FFmpeg
matroskadec.c
Go to the documentation of this file.
1 /*
2  * Matroska file demuxer
3  * Copyright (c) 2003-2008 The FFmpeg Project
4  *
5  * This file is part of FFmpeg.
6  *
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.
11  *
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.
16  *
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
20  */
21 
22 /**
23  * @file
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/
29  */
30 
31 #include "config.h"
32 #include "config_components.h"
33 
34 #include <inttypes.h>
35 #include <stdio.h>
36 
37 #include "libavutil/avstring.h"
38 #include "libavutil/base64.h"
39 #include "libavutil/bprint.h"
40 #include "libavutil/dict.h"
42 #include "libavutil/display.h"
44 #include "libavutil/intfloat.h"
45 #include "libavutil/intreadwrite.h"
46 #include "libavutil/lzo.h"
48 #include "libavutil/mathematics.h"
49 #include "libavutil/opt.h"
50 #include "libavutil/pixdesc.h"
52 #include "libavutil/spherical.h"
53 
54 #include "libavcodec/bytestream.h"
55 #include "libavcodec/defs.h"
56 #include "libavcodec/flac.h"
57 #include "libavcodec/mpeg4audio.h"
59 
60 #include "avformat.h"
61 #include "avio_internal.h"
62 #include "demux.h"
63 #include "dovi_isom.h"
64 #include "internal.h"
65 #include "isom.h"
66 #include "matroska.h"
67 #include "oggdec.h"
68 /* For ff_codec_get_id(). */
69 #include "riff.h"
70 #include "rmsipr.h"
71 
72 #if CONFIG_BZLIB
73 #include <bzlib.h>
74 #endif
75 #if CONFIG_ZLIB
76 #include <zlib.h>
77 #endif
78 
79 #include "qtpalette.h"
80 
81 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
82 #define NEEDS_CHECKING 2 /* Indicates that some error checks
83  * still need to be performed */
84 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
85  * syntax level used for parsing ended. */
86 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
87  * of unkown, potentially damaged data is encountered,
88  * it is considered an error. */
89 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
90  * to this many bytes of unknown data for the
91  * SKIP_THRESHOLD check. */
92 
93 typedef enum {
99  EBML_UTF8,
100  EBML_BIN,
102  EBML_LEVEL1,
106 
107 typedef struct CountedElement {
108  union {
109  uint64_t u;
110  int64_t i;
111  double f;
112  char *s;
113  } el;
114  unsigned count;
116 
117 typedef const struct EbmlSyntax {
118  uint32_t id;
119  uint8_t type;
120  uint8_t is_counted;
122  size_t data_offset;
123  union {
124  int64_t i;
125  uint64_t u;
126  double f;
127  const char *s;
128  const struct EbmlSyntax *n;
129  } def;
130 } EbmlSyntax;
131 
132 typedef struct EbmlList {
133  int nb_elem;
134  unsigned int alloc_elem_size;
135  void *elem;
137 
138 typedef struct EbmlBin {
139  int size;
141  uint8_t *data;
142  int64_t pos;
144 
145 typedef struct Ebml {
146  uint64_t version;
147  uint64_t max_size;
148  uint64_t id_length;
149  char *doctype;
150  uint64_t doctype_version;
151 } Ebml;
152 
153 typedef struct MatroskaTrackCompression {
154  uint64_t algo;
157 
158 typedef struct MatroskaTrackEncryption {
159  uint64_t algo;
162 
163 typedef struct MatroskaTrackEncoding {
164  uint64_t scope;
165  uint64_t type;
169 
170 typedef struct MatroskaMasteringMeta {
171  double r_x;
172  double r_y;
173  double g_x;
174  double g_y;
175  double b_x;
176  double b_y;
177  double white_x;
178  double white_y;
182 
183 typedef struct MatroskaTrackVideoColor {
186  uint64_t chroma_sub_horz;
187  uint64_t chroma_sub_vert;
188  uint64_t cb_sub_horz;
189  uint64_t cb_sub_vert;
192  uint64_t range;
193  uint64_t transfer_characteristics;
194  uint64_t primaries;
195  uint64_t max_cll;
196  uint64_t max_fall;
199 
200 typedef struct MatroskaTrackVideoProjection {
201  uint64_t type;
202  EbmlBin private;
203  double yaw;
204  double pitch;
205  double roll;
207 
208 typedef struct MatroskaTrackVideo {
209  double frame_rate;
210  uint64_t display_width;
211  uint64_t display_height;
212  uint64_t pixel_width;
213  uint64_t pixel_height;
215  uint64_t display_unit;
216  uint64_t interlaced;
217  uint64_t field_order;
218  uint64_t stereo_mode;
219  uint64_t alpha_mode;
223 
224 typedef struct MatroskaTrackAudio {
225  double samplerate;
227  uint64_t bitdepth;
228  uint64_t channels;
229 
230  /* real audio header (extracted from extradata) */
233  int frame_size;
234  int sub_packet_size;
236  int pkt_cnt;
237  uint64_t buf_timecode;
238  uint8_t *buf;
240 
241 typedef struct MatroskaTrackPlane {
242  uint64_t uid;
243  uint64_t type;
245 
246 typedef struct MatroskaTrackOperation {
249 
250 typedef struct MatroskaBlockAdditionMapping {
251  uint64_t value;
252  char *name;
253  uint64_t type;
256 
257 typedef struct MatroskaTrack {
258  uint64_t num;
259  uint64_t uid;
260  uint64_t type;
261  char *name;
262  char *codec_id;
264  char *language;
265  double time_scale;
267  uint64_t flag_default;
268  uint64_t flag_forced;
269  uint64_t flag_comment;
274  uint64_t seek_preroll;
279  uint64_t codec_delay;
281 
282  AVStream *stream;
283  int64_t end_timecode;
285  int needs_decoding;
286  uint64_t max_block_additional_id;
288 
292 
293 typedef struct MatroskaAttachment {
294  uint64_t uid;
295  char *filename;
296  char *description;
297  char *mime;
299 
302 
303 typedef struct MatroskaChapter {
304  uint64_t start;
305  uint64_t end;
306  uint64_t uid;
307  char *title;
308 
311 
312 typedef struct MatroskaIndexPos {
313  uint64_t track;
314  uint64_t pos;
316 
317 typedef struct MatroskaIndex {
318  uint64_t time;
321 
322 typedef struct MatroskaTag {
323  char *name;
324  char *string;
325  char *lang;
326  uint64_t def;
329 
330 typedef struct MatroskaTagTarget {
331  char *type;
332  uint64_t typevalue;
333  uint64_t trackuid;
334  uint64_t chapteruid;
335  uint64_t attachuid;
337 
338 typedef struct MatroskaTags {
340  EbmlList tag;
341 } MatroskaTags;
342 
343 typedef struct MatroskaSeekhead {
344  uint64_t id;
345  uint64_t pos;
347 
348 typedef struct MatroskaLevel {
349  uint64_t start;
350  uint64_t length;
351 } MatroskaLevel;
352 
353 typedef struct MatroskaBlockMore {
354  uint64_t additional_id;
357 
358 typedef struct MatroskaBlock {
359  uint64_t duration;
361  uint64_t non_simple;
365 } MatroskaBlock;
366 
367 typedef struct MatroskaCluster {
369  uint64_t timecode;
370  int64_t pos;
372 
373 typedef struct MatroskaLevel1Element {
374  int64_t pos;
375  uint32_t id;
376  int parsed;
378 
379 typedef struct MatroskaDemuxContext {
380  const AVClass *class;
382 
383  /* EBML stuff */
386  uint32_t current_id;
387  int64_t resync_pos;
389 
390  uint64_t time_scale;
391  double duration;
392  char *title;
393  char *muxingapp;
398  EbmlList index;
399  EbmlList tags;
401 
402  /* byte position of the segment inside the stream */
403  int64_t segment_start;
404 
405  /* This packet coincides with FFFormatContext.parse_pkt
406  * and is not owned by us. */
407  AVPacket *pkt;
408 
409  /* the packet queue */
411 
412  int done;
413 
414  /* What to skip before effectively reading a packet. */
415  int skip_to_keyframe;
417 
418  /* File has a CUES element, but we defer parsing until it is needed. */
420 
421  /* Level1 elements and whether they were read yet */
423  int num_level1_elems;
424 
426 
427  int is_webm;
428 
429  /* WebM DASH Manifest live flag */
430  int is_live;
431 
432  /* Bandwidth value for WebM DASH Manifest */
433  int bandwidth;
435 
436 #define CHILD_OF(parent) { .def = { .n = parent } }
437 
438 // The following forward declarations need their size because
439 // a tentative definition with internal linkage must not be an
440 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
441 // Removing the sizes breaks MSVC.
448 
449 static EbmlSyntax ebml_header[] = {
450  { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
451  { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, 0, offsetof(Ebml, max_size), { .u = 8 } },
452  { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, 0, offsetof(Ebml, id_length), { .u = 4 } },
453  { EBML_ID_DOCTYPE, EBML_STR, 0, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
454  { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
458 };
459 
461  { EBML_ID_HEADER, EBML_NEST, 0, 0, 0, { .n = ebml_header } },
463  { 0 }
464 };
465 
466 static EbmlSyntax matroska_info[] = {
467  { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
469  { MATROSKA_ID_TITLE, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, title) },
471  { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, muxingapp) },
472  { MATROSKA_ID_DATEUTC, EBML_BIN, 0, 0, offsetof(MatroskaDemuxContext, date_utc) },
475 };
476 
484  { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_x) },
485  { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_y) },
486  { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 1, 0, offsetof(MatroskaMasteringMeta, min_luminance) },
487  { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, max_luminance) },
489 };
490 
492  { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
493  { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u = 0 } },
494  { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz) },
495  { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert) },
496  { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz) },
497  { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert) },
504  { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_fall) },
507 };
508 
512  { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f = 0.0 } },
513  { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f = 0.0 } },
514  { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f = 0.0 } },
516 };
517 
518 static EbmlSyntax matroska_track_video[] = {
519  { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideo, frame_rate) },
520  { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
521  { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
522  { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_width) },
523  { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_height) },
524  { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideo, color_space) },
525  { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, alpha_mode), { .u = 0 } },
538 };
539 
540 static EbmlSyntax matroska_track_audio[] = {
541  { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
542  { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
543  { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, bitdepth) },
544  { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
546 };
547 
552 };
553 
555  { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
563 };
565  { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
566  { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
571 };
572 
576 };
577 
578 static EbmlSyntax matroska_track_plane[] = {
582 };
583 
585  { MATROSKA_ID_TRACKPLANE, EBML_NEST, 0, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
587 };
588 
592 };
593 
600 };
601 
602 static EbmlSyntax matroska_track[] = {
603  { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, 0, offsetof(MatroskaTrack, num) },
604  { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTrack, name) },
605  { MATROSKA_ID_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, uid) },
606  { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrack, type) },
607  { MATROSKA_ID_CODECID, EBML_STR, 0, 0, offsetof(MatroskaTrack, codec_id) },
608  { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrack, codec_priv) },
609  { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, codec_delay), { .u = 0 } },
610  { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
611  { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, 0, offsetof(MatroskaTrack, default_duration) },
612  { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
613  { MATROSKA_ID_TRACKFLAGCOMMENTARY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_comment), { .u = 0 } },
614  { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
615  { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
616  { MATROSKA_ID_TRACKFLAGHEARINGIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_hearingimpaired), { .u = 0 } },
617  { MATROSKA_ID_TRACKFLAGVISUALIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_visualimpaired), { .u = 0 } },
618  { MATROSKA_ID_TRACKFLAGTEXTDESCRIPTIONS, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_textdescriptions), { .u = 0 } },
619  { MATROSKA_ID_TRACKFLAGORIGINAL, EBML_UINT, 1, 0, offsetof(MatroskaTrack, flag_original), {.u = 0 } },
621  { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
622  { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
624  { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, max_block_additional_id), { .u = 0 } },
625  { MATROSKA_ID_TRACKBLKADDMAPPING, EBML_NEST, 0, sizeof(MatroskaBlockAdditionMapping), offsetof(MatroskaTrack, block_addition_mappings), { .n = matroska_block_addition_mapping } },
626  { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, 0, offsetof(MatroskaTrack, seek_preroll), { .u = 0 } },
636 };
637 
638 static EbmlSyntax matroska_tracks[] = {
639  { MATROSKA_ID_TRACKENTRY, EBML_NEST, 0, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
641 };
642 
643 static EbmlSyntax matroska_attachment[] = {
644  { MATROSKA_ID_FILEUID, EBML_UINT, 0, 0, offsetof(MatroskaAttachment, uid) },
645  { MATROSKA_ID_FILENAME, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, filename) },
646  { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, 0, offsetof(MatroskaAttachment, mime) },
647  { MATROSKA_ID_FILEDATA, EBML_BIN, 0, 0, offsetof(MatroskaAttachment, bin) },
650 };
651 
652 static EbmlSyntax matroska_attachments[] = {
653  { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, 0, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
655 };
656 
658  { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaChapter, title) },
662 };
663 
665  { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
666  { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
667  { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaChapter, uid) },
674 };
675 
676 static EbmlSyntax matroska_chapter[] = {
677  { MATROSKA_ID_CHAPTERATOM, EBML_NEST, 0, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
683 };
684 
685 static EbmlSyntax matroska_chapters[] = {
686  { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter } },
688 };
689 
690 static EbmlSyntax matroska_index_pos[] = {
691  { MATROSKA_ID_CUETRACK, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, track) },
697 };
698 
699 static EbmlSyntax matroska_index_entry[] = {
700  { MATROSKA_ID_CUETIME, EBML_UINT, 0, 0, offsetof(MatroskaIndex, time) },
703 };
704 
705 static EbmlSyntax matroska_index[] = {
708 };
709 
710 static EbmlSyntax matroska_simpletag[] = {
711  { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTag, name) },
712  { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaTag, string) },
713  { MATROSKA_ID_TAGLANG, EBML_STR, 0, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
714  { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
715  { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
716  { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
718 };
719 
720 static EbmlSyntax matroska_tagtargets[] = {
722  { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
723  { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, trackuid), { .u = 0 } },
724  { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, chapteruid), { .u = 0 } },
725  { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, attachuid), { .u = 0 } },
727 };
728 
729 static EbmlSyntax matroska_tag[] = {
730  { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
731  { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
733 };
734 
735 static EbmlSyntax matroska_tags[] = {
736  { MATROSKA_ID_TAG, EBML_NEST, 0, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
738 };
739 
741  { MATROSKA_ID_SEEKID, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, id) },
742  { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
744 };
745 
746 static EbmlSyntax matroska_seekhead[] = {
747  { MATROSKA_ID_SEEKENTRY, EBML_NEST, 0, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
749 };
750 
751 static EbmlSyntax matroska_segment[] = {
753  { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, 0, { .n = matroska_info } },
754  { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tracks } },
756  { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_chapters } },
757  { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, 0, { .n = matroska_index } },
758  { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tags } },
759  { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, 0, { .n = matroska_seekhead } },
760  { 0 } /* We don't want to go back to level 0, so don't add the parent. */
761 };
762 
763 static EbmlSyntax matroska_segments[] = {
764  { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, 0, { .n = matroska_segment } },
765  { 0 }
766 };
767 
768 static EbmlSyntax matroska_blockmore[] = {
769  { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, 0, offsetof(MatroskaBlockMore,additional_id), { .u = MATROSKA_BLOCK_ADD_ID_OPAQUE } },
770  { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, 0, offsetof(MatroskaBlockMore,additional) },
772 };
773 
775  { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, sizeof(MatroskaBlockMore), offsetof(MatroskaBlock, blockmore), { .n = matroska_blockmore } },
777 };
778 
779 static EbmlSyntax matroska_blockgroup[] = {
780  { MATROSKA_ID_BLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
783  { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, 0, offsetof(MatroskaBlock, discard_padding) },
784  { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 1, 0, offsetof(MatroskaBlock, reference) },
786  { 1, EBML_UINT, 0, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
788 };
789 
790 // The following array contains SimpleBlock and BlockGroup twice
791 // in order to reuse the other values for matroska_cluster_enter.
793  { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
794  { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, 0, { .n = matroska_blockgroup } },
795  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, 0, offsetof(MatroskaCluster, timecode) },
801 };
802 
804  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, 0, { .n = &matroska_cluster_parsing[2] } },
805  { 0 }
806 };
807 #undef CHILD_OF
808 
809 static const CodecMime mkv_image_mime_tags[] = {
810  {"image/gif" , AV_CODEC_ID_GIF},
811  {"image/jpeg" , AV_CODEC_ID_MJPEG},
812  {"image/png" , AV_CODEC_ID_PNG},
813  {"image/tiff" , AV_CODEC_ID_TIFF},
814 
815  {"" , AV_CODEC_ID_NONE}
816 };
817 
818 static const CodecMime mkv_mime_tags[] = {
819  {"application/x-truetype-font", AV_CODEC_ID_TTF},
820  {"application/x-font" , AV_CODEC_ID_TTF},
821  {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
822  {"binary" , AV_CODEC_ID_BIN_DATA},
823 
824  {"" , AV_CODEC_ID_NONE}
825 };
826 
828  "left",
829  "right",
830  "background",
831 };
832 
833 static const char *const matroska_doctypes[] = { "matroska", "webm" };
834 
835 /*
836  * This function prepares the status for parsing of level 1 elements.
837  */
838 static int matroska_reset_status(MatroskaDemuxContext *matroska,
839  uint32_t id, int64_t position)
840 {
841  int64_t err = 0;
842  if (position >= 0) {
843  err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
844  if (err > 0)
845  err = 0;
846  } else
847  position = avio_tell(matroska->ctx->pb);
848 
849  matroska->current_id = id;
850  matroska->num_levels = 1;
851  matroska->unknown_count = 0;
852  matroska->resync_pos = position;
853  if (id)
854  matroska->resync_pos -= (av_log2(id) + 7) / 8;
855 
856  return err;
857 }
858 
859 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
860 {
861  AVIOContext *pb = matroska->ctx->pb;
862  uint32_t id;
863 
864  /* Try to seek to the last position to resync from. If this doesn't work,
865  * we resync from the earliest position available: The start of the buffer. */
866  if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
867  av_log(matroska->ctx, AV_LOG_WARNING,
868  "Seek to desired resync point failed. Seeking to "
869  "earliest point available instead.\n");
870  avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
871  last_pos + 1), SEEK_SET);
872  }
873 
874  id = avio_rb32(pb);
875 
876  // try to find a toplevel element
877  while (!avio_feof(pb)) {
878  if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
879  id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
881  id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
882  /* Prepare the context for parsing of a level 1 element. */
883  matroska_reset_status(matroska, id, -1);
884  /* Given that we are here means that an error has occurred,
885  * so treat the segment as unknown length in order not to
886  * discard valid data that happens to be beyond the designated
887  * end of the segment. */
888  matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
889  return 0;
890  }
891  id = (id << 8) | avio_r8(pb);
892  }
893 
894  matroska->done = 1;
895  return pb->error ? pb->error : AVERROR_EOF;
896 }
897 
898 /*
899  * Read: an "EBML number", which is defined as a variable-length
900  * array of bytes. The first byte indicates the length by giving a
901  * number of 0-bits followed by a one. The position of the first
902  * "one" bit inside the first byte indicates the length of this
903  * number.
904  * Returns: number of bytes read, < 0 on error
905  */
906 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
907  int max_size, uint64_t *number, int eof_forbidden)
908 {
909  int read, n = 1;
910  uint64_t total;
911  int64_t pos;
912 
913  /* The first byte tells us the length in bytes - except when it is zero. */
914  total = avio_r8(pb);
915  if (pb->eof_reached)
916  goto err;
917 
918  /* get the length of the EBML number */
919  read = 8 - ff_log2_tab[total];
920 
921  if (!total || read > max_size) {
922  pos = avio_tell(pb) - 1;
923  if (!total) {
924  av_log(matroska->ctx, AV_LOG_ERROR,
925  "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
926  "of an EBML number\n", pos, pos);
927  } else {
928  av_log(matroska->ctx, AV_LOG_ERROR,
929  "Length %d indicated by an EBML number's first byte 0x%02x "
930  "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
931  read, (uint8_t) total, pos, pos, max_size);
932  }
933  return AVERROR_INVALIDDATA;
934  }
935 
936  /* read out length */
937  total ^= 1 << ff_log2_tab[total];
938  while (n++ < read)
939  total = (total << 8) | avio_r8(pb);
940 
941  if (pb->eof_reached) {
942  eof_forbidden = 1;
943  goto err;
944  }
945 
946  *number = total;
947 
948  return read;
949 
950 err:
951  pos = avio_tell(pb);
952  if (pb->error) {
953  av_log(matroska->ctx, AV_LOG_ERROR,
954  "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
955  pos, pos);
956  return pb->error;
957  }
958  if (eof_forbidden) {
959  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
960  "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
961  return AVERROR(EIO);
962  }
963  return AVERROR_EOF;
964 }
965 
966 /**
967  * Read a EBML length value.
968  * This needs special handling for the "unknown length" case which has multiple
969  * encodings.
970  */
971 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
972  uint64_t *number)
973 {
974  int res = ebml_read_num(matroska, pb, 8, number, 1);
975  if (res > 0 && *number + 1 == 1ULL << (7 * res))
976  *number = EBML_UNKNOWN_LENGTH;
977  return res;
978 }
979 
980 /*
981  * Read the next element as an unsigned int.
982  * Returns NEEDS_CHECKING unless size == 0.
983  */
984 static int ebml_read_uint(AVIOContext *pb, int size,
985  uint64_t default_value, uint64_t *num)
986 {
987  int n = 0;
988 
989  if (size == 0) {
990  *num = default_value;
991  return 0;
992  }
993  /* big-endian ordering; build up number */
994  *num = 0;
995  while (n++ < size)
996  *num = (*num << 8) | avio_r8(pb);
997 
998  return NEEDS_CHECKING;
999 }
1000 
1001 /*
1002  * Read the next element as a signed int.
1003  * Returns NEEDS_CHECKING unless size == 0.
1004  */
1005 static int ebml_read_sint(AVIOContext *pb, int size,
1006  int64_t default_value, int64_t *num)
1007 {
1008  int n = 1;
1009 
1010  if (size == 0) {
1011  *num = default_value;
1012  return 0;
1013  } else {
1014  *num = sign_extend(avio_r8(pb), 8);
1015 
1016  /* big-endian ordering; build up number */
1017  while (n++ < size)
1018  *num = ((uint64_t)*num << 8) | avio_r8(pb);
1019  }
1020 
1021  return NEEDS_CHECKING;
1023 
1024 /*
1025  * Read the next element as a float.
1026  * Returns 0 if size == 0, NEEDS_CHECKING or < 0 on obvious failure.
1027  */
1028 static int ebml_read_float(AVIOContext *pb, int size,
1029  double default_value, double *num)
1030 {
1031  if (size == 0) {
1032  *num = default_value;
1033  return 0;
1034  } else if (size == 4) {
1035  *num = av_int2float(avio_rb32(pb));
1036  } else if (size == 8) {
1037  *num = av_int2double(avio_rb64(pb));
1038  } else
1039  return AVERROR_INVALIDDATA;
1040 
1041  return NEEDS_CHECKING;
1043 
1044 /*
1045  * Read the next element as an ASCII string.
1046  * 0 is success, < 0 or NEEDS_CHECKING is failure.
1047  */
1048 static int ebml_read_ascii(AVIOContext *pb, int size,
1049  const char *default_value, char **str)
1050 {
1051  char *res;
1052  int ret;
1053 
1054  if (size == 0 && default_value) {
1055  res = av_strdup(default_value);
1056  if (!res)
1057  return AVERROR(ENOMEM);
1058  } else {
1059  /* EBML strings are usually not 0-terminated, so we allocate one
1060  * byte more, read the string and NUL-terminate it ourselves. */
1061  if (!(res = av_malloc(size + 1)))
1062  return AVERROR(ENOMEM);
1063  if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
1064  av_free(res);
1065  return ret < 0 ? ret : NEEDS_CHECKING;
1066  }
1067  (res)[size] = '\0';
1068  }
1069  av_free(*str);
1070  *str = res;
1071 
1072  return 0;
1074 
1075 /*
1076  * Read the next element as binary data.
1077  * 0 is success, < 0 or NEEDS_CHECKING is failure.
1078  */
1079 static int ebml_read_binary(AVIOContext *pb, int length,
1080  int64_t pos, EbmlBin *bin)
1081 {
1082  int ret;
1083 
1085  if (ret < 0)
1086  return ret;
1087  memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1088 
1089  bin->data = bin->buf->data;
1090  bin->size = length;
1091  bin->pos = pos;
1092  if ((ret = avio_read(pb, bin->data, length)) != length) {
1093  av_buffer_unref(&bin->buf);
1094  bin->data = NULL;
1095  bin->size = 0;
1096  return ret < 0 ? ret : NEEDS_CHECKING;
1097  }
1098 
1099  return 0;
1100 }
1102 /*
1103  * Read the next element, but only the header. The contents
1104  * are supposed to be sub-elements which can be read separately.
1105  * 0 is success, < 0 is failure.
1106  */
1107 static int ebml_read_master(MatroskaDemuxContext *matroska,
1108  uint64_t length, int64_t pos)
1109 {
1111 
1112  if (matroska->num_levels >= EBML_MAX_DEPTH) {
1113  av_log(matroska->ctx, AV_LOG_ERROR,
1114  "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1115  return AVERROR(ENOSYS);
1116  }
1117 
1118  level = &matroska->levels[matroska->num_levels++];
1119  level->start = pos;
1120  level->length = length;
1121 
1122  return 0;
1124 
1125 /*
1126  * Read a signed "EBML number"
1127  * Return: number of bytes processed, < 0 on error
1128  */
1129 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1130  AVIOContext *pb, int64_t *num)
1131 {
1132  uint64_t unum;
1133  int res;
1134 
1135  /* read as unsigned number first */
1136  if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1137  return res;
1138 
1139  /* make signed (weird way) */
1140  *num = unum - ((1LL << (7 * res - 1)) - 1);
1141 
1142  return res;
1143 }
1144 
1145 static int ebml_parse(MatroskaDemuxContext *matroska,
1146  EbmlSyntax *syntax, void *data);
1147 
1148 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1149 {
1150  int i;
1151 
1152  // Whoever touches this should be aware of the duplication
1153  // existing in matroska_cluster_parsing.
1154  for (i = 0; syntax[i].id; i++)
1155  if (id == syntax[i].id)
1156  break;
1157 
1158  return &syntax[i];
1159 }
1160 
1161 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1162  void *data)
1163 {
1164  int res;
1165 
1166  if (data) {
1167  for (int i = 0; syntax[i].id; i++) {
1168  void *dst = (char *)data + syntax[i].data_offset;
1169  switch (syntax[i].type) {
1170  case EBML_UINT:
1171  *(uint64_t *)dst = syntax[i].def.u;
1172  break;
1173  case EBML_SINT:
1174  *(int64_t *) dst = syntax[i].def.i;
1175  break;
1176  case EBML_FLOAT:
1177  *(double *) dst = syntax[i].def.f;
1178  break;
1179  case EBML_STR:
1180  case EBML_UTF8:
1181  // the default may be NULL
1182  if (syntax[i].def.s) {
1183  *(char**)dst = av_strdup(syntax[i].def.s);
1184  if (!*(char**)dst)
1185  return AVERROR(ENOMEM);
1186  }
1187  break;
1188  }
1189  }
1190 
1191  if (!matroska->levels[matroska->num_levels - 1].length) {
1192  matroska->num_levels--;
1193  return 0;
1194  }
1195  }
1196 
1197  do {
1198  res = ebml_parse(matroska, syntax, data);
1199  } while (!res);
1200 
1201  return res == LEVEL_ENDED ? 0 : res;
1202 }
1203 
1204 static int is_ebml_id_valid(uint32_t id)
1205 {
1206  // Due to endian nonsense in Matroska, the highest byte with any bits set
1207  // will contain the leading length bit. This bit in turn identifies the
1208  // total byte length of the element by its position within the byte.
1209  unsigned int bits = av_log2(id);
1210  return id && (bits + 7) / 8 == (8 - bits % 8);
1212 
1213 /*
1214  * Allocate and return the entry for the level1 element with the given ID. If
1215  * an entry already exists, return the existing entry.
1216  */
1218  uint32_t id, int64_t pos)
1219 {
1220  int i;
1221  MatroskaLevel1Element *elem;
1222 
1223  if (!is_ebml_id_valid(id))
1224  return NULL;
1225 
1226  // Some files link to all clusters; useless.
1227  if (id == MATROSKA_ID_CLUSTER)
1228  return NULL;
1229 
1230  // There can be multiple SeekHeads and Tags.
1231  for (i = 0; i < matroska->num_level1_elems; i++) {
1232  if (matroska->level1_elems[i].id == id) {
1233  if (matroska->level1_elems[i].pos == pos ||
1234  id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1235  return &matroska->level1_elems[i];
1236  }
1237  }
1238 
1239  // Only a completely broken file would have more elements.
1240  if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1241  av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1242  return NULL;
1243  }
1244 
1245  elem = &matroska->level1_elems[matroska->num_level1_elems++];
1246  *elem = (MatroskaLevel1Element){.id = id};
1247 
1248  return elem;
1249 }
1250 
1251 static int ebml_parse(MatroskaDemuxContext *matroska,
1252  EbmlSyntax *syntax, void *data)
1253 {
1254  static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1255  // Forbid unknown-length EBML_NONE elements.
1257  [EBML_UINT] = 8,
1258  [EBML_SINT] = 8,
1259  [EBML_FLOAT] = 8,
1260  // max. 16 MB for strings
1261  [EBML_STR] = 0x1000000,
1262  [EBML_UTF8] = 0x1000000,
1263  // max. 256 MB for binary data
1264  [EBML_BIN] = 0x10000000,
1265  // no limits for anything else
1266  };
1267  AVIOContext *pb = matroska->ctx->pb;
1268  uint32_t id;
1269  uint64_t length;
1270  int64_t pos = avio_tell(pb), pos_alt;
1271  int res, update_pos = 1, level_check;
1272  MatroskaLevel1Element *level1_elem;
1273  MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1274 
1275  if (!matroska->current_id) {
1276  uint64_t id;
1277  res = ebml_read_num(matroska, pb, 4, &id, 0);
1278  if (res < 0) {
1279  if (pb->eof_reached && res == AVERROR_EOF) {
1280  if (matroska->is_live)
1281  // in live mode, finish parsing if EOF is reached.
1282  return 1;
1283  if (level && pos == avio_tell(pb)) {
1284  if (level->length == EBML_UNKNOWN_LENGTH) {
1285  // Unknown-length levels automatically end at EOF.
1286  matroska->num_levels--;
1287  return LEVEL_ENDED;
1288  } else {
1289  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1290  "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1291  }
1292  }
1293  }
1294  return res;
1295  }
1296  matroska->current_id = id | 1 << 7 * res;
1297  pos_alt = pos + res;
1298  } else {
1299  pos_alt = pos;
1300  pos -= (av_log2(matroska->current_id) + 7) / 8;
1301  }
1302 
1303  id = matroska->current_id;
1304 
1305  syntax = ebml_parse_id(syntax, id);
1306  if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1307  if (level && level->length == EBML_UNKNOWN_LENGTH) {
1308  // Unknown-length levels end when an element from an upper level
1309  // in the hierarchy is encountered.
1310  while (syntax->def.n) {
1311  syntax = ebml_parse_id(syntax->def.n, id);
1312  if (syntax->id) {
1313  matroska->num_levels--;
1314  return LEVEL_ENDED;
1315  }
1316  };
1317  }
1318 
1319  av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1320  "%"PRId64"\n", id, pos);
1321  update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1322  }
1323 
1324  if (data) {
1325  data = (char *) data + syntax->data_offset;
1326  if (syntax->list_elem_size) {
1327  EbmlList *list = data;
1328  void *newelem;
1329 
1330  if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1331  return AVERROR(ENOMEM);
1332  newelem = av_fast_realloc(list->elem,
1333  &list->alloc_elem_size,
1334  (list->nb_elem + 1) * syntax->list_elem_size);
1335  if (!newelem)
1336  return AVERROR(ENOMEM);
1337  list->elem = newelem;
1338  data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1339  memset(data, 0, syntax->list_elem_size);
1340  list->nb_elem++;
1341  }
1342  }
1343 
1344  if (syntax->type != EBML_STOP) {
1345  matroska->current_id = 0;
1346  if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1347  return res;
1348 
1349  pos_alt += res;
1350 
1351  if (matroska->num_levels > 0) {
1352  if (length != EBML_UNKNOWN_LENGTH &&
1353  level->length != EBML_UNKNOWN_LENGTH) {
1354  uint64_t elem_end = pos_alt + length,
1355  level_end = level->start + level->length;
1356 
1357  if (elem_end < level_end) {
1358  level_check = 0;
1359  } else if (elem_end == level_end) {
1360  level_check = LEVEL_ENDED;
1361  } else {
1362  av_log(matroska->ctx, AV_LOG_ERROR,
1363  "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1364  "containing master element ending at 0x%"PRIx64"\n",
1365  pos, elem_end, level_end);
1366  return AVERROR_INVALIDDATA;
1367  }
1368  } else if (length != EBML_UNKNOWN_LENGTH) {
1369  level_check = 0;
1370  } else if (level->length != EBML_UNKNOWN_LENGTH) {
1371  av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1372  "at 0x%"PRIx64" inside parent with finite size\n", pos);
1373  return AVERROR_INVALIDDATA;
1374  } else {
1375  level_check = 0;
1376  if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1377  || syntax->type == EBML_NEST)) {
1378  // According to the current specifications only clusters and
1379  // segments are allowed to be unknown-length. We also accept
1380  // other unknown-length master elements.
1381  av_log(matroska->ctx, AV_LOG_WARNING,
1382  "Found unknown-length element 0x%"PRIX32" other than "
1383  "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1384  "parsing will nevertheless be attempted.\n", id, pos);
1385  update_pos = -1;
1386  }
1387  }
1388  } else
1389  level_check = 0;
1390 
1391  if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1392  if (length != EBML_UNKNOWN_LENGTH) {
1393  av_log(matroska->ctx, AV_LOG_ERROR,
1394  "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1395  "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1396  length, max_lengths[syntax->type], id, pos);
1397  } else if (syntax->type != EBML_NONE) {
1398  av_log(matroska->ctx, AV_LOG_ERROR,
1399  "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1400  "unknown length, yet the length of an element of its "
1401  "type must be known.\n", id, pos);
1402  } else {
1403  av_log(matroska->ctx, AV_LOG_ERROR,
1404  "Found unknown-length element with ID 0x%"PRIX32" at "
1405  "pos. 0x%"PRIx64" for which no syntax for parsing is "
1406  "available.\n", id, pos);
1407  }
1408  return AVERROR_INVALIDDATA;
1409  }
1410 
1411  if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1412  // Loosing sync will likely manifest itself as encountering unknown
1413  // elements which are not reliably distinguishable from elements
1414  // belonging to future extensions of the format.
1415  // We use a heuristic to detect such situations: If the current
1416  // element is not expected at the current syntax level and there
1417  // were only a few unknown elements in a row, then the element is
1418  // skipped or considered defective based upon the length of the
1419  // current element (i.e. how much would be skipped); if there were
1420  // more than a few skipped elements in a row and skipping the current
1421  // element would lead us more than SKIP_THRESHOLD away from the last
1422  // known good position, then it is inferred that an error occurred.
1423  // The dependency on the number of unknown elements in a row exists
1424  // because the distance to the last known good position is
1425  // automatically big if the last parsed element was big.
1426  // In both cases, each unknown element is considered equivalent to
1427  // UNKNOWN_EQUIV of skipped bytes for the check.
1428  // The whole check is only done for non-seekable output, because
1429  // in this situation skipped data can't simply be rechecked later.
1430  // This is especially important when using unkown length elements
1431  // as the check for whether a child exceeds its containing master
1432  // element is not effective in this situation.
1433  if (update_pos) {
1434  matroska->unknown_count = 0;
1435  } else {
1436  int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1437 
1438  if (matroska->unknown_count > 3)
1439  dist += pos_alt - matroska->resync_pos;
1440 
1441  if (dist > SKIP_THRESHOLD) {
1442  av_log(matroska->ctx, AV_LOG_ERROR,
1443  "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1444  "length 0x%"PRIx64" considered as invalid data. Last "
1445  "known good position 0x%"PRIx64", %d unknown elements"
1446  " in a row\n", id, pos, length, matroska->resync_pos,
1447  matroska->unknown_count);
1448  return AVERROR_INVALIDDATA;
1449  }
1450  }
1451  }
1452 
1453  if (update_pos > 0) {
1454  // We have found an element that is allowed at this place
1455  // in the hierarchy and it passed all checks, so treat the beginning
1456  // of the element as the "last known good" position.
1457  matroska->resync_pos = pos;
1458  }
1459 
1460  if (!data && length != EBML_UNKNOWN_LENGTH)
1461  goto skip;
1462  }
1463 
1464  switch (syntax->type) {
1465  case EBML_UINT:
1466  res = ebml_read_uint(pb, length, syntax->def.u, data);
1467  break;
1468  case EBML_SINT:
1469  res = ebml_read_sint(pb, length, syntax->def.i, data);
1470  break;
1471  case EBML_FLOAT:
1472  res = ebml_read_float(pb, length, syntax->def.f, data);
1473  break;
1474  case EBML_STR:
1475  case EBML_UTF8:
1476  res = ebml_read_ascii(pb, length, syntax->def.s, data);
1477  break;
1478  case EBML_BIN:
1479  res = ebml_read_binary(pb, length, pos_alt, data);
1480  break;
1481  case EBML_LEVEL1:
1482  case EBML_NEST:
1483  if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1484  return res;
1485  if (id == MATROSKA_ID_SEGMENT)
1486  matroska->segment_start = pos_alt;
1487  if (id == MATROSKA_ID_CUES)
1488  matroska->cues_parsing_deferred = 0;
1489  if (syntax->type == EBML_LEVEL1 &&
1490  (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1491  if (!level1_elem->pos) {
1492  // Zero is not a valid position for a level 1 element.
1493  level1_elem->pos = pos;
1494  } else if (level1_elem->pos != pos)
1495  av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1496  level1_elem->parsed = 1;
1497  }
1498  if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1499  return res;
1500  break;
1501  case EBML_STOP:
1502  return 1;
1503  skip:
1504  default:
1505  if (length) {
1506  int64_t res2;
1507  if (ffio_limit(pb, length) != length) {
1508  // ffio_limit emits its own error message,
1509  // so we don't have to.
1510  return AVERROR(EIO);
1511  }
1512  if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1513  // avio_skip might take us past EOF. We check for this
1514  // by skipping only length - 1 bytes, reading a byte and
1515  // checking the error flags. This is done in order to check
1516  // that the element has been properly skipped even when
1517  // no filesize (that ffio_limit relies on) is available.
1518  avio_r8(pb);
1519  res = NEEDS_CHECKING;
1520  } else
1521  res = res2;
1522  } else
1523  res = 0;
1524  }
1525  if (res) {
1526  if (res == NEEDS_CHECKING) {
1527  if (pb->eof_reached) {
1528  if (pb->error)
1529  res = pb->error;
1530  else
1531  res = AVERROR_EOF;
1532  } else
1533  goto level_check;
1534  }
1535 
1536  if (res == AVERROR_INVALIDDATA)
1537  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1538  else if (res == AVERROR(EIO))
1539  av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1540  else if (res == AVERROR_EOF) {
1541  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1542  res = AVERROR(EIO);
1543  }
1544 
1545  return res;
1546  }
1547 
1548 level_check:
1549  if (syntax->is_counted && data) {
1550  CountedElement *elem = data;
1551  if (elem->count != UINT_MAX)
1552  elem->count++;
1553  }
1554 
1555  if (level_check == LEVEL_ENDED && matroska->num_levels) {
1556  level = &matroska->levels[matroska->num_levels - 1];
1557  pos = avio_tell(pb);
1558 
1559  // Given that pos >= level->start no check for
1560  // level->length != EBML_UNKNOWN_LENGTH is necessary.
1561  while (matroska->num_levels && pos == level->start + level->length) {
1562  matroska->num_levels--;
1563  level--;
1564  }
1565  }
1566 
1567  return level_check;
1568 }
1569 
1570 static void ebml_free(EbmlSyntax *syntax, void *data)
1571 {
1572  int i, j;
1573  for (i = 0; syntax[i].id; i++) {
1574  void *data_off = (char *) data + syntax[i].data_offset;
1575  switch (syntax[i].type) {
1576  case EBML_STR:
1577  case EBML_UTF8:
1578  av_freep(data_off);
1579  break;
1580  case EBML_BIN:
1581  av_buffer_unref(&((EbmlBin *) data_off)->buf);
1582  break;
1583  case EBML_LEVEL1:
1584  case EBML_NEST:
1585  if (syntax[i].list_elem_size) {
1586  EbmlList *list = data_off;
1587  char *ptr = list->elem;
1588  for (j = 0; j < list->nb_elem;
1589  j++, ptr += syntax[i].list_elem_size)
1590  ebml_free(syntax[i].def.n, ptr);
1591  av_freep(&list->elem);
1592  list->nb_elem = 0;
1593  list->alloc_elem_size = 0;
1594  } else
1595  ebml_free(syntax[i].def.n, data_off);
1596  default:
1597  break;
1598  }
1599  }
1600 }
1601 
1602 /*
1603  * Autodetecting...
1604  */
1605 static int matroska_probe(const AVProbeData *p)
1606 {
1607  uint64_t total = 0;
1608  int len_mask = 0x80, size = 1, n = 1, i;
1609 
1610  /* EBML header? */
1611  if (AV_RB32(p->buf) != EBML_ID_HEADER)
1612  return 0;
1613 
1614  /* length of header */
1615  total = p->buf[4];
1616  while (size <= 8 && !(total & len_mask)) {
1617  size++;
1618  len_mask >>= 1;
1619  }
1620  if (size > 8)
1621  return 0;
1622  total &= (len_mask - 1);
1623  while (n < size)
1624  total = (total << 8) | p->buf[4 + n++];
1625 
1626  if (total + 1 == 1ULL << (7 * size)){
1627  /* Unknown-length header - simply parse the whole buffer. */
1628  total = p->buf_size - 4 - size;
1629  } else {
1630  /* Does the probe data contain the whole header? */
1631  if (p->buf_size < 4 + size + total)
1632  return 0;
1633  }
1634 
1635  /* The header should contain a known document type. For now,
1636  * we don't parse the whole header but simply check for the
1637  * availability of that array of characters inside the header.
1638  * Not fully fool-proof, but good enough. */
1639  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1640  size_t probelen = strlen(matroska_doctypes[i]);
1641  if (total < probelen)
1642  continue;
1643  for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1644  if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1645  return AVPROBE_SCORE_MAX;
1646  }
1647 
1648  // probably valid EBML header but no recognized doctype
1649  return AVPROBE_SCORE_EXTENSION;
1650 }
1651 
1653  uint64_t num)
1654 {
1655  MatroskaTrack *tracks = matroska->tracks.elem;
1656  int i;
1657 
1658  for (i = 0; i < matroska->tracks.nb_elem; i++)
1659  if (tracks[i].num == num)
1660  return &tracks[i];
1661 
1662  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1663  return NULL;
1664 }
1665 
1666 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1667  MatroskaTrack *track)
1668 {
1669  MatroskaTrackEncoding *encodings = track->encodings.elem;
1670  uint8_t *data = *buf;
1671  int isize = *buf_size;
1672  uint8_t *pkt_data = NULL;
1673  uint8_t av_unused *newpktdata;
1674  int pkt_size = isize;
1675  int result = 0;
1676  int olen;
1677 
1678  if (pkt_size >= 10000000U)
1679  return AVERROR_INVALIDDATA;
1680 
1681  switch (encodings[0].compression.algo) {
1683  {
1684  int header_size = encodings[0].compression.settings.size;
1685  uint8_t *header = encodings[0].compression.settings.data;
1686 
1687  if (header_size && !header) {
1688  av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1689  return -1;
1690  }
1691 
1692  if (!header_size)
1693  return 0;
1694 
1695  pkt_size = isize + header_size;
1696  pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1697  if (!pkt_data)
1698  return AVERROR(ENOMEM);
1699 
1700  memcpy(pkt_data, header, header_size);
1701  memcpy(pkt_data + header_size, data, isize);
1702  break;
1703  }
1705  do {
1706  int insize = isize;
1707  olen = pkt_size *= 3;
1708  newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1710  if (!newpktdata) {
1711  result = AVERROR(ENOMEM);
1712  goto failed;
1713  }
1714  pkt_data = newpktdata;
1715  result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1716  } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1717  if (result) {
1719  goto failed;
1720  }
1721  pkt_size -= olen;
1722  break;
1723 #if CONFIG_ZLIB
1725  {
1726  z_stream zstream = { 0 };
1727  if (!pkt_size || inflateInit(&zstream) != Z_OK)
1728  return -1;
1729  zstream.next_in = data;
1730  zstream.avail_in = isize;
1731  do {
1732  pkt_size *= 3;
1733  newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1734  if (!newpktdata) {
1735  inflateEnd(&zstream);
1736  result = AVERROR(ENOMEM);
1737  goto failed;
1738  }
1739  pkt_data = newpktdata;
1740  zstream.avail_out = pkt_size - zstream.total_out;
1741  zstream.next_out = pkt_data + zstream.total_out;
1742  result = inflate(&zstream, Z_NO_FLUSH);
1743  } while (result == Z_OK && pkt_size < 10000000);
1744  pkt_size = zstream.total_out;
1745  inflateEnd(&zstream);
1746  if (result != Z_STREAM_END) {
1747  if (result == Z_MEM_ERROR)
1748  result = AVERROR(ENOMEM);
1749  else
1751  goto failed;
1752  }
1753  break;
1754  }
1755 #endif
1756 #if CONFIG_BZLIB
1758  {
1759  bz_stream bzstream = { 0 };
1760  if (!pkt_size || BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1761  return -1;
1762  bzstream.next_in = data;
1763  bzstream.avail_in = isize;
1764  do {
1765  pkt_size *= 3;
1766  newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1767  if (!newpktdata) {
1768  BZ2_bzDecompressEnd(&bzstream);
1769  result = AVERROR(ENOMEM);
1770  goto failed;
1771  }
1772  pkt_data = newpktdata;
1773  bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1774  bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1775  result = BZ2_bzDecompress(&bzstream);
1776  } while (result == BZ_OK && pkt_size < 10000000);
1777  pkt_size = bzstream.total_out_lo32;
1778  BZ2_bzDecompressEnd(&bzstream);
1779  if (result != BZ_STREAM_END) {
1780  if (result == BZ_MEM_ERROR)
1781  result = AVERROR(ENOMEM);
1782  else
1784  goto failed;
1785  }
1786  break;
1787  }
1788 #endif
1789  default:
1790  return AVERROR_INVALIDDATA;
1791  }
1792 
1793  memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1794 
1795  *buf = pkt_data;
1796  *buf_size = pkt_size;
1797  return 0;
1799 failed:
1800  av_free(pkt_data);
1801  return result;
1802 }
1803 
1805  AVDictionary **metadata, char *prefix)
1806 {
1807  MatroskaTag *tags = list->elem;
1808  char key[1024];
1809  int i;
1810 
1811  for (i = 0; i < list->nb_elem; i++) {
1812  const char *lang = tags[i].lang &&
1813  strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1814 
1815  if (!tags[i].name) {
1816  av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1817  continue;
1818  }
1819  if (prefix)
1820  snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1821  else
1822  av_strlcpy(key, tags[i].name, sizeof(key));
1823  if (tags[i].def || !lang) {
1824  av_dict_set(metadata, key, tags[i].string, 0);
1825  if (tags[i].sub.nb_elem)
1826  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1827  }
1828  if (lang) {
1829  av_strlcat(key, "-", sizeof(key));
1830  av_strlcat(key, lang, sizeof(key));
1831  av_dict_set(metadata, key, tags[i].string, 0);
1832  if (tags[i].sub.nb_elem)
1833  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1834  }
1835  }
1837 }
1838 
1840 {
1841  MatroskaDemuxContext *matroska = s->priv_data;
1842  MatroskaTags *tags = matroska->tags.elem;
1843  int i, j;
1844 
1845  for (i = 0; i < matroska->tags.nb_elem; i++) {
1846  if (tags[i].target.attachuid) {
1847  MatroskaAttachment *attachment = matroska->attachments.elem;
1848  int found = 0;
1849  for (j = 0; j < matroska->attachments.nb_elem; j++) {
1850  if (attachment[j].uid == tags[i].target.attachuid &&
1851  attachment[j].stream) {
1852  matroska_convert_tag(s, &tags[i].tag,
1853  &attachment[j].stream->metadata, NULL);
1854  found = 1;
1855  }
1856  }
1857  if (!found) {
1859  "The tags at index %d refer to a "
1860  "non-existent attachment %"PRId64".\n",
1861  i, tags[i].target.attachuid);
1862  }
1863  } else if (tags[i].target.chapteruid) {
1864  MatroskaChapter *chapter = matroska->chapters.elem;
1865  int found = 0;
1866  for (j = 0; j < matroska->chapters.nb_elem; j++) {
1867  if (chapter[j].uid == tags[i].target.chapteruid &&
1868  chapter[j].chapter) {
1869  matroska_convert_tag(s, &tags[i].tag,
1870  &chapter[j].chapter->metadata, NULL);
1871  found = 1;
1872  }
1873  }
1874  if (!found) {
1876  "The tags at index %d refer to a non-existent chapter "
1877  "%"PRId64".\n",
1878  i, tags[i].target.chapteruid);
1879  }
1880  } else if (tags[i].target.trackuid) {
1881  MatroskaTrack *track = matroska->tracks.elem;
1882  int found = 0;
1883  for (j = 0; j < matroska->tracks.nb_elem; j++) {
1884  if (track[j].uid == tags[i].target.trackuid &&
1885  track[j].stream) {
1886  matroska_convert_tag(s, &tags[i].tag,
1887  &track[j].stream->metadata, NULL);
1888  found = 1;
1889  }
1890  }
1891  if (!found) {
1893  "The tags at index %d refer to a non-existent track "
1894  "%"PRId64".\n",
1895  i, tags[i].target.trackuid);
1896  }
1897  } else {
1898  matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1899  tags[i].target.type);
1900  }
1901  }
1902 }
1903 
1905  int64_t pos)
1906 {
1907  uint32_t saved_id = matroska->current_id;
1908  int64_t before_pos = avio_tell(matroska->ctx->pb);
1909  int ret = 0;
1910  int ret2;
1911 
1912  /* seek */
1913  if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1914  /* We don't want to lose our seekhead level, so we add
1915  * a dummy. This is a crude hack. */
1916  if (matroska->num_levels == EBML_MAX_DEPTH) {
1917  av_log(matroska->ctx, AV_LOG_INFO,
1918  "Max EBML element depth (%d) reached, "
1919  "cannot parse further.\n", EBML_MAX_DEPTH);
1921  } else {
1922  matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1923  matroska->num_levels++;
1924  matroska->current_id = 0;
1925 
1926  ret = ebml_parse(matroska, matroska_segment, matroska);
1927  if (ret == LEVEL_ENDED) {
1928  /* This can only happen if the seek brought us beyond EOF. */
1929  ret = AVERROR_EOF;
1930  }
1931  }
1932  }
1933  /* Seek back - notice that in all instances where this is used
1934  * it is safe to set the level to 1. */
1935  ret2 = matroska_reset_status(matroska, saved_id, before_pos);
1936  if (ret >= 0)
1937  ret = ret2;
1938 
1939  return ret;
1940 }
1941 
1942 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1943 {
1944  EbmlList *seekhead_list = &matroska->seekhead;
1945  int i;
1946 
1947  // we should not do any seeking in the streaming case
1948  if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1949  return;
1950 
1951  for (i = 0; i < seekhead_list->nb_elem; i++) {
1952  MatroskaSeekhead *seekheads = seekhead_list->elem;
1953  uint32_t id = seekheads[i].id;
1954  int64_t pos = seekheads[i].pos + matroska->segment_start;
1955  MatroskaLevel1Element *elem;
1956 
1957  if (id != seekheads[i].id || pos < matroska->segment_start)
1958  continue;
1959 
1960  elem = matroska_find_level1_elem(matroska, id, pos);
1961  if (!elem || elem->parsed)
1962  continue;
1963 
1964  elem->pos = pos;
1965 
1966  // defer cues parsing until we actually need cue data.
1967  if (id == MATROSKA_ID_CUES)
1968  continue;
1969 
1970  if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1971  // mark index as broken
1972  matroska->cues_parsing_deferred = -1;
1973  break;
1974  }
1975 
1976  elem->parsed = 1;
1977  }
1978 }
1979 
1980 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1981 {
1982  EbmlList *index_list;
1984  uint64_t index_scale = 1;
1985  int i, j;
1986 
1987  if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1988  return;
1989 
1990  index_list = &matroska->index;
1991  index = index_list->elem;
1992  if (index_list->nb_elem < 2)
1993  return;
1994  if (index[1].time > 1E14 / matroska->time_scale) {
1995  av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1996  return;
1997  }
1998  for (i = 0; i < index_list->nb_elem; i++) {
1999  EbmlList *pos_list = &index[i].pos;
2000  MatroskaIndexPos *pos = pos_list->elem;
2001  for (j = 0; j < pos_list->nb_elem; j++) {
2002  MatroskaTrack *track = matroska_find_track_by_num(matroska,
2003  pos[j].track);
2004  if (track && track->stream)
2005  av_add_index_entry(track->stream,
2006  pos[j].pos + matroska->segment_start,
2007  index[i].time / index_scale, 0, 0,
2009  }
2010  }
2011 }
2012 
2013 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
2014  int i;
2015 
2016  if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
2017  return;
2018 
2019  for (i = 0; i < matroska->num_level1_elems; i++) {
2020  MatroskaLevel1Element *elem = &matroska->level1_elems[i];
2021  if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
2022  if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
2023  matroska->cues_parsing_deferred = -1;
2024  elem->parsed = 1;
2025  break;
2026  }
2027  }
2028 
2029  matroska_add_index_entries(matroska);
2030 }
2031 
2033  unsigned nb_encodings,
2034  MatroskaTrack *track,
2035  char **key_id_base64, void *logctx)
2036 {
2037  if (nb_encodings > 1) {
2038  av_log(logctx, AV_LOG_ERROR,
2039  "Multiple combined encodings not supported");
2040  return 0;
2041  }
2042  if (!nb_encodings)
2043  return 0;
2044  if (encodings->type) {
2045  if (encodings->encryption.key_id.size > 0) {
2046  /* Save the encryption key id to be stored later
2047  * as a metadata tag. */
2048  const int b64_size = AV_BASE64_SIZE(encodings->encryption.key_id.size);
2049  *key_id_base64 = av_malloc(b64_size);
2050  if (!*key_id_base64)
2051  return AVERROR(ENOMEM);
2052 
2053  av_base64_encode(*key_id_base64, b64_size,
2054  encodings->encryption.key_id.data,
2055  encodings->encryption.key_id.size);
2056  } else {
2057  encodings->scope = 0;
2058  av_log(logctx, AV_LOG_ERROR, "Unsupported encoding type");
2059  }
2060  } else if (
2061 #if CONFIG_ZLIB
2063 #endif
2064 #if CONFIG_BZLIB
2066 #endif
2069  encodings->scope = 0;
2070  av_log(logctx, AV_LOG_ERROR, "Unsupported encoding type");
2071  } else if (track->codec_priv.size && encodings[0].scope & 2) {
2072  uint8_t *codec_priv = track->codec_priv.data;
2073  int ret = matroska_decode_buffer(&track->codec_priv.data,
2074  &track->codec_priv.size,
2075  track);
2076  if (ret < 0) {
2077  track->codec_priv.data = NULL;
2078  track->codec_priv.size = 0;
2079  av_log(logctx, AV_LOG_ERROR,
2080  "Failed to decode codec private data\n");
2081  }
2082 
2083  if (codec_priv != track->codec_priv.data) {
2084  av_buffer_unref(&track->codec_priv.buf);
2085  if (track->codec_priv.data) {
2086  track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2088  NULL, NULL, 0);
2089  if (!track->codec_priv.buf) {
2090  av_freep(&track->codec_priv.data);
2091  track->codec_priv.size = 0;
2092  return AVERROR(ENOMEM);
2093  }
2094  }
2095  }
2096  }
2097  track->needs_decoding = !encodings->type &&
2098  encodings->scope & 1 &&
2099  (encodings->compression.algo !=
2101  encodings->compression.settings.size);
2102 
2103  return 0;
2104 }
2105 
2106 static int matroska_aac_profile(char *codec_id)
2107 {
2108  static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
2109  int profile;
2110 
2112  if (strstr(codec_id, aac_profiles[profile]))
2113  break;
2114  return profile + 1;
2115 }
2116 
2117 static int matroska_aac_sri(int samplerate)
2118 {
2119  int sri;
2120 
2121  for (sri = 0; sri < FF_ARRAY_ELEMS(ff_mpeg4audio_sample_rates); sri++)
2122  if (ff_mpeg4audio_sample_rates[sri] == samplerate)
2123  break;
2124  return sri;
2125 }
2126 
2127 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
2128 {
2129  /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
2130  avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
2131 }
2132 
2134  MatroskaTrack *track,
2135  int *offset)
2136 {
2137  AVStream *st = track->stream;
2138  uint8_t *p = track->codec_priv.data;
2139  int size = track->codec_priv.size;
2140 
2141  if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
2142  av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
2143  track->codec_priv.size = 0;
2144  return 0;
2145  }
2146  *offset = 8;
2147  track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
2148 
2149  p += track->codec_priv.size;
2150  size -= track->codec_priv.size;
2151 
2152  /* parse the remaining metadata blocks if present */
2153  while (size >= 4) {
2154  int block_last, block_type, block_size;
2155 
2156  flac_parse_block_header(p, &block_last, &block_type, &block_size);
2157 
2158  p += 4;
2159  size -= 4;
2160  if (block_size > size)
2161  return 0;
2162 
2163  /* check for the channel mask */
2164  if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2165  AVDictionary *dict = NULL;
2166  AVDictionaryEntry *chmask;
2167 
2168  ff_vorbis_comment(s, &dict, p, block_size, 0);
2169  chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2170  if (chmask) {
2171  uint64_t mask = strtol(chmask->value, NULL, 0);
2172  if (!mask || mask & ~0x3ffffULL) {
2174  "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2175  } else
2177  }
2178  av_dict_free(&dict);
2179  }
2180 
2181  p += block_size;
2182  size -= block_size;
2183  }
2184 
2185  return 0;
2186 }
2187 
2188 static int mkv_field_order(const MatroskaDemuxContext *matroska, uint64_t field_order)
2189 {
2190  int minor, micro, bttb = 0;
2191 
2192  /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2193  * this function, and fixed in 57.52 */
2194  if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, &micro) == 2)
2195  bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2196 
2197  switch (field_order) {
2199  return AV_FIELD_PROGRESSIVE;
2201  return AV_FIELD_UNKNOWN;
2203  return AV_FIELD_TT;
2205  return AV_FIELD_BB;
2207  return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2209  return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2210  default:
2211  return AV_FIELD_UNKNOWN;
2212  }
2213 }
2214 
2215 static void mkv_stereo_mode_display_mul(int stereo_mode,
2216  int *h_width, int *h_height)
2217 {
2218  switch (stereo_mode) {
2224  break;
2229  *h_width = 2;
2230  break;
2235  *h_height = 2;
2236  break;
2237  }
2238 }
2239 
2240 static int mkv_stereo3d_conv(AVStream *st, MatroskaVideoStereoModeType stereo_mode)
2241 {
2242  static const struct {
2243  char type;
2244  char flags;
2245  } stereo_mode_conv [] = {
2246 #define STEREO_MODE_CONV(STEREOMODETYPE, STEREO3DTYPE, FLAGS, WDIV, HDIV, WEBM) \
2247  [(STEREOMODETYPE)] = { .type = (STEREO3DTYPE), .flags = (FLAGS) },
2248 #define NOTHING(STEREOMODETYPE, WDIV, HDIV, WEBM)
2250  };
2251  AVStereo3D *stereo;
2252 
2253  stereo = av_stereo3d_alloc();
2254  if (!stereo)
2255  return AVERROR(ENOMEM);
2256 
2257  stereo->type = stereo_mode_conv[stereo_mode].type;
2258  stereo->flags = stereo_mode_conv[stereo_mode].flags;
2259 
2261  AV_PKT_DATA_STEREO3D, stereo, sizeof(*stereo), 0)) {
2262  av_freep(&stereo);
2263  return AVERROR(ENOMEM);
2264  }
2265 
2266  return 0;
2267 }
2268 
2269 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2270  const MatroskaTrackVideoColor *color = track->video.color.elem;
2271  const MatroskaMasteringMeta *mastering_meta;
2272  int has_mastering_primaries, has_mastering_luminance;
2273 
2274  if (!track->video.color.nb_elem)
2275  return 0;
2276 
2277  mastering_meta = &color->mastering_meta;
2278  // Mastering primaries are CIE 1931 coords, and must be > 0.
2279  has_mastering_primaries =
2280  mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2281  mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2282  mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2283  mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2284  has_mastering_luminance = mastering_meta->max_luminance >
2285  mastering_meta->min_luminance.el.f &&
2286  mastering_meta->min_luminance.el.f >= 0 &&
2287  mastering_meta->min_luminance.count;
2288 
2289  if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2290  st->codecpar->color_space = color->matrix_coefficients;
2291  if (color->primaries != AVCOL_PRI_RESERVED &&
2292  color->primaries != AVCOL_PRI_RESERVED0)
2293  st->codecpar->color_primaries = color->primaries;
2294  if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2295  color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2296  st->codecpar->color_trc = color->transfer_characteristics;
2297  if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2298  color->range <= AVCOL_RANGE_JPEG)
2299  st->codecpar->color_range = color->range;
2300  if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2301  color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2302  color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2303  color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2304  st->codecpar->chroma_location =
2305  av_chroma_location_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2306  (color->chroma_siting_vert - 1) << 7);
2307  }
2308  if (color->max_cll && color->max_fall) {
2309  size_t size = 0;
2311  if (!metadata)
2312  return AVERROR(ENOMEM);
2314  AV_PKT_DATA_CONTENT_LIGHT_LEVEL, metadata, size, 0)) {
2315  av_freep(&metadata);
2316  return AVERROR(ENOMEM);
2317  }
2318  metadata->MaxCLL = color->max_cll;
2319  metadata->MaxFALL = color->max_fall;
2320  }
2321 
2322  if (has_mastering_primaries || has_mastering_luminance) {
2323  AVMasteringDisplayMetadata *metadata;
2327  sizeof(AVMasteringDisplayMetadata), 0);
2328  if (!sd)
2329  return AVERROR(ENOMEM);
2330  metadata = (AVMasteringDisplayMetadata*)sd->data;
2331  memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2332  if (has_mastering_primaries) {
2333  metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2334  metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2335  metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2336  metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2337  metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2338  metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2339  metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2340  metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2341  metadata->has_primaries = 1;
2342  }
2343  if (has_mastering_luminance) {
2344  metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2345  metadata->min_luminance = av_d2q(mastering_meta->min_luminance.el.f, INT_MAX);
2346  metadata->has_luminance = 1;
2347  }
2348  }
2349  return 0;
2350 }
2351 
2352 static int mkv_create_display_matrix(AVStream *st,
2353  const MatroskaTrackVideoProjection *proj,
2354  void *logctx)
2355 {
2356  AVPacketSideData *sd;
2357  double pitch = proj->pitch, yaw = proj->yaw, roll = proj->roll;
2358  int32_t *matrix;
2359  int hflip;
2360 
2361  if (pitch == 0.0 && yaw == 0.0 && roll == 0.0)
2362  return 0;
2363 
2364  /* Note: The following constants are exactly representable
2365  * as floating-point numbers. */
2366  if (pitch != 0.0 || (yaw != 0.0 && yaw != 180.0 && yaw != -180.0) ||
2367  isnan(roll)) {
2368  av_log(logctx, AV_LOG_WARNING, "Ignoring non-2D rectangular "
2369  "projection in stream %u (yaw %f, pitch %f, roll %f)\n",
2370  st->index, yaw, pitch, roll);
2371  return 0;
2372  }
2376  9 * sizeof(*matrix), 0);
2377  if (!sd)
2378  return AVERROR(ENOMEM);
2379  matrix = (int32_t*)sd->data;
2380 
2381  hflip = yaw != 0.0;
2382  /* ProjectionPoseRoll is in the counter-clockwise direction
2383  * whereas av_display_rotation_set() expects its argument
2384  * to be oriented clockwise, so we need to negate roll.
2385  * Furthermore, if hflip is set, we need to negate it again
2386  * to account for the fact that the Matroska specifications
2387  * require the yaw rotation to be applied first. */
2388  av_display_rotation_set(matrix, roll * (2 * hflip - 1));
2389  av_display_matrix_flip(matrix, hflip, 0);
2390 
2391  return 0;
2392 }
2393 
2394 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2395  void *logctx)
2396 {
2397  AVSphericalMapping *spherical;
2398  const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2399  const uint8_t *priv_data = mkv_projection->private.data;
2400  enum AVSphericalProjection projection;
2401  size_t spherical_size;
2402  uint32_t l = 0, t = 0, r = 0, b = 0;
2403  uint32_t padding = 0;
2404 
2405  if (mkv_projection->private.size && priv_data[0] != 0) {
2406  av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2407  return 0;
2408  }
2409 
2410  switch (track->video.projection.type) {
2412  return mkv_create_display_matrix(st, mkv_projection, logctx);
2414  if (track->video.projection.private.size == 20) {
2415  t = AV_RB32(priv_data + 4);
2416  b = AV_RB32(priv_data + 8);
2417  l = AV_RB32(priv_data + 12);
2418  r = AV_RB32(priv_data + 16);
2419 
2420  if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2421  av_log(logctx, AV_LOG_ERROR,
2422  "Invalid bounding rectangle coordinates "
2423  "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2424  l, t, r, b);
2425  return AVERROR_INVALIDDATA;
2426  }
2427  } else if (track->video.projection.private.size != 0) {
2428  av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2429  return AVERROR_INVALIDDATA;
2430  }
2431 
2432  if (l || t || r || b)
2433  projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2434  else
2435  projection = AV_SPHERICAL_EQUIRECTANGULAR;
2436  break;
2438  if (track->video.projection.private.size < 4) {
2439  av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2440  return AVERROR_INVALIDDATA;
2441  } else if (track->video.projection.private.size == 12) {
2442  uint32_t layout = AV_RB32(priv_data + 4);
2443  if (layout) {
2444  av_log(logctx, AV_LOG_WARNING,
2445  "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2446  return 0;
2447  }
2448  projection = AV_SPHERICAL_CUBEMAP;
2449  padding = AV_RB32(priv_data + 8);
2450  } else {
2451  av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2452  return AVERROR_INVALIDDATA;
2453  }
2454  break;
2455  default:
2456  av_log(logctx, AV_LOG_WARNING,
2457  "Unknown spherical metadata type %"PRIu64"\n",
2458  track->video.projection.type);
2459  return 0;
2460  }
2461 
2462  spherical = av_spherical_alloc(&spherical_size);
2463  if (!spherical)
2464  return AVERROR(ENOMEM);
2465 
2466  spherical->projection = projection;
2467 
2468  spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2469  spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2470  spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2471 
2472  spherical->padding = padding;
2473 
2474  spherical->bound_left = l;
2475  spherical->bound_top = t;
2476  spherical->bound_right = r;
2477  spherical->bound_bottom = b;
2478 
2480  AV_PKT_DATA_SPHERICAL, spherical, spherical_size, 0)) {
2481  av_freep(&spherical);
2482  return AVERROR(ENOMEM);
2483  }
2484 
2485  return 0;
2486 }
2487 
2489  EbmlBin *bin)
2490 {
2491  return ff_isom_parse_dvcc_dvvc(s, st, bin->data, bin->size);
2492 }
2493 
2495 {
2496  const EbmlList *mappings_list = &track->block_addition_mappings;
2497  MatroskaBlockAdditionMapping *mappings = mappings_list->elem;
2498  int ret;
2499 
2500  for (int i = 0; i < mappings_list->nb_elem; i++) {
2501  MatroskaBlockAdditionMapping *mapping = &mappings[i];
2502  uint64_t type = mapping->type;
2503 
2504  switch (mapping->type) {
2507  "Explicit block Addition Mapping type \"Use BlockAddIDValue\", value %"PRIu64","
2508  " name \"%s\" found.\n", mapping->value, mapping->name ? mapping->name : "");
2510  // fall-through
2513  if (mapping->value != type) {
2514  int strict = s->strict_std_compliance >= FF_COMPLIANCE_STRICT;
2515  av_log(s, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
2516  "Invalid Block Addition Value 0x%"PRIx64" for Block Addition Mapping Type "
2517  "0x%"PRIx64", name \"%s\"\n", mapping->value, mapping->type,
2518  mapping->name ? mapping->name : "");
2519  if (strict)
2520  return AVERROR_INVALIDDATA;
2521  }
2522  break;
2525  if ((ret = mkv_parse_dvcc_dvvc(s, st, track, &mapping->extradata)) < 0)
2526  return ret;
2527 
2528  break;
2529  default:
2531  "Unknown Block Addition Mapping type 0x%"PRIx64", value %"PRIu64", name \"%s\"\n",
2532  mapping->type, mapping->value, mapping->name ? mapping->name : "");
2533  if (mapping->value < 2) {
2534  int strict = s->strict_std_compliance >= FF_COMPLIANCE_STRICT;
2535  av_log(s, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
2536  "Invalid Block Addition value 0x%"PRIu64" for unknown Block Addition Mapping "
2537  "type %"PRIx64", name \"%s\"\n", mapping->value, mapping->type,
2538  mapping->name ? mapping->name : "");
2539  if (strict)
2540  return AVERROR_INVALIDDATA;
2541  }
2542  break;
2543  }
2544  }
2545 
2546  return 0;
2547 }
2548 
2549 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2550 {
2551  const AVCodecTag *codec_tags;
2552 
2553  codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2555 
2556  /* Normalize noncompliant private data that starts with the fourcc
2557  * by expanding/shifting the data by 4 bytes and storing the data
2558  * size at the start. */
2559  if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2560  int ret = av_buffer_realloc(&track->codec_priv.buf,
2562  if (ret < 0)
2563  return ret;
2564 
2565  track->codec_priv.data = track->codec_priv.buf->data;
2566  memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2567  track->codec_priv.size += 4;
2568  AV_WB32(track->codec_priv.data, track->codec_priv.size);
2569  }
2570 
2571  *fourcc = AV_RL32(track->codec_priv.data + 4);
2572  *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2573 
2574  return 0;
2576 
2577 /* An enum with potential return values of the functions for parsing a track.
2578  * Apart from that all these functions can also indicate ordinary errors via
2579  * negative return values. */
2580 enum {
2581  SKIP_TRACK = 1,
2582 };
2583 
2584 #define AAC_MAX_EXTRADATA_SIZE 5
2585 #define TTA_EXTRADATA_SIZE 22
2586 #define WAVPACK_EXTRADATA_SIZE 2
2587 /* Performs the codec-specific part of parsing an audio track. */
2588 static int mka_parse_audio_codec(MatroskaTrack *track, AVCodecParameters *par,
2589  const MatroskaDemuxContext *matroska,
2590  AVFormatContext *s, int *extradata_offset)
2591 {
2592  uint8_t extradata[FFMAX3(AAC_MAX_EXTRADATA_SIZE,
2595  int extradata_size = 0; // > 0 means that the extradata buffer is used
2596  int ret;
2597 
2598  if (!strcmp(track->codec_id, "A_MS/ACM") &&
2599  track->codec_priv.size >= 14) {
2600  FFIOContext b;
2602  track->codec_priv.size);
2603  ret = ff_get_wav_header(s, &b.pub, par,
2604  track->codec_priv.size, 0);
2605  if (ret < 0)
2606  return ret;
2607  *extradata_offset = FFMIN(track->codec_priv.size, 18);
2608  return 0;
2609  } else if (!strcmp(track->codec_id, "A_QUICKTIME") &&
2610  /* Normally 36, but allow noncompliant private data */
2611  track->codec_priv.size >= 32) {
2612  enum AVCodecID codec_id;
2613  uint32_t fourcc;
2614  uint16_t sample_size;
2615 
2616  ret = get_qt_codec(track, &fourcc, &codec_id);
2617  if (ret < 0)
2618  return ret;
2619  sample_size = AV_RB16(track->codec_priv.data + 26);
2620  if (fourcc == 0) {
2621  if (sample_size == 8) {
2622  fourcc = MKTAG('r','a','w',' ');
2624  } else if (sample_size == 16) {
2625  fourcc = MKTAG('t','w','o','s');
2627  }
2628  }
2629  if ((fourcc == MKTAG('t','w','o','s') ||
2630  fourcc == MKTAG('s','o','w','t')) && sample_size == 8)
2632  par->codec_id = codec_id;
2633  par->codec_tag = fourcc;
2634  return 0;
2635  }
2636 
2637  switch (par->codec_id) {
2638  case AV_CODEC_ID_PCM_S16BE:
2639  switch (track->audio.bitdepth) {
2640  case 8:
2642  break;
2643  case 24:
2645  break;
2646  case 32:
2648  break;
2649  }
2650  break;
2651  case AV_CODEC_ID_PCM_S16LE:
2652  switch (track->audio.bitdepth) {
2653  case 8:
2655  break;
2656  case 24:
2658  break;
2659  case 32:
2661  break;
2662  }
2663  break;
2664  case AV_CODEC_ID_PCM_F32LE:
2665  if (track->audio.bitdepth == 64)
2667  break;
2668  case AV_CODEC_ID_AAC:
2669  if (!track->codec_priv.size) {
2670  int profile = matroska_aac_profile(track->codec_id);
2671  int sri = matroska_aac_sri(track->audio.samplerate);
2672 
2673  extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2674  extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2675  if (strstr(track->codec_id, "SBR")) {
2676  sri = matroska_aac_sri(track->audio.out_samplerate);
2677  extradata[2] = 0x56;
2678  extradata[3] = 0xE5;
2679  extradata[4] = 0x80 | (sri << 3);
2680  extradata_size = 5;
2681  } else
2682  extradata_size = 2;
2683  }
2684  break;
2685  case AV_CODEC_ID_ALAC:
2686  if (track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2687  /* Only ALAC's magic cookie is stored in Matroska's track headers.
2688  * Create the "atom size", "tag", and "tag version" fields the
2689  * decoder expects manually. */
2690  ret = ff_alloc_extradata(par, 12 + track->codec_priv.size);
2691  if (ret < 0)
2692  return ret;
2693  AV_WB32(par->extradata, par->extradata_size);
2694  AV_WB32(&par->extradata[4], MKBETAG('a', 'l', 'a', 'c'));
2695  AV_WB32(&par->extradata[8], 0);
2696  memcpy(&par->extradata[12], track->codec_priv.data,
2697  track->codec_priv.size);
2698  }
2699  break;
2700  case AV_CODEC_ID_TTA:
2701  {
2702  uint8_t *ptr;
2703  if (track->audio.channels > UINT16_MAX ||
2704  track->audio.bitdepth > UINT16_MAX) {
2705  av_log(matroska->ctx, AV_LOG_WARNING,
2706  "Too large audio channel number %"PRIu64
2707  " or bitdepth %"PRIu64". Skipping track.\n",
2708  track->audio.channels, track->audio.bitdepth);
2709  if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2710  return AVERROR_INVALIDDATA;
2711  else
2712  return SKIP_TRACK;
2713  }
2714  if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2715  return AVERROR_INVALIDDATA;
2716  extradata_size = TTA_EXTRADATA_SIZE;
2717  ptr = extradata;
2718  bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2719  bytestream_put_le16(&ptr, 1);
2720  bytestream_put_le16(&ptr, track->audio.channels);
2721  bytestream_put_le16(&ptr, track->audio.bitdepth);
2722  bytestream_put_le32(&ptr, track->audio.out_samplerate);
2723  bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2724  track->audio.out_samplerate,
2725  AV_TIME_BASE * 1000));
2726  break;
2727  }
2728  case AV_CODEC_ID_RA_144:
2729  track->audio.out_samplerate = 8000;
2730  track->audio.channels = 1;
2731  break;
2732  case AV_CODEC_ID_RA_288:
2733  case AV_CODEC_ID_COOK:
2734  case AV_CODEC_ID_ATRAC3:
2735  case AV_CODEC_ID_SIPR:
2736  {
2737  const uint8_t *ptr = track->codec_priv.data;
2738  int flavor;
2739 
2740  if (!track->codec_priv.size)
2741  break;
2742 
2743  if (track->codec_priv.size < 46)
2744  return AVERROR_INVALIDDATA;
2745  ptr += 22;
2746  flavor = bytestream_get_be16(&ptr);
2747  track->audio.coded_framesize = bytestream_get_be32(&ptr);
2748  ptr += 12;
2749  track->audio.sub_packet_h = bytestream_get_be16(&ptr);
2750  track->audio.frame_size = bytestream_get_be16(&ptr);
2751  track->audio.sub_packet_size = bytestream_get_be16(&ptr);
2752  if (track->audio.coded_framesize <= 0 ||
2753  track->audio.sub_packet_h <= 0 ||
2754  track->audio.frame_size <= 0)
2755  return AVERROR_INVALIDDATA;
2756 
2757  if (par->codec_id == AV_CODEC_ID_RA_288) {
2758  if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2759  != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2760  return AVERROR_INVALIDDATA;
2761  par->block_align = track->audio.coded_framesize;
2762  track->codec_priv.size = 0;
2763  } else {
2764  if (par->codec_id == AV_CODEC_ID_SIPR) {
2765  static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2766  if (flavor > 3)
2767  return AVERROR_INVALIDDATA;
2768  track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2769  par->bit_rate = sipr_bit_rate[flavor];
2770  } else if (track->audio.sub_packet_size <= 0 ||
2771  track->audio.frame_size % track->audio.sub_packet_size)
2772  return AVERROR_INVALIDDATA;
2773  par->block_align = track->audio.sub_packet_size;
2774  *extradata_offset = 78;
2775  }
2776  track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2777  track->audio.frame_size);
2778  if (!track->audio.buf)
2779  return AVERROR(ENOMEM);
2780  break;
2781  }
2782  case AV_CODEC_ID_ATRAC1:
2783  /* ATRAC1 uses a constant frame size.
2784  * Typical ATRAC1 streams are either mono or stereo.
2785  * At most, ATRAC1 was used to store 8 channels of audio. */
2786  if (track->audio.channels > 8)
2787  return AVERROR_INVALIDDATA;
2788  par->block_align = track->audio.channels * 212;
2789  break;
2790  case AV_CODEC_ID_FLAC:
2791  if (track->codec_priv.size) {
2792  ret = matroska_parse_flac(s, track, extradata_offset);
2793  if (ret < 0)
2794  return ret;
2795  }
2796  break;
2797  case AV_CODEC_ID_WAVPACK:
2798  if (track->codec_priv.size < 2) {
2799  av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2800  "in absence of valid CodecPrivate.\n");
2801  extradata_size = WAVPACK_EXTRADATA_SIZE;
2802  AV_WL16(extradata, 0x410);
2803  }
2804  break;
2805  }
2806 
2807  if (extradata_size > 0) {
2808  ret = ff_alloc_extradata(par, extradata_size);
2809  if (ret < 0)
2810  return ret;
2811  memcpy(par->extradata, extradata, extradata_size);
2812  }
2813 
2814  return 0;
2815 }
2816 
2817 /* Performs the generic part of parsing an audio track. */
2818 static int mka_parse_audio(MatroskaTrack *track, AVStream *st,
2819  AVCodecParameters *par,
2820  const MatroskaDemuxContext *matroska,
2821  AVFormatContext *s, int *extradata_offset)
2822 {
2823  FFStream *const sti = ffstream(st);
2824  int ret;
2825 
2826  ret = mka_parse_audio_codec(track, par, matroska,
2827  s, extradata_offset);
2828  if (ret)
2829  return ret;
2830 
2832  par->sample_rate = track->audio.out_samplerate;
2833  // channel layout may be already set by codec private checks above
2834  if (!av_channel_layout_check(&par->ch_layout)) {
2836  par->ch_layout.nb_channels = track->audio.channels;
2837  }
2838  if (!par->bits_per_coded_sample)
2839  par->bits_per_coded_sample = track->audio.bitdepth;
2840  if (par->codec_id == AV_CODEC_ID_MP3 ||
2841  par->codec_id == AV_CODEC_ID_MLP ||
2842  par->codec_id == AV_CODEC_ID_TRUEHD)
2844  else if (par->codec_id != AV_CODEC_ID_AAC)
2846  if (track->codec_delay > 0) {
2848  (AVRational){1, 1000000000},
2849  (AVRational){1, par->codec_id == AV_CODEC_ID_OPUS ?
2850  48000 : par->sample_rate});
2851  }
2852  if (track->seek_preroll > 0) {
2853  par->seek_preroll = av_rescale_q(track->seek_preroll,
2854  (AVRational){1, 1000000000},
2855  (AVRational){1, par->sample_rate});
2856  }
2857 
2858  return 0;
2859 }
2860 
2861 /* Performs the codec-specific part of parsing a video track. */
2862 static int mkv_parse_video_codec(MatroskaTrack *track, AVCodecParameters *par,
2863  const MatroskaDemuxContext *matroska,
2864  int *extradata_offset)
2865 {
2866  if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2867  track->codec_priv.size >= 40) {
2868  track->ms_compat = 1;
2869  par->bits_per_coded_sample = AV_RL16(track->codec_priv.data + 14);
2870  par->codec_tag = AV_RL32(track->codec_priv.data + 16);
2872  par->codec_tag);
2873  if (!par->codec_id)
2875  par->codec_tag);
2876  *extradata_offset = 40;
2877  return 0;
2878  } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2879  track->codec_priv.size >= 21) {
2880  enum AVCodecID codec_id;
2881  uint32_t fourcc;
2882  int ret = get_qt_codec(track, &fourcc, &codec_id);
2883  if (ret < 0)
2884  return ret;
2885  if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2886  fourcc = MKTAG('S','V','Q','3');
2888  }
2889  par->codec_id = codec_id;
2890  if (codec_id == AV_CODEC_ID_NONE)
2891  av_log(matroska->ctx, AV_LOG_ERROR,
2892  "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2893  if (track->codec_priv.size >= 86) {
2894  FFIOContext b;
2895  unsigned bit_depth = AV_RB16(track->codec_priv.data + 82);
2897  track->codec_priv.size);
2898  if (ff_get_qtpalette(codec_id, &b.pub, track->palette)) {
2899  bit_depth &= 0x1F;
2900  track->has_palette = 1;
2901  }
2903  }
2904  par->codec_tag = fourcc;
2905  return 0;
2906  }
2907 
2908  switch (par->codec_id) {
2909  case AV_CODEC_ID_RV10:
2910  case AV_CODEC_ID_RV20:
2911  case AV_CODEC_ID_RV30:
2912  case AV_CODEC_ID_RV40:
2913  *extradata_offset = 26;
2914  break;
2915  case AV_CODEC_ID_PRORES:
2916  if (track->codec_priv.size == 4)
2917  par->codec_tag = AV_RL32(track->codec_priv.data);
2918  break;
2919  case AV_CODEC_ID_VP9:
2920  /* we don't need any value stored in CodecPrivate.
2921  * make sure that it's not exported as extradata. */
2922  track->codec_priv.size = 0;
2923  break;
2924  }
2925 
2926  return 0;
2927 }
2928 
2929 /* Performs the generic part of parsing a video track. */
2930 static int mkv_parse_video(MatroskaTrack *track, AVStream *st,
2931  AVCodecParameters *par,
2932  const MatroskaDemuxContext *matroska,
2933  int *extradata_offset)
2934 {
2935  FFStream *const sti = ffstream(st);
2937  int display_width_mul = 1;
2938  int display_height_mul = 1;
2939  int ret;
2940 
2941  if (track->video.color_space.size == 4)
2942  par->codec_tag = AV_RL32(track->video.color_space.data);
2943 
2944  ret = mkv_parse_video_codec(track, par, matroska,
2945  extradata_offset);
2946  if (ret < 0)
2947  return ret;
2948 
2950  par->width = track->video.pixel_width;
2951  par->height = track->video.pixel_height;
2952 
2954  par->field_order = mkv_field_order(matroska, track->video.field_order);
2957 
2960  &display_width_mul, &display_height_mul);
2961 
2963  if (track->video.display_width && track->video.display_height &&
2964  par->height < INT64_MAX / track->video.display_width / display_width_mul &&
2965  par->width < INT64_MAX / track->video.display_height / display_height_mul)
2967  &st->sample_aspect_ratio.den,
2968  par->height * track->video.display_width * display_width_mul,
2969  par->width * track->video.display_height * display_height_mul,
2970  INT_MAX);
2971  }
2972  if (par->codec_id != AV_CODEC_ID_HEVC)
2974 
2975  if (track->default_duration) {
2976  int div = track->default_duration <= INT64_MAX ? 1 : 2;
2978  1000000000 / div, track->default_duration / div, 30000);
2979 #if FF_API_R_FRAME_RATE
2980  if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2981  && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2982  st->r_frame_rate = st->avg_frame_rate;
2983 #endif
2984  }
2985 
2986  /* export stereo mode flag as metadata tag */
2988  av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2989 
2990  /* export alpha mode flag as metadata tag */
2991  if (track->video.alpha_mode)
2992  av_dict_set_int(&st->metadata, "alpha_mode", 1, 0);
2993 
2994  /* if we have virtual track, mark the real tracks */
2996  for (int j = 0; j < track->operation.combine_planes.nb_elem; j++) {
2997  MatroskaTrack *tracks = matroska->tracks.elem;
2998  char buf[32];
3000  continue;
3001  snprintf(buf, sizeof(buf), "%s_%d",
3003  for (int k = 0; k < matroska->tracks.nb_elem; k++)
3004  if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
3005  av_dict_set(&tracks[k].stream->metadata,
3006  "stereo_mode", buf, 0);
3007  break;
3008  }
3009  }
3010  // add stream level stereo3d side data if it is a supported format
3014  int ret = mkv_stereo3d_conv(st, track->video.stereo_mode);
3015  if (ret < 0)
3016  return ret;
3017  }
3018 
3019  ret = mkv_parse_video_color(st, track);
3020  if (ret < 0)
3021  return ret;
3022  ret = mkv_parse_video_projection(st, track, matroska->ctx);
3023  if (ret < 0)
3024  return ret;
3025 
3026  return 0;
3027 }
3028 
3029 /* Performs the codec-specific part of parsing a subtitle track. */
3030 static int mkv_parse_subtitle_codec(MatroskaTrack *track, AVStream *st,
3031  AVCodecParameters *par,
3032  const MatroskaDemuxContext *matroska)
3033 {
3034  switch (par->codec_id) {
3036  if (track->codec_priv.size == 3) {
3037  int component_tag = track->codec_priv.data[0];
3038  int data_component_id = AV_RB16(track->codec_priv.data + 1);
3039 
3040  switch (data_component_id) {
3041  case 0x0008:
3042  // [0x30..0x37] are component tags utilized for
3043  // non-mobile captioning service ("profile A").
3044  if (component_tag >= 0x30 && component_tag <= 0x37) {
3046  }
3047  break;
3048  case 0x0012:
3049  // component tag 0x87 signifies a mobile/partial reception
3050  // (1seg) captioning service ("profile C").
3051  if (component_tag == 0x87) {
3053  }
3054  break;
3055  default:
3056  break;
3057  }
3058 
3059  if (par->profile == AV_PROFILE_UNKNOWN)
3060  av_log(matroska->ctx, AV_LOG_WARNING,
3061  "Unknown ARIB caption profile utilized: %02x / %04x\n",
3062  component_tag, data_component_id);
3063 
3064  track->codec_priv.size = 0;
3065  }
3066  break;
3067  case AV_CODEC_ID_WEBVTT:
3068  if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
3070  } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
3072  } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
3074  }
3075  break;
3076  }
3077 
3078  return 0;
3079 }
3080 
3082 {
3083  MatroskaDemuxContext *matroska = s->priv_data;
3084  MatroskaTrack *tracks = matroska->tracks.elem;
3085  int i, j, ret;
3086 
3087  for (i = 0; i < matroska->tracks.nb_elem; i++) {
3088  MatroskaTrack *track = &tracks[i];
3090  AVCodecParameters *par;
3092  int extradata_offset = 0;
3093  AVStream *st;
3094  char* key_id_base64 = NULL;
3095 
3096  /* Apply some sanity checks. */
3097  if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
3098  track->type != MATROSKA_TRACK_TYPE_AUDIO &&
3099  track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
3100  track->type != MATROSKA_TRACK_TYPE_METADATA) {
3101  av_log(matroska->ctx, AV_LOG_INFO,
3102  "Unknown or unsupported track type %"PRIu64"\n",
3103  track->type);
3104  continue;
3105  }
3106  if (!track->codec_id)
3107  continue;
3108 
3109  if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
3110  || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
3111  || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
3112  || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
3113  ) {
3114  av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
3115  continue;
3116  }
3117 
3118  if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
3119  isnan(track->audio.samplerate)) {
3120  av_log(matroska->ctx, AV_LOG_WARNING,
3121  "Invalid sample rate %f, defaulting to 8000 instead.\n",
3122  track->audio.samplerate);
3123  track->audio.samplerate = 8000;
3124  }
3125 
3126  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
3127  if (!track->default_duration && track->video.frame_rate > 0) {
3128  double default_duration = 1000000000 / track->video.frame_rate;
3129  if (default_duration > UINT64_MAX || default_duration < 0) {
3130  av_log(matroska->ctx, AV_LOG_WARNING,
3131  "Invalid frame rate %e. Cannot calculate default duration.\n",
3132  track->video.frame_rate);
3133  } else {
3134  track->default_duration = default_duration;
3135  }
3136  }
3137  if (track->video.display_width == -1)
3138  track->video.display_width = track->video.pixel_width;
3139  if (track->video.display_height == -1)
3140  track->video.display_height = track->video.pixel_height;
3141  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
3142  if (!track->audio.out_samplerate)
3143  track->audio.out_samplerate = track->audio.samplerate;
3144  }
3146  track->encodings.nb_elem,
3147  track, &key_id_base64, matroska->ctx);
3148  if (ret < 0)
3149  return ret;
3150 
3151  for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
3152  if (av_strstart(track->codec_id, ff_mkv_codec_tags[j].str, NULL)) {
3154  break;
3155  }
3156  }
3157 
3158  st = track->stream = avformat_new_stream(s, NULL);
3159  if (!st) {
3160  av_free(key_id_base64);
3161  return AVERROR(ENOMEM);
3162  }
3163  par = st->codecpar;
3164 
3165  par->codec_id = codec_id;
3166 
3167  if (track->flag_default)
3169  if (track->flag_forced)
3171  if (track->flag_comment)
3173  if (track->flag_hearingimpaired)
3175  if (track->flag_visualimpaired)
3177  if (track->flag_original.count > 0)
3180 
3181  if (key_id_base64) {
3182  /* export encryption key id as base64 metadata tag */
3183  av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
3185  }
3186 
3187  if (strcmp(track->language, "und"))
3188  av_dict_set(&st->metadata, "language", track->language, 0);
3189  av_dict_set(&st->metadata, "title", track->name, 0);
3190 
3191  if (track->time_scale < 0.01) {
3192  av_log(matroska->ctx, AV_LOG_WARNING,
3193  "Track TimestampScale too small %f, assuming 1.0.\n",
3194  track->time_scale);
3195  track->time_scale = 1.0;
3196  }
3197  avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
3198  1000 * 1000 * 1000); /* 64 bit pts in ns */
3199 
3200  /* convert the delay from ns to the track timebase */
3202  (AVRational){ 1, 1000000000 },
3203  st->time_base);
3204 
3205  type = track->type;
3206  if (par->codec_id == AV_CODEC_ID_WEBVTT)
3208  switch (type) {
3210  ret = mka_parse_audio(track, st, par, matroska,
3211  s, &extradata_offset);
3212  if (ret < 0)
3213  return ret;
3214  if (ret == SKIP_TRACK)
3215  continue;
3216  break;
3218  ret = mkv_parse_video(track, st, par, matroska, &extradata_offset);
3219  if (ret < 0)
3220  return ret;
3221  break;
3223  ret = mkv_parse_subtitle_codec(track, st, par, matroska);
3224  if (ret < 0)
3225  return ret;
3227 
3228  if (track->flag_textdescriptions)
3230  break;
3231  }
3232 
3233  if (par->codec_id == AV_CODEC_ID_NONE)
3234  av_log(matroska->ctx, AV_LOG_INFO,
3235  "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
3236 
3237  if (!par->extradata && track->codec_priv.size > extradata_offset) {
3238  const uint8_t *src = track->codec_priv.data + extradata_offset;
3239  unsigned extra_size = track->codec_priv.size - extradata_offset;
3240  ret = ff_alloc_extradata(par, extra_size);
3241  if (ret < 0)
3242  return ret;
3243  memcpy(par->extradata, src, extra_size);
3244  }
3245 
3246  ret = mkv_parse_block_addition_mappings(s, st, track);
3247  if (ret < 0)
3248  return ret;
3249  }
3250 
3251  return 0;
3252 }
3253 
3255 {
3256  FFFormatContext *const si = ffformatcontext(s);
3257  MatroskaDemuxContext *matroska = s->priv_data;
3258  EbmlList *attachments_list = &matroska->attachments;
3259  EbmlList *chapters_list = &matroska->chapters;
3260  MatroskaAttachment *attachments;
3261  MatroskaChapter *chapters;
3262  uint64_t max_start = 0;
3263  int64_t pos;
3264  Ebml ebml = { 0 };
3265  int i, j, res;
3266 
3267  matroska->ctx = s;
3268  matroska->cues_parsing_deferred = 1;
3269 
3270  /* First read the EBML header. */
3271  if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
3272  av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
3273  ebml_free(ebml_syntax, &ebml);
3274  return AVERROR_INVALIDDATA;
3275  }
3276  if (ebml.version > EBML_VERSION ||
3277  ebml.max_size > sizeof(uint64_t) ||
3278  ebml.id_length > sizeof(uint32_t) ||
3279  ebml.doctype_version > 3) {
3281  "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
3282  ebml.version, ebml.doctype, ebml.doctype_version);
3283  ebml_free(ebml_syntax, &ebml);
3284  return AVERROR_PATCHWELCOME;
3285  } else if (ebml.doctype_version == 3) {
3286  av_log(matroska->ctx, AV_LOG_WARNING,
3287  "EBML header using unsupported features\n"
3288  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
3289  ebml.version, ebml.doctype, ebml.doctype_version);
3290  }
3291  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
3292  if (!strcmp(ebml.doctype, matroska_doctypes[i]))
3293  break;
3295  av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
3296  if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
3297  ebml_free(ebml_syntax, &ebml);
3298  return AVERROR_INVALIDDATA;
3299  }
3300  }
3301  matroska->is_webm = !strcmp(ebml.doctype, "webm");
3302 
3303  ebml_free(ebml_syntax, &ebml);
3304 
3305  matroska->pkt = si->parse_pkt;
3306 
3307  /* The next thing is a segment. */
3308  pos = avio_tell(matroska->ctx->pb);
3309  res = ebml_parse(matroska, matroska_segments, matroska);
3310  // Try resyncing until we find an EBML_STOP type element.
3311  while (res != 1) {
3312  res = matroska_resync(matroska, pos);
3313  if (res < 0)
3314  return res;
3315  pos = avio_tell(matroska->ctx->pb);
3316  res = ebml_parse(matroska, matroska_segment, matroska);
3317  if (res == AVERROR(EIO)) // EOF is translated to EIO, this exists the loop on EOF
3318  return res;
3319  }
3320  /* Set data_offset as it might be needed later by seek_frame_generic. */
3321  if (matroska->current_id == MATROSKA_ID_CLUSTER)
3322  si->data_offset = avio_tell(matroska->ctx->pb) - 4;
3323  matroska_execute_seekhead(matroska);
3324 
3325  if (!matroska->time_scale)
3326  matroska->time_scale = 1000000;
3327  if (isnan(matroska->duration))
3328  matroska->duration = 0;
3329  if (matroska->duration)
3330  matroska->ctx->duration = matroska->duration * matroska->time_scale *
3331  1000 / AV_TIME_BASE;
3332  av_dict_set(&s->metadata, "title", matroska->title, 0);
3333  av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
3334 
3335  if (matroska->date_utc.size == 8)
3336  matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
3337 
3338  res = matroska_parse_tracks(s);
3339  if (res < 0)
3340  return res;
3341 
3342  attachments = attachments_list->elem;
3343  for (j = 0; j < attachments_list->nb_elem; j++) {
3344  if (!(attachments[j].filename && attachments[j].mime &&
3345  attachments[j].bin.data && attachments[j].bin.size > 0)) {
3346  av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
3347  } else {
3349  if (!st)
3350  break;
3351  av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
3352  av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
3353  if (attachments[j].description)
3354  av_dict_set(&st->metadata, "title", attachments[j].description, 0);
3356 
3357  for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3358  if (av_strstart(attachments[j].mime, mkv_image_mime_tags[i].str, NULL)) {
3360  break;
3361  }
3362  }
3363 
3364  attachments[j].stream = st;
3365 
3366  if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
3367  res = ff_add_attached_pic(s, st, NULL, &attachments[j].bin.buf, 0);
3368  if (res < 0)
3369  return res;
3370  } else {
3372  if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
3373  break;
3374  memcpy(st->codecpar->extradata, attachments[j].bin.data,
3375  attachments[j].bin.size);
3376 
3377  for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3378  if (av_strstart(attachments[j].mime, mkv_mime_tags[i].str, NULL)) {
3380  break;
3381  }
3382  }
3383  }
3384  }
3385  }
3386 
3387  chapters = chapters_list->elem;
3388  for (i = 0; i < chapters_list->nb_elem; i++)
3389  if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
3390  (max_start == 0 || chapters[i].start > max_start)) {
3391  chapters[i].chapter =
3392  avpriv_new_chapter(s, chapters[i].uid,
3393  (AVRational) { 1, 1000000000 },
3394  chapters[i].start, chapters[i].end,
3395  chapters[i].title);
3396  max_start = chapters[i].start;
3397  }
3398 
3399  matroska_add_index_entries(matroska);
3400 
3402 
3403  return 0;
3405 
3406 /*
3407  * Put one packet in an application-supplied AVPacket struct.
3408  * Returns 0 on success or -1 on failure.
3409  */
3410 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3411  AVPacket *pkt)
3412 {
3413  if (matroska->queue.head) {
3414  MatroskaTrack *tracks = matroska->tracks.elem;
3415  MatroskaTrack *track;
3416 
3417  avpriv_packet_list_get(&matroska->queue, pkt);
3418  track = &tracks[pkt->stream_index];
3419  if (track->has_palette) {
3421  if (!pal) {
3422  av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3423  } else {
3424  memcpy(pal, track->palette, AVPALETTE_SIZE);
3425  }
3426  track->has_palette = 0;
3427  }
3428  return 0;
3429  }
3430 
3431  return -1;
3432 }
3433 
3434 /*
3435  * Free all packets in our internal queue.
3436  */
3437 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3438 {
3439  avpriv_packet_list_free(&matroska->queue);
3440 }
3441 
3442 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3443  int size, int type, AVIOContext *pb,
3444  uint32_t lace_size[256], int *laces)
3445 {
3446  int n;
3447  uint8_t *data = *buf;
3448 
3449  if (!type) {
3450  *laces = 1;
3451  lace_size[0] = size;
3452  return 0;
3453  }
3454 
3455  if (size <= 0)
3456  return AVERROR_INVALIDDATA;
3457 
3458  *laces = *data + 1;
3459  data += 1;
3460  size -= 1;
3461 
3462  switch (type) {
3463  case 0x1: /* Xiph lacing */
3464  {
3465  uint8_t temp;
3466  uint32_t total = 0;
3467  for (n = 0; n < *laces - 1; n++) {
3468  lace_size[n] = 0;
3469 
3470  do {
3471  if (size <= total)
3472  return AVERROR_INVALIDDATA;
3473  temp = *data;
3474  total += temp;
3475  lace_size[n] += temp;
3476  data += 1;
3477  size -= 1;
3478  } while (temp == 0xff);
3479  }
3480  if (size < total)
3481  return AVERROR_INVALIDDATA;
3482 
3483  lace_size[n] = size - total;
3484  break;
3485  }
3486 
3487  case 0x2: /* fixed-size lacing */
3488  if (size % (*laces))
3489  return AVERROR_INVALIDDATA;
3490  for (n = 0; n < *laces; n++)
3491  lace_size[n] = size / *laces;
3492  break;
3493 
3494  case 0x3: /* EBML lacing */
3495  {
3496  uint64_t num;
3497  uint64_t total;
3498  int offset;
3499 
3500  avio_skip(pb, 4);
3501 
3502  n = ebml_read_num(matroska, pb, 8, &num, 1);
3503  if (n < 0)
3504  return n;
3505  if (num > INT_MAX)
3506  return AVERROR_INVALIDDATA;
3507 
3508  total = lace_size[0] = num;
3509  offset = n;
3510  for (n = 1; n < *laces - 1; n++) {
3511  int64_t snum;
3512  int r;
3513  r = matroska_ebmlnum_sint(matroska, pb, &snum);
3514  if (r < 0)
3515  return r;
3516  if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3517  return AVERROR_INVALIDDATA;
3518 
3519  lace_size[n] = lace_size[n - 1] + snum;
3520  total += lace_size[n];
3521  offset += r;
3522  }
3523  data += offset;
3524  size -= offset;
3525  if (size < total)
3526  return AVERROR_INVALIDDATA;
3527 
3528  lace_size[*laces - 1] = size - total;
3529  break;
3530  }
3531  }
3533  *buf = data;
3534 
3535  return 0;
3536 }
3537 
3538 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3539  MatroskaTrack *track, AVStream *st,
3540  uint8_t *data, int size, uint64_t timecode,
3541  int64_t pos)
3542 {
3543  const int a = st->codecpar->block_align;
3544  const int sps = track->audio.sub_packet_size;
3545  const int cfs = track->audio.coded_framesize;
3546  const int h = track->audio.sub_packet_h;
3547  const int w = track->audio.frame_size;
3548  int y = track->audio.sub_packet_cnt;
3549  int x;
3550 
3551  if (!track->audio.pkt_cnt) {
3552  if (track->audio.sub_packet_cnt == 0)
3553  track->audio.buf_timecode = timecode;
3554  if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3555  if (size < cfs * h / 2) {
3556  av_log(matroska->ctx, AV_LOG_ERROR,
3557  "Corrupt int4 RM-style audio packet size\n");
3558  return AVERROR_INVALIDDATA;
3559  }
3560  for (x = 0; x < h / 2; x++)
3561  memcpy(track->audio.buf + x * 2 * w + y * cfs,
3562  data + x * cfs, cfs);
3563  } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3564  if (size < w) {
3565  av_log(matroska->ctx, AV_LOG_ERROR,
3566  "Corrupt sipr RM-style audio packet size\n");
3567  return AVERROR_INVALIDDATA;
3568  }
3569  memcpy(track->audio.buf + y * w, data, w);
3570  } else {
3571  if (size < w) {
3572  av_log(matroska->ctx, AV_LOG_ERROR,
3573  "Corrupt generic RM-style audio packet size\n");
3574  return AVERROR_INVALIDDATA;
3575  }
3576  for (x = 0; x < w / sps; x++)
3577  memcpy(track->audio.buf +
3578  sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3579  data + x * sps, sps);
3580  }
3581 
3582  if (++track->audio.sub_packet_cnt >= h) {
3583  if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3584  ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3585  track->audio.sub_packet_cnt = 0;
3586  track->audio.pkt_cnt = h * w / a;
3587  }
3588  }
3589 
3590  while (track->audio.pkt_cnt) {
3591  int ret;
3592  AVPacket *pkt = matroska->pkt;
3593 
3594  ret = av_new_packet(pkt, a);
3595  if (ret < 0) {
3596  return ret;
3597  }
3598  memcpy(pkt->data,
3599  track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3600  a);
3601  pkt->pts = track->audio.buf_timecode;
3603  pkt->pos = pos;
3604  pkt->stream_index = st->index;
3605  ret = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
3606  if (ret < 0) {
3608  return AVERROR(ENOMEM);
3609  }
3610  }
3611 
3612  return 0;
3613 }
3614 
3615 /* reconstruct full wavpack blocks from mangled matroska ones */
3616 static int matroska_parse_wavpack(MatroskaTrack *track,
3617  uint8_t **data, int *size)
3618 {
3619  uint8_t *dst = NULL;
3620  uint8_t *src = *data;
3621  int dstlen = 0;
3622  int srclen = *size;
3623  uint32_t samples;
3624  uint16_t ver;
3625  int ret, offset = 0;
3626 
3627  if (srclen < 12)
3628  return AVERROR_INVALIDDATA;
3629 
3630  av_assert1(track->stream->codecpar->extradata_size >= 2);
3631  ver = AV_RL16(track->stream->codecpar->extradata);
3632 
3633  samples = AV_RL32(src);
3634  src += 4;
3635  srclen -= 4;
3636 
3637  while (srclen >= 8) {
3638  int multiblock;
3639  uint32_t blocksize;
3640  uint8_t *tmp;
3641 
3642  uint32_t flags = AV_RL32(src);
3643  uint32_t crc = AV_RL32(src + 4);
3644  src += 8;
3645  srclen -= 8;
3646 
3647  multiblock = (flags & 0x1800) != 0x1800;
3648  if (multiblock) {
3649  if (srclen < 4) {
3651  goto fail;
3652  }
3653  blocksize = AV_RL32(src);
3654  src += 4;
3655  srclen -= 4;
3656  } else
3657  blocksize = srclen;
3658 
3659  if (blocksize > srclen) {
3661  goto fail;
3662  }
3663 
3664  tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3665  if (!tmp) {
3666  ret = AVERROR(ENOMEM);
3667  goto fail;
3668  }
3669  dst = tmp;
3670  dstlen += blocksize + 32;
3671 
3672  AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3673  AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3674  AV_WL16(dst + offset + 8, ver); // version
3675  AV_WL16(dst + offset + 10, 0); // track/index_no
3676  AV_WL32(dst + offset + 12, 0); // total samples
3677  AV_WL32(dst + offset + 16, 0); // block index
3678  AV_WL32(dst + offset + 20, samples); // number of samples
3679  AV_WL32(dst + offset + 24, flags); // flags
3680  AV_WL32(dst + offset + 28, crc); // crc
3681  memcpy(dst + offset + 32, src, blocksize); // block data
3682 
3683  src += blocksize;
3684  srclen -= blocksize;
3685  offset += blocksize + 32;
3686  }
3687 
3688  memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3689 
3690  *data = dst;
3691  *size = dstlen;
3692 
3693  return 0;
3695 fail:
3696  av_freep(&dst);
3697  return ret;
3698 }
3699 
3700 static int matroska_parse_prores(MatroskaTrack *track,
3701  uint8_t **data, int *size)
3702 {
3703  uint8_t *dst;
3704  int dstlen = *size + 8;
3705 
3706  dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3707  if (!dst)
3708  return AVERROR(ENOMEM);
3709 
3710  AV_WB32(dst, dstlen);
3711  AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3712  memcpy(dst + 8, *data, dstlen - 8);
3713  memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3714 
3715  *data = dst;
3716  *size = dstlen;
3717 
3718  return 0;
3719 }
3720 
3721 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3722  MatroskaTrack *track,
3723  AVStream *st,
3724  uint8_t *data, int data_len,
3725  uint64_t timecode,
3726  uint64_t duration,
3727  int64_t pos)
3728 {
3729  AVPacket *pkt = matroska->pkt;
3730  uint8_t *id, *settings, *text, *buf;
3731  int id_len, settings_len, text_len;
3732  uint8_t *p, *q;
3733  int err;
3734 
3735  if (data_len <= 0)
3736  return AVERROR_INVALIDDATA;
3737 
3738  p = data;
3739  q = data + data_len;
3740 
3741  id = p;
3742  id_len = -1;
3743  while (p < q) {
3744  if (*p == '\r' || *p == '\n') {
3745  id_len = p - id;
3746  if (*p == '\r')
3747  p++;
3748  break;
3749  }
3750  p++;
3751  }
3752 
3753  if (p >= q || *p != '\n')
3754  return AVERROR_INVALIDDATA;
3755  p++;
3756 
3757  settings = p;
3758  settings_len = -1;
3759  while (p < q) {
3760  if (*p == '\r' || *p == '\n') {
3761  settings_len = p - settings;
3762  if (*p == '\r')
3763  p++;
3764  break;
3765  }
3766  p++;
3767  }
3768 
3769  if (p >= q || *p != '\n')
3770  return AVERROR_INVALIDDATA;
3771  p++;
3772 
3773  text = p;
3774  text_len = q - p;
3775  while (text_len > 0) {
3776  const int len = text_len - 1;
3777  const uint8_t c = p[len];
3778  if (c != '\r' && c != '\n')
3779  break;
3780  text_len = len;
3781  }
3782 
3783  if (text_len <= 0)
3784  return AVERROR_INVALIDDATA;
3785 
3786  err = av_new_packet(pkt, text_len);
3787  if (err < 0) {
3788  return err;
3789  }
3790 
3791  memcpy(pkt->data, text, text_len);
3792 
3793  if (id_len > 0) {
3796  id_len);
3797  if (!buf) {
3799  return AVERROR(ENOMEM);
3800  }
3801  memcpy(buf, id, id_len);
3802  }
3803 
3804  if (settings_len > 0) {
3807  settings_len);
3808  if (!buf) {
3810  return AVERROR(ENOMEM);
3811  }
3812  memcpy(buf, settings, settings_len);
3813  }
3814 
3815  // Do we need this for subtitles?
3816  // pkt->flags = AV_PKT_FLAG_KEY;
3817 
3818  pkt->stream_index = st->index;
3819  pkt->pts = timecode;
3820 
3821  // Do we need this for subtitles?
3822  // pkt->dts = timecode;
3823 
3824  pkt->duration = duration;
3825  pkt->pos = pos;
3826 
3827  err = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
3828  if (err < 0) {
3830  return AVERROR(ENOMEM);
3831  }
3832 
3833  return 0;
3834 }
3835 
3837  MatroskaTrack *track, AVPacket *pkt,
3838  const uint8_t *data, int size, uint64_t id)
3839 {
3840  const EbmlList *mappings_list = &track->block_addition_mappings;
3841  MatroskaBlockAdditionMapping *mappings = mappings_list->elem, *mapping = NULL;
3842  uint8_t *side_data;
3843  int res;
3844 
3845  if (!matroska->is_webm && track->max_block_additional_id && id > track->max_block_additional_id) {
3846  int strict = matroska->ctx->strict_std_compliance >= FF_COMPLIANCE_STRICT;
3847  av_log(matroska->ctx, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
3848  "BlockAddID %"PRIu64" is higher than the reported MaxBlockAdditionID %"PRIu64" "
3849  "for Track with TrackNumber %"PRIu64"\n", id, track->max_block_additional_id,
3850  track->num);
3851  if (strict)
3852  return AVERROR_INVALIDDATA;
3853  }
3854 
3855  for (int i = 0; i < mappings_list->nb_elem; i++) {
3856  if (id != mappings[i].value)
3857  continue;
3858  mapping = &mappings[i];
3859  break;
3860  }
3861 
3862  if (id != 1 && !matroska->is_webm && !mapping) {
3863  av_log(matroska->ctx, AV_LOG_WARNING, "BlockAddID %"PRIu64" has no mapping. Skipping\n", id);
3864  return 0;
3865  }
3866 
3867  if (mapping && mapping->type)
3868  id = mapping->type;
3869 
3870  switch (id) {
3872  GetByteContext bc;
3873  int country_code, provider_code;
3874  int provider_oriented_code, application_identifier;
3875  size_t hdrplus_size;
3876  AVDynamicHDRPlus *hdrplus;
3877 
3878  if (size < 6)
3879  break; //ignore
3880 
3881  bytestream2_init(&bc, data, size);
3882 
3883  /* ITU-T T.35 metadata */
3884  country_code = bytestream2_get_byteu(&bc);
3885  provider_code = bytestream2_get_be16u(&bc);
3886 
3887  if (country_code != 0xB5 || provider_code != 0x3C)
3888  break; // ignore
3889 
3890  provider_oriented_code = bytestream2_get_be16u(&bc);
3891  application_identifier = bytestream2_get_byteu(&bc);
3892 
3893  if (provider_oriented_code != 1 || application_identifier != 4)
3894  break; // ignore
3895 
3896  hdrplus = av_dynamic_hdr_plus_alloc(&hdrplus_size);
3897  if (!hdrplus)
3898  return AVERROR(ENOMEM);
3899 
3900  if ((res = av_dynamic_hdr_plus_from_t35(hdrplus, bc.buffer,
3901  bytestream2_get_bytes_left(&bc))) < 0 ||
3903  (uint8_t *)hdrplus, hdrplus_size)) < 0) {
3904  av_free(hdrplus);
3905  return res;
3906  }
3907 
3908  return 0;
3909  }
3910  default:
3911  break;
3912  }
3913 
3915  size + (size_t)8);
3916  if (!side_data)
3917  return AVERROR(ENOMEM);
3918 
3919  AV_WB64(side_data, id);
3920  memcpy(side_data + 8, data, size);
3921 
3922  return 0;
3923 }
3924 
3925 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3926  MatroskaTrack *track, AVStream *st,
3927  AVBufferRef *buf, uint8_t *data, int pkt_size,
3928  uint64_t timecode, uint64_t lace_duration,
3929  int64_t pos, int is_keyframe,
3930  MatroskaBlockMore *blockmore, int nb_blockmore,
3931  int64_t discard_padding)
3932 {
3933  uint8_t *pkt_data = data;
3934  int res = 0;
3935  AVPacket *pkt = matroska->pkt;
3936 
3937  if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3938  res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3939  if (res < 0) {
3940  av_log(matroska->ctx, AV_LOG_ERROR,
3941  "Error parsing a wavpack block.\n");
3942  goto fail;
3943  }
3944  if (!buf)
3945  av_freep(&data);
3946  buf = NULL;
3947  }
3948 
3949  if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3950  AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3951  res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3952  if (res < 0) {
3953  av_log(matroska->ctx, AV_LOG_ERROR,
3954  "Error parsing a prores block.\n");
3955  goto fail;
3956  }
3957  if (!buf)
3958  av_freep(&data);
3959  buf = NULL;
3960  }
3961 
3962  if (!pkt_size && !nb_blockmore)
3963  goto no_output;
3964 
3965  if (!matroska->is_webm && nb_blockmore && !track->max_block_additional_id) {
3966  int strict = matroska->ctx->strict_std_compliance >= FF_COMPLIANCE_STRICT;
3967  av_log(matroska->ctx, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
3968  "Unexpected BlockAdditions found in a Block from Track with TrackNumber %"PRIu64" "
3969  "where MaxBlockAdditionID is 0\n", track->num);
3970  if (strict) {
3971  res = AVERROR_INVALIDDATA;
3972  goto fail;
3973  }
3974  }
3975 
3976  if (!buf)
3977  pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3978  NULL, NULL, 0);
3979  else
3980  pkt->buf = av_buffer_ref(buf);
3981 
3982  if (!pkt->buf) {
3983  res = AVERROR(ENOMEM);
3984  goto fail;
3985  }
3986 
3987  pkt->data = pkt_data;
3988  pkt->size = pkt_size;
3989  pkt->flags = is_keyframe;
3990  pkt->stream_index = st->index;
3991 
3992  for (int i = 0; i < nb_blockmore; i++) {
3993  MatroskaBlockMore *more = &blockmore[i];
3994 
3995  if (!more->additional.size)
3996  continue;
3997 
3998  res = matroska_parse_block_additional(matroska, track, pkt, more->additional.data,
3999  more->additional.size, more->additional_id);
4000  if (res < 0) {
4002  return res;
4003  }
4004  }
4005 
4006  if (discard_padding) {
4007  uint8_t *side_data = av_packet_new_side_data(pkt,
4009  10);
4010  if (!side_data) {
4012  return AVERROR(ENOMEM);
4013  }
4014  discard_padding = av_rescale_q(discard_padding,
4015  (AVRational){1, 1000000000},
4016  (AVRational){1, st->codecpar->sample_rate});
4017  if (discard_padding > 0) {
4018  AV_WL32(side_data + 4, discard_padding);
4019  } else {
4020  AV_WL32(side_data, -discard_padding);
4021  }
4022  }
4023 
4024  if (track->ms_compat)
4025  pkt->dts = timecode;
4026  else
4027  pkt->pts = timecode;
4028  pkt->pos = pos;
4029  pkt->duration = lace_duration;
4030 
4031  res = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
4032  if (res < 0) {
4034  return AVERROR(ENOMEM);
4035  }
4036 
4037  return 0;
4038 
4039 no_output:
4041  if (!buf)
4042  av_free(pkt_data);
4043  return res;
4044 }
4045 
4046 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
4047  int size, int64_t pos, uint64_t cluster_time,
4048  uint64_t block_duration, int is_keyframe,
4049  MatroskaBlockMore *blockmore, int nb_blockmore,
4050  int64_t cluster_pos, int64_t discard_padding)
4051 {
4052  uint64_t timecode = AV_NOPTS_VALUE;
4053  MatroskaTrack *track;
4054  FFIOContext pb;
4055  int res = 0;
4056  AVStream *st;
4057  int16_t block_time;
4058  uint32_t lace_size[256];
4059  int n, flags, laces = 0;
4060  uint64_t num;
4061  int trust_default_duration;
4062 
4063  av_assert1(buf);
4064 
4066 
4067  if ((n = ebml_read_num(matroska, &pb.pub, 8, &num, 1)) < 0)
4068  return n;
4069  data += n;
4070  size -= n;
4071 
4072  track = matroska_find_track_by_num(matroska, num);
4073  if (!track || size < 3)
4074  return AVERROR_INVALIDDATA;
4075 
4076  if (!(st = track->stream)) {
4077  av_log(matroska->ctx, AV_LOG_VERBOSE,
4078  "No stream associated to TrackNumber %"PRIu64". "
4079  "Ignoring Block with this TrackNumber.\n", num);
4080  return 0;
4081  }
4082 
4083  if (st->discard >= AVDISCARD_ALL)
4084  return res;
4085  if (block_duration > INT64_MAX)
4086  block_duration = INT64_MAX;
4087 
4088  block_time = sign_extend(AV_RB16(data), 16);
4089  data += 2;
4090  flags = *data++;
4091  size -= 3;
4092  if (is_keyframe == -1)
4093  is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
4094 
4095  if (cluster_time != (uint64_t) -1 &&
4096  (block_time >= 0 || cluster_time >= -block_time)) {
4097  uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
4098  timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
4099  if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
4100  timecode < track->end_timecode)
4101  is_keyframe = 0; /* overlapping subtitles are not key frame */
4102  if (is_keyframe) {
4103  ff_reduce_index(matroska->ctx, st->index);
4104  av_add_index_entry(st, cluster_pos, timecode, 0, 0,
4106  }
4107  }
4108 
4109  if (matroska->skip_to_keyframe &&
4110  track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
4111  // Compare signed timecodes. Timecode may be negative due to codec delay
4112  // offset. We don't support timestamps greater than int64_t anyway - see
4113  // AVPacket's pts.
4114  if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
4115  return res;
4116  if (is_keyframe)
4117  matroska->skip_to_keyframe = 0;
4118  else if (!ffstream(st)->skip_to_keyframe) {
4119  av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
4120  matroska->skip_to_keyframe = 0;
4121  }
4122  }
4123 
4124  res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
4125  &pb.pub, lace_size, &laces);
4126  if (res < 0) {
4127  av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
4128  return res;
4129  }
4130 
4131  trust_default_duration = track->default_duration != 0;
4132  if (track->audio.samplerate == 8000 && trust_default_duration) {
4133  // If this is needed for more codecs, then add them here
4134  if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
4135  if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
4136  trust_default_duration = 0;
4137  }
4138  }
4139 
4140  if (!block_duration && trust_default_duration)
4141  block_duration = track->default_duration * laces / matroska->time_scale;
4142 
4143  if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
4144  track->end_timecode =
4145  FFMAX(track->end_timecode, timecode + block_duration);
4146 
4147  for (n = 0; n < laces; n++) {
4148  int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
4149  uint8_t *out_data = data;
4150  int out_size = lace_size[n];
4151 
4152  if (track->needs_decoding) {
4153  res = matroska_decode_buffer(&out_data, &out_size, track);
4154  if (res < 0)
4155  return res;
4156  /* Given that we are here means that out_data is no longer
4157  * owned by buf, so set it to NULL. This depends upon
4158  * zero-length header removal compression being ignored. */
4159  av_assert1(out_data != data);
4160  buf = NULL;
4161  }
4162 
4163  if (track->audio.buf) {
4164  res = matroska_parse_rm_audio(matroska, track, st,
4165  out_data, out_size,
4166  timecode, pos);
4167  if (!buf)
4168  av_free(out_data);
4169  if (res)
4170  return res;
4171  } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
4172  res = matroska_parse_webvtt(matroska, track, st,
4173  out_data, out_size,
4174  timecode, lace_duration,
4175  pos);
4176  if (!buf)
4177  av_free(out_data);
4178  if (res)
4179  return res;
4180  } else {
4181  res = matroska_parse_frame(matroska, track, st, buf, out_data,
4182  out_size, timecode, lace_duration,
4183  pos, !n ? is_keyframe : 0,
4184  blockmore, nb_blockmore,
4185  discard_padding);
4186  if (res)
4187  return res;
4188  }
4189 
4190  if (timecode != AV_NOPTS_VALUE)
4191  timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
4192  data += lace_size[n];
4193  }
4194 
4195  return 0;
4196 }
4197 
4198 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
4199 {
4200  MatroskaCluster *cluster = &matroska->current_cluster;
4201  MatroskaBlock *block = &cluster->block;
4202  int res;
4203 
4204  av_assert0(matroska->num_levels <= 2);
4205 
4206  if (matroska->num_levels == 1) {
4207  res = ebml_parse(matroska, matroska_segment, NULL);
4208 
4209  if (res == 1) {
4210  /* Found a cluster: subtract the size of the ID already read. */
4211  cluster->pos = avio_tell(matroska->ctx->pb) - 4;
4212 
4213  res = ebml_parse(matroska, matroska_cluster_enter, cluster);
4214  if (res < 0)
4215  return res;
4216  }
4217  }
4218 
4219  if (matroska->num_levels == 2) {
4220  /* We are inside a cluster. */
4221  res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
4222 
4223  if (res >= 0 && block->bin.size > 0) {
4224  int is_keyframe = block->non_simple ? block->reference.count == 0 : -1;
4225 
4226  res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
4227  block->bin.size, block->bin.pos,
4228  cluster->timecode, block->duration,
4229  is_keyframe, block->blockmore.elem,
4230  block->blockmore.nb_elem, cluster->pos,
4231  block->discard_padding);
4232  }
4233 
4235  memset(block, 0, sizeof(*block));
4236  } else if (!matroska->num_levels) {
4237  if (!avio_feof(matroska->ctx->pb)) {
4238  avio_r8(matroska->ctx->pb);
4239  if (!avio_feof(matroska->ctx->pb)) {
4240  av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
4241  "end of segment.\n");
4242  return AVERROR_INVALIDDATA;
4243  }
4244  }
4245  matroska->done = 1;
4246  return AVERROR_EOF;
4247  }
4248 
4249  return res;
4250 }
4251 
4253 {
4254  MatroskaDemuxContext *matroska = s->priv_data;
4255  int ret = 0;
4256 
4257  if (matroska->resync_pos == -1) {
4258  // This can only happen if generic seeking has been used.
4259  matroska->resync_pos = avio_tell(s->pb);
4260  }
4261 
4262  while (matroska_deliver_packet(matroska, pkt)) {
4263  if (matroska->done)
4264  return (ret < 0) ? ret : AVERROR_EOF;
4265  if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
4266  ret = matroska_resync(matroska, matroska->resync_pos);
4267  }
4268 
4269  return 0;
4270 }
4271 
4272 static int matroska_read_seek(AVFormatContext *s, int stream_index,
4273  int64_t timestamp, int flags)
4274 {
4275  MatroskaDemuxContext *matroska = s->priv_data;
4276  MatroskaTrack *tracks = NULL;
4277  AVStream *st = s->streams[stream_index];
4278  FFStream *const sti = ffstream(st);
4279  int i, index;
4280 
4281  /* Parse the CUES now since we need the index data to seek. */
4282  if (matroska->cues_parsing_deferred > 0) {
4283  matroska->cues_parsing_deferred = 0;
4284  matroska_parse_cues(matroska);
4285  }
4286 
4287  if (!sti->nb_index_entries)
4288  goto err;
4289  timestamp = FFMAX(timestamp, sti->index_entries[0].timestamp);
4290 
4291  if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 ||
4292  index == sti->nb_index_entries - 1) {
4293  matroska_reset_status(matroska, 0, sti->index_entries[sti->nb_index_entries - 1].pos);
4294  while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 ||
4295  index == sti->nb_index_entries - 1) {
4296  matroska_clear_queue(matroska);
4297  if (matroska_parse_cluster(matroska) < 0)
4298  break;
4299  }
4300  }
4301 
4302  matroska_clear_queue(matroska);
4303  if (index < 0 || (matroska->cues_parsing_deferred < 0 &&
4304  index == sti->nb_index_entries - 1))
4305  goto err;
4306 
4307  tracks = matroska->tracks.elem;
4308  for (i = 0; i < matroska->tracks.nb_elem; i++) {
4309  tracks[i].audio.pkt_cnt = 0;
4310  tracks[i].audio.sub_packet_cnt = 0;
4311  tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
4312  tracks[i].end_timecode = 0;
4313  }
4314 
4315  /* We seek to a level 1 element, so set the appropriate status. */
4316  matroska_reset_status(matroska, 0, sti->index_entries[index].pos);
4317  if (flags & AVSEEK_FLAG_ANY) {
4318  sti->skip_to_keyframe = 0;
4319  matroska->skip_to_timecode = timestamp;
4320  } else {
4321  sti->skip_to_keyframe = 1;
4322  matroska->skip_to_timecode = sti->index_entries[index].timestamp;
4323  }
4324  matroska->skip_to_keyframe = 1;
4325  matroska->done = 0;
4327  return 0;
4328 err:
4329  // slightly hackish but allows proper fallback to
4330  // the generic seeking code.
4331  matroska_reset_status(matroska, 0, -1);
4332  matroska->resync_pos = -1;
4333  matroska_clear_queue(matroska);
4335  matroska->skip_to_keyframe = 0;
4336  matroska->done = 0;
4337  return -1;
4338 }
4339 
4341 {
4342  MatroskaDemuxContext *matroska = s->priv_data;
4343  MatroskaTrack *tracks = matroska->tracks.elem;
4344  int n;
4345 
4346  matroska_clear_queue(matroska);
4347 
4348  for (n = 0; n < matroska->tracks.nb_elem; n++)
4349  if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
4350  av_freep(&tracks[n].audio.buf);
4351  ebml_free(matroska_segment, matroska);
4352 
4353  return 0;
4354 }
4355 
4356 #if CONFIG_WEBM_DASH_MANIFEST_DEMUXER
4357 typedef struct {
4358  int64_t start_time_ns;
4359  int64_t end_time_ns;
4360  int64_t start_offset;
4361  int64_t end_offset;
4362 } CueDesc;
4363 
4364 /* This function searches all the Cues and returns the CueDesc corresponding to
4365  * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
4366  * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration or
4367  * if an error occurred.
4368  */
4369 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
4370  MatroskaDemuxContext *matroska = s->priv_data;
4371  FFStream *const sti = ffstream(s->streams[0]);
4372  AVIndexEntry *const index_entries = sti->index_entries;
4373  int nb_index_entries = sti->nb_index_entries;
4374  CueDesc cue_desc;
4375  int i;
4376 
4377  if (ts >= (int64_t)(matroska->duration * matroska->time_scale))
4378  return (CueDesc) {-1, -1, -1, -1};
4379  for (i = 1; i < nb_index_entries; i++) {
4380  if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
4381  index_entries[i].timestamp * matroska->time_scale > ts) {
4382  break;
4383  }
4384  }
4385  --i;
4386  if (index_entries[i].timestamp > matroska->duration)
4387  return (CueDesc) {-1, -1, -1, -1};
4388  cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
4389  cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
4390  if (i != nb_index_entries - 1) {
4391  cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
4392  cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
4393  } else {
4394  cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
4395  // FIXME: this needs special handling for files where Cues appear
4396  // before Clusters. the current logic assumes Cues appear after
4397  // Clusters.
4398  cue_desc.end_offset = cues_start - matroska->segment_start;
4399  }
4400  return cue_desc;
4401 }
4402 
4403 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
4404 {
4405  MatroskaDemuxContext *matroska = s->priv_data;
4406  AVStream *const st = s->streams[0];