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;
386 /* the packet queue */
388 PacketList *queue_end;
392 /* What to skip before effectively reading a packet. */
393 int skip_to_keyframe;
394 uint64_t skip_to_timecode;
396 /* File has a CUES element, but we defer parsing until it is needed. */
397 int cues_parsing_deferred;
399 /* Level1 elements and whether they were read yet */
400 MatroskaLevel1Element level1_elems[64];
401 int num_level1_elems;
403 MatroskaCluster current_cluster;
405 /* WebM DASH Manifest live flag */
408 /* Bandwidth value for WebM DASH Manifest */
410 } MatroskaDemuxContext;
412 #define CHILD_OF(parent) { .def = { .n = parent } }
414 // The following forward declarations need their size because
415 // a tentative definition with internal linkage must not be an
416 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
417 // Removing the sizes breaks MSVC.
418 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
419 matroska_track[32], matroska_track_encoding[6], matroska_track_encodings[2],
420 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_tracks[2],
421 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
422 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
423 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
425 static EbmlSyntax ebml_header[] = {
426 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
427 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, 0, offsetof(Ebml, max_size), { .u = 8 } },
428 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, 0, offsetof(Ebml, id_length), { .u = 4 } },
429 { EBML_ID_DOCTYPE, EBML_STR, 0, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
430 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
431 { EBML_ID_EBMLVERSION, EBML_NONE },
432 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
433 CHILD_OF(ebml_syntax)
436 static EbmlSyntax ebml_syntax[] = {
437 { EBML_ID_HEADER, EBML_NEST, 0, 0, 0, { .n = ebml_header } },
438 { MATROSKA_ID_SEGMENT, EBML_STOP },
442 static EbmlSyntax matroska_info[] = {
443 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
444 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, 0, offsetof(MatroskaDemuxContext, duration) },
445 { MATROSKA_ID_TITLE, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, title) },
446 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
447 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, muxingapp) },
448 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, 0, offsetof(MatroskaDemuxContext, date_utc) },
449 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
450 CHILD_OF(matroska_segment)
453 static EbmlSyntax matroska_mastering_meta[] = {
454 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_x) },
455 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_y) },
456 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_x) },
457 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_y) },
458 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_x) },
459 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_y) },
460 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_x) },
461 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_y) },
462 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 1, 0, offsetof(MatroskaMasteringMeta, min_luminance) },
463 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, max_luminance) },
464 CHILD_OF(matroska_track_video_color)
467 static EbmlSyntax matroska_track_video_color[] = {
468 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
469 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u = 0 } },
470 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz) },
471 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert) },
472 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz) },
473 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert) },
474 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
475 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
476 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
477 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
478 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
479 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_cll) },
480 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_fall) },
481 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
482 CHILD_OF(matroska_track_video)
485 static EbmlSyntax matroska_track_video_projection[] = {
486 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
487 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideoProjection, private) },
488 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f = 0.0 } },
489 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f = 0.0 } },
490 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f = 0.0 } },
491 CHILD_OF(matroska_track_video)
494 static EbmlSyntax matroska_track_video[] = {
495 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideo, frame_rate) },
496 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
497 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
498 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_width) },
499 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_height) },
500 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideo, color_space) },
501 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, alpha_mode), { .u = 0 } },
502 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, 0, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
503 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
504 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
505 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
506 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
507 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
508 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
509 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
510 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
511 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
512 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
513 CHILD_OF(matroska_track)
516 static EbmlSyntax matroska_track_audio[] = {
517 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
518 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
519 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, bitdepth) },
520 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
521 CHILD_OF(matroska_track)
524 static EbmlSyntax matroska_track_encoding_compression[] = {
525 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackCompression, algo), { .u = MATROSKA_TRACK_ENCODING_COMP_ZLIB } },
526 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, 0, offsetof(MatroskaTrackCompression, settings) },
527 CHILD_OF(matroska_track_encoding)
530 static EbmlSyntax matroska_track_encoding_encryption[] = {
531 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
532 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, 0, offsetof(MatroskaTrackEncryption,key_id) },
533 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
534 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
535 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
536 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
537 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
538 CHILD_OF(matroska_track_encoding)
540 static EbmlSyntax matroska_track_encoding[] = {
541 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
542 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
543 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
544 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
545 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
546 CHILD_OF(matroska_track_encodings)
549 static EbmlSyntax matroska_track_encodings[] = {
550 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, 0, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
551 CHILD_OF(matroska_track)
554 static EbmlSyntax matroska_track_plane[] = {
555 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,uid) },
556 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,type) },
557 CHILD_OF(matroska_track_combine_planes)
560 static EbmlSyntax matroska_track_combine_planes[] = {
561 { MATROSKA_ID_TRACKPLANE, EBML_NEST, 0, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
562 CHILD_OF(matroska_track_operation)
565 static EbmlSyntax matroska_track_operation[] = {
566 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, 0, {.n = matroska_track_combine_planes} },
567 CHILD_OF(matroska_track)
570 static EbmlSyntax matroska_track[] = {
571 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, 0, offsetof(MatroskaTrack, num) },
572 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTrack, name) },
573 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, uid) },
574 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrack, type) },
575 { MATROSKA_ID_CODECID, EBML_STR, 0, 0, offsetof(MatroskaTrack, codec_id) },
576 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrack, codec_priv) },
577 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, codec_delay), { .u = 0 } },
578 { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
579 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, 0, offsetof(MatroskaTrack, default_duration) },
580 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
581 { MATROSKA_ID_TRACKFLAGCOMMENTARY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_comment), { .u = 0 } },
582 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
583 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
584 { MATROSKA_ID_TRACKFLAGHEARINGIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_hearingimpaired), { .u = 0 } },
585 { MATROSKA_ID_TRACKFLAGVISUALIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_visualimpaired), { .u = 0 } },
586 { MATROSKA_ID_TRACKFLAGTEXTDESCRIPTIONS, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_textdescriptions), { .u = 0 } },
587 { MATROSKA_ID_TRACKFLAGORIGINAL, EBML_UINT, 1, 0, offsetof(MatroskaTrack, flag_original), {.u = 0 } },
588 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
589 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
590 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
591 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, 0, { .n = matroska_track_encodings } },
592 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, max_block_additional_id), { .u = 0 } },
593 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, 0, offsetof(MatroskaTrack, seek_preroll), { .u = 0 } },
594 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
595 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
596 { MATROSKA_ID_CODECNAME, EBML_NONE },
597 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
598 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
599 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
600 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
601 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
602 CHILD_OF(matroska_tracks)
605 static EbmlSyntax matroska_tracks[] = {
606 { MATROSKA_ID_TRACKENTRY, EBML_NEST, 0, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
607 CHILD_OF(matroska_segment)
610 static EbmlSyntax matroska_attachment[] = {
611 { MATROSKA_ID_FILEUID, EBML_UINT, 0, 0, offsetof(MatroskaAttachment, uid) },
612 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, filename) },
613 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, 0, offsetof(MatroskaAttachment, mime) },
614 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, 0, offsetof(MatroskaAttachment, bin) },
615 { MATROSKA_ID_FILEDESC, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, description) },
616 CHILD_OF(matroska_attachments)
619 static EbmlSyntax matroska_attachments[] = {
620 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, 0, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
621 CHILD_OF(matroska_segment)
624 static EbmlSyntax matroska_chapter_display[] = {
625 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaChapter, title) },
626 { MATROSKA_ID_CHAPLANG, EBML_NONE },
627 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
628 CHILD_OF(matroska_chapter_entry)
631 static EbmlSyntax matroska_chapter_entry[] = {
632 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
633 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
634 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaChapter, uid) },
635 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter_display } },
636 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
637 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
638 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
639 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
640 CHILD_OF(matroska_chapter)
643 static EbmlSyntax matroska_chapter[] = {
644 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, 0, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
645 { MATROSKA_ID_EDITIONUID, EBML_NONE },
646 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
647 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
648 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
649 CHILD_OF(matroska_chapters)
652 static EbmlSyntax matroska_chapters[] = {
653 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter } },
654 CHILD_OF(matroska_segment)
657 static EbmlSyntax matroska_index_pos[] = {
658 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, track) },
659 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, pos) },
660 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
661 { MATROSKA_ID_CUEDURATION, EBML_NONE },
662 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
663 CHILD_OF(matroska_index_entry)
666 static EbmlSyntax matroska_index_entry[] = {
667 { MATROSKA_ID_CUETIME, EBML_UINT, 0, 0, offsetof(MatroskaIndex, time) },
668 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, 0, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
669 CHILD_OF(matroska_index)
672 static EbmlSyntax matroska_index[] = {
673 { MATROSKA_ID_POINTENTRY, EBML_NEST, 0, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
674 CHILD_OF(matroska_segment)
677 static EbmlSyntax matroska_simpletag[] = {
678 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTag, name) },
679 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaTag, string) },
680 { MATROSKA_ID_TAGLANG, EBML_STR, 0, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
681 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
682 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
683 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
684 CHILD_OF(matroska_tag)
687 static EbmlSyntax matroska_tagtargets[] = {
688 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, 0, offsetof(MatroskaTagTarget, type) },
689 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
690 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, trackuid), { .u = 0 } },
691 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, chapteruid), { .u = 0 } },
692 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, attachuid), { .u = 0 } },
693 CHILD_OF(matroska_tag)
696 static EbmlSyntax matroska_tag[] = {
697 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
698 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
699 CHILD_OF(matroska_tags)
702 static EbmlSyntax matroska_tags[] = {
703 { MATROSKA_ID_TAG, EBML_NEST, 0, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
704 CHILD_OF(matroska_segment)
707 static EbmlSyntax matroska_seekhead_entry[] = {
708 { MATROSKA_ID_SEEKID, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, id) },
709 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
710 CHILD_OF(matroska_seekhead)
713 static EbmlSyntax matroska_seekhead[] = {
714 { MATROSKA_ID_SEEKENTRY, EBML_NEST, 0, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
715 CHILD_OF(matroska_segment)
718 static EbmlSyntax matroska_segment[] = {
719 { MATROSKA_ID_CLUSTER, EBML_STOP },
720 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, 0, { .n = matroska_info } },
721 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tracks } },
722 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_attachments } },
723 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_chapters } },
724 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, 0, { .n = matroska_index } },
725 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tags } },
726 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, 0, { .n = matroska_seekhead } },
727 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
730 static EbmlSyntax matroska_segments[] = {
731 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, 0, { .n = matroska_segment } },
735 static EbmlSyntax matroska_blockmore[] = {
736 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
737 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, 0, offsetof(MatroskaBlock,additional) },
738 CHILD_OF(matroska_blockadditions)
741 static EbmlSyntax matroska_blockadditions[] = {
742 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, 0, {.n = matroska_blockmore} },
743 CHILD_OF(matroska_blockgroup)
746 static EbmlSyntax matroska_blockgroup[] = {
747 { MATROSKA_ID_BLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
748 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, 0, { .n = matroska_blockadditions} },
749 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, 0, offsetof(MatroskaBlock, duration) },
750 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, 0, offsetof(MatroskaBlock, discard_padding) },
751 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 1, 0, offsetof(MatroskaBlock, reference) },
752 { MATROSKA_ID_CODECSTATE, EBML_NONE },
753 { 1, EBML_UINT, 0, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
754 CHILD_OF(matroska_cluster_parsing)
757 // The following array contains SimpleBlock and BlockGroup twice
758 // in order to reuse the other values for matroska_cluster_enter.
759 static EbmlSyntax matroska_cluster_parsing[] = {
760 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
761 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, 0, { .n = matroska_blockgroup } },
762 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, 0, offsetof(MatroskaCluster, timecode) },
763 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
764 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
765 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
766 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
767 CHILD_OF(matroska_segment)
770 static EbmlSyntax matroska_cluster_enter[] = {
771 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, 0, { .n = &matroska_cluster_parsing[2] } },
776 static const CodecMime mkv_image_mime_tags[] = {
777 {"image/gif" , AV_CODEC_ID_GIF},
778 {"image/jpeg" , AV_CODEC_ID_MJPEG},
779 {"image/png" , AV_CODEC_ID_PNG},
780 {"image/tiff" , AV_CODEC_ID_TIFF},
782 {"" , AV_CODEC_ID_NONE}
785 static const CodecMime mkv_mime_tags[] = {
786 {"text/plain" , AV_CODEC_ID_TEXT},
787 {"application/x-truetype-font", AV_CODEC_ID_TTF},
788 {"application/x-font" , AV_CODEC_ID_TTF},
789 {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
790 {"binary" , AV_CODEC_ID_BIN_DATA},
792 {"" , AV_CODEC_ID_NONE}
795 static const char *const matroska_doctypes[] = { "matroska", "webm" };
797 static int matroska_read_close(AVFormatContext *s);
800 * This function prepares the status for parsing of level 1 elements.
802 static int matroska_reset_status(MatroskaDemuxContext *matroska,
803 uint32_t id, int64_t position)
806 int64_t err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
811 matroska->current_id = id;
812 matroska->num_levels = 1;
813 matroska->unknown_count = 0;
814 matroska->resync_pos = avio_tell(matroska->ctx->pb);
816 matroska->resync_pos -= (av_log2(id) + 7) / 8;
821 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
823 AVIOContext *pb = matroska->ctx->pb;
826 /* Try to seek to the last position to resync from. If this doesn't work,
827 * we resync from the earliest position available: The start of the buffer. */
828 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
829 av_log(matroska->ctx, AV_LOG_WARNING,
830 "Seek to desired resync point failed. Seeking to "
831 "earliest point available instead.\n");
832 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
833 last_pos + 1), SEEK_SET);
838 // try to find a toplevel element
839 while (!avio_feof(pb)) {
840 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
841 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
842 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
843 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
844 /* Prepare the context for parsing of a level 1 element. */
845 matroska_reset_status(matroska, id, -1);
846 /* Given that we are here means that an error has occurred,
847 * so treat the segment as unknown length in order not to
848 * discard valid data that happens to be beyond the designated
849 * end of the segment. */
850 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
853 id = (id << 8) | avio_r8(pb);
857 return pb->error ? pb->error : AVERROR_EOF;
861 * Read: an "EBML number", which is defined as a variable-length
862 * array of bytes. The first byte indicates the length by giving a
863 * number of 0-bits followed by a one. The position of the first
864 * "one" bit inside the first byte indicates the length of this
866 * Returns: number of bytes read, < 0 on error
868 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
869 int max_size, uint64_t *number, int eof_forbidden)
875 /* The first byte tells us the length in bytes - except when it is zero. */
880 /* get the length of the EBML number */
881 read = 8 - ff_log2_tab[total];
883 if (!total || read > max_size) {
884 pos = avio_tell(pb) - 1;
886 av_log(matroska->ctx, AV_LOG_ERROR,
887 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
888 "of an EBML number\n", pos, pos);
890 av_log(matroska->ctx, AV_LOG_ERROR,
891 "Length %d indicated by an EBML number's first byte 0x%02x "
892 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
893 read, (uint8_t) total, pos, pos, max_size);
895 return AVERROR_INVALIDDATA;
898 /* read out length */
899 total ^= 1 << ff_log2_tab[total];
901 total = (total << 8) | avio_r8(pb);
903 if (pb->eof_reached) {
915 av_log(matroska->ctx, AV_LOG_ERROR,
916 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
921 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
922 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
929 * Read a EBML length value.
930 * This needs special handling for the "unknown length" case which has multiple
933 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
936 int res = ebml_read_num(matroska, pb, 8, number, 1);
937 if (res > 0 && *number + 1 == 1ULL << (7 * res))
938 *number = EBML_UNKNOWN_LENGTH;
943 * Read the next element as an unsigned int.
944 * Returns NEEDS_CHECKING unless size == 0.
946 static int ebml_read_uint(AVIOContext *pb, int size,
947 uint64_t default_value, uint64_t *num)
952 *num = default_value;
955 /* big-endian ordering; build up number */
958 *num = (*num << 8) | avio_r8(pb);
960 return NEEDS_CHECKING;
964 * Read the next element as a signed int.
965 * Returns NEEDS_CHECKING unless size == 0.
967 static int ebml_read_sint(AVIOContext *pb, int size,
968 int64_t default_value, int64_t *num)
973 *num = default_value;
976 *num = sign_extend(avio_r8(pb), 8);
978 /* big-endian ordering; build up number */
980 *num = ((uint64_t)*num << 8) | avio_r8(pb);
983 return NEEDS_CHECKING;
987 * Read the next element as a float.
988 * Returns 0 if size == 0, NEEDS_CHECKING or < 0 on obvious failure.
990 static int ebml_read_float(AVIOContext *pb, int size,
991 double default_value, double *num)
994 *num = default_value;
996 } else if (size == 4) {
997 *num = av_int2float(avio_rb32(pb));
998 } else if (size == 8) {
999 *num = av_int2double(avio_rb64(pb));
1001 return AVERROR_INVALIDDATA;
1003 return NEEDS_CHECKING;
1007 * Read the next element as an ASCII string.
1008 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1010 static int ebml_read_ascii(AVIOContext *pb, int size,
1011 const char *default_value, char **str)
1016 if (size == 0 && default_value) {
1017 res = av_strdup(default_value);
1019 return AVERROR(ENOMEM);
1021 /* EBML strings are usually not 0-terminated, so we allocate one
1022 * byte more, read the string and NUL-terminate it ourselves. */
1023 if (!(res = av_malloc(size + 1)))
1024 return AVERROR(ENOMEM);
1025 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
1027 return ret < 0 ? ret : NEEDS_CHECKING;
1038 * Read the next element as binary data.
1039 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1041 static int ebml_read_binary(AVIOContext *pb, int length,
1042 int64_t pos, EbmlBin *bin)
1046 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1049 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1051 bin->data = bin->buf->data;
1054 if ((ret = avio_read(pb, bin->data, length)) != length) {
1055 av_buffer_unref(&bin->buf);
1058 return ret < 0 ? ret : NEEDS_CHECKING;
1065 * Read the next element, but only the header. The contents
1066 * are supposed to be sub-elements which can be read separately.
1067 * 0 is success, < 0 is failure.
1069 static int ebml_read_master(MatroskaDemuxContext *matroska,
1070 uint64_t length, int64_t pos)
1072 MatroskaLevel *level;
1074 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1075 av_log(matroska->ctx, AV_LOG_ERROR,
1076 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1077 return AVERROR(ENOSYS);
1080 level = &matroska->levels[matroska->num_levels++];
1082 level->length = length;
1088 * Read a signed "EBML number"
1089 * Return: number of bytes processed, < 0 on error
1091 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1092 AVIOContext *pb, int64_t *num)
1097 /* read as unsigned number first */
1098 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1101 /* make signed (weird way) */
1102 *num = unum - ((1LL << (7 * res - 1)) - 1);
1107 static int ebml_parse(MatroskaDemuxContext *matroska,
1108 EbmlSyntax *syntax, void *data);
1110 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1114 // Whoever touches this should be aware of the duplication
1115 // existing in matroska_cluster_parsing.
1116 for (i = 0; syntax[i].id; i++)
1117 if (id == syntax[i].id)
1123 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1129 for (int i = 0; syntax[i].id; i++) {
1130 void *dst = (char *)data + syntax[i].data_offset;
1131 switch (syntax[i].type) {
1133 *(uint64_t *)dst = syntax[i].def.u;
1136 *(int64_t *) dst = syntax[i].def.i;
1139 *(double *) dst = syntax[i].def.f;
1143 // the default may be NULL
1144 if (syntax[i].def.s) {
1145 *(char**)dst = av_strdup(syntax[i].def.s);
1147 return AVERROR(ENOMEM);
1153 if (!matroska->levels[matroska->num_levels - 1].length) {
1154 matroska->num_levels--;
1160 res = ebml_parse(matroska, syntax, data);
1163 return res == LEVEL_ENDED ? 0 : res;
1166 static int is_ebml_id_valid(uint32_t id)
1168 // Due to endian nonsense in Matroska, the highest byte with any bits set
1169 // will contain the leading length bit. This bit in turn identifies the
1170 // total byte length of the element by its position within the byte.
1171 unsigned int bits = av_log2(id);
1172 return id && (bits + 7) / 8 == (8 - bits % 8);
1176 * Allocate and return the entry for the level1 element with the given ID. If
1177 * an entry already exists, return the existing entry.
1179 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1180 uint32_t id, int64_t pos)
1183 MatroskaLevel1Element *elem;
1185 if (!is_ebml_id_valid(id))
1188 // Some files link to all clusters; useless.
1189 if (id == MATROSKA_ID_CLUSTER)
1192 // There can be multiple SeekHeads and Tags.
1193 for (i = 0; i < matroska->num_level1_elems; i++) {
1194 if (matroska->level1_elems[i].id == id) {
1195 if (matroska->level1_elems[i].pos == pos ||
1196 id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1197 return &matroska->level1_elems[i];
1201 // Only a completely broken file would have more elements.
1202 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1203 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1207 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1208 *elem = (MatroskaLevel1Element){.id = id};
1213 static int ebml_parse(MatroskaDemuxContext *matroska,
1214 EbmlSyntax *syntax, void *data)
1216 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1217 // Forbid unknown-length EBML_NONE elements.
1218 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1222 // max. 16 MB for strings
1223 [EBML_STR] = 0x1000000,
1224 [EBML_UTF8] = 0x1000000,
1225 // max. 256 MB for binary data
1226 [EBML_BIN] = 0x10000000,
1227 // no limits for anything else
1229 AVIOContext *pb = matroska->ctx->pb;
1232 int64_t pos = avio_tell(pb), pos_alt;
1233 int res, update_pos = 1, level_check;
1234 MatroskaLevel1Element *level1_elem;
1235 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1237 if (!matroska->current_id) {
1239 res = ebml_read_num(matroska, pb, 4, &id, 0);
1241 if (pb->eof_reached && res == AVERROR_EOF) {
1242 if (matroska->is_live)
1243 // in live mode, finish parsing if EOF is reached.
1245 if (level && pos == avio_tell(pb)) {
1246 if (level->length == EBML_UNKNOWN_LENGTH) {
1247 // Unknown-length levels automatically end at EOF.
1248 matroska->num_levels--;
1251 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1252 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1258 matroska->current_id = id | 1 << 7 * res;
1259 pos_alt = pos + res;
1262 pos -= (av_log2(matroska->current_id) + 7) / 8;
1265 id = matroska->current_id;
1267 syntax = ebml_parse_id(syntax, id);
1268 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1269 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1270 // Unknown-length levels end when an element from an upper level
1271 // in the hierarchy is encountered.
1272 while (syntax->def.n) {
1273 syntax = ebml_parse_id(syntax->def.n, id);
1275 matroska->num_levels--;
1281 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1282 "%"PRId64"\n", id, pos);
1283 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1287 data = (char *) data + syntax->data_offset;
1288 if (syntax->list_elem_size) {
1289 EbmlList *list = data;
1292 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1293 return AVERROR(ENOMEM);
1294 newelem = av_fast_realloc(list->elem,
1295 &list->alloc_elem_size,
1296 (list->nb_elem + 1) * syntax->list_elem_size);
1298 return AVERROR(ENOMEM);
1299 list->elem = newelem;
1300 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1301 memset(data, 0, syntax->list_elem_size);
1306 if (syntax->type != EBML_STOP) {
1307 matroska->current_id = 0;
1308 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1313 if (matroska->num_levels > 0) {
1314 if (length != EBML_UNKNOWN_LENGTH &&
1315 level->length != EBML_UNKNOWN_LENGTH) {
1316 uint64_t elem_end = pos_alt + length,
1317 level_end = level->start + level->length;
1319 if (elem_end < level_end) {
1321 } else if (elem_end == level_end) {
1322 level_check = LEVEL_ENDED;
1324 av_log(matroska->ctx, AV_LOG_ERROR,
1325 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1326 "containing master element ending at 0x%"PRIx64"\n",
1327 pos, elem_end, level_end);
1328 return AVERROR_INVALIDDATA;
1330 } else if (length != EBML_UNKNOWN_LENGTH) {
1332 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1333 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1334 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1335 return AVERROR_INVALIDDATA;
1338 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1339 || syntax->type == EBML_NEST)) {
1340 // According to the current specifications only clusters and
1341 // segments are allowed to be unknown-length. We also accept
1342 // other unknown-length master elements.
1343 av_log(matroska->ctx, AV_LOG_WARNING,
1344 "Found unknown-length element 0x%"PRIX32" other than "
1345 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1346 "parsing will nevertheless be attempted.\n", id, pos);
1353 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1354 if (length != EBML_UNKNOWN_LENGTH) {
1355 av_log(matroska->ctx, AV_LOG_ERROR,
1356 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1357 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1358 length, max_lengths[syntax->type], id, pos);
1359 } else if (syntax->type != EBML_NONE) {
1360 av_log(matroska->ctx, AV_LOG_ERROR,
1361 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1362 "unknown length, yet the length of an element of its "
1363 "type must be known.\n", id, pos);
1365 av_log(matroska->ctx, AV_LOG_ERROR,
1366 "Found unknown-length element with ID 0x%"PRIX32" at "
1367 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1368 "available.\n", id, pos);
1370 return AVERROR_INVALIDDATA;
1373 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1374 // Loosing sync will likely manifest itself as encountering unknown
1375 // elements which are not reliably distinguishable from elements
1376 // belonging to future extensions of the format.
1377 // We use a heuristic to detect such situations: If the current
1378 // element is not expected at the current syntax level and there
1379 // were only a few unknown elements in a row, then the element is
1380 // skipped or considered defective based upon the length of the
1381 // current element (i.e. how much would be skipped); if there were
1382 // more than a few skipped elements in a row and skipping the current
1383 // element would lead us more than SKIP_THRESHOLD away from the last
1384 // known good position, then it is inferred that an error occurred.
1385 // The dependency on the number of unknown elements in a row exists
1386 // because the distance to the last known good position is
1387 // automatically big if the last parsed element was big.
1388 // In both cases, each unknown element is considered equivalent to
1389 // UNKNOWN_EQUIV of skipped bytes for the check.
1390 // The whole check is only done for non-seekable output, because
1391 // in this situation skipped data can't simply be rechecked later.
1392 // This is especially important when using unkown length elements
1393 // as the check for whether a child exceeds its containing master
1394 // element is not effective in this situation.
1396 matroska->unknown_count = 0;
1398 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1400 if (matroska->unknown_count > 3)
1401 dist += pos_alt - matroska->resync_pos;
1403 if (dist > SKIP_THRESHOLD) {
1404 av_log(matroska->ctx, AV_LOG_ERROR,
1405 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1406 "length 0x%"PRIx64" considered as invalid data. Last "
1407 "known good position 0x%"PRIx64", %d unknown elements"
1408 " in a row\n", id, pos, length, matroska->resync_pos,
1409 matroska->unknown_count);
1410 return AVERROR_INVALIDDATA;
1415 if (update_pos > 0) {
1416 // We have found an element that is allowed at this place
1417 // in the hierarchy and it passed all checks, so treat the beginning
1418 // of the element as the "last known good" position.
1419 matroska->resync_pos = pos;
1422 if (!data && length != EBML_UNKNOWN_LENGTH)
1426 switch (syntax->type) {
1428 res = ebml_read_uint(pb, length, syntax->def.u, data);
1431 res = ebml_read_sint(pb, length, syntax->def.i, data);
1434 res = ebml_read_float(pb, length, syntax->def.f, data);
1438 res = ebml_read_ascii(pb, length, syntax->def.s, data);
1441 res = ebml_read_binary(pb, length, pos_alt, data);
1445 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1447 if (id == MATROSKA_ID_SEGMENT)
1448 matroska->segment_start = pos_alt;
1449 if (id == MATROSKA_ID_CUES)
1450 matroska->cues_parsing_deferred = 0;
1451 if (syntax->type == EBML_LEVEL1 &&
1452 (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1453 if (!level1_elem->pos) {
1454 // Zero is not a valid position for a level 1 element.
1455 level1_elem->pos = pos;
1456 } else if (level1_elem->pos != pos)
1457 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1458 level1_elem->parsed = 1;
1460 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1469 if (ffio_limit(pb, length) != length) {
1470 // ffio_limit emits its own error message,
1471 // so we don't have to.
1472 return AVERROR(EIO);
1474 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1475 // avio_skip might take us past EOF. We check for this
1476 // by skipping only length - 1 bytes, reading a byte and
1477 // checking the error flags. This is done in order to check
1478 // that the element has been properly skipped even when
1479 // no filesize (that ffio_limit relies on) is available.
1481 res = NEEDS_CHECKING;
1488 if (res == NEEDS_CHECKING) {
1489 if (pb->eof_reached) {
1498 if (res == AVERROR_INVALIDDATA)
1499 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1500 else if (res == AVERROR(EIO))
1501 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1502 else if (res == AVERROR_EOF) {
1503 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1511 if (syntax->is_counted && data) {
1512 CountedElement *elem = data;
1513 if (elem->count != UINT_MAX)
1517 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1518 level = &matroska->levels[matroska->num_levels - 1];
1519 pos = avio_tell(pb);
1521 // Given that pos >= level->start no check for
1522 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1523 while (matroska->num_levels && pos == level->start + level->length) {
1524 matroska->num_levels--;
1532 static void ebml_free(EbmlSyntax *syntax, void *data)
1535 for (i = 0; syntax[i].id; i++) {
1536 void *data_off = (char *) data + syntax[i].data_offset;
1537 switch (syntax[i].type) {
1543 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1547 if (syntax[i].list_elem_size) {
1548 EbmlList *list = data_off;
1549 char *ptr = list->elem;
1550 for (j = 0; j < list->nb_elem;
1551 j++, ptr += syntax[i].list_elem_size)
1552 ebml_free(syntax[i].def.n, ptr);
1553 av_freep(&list->elem);
1555 list->alloc_elem_size = 0;
1557 ebml_free(syntax[i].def.n, data_off);
1567 static int matroska_probe(const AVProbeData *p)
1570 int len_mask = 0x80, size = 1, n = 1, i;
1573 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1576 /* length of header */
1578 while (size <= 8 && !(total & len_mask)) {
1584 total &= (len_mask - 1);
1586 total = (total << 8) | p->buf[4 + n++];
1588 if (total + 1 == 1ULL << (7 * size)){
1589 /* Unknown-length header - simply parse the whole buffer. */
1590 total = p->buf_size - 4 - size;
1592 /* Does the probe data contain the whole header? */
1593 if (p->buf_size < 4 + size + total)
1597 /* The header should contain a known document type. For now,
1598 * we don't parse the whole header but simply check for the
1599 * availability of that array of characters inside the header.
1600 * Not fully fool-proof, but good enough. */
1601 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1602 size_t probelen = strlen(matroska_doctypes[i]);
1603 if (total < probelen)
1605 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1606 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1607 return AVPROBE_SCORE_MAX;
1610 // probably valid EBML header but no recognized doctype
1611 return AVPROBE_SCORE_EXTENSION;
1614 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1617 MatroskaTrack *tracks = matroska->tracks.elem;
1620 for (i = 0; i < matroska->tracks.nb_elem; i++)
1621 if (tracks[i].num == num)
1624 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1628 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1629 MatroskaTrack *track)
1631 MatroskaTrackEncoding *encodings = track->encodings.elem;
1632 uint8_t *data = *buf;
1633 int isize = *buf_size;
1634 uint8_t *pkt_data = NULL;
1635 uint8_t av_unused *newpktdata;
1636 int pkt_size = isize;
1640 if (pkt_size >= 10000000U)
1641 return AVERROR_INVALIDDATA;
1643 switch (encodings[0].compression.algo) {
1644 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1646 int header_size = encodings[0].compression.settings.size;
1647 uint8_t *header = encodings[0].compression.settings.data;
1649 if (header_size && !header) {
1650 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1657 pkt_size = isize + header_size;
1658 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1660 return AVERROR(ENOMEM);
1662 memcpy(pkt_data, header, header_size);
1663 memcpy(pkt_data + header_size, data, isize);
1667 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1670 olen = pkt_size *= 3;
1671 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1672 + AV_INPUT_BUFFER_PADDING_SIZE);
1674 result = AVERROR(ENOMEM);
1677 pkt_data = newpktdata;
1678 result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1679 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1681 result = AVERROR_INVALIDDATA;
1688 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1690 z_stream zstream = { 0 };
1691 if (inflateInit(&zstream) != Z_OK)
1693 zstream.next_in = data;
1694 zstream.avail_in = isize;
1697 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1699 inflateEnd(&zstream);
1700 result = AVERROR(ENOMEM);
1703 pkt_data = newpktdata;
1704 zstream.avail_out = pkt_size - zstream.total_out;
1705 zstream.next_out = pkt_data + zstream.total_out;
1706 result = inflate(&zstream, Z_NO_FLUSH);
1707 } while (result == Z_OK && pkt_size < 10000000);
1708 pkt_size = zstream.total_out;
1709 inflateEnd(&zstream);
1710 if (result != Z_STREAM_END) {
1711 if (result == Z_MEM_ERROR)
1712 result = AVERROR(ENOMEM);
1714 result = AVERROR_INVALIDDATA;
1721 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1723 bz_stream bzstream = { 0 };
1724 if (BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1726 bzstream.next_in = data;
1727 bzstream.avail_in = isize;
1730 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1732 BZ2_bzDecompressEnd(&bzstream);
1733 result = AVERROR(ENOMEM);
1736 pkt_data = newpktdata;
1737 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1738 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1739 result = BZ2_bzDecompress(&bzstream);
1740 } while (result == BZ_OK && pkt_size < 10000000);
1741 pkt_size = bzstream.total_out_lo32;
1742 BZ2_bzDecompressEnd(&bzstream);
1743 if (result != BZ_STREAM_END) {
1744 if (result == BZ_MEM_ERROR)
1745 result = AVERROR(ENOMEM);
1747 result = AVERROR_INVALIDDATA;
1754 return AVERROR_INVALIDDATA;
1757 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1760 *buf_size = pkt_size;
1768 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1769 AVDictionary **metadata, char *prefix)
1771 MatroskaTag *tags = list->elem;
1775 for (i = 0; i < list->nb_elem; i++) {
1776 const char *lang = tags[i].lang &&
1777 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1779 if (!tags[i].name) {
1780 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1784 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1786 av_strlcpy(key, tags[i].name, sizeof(key));
1787 if (tags[i].def || !lang) {
1788 av_dict_set(metadata, key, tags[i].string, 0);
1789 if (tags[i].sub.nb_elem)
1790 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1793 av_strlcat(key, "-", sizeof(key));
1794 av_strlcat(key, lang, sizeof(key));
1795 av_dict_set(metadata, key, tags[i].string, 0);
1796 if (tags[i].sub.nb_elem)
1797 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1800 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1803 static void matroska_convert_tags(AVFormatContext *s)
1805 MatroskaDemuxContext *matroska = s->priv_data;
1806 MatroskaTags *tags = matroska->tags.elem;
1809 for (i = 0; i < matroska->tags.nb_elem; i++) {
1810 if (tags[i].target.attachuid) {
1811 MatroskaAttachment *attachment = matroska->attachments.elem;
1813 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1814 if (attachment[j].uid == tags[i].target.attachuid &&
1815 attachment[j].stream) {
1816 matroska_convert_tag(s, &tags[i].tag,
1817 &attachment[j].stream->metadata, NULL);
1822 av_log(s, AV_LOG_WARNING,
1823 "The tags at index %d refer to a "
1824 "non-existent attachment %"PRId64".\n",
1825 i, tags[i].target.attachuid);
1827 } else if (tags[i].target.chapteruid) {
1828 MatroskaChapter *chapter = matroska->chapters.elem;
1830 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1831 if (chapter[j].uid == tags[i].target.chapteruid &&
1832 chapter[j].chapter) {
1833 matroska_convert_tag(s, &tags[i].tag,
1834 &chapter[j].chapter->metadata, NULL);
1839 av_log(s, AV_LOG_WARNING,
1840 "The tags at index %d refer to a non-existent chapter "
1842 i, tags[i].target.chapteruid);
1844 } else if (tags[i].target.trackuid) {
1845 MatroskaTrack *track = matroska->tracks.elem;
1847 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1848 if (track[j].uid == tags[i].target.trackuid &&
1850 matroska_convert_tag(s, &tags[i].tag,
1851 &track[j].stream->metadata, NULL);
1856 av_log(s, AV_LOG_WARNING,
1857 "The tags at index %d refer to a non-existent track "
1859 i, tags[i].target.trackuid);
1862 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1863 tags[i].target.type);
1868 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1871 uint32_t saved_id = matroska->current_id;
1872 int64_t before_pos = avio_tell(matroska->ctx->pb);
1876 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1877 /* We don't want to lose our seekhead level, so we add
1878 * a dummy. This is a crude hack. */
1879 if (matroska->num_levels == EBML_MAX_DEPTH) {
1880 av_log(matroska->ctx, AV_LOG_INFO,
1881 "Max EBML element depth (%d) reached, "
1882 "cannot parse further.\n", EBML_MAX_DEPTH);
1883 ret = AVERROR_INVALIDDATA;
1885 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1886 matroska->num_levels++;
1887 matroska->current_id = 0;
1889 ret = ebml_parse(matroska, matroska_segment, matroska);
1890 if (ret == LEVEL_ENDED) {
1891 /* This can only happen if the seek brought us beyond EOF. */
1896 /* Seek back - notice that in all instances where this is used
1897 * it is safe to set the level to 1. */
1898 matroska_reset_status(matroska, saved_id, before_pos);
1903 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1905 EbmlList *seekhead_list = &matroska->seekhead;
1908 // we should not do any seeking in the streaming case
1909 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1912 for (i = 0; i < seekhead_list->nb_elem; i++) {
1913 MatroskaSeekhead *seekheads = seekhead_list->elem;
1914 uint32_t id = seekheads[i].id;
1915 int64_t pos = seekheads[i].pos + matroska->segment_start;
1916 MatroskaLevel1Element *elem;
1918 if (id != seekheads[i].id || pos < matroska->segment_start)
1921 elem = matroska_find_level1_elem(matroska, id, pos);
1922 if (!elem || elem->parsed)
1927 // defer cues parsing until we actually need cue data.
1928 if (id == MATROSKA_ID_CUES)
1931 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1932 // mark index as broken
1933 matroska->cues_parsing_deferred = -1;
1941 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1943 EbmlList *index_list;
1944 MatroskaIndex *index;
1945 uint64_t index_scale = 1;
1948 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1951 index_list = &matroska->index;
1952 index = index_list->elem;
1953 if (index_list->nb_elem < 2)
1955 if (index[1].time > 1E14 / matroska->time_scale) {
1956 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1959 for (i = 0; i < index_list->nb_elem; i++) {
1960 EbmlList *pos_list = &index[i].pos;
1961 MatroskaIndexPos *pos = pos_list->elem;
1962 for (j = 0; j < pos_list->nb_elem; j++) {
1963 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1965 if (track && track->stream)
1966 av_add_index_entry(track->stream,
1967 pos[j].pos + matroska->segment_start,
1968 index[i].time / index_scale, 0, 0,
1974 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1977 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1980 for (i = 0; i < matroska->num_level1_elems; i++) {
1981 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1982 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1983 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1984 matroska->cues_parsing_deferred = -1;
1990 matroska_add_index_entries(matroska);
1993 static int matroska_aac_profile(char *codec_id)
1995 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1998 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
1999 if (strstr(codec_id, aac_profiles[profile]))
2004 static int matroska_aac_sri(int samplerate)
2008 for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
2009 if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
2014 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
2016 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
2017 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
2020 static int matroska_parse_flac(AVFormatContext *s,
2021 MatroskaTrack *track,
2024 AVStream *st = track->stream;
2025 uint8_t *p = track->codec_priv.data;
2026 int size = track->codec_priv.size;
2028 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
2029 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
2030 track->codec_priv.size = 0;
2034 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
2036 p += track->codec_priv.size;
2037 size -= track->codec_priv.size;
2039 /* parse the remaining metadata blocks if present */
2041 int block_last, block_type, block_size;
2043 flac_parse_block_header(p, &block_last, &block_type, &block_size);
2047 if (block_size > size)
2050 /* check for the channel mask */
2051 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2052 AVDictionary *dict = NULL;
2053 AVDictionaryEntry *chmask;
2055 ff_vorbis_comment(s, &dict, p, block_size, 0);
2056 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2058 uint64_t mask = strtol(chmask->value, NULL, 0);
2059 if (!mask || mask & ~0x3ffffULL) {
2060 av_log(s, AV_LOG_WARNING,
2061 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2063 st->codecpar->channel_layout = mask;
2065 av_dict_free(&dict);
2075 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2077 int minor, micro, bttb = 0;
2079 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2080 * this function, and fixed in 57.52 */
2081 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, µ) == 2)
2082 bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2084 switch (field_order) {
2085 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2086 return AV_FIELD_PROGRESSIVE;
2087 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2088 return AV_FIELD_UNKNOWN;
2089 case MATROSKA_VIDEO_FIELDORDER_TT:
2091 case MATROSKA_VIDEO_FIELDORDER_BB:
2093 case MATROSKA_VIDEO_FIELDORDER_BT:
2094 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2095 case MATROSKA_VIDEO_FIELDORDER_TB:
2096 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2098 return AV_FIELD_UNKNOWN;
2102 static void mkv_stereo_mode_display_mul(int stereo_mode,
2103 int *h_width, int *h_height)
2105 switch (stereo_mode) {
2106 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2107 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2108 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2109 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2110 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2112 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2113 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2114 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2115 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2118 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2119 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2120 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2121 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2127 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2128 const MatroskaTrackVideoColor *color = track->video.color.elem;
2129 const MatroskaMasteringMeta *mastering_meta;
2130 int has_mastering_primaries, has_mastering_luminance;
2132 if (!track->video.color.nb_elem)
2135 mastering_meta = &color->mastering_meta;
2136 // Mastering primaries are CIE 1931 coords, and must be > 0.
2137 has_mastering_primaries =
2138 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2139 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2140 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2141 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2142 has_mastering_luminance = mastering_meta->max_luminance >
2143 mastering_meta->min_luminance.el.f &&
2144 mastering_meta->min_luminance.el.f >= 0 &&
2145 mastering_meta->min_luminance.count;
2147 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2148 st->codecpar->color_space = color->matrix_coefficients;
2149 if (color->primaries != AVCOL_PRI_RESERVED &&
2150 color->primaries != AVCOL_PRI_RESERVED0)
2151 st->codecpar->color_primaries = color->primaries;
2152 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2153 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2154 st->codecpar->color_trc = color->transfer_characteristics;
2155 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2156 color->range <= AVCOL_RANGE_JPEG)
2157 st->codecpar->color_range = color->range;
2158 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2159 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2160 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2161 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2162 st->codecpar->chroma_location =
2163 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2164 (color->chroma_siting_vert - 1) << 7);
2166 if (color->max_cll && color->max_fall) {
2169 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2171 return AVERROR(ENOMEM);
2172 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2173 (uint8_t *)metadata, size);
2175 av_freep(&metadata);
2178 metadata->MaxCLL = color->max_cll;
2179 metadata->MaxFALL = color->max_fall;
2182 if (has_mastering_primaries || has_mastering_luminance) {
2183 AVMasteringDisplayMetadata *metadata =
2184 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2185 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2186 sizeof(AVMasteringDisplayMetadata));
2188 return AVERROR(ENOMEM);
2190 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2191 if (has_mastering_primaries) {
2192 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2193 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2194 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2195 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2196 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2197 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2198 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2199 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2200 metadata->has_primaries = 1;
2202 if (has_mastering_luminance) {
2203 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2204 metadata->min_luminance = av_d2q(mastering_meta->min_luminance.el.f, INT_MAX);
2205 metadata->has_luminance = 1;
2211 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2214 AVSphericalMapping *spherical;
2215 const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2216 const uint8_t *priv_data = mkv_projection->private.data;
2217 enum AVSphericalProjection projection;
2218 size_t spherical_size;
2219 uint32_t l = 0, t = 0, r = 0, b = 0;
2220 uint32_t padding = 0;
2223 if (mkv_projection->private.size && priv_data[0] != 0) {
2224 av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2228 switch (track->video.projection.type) {
2229 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2230 if (track->video.projection.private.size == 20) {
2231 t = AV_RB32(priv_data + 4);
2232 b = AV_RB32(priv_data + 8);
2233 l = AV_RB32(priv_data + 12);
2234 r = AV_RB32(priv_data + 16);
2236 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2237 av_log(logctx, AV_LOG_ERROR,
2238 "Invalid bounding rectangle coordinates "
2239 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2241 return AVERROR_INVALIDDATA;
2243 } else if (track->video.projection.private.size != 0) {
2244 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2245 return AVERROR_INVALIDDATA;
2248 if (l || t || r || b)
2249 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2251 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2253 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2254 if (track->video.projection.private.size < 4) {
2255 av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2256 return AVERROR_INVALIDDATA;
2257 } else if (track->video.projection.private.size == 12) {
2258 uint32_t layout = AV_RB32(priv_data + 4);
2260 av_log(logctx, AV_LOG_WARNING,
2261 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2264 projection = AV_SPHERICAL_CUBEMAP;
2265 padding = AV_RB32(priv_data + 8);
2267 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2268 return AVERROR_INVALIDDATA;
2271 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2272 /* No Spherical metadata */
2275 av_log(logctx, AV_LOG_WARNING,
2276 "Unknown spherical metadata type %"PRIu64"\n",
2277 track->video.projection.type);
2281 spherical = av_spherical_alloc(&spherical_size);
2283 return AVERROR(ENOMEM);
2285 spherical->projection = projection;
2287 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2288 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2289 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2291 spherical->padding = padding;
2293 spherical->bound_left = l;
2294 spherical->bound_top = t;
2295 spherical->bound_right = r;
2296 spherical->bound_bottom = b;
2298 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2301 av_freep(&spherical);
2308 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2310 const AVCodecTag *codec_tags;
2312 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2313 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2315 /* Normalize noncompliant private data that starts with the fourcc
2316 * by expanding/shifting the data by 4 bytes and storing the data
2317 * size at the start. */
2318 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2319 int ret = av_buffer_realloc(&track->codec_priv.buf,
2320 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2324 track->codec_priv.data = track->codec_priv.buf->data;
2325 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2326 track->codec_priv.size += 4;
2327 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2330 *fourcc = AV_RL32(track->codec_priv.data + 4);
2331 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2336 static int matroska_parse_tracks(AVFormatContext *s)
2338 MatroskaDemuxContext *matroska = s->priv_data;
2339 MatroskaTrack *tracks = matroska->tracks.elem;
2344 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2345 MatroskaTrack *track = &tracks[i];
2346 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2347 EbmlList *encodings_list = &track->encodings;
2348 MatroskaTrackEncoding *encodings = encodings_list->elem;
2349 uint8_t *extradata = NULL;
2350 int extradata_size = 0;
2351 int extradata_offset = 0;
2352 uint32_t fourcc = 0;
2354 char* key_id_base64 = NULL;
2357 /* Apply some sanity checks. */
2358 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2359 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2360 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2361 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2362 av_log(matroska->ctx, AV_LOG_INFO,
2363 "Unknown or unsupported track type %"PRIu64"\n",
2367 if (!track->codec_id)
2370 if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
2371 || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
2372 || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2373 || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2375 av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
2379 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2380 isnan(track->audio.samplerate)) {
2381 av_log(matroska->ctx, AV_LOG_WARNING,
2382 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2383 track->audio.samplerate);
2384 track->audio.samplerate = 8000;
2387 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2388 if (!track->default_duration && track->video.frame_rate > 0) {
2389 double default_duration = 1000000000 / track->video.frame_rate;
2390 if (default_duration > UINT64_MAX || default_duration < 0) {
2391 av_log(matroska->ctx, AV_LOG_WARNING,
2392 "Invalid frame rate %e. Cannot calculate default duration.\n",
2393 track->video.frame_rate);
2395 track->default_duration = default_duration;
2398 if (track->video.display_width == -1)
2399 track->video.display_width = track->video.pixel_width;
2400 if (track->video.display_height == -1)
2401 track->video.display_height = track->video.pixel_height;
2402 if (track->video.color_space.size == 4)
2403 fourcc = AV_RL32(track->video.color_space.data);
2404 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2405 if (!track->audio.out_samplerate)
2406 track->audio.out_samplerate = track->audio.samplerate;
2408 if (encodings_list->nb_elem > 1) {
2409 av_log(matroska->ctx, AV_LOG_ERROR,
2410 "Multiple combined encodings not supported");
2411 } else if (encodings_list->nb_elem == 1) {
2412 if (encodings[0].type) {
2413 if (encodings[0].encryption.key_id.size > 0) {
2414 /* Save the encryption key id to be stored later as a
2416 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2417 key_id_base64 = av_malloc(b64_size);
2418 if (key_id_base64 == NULL)
2419 return AVERROR(ENOMEM);
2421 av_base64_encode(key_id_base64, b64_size,
2422 encodings[0].encryption.key_id.data,
2423 encodings[0].encryption.key_id.size);
2425 encodings[0].scope = 0;
2426 av_log(matroska->ctx, AV_LOG_ERROR,
2427 "Unsupported encoding type");
2431 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2434 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2437 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2439 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2440 encodings[0].scope = 0;
2441 av_log(matroska->ctx, AV_LOG_ERROR,
2442 "Unsupported encoding type");
2443 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2444 uint8_t *codec_priv = track->codec_priv.data;
2445 int ret = matroska_decode_buffer(&track->codec_priv.data,
2446 &track->codec_priv.size,
2449 track->codec_priv.data = NULL;
2450 track->codec_priv.size = 0;
2451 av_log(matroska->ctx, AV_LOG_ERROR,
2452 "Failed to decode codec private data\n");
2455 if (codec_priv != track->codec_priv.data) {
2456 av_buffer_unref(&track->codec_priv.buf);
2457 if (track->codec_priv.data) {
2458 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2459 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2461 if (!track->codec_priv.buf) {
2462 av_freep(&track->codec_priv.data);
2463 track->codec_priv.size = 0;
2464 return AVERROR(ENOMEM);
2470 track->needs_decoding = encodings && !encodings[0].type &&
2471 encodings[0].scope & 1 &&
2472 (encodings[0].compression.algo !=
2473 MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP ||
2474 encodings[0].compression.settings.size);
2476 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2477 if (av_strstart(track->codec_id, ff_mkv_codec_tags[j].str, NULL)) {
2478 codec_id = ff_mkv_codec_tags[j].id;
2483 st = track->stream = avformat_new_stream(s, NULL);
2485 av_free(key_id_base64);
2486 return AVERROR(ENOMEM);
2489 if (key_id_base64) {
2490 /* export encryption key id as base64 metadata tag */
2491 av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2492 AV_DICT_DONT_STRDUP_VAL);
2495 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2496 track->codec_priv.size >= 40 &&
2497 track->codec_priv.data) {
2498 track->ms_compat = 1;
2499 bit_depth = AV_RL16(track->codec_priv.data + 14);
2500 fourcc = AV_RL32(track->codec_priv.data + 16);
2501 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2504 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2506 extradata_offset = 40;
2507 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2508 track->codec_priv.size >= 14 &&
2509 track->codec_priv.data) {
2511 ffio_init_context(&b, track->codec_priv.data,
2512 track->codec_priv.size,
2513 0, NULL, NULL, NULL, NULL);
2514 ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2517 codec_id = st->codecpar->codec_id;
2518 fourcc = st->codecpar->codec_tag;
2519 extradata_offset = FFMIN(track->codec_priv.size, 18);
2520 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2521 /* Normally 36, but allow noncompliant private data */
2522 && (track->codec_priv.size >= 32)
2523 && (track->codec_priv.data)) {
2524 uint16_t sample_size;
2525 int ret = get_qt_codec(track, &fourcc, &codec_id);
2528 sample_size = AV_RB16(track->codec_priv.data + 26);
2530 if (sample_size == 8) {
2531 fourcc = MKTAG('r','a','w',' ');
2532 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2533 } else if (sample_size == 16) {
2534 fourcc = MKTAG('t','w','o','s');
2535 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2538 if ((fourcc == MKTAG('t','w','o','s') ||
2539 fourcc == MKTAG('s','o','w','t')) &&
2541 codec_id = AV_CODEC_ID_PCM_S8;
2542 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2543 (track->codec_priv.size >= 21) &&
2544 (track->codec_priv.data)) {
2545 int ret = get_qt_codec(track, &fourcc, &codec_id);
2548 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2549 fourcc = MKTAG('S','V','Q','3');
2550 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2552 if (codec_id == AV_CODEC_ID_NONE)
2553 av_log(matroska->ctx, AV_LOG_ERROR,
2554 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2555 if (track->codec_priv.size >= 86) {
2556 bit_depth = AV_RB16(track->codec_priv.data + 82);
2557 ffio_init_context(&b, track->codec_priv.data,
2558 track->codec_priv.size,
2559 0, NULL, NULL, NULL, NULL);
2560 if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2562 track->has_palette = 1;
2565 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2566 switch (track->audio.bitdepth) {
2568 codec_id = AV_CODEC_ID_PCM_U8;
2571 codec_id = AV_CODEC_ID_PCM_S24BE;
2574 codec_id = AV_CODEC_ID_PCM_S32BE;
2577 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2578 switch (track->audio.bitdepth) {
2580 codec_id = AV_CODEC_ID_PCM_U8;
2583 codec_id = AV_CODEC_ID_PCM_S24LE;
2586 codec_id = AV_CODEC_ID_PCM_S32LE;
2589 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2590 track->audio.bitdepth == 64) {
2591 codec_id = AV_CODEC_ID_PCM_F64LE;
2592 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2593 int profile = matroska_aac_profile(track->codec_id);
2594 int sri = matroska_aac_sri(track->audio.samplerate);
2595 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2597 return AVERROR(ENOMEM);
2598 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2599 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2600 if (strstr(track->codec_id, "SBR")) {
2601 sri = matroska_aac_sri(track->audio.out_samplerate);
2602 extradata[2] = 0x56;
2603 extradata[3] = 0xE5;
2604 extradata[4] = 0x80 | (sri << 3);
2608 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2609 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2610 * Create the "atom size", "tag", and "tag version" fields the
2611 * decoder expects manually. */
2612 extradata_size = 12 + track->codec_priv.size;
2613 extradata = av_mallocz(extradata_size +
2614 AV_INPUT_BUFFER_PADDING_SIZE);
2616 return AVERROR(ENOMEM);
2617 AV_WB32(extradata, extradata_size);
2618 memcpy(&extradata[4], "alac", 4);
2619 AV_WB32(&extradata[8], 0);
2620 memcpy(&extradata[12], track->codec_priv.data,
2621 track->codec_priv.size);
2622 } else if (codec_id == AV_CODEC_ID_TTA) {
2624 if (track->audio.channels > UINT16_MAX ||
2625 track->audio.bitdepth > UINT16_MAX) {
2626 av_log(matroska->ctx, AV_LOG_WARNING,
2627 "Too large audio channel number %"PRIu64
2628 " or bitdepth %"PRIu64". Skipping track.\n",
2629 track->audio.channels, track->audio.bitdepth);
2630 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2631 return AVERROR_INVALIDDATA;
2635 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2636 return AVERROR_INVALIDDATA;
2637 extradata_size = 22;
2638 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2640 return AVERROR(ENOMEM);
2642 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2643 bytestream_put_le16(&ptr, 1);
2644 bytestream_put_le16(&ptr, track->audio.channels);
2645 bytestream_put_le16(&ptr, track->audio.bitdepth);
2646 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2647 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2648 track->audio.out_samplerate,
2649 AV_TIME_BASE * 1000));
2650 } else if (codec_id == AV_CODEC_ID_RV10 ||
2651 codec_id == AV_CODEC_ID_RV20 ||
2652 codec_id == AV_CODEC_ID_RV30 ||
2653 codec_id == AV_CODEC_ID_RV40) {
2654 extradata_offset = 26;
2655 } else if (codec_id == AV_CODEC_ID_RA_144) {
2656 track->audio.out_samplerate = 8000;
2657 track->audio.channels = 1;
2658 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2659 codec_id == AV_CODEC_ID_COOK ||
2660 codec_id == AV_CODEC_ID_ATRAC3 ||
2661 codec_id == AV_CODEC_ID_SIPR)
2662 && track->codec_priv.data) {
2665 ffio_init_context(&b, track->codec_priv.data,
2666 track->codec_priv.size,
2667 0, NULL, NULL, NULL, NULL);
2669 flavor = avio_rb16(&b);
2670 track->audio.coded_framesize = avio_rb32(&b);
2672 track->audio.sub_packet_h = avio_rb16(&b);
2673 track->audio.frame_size = avio_rb16(&b);
2674 track->audio.sub_packet_size = avio_rb16(&b);
2675 if (track->audio.coded_framesize <= 0 ||
2676 track->audio.sub_packet_h <= 0 ||
2677 track->audio.frame_size <= 0)
2678 return AVERROR_INVALIDDATA;
2680 if (codec_id == AV_CODEC_ID_RA_288) {
2681 if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2682 != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2683 return AVERROR_INVALIDDATA;
2684 st->codecpar->block_align = track->audio.coded_framesize;
2685 track->codec_priv.size = 0;
2687 if (codec_id == AV_CODEC_ID_SIPR) {
2688 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2690 return AVERROR_INVALIDDATA;
2691 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2692 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2693 } else if (track->audio.sub_packet_size <= 0 ||
2694 track->audio.frame_size % track->audio.sub_packet_size)
2695 return AVERROR_INVALIDDATA;
2696 st->codecpar->block_align = track->audio.sub_packet_size;
2697 extradata_offset = 78;
2699 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2700 track->audio.frame_size);
2701 if (!track->audio.buf)
2702 return AVERROR(ENOMEM);
2703 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2704 ret = matroska_parse_flac(s, track, &extradata_offset);
2707 } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2708 av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2709 "in absence of valid CodecPrivate.\n");
2711 extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2713 return AVERROR(ENOMEM);
2714 AV_WL16(extradata, 0x410);
2715 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2716 fourcc = AV_RL32(track->codec_priv.data);
2717 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2718 /* we don't need any value stored in CodecPrivate.
2719 make sure that it's not exported as extradata. */
2720 track->codec_priv.size = 0;
2721 } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2722 /* For now, propagate only the OBUs, if any. Once libavcodec is
2723 updated to handle isobmff style extradata this can be removed. */
2724 extradata_offset = 4;
2726 track->codec_priv.size -= extradata_offset;
2728 if (codec_id == AV_CODEC_ID_NONE)
2729 av_log(matroska->ctx, AV_LOG_INFO,
2730 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2732 if (track->time_scale < 0.01) {
2733 av_log(matroska->ctx, AV_LOG_WARNING,
2734 "Track TimestampScale too small %f, assuming 1.0.\n",
2736 track->time_scale = 1.0;
2738 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2739 1000 * 1000 * 1000); /* 64 bit pts in ns */
2741 /* convert the delay from ns to the track timebase */
2742 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2743 (AVRational){ 1, 1000000000 },
2746 st->codecpar->codec_id = codec_id;
2748 if (strcmp(track->language, "und"))
2749 av_dict_set(&st->metadata, "language", track->language, 0);
2750 av_dict_set(&st->metadata, "title", track->name, 0);
2752 if (track->flag_default)
2753 st->disposition |= AV_DISPOSITION_DEFAULT;
2754 if (track->flag_forced)
2755 st->disposition |= AV_DISPOSITION_FORCED;
2756 if (track->flag_comment)
2757 st->disposition |= AV_DISPOSITION_COMMENT;
2758 if (track->flag_hearingimpaired)
2759 st->disposition |= AV_DISPOSITION_HEARING_IMPAIRED;
2760 if (track->flag_visualimpaired)
2761 st->disposition |= AV_DISPOSITION_VISUAL_IMPAIRED;
2762 if (track->flag_original.count > 0)
2763 st->disposition |= track->flag_original.el.u ? AV_DISPOSITION_ORIGINAL
2764 : AV_DISPOSITION_DUB;
2766 if (!st->codecpar->extradata) {
2768 st->codecpar->extradata = extradata;
2769 st->codecpar->extradata_size = extradata_size;
2770 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2771 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2772 return AVERROR(ENOMEM);
2773 memcpy(st->codecpar->extradata,
2774 track->codec_priv.data + extradata_offset,
2775 track->codec_priv.size);
2779 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2780 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2781 int display_width_mul = 1;
2782 int display_height_mul = 1;
2784 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2785 st->codecpar->codec_tag = fourcc;
2787 st->codecpar->bits_per_coded_sample = bit_depth;
2788 st->codecpar->width = track->video.pixel_width;
2789 st->codecpar->height = track->video.pixel_height;
2791 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2792 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2793 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2794 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2796 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2797 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2799 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2800 av_reduce(&st->sample_aspect_ratio.num,
2801 &st->sample_aspect_ratio.den,
2802 st->codecpar->height * track->video.display_width * display_width_mul,
2803 st->codecpar->width * track->video.display_height * display_height_mul,
2806 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2807 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2809 if (track->default_duration) {
2810 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2811 1000000000, track->default_duration, 30000);
2812 #if FF_API_R_FRAME_RATE
2813 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2814 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2815 st->r_frame_rate = st->avg_frame_rate;
2819 /* export stereo mode flag as metadata tag */
2820 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2821 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2823 /* export alpha mode flag as metadata tag */
2824 if (track->video.alpha_mode)
2825 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2827 /* if we have virtual track, mark the real tracks */
2828 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2830 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2832 snprintf(buf, sizeof(buf), "%s_%d",
2833 ff_matroska_video_stereo_plane[planes[j].type], i);
2834 for (k=0; k < matroska->tracks.nb_elem; k++)
2835 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2836 av_dict_set(&tracks[k].stream->metadata,
2837 "stereo_mode", buf, 0);
2841 // add stream level stereo3d side data if it is a supported format
2842 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2843 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2844 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2849 ret = mkv_parse_video_color(st, track);
2852 ret = mkv_parse_video_projection(st, track, matroska->ctx);
2855 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2856 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2857 st->codecpar->codec_tag = fourcc;
2858 st->codecpar->sample_rate = track->audio.out_samplerate;
2859 st->codecpar->channels = track->audio.channels;
2860 if (!st->codecpar->bits_per_coded_sample)
2861 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2862 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2863 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2864 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2865 st->need_parsing = AVSTREAM_PARSE_FULL;
2866 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2867 st->need_parsing = AVSTREAM_PARSE_HEADERS;
2868 if (track->codec_delay > 0) {
2869 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2870 (AVRational){1, 1000000000},
2871 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2872 48000 : st->codecpar->sample_rate});
2874 if (track->seek_preroll > 0) {
2875 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2876 (AVRational){1, 1000000000},
2877 (AVRational){1, st->codecpar->sample_rate});
2879 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2880 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2882 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2883 st->disposition |= AV_DISPOSITION_CAPTIONS;
2884 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2885 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2886 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2887 st->disposition |= AV_DISPOSITION_METADATA;
2889 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2890 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2892 if (track->flag_textdescriptions)
2893 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2900 static int matroska_read_header(AVFormatContext *s)
2902 MatroskaDemuxContext *matroska = s->priv_data;
2903 EbmlList *attachments_list = &matroska->attachments;
2904 EbmlList *chapters_list = &matroska->chapters;
2905 MatroskaAttachment *attachments;
2906 MatroskaChapter *chapters;
2907 uint64_t max_start = 0;
2913 matroska->cues_parsing_deferred = 1;
2915 /* First read the EBML header. */
2916 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2917 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2918 ebml_free(ebml_syntax, &ebml);
2919 return AVERROR_INVALIDDATA;
2921 if (ebml.version > EBML_VERSION ||
2922 ebml.max_size > sizeof(uint64_t) ||
2923 ebml.id_length > sizeof(uint32_t) ||
2924 ebml.doctype_version > 3) {
2925 avpriv_report_missing_feature(matroska->ctx,
2926 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2927 ebml.version, ebml.doctype, ebml.doctype_version);
2928 ebml_free(ebml_syntax, &ebml);
2929 return AVERROR_PATCHWELCOME;
2930 } else if (ebml.doctype_version == 3) {
2931 av_log(matroska->ctx, AV_LOG_WARNING,
2932 "EBML header using unsupported features\n"
2933 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2934 ebml.version, ebml.doctype, ebml.doctype_version);
2936 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2937 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2939 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
2940 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2941 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2942 ebml_free(ebml_syntax, &ebml);
2943 return AVERROR_INVALIDDATA;
2946 ebml_free(ebml_syntax, &ebml);
2948 matroska->pkt = av_packet_alloc();
2950 return AVERROR(ENOMEM);
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 AVPacket *pkt = &st->attached_pic;
3012 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
3013 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
3015 av_packet_unref(pkt);
3016 pkt->buf = attachments[j].bin.buf;
3017 attachments[j].bin.buf = NULL;
3018 pkt->data = attachments[j].bin.data;
3019 pkt->size = attachments[j].bin.size;
3020 pkt->stream_index = st->index;
3021 pkt->flags |= AV_PKT_FLAG_KEY;
3023 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
3024 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
3026 memcpy(st->codecpar->extradata, attachments[j].bin.data,
3027 attachments[j].bin.size);
3029 for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3030 if (av_strstart(attachments[j].mime, mkv_mime_tags[i].str, NULL)) {
3031 st->codecpar->codec_id = mkv_mime_tags[i].id;
3039 chapters = chapters_list->elem;
3040 for (i = 0; i < chapters_list->nb_elem; i++)
3041 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
3042 (max_start == 0 || chapters[i].start > max_start)) {
3043 chapters[i].chapter =
3044 avpriv_new_chapter(s, chapters[i].uid,
3045 (AVRational) { 1, 1000000000 },
3046 chapters[i].start, chapters[i].end,
3048 max_start = chapters[i].start;
3051 matroska_add_index_entries(matroska);
3053 matroska_convert_tags(s);
3057 matroska_read_close(s);
3062 * Put one packet in an application-supplied AVPacket struct.
3063 * Returns 0 on success or -1 on failure.
3065 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3068 if (matroska->queue) {
3069 MatroskaTrack *tracks = matroska->tracks.elem;
3070 MatroskaTrack *track;
3072 avpriv_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
3073 track = &tracks[pkt->stream_index];
3074 if (track->has_palette) {
3075 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
3077 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3079 memcpy(pal, track->palette, AVPALETTE_SIZE);
3081 track->has_palette = 0;
3090 * Free all packets in our internal queue.
3092 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3094 avpriv_packet_list_free(&matroska->queue, &matroska->queue_end);
3097 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3098 int size, int type, AVIOContext *pb,
3099 uint32_t lace_size[256], int *laces)
3102 uint8_t *data = *buf;
3106 lace_size[0] = size;
3111 return AVERROR_INVALIDDATA;
3118 case 0x1: /* Xiph lacing */
3122 for (n = 0; n < *laces - 1; n++) {
3127 return AVERROR_INVALIDDATA;
3130 lace_size[n] += temp;
3133 } while (temp == 0xff);
3136 return AVERROR_INVALIDDATA;
3138 lace_size[n] = size - total;
3142 case 0x2: /* fixed-size lacing */
3143 if (size % (*laces))
3144 return AVERROR_INVALIDDATA;
3145 for (n = 0; n < *laces; n++)
3146 lace_size[n] = size / *laces;
3149 case 0x3: /* EBML lacing */
3157 n = ebml_read_num(matroska, pb, 8, &num, 1);
3161 return AVERROR_INVALIDDATA;
3163 total = lace_size[0] = num;
3165 for (n = 1; n < *laces - 1; n++) {
3168 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3171 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3172 return AVERROR_INVALIDDATA;
3174 lace_size[n] = lace_size[n - 1] + snum;
3175 total += lace_size[n];
3181 return AVERROR_INVALIDDATA;
3183 lace_size[*laces - 1] = size - total;
3193 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3194 MatroskaTrack *track, AVStream *st,
3195 uint8_t *data, int size, uint64_t timecode,
3198 const int a = st->codecpar->block_align;
3199 const int sps = track->audio.sub_packet_size;
3200 const int cfs = track->audio.coded_framesize;
3201 const int h = track->audio.sub_packet_h;
3202 const int w = track->audio.frame_size;
3203 int y = track->audio.sub_packet_cnt;
3206 if (!track->audio.pkt_cnt) {
3207 if (track->audio.sub_packet_cnt == 0)
3208 track->audio.buf_timecode = timecode;
3209 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3210 if (size < cfs * h / 2) {
3211 av_log(matroska->ctx, AV_LOG_ERROR,
3212 "Corrupt int4 RM-style audio packet size\n");
3213 return AVERROR_INVALIDDATA;
3215 for (x = 0; x < h / 2; x++)
3216 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3217 data + x * cfs, cfs);
3218 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3220 av_log(matroska->ctx, AV_LOG_ERROR,
3221 "Corrupt sipr RM-style audio packet size\n");
3222 return AVERROR_INVALIDDATA;
3224 memcpy(track->audio.buf + y * w, data, w);
3227 av_log(matroska->ctx, AV_LOG_ERROR,
3228 "Corrupt generic RM-style audio packet size\n");
3229 return AVERROR_INVALIDDATA;
3231 for (x = 0; x < w / sps; x++)
3232 memcpy(track->audio.buf +
3233 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3234 data + x * sps, sps);
3237 if (++track->audio.sub_packet_cnt >= h) {
3238 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3239 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3240 track->audio.sub_packet_cnt = 0;
3241 track->audio.pkt_cnt = h * w / a;
3245 while (track->audio.pkt_cnt) {
3247 AVPacket *pkt = matroska->pkt;
3249 ret = av_new_packet(pkt, a);
3254 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3256 pkt->pts = track->audio.buf_timecode;
3257 track->audio.buf_timecode = AV_NOPTS_VALUE;
3259 pkt->stream_index = st->index;
3260 ret = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3262 av_packet_unref(pkt);
3263 return AVERROR(ENOMEM);
3270 /* reconstruct full wavpack blocks from mangled matroska ones */
3271 static int matroska_parse_wavpack(MatroskaTrack *track,
3272 uint8_t **data, int *size)
3274 uint8_t *dst = NULL;
3275 uint8_t *src = *data;
3280 int ret, offset = 0;
3283 return AVERROR_INVALIDDATA;
3285 av_assert1(track->stream->codecpar->extradata_size >= 2);
3286 ver = AV_RL16(track->stream->codecpar->extradata);
3288 samples = AV_RL32(src);
3292 while (srclen >= 8) {
3297 uint32_t flags = AV_RL32(src);
3298 uint32_t crc = AV_RL32(src + 4);
3302 multiblock = (flags & 0x1800) != 0x1800;
3305 ret = AVERROR_INVALIDDATA;
3308 blocksize = AV_RL32(src);
3314 if (blocksize > srclen) {
3315 ret = AVERROR_INVALIDDATA;
3319 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3321 ret = AVERROR(ENOMEM);
3325 dstlen += blocksize + 32;
3327 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3328 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3329 AV_WL16(dst + offset + 8, ver); // version
3330 AV_WL16(dst + offset + 10, 0); // track/index_no
3331 AV_WL32(dst + offset + 12, 0); // total samples
3332 AV_WL32(dst + offset + 16, 0); // block index
3333 AV_WL32(dst + offset + 20, samples); // number of samples
3334 AV_WL32(dst + offset + 24, flags); // flags
3335 AV_WL32(dst + offset + 28, crc); // crc
3336 memcpy(dst + offset + 32, src, blocksize); // block data
3339 srclen -= blocksize;
3340 offset += blocksize + 32;
3343 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3355 static int matroska_parse_prores(MatroskaTrack *track,
3356 uint8_t **data, int *size)
3359 int dstlen = *size + 8;
3361 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3363 return AVERROR(ENOMEM);
3365 AV_WB32(dst, dstlen);
3366 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3367 memcpy(dst + 8, *data, dstlen - 8);
3368 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3376 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3377 MatroskaTrack *track,
3379 uint8_t *data, int data_len,
3384 AVPacket *pkt = matroska->pkt;
3385 uint8_t *id, *settings, *text, *buf;
3386 int id_len, settings_len, text_len;
3391 return AVERROR_INVALIDDATA;
3394 q = data + data_len;
3399 if (*p == '\r' || *p == '\n') {
3408 if (p >= q || *p != '\n')
3409 return AVERROR_INVALIDDATA;
3415 if (*p == '\r' || *p == '\n') {
3416 settings_len = p - settings;
3424 if (p >= q || *p != '\n')
3425 return AVERROR_INVALIDDATA;
3430 while (text_len > 0) {
3431 const int len = text_len - 1;
3432 const uint8_t c = p[len];
3433 if (c != '\r' && c != '\n')
3439 return AVERROR_INVALIDDATA;
3441 err = av_new_packet(pkt, text_len);
3446 memcpy(pkt->data, text, text_len);
3449 buf = av_packet_new_side_data(pkt,
3450 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3453 av_packet_unref(pkt);
3454 return AVERROR(ENOMEM);
3456 memcpy(buf, id, id_len);
3459 if (settings_len > 0) {
3460 buf = av_packet_new_side_data(pkt,
3461 AV_PKT_DATA_WEBVTT_SETTINGS,
3464 av_packet_unref(pkt);
3465 return AVERROR(ENOMEM);
3467 memcpy(buf, settings, settings_len);
3470 // Do we need this for subtitles?
3471 // pkt->flags = AV_PKT_FLAG_KEY;
3473 pkt->stream_index = st->index;
3474 pkt->pts = timecode;
3476 // Do we need this for subtitles?
3477 // pkt->dts = timecode;
3479 pkt->duration = duration;
3482 err = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3484 av_packet_unref(pkt);
3485 return AVERROR(ENOMEM);
3491 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3492 MatroskaTrack *track, AVStream *st,
3493 AVBufferRef *buf, uint8_t *data, int pkt_size,
3494 uint64_t timecode, uint64_t lace_duration,
3495 int64_t pos, int is_keyframe,
3496 uint8_t *additional, uint64_t additional_id, int additional_size,
3497 int64_t discard_padding)
3499 uint8_t *pkt_data = data;
3501 AVPacket *pkt = matroska->pkt;
3503 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3504 res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3506 av_log(matroska->ctx, AV_LOG_ERROR,
3507 "Error parsing a wavpack block.\n");
3515 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3516 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3517 res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3519 av_log(matroska->ctx, AV_LOG_ERROR,
3520 "Error parsing a prores block.\n");
3528 if (!pkt_size && !additional_size)
3531 av_packet_unref(pkt);
3533 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3536 pkt->buf = av_buffer_ref(buf);
3539 res = AVERROR(ENOMEM);
3543 pkt->data = pkt_data;
3544 pkt->size = pkt_size;
3545 pkt->flags = is_keyframe;
3546 pkt->stream_index = st->index;
3548 if (additional_size > 0) {
3549 uint8_t *side_data = av_packet_new_side_data(pkt,
3550 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3551 additional_size + 8);
3553 av_packet_unref(pkt);
3554 return AVERROR(ENOMEM);
3556 AV_WB64(side_data, additional_id);
3557 memcpy(side_data + 8, additional, additional_size);
3560 if (discard_padding) {
3561 uint8_t *side_data = av_packet_new_side_data(pkt,
3562 AV_PKT_DATA_SKIP_SAMPLES,
3565 av_packet_unref(pkt);
3566 return AVERROR(ENOMEM);
3568 discard_padding = av_rescale_q(discard_padding,
3569 (AVRational){1, 1000000000},
3570 (AVRational){1, st->codecpar->sample_rate});
3571 if (discard_padding > 0) {
3572 AV_WL32(side_data + 4, discard_padding);
3574 AV_WL32(side_data, -discard_padding);
3578 if (track->ms_compat)
3579 pkt->dts = timecode;
3581 pkt->pts = timecode;
3583 pkt->duration = lace_duration;
3585 #if FF_API_CONVERGENCE_DURATION
3586 FF_DISABLE_DEPRECATION_WARNINGS
3587 if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3588 pkt->convergence_duration = lace_duration;
3590 FF_ENABLE_DEPRECATION_WARNINGS
3593 res = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3595 av_packet_unref(pkt);
3596 return AVERROR(ENOMEM);
3608 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3609 int size, int64_t pos, uint64_t cluster_time,
3610 uint64_t block_duration, int is_keyframe,
3611 uint8_t *additional, uint64_t additional_id, int additional_size,
3612 int64_t cluster_pos, int64_t discard_padding)
3614 uint64_t timecode = AV_NOPTS_VALUE;
3615 MatroskaTrack *track;
3620 uint32_t lace_size[256];
3621 int n, flags, laces = 0;
3623 int trust_default_duration;
3625 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3627 if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3632 track = matroska_find_track_by_num(matroska, num);
3633 if (!track || size < 3)
3634 return AVERROR_INVALIDDATA;
3636 if (!(st = track->stream)) {
3637 av_log(matroska->ctx, AV_LOG_VERBOSE,
3638 "No stream associated to TrackNumber %"PRIu64". "
3639 "Ignoring Block with this TrackNumber.\n", num);
3643 if (st->discard >= AVDISCARD_ALL)
3645 if (block_duration > INT64_MAX)
3646 block_duration = INT64_MAX;
3648 block_time = sign_extend(AV_RB16(data), 16);
3652 if (is_keyframe == -1)
3653 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3655 if (cluster_time != (uint64_t) -1 &&
3656 (block_time >= 0 || cluster_time >= -block_time)) {
3657 uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
3658 timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
3659 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3660 timecode < track->end_timecode)
3661 is_keyframe = 0; /* overlapping subtitles are not key frame */
3663 ff_reduce_index(matroska->ctx, st->index);
3664 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3669 if (matroska->skip_to_keyframe &&
3670 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3671 // Compare signed timecodes. Timecode may be negative due to codec delay
3672 // offset. We don't support timestamps greater than int64_t anyway - see
3674 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3677 matroska->skip_to_keyframe = 0;
3678 else if (!st->internal->skip_to_keyframe) {
3679 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3680 matroska->skip_to_keyframe = 0;
3684 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3685 &pb, lace_size, &laces);
3687 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3691 trust_default_duration = track->default_duration != 0;
3692 if (track->audio.samplerate == 8000 && trust_default_duration) {
3693 // If this is needed for more codecs, then add them here
3694 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3695 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3696 trust_default_duration = 0;
3700 if (!block_duration && trust_default_duration)
3701 block_duration = track->default_duration * laces / matroska->time_scale;
3703 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3704 track->end_timecode =
3705 FFMAX(track->end_timecode, timecode + block_duration);
3707 for (n = 0; n < laces; n++) {
3708 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3709 uint8_t *out_data = data;
3710 int out_size = lace_size[n];
3712 if (track->needs_decoding) {
3713 res = matroska_decode_buffer(&out_data, &out_size, track);
3716 /* Given that we are here means that out_data is no longer
3717 * owned by buf, so set it to NULL. This depends upon
3718 * zero-length header removal compression being ignored. */
3719 av_assert1(out_data != data);
3723 if (track->audio.buf) {
3724 res = matroska_parse_rm_audio(matroska, track, st,
3731 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3732 res = matroska_parse_webvtt(matroska, track, st,
3734 timecode, lace_duration,
3741 res = matroska_parse_frame(matroska, track, st, buf, out_data,
3742 out_size, timecode, lace_duration,
3743 pos, !n ? is_keyframe : 0,
3744 additional, additional_id, additional_size,
3750 if (timecode != AV_NOPTS_VALUE)
3751 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3752 data += lace_size[n];
3758 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3760 MatroskaCluster *cluster = &matroska->current_cluster;
3761 MatroskaBlock *block = &cluster->block;
3764 av_assert0(matroska->num_levels <= 2);
3766 if (matroska->num_levels == 1) {
3767 res = ebml_parse(matroska, matroska_segment, NULL);
3770 /* Found a cluster: subtract the size of the ID already read. */
3771 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3773 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3779 if (matroska->num_levels == 2) {
3780 /* We are inside a cluster. */
3781 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3783 if (res >= 0 && block->bin.size > 0) {
3784 int is_keyframe = block->non_simple ? block->reference.count == 0 : -1;
3785 uint8_t* additional = block->additional.size > 0 ?
3786 block->additional.data : NULL;
3788 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3789 block->bin.size, block->bin.pos,
3790 cluster->timecode, block->duration,
3791 is_keyframe, additional, block->additional_id,
3792 block->additional.size, cluster->pos,
3793 block->discard_padding);
3796 ebml_free(matroska_blockgroup, block);
3797 memset(block, 0, sizeof(*block));
3798 } else if (!matroska->num_levels) {
3799 if (!avio_feof(matroska->ctx->pb)) {
3800 avio_r8(matroska->ctx->pb);
3801 if (!avio_feof(matroska->ctx->pb)) {
3802 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3803 "end of segment.\n");
3804 return AVERROR_INVALIDDATA;
3814 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3816 MatroskaDemuxContext *matroska = s->priv_data;
3819 if (matroska->resync_pos == -1) {
3820 // This can only happen if generic seeking has been used.
3821 matroska->resync_pos = avio_tell(s->pb);
3824 while (matroska_deliver_packet(matroska, pkt)) {
3826 return (ret < 0) ? ret : AVERROR_EOF;
3827 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3828 ret = matroska_resync(matroska, matroska->resync_pos);
3834 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3835 int64_t timestamp, int flags)
3837 MatroskaDemuxContext *matroska = s->priv_data;
3838 MatroskaTrack *tracks = NULL;
3839 AVStream *st = s->streams[stream_index];
3842 /* Parse the CUES now since we need the index data to seek. */
3843 if (matroska->cues_parsing_deferred > 0) {
3844 matroska->cues_parsing_deferred = 0;
3845 matroska_parse_cues(matroska);
3848 if (!st->internal->nb_index_entries)
3850 timestamp = FFMAX(timestamp, st->internal->index_entries[0].timestamp);
3852 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3853 matroska_reset_status(matroska, 0, st->internal->index_entries[st->internal->nb_index_entries - 1].pos);
3854 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->internal->nb_index_entries - 1) {
3855 matroska_clear_queue(matroska);
3856 if (matroska_parse_cluster(matroska) < 0)
3861 matroska_clear_queue(matroska);
3862 if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->internal->nb_index_entries - 1))
3865 tracks = matroska->tracks.elem;
3866 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3867 tracks[i].audio.pkt_cnt = 0;
3868 tracks[i].audio.sub_packet_cnt = 0;
3869 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3870 tracks[i].end_timecode = 0;
3873 /* We seek to a level 1 element, so set the appropriate status. */
3874 matroska_reset_status(matroska, 0, st->internal->index_entries[index].pos);
3875 if (flags & AVSEEK_FLAG_ANY) {
3876 st->internal->skip_to_keyframe = 0;
3877 matroska->skip_to_timecode = timestamp;
3879 st->internal->skip_to_keyframe = 1;
3880 matroska->skip_to_timecode = st->internal->index_entries[index].timestamp;
3882 matroska->skip_to_keyframe = 1;
3884 ff_update_cur_dts(s, st, st->internal->index_entries[index].timestamp);
3887 // slightly hackish but allows proper fallback to
3888 // the generic seeking code.
3889 matroska_reset_status(matroska, 0, -1);
3890 matroska->resync_pos = -1;
3891 matroska_clear_queue(matroska);
3892 st->internal->skip_to_keyframe =
3893 matroska->skip_to_keyframe = 0;
3898 static int matroska_read_close(AVFormatContext *s)
3900 MatroskaDemuxContext *matroska = s->priv_data;
3901 MatroskaTrack *tracks = matroska->tracks.elem;
3904 matroska_clear_queue(matroska);
3905 av_packet_free(&matroska->pkt);
3907 for (n = 0; n < matroska->tracks.nb_elem; n++)
3908 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3909 av_freep(&tracks[n].audio.buf);
3910 ebml_free(matroska_segment, matroska);
3916 int64_t start_time_ns;
3917 int64_t end_time_ns;
3918 int64_t start_offset;
3922 /* This function searches all the Cues and returns the CueDesc corresponding to
3923 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3924 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3926 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3927 MatroskaDemuxContext *matroska = s->priv_data;
3930 int nb_index_entries = s->streams[0]->internal->nb_index_entries;
3931 AVIndexEntry *index_entries = s->streams[0]->internal->index_entries;
3932 if (ts >= matroska->duration * matroska->time_scale) return (CueDesc) {-1, -1, -1, -1};
3933 for (i = 1; i < nb_index_entries; i++) {
3934 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3935 index_entries[i].timestamp * matroska->time_scale > ts) {
3940 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3941 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3942 if (i != nb_index_entries - 1) {
3943 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3944 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3946 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3947 // FIXME: this needs special handling for files where Cues appear
3948 // before Clusters. the current logic assumes Cues appear after
3950 cue_desc.end_offset = cues_start - matroska->segment_start;
3955 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
3957 MatroskaDemuxContext *matroska = s->priv_data;
3958 uint32_t id = matroska->current_id;
3959 int64_t cluster_pos, before_pos;
3961 if (s->streams[0]->internal->nb_index_entries <= 0) return 0;
3962 // seek to the first cluster using cues.
3963 index = av_index_search_timestamp(s->streams[0], 0, 0);
3964 if (index < 0) return 0;
3965 cluster_pos = s->streams[0]->internal->index_entries[index].pos;
3966 before_pos = avio_tell(s->pb);
3968 uint64_t cluster_id, cluster_length;
3971 avio_seek(s->pb, cluster_pos, SEEK_SET);
3972 // read cluster id and length
3973 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3974 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3976 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3980 matroska_reset_status(matroska, 0, cluster_pos);
3981 matroska_clear_queue(matroska);
3982 if (matroska_parse_cluster(matroska) < 0 ||
3986 pkt = &matroska->queue->pkt;
3987 // 4 + read is the length of the cluster id and the cluster length field.
3988 cluster_pos += 4 + read + cluster_length;
3989 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3995 /* Restore the status after matroska_read_header: */
3996 matroska_reset_status(matroska, id, before_pos);
4001 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
4002 double min_buffer, double* buffer,
4003 double* sec_to_download, AVFormatContext *s,
4006 double nano_seconds_per_second = 1000000000.0;
4007 double time_sec = time_ns / nano_seconds_per_second;
4009 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
4010 int64_t end_time_ns = time_ns + time_to_search_ns;
4011 double sec_downloaded = 0.0;
4012 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
4013 if (desc_curr.start_time_ns == -1)
4015 *sec_to_download = 0.0;
4017 // Check for non cue start time.
4018 if (time_ns > desc_curr.start_time_ns) {
4019 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
4020 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
4021 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
4022 double timeToDownload = (cueBytes * 8.0) / bps;
4024 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
4025 *sec_to_download += timeToDownload;
4027 // Check if the search ends within the first cue.
4028 if (desc_curr.end_time_ns >= end_time_ns) {
4029 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4030 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4031 sec_downloaded = percent_to_sub * sec_downloaded;
4032 *sec_to_download = percent_to_sub * *sec_to_download;
4035 if ((sec_downloaded + *buffer) <= min_buffer) {
4039 // Get the next Cue.
4040 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4043 while (desc_curr.start_time_ns != -1) {
4044 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
4045 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
4046 double desc_sec = desc_ns / nano_seconds_per_second;
4047 double bits = (desc_bytes * 8.0);
4048 double time_to_download = bits / bps;
4050 sec_downloaded += desc_sec - time_to_download;
4051 *sec_to_download += time_to_download;
4053 if (desc_curr.end_time_ns >= end_time_ns) {
4054 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4055 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4056 sec_downloaded = percent_to_sub * sec_downloaded;
4057 *sec_to_download = percent_to_sub * *sec_to_download;
4059 if ((sec_downloaded + *buffer) <= min_buffer)
4064 if ((sec_downloaded + *buffer) <= min_buffer) {
4069 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4071 *buffer = *buffer + sec_downloaded;
4075 /* This function computes the bandwidth of the WebM file with the help of
4076 * buffer_size_after_time_downloaded() function. Both of these functions are
4077 * adapted from WebM Tools project and are adapted to work with FFmpeg's
4078 * Matroska parsing mechanism.
4080 * Returns the bandwidth of the file on success; -1 on error.
4082 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4084 MatroskaDemuxContext *matroska = s->priv_data;
4085 AVStream *st = s->streams[0];
4086 double bandwidth = 0.0;
4089 for (i = 0; i < st->internal->nb_index_entries; i++) {
4090 int64_t prebuffer_ns = 1000000000;
4091 int64_t time_ns = st->internal->index_entries[i].timestamp * matroska->time_scale;
4092 double nano_seconds_per_second = 1000000000.0;
4093 int64_t prebuffered_ns = time_ns + prebuffer_ns;
4094 double prebuffer_bytes = 0.0;
4095 int64_t temp_prebuffer_ns = prebuffer_ns;
4096 int64_t pre_bytes, pre_ns;
4097 double pre_sec, prebuffer, bits_per_second;
4098 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4100 // Start with the first Cue.
4101 CueDesc desc_end = desc_beg;
4103 // Figure out how much data we have downloaded for the prebuffer. This will
4104 // be used later to adjust the bits per sample to try.
4105 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4106 // Prebuffered the entire Cue.
4107 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4108 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4109 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4111 if (desc_end.start_time_ns == -1) {
4112 // The prebuffer is larger than the duration.
4113 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4115 bits_per_second = 0.0;
4117 // The prebuffer ends in the last Cue. Estimate how much data was
4119 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4120 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4121 pre_sec = pre_ns / nano_seconds_per_second;
4123 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4125 prebuffer = prebuffer_ns / nano_seconds_per_second;
4127 // Set this to 0.0 in case our prebuffer buffers the entire video.
4128 bits_per_second = 0.0;
4130 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4131 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4132 double desc_sec = desc_ns / nano_seconds_per_second;
4133 double calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4135 // Drop the bps by the percentage of bytes buffered.
4136 double percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4137 double mod_bits_per_second = calc_bits_per_second * percent;
4139 if (prebuffer < desc_sec) {
4141 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4143 // Add 1 so the bits per second should be a little bit greater than file
4145 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4146 const double min_buffer = 0.0;
4147 double buffer = prebuffer;
4148 double sec_to_download = 0.0;
4150 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4151 min_buffer, &buffer, &sec_to_download,
4155 } else if (rv == 0) {
4156 bits_per_second = (double)(bps);
4161 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4162 } while (desc_end.start_time_ns != -1);
4164 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4166 return (int64_t)bandwidth;
4169 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4171 MatroskaDemuxContext *matroska = s->priv_data;
4172 EbmlList *seekhead_list = &matroska->seekhead;
4173 MatroskaSeekhead *seekhead = seekhead_list->elem;
4175 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4179 // determine cues start and end positions
4180 for (i = 0; i < seekhead_list->nb_elem; i++)
4181 if (seekhead[i].id == MATROSKA_ID_CUES)
4184 if (i >= seekhead_list->nb_elem) return -1;
4186 before_pos = avio_tell(matroska->ctx->pb);
4187 cues_start = seekhead[i].pos + matroska->segment_start;
4188 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4189 // cues_end is computed as cues_start + cues_length + length of the
4190 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4191 // cues_end is inclusive and the above sum is reduced by 1.
4192 uint64_t cues_length, cues_id;
4194 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4195 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4196 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4197 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4200 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4202 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4203 if (cues_start == -1 || cues_end == -1) return -1;
4206 matroska_parse_cues(matroska);
4209 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4212 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4214 // if the file has cues at the start, fix up the init range so that
4215 // it does not include it
4216 if (cues_start <= init_range)
4217 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4220 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4221 if (bandwidth < 0) return -1;
4222 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4224 // check if all clusters start with key frames
4225 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4227 // store cue point timestamps as a comma separated list for checking subsegment alignment in
4228 // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4229 buf = av_malloc_array(s->streams[0]->internal->nb_index_entries, 20);
4230 if (!buf) return -1;
4232 for (i = 0; i < s->streams[0]->internal->nb_index_entries; i++) {
4233 int ret = snprintf(buf + end, 20,
4234 "%" PRId64"%s", s->streams[0]->internal->index_entries[i].timestamp,
4235 i != s->streams[0]->internal->nb_index_entries - 1 ? "," : "");
4236 if (ret <= 0 || (ret == 20 && i == s->streams[0]->internal->nb_index_entries - 1)) {
4237 av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4239 return AVERROR_INVALIDDATA;
4243 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4244 buf, AV_DICT_DONT_STRDUP_VAL);
4249 static int webm_dash_manifest_read_header(AVFormatContext *s)
4252 int ret = matroska_read_header(s);
4254 MatroskaTrack *tracks;
4255 MatroskaDemuxContext *matroska = s->priv_data;
4257 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4260 if (!matroska->tracks.nb_elem || !s->nb_streams) {
4261 av_log(s, AV_LOG_ERROR, "No track found\n");
4262 ret = AVERROR_INVALIDDATA;
4266 if (!matroska->is_live) {
4267 buf = av_asprintf("%g", matroska->duration);
4269 ret = AVERROR(ENOMEM);
4272 av_dict_set(&s->streams[0]->metadata, DURATION,
4273 buf, AV_DICT_DONT_STRDUP_VAL);
4275 // initialization range
4276 // 5 is the offset of Cluster ID.
4277 init_range = avio_tell(s->pb) - 5;
4278 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4281 // basename of the file
4282 buf = strrchr(s->url, '/');
4283 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4286 tracks = matroska->tracks.elem;
4287 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4289 // parse the cues and populate Cue related fields
4290 if (!matroska->is_live) {
4291 ret = webm_dash_manifest_cues(s, init_range);
4293 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4298 // use the bandwidth from the command line if it was provided
4299 if (matroska->bandwidth > 0) {
4300 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4301 matroska->bandwidth, 0);
4305 matroska_read_close(s);
4309 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4314 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4315 static const AVOption options[] = {
4316 { "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 },
4317 { "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 },
4321 static const AVClass webm_dash_class = {
4322 .class_name = "WebM DASH Manifest demuxer",
4323 .item_name = av_default_item_name,
4325 .version = LIBAVUTIL_VERSION_INT,
4328 AVInputFormat ff_matroska_demuxer = {
4329 .name = "matroska,webm",
4330 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4331 .extensions = "mkv,mk3d,mka,mks,webm",
4332 .priv_data_size = sizeof(MatroskaDemuxContext),
4333 .read_probe = matroska_probe,
4334 .read_header = matroska_read_header,
4335 .read_packet = matroska_read_packet,
4336 .read_close = matroska_read_close,
4337 .read_seek = matroska_read_seek,
4338 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4341 AVInputFormat ff_webm_dash_manifest_demuxer = {
4342 .name = "webm_dash_manifest",
4343 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4344 .priv_data_size = sizeof(MatroskaDemuxContext),
4345 .read_header = webm_dash_manifest_read_header,
4346 .read_packet = webm_dash_manifest_read_packet,
4347 .read_close = matroska_read_close,
4348 .priv_class = &webm_dash_class,