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 unless size == 0.
934 static int ebml_read_uint(AVIOContext *pb, int size,
935 uint64_t default_value, uint64_t *num)
940 *num = default_value;
943 /* big-endian ordering; build up number */
946 *num = (*num << 8) | avio_r8(pb);
948 return NEEDS_CHECKING;
952 * Read the next element as a signed int.
953 * Returns NEEDS_CHECKING unless size == 0.
955 static int ebml_read_sint(AVIOContext *pb, int size,
956 int64_t default_value, int64_t *num)
961 *num = default_value;
964 *num = sign_extend(avio_r8(pb), 8);
966 /* big-endian ordering; build up number */
968 *num = ((uint64_t)*num << 8) | avio_r8(pb);
971 return NEEDS_CHECKING;
975 * Read the next element as a float.
976 * Returns 0 if size == 0, NEEDS_CHECKING or < 0 on obvious failure.
978 static int ebml_read_float(AVIOContext *pb, int size,
979 double default_value, double *num)
982 *num = default_value;
984 } else if (size == 4) {
985 *num = av_int2float(avio_rb32(pb));
986 } else if (size == 8) {
987 *num = av_int2double(avio_rb64(pb));
989 return AVERROR_INVALIDDATA;
991 return NEEDS_CHECKING;
995 * Read the next element as an ASCII string.
996 * 0 is success, < 0 or NEEDS_CHECKING is failure.
998 static int ebml_read_ascii(AVIOContext *pb, int size,
999 const char *default_value, char **str)
1004 if (size == 0 && default_value) {
1005 res = av_strdup(default_value);
1007 return AVERROR(ENOMEM);
1009 /* EBML strings are usually not 0-terminated, so we allocate one
1010 * byte more, read the string and NUL-terminate it ourselves. */
1011 if (!(res = av_malloc(size + 1)))
1012 return AVERROR(ENOMEM);
1013 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
1015 return ret < 0 ? ret : NEEDS_CHECKING;
1026 * Read the next element as binary data.
1027 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1029 static int ebml_read_binary(AVIOContext *pb, int length,
1030 int64_t pos, EbmlBin *bin)
1034 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1037 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1039 bin->data = bin->buf->data;
1042 if ((ret = avio_read(pb, bin->data, length)) != length) {
1043 av_buffer_unref(&bin->buf);
1046 return ret < 0 ? ret : NEEDS_CHECKING;
1053 * Read the next element, but only the header. The contents
1054 * are supposed to be sub-elements which can be read separately.
1055 * 0 is success, < 0 is failure.
1057 static int ebml_read_master(MatroskaDemuxContext *matroska,
1058 uint64_t length, int64_t pos)
1060 MatroskaLevel *level;
1062 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1063 av_log(matroska->ctx, AV_LOG_ERROR,
1064 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1065 return AVERROR(ENOSYS);
1068 level = &matroska->levels[matroska->num_levels++];
1070 level->length = length;
1076 * Read a signed "EBML number"
1077 * Return: number of bytes processed, < 0 on error
1079 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1080 AVIOContext *pb, int64_t *num)
1085 /* read as unsigned number first */
1086 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1089 /* make signed (weird way) */
1090 *num = unum - ((1LL << (7 * res - 1)) - 1);
1095 static int ebml_parse(MatroskaDemuxContext *matroska,
1096 EbmlSyntax *syntax, void *data);
1098 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1102 // Whoever touches this should be aware of the duplication
1103 // existing in matroska_cluster_parsing.
1104 for (i = 0; syntax[i].id; i++)
1105 if (id == syntax[i].id)
1111 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1117 for (int i = 0; syntax[i].id; i++) {
1118 void *dst = (char *)data + syntax[i].data_offset;
1119 switch (syntax[i].type) {
1121 *(uint64_t *)dst = syntax[i].def.u;
1124 *(int64_t *) dst = syntax[i].def.i;
1127 *(double *) dst = syntax[i].def.f;
1131 // the default may be NULL
1132 if (syntax[i].def.s) {
1133 *(char**)dst = av_strdup(syntax[i].def.s);
1135 return AVERROR(ENOMEM);
1141 if (!matroska->levels[matroska->num_levels - 1].length) {
1142 matroska->num_levels--;
1148 res = ebml_parse(matroska, syntax, data);
1151 return res == LEVEL_ENDED ? 0 : res;
1154 static int is_ebml_id_valid(uint32_t id)
1156 // Due to endian nonsense in Matroska, the highest byte with any bits set
1157 // will contain the leading length bit. This bit in turn identifies the
1158 // total byte length of the element by its position within the byte.
1159 unsigned int bits = av_log2(id);
1160 return id && (bits + 7) / 8 == (8 - bits % 8);
1164 * Allocate and return the entry for the level1 element with the given ID. If
1165 * an entry already exists, return the existing entry.
1167 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1168 uint32_t id, int64_t pos)
1171 MatroskaLevel1Element *elem;
1173 if (!is_ebml_id_valid(id))
1176 // Some files link to all clusters; useless.
1177 if (id == MATROSKA_ID_CLUSTER)
1180 // There can be multiple SeekHeads and Tags.
1181 for (i = 0; i < matroska->num_level1_elems; i++) {
1182 if (matroska->level1_elems[i].id == id) {
1183 if (matroska->level1_elems[i].pos == pos ||
1184 id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1185 return &matroska->level1_elems[i];
1189 // Only a completely broken file would have more elements.
1190 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1191 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1195 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1196 *elem = (MatroskaLevel1Element){.id = id};
1201 static int ebml_parse(MatroskaDemuxContext *matroska,
1202 EbmlSyntax *syntax, void *data)
1204 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1205 // Forbid unknown-length EBML_NONE elements.
1206 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1210 // max. 16 MB for strings
1211 [EBML_STR] = 0x1000000,
1212 [EBML_UTF8] = 0x1000000,
1213 // max. 256 MB for binary data
1214 [EBML_BIN] = 0x10000000,
1215 // no limits for anything else
1217 AVIOContext *pb = matroska->ctx->pb;
1220 int64_t pos = avio_tell(pb), pos_alt;
1221 int res, update_pos = 1, level_check;
1222 MatroskaLevel1Element *level1_elem;
1223 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1225 if (!matroska->current_id) {
1227 res = ebml_read_num(matroska, pb, 4, &id, 0);
1229 if (pb->eof_reached && res == AVERROR_EOF) {
1230 if (matroska->is_live)
1231 // in live mode, finish parsing if EOF is reached.
1233 if (level && pos == avio_tell(pb)) {
1234 if (level->length == EBML_UNKNOWN_LENGTH) {
1235 // Unknown-length levels automatically end at EOF.
1236 matroska->num_levels--;
1239 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1240 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1246 matroska->current_id = id | 1 << 7 * res;
1247 pos_alt = pos + res;
1250 pos -= (av_log2(matroska->current_id) + 7) / 8;
1253 id = matroska->current_id;
1255 syntax = ebml_parse_id(syntax, id);
1256 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1257 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1258 // Unknown-length levels end when an element from an upper level
1259 // in the hierarchy is encountered.
1260 while (syntax->def.n) {
1261 syntax = ebml_parse_id(syntax->def.n, id);
1263 matroska->num_levels--;
1269 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1270 "%"PRId64"\n", id, pos);
1271 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1275 data = (char *) data + syntax->data_offset;
1276 if (syntax->list_elem_size) {
1277 EbmlList *list = data;
1280 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1281 return AVERROR(ENOMEM);
1282 newelem = av_fast_realloc(list->elem,
1283 &list->alloc_elem_size,
1284 (list->nb_elem + 1) * syntax->list_elem_size);
1286 return AVERROR(ENOMEM);
1287 list->elem = newelem;
1288 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1289 memset(data, 0, syntax->list_elem_size);
1294 if (syntax->type != EBML_STOP) {
1295 matroska->current_id = 0;
1296 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1301 if (matroska->num_levels > 0) {
1302 if (length != EBML_UNKNOWN_LENGTH &&
1303 level->length != EBML_UNKNOWN_LENGTH) {
1304 uint64_t elem_end = pos_alt + length,
1305 level_end = level->start + level->length;
1307 if (elem_end < level_end) {
1309 } else if (elem_end == level_end) {
1310 level_check = LEVEL_ENDED;
1312 av_log(matroska->ctx, AV_LOG_ERROR,
1313 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1314 "containing master element ending at 0x%"PRIx64"\n",
1315 pos, elem_end, level_end);
1316 return AVERROR_INVALIDDATA;
1318 } else if (length != EBML_UNKNOWN_LENGTH) {
1320 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1321 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1322 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1323 return AVERROR_INVALIDDATA;
1326 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1327 || syntax->type == EBML_NEST)) {
1328 // According to the current specifications only clusters and
1329 // segments are allowed to be unknown-length. We also accept
1330 // other unknown-length master elements.
1331 av_log(matroska->ctx, AV_LOG_WARNING,
1332 "Found unknown-length element 0x%"PRIX32" other than "
1333 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1334 "parsing will nevertheless be attempted.\n", id, pos);
1341 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1342 if (length != EBML_UNKNOWN_LENGTH) {
1343 av_log(matroska->ctx, AV_LOG_ERROR,
1344 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1345 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1346 length, max_lengths[syntax->type], id, pos);
1347 } else if (syntax->type != EBML_NONE) {
1348 av_log(matroska->ctx, AV_LOG_ERROR,
1349 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1350 "unknown length, yet the length of an element of its "
1351 "type must be known.\n", id, pos);
1353 av_log(matroska->ctx, AV_LOG_ERROR,
1354 "Found unknown-length element with ID 0x%"PRIX32" at "
1355 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1356 "available.\n", id, pos);
1358 return AVERROR_INVALIDDATA;
1361 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1362 // Loosing sync will likely manifest itself as encountering unknown
1363 // elements which are not reliably distinguishable from elements
1364 // belonging to future extensions of the format.
1365 // We use a heuristic to detect such situations: If the current
1366 // element is not expected at the current syntax level and there
1367 // were only a few unknown elements in a row, then the element is
1368 // skipped or considered defective based upon the length of the
1369 // current element (i.e. how much would be skipped); if there were
1370 // more than a few skipped elements in a row and skipping the current
1371 // element would lead us more than SKIP_THRESHOLD away from the last
1372 // known good position, then it is inferred that an error occurred.
1373 // The dependency on the number of unknown elements in a row exists
1374 // because the distance to the last known good position is
1375 // automatically big if the last parsed element was big.
1376 // In both cases, each unknown element is considered equivalent to
1377 // UNKNOWN_EQUIV of skipped bytes for the check.
1378 // The whole check is only done for non-seekable output, because
1379 // in this situation skipped data can't simply be rechecked later.
1380 // This is especially important when using unkown length elements
1381 // as the check for whether a child exceeds its containing master
1382 // element is not effective in this situation.
1384 matroska->unknown_count = 0;
1386 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1388 if (matroska->unknown_count > 3)
1389 dist += pos_alt - matroska->resync_pos;
1391 if (dist > SKIP_THRESHOLD) {
1392 av_log(matroska->ctx, AV_LOG_ERROR,
1393 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1394 "length 0x%"PRIx64" considered as invalid data. Last "
1395 "known good position 0x%"PRIx64", %d unknown elements"
1396 " in a row\n", id, pos, length, matroska->resync_pos,
1397 matroska->unknown_count);
1398 return AVERROR_INVALIDDATA;
1403 if (update_pos > 0) {
1404 // We have found an element that is allowed at this place
1405 // in the hierarchy and it passed all checks, so treat the beginning
1406 // of the element as the "last known good" position.
1407 matroska->resync_pos = pos;
1410 if (!data && length != EBML_UNKNOWN_LENGTH)
1414 switch (syntax->type) {
1416 res = ebml_read_uint(pb, length, syntax->def.u, data);
1419 res = ebml_read_sint(pb, length, syntax->def.i, data);
1422 res = ebml_read_float(pb, length, syntax->def.f, data);
1426 res = ebml_read_ascii(pb, length, syntax->def.s, data);
1429 res = ebml_read_binary(pb, length, pos_alt, data);
1433 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1435 if (id == MATROSKA_ID_SEGMENT)
1436 matroska->segment_start = pos_alt;
1437 if (id == MATROSKA_ID_CUES)
1438 matroska->cues_parsing_deferred = 0;
1439 if (syntax->type == EBML_LEVEL1 &&
1440 (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1441 if (!level1_elem->pos) {
1442 // Zero is not a valid position for a level 1 element.
1443 level1_elem->pos = pos;
1444 } else if (level1_elem->pos != pos)
1445 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1446 level1_elem->parsed = 1;
1448 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1457 if (ffio_limit(pb, length) != length) {
1458 // ffio_limit emits its own error message,
1459 // so we don't have to.
1460 return AVERROR(EIO);
1462 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1463 // avio_skip might take us past EOF. We check for this
1464 // by skipping only length - 1 bytes, reading a byte and
1465 // checking the error flags. This is done in order to check
1466 // that the element has been properly skipped even when
1467 // no filesize (that ffio_limit relies on) is available.
1469 res = NEEDS_CHECKING;
1476 if (res == NEEDS_CHECKING) {
1477 if (pb->eof_reached) {
1486 if (res == AVERROR_INVALIDDATA)
1487 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1488 else if (res == AVERROR(EIO))
1489 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1490 else if (res == AVERROR_EOF) {
1491 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1499 if (syntax->is_counted && data) {
1500 CountedElement *elem = data;
1501 if (elem->count != UINT_MAX)
1505 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1506 level = &matroska->levels[matroska->num_levels - 1];
1507 pos = avio_tell(pb);
1509 // Given that pos >= level->start no check for
1510 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1511 while (matroska->num_levels && pos == level->start + level->length) {
1512 matroska->num_levels--;
1520 static void ebml_free(EbmlSyntax *syntax, void *data)
1523 for (i = 0; syntax[i].id; i++) {
1524 void *data_off = (char *) data + syntax[i].data_offset;
1525 switch (syntax[i].type) {
1531 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1535 if (syntax[i].list_elem_size) {
1536 EbmlList *list = data_off;
1537 char *ptr = list->elem;
1538 for (j = 0; j < list->nb_elem;
1539 j++, ptr += syntax[i].list_elem_size)
1540 ebml_free(syntax[i].def.n, ptr);
1541 av_freep(&list->elem);
1543 list->alloc_elem_size = 0;
1545 ebml_free(syntax[i].def.n, data_off);
1555 static int matroska_probe(const AVProbeData *p)
1558 int len_mask = 0x80, size = 1, n = 1, i;
1561 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1564 /* length of header */
1566 while (size <= 8 && !(total & len_mask)) {
1572 total &= (len_mask - 1);
1574 total = (total << 8) | p->buf[4 + n++];
1576 if (total + 1 == 1ULL << (7 * size)){
1577 /* Unknown-length header - simply parse the whole buffer. */
1578 total = p->buf_size - 4 - size;
1580 /* Does the probe data contain the whole header? */
1581 if (p->buf_size < 4 + size + total)
1585 /* The header should contain a known document type. For now,
1586 * we don't parse the whole header but simply check for the
1587 * availability of that array of characters inside the header.
1588 * Not fully fool-proof, but good enough. */
1589 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1590 size_t probelen = strlen(matroska_doctypes[i]);
1591 if (total < probelen)
1593 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1594 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1595 return AVPROBE_SCORE_MAX;
1598 // probably valid EBML header but no recognized doctype
1599 return AVPROBE_SCORE_EXTENSION;
1602 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1605 MatroskaTrack *tracks = matroska->tracks.elem;
1608 for (i = 0; i < matroska->tracks.nb_elem; i++)
1609 if (tracks[i].num == num)
1612 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1616 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1617 MatroskaTrack *track)
1619 MatroskaTrackEncoding *encodings = track->encodings.elem;
1620 uint8_t *data = *buf;
1621 int isize = *buf_size;
1622 uint8_t *pkt_data = NULL;
1623 uint8_t av_unused *newpktdata;
1624 int pkt_size = isize;
1628 if (pkt_size >= 10000000U)
1629 return AVERROR_INVALIDDATA;
1631 switch (encodings[0].compression.algo) {
1632 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1634 int header_size = encodings[0].compression.settings.size;
1635 uint8_t *header = encodings[0].compression.settings.data;
1637 if (header_size && !header) {
1638 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1645 pkt_size = isize + header_size;
1646 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1648 return AVERROR(ENOMEM);
1650 memcpy(pkt_data, header, header_size);
1651 memcpy(pkt_data + header_size, data, isize);
1655 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1658 olen = pkt_size *= 3;
1659 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1660 + AV_INPUT_BUFFER_PADDING_SIZE);
1662 result = AVERROR(ENOMEM);
1665 pkt_data = newpktdata;
1666 result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1667 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1669 result = AVERROR_INVALIDDATA;
1676 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1678 z_stream zstream = { 0 };
1679 if (inflateInit(&zstream) != Z_OK)
1681 zstream.next_in = data;
1682 zstream.avail_in = isize;
1685 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1687 inflateEnd(&zstream);
1688 result = AVERROR(ENOMEM);
1691 pkt_data = newpktdata;
1692 zstream.avail_out = pkt_size - zstream.total_out;
1693 zstream.next_out = pkt_data + zstream.total_out;
1694 result = inflate(&zstream, Z_NO_FLUSH);
1695 } while (result == Z_OK && pkt_size < 10000000);
1696 pkt_size = zstream.total_out;
1697 inflateEnd(&zstream);
1698 if (result != Z_STREAM_END) {
1699 if (result == Z_MEM_ERROR)
1700 result = AVERROR(ENOMEM);
1702 result = AVERROR_INVALIDDATA;
1709 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1711 bz_stream bzstream = { 0 };
1712 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1714 bzstream.next_in = data;
1715 bzstream.avail_in = isize;
1718 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1720 BZ2_bzDecompressEnd(&bzstream);
1721 result = AVERROR(ENOMEM);
1724 pkt_data = newpktdata;
1725 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1726 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1727 result = BZ2_bzDecompress(&bzstream);
1728 } while (result == BZ_OK && pkt_size < 10000000);
1729 pkt_size = bzstream.total_out_lo32;
1730 BZ2_bzDecompressEnd(&bzstream);
1731 if (result != BZ_STREAM_END) {
1732 if (result == BZ_MEM_ERROR)
1733 result = AVERROR(ENOMEM);
1735 result = AVERROR_INVALIDDATA;
1742 return AVERROR_INVALIDDATA;
1745 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1748 *buf_size = pkt_size;
1756 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1757 AVDictionary **metadata, char *prefix)
1759 MatroskaTag *tags = list->elem;
1763 for (i = 0; i < list->nb_elem; i++) {
1764 const char *lang = tags[i].lang &&
1765 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1767 if (!tags[i].name) {
1768 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1772 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1774 av_strlcpy(key, tags[i].name, sizeof(key));
1775 if (tags[i].def || !lang) {
1776 av_dict_set(metadata, key, tags[i].string, 0);
1777 if (tags[i].sub.nb_elem)
1778 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1781 av_strlcat(key, "-", sizeof(key));
1782 av_strlcat(key, lang, sizeof(key));
1783 av_dict_set(metadata, key, tags[i].string, 0);
1784 if (tags[i].sub.nb_elem)
1785 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1788 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1791 static void matroska_convert_tags(AVFormatContext *s)
1793 MatroskaDemuxContext *matroska = s->priv_data;
1794 MatroskaTags *tags = matroska->tags.elem;
1797 for (i = 0; i < matroska->tags.nb_elem; i++) {
1798 if (tags[i].target.attachuid) {
1799 MatroskaAttachment *attachment = matroska->attachments.elem;
1801 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1802 if (attachment[j].uid == tags[i].target.attachuid &&
1803 attachment[j].stream) {
1804 matroska_convert_tag(s, &tags[i].tag,
1805 &attachment[j].stream->metadata, NULL);
1810 av_log(s, AV_LOG_WARNING,
1811 "The tags at index %d refer to a "
1812 "non-existent attachment %"PRId64".\n",
1813 i, tags[i].target.attachuid);
1815 } else if (tags[i].target.chapteruid) {
1816 MatroskaChapter *chapter = matroska->chapters.elem;
1818 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1819 if (chapter[j].uid == tags[i].target.chapteruid &&
1820 chapter[j].chapter) {
1821 matroska_convert_tag(s, &tags[i].tag,
1822 &chapter[j].chapter->metadata, NULL);
1827 av_log(s, AV_LOG_WARNING,
1828 "The tags at index %d refer to a non-existent chapter "
1830 i, tags[i].target.chapteruid);
1832 } else if (tags[i].target.trackuid) {
1833 MatroskaTrack *track = matroska->tracks.elem;
1835 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1836 if (track[j].uid == tags[i].target.trackuid &&
1838 matroska_convert_tag(s, &tags[i].tag,
1839 &track[j].stream->metadata, NULL);
1844 av_log(s, AV_LOG_WARNING,
1845 "The tags at index %d refer to a non-existent track "
1847 i, tags[i].target.trackuid);
1850 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1851 tags[i].target.type);
1856 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1859 uint32_t saved_id = matroska->current_id;
1860 int64_t before_pos = avio_tell(matroska->ctx->pb);
1864 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1865 /* We don't want to lose our seekhead level, so we add
1866 * a dummy. This is a crude hack. */
1867 if (matroska->num_levels == EBML_MAX_DEPTH) {
1868 av_log(matroska->ctx, AV_LOG_INFO,
1869 "Max EBML element depth (%d) reached, "
1870 "cannot parse further.\n", EBML_MAX_DEPTH);
1871 ret = AVERROR_INVALIDDATA;
1873 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1874 matroska->num_levels++;
1875 matroska->current_id = 0;
1877 ret = ebml_parse(matroska, matroska_segment, matroska);
1878 if (ret == LEVEL_ENDED) {
1879 /* This can only happen if the seek brought us beyond EOF. */
1884 /* Seek back - notice that in all instances where this is used
1885 * it is safe to set the level to 1. */
1886 matroska_reset_status(matroska, saved_id, before_pos);
1891 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1893 EbmlList *seekhead_list = &matroska->seekhead;
1896 // we should not do any seeking in the streaming case
1897 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1900 for (i = 0; i < seekhead_list->nb_elem; i++) {
1901 MatroskaSeekhead *seekheads = seekhead_list->elem;
1902 uint32_t id = seekheads[i].id;
1903 int64_t pos = seekheads[i].pos + matroska->segment_start;
1904 MatroskaLevel1Element *elem;
1906 if (id != seekheads[i].id || pos < matroska->segment_start)
1909 elem = matroska_find_level1_elem(matroska, id, pos);
1910 if (!elem || elem->parsed)
1915 // defer cues parsing until we actually need cue data.
1916 if (id == MATROSKA_ID_CUES)
1919 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1920 // mark index as broken
1921 matroska->cues_parsing_deferred = -1;
1929 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1931 EbmlList *index_list;
1932 MatroskaIndex *index;
1933 uint64_t index_scale = 1;
1936 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1939 index_list = &matroska->index;
1940 index = index_list->elem;
1941 if (index_list->nb_elem < 2)
1943 if (index[1].time > 1E14 / matroska->time_scale) {
1944 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1947 for (i = 0; i < index_list->nb_elem; i++) {
1948 EbmlList *pos_list = &index[i].pos;
1949 MatroskaIndexPos *pos = pos_list->elem;
1950 for (j = 0; j < pos_list->nb_elem; j++) {
1951 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1953 if (track && track->stream)
1954 av_add_index_entry(track->stream,
1955 pos[j].pos + matroska->segment_start,
1956 index[i].time / index_scale, 0, 0,
1962 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1965 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1968 for (i = 0; i < matroska->num_level1_elems; i++) {
1969 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1970 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1971 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1972 matroska->cues_parsing_deferred = -1;
1978 matroska_add_index_entries(matroska);
1981 static int matroska_aac_profile(char *codec_id)
1983 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1986 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1987 if (strstr(codec_id, aac_profiles[profile]))
1992 static int matroska_aac_sri(int samplerate)
1996 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1997 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
2002 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
2004 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
2005 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
2008 static int matroska_parse_flac(AVFormatContext *s,
2009 MatroskaTrack *track,
2012 AVStream *st = track->stream;
2013 uint8_t *p = track->codec_priv.data;
2014 int size = track->codec_priv.size;
2016 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
2017 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
2018 track->codec_priv.size = 0;
2022 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
2024 p += track->codec_priv.size;
2025 size -= track->codec_priv.size;
2027 /* parse the remaining metadata blocks if present */
2029 int block_last, block_type, block_size;
2031 flac_parse_block_header(p, &block_last, &block_type, &block_size);
2035 if (block_size > size)
2038 /* check for the channel mask */
2039 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2040 AVDictionary *dict = NULL;
2041 AVDictionaryEntry *chmask;
2043 ff_vorbis_comment(s, &dict, p, block_size, 0);
2044 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2046 uint64_t mask = strtol(chmask->value, NULL, 0);
2047 if (!mask || mask & ~0x3ffffULL) {
2048 av_log(s, AV_LOG_WARNING,
2049 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2051 st->codecpar->channel_layout = mask;
2053 av_dict_free(&dict);
2063 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2065 int minor, micro, bttb = 0;
2067 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2068 * this function, and fixed in 57.52 */
2069 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, µ) == 2)
2070 bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2072 switch (field_order) {
2073 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2074 return AV_FIELD_PROGRESSIVE;
2075 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2076 return AV_FIELD_UNKNOWN;
2077 case MATROSKA_VIDEO_FIELDORDER_TT:
2079 case MATROSKA_VIDEO_FIELDORDER_BB:
2081 case MATROSKA_VIDEO_FIELDORDER_BT:
2082 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2083 case MATROSKA_VIDEO_FIELDORDER_TB:
2084 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2086 return AV_FIELD_UNKNOWN;
2090 static void mkv_stereo_mode_display_mul(int stereo_mode,
2091 int *h_width, int *h_height)
2093 switch (stereo_mode) {
2094 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2095 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2096 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2097 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2098 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2100 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2101 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2102 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2103 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2106 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2107 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2108 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2109 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2115 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2116 const MatroskaTrackVideoColor *color = track->video.color.elem;
2117 const MatroskaMasteringMeta *mastering_meta;
2118 int has_mastering_primaries, has_mastering_luminance;
2120 if (!track->video.color.nb_elem)
2123 mastering_meta = &color->mastering_meta;
2124 // Mastering primaries are CIE 1931 coords, and must be > 0.
2125 has_mastering_primaries =
2126 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2127 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2128 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2129 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2130 has_mastering_luminance = mastering_meta->max_luminance >
2131 mastering_meta->min_luminance.el.f &&
2132 mastering_meta->min_luminance.el.f >= 0 &&
2133 mastering_meta->min_luminance.count;
2135 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2136 st->codecpar->color_space = color->matrix_coefficients;
2137 if (color->primaries != AVCOL_PRI_RESERVED &&
2138 color->primaries != AVCOL_PRI_RESERVED0)
2139 st->codecpar->color_primaries = color->primaries;
2140 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2141 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2142 st->codecpar->color_trc = color->transfer_characteristics;
2143 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2144 color->range <= AVCOL_RANGE_JPEG)
2145 st->codecpar->color_range = color->range;
2146 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2147 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2148 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2149 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2150 st->codecpar->chroma_location =
2151 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2152 (color->chroma_siting_vert - 1) << 7);
2154 if (color->max_cll && color->max_fall) {
2157 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2159 return AVERROR(ENOMEM);
2160 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2161 (uint8_t *)metadata, size);
2163 av_freep(&metadata);
2166 metadata->MaxCLL = color->max_cll;
2167 metadata->MaxFALL = color->max_fall;
2170 if (has_mastering_primaries || has_mastering_luminance) {
2171 AVMasteringDisplayMetadata *metadata =
2172 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2173 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2174 sizeof(AVMasteringDisplayMetadata));
2176 return AVERROR(ENOMEM);
2178 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2179 if (has_mastering_primaries) {
2180 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2181 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2182 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2183 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2184 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2185 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2186 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2187 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2188 metadata->has_primaries = 1;
2190 if (has_mastering_luminance) {
2191 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2192 metadata->min_luminance = av_d2q(mastering_meta->min_luminance.el.f, INT_MAX);
2193 metadata->has_luminance = 1;
2199 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2202 AVSphericalMapping *spherical;
2203 const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2204 const uint8_t *priv_data = mkv_projection->private.data;
2205 enum AVSphericalProjection projection;
2206 size_t spherical_size;
2207 uint32_t l = 0, t = 0, r = 0, b = 0;
2208 uint32_t padding = 0;
2211 if (mkv_projection->private.size && priv_data[0] != 0) {
2212 av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2216 switch (track->video.projection.type) {
2217 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2218 if (track->video.projection.private.size == 20) {
2219 t = AV_RB32(priv_data + 4);
2220 b = AV_RB32(priv_data + 8);
2221 l = AV_RB32(priv_data + 12);
2222 r = AV_RB32(priv_data + 16);
2224 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2225 av_log(logctx, AV_LOG_ERROR,
2226 "Invalid bounding rectangle coordinates "
2227 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2229 return AVERROR_INVALIDDATA;
2231 } else if (track->video.projection.private.size != 0) {
2232 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2233 return AVERROR_INVALIDDATA;
2236 if (l || t || r || b)
2237 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2239 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2241 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2242 if (track->video.projection.private.size < 4) {
2243 av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2244 return AVERROR_INVALIDDATA;
2245 } else if (track->video.projection.private.size == 12) {
2246 uint32_t layout = AV_RB32(priv_data + 4);
2248 av_log(logctx, AV_LOG_WARNING,
2249 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2252 projection = AV_SPHERICAL_CUBEMAP;
2253 padding = AV_RB32(priv_data + 8);
2255 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2256 return AVERROR_INVALIDDATA;
2259 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2260 /* No Spherical metadata */
2263 av_log(logctx, AV_LOG_WARNING,
2264 "Unknown spherical metadata type %"PRIu64"\n",
2265 track->video.projection.type);
2269 spherical = av_spherical_alloc(&spherical_size);
2271 return AVERROR(ENOMEM);
2273 spherical->projection = projection;
2275 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2276 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2277 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2279 spherical->padding = padding;
2281 spherical->bound_left = l;
2282 spherical->bound_top = t;
2283 spherical->bound_right = r;
2284 spherical->bound_bottom = b;
2286 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2289 av_freep(&spherical);
2296 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2298 const AVCodecTag *codec_tags;
2300 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2301 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2303 /* Normalize noncompliant private data that starts with the fourcc
2304 * by expanding/shifting the data by 4 bytes and storing the data
2305 * size at the start. */
2306 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2307 int ret = av_buffer_realloc(&track->codec_priv.buf,
2308 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2312 track->codec_priv.data = track->codec_priv.buf->data;
2313 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2314 track->codec_priv.size += 4;
2315 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2318 *fourcc = AV_RL32(track->codec_priv.data + 4);
2319 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2324 static int matroska_parse_tracks(AVFormatContext *s)
2326 MatroskaDemuxContext *matroska = s->priv_data;
2327 MatroskaTrack *tracks = matroska->tracks.elem;
2332 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2333 MatroskaTrack *track = &tracks[i];
2334 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2335 EbmlList *encodings_list = &track->encodings;
2336 MatroskaTrackEncoding *encodings = encodings_list->elem;
2337 uint8_t *extradata = NULL;
2338 int extradata_size = 0;
2339 int extradata_offset = 0;
2340 uint32_t fourcc = 0;
2342 char* key_id_base64 = NULL;
2345 /* Apply some sanity checks. */
2346 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2347 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2348 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2349 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2350 av_log(matroska->ctx, AV_LOG_INFO,
2351 "Unknown or unsupported track type %"PRIu64"\n",
2355 if (!track->codec_id)
2358 if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
2359 || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
2360 || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2361 || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2363 av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
2367 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2368 isnan(track->audio.samplerate)) {
2369 av_log(matroska->ctx, AV_LOG_WARNING,
2370 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2371 track->audio.samplerate);
2372 track->audio.samplerate = 8000;
2375 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2376 if (!track->default_duration && track->video.frame_rate > 0) {
2377 double default_duration = 1000000000 / track->video.frame_rate;
2378 if (default_duration > UINT64_MAX || default_duration < 0) {
2379 av_log(matroska->ctx, AV_LOG_WARNING,
2380 "Invalid frame rate %e. Cannot calculate default duration.\n",
2381 track->video.frame_rate);
2383 track->default_duration = default_duration;
2386 if (track->video.display_width == -1)
2387 track->video.display_width = track->video.pixel_width;
2388 if (track->video.display_height == -1)
2389 track->video.display_height = track->video.pixel_height;
2390 if (track->video.color_space.size == 4)
2391 fourcc = AV_RL32(track->video.color_space.data);
2392 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2393 if (!track->audio.out_samplerate)
2394 track->audio.out_samplerate = track->audio.samplerate;
2396 if (encodings_list->nb_elem > 1) {
2397 av_log(matroska->ctx, AV_LOG_ERROR,
2398 "Multiple combined encodings not supported");
2399 } else if (encodings_list->nb_elem == 1) {
2400 if (encodings[0].type) {
2401 if (encodings[0].encryption.key_id.size > 0) {
2402 /* Save the encryption key id to be stored later as a
2404 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2405 key_id_base64 = av_malloc(b64_size);
2406 if (key_id_base64 == NULL)
2407 return AVERROR(ENOMEM);
2409 av_base64_encode(key_id_base64, b64_size,
2410 encodings[0].encryption.key_id.data,
2411 encodings[0].encryption.key_id.size);
2413 encodings[0].scope = 0;
2414 av_log(matroska->ctx, AV_LOG_ERROR,
2415 "Unsupported encoding type");
2419 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2422 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2425 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2427 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2428 encodings[0].scope = 0;
2429 av_log(matroska->ctx, AV_LOG_ERROR,
2430 "Unsupported encoding type");
2431 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2432 uint8_t *codec_priv = track->codec_priv.data;
2433 int ret = matroska_decode_buffer(&track->codec_priv.data,
2434 &track->codec_priv.size,
2437 track->codec_priv.data = NULL;
2438 track->codec_priv.size = 0;
2439 av_log(matroska->ctx, AV_LOG_ERROR,
2440 "Failed to decode codec private data\n");
2443 if (codec_priv != track->codec_priv.data) {
2444 av_buffer_unref(&track->codec_priv.buf);
2445 if (track->codec_priv.data) {
2446 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2447 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2449 if (!track->codec_priv.buf) {
2450 av_freep(&track->codec_priv.data);
2451 track->codec_priv.size = 0;
2452 return AVERROR(ENOMEM);
2458 track->needs_decoding = encodings && !encodings[0].type &&
2459 encodings[0].scope & 1 &&
2460 (encodings[0].compression.algo !=
2461 MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP ||
2462 encodings[0].compression.settings.size);
2464 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2465 if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2466 strlen(ff_mkv_codec_tags[j].str))) {
2467 codec_id = ff_mkv_codec_tags[j].id;
2472 st = track->stream = avformat_new_stream(s, NULL);
2474 av_free(key_id_base64);
2475 return AVERROR(ENOMEM);
2478 if (key_id_base64) {
2479 /* export encryption key id as base64 metadata tag */
2480 av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2481 AV_DICT_DONT_STRDUP_VAL);
2484 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2485 track->codec_priv.size >= 40 &&
2486 track->codec_priv.data) {
2487 track->ms_compat = 1;
2488 bit_depth = AV_RL16(track->codec_priv.data + 14);
2489 fourcc = AV_RL32(track->codec_priv.data + 16);
2490 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2493 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2495 extradata_offset = 40;
2496 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2497 track->codec_priv.size >= 14 &&
2498 track->codec_priv.data) {
2500 ffio_init_context(&b, track->codec_priv.data,
2501 track->codec_priv.size,
2502 0, NULL, NULL, NULL, NULL);
2503 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2506 codec_id = st->codecpar->codec_id;
2507 fourcc = st->codecpar->codec_tag;
2508 extradata_offset = FFMIN(track->codec_priv.size, 18);
2509 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2510 /* Normally 36, but allow noncompliant private data */
2511 && (track->codec_priv.size >= 32)
2512 && (track->codec_priv.data)) {
2513 uint16_t sample_size;
2514 int ret = get_qt_codec(track, &fourcc, &codec_id);
2517 sample_size = AV_RB16(track->codec_priv.data + 26);
2519 if (sample_size == 8) {
2520 fourcc = MKTAG('r','a','w',' ');
2521 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2522 } else if (sample_size == 16) {
2523 fourcc = MKTAG('t','w','o','s');
2524 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2527 if ((fourcc == MKTAG('t','w','o','s') ||
2528 fourcc == MKTAG('s','o','w','t')) &&
2530 codec_id = AV_CODEC_ID_PCM_S8;
2531 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2532 (track->codec_priv.size >= 21) &&
2533 (track->codec_priv.data)) {
2534 int ret = get_qt_codec(track, &fourcc, &codec_id);
2537 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2538 fourcc = MKTAG('S','V','Q','3');
2539 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2541 if (codec_id == AV_CODEC_ID_NONE)
2542 av_log(matroska->ctx, AV_LOG_ERROR,
2543 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2544 if (track->codec_priv.size >= 86) {
2545 bit_depth = AV_RB16(track->codec_priv.data + 82);
2546 ffio_init_context(&b, track->codec_priv.data,
2547 track->codec_priv.size,
2548 0, NULL, NULL, NULL, NULL);
2549 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2551 track->has_palette = 1;
2554 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2555 switch (track->audio.bitdepth) {
2557 codec_id = AV_CODEC_ID_PCM_U8;
2560 codec_id = AV_CODEC_ID_PCM_S24BE;
2563 codec_id = AV_CODEC_ID_PCM_S32BE;
2566 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2567 switch (track->audio.bitdepth) {
2569 codec_id = AV_CODEC_ID_PCM_U8;
2572 codec_id = AV_CODEC_ID_PCM_S24LE;
2575 codec_id = AV_CODEC_ID_PCM_S32LE;
2578 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2579 track->audio.bitdepth == 64) {
2580 codec_id = AV_CODEC_ID_PCM_F64LE;
2581 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2582 int profile = matroska_aac_profile(track->codec_id);
2583 int sri = matroska_aac_sri(track->audio.samplerate);
2584 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2586 return AVERROR(ENOMEM);
2587 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2588 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2589 if (strstr(track->codec_id, "SBR")) {
2590 sri = matroska_aac_sri(track->audio.out_samplerate);
2591 extradata[2] = 0x56;
2592 extradata[3] = 0xE5;
2593 extradata[4] = 0x80 | (sri << 3);
2597 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2598 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2599 * Create the "atom size", "tag", and "tag version" fields the
2600 * decoder expects manually. */
2601 extradata_size = 12 + track->codec_priv.size;
2602 extradata = av_mallocz(extradata_size +
2603 AV_INPUT_BUFFER_PADDING_SIZE);
2605 return AVERROR(ENOMEM);
2606 AV_WB32(extradata, extradata_size);
2607 memcpy(&extradata[4], "alac", 4);
2608 AV_WB32(&extradata[8], 0);
2609 memcpy(&extradata[12], track->codec_priv.data,
2610 track->codec_priv.size);
2611 } else if (codec_id == AV_CODEC_ID_TTA) {
2613 if (track->audio.channels > UINT16_MAX ||
2614 track->audio.bitdepth > UINT16_MAX) {
2615 av_log(matroska->ctx, AV_LOG_WARNING,
2616 "Too large audio channel number %"PRIu64
2617 " or bitdepth %"PRIu64". Skipping track.\n",
2618 track->audio.channels, track->audio.bitdepth);
2619 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2620 return AVERROR_INVALIDDATA;
2624 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2625 return AVERROR_INVALIDDATA;
2626 extradata_size = 22;
2627 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2629 return AVERROR(ENOMEM);
2631 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2632 bytestream_put_le16(&ptr, 1);
2633 bytestream_put_le16(&ptr, track->audio.channels);
2634 bytestream_put_le16(&ptr, track->audio.bitdepth);
2635 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2636 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2637 track->audio.out_samplerate,
2638 AV_TIME_BASE * 1000));
2639 } else if (codec_id == AV_CODEC_ID_RV10 ||
2640 codec_id == AV_CODEC_ID_RV20 ||
2641 codec_id == AV_CODEC_ID_RV30 ||
2642 codec_id == AV_CODEC_ID_RV40) {
2643 extradata_offset = 26;
2644 } else if (codec_id == AV_CODEC_ID_RA_144) {
2645 track->audio.out_samplerate = 8000;
2646 track->audio.channels = 1;
2647 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2648 codec_id == AV_CODEC_ID_COOK ||
2649 codec_id == AV_CODEC_ID_ATRAC3 ||
2650 codec_id == AV_CODEC_ID_SIPR)
2651 && track->codec_priv.data) {
2654 ffio_init_context(&b, track->codec_priv.data,
2655 track->codec_priv.size,
2656 0, NULL, NULL, NULL, NULL);
2658 flavor = avio_rb16(&b);
2659 track->audio.coded_framesize = avio_rb32(&b);
2661 track->audio.sub_packet_h = avio_rb16(&b);
2662 track->audio.frame_size = avio_rb16(&b);
2663 track->audio.sub_packet_size = avio_rb16(&b);
2664 if (track->audio.coded_framesize <= 0 ||
2665 track->audio.sub_packet_h <= 0 ||
2666 track->audio.frame_size <= 0)
2667 return AVERROR_INVALIDDATA;
2669 if (codec_id == AV_CODEC_ID_RA_288) {
2670 if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2671 != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2672 return AVERROR_INVALIDDATA;
2673 st->codecpar->block_align = track->audio.coded_framesize;
2674 track->codec_priv.size = 0;
2676 if (codec_id == AV_CODEC_ID_SIPR) {
2677 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2679 return AVERROR_INVALIDDATA;
2680 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2681 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2682 } else if (track->audio.sub_packet_size <= 0 ||
2683 track->audio.frame_size % track->audio.sub_packet_size)
2684 return AVERROR_INVALIDDATA;
2685 st->codecpar->block_align = track->audio.sub_packet_size;
2686 extradata_offset = 78;
2688 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2689 track->audio.frame_size);
2690 if (!track->audio.buf)
2691 return AVERROR(ENOMEM);
2692 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2693 ret = matroska_parse_flac(s, track, &extradata_offset);
2696 } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2697 av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2698 "in absence of valid CodecPrivate.\n");
2700 extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2702 return AVERROR(ENOMEM);
2703 AV_WL16(extradata, 0x410);
2704 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2705 fourcc = AV_RL32(track->codec_priv.data);
2706 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2707 /* we don't need any value stored in CodecPrivate.
2708 make sure that it's not exported as extradata. */
2709 track->codec_priv.size = 0;
2710 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2711 /* For now, propagate only the OBUs, if any. Once libavcodec is
2712 updated to handle isobmff style extradata this can be removed. */
2713 extradata_offset = 4;
2715 track->codec_priv.size -= extradata_offset;
2717 if (codec_id == AV_CODEC_ID_NONE)
2718 av_log(matroska->ctx, AV_LOG_INFO,
2719 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2721 if (track->time_scale < 0.01) {
2722 av_log(matroska->ctx, AV_LOG_WARNING,
2723 "Track TimestampScale too small %f, assuming 1.0.\n",
2725 track->time_scale = 1.0;
2727 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2728 1000 * 1000 * 1000); /* 64 bit pts in ns */
2730 /* convert the delay from ns to the track timebase */
2731 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2732 (AVRational){ 1, 1000000000 },
2735 st->codecpar->codec_id = codec_id;
2737 if (strcmp(track->language, "und"))
2738 av_dict_set(&st->metadata, "language", track->language, 0);
2739 av_dict_set(&st->metadata, "title", track->name, 0);
2741 if (track->flag_default)
2742 st->disposition |= AV_DISPOSITION_DEFAULT;
2743 if (track->flag_forced)
2744 st->disposition |= AV_DISPOSITION_FORCED;
2746 if (!st->codecpar->extradata) {
2748 st->codecpar->extradata = extradata;
2749 st->codecpar->extradata_size = extradata_size;
2750 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2751 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2752 return AVERROR(ENOMEM);
2753 memcpy(st->codecpar->extradata,
2754 track->codec_priv.data + extradata_offset,
2755 track->codec_priv.size);
2759 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2760 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2761 int display_width_mul = 1;
2762 int display_height_mul = 1;
2764 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2765 st->codecpar->codec_tag = fourcc;
2767 st->codecpar->bits_per_coded_sample = bit_depth;
2768 st->codecpar->width = track->video.pixel_width;
2769 st->codecpar->height = track->video.pixel_height;
2771 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2772 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2773 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2774 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2776 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2777 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2779 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2780 av_reduce(&st->sample_aspect_ratio.num,
2781 &st->sample_aspect_ratio.den,
2782 st->codecpar->height * track->video.display_width * display_width_mul,
2783 st->codecpar->width * track->video.display_height * display_height_mul,
2786 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2787 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2789 if (track->default_duration) {
2790 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2791 1000000000, track->default_duration, 30000);
2792 #if FF_API_R_FRAME_RATE
2793 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2794 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2795 st->r_frame_rate = st->avg_frame_rate;
2799 /* export stereo mode flag as metadata tag */
2800 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2801 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2803 /* export alpha mode flag as metadata tag */
2804 if (track->video.alpha_mode)
2805 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2807 /* if we have virtual track, mark the real tracks */
2808 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2810 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2812 snprintf(buf, sizeof(buf), "%s_%d",
2813 ff_matroska_video_stereo_plane[planes[j].type], i);
2814 for (k=0; k < matroska->tracks.nb_elem; k++)
2815 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2816 av_dict_set(&tracks[k].stream->metadata,
2817 "stereo_mode", buf, 0);
2821 // add stream level stereo3d side data if it is a supported format
2822 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2823 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2824 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2829 ret = mkv_parse_video_color(st, track);
2832 ret = mkv_parse_video_projection(st, track, matroska->ctx);
2835 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2836 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2837 st->codecpar->codec_tag = fourcc;
2838 st->codecpar->sample_rate = track->audio.out_samplerate;
2839 st->codecpar->channels = track->audio.channels;
2840 if (!st->codecpar->bits_per_coded_sample)
2841 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2842 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2843 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2844 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2845 st->need_parsing = AVSTREAM_PARSE_FULL;
2846 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2847 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2848 if (track->codec_delay > 0) {
2849 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2850 (AVRational){1, 1000000000},
2851 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2852 48000 : st->codecpar->sample_rate});
2854 if (track->seek_preroll > 0) {
2855 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2856 (AVRational){1, 1000000000},
2857 (AVRational){1, st->codecpar->sample_rate});
2859 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2860 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2862 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2863 st->disposition |= AV_DISPOSITION_CAPTIONS;
2864 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2865 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2866 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2867 st->disposition |= AV_DISPOSITION_METADATA;
2869 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2870 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2877 static int matroska_read_header(AVFormatContext *s)
2879 MatroskaDemuxContext *matroska = s->priv_data;
2880 EbmlList *attachments_list = &matroska->attachments;
2881 EbmlList *chapters_list = &matroska->chapters;
2882 MatroskaAttachment *attachments;
2883 MatroskaChapter *chapters;
2884 uint64_t max_start = 0;
2890 matroska->cues_parsing_deferred = 1;
2892 /* First read the EBML header. */
2893 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2894 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2895 ebml_free(ebml_syntax, &ebml);
2896 return AVERROR_INVALIDDATA;
2898 if (ebml.version > EBML_VERSION ||
2899 ebml.max_size > sizeof(uint64_t) ||
2900 ebml.id_length > sizeof(uint32_t) ||
2901 ebml.doctype_version > 3) {
2902 avpriv_report_missing_feature(matroska->ctx,
2903 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2904 ebml.version, ebml.doctype, ebml.doctype_version);
2905 ebml_free(ebml_syntax, &ebml);
2906 return AVERROR_PATCHWELCOME;
2907 } else if (ebml.doctype_version == 3) {
2908 av_log(matroska->ctx, AV_LOG_WARNING,
2909 "EBML header using unsupported features\n"
2910 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2911 ebml.version, ebml.doctype, ebml.doctype_version);
2913 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2914 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2916 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2917 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2918 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2919 ebml_free(ebml_syntax, &ebml);
2920 return AVERROR_INVALIDDATA;
2923 ebml_free(ebml_syntax, &ebml);
2925 /* The next thing is a segment. */
2926 pos = avio_tell(matroska->ctx->pb);
2927 res = ebml_parse(matroska, matroska_segments, matroska);
2928 // Try resyncing until we find an EBML_STOP type element.
2930 res = matroska_resync(matroska, pos);
2933 pos = avio_tell(matroska->ctx->pb);
2934 res = ebml_parse(matroska, matroska_segment, matroska);
2936 /* Set data_offset as it might be needed later by seek_frame_generic. */
2937 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2938 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2939 matroska_execute_seekhead(matroska);
2941 if (!matroska->time_scale)
2942 matroska->time_scale = 1000000;
2943 if (matroska->duration)
2944 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2945 1000 / AV_TIME_BASE;
2946 av_dict_set(&s->metadata, "title", matroska->title, 0);
2947 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2949 if (matroska->date_utc.size == 8)
2950 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2952 res = matroska_parse_tracks(s);
2956 attachments = attachments_list->elem;
2957 for (j = 0; j < attachments_list->nb_elem; j++) {
2958 if (!(attachments[j].filename && attachments[j].mime &&
2959 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2960 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2962 AVStream *st = avformat_new_stream(s, NULL);
2965 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2966 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2967 if (attachments[j].description)
2968 av_dict_set(&st->metadata, "title", attachments[j].description, 0);
2969 st->codecpar->codec_id = AV_CODEC_ID_NONE;
2971 for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2972 if (!strncmp(mkv_image_mime_tags[i].str, attachments[j].mime,
2973 strlen(mkv_image_mime_tags[i].str))) {
2974 st->codecpar->codec_id = mkv_image_mime_tags[i].id;
2979 attachments[j].stream = st;
2981 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2982 AVPacket *pkt = &st->attached_pic;
2984 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
2985 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2987 av_init_packet(pkt);
2988 pkt->buf = attachments[j].bin.buf;
2989 attachments[j].bin.buf = NULL;
2990 pkt->data = attachments[j].bin.data;
2991 pkt->size = attachments[j].bin.size;
2992 pkt->stream_index = st->index;
2993 pkt->flags |= AV_PKT_FLAG_KEY;
2995 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
2996 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2998 memcpy(st->codecpar->extradata, attachments[j].bin.data,
2999 attachments[j].bin.size);
3001 for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3002 if (!strncmp(mkv_mime_tags[i].str, attachments[j].mime,
3003 strlen(mkv_mime_tags[i].str))) {
3004 st->codecpar->codec_id = mkv_mime_tags[i].id;
3012 chapters = chapters_list->elem;
3013 for (i = 0; i < chapters_list->nb_elem; i++)
3014 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
3015 (max_start == 0 || chapters[i].start > max_start)) {
3016 chapters[i].chapter =
3017 avpriv_new_chapter(s, chapters[i].uid,
3018 (AVRational) { 1, 1000000000 },
3019 chapters[i].start, chapters[i].end,
3021 max_start = chapters[i].start;
3024 matroska_add_index_entries(matroska);
3026 matroska_convert_tags(s);
3030 matroska_read_close(s);
3035 * Put one packet in an application-supplied AVPacket struct.
3036 * Returns 0 on success or -1 on failure.
3038 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3041 if (matroska->queue) {
3042 MatroskaTrack *tracks = matroska->tracks.elem;
3043 MatroskaTrack *track;
3045 avpriv_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
3046 track = &tracks[pkt->stream_index];
3047 if (track->has_palette) {
3048 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
3050 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3052 memcpy(pal, track->palette, AVPALETTE_SIZE);
3054 track->has_palette = 0;
3063 * Free all packets in our internal queue.
3065 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3067 avpriv_packet_list_free(&matroska->queue, &matroska->queue_end);
3070 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3071 int size, int type, AVIOContext *pb,
3072 uint32_t lace_size[256], int *laces)
3075 uint8_t *data = *buf;
3079 lace_size[0] = size;
3084 return AVERROR_INVALIDDATA;
3091 case 0x1: /* Xiph lacing */
3095 for (n = 0; n < *laces - 1; n++) {
3100 return AVERROR_INVALIDDATA;
3103 lace_size[n] += temp;
3106 } while (temp == 0xff);
3109 return AVERROR_INVALIDDATA;
3111 lace_size[n] = size - total;
3115 case 0x2: /* fixed-size lacing */
3116 if (size % (*laces))
3117 return AVERROR_INVALIDDATA;
3118 for (n = 0; n < *laces; n++)
3119 lace_size[n] = size / *laces;
3122 case 0x3: /* EBML lacing */
3130 n = ebml_read_num(matroska, pb, 8, &num, 1);
3134 return AVERROR_INVALIDDATA;
3136 total = lace_size[0] = num;
3138 for (n = 1; n < *laces - 1; n++) {
3141 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3144 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3145 return AVERROR_INVALIDDATA;
3147 lace_size[n] = lace_size[n - 1] + snum;
3148 total += lace_size[n];
3154 return AVERROR_INVALIDDATA;
3156 lace_size[*laces - 1] = size - total;
3166 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3167 MatroskaTrack *track, AVStream *st,
3168 uint8_t *data, int size, uint64_t timecode,
3171 const int a = st->codecpar->block_align;
3172 const int sps = track->audio.sub_packet_size;
3173 const int cfs = track->audio.coded_framesize;
3174 const int h = track->audio.sub_packet_h;
3175 const int w = track->audio.frame_size;
3176 int y = track->audio.sub_packet_cnt;
3179 if (!track->audio.pkt_cnt) {
3180 if (track->audio.sub_packet_cnt == 0)
3181 track->audio.buf_timecode = timecode;
3182 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3183 if (size < cfs * h / 2) {
3184 av_log(matroska->ctx, AV_LOG_ERROR,
3185 "Corrupt int4 RM-style audio packet size\n");
3186 return AVERROR_INVALIDDATA;
3188 for (x = 0; x < h / 2; x++)
3189 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3190 data + x * cfs, cfs);
3191 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3193 av_log(matroska->ctx, AV_LOG_ERROR,
3194 "Corrupt sipr RM-style audio packet size\n");
3195 return AVERROR_INVALIDDATA;
3197 memcpy(track->audio.buf + y * w, data, w);
3200 av_log(matroska->ctx, AV_LOG_ERROR,
3201 "Corrupt generic RM-style audio packet size\n");
3202 return AVERROR_INVALIDDATA;
3204 for (x = 0; x < w / sps; x++)
3205 memcpy(track->audio.buf +
3206 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3207 data + x * sps, sps);
3210 if (++track->audio.sub_packet_cnt >= h) {
3211 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3212 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3213 track->audio.sub_packet_cnt = 0;
3214 track->audio.pkt_cnt = h * w / a;
3218 while (track->audio.pkt_cnt) {
3220 AVPacket pktl, *pkt = &pktl;
3222 ret = av_new_packet(pkt, a);
3227 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3229 pkt->pts = track->audio.buf_timecode;
3230 track->audio.buf_timecode = AV_NOPTS_VALUE;
3232 pkt->stream_index = st->index;
3233 ret = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3235 av_packet_unref(pkt);
3236 return AVERROR(ENOMEM);
3243 /* reconstruct full wavpack blocks from mangled matroska ones */
3244 static int matroska_parse_wavpack(MatroskaTrack *track,
3245 uint8_t **data, int *size)
3247 uint8_t *dst = NULL;
3248 uint8_t *src = *data;
3253 int ret, offset = 0;
3256 return AVERROR_INVALIDDATA;
3258 av_assert1(track->stream->codecpar->extradata_size >= 2);
3259 ver = AV_RL16(track->stream->codecpar->extradata);
3261 samples = AV_RL32(src);
3265 while (srclen >= 8) {
3270 uint32_t flags = AV_RL32(src);
3271 uint32_t crc = AV_RL32(src + 4);
3275 multiblock = (flags & 0x1800) != 0x1800;
3278 ret = AVERROR_INVALIDDATA;
3281 blocksize = AV_RL32(src);
3287 if (blocksize > srclen) {
3288 ret = AVERROR_INVALIDDATA;
3292 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3294 ret = AVERROR(ENOMEM);
3298 dstlen += blocksize + 32;
3300 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3301 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3302 AV_WL16(dst + offset + 8, ver); // version
3303 AV_WL16(dst + offset + 10, 0); // track/index_no
3304 AV_WL32(dst + offset + 12, 0); // total samples
3305 AV_WL32(dst + offset + 16, 0); // block index
3306 AV_WL32(dst + offset + 20, samples); // number of samples
3307 AV_WL32(dst + offset + 24, flags); // flags
3308 AV_WL32(dst + offset + 28, crc); // crc
3309 memcpy(dst + offset + 32, src, blocksize); // block data
3312 srclen -= blocksize;
3313 offset += blocksize + 32;
3316 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3328 static int matroska_parse_prores(MatroskaTrack *track,
3329 uint8_t **data, int *size)
3332 int dstlen = *size + 8;
3334 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3336 return AVERROR(ENOMEM);
3338 AV_WB32(dst, dstlen);
3339 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3340 memcpy(dst + 8, *data, dstlen - 8);
3341 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3349 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3350 MatroskaTrack *track,
3352 uint8_t *data, int data_len,
3357 AVPacket pktl, *pkt = &pktl;
3358 uint8_t *id, *settings, *text, *buf;
3359 int id_len, settings_len, text_len;
3364 return AVERROR_INVALIDDATA;
3367 q = data + data_len;
3372 if (*p == '\r' || *p == '\n') {
3381 if (p >= q || *p != '\n')
3382 return AVERROR_INVALIDDATA;
3388 if (*p == '\r' || *p == '\n') {
3389 settings_len = p - settings;
3397 if (p >= q || *p != '\n')
3398 return AVERROR_INVALIDDATA;
3403 while (text_len > 0) {
3404 const int len = text_len - 1;
3405 const uint8_t c = p[len];
3406 if (c != '\r' && c != '\n')
3412 return AVERROR_INVALIDDATA;
3414 err = av_new_packet(pkt, text_len);
3419 memcpy(pkt->data, text, text_len);
3422 buf = av_packet_new_side_data(pkt,
3423 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3426 av_packet_unref(pkt);
3427 return AVERROR(ENOMEM);
3429 memcpy(buf, id, id_len);
3432 if (settings_len > 0) {
3433 buf = av_packet_new_side_data(pkt,
3434 AV_PKT_DATA_WEBVTT_SETTINGS,
3437 av_packet_unref(pkt);
3438 return AVERROR(ENOMEM);
3440 memcpy(buf, settings, settings_len);
3443 // Do we need this for subtitles?
3444 // pkt->flags = AV_PKT_FLAG_KEY;
3446 pkt->stream_index = st->index;
3447 pkt->pts = timecode;
3449 // Do we need this for subtitles?
3450 // pkt->dts = timecode;
3452 pkt->duration = duration;
3455 err = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3457 av_packet_unref(pkt);
3458 return AVERROR(ENOMEM);
3464 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3465 MatroskaTrack *track, AVStream *st,
3466 AVBufferRef *buf, uint8_t *data, int pkt_size,
3467 uint64_t timecode, uint64_t lace_duration,
3468 int64_t pos, int is_keyframe,
3469 uint8_t *additional, uint64_t additional_id, int additional_size,
3470 int64_t discard_padding)
3472 uint8_t *pkt_data = data;
3474 AVPacket pktl, *pkt = &pktl;
3476 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3477 res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3479 av_log(matroska->ctx, AV_LOG_ERROR,
3480 "Error parsing a wavpack block.\n");
3488 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3489 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3490 res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3492 av_log(matroska->ctx, AV_LOG_ERROR,
3493 "Error parsing a prores block.\n");
3501 if (!pkt_size && !additional_size)
3504 av_init_packet(pkt);
3506 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3509 pkt->buf = av_buffer_ref(buf);
3512 res = AVERROR(ENOMEM);
3516 pkt->data = pkt_data;
3517 pkt->size = pkt_size;
3518 pkt->flags = is_keyframe;
3519 pkt->stream_index = st->index;
3521 if (additional_size > 0) {
3522 uint8_t *side_data = av_packet_new_side_data(pkt,
3523 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3524 additional_size + 8);
3526 av_packet_unref(pkt);
3527 return AVERROR(ENOMEM);
3529 AV_WB64(side_data, additional_id);
3530 memcpy(side_data + 8, additional, additional_size);
3533 if (discard_padding) {
3534 uint8_t *side_data = av_packet_new_side_data(pkt,
3535 AV_PKT_DATA_SKIP_SAMPLES,
3538 av_packet_unref(pkt);
3539 return AVERROR(ENOMEM);
3541 discard_padding = av_rescale_q(discard_padding,
3542 (AVRational){1, 1000000000},
3543 (AVRational){1, st->codecpar->sample_rate});
3544 if (discard_padding > 0) {
3545 AV_WL32(side_data + 4, discard_padding);
3547 AV_WL32(side_data, -discard_padding);
3551 if (track->ms_compat)
3552 pkt->dts = timecode;
3554 pkt->pts = timecode;
3556 pkt->duration = lace_duration;
3558 #if FF_API_CONVERGENCE_DURATION
3559 FF_DISABLE_DEPRECATION_WARNINGS
3560 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3561 pkt->convergence_duration = lace_duration;
3563 FF_ENABLE_DEPRECATION_WARNINGS
3566 res = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3568 av_packet_unref(pkt);
3569 return AVERROR(ENOMEM);
3581 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3582 int size, int64_t pos, uint64_t cluster_time,
3583 uint64_t block_duration, int is_keyframe,
3584 uint8_t *additional, uint64_t additional_id, int additional_size,
3585 int64_t cluster_pos, int64_t discard_padding)
3587 uint64_t timecode = AV_NOPTS_VALUE;
3588 MatroskaTrack *track;
3593 uint32_t lace_size[256];
3594 int n, flags, laces = 0;
3596 int trust_default_duration;
3598 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3600 if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3605 track = matroska_find_track_by_num(matroska, num);
3606 if (!track || size < 3)
3607 return AVERROR_INVALIDDATA;
3609 if (!(st = track->stream)) {
3610 av_log(matroska->ctx, AV_LOG_VERBOSE,
3611 "No stream associated to TrackNumber %"PRIu64". "
3612 "Ignoring Block with this TrackNumber.\n", num);
3616 if (st->discard >= AVDISCARD_ALL)
3618 if (block_duration > INT64_MAX)
3619 block_duration = INT64_MAX;
3621 block_time = sign_extend(AV_RB16(data), 16);
3625 if (is_keyframe == -1)
3626 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3628 if (cluster_time != (uint64_t) -1 &&
3629 (block_time >= 0 || cluster_time >= -block_time)) {
3630 uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
3631 timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
3632 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3633 timecode < track->end_timecode)
3634 is_keyframe = 0; /* overlapping subtitles are not key frame */
3636 ff_reduce_index(matroska->ctx, st->index);
3637 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3642 if (matroska->skip_to_keyframe &&
3643 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3644 // Compare signed timecodes. Timecode may be negative due to codec delay
3645 // offset. We don't support timestamps greater than int64_t anyway - see
3647 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3650 matroska->skip_to_keyframe = 0;
3651 else if (!st->internal->skip_to_keyframe) {
3652 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3653 matroska->skip_to_keyframe = 0;
3657 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3658 &pb, lace_size, &laces);
3660 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3664 trust_default_duration = track->default_duration != 0;
3665 if (track->audio.samplerate == 8000 && trust_default_duration) {
3666 // If this is needed for more codecs, then add them here
3667 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3668 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3669 trust_default_duration = 0;
3673 if (!block_duration && trust_default_duration)
3674 block_duration = track->default_duration * laces / matroska->time_scale;
3676 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3677 track->end_timecode =
3678 FFMAX(track->end_timecode, timecode + block_duration);
3680 for (n = 0; n < laces; n++) {
3681 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3682 uint8_t *out_data = data;
3683 int out_size = lace_size[n];
3685 if (track->needs_decoding) {
3686 res = matroska_decode_buffer(&out_data, &out_size, track);
3689 /* Given that we are here means that out_data is no longer
3690 * owned by buf, so set it to NULL. This depends upon
3691 * zero-length header removal compression being ignored. */
3692 av_assert1(out_data != data);
3696 if (track->audio.buf) {
3697 res = matroska_parse_rm_audio(matroska, track, st,
3704 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3705 res = matroska_parse_webvtt(matroska, track, st,
3707 timecode, lace_duration,
3714 res = matroska_parse_frame(matroska, track, st, buf, out_data,
3715 out_size, timecode, lace_duration,
3716 pos, !n ? is_keyframe : 0,
3717 additional, additional_id, additional_size,
3723 if (timecode != AV_NOPTS_VALUE)
3724 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3725 data += lace_size[n];
3731 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3733 MatroskaCluster *cluster = &matroska->current_cluster;
3734 MatroskaBlock *block = &cluster->block;
3737 av_assert0(matroska->num_levels <= 2);
3739 if (matroska->num_levels == 1) {
3740 res = ebml_parse(matroska, matroska_segment, NULL);
3743 /* Found a cluster: subtract the size of the ID already read. */
3744 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3746 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3752 if (matroska->num_levels == 2) {
3753 /* We are inside a cluster. */
3754 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3756 if (res >= 0 && block->bin.size > 0) {
3757 int is_keyframe = block->non_simple ? block->reference.count == 0 : -1;
3758 uint8_t* additional = block->additional.size > 0 ?
3759 block->additional.data : NULL;
3761 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3762 block->bin.size, block->bin.pos,
3763 cluster->timecode, block->duration,
3764 is_keyframe, additional, block->additional_id,
3765 block->additional.size, cluster->pos,
3766 block->discard_padding);
3769 ebml_free(matroska_blockgroup, block);
3770 memset(block, 0, sizeof(*block));
3771 } else if (!matroska->num_levels) {
3772 if (!avio_feof(matroska->ctx->pb)) {
3773 avio_r8(matroska->ctx->pb);
3774 if (!avio_feof(matroska->ctx->pb)) {
3775 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3776 "end of segment.\n");
3777 return AVERROR_INVALIDDATA;
3787 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3789 MatroskaDemuxContext *matroska = s->priv_data;
3792 if (matroska->resync_pos == -1) {
3793 // This can only happen if generic seeking has been used.
3794 matroska->resync_pos = avio_tell(s->pb);
3797 while (matroska_deliver_packet(matroska, pkt)) {
3799 return (ret < 0) ? ret : AVERROR_EOF;
3800 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3801 ret = matroska_resync(matroska, matroska->resync_pos);
3807 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3808 int64_t timestamp, int flags)
3810 MatroskaDemuxContext *matroska = s->priv_data;
3811 MatroskaTrack *tracks = NULL;
3812 AVStream *st = s->streams[stream_index];
3815 /* Parse the CUES now since we need the index data to seek. */
3816 if (matroska->cues_parsing_deferred > 0) {
3817 matroska->cues_parsing_deferred = 0;
3818 matroska_parse_cues(matroska);
3821 if (!st->internal->nb_index_entries)
3823 timestamp = FFMAX(timestamp, st->internal->index_entries[0].timestamp);
3825 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3826 matroska_reset_status(matroska, 0, st->internal->index_entries[st->internal->nb_index_entries - 1].pos);
3827 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3828 matroska_clear_queue(matroska);
3829 if (matroska_parse_cluster(matroska) < 0)
3834 matroska_clear_queue(matroska);
3835 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->internal->nb_index_entries - 1))
3838 tracks = matroska->tracks.elem;
3839 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3840 tracks[i].audio.pkt_cnt = 0;
3841 tracks[i].audio.sub_packet_cnt = 0;
3842 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3843 tracks[i].end_timecode = 0;
3846 /* We seek to a level 1 element, so set the appropriate status. */
3847 matroska_reset_status(matroska, 0, st->internal->index_entries[index].pos);
3848 if (flags & AVSEEK_FLAG_ANY) {
3849 st->internal->skip_to_keyframe = 0;
3850 matroska->skip_to_timecode = timestamp;
3852 st->internal->skip_to_keyframe = 1;
3853 matroska->skip_to_timecode = st->internal->index_entries[index].timestamp;
3855 matroska->skip_to_keyframe = 1;
3857 ff_update_cur_dts(s, st, st->internal->index_entries[index].timestamp);
3860 // slightly hackish but allows proper fallback to
3861 // the generic seeking code.
3862 matroska_reset_status(matroska, 0, -1);
3863 matroska->resync_pos = -1;
3864 matroska_clear_queue(matroska);
3865 st->internal->skip_to_keyframe =
3866 matroska->skip_to_keyframe = 0;
3871 static int matroska_read_close(AVFormatContext *s)
3873 MatroskaDemuxContext *matroska = s->priv_data;
3874 MatroskaTrack *tracks = matroska->tracks.elem;
3877 matroska_clear_queue(matroska);
3879 for (n = 0; n < matroska->tracks.nb_elem; n++)
3880 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3881 av_freep(&tracks[n].audio.buf);
3882 ebml_free(matroska_segment, matroska);
3888 int64_t start_time_ns;
3889 int64_t end_time_ns;
3890 int64_t start_offset;
3894 /* This function searches all the Cues and returns the CueDesc corresponding to
3895 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3896 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3898 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3899 MatroskaDemuxContext *matroska = s->priv_data;
3902 int nb_index_entries = s->streams[0]->internal->nb_index_entries;
3903 AVIndexEntry *index_entries = s->streams[0]->internal->index_entries;
3904 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3905 for (i = 1; i < nb_index_entries; i++) {
3906 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3907 index_entries[i].timestamp * matroska->time_scale > ts) {
3912 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3913 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3914 if (i != nb_index_entries - 1) {
3915 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3916 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3918 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3919 // FIXME: this needs special handling for files where Cues appear
3920 // before Clusters. the current logic assumes Cues appear after
3922 cue_desc.end_offset = cues_start - matroska->segment_start;
3927 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3929 MatroskaDemuxContext *matroska = s->priv_data;
3930 uint32_t id = matroska->current_id;
3931 int64_t cluster_pos, before_pos;
3933 if (s->streams[0]->internal->nb_index_entries <= 0) return 0;
3934 // seek to the first cluster using cues.
3935 index = av_index_search_timestamp(s->streams[0], 0, 0);
3936 if (index < 0) return 0;
3937 cluster_pos = s->streams[0]->internal->index_entries[index].pos;
3938 before_pos = avio_tell(s->pb);
3940 uint64_t cluster_id, cluster_length;
3943 avio_seek(s->pb, cluster_pos, SEEK_SET);
3944 // read cluster id and length
3945 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3946 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3948 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3952 matroska_reset_status(matroska, 0, cluster_pos);
3953 matroska_clear_queue(matroska);
3954 if (matroska_parse_cluster(matroska) < 0 ||
3958 pkt = &matroska->queue->pkt;
3959 // 4 + read is the length of the cluster id and the cluster length field.
3960 cluster_pos += 4 + read + cluster_length;
3961 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3967 /* Restore the status after matroska_read_header: */
3968 matroska_reset_status(matroska, id, before_pos);
3973 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3974 double min_buffer, double* buffer,
3975 double* sec_to_download, AVFormatContext *s,
3978 double nano_seconds_per_second = 1000000000.0;
3979 double time_sec = time_ns / nano_seconds_per_second;
3981 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3982 int64_t end_time_ns = time_ns + time_to_search_ns;
3983 double sec_downloaded = 0.0;
3984 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3985 if (desc_curr.start_time_ns == -1)
3987 *sec_to_download = 0.0;
3989 // Check for non cue start time.
3990 if (time_ns > desc_curr.start_time_ns) {
3991 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3992 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3993 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3994 double timeToDownload = (cueBytes * 8.0) / bps;
3996 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3997 *sec_to_download += timeToDownload;
3999 // Check if the search ends within the first cue.
4000 if (desc_curr.end_time_ns >= end_time_ns) {
4001 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4002 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4003 sec_downloaded = percent_to_sub * sec_downloaded;
4004 *sec_to_download = percent_to_sub * *sec_to_download;
4007 if ((sec_downloaded + *buffer) <= min_buffer) {
4011 // Get the next Cue.
4012 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4015 while (desc_curr.start_time_ns != -1) {
4016 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
4017 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
4018 double desc_sec = desc_ns / nano_seconds_per_second;
4019 double bits = (desc_bytes * 8.0);
4020 double time_to_download = bits / bps;
4022 sec_downloaded += desc_sec - time_to_download;
4023 *sec_to_download += time_to_download;
4025 if (desc_curr.end_time_ns >= end_time_ns) {
4026 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4027 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4028 sec_downloaded = percent_to_sub * sec_downloaded;
4029 *sec_to_download = percent_to_sub * *sec_to_download;
4031 if ((sec_downloaded + *buffer) <= min_buffer)
4036 if ((sec_downloaded + *buffer) <= min_buffer) {
4041 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4043 *buffer = *buffer + sec_downloaded;
4047 /* This function computes the bandwidth of the WebM file with the help of
4048 * buffer_size_after_time_downloaded() function. Both of these functions are
4049 * adapted from WebM Tools project and are adapted to work with FFmpeg's
4050 * Matroska parsing mechanism.
4052 * Returns the bandwidth of the file on success; -1 on error.
4054 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4056 MatroskaDemuxContext *matroska = s->priv_data;
4057 AVStream *st = s->streams[0];
4058 double bandwidth = 0.0;
4061 for (i = 0; i < st->internal->nb_index_entries; i++) {
4062 int64_t prebuffer_ns = 1000000000;
4063 int64_t time_ns = st->internal->index_entries[i].timestamp * matroska->time_scale;
4064 double nano_seconds_per_second = 1000000000.0;
4065 int64_t prebuffered_ns = time_ns + prebuffer_ns;
4066 double prebuffer_bytes = 0.0;
4067 int64_t temp_prebuffer_ns = prebuffer_ns;
4068 int64_t pre_bytes, pre_ns;
4069 double pre_sec, prebuffer, bits_per_second;
4070 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4072 // Start with the first Cue.
4073 CueDesc desc_end = desc_beg;
4075 // Figure out how much data we have downloaded for the prebuffer. This will
4076 // be used later to adjust the bits per sample to try.
4077 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4078 // Prebuffered the entire Cue.
4079 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4080 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4081 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4083 if (desc_end.start_time_ns == -1) {
4084 // The prebuffer is larger than the duration.
4085 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4087 bits_per_second = 0.0;
4089 // The prebuffer ends in the last Cue. Estimate how much data was
4091 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4092 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4093 pre_sec = pre_ns / nano_seconds_per_second;
4095 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4097 prebuffer = prebuffer_ns / nano_seconds_per_second;
4099 // Set this to 0.0 in case our prebuffer buffers the entire video.
4100 bits_per_second = 0.0;
4102 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4103 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4104 double desc_sec = desc_ns / nano_seconds_per_second;
4105 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4107 // Drop the bps by the percentage of bytes buffered.
4108 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4109 double mod_bits_per_second = calc_bits_per_second * percent;
4111 if (prebuffer < desc_sec) {
4113 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4115 // Add 1 so the bits per second should be a little bit greater than file
4117 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4118 const double min_buffer = 0.0;
4119 double buffer = prebuffer;
4120 double sec_to_download = 0.0;
4122 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4123 min_buffer, &buffer, &sec_to_download,
4127 } else if (rv == 0) {
4128 bits_per_second = (double)(bps);
4133 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4134 } while (desc_end.start_time_ns != -1);
4136 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4138 return (int64_t)bandwidth;
4141 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4143 MatroskaDemuxContext *matroska = s->priv_data;
4144 EbmlList *seekhead_list = &matroska->seekhead;
4145 MatroskaSeekhead *seekhead = seekhead_list->elem;
4147 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4151 // determine cues start and end positions
4152 for (i = 0; i < seekhead_list->nb_elem; i++)
4153 if (seekhead[i].id == MATROSKA_ID_CUES)
4156 if (i >= seekhead_list->nb_elem) return -1;
4158 before_pos = avio_tell(matroska->ctx->pb);
4159 cues_start = seekhead[i].pos + matroska->segment_start;
4160 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4161 // cues_end is computed as cues_start + cues_length + length of the
4162 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4163 // cues_end is inclusive and the above sum is reduced by 1.
4164 uint64_t cues_length, cues_id;
4166 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4167 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4168 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4169 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4172 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4174 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4175 if (cues_start == -1 || cues_end == -1) return -1;
4178 matroska_parse_cues(matroska);
4181 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4184 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4186 // if the file has cues at the start, fix up the init range so that
4187 // it does not include it
4188 if (cues_start <= init_range)
4189 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4192 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4193 if (bandwidth < 0) return -1;
4194 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4196 // check if all clusters start with key frames
4197 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4199 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4200 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4201 buf = av_malloc_array(s->streams[0]->internal->nb_index_entries, 20);
4202 if (!buf) return -1;
4204 for (i = 0; i < s->streams[0]->internal->nb_index_entries; i++) {
4205 int ret = snprintf(buf + end, 20,
4206 "%" PRId64"%s", s->streams[0]->internal->index_entries[i].timestamp,
4207 i != s->streams[0]->internal->nb_index_entries - 1 ? "," : "");
4208 if (ret <= 0 || (ret == 20 && i == s->streams[0]->internal->nb_index_entries - 1)) {
4209 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4211 return AVERROR_INVALIDDATA;
4215 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4216 buf, AV_DICT_DONT_STRDUP_VAL);
4221 static int webm_dash_manifest_read_header(AVFormatContext *s)
4224 int ret = matroska_read_header(s);
4226 MatroskaTrack *tracks;
4227 MatroskaDemuxContext *matroska = s->priv_data;
4229 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4232 if (!matroska->tracks.nb_elem || !s->nb_streams) {
4233 av_log(s, AV_LOG_ERROR, "No track found\n");
4234 ret = AVERROR_INVALIDDATA;
4238 if (!matroska->is_live) {
4239 buf = av_asprintf("%g", matroska->duration);
4241 ret = AVERROR(ENOMEM);
4244 av_dict_set(&s->streams[0]->metadata, DURATION,
4245 buf, AV_DICT_DONT_STRDUP_VAL);
4247 // initialization range
4248 // 5 is the offset of Cluster ID.
4249 init_range = avio_tell(s->pb) - 5;
4250 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4253 // basename of the file
4254 buf = strrchr(s->url, '/');
4255 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4258 tracks = matroska->tracks.elem;
4259 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4261 // parse the cues and populate Cue related fields
4262 if (!matroska->is_live) {
4263 ret = webm_dash_manifest_cues(s, init_range);
4265 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4270 // use the bandwidth from the command line if it was provided
4271 if (matroska->bandwidth > 0) {
4272 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4273 matroska->bandwidth, 0);
4277 matroska_read_close(s);
4281 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4286 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4287 static const AVOption options[] = {
4288 { "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 },
4289 { "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 },
4293 static const AVClass webm_dash_class = {
4294 .class_name = "WebM DASH Manifest demuxer",
4295 .item_name = av_default_item_name,
4297 .version = LIBAVUTIL_VERSION_INT,
4300 AVInputFormat ff_matroska_demuxer = {
4301 .name = "matroska,webm",
4302 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4303 .extensions = "mkv,mk3d,mka,mks",
4304 .priv_data_size = sizeof(MatroskaDemuxContext),
4305 .read_probe = matroska_probe,
4306 .read_header = matroska_read_header,
4307 .read_packet = matroska_read_packet,
4308 .read_close = matroska_read_close,
4309 .read_seek = matroska_read_seek,
4310 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4313 AVInputFormat ff_webm_dash_manifest_demuxer = {
4314 .name = "webm_dash_manifest",
4315 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4316 .priv_data_size = sizeof(MatroskaDemuxContext),
4317 .read_header = webm_dash_manifest_read_header,
4318 .read_packet = webm_dash_manifest_read_packet,
4319 .read_close = matroska_read_close,
4320 .priv_class = &webm_dash_class,