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 
33 #include <inttypes.h>
34 #include <stdio.h>
35 
36 #include "libavutil/avstring.h"
37 #include "libavutil/base64.h"
38 #include "libavutil/dict.h"
39 #include "libavutil/intfloat.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/lzo.h"
43 #include "libavutil/mathematics.h"
44 #include "libavutil/opt.h"
46 #include "libavutil/spherical.h"
47 
48 #include "libavcodec/bytestream.h"
49 #include "libavcodec/flac.h"
50 #include "libavcodec/mpeg4audio.h"
52 
53 #include "avformat.h"
54 #include "avio_internal.h"
55 #include "internal.h"
56 #include "isom.h"
57 #include "matroska.h"
58 #include "oggdec.h"
59 /* For ff_codec_get_id(). */
60 #include "riff.h"
61 #include "rmsipr.h"
62 
63 #if CONFIG_BZLIB
64 #include <bzlib.h>
65 #endif
66 #if CONFIG_ZLIB
67 #include <zlib.h>
68 #endif
69 
70 #include "qtpalette.h"
71 
72 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
73 #define NEEDS_CHECKING 2 /* Indicates that some error checks
74  * still need to be performed */
75 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
76  * syntax level used for parsing ended. */
77 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
78  * of unkown, potentially damaged data is encountered,
79  * it is considered an error. */
80 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
81  * to this many bytes of unknown data for the
82  * SKIP_THRESHOLD check. */
83 
84 typedef enum {
90  EBML_UTF8,
91  EBML_BIN,
97 
98 typedef struct CountedElement {
99  union {
100  uint64_t u;
101  int64_t i;
102  double f;
103  char *s;
104  } el;
105  unsigned count;
107 
108 typedef const struct EbmlSyntax {
109  uint32_t id;
113  size_t data_offset;
114  union {
115  int64_t i;
116  uint64_t u;
117  double f;
118  const char *s;
119  const struct EbmlSyntax *n;
120  } def;
121 } EbmlSyntax;
122 
123 typedef struct EbmlList {
124  int nb_elem;
125  unsigned int alloc_elem_size;
126  void *elem;
128 
129 typedef struct EbmlBin {
130  int size;
133  int64_t pos;
135 
136 typedef struct Ebml {
137  uint64_t version;
138  uint64_t max_size;
139  uint64_t id_length;
140  char *doctype;
141  uint64_t doctype_version;
142 } Ebml;
143 
144 typedef struct MatroskaTrackCompression {
145  uint64_t algo;
148 
149 typedef struct MatroskaTrackEncryption {
150  uint64_t algo;
153 
154 typedef struct MatroskaTrackEncoding {
155  uint64_t scope;
156  uint64_t type;
160 
161 typedef struct MatroskaMasteringMeta {
162  double r_x;
163  double r_y;
164  double g_x;
165  double g_y;
166  double b_x;
167  double b_y;
168  double white_x;
169  double white_y;
173 
174 typedef struct MatroskaTrackVideoColor {
177  uint64_t chroma_sub_horz;
178  uint64_t chroma_sub_vert;
179  uint64_t cb_sub_horz;
180  uint64_t cb_sub_vert;
183  uint64_t range;
184  uint64_t transfer_characteristics;
185  uint64_t primaries;
186  uint64_t max_cll;
187  uint64_t max_fall;
190 
191 typedef struct MatroskaTrackVideoProjection {
192  uint64_t type;
193  EbmlBin private;
194  double yaw;
195  double pitch;
196  double roll;
198 
199 typedef struct MatroskaTrackVideo {
200  double frame_rate;
201  uint64_t display_width;
202  uint64_t display_height;
203  uint64_t pixel_width;
204  uint64_t pixel_height;
206  uint64_t display_unit;
207  uint64_t interlaced;
208  uint64_t field_order;
209  uint64_t stereo_mode;
210  uint64_t alpha_mode;
214 
215 typedef struct MatroskaTrackAudio {
216  double samplerate;
218  uint64_t bitdepth;
219  uint64_t channels;
220 
221  /* real audio header (extracted from extradata) */
224  int frame_size;
225  int sub_packet_size;
227  int pkt_cnt;
228  uint64_t buf_timecode;
229  uint8_t *buf;
231 
232 typedef struct MatroskaTrackPlane {
233  uint64_t uid;
234  uint64_t type;
236 
237 typedef struct MatroskaTrackOperation {
240 
241 typedef struct MatroskaTrack {
242  uint64_t num;
243  uint64_t uid;
244  uint64_t type;
245  char *name;
246  char *codec_id;
248  char *language;
249  double time_scale;
251  uint64_t flag_default;
252  uint64_t flag_forced;
253  uint64_t flag_comment;
258  uint64_t seek_preroll;
263  uint64_t codec_delay;
265 
267  int64_t end_timecode;
268  int ms_compat;
269  int needs_decoding;
271 
275 
276 typedef struct MatroskaAttachment {
277  uint64_t uid;
278  char *filename;
279  char *description;
280  char *mime;
282 
285 
286 typedef struct MatroskaChapter {
287  uint64_t start;
288  uint64_t end;
289  uint64_t uid;
290  char *title;
291 
294 
295 typedef struct MatroskaIndexPos {
296  uint64_t track;
297  uint64_t pos;
299 
300 typedef struct MatroskaIndex {
301  uint64_t time;
304 
305 typedef struct MatroskaTag {
306  char *name;
307  char *string;
308  char *lang;
309  uint64_t def;
312 
313 typedef struct MatroskaTagTarget {
314  char *type;
315  uint64_t typevalue;
316  uint64_t trackuid;
317  uint64_t chapteruid;
318  uint64_t attachuid;
320 
321 typedef struct MatroskaTags {
323  EbmlList tag;
324 } MatroskaTags;
325 
326 typedef struct MatroskaSeekhead {
327  uint64_t id;
328  uint64_t pos;
330 
331 typedef struct MatroskaLevel {
332  uint64_t start;
333  uint64_t length;
335 
336 typedef struct MatroskaBlock {
337  uint64_t duration;
339  uint64_t non_simple;
341  uint64_t additional_id;
344 } MatroskaBlock;
345 
346 typedef struct MatroskaCluster {
348  uint64_t timecode;
349  int64_t pos;
351 
352 typedef struct MatroskaLevel1Element {
353  int64_t pos;
354  uint32_t id;
355  int parsed;
357 
358 typedef struct MatroskaDemuxContext {
359  const AVClass *class;
361 
362  /* EBML stuff */
365  uint32_t current_id;
366  int64_t resync_pos;
368 
369  uint64_t time_scale;
370  double duration;
371  char *title;
372  char *muxingapp;
377  EbmlList index;
380 
381  /* byte position of the segment inside the stream */
382  int64_t segment_start;
383 
385 
386  /* the packet queue */
389 
390  int done;
391 
392  /* What to skip before effectively reading a packet. */
393  int skip_to_keyframe;
395 
396  /* File has a CUES element, but we defer parsing until it is needed. */
398 
399  /* Level1 elements and whether they were read yet */
401  int num_level1_elems;
402 
404 
405  /* WebM DASH Manifest live flag */
406  int is_live;
407 
408  /* Bandwidth value for WebM DASH Manifest */
409  int bandwidth;
411 
412 #define CHILD_OF(parent) { .def = { .n = parent } }
413 
414 // The following forward declarations need their size because
415 // a tentative definition with internal linkage must not be an
416 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
417 // Removing the sizes breaks MSVC.
424 
425 static EbmlSyntax ebml_header[] = {
426  { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
427  { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, 0, offsetof(Ebml, max_size), { .u = 8 } },
428  { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, 0, offsetof(Ebml, id_length), { .u = 4 } },
429  { EBML_ID_DOCTYPE, EBML_STR, 0, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
430  { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
434 };
435 
437  { EBML_ID_HEADER, EBML_NEST, 0, 0, 0, { .n = ebml_header } },
439  { 0 }
440 };
441 
442 static EbmlSyntax matroska_info[] = {
443  { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
445  { MATROSKA_ID_TITLE, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, title) },
447  { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, muxingapp) },
448  { MATROSKA_ID_DATEUTC, EBML_BIN, 0, 0, offsetof(MatroskaDemuxContext, date_utc) },
451 };
452 
460  { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_x) },
461  { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_y) },
462  { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 1, 0, offsetof(MatroskaMasteringMeta, min_luminance) },
463  { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, max_luminance) },
465 };
466 
468  { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
469  { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u = 0 } },
470  { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz) },
471  { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert) },
472  { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz) },
473  { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert) },
480  { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_fall) },
483 };
484 
488  { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f = 0.0 } },
489  { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f = 0.0 } },
490  { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f = 0.0 } },
492 };
493 
494 static EbmlSyntax matroska_track_video[] = {
495  { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideo, frame_rate) },
496  { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
497  { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
498  { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_width) },
499  { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_height) },
500  { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideo, color_space) },
501  { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, alpha_mode), { .u = 0 } },
514 };
515 
516 static EbmlSyntax matroska_track_audio[] = {
517  { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
518  { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
519  { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, bitdepth) },
520  { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
522 };
523 
528 };
529 
531  { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
539 };
541  { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
542  { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
547 };
548 
552 };
553 
554 static EbmlSyntax matroska_track_plane[] = {
558 };
559 
561  { MATROSKA_ID_TRACKPLANE, EBML_NEST, 0, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
563 };
564 
568 };
569 
570 static EbmlSyntax matroska_track[] = {
571  { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, 0, offsetof(MatroskaTrack, num) },
572  { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTrack, name) },
573  { MATROSKA_ID_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, uid) },
574  { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrack, type) },
575  { MATROSKA_ID_CODECID, EBML_STR, 0, 0, offsetof(MatroskaTrack, codec_id) },
576  { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrack, codec_priv) },
577  { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, codec_delay), { .u = 0 } },
578  { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
579  { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, 0, offsetof(MatroskaTrack, default_duration) },
580  { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
581  { MATROSKA_ID_TRACKFLAGCOMMENTARY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_comment), { .u = 0 } },
582  { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
583  { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
584  { MATROSKA_ID_TRACKFLAGHEARINGIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_hearingimpaired), { .u = 0 } },
585  { MATROSKA_ID_TRACKFLAGVISUALIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_visualimpaired), { .u = 0 } },
586  { MATROSKA_ID_TRACKFLAGTEXTDESCRIPTIONS, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_textdescriptions), { .u = 0 } },
587  { MATROSKA_ID_TRACKFLAGORIGINAL, EBML_UINT, 1, 0, offsetof(MatroskaTrack, flag_original), {.u = 0 } },
589  { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
590  { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
592  { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, max_block_additional_id), { .u = 0 } },
593  { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, 0, offsetof(MatroskaTrack, seek_preroll), { .u = 0 } },
603 };
604 
605 static EbmlSyntax matroska_tracks[] = {
606  { MATROSKA_ID_TRACKENTRY, EBML_NEST, 0, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
608 };
609 
610 static EbmlSyntax matroska_attachment[] = {
611  { MATROSKA_ID_FILEUID, EBML_UINT, 0, 0, offsetof(MatroskaAttachment, uid) },
612  { MATROSKA_ID_FILENAME, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, filename) },
613  { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, 0, offsetof(MatroskaAttachment, mime) },
614  { MATROSKA_ID_FILEDATA, EBML_BIN, 0, 0, offsetof(MatroskaAttachment, bin) },
617 };
618 
619 static EbmlSyntax matroska_attachments[] = {
620  { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, 0, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
622 };
623 
625  { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaChapter, title) },
629 };
630 
632  { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
633  { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
634  { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaChapter, uid) },
641 };
642 
643 static EbmlSyntax matroska_chapter[] = {
644  { MATROSKA_ID_CHAPTERATOM, EBML_NEST, 0, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
650 };
651 
652 static EbmlSyntax matroska_chapters[] = {
653  { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter } },
655 };
656 
657 static EbmlSyntax matroska_index_pos[] = {
658  { MATROSKA_ID_CUETRACK, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, track) },
664 };
665 
666 static EbmlSyntax matroska_index_entry[] = {
667  { MATROSKA_ID_CUETIME, EBML_UINT, 0, 0, offsetof(MatroskaIndex, time) },
670 };
671 
672 static EbmlSyntax matroska_index[] = {
675 };
676 
677 static EbmlSyntax matroska_simpletag[] = {
678  { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTag, name) },
679  { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaTag, string) },
680  { MATROSKA_ID_TAGLANG, EBML_STR, 0, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
681  { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
682  { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
683  { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
685 };
686 
687 static EbmlSyntax matroska_tagtargets[] = {
689  { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
690  { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, trackuid), { .u = 0 } },
691  { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, chapteruid), { .u = 0 } },
692  { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, attachuid), { .u = 0 } },
694 };
695 
696 static EbmlSyntax matroska_tag[] = {
697  { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
698  { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
700 };
701 
702 static EbmlSyntax matroska_tags[] = {
703  { MATROSKA_ID_TAG, EBML_NEST, 0, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
705 };
706 
708  { MATROSKA_ID_SEEKID, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, id) },
709  { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
711 };
712 
713 static EbmlSyntax matroska_seekhead[] = {
714  { MATROSKA_ID_SEEKENTRY, EBML_NEST, 0, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
716 };
717 
718 static EbmlSyntax matroska_segment[] = {
720  { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, 0, { .n = matroska_info } },
721  { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tracks } },
723  { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_chapters } },
724  { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, 0, { .n = matroska_index } },
725  { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tags } },
726  { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, 0, { .n = matroska_seekhead } },
727  { 0 } /* We don't want to go back to level 0, so don't add the parent. */
728 };
729 
730 static EbmlSyntax matroska_segments[] = {
731  { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, 0, { .n = matroska_segment } },
732  { 0 }
733 };
734 
735 static EbmlSyntax matroska_blockmore[] = {
736  { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
737  { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, 0, offsetof(MatroskaBlock,additional) },
739 };
740 
742  { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, 0, {.n = matroska_blockmore} },
744 };
745 
746 static EbmlSyntax matroska_blockgroup[] = {
747  { MATROSKA_ID_BLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
750  { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, 0, offsetof(MatroskaBlock, discard_padding) },
751  { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 1, 0, offsetof(MatroskaBlock, reference) },
753  { 1, EBML_UINT, 0, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
755 };
756 
757 // The following array contains SimpleBlock and BlockGroup twice
758 // in order to reuse the other values for matroska_cluster_enter.
760  { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
761  { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, 0, { .n = matroska_blockgroup } },
762  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, 0, offsetof(MatroskaCluster, timecode) },
768 };
769 
771  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, 0, { .n = &matroska_cluster_parsing[2] } },
772  { 0 }
773 };
774 #undef CHILD_OF
775 
776 static const CodecMime mkv_image_mime_tags[] = {
777  {"image/gif" , AV_CODEC_ID_GIF},
778  {"image/jpeg" , AV_CODEC_ID_MJPEG},
779  {"image/png" , AV_CODEC_ID_PNG},
780  {"image/tiff" , AV_CODEC_ID_TIFF},
781 
782  {"" , AV_CODEC_ID_NONE}
783 };
784 
785 static const CodecMime mkv_mime_tags[] = {
786  {"text/plain" , AV_CODEC_ID_TEXT},
787  {"application/x-truetype-font", AV_CODEC_ID_TTF},
788  {"application/x-font" , AV_CODEC_ID_TTF},
789  {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
790  {"binary" , AV_CODEC_ID_BIN_DATA},
791 
792  {"" , AV_CODEC_ID_NONE}
793 };
794 
795 static const char *const matroska_doctypes[] = { "matroska", "webm" };
796 
798 
799 /*
800  * This function prepares the status for parsing of level 1 elements.
801  */
802 static int matroska_reset_status(MatroskaDemuxContext *matroska,
803  uint32_t id, int64_t position)
804 {
805  int64_t err = 0;
806  if (position >= 0) {
807  err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
808  if (err > 0)
809  err = 0;
810  } else
811  position = avio_tell(matroska->ctx->pb);
812 
813  matroska->current_id = id;
814  matroska->num_levels = 1;
815  matroska->unknown_count = 0;
816  matroska->resync_pos = position;
817  if (id)
818  matroska->resync_pos -= (av_log2(id) + 7) / 8;
819 
820  return err;
821 }
822 
823 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
824 {
825  AVIOContext *pb = matroska->ctx->pb;
826  uint32_t id;
827 
828  /* Try to seek to the last position to resync from. If this doesn't work,
829  * we resync from the earliest position available: The start of the buffer. */
830  if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
831  av_log(matroska->ctx, AV_LOG_WARNING,
832  "Seek to desired resync point failed. Seeking to "
833  "earliest point available instead.\n");
834  avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
835  last_pos + 1), SEEK_SET);
836  }
837 
838  id = avio_rb32(pb);
839 
840  // try to find a toplevel element
841  while (!avio_feof(pb)) {
842  if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
843  id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
845  id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
846  /* Prepare the context for parsing of a level 1 element. */
847  matroska_reset_status(matroska, id, -1);
848  /* Given that we are here means that an error has occurred,
849  * so treat the segment as unknown length in order not to
850  * discard valid data that happens to be beyond the designated
851  * end of the segment. */
852  matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
853  return 0;
854  }
855  id = (id << 8) | avio_r8(pb);
856  }
857 
858  matroska->done = 1;
859  return pb->error ? pb->error : AVERROR_EOF;
860 }
861 
862 /*
863  * Read: an "EBML number", which is defined as a variable-length
864  * array of bytes. The first byte indicates the length by giving a
865  * number of 0-bits followed by a one. The position of the first
866  * "one" bit inside the first byte indicates the length of this
867  * number.
868  * Returns: number of bytes read, < 0 on error
869  */
870 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
871  int max_size, uint64_t *number, int eof_forbidden)
872 {
873  int read, n = 1;
874  uint64_t total;
875  int64_t pos;
876 
877  /* The first byte tells us the length in bytes - except when it is zero. */
878  total = avio_r8(pb);
879  if (pb->eof_reached)
880  goto err;
881 
882  /* get the length of the EBML number */
883  read = 8 - ff_log2_tab[total];
884 
885  if (!total || read > max_size) {
886  pos = avio_tell(pb) - 1;
887  if (!total) {
888  av_log(matroska->ctx, AV_LOG_ERROR,
889  "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
890  "of an EBML number\n", pos, pos);
891  } else {
892  av_log(matroska->ctx, AV_LOG_ERROR,
893  "Length %d indicated by an EBML number's first byte 0x%02x "
894  "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
895  read, (uint8_t) total, pos, pos, max_size);
896  }
897  return AVERROR_INVALIDDATA;
898  }
899 
900  /* read out length */
901  total ^= 1 << ff_log2_tab[total];
902  while (n++ < read)
903  total = (total << 8) | avio_r8(pb);
904 
905  if (pb->eof_reached) {
906  eof_forbidden = 1;
907  goto err;
908  }
909 
910  *number = total;
911 
912  return read;
913 
914 err:
915  pos = avio_tell(pb);
916  if (pb->error) {
917  av_log(matroska->ctx, AV_LOG_ERROR,
918  "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
919  pos, pos);
920  return pb->error;
921  }
922  if (eof_forbidden) {
923  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
924  "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
925  return AVERROR(EIO);
926  }
927  return AVERROR_EOF;
928 }
929 
930 /**
931  * Read a EBML length value.
932  * This needs special handling for the "unknown length" case which has multiple
933  * encodings.
934  */
935 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
936  uint64_t *number)
937 {
938  int res = ebml_read_num(matroska, pb, 8, number, 1);
939  if (res > 0 && *number + 1 == 1ULL << (7 * res))
940  *number = EBML_UNKNOWN_LENGTH;
941  return res;
942 }
943 
944 /*
945  * Read the next element as an unsigned int.
946  * Returns NEEDS_CHECKING unless size == 0.
947  */
948 static int ebml_read_uint(AVIOContext *pb, int size,
949  uint64_t default_value, uint64_t *num)
950 {
951  int n = 0;
952 
953  if (size == 0) {
954  *num = default_value;
955  return 0;
956  }
957  /* big-endian ordering; build up number */
958  *num = 0;
959  while (n++ < size)
960  *num = (*num << 8) | avio_r8(pb);
961 
962  return NEEDS_CHECKING;
963 }
964 
965 /*
966  * Read the next element as a signed int.
967  * Returns NEEDS_CHECKING unless size == 0.
968  */
969 static int ebml_read_sint(AVIOContext *pb, int size,
970  int64_t default_value, int64_t *num)
971 {
972  int n = 1;
973 
974  if (size == 0) {
975  *num = default_value;
976  return 0;
977  } else {
978  *num = sign_extend(avio_r8(pb), 8);
979 
980  /* big-endian ordering; build up number */
981  while (n++ < size)
982  *num = ((uint64_t)*num << 8) | avio_r8(pb);
983  }
984 
985  return NEEDS_CHECKING;
986 }
987 
988 /*
989  * Read the next element as a float.
990  * Returns 0 if size == 0, NEEDS_CHECKING or < 0 on obvious failure.
991  */
992 static int ebml_read_float(AVIOContext *pb, int size,
993  double default_value, double *num)
994 {
995  if (size == 0) {
996  *num = default_value;
997  return 0;
998  } else if (size == 4) {
999  *num = av_int2float(avio_rb32(pb));
1000  } else if (size == 8) {
1001  *num = av_int2double(avio_rb64(pb));
1002  } else
1003  return AVERROR_INVALIDDATA;
1004 
1005  return NEEDS_CHECKING;
1007 
1008 /*
1009  * Read the next element as an ASCII string.
1010  * 0 is success, < 0 or NEEDS_CHECKING is failure.
1011  */
1012 static int ebml_read_ascii(AVIOContext *pb, int size,
1013  const char *default_value, char **str)
1014 {
1015  char *res;
1016  int ret;
1017 
1018  if (size == 0 && default_value) {
1019  res = av_strdup(default_value);
1020  if (!res)
1021  return AVERROR(ENOMEM);
1022  } else {
1023  /* EBML strings are usually not 0-terminated, so we allocate one
1024  * byte more, read the string and NUL-terminate it ourselves. */
1025  if (!(res = av_malloc(size + 1)))
1026  return AVERROR(ENOMEM);
1027  if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
1028  av_free(res);
1029  return ret < 0 ? ret : NEEDS_CHECKING;
1030  }
1031  (res)[size] = '\0';
1032  }
1033  av_free(*str);
1034  *str = res;
1035 
1036  return 0;
1038 
1039 /*
1040  * Read the next element as binary data.
1041  * 0 is success, < 0 or NEEDS_CHECKING is failure.
1042  */
1043 static int ebml_read_binary(AVIOContext *pb, int length,
1044  int64_t pos, EbmlBin *bin)
1045 {
1046  int ret;
1047 
1049  if (ret < 0)
1050  return ret;
1051  memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1052 
1053  bin->data = bin->buf->data;
1054  bin->size = length;
1055  bin->pos = pos;
1056  if ((ret = avio_read(pb, bin->data, length)) != length) {
1057  av_buffer_unref(&bin->buf);
1058  bin->data = NULL;
1059  bin->size = 0;
1060  return ret < 0 ? ret : NEEDS_CHECKING;
1061  }
1062 
1063  return 0;
1064 }
1066 /*
1067  * Read the next element, but only the header. The contents
1068  * are supposed to be sub-elements which can be read separately.
1069  * 0 is success, < 0 is failure.
1070  */
1071 static int ebml_read_master(MatroskaDemuxContext *matroska,
1072  uint64_t length, int64_t pos)
1073 {
1075 
1076  if (matroska->num_levels >= EBML_MAX_DEPTH) {
1077  av_log(matroska->ctx, AV_LOG_ERROR,
1078  "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1079  return AVERROR(ENOSYS);
1080  }
1081 
1082  level = &matroska->levels[matroska->num_levels++];
1083  level->start = pos;
1084  level->length = length;
1085 
1086  return 0;
1088 
1089 /*
1090  * Read a signed "EBML number"
1091  * Return: number of bytes processed, < 0 on error
1092  */
1093 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1094  AVIOContext *pb, int64_t *num)
1095 {
1096  uint64_t unum;
1097  int res;
1098 
1099  /* read as unsigned number first */
1100  if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1101  return res;
1102 
1103  /* make signed (weird way) */
1104  *num = unum - ((1LL << (7 * res - 1)) - 1);
1105 
1106  return res;
1107 }
1108 
1109 static int ebml_parse(MatroskaDemuxContext *matroska,
1110  EbmlSyntax *syntax, void *data);
1111 
1112 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1113 {
1114  int i;
1115 
1116  // Whoever touches this should be aware of the duplication
1117  // existing in matroska_cluster_parsing.
1118  for (i = 0; syntax[i].id; i++)
1119  if (id == syntax[i].id)
1120  break;
1121 
1122  return &syntax[i];
1123 }
1124 
1125 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1126  void *data)
1127 {
1128  int res;
1129 
1130  if (data) {
1131  for (int i = 0; syntax[i].id; i++) {
1132  void *dst = (char *)data + syntax[i].data_offset;
1133  switch (syntax[i].type) {
1134  case EBML_UINT:
1135  *(uint64_t *)dst = syntax[i].def.u;
1136  break;
1137  case EBML_SINT:
1138  *(int64_t *) dst = syntax[i].def.i;
1139  break;
1140  case EBML_FLOAT:
1141  *(double *) dst = syntax[i].def.f;
1142  break;
1143  case EBML_STR:
1144  case EBML_UTF8:
1145  // the default may be NULL
1146  if (syntax[i].def.s) {
1147  *(char**)dst = av_strdup(syntax[i].def.s);
1148  if (!*(char**)dst)
1149  return AVERROR(ENOMEM);
1150  }
1151  break;
1152  }
1153  }
1154 
1155  if (!matroska->levels[matroska->num_levels - 1].length) {
1156  matroska->num_levels--;
1157  return 0;
1158  }
1159  }
1160 
1161  do {
1162  res = ebml_parse(matroska, syntax, data);
1163  } while (!res);
1164 
1165  return res == LEVEL_ENDED ? 0 : res;
1166 }
1167 
1168 static int is_ebml_id_valid(uint32_t id)
1169 {
1170  // Due to endian nonsense in Matroska, the highest byte with any bits set
1171  // will contain the leading length bit. This bit in turn identifies the
1172  // total byte length of the element by its position within the byte.
1173  unsigned int bits = av_log2(id);
1174  return id && (bits + 7) / 8 == (8 - bits % 8);
1176 
1177 /*
1178  * Allocate and return the entry for the level1 element with the given ID. If
1179  * an entry already exists, return the existing entry.
1180  */
1182  uint32_t id, int64_t pos)
1183 {
1184  int i;
1185  MatroskaLevel1Element *elem;
1186 
1187  if (!is_ebml_id_valid(id))
1188  return NULL;
1189 
1190  // Some files link to all clusters; useless.
1191  if (id == MATROSKA_ID_CLUSTER)
1192  return NULL;
1193 
1194  // There can be multiple SeekHeads and Tags.
1195  for (i = 0; i < matroska->num_level1_elems; i++) {
1196  if (matroska->level1_elems[i].id == id) {
1197  if (matroska->level1_elems[i].pos == pos ||
1198  id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1199  return &matroska->level1_elems[i];
1200  }
1201  }
1202 
1203  // Only a completely broken file would have more elements.
1204  if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1205  av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1206  return NULL;
1207  }
1208 
1209  elem = &matroska->level1_elems[matroska->num_level1_elems++];
1210  *elem = (MatroskaLevel1Element){.id = id};
1211 
1212  return elem;
1213 }
1214 
1215 static int ebml_parse(MatroskaDemuxContext *matroska,
1216  EbmlSyntax *syntax, void *data)
1217 {
1218  static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1219  // Forbid unknown-length EBML_NONE elements.
1221  [EBML_UINT] = 8,
1222  [EBML_SINT] = 8,
1223  [EBML_FLOAT] = 8,
1224  // max. 16 MB for strings
1225  [EBML_STR] = 0x1000000,
1226  [EBML_UTF8] = 0x1000000,
1227  // max. 256 MB for binary data
1228  [EBML_BIN] = 0x10000000,
1229  // no limits for anything else
1230  };
1231  AVIOContext *pb = matroska->ctx->pb;
1232  uint32_t id;
1233  uint64_t length;
1234  int64_t pos = avio_tell(pb), pos_alt;
1235  int res, update_pos = 1, level_check;
1236  MatroskaLevel1Element *level1_elem;
1237  MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1238 
1239  if (!matroska->current_id) {
1240  uint64_t id;
1241  res = ebml_read_num(matroska, pb, 4, &id, 0);
1242  if (res < 0) {
1243  if (pb->eof_reached && res == AVERROR_EOF) {
1244  if (matroska->is_live)
1245  // in live mode, finish parsing if EOF is reached.
1246  return 1;
1247  if (level && pos == avio_tell(pb)) {
1248  if (level->length == EBML_UNKNOWN_LENGTH) {
1249  // Unknown-length levels automatically end at EOF.
1250  matroska->num_levels--;
1251  return LEVEL_ENDED;
1252  } else {
1253  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1254  "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1255  }
1256  }
1257  }
1258  return res;
1259  }
1260  matroska->current_id = id | 1 << 7 * res;
1261  pos_alt = pos + res;
1262  } else {
1263  pos_alt = pos;
1264  pos -= (av_log2(matroska->current_id) + 7) / 8;
1265  }
1266 
1267  id = matroska->current_id;
1268 
1269  syntax = ebml_parse_id(syntax, id);
1270  if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1271  if (level && level->length == EBML_UNKNOWN_LENGTH) {
1272  // Unknown-length levels end when an element from an upper level
1273  // in the hierarchy is encountered.
1274  while (syntax->def.n) {
1275  syntax = ebml_parse_id(syntax->def.n, id);
1276  if (syntax->id) {
1277  matroska->num_levels--;
1278  return LEVEL_ENDED;
1279  }
1280  };
1281  }
1282 
1283  av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1284  "%"PRId64"\n", id, pos);
1285  update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1286  }
1287 
1288  if (data) {
1289  data = (char *) data + syntax->data_offset;
1290  if (syntax->list_elem_size) {
1291  EbmlList *list = data;
1292  void *newelem;
1293 
1294  if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1295  return AVERROR(ENOMEM);
1296  newelem = av_fast_realloc(list->elem,
1297  &list->alloc_elem_size,
1298  (list->nb_elem + 1) * syntax->list_elem_size);
1299  if (!newelem)
1300  return AVERROR(ENOMEM);
1301  list->elem = newelem;
1302  data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1303  memset(data, 0, syntax->list_elem_size);
1304  list->nb_elem++;
1305  }
1306  }
1307 
1308  if (syntax->type != EBML_STOP) {
1309  matroska->current_id = 0;
1310  if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1311  return res;
1312 
1313  pos_alt += res;
1314 
1315  if (matroska->num_levels > 0) {
1316  if (length != EBML_UNKNOWN_LENGTH &&
1317  level->length != EBML_UNKNOWN_LENGTH) {
1318  uint64_t elem_end = pos_alt + length,
1319  level_end = level->start + level->length;
1320 
1321  if (elem_end < level_end) {
1322  level_check = 0;
1323  } else if (elem_end == level_end) {
1324  level_check = LEVEL_ENDED;
1325  } else {
1326  av_log(matroska->ctx, AV_LOG_ERROR,
1327  "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1328  "containing master element ending at 0x%"PRIx64"\n",
1329  pos, elem_end, level_end);
1330  return AVERROR_INVALIDDATA;
1331  }
1332  } else if (length != EBML_UNKNOWN_LENGTH) {
1333  level_check = 0;
1334  } else if (level->length != EBML_UNKNOWN_LENGTH) {
1335  av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1336  "at 0x%"PRIx64" inside parent with finite size\n", pos);
1337  return AVERROR_INVALIDDATA;
1338  } else {
1339  level_check = 0;
1340  if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1341  || syntax->type == EBML_NEST)) {
1342  // According to the current specifications only clusters and
1343  // segments are allowed to be unknown-length. We also accept
1344  // other unknown-length master elements.
1345  av_log(matroska->ctx, AV_LOG_WARNING,
1346  "Found unknown-length element 0x%"PRIX32" other than "
1347  "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1348  "parsing will nevertheless be attempted.\n", id, pos);
1349  update_pos = -1;
1350  }
1351  }
1352  } else
1353  level_check = 0;
1354 
1355  if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1356  if (length != EBML_UNKNOWN_LENGTH) {
1357  av_log(matroska->ctx, AV_LOG_ERROR,
1358  "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1359  "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1360  length, max_lengths[syntax->type], id, pos);
1361  } else if (syntax->type != EBML_NONE) {
1362  av_log(matroska->ctx, AV_LOG_ERROR,
1363  "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1364  "unknown length, yet the length of an element of its "
1365  "type must be known.\n", id, pos);
1366  } else {
1367  av_log(matroska->ctx, AV_LOG_ERROR,
1368  "Found unknown-length element with ID 0x%"PRIX32" at "
1369  "pos. 0x%"PRIx64" for which no syntax for parsing is "
1370  "available.\n", id, pos);
1371  }
1372  return AVERROR_INVALIDDATA;
1373  }
1374 
1375  if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1376  // Loosing sync will likely manifest itself as encountering unknown
1377  // elements which are not reliably distinguishable from elements
1378  // belonging to future extensions of the format.
1379  // We use a heuristic to detect such situations: If the current
1380  // element is not expected at the current syntax level and there
1381  // were only a few unknown elements in a row, then the element is
1382  // skipped or considered defective based upon the length of the
1383  // current element (i.e. how much would be skipped); if there were
1384  // more than a few skipped elements in a row and skipping the current
1385  // element would lead us more than SKIP_THRESHOLD away from the last
1386  // known good position, then it is inferred that an error occurred.
1387  // The dependency on the number of unknown elements in a row exists
1388  // because the distance to the last known good position is
1389  // automatically big if the last parsed element was big.
1390  // In both cases, each unknown element is considered equivalent to
1391  // UNKNOWN_EQUIV of skipped bytes for the check.
1392  // The whole check is only done for non-seekable output, because
1393  // in this situation skipped data can't simply be rechecked later.
1394  // This is especially important when using unkown length elements
1395  // as the check for whether a child exceeds its containing master
1396  // element is not effective in this situation.
1397  if (update_pos) {
1398  matroska->unknown_count = 0;
1399  } else {
1400  int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1401 
1402  if (matroska->unknown_count > 3)
1403  dist += pos_alt - matroska->resync_pos;
1404 
1405  if (dist > SKIP_THRESHOLD) {
1406  av_log(matroska->ctx, AV_LOG_ERROR,
1407  "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1408  "length 0x%"PRIx64" considered as invalid data. Last "
1409  "known good position 0x%"PRIx64", %d unknown elements"
1410  " in a row\n", id, pos, length, matroska->resync_pos,
1411  matroska->unknown_count);
1412  return AVERROR_INVALIDDATA;
1413  }
1414  }
1415  }
1416 
1417  if (update_pos > 0) {
1418  // We have found an element that is allowed at this place
1419  // in the hierarchy and it passed all checks, so treat the beginning
1420  // of the element as the "last known good" position.
1421  matroska->resync_pos = pos;
1422  }
1423 
1424  if (!data && length != EBML_UNKNOWN_LENGTH)
1425  goto skip;
1426  }
1427 
1428  switch (syntax->type) {
1429  case EBML_UINT:
1430  res = ebml_read_uint(pb, length, syntax->def.u, data);
1431  break;
1432  case EBML_SINT:
1433  res = ebml_read_sint(pb, length, syntax->def.i, data);
1434  break;
1435  case EBML_FLOAT:
1436  res = ebml_read_float(pb, length, syntax->def.f, data);
1437  break;
1438  case EBML_STR:
1439  case EBML_UTF8:
1440  res = ebml_read_ascii(pb, length, syntax->def.s, data);
1441  break;
1442  case EBML_BIN:
1443  res = ebml_read_binary(pb, length, pos_alt, data);
1444  break;
1445  case EBML_LEVEL1:
1446  case EBML_NEST:
1447  if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1448  return res;
1449  if (id == MATROSKA_ID_SEGMENT)
1450  matroska->segment_start = pos_alt;
1451  if (id == MATROSKA_ID_CUES)
1452  matroska->cues_parsing_deferred = 0;
1453  if (syntax->type == EBML_LEVEL1 &&
1454  (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1455  if (!level1_elem->pos) {
1456  // Zero is not a valid position for a level 1 element.
1457  level1_elem->pos = pos;
1458  } else if (level1_elem->pos != pos)
1459  av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1460  level1_elem->parsed = 1;
1461  }
1462  if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1463  return res;
1464  break;
1465  case EBML_STOP:
1466  return 1;
1467  skip:
1468  default:
1469  if (length) {
1470  int64_t res2;
1471  if (ffio_limit(pb, length) != length) {
1472  // ffio_limit emits its own error message,
1473  // so we don't have to.
1474  return AVERROR(EIO);
1475  }
1476  if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1477  // avio_skip might take us past EOF. We check for this
1478  // by skipping only length - 1 bytes, reading a byte and
1479  // checking the error flags. This is done in order to check
1480  // that the element has been properly skipped even when
1481  // no filesize (that ffio_limit relies on) is available.
1482  avio_r8(pb);
1483  res = NEEDS_CHECKING;
1484  } else
1485  res = res2;
1486  } else
1487  res = 0;
1488  }
1489  if (res) {
1490  if (res == NEEDS_CHECKING) {
1491  if (pb->eof_reached) {
1492  if (pb->error)
1493  res = pb->error;
1494  else
1495  res = AVERROR_EOF;
1496  } else
1497  goto level_check;
1498  }
1499 
1500  if (res == AVERROR_INVALIDDATA)
1501  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1502  else if (res == AVERROR(EIO))
1503  av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1504  else if (res == AVERROR_EOF) {
1505  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1506  res = AVERROR(EIO);
1507  }
1508 
1509  return res;
1510  }
1511 
1512 level_check:
1513  if (syntax->is_counted && data) {
1514  CountedElement *elem = data;
1515  if (elem->count != UINT_MAX)
1516  elem->count++;
1517  }
1518 
1519  if (level_check == LEVEL_ENDED && matroska->num_levels) {
1520  level = &matroska->levels[matroska->num_levels - 1];
1521  pos = avio_tell(pb);
1522 
1523  // Given that pos >= level->start no check for
1524  // level->length != EBML_UNKNOWN_LENGTH is necessary.
1525  while (matroska->num_levels && pos == level->start + level->length) {
1526  matroska->num_levels--;
1527  level--;
1528  }
1529  }
1530 
1531  return level_check;
1532 }
1533 
1534 static void ebml_free(EbmlSyntax *syntax, void *data)
1535 {
1536  int i, j;
1537  for (i = 0; syntax[i].id; i++) {
1538  void *data_off = (char *) data + syntax[i].data_offset;
1539  switch (syntax[i].type) {
1540  case EBML_STR:
1541  case EBML_UTF8:
1542  av_freep(data_off);
1543  break;
1544  case EBML_BIN:
1545  av_buffer_unref(&((EbmlBin *) data_off)->buf);
1546  break;
1547  case EBML_LEVEL1:
1548  case EBML_NEST:
1549  if (syntax[i].list_elem_size) {
1550  EbmlList *list = data_off;
1551  char *ptr = list->elem;
1552  for (j = 0; j < list->nb_elem;
1553  j++, ptr += syntax[i].list_elem_size)
1554  ebml_free(syntax[i].def.n, ptr);
1555  av_freep(&list->elem);
1556  list->nb_elem = 0;
1557  list->alloc_elem_size = 0;
1558  } else
1559  ebml_free(syntax[i].def.n, data_off);
1560  default:
1561  break;
1562  }
1563  }
1564 }
1565 
1566 /*
1567  * Autodetecting...
1568  */
1569 static int matroska_probe(const AVProbeData *p)
1570 {
1571  uint64_t total = 0;
1572  int len_mask = 0x80, size = 1, n = 1, i;
1573 
1574  /* EBML header? */
1575  if (AV_RB32(p->buf) != EBML_ID_HEADER)
1576  return 0;
1577 
1578  /* length of header */
1579  total = p->buf[4];
1580  while (size <= 8 && !(total & len_mask)) {
1581  size++;
1582  len_mask >>= 1;
1583  }
1584  if (size > 8)
1585  return 0;
1586  total &= (len_mask - 1);
1587  while (n < size)
1588  total = (total << 8) | p->buf[4 + n++];
1589 
1590  if (total + 1 == 1ULL << (7 * size)){
1591  /* Unknown-length header - simply parse the whole buffer. */
1592  total = p->buf_size - 4 - size;
1593  } else {
1594  /* Does the probe data contain the whole header? */
1595  if (p->buf_size < 4 + size + total)
1596  return 0;
1597  }
1598 
1599  /* The header should contain a known document type. For now,
1600  * we don't parse the whole header but simply check for the
1601  * availability of that array of characters inside the header.
1602  * Not fully fool-proof, but good enough. */
1603  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1604  size_t probelen = strlen(matroska_doctypes[i]);
1605  if (total < probelen)
1606  continue;
1607  for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1608  if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1609  return AVPROBE_SCORE_MAX;
1610  }
1611 
1612  // probably valid EBML header but no recognized doctype
1613  return AVPROBE_SCORE_EXTENSION;
1614 }
1615 
1617  uint64_t num)
1618 {
1619  MatroskaTrack *tracks = matroska->tracks.elem;
1620  int i;
1621 
1622  for (i = 0; i < matroska->tracks.nb_elem; i++)
1623  if (tracks[i].num == num)
1624  return &tracks[i];
1625 
1626  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1627  return NULL;
1628 }
1629 
1630 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1631  MatroskaTrack *track)
1632 {
1633  MatroskaTrackEncoding *encodings = track->encodings.elem;
1634  uint8_t *data = *buf;
1635  int isize = *buf_size;
1636  uint8_t *pkt_data = NULL;
1637  uint8_t av_unused *newpktdata;
1638  int pkt_size = isize;
1639  int result = 0;
1640  int olen;
1641 
1642  if (pkt_size >= 10000000U)
1643  return AVERROR_INVALIDDATA;
1644 
1645  switch (encodings[0].compression.algo) {
1647  {
1648  int header_size = encodings[0].compression.settings.size;
1649  uint8_t *header = encodings[0].compression.settings.data;
1650 
1651  if (header_size && !header) {
1652  av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1653  return -1;
1654  }
1655 
1656  if (!header_size)
1657  return 0;
1658 
1659  pkt_size = isize + header_size;
1660  pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1661  if (!pkt_data)
1662  return AVERROR(ENOMEM);
1663 
1664  memcpy(pkt_data, header, header_size);
1665  memcpy(pkt_data + header_size, data, isize);
1666  break;
1667  }
1668 #if CONFIG_LZO
1670  do {
1671  int insize = isize;
1672  olen = pkt_size *= 3;
1673  newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1675  if (!newpktdata) {
1676  result = AVERROR(ENOMEM);
1677  goto failed;
1678  }
1679  pkt_data = newpktdata;
1680  result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1681  } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1682  if (result) {
1684  goto failed;
1685  }
1686  pkt_size -= olen;
1687  break;
1688 #endif
1689 #if CONFIG_ZLIB
1691  {
1692  z_stream zstream = { 0 };
1693  if (!pkt_size || inflateInit(&zstream) != Z_OK)
1694  return -1;
1695  zstream.next_in = data;
1696  zstream.avail_in = isize;
1697  do {
1698  pkt_size *= 3;
1699  newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1700  if (!newpktdata) {
1701  inflateEnd(&zstream);
1702  result = AVERROR(ENOMEM);
1703  goto failed;
1704  }
1705  pkt_data = newpktdata;
1706  zstream.avail_out = pkt_size - zstream.total_out;
1707  zstream.next_out = pkt_data + zstream.total_out;
1708  result = inflate(&zstream, Z_NO_FLUSH);
1709  } while (result == Z_OK && pkt_size < 10000000);
1710  pkt_size = zstream.total_out;
1711  inflateEnd(&zstream);
1712  if (result != Z_STREAM_END) {
1713  if (result == Z_MEM_ERROR)
1714  result = AVERROR(ENOMEM);
1715  else
1717  goto failed;
1718  }
1719  break;
1720  }
1721 #endif
1722 #if CONFIG_BZLIB
1724  {
1725  bz_stream bzstream = { 0 };
1726  if (!pkt_size || BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1727  return -1;
1728  bzstream.next_in = data;
1729  bzstream.avail_in = isize;
1730  do {
1731  pkt_size *= 3;
1732  newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1733  if (!newpktdata) {
1734  BZ2_bzDecompressEnd(&bzstream);
1735  result = AVERROR(ENOMEM);
1736  goto failed;
1737  }
1738  pkt_data = newpktdata;
1739  bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1740  bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1741  result = BZ2_bzDecompress(&bzstream);
1742  } while (result == BZ_OK && pkt_size < 10000000);
1743  pkt_size = bzstream.total_out_lo32;
1744  BZ2_bzDecompressEnd(&bzstream);
1745  if (result != BZ_STREAM_END) {
1746  if (result == BZ_MEM_ERROR)
1747  result = AVERROR(ENOMEM);
1748  else
1750  goto failed;
1751  }
1752  break;
1753  }
1754 #endif
1755  default:
1756  return AVERROR_INVALIDDATA;
1757  }
1758 
1759  memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1760 
1761  *buf = pkt_data;
1762  *buf_size = pkt_size;
1763  return 0;
1765 failed:
1766  av_free(pkt_data);
1767  return result;
1768 }
1769 
1771  AVDictionary **metadata, char *prefix)
1772 {
1773  MatroskaTag *tags = list->elem;
1774  char key[1024];
1775  int i;
1776 
1777  for (i = 0; i < list->nb_elem; i++) {
1778  const char *lang = tags[i].lang &&
1779  strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1780 
1781  if (!tags[i].name) {
1782  av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1783  continue;
1784  }
1785  if (prefix)
1786  snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1787  else
1788  av_strlcpy(key, tags[i].name, sizeof(key));
1789  if (tags[i].def || !lang) {
1790  av_dict_set(metadata, key, tags[i].string, 0);
1791  if (tags[i].sub.nb_elem)
1792  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1793  }
1794  if (lang) {
1795  av_strlcat(key, "-", sizeof(key));
1796  av_strlcat(key, lang, sizeof(key));
1797  av_dict_set(metadata, key, tags[i].string, 0);
1798  if (tags[i].sub.nb_elem)
1799  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1800  }
1801  }
1803 }
1804 
1806 {
1807  MatroskaDemuxContext *matroska = s->priv_data;
1808  MatroskaTags *tags = matroska->tags.elem;
1809  int i, j;
1810 
1811  for (i = 0; i < matroska->tags.nb_elem; i++) {
1812  if (tags[i].target.attachuid) {
1813  MatroskaAttachment *attachment = matroska->attachments.elem;
1814  int found = 0;
1815  for (j = 0; j < matroska->attachments.nb_elem; j++) {
1816  if (attachment[j].uid == tags[i].target.attachuid &&
1817  attachment[j].stream) {
1818  matroska_convert_tag(s, &tags[i].tag,
1819  &attachment[j].stream->metadata, NULL);
1820  found = 1;
1821  }
1822  }
1823  if (!found) {
1825  "The tags at index %d refer to a "
1826  "non-existent attachment %"PRId64".\n",
1827  i, tags[i].target.attachuid);
1828  }
1829  } else if (tags[i].target.chapteruid) {
1830  MatroskaChapter *chapter = matroska->chapters.elem;
1831  int found = 0;
1832  for (j = 0; j < matroska->chapters.nb_elem; j++) {
1833  if (chapter[j].uid == tags[i].target.chapteruid &&
1834  chapter[j].chapter) {
1835  matroska_convert_tag(s, &tags[i].tag,
1836  &chapter[j].chapter->metadata, NULL);
1837  found = 1;
1838  }
1839  }
1840  if (!found) {
1842  "The tags at index %d refer to a non-existent chapter "
1843  "%"PRId64".\n",
1844  i, tags[i].target.chapteruid);
1845  }
1846  } else if (tags[i].target.trackuid) {
1847  MatroskaTrack *track = matroska->tracks.elem;
1848  int found = 0;
1849  for (j = 0; j < matroska->tracks.nb_elem; j++) {
1850  if (track[j].uid == tags[i].target.trackuid &&
1851  track[j].stream) {
1852  matroska_convert_tag(s, &tags[i].tag,
1853  &track[j].stream->metadata, NULL);
1854  found = 1;
1855  }
1856  }
1857  if (!found) {
1859  "The tags at index %d refer to a non-existent track "
1860  "%"PRId64".\n",
1861  i, tags[i].target.trackuid);
1862  }
1863  } else {
1864  matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1865  tags[i].target.type);
1866  }
1867  }
1868 }
1869 
1871  int64_t pos)
1872 {
1873  uint32_t saved_id = matroska->current_id;
1874  int64_t before_pos = avio_tell(matroska->ctx->pb);
1875  int ret = 0;
1876  int ret2;
1877 
1878  /* seek */
1879  if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1880  /* We don't want to lose our seekhead level, so we add
1881  * a dummy. This is a crude hack. */
1882  if (matroska->num_levels == EBML_MAX_DEPTH) {
1883  av_log(matroska->ctx, AV_LOG_INFO,
1884  "Max EBML element depth (%d) reached, "
1885  "cannot parse further.\n", EBML_MAX_DEPTH);
1887  } else {
1888  matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1889  matroska->num_levels++;
1890  matroska->current_id = 0;
1891 
1892  ret = ebml_parse(matroska, matroska_segment, matroska);
1893  if (ret == LEVEL_ENDED) {
1894  /* This can only happen if the seek brought us beyond EOF. */
1895  ret = AVERROR_EOF;
1896  }
1897  }
1898  }
1899  /* Seek back - notice that in all instances where this is used
1900  * it is safe to set the level to 1. */
1901  ret2 = matroska_reset_status(matroska, saved_id, before_pos);
1902  if (ret >= 0)
1903  ret = ret2;
1904 
1905  return ret;
1906 }
1907 
1908 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1909 {
1910  EbmlList *seekhead_list = &matroska->seekhead;
1911  int i;
1912 
1913  // we should not do any seeking in the streaming case
1914  if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1915  return;
1916 
1917  for (i = 0; i < seekhead_list->nb_elem; i++) {
1918  MatroskaSeekhead *seekheads = seekhead_list->elem;
1919  uint32_t id = seekheads[i].id;
1920  int64_t pos = seekheads[i].pos + matroska->segment_start;
1921  MatroskaLevel1Element *elem;
1922 
1923  if (id != seekheads[i].id || pos < matroska->segment_start)
1924  continue;
1925 
1926  elem = matroska_find_level1_elem(matroska, id, pos);
1927  if (!elem || elem->parsed)
1928  continue;
1929 
1930  elem->pos = pos;
1931 
1932  // defer cues parsing until we actually need cue data.
1933  if (id == MATROSKA_ID_CUES)
1934  continue;
1935 
1936  if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1937  // mark index as broken
1938  matroska->cues_parsing_deferred = -1;
1939  break;
1940  }
1941 
1942  elem->parsed = 1;
1943  }
1944 }
1945 
1946 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1947 {
1948  EbmlList *index_list;
1950  uint64_t index_scale = 1;
1951  int i, j;
1952 
1953  if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1954  return;
1955 
1956  index_list = &matroska->index;
1957  index = index_list->elem;
1958  if (index_list->nb_elem < 2)
1959  return;
1960  if (index[1].time > 1E14 / matroska->time_scale) {
1961  av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1962  return;
1963  }
1964  for (i = 0; i < index_list->nb_elem; i++) {
1965  EbmlList *pos_list = &index[i].pos;
1966  MatroskaIndexPos *pos = pos_list->elem;
1967  for (j = 0; j < pos_list->nb_elem; j++) {
1968  MatroskaTrack *track = matroska_find_track_by_num(matroska,
1969  pos[j].track);
1970  if (track && track->stream)
1971  av_add_index_entry(track->stream,
1972  pos[j].pos + matroska->segment_start,
1973  index[i].time / index_scale, 0, 0,
1975  }
1976  }
1977 }
1978 
1979 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1980  int i;
1981 
1982  if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1983  return;
1984 
1985  for (i = 0; i < matroska->num_level1_elems; i++) {
1986  MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1987  if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1988  if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1989  matroska->cues_parsing_deferred = -1;
1990  elem->parsed = 1;
1991  break;
1992  }
1993  }
1994 
1995  matroska_add_index_entries(matroska);
1996 }
1997 
1998 static int matroska_aac_profile(char *codec_id)
1999 {
2000  static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
2001  int profile;
2002 
2004  if (strstr(codec_id, aac_profiles[profile]))
2005  break;
2006  return profile + 1;
2007 }
2008 
2009 static int matroska_aac_sri(int samplerate)
2010 {
2011  int sri;
2012 
2013  for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
2014  if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
2015  break;
2016  return sri;
2017 }
2018 
2019 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
2020 {
2021  /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
2022  avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
2023 }
2024 
2026  MatroskaTrack *track,
2027  int *offset)
2028 {
2029  AVStream *st = track->stream;
2030  uint8_t *p = track->codec_priv.data;
2031  int size = track->codec_priv.size;
2032 
2033  if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
2034  av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
2035  track->codec_priv.size = 0;
2036  return 0;
2037  }
2038  *offset = 8;
2039  track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
2040 
2041  p += track->codec_priv.size;
2042  size -= track->codec_priv.size;
2043 
2044  /* parse the remaining metadata blocks if present */
2045  while (size >= 4) {
2046  int block_last, block_type, block_size;
2047 
2048  flac_parse_block_header(p, &block_last, &block_type, &block_size);
2049 
2050  p += 4;
2051  size -= 4;
2052  if (block_size > size)
2053  return 0;
2054 
2055  /* check for the channel mask */
2056  if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2057  AVDictionary *dict = NULL;
2058  AVDictionaryEntry *chmask;
2059 
2060  ff_vorbis_comment(s, &dict, p, block_size, 0);
2061  chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2062  if (chmask) {
2063  uint64_t mask = strtol(chmask->value, NULL, 0);
2064  if (!mask || mask & ~0x3ffffULL) {
2066  "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2067  } else
2068  st->codecpar->channel_layout = mask;
2069  }
2070  av_dict_free(&dict);
2071  }
2072 
2073  p += block_size;
2074  size -= block_size;
2075  }
2076 
2077  return 0;
2078 }
2079 
2080 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2081 {
2082  int minor, micro, bttb = 0;
2083 
2084  /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2085  * this function, and fixed in 57.52 */
2086  if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, &micro) == 2)
2087  bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2088 
2089  switch (field_order) {
2091  return AV_FIELD_PROGRESSIVE;
2093  return AV_FIELD_UNKNOWN;
2095  return AV_FIELD_TT;
2097  return AV_FIELD_BB;
2099  return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2101  return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2102  default:
2103  return AV_FIELD_UNKNOWN;
2104  }
2105 }
2106 
2107 static void mkv_stereo_mode_display_mul(int stereo_mode,
2108  int *h_width, int *h_height)
2109 {
2110  switch (stereo_mode) {
2116  break;
2121  *h_width = 2;
2122  break;
2127  *h_height = 2;
2128  break;
2129  }
2130 }
2131 
2132 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2133  const MatroskaTrackVideoColor *color = track->video.color.elem;
2134  const MatroskaMasteringMeta *mastering_meta;
2135  int has_mastering_primaries, has_mastering_luminance;
2136 
2137  if (!track->video.color.nb_elem)
2138  return 0;
2139 
2140  mastering_meta = &color->mastering_meta;
2141  // Mastering primaries are CIE 1931 coords, and must be > 0.
2142  has_mastering_primaries =
2143  mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2144  mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2145  mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2146  mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2147  has_mastering_luminance = mastering_meta->max_luminance >
2148  mastering_meta->min_luminance.el.f &&
2149  mastering_meta->min_luminance.el.f >= 0 &&
2150  mastering_meta->min_luminance.count;
2151 
2152  if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2153  st->codecpar->color_space = color->matrix_coefficients;
2154  if (color->primaries != AVCOL_PRI_RESERVED &&
2155  color->primaries != AVCOL_PRI_RESERVED0)
2156  st->codecpar->color_primaries = color->primaries;
2157  if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2158  color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2159  st->codecpar->color_trc = color->transfer_characteristics;
2160  if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2161  color->range <= AVCOL_RANGE_JPEG)
2162  st->codecpar->color_range = color->range;
2163  if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2164  color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2165  color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2166  color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2167  st->codecpar->chroma_location =
2168  avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2169  (color->chroma_siting_vert - 1) << 7);
2170  }
2171  if (color->max_cll && color->max_fall) {
2172  size_t size = 0;
2173  int ret;
2175  if (!metadata)
2176  return AVERROR(ENOMEM);
2178  (uint8_t *)metadata, size);
2179  if (ret < 0) {
2180  av_freep(&metadata);
2181  return ret;
2182  }
2183  metadata->MaxCLL = color->max_cll;
2184  metadata->MaxFALL = color->max_fall;
2185  }
2186 
2187  if (has_mastering_primaries || has_mastering_luminance) {
2188  AVMasteringDisplayMetadata *metadata =
2191  sizeof(AVMasteringDisplayMetadata));
2192  if (!metadata) {
2193  return AVERROR(ENOMEM);
2194  }
2195  memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2196  if (has_mastering_primaries) {
2197  metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2198  metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2199  metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2200  metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2201  metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2202  metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2203  metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2204  metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2205  metadata->has_primaries = 1;
2206  }
2207  if (has_mastering_luminance) {
2208  metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2209  metadata->min_luminance = av_d2q(mastering_meta->min_luminance.el.f, INT_MAX);
2210  metadata->has_luminance = 1;
2211  }
2212  }
2213  return 0;
2214 }
2215 
2216 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2217  void *logctx)
2218 {
2219  AVSphericalMapping *spherical;
2220  const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2221  const uint8_t *priv_data = mkv_projection->private.data;
2222  enum AVSphericalProjection projection;
2223  size_t spherical_size;
2224  uint32_t l = 0, t = 0, r = 0, b = 0;
2225  uint32_t padding = 0;
2226  int ret;
2227 
2228  if (mkv_projection->private.size && priv_data[0] != 0) {
2229  av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2230  return 0;
2231  }
2232 
2233  switch (track->video.projection.type) {
2235  if (track->video.projection.private.size == 20) {
2236  t = AV_RB32(priv_data + 4);
2237  b = AV_RB32(priv_data + 8);
2238  l = AV_RB32(priv_data + 12);
2239  r = AV_RB32(priv_data + 16);
2240 
2241  if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2242  av_log(logctx, AV_LOG_ERROR,
2243  "Invalid bounding rectangle coordinates "
2244  "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2245  l, t, r, b);
2246  return AVERROR_INVALIDDATA;
2247  }
2248  } else if (track->video.projection.private.size != 0) {
2249  av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2250  return AVERROR_INVALIDDATA;
2251  }
2252 
2253  if (l || t || r || b)
2254  projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2255  else
2256  projection = AV_SPHERICAL_EQUIRECTANGULAR;
2257  break;
2259  if (track->video.projection.private.size < 4) {
2260  av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2261  return AVERROR_INVALIDDATA;
2262  } else if (track->video.projection.private.size == 12) {
2263  uint32_t layout = AV_RB32(priv_data + 4);
2264  if (layout) {
2265  av_log(logctx, AV_LOG_WARNING,
2266  "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2267  return 0;
2268  }
2269  projection = AV_SPHERICAL_CUBEMAP;
2270  padding = AV_RB32(priv_data + 8);
2271  } else {
2272  av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2273  return AVERROR_INVALIDDATA;
2274  }
2275  break;
2277  /* No Spherical metadata */
2278  return 0;
2279  default:
2280  av_log(logctx, AV_LOG_WARNING,
2281  "Unknown spherical metadata type %"PRIu64"\n",
2282  track->video.projection.type);
2283  return 0;
2284  }
2285 
2286  spherical = av_spherical_alloc(&spherical_size);
2287  if (!spherical)
2288  return AVERROR(ENOMEM);
2289 
2290  spherical->projection = projection;
2291 
2292  spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2293  spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2294  spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2295 
2296  spherical->padding = padding;
2297 
2298  spherical->bound_left = l;
2299  spherical->bound_top = t;
2300  spherical->bound_right = r;
2301  spherical->bound_bottom = b;
2302 
2304  spherical_size);
2305  if (ret < 0) {
2306  av_freep(&spherical);
2307  return ret;
2308  }
2309 
2310  return 0;
2311 }
2312 
2313 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2314 {
2315  const AVCodecTag *codec_tags;
2316 
2317  codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2319 
2320  /* Normalize noncompliant private data that starts with the fourcc
2321  * by expanding/shifting the data by 4 bytes and storing the data
2322  * size at the start. */
2323  if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2324  int ret = av_buffer_realloc(&track->codec_priv.buf,
2326  if (ret < 0)
2327  return ret;
2328 
2329  track->codec_priv.data = track->codec_priv.buf->data;
2330  memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2331  track->codec_priv.size += 4;
2332  AV_WB32(track->codec_priv.data, track->codec_priv.size);
2333  }
2334 
2335  *fourcc = AV_RL32(track->codec_priv.data + 4);
2336  *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2337 
2338  return 0;
2339 }
2340 
2342 {
2343  MatroskaDemuxContext *matroska = s->priv_data;
2344  MatroskaTrack *tracks = matroska->tracks.elem;
2345  AVStream *st;
2346  int i, j, ret;
2347  int k;
2348 
2349  for (i = 0; i < matroska->tracks.nb_elem; i++) {
2350  MatroskaTrack *track = &tracks[i];
2352  EbmlList *encodings_list = &track->encodings;
2353  MatroskaTrackEncoding *encodings = encodings_list->elem;
2354  uint8_t *extradata = NULL;
2355  int extradata_size = 0;
2356  int extradata_offset = 0;
2357  uint32_t fourcc = 0;
2358  AVIOContext b;
2359  char* key_id_base64 = NULL;
2360  int bit_depth = -1;
2361 
2362  /* Apply some sanity checks. */
2363  if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2364  track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2365  track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2366  track->type != MATROSKA_TRACK_TYPE_METADATA) {
2367  av_log(matroska->ctx, AV_LOG_INFO,
2368  "Unknown or unsupported track type %"PRIu64"\n",
2369  track->type);
2370  continue;
2371  }
2372  if (!track->codec_id)
2373  continue;
2374 
2375  if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
2376  || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
2377  || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2378  || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2379  ) {
2380  av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
2381  continue;
2382  }
2383 
2384  if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2385  isnan(track->audio.samplerate)) {
2386  av_log(matroska->ctx, AV_LOG_WARNING,
2387  "Invalid sample rate %f, defaulting to 8000 instead.\n",
2388  track->audio.samplerate);
2389  track->audio.samplerate = 8000;
2390  }
2391 
2392  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2393  if (!track->default_duration && track->video.frame_rate > 0) {
2394  double default_duration = 1000000000 / track->video.frame_rate;
2395  if (default_duration > UINT64_MAX || default_duration < 0) {
2396  av_log(matroska->ctx, AV_LOG_WARNING,
2397  "Invalid frame rate %e. Cannot calculate default duration.\n",
2398  track->video.frame_rate);
2399  } else {
2400  track->default_duration = default_duration;
2401  }
2402  }
2403  if (track->video.display_width == -1)
2404  track->video.display_width = track->video.pixel_width;
2405  if (track->video.display_height == -1)
2406  track->video.display_height = track->video.pixel_height;
2407  if (track->video.color_space.size == 4)
2408  fourcc = AV_RL32(track->video.color_space.data);
2409  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2410  if (!track->audio.out_samplerate)
2411  track->audio.out_samplerate = track->audio.samplerate;
2412  }
2413  if (encodings_list->nb_elem > 1) {
2414  av_log(matroska->ctx, AV_LOG_ERROR,
2415  "Multiple combined encodings not supported");
2416  } else if (encodings_list->nb_elem == 1) {
2417  if (encodings[0].type) {
2418  if (encodings[0].encryption.key_id.size > 0) {
2419  /* Save the encryption key id to be stored later as a
2420  metadata tag. */
2421  const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2422  key_id_base64 = av_malloc(b64_size);
2423  if (key_id_base64 == NULL)
2424  return AVERROR(ENOMEM);
2425 
2426  av_base64_encode(key_id_base64, b64_size,
2427  encodings[0].encryption.key_id.data,
2428  encodings[0].encryption.key_id.size);
2429  } else {
2430  encodings[0].scope = 0;
2431  av_log(matroska->ctx, AV_LOG_ERROR,
2432  "Unsupported encoding type");
2433  }
2434  } else if (
2435 #if CONFIG_ZLIB
2436  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2437 #endif
2438 #if CONFIG_BZLIB
2439  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2440 #endif
2441 #if CONFIG_LZO
2442  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2443 #endif
2444  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2445  encodings[0].scope = 0;
2446  av_log(matroska->ctx, AV_LOG_ERROR,
2447  "Unsupported encoding type");
2448  } else if (track->codec_priv.size && encodings[0].scope & 2) {
2449  uint8_t *codec_priv = track->codec_priv.data;
2450  int ret = matroska_decode_buffer(&track->codec_priv.data,
2451  &track->codec_priv.size,
2452  track);
2453  if (ret < 0) {
2454  track->codec_priv.data = NULL;
2455  track->codec_priv.size = 0;
2456  av_log(matroska->ctx, AV_LOG_ERROR,
2457  "Failed to decode codec private data\n");
2458  }
2459 
2460  if (codec_priv != track->codec_priv.data) {
2461  av_buffer_unref(&track->codec_priv.buf);
2462  if (track->codec_priv.data) {
2463  track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2465  NULL, NULL, 0);
2466  if (!track->codec_priv.buf) {
2467  av_freep(&track->codec_priv.data);
2468  track->codec_priv.size = 0;
2469  return AVERROR(ENOMEM);
2470  }
2471  }
2472  }
2473  }
2474  }
2475  track->needs_decoding = encodings && !encodings[0].type &&
2476  encodings[0].scope & 1 &&
2477  (encodings[0].compression.algo !=
2479  encodings[0].compression.settings.size);
2480 
2481  for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2482  if (av_strstart(track->codec_id, ff_mkv_codec_tags[j].str, NULL)) {
2484  break;
2485  }
2486  }
2487 
2488  st = track->stream = avformat_new_stream(s, NULL);
2489  if (!st) {
2490  av_free(key_id_base64);
2491  return AVERROR(ENOMEM);
2492  }
2493 
2494  if (key_id_base64) {
2495  /* export encryption key id as base64 metadata tag */
2496  av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2498  }
2499 
2500  if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2501  track->codec_priv.size >= 40 &&
2502  track->codec_priv.data) {
2503  track->ms_compat = 1;
2504  bit_depth = AV_RL16(track->codec_priv.data + 14);
2505  fourcc = AV_RL32(track->codec_priv.data + 16);
2507  fourcc);
2508  if (!codec_id)
2510  fourcc);
2511  extradata_offset = 40;
2512  } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2513  track->codec_priv.size >= 14 &&
2514  track->codec_priv.data) {
2515  int ret;
2517  track->codec_priv.size,
2518  0, NULL, NULL, NULL, NULL);
2519  ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2520  if (ret < 0)
2521  return ret;
2522  codec_id = st->codecpar->codec_id;
2523  fourcc = st->codecpar->codec_tag;
2524  extradata_offset = FFMIN(track->codec_priv.size, 18);
2525  } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2526  /* Normally 36, but allow noncompliant private data */
2527  && (track->codec_priv.size >= 32)
2528  && (track->codec_priv.data)) {
2529  uint16_t sample_size;
2530  int ret = get_qt_codec(track, &fourcc, &codec_id);
2531  if (ret < 0)
2532  return ret;
2533  sample_size = AV_RB16(track->codec_priv.data + 26);
2534  if (fourcc == 0) {
2535  if (sample_size == 8) {
2536  fourcc = MKTAG('r','a','w',' ');
2538  } else if (sample_size == 16) {
2539  fourcc = MKTAG('t','w','o','s');
2541  }
2542  }
2543  if ((fourcc == MKTAG('t','w','o','s') ||
2544  fourcc == MKTAG('s','o','w','t')) &&
2545  sample_size == 8)
2547  } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2548  (track->codec_priv.size >= 21) &&
2549  (track->codec_priv.data)) {
2550  int ret = get_qt_codec(track, &fourcc, &codec_id);
2551  if (ret < 0)
2552  return ret;
2553  if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2554  fourcc = MKTAG('S','V','Q','3');
2556  }
2557  if (codec_id == AV_CODEC_ID_NONE)
2558  av_log(matroska->ctx, AV_LOG_ERROR,
2559  "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2560  if (track->codec_priv.size >= 86) {
2561  bit_depth = AV_RB16(track->codec_priv.data + 82);
2563  track->codec_priv.size,
2564  0, NULL, NULL, NULL, NULL);
2565  if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2566  bit_depth &= 0x1F;
2567  track->has_palette = 1;
2568  }
2569  }
2570  } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2571  switch (track->audio.bitdepth) {
2572  case 8:
2574  break;
2575  case 24:
2577  break;
2578  case 32:
2580  break;
2581  }
2582  } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2583  switch (track->audio.bitdepth) {
2584  case 8:
2586  break;
2587  case 24:
2589  break;
2590  case 32:
2592  break;
2593  }
2594  } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2595  track->audio.bitdepth == 64) {
2597  } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2598  int profile = matroska_aac_profile(track->codec_id);
2599  int sri = matroska_aac_sri(track->audio.samplerate);
2600  extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2601  if (!extradata)
2602  return AVERROR(ENOMEM);
2603  extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2604  extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2605  if (strstr(track->codec_id, "SBR")) {
2606  sri = matroska_aac_sri(track->audio.out_samplerate);
2607  extradata[2] = 0x56;
2608  extradata[3] = 0xE5;
2609  extradata[4] = 0x80 | (sri << 3);
2610  extradata_size = 5;
2611  } else
2612  extradata_size = 2;
2613  } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2614  /* Only ALAC's magic cookie is stored in Matroska's track headers.
2615  * Create the "atom size", "tag", and "tag version" fields the
2616  * decoder expects manually. */
2617  extradata_size = 12 + track->codec_priv.size;
2618  extradata = av_mallocz(extradata_size +
2620  if (!extradata)
2621  return AVERROR(ENOMEM);
2622  AV_WB32(extradata, extradata_size);
2623  memcpy(&extradata[4], "alac", 4);
2624  AV_WB32(&extradata[8], 0);
2625  memcpy(&extradata[12], track->codec_priv.data,
2626  track->codec_priv.size);
2627  } else if (codec_id == AV_CODEC_ID_TTA) {
2628  uint8_t *ptr;
2629  if (track->audio.channels > UINT16_MAX ||
2630  track->audio.bitdepth > UINT16_MAX) {
2631  av_log(matroska->ctx, AV_LOG_WARNING,
2632  "Too large audio channel number %"PRIu64
2633  " or bitdepth %"PRIu64". Skipping track.\n",
2634  track->audio.channels, track->audio.bitdepth);
2635  if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2636  return AVERROR_INVALIDDATA;
2637  else
2638  continue;
2639  }
2640  if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2641  return AVERROR_INVALIDDATA;
2642  extradata_size = 22;
2643  extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2644  if (!extradata)
2645  return AVERROR(ENOMEM);
2646  ptr = extradata;
2647  bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2648  bytestream_put_le16(&ptr, 1);
2649  bytestream_put_le16(&ptr, track->audio.channels);
2650  bytestream_put_le16(&ptr, track->audio.bitdepth);
2651  bytestream_put_le32(&ptr, track->audio.out_samplerate);
2652  bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2653  track->audio.out_samplerate,
2654  AV_TIME_BASE * 1000));
2655  } else if (codec_id == AV_CODEC_ID_RV10 ||
2659  extradata_offset = 26;
2660  } else if (codec_id == AV_CODEC_ID_RA_144) {
2661  track->audio.out_samplerate = 8000;
2662  track->audio.channels = 1;
2663  } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2667  && track->codec_priv.data) {
2668  int flavor;
2669 
2671  track->codec_priv.size,
2672  0, NULL, NULL, NULL, NULL);
2673  avio_skip(&b, 22);
2674  flavor = avio_rb16(&b);
2675  track->audio.coded_framesize = avio_rb32(&b);
2676  avio_skip(&b, 12);
2677  track->audio.sub_packet_h = avio_rb16(&b);
2678  track->audio.frame_size = avio_rb16(&b);
2679  track->audio.sub_packet_size = avio_rb16(&b);
2680  if (track->audio.coded_framesize <= 0 ||
2681  track->audio.sub_packet_h <= 0 ||
2682  track->audio.frame_size <= 0)
2683  return AVERROR_INVALIDDATA;
2684 
2685  if (codec_id == AV_CODEC_ID_RA_288) {
2686  if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2687  != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2688  return AVERROR_INVALIDDATA;
2689  st->codecpar->block_align = track->audio.coded_framesize;
2690  track->codec_priv.size = 0;
2691  } else {
2692  if (codec_id == AV_CODEC_ID_SIPR) {
2693  static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2694  if (flavor > 3)
2695  return AVERROR_INVALIDDATA;
2696  track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2697  st->codecpar->bit_rate = sipr_bit_rate[flavor];
2698  } else if (track->audio.sub_packet_size <= 0 ||
2699  track->audio.frame_size % track->audio.sub_packet_size)
2700  return AVERROR_INVALIDDATA;
2701  st->codecpar->block_align = track->audio.sub_packet_size;
2702  extradata_offset = 78;
2703  }
2704  track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2705  track->audio.frame_size);
2706  if (!track->audio.buf)
2707  return AVERROR(ENOMEM);
2708  } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2709  ret = matroska_parse_flac(s, track, &extradata_offset);
2710  if (ret < 0)
2711  return ret;
2712  } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2713  av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2714  "in absence of valid CodecPrivate.\n");
2715  extradata_size = 2;
2716  extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2717  if (!extradata)
2718  return AVERROR(ENOMEM);
2719  AV_WL16(extradata, 0x410);
2720  } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2721  fourcc = AV_RL32(track->codec_priv.data);
2722  } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2723  /* we don't need any value stored in CodecPrivate.
2724  make sure that it's not exported as extradata. */
2725  track->codec_priv.size = 0;
2726  } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2727  /* For now, propagate only the OBUs, if any. Once libavcodec is
2728  updated to handle isobmff style extradata this can be removed. */
2729  extradata_offset = 4;
2730  }
2731  track->codec_priv.size -= extradata_offset;
2732 
2733  if (codec_id == AV_CODEC_ID_NONE)
2734  av_log(matroska->ctx, AV_LOG_INFO,
2735  "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2736 
2737  if (track->time_scale < 0.01) {
2738  av_log(matroska->ctx, AV_LOG_WARNING,
2739  "Track TimestampScale too small %f, assuming 1.0.\n",
2740  track->time_scale);
2741  track->time_scale = 1.0;
2742  }
2743  avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2744  1000 * 1000 * 1000); /* 64 bit pts in ns */
2745 
2746  /* convert the delay from ns to the track timebase */
2748  (AVRational){ 1, 1000000000 },
2749  st->time_base);
2750 
2751  st->codecpar->codec_id = codec_id;
2752 
2753  if (strcmp(track->language, "und"))
2754  av_dict_set(&st->metadata, "language", track->language, 0);
2755  av_dict_set(&st->metadata, "title", track->name, 0);
2756 
2757  if (track->flag_default)
2759  if (track->flag_forced)
2761  if (track->flag_comment)
2763  if (track->flag_hearingimpaired)
2765  if (track->flag_visualimpaired)
2767  if (track->flag_original.count > 0)
2770 
2771  if (!st->codecpar->extradata) {
2772  if (extradata) {
2773  st->codecpar->extradata = extradata;
2774  st->codecpar->extradata_size = extradata_size;
2775  } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2776  if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2777  return AVERROR(ENOMEM);
2778  memcpy(st->codecpar->extradata,
2779  track->codec_priv.data + extradata_offset,
2780  track->codec_priv.size);
2781  }
2782  }
2783 
2784  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2786  int display_width_mul = 1;
2787  int display_height_mul = 1;
2788 
2790  st->codecpar->codec_tag = fourcc;
2791  if (bit_depth >= 0)
2793  st->codecpar->width = track->video.pixel_width;
2794  st->codecpar->height = track->video.pixel_height;
2795 
2797  st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2800 
2802  mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2803 
2806  &st->sample_aspect_ratio.den,
2807  st->codecpar->height * track->video.display_width * display_width_mul,
2808  st->codecpar->width * track->video.display_height * display_height_mul,
2809  INT_MAX);
2810  }
2811  if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2813 
2814  if (track->default_duration) {
2815  int div = track->default_duration <= INT64_MAX ? 1 : 2;
2817  1000000000 / div, track->default_duration / div, 30000);
2818 #if FF_API_R_FRAME_RATE
2819  if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2820  && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2821  st->r_frame_rate = st->avg_frame_rate;
2822 #endif
2823  }
2824 
2825  /* export stereo mode flag as metadata tag */
2827  av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2828 
2829  /* export alpha mode flag as metadata tag */
2830  if (track->video.alpha_mode)
2831  av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2832 
2833  /* if we have virtual track, mark the real tracks */
2834  for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2835  char buf[32];
2837  continue;
2838  snprintf(buf, sizeof(buf), "%s_%d",
2840  for (k=0; k < matroska->tracks.nb_elem; k++)
2841  if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2842  av_dict_set(&tracks[k].stream->metadata,
2843  "stereo_mode", buf, 0);
2844  break;
2845  }
2846  }
2847  // add stream level stereo3d side data if it is a supported format
2849  track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2850  int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2851  if (ret < 0)
2852  return ret;
2853  }
2854 
2855  ret = mkv_parse_video_color(st, track);
2856  if (ret < 0)
2857  return ret;
2858  ret = mkv_parse_video_projection(st, track, matroska->ctx);
2859  if (ret < 0)
2860  return ret;
2861  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2863  st->codecpar->codec_tag = fourcc;
2864  st->codecpar->sample_rate = track->audio.out_samplerate;
2865  st->codecpar->channels = track->audio.channels;
2866  if (!st->codecpar->bits_per_coded_sample)
2868  if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2869  st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2872  else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2874  if (track->codec_delay > 0) {
2876  (AVRational){1, 1000000000},
2877  (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2878  48000 : st->codecpar->sample_rate});
2879  }
2880  if (track->seek_preroll > 0) {
2882  (AVRational){1, 1000000000},
2883  (AVRational){1, st->codecpar->sample_rate});
2884  }
2885  } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2887 
2888  if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2890  } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2892  } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2894  }
2895  } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2897 
2898  if (track->flag_textdescriptions)
2900  }
2901  }
2902 
2903  return 0;
2904 }
2905 
2907 {
2908  MatroskaDemuxContext *matroska = s->priv_data;
2909  EbmlList *attachments_list = &matroska->attachments;
2910  EbmlList *chapters_list = &matroska->chapters;
2911  MatroskaAttachment *attachments;
2912  MatroskaChapter *chapters;
2913  uint64_t max_start = 0;
2914  int64_t pos;
2915  Ebml ebml = { 0 };
2916  int i, j, res;
2917 
2918  matroska->ctx = s;
2919  matroska->cues_parsing_deferred = 1;
2920 
2921  /* First read the EBML header. */
2922  if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2923  av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2924  ebml_free(ebml_syntax, &ebml);
2925  return AVERROR_INVALIDDATA;
2926  }
2927  if (ebml.version > EBML_VERSION ||
2928  ebml.max_size > sizeof(uint64_t) ||
2929  ebml.id_length > sizeof(uint32_t) ||
2930  ebml.doctype_version > 3) {
2932  "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2933  ebml.version, ebml.doctype, ebml.doctype_version);
2934  ebml_free(ebml_syntax, &ebml);
2935  return AVERROR_PATCHWELCOME;
2936  } else if (ebml.doctype_version == 3) {
2937  av_log(matroska->ctx, AV_LOG_WARNING,
2938  "EBML header using unsupported features\n"
2939  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2940  ebml.version, ebml.doctype, ebml.doctype_version);
2941  }
2942  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2943  if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2944  break;
2946  av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2947  if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2948  ebml_free(ebml_syntax, &ebml);
2949  return AVERROR_INVALIDDATA;
2950  }
2951  }
2952  ebml_free(ebml_syntax, &ebml);
2953 
2954  matroska->pkt = av_packet_alloc();
2955  if (!matroska->pkt)
2956  return AVERROR(ENOMEM);
2957 
2958  /* The next thing is a segment. */
2959  pos = avio_tell(matroska->ctx->pb);
2960  res = ebml_parse(matroska, matroska_segments, matroska);
2961  // Try resyncing until we find an EBML_STOP type element.
2962  while (res != 1) {
2963  res = matroska_resync(matroska, pos);
2964  if (res < 0)
2965  goto fail;
2966  pos = avio_tell(matroska->ctx->pb);
2967  res = ebml_parse(matroska, matroska_segment, matroska);
2968  if (res == AVERROR(EIO)) // EOF is translated to EIO, this exists the loop on EOF
2969  goto fail;
2970  }
2971  /* Set data_offset as it might be needed later by seek_frame_generic. */
2972  if (matroska->current_id == MATROSKA_ID_CLUSTER)
2973  s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2974  matroska_execute_seekhead(matroska);
2975 
2976  if (!matroska->time_scale)
2977  matroska->time_scale = 1000000;
2978  if (isnan(matroska->duration))
2979  matroska->duration = 0;
2980  if (matroska->duration)
2981  matroska->ctx->duration = matroska->duration * matroska->time_scale *
2982  1000 / AV_TIME_BASE;
2983  av_dict_set(&s->metadata, "title", matroska->title, 0);
2984  av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2985 
2986  if (matroska->date_utc.size == 8)
2987  matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2988 
2989  res = matroska_parse_tracks(s);
2990  if (res < 0)
2991  goto fail;
2992 
2993  attachments = attachments_list->elem;
2994  for (j = 0; j < attachments_list->nb_elem; j++) {
2995  if (!(attachments[j].filename && attachments[j].mime &&
2996  attachments[j].bin.data && attachments[j].bin.size > 0)) {
2997  av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2998  } else {
3000  if (!st)
3001  break;
3002  av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
3003  av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
3004  if (attachments[j].description)
3005  av_dict_set(&st->metadata, "title", attachments[j].description, 0);
3007 
3008  for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3009  if (av_strstart(attachments[j].mime, mkv_image_mime_tags[i].str, NULL)) {
3011  break;
3012  }
3013  }
3014 
3015  attachments[j].stream = st;
3016 
3017  if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
3018  AVPacket *pkt = &st->attached_pic;
3019 
3022 
3024  pkt->buf = attachments[j].bin.buf;
3025  attachments[j].bin.buf = NULL;
3026  pkt->data = attachments[j].bin.data;
3027  pkt->size = attachments[j].bin.size;
3028  pkt->stream_index = st->index;
3030  } else {
3032  if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
3033  break;
3034  memcpy(st->codecpar->extradata, attachments[j].bin.data,
3035  attachments[j].bin.size);
3036 
3037  for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3038  if (av_strstart(attachments[j].mime, mkv_mime_tags[i].str, NULL)) {
3040  break;
3041  }
3042  }
3043  }
3044  }
3045  }
3046 
3047  chapters = chapters_list->elem;
3048  for (i = 0; i < chapters_list->nb_elem; i++)
3049  if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
3050  (max_start == 0 || chapters[i].start > max_start)) {
3051  chapters[i].chapter =
3052  avpriv_new_chapter(s, chapters[i].uid,
3053  (AVRational) { 1, 1000000000 },
3054  chapters[i].start, chapters[i].end,
3055  chapters[i].title);
3056  max_start = chapters[i].start;
3057  }
3058 
3059  matroska_add_index_entries(matroska);
3060 
3062 
3063  return 0;
3064 fail:
3066  return res;
3068 
3069 /*
3070  * Put one packet in an application-supplied AVPacket struct.
3071  * Returns 0 on success or -1 on failure.
3072  */
3073 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3074  AVPacket *pkt)
3075 {
3076  if (matroska->queue) {
3077  MatroskaTrack *tracks = matroska->tracks.elem;
3078  MatroskaTrack *track;
3079 
3080  avpriv_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
3081  track = &tracks[pkt->stream_index];
3082  if (track->has_palette) {
3084  if (!pal) {
3085  av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3086  } else {
3087  memcpy(pal, track->palette, AVPALETTE_SIZE);
3088  }
3089  track->has_palette = 0;
3090  }
3091  return 0;
3092  }
3093 
3094  return -1;
3095 }
3096 
3097 /*
3098  * Free all packets in our internal queue.
3099  */
3100 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3101 {
3102  avpriv_packet_list_free(&matroska->queue, &matroska->queue_end);
3103 }
3104 
3105 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3106  int size, int type, AVIOContext *pb,
3107  uint32_t lace_size[256], int *laces)
3108 {
3109  int n;
3110  uint8_t *data = *buf;
3111 
3112  if (!type) {
3113  *laces = 1;
3114  lace_size[0] = size;
3115  return 0;
3116  }
3117 
3118  if (size <= 0)
3119  return AVERROR_INVALIDDATA;
3120 
3121  *laces = *data + 1;
3122  data += 1;
3123  size -= 1;
3124 
3125  switch (type) {
3126  case 0x1: /* Xiph lacing */
3127  {
3128  uint8_t temp;
3129  uint32_t total = 0;
3130  for (n = 0; n < *laces - 1; n++) {
3131  lace_size[n] = 0;
3132 
3133  do {
3134  if (size <= total)
3135  return AVERROR_INVALIDDATA;
3136  temp = *data;
3137  total += temp;
3138  lace_size[n] += temp;
3139  data += 1;
3140  size -= 1;
3141  } while (temp == 0xff);
3142  }
3143  if (size < total)
3144  return AVERROR_INVALIDDATA;
3145 
3146  lace_size[n] = size - total;
3147  break;
3148  }
3149 
3150  case 0x2: /* fixed-size lacing */
3151  if (size % (*laces))
3152  return AVERROR_INVALIDDATA;
3153  for (n = 0; n < *laces; n++)
3154  lace_size[n] = size / *laces;
3155  break;
3156 
3157  case 0x3: /* EBML lacing */
3158  {
3159  uint64_t num;
3160  uint64_t total;
3161  int offset;
3162 
3163  avio_skip(pb, 4);
3164 
3165  n = ebml_read_num(matroska, pb, 8, &num, 1);
3166  if (n < 0)
3167  return n;
3168  if (num > INT_MAX)
3169  return AVERROR_INVALIDDATA;
3170 
3171  total = lace_size[0] = num;
3172  offset = n;
3173  for (n = 1; n < *laces - 1; n++) {
3174  int64_t snum;
3175  int r;
3176  r = matroska_ebmlnum_sint(matroska, pb, &snum);
3177  if (r < 0)
3178  return r;
3179  if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3180  return AVERROR_INVALIDDATA;
3181 
3182  lace_size[n] = lace_size[n - 1] + snum;
3183  total += lace_size[n];
3184  offset += r;
3185  }
3186  data += offset;
3187  size -= offset;
3188  if (size < total)
3189  return AVERROR_INVALIDDATA;
3190 
3191  lace_size[*laces - 1] = size - total;
3192  break;
3193  }
3194  }
3196  *buf = data;
3197 
3198  return 0;
3199 }
3200 
3201 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3202  MatroskaTrack *track, AVStream *st,
3203  uint8_t *data, int size, uint64_t timecode,
3204  int64_t pos)
3205 {
3206  const int a = st->codecpar->block_align;
3207  const int sps = track->audio.sub_packet_size;
3208  const int cfs = track->audio.coded_framesize;
3209  const int h = track->audio.sub_packet_h;
3210  const int w = track->audio.frame_size;
3211  int y = track->audio.sub_packet_cnt;
3212  int x;
3213 
3214  if (!track->audio.pkt_cnt) {
3215  if (track->audio.sub_packet_cnt == 0)
3216  track->audio.buf_timecode = timecode;
3217  if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3218  if (size < cfs * h / 2) {
3219  av_log(matroska->ctx, AV_LOG_ERROR,
3220  "Corrupt int4 RM-style audio packet size\n");
3221  return AVERROR_INVALIDDATA;
3222  }
3223  for (x = 0; x < h / 2; x++)
3224  memcpy(track->audio.buf + x * 2 * w + y * cfs,
3225  data + x * cfs, cfs);
3226  } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3227  if (size < w) {
3228  av_log(matroska->ctx, AV_LOG_ERROR,
3229  "Corrupt sipr RM-style audio packet size\n");
3230  return AVERROR_INVALIDDATA;
3231  }
3232  memcpy(track->audio.buf + y * w, data, w);
3233  } else {
3234  if (size < w) {
3235  av_log(matroska->ctx, AV_LOG_ERROR,
3236  "Corrupt generic RM-style audio packet size\n");
3237  return AVERROR_INVALIDDATA;
3238  }
3239  for (x = 0; x < w / sps; x++)
3240  memcpy(track->audio.buf +
3241  sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3242  data + x * sps, sps);
3243  }
3244 
3245  if (++track->audio.sub_packet_cnt >= h) {
3246  if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3247  ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3248  track->audio.sub_packet_cnt = 0;
3249  track->audio.pkt_cnt = h * w / a;
3250  }
3251  }
3252 
3253  while (track->audio.pkt_cnt) {
3254  int ret;
3255  AVPacket *pkt = matroska->pkt;
3256 
3257  ret = av_new_packet(pkt, a);
3258  if (ret < 0) {
3259  return ret;
3260  }
3261  memcpy(pkt->data,
3262  track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3263  a);
3264  pkt->pts = track->audio.buf_timecode;
3266  pkt->pos = pos;
3267  pkt->stream_index = st->index;
3268  ret = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3269  if (ret < 0) {
3271  return AVERROR(ENOMEM);
3272  }
3273  }
3274 
3275  return 0;
3276 }
3277 
3278 /* reconstruct full wavpack blocks from mangled matroska ones */
3279 static int matroska_parse_wavpack(MatroskaTrack *track,
3280  uint8_t **data, int *size)
3281 {
3282  uint8_t *dst = NULL;
3283  uint8_t *src = *data;
3284  int dstlen = 0;
3285  int srclen = *size;
3286  uint32_t samples;
3287  uint16_t ver;
3288  int ret, offset = 0;
3289 
3290  if (srclen < 12)
3291  return AVERROR_INVALIDDATA;
3292 
3293  av_assert1(track->stream->codecpar->extradata_size >= 2);
3294  ver = AV_RL16(track->stream->codecpar->extradata);
3295 
3296  samples = AV_RL32(src);
3297  src += 4;
3298  srclen -= 4;
3299 
3300  while (srclen >= 8) {
3301  int multiblock;
3302  uint32_t blocksize;
3303  uint8_t *tmp;
3304 
3305  uint32_t flags = AV_RL32(src);
3306  uint32_t crc = AV_RL32(src + 4);
3307  src += 8;
3308  srclen -= 8;
3309 
3310  multiblock = (flags & 0x1800) != 0x1800;
3311  if (multiblock) {
3312  if (srclen < 4) {
3314  goto fail;
3315  }
3316  blocksize = AV_RL32(src);
3317  src += 4;
3318  srclen -= 4;
3319  } else
3320  blocksize = srclen;
3321 
3322  if (blocksize > srclen) {
3324  goto fail;
3325  }
3326 
3327  tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3328  if (!tmp) {
3329  ret = AVERROR(ENOMEM);
3330  goto fail;
3331  }
3332  dst = tmp;
3333  dstlen += blocksize + 32;
3334 
3335  AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3336  AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3337  AV_WL16(dst + offset + 8, ver); // version
3338  AV_WL16(dst + offset + 10, 0); // track/index_no
3339  AV_WL32(dst + offset + 12, 0); // total samples
3340  AV_WL32(dst + offset + 16, 0); // block index
3341  AV_WL32(dst + offset + 20, samples); // number of samples
3342  AV_WL32(dst + offset + 24, flags); // flags
3343  AV_WL32(dst + offset + 28, crc); // crc
3344  memcpy(dst + offset + 32, src, blocksize); // block data
3345 
3346  src += blocksize;
3347  srclen -= blocksize;
3348  offset += blocksize + 32;
3349  }
3350 
3351  memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3352 
3353  *data = dst;
3354  *size = dstlen;
3355 
3356  return 0;
3358 fail:
3359  av_freep(&dst);
3360  return ret;
3361 }
3362 
3363 static int matroska_parse_prores(MatroskaTrack *track,
3364  uint8_t **data, int *size)
3365 {
3366  uint8_t *dst;
3367  int dstlen = *size + 8;
3368 
3369  dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3370  if (!dst)
3371  return AVERROR(ENOMEM);
3372 
3373  AV_WB32(dst, dstlen);
3374  AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3375  memcpy(dst + 8, *data, dstlen - 8);
3376  memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3377 
3378  *data = dst;
3379  *size = dstlen;
3380 
3381  return 0;
3382 }
3383 
3384 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3385  MatroskaTrack *track,
3386  AVStream *st,
3387  uint8_t *data, int data_len,
3388  uint64_t timecode,
3389  uint64_t duration,
3390  int64_t pos)
3391 {
3392  AVPacket *pkt = matroska->pkt;
3393  uint8_t *id, *settings, *text, *buf;
3394  int id_len, settings_len, text_len;
3395  uint8_t *p, *q;
3396  int err;
3397 
3398  if (data_len <= 0)
3399  return AVERROR_INVALIDDATA;
3400 
3401  p = data;
3402  q = data + data_len;
3403 
3404  id = p;
3405  id_len = -1;
3406  while (p < q) {
3407  if (*p == '\r' || *p == '\n') {
3408  id_len = p - id;
3409  if (*p == '\r')
3410  p++;
3411  break;
3412  }
3413  p++;
3414  }
3415 
3416  if (p >= q || *p != '\n')
3417  return AVERROR_INVALIDDATA;
3418  p++;
3419 
3420  settings = p;
3421  settings_len = -1;
3422  while (p < q) {
3423  if (*p == '\r' || *p == '\n') {
3424  settings_len = p - settings;
3425  if (*p == '\r')
3426  p++;
3427  break;
3428  }
3429  p++;
3430  }
3431 
3432  if (p >= q || *p != '\n')
3433  return AVERROR_INVALIDDATA;
3434  p++;
3435 
3436  text = p;
3437  text_len = q - p;
3438  while (text_len > 0) {
3439  const int len = text_len - 1;
3440  const uint8_t c = p[len];
3441  if (c != '\r' && c != '\n')
3442  break;
3443  text_len = len;
3444  }
3445 
3446  if (text_len <= 0)
3447  return AVERROR_INVALIDDATA;
3448 
3449  err = av_new_packet(pkt, text_len);
3450  if (err < 0) {
3451  return err;
3452  }
3453 
3454  memcpy(pkt->data, text, text_len);
3455 
3456  if (id_len > 0) {
3459  id_len);
3460  if (!buf) {
3462  return AVERROR(ENOMEM);
3463  }
3464  memcpy(buf, id, id_len);
3465  }
3466 
3467  if (settings_len > 0) {
3470  settings_len);
3471  if (!buf) {
3473  return AVERROR(ENOMEM);
3474  }
3475  memcpy(buf, settings, settings_len);
3476  }
3477 
3478  // Do we need this for subtitles?
3479  // pkt->flags = AV_PKT_FLAG_KEY;
3480 
3481  pkt->stream_index = st->index;
3482  pkt->pts = timecode;
3483 
3484  // Do we need this for subtitles?
3485  // pkt->dts = timecode;
3486 
3487  pkt->duration = duration;
3488  pkt->pos = pos;
3489 
3490  err = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3491  if (err < 0) {
3493  return AVERROR(ENOMEM);
3494  }
3495 
3496  return 0;
3497 }
3498 
3499 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3500  MatroskaTrack *track, AVStream *st,
3501  AVBufferRef *buf, uint8_t *data, int pkt_size,
3502  uint64_t timecode, uint64_t lace_duration,
3503  int64_t pos, int is_keyframe,
3504  uint8_t *additional, uint64_t additional_id, int additional_size,
3505  int64_t discard_padding)
3506 {
3507  uint8_t *pkt_data = data;
3508  int res = 0;
3509  AVPacket *pkt = matroska->pkt;
3510 
3511  if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3512  res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3513  if (res < 0) {
3514  av_log(matroska->ctx, AV_LOG_ERROR,
3515  "Error parsing a wavpack block.\n");
3516  goto fail;
3517  }
3518  if (!buf)
3519  av_freep(&data);
3520  buf = NULL;
3521  }
3522 
3523  if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3524  AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3525  res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3526  if (res < 0) {
3527  av_log(matroska->ctx, AV_LOG_ERROR,
3528  "Error parsing a prores block.\n");
3529  goto fail;
3530  }
3531  if (!buf)
3532  av_freep(&data);
3533  buf = NULL;
3534  }
3535 
3536  if (!pkt_size && !additional_size)
3537  goto no_output;
3538 
3540  if (!buf)
3541  pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3542  NULL, NULL, 0);
3543  else
3544  pkt->buf = av_buffer_ref(buf);
3545 
3546  if (!pkt->buf) {
3547  res = AVERROR(ENOMEM);
3548  goto fail;
3549  }
3550 
3551  pkt->data = pkt_data;
3552  pkt->size = pkt_size;
3553  pkt->flags = is_keyframe;
3554  pkt->stream_index = st->index;
3555 
3556  if (additional_size > 0) {
3557  uint8_t *side_data = av_packet_new_side_data(pkt,
3559  additional_size + 8);
3560  if (!side_data) {
3562  return AVERROR(ENOMEM);
3563  }
3564  AV_WB64(side_data, additional_id);
3565  memcpy(side_data + 8, additional, additional_size);
3566  }
3567 
3568  if (discard_padding) {
3569  uint8_t *side_data = av_packet_new_side_data(pkt,
3571  10);
3572  if (!side_data) {
3574  return AVERROR(ENOMEM);
3575  }
3576  discard_padding = av_rescale_q(discard_padding,
3577  (AVRational){1, 1000000000},
3578  (AVRational){1, st->codecpar->sample_rate});
3579  if (discard_padding > 0) {
3580  AV_WL32(side_data + 4, discard_padding);
3581  } else {
3582  AV_WL32(side_data, -discard_padding);
3583  }
3584  }
3585 
3586  if (track->ms_compat)
3587  pkt->dts = timecode;
3588  else
3589  pkt->pts = timecode;
3590  pkt->pos = pos;
3591  pkt->duration = lace_duration;
3592 
3593 #if FF_API_CONVERGENCE_DURATION
3595  if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3596  pkt->convergence_duration = lace_duration;
3597  }
3599 #endif
3600 
3601  res = avpriv_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, NULL, 0);
3602  if (res < 0) {
3604  return AVERROR(ENOMEM);
3605  }
3606 
3607  return 0;
3608 
3609 no_output:
3611  if (!buf)
3612  av_free(pkt_data);
3613  return res;
3614 }
3615 
3617  int size, int64_t pos, uint64_t cluster_time,
3618  uint64_t block_duration, int is_keyframe,
3619  uint8_t *additional, uint64_t additional_id, int additional_size,
3620  int64_t cluster_pos, int64_t discard_padding)
3621 {
3622  uint64_t timecode = AV_NOPTS_VALUE;
3623  MatroskaTrack *track;
3624  AVIOContext pb;
3625  int res = 0;
3626  AVStream *st;
3627  int16_t block_time;
3628  uint32_t lace_size[256];
3629  int n, flags, laces = 0;
3630  uint64_t num;
3631  int trust_default_duration;
3632 
3633  ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3634 
3635  if ((n = ebml_read_num(matroska, &pb, 8, &num, 1)) < 0)
3636  return n;
3637  data += n;
3638  size -= n;
3639 
3640  track = matroska_find_track_by_num(matroska, num);
3641  if (!track || size < 3)
3642  return AVERROR_INVALIDDATA;
3643 
3644  if (!(st = track->stream)) {
3645  av_log(matroska->ctx, AV_LOG_VERBOSE,
3646  "No stream associated to TrackNumber %"PRIu64". "
3647  "Ignoring Block with this TrackNumber.\n", num);
3648  return 0;
3649  }
3650 
3651  if (st->discard >= AVDISCARD_ALL)
3652  return res;
3653  if (block_duration > INT64_MAX)
3654  block_duration = INT64_MAX;
3655 
3656  block_time = sign_extend(AV_RB16(data), 16);
3657  data += 2;
3658  flags = *data++;
3659  size -= 3;
3660  if (is_keyframe == -1)
3661  is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3662 
3663  if (cluster_time != (uint64_t) -1 &&
3664  (block_time >= 0 || cluster_time >= -block_time)) {
3665  uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
3666  timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
3667  if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3668  timecode < track->end_timecode)
3669  is_keyframe = 0; /* overlapping subtitles are not key frame */
3670  if (is_keyframe) {
3671  ff_reduce_index(matroska->ctx, st->index);
3672  av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3674  }
3675  }
3676 
3677  if (matroska->skip_to_keyframe &&
3678  track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3679  // Compare signed timecodes. Timecode may be negative due to codec delay
3680  // offset. We don't support timestamps greater than int64_t anyway - see
3681  // AVPacket's pts.
3682  if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3683  return res;
3684  if (is_keyframe)
3685  matroska->skip_to_keyframe = 0;
3686  else if (!st->internal->skip_to_keyframe) {
3687  av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3688  matroska->skip_to_keyframe = 0;
3689  }
3690  }
3691 
3692  res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3693  &pb, lace_size, &laces);
3694  if (res < 0) {
3695  av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3696  return res;
3697  }
3698 
3699  trust_default_duration = track->default_duration != 0;
3700  if (track->audio.samplerate == 8000 && trust_default_duration) {
3701  // If this is needed for more codecs, then add them here
3702  if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3703  if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3704  trust_default_duration = 0;
3705  }
3706  }
3707 
3708  if (!block_duration && trust_default_duration)
3709  block_duration = track->default_duration * laces / matroska->time_scale;
3710 
3711  if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3712  track->end_timecode =
3713  FFMAX(track->end_timecode, timecode + block_duration);
3714 
3715  for (n = 0; n < laces; n++) {
3716  int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3717  uint8_t *out_data = data;
3718  int out_size = lace_size[n];
3719 
3720  if (track->needs_decoding) {
3721  res = matroska_decode_buffer(&out_data, &out_size, track);
3722  if (res < 0)
3723  return res;
3724  /* Given that we are here means that out_data is no longer
3725  * owned by buf, so set it to NULL. This depends upon
3726  * zero-length header removal compression being ignored. */
3727  av_assert1(out_data != data);
3728  buf = NULL;
3729  }
3730 
3731  if (track->audio.buf) {
3732  res = matroska_parse_rm_audio(matroska, track, st,
3733  out_data, out_size,
3734  timecode, pos);
3735  if (!buf)
3736  av_free(out_data);
3737  if (res)
3738  return res;
3739  } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3740  res = matroska_parse_webvtt(matroska, track, st,
3741  out_data, out_size,
3742  timecode, lace_duration,
3743  pos);
3744  if (!buf)
3745  av_free(out_data);
3746  if (res)
3747  return res;
3748  } else {
3749  res = matroska_parse_frame(matroska, track, st, buf, out_data,
3750  out_size, timecode, lace_duration,
3751  pos, !n ? is_keyframe : 0,
3752  additional, additional_id, additional_size,
3753  discard_padding);
3754  if (res)
3755  return res;
3756  }
3757 
3758  if (timecode != AV_NOPTS_VALUE)
3759  timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3760  data += lace_size[n];
3761  }
3762 
3763  return 0;
3764 }
3765 
3766 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3767 {
3768  MatroskaCluster *cluster = &matroska->current_cluster;
3769  MatroskaBlock *block = &cluster->block;
3770  int res;
3771 
3772  av_assert0(matroska->num_levels <= 2);
3773 
3774  if (matroska->num_levels == 1) {
3775  res = ebml_parse(matroska, matroska_segment, NULL);
3776 
3777  if (res == 1) {
3778  /* Found a cluster: subtract the size of the ID already read. */
3779  cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3780 
3781  res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3782  if (res < 0)
3783  return res;
3784  }
3785  }
3786 
3787  if (matroska->num_levels == 2) {
3788  /* We are inside a cluster. */
3789  res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3790 
3791  if (res >= 0 && block->bin.size > 0) {
3792  int is_keyframe = block->non_simple ? block->reference.count == 0 : -1;
3793  uint8_t* additional = block->additional.size > 0 ?
3794  block->additional.data : NULL;
3795 
3796  res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3797  block->bin.size, block->bin.pos,
3798  cluster->timecode, block->duration,
3799  is_keyframe, additional, block->additional_id,
3800  block->additional.size, cluster->pos,
3801  block->discard_padding);
3802  }
3803 
3805  memset(block, 0, sizeof(*block));
3806  } else if (!matroska->num_levels) {
3807  if (!avio_feof(matroska->ctx->pb)) {
3808  avio_r8(matroska->ctx->pb);
3809  if (!avio_feof(matroska->ctx->pb)) {
3810  av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3811  "end of segment.\n");
3812  return AVERROR_INVALIDDATA;
3813  }
3814  }
3815  matroska->done = 1;
3816  return AVERROR_EOF;
3817  }
3818 
3819  return res;
3820 }
3821 
3823 {
3824  MatroskaDemuxContext *matroska = s->priv_data;
3825  int ret = 0;
3826 
3827  if (matroska->resync_pos == -1) {
3828  // This can only happen if generic seeking has been used.
3829  matroska->resync_pos = avio_tell(s->pb);
3830  }
3831 
3832  while (matroska_deliver_packet(matroska, pkt)) {
3833  if (matroska->done)
3834  return (ret < 0) ? ret : AVERROR_EOF;
3835  if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3836  ret = matroska_resync(matroska, matroska->resync_pos);
3837  }
3838 
3839  return 0;
3840 }
3841 
3842 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3843  int64_t timestamp, int flags)
3844 {
3845  MatroskaDemuxContext *matroska = s->priv_data;
3846  MatroskaTrack *tracks = NULL;
3847  AVStream *st = s->streams[stream_index];
3848  int i, index;
3849 
3850  /* Parse the CUES now since we need the index data to seek. */
3851  if (matroska->cues_parsing_deferred > 0) {
3852  matroska->cues_parsing_deferred = 0;
3853  matroska_parse_cues(matroska);
3854  }
3855 
3856  if (!st->nb_index_entries)
3857  goto err;
3858  timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3859 
3860  if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3861  matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3862  while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3863  matroska_clear_queue(matroska);
3864  if (matroska_parse_cluster(matroska) < 0)
3865  break;
3866  }
3867  }
3868 
3869  matroska_clear_queue(matroska);
3870  if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3871  goto err;
3872 
3873  tracks = matroska->tracks.elem;
3874  for (i = 0; i < matroska->tracks.nb_elem; i++) {
3875  tracks[i].audio.pkt_cnt = 0;
3876  tracks[i].audio.sub_packet_cnt = 0;
3877  tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3878  tracks[i].end_timecode = 0;
3879  }
3880 
3881  /* We seek to a level 1 element, so set the appropriate status. */
3882  matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3883  if (flags & AVSEEK_FLAG_ANY) {
3884  st->internal->skip_to_keyframe = 0;
3885  matroska->skip_to_timecode = timestamp;
3886  } else {
3887  st->internal->skip_to_keyframe = 1;
3888  matroska->skip_to_timecode = st->index_entries[index].timestamp;
3889  }
3890  matroska->skip_to_keyframe = 1;
3891  matroska->done = 0;
3893  return 0;
3894 err:
3895  // slightly hackish but allows proper fallback to
3896  // the generic seeking code.
3897  matroska_reset_status(matroska, 0, -1);
3898  matroska->resync_pos = -1;
3899  matroska_clear_queue(matroska);
3901  matroska->skip_to_keyframe = 0;
3902  matroska->done = 0;
3903  return -1;
3904 }
3905 
3907 {
3908  MatroskaDemuxContext *matroska = s->priv_data;
3909  MatroskaTrack *tracks = matroska->tracks.elem;
3910  int n;
3911 
3912  matroska_clear_queue(matroska);
3913  av_packet_free(&matroska->pkt);
3914 
3915  for (n = 0; n < matroska->tracks.nb_elem; n++)
3916  if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3917  av_freep(&tracks[n].audio.buf);
3920  return 0;
3922 
3923 typedef struct {
3924  int64_t start_time_ns;
3925  int64_t end_time_ns;
3926  int64_t start_offset;
3927  int64_t end_offset;
3929 
3930 /* This function searches all the Cues and returns the CueDesc corresponding to
3931  * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3932  * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3933  */
3934 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3935  MatroskaDemuxContext *matroska = s->priv_data;
3936  CueDesc cue_desc;
3937  int i;
3938  int nb_index_entries = s->streams[0]->nb_index_entries;
3939  AVIndexEntry *index_entries = s->streams[0]->index_entries;
3940 
3941  if (ts >= (int64_t)(matroska->duration * matroska->time_scale))
3942  return (CueDesc) {-1, -1, -1, -1};
3943  for (i = 1; i < nb_index_entries; i++) {
3944  if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3945  index_entries[i].timestamp * matroska->time_scale > ts) {
3946  break;
3947  }
3948  }
3949  --i;
3950  cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3951  cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3952  if (i != nb_index_entries - 1) {
3953  cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3954  cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3955  } else {
3956  cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3957  // FIXME: this needs special handling for files where Cues appear
3958  // before Clusters. the current logic assumes Cues appear after
3959  // Clusters.
3960  cue_desc.end_offset = cues_start - matroska->segment_start;
3961  }
3962  return cue_desc;
3963 }
3964 
3966 {
3967  MatroskaDemuxContext *matroska = s->priv_data;
3968  uint32_t id = matroska->current_id;
3969  int64_t cluster_pos, before_pos;
3970  int index, rv = 1;
3971  if (s->streams[0]->nb_index_entries <= 0) return 0;
3972  // seek to the first cluster using cues.
3973  index = av_index_search_timestamp(s->streams[0], 0, 0);
3974  if (index < 0) return 0;
3975  cluster_pos = s->streams[0]->index_entries[index].pos;
3976  before_pos = avio_tell(s->pb);
3977  while (1) {
3978  uint64_t cluster_id, cluster_length;
3979  int read;
3980  AVPacket *pkt;
3981  avio_seek(s->pb, cluster_pos, SEEK_SET);
3982  // read cluster id and length
3983  read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3984  if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3985  break;
3986  read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3987  if (read < 0)
3988  break;
3989 
3990  matroska_reset_status(matroska, 0, cluster_pos);
3991  matroska_clear_queue(matroska);
3992  if (matroska_parse_cluster(matroska) < 0 ||
3993  !matroska->queue) {
3994  break;
3995  }
3996  pkt = &matroska->queue->pkt;
3997  // 4 + read is the length of the cluster id and the cluster length field.
3998  cluster_pos += 4 + read + cluster_length;
3999  if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
4000  rv = 0;
4001  break;
4002  }
4003  }
4004 
4005  /* Restore the status after matroska_read_header: */
4006  matroska_reset_status(matroska, id, before_pos);
4007 
4008  return rv;
4009 }
4010 
4011 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
4012  double min_buffer, double* buffer,
4013  double* sec_to_download, AVFormatContext *s,
4014  int64_t cues_start)
4015 {
4016  double nano_seconds_per_second = 1000000000.0;
4017  double time_sec = time_ns / nano_seconds_per_second;
4018  int rv = 0;
4019  int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
4020  int64_t end_time_ns = time_ns + time_to_search_ns;
4021  double sec_downloaded = 0.0;
4022  CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
4023  if (desc_curr.start_time_ns == -1)
4024  return -1;
4025  *sec_to_download = 0.0;
4026 
4027  // Check for non cue start time.
4028  if (time_ns > desc_curr.start_time_ns) {
4029  int64_t cue_nano = desc_curr.end_time_ns - time_ns;
4030  double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
4031  double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
4032  double timeToDownload = (cueBytes * 8.0) / bps;
4033 
4034  sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
4035  *sec_to_download += timeToDownload;
4036 
4037  // Check if the search ends within the first cue.
4038  if (desc_curr.end_time_ns >= end_time_ns) {
4039  double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4040  double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4041  sec_downloaded = percent_to_sub * sec_downloaded;
4042  *sec_to_download = percent_to_sub * *sec_to_download;
4043  }
4044 
4045  if ((sec_downloaded + *buffer) <= min_buffer) {
4046  return 1;
4047  }
4048 
4049  // Get the next Cue.
4050  desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4051  }
4052 
4053  while (desc_curr.start_time_ns != -1) {
4054  int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
4055  int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
4056  double desc_sec = desc_ns / nano_seconds_per_second;
4057  double bits = (desc_bytes * 8.0);
4058  double time_to_download = bits / bps;
4059 
4060  sec_downloaded += desc_sec - time_to_download;
4061  *sec_to_download += time_to_download;
4062 
4063  if (desc_curr.end_time_ns >= end_time_ns) {
4064  double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4065  double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4066  sec_downloaded = percent_to_sub * sec_downloaded;
4067  *sec_to_download = percent_to_sub * *sec_to_download;
4068 
4069  if ((sec_downloaded + *buffer) <= min_buffer)
4070  rv = 1;
4071  break;
4072  }
4073 
4074  if ((sec_downloaded + *buffer) <= min_buffer) {
4075  rv = 1;
4076  break;
4077  }
4078 
4079  desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4080  }
4081  *buffer = *buffer + sec_downloaded;
4082  return rv;
4083 }
4084 
4085 /* This function computes the bandwidth of the WebM file with the help of
4086  * buffer_size_after_time_downloaded() function. Both of these functions are
4087  * adapted from WebM Tools project and are adapted to work with FFmpeg's
4088  * Matroska parsing mechanism.
4089  *
4090  * Returns the bandwidth of the file on success; -1 on error.
4091  * */
4092 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4093 {
4094  MatroskaDemuxContext *matroska = s->priv_data;
4095  AVStream *st = s->streams[0];
4096  double bandwidth = 0.0;
4097  int i;
4098 
4099  for (i = 0; i < st->nb_index_entries; i++) {
4100  int64_t prebuffer_ns = 1000000000;
4101  int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
4102  double nano_seconds_per_second = 1000000000.0;
4103  int64_t prebuffered_ns = time_ns + prebuffer_ns;
4104  double prebuffer_bytes = 0.0;
4105  int64_t temp_prebuffer_ns = prebuffer_ns;
4106  int64_t pre_bytes, pre_ns;
4107  double pre_sec, prebuffer, bits_per_second;
4108  CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4109 
4110  // Start with the first Cue.
4111  CueDesc desc_end = desc_beg;
4112 
4113  // Figure out how much data we have downloaded for the prebuffer. This will
4114  // be used later to adjust the bits per sample to try.
4115  while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4116  // Prebuffered the entire Cue.
4117  prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4118  temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4119  desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4120  }
4121  if (desc_end.start_time_ns == -1) {
4122  // The prebuffer is larger than the duration.
4123  if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4124  return -1;
4125  bits_per_second = 0.0;
4126  } else {
4127  // The prebuffer ends in the last Cue. Estimate how much data was
4128  // prebuffered.
4129  pre_bytes = desc_end.end_offset - desc_end.start_offset;
4130  pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4131  if (pre_ns <= 0)
4132  return -1;
4133  pre_sec = pre_ns / nano_seconds_per_second;
4134  prebuffer_bytes +=
4135  pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4136 
4137  prebuffer = prebuffer_ns / nano_seconds_per_second;
4138 
4139  // Set this to 0.0 in case our prebuffer buffers the entire video.
4140  bits_per_second = 0.0;
4141  do {
4142  int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4143  int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4144  double desc_sec, calc_bits_per_second, percent, mod_bits_per_second;
4145  if (desc_bytes <= 0)
4146  return -1;
4147 
4148  desc_sec = desc_ns / nano_seconds_per_second;
4149  calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4150 
4151  // Drop the bps by the percentage of bytes buffered.
4152  percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4153  mod_bits_per_second = calc_bits_per_second * percent;
4154 
4155  if (prebuffer < desc_sec) {
4156  double search_sec =
4157  (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4158 
4159  // Add 1 so the bits per second should be a little bit greater than file
4160  // datarate.
4161  int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4162  const double min_buffer = 0.0;
4163  double buffer = prebuffer;
4164  double sec_to_download = 0.0;
4165 
4166  int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4167  min_buffer, &buffer, &sec_to_download,
4168  s, cues_start);
4169  if (rv < 0) {
4170  return -1;
4171  } else if (rv == 0) {
4172  bits_per_second = (double)(bps);
4173  break;
4174  }
4175  }
4176 
4177  desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4178  } while (desc_end.start_time_ns != -1);
4179  }
4180  if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4181  }
4182  return (int64_t)bandwidth;
4183 }
4184 
4185 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4186 {
4187  MatroskaDemuxContext *matroska = s->priv_data;
4188  EbmlList *seekhead_list = &matroska->seekhead;
4189  MatroskaSeekhead *seekhead = seekhead_list->elem;
4190  char *buf;
4191  int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4192  int i;
4193  int end = 0;
4194 
4195  // determine cues start and end positions
4196  for (i = 0; i < seekhead_list->nb_elem; i++)
4197  if (seekhead[i].id == MATROSKA_ID_CUES)
4198  break;
4199 
4200  if (i >= seekhead_list->nb_elem) return -1;
4201 
4202  before_pos = avio_tell(matroska->ctx->pb);
4203  cues_start = seekhead[i].pos + matroska->segment_start;
4204  if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4205  // cues_end is computed as cues_start + cues_length + length of the
4206  // Cues element ID (i.e. 4) + EBML length of the Cues element.
4207  // cues_end is inclusive and the above sum is reduced by 1.
4208  uint64_t cues_length, cues_id;
4209  int bytes_read;
4210  bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4211  if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4212  return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4213  bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4214  if (bytes_read < 0)
4215  return bytes_read;
4216  cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4217  }
4218  avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4219  if (cues_start == -1 || cues_end == -1) return -1;
4220 
4221  // parse the cues
4222  matroska_parse_cues(matroska);
4223 
4224  // cues start
4225  av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4226 
4227  // cues end
4228  av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4229 
4230  // if the file has cues at the start, fix up the init range so that
4231  // it does not include it
4232  if (cues_start <= init_range)
4233  av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4234 
4235  // bandwidth
4236  bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4237  if (bandwidth < 0) return -1;
4238  av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4239 
4240  // check if all clusters start with key frames
4242 
4243  // store cue point timestamps as a comma separated list for checking subsegment alignment in
4244  // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4245  buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4246  if (!buf) return -1;
4247  strcpy(buf, "");
4248  for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4249  int ret = snprintf(buf + end, 20,
4250  "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4251  i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4252  if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4253  av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4254  av_free(buf);
4255  return AVERROR_INVALIDDATA;
4256  }
4257  end += ret;
4258  }
4259  av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4261 
4262  return 0;
4263 }
4264 
4266 {
4267  char *buf;
4268  int ret = matroska_read_header(s);
4269  int64_t init_range;
4270  MatroskaTrack *tracks;
4271  MatroskaDemuxContext *matroska = s->priv_data;
4272  if (ret) {
4273  av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4274  return -1;
4275  }
4276  if (!matroska->tracks.nb_elem || !s->nb_streams) {
4277  av_log(s, AV_LOG_ERROR, "No track found\n");
4279  goto fail;
4280  }
4281 
4282  if (!matroska->is_live) {
4283  buf = av_asprintf("%g", matroska->duration);
4284  if (!buf) {
4285  ret = AVERROR(ENOMEM);
4286  goto fail;
4287  }
4288  av_dict_set(&s->streams[0]->metadata, DURATION,
4290 
4291  // initialization range
4292  // 5 is the offset of Cluster ID.
4293  init_range = avio_tell(s->pb) - 5;
4294  av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4295  }
4296 
4297  // basename of the file
4298  buf = strrchr(s->url, '/');
4299  av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4300 
4301  // track number
4302  tracks = matroska->tracks.elem;
4303  av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4304 
4305  // parse the cues and populate Cue related fields
4306  if (!matroska->is_live) {
4307  ret = webm_dash_manifest_cues(s, init_range);
4308  if (ret < 0) {
4309  av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4310  goto fail;
4311  }
4312  }
4313 
4314  // use the bandwidth from the command line if it was provided
4315  if (matroska->bandwidth > 0) {
4316  av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4317  matroska->bandwidth, 0);
4318  }
4319  return 0;
4320 fail:
4322  return ret;
4323 }
4326 {
4327  return AVERROR_EOF;
4328 }
4329 
4330 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4331 static const AVOption options[] = {
4332  { "live", "flag indicating that the input is a live file that only has the headers.", OFFSET(is_live), AV_OPT_TYPE_BOOL, {.i64 = 0}, 0, 1, AV_OPT_FLAG_DECODING_PARAM },
4333  { "bandwidth", "bandwidth of this stream to be specified in the DASH manifest.", OFFSET(bandwidth), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, AV_OPT_FLAG_DECODING_PARAM },
4334  { NULL },
4335 };
4336 
4337 static const AVClass webm_dash_class = {
4338  .class_name = "WebM DASH Manifest demuxer",
4339  .item_name = av_default_item_name,
4340  .option = options,
4341  .version = LIBAVUTIL_VERSION_INT,
4342 };
4343 
4345  .name = "matroska,webm",
4346  .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4347  .extensions = "mkv,mk3d,mka,mks,webm",
4348  .priv_data_size = sizeof(MatroskaDemuxContext),
4354  .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matros