2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The FFmpeg Project
5 * This file is part of FFmpeg.
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 * Matroska file demuxer
25 * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28 * @see specs available on the Matroska project page: http://www.matroska.org/
36 #include "libavutil/avstring.h"
37 #include "libavutil/base64.h"
38 #include "libavutil/dict.h"
39 #include "libavutil/intfloat.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/lzo.h"
42 #include "libavutil/mastering_display_metadata.h"
43 #include "libavutil/mathematics.h"
44 #include "libavutil/opt.h"
45 #include "libavutil/time_internal.h"
46 #include "libavutil/spherical.h"
48 #include "libavcodec/bytestream.h"
49 #include "libavcodec/flac.h"
50 #include "libavcodec/mpeg4audio.h"
51 #include "libavcodec/packet_internal.h"
54 #include "avio_internal.h"
59 /* For ff_codec_get_id(). */
70 #include "qtpalette.h"
72 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
73 #define NEEDS_CHECKING 2 /* Indicates that some error checks
74 * still need to be performed */
75 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
76 * syntax level used for parsing ended. */
77 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
78 * of unkown, potentially damaged data is encountered,
79 * it is considered an error. */
80 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
81 * to this many bytes of unknown data for the
82 * SKIP_THRESHOLD check. */
98 typedef struct CountedElement {
108 typedef const struct EbmlSyntax {
112 size_t list_elem_size;
119 const struct EbmlSyntax *n;
123 typedef struct EbmlList {
125 unsigned int alloc_elem_size;
129 typedef struct EbmlBin {
136 typedef struct Ebml {
141 uint64_t doctype_version;
144 typedef struct MatroskaTrackCompression {
147 } MatroskaTrackCompression;
149 typedef struct MatroskaTrackEncryption {
152 } MatroskaTrackEncryption;
154 typedef struct MatroskaTrackEncoding {
157 MatroskaTrackCompression compression;
158 MatroskaTrackEncryption encryption;
159 } MatroskaTrackEncoding;
161 typedef struct MatroskaMasteringMeta {
170 double max_luminance;
171 CountedElement min_luminance;
172 } MatroskaMasteringMeta;
174 typedef struct MatroskaTrackVideoColor {
175 uint64_t matrix_coefficients;
176 uint64_t bits_per_channel;
177 uint64_t chroma_sub_horz;
178 uint64_t chroma_sub_vert;
179 uint64_t cb_sub_horz;
180 uint64_t cb_sub_vert;
181 uint64_t chroma_siting_horz;
182 uint64_t chroma_siting_vert;
184 uint64_t transfer_characteristics;
188 MatroskaMasteringMeta mastering_meta;
189 } MatroskaTrackVideoColor;
191 typedef struct MatroskaTrackVideoProjection {
197 } MatroskaTrackVideoProjection;
199 typedef struct MatroskaTrackVideo {
201 uint64_t display_width;
202 uint64_t display_height;
203 uint64_t pixel_width;
204 uint64_t pixel_height;
206 uint64_t display_unit;
208 uint64_t field_order;
209 uint64_t stereo_mode;
212 MatroskaTrackVideoProjection projection;
213 } MatroskaTrackVideo;
215 typedef struct MatroskaTrackAudio {
217 double out_samplerate;
221 /* real audio header (extracted from extradata) */
228 uint64_t buf_timecode;
230 } MatroskaTrackAudio;
232 typedef struct MatroskaTrackPlane {
235 } MatroskaTrackPlane;
237 typedef struct MatroskaTrackOperation {
238 EbmlList combine_planes;
239 } MatroskaTrackOperation;
241 typedef struct MatroskaTrack {
250 uint64_t default_duration;
251 uint64_t flag_default;
252 uint64_t flag_forced;
253 uint64_t flag_comment;
254 uint64_t flag_hearingimpaired;
255 uint64_t flag_visualimpaired;
256 uint64_t flag_textdescriptions;
257 CountedElement flag_original;
258 uint64_t seek_preroll;
259 MatroskaTrackVideo video;
260 MatroskaTrackAudio audio;
261 MatroskaTrackOperation operation;
263 uint64_t codec_delay;
264 uint64_t codec_delay_in_track_tb;
267 int64_t end_timecode;
270 uint64_t max_block_additional_id;
272 uint32_t palette[AVPALETTE_COUNT];
276 typedef struct MatroskaAttachment {
284 } MatroskaAttachment;
286 typedef struct MatroskaChapter {
295 typedef struct MatroskaIndexPos {
300 typedef struct MatroskaIndex {
305 typedef struct MatroskaTag {
313 typedef struct MatroskaTagTarget {
321 typedef struct MatroskaTags {
322 MatroskaTagTarget target;
326 typedef struct MatroskaSeekhead {
331 typedef struct MatroskaLevel {
336 typedef struct MatroskaBlock {
338 CountedElement reference;
341 uint64_t additional_id;
343 int64_t discard_padding;
346 typedef struct MatroskaCluster {
352 typedef struct MatroskaLevel1Element {
356 } MatroskaLevel1Element;
358 typedef struct MatroskaDemuxContext {
359 const AVClass *class;
360 AVFormatContext *ctx;
363 MatroskaLevel levels[EBML_MAX_DEPTH];
375 EbmlList attachments;
381 /* byte position of the segment inside the stream */
382 int64_t segment_start;
384 /* This packet coincides with AVFormatInternal.parse_pkt
385 * and is not owned by us. */
388 /* the packet queue */
390 PacketList *queue_end;
394 /* What to skip before effectively reading a packet. */
395 int skip_to_keyframe;
396 uint64_t skip_to_timecode;
398 /* File has a CUES element, but we defer parsing until it is needed. */
399 int cues_parsing_deferred;
401 /* Level1 elements and whether they were read yet */
402 MatroskaLevel1Element level1_elems[64];
403 int num_level1_elems;
405 MatroskaCluster current_cluster;
407 /* WebM DASH Manifest live flag */
410 /* Bandwidth value for WebM DASH Manifest */
412 } MatroskaDemuxContext;
414 #define CHILD_OF(parent) { .def = { .n = parent } }
416 // The following forward declarations need their size because
417 // a tentative definition with internal linkage must not be an
418 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
419 // Removing the sizes breaks MSVC.
420 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
421 matroska_track[32], matroska_track_encoding[6], matroska_track_encodings[2],
422 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_tracks[2],
423 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
424 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
425 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
427 static EbmlSyntax ebml_header[] = {
428 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
429 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, 0, offsetof(Ebml, max_size), { .u = 8 } },
430 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, 0, offsetof(Ebml, id_length), { .u = 4 } },
431 { EBML_ID_DOCTYPE, EBML_STR, 0, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
432 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
433 { EBML_ID_EBMLVERSION, EBML_NONE },
434 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
435 CHILD_OF(ebml_syntax)
438 static EbmlSyntax ebml_syntax[] = {
439 { EBML_ID_HEADER, EBML_NEST, 0, 0, 0, { .n = ebml_header } },
440 { MATROSKA_ID_SEGMENT, EBML_STOP },
444 static EbmlSyntax matroska_info[] = {
445 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
446 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, 0, offsetof(MatroskaDemuxContext, duration) },
447 { MATROSKA_ID_TITLE, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, title) },
448 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
449 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, muxingapp) },
450 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, 0, offsetof(MatroskaDemuxContext, date_utc) },
451 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
452 CHILD_OF(matroska_segment)
455 static EbmlSyntax matroska_mastering_meta[] = {
456 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_x) },
457 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_y) },
458 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_x) },
459 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_y) },
460 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_x) },
461 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_y) },
462 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_x) },
463 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_y) },
464 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 1, 0, offsetof(MatroskaMasteringMeta, min_luminance) },
465 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, max_luminance) },
466 CHILD_OF(matroska_track_video_color)
469 static EbmlSyntax matroska_track_video_color[] = {
470 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
471 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u = 0 } },
472 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz) },
473 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert) },
474 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz) },
475 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert) },
476 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
477 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
478 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
479 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
480 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
481 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_cll) },
482 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_fall) },
483 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
484 CHILD_OF(matroska_track_video)
487 static EbmlSyntax matroska_track_video_projection[] = {
488 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
489 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideoProjection, private) },
490 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f = 0.0 } },
491 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f = 0.0 } },
492 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f = 0.0 } },
493 CHILD_OF(matroska_track_video)
496 static EbmlSyntax matroska_track_video[] = {
497 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideo, frame_rate) },
498 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
499 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
500 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_width) },
501 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_height) },
502 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideo, color_space) },
503 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, alpha_mode), { .u = 0 } },
504 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, 0, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
505 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
506 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
507 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
508 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
509 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
510 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
511 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
512 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
513 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
514 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
515 CHILD_OF(matroska_track)
518 static EbmlSyntax matroska_track_audio[] = {
519 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
520 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
521 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, bitdepth) },
522 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
523 CHILD_OF(matroska_track)
526 static EbmlSyntax matroska_track_encoding_compression[] = {
527 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackCompression, algo), { .u = MATROSKA_TRACK_ENCODING_COMP_ZLIB } },
528 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, 0, offsetof(MatroskaTrackCompression, settings) },
529 CHILD_OF(matroska_track_encoding)
532 static EbmlSyntax matroska_track_encoding_encryption[] = {
533 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
534 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, 0, offsetof(MatroskaTrackEncryption,key_id) },
535 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
536 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
537 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
538 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
539 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
540 CHILD_OF(matroska_track_encoding)
542 static EbmlSyntax matroska_track_encoding[] = {
543 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
544 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
545 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
546 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
547 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
548 CHILD_OF(matroska_track_encodings)
551 static EbmlSyntax matroska_track_encodings[] = {
552 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, 0, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
553 CHILD_OF(matroska_track)
556 static EbmlSyntax matroska_track_plane[] = {
557 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,uid) },
558 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,type) },
559 CHILD_OF(matroska_track_combine_planes)
562 static EbmlSyntax matroska_track_combine_planes[] = {
563 { MATROSKA_ID_TRACKPLANE, EBML_NEST, 0, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
564 CHILD_OF(matroska_track_operation)
567 static EbmlSyntax matroska_track_operation[] = {
568 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, 0, {.n = matroska_track_combine_planes} },
569 CHILD_OF(matroska_track)
572 static EbmlSyntax matroska_track[] = {
573 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, 0, offsetof(MatroskaTrack, num) },
574 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTrack, name) },
575 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, uid) },
576 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrack, type) },
577 { MATROSKA_ID_CODECID, EBML_STR, 0, 0, offsetof(MatroskaTrack, codec_id) },
578 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrack, codec_priv) },
579 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, codec_delay), { .u = 0 } },
580 { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
581 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, 0, offsetof(MatroskaTrack, default_duration) },
582 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
583 { MATROSKA_ID_TRACKFLAGCOMMENTARY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_comment), { .u = 0 } },
584 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
585 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
586 { MATROSKA_ID_TRACKFLAGHEARINGIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_hearingimpaired), { .u = 0 } },
587 { MATROSKA_ID_TRACKFLAGVISUALIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_visualimpaired), { .u = 0 } },
588 { MATROSKA_ID_TRACKFLAGTEXTDESCRIPTIONS, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_textdescriptions), { .u = 0 } },
589 { MATROSKA_ID_TRACKFLAGORIGINAL, EBML_UINT, 1, 0, offsetof(MatroskaTrack, flag_original), {.u = 0 } },
590 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
591 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
592 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
593 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, 0, { .n = matroska_track_encodings } },
594 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, max_block_additional_id), { .u = 0 } },
595 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, 0, offsetof(MatroskaTrack, seek_preroll), { .u = 0 } },
596 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
597 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
598 { MATROSKA_ID_CODECNAME, EBML_NONE },
599 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
600 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
601 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
602 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
603 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
604 CHILD_OF(matroska_tracks)
607 static EbmlSyntax matroska_tracks[] = {
608 { MATROSKA_ID_TRACKENTRY, EBML_NEST, 0, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
609 CHILD_OF(matroska_segment)
612 static EbmlSyntax matroska_attachment[] = {
613 { MATROSKA_ID_FILEUID, EBML_UINT, 0, 0, offsetof(MatroskaAttachment, uid) },
614 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, filename) },
615 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, 0, offsetof(MatroskaAttachment, mime) },
616 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, 0, offsetof(MatroskaAttachment, bin) },
617 { MATROSKA_ID_FILEDESC, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, description) },
618 CHILD_OF(matroska_attachments)
621 static EbmlSyntax matroska_attachments[] = {
622 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, 0, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
623 CHILD_OF(matroska_segment)
626 static EbmlSyntax matroska_chapter_display[] = {
627 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaChapter, title) },
628 { MATROSKA_ID_CHAPLANG, EBML_NONE },
629 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
630 CHILD_OF(matroska_chapter_entry)
633 static EbmlSyntax matroska_chapter_entry[] = {
634 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
635 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
636 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaChapter, uid) },
637 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter_display } },
638 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
639 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
640 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
641 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
642 CHILD_OF(matroska_chapter)
645 static EbmlSyntax matroska_chapter[] = {
646 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, 0, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
647 { MATROSKA_ID_EDITIONUID, EBML_NONE },
648 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
649 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
650 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
651 CHILD_OF(matroska_chapters)
654 static EbmlSyntax matroska_chapters[] = {
655 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter } },
656 CHILD_OF(matroska_segment)
659 static EbmlSyntax matroska_index_pos[] = {
660 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, track) },
661 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, pos) },
662 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
663 { MATROSKA_ID_CUEDURATION, EBML_NONE },
664 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
665 CHILD_OF(matroska_index_entry)
668 static EbmlSyntax matroska_index_entry[] = {
669 { MATROSKA_ID_CUETIME, EBML_UINT, 0, 0, offsetof(MatroskaIndex, time) },
670 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, 0, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
671 CHILD_OF(matroska_index)
674 static EbmlSyntax matroska_index[] = {
675 { MATROSKA_ID_POINTENTRY, EBML_NEST, 0, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
676 CHILD_OF(matroska_segment)
679 static EbmlSyntax matroska_simpletag[] = {
680 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTag, name) },
681 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaTag, string) },
682 { MATROSKA_ID_TAGLANG, EBML_STR, 0, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
683 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
684 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
685 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
686 CHILD_OF(matroska_tag)
689 static EbmlSyntax matroska_tagtargets[] = {
690 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, 0, offsetof(MatroskaTagTarget, type) },
691 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
692 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, trackuid), { .u = 0 } },
693 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, chapteruid), { .u = 0 } },
694 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, attachuid), { .u = 0 } },
695 CHILD_OF(matroska_tag)
698 static EbmlSyntax matroska_tag[] = {
699 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
700 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
701 CHILD_OF(matroska_tags)
704 static EbmlSyntax matroska_tags[] = {
705 { MATROSKA_ID_TAG, EBML_NEST, 0, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
706 CHILD_OF(matroska_segment)
709 static EbmlSyntax matroska_seekhead_entry[] = {
710 { MATROSKA_ID_SEEKID, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, id) },
711 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
712 CHILD_OF(matroska_seekhead)
715 static EbmlSyntax matroska_seekhead[] = {
716 { MATROSKA_ID_SEEKENTRY, EBML_NEST, 0, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
717 CHILD_OF(matroska_segment)
720 static EbmlSyntax matroska_segment[] = {
721 { MATROSKA_ID_CLUSTER, EBML_STOP },
722 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, 0, { .n = matroska_info } },
723 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tracks } },
724 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_attachments } },
725 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_chapters } },
726 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, 0, { .n = matroska_index } },
727 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tags } },
728 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, 0, { .n = matroska_seekhead } },
729 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
732 static EbmlSyntax matroska_segments[] = {
733 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, 0, { .n = matroska_segment } },
737 static EbmlSyntax matroska_blockmore[] = {
738 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
739 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, 0, offsetof(MatroskaBlock,additional) },
740 CHILD_OF(matroska_blockadditions)
743 static EbmlSyntax matroska_blockadditions[] = {
744 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, 0, {.n = matroska_blockmore} },
745 CHILD_OF(matroska_blockgroup)
748 static EbmlSyntax matroska_blockgroup[] = {
749 { MATROSKA_ID_BLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
750 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, 0, { .n = matroska_blockadditions} },
751 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, 0, offsetof(MatroskaBlock, duration) },
752 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, 0, offsetof(MatroskaBlock, discard_padding) },
753 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 1, 0, offsetof(MatroskaBlock, reference) },
754 { MATROSKA_ID_CODECSTATE, EBML_NONE },
755 { 1, EBML_UINT, 0, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
756 CHILD_OF(matroska_cluster_parsing)
759 // The following array contains SimpleBlock and BlockGroup twice
760 // in order to reuse the other values for matroska_cluster_enter.
761 static EbmlSyntax matroska_cluster_parsing[] = {
762 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
763 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, 0, { .n = matroska_blockgroup } },
764 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, 0, offsetof(MatroskaCluster, timecode) },
765 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
766 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
767 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
768 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
769 CHILD_OF(matroska_segment)
772 static EbmlSyntax matroska_cluster_enter[] = {
773 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, 0, { .n = &matroska_cluster_parsing[2] } },
778 static const CodecMime mkv_image_mime_tags[] = {
779 {"image/gif" , AV_CODEC_ID_GIF},
780 {"image/jpeg" , AV_CODEC_ID_MJPEG},
781 {"image/png" , AV_CODEC_ID_PNG},
782 {"image/tiff" , AV_CODEC_ID_TIFF},
784 {"" , AV_CODEC_ID_NONE}
787 static const CodecMime mkv_mime_tags[] = {
788 {"text/plain" , AV_CODEC_ID_TEXT},
789 {"application/x-truetype-font", AV_CODEC_ID_TTF},
790 {"application/x-font" , AV_CODEC_ID_TTF},
791 {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
792 {"binary" , AV_CODEC_ID_BIN_DATA},
794 {"" , AV_CODEC_ID_NONE}
797 static const char *const matroska_doctypes[] = { "matroska", "webm" };
799 static int matroska_read_close(AVFormatContext *s);
802 * This function prepares the status for parsing of level 1 elements.
804 static int matroska_reset_status(MatroskaDemuxContext *matroska,
805 uint32_t id, int64_t position)
808 int64_t err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
813 matroska->current_id = id;
814 matroska->num_levels = 1;
815 matroska->unknown_count = 0;
816 matroska->resync_pos = avio_tell(matroska->ctx->pb);
818 matroska->resync_pos -= (av_log2(id) + 7) / 8;
823 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
825 AVIOContext *pb = matroska->ctx->pb;
828 /* Try to seek to the last position to resync from. If this doesn't work,
829 * we resync from the earliest position available: The start of the buffer. */
830 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
831 av_log(matroska->ctx, AV_LOG_WARNING,
832 "Seek to desired resync point failed. Seeking to "
833 "earliest point available instead.\n");
834 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
835 last_pos + 1), SEEK_SET);
840 // try to find a toplevel element
841 while (!avio_feof(pb)) {
842 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
843 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
844 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
845 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
846 /* Prepare the context for parsing of a level 1 element. */
847 matroska_reset_status(matroska, id, -1);
848 /* Given that we are here means that an error has occurred,
849 * so treat the segment as unknown length in order not to
850 * discard valid data that happens to be beyond the designated
851 * end of the segment. */
852 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
855 id = (id << 8) | avio_r8(pb);
859 return pb->error ? pb->error : AVERROR_EOF;
863 * Read: an "EBML number", which is defined as a variable-length
864 * array of bytes. The first byte indicates the length by giving a
865 * number of 0-bits followed by a one. The position of the first
866 * "one" bit inside the first byte indicates the length of this
868 * Returns: number of bytes read, < 0 on error
870 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
871 int max_size, uint64_t *number, int eof_forbidden)
877 /* The first byte tells us the length in bytes - except when it is zero. */
882 /* get the length of the EBML number */
883 read = 8 - ff_log2_tab[total];
885 if (!total || read > max_size) {
886 pos = avio_tell(pb) - 1;
888 av_log(matroska->ctx, AV_LOG_ERROR,
889 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
890 "of an EBML number\n", pos, pos);
892 av_log(matroska->ctx, AV_LOG_ERROR,
893 "Length %d indicated by an EBML number's first byte 0x%02x "
894 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
895 read, (uint8_t) total, pos, pos, max_size);
897 return AVERROR_INVALIDDATA;
900 /* read out length */
901 total ^= 1 << ff_log2_tab[total];
903 total = (total << 8) | avio_r8(pb);
905 if (pb->eof_reached) {
917 av_log(matroska->ctx, AV_LOG_ERROR,
918 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
923 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
924 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
931 * Read a EBML length value.
932 * This needs special handling for the "unknown length" case which has multiple
935 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
938 int res = ebml_read_num(matroska, pb, 8, number, 1);
939 if (res > 0 && *number + 1 == 1ULL << (7 * res))
940 *number = EBML_UNKNOWN_LENGTH;
945 * Read the next element as an unsigned int.
946 * Returns NEEDS_CHECKING unless size == 0.
948 static int ebml_read_uint(AVIOContext *pb, int size,
949 uint64_t default_value, uint64_t *num)
954 *num = default_value;
957 /* big-endian ordering; build up number */
960 *num = (*num << 8) | avio_r8(pb);
962 return NEEDS_CHECKING;
966 * Read the next element as a signed int.
967 * Returns NEEDS_CHECKING unless size == 0.
969 static int ebml_read_sint(AVIOContext *pb, int size,
970 int64_t default_value, int64_t *num)
975 *num = default_value;
978 *num = sign_extend(avio_r8(pb), 8);
980 /* big-endian ordering; build up number */
982 *num = ((uint64_t)*num << 8) | avio_r8(pb);
985 return NEEDS_CHECKING;
989 * Read the next element as a float.
990 * Returns 0 if size == 0, NEEDS_CHECKING or < 0 on obvious failure.
992 static int ebml_read_float(AVIOContext *pb, int size,
993 double default_value, double *num)
996 *num = default_value;
998 } else if (size == 4) {
999 *num = av_int2float(avio_rb32(pb));
1000 } else if (size == 8) {
1001 *num = av_int2double(avio_rb64(pb));
1003 return AVERROR_INVALIDDATA;
1005 return NEEDS_CHECKING;
1009 * Read the next element as an ASCII string.
1010 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1012 static int ebml_read_ascii(AVIOContext *pb, int size,
1013 const char *default_value, char **str)
1018 if (size == 0 && default_value) {
1019 res = av_strdup(default_value);
1021 return AVERROR(ENOMEM);
1023 /* EBML strings are usually not 0-terminated, so we allocate one
1024 * byte more, read the string and NUL-terminate it ourselves. */
1025 if (!(res = av_malloc(size + 1)))
1026 return AVERROR(ENOMEM);
1027 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
1029 return ret < 0 ? ret : NEEDS_CHECKING;
1040 * Read the next element as binary data.
1041 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1043 static int ebml_read_binary(AVIOContext *pb, int length,
1044 int64_t pos, EbmlBin *bin)
1048 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1051 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1053 bin->data = bin->buf->data;
1056 if ((ret = avio_read(pb, bin->data, length)) != length) {
1057 av_buffer_unref(&bin->buf);
1060 return ret < 0 ? ret : NEEDS_CHECKING;
1067 * Read the next element, but only the header. The contents
1068 * are supposed to be sub-elements which can be read separately.
1069 * 0 is success, < 0 is failure.
1071 static int ebml_read_master(MatroskaDemuxContext *matroska,
1072 uint64_t length, int64_t pos)
1074 MatroskaLevel *level;
1076 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1077 av_log(matroska->ctx, AV_LOG_ERROR,
1078 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1079 return AVERROR(ENOSYS);
1082 level = &matroska->levels[matroska->num_levels++];
1084 level->length = length;
1090 * Read a signed "EBML number"
1091 * Return: number of bytes processed, < 0 on error
1093 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1094 AVIOContext *pb, int64_t *num)
1099 /* read as unsigned number first */
1100 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1103 /* make signed (weird way) */
1104 *num = unum - ((1LL << (7 * res - 1)) - 1);
1109 static int ebml_parse(MatroskaDemuxContext *matroska,
1110 EbmlSyntax *syntax, void *data);
1112 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1116 // Whoever touches this should be aware of the duplication
1117 // existing in matroska_cluster_parsing.
1118 for (i = 0; syntax[i].id; i++)
1119 if (id == syntax[i].id)
1125 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1131 for (int i = 0; syntax[i].id; i++) {
1132 void *dst = (char *)data + syntax[i].data_offset;
1133 switch (syntax[i].type) {
1135 *(uint64_t *)dst = syntax[i].def.u;
1138 *(int64_t *) dst = syntax[i].def.i;
1141 *(double *) dst = syntax[i].def.f;
1145 // the default may be NULL
1146 if (syntax[i].def.s) {
1147 *(char**)dst = av_strdup(syntax[i].def.s);
1149 return AVERROR(ENOMEM);
1155 if (!matroska->levels[matroska->num_levels - 1].length) {
1156 matroska->num_levels--;
1162 res = ebml_parse(matroska, syntax, data);
1165 return res == LEVEL_ENDED ? 0 : res;
1168 static int is_ebml_id_valid(uint32_t id)
1170 // Due to endian nonsense in Matroska, the highest byte with any bits set
1171 // will contain the leading length bit. This bit in turn identifies the
1172 // total byte length of the element by its position within the byte.
1173 unsigned int bits = av_log2(id);
1174 return id && (bits + 7) / 8 == (8 - bits % 8);
1178 * Allocate and return the entry for the level1 element with the given ID. If
1179 * an entry already exists, return the existing entry.
1181 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1182 uint32_t id, int64_t pos)
1185 MatroskaLevel1Element *elem;
1187 if (!is_ebml_id_valid(id))
1190 // Some files link to all clusters; useless.
1191 if (id == MATROSKA_ID_CLUSTER)
1194 // There can be multiple SeekHeads and Tags.
1195 for (i = 0; i < matroska->num_level1_elems; i++) {
1196 if (matroska->level1_elems[i].id == id) {
1197 if (matroska->level1_elems[i].pos == pos ||
1198 id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1199 return &matroska->level1_elems[i];
1203 // Only a completely broken file would have more elements.
1204 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1205 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1209 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1210 *elem = (MatroskaLevel1Element){.id = id};
1215 static int ebml_parse(MatroskaDemuxContext *matroska,
1216 EbmlSyntax *syntax, void *data)
1218 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1219 // Forbid unknown-length EBML_NONE elements.
1220 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1224 // max. 16 MB for strings
1225 [EBML_STR] = 0x1000000,
1226 [EBML_UTF8] = 0x1000000,
1227 // max. 256 MB for binary data
1228 [EBML_BIN] = 0x10000000,
1229 // no limits for anything else
1231 AVIOContext *pb = matroska->ctx->pb;
1234 int64_t pos = avio_tell(pb), pos_alt;
1235 int res, update_pos = 1, level_check;
1236 MatroskaLevel1Element *level1_elem;
1237 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1239 if (!matroska->current_id) {
1241 res = ebml_read_num(matroska, pb, 4, &id, 0);
1243 if (pb->eof_reached && res == AVERROR_EOF) {
1244 if (matroska->is_live)
1245 // in live mode, finish parsing if EOF is reached.
1247 if (level && pos == avio_tell(pb)) {
1248 if (level->length == EBML_UNKNOWN_LENGTH) {
1249 // Unknown-length levels automatically end at EOF.
1250 matroska->num_levels--;
1253 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1254 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1260 matroska->current_id = id | 1 << 7 * res;
1261 pos_alt = pos + res;
1264 pos -= (av_log2(matroska->current_id) + 7) / 8;
1267 id = matroska->current_id;
1269 syntax = ebml_parse_id(syntax, id);
1270 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1271 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1272 // Unknown-length levels end when an element from an upper level
1273 // in the hierarchy is encountered.
1274 while (syntax->def.n) {
1275 syntax = ebml_parse_id(syntax->def.n, id);
1277 matroska->num_levels--;
1283 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1284 "%"PRId64"\n", id, pos);
1285 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1289 data = (char *) data + syntax->data_offset;
1290 if (syntax->list_elem_size) {
1291 EbmlList *list = data;
1294 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1295 return AVERROR(ENOMEM);
1296 newelem = av_fast_realloc(list->elem,
1297 &list->alloc_elem_size,
1298 (list->nb_elem + 1) * syntax->list_elem_size);
1300 return AVERROR(ENOMEM);
1301 list->elem = newelem;
1302 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1303 memset(data, 0, syntax->list_elem_size);
1308 if (syntax->type != EBML_STOP) {
1309 matroska->current_id = 0;
1310 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1315 if (matroska->num_levels > 0) {
1316 if (length != EBML_UNKNOWN_LENGTH &&
1317 level->length != EBML_UNKNOWN_LENGTH) {
1318 uint64_t elem_end = pos_alt + length,
1319 level_end = level->start + level->length;
1321 if (elem_end < level_end) {
1323 } else if (elem_end == level_end) {
1324 level_check = LEVEL_ENDED;
1326 av_log(matroska->ctx, AV_LOG_ERROR,
1327 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1328 "containing master element ending at 0x%"PRIx64"\n",
1329 pos, elem_end, level_end);
1330 return AVERROR_INVALIDDATA;
1332 } else if (length != EBML_UNKNOWN_LENGTH) {
1334 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1335 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1336 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1337 return AVERROR_INVALIDDATA;
1340 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1341 || syntax->type == EBML_NEST)) {
1342 // According to the current specifications only clusters and
1343 // segments are allowed to be unknown-length. We also accept
1344 // other unknown-length master elements.
1345 av_log(matroska->ctx, AV_LOG_WARNING,
1346 "Found unknown-length element 0x%"PRIX32" other than "
1347 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1348 "parsing will nevertheless be attempted.\n", id, pos);
1355 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1356 if (length != EBML_UNKNOWN_LENGTH) {
1357 av_log(matroska->ctx, AV_LOG_ERROR,
1358 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1359 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1360 length, max_lengths[syntax->type], id, pos);
1361 } else if (syntax->type != EBML_NONE) {
1362 av_log(matroska->ctx, AV_LOG_ERROR,
1363 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1364 "unknown length, yet the length of an element of its "
1365 "type must be known.\n", id, pos);
1367 av_log(matroska->ctx, AV_LOG_ERROR,
1368 "Found unknown-length element with ID 0x%"PRIX32" at "
1369 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1370 "available.\n", id, pos);
1372 return AVERROR_INVALIDDATA;
1375 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1376 // Loosing sync will likely manifest itself as encountering unknown
1377 // elements which are not reliably distinguishable from elements
1378 // belonging to future extensions of the format.
1379 // We use a heuristic to detect such situations: If the current
1380 // element is not expected at the current syntax level and there
1381 // were only a few unknown elements in a row, then the element is
1382 // skipped or considered defective based upon the length of the
1383 // current element (i.e. how much would be skipped); if there were
1384 // more than a few skipped elements in a row and skipping the current
1385 // element would lead us more than SKIP_THRESHOLD away from the last
1386 // known good position, then it is inferred that an error occurred.
1387 // The dependency on the number of unknown elements in a row exists
1388 // because the distance to the last known good position is
1389 // automatically big if the last parsed element was big.
1390 // In both cases, each unknown element is considered equivalent to
1391 // UNKNOWN_EQUIV of skipped bytes for the check.
1392 // The whole check is only done for non-seekable output, because
1393 // in this situation skipped data can't simply be rechecked later.
1394 // This is especially important when using unkown length elements
1395 // as the check for whether a child exceeds its containing master
1396 // element is not effective in this situation.
1398 matroska->unknown_count = 0;
1400 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1402 if (matroska->unknown_count > 3)
1403 dist += pos_alt - matroska->resync_pos;
1405 if (dist > SKIP_THRESHOLD) {
1406 av_log(matroska->ctx, AV_LOG_ERROR,
1407 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1408 "length 0x%"PRIx64" considered as invalid data. Last "
1409 "known good position 0x%"PRIx64", %d unknown elements"
1410 " in a row\n", id, pos, length, matroska->resync_pos,
1411 matroska->unknown_count);
1412 return AVERROR_INVALIDDATA;
1417 if (update_pos > 0) {
1418 // We have found an element that is allowed at this place
1419 // in the hierarchy and it passed all checks, so treat the beginning
1420 // of the element as the "last known good" position.
1421 matroska->resync_pos = pos;
1424 if (!data && length != EBML_UNKNOWN_LENGTH)
1428 switch (syntax->type) {
1430 res = ebml_read_uint(pb, length, syntax->def.u, data);
1433 res = ebml_read_sint(pb, length, syntax->def.i, data);
1436 res = ebml_read_float(pb, length, syntax->def.f, data);
1440 res = ebml_read_ascii(pb, length, syntax->def.s, data);
1443 res = ebml_read_binary(pb, length, pos_alt, data);
1447 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1449 if (id == MATROSKA_ID_SEGMENT)
1450 matroska->segment_start = pos_alt;
1451 if (id == MATROSKA_ID_CUES)
1452 matroska->cues_parsing_deferred = 0;
1453 if (syntax->type == EBML_LEVEL1 &&
1454 (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1455 if (!level1_elem->pos) {
1456 // Zero is not a valid position for a level 1 element.
1457 level1_elem->pos = pos;
1458 } else if (level1_elem->pos != pos)
1459 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1460 level1_elem->parsed = 1;
1462 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1471 if (ffio_limit(pb, length) != length) {
1472 // ffio_limit emits its own error message,
1473 // so we don't have to.
1474 return AVERROR(EIO);
1476 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1477 // avio_skip might take us past EOF. We check for this
1478 // by skipping only length - 1 bytes, reading a byte and
1479 // checking the error flags. This is done in order to check
1480 // that the element has been properly skipped even when
1481 // no filesize (that ffio_limit relies on) is available.
1483 res = NEEDS_CHECKING;
1490 if (res == NEEDS_CHECKING) {
1491 if (pb->eof_reached) {
1500 if (res == AVERROR_INVALIDDATA)
1501 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1502 else if (res == AVERROR(EIO))
1503 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1504 else if (res == AVERROR_EOF) {
1505 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1513 if (syntax->is_counted && data) {
1514 CountedElement *elem = data;
1515 if (elem->count != UINT_MAX)
1519 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1520 level = &matroska->levels[matroska->num_levels - 1];
1521 pos = avio_tell(pb);
1523 // Given that pos >= level->start no check for
1524 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1525 while (matroska->num_levels && pos == level->start + level->length) {
1526 matroska->num_levels--;
1534 static void ebml_free(EbmlSyntax *syntax, void *data)
1537 for (i = 0; syntax[i].id; i++) {
1538 void *data_off = (char *) data + syntax[i].data_offset;
1539 switch (syntax[i].type) {
1545 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1549 if (syntax[i].list_elem_size) {
1550 EbmlList *list = data_off;
1551 char *ptr = list->elem;
1552 for (j = 0; j < list->nb_elem;
1553 j++, ptr += syntax[i].list_elem_size)
1554 ebml_free(syntax[i].def.n, ptr);
1555 av_freep(&list->elem);
1557 list->alloc_elem_size = 0;
1559 ebml_free(syntax[i].def.n, data_off);
1569 static int matroska_probe(const AVProbeData *p)
1572 int len_mask = 0x80, size = 1, n = 1, i;
1575 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1578 /* length of header */
1580 while (size <= 8 && !(total & len_mask)) {
1586 total &= (len_mask - 1);
1588 total = (total << 8) | p->buf[4 + n++];
1590 if (total + 1 == 1ULL << (7 * size)){
1591 /* Unknown-length header - simply parse the whole buffer. */
1592 total = p->buf_size - 4 - size;
1594 /* Does the probe data contain the whole header? */
1595 if (p->buf_size < 4 + size + total)
1599 /* The header should contain a known document type. For now,
1600 * we don't parse the whole header but simply check for the
1601 * availability of that array of characters inside the header.
1602 * Not fully fool-proof, but good enough. */
1603 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1604 size_t probelen = strlen(matroska_doctypes[i]);
1605 if (total < probelen)
1607 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1608 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1609 return AVPROBE_SCORE_MAX;
1612 // probably valid EBML header but no recognized doctype
1613 return AVPROBE_SCORE_EXTENSION;
1616 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1619 MatroskaTrack *tracks = matroska->tracks.elem;
1622 for (i = 0; i < matroska->tracks.nb_elem; i++)
1623 if (tracks[i].num == num)
1626 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1630 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1631 MatroskaTrack *track)
1633 MatroskaTrackEncoding *encodings = track->encodings.elem;
1634 uint8_t *data = *buf;
1635 int isize = *buf_size;
1636 uint8_t *pkt_data = NULL;
1637 uint8_t av_unused *newpktdata;
1638 int pkt_size = isize;
1642 if (pkt_size >= 10000000U)
1643 return AVERROR_INVALIDDATA;
1645 switch (encodings[0].compression.algo) {
1646 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1648 int header_size = encodings[0].compression.settings.size;
1649 uint8_t *header = encodings[0].compression.settings.data;
1651 if (header_size && !header) {
1652 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1659 pkt_size = isize + header_size;
1660 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1662 return AVERROR(ENOMEM);
1664 memcpy(pkt_data, header, header_size);
1665 memcpy(pkt_data + header_size, data, isize);
1669 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1672 olen = pkt_size *= 3;
1673 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1674 + AV_INPUT_BUFFER_PADDING_SIZE);
1676 result = AVERROR(ENOMEM);
1679 pkt_data = newpktdata;
1680 result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1681 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1683 result = AVERROR_INVALIDDATA;
1690 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1692 z_stream zstream = { 0 };
1693 if (inflateInit(&zstream) != Z_OK)
1695 zstream.next_in = data;
1696 zstream.avail_in = isize;
1699 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1701 inflateEnd(&zstream);
1702 result = AVERROR(ENOMEM);
1705 pkt_data = newpktdata;
1706 zstream.avail_out = pkt_size - zstream.total_out;
1707 zstream.next_out = pkt_data + zstream.total_out;
1708 result = inflate(&zstream, Z_NO_FLUSH);
1709 } while (result == Z_OK && pkt_size < 10000000);
1710 pkt_size = zstream.total_out;
1711 inflateEnd(&zstream);
1712 if (result != Z_STREAM_END) {
1713 if (result == Z_MEM_ERROR)
1714 result = AVERROR(ENOMEM);
1716 result = AVERROR_INVALIDDATA;
1723 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1725 bz_stream bzstream = { 0 };
1726 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1728 bzstream.next_in = data;
1729 bzstream.avail_in = isize;
1732 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1734 BZ2_bzDecompressEnd(&bzstream);
1735 result = AVERROR(ENOMEM);
1738 pkt_data = newpktdata;
1739 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1740 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1741 result = BZ2_bzDecompress(&bzstream);
1742 } while (result == BZ_OK && pkt_size < 10000000);
1743 pkt_size = bzstream.total_out_lo32;
1744 BZ2_bzDecompressEnd(&bzstream);
1745 if (result != BZ_STREAM_END) {
1746 if (result == BZ_MEM_ERROR)
1747 result = AVERROR(ENOMEM);
1749 result = AVERROR_INVALIDDATA;
1756 return AVERROR_INVALIDDATA;
1759 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1762 *buf_size = pkt_size;
1770 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1771 AVDictionary **metadata, char *prefix)
1773 MatroskaTag *tags = list->elem;
1777 for (i = 0; i < list->nb_elem; i++) {
1778 const char *lang = tags[i].lang &&
1779 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1781 if (!tags[i].name) {
1782 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1786 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1788 av_strlcpy(key, tags[i].name, sizeof(key));
1789 if (tags[i].def || !lang) {
1790 av_dict_set(metadata, key, tags[i].string, 0);
1791 if (tags[i].sub.nb_elem)
1792 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1795 av_strlcat(key, "-", sizeof(key));
1796 av_strlcat(key, lang, sizeof(key));
1797 av_dict_set(metadata, key, tags[i].string, 0);
1798 if (tags[i].sub.nb_elem)
1799 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1802 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1805 static void matroska_convert_tags(AVFormatContext *s)
1807 MatroskaDemuxContext *matroska = s->priv_data;
1808 MatroskaTags *tags = matroska->tags.elem;
1811 for (i = 0; i < matroska->tags.nb_elem; i++) {
1812 if (tags[i].target.attachuid) {
1813 MatroskaAttachment *attachment = matroska->attachments.elem;
1815 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1816 if (attachment[j].uid == tags[i].target.attachuid &&
1817 attachment[j].stream) {
1818 matroska_convert_tag(s, &tags[i].tag,
1819 &attachment[j].stream->metadata, NULL);
1824 av_log(s, AV_LOG_WARNING,
1825 "The tags at index %d refer to a "
1826 "non-existent attachment %"PRId64".\n",
1827 i, tags[i].target.attachuid);
1829 } else if (tags[i].target.chapteruid) {
1830 MatroskaChapter *chapter = matroska->chapters.elem;
1832 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1833 if (chapter[j].uid == tags[i].target.chapteruid &&
1834 chapter[j].chapter) {
1835 matroska_convert_tag(s, &tags[i].tag,
1836 &chapter[j].chapter->metadata, NULL);
1841 av_log(s, AV_LOG_WARNING,
1842 "The tags at index %d refer to a non-existent chapter "
1844 i, tags[i].target.chapteruid);
1846 } else if (tags[i].target.trackuid) {
1847 MatroskaTrack *track = matroska->tracks.elem;
1849 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1850 if (track[j].uid == tags[i].target.trackuid &&
1852 matroska_convert_tag(s, &tags[i].tag,
1853 &track[j].stream->metadata, NULL);
1858 av_log(s, AV_LOG_WARNING,
1859 "The tags at index %d refer to a non-existent track "
1861 i, tags[i].target.trackuid);
1864 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1865 tags[i].target.type);
1870 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1873 uint32_t saved_id = matroska->current_id;
1874 int64_t before_pos = avio_tell(matroska->ctx->pb);
1878 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1879 /* We don't want to lose our seekhead level, so we add
1880 * a dummy. This is a crude hack. */
1881 if (matroska->num_levels == EBML_MAX_DEPTH) {
1882 av_log(matroska->ctx, AV_LOG_INFO,
1883 "Max EBML element depth (%d) reached, "
1884 "cannot parse further.\n", EBML_MAX_DEPTH);
1885 ret = AVERROR_INVALIDDATA;
1887 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1888 matroska->num_levels++;
1889 matroska->current_id = 0;
1891 ret = ebml_parse(matroska, matroska_segment, matroska);
1892 if (ret == LEVEL_ENDED) {
1893 /* This can only happen if the seek brought us beyond EOF. */
1898 /* Seek back - notice that in all instances where this is used
1899 * it is safe to set the level to 1. */
1900 matroska_reset_status(matroska, saved_id, before_pos);
1905 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1907 EbmlList *seekhead_list = &matroska->seekhead;
1910 // we should not do any seeking in the streaming case
1911 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1914 for (i = 0; i < seekhead_list->nb_elem; i++) {
1915 MatroskaSeekhead *seekheads = seekhead_list->elem;
1916 uint32_t id = seekheads[i].id;
1917 int64_t pos = seekheads[i].pos + matroska->segment_start;
1918 MatroskaLevel1Element *elem;
1920 if (id != seekheads[i].id || pos < matroska->segment_start)
1923 elem = matroska_find_level1_elem(matroska, id, pos);
1924 if (!elem || elem->parsed)
1929 // defer cues parsing until we actually need cue data.
1930 if (id == MATROSKA_ID_CUES)
1933 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1934 // mark index as broken
1935 matroska->cues_parsing_deferred = -1;
1943 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1945 EbmlList *index_list;
1946 MatroskaIndex *index;
1947 uint64_t index_scale = 1;
1950 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1953 index_list = &matroska->index;
1954 index = index_list->elem;
1955 if (index_list->nb_elem < 2)
1957 if (index[1].time > 1E14 / matroska->time_scale) {
1958 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1961 for (i = 0; i < index_list->nb_elem; i++) {
1962 EbmlList *pos_list = &index[i].pos;
1963 MatroskaIndexPos *pos = pos_list->elem;
1964 for (j = 0; j < pos_list->nb_elem; j++) {
1965 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1967 if (track && track->stream)
1968 av_add_index_entry(track->stream,
1969 pos[j].pos + matroska->segment_start,
1970 index[i].time / index_scale, 0, 0,
1976 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1979 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1982 for (i = 0; i < matroska->num_level1_elems; i++) {
1983 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1984 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1985 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1986 matroska->cues_parsing_deferred = -1;
1992 matroska_add_index_entries(matroska);
1995 static int matroska_aac_profile(char *codec_id)
1997 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
2000 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
2001 if (strstr(codec_id, aac_profiles[profile]))
2006 static int matroska_aac_sri(int samplerate)
2010 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
2011 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
2016 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
2018 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
2019 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
2022 static int matroska_parse_flac(AVFormatContext *s,
2023 MatroskaTrack *track,
2026 AVStream *st = track->stream;
2027 uint8_t *p = track->codec_priv.data;
2028 int size = track->codec_priv.size;
2030 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
2031 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
2032 track->codec_priv.size = 0;
2036 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
2038 p += track->codec_priv.size;
2039 size -= track->codec_priv.size;
2041 /* parse the remaining metadata blocks if present */
2043 int block_last, block_type, block_size;
2045 flac_parse_block_header(p, &block_last, &block_type, &block_size);
2049 if (block_size > size)
2052 /* check for the channel mask */
2053 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2054 AVDictionary *dict = NULL;
2055 AVDictionaryEntry *chmask;
2057 ff_vorbis_comment(s, &dict, p, block_size, 0);
2058 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2060 uint64_t mask = strtol(chmask->value, NULL, 0);
2061 if (!mask || mask & ~0x3ffffULL) {
2062 av_log(s, AV_LOG_WARNING,
2063 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2065 st->codecpar->channel_layout = mask;
2067 av_dict_free(&dict);
2077 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2079 int minor, micro, bttb = 0;
2081 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2082 * this function, and fixed in 57.52 */
2083 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, µ) == 2)
2084 bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2086 switch (field_order) {
2087 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2088 return AV_FIELD_PROGRESSIVE;
2089 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2090 return AV_FIELD_UNKNOWN;
2091 case MATROSKA_VIDEO_FIELDORDER_TT:
2093 case MATROSKA_VIDEO_FIELDORDER_BB:
2095 case MATROSKA_VIDEO_FIELDORDER_BT:
2096 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2097 case MATROSKA_VIDEO_FIELDORDER_TB:
2098 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2100 return AV_FIELD_UNKNOWN;
2104 static void mkv_stereo_mode_display_mul(int stereo_mode,
2105 int *h_width, int *h_height)
2107 switch (stereo_mode) {
2108 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2109 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2110 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2111 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2112 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2114 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2115 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2116 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2117 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2120 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2121 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2122 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2123 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2129 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2130 const MatroskaTrackVideoColor *color = track->video.color.elem;
2131 const MatroskaMasteringMeta *mastering_meta;
2132 int has_mastering_primaries, has_mastering_luminance;
2134 if (!track->video.color.nb_elem)
2137 mastering_meta = &color->mastering_meta;
2138 // Mastering primaries are CIE 1931 coords, and must be > 0.
2139 has_mastering_primaries =
2140 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2141 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2142 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2143 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2144 has_mastering_luminance = mastering_meta->max_luminance >
2145 mastering_meta->min_luminance.el.f &&
2146 mastering_meta->min_luminance.el.f >= 0 &&
2147 mastering_meta->min_luminance.count;
2149 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2150 st->codecpar->color_space = color->matrix_coefficients;
2151 if (color->primaries != AVCOL_PRI_RESERVED &&
2152 color->primaries != AVCOL_PRI_RESERVED0)
2153 st->codecpar->color_primaries = color->primaries;
2154 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2155 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2156 st->codecpar->color_trc = color->transfer_characteristics;
2157 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2158 color->range <= AVCOL_RANGE_JPEG)
2159 st->codecpar->color_range = color->range;
2160 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2161 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2162 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2163 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2164 st->codecpar->chroma_location =
2165 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2166 (color->chroma_siting_vert - 1) << 7);
2168 if (color->max_cll && color->max_fall) {
2171 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2173 return AVERROR(ENOMEM);
2174 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2175 (uint8_t *)metadata, size);
2177 av_freep(&metadata);
2180 metadata->MaxCLL = color->max_cll;
2181 metadata->MaxFALL = color->max_fall;
2184 if (has_mastering_primaries || has_mastering_luminance) {
2185 AVMasteringDisplayMetadata *metadata =
2186 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2187 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2188 sizeof(AVMasteringDisplayMetadata));
2190 return AVERROR(ENOMEM);
2192 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2193 if (has_mastering_primaries) {
2194 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2195 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2196 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2197 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2198 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2199 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2200 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2201 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2202 metadata->has_primaries = 1;
2204 if (has_mastering_luminance) {
2205 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2206 metadata->min_luminance = av_d2q(mastering_meta->min_luminance.el.f, INT_MAX);
2207 metadata->has_luminance = 1;
2213 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2216 AVSphericalMapping *spherical;
2217 const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2218 const uint8_t *priv_data = mkv_projection->private.data;
2219 enum AVSphericalProjection projection;
2220 size_t spherical_size;
2221 uint32_t l = 0, t = 0, r = 0, b = 0;
2222 uint32_t padding = 0;
2225 if (mkv_projection->private.size && priv_data[0] != 0) {
2226 av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2230 switch (track->video.projection.type) {
2231 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2232 if (track->video.projection.private.size == 20) {
2233 t = AV_RB32(priv_data + 4);
2234 b = AV_RB32(priv_data + 8);
2235 l = AV_RB32(priv_data + 12);
2236 r = AV_RB32(priv_data + 16);
2238 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2239 av_log(logctx, AV_LOG_ERROR,
2240 "Invalid bounding rectangle coordinates "
2241 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2243 return AVERROR_INVALIDDATA;
2245 } else if (track->video.projection.private.size != 0) {
2246 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2247 return AVERROR_INVALIDDATA;
2250 if (l || t || r || b)
2251 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2253 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2255 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2256 if (track->video.projection.private.size < 4) {
2257 av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2258 return AVERROR_INVALIDDATA;
2259 } else if (track->video.projection.private.size == 12) {
2260 uint32_t layout = AV_RB32(priv_data + 4);
2262 av_log(logctx, AV_LOG_WARNING,
2263 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2266 projection = AV_SPHERICAL_CUBEMAP;
2267 padding = AV_RB32(priv_data + 8);
2269 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2270 return AVERROR_INVALIDDATA;
2273 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2274 /* No Spherical metadata */
2277 av_log(logctx, AV_LOG_WARNING,
2278 "Unknown spherical metadata type %"PRIu64"\n",
2279 track->video.projection.type);
2283 spherical = av_spherical_alloc(&spherical_size);
2285 return AVERROR(ENOMEM);
2287 spherical->projection = projection;
2289 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2290 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2291 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2293 spherical->padding = padding;
2295 spherical->bound_left = l;
2296 spherical->bound_top = t;
2297 spherical->bound_right = r;
2298 spherical->bound_bottom = b;
2300 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2303 av_freep(&spherical);
2310 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2312 const AVCodecTag *codec_tags;
2314 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2315 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2317 /* Normalize noncompliant private data that starts with the fourcc
2318 * by expanding/shifting the data by 4 bytes and storing the data
2319 * size at the start. */
2320 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2321 int ret = av_buffer_realloc(&track->codec_priv.buf,
2322 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2326 track->codec_priv.data = track->codec_priv.buf->data;
2327 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2328 track->codec_priv.size += 4;
2329 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2332 *fourcc = AV_RL32(track->codec_priv.data + 4);
2333 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2338 static int matroska_parse_tracks(AVFormatContext *s)
2340 MatroskaDemuxContext *matroska = s->priv_data;
2341 MatroskaTrack *tracks = matroska->tracks.elem;
2346 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2347 MatroskaTrack *track = &tracks[i];
2348 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2349 EbmlList *encodings_list = &track->encodings;
2350 MatroskaTrackEncoding *encodings = encodings_list->elem;
2351 uint8_t *extradata = NULL;
2352 int extradata_size = 0;
2353 int extradata_offset = 0;
2354 uint32_t fourcc = 0;
2356 char* key_id_base64 = NULL;
2359 /* Apply some sanity checks. */
2360 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2361 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2362 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2363 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2364 av_log(matroska->ctx, AV_LOG_INFO,
2365 "Unknown or unsupported track type %"PRIu64"\n",
2369 if (!track->codec_id)
2372 if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
2373 || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
2374 || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2375 || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2377 av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
2381 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2382 isnan(track->audio.samplerate)) {
2383 av_log(matroska->ctx, AV_LOG_WARNING,
2384 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2385 track->audio.samplerate);
2386 track->audio.samplerate = 8000;
2389 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2390 if (!track->default_duration && track->video.frame_rate > 0) {
2391 double default_duration = 1000000000 / track->video.frame_rate;
2392 if (default_duration > UINT64_MAX || default_duration < 0) {
2393 av_log(matroska->ctx, AV_LOG_WARNING,
2394 "Invalid frame rate %e. Cannot calculate default duration.\n",
2395 track->video.frame_rate);
2397 track->default_duration = default_duration;
2400 if (track->video.display_width == -1)
2401 track->video.display_width = track->video.pixel_width;
2402 if (track->video.display_height == -1)
2403 track->video.display_height = track->video.pixel_height;
2404 if (track->video.color_space.size == 4)
2405 fourcc = AV_RL32(track->video.color_space.data);
2406 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2407 if (!track->audio.out_samplerate)
2408 track->audio.out_samplerate = track->audio.samplerate;
2410 if (encodings_list->nb_elem > 1) {
2411 av_log(matroska->ctx, AV_LOG_ERROR,
2412 "Multiple combined encodings not supported");
2413 } else if (encodings_list->nb_elem == 1) {
2414 if (encodings[0].type) {
2415 if (encodings[0].encryption.key_id.size > 0) {
2416 /* Save the encryption key id to be stored later as a
2418 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2419 key_id_base64 = av_malloc(b64_size);
2420 if (key_id_base64 == NULL)
2421 return AVERROR(ENOMEM);
2423 av_base64_encode(key_id_base64, b64_size,
2424 encodings[0].encryption.key_id.data,
2425 encodings[0].encryption.key_id.size);
2427 encodings[0].scope = 0;
2428 av_log(matroska->ctx, AV_LOG_ERROR,
2429 "Unsupported encoding type");
2433 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2436 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2439 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2441 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2442 encodings[0].scope = 0;
2443 av_log(matroska->ctx, AV_LOG_ERROR,
2444 "Unsupported encoding type");
2445 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2446 uint8_t *codec_priv = track->codec_priv.data;
2447 int ret = matroska_decode_buffer(&track->codec_priv.data,
2448 &track->codec_priv.size,
2451 track->codec_priv.data = NULL;
2452 track->codec_priv.size = 0;
2453 av_log(matroska->ctx, AV_LOG_ERROR,
2454 "Failed to decode codec private data\n");
2457 if (codec_priv != track->codec_priv.data) {
2458 av_buffer_unref(&track->codec_priv.buf);
2459 if (track->codec_priv.data) {
2460 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2461 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2463 if (!track->codec_priv.buf) {
2464 av_freep(&track->codec_priv.data);
2465 track->codec_priv.size = 0;
2466 return AVERROR(ENOMEM);
2472 track->needs_decoding = encodings && !encodings[0].type &&
2473 encodings[0].scope & 1 &&
2474 (encodings[0].compression.algo !=
2475 MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP ||
2476 encodings[0].compression.settings.size);
2478 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2479 if (av_strstart(track->codec_id, ff_mkv_codec_tags[j].str, NULL)) {
2480 codec_id = ff_mkv_codec_tags[j].id;
2485 st = track->stream = avformat_new_stream(s, NULL);
2487 av_free(key_id_base64);
2488 return AVERROR(ENOMEM);
2491 if (key_id_base64) {
2492 /* export encryption key id as base64 metadata tag */
2493 av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2494 AV_DICT_DONT_STRDUP_VAL);
2497 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2498 track->codec_priv.size >= 40 &&
2499 track->codec_priv.data) {
2500 track->ms_compat = 1;
2501 bit_depth = AV_RL16(track->codec_priv.data + 14);
2502 fourcc = AV_RL32(track->codec_priv.data + 16);
2503 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2506 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2508 extradata_offset = 40;
2509 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2510 track->codec_priv.size >= 14 &&
2511 track->codec_priv.data) {
2513 ffio_init_context(&b, track->codec_priv.data,
2514 track->codec_priv.size,
2515 0, NULL, NULL, NULL, NULL);
2516 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2519 codec_id = st->codecpar->codec_id;
2520 fourcc = st->codecpar->codec_tag;
2521 extradata_offset = FFMIN(track->codec_priv.size, 18);
2522 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2523 /* Normally 36, but allow noncompliant private data */
2524 && (track->codec_priv.size >= 32)
2525 && (track->codec_priv.data)) {
2526 uint16_t sample_size;
2527 int ret = get_qt_codec(track, &fourcc, &codec_id);
2530 sample_size = AV_RB16(track->codec_priv.data + 26);
2532 if (sample_size == 8) {
2533 fourcc = MKTAG('r','a','w',' ');
2534 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2535 } else if (sample_size == 16) {
2536 fourcc = MKTAG('t','w','o','s');
2537 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2540 if ((fourcc == MKTAG('t','w','o','s') ||
2541 fourcc == MKTAG('s','o','w','t')) &&
2543 codec_id = AV_CODEC_ID_PCM_S8;
2544 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2545 (track->codec_priv.size >= 21) &&
2546 (track->codec_priv.data)) {
2547 int ret = get_qt_codec(track, &fourcc, &codec_id);
2550 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2551 fourcc = MKTAG('S','V','Q','3');
2552 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2554 if (codec_id == AV_CODEC_ID_NONE)
2555 av_log(matroska->ctx, AV_LOG_ERROR,
2556 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2557 if (track->codec_priv.size >= 86) {
2558 bit_depth = AV_RB16(track->codec_priv.data + 82);
2559 ffio_init_context(&b, track->codec_priv.data,
2560 track->codec_priv.size,
2561 0, NULL, NULL, NULL, NULL);
2562 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2564 track->has_palette = 1;
2567 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2568 switch (track->audio.bitdepth) {
2570 codec_id = AV_CODEC_ID_PCM_U8;
2573 codec_id = AV_CODEC_ID_PCM_S24BE;
2576 codec_id = AV_CODEC_ID_PCM_S32BE;
2579 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2580 switch (track->audio.bitdepth) {
2582 codec_id = AV_CODEC_ID_PCM_U8;
2585 codec_id = AV_CODEC_ID_PCM_S24LE;
2588 codec_id = AV_CODEC_ID_PCM_S32LE;
2591 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2592 track->audio.bitdepth == 64) {
2593 codec_id = AV_CODEC_ID_PCM_F64LE;
2594 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2595 int profile = matroska_aac_profile(track->codec_id);
2596 int sri = matroska_aac_sri(track->audio.samplerate);
2597 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2599 return AVERROR(ENOMEM);
2600 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2601 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2602 if (strstr(track->codec_id, "SBR")) {
2603 sri = matroska_aac_sri(track->audio.out_samplerate);
2604 extradata[2] = 0x56;
2605 extradata[3] = 0xE5;
2606 extradata[4] = 0x80 | (sri << 3);
2610 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2611 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2612 * Create the "atom size", "tag", and "tag version" fields the
2613 * decoder expects manually. */
2614 extradata_size = 12 + track->codec_priv.size;
2615 extradata = av_mallocz(extradata_size +
2616 AV_INPUT_BUFFER_PADDING_SIZE);
2618 return AVERROR(ENOMEM);
2619 AV_WB32(extradata, extradata_size);
2620 memcpy(&extradata[4], "alac", 4);
2621 AV_WB32(&extradata[8], 0);
2622 memcpy(&extradata[12], track->codec_priv.data,
2623 track->codec_priv.size);
2624 } else if (codec_id == AV_CODEC_ID_TTA) {
2626 if (track->audio.channels > UINT16_MAX ||
2627 track->audio.bitdepth > UINT16_MAX) {
2628 av_log(matroska->ctx, AV_LOG_WARNING,
2629 "Too large audio channel number %"PRIu64
2630 " or bitdepth %"PRIu64". Skipping track.\n",
2631 track->audio.channels, track->audio.bitdepth);
2632 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2633 return AVERROR_INVALIDDATA;
2637 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2638 return AVERROR_INVALIDDATA;
2639 extradata_size = 22;
2640 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2642 return AVERROR(ENOMEM);
2644 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2645 bytestream_put_le16(&ptr, 1);
2646 bytestream_put_le16(&ptr, track->audio.channels);
2647 bytestream_put_le16(&ptr, track->audio.bitdepth);
2648 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2649 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2650 track->audio.out_samplerate,
2651 AV_TIME_BASE * 1000));
2652 } else if (codec_id == AV_CODEC_ID_RV10 ||
2653 codec_id == AV_CODEC_ID_RV20 ||
2654 codec_id == AV_CODEC_ID_RV30 ||
2655 codec_id == AV_CODEC_ID_RV40) {
2656 extradata_offset = 26;
2657 } else if (codec_id == AV_CODEC_ID_RA_144) {
2658 track->audio.out_samplerate = 8000;
2659 track->audio.channels = 1;
2660 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2661 codec_id == AV_CODEC_ID_COOK ||
2662 codec_id == AV_CODEC_ID_ATRAC3 ||
2663 codec_id == AV_CODEC_ID_SIPR)
2664 && track->codec_priv.data) {
2667 ffio_init_context(&b, track->codec_priv.data,
2668 track->codec_priv.size,
2669 0, NULL, NULL, NULL, NULL);
2671 flavor = avio_rb16(&b);
2672 track->audio.coded_framesize = avio_rb32(&b);
2674 track->audio.sub_packet_h = avio_rb16(&b);
2675 track->audio.frame_size = avio_rb16(&b);
2676 track->audio.sub_packet_size = avio_rb16(&b);
2677 if (track->audio.coded_framesize <= 0 ||
2678 track->audio.sub_packet_h <= 0 ||
2679 track->audio.frame_size <= 0)
2680 return AVERROR_INVALIDDATA;
2682 if (codec_id == AV_CODEC_ID_RA_288) {
2683 if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2684 != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2685 return AVERROR_INVALIDDATA;
2686 st->codecpar->block_align = track->audio.coded_framesize;
2687 track->codec_priv.size = 0;
2689 if (codec_id == AV_CODEC_ID_SIPR) {
2690 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2692 return AVERROR_INVALIDDATA;
2693 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2694 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2695 } else if (track->audio.sub_packet_size <= 0 ||
2696 track->audio.frame_size % track->audio.sub_packet_size)
2697 return AVERROR_INVALIDDATA;
2698 st->codecpar->block_align = track->audio.sub_packet_size;
2699 extradata_offset = 78;
2701 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2702 track->audio.frame_size);
2703 if (!track->audio.buf)
2704 return AVERROR(ENOMEM);
2705 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2706 ret = matroska_parse_flac(s, track, &extradata_offset);
2709 } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2710 av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2711 "in absence of valid CodecPrivate.\n");
2713 extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2715 return AVERROR(ENOMEM);
2716 AV_WL16(extradata, 0x410);
2717 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2718 fourcc = AV_RL32(track->codec_priv.data);
2719 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2720 /* we don't need any value stored in CodecPrivate.
2721 make sure that it's not exported as extradata. */
2722 track->codec_priv.size = 0;
2723 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2724 /* For now, propagate only the OBUs, if any. Once libavcodec is
2725 updated to handle isobmff style extradata this can be removed. */
2726 extradata_offset = 4;
2728 track->codec_priv.size -= extradata_offset;
2730 if (codec_id == AV_CODEC_ID_NONE)
2731 av_log(matroska->ctx, AV_LOG_INFO,
2732 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2734 if (track->time_scale < 0.01) {
2735 av_log(matroska->ctx, AV_LOG_WARNING,
2736 "Track TimestampScale too small %f, assuming 1.0.\n",
2738 track->time_scale = 1.0;
2740 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2741 1000 * 1000 * 1000); /* 64 bit pts in ns */
2743 /* convert the delay from ns to the track timebase */
2744 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2745 (AVRational){ 1, 1000000000 },
2748 st->codecpar->codec_id = codec_id;
2750 if (strcmp(track->language, "und"))
2751 av_dict_set(&st->metadata, "language", track->language, 0);
2752 av_dict_set(&st->metadata, "title", track->name, 0);
2754 if (track->flag_default)
2755 st->disposition |= AV_DISPOSITION_DEFAULT;
2756 if (track->flag_forced)
2757 st->disposition |= AV_DISPOSITION_FORCED;
2758 if (track->flag_comment)
2759 st->disposition |= AV_DISPOSITION_COMMENT;
2760 if (track->flag_hearingimpaired)
2761 st->disposition |= AV_DISPOSITION_HEARING_IMPAIRED;
2762 if (track->flag_visualimpaired)
2763 st->disposition |= AV_DISPOSITION_VISUAL_IMPAIRED;
2764 if (track->flag_original.count > 0)
2765 st->disposition |= track->flag_original.el.u ? AV_DISPOSITION_ORIGINAL
2766 : AV_DISPOSITION_DUB;
2768 if (!st->codecpar->extradata) {
2770 st->codecpar->extradata = extradata;
2771 st->codecpar->extradata_size = extradata_size;
2772 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2773 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2774 return AVERROR(ENOMEM);
2775 memcpy(st->codecpar->extradata,
2776 track->codec_priv.data + extradata_offset,
2777 track->codec_priv.size);
2781 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2782 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2783 int display_width_mul = 1;
2784 int display_height_mul = 1;
2786 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2787 st->codecpar->codec_tag = fourcc;
2789 st->codecpar->bits_per_coded_sample = bit_depth;
2790 st->codecpar->width = track->video.pixel_width;
2791 st->codecpar->height = track->video.pixel_height;
2793 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2794 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2795 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2796 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2798 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2799 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2801 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2802 av_reduce(&st->sample_aspect_ratio.num,
2803 &st->sample_aspect_ratio.den,
2804 st->codecpar->height * track->video.display_width * display_width_mul,
2805 st->codecpar->width * track->video.display_height * display_height_mul,
2808 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2809 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2811 if (track->default_duration) {
2812 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2813 1000000000, track->default_duration, 30000);
2814 #if FF_API_R_FRAME_RATE
2815 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2816 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2817 st->r_frame_rate = st->avg_frame_rate;
2821 /* export stereo mode flag as metadata tag */
2822 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2823 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2825 /* export alpha mode flag as metadata tag */
2826 if (track->video.alpha_mode)
2827 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2829 /* if we have virtual track, mark the real tracks */
2830 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2832 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2834 snprintf(buf, sizeof(buf), "%s_%d",
2835 ff_matroska_video_stereo_plane[planes[j].type], i);
2836 for (k=0; k < matroska->tracks.nb_elem; k++)
2837 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2838 av_dict_set(&tracks[k].stream->metadata,
2839 "stereo_mode", buf, 0);
2843 // add stream level stereo3d side data if it is a supported format
2844 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2845 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2846 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2851 ret = mkv_parse_video_color(st, track);
2854 ret = mkv_parse_video_projection(st, track, matroska->ctx);
2857 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2858 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2859 st->codecpar->codec_tag = fourcc;
2860 st->codecpar->sample_rate = track->audio.out_samplerate;
2861 st->codecpar->channels = track->audio.channels;
2862 if (!st->codecpar->bits_per_coded_sample)
2863 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2864 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2865 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2866 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2867 st->need_parsing = AVSTREAM_PARSE_FULL;
2868 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2869 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2870 if (track->codec_delay > 0) {
2871 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2872 (AVRational){1, 1000000000},
2873 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2874 48000 : st->codecpar->sample_rate});
2876 if (track->seek_preroll > 0) {
2877 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2878 (AVRational){1, 1000000000},
2879 (AVRational){1, st->codecpar->sample_rate});
2881 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2882 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2884 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2885 st->disposition |= AV_DISPOSITION_CAPTIONS;
2886 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2887 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2888 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2889 st->disposition |= AV_DISPOSITION_METADATA;
2891 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2892 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2894 if (track->flag_textdescriptions)
2895 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2902 static int matroska_read_header(AVFormatContext *s)
2904 MatroskaDemuxContext *matroska = s->priv_data;
2905 EbmlList *attachments_list = &matroska->attachments;
2906 EbmlList *chapters_list = &matroska->chapters;
2907 MatroskaAttachment *attachments;
2908 MatroskaChapter *chapters;
2909 uint64_t max_start = 0;
2915 matroska->cues_parsing_deferred = 1;
2917 /* First read the EBML header. */
2918 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2919 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2920 ebml_free(ebml_syntax, &ebml);
2921 return AVERROR_INVALIDDATA;
2923 if (ebml.version > EBML_VERSION ||
2924 ebml.max_size > sizeof(uint64_t) ||
2925 ebml.id_length > sizeof(uint32_t) ||
2926 ebml.doctype_version > 3) {
2927 avpriv_report_missing_feature(matroska->ctx,
2928 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2929 ebml.version, ebml.doctype, ebml.doctype_version);
2930 ebml_free(ebml_syntax, &ebml);
2931 return AVERROR_PATCHWELCOME;
2932 } else if (ebml.doctype_version == 3) {
2933 av_log(matroska->ctx, AV_LOG_WARNING,
2934 "EBML header using unsupported features\n"
2935 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2936 ebml.version, ebml.doctype, ebml.doctype_version);
2938 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2939 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2941 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2942 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2943 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2944 ebml_free(ebml_syntax, &ebml);
2945 return AVERROR_INVALIDDATA;
2948 ebml_free(ebml_syntax, &ebml);
2950 matroska->pkt = s->internal->parse_pkt;
2952 /* The next thing is a segment. */
2953 pos = avio_tell(matroska->ctx->pb);
2954 res = ebml_parse(matroska, matroska_segments, matroska);
2955 // Try resyncing until we find an EBML_STOP type element.
2957 res = matroska_resync(matroska, pos);
2960 pos = avio_tell(matroska->ctx->pb);
2961 res = ebml_parse(matroska, matroska_segment, matroska);
2962 if (res == AVERROR(EIO)) // EOF is translated to EIO, this exists the loop on EOF
2965 /* Set data_offset as it might be needed later by seek_frame_generic. */
2966 if (matroska->current_id == MATROSKA_ID_CLUSTER)
2967 s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2968 matroska_execute_seekhead(matroska);
2970 if (!matroska->time_scale)
2971 matroska->time_scale = 1000000;
2972 if (matroska->duration)
2973 matroska->ctx->duration = matroska->duration * matroska->time_scale *
2974 1000 / AV_TIME_BASE;
2975 av_dict_set(&s->metadata, "title", matroska->title, 0);
2976 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2978 if (matroska->date_utc.size == 8)
2979 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2981 res = matroska_parse_tracks(s);
2985 attachments = attachments_list->elem;
2986 for (j = 0; j < attachments_list->nb_elem; j++) {
2987 if (!(attachments[j].filename && attachments[j].mime &&
2988 attachments[j].bin.data && attachments[j].bin.size > 0)) {
2989 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2991 AVStream *st = avformat_new_stream(s, NULL);
2994 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2995 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2996 if (attachments[j].description)
2997 av_dict_set(&st->metadata, "title", attachments[j].description, 0);
2998 st->codecpar->codec_id = AV_CODEC_ID_NONE;
3000 for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3001 if (av_strstart(attachments[j].mime, mkv_image_mime_tags[i].str, NULL)) {
3002 st->codecpar->codec_id = mkv_image_mime_tags[i].id;
3007 attachments[j].stream = st;
3009 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
3010 res = ff_add_attached_pic(s, st, NULL, &attachments[j].bin.buf, 0);
3014 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
3015 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
3017 memcpy(st->codecpar->extradata, attachments[j].bin.data,
3018 attachments[j].bin.size);
3020 for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3021 if (av_strstart(attachments[j].mime, mkv_mime_tags[i].str, NULL)) {
3022 st->codecpar->codec_id = mkv_mime_tags[i].id;
3030 chapters = chapters_list->elem;
3031 for (i = 0; i < chapters_list->nb_elem; i++)
3032 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
3033 (max_start == 0 || chapters[i].start > max_start)) {
3034 chapters[i].chapter =
3035 avpriv_new_chapter(s, chapters[i].uid,
3036 (AVRational) { 1, 1000000000 },
3037 chapters[i].start, chapters[i].end,
3039 max_start = chapters[i].start;
3042 matroska_add_index_entries(matroska);
3044 matroska_convert_tags(s);
3048 matroska_read_close(s);
3053 * Put one packet in an application-supplied AVPacket struct.
3054 * Returns 0 on success or -1 on failure.
3056 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3059 if (matroska->queue) {
3060 MatroskaTrack *tracks = matroska->tracks.elem;
3061 MatroskaTrack *track;
3063 avpriv_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
3064 track = &tracks[pkt->stream_index];
3065 if (track->has_palette) {
3066 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
3068 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3070 memcpy(pal, track->palette, AVPALETTE_SIZE);
3072 track->has_palette = 0;
3081 * Free all packets in our internal queue.
3083 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3085 avpriv_packet_list_free(&matroska->queue, &matroska->queue_end);
3088 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3089 int size, int type, AVIOContext *pb,
3090 uint32_t lace_size[256], int *laces)
3093 uint8_t *data = *buf;
3097 lace_size[0] = size;
3102 return AVERROR_INVALIDDATA;
3109 case 0x1: /* Xiph lacing */
3113 for (n = 0; n < *laces - 1; n++) {
3118 return AVERROR_INVALIDDATA;
3121 lace_size[n] += temp;
3124 } while (temp == 0xff);
3127 return AVERROR_INVALIDDATA;
3129 lace_size[n] = size - total;
3133 case 0x2: /* fixed-size lacing */
3134 if (size % (*laces))
3135 return AVERROR_INVALIDDATA;
3136 for (n = 0; n < *laces; n++)
3137 lace_size[n] = size / *laces;
3140 case 0x3: /* EBML lacing */
3148 n = ebml_read_num(matroska, pb, 8, &num, 1);
3152 return AVERROR_INVALIDDATA;
3154 total = lace_size[0] = num;
3156 for (n = 1; n < *laces - 1; n++) {
3159 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3162 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3163 return AVERROR_INVALIDDATA;
3165 lace_size[n] = lace_size[n - 1] + snum;
3166 total += lace_size[n];
3172 return AVERROR_INVALIDDATA;
3174 lace_size[*laces - 1] = size - total;
3184 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3185 MatroskaTrack *track, AVStream *st,
3186 uint8_t *data, int size, uint64_t timecode,
3189 const int a = st->codecpar->block_align;
3190 const int sps = track->audio.sub_packet_size;
3191 const int cfs = track->audio.coded_framesize;
3192 const int h = track->audio.sub_packet_h;
3193 const int w = track->audio.frame_size;
3194 int y = track->audio.sub_packet_cnt;
3197 if (!track->audio.pkt_cnt) {
3198 if (track->audio.sub_packet_cnt == 0)
3199 track->audio.buf_timecode = timecode;
3200 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3201 if (size < cfs * h / 2) {
3202 av_log(matroska->ctx, AV_LOG_ERROR,
3203 "Corrupt int4 RM-style audio packet size\n");
3204 return AVERROR_INVALIDDATA;
3206 for (x = 0; x < h / 2; x++)
3207 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3208 data + x * cfs, cfs);
3209 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3211 av_log(matroska->ctx, AV_LOG_ERROR,
3212 "Corrupt sipr RM-style audio packet size\n");
3213 return AVERROR_INVALIDDATA;
3215 memcpy(track->audio.buf + y * w, data, w);
3218 av_log(matroska->ctx, AV_LOG_ERROR,
3219 "Corrupt generic RM-style audio packet size\n");
3220 return AVERROR_INVALIDDATA;
3222 for (x = 0; x < w / sps; x++)
3223 memcpy(track->audio.buf +
3224 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3225 data + x * sps, sps);
3228 if (++track->audio.sub_packet_cnt >= h) {
3229 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3230 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3231 track->audio.sub_packet_cnt = 0;
3232 track->audio.pkt_cnt = h * w / a;
3236 while (track->audio.pkt_cnt) {
3238 AVPacket *pkt = matroska->pkt;
3240 ret = av_new_packet(pkt, a);
3245 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3247 pkt->pts = track->audio.buf_timecode;
3248 track->audio.buf_timecode = AV_NOPTS_VALUE;
3250 pkt->stream_index = st->index;
3251 ret = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3253 av_packet_unref(pkt);
3254 return AVERROR(ENOMEM);
3261 /* reconstruct full wavpack blocks from mangled matroska ones */
3262 static int matroska_parse_wavpack(MatroskaTrack *track,
3263 uint8_t **data, int *size)
3265 uint8_t *dst = NULL;
3266 uint8_t *src = *data;
3271 int ret, offset = 0;
3274 return AVERROR_INVALIDDATA;
3276 av_assert1(track->stream->codecpar->extradata_size >= 2);
3277 ver = AV_RL16(track->stream->codecpar->extradata);
3279 samples = AV_RL32(src);
3283 while (srclen >= 8) {
3288 uint32_t flags = AV_RL32(src);
3289 uint32_t crc = AV_RL32(src + 4);
3293 multiblock = (flags & 0x1800) != 0x1800;
3296 ret = AVERROR_INVALIDDATA;
3299 blocksize = AV_RL32(src);
3305 if (blocksize > srclen) {
3306 ret = AVERROR_INVALIDDATA;
3310 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3312 ret = AVERROR(ENOMEM);
3316 dstlen += blocksize + 32;
3318 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3319 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3320 AV_WL16(dst + offset + 8, ver); // version
3321 AV_WL16(dst + offset + 10, 0); // track/index_no
3322 AV_WL32(dst + offset + 12, 0); // total samples
3323 AV_WL32(dst + offset + 16, 0); // block index
3324 AV_WL32(dst + offset + 20, samples); // number of samples
3325 AV_WL32(dst + offset + 24, flags); // flags
3326 AV_WL32(dst + offset + 28, crc); // crc
3327 memcpy(dst + offset + 32, src, blocksize); // block data
3330 srclen -= blocksize;
3331 offset += blocksize + 32;
3334 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3346 static int matroska_parse_prores(MatroskaTrack *track,
3347 uint8_t **data, int *size)
3350 int dstlen = *size + 8;
3352 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3354 return AVERROR(ENOMEM);
3356 AV_WB32(dst, dstlen);
3357 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3358 memcpy(dst + 8, *data, dstlen - 8);
3359 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3367 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3368 MatroskaTrack *track,
3370 uint8_t *data, int data_len,
3375 AVPacket *pkt = matroska->pkt;
3376 uint8_t *id, *settings, *text, *buf;
3377 int id_len, settings_len, text_len;
3382 return AVERROR_INVALIDDATA;
3385 q = data + data_len;
3390 if (*p == '\r' || *p == '\n') {
3399 if (p >= q || *p != '\n')
3400 return AVERROR_INVALIDDATA;
3406 if (*p == '\r' || *p == '\n') {
3407 settings_len = p - settings;
3415 if (p >= q || *p != '\n')
3416 return AVERROR_INVALIDDATA;
3421 while (text_len > 0) {
3422 const int len = text_len - 1;
3423 const uint8_t c = p[len];
3424 if (c != '\r' && c != '\n')
3430 return AVERROR_INVALIDDATA;
3432 err = av_new_packet(pkt, text_len);
3437 memcpy(pkt->data, text, text_len);
3440 buf = av_packet_new_side_data(pkt,
3441 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3444 av_packet_unref(pkt);
3445 return AVERROR(ENOMEM);
3447 memcpy(buf, id, id_len);
3450 if (settings_len > 0) {
3451 buf = av_packet_new_side_data(pkt,
3452 AV_PKT_DATA_WEBVTT_SETTINGS,
3455 av_packet_unref(pkt);
3456 return AVERROR(ENOMEM);
3458 memcpy(buf, settings, settings_len);
3461 // Do we need this for subtitles?
3462 // pkt->flags = AV_PKT_FLAG_KEY;
3464 pkt->stream_index = st->index;
3465 pkt->pts = timecode;
3467 // Do we need this for subtitles?
3468 // pkt->dts = timecode;
3470 pkt->duration = duration;
3473 err = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3475 av_packet_unref(pkt);
3476 return AVERROR(ENOMEM);
3482 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3483 MatroskaTrack *track, AVStream *st,
3484 AVBufferRef *buf, uint8_t *data, int pkt_size,
3485 uint64_t timecode, uint64_t lace_duration,
3486 int64_t pos, int is_keyframe,
3487 uint8_t *additional, uint64_t additional_id, int additional_size,
3488 int64_t discard_padding)
3490 uint8_t *pkt_data = data;
3492 AVPacket *pkt = matroska->pkt;
3494 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3495 res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3497 av_log(matroska->ctx, AV_LOG_ERROR,
3498 "Error parsing a wavpack block.\n");
3506 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3507 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3508 res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3510 av_log(matroska->ctx, AV_LOG_ERROR,
3511 "Error parsing a prores block.\n");
3519 if (!pkt_size && !additional_size)
3523 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3526 pkt->buf = av_buffer_ref(buf);
3529 res = AVERROR(ENOMEM);
3533 pkt->data = pkt_data;
3534 pkt->size = pkt_size;
3535 pkt->flags = is_keyframe;
3536 pkt->stream_index = st->index;
3538 if (additional_size > 0) {
3539 uint8_t *side_data = av_packet_new_side_data(pkt,
3540 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3541 additional_size + 8);
3543 av_packet_unref(pkt);
3544 return AVERROR(ENOMEM);
3546 AV_WB64(side_data, additional_id);
3547 memcpy(side_data + 8, additional, additional_size);
3550 if (discard_padding) {
3551 uint8_t *side_data = av_packet_new_side_data(pkt,
3552 AV_PKT_DATA_SKIP_SAMPLES,
3555 av_packet_unref(pkt);
3556 return AVERROR(ENOMEM);
3558 discard_padding = av_rescale_q(discard_padding,
3559 (AVRational){1, 1000000000},
3560 (AVRational){1, st->codecpar->sample_rate});
3561 if (discard_padding > 0) {
3562 AV_WL32(side_data + 4, discard_padding);
3564 AV_WL32(side_data, -discard_padding);
3568 if (track->ms_compat)
3569 pkt->dts = timecode;
3571 pkt->pts = timecode;
3573 pkt->duration = lace_duration;
3575 #if FF_API_CONVERGENCE_DURATION
3576 FF_DISABLE_DEPRECATION_WARNINGS
3577 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3578 pkt->convergence_duration = lace_duration;
3580 FF_ENABLE_DEPRECATION_WARNINGS
3583 res = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3585 av_packet_unref(pkt);
3586 return AVERROR(ENOMEM);
3598 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3599 int size, int64_t pos, uint64_t cluster_time,
3600 uint64_t block_duration, int is_keyframe,
3601 uint8_t *additional, uint64_t additional_id, int additional_size,
3602 int64_t cluster_pos, int64_t discard_padding)
3604 uint64_t timecode = AV_NOPTS_VALUE;
3605 MatroskaTrack *track;
3610 uint32_t lace_size[256];
3611 int n, flags, laces = 0;
3613 int trust_default_duration;
3615 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3617 if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3622 track = matroska_find_track_by_num(matroska, num);
3623 if (!track || size < 3)
3624 return AVERROR_INVALIDDATA;
3626 if (!(st = track->stream)) {
3627 av_log(matroska->ctx, AV_LOG_VERBOSE,
3628 "No stream associated to TrackNumber %"PRIu64". "
3629 "Ignoring Block with this TrackNumber.\n", num);
3633 if (st->discard >= AVDISCARD_ALL)
3635 if (block_duration > INT64_MAX)
3636 block_duration = INT64_MAX;
3638 block_time = sign_extend(AV_RB16(data), 16);
3642 if (is_keyframe == -1)
3643 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3645 if (cluster_time != (uint64_t) -1 &&
3646 (block_time >= 0 || cluster_time >= -block_time)) {
3647 uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
3648 timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
3649 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3650 timecode < track->end_timecode)
3651 is_keyframe = 0; /* overlapping subtitles are not key frame */
3653 ff_reduce_index(matroska->ctx, st->index);
3654 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3659 if (matroska->skip_to_keyframe &&
3660 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3661 // Compare signed timecodes. Timecode may be negative due to codec delay
3662 // offset. We don't support timestamps greater than int64_t anyway - see
3664 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3667 matroska->skip_to_keyframe = 0;
3668 else if (!st->internal->skip_to_keyframe) {
3669 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3670 matroska->skip_to_keyframe = 0;
3674 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3675 &pb, lace_size, &laces);
3677 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3681 trust_default_duration = track->default_duration != 0;
3682 if (track->audio.samplerate == 8000 && trust_default_duration) {
3683 // If this is needed for more codecs, then add them here
3684 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3685 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3686 trust_default_duration = 0;
3690 if (!block_duration && trust_default_duration)
3691 block_duration = track->default_duration * laces / matroska->time_scale;
3693 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3694 track->end_timecode =
3695 FFMAX(track->end_timecode, timecode + block_duration);
3697 for (n = 0; n < laces; n++) {
3698 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3699 uint8_t *out_data = data;
3700 int out_size = lace_size[n];
3702 if (track->needs_decoding) {
3703 res = matroska_decode_buffer(&out_data, &out_size, track);
3706 /* Given that we are here means that out_data is no longer
3707 * owned by buf, so set it to NULL. This depends upon
3708 * zero-length header removal compression being ignored. */
3709 av_assert1(out_data != data);
3713 if (track->audio.buf) {
3714 res = matroska_parse_rm_audio(matroska, track, st,
3721 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3722 res = matroska_parse_webvtt(matroska, track, st,
3724 timecode, lace_duration,
3731 res = matroska_parse_frame(matroska, track, st, buf, out_data,
3732 out_size, timecode, lace_duration,
3733 pos, !n ? is_keyframe : 0,
3734 additional, additional_id, additional_size,
3740 if (timecode != AV_NOPTS_VALUE)
3741 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3742 data += lace_size[n];
3748 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3750 MatroskaCluster *cluster = &matroska->current_cluster;
3751 MatroskaBlock *block = &cluster->block;
3754 av_assert0(matroska->num_levels <= 2);
3756 if (matroska->num_levels == 1) {
3757 res = ebml_parse(matroska, matroska_segment, NULL);
3760 /* Found a cluster: subtract the size of the ID already read. */
3761 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3763 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3769 if (matroska->num_levels == 2) {
3770 /* We are inside a cluster. */
3771 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3773 if (res >= 0 && block->bin.size > 0) {
3774 int is_keyframe = block->non_simple ? block->reference.count == 0 : -1;
3775 uint8_t* additional = block->additional.size > 0 ?
3776 block->additional.data : NULL;
3778 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3779 block->bin.size, block->bin.pos,
3780 cluster->timecode, block->duration,
3781 is_keyframe, additional, block->additional_id,
3782 block->additional.size, cluster->pos,
3783 block->discard_padding);
3786 ebml_free(matroska_blockgroup, block);
3787 memset(block, 0, sizeof(*block));
3788 } else if (!matroska->num_levels) {
3789 if (!avio_feof(matroska->ctx->pb)) {
3790 avio_r8(matroska->ctx->pb);
3791 if (!avio_feof(matroska->ctx->pb)) {
3792 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3793 "end of segment.\n");
3794 return AVERROR_INVALIDDATA;
3804 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3806 MatroskaDemuxContext *matroska = s->priv_data;
3809 if (matroska->resync_pos == -1) {
3810 // This can only happen if generic seeking has been used.
3811 matroska->resync_pos = avio_tell(s->pb);
3814 while (matroska_deliver_packet(matroska, pkt)) {
3816 return (ret < 0) ? ret : AVERROR_EOF;
3817 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3818 ret = matroska_resync(matroska, matroska->resync_pos);
3824 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3825 int64_t timestamp, int flags)
3827 MatroskaDemuxContext *matroska = s->priv_data;
3828 MatroskaTrack *tracks = NULL;
3829 AVStream *st = s->streams[stream_index];
3832 /* Parse the CUES now since we need the index data to seek. */
3833 if (matroska->cues_parsing_deferred > 0) {
3834 matroska->cues_parsing_deferred = 0;
3835 matroska_parse_cues(matroska);
3838 if (!st->internal->nb_index_entries)
3840 timestamp = FFMAX(timestamp, st->internal->index_entries[0].timestamp);
3842 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3843 matroska_reset_status(matroska, 0, st->internal->index_entries[st->internal->nb_index_entries - 1].pos);
3844 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3845 matroska_clear_queue(matroska);
3846 if (matroska_parse_cluster(matroska) < 0)
3851 matroska_clear_queue(matroska);
3852 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->internal->nb_index_entries - 1))
3855 tracks = matroska->tracks.elem;
3856 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3857 tracks[i].audio.pkt_cnt = 0;
3858 tracks[i].audio.sub_packet_cnt = 0;
3859 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3860 tracks[i].end_timecode = 0;
3863 /* We seek to a level 1 element, so set the appropriate status. */
3864 matroska_reset_status(matroska, 0, st->internal->index_entries[index].pos);
3865 if (flags & AVSEEK_FLAG_ANY) {
3866 st->internal->skip_to_keyframe = 0;
3867 matroska->skip_to_timecode = timestamp;
3869 st->internal->skip_to_keyframe = 1;
3870 matroska->skip_to_timecode = st->internal->index_entries[index].timestamp;
3872 matroska->skip_to_keyframe = 1;
3874 ff_update_cur_dts(s, st, st->internal->index_entries[index].timestamp);
3877 // slightly hackish but allows proper fallback to
3878 // the generic seeking code.
3879 matroska_reset_status(matroska, 0, -1);
3880 matroska->resync_pos = -1;
3881 matroska_clear_queue(matroska);
3882 st->internal->skip_to_keyframe =
3883 matroska->skip_to_keyframe = 0;
3888 static int matroska_read_close(AVFormatContext *s)
3890 MatroskaDemuxContext *matroska = s->priv_data;
3891 MatroskaTrack *tracks = matroska->tracks.elem;
3894 matroska_clear_queue(matroska);
3896 for (n = 0; n < matroska->tracks.nb_elem; n++)
3897 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3898 av_freep(&tracks[n].audio.buf);
3899 ebml_free(matroska_segment, matroska);
3905 int64_t start_time_ns;
3906 int64_t end_time_ns;
3907 int64_t start_offset;
3911 /* This function searches all the Cues and returns the CueDesc corresponding to
3912 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3913 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3915 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3916 MatroskaDemuxContext *matroska = s->priv_data;
3919 int nb_index_entries = s->streams[0]->internal->nb_index_entries;
3920 AVIndexEntry *index_entries = s->streams[0]->internal->index_entries;
3921 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3922 for (i = 1; i < nb_index_entries; i++) {
3923 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3924 index_entries[i].timestamp * matroska->time_scale > ts) {
3929 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3930 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3931 if (i != nb_index_entries - 1) {
3932 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3933 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3935 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3936 // FIXME: this needs special handling for files where Cues appear
3937 // before Clusters. the current logic assumes Cues appear after
3939 cue_desc.end_offset = cues_start - matroska->segment_start;
3944 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3946 MatroskaDemuxContext *matroska = s->priv_data;
3947 uint32_t id = matroska->current_id;
3948 int64_t cluster_pos, before_pos;
3950 if (s->streams[0]->internal->nb_index_entries <= 0) return 0;
3951 // seek to the first cluster using cues.
3952 index = av_index_search_timestamp(s->streams[0], 0, 0);
3953 if (index < 0) return 0;
3954 cluster_pos = s->streams[0]->internal->index_entries[index].pos;
3955 before_pos = avio_tell(s->pb);
3957 uint64_t cluster_id, cluster_length;
3960 avio_seek(s->pb, cluster_pos, SEEK_SET);
3961 // read cluster id and length
3962 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3963 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3965 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3969 matroska_reset_status(matroska, 0, cluster_pos);
3970 matroska_clear_queue(matroska);
3971 if (matroska_parse_cluster(matroska) < 0 ||
3975 pkt = &matroska->queue->pkt;
3976 // 4 + read is the length of the cluster id and the cluster length field.
3977 cluster_pos += 4 + read + cluster_length;
3978 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3984 /* Restore the status after matroska_read_header: */
3985 matroska_reset_status(matroska, id, before_pos);
3990 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3991 double min_buffer, double* buffer,
3992 double* sec_to_download, AVFormatContext *s,
3995 double nano_seconds_per_second = 1000000000.0;
3996 double time_sec = time_ns / nano_seconds_per_second;
3998 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3999 int64_t end_time_ns = time_ns + time_to_search_ns;
4000 double sec_downloaded = 0.0;
4001 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
4002 if (desc_curr.start_time_ns == -1)
4004 *sec_to_download = 0.0;
4006 // Check for non cue start time.
4007 if (time_ns > desc_curr.start_time_ns) {
4008 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
4009 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
4010 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
4011 double timeToDownload = (cueBytes * 8.0) / bps;
4013 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
4014 *sec_to_download += timeToDownload;
4016 // Check if the search ends within the first cue.
4017 if (desc_curr.end_time_ns >= end_time_ns) {
4018 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4019 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4020 sec_downloaded = percent_to_sub * sec_downloaded;
4021 *sec_to_download = percent_to_sub * *sec_to_download;
4024 if ((sec_downloaded + *buffer) <= min_buffer) {
4028 // Get the next Cue.
4029 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4032 while (desc_curr.start_time_ns != -1) {
4033 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
4034 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
4035 double desc_sec = desc_ns / nano_seconds_per_second;
4036 double bits = (desc_bytes * 8.0);
4037 double time_to_download = bits / bps;
4039 sec_downloaded += desc_sec - time_to_download;
4040 *sec_to_download += time_to_download;
4042 if (desc_curr.end_time_ns >= end_time_ns) {
4043 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4044 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4045 sec_downloaded = percent_to_sub * sec_downloaded;
4046 *sec_to_download = percent_to_sub * *sec_to_download;
4048 if ((sec_downloaded + *buffer) <= min_buffer)
4053 if ((sec_downloaded + *buffer) <= min_buffer) {
4058 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4060 *buffer = *buffer + sec_downloaded;
4064 /* This function computes the bandwidth of the WebM file with the help of
4065 * buffer_size_after_time_downloaded() function. Both of these functions are
4066 * adapted from WebM Tools project and are adapted to work with FFmpeg's
4067 * Matroska parsing mechanism.
4069 * Returns the bandwidth of the file on success; -1 on error.
4071 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4073 MatroskaDemuxContext *matroska = s->priv_data;
4074 AVStream *st = s->streams[0];
4075 double bandwidth = 0.0;
4078 for (i = 0; i < st->internal->nb_index_entries; i++) {
4079 int64_t prebuffer_ns = 1000000000;
4080 int64_t time_ns = st->internal->index_entries[i].timestamp * matroska->time_scale;
4081 double nano_seconds_per_second = 1000000000.0;
4082 int64_t prebuffered_ns = time_ns + prebuffer_ns;
4083 double prebuffer_bytes = 0.0;
4084 int64_t temp_prebuffer_ns = prebuffer_ns;
4085 int64_t pre_bytes, pre_ns;
4086 double pre_sec, prebuffer, bits_per_second;
4087 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4089 // Start with the first Cue.
4090 CueDesc desc_end = desc_beg;
4092 // Figure out how much data we have downloaded for the prebuffer. This will
4093 // be used later to adjust the bits per sample to try.
4094 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4095 // Prebuffered the entire Cue.
4096 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4097 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4098 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4100 if (desc_end.start_time_ns == -1) {
4101 // The prebuffer is larger than the duration.
4102 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4104 bits_per_second = 0.0;
4106 // The prebuffer ends in the last Cue. Estimate how much data was
4108 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4109 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4110 pre_sec = pre_ns / nano_seconds_per_second;
4112 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4114 prebuffer = prebuffer_ns / nano_seconds_per_second;
4116 // Set this to 0.0 in case our prebuffer buffers the entire video.
4117 bits_per_second = 0.0;
4119 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4120 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4121 double desc_sec = desc_ns / nano_seconds_per_second;
4122 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4124 // Drop the bps by the percentage of bytes buffered.
4125 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4126 double mod_bits_per_second = calc_bits_per_second * percent;
4128 if (prebuffer < desc_sec) {
4130 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4132 // Add 1 so the bits per second should be a little bit greater than file
4134 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4135 const double min_buffer = 0.0;
4136 double buffer = prebuffer;
4137 double sec_to_download = 0.0;
4139 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4140 min_buffer, &buffer, &sec_to_download,
4144 } else if (rv == 0) {
4145 bits_per_second = (double)(bps);
4150 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4151 } while (desc_end.start_time_ns != -1);
4153 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4155 return (int64_t)bandwidth;
4158 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4160 MatroskaDemuxContext *matroska = s->priv_data;
4161 EbmlList *seekhead_list = &matroska->seekhead;
4162 MatroskaSeekhead *seekhead = seekhead_list->elem;
4164 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4168 // determine cues start and end positions
4169 for (i = 0; i < seekhead_list->nb_elem; i++)
4170 if (seekhead[i].id == MATROSKA_ID_CUES)
4173 if (i >= seekhead_list->nb_elem) return -1;
4175 before_pos = avio_tell(matroska->ctx->pb);
4176 cues_start = seekhead[i].pos + matroska->segment_start;
4177 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4178 // cues_end is computed as cues_start + cues_length + length of the
4179 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4180 // cues_end is inclusive and the above sum is reduced by 1.
4181 uint64_t cues_length, cues_id;
4183 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4184 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4185 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4186 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4189 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4191 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4192 if (cues_start == -1 || cues_end == -1) return -1;
4195 matroska_parse_cues(matroska);
4198 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4201 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4203 // if the file has cues at the start, fix up the init range so that
4204 // it does not include it
4205 if (cues_start <= init_range)
4206 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4209 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4210 if (bandwidth < 0) return -1;
4211 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4213 // check if all clusters start with key frames
4214 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4216 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4217 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4218 buf = av_malloc_array(s->streams[0]->internal->nb_index_entries, 20);
4219 if (!buf) return -1;
4221 for (i = 0; i < s->streams[0]->internal->nb_index_entries; i++) {
4222 int ret = snprintf(buf + end, 20,
4223 "%" PRId64"%s", s->streams[0]->internal->index_entries[i].timestamp,
4224 i != s->streams[0]->internal->nb_index_entries - 1 ? "," : "");
4225 if (ret <= 0 || (ret == 20 && i == s->streams[0]->internal->nb_index_entries - 1)) {
4226 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4228 return AVERROR_INVALIDDATA;
4232 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4233 buf, AV_DICT_DONT_STRDUP_VAL);
4238 static int webm_dash_manifest_read_header(AVFormatContext *s)
4241 int ret = matroska_read_header(s);
4243 MatroskaTrack *tracks;
4244 MatroskaDemuxContext *matroska = s->priv_data;
4246 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4249 if (!matroska->tracks.nb_elem || !s->nb_streams) {
4250 av_log(s, AV_LOG_ERROR, "No track found\n");
4251 ret = AVERROR_INVALIDDATA;
4255 if (!matroska->is_live) {
4256 buf = av_asprintf("%g", matroska->duration);
4258 ret = AVERROR(ENOMEM);
4261 av_dict_set(&s->streams[0]->metadata, DURATION,
4262 buf, AV_DICT_DONT_STRDUP_VAL);
4264 // initialization range
4265 // 5 is the offset of Cluster ID.
4266 init_range = avio_tell(s->pb) - 5;
4267 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4270 // basename of the file
4271 buf = strrchr(s->url, '/');
4272 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4275 tracks = matroska->tracks.elem;
4276 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4278 // parse the cues and populate Cue related fields
4279 if (!matroska->is_live) {
4280 ret = webm_dash_manifest_cues(s, init_range);
4282 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4287 // use the bandwidth from the command line if it was provided
4288 if (matroska->bandwidth > 0) {
4289 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4290 matroska->bandwidth, 0);
4294 matroska_read_close(s);
4298 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4303 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4304 static const AVOption options[] = {
4305 { "live", "flag indicating that the input is a live file that only has the headers.", OFFSET(is_live), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_DECODING_PARAM },
4306 { "bandwidth", "bandwidth of this stream to be specified in the DASH manifest.", OFFSET(bandwidth), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, AV_OPT_FLAG_DECODING_PARAM },
4310 static const AVClass webm_dash_class = {
4311 .class_name = "WebM DASH Manifest demuxer",
4312 .item_name = av_default_item_name,
4314 .version = LIBAVUTIL_VERSION_INT,
4317 AVInputFormat ff_matroska_demuxer = {
4318 .name = "matroska,webm",
4319 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4320 .extensions = "mkv,mk3d,mka,mks,webm",
4321 .priv_data_size = sizeof(MatroskaDemuxContext),
4322 .read_probe = matroska_probe,
4323 .read_header = matroska_read_header,
4324 .read_packet = matroska_read_packet,
4325 .read_close = matroska_read_close,
4326 .read_seek = matroska_read_seek,
4327 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4330 AVInputFormat ff_webm_dash_manifest_demuxer = {
4331 .name = "webm_dash_manifest",
4332 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4333 .priv_data_size = sizeof(MatroskaDemuxContext),
4334 .read_header = webm_dash_manifest_read_header,
4335 .read_packet = webm_dash_manifest_read_packet,
4336 .read_close = matroska_read_close,
4337 .priv_class = &webm_dash_class,