t lace_size[256], int *laces)
3058 uint8_t *data = *buf;
3062 lace_size[0] = size;
3067 return AVERROR_INVALIDDATA;
3074 case 0x1: /* Xiph lacing */
3078 for (n = 0; n < *laces - 1; n++) {
3083 return AVERROR_INVALIDDATA;
3086 lace_size[n] += temp;
3089 } while (temp == 0xff);
3092 return AVERROR_INVALIDDATA;
3094 lace_size[n] = size - total;
3098 case 0x2: /* fixed-size lacing */
3099 if (size % (*laces))
3100 return AVERROR_INVALIDDATA;
3101 for (n = 0; n < *laces; n++)
3102 lace_size[n] = size / *laces;
3105 case 0x3: /* EBML lacing */
3113 n = ebml_read_num(matroska, pb, 8, &num, 1);
3117 return AVERROR_INVALIDDATA;
3119 total = lace_size[0] = num;
3121 for (n = 1; n < *laces - 1; n++) {
3124 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3127 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3128 return AVERROR_INVALIDDATA;
3130 lace_size[n] = lace_size[n - 1] + snum;
3131 total += lace_size[n];
3137 return AVERROR_INVALIDDATA;
3139 lace_size[*laces - 1] = size - total;
3149 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3150 MatroskaTrack *track, AVStream *st,
3151 uint8_t *data, int size, uint64_t timecode,
3154 const int a = st->codecpar->block_align;
3155 const int sps = track->audio.sub_packet_size;
3156 const int cfs = track->audio.coded_framesize;
3157 const int h = track->audio.sub_packet_h;
3158 const int w = track->audio.frame_size;
3159 int y = track->audio.sub_packet_cnt;
3162 if (!track->audio.pkt_cnt) {
3163 if (track->audio.sub_packet_cnt == 0)
3164 track->audio.buf_timecode = timecode;
3165 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3166 if (size < cfs * h / 2) {
3167 av_log(matroska->ctx, AV_LOG_ERROR,
3168 "Corrupt int4 RM-style audio packet size\n");
3169 return AVERROR_INVALIDDATA;
3171 for (x = 0; x < h / 2; x++)
3172 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3173 data + x * cfs, cfs);
3174 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3176 av_log(matroska->ctx, AV_LOG_ERROR,
3177 "Corrupt sipr RM-style audio packet size\n");
3178 return AVERROR_INVALIDDATA;
3180 memcpy(track->audio.buf + y * w, data, w);
3183 av_log(matroska->ctx, AV_LOG_ERROR,
3184 "Corrupt generic RM-style audio packet size\n");
3185 return AVERROR_INVALIDDATA;
3187 for (x = 0; x < w / sps; x++)
3188 memcpy(track->audio.buf +
3189 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3190 data + x * sps, sps);
3193 if (++track->audio.sub_packet_cnt >= h) {
3194 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3195 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3196 track->audio.sub_packet_cnt = 0;
3197 track->audio.pkt_cnt = h * w / a;
3201 while (track->audio.pkt_cnt) {
3203 AVPacket pktl, *pkt = &pktl;
3205 ret = av_new_packet(pkt, a);
3210 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3212 pkt->pts = track->audio.buf_timecode;
3213 track->audio.buf_timecode = AV_NOPTS_VALUE;
3215 pkt->stream_index = st->index;
3216 ret = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3218 av_packet_unref(pkt);
3219 return AVERROR(ENOMEM);
3226 /* reconstruct full wavpack blocks from mangled matroska ones */
3227 static int matroska_parse_wavpack(MatroskaTrack *track,
3228 uint8_t **data, int *size)
3230 uint8_t *dst = NULL;
3231 uint8_t *src = *data;
3236 int ret, offset = 0;
3239 return AVERROR_INVALIDDATA;
3241 av_assert1(track->stream->codecpar->extradata_size >= 2);
3242 ver = AV_RL16(track->stream->codecpar->extradata);
3244 samples = AV_RL32(src);
3248 while (srclen >= 8) {
3253 uint32_t flags = AV_RL32(src);
3254 uint32_t crc = AV_RL32(src + 4);
3258 multiblock = (flags & 0x1800) != 0x1800;
3261 ret = AVERROR_INVALIDDATA;
3264 blocksize = AV_RL32(src);
3270 if (blocksize > srclen) {
3271 ret = AVERROR_INVALIDDATA;
3275 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3277 ret = AVERROR(ENOMEM);
3281 dstlen += blocksize + 32;
3283 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3284 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3285 AV_WL16(dst + offset + 8, ver); // version
3286 AV_WL16(dst + offset + 10, 0); // track/index_no
3287 AV_WL32(dst + offset + 12, 0); // total samples
3288 AV_WL32(dst + offset + 16, 0); // block index
3289 AV_WL32(dst + offset + 20, samples); // number of samples
3290 AV_WL32(dst + offset + 24, flags); // flags
3291 AV_WL32(dst + offset + 28, crc); // crc
3292 memcpy(dst + offset + 32, src, blocksize); // block data
3295 srclen -= blocksize;
3296 offset += blocksize + 32;
3299 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3311 static int matroska_parse_prores(MatroskaTrack *track,
3312 uint8_t **data, int *size)
3315 int dstlen = *size + 8;
3317 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3319 return AVERROR(ENOMEM);
3321 AV_WB32(dst, dstlen);
3322 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3323 memcpy(dst + 8, *data, dstlen - 8);
3324 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3332 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3333 MatroskaTrack *track,
3335 uint8_t *data, int data_len,
3340 AVPacket pktl, *pkt = &pktl;
3341 uint8_t *id, *settings, *text, *buf;
3342 int id_len, settings_len, text_len;
3347 return AVERROR_INVALIDDATA;
3350 q = data + data_len;
3355 if (*p == '\r' || *p == '\n') {
3364 if (p >= q || *p != '\n')
3365 return AVERROR_INVALIDDATA;
3371 if (*p == '\r' || *p == '\n') {
3372 settings_len = p - settings;
3380 if (p >= q || *p != '\n')
3381 return AVERROR_INVALIDDATA;
3386 while (text_len > 0) {
3387 const int len = text_len - 1;
3388 const uint8_t c = p[len];
3389 if (c != '\r' && c != '\n')
3395 return AVERROR_INVALIDDATA;
3397 err = av_new_packet(pkt, text_len);
3402 memcpy(pkt->data, text, text_len);
3405 buf = av_packet_new_side_data(pkt,
3406 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3409 av_packet_unref(pkt);
3410 return AVERROR(ENOMEM);
3412 memcpy(buf, id, id_len);
3415 if (settings_len > 0) {
3416 buf = av_packet_new_side_data(pkt,
3417 AV_PKT_DATA_WEBVTT_SETTINGS,
3420 av_packet_unref(pkt);
3421 return AVERROR(ENOMEM);
3423 memcpy(buf, settings, settings_len);
3426 // Do we need this for subtitles?
3427 // pkt->flags = AV_PKT_FLAG_KEY;
3429 pkt->stream_index = st->index;
3430 pkt->pts = timecode;
3432 // Do we need this for subtitles?
3433 // pkt->dts = timecode;
3435 pkt->duration = duration;
3438 err = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3440 av_packet_unref(pkt);
3441 return AVERROR(ENOMEM);
3447 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3448 MatroskaTrack *track, AVStream *st,
3449 AVBufferRef *buf, uint8_t *data, int pkt_size,
3450 uint64_t timecode, uint64_t lace_duration,
3451 int64_t pos, int is_keyframe,
3452 uint8_t *additional, uint64_t additional_id, int additional_size,
3453 int64_t discard_padding)
3455 uint8_t *pkt_data = data;
3457 AVPacket pktl, *pkt = &pktl;
3459 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3460 res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3462 av_log(matroska->ctx, AV_LOG_ERROR,
3463 "Error parsing a wavpack block.\n");
3471 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3472 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3473 res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3475 av_log(matroska->ctx, AV_LOG_ERROR,
3476 "Error parsing a prores block.\n");
3484 if (!pkt_size && !additional_size)
3487 av_init_packet(pkt);
3489 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3492 pkt->buf = av_buffer_ref(buf);
3495 res = AVERROR(ENOMEM);
3499 pkt->data = pkt_data;
3500 pkt->size = pkt_size;
3501 pkt->flags = is_keyframe;
3502 pkt->stream_index = st->index;
3504 if (additional_size > 0) {
3505 uint8_t *side_data = av_packet_new_side_data(pkt,
3506 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3507 additional_size + 8);
3509 av_packet_unref(pkt);
3510 return AVERROR(ENOMEM);
3512 AV_WB64(side_data, additional_id);
3513 memcpy(side_data + 8, additional, additional_size);
3516 if (discard_padding) {
3517 uint8_t *side_data = av_packet_new_side_data(pkt,
3518 AV_PKT_DATA_SKIP_SAMPLES,
3521 av_packet_unref(pkt);
3522 return AVERROR(ENOMEM);
3524 discard_padding = av_rescale_q(discard_padding,
3525 (AVRational){1, 1000000000},
3526 (AVRational){1, st->codecpar->sample_rate});
3527 if (discard_padding > 0) {
3528 AV_WL32(side_data + 4, discard_padding);
3530 AV_WL32(side_data, -discard_padding);
3534 if (track->ms_compat)
3535 pkt->dts = timecode;
3537 pkt->pts = timecode;
3539 pkt->duration = lace_duration;
3541 #if FF_API_CONVERGENCE_DURATION
3542 FF_DISABLE_DEPRECATION_WARNINGS
3543 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3544 pkt->convergence_duration = lace_duration;
3546 FF_ENABLE_DEPRECATION_WARNINGS
3549 res = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3551 av_packet_unref(pkt);
3552 return AVERROR(ENOMEM);
3564 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3565 int size, int64_t pos, uint64_t cluster_time,
3566 uint64_t block_duration, int is_keyframe,
3567 uint8_t *additional, uint64_t additional_id, int additional_size,
3568 int64_t cluster_pos, int64_t discard_padding)
3570 uint64_t timecode = AV_NOPTS_VALUE;
3571 MatroskaTrack *track;
3576 uint32_t lace_size[256];
3577 int n, flags, laces = 0;
3579 int trust_default_duration;
3581 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3583 if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3588 track = matroska_find_track_by_num(matroska, num);
3589 if (!track || size < 3)
3590 return AVERROR_INVALIDDATA;
3592 if (!(st = track->stream)) {
3593 av_log(matroska->ctx, AV_LOG_VERBOSE,
3594 "No stream associated to TrackNumber %"PRIu64". "
3595 "Ignoring Block with this TrackNumber.\n", num);
3599 if (st->discard >= AVDISCARD_ALL)
3601 if (block_duration > INT64_MAX)
3602 block_duration = INT64_MAX;
3604 block_time = sign_extend(AV_RB16(data), 16);
3608 if (is_keyframe == -1)
3609 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3611 if (cluster_time != (uint64_t) -1 &&
3612 (block_time >= 0 || cluster_time >= -block_time)) {
3613 uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
3614 timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
3615 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3616 timecode < track->end_timecode)
3617 is_keyframe = 0; /* overlapping subtitles are not key frame */
3619 ff_reduce_index(matroska->ctx, st->index);
3620 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3625 if (matroska->skip_to_keyframe &&
3626 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3627 // Compare signed timecodes. Timecode may be negative due to codec delay
3628 // offset. We don't support timestamps greater than int64_t anyway - see
3630 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3633 matroska->skip_to_keyframe = 0;
3634 else if (!st->internal->skip_to_keyframe) {
3635 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3636 matroska->skip_to_keyframe = 0;
3640 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3641 &pb, lace_size, &laces);
3643 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3647 trust_default_duration = track->default_duration != 0;
3648 if (track->audio.samplerate == 8000 && trust_default_duration) {
3649 // If this is needed for more codecs, then add them here
3650 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3651 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3652 trust_default_duration = 0;
3656 if (!block_duration && trust_default_duration)
3657 block_duration = track->default_duration * laces / matroska->time_scale;
3659 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3660 track->end_timecode =
3661 FFMAX(track->end_timecode, timecode + block_duration);
3663 for (n = 0; n < laces; n++) {
3664 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3665 uint8_t *out_data = data;
3666 int out_size = lace_size[n];
3668 if (track->needs_decoding) {
3669 res = matroska_decode_buffer(&out_data, &out_size, track);
3672 /* Given that we are here means that out_data is no longer
3673 * owned by buf, so set it to NULL. This depends upon
3674 * zero-length header removal compression being ignored. */
3675 av_assert1(out_data != data);
3679 if (track->audio.buf) {
3680 res = matroska_parse_rm_audio(matroska, track, st,
3687 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3688 res = matroska_parse_webvtt(matroska, track, st,
3690 timecode, lace_duration,
3697 res = matroska_parse_frame(matroska, track, st, buf, out_data,
3698 out_size, timecode, lace_duration,
3699 pos, !n ? is_keyframe : 0,
3700 additional, additional_id, additional_size,
3706 if (timecode != AV_NOPTS_VALUE)
3707 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3708 data += lace_size[n];
3714 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3716 MatroskaCluster *cluster = &matroska->current_cluster;
3717 MatroskaBlock *block = &cluster->block;
3720 av_assert0(matroska->num_levels <= 2);
3722 if (matroska->num_levels == 1) {
3723 res = ebml_parse(matroska, matroska_segment, NULL);
3726 /* Found a cluster: subtract the size of the ID already read. */
3727 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3729 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3735 if (matroska->num_levels == 2) {
3736 /* We are inside a cluster. */
3737 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3739 if (res >= 0 && block->bin.size > 0) {
3740 int is_keyframe = block->non_simple ? block->reference.count == 0 : -1;
3741 uint8_t* additional = block->additional.size > 0 ?
3742 block->additional.data : NULL;
3744 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3745 block->bin.size, block->bin.pos,
3746 cluster->timecode, block->duration,
3747 is_keyframe, additional, block->additional_id,
3748 block->additional.size, cluster->pos,
3749 block->discard_padding);
3752 ebml_free(matroska_blockgroup, block);
3753 memset(block, 0, sizeof(*block));
3754 } else if (!matroska->num_levels) {
3755 if (!avio_feof(matroska->ctx->pb)) {
3756 avio_r8(matroska->ctx->pb);
3757 if (!avio_feof(matroska->ctx->pb)) {
3758 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3759 "end of segment.\n");
3760 return AVERROR_INVALIDDATA;
3770 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3772 MatroskaDemuxContext *matroska = s->priv_data;
3775 if (matroska->resync_pos == -1) {
3776 // This can only happen if generic seeking has been used.
3777 matroska->resync_pos = avio_tell(s->pb);
3780 while (matroska_deliver_packet(matroska, pkt)) {
3782 return (ret < 0) ? ret : AVERROR_EOF;
3783 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3784 ret = matroska_resync(matroska, matroska->resync_pos);
3790 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3791 int64_t timestamp, int flags)
3793 MatroskaDemuxContext *matroska = s->priv_data;
3794 MatroskaTrack *tracks = NULL;
3795 AVStream *st = s->streams[stream_index];
3798 /* Parse the CUES now since we need the index data to seek. */
3799 if (matroska->cues_parsing_deferred > 0) {
3800 matroska->cues_parsing_deferred = 0;
3801 matroska_parse_cues(matroska);
3804 if (!st->internal->nb_index_entries)
3806 timestamp = FFMAX(timestamp, st->internal->index_entries[0].timestamp);
3808 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3809 matroska_reset_status(matroska, 0, st->internal->index_entries[st->internal->nb_index_entries - 1].pos);
3810 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3811 matroska_clear_queue(matroska);
3812 if (matroska_parse_cluster(matroska) < 0)
3817 matroska_clear_queue(matroska);
3818 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->internal->nb_index_entries - 1))
3821 tracks = matroska->tracks.elem;
3822 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3823 tracks[i].audio.pkt_cnt = 0;
3824 tracks[i].audio.sub_packet_cnt = 0;
3825 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3826 tracks[i].end_timecode = 0;
3829 /* We seek to a level 1 element, so set the appropriate status. */
3830 matroska_reset_status(matroska, 0, st->internal->index_entries[index].pos);
3831 if (flags & AVSEEK_FLAG_ANY) {
3832 st->internal->skip_to_keyframe = 0;
3833 matroska->skip_to_timecode = timestamp;
3835 st->internal->skip_to_keyframe = 1;
3836 matroska->skip_to_timecode = st->internal->index_entries[index].timestamp;
3838 matroska->skip_to_keyframe = 1;
3840 ff_update_cur_dts(s, st, st->internal->index_entries[index].timestamp);
3843 // slightly hackish but allows proper fallback to
3844 // the generic seeking code.
3845 matroska_reset_status(matroska, 0, -1);
3846 matroska->resync_pos = -1;
3847 matroska_clear_queue(matroska);
3848 st->internal->skip_to_keyframe =
3849 matroska->skip_to_keyframe = 0;
3854 static int matroska_read_close(AVFormatContext *s)
3856 MatroskaDemuxContext *matroska = s->priv_data;
3857 MatroskaTrack *tracks = matroska->tracks.elem;
3860 matroska_clear_queue(matroska);
3862 for (n = 0; n < matroska->tracks.nb_elem; n++)
3863 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3864 av_freep(&tracks[n].audio.buf);
3865 ebml_free(matroska_segment, matroska);
3871 int64_t start_time_ns;
3872 int64_t end_time_ns;
3873 int64_t start_offset;
3877 /* This function searches all the Cues and returns the CueDesc corresponding to
3878 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3879 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3881 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3882 MatroskaDemuxContext *matroska = s->priv_data;
3885 int nb_index_entries = s->streams[0]->internal->nb_index_entries;
3886 AVIndexEntry *index_entries = s->streams[0]->internal->index_entries;
3887 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3888 for (i = 1; i < nb_index_entries; i++) {
3889 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3890 index_entries[i].timestamp * matroska->time_scale > ts) {
3895 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3896 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3897 if (i != nb_index_entries - 1) {
3898 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3899 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3901 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3902 // FIXME: this needs special handling for files where Cues appear
3903 // before Clusters. the current logic assumes Cues appear after
3905 cue_desc.end_offset = cues_start - matroska->segment_start;
3910 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3912 MatroskaDemuxContext *matroska = s->priv_data;
3913 uint32_t id = matroska->current_id;
3914 int64_t cluster_pos, before_pos;
3916 if (s->streams[0]->internal->nb_index_entries <= 0) return 0;
3917 // seek to the first cluster using cues.
3918 index = av_index_search_timestamp(s->streams[0], 0, 0);
3919 if (index < 0) return 0;
3920 cluster_pos = s->streams[0]->internal->index_entries[index].pos;
3921 before_pos = avio_tell(s->pb);
3923 uint64_t cluster_id, cluster_length;
3926 avio_seek(s->pb, cluster_pos, SEEK_SET);
3927 // read cluster id and length
3928 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3929 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3931 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3935 matroska_reset_status(matroska, 0, cluster_pos);
3936 matroska_clear_queue(matroska);
3937 if (matroska_parse_cluster(matroska) < 0 ||
3941 pkt = &matroska->queue->pkt;
3942 // 4 + read is the length of the cluster id and the cluster length field.
3943 cluster_pos += 4 + read + cluster_length;
3944 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3950 /* Restore the status after matroska_read_header: */
3951 matroska_reset_status(matroska, id, before_pos);
3956 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3957 double min_buffer, double* buffer,
3958 double* sec_to_download, AVFormatContext *s,
3961 double nano_seconds_per_second = 1000000000.0;
3962 double time_sec = time_ns / nano_seconds_per_second;
3964 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3965 int64_t end_time_ns = time_ns + time_to_search_ns;
3966 double sec_downloaded = 0.0;
3967 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3968 if (desc_curr.start_time_ns == -1)
3970 *sec_to_download = 0.0;
3972 // Check for non cue start time.
3973 if (time_ns > desc_curr.start_time_ns) {
3974 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3975 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3976 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3977 double timeToDownload = (cueBytes * 8.0) / bps;
3979 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3980 *sec_to_download += timeToDownload;
3982 // Check if the search ends within the first cue.
3983 if (desc_curr.end_time_ns >= end_time_ns) {
3984 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3985 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3986 sec_downloaded = percent_to_sub * sec_downloaded;
3987 *sec_to_download = percent_to_sub * *sec_to_download;
3990 if ((sec_downloaded + *buffer) <= min_buffer) {
3994 // Get the next Cue.
3995 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3998 while (desc_curr.start_time_ns != -1) {
3999 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
4000 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
4001 double desc_sec = desc_ns / nano_seconds_per_second;
4002 double bits = (desc_bytes * 8.0);
4003 double time_to_download = bits / bps;
4005 sec_downloaded += desc_sec - time_to_download;
4006 *sec_to_download += time_to_download;
4008 if (desc_curr.end_time_ns >= end_time_ns) {
4009 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4010 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4011 sec_downloaded = percent_to_sub * sec_downloaded;
4012 *sec_to_download = percent_to_sub * *sec_to_download;
4014 if ((sec_downloaded + *buffer) <= min_buffer)
4019 if ((sec_downloaded + *buffer) <= min_buffer) {
4024 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4026 *buffer = *buffer + sec_downloaded;
4030 /* This function computes the bandwidth of the WebM file with the help of
4031 * buffer_size_after_time_downloaded() function. Both of these functions are
4032 * adapted from WebM Tools project and are adapted to work with FFmpeg's
4033 * Matroska parsing mechanism.
4035 * Returns the bandwidth of the file on success; -1 on error.
4037 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4039 MatroskaDemuxContext *matroska = s->priv_data;
4040 AVStream *st = s->streams[0];
4041 double bandwidth = 0.0;
4044 for (i = 0; i < st->internal->nb_index_entries; i++) {
4045 int64_t prebuffer_ns = 1000000000;
4046 int64_t time_ns = st->internal->index_entries[i].timestamp * matroska->time_scale;
4047 double nano_seconds_per_second = 1000000000.0;
4048 int64_t prebuffered_ns = time_ns + prebuffer_ns;
4049 double prebuffer_bytes = 0.0;
4050 int64_t temp_prebuffer_ns = prebuffer_ns;
4051 int64_t pre_bytes, pre_ns;
4052 double pre_sec, prebuffer, bits_per_second;
4053 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4055 // Start with the first Cue.
4056 CueDesc desc_end = desc_beg;
4058 // Figure out how much data we have downloaded for the prebuffer. This will
4059 // be used later to adjust the bits per sample to try.
4060 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4061 // Prebuffered the entire Cue.
4062 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4063 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4064 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4066 if (desc_end.start_time_ns == -1) {
4067 // The prebuffer is larger than the duration.
4068 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4070 bits_per_second = 0.0;
4072 // The prebuffer ends in the last Cue. Estimate how much data was
4074 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4075 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4076 pre_sec = pre_ns / nano_seconds_per_second;
4078 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4080 prebuffer = prebuffer_ns / nano_seconds_per_second;
4082 // Set this to 0.0 in case our prebuffer buffers the entire video.
4083 bits_per_second = 0.0;
4085 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4086 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4087 double desc_sec = desc_ns / nano_seconds_per_second;
4088 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4090 // Drop the bps by the percentage of bytes buffered.
4091 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4092 double mod_bits_per_second = calc_bits_per_second * percent;
4094 if (prebuffer < desc_sec) {
4096 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4098 // Add 1 so the bits per second should be a little bit greater than file
4100 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4101 const double min_buffer = 0.0;
4102 double buffer = prebuffer;
4103 double sec_to_download = 0.0;
4105 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4106 min_buffer, &buffer, &sec_to_download,
4110 } else if (rv == 0) {
4111 bits_per_second = (double)(bps);
4116 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4117 } while (desc_end.start_time_ns != -1);
4119 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4121 return (int64_t)bandwidth;
4124 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4126 MatroskaDemuxContext *matroska = s->priv_data;
4127 EbmlList *seekhead_list = &matroska->seekhead;
4128 MatroskaSeekhead *seekhead = seekhead_list->elem;
4130 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4134 // determine cues start and end positions
4135 for (i = 0; i < seekhead_list->nb_elem; i++)
4136 if (seekhead[i].id == MATROSKA_ID_CUES)
4139 if (i >= seekhead_list->nb_elem) return -1;
4141 before_pos = avio_tell(matroska->ctx->pb);
4142 cues_start = seekhead[i].pos + matroska->segment_start;
4143 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4144 // cues_end is computed as cues_start + cues_length + length of the
4145 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4146 // cues_end is inclusive and the above sum is reduced by 1.
4147 uint64_t cues_length, cues_id;
4149 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4150 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4151 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4152 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4155 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4157 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4158 if (cues_start == -1 || cues_end == -1) return -1;
4161 matroska_parse_cues(matroska);
4164 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4167 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4169 // if the file has cues at the start, fix up the init range so that
4170 // it does not include it
4171 if (cues_start <= init_range)
4172 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4175 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4176 if (bandwidth < 0) return -1;
4177 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4179 // check if all clusters start with key frames
4180 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4182 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4183 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4184 buf = av_malloc_array(s->streams[0]->internal->nb_index_entries, 20);
4185 if (!buf) return -1;
4187 for (i = 0; i < s->streams[0]->internal->nb_index_entries; i++) {
4188 int ret = snprintf(buf + end, 20,
4189 "%" PRId64"%s", s->streams[0]->internal->index_entries[i].timestamp,
4190 i != s->streams[0]->internal->nb_index_entries - 1 ? "," : "");
4191 if (ret <= 0 || (ret == 20 && i == s->streams[0]->internal->nb_index_entries - 1)) {
4192 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4194 return AVERROR_INVALIDDATA;
4198 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4199 buf, AV_DICT_DONT_STRDUP_VAL);
4204 static int webm_dash_manifest_read_header(AVFormatContext *s)
4207 int ret = matroska_read_header(s);
4209 MatroskaTrack *tracks;
4210 MatroskaDemuxContext *matroska = s->priv_data;
4212 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4215 if (!matroska->tracks.nb_elem || !s->nb_streams) {
4216 av_log(s, AV_LOG_ERROR, "No track found\n");
4217 ret = AVERROR_INVALIDDATA;
4221 if (!matroska->is_live) {
4222 buf = av_asprintf("%g", matroska->duration);
4224 ret = AVERROR(ENOMEM);
4227 av_dict_set(&s->streams[0]->metadata, DURATION,
4228 buf, AV_DICT_DONT_STRDUP_VAL);
4230 // initialization range
4231 // 5 is the offset of Cluster ID.
4232 init_range = avio_tell(s->pb) - 5;
4233 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4236 // basename of the file
4237 buf = strrchr(s->url, '/');
4238 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4241 tracks = matroska->tracks.elem;
4242 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4244 // parse the cues and populate Cue related fields
4245 if (!matroska->is_live) {
4246 ret = webm_dash_manifest_cues(s, init_range);
4248 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4253 // use the bandwidth from the command line if it was provided
4254 if (matroska->bandwidth > 0) {
4255 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4256 matroska->bandwidth, 0);
4260 matroska_read_close(s);
4264 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4269 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4270 static const AVOption options[] = {
4271 { "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 },
4272 { "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 },
4276 static const AVClass webm_dash_class = {
4277 .class_name = "WebM DASH Manifest demuxer",
4278 .item_name = av_default_item_name,
4280 .version = LIBAVUTIL_VERSION_INT,
4283 AVInputFormat ff_matroska_demuxer = {
4284 .name = "matroska,webm",
4285 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4286 .extensions = "mkv,mk3d,mka,mks",
4287 .priv_data_size = sizeof(MatroskaDemuxContext),
4288 .read_probe = matroska_probe,
4289 .read_header = matroska_read_header,
4290 .read_packet = matroska_read_packet,
4291 .read_close = matroska_read_close,
4292 .read_seek = matroska_read_seek,
4293 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4296 AVInputFormat ff_webm_dash_manifest_demuxer = {
4297 .name = "webm_dash_manifest",
4298 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4299 .priv_data_size = sizeof(MatroskaDemuxContext),
4300 .read_header = webm_dash_manifest_read_header,
4301 .read_packet = webm_dash_manifest_read_packet,
4302 .read_close = matroska_read_close,
4303 .priv_class = &webm_dash_class,