projects / ffmpeg.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
history | raw | HEAD
avfilter/avfiltergraph: fix constant string comparision
[ffmpeg.git] / libavformat / matroskadec.c
1 /*
2 * Matroska file demuxer
3 * Copyright (c) 2003-2008 The FFmpeg Project
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 /**
23 * @file
24 * Matroska file demuxer
25 * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26 * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27 * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28 * @see specs available on the Matroska project page: http://www.matroska.org/
29 */
30
31 #include "config.h"
32 #include "config_components.h"
33
34 #include <inttypes.h>
35 #include <stdio.h>
36
37 #include "libavutil/avstring.h"
38 #include "libavutil/base64.h"
39 #include "libavutil/bprint.h"
40 #include "libavutil/dict.h"
41 #include "libavutil/display.h"
42 #include "libavutil/hdr_dynamic_metadata.h"
43 #include "libavutil/intfloat.h"
44 #include "libavutil/intreadwrite.h"
45 #include "libavutil/lzo.h"
46 #include "libavutil/mastering_display_metadata.h"
47 #include "libavutil/mathematics.h"
48 #include "libavutil/mem.h"
49 #include "libavutil/opt.h"
50 #include "libavutil/pixdesc.h"
51 #include "libavutil/time_internal.h"
52 #include "libavutil/spherical.h"
53
54 #include "libavcodec/bytestream.h"
55 #include "libavcodec/defs.h"
56 #include "libavcodec/flac.h"
57 #include "libavcodec/itut35.h"
58 #include "libavcodec/mpeg4audio.h"
59 #include "libavcodec/packet_internal.h"
60
61 #include "avformat.h"
62 #include "avio_internal.h"
63 #include "demux.h"
64 #include "dovi_isom.h"
65 #include "internal.h"
66 #include "isom.h"
67 #include "matroska.h"
68 #include "oggdec.h"
69 /* For ff_codec_get_id(). */
70 #include "riff.h"
71 #include "rmsipr.h"
72
73 #if CONFIG_BZLIB
74 #include <bzlib.h>
75 #endif
76 #if CONFIG_ZLIB
77 #include <zlib.h>
78 #endif
79
80 #include "qtpalette.h"
81
82 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
83 #define NEEDS_CHECKING 2 /* Indicates that some error checks
84 * still need to be performed */
85 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
86 * syntax level used for parsing ended. */
87 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
88 * of unknown, potentially damaged data is encountered,
89 * it is considered an error. */
90 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
91 * to this many bytes of unknown data for the
92 * SKIP_THRESHOLD check. */
93
94 typedef enum {
95 EBML_NONE,
96 EBML_UINT,
97 EBML_SINT,
98 EBML_FLOAT,
99 EBML_STR,
100 EBML_UTF8,
101 EBML_BIN,
102 EBML_NEST,
103 EBML_LEVEL1,
104 EBML_STOP,
105 EBML_TYPE_COUNT
106 } EbmlType;
107
108 typedef struct CountedElement {
109 union {
110 uint64_t u;
111 int64_t i;
112 double f;
113 char *s;
114 } el;
115 unsigned count;
116 } CountedElement;
117
118 typedef const struct EbmlSyntax {
119 uint32_t id;
120 uint8_t type;
121 uint8_t is_counted;
122 size_t list_elem_size;
123 size_t data_offset;
124 union {
125 int64_t i;
126 uint64_t u;
127 double f;
128 const char *s;
129 const struct EbmlSyntax *n;
130 } def;
131 } EbmlSyntax;
132
133 typedef struct EbmlList {
134 int nb_elem;
135 unsigned int alloc_elem_size;
136 void *elem;
137 } EbmlList;
138
139 typedef struct EbmlBin {
140 int size;
141 AVBufferRef *buf;
142 uint8_t *data;
143 int64_t pos;
144 } EbmlBin;
145
146 typedef struct Ebml {
147 uint64_t version;
148 uint64_t max_size;
149 uint64_t id_length;
150 char *doctype;
151 uint64_t doctype_version;
152 } Ebml;
153
154 typedef struct MatroskaTrackCompression {
155 uint64_t algo;
156 EbmlBin settings;
157 } MatroskaTrackCompression;
158
159 typedef struct MatroskaTrackEncryption {
160 uint64_t algo;
161 EbmlBin key_id;
162 } MatroskaTrackEncryption;
163
164 typedef struct MatroskaTrackEncoding {
165 uint64_t scope;
166 uint64_t type;
167 MatroskaTrackCompression compression;
168 MatroskaTrackEncryption encryption;
169 } MatroskaTrackEncoding;
170
171 typedef struct MatroskaMasteringMeta {
172 double r_x;
173 double r_y;
174 double g_x;
175 double g_y;
176 double b_x;
177 double b_y;
178 double white_x;
179 double white_y;
180 double max_luminance;
181 CountedElement min_luminance;
182 } MatroskaMasteringMeta;
183
184 typedef struct MatroskaTrackVideoColor {
185 uint64_t matrix_coefficients;
186 uint64_t bits_per_channel;
187 uint64_t chroma_sub_horz;
188 uint64_t chroma_sub_vert;
189 uint64_t cb_sub_horz;
190 uint64_t cb_sub_vert;
191 uint64_t chroma_siting_horz;
192 uint64_t chroma_siting_vert;
193 uint64_t range;
194 uint64_t transfer_characteristics;
195 uint64_t primaries;
196 uint64_t max_cll;
197 uint64_t max_fall;
198 MatroskaMasteringMeta mastering_meta;
199 } MatroskaTrackVideoColor;
200
201 typedef struct MatroskaTrackVideoProjection {
202 uint64_t type;
203 EbmlBin private;
204 double yaw;
205 double pitch;
206 double roll;
207 } MatroskaTrackVideoProjection;
208
209 typedef struct MatroskaTrackVideo {
210 double frame_rate;
211 uint64_t display_width;
212 uint64_t display_height;
213 uint64_t pixel_width;
214 uint64_t pixel_height;
215 uint64_t cropped_width;
216 uint64_t cropped_height;
217 EbmlBin color_space;
218 uint64_t pixel_cropt;
219 uint64_t pixel_cropl;
220 uint64_t pixel_cropb;
221 uint64_t pixel_cropr;
222 uint64_t display_unit;
223 uint64_t interlaced;
224 uint64_t field_order;
225 uint64_t stereo_mode;
226 uint64_t alpha_mode;
227 EbmlList color;
228 MatroskaTrackVideoProjection projection;
229 } MatroskaTrackVideo;
230
231 typedef struct MatroskaTrackAudio {
232 double samplerate;
233 double out_samplerate;
234 uint64_t bitdepth;
235 uint64_t channels;
236
237 /* real audio header (extracted from extradata) */
238 int coded_framesize;
239 int sub_packet_h;
240 int frame_size;
241 int sub_packet_size;
242 int sub_packet_cnt;
243 int pkt_cnt;
244 uint64_t buf_timecode;
245 uint8_t *buf;
246 } MatroskaTrackAudio;
247
248 typedef struct MatroskaTrackPlane {
249 uint64_t uid;
250 uint64_t type;
251 } MatroskaTrackPlane;
252
253 typedef struct MatroskaTrackOperation {
254 EbmlList combine_planes;
255 } MatroskaTrackOperation;
256
257 typedef struct MatroskaBlockAdditionMapping {
258 uint64_t value;
259 char *name;
260 uint64_t type;
261 EbmlBin extradata;
262 } MatroskaBlockAdditionMapping;
263
264 typedef struct MatroskaTrack {
265 uint64_t num;
266 uint64_t uid;
267 uint64_t type;
268 char *name;
269 char *codec_id;
270 EbmlBin codec_priv;
271 char *language;
272 double time_scale;
273 uint64_t default_duration;
274 uint64_t flag_default;
275 uint64_t flag_forced;
276 uint64_t flag_comment;
277 uint64_t flag_hearingimpaired;
278 uint64_t flag_visualimpaired;
279 uint64_t flag_textdescriptions;
280 CountedElement flag_original;
281 uint64_t seek_preroll;
282 MatroskaTrackVideo video;
283 MatroskaTrackAudio audio;
284 MatroskaTrackOperation operation;
285 EbmlList encodings;
286 uint64_t codec_delay;
287 uint64_t codec_delay_in_track_tb;
288
289 AVStream *stream;
290 int64_t end_timecode;
291 int ms_compat;
292 int needs_decoding;
293 uint64_t max_block_additional_id;
294 EbmlList block_addition_mappings;
295
296 uint32_t palette[AVPALETTE_COUNT];
297 int has_palette;
298 } MatroskaTrack;
299
300 typedef struct MatroskaAttachment {
301 uint64_t uid;
302 char *filename;
303 char *description;
304 char *mime;
305 EbmlBin bin;
306
307 AVStream *stream;
308 } MatroskaAttachment;
309
310 typedef struct MatroskaChapter {
311 uint64_t start;
312 uint64_t end;
313 uint64_t uid;
314 char *title;
315
316 AVChapter *chapter;
317 } MatroskaChapter;
318
319 typedef struct MatroskaIndexPos {
320 uint64_t track;
321 uint64_t pos;
322 } MatroskaIndexPos;
323
324 typedef struct MatroskaIndex {
325 uint64_t time;
326 EbmlList pos;
327 } MatroskaIndex;
328
329 typedef struct MatroskaTag {
330 char *name;
331 char *string;
332 char *lang;
333 uint64_t def;
334 EbmlList sub;
335 } MatroskaTag;
336
337 typedef struct MatroskaTagTarget {
338 char *type;
339 uint64_t typevalue;
340 uint64_t trackuid;
341 uint64_t chapteruid;
342 uint64_t attachuid;
343 } MatroskaTagTarget;
344
345 typedef struct MatroskaTags {
346 MatroskaTagTarget target;
347 EbmlList tag;
348 } MatroskaTags;
349
350 typedef struct MatroskaSeekhead {
351 uint64_t id;
352 uint64_t pos;
353 } MatroskaSeekhead;
354
355 typedef struct MatroskaLevel {
356 uint64_t start;
357 uint64_t length;
358 } MatroskaLevel;
359
360 typedef struct MatroskaBlockMore {
361 uint64_t additional_id;
362 EbmlBin additional;
363 } MatroskaBlockMore;
364
365 typedef struct MatroskaBlock {
366 uint64_t duration;
367 CountedElement reference;
368 uint64_t non_simple;
369 EbmlBin bin;
370 EbmlList blockmore;
371 int64_t discard_padding;
372 } MatroskaBlock;
373
374 typedef struct MatroskaCluster {
375 MatroskaBlock block;
376 uint64_t timecode;
377 int64_t pos;
378 } MatroskaCluster;
379
380 typedef struct MatroskaLevel1Element {
381 int64_t pos;
382 uint32_t id;
383 int parsed;
384 } MatroskaLevel1Element;
385
386 typedef struct MatroskaDemuxContext {
387 const AVClass *class;
388 AVFormatContext *ctx;
389
390 /* EBML stuff */
391 MatroskaLevel levels[EBML_MAX_DEPTH];
392 int num_levels;
393 uint32_t current_id;
394 int64_t resync_pos;
395 int unknown_count;
396
397 uint64_t time_scale;
398 double duration;
399 char *title;
400 char *muxingapp;
401 EbmlBin date_utc;
402 EbmlList tracks;
403 EbmlList attachments;
404 EbmlList chapters;
405 EbmlList index;
406 EbmlList tags;
407 EbmlList seekhead;
408
409 /* byte position of the segment inside the stream */
410 int64_t segment_start;
411
412 /* This packet coincides with FFFormatContext.parse_pkt
413 * and is not owned by us. */
414 AVPacket *pkt;
415
416 /* the packet queue */
417 PacketList queue;
418
419 int done;
420
421 /* What to skip before effectively reading a packet. */
422 int skip_to_keyframe;
423 uint64_t skip_to_timecode;
424
425 /* File has a CUES element, but we defer parsing until it is needed. */
426 int cues_parsing_deferred;
427
428 /* Level1 elements and whether they were read yet */
429 MatroskaLevel1Element level1_elems[64];
430 int num_level1_elems;
431
432 MatroskaCluster current_cluster;
433
434 int is_webm;
435
436 /* WebM DASH Manifest live flag */
437 int is_live;
438
439 /* Bandwidth value for WebM DASH Manifest */
440 int bandwidth;
441 } MatroskaDemuxContext;
442
443 #define CHILD_OF(parent) { .def = { .n = parent } }
444
445 // The following forward declarations need their size because
446 // a tentative definition with internal linkage must not be an
447 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
448 // Removing the sizes breaks MSVC.
449 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
450 matroska_track[33], matroska_track_encoding[6], matroska_track_encodings[2],
451 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_block_addition_mapping[5], matroska_tracks[2],
452 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
453 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
454 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
455
456 static EbmlSyntax ebml_header[] = {
457 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
458 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, 0, offsetof(Ebml, max_size), { .u = 8 } },
459 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, 0, offsetof(Ebml, id_length), { .u = 4 } },
460 { EBML_ID_DOCTYPE, EBML_STR, 0, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
461 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
462 { EBML_ID_EBMLVERSION, EBML_NONE },
463 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
464 CHILD_OF(ebml_syntax)
465 };
466
467 static EbmlSyntax ebml_syntax[] = {
468 { EBML_ID_HEADER, EBML_NEST, 0, 0, 0, { .n = ebml_header } },
469 { MATROSKA_ID_SEGMENT, EBML_STOP },
470 { 0 }
471 };
472
473 static EbmlSyntax matroska_info[] = {
474 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
475 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, 0, offsetof(MatroskaDemuxContext, duration) },
476 { MATROSKA_ID_TITLE, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, title) },
477 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
478 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, muxingapp) },
479 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, 0, offsetof(MatroskaDemuxContext, date_utc) },
480 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
481 CHILD_OF(matroska_segment)
482 };
483
484 static EbmlSyntax matroska_mastering_meta[] = {
485 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_x) },
486 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_y) },
487 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_x) },
488 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_y) },
489 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_x) },
490 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_y) },
491 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_x) },
492 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_y) },
493 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 1, 0, offsetof(MatroskaMasteringMeta, min_luminance) },
494 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, max_luminance) },
495 CHILD_OF(matroska_track_video_color)
496 };
497
498 static EbmlSyntax matroska_track_video_color[] = {
499 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
500 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u = 0 } },
501 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz) },
502 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert) },
503 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz) },
504 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert) },
505 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
506 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
507 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
508 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
509 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
510 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_cll) },
511 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_fall) },
512 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
513 CHILD_OF(matroska_track_video)
514 };
515
516 static EbmlSyntax matroska_track_video_projection[] = {
517 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
518 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideoProjection, private) },
519 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f = 0.0 } },
520 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f = 0.0 } },
521 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f = 0.0 } },
522 CHILD_OF(matroska_track_video)
523 };
524
525 static EbmlSyntax matroska_track_video[] = {
526 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideo, frame_rate) },
527 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
528 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
529 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_width) },
530 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_height) },
531 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideo, color_space) },
532 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, alpha_mode), { .u = 0 } },
533 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, 0, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
534 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
535 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_cropb), {.u = 0 } },
536 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_cropt), {.u = 0 } },
537 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_cropl), {.u = 0 } },
538 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_cropr), {.u = 0 } },
539 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
540 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
541 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
542 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
543 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
544 CHILD_OF(matroska_track)
545 };
546
547 static EbmlSyntax matroska_track_audio[] = {
548 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
549 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
550 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, bitdepth) },
551 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
552 CHILD_OF(matroska_track)
553 };
554
555 static EbmlSyntax matroska_track_encoding_compression[] = {
556 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackCompression, algo), { .u = MATROSKA_TRACK_ENCODING_COMP_ZLIB } },
557 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, 0, offsetof(MatroskaTrackCompression, settings) },
558 CHILD_OF(matroska_track_encoding)
559 };
560
561 static EbmlSyntax matroska_track_encoding_encryption[] = {
562 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
563 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, 0, offsetof(MatroskaTrackEncryption,key_id) },
564 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
565 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
566 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
567 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
568 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
569 CHILD_OF(matroska_track_encoding)
570 };
571 static EbmlSyntax matroska_track_encoding[] = {
572 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
573 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
574 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
575 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
576 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
577 CHILD_OF(matroska_track_encodings)
578 };
579
580 static EbmlSyntax matroska_track_encodings[] = {
581 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, 0, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
582 CHILD_OF(matroska_track)
583 };
584
585 static EbmlSyntax matroska_track_plane[] = {
586 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,uid) },
587 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,type) },
588 CHILD_OF(matroska_track_combine_planes)
589 };
590
591 static EbmlSyntax matroska_track_combine_planes[] = {
592 { MATROSKA_ID_TRACKPLANE, EBML_NEST, 0, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
593 CHILD_OF(matroska_track_operation)
594 };
595
596 static EbmlSyntax matroska_track_operation[] = {
597 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, 0, {.n = matroska_track_combine_planes} },
598 CHILD_OF(matroska_track)
599 };
600
601 static EbmlSyntax matroska_block_addition_mapping[] = {
602 { MATROSKA_ID_BLKADDIDVALUE, EBML_UINT, 0, 0, offsetof(MatroskaBlockAdditionMapping, value) },
603 { MATROSKA_ID_BLKADDIDNAME, EBML_STR, 0, 0, offsetof(MatroskaBlockAdditionMapping, name) },
604 { MATROSKA_ID_BLKADDIDTYPE, EBML_UINT, 0, 0, offsetof(MatroskaBlockAdditionMapping, type), { .u = MATROSKA_BLOCK_ADD_ID_TYPE_DEFAULT } },
605 { MATROSKA_ID_BLKADDIDEXTRADATA, EBML_BIN, 0, 0, offsetof(MatroskaBlockAdditionMapping, extradata) },
606 CHILD_OF(matroska_track)
607 };
608
609 static EbmlSyntax matroska_track[] = {
610 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, 0, offsetof(MatroskaTrack, num) },
611 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTrack, name) },
612 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, uid) },
613 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrack, type) },
614 { MATROSKA_ID_CODECID, EBML_STR, 0, 0, offsetof(MatroskaTrack, codec_id) },
615 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrack, codec_priv) },
616 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, codec_delay), { .u = 0 } },
617 { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
618 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, 0, offsetof(MatroskaTrack, default_duration) },
619 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
620 { MATROSKA_ID_TRACKFLAGCOMMENTARY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_comment), { .u = 0 } },
621 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
622 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
623 { MATROSKA_ID_TRACKFLAGHEARINGIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_hearingimpaired), { .u = 0 } },
624 { MATROSKA_ID_TRACKFLAGVISUALIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_visualimpaired), { .u = 0 } },
625 { MATROSKA_ID_TRACKFLAGTEXTDESCRIPTIONS, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_textdescriptions), { .u = 0 } },
626 { MATROSKA_ID_TRACKFLAGORIGINAL, EBML_UINT, 1, 0, offsetof(MatroskaTrack, flag_original), {.u = 0 } },
627 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
628 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
629 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
630 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, 0, { .n = matroska_track_encodings } },
631 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, max_block_additional_id), { .u = 0 } },
632 { MATROSKA_ID_TRACKBLKADDMAPPING, EBML_NEST, 0, sizeof(MatroskaBlockAdditionMapping), offsetof(MatroskaTrack, block_addition_mappings), { .n = matroska_block_addition_mapping } },
633 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, 0, offsetof(MatroskaTrack, seek_preroll), { .u = 0 } },
634 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
635 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
636 { MATROSKA_ID_CODECNAME, EBML_NONE },
637 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
638 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
639 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
640 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
641 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
642 CHILD_OF(matroska_tracks)
643 };
644
645 static EbmlSyntax matroska_tracks[] = {
646 { MATROSKA_ID_TRACKENTRY, EBML_NEST, 0, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
647 CHILD_OF(matroska_segment)
648 };
649
650 static EbmlSyntax matroska_attachment[] = {
651 { MATROSKA_ID_FILEUID, EBML_UINT, 0, 0, offsetof(MatroskaAttachment, uid) },
652 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, filename) },
653 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, 0, offsetof(MatroskaAttachment, mime) },
654 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, 0, offsetof(MatroskaAttachment, bin) },
655 { MATROSKA_ID_FILEDESC, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, description) },
656 CHILD_OF(matroska_attachments)
657 };
658
659 static EbmlSyntax matroska_attachments[] = {
660 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, 0, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
661 CHILD_OF(matroska_segment)
662 };
663
664 static EbmlSyntax matroska_chapter_display[] = {
665 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaChapter, title) },
666 { MATROSKA_ID_CHAPLANG, EBML_NONE },
667 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
668 CHILD_OF(matroska_chapter_entry)
669 };
670
671 static EbmlSyntax matroska_chapter_entry[] = {
672 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
673 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
674 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaChapter, uid) },
675 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter_display } },
676 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
677 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
678 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
679 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
680 CHILD_OF(matroska_chapter)
681 };
682
683 static EbmlSyntax matroska_chapter[] = {
684 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, 0, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
685 { MATROSKA_ID_EDITIONUID, EBML_NONE },
686 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
687 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
688 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
689 CHILD_OF(matroska_chapters)
690 };
691
692 static EbmlSyntax matroska_chapters[] = {
693 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter } },
694 CHILD_OF(matroska_segment)
695 };
696
697 static EbmlSyntax matroska_index_pos[] = {
698 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, track) },
699 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, pos) },
700 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
701 { MATROSKA_ID_CUEDURATION, EBML_NONE },
702 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
703 CHILD_OF(matroska_index_entry)
704 };
705
706 static EbmlSyntax matroska_index_entry[] = {
707 { MATROSKA_ID_CUETIME, EBML_UINT, 0, 0, offsetof(MatroskaIndex, time) },
708 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, 0, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
709 CHILD_OF(matroska_index)
710 };
711
712 static EbmlSyntax matroska_index[] = {
713 { MATROSKA_ID_POINTENTRY, EBML_NEST, 0, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
714 CHILD_OF(matroska_segment)
715 };
716
717 static EbmlSyntax matroska_simpletag[] = {
718 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTag, name) },
719 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaTag, string) },
720 { MATROSKA_ID_TAGLANG, EBML_STR, 0, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
721 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
722 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
723 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
724 CHILD_OF(matroska_tag)
725 };
726
727 static EbmlSyntax matroska_tagtargets[] = {
728 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, 0, offsetof(MatroskaTagTarget, type) },
729 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
730 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, trackuid), { .u = 0 } },
731 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, chapteruid), { .u = 0 } },
732 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, attachuid), { .u = 0 } },
733 CHILD_OF(matroska_tag)
734 };
735
736 static EbmlSyntax matroska_tag[] = {
737 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
738 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
739 CHILD_OF(matroska_tags)
740 };
741
742 static EbmlSyntax matroska_tags[] = {
743 { MATROSKA_ID_TAG, EBML_NEST, 0, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
744 CHILD_OF(matroska_segment)
745 };
746
747 static EbmlSyntax matroska_seekhead_entry[] = {
748 { MATROSKA_ID_SEEKID, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, id) },
749 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
750 CHILD_OF(matroska_seekhead)
751 };
752
753 static EbmlSyntax matroska_seekhead[] = {
754 { MATROSKA_ID_SEEKENTRY, EBML_NEST, 0, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
755 CHILD_OF(matroska_segment)
756 };
757
758 static EbmlSyntax matroska_segment[] = {
759 { MATROSKA_ID_CLUSTER, EBML_STOP },
760 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, 0, { .n = matroska_info } },
761 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tracks } },
762 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_attachments } },
763 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_chapters } },
764 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, 0, { .n = matroska_index } },
765 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tags } },
766 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, 0, { .n = matroska_seekhead } },
767 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
768 };
769
770 static EbmlSyntax matroska_segments[] = {
771 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, 0, { .n = matroska_segment } },
772 { 0 }
773 };
774
775 static EbmlSyntax matroska_blockmore[] = {
776 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, 0, offsetof(MatroskaBlockMore,additional_id), { .u = MATROSKA_BLOCK_ADD_ID_OPAQUE } },
777 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, 0, offsetof(MatroskaBlockMore,additional) },
778 CHILD_OF(matroska_blockadditions)
779 };
780
781 static EbmlSyntax matroska_blockadditions[] = {
782 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, sizeof(MatroskaBlockMore), offsetof(MatroskaBlock, blockmore), { .n = matroska_blockmore } },
783 CHILD_OF(matroska_blockgroup)
784 };
785
786 static EbmlSyntax matroska_blockgroup[] = {
787 { MATROSKA_ID_BLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
788 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, 0, { .n = matroska_blockadditions} },
789 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, 0, offsetof(MatroskaBlock, duration) },
790 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, 0, offsetof(MatroskaBlock, discard_padding) },
791 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 1, 0, offsetof(MatroskaBlock, reference) },
792 { MATROSKA_ID_CODECSTATE, EBML_NONE },
793 { 1, EBML_UINT, 0, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
794 CHILD_OF(matroska_cluster_parsing)
795 };
796
797 // The following array contains SimpleBlock and BlockGroup twice
798 // in order to reuse the other values for matroska_cluster_enter.
799 static EbmlSyntax matroska_cluster_parsing[] = {
800 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
801 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, 0, { .n = matroska_blockgroup } },
802 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, 0, offsetof(MatroskaCluster, timecode) },
803 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
804 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
805 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
806 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
807 CHILD_OF(matroska_segment)
808 };
809
810 static EbmlSyntax matroska_cluster_enter[] = {
811 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, 0, { .n = &matroska_cluster_parsing[2] } },
812 { 0 }
813 };
814 #undef CHILD_OF
815
816 static const CodecMime mkv_image_mime_tags[] = {
817 {"image/gif" , AV_CODEC_ID_GIF},
818 {"image/jpeg" , AV_CODEC_ID_MJPEG},
819 {"image/png" , AV_CODEC_ID_PNG},
820 {"image/tiff" , AV_CODEC_ID_TIFF},
821
822 {"" , AV_CODEC_ID_NONE}
823 };
824
825 static const CodecMime mkv_mime_tags[] = {
826 {"application/x-truetype-font", AV_CODEC_ID_TTF},
827 {"application/x-font" , AV_CODEC_ID_TTF},
828 {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
829 {"binary" , AV_CODEC_ID_BIN_DATA},
830
831 {"" , AV_CODEC_ID_NONE}
832 };
833
834 static const char * const matroska_video_stereo_plane[MATROSKA_VIDEO_STEREO_PLANE_COUNT] = {
835 "left",
836 "right",
837 "background",
838 };
839
840 static const char *const matroska_doctypes[] = { "matroska", "webm" };
841
842 /*
843 * This function prepares the status for parsing of level 1 elements.
844 */
845 static int matroska_reset_status(MatroskaDemuxContext *matroska,
846 uint32_t id, int64_t position)
847 {
848 int64_t err = 0;
849 if (position >= 0) {
850 err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
851 if (err > 0)
852 err = 0;
853 } else
854 position = avio_tell(matroska->ctx->pb);
855
856 matroska->current_id = id;
857 matroska->num_levels = 1;
858 matroska->unknown_count = 0;
859 matroska->resync_pos = position;
860 if (id)
861 matroska->resync_pos -= (av_log2(id) + 7) / 8;
862
863 return err;
864 }
865
866 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
867 {
868 AVIOContext *pb = matroska->ctx->pb;
869 uint32_t id;
870
871 /* Try to seek to the last position to resync from. If this doesn't work,
872 * we resync from the earliest position available: The start of the buffer. */
873 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
874 av_log(matroska->ctx, AV_LOG_WARNING,
875 "Seek to desired resync point failed. Seeking to "
876 "earliest point available instead.\n");
877 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
878 last_pos + 1), SEEK_SET);
879 }
880
881 id = avio_rb32(pb);
882
883 // try to find a toplevel element
884 while (!avio_feof(pb)) {
885 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
886 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
887 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
888 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
889 /* Prepare the context for parsing of a level 1 element. */
890 matroska_reset_status(matroska, id, -1);
891 /* Given that we are here means that an error has occurred,
892 * so treat the segment as unknown length in order not to
893 * discard valid data that happens to be beyond the designated
894 * end of the segment. */
895 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
896 return 0;
897 }
898 id = (id << 8) | avio_r8(pb);
899 }
900
901 matroska->done = 1;
902 return pb->error ? pb->error : AVERROR_EOF;
903 }
904
905 /*
906 * Read: an "EBML number", which is defined as a variable-length
907 * array of bytes. The first byte indicates the length by giving a
908 * number of 0-bits followed by a one. The position of the first
909 * "one" bit inside the first byte indicates the length of this
910 * number.
911 * Returns: number of bytes read, < 0 on error
912 */
913 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
914 int max_size, uint64_t *number, int eof_forbidden)
915 {
916 int read, n = 1;
917 uint64_t total;
918 int64_t pos;
919
920 /* The first byte tells us the length in bytes - except when it is zero. */
921 total = avio_r8(pb);
922 if (pb->eof_reached)
923 goto err;
924
925 /* get the length of the EBML number */
926 read = 8 - ff_log2_tab[total];
927
928 if (!total || read > max_size) {
929 pos = avio_tell(pb) - 1;
930 if (!total) {
931 av_log(matroska->ctx, AV_LOG_ERROR,
932 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
933 "of an EBML number\n", pos, pos);
934 } else {
935 av_log(matroska->ctx, AV_LOG_ERROR,
936 "Length %d indicated by an EBML number's first byte 0x%02x "
937 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
938 read, (uint8_t) total, pos, pos, max_size);
939 }
940 return AVERROR_INVALIDDATA;
941 }
942
943 /* read out length */
944 total ^= 1 << ff_log2_tab[total];
945 while (n++ < read)
946 total = (total << 8) | avio_r8(pb);
947
948 if (pb->eof_reached) {
949 eof_forbidden = 1;
950 goto err;
951 }
952
953 *number = total;
954
955 return read;
956
957 err:
958 pos = avio_tell(pb);
959 if (pb->error) {
960 av_log(matroska->ctx, AV_LOG_ERROR,
961 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
962 pos, pos);
963 return pb->error;
964 }
965 if (eof_forbidden) {
966 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
967 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
968 return AVERROR(EIO);
969 }
970 return AVERROR_EOF;
971 }
972
973 /**
974 * Read a EBML length value.
975 * This needs special handling for the "unknown length" case which has multiple
976 * encodings.
977 */
978 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
979 uint64_t *number)
980 {
981 int res = ebml_read_num(matroska, pb, 8, number, 1);
982 if (res > 0 && *number + 1 == 1ULL << (7 * res))
983 *number = EBML_UNKNOWN_LENGTH;
984 return res;
985 }
986
987 /*
988 * Read the next element as an unsigned int.
989 * Returns NEEDS_CHECKING unless size == 0.
990 */
991 static int ebml_read_uint(AVIOContext *pb, int size,
992 uint64_t default_value, uint64_t *num)
993 {
994 int n = 0;
995
996 if (size == 0) {
997 *num = default_value;
998 return 0;
999 }
1000 /* big-endian ordering; build up number */
1001 *num = 0;
1002 while (n++ < size)
1003 *num = (*num << 8) | avio_r8(pb);
1004
1005 return NEEDS_CHECKING;
1006 }
1007
1008 /*
1009 * Read the next element as a signed int.
1010 * Returns NEEDS_CHECKING unless size == 0.
1011 */
1012 static int ebml_read_sint(AVIOContext *pb, int size,
1013 int64_t default_value, int64_t *num)
1014 {
1015 int n = 1;
1016
1017 if (size == 0) {
1018 *num = default_value;
1019 return 0;
1020 } else {
1021 *num = sign_extend(avio_r8(pb), 8);
1022
1023 /* big-endian ordering; build up number */
1024 while (n++ < size)
1025 *num = ((uint64_t)*num << 8) | avio_r8(pb);
1026 }
1027
1028 return NEEDS_CHECKING;
1029 }
1030
1031 /*
1032 * Read the next element as a float.
1033 * Returns 0 if size == 0, NEEDS_CHECKING or < 0 on obvious failure.
1034 */
1035 static int ebml_read_float(AVIOContext *pb, int size,
1036 double default_value, double *num)
1037 {
1038 if (size == 0) {
1039 *num = default_value;
1040 return 0;
1041 } else if (size == 4) {
1042 *num = av_int2float(avio_rb32(pb));
1043 } else if (size == 8) {
1044 *num = av_int2double(avio_rb64(pb));
1045 } else
1046 return AVERROR_INVALIDDATA;
1047
1048 return NEEDS_CHECKING;
1049 }
1050
1051 /*
1052 * Read the next element as an ASCII string.
1053 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1054 */
1055 static int ebml_read_ascii(AVIOContext *pb, int size,
1056 const char *default_value, char **str)
1057 {
1058 char *res;
1059 int ret;
1060
1061 if (size == 0 && default_value) {
1062 res = av_strdup(default_value);
1063 if (!res)
1064 return AVERROR(ENOMEM);
1065 } else {
1066 /* EBML strings are usually not 0-terminated, so we allocate one
1067 * byte more, read the string and NUL-terminate it ourselves. */
1068 if (!(res = av_malloc(size + 1)))
1069 return AVERROR(ENOMEM);
1070 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
1071 av_free(res);
1072 return ret < 0 ? ret : NEEDS_CHECKING;
1073 }
1074 (res)[size] = '\0';
1075 }
1076 av_free(*str);
1077 *str = res;
1078
1079 return 0;
1080 }
1081
1082 /*
1083 * Read the next element as binary data.
1084 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1085 */
1086 static int ebml_read_binary(AVIOContext *pb, int length,
1087 int64_t pos, EbmlBin *bin)
1088 {
1089 int ret;
1090
1091 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1092 if (ret < 0)
1093 return ret;
1094 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1095
1096 bin->data = bin->buf->data;
1097 bin->size = length;
1098 bin->pos = pos;
1099 if ((ret = avio_read(pb, bin->data, length)) != length) {
1100 av_buffer_unref(&bin->buf);
1101 bin->data = NULL;
1102 bin->size = 0;
1103 return ret < 0 ? ret : NEEDS_CHECKING;
1104 }
1105
1106 return 0;
1107 }
1108
1109 /*
1110 * Read the next element, but only the header. The contents
1111 * are supposed to be sub-elements which can be read separately.
1112 * 0 is success, < 0 is failure.
1113 */
1114 static int ebml_read_master(MatroskaDemuxContext *matroska,
1115 uint64_t length, int64_t pos)
1116 {
1117 MatroskaLevel *level;
1118
1119 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1120 av_log(matroska->ctx, AV_LOG_ERROR,
1121 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1122 return AVERROR(ENOSYS);
1123 }
1124
1125 level = &matroska->levels[matroska->num_levels++];
1126 level->start = pos;
1127 level->length = length;
1128
1129 return 0;
1130 }
1131
1132 /*
1133 * Read a signed "EBML number"
1134 * Return: number of bytes processed, < 0 on error
1135 */
1136 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1137 AVIOContext *pb, int64_t *num)
1138 {
1139 uint64_t unum;
1140 int res;
1141
1142 /* read as unsigned number first */
1143 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1144 return res;
1145
1146 /* make signed (weird way) */
1147 *num = unum - ((1LL << (7 * res - 1)) - 1);
1148
1149 return res;
1150 }
1151
1152 static int ebml_parse(MatroskaDemuxContext *matroska,
1153 EbmlSyntax *syntax, void *data);
1154
1155 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1156 {
1157 int i;
1158
1159 // Whoever touches this should be aware of the duplication
1160 // existing in matroska_cluster_parsing.
1161 for (i = 0; syntax[i].id; i++)
1162 if (id == syntax[i].id)
1163 break;
1164
1165 return &syntax[i];
1166 }
1167
1168 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1169 void *data)
1170 {
1171 int res;
1172
1173 if (data) {
1174 for (int i = 0; syntax[i].id; i++) {
1175 void *dst = (char *)data + syntax[i].data_offset;
1176 switch (syntax[i].type) {
1177 case EBML_UINT:
1178 *(uint64_t *)dst = syntax[i].def.u;
1179 break;
1180 case EBML_SINT:
1181 *(int64_t *) dst = syntax[i].def.i;
1182 break;
1183 case EBML_FLOAT:
1184 *(double *) dst = syntax[i].def.f;
1185 break;
1186 case EBML_STR:
1187 case EBML_UTF8:
1188 // the default may be NULL
1189 if (syntax[i].def.s) {
1190 *(char**)dst = av_strdup(syntax[i].def.s);
1191 if (!*(char**)dst)
1192 return AVERROR(ENOMEM);
1193 }
1194 break;
1195 }
1196 }
1197
1198 if (!matroska->levels[matroska->num_levels - 1].length) {
1199 matroska->num_levels--;
1200 return 0;
1201 }
1202 }
1203
1204 do {
1205 res = ebml_parse(matroska, syntax, data);
1206 } while (!res);
1207
1208 return res == LEVEL_ENDED ? 0 : res;
1209 }
1210
1211 static int is_ebml_id_valid(uint32_t id)
1212 {
1213 // Due to endian nonsense in Matroska, the highest byte with any bits set
1214 // will contain the leading length bit. This bit in turn identifies the
1215 // total byte length of the element by its position within the byte.
1216 unsigned int bits = av_log2(id);
1217 return id && (bits + 7) / 8 == (8 - bits % 8);
1218 }
1219
1220 /*
1221 * Allocate and return the entry for the level1 element with the given ID. If
1222 * an entry already exists, return the existing entry.
1223 */
1224 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1225 uint32_t id, int64_t pos)
1226 {
1227 int i;
1228 MatroskaLevel1Element *elem;
1229
1230 if (!is_ebml_id_valid(id))
1231 return NULL;
1232
1233 // Some files link to all clusters; useless.
1234 if (id == MATROSKA_ID_CLUSTER)
1235 return NULL;
1236
1237 // There can be multiple SeekHeads and Tags.
1238 for (i = 0; i < matroska->num_level1_elems; i++) {
1239 if (matroska->level1_elems[i].id == id) {
1240 if (matroska->level1_elems[i].pos == pos ||
1241 id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1242 return &matroska->level1_elems[i];
1243 }
1244 }
1245
1246 // Only a completely broken file would have more elements.
1247 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1248 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1249 return NULL;
1250 }
1251
1252 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1253 *elem = (MatroskaLevel1Element){.id = id};
1254
1255 return elem;
1256 }
1257
1258 static int ebml_parse(MatroskaDemuxContext *matroska,
1259 EbmlSyntax *syntax, void *data)
1260 {
1261 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1262 // Forbid unknown-length EBML_NONE elements.
1263 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1264 [EBML_UINT] = 8,
1265 [EBML_SINT] = 8,
1266 [EBML_FLOAT] = 8,
1267 // max. 16 MB for strings
1268 [EBML_STR] = 0x1000000,
1269 [EBML_UTF8] = 0x1000000,
1270 // max. 256 MB for binary data
1271 [EBML_BIN] = 0x10000000,
1272 // no limits for anything else
1273 };
1274 AVIOContext *pb = matroska->ctx->pb;
1275 uint32_t id;
1276 uint64_t length;
1277 int64_t pos = avio_tell(pb), pos_alt;
1278 int res, update_pos = 1, level_check;
1279 MatroskaLevel1Element *level1_elem;
1280 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1281
1282 if (!matroska->current_id) {
1283 uint64_t id;
1284 res = ebml_read_num(matroska, pb, 4, &id, 0);
1285 if (res < 0) {
1286 if (pb->eof_reached && res == AVERROR_EOF) {
1287 if (matroska->is_live)
1288 // in live mode, finish parsing if EOF is reached.
1289 return 1;
1290 if (level && pos == avio_tell(pb)) {
1291 if (level->length == EBML_UNKNOWN_LENGTH) {
1292 // Unknown-length levels automatically end at EOF.
1293 matroska->num_levels--;
1294 return LEVEL_ENDED;
1295 } else {
1296 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1297 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1298 }
1299 }
1300 }
1301 return res;
1302 }
1303 matroska->current_id = id | 1 << 7 * res;
1304 pos_alt = pos + res;
1305 } else {
1306 pos_alt = pos;
1307 pos -= (av_log2(matroska->current_id) + 7) / 8;
1308 }
1309
1310 id = matroska->current_id;
1311
1312 syntax = ebml_parse_id(syntax, id);
1313 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1314 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1315 // Unknown-length levels end when an element from an upper level
1316 // in the hierarchy is encountered.
1317 while (syntax->def.n) {
1318 syntax = ebml_parse_id(syntax->def.n, id);
1319 if (syntax->id) {
1320 matroska->num_levels--;
1321 return LEVEL_ENDED;
1322 }
1323 // We have not encountered a known element; syntax is a sentinel.
1324 av_assert1(syntax->type == EBML_NONE);
1325 };
1326 }
1327
1328 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1329 "%"PRId64"\n", id, pos);
1330 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1331 }
1332
1333 if (data) {
1334 data = (char *) data + syntax->data_offset;
1335 if (syntax->list_elem_size) {
1336 EbmlList *list = data;
1337 void *newelem;
1338
1339 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1340 return AVERROR(ENOMEM);
1341 newelem = av_fast_realloc(list->elem,
1342 &list->alloc_elem_size,
1343 (list->nb_elem + 1) * syntax->list_elem_size);
1344 if (!newelem)
1345 return AVERROR(ENOMEM);
1346 list->elem = newelem;
1347 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1348 memset(data, 0, syntax->list_elem_size);
1349 list->nb_elem++;
1350 }
1351 }
1352
1353 if (syntax->type != EBML_STOP) {
1354 matroska->current_id = 0;
1355 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1356 return res;
1357
1358 pos_alt += res;
1359
1360 if (matroska->num_levels > 0) {
1361 if (length != EBML_UNKNOWN_LENGTH &&
1362 level->length != EBML_UNKNOWN_LENGTH) {
1363 uint64_t elem_end = pos_alt + length,
1364 level_end = level->start + level->length;
1365
1366 if (elem_end < level_end) {
1367 level_check = 0;
1368 } else if (elem_end == level_end) {
1369 level_check = LEVEL_ENDED;
1370 } else {
1371 av_log(matroska->ctx, AV_LOG_ERROR,
1372 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1373 "containing master element ending at 0x%"PRIx64"\n",
1374 pos, elem_end, level_end);
1375 return AVERROR_INVALIDDATA;
1376 }
1377 } else if (length != EBML_UNKNOWN_LENGTH) {
1378 level_check = 0;
1379 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1380 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1381 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1382 return AVERROR_INVALIDDATA;
1383 } else {
1384 level_check = 0;
1385 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1386 || syntax->type == EBML_NEST)) {
1387 // According to the current specifications only clusters and
1388 // segments are allowed to be unknown-length. We also accept
1389 // other unknown-length master elements.
1390 av_log(matroska->ctx, AV_LOG_WARNING,
1391 "Found unknown-length element 0x%"PRIX32" other than "
1392 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1393 "parsing will nevertheless be attempted.\n", id, pos);
1394 update_pos = -1;
1395 }
1396 }
1397 } else
1398 level_check = 0;
1399
1400 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1401 if (length != EBML_UNKNOWN_LENGTH) {
1402 av_log(matroska->ctx, AV_LOG_ERROR,
1403 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1404 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1405 length, max_lengths[syntax->type], id, pos);
1406 } else if (syntax->type != EBML_NONE) {
1407 av_log(matroska->ctx, AV_LOG_ERROR,
1408 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1409 "unknown length, yet the length of an element of its "
1410 "type must be known.\n", id, pos);
1411 } else {
1412 av_log(matroska->ctx, AV_LOG_ERROR,
1413 "Found unknown-length element with ID 0x%"PRIX32" at "
1414 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1415 "available.\n", id, pos);
1416 }
1417 return AVERROR_INVALIDDATA;
1418 }
1419
1420 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1421 // Losing sync will likely manifest itself as encountering unknown
1422 // elements which are not reliably distinguishable from elements
1423 // belonging to future extensions of the format.
1424 // We use a heuristic to detect such situations: If the current
1425 // element is not expected at the current syntax level and there
1426 // were only a few unknown elements in a row, then the element is
1427 // skipped or considered defective based upon the length of the
1428 // current element (i.e. how much would be skipped); if there were
1429 // more than a few skipped elements in a row and skipping the current
1430 // element would lead us more than SKIP_THRESHOLD away from the last
1431 // known good position, then it is inferred that an error occurred.
1432 // The dependency on the number of unknown elements in a row exists
1433 // because the distance to the last known good position is
1434 // automatically big if the last parsed element was big.
1435 // In both cases, each unknown element is considered equivalent to
1436 // UNKNOWN_EQUIV of skipped bytes for the check.
1437 // The whole check is only done for non-seekable output, because
1438 // in this situation skipped data can't simply be rechecked later.
1439 // This is especially important when using unknown length elements
1440 // as the check for whether a child exceeds its containing master
1441 // element is not effective in this situation.
1442 if (update_pos) {
1443 matroska->unknown_count = 0;
1444 } else {
1445 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1446
1447 if (matroska->unknown_count > 3)
1448 dist += pos_alt - matroska->resync_pos;
1449
1450 if (dist > SKIP_THRESHOLD) {
1451 av_log(matroska->ctx, AV_LOG_ERROR,
1452 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1453 "length 0x%"PRIx64" considered as invalid data. Last "
1454 "known good position 0x%"PRIx64", %d unknown elements"
1455 " in a row\n", id, pos, length, matroska->resync_pos,
1456 matroska->unknown_count);
1457 return AVERROR_INVALIDDATA;
1458 }
1459 }
1460 }
1461
1462 if (update_pos > 0) {
1463 // We have found an element that is allowed at this place
1464 // in the hierarchy and it passed all checks, so treat the beginning
1465 // of the element as the "last known good" position.
1466 matroska->resync_pos = pos;
1467 }
1468
1469 if (!data && length != EBML_UNKNOWN_LENGTH)
1470 goto skip;
1471 }
1472
1473 switch (syntax->type) {
1474 case EBML_UINT:
1475 res = ebml_read_uint(pb, length, syntax->def.u, data);
1476 break;
1477 case EBML_SINT:
1478 res = ebml_read_sint(pb, length, syntax->def.i, data);
1479 break;
1480 case EBML_FLOAT:
1481 res = ebml_read_float(pb, length, syntax->def.f, data);
1482 break;
1483 case EBML_STR:
1484 case EBML_UTF8:
1485 res = ebml_read_ascii(pb, length, syntax->def.s, data);
1486 break;
1487 case EBML_BIN:
1488 res = ebml_read_binary(pb, length, pos_alt, data);
1489 break;
1490 case EBML_LEVEL1:
1491 case EBML_NEST:
1492 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1493 return res;
1494 if (id == MATROSKA_ID_SEGMENT)
1495 matroska->segment_start = pos_alt;
1496 if (id == MATROSKA_ID_CUES)
1497 matroska->cues_parsing_deferred = 0;
1498 if (syntax->type == EBML_LEVEL1 &&
1499 (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1500 if (!level1_elem->pos) {
1501 // Zero is not a valid position for a level 1 element.
1502 level1_elem->pos = pos;
1503 } else if (level1_elem->pos != pos)
1504 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1505 level1_elem->parsed = 1;
1506 }
1507 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1508 return res;
1509 break;
1510 case EBML_STOP:
1511 return 1;
1512 skip:
1513 default:
1514 if (length) {
1515 int64_t res2;
1516 if (ffio_limit(pb, length) != length) {
1517 // ffio_limit emits its own error message,
1518 // so we don't have to.
1519 return AVERROR(EIO);
1520 }
1521 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1522 // avio_skip might take us past EOF. We check for this
1523 // by skipping only length - 1 bytes, reading a byte and
1524 // checking the error flags. This is done in order to check
1525 // that the element has been properly skipped even when
1526 // no filesize (that ffio_limit relies on) is available.
1527 avio_r8(pb);
1528 res = NEEDS_CHECKING;
1529 } else
1530 res = res2;
1531 } else
1532 res = 0;
1533 }
1534 if (res) {
1535 if (res == NEEDS_CHECKING) {
1536 if (pb->eof_reached) {
1537 if (pb->error)
1538 res = pb->error;
1539 else
1540 res = AVERROR_EOF;
1541 } else
1542 goto level_check;
1543 }
1544
1545 if (res == AVERROR_INVALIDDATA)
1546 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1547 else if (res == AVERROR(EIO))
1548 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1549 else if (res == AVERROR_EOF) {
1550 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1551 res = AVERROR(EIO);
1552 }
1553
1554 return res;
1555 }
1556
1557 level_check:
1558 if (syntax->is_counted && data) {
1559 CountedElement *elem = data;
1560 if (elem->count != UINT_MAX)
1561 elem->count++;
1562 }
1563
1564 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1565 level = &matroska->levels[matroska->num_levels - 1];
1566 pos = avio_tell(pb);
1567
1568 // Given that pos >= level->start no check for
1569 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1570 while (matroska->num_levels && pos == level->start + level->length) {
1571 matroska->num_levels--;
1572 level--;
1573 }
1574 }
1575
1576 return level_check;
1577 }
1578
1579 static void ebml_free(EbmlSyntax *syntax, void *data)
1580 {
1581 int i, j;
1582 for (i = 0; syntax[i].id; i++) {
1583 void *data_off = (char *) data + syntax[i].data_offset;
1584 switch (syntax[i].type) {
1585 case EBML_STR:
1586 case EBML_UTF8:
1587 av_freep(data_off);
1588 break;
1589 case EBML_BIN:
1590 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1591 break;
1592 case EBML_LEVEL1:
1593 case EBML_NEST:
1594 if (syntax[i].list_elem_size) {
1595 EbmlList *list = data_off;
1596 char *ptr = list->elem;
1597 for (j = 0; j < list->nb_elem;
1598 j++, ptr += syntax[i].list_elem_size)
1599 ebml_free(syntax[i].def.n, ptr);
1600 av_freep(&list->elem);
1601 list->nb_elem = 0;
1602 list->alloc_elem_size = 0;
1603 } else
1604 ebml_free(syntax[i].def.n, data_off);
1605 default:
1606 break;
1607 }
1608 }
1609 }
1610
1611 /*
1612 * Autodetecting...
1613 */
1614 static int matroska_probe(const AVProbeData *p)
1615 {
1616 uint64_t total = 0;
1617 int len_mask = 0x80, size = 1, n = 1, i;
1618
1619 /* EBML header? */
1620 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1621 return 0;
1622
1623 /* length of header */
1624 total = p->buf[4];
1625 while (size <= 8 && !(total & len_mask)) {
1626 size++;
1627 len_mask >>= 1;
1628 }
1629 if (size > 8)
1630 return 0;
1631 total &= (len_mask - 1);
1632 while (n < size)
1633 total = (total << 8) | p->buf[4 + n++];
1634
1635 if (total + 1 == 1ULL << (7 * size)){
1636 /* Unknown-length header - simply parse the whole buffer. */
1637 total = p->buf_size - 4 - size;
1638 } else {
1639 /* Does the probe data contain the whole header? */
1640 if (p->buf_size < 4 + size + total)
1641 return 0;
1642 }
1643
1644 /* The header should contain a known document type. For now,
1645 * we don't parse the whole header but simply check for the
1646 * availability of that array of characters inside the header.
1647 * Not fully fool-proof, but good enough. */
1648 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1649 size_t probelen = strlen(matroska_doctypes[i]);
1650 if (total < probelen)
1651 continue;
1652 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1653 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1654 return AVPROBE_SCORE_MAX;
1655 }
1656
1657 // probably valid EBML header but no recognized doctype
1658 return AVPROBE_SCORE_EXTENSION;
1659 }
1660
1661 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1662 uint64_t num)
1663 {
1664 MatroskaTrack *tracks = matroska->tracks.elem;
1665 int i;
1666
1667 for (i = 0; i < matroska->tracks.nb_elem; i++)
1668 if (tracks[i].num == num)
1669 return &tracks[i];
1670
1671 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1672 return NULL;
1673 }
1674
1675 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1676 MatroskaTrack *track)
1677 {
1678 MatroskaTrackEncoding *encodings = track->encodings.elem;
1679 uint8_t *data = *buf;
1680 int isize = *buf_size;
1681 uint8_t *pkt_data = NULL;
1682 av_unused uint8_t *newpktdata;
1683 int pkt_size = isize;
1684 int result = 0;
1685 int olen;
1686
1687 if (pkt_size >= 10000000U)
1688 return AVERROR_INVALIDDATA;
1689
1690 switch (encodings[0].compression.algo) {
1691 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1692 {
1693 int header_size = encodings[0].compression.settings.size;
1694 uint8_t *header = encodings[0].compression.settings.data;
1695
1696 if (header_size && !header) {
1697 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1698 return -1;
1699 }
1700
1701 if (!header_size)
1702 return 0;
1703
1704 pkt_size = isize + header_size;
1705 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1706 if (!pkt_data)
1707 return AVERROR(ENOMEM);
1708
1709 memcpy(pkt_data, header, header_size);
1710 memcpy(pkt_data + header_size, data, isize);
1711 break;
1712 }
1713 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1714 do {
1715 int insize = isize;
1716 olen = pkt_size *= 3;
1717 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1718 + AV_INPUT_BUFFER_PADDING_SIZE);
1719 if (!newpktdata) {
1720 result = AVERROR(ENOMEM);
1721 goto failed;
1722 }
1723 pkt_data = newpktdata;
1724 result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1725 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1726 if (result) {
1727 result = AVERROR_INVALIDDATA;
1728 goto failed;
1729 }
1730 pkt_size -= olen;
1731 break;
1732 #if CONFIG_ZLIB
1733 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1734 {
1735 z_stream zstream = { 0 };
1736 if (!pkt_size || inflateInit(&zstream) != Z_OK)
1737 return -1;
1738 zstream.next_in = data;
1739 zstream.avail_in = isize;
1740 do {
1741 pkt_size *= 3;
1742 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1743 if (!newpktdata) {
1744 inflateEnd(&zstream);
1745 result = AVERROR(ENOMEM);
1746 goto failed;
1747 }
1748 pkt_data = newpktdata;
1749 zstream.avail_out = pkt_size - zstream.total_out;
1750 zstream.next_out = pkt_data + zstream.total_out;
1751 result = inflate(&zstream, Z_NO_FLUSH);
1752 } while (result == Z_OK && pkt_size < 10000000);
1753 pkt_size = zstream.total_out;
1754 inflateEnd(&zstream);
1755 if (result != Z_STREAM_END) {
1756 if (result == Z_MEM_ERROR)
1757 result = AVERROR(ENOMEM);
1758 else
1759 result = AVERROR_INVALIDDATA;
1760 goto failed;
1761 }
1762 break;
1763 }
1764 #endif
1765 #if CONFIG_BZLIB
1766 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1767 {
1768 bz_stream bzstream = { 0 };
1769 if (!pkt_size || BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1770 return -1;
1771 bzstream.next_in = data;
1772 bzstream.avail_in = isize;
1773 do {
1774 pkt_size *= 3;
1775 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1776 if (!newpktdata) {
1777 BZ2_bzDecompressEnd(&bzstream);
1778 result = AVERROR(ENOMEM);
1779 goto failed;
1780 }
1781 pkt_data = newpktdata;
1782 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1783 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1784 result = BZ2_bzDecompress(&bzstream);
1785 } while (result == BZ_OK && pkt_size < 10000000);
1786 pkt_size = bzstream.total_out_lo32;
1787 BZ2_bzDecompressEnd(&bzstream);
1788 if (result != BZ_STREAM_END) {
1789 if (result == BZ_MEM_ERROR)
1790 result = AVERROR(ENOMEM);
1791 else
1792 result = AVERROR_INVALIDDATA;
1793 goto failed;
1794 }
1795 break;
1796 }
1797 #endif
1798 default:
1799 return AVERROR_INVALIDDATA;
1800 }
1801
1802 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1803
1804 *buf = pkt_data;
1805 *buf_size = pkt_size;
1806 return 0;
1807
1808 failed:
1809 av_free(pkt_data);
1810 return result;
1811 }
1812
1813 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1814 AVDictionary **metadata, char *prefix)
1815 {
1816 MatroskaTag *tags = list->elem;
1817 char key[1024];
1818 int i;
1819
1820 for (i = 0; i < list->nb_elem; i++) {
1821 const char *lang = tags[i].lang &&
1822 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1823
1824 if (!tags[i].name) {
1825 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1826 continue;
1827 }
1828 if (prefix)
1829 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1830 else
1831 av_strlcpy(key, tags[i].name, sizeof(key));
1832 if (tags[i].def || !lang) {
1833 av_dict_set(metadata, key, tags[i].string, 0);
1834 if (tags[i].sub.nb_elem)
1835 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1836 }
1837 if (lang) {
1838 av_strlcat(key, "-", sizeof(key));
1839 av_strlcat(key, lang, sizeof(key));
1840 av_dict_set(metadata, key, tags[i].string, 0);
1841 if (tags[i].sub.nb_elem)
1842 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1843 }
1844 }
1845 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1846 }
1847
1848 static void matroska_convert_tags(AVFormatContext *s)
1849 {
1850 MatroskaDemuxContext *matroska = s->priv_data;
1851 MatroskaTags *tags = matroska->tags.elem;
1852 int i, j;
1853
1854 for (i = 0; i < matroska->tags.nb_elem; i++) {
1855 if (tags[i].target.attachuid) {
1856 MatroskaAttachment *attachment = matroska->attachments.elem;
1857 int found = 0;
1858 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1859 if (attachment[j].uid == tags[i].target.attachuid &&
1860 attachment[j].stream) {
1861 matroska_convert_tag(s, &tags[i].tag,
1862 &attachment[j].stream->metadata, NULL);
1863 found = 1;
1864 }
1865 }
1866 if (!found) {
1867 av_log(s, AV_LOG_WARNING,
1868 "The tags at index %d refer to a "
1869 "non-existent attachment %"PRId64".\n",
1870 i, tags[i].target.attachuid);
1871 }
1872 } else if (tags[i].target.chapteruid) {
1873 MatroskaChapter *chapter = matroska->chapters.elem;
1874 int found = 0;
1875 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1876 if (chapter[j].uid == tags[i].target.chapteruid &&
1877 chapter[j].chapter) {
1878 matroska_convert_tag(s, &tags[i].tag,
1879 &chapter[j].chapter->metadata, NULL);
1880 found = 1;
1881 }
1882 }
1883 if (!found) {
1884 av_log(s, AV_LOG_WARNING,
1885 "The tags at index %d refer to a non-existent chapter "
1886 "%"PRId64".\n",
1887 i, tags[i].target.chapteruid);
1888 }
1889 } else if (tags[i].target.trackuid) {
1890 MatroskaTrack *track = matroska->tracks.elem;
1891 int found = 0;
1892 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1893 if (track[j].uid == tags[i].target.trackuid &&
1894 track[j].stream) {
1895 matroska_convert_tag(s, &tags[i].tag,
1896 &track[j].stream->metadata, NULL);
1897 found = 1;
1898 }
1899 }
1900 if (!found) {
1901 av_log(s, AV_LOG_WARNING,
1902 "The tags at index %d refer to a non-existent track "
1903 "%"PRId64".\n",
1904 i, tags[i].target.trackuid);
1905 }
1906 } else {
1907 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1908 tags[i].target.type);
1909 }
1910 }
1911 }
1912
1913 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1914 int64_t pos)
1915 {
1916 uint32_t saved_id = matroska->current_id;
1917 int64_t before_pos = avio_tell(matroska->ctx->pb);
1918 int ret = 0;
1919 int ret2;
1920
1921 /* seek */
1922 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1923 /* We don't want to lose our seekhead level, so we add
1924 * a dummy. This is a crude hack. */
1925 if (matroska->num_levels == EBML_MAX_DEPTH) {
1926 av_log(matroska->ctx, AV_LOG_INFO,
1927 "Max EBML element depth (%d) reached, "
1928 "cannot parse further.\n", EBML_MAX_DEPTH);
1929 ret = AVERROR_INVALIDDATA;
1930 } else {
1931 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1932 matroska->num_levels++;
1933 matroska->current_id = 0;
1934
1935 ret = ebml_parse(matroska, matroska_segment, matroska);
1936 if (ret == LEVEL_ENDED) {
1937 /* This can only happen if the seek brought us beyond EOF. */
1938 ret = AVERROR_EOF;
1939 }
1940 }
1941 }
1942 /* Seek back - notice that in all instances where this is used
1943 * it is safe to set the level to 1. */
1944 ret2 = matroska_reset_status(matroska, saved_id, before_pos);
1945 if (ret >= 0)
1946 ret = ret2;
1947
1948 return ret;
1949 }
1950
1951 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1952 {
1953 EbmlList *seekhead_list = &matroska->seekhead;
1954 int i;
1955
1956 // we should not do any seeking in the streaming case
1957 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1958 return;
1959
1960 for (i = 0; i < seekhead_list->nb_elem; i++) {
1961 MatroskaSeekhead *seekheads = seekhead_list->elem;
1962 uint32_t id = seekheads[i].id;
1963 int64_t pos = seekheads[i].pos + matroska->segment_start;
1964 MatroskaLevel1Element *elem;
1965
1966 if (id != seekheads[i].id || pos < matroska->segment_start)
1967 continue;
1968
1969 elem = matroska_find_level1_elem(matroska, id, pos);
1970 if (!elem || elem->parsed)
1971 continue;
1972
1973 elem->pos = pos;
1974
1975 // defer cues parsing until we actually need cue data.
1976 if (id == MATROSKA_ID_CUES)
1977 continue;
1978
1979 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1980 // mark index as broken
1981 matroska->cues_parsing_deferred = -1;
1982 break;
1983 }
1984
1985 elem->parsed = 1;
1986 }
1987 }
1988
1989 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1990 {
1991 EbmlList *index_list;
1992 MatroskaIndex *index;
1993 uint64_t index_scale = 1;
1994 int i, j;
1995
1996 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1997 return;
1998
1999 index_list = &matroska->index;
2000 index = index_list->elem;
2001 if (index_list->nb_elem < 2)
2002 return;
2003 if (index[1].time > 1E14 / matroska->time_scale) {
2004 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
2005 return;
2006 }
2007 for (i = 0; i < index_list->nb_elem; i++) {
2008 EbmlList *pos_list = &index[i].pos;
2009 MatroskaIndexPos *pos = pos_list->elem;
2010 for (j = 0; j < pos_list->nb_elem; j++) {
2011 MatroskaTrack *track = matroska_find_track_by_num(matroska,
2012 pos[j].track);
2013 if (track && track->stream)
2014 av_add_index_entry(track->stream,
2015 pos[j].pos + matroska->segment_start,
2016 index[i].time / index_scale, 0, 0,
2017 AVINDEX_KEYFRAME);
2018 }
2019 }
2020 }
2021
2022 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
2023 int i;
2024
2025 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
2026 return;
2027
2028 for (i = 0; i < matroska->num_level1_elems; i++) {
2029 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
2030 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
2031 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
2032 matroska->cues_parsing_deferred = -1;
2033 elem->parsed = 1;
2034 break;
2035 }
2036 }
2037
2038 matroska_add_index_entries(matroska);
2039 }
2040
2041 static int matroska_parse_content_encodings(MatroskaTrackEncoding *encodings,
2042 unsigned nb_encodings,
2043 MatroskaTrack *track,
2044 char **key_id_base64, void *logctx)
2045 {
2046 if (nb_encodings > 1) {
2047 av_log(logctx, AV_LOG_ERROR,
2048 "Multiple combined encodings not supported\n");
2049 return 0;
2050 }
2051 if (!nb_encodings)
2052 return 0;
2053 if (encodings->type) {
2054 if (encodings->encryption.key_id.size > 0) {
2055 /* Save the encryption key id to be stored later
2056 * as a metadata tag. */
2057 const int b64_size = AV_BASE64_SIZE(encodings->encryption.key_id.size);
2058 *key_id_base64 = av_malloc(b64_size);
2059 if (!*key_id_base64)
2060 return AVERROR(ENOMEM);
2061
2062 av_base64_encode(*key_id_base64, b64_size,
2063 encodings->encryption.key_id.data,
2064 encodings->encryption.key_id.size);
2065 } else {
2066 encodings->scope = 0;
2067 av_log(logctx, AV_LOG_ERROR, "Unsupported encoding type\n");
2068 }
2069 } else if (
2070 #if CONFIG_ZLIB
2071 encodings->compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2072 #endif
2073 #if CONFIG_BZLIB
2074 encodings->compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2075 #endif
2076 encodings->compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2077 encodings->compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2078 encodings->scope = 0;
2079 av_log(logctx, AV_LOG_ERROR, "Unsupported encoding type\n");
2080 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2081 uint8_t *codec_priv = track->codec_priv.data;
2082 int ret = matroska_decode_buffer(&track->codec_priv.data,
2083 &track->codec_priv.size,
2084 track);
2085 if (ret < 0) {
2086 track->codec_priv.data = NULL;
2087 track->codec_priv.size = 0;
2088 av_log(logctx, AV_LOG_ERROR,
2089 "Failed to decode codec private data\n");
2090 }
2091
2092 if (codec_priv != track->codec_priv.data) {
2093 av_buffer_unref(&track->codec_priv.buf);
2094 if (track->codec_priv.data) {
2095 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2096 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2097 NULL, NULL, 0);
2098 if (!track->codec_priv.buf) {
2099 av_freep(&track->codec_priv.data);
2100 track->codec_priv.size = 0;
2101 return AVERROR(ENOMEM);
2102 }
2103 }
2104 }
2105 }
2106 track->needs_decoding = !encodings->type &&
2107 encodings->scope & 1 &&
2108 (encodings->compression.algo !=
2109 MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP ||
2110 encodings->compression.settings.size);
2111
2112 return 0;
2113 }
2114
2115 static int matroska_aac_profile(char *codec_id)
2116 {
2117 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
2118 int profile;
2119
2120 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
2121 if (strstr(codec_id, aac_profiles[profile]))
2122 break;
2123 return profile + 1;
2124 }
2125
2126 static int matroska_aac_sri(int samplerate)
2127 {
2128 int sri;
2129
2130 for (sri = 0; sri < FF_ARRAY_ELEMS(ff_mpeg4audio_sample_rates); sri++)
2131 if (ff_mpeg4audio_sample_rates[sri] == samplerate)
2132 break;
2133 return sri;
2134 }
2135
2136 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
2137 {
2138 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
2139 ff_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
2140 }
2141
2142 static int matroska_parse_flac(AVFormatContext *s,
2143 MatroskaTrack *track,
2144 int *offset)
2145 {
2146 AVStream *st = track->stream;
2147 uint8_t *p = track->codec_priv.data;
2148 int size = track->codec_priv.size;
2149
2150 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
2151 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
2152 track->codec_priv.size = 0;
2153 return 0;
2154 }
2155 *offset = 8;
2156 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
2157
2158 p += track->codec_priv.size;
2159 size -= track->codec_priv.size;
2160
2161 /* parse the remaining metadata blocks if present */
2162 while (size >= 4) {
2163 int block_last, block_type, block_size;
2164
2165 flac_parse_block_header(p, &block_last, &block_type, &block_size);
2166
2167 p += 4;
2168 size -= 4;
2169 if (block_size > size)
2170 return 0;
2171
2172 /* check for the channel mask */
2173 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2174 AVDictionary *dict = NULL;
2175 AVDictionaryEntry *chmask;
2176
2177 ff_vorbis_comment(s, &dict, p, block_size, 0);
2178 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2179 if (chmask) {
2180 uint64_t mask = strtol(chmask->value, NULL, 0);
2181 if (!mask || mask & ~0x3ffffULL) {
2182 av_log(s, AV_LOG_WARNING,
2183 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2184 } else
2185 av_channel_layout_from_mask(&st->codecpar->ch_layout, mask);
2186 }
2187 av_dict_free(&dict);
2188 }
2189
2190 p += block_size;
2191 size -= block_size;
2192 }
2193
2194 return 0;
2195 }
2196
2197 static int mkv_field_order(const MatroskaDemuxContext *matroska, uint64_t field_order)
2198 {
2199 int minor, micro, bttb = 0;
2200
2201 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2202 * this function, and fixed in 57.52 */
2203 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, &micro) == 2)
2204 bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2205
2206 switch (field_order) {
2207 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2208 return AV_FIELD_PROGRESSIVE;
2209 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2210 return AV_FIELD_UNKNOWN;
2211 case MATROSKA_VIDEO_FIELDORDER_TT:
2212 return AV_FIELD_TT;
2213 case MATROSKA_VIDEO_FIELDORDER_BB:
2214 return AV_FIELD_BB;
2215 case MATROSKA_VIDEO_FIELDORDER_BT:
2216 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2217 case MATROSKA_VIDEO_FIELDORDER_TB:
2218 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2219 default:
2220 return AV_FIELD_UNKNOWN;
2221 }
2222 }
2223
2224 static void mkv_stereo_mode_display_mul(int stereo_mode,
2225 int *h_width, int *h_height)
2226 {
2227 switch (stereo_mode) {
2228 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2229 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2230 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2231 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2232 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2233 break;
2234 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2235 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2236 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2237 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2238 *h_width = 2;
2239 break;
2240 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2241 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2242 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2243 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2244 *h_height = 2;
2245 break;
2246 }
2247 }
2248
2249 static int mkv_stereo3d_conv(AVStream *st, MatroskaVideoStereoModeType stereo_mode)
2250 {
2251 static const struct {
2252 char type;
2253 char flags;
2254 } stereo_mode_conv [] = {
2255 #define STEREO_MODE_CONV(STEREOMODETYPE, STEREO3DTYPE, FLAGS, WDIV, HDIV, WEBM) \
2256 [(STEREOMODETYPE)] = { .type = (STEREO3DTYPE), .flags = (FLAGS) },
2257 #define NOTHING(STEREOMODETYPE, WDIV, HDIV, WEBM)
2258 STEREOMODE_STEREO3D_MAPPING(STEREO_MODE_CONV, NOTHING)
2259 };
2260 AVStereo3D *stereo;
2261 size_t size;
2262
2263 stereo = av_stereo3d_alloc_size(&size);
2264 if (!stereo)
2265 return AVERROR(ENOMEM);
2266
2267 stereo->type = stereo_mode_conv[stereo_mode].type;
2268 stereo->flags = stereo_mode_conv[stereo_mode].flags;
2269
2270 if (!av_packet_side_data_add(&st->codecpar->coded_side_data, &st->codecpar->nb_coded_side_data,
2271 AV_PKT_DATA_STEREO3D, stereo, size, 0)) {
2272 av_freep(&stereo);
2273 return AVERROR(ENOMEM);
2274 }
2275
2276 return 0;
2277 }
2278
2279 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2280 const MatroskaTrackVideoColor *color = track->video.color.elem;
2281 const MatroskaMasteringMeta *mastering_meta;
2282 int has_mastering_primaries, has_mastering_luminance;
2283
2284 if (!track->video.color.nb_elem)
2285 return 0;
2286
2287 mastering_meta = &color->mastering_meta;
2288 // Mastering primaries are CIE 1931 coords, and must be > 0.
2289 has_mastering_primaries =
2290 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2291 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2292 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2293 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2294 has_mastering_luminance = mastering_meta->max_luminance >
2295 mastering_meta->min_luminance.el.f &&
2296 mastering_meta->min_luminance.el.f >= 0 &&
2297 mastering_meta->min_luminance.count;
2298
2299 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2300 st->codecpar->color_space = color->matrix_coefficients;
2301 if (color->primaries != AVCOL_PRI_RESERVED &&
2302 color->primaries != AVCOL_PRI_RESERVED0)
2303 st->codecpar->color_primaries = color->primaries;
2304 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2305 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2306 st->codecpar->color_trc = color->transfer_characteristics;
2307 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2308 color->range <= AVCOL_RANGE_JPEG)
2309 st->codecpar->color_range = color->range;
2310 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2311 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2312 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2313 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2314 st->codecpar->chroma_location =
2315 av_chroma_location_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2316 (color->chroma_siting_vert - 1) << 7);
2317 }
2318 if (color->max_cll && color->max_fall) {
2319 size_t size = 0;
2320 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2321 if (!metadata)
2322 return AVERROR(ENOMEM);
2323 if (!av_packet_side_data_add(&st->codecpar->coded_side_data, &st->codecpar->nb_coded_side_data,
2324 AV_PKT_DATA_CONTENT_LIGHT_LEVEL, metadata, size, 0)) {
2325 av_freep(&metadata);
2326 return AVERROR(ENOMEM);
2327 }
2328 metadata->MaxCLL = color->max_cll;
2329 metadata->MaxFALL = color->max_fall;
2330 }
2331
2332 if (has_mastering_primaries || has_mastering_luminance) {
2333 size_t size = 0;
2334 AVMasteringDisplayMetadata *metadata = av_mastering_display_metadata_alloc_size(&size);
2335 if (!metadata)
2336 return AVERROR(ENOMEM);
2337 if (!av_packet_side_data_add(&st->codecpar->coded_side_data, &st->codecpar->nb_coded_side_data,
2338 AV_PKT_DATA_MASTERING_DISPLAY_METADATA, metadata, size, 0)) {
2339 av_freep(&metadata);
2340 return AVERROR(ENOMEM);
2341 }
2342 if (has_mastering_primaries) {
2343 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2344 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2345 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2346 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2347 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2348 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2349 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2350 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2351 metadata->has_primaries = 1;
2352 }
2353 if (has_mastering_luminance) {
2354 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2355 metadata->min_luminance = av_d2q(mastering_meta->min_luminance.el.f, INT_MAX);
2356 metadata->has_luminance = 1;
2357 }
2358 }
2359 return 0;
2360 }
2361
2362 static int mkv_create_display_matrix(AVStream *st,
2363 const MatroskaTrackVideoProjection *proj,
2364 void *logctx)
2365 {
2366 AVPacketSideData *sd;
2367 double pitch = proj->pitch, yaw = proj->yaw, roll = proj->roll;
2368 int32_t *matrix;
2369 int hflip;
2370
2371 if (pitch == 0.0 && yaw == 0.0 && roll == 0.0)
2372 return 0;
2373
2374 /* Note: The following constants are exactly representable
2375 * as floating-point numbers. */
2376 if (pitch != 0.0 || (yaw != 0.0 && yaw != 180.0 && yaw != -180.0) ||
2377 isnan(roll)) {
2378 av_log(logctx, AV_LOG_WARNING, "Ignoring non-2D rectangular "
2379 "projection in stream %u (yaw %f, pitch %f, roll %f)\n",
2380 st->index, yaw, pitch, roll);
2381 return 0;
2382 }
2383 sd = av_packet_side_data_new(&st->codecpar->coded_side_data,
2384 &st->codecpar->nb_coded_side_data,
2385 AV_PKT_DATA_DISPLAYMATRIX,
2386 9 * sizeof(*matrix), 0);
2387 if (!sd)
2388 return AVERROR(ENOMEM);
2389 matrix = (int32_t*)sd->data;
2390
2391 hflip = yaw != 0.0;
2392 /* ProjectionPoseRoll is in the counter-clockwise direction
2393 * whereas av_display_rotation_set() expects its argument
2394 * to be oriented clockwise, so we need to negate roll.
2395 * Furthermore, if hflip is set, we need to negate it again
2396 * to account for the fact that the Matroska specifications
2397 * require the yaw rotation to be applied first. */
2398 av_display_rotation_set(matrix, roll * (2 * hflip - 1));
2399 av_display_matrix_flip(matrix, hflip, 0);
2400
2401 return 0;
2402 }
2403
2404 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2405 void *logctx)
2406 {
2407 AVSphericalMapping *spherical;
2408 const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2409 const uint8_t *priv_data = mkv_projection->private.data;
2410 enum AVSphericalProjection projection;
2411 size_t spherical_size;
2412 uint32_t l = 0, t = 0, r = 0, b = 0;
2413 uint32_t padding = 0;
2414
2415 if (mkv_projection->private.size && priv_data[0] != 0) {
2416 av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2417 return 0;
2418 }
2419
2420 switch (track->video.projection.type) {
2421 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2422 return mkv_create_display_matrix(st, mkv_projection, logctx);
2423 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2424 if (track->video.projection.private.size == 20) {
2425 t = AV_RB32(priv_data + 4);
2426 b = AV_RB32(priv_data + 8);
2427 l = AV_RB32(priv_data + 12);
2428 r = AV_RB32(priv_data + 16);
2429
2430 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2431 av_log(logctx, AV_LOG_ERROR,
2432 "Invalid bounding rectangle coordinates "
2433 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2434 l, t, r, b);
2435 return AVERROR_INVALIDDATA;
2436 }
2437 } else if (track->video.projection.private.size != 0) {
2438 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2439 return AVERROR_INVALIDDATA;
2440 }
2441
2442 if (l || t || r || b)
2443 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2444 else
2445 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2446 break;
2447 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2448 if (track->video.projection.private.size < 4) {
2449 av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2450 return AVERROR_INVALIDDATA;
2451 } else if (track->video.projection.private.size == 12) {
2452 uint32_t layout = AV_RB32(priv_data + 4);
2453 if (layout) {
2454 av_log(logctx, AV_LOG_WARNING,
2455 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2456 return 0;
2457 }
2458 projection = AV_SPHERICAL_CUBEMAP;
2459 padding = AV_RB32(priv_data + 8);
2460 } else {
2461 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2462 return AVERROR_INVALIDDATA;
2463 }
2464 break;
2465 default:
2466 av_log(logctx, AV_LOG_WARNING,
2467 "Unknown spherical metadata type %"PRIu64"\n",
2468 track->video.projection.type);
2469 return 0;
2470 }
2471
2472 spherical = av_spherical_alloc(&spherical_size);
2473 if (!spherical)
2474 return AVERROR(ENOMEM);
2475
2476 spherical->projection = projection;
2477
2478 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2479 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2480 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2481
2482 spherical->padding = padding;
2483
2484 spherical->bound_left = l;
2485 spherical->bound_top = t;
2486 spherical->bound_right = r;
2487 spherical->bound_bottom = b;
2488
2489 if (!av_packet_side_data_add(&st->codecpar->coded_side_data, &st->codecpar->nb_coded_side_data,
2490 AV_PKT_DATA_SPHERICAL, spherical, spherical_size, 0)) {
2491 av_freep(&spherical);
2492 return AVERROR(ENOMEM);
2493 }
2494
2495 return 0;
2496 }
2497
2498 static int mkv_parse_dvcc_dvvc(AVFormatContext *s, AVStream *st, const MatroskaTrack *track,
2499 EbmlBin *bin)
2500 {
2501 return ff_isom_parse_dvcc_dvvc(s, st, bin->data, bin->size);
2502 }
2503
2504 static int mkv_parse_block_addition_mappings(AVFormatContext *s, AVStream *st, MatroskaTrack *track)
2505 {
2506 const EbmlList *mappings_list = &track->block_addition_mappings;
2507 MatroskaBlockAdditionMapping *mappings = mappings_list->elem;
2508 int ret;
2509
2510 for (int i = 0; i < mappings_list->nb_elem; i++) {
2511 MatroskaBlockAdditionMapping *mapping = &mappings[i];
2512 uint64_t type = mapping->type;
2513
2514 switch (mapping->type) {
2515 case MATROSKA_BLOCK_ADD_ID_TYPE_DEFAULT:
2516 av_log(s, AV_LOG_DEBUG,
2517 "Explicit block Addition Mapping type \"Use BlockAddIDValue\", value %"PRIu64","
2518 " name \"%s\" found.\n", mapping->value, mapping->name ? mapping->name : "");
2519 type = MATROSKA_BLOCK_ADD_ID_TYPE_OPAQUE;
2520 // fall-through
2521 case MATROSKA_BLOCK_ADD_ID_TYPE_OPAQUE:
2522 case MATROSKA_BLOCK_ADD_ID_TYPE_ITU_T_T35:
2523 if (mapping->value != type) {
2524 int strict = s->strict_std_compliance >= FF_COMPLIANCE_STRICT;
2525 av_log(s, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
2526 "Invalid Block Addition Value 0x%"PRIx64" for Block Addition Mapping Type "
2527 "0x%"PRIx64", name \"%s\"\n", mapping->value, mapping->type,
2528 mapping->name ? mapping->name : "");
2529 if (strict)
2530 return AVERROR_INVALIDDATA;
2531 }
2532 break;
2533 case MATROSKA_BLOCK_ADD_ID_TYPE_DVCC:
2534 case MATROSKA_BLOCK_ADD_ID_TYPE_DVVC:
2535 if ((ret = mkv_parse_dvcc_dvvc(s, st, track, &mapping->extradata)) < 0)
2536 return ret;
2537
2538 break;
2539 default:
2540 av_log(s, AV_LOG_DEBUG,
2541 "Unknown Block Addition Mapping type 0x%"PRIx64", value %"PRIu64", name \"%s\"\n",
2542 mapping->type, mapping->value, mapping->name ? mapping->name : "");
2543 if (mapping->value < 2) {
2544 int strict = s->strict_std_compliance >= FF_COMPLIANCE_STRICT;
2545 av_log(s, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
2546 "Invalid Block Addition value 0x%"PRIu64" for unknown Block Addition Mapping "
2547 "type %"PRIx64", name \"%s\"\n", mapping->value, mapping->type,
2548 mapping->name ? mapping->name : "");
2549 if (strict)
2550 return AVERROR_INVALIDDATA;
2551 }
2552 break;
2553 }
2554 }
2555
2556 return 0;
2557 }
2558
2559 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2560 {
2561 const AVCodecTag *codec_tags;
2562
2563 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2564 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2565
2566 /* Normalize noncompliant private data that starts with the fourcc
2567 * by expanding/shifting the data by 4 bytes and storing the data
2568 * size at the start. */
2569 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2570 int ret = av_buffer_realloc(&track->codec_priv.buf,
2571 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2572 if (ret < 0)
2573 return ret;
2574
2575 track->codec_priv.data = track->codec_priv.buf->data;
2576 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2577 track->codec_priv.size += 4;
2578 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2579 }
2580
2581 *fourcc = AV_RL32(track->codec_priv.data + 4);
2582 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2583
2584 return 0;
2585 }
2586
2587 /* An enum with potential return values of the functions for parsing a track.
2588 * Apart from that all these functions can also indicate ordinary errors via
2589 * negative return values. */
2590 enum {
2591 SKIP_TRACK = 1,
2592 };
2593
2594 #define AAC_MAX_EXTRADATA_SIZE 5
2595 #define TTA_EXTRADATA_SIZE 22
2596 #define WAVPACK_EXTRADATA_SIZE 2
2597 /* Performs the codec-specific part of parsing an audio track. */
2598 static int mka_parse_audio_codec(MatroskaTrack *track, AVCodecParameters *par,
2599 const MatroskaDemuxContext *matroska,
2600 AVFormatContext *s, int *extradata_offset)
2601 {
2602 uint8_t extradata[FFMAX3(AAC_MAX_EXTRADATA_SIZE,
2603 TTA_EXTRADATA_SIZE,
2604 WAVPACK_EXTRADATA_SIZE)];
2605 int extradata_size = 0; // > 0 means that the extradata buffer is used
2606 int ret;
2607
2608 if (!strcmp(track->codec_id, "A_MS/ACM") &&
2609 track->codec_priv.size >= 14) {
2610 FFIOContext b;
2611 ffio_init_read_context(&b, track->codec_priv.data,
2612 track->codec_priv.size);
2613 ret = ff_get_wav_header(s, &b.pub, par,
2614 track->codec_priv.size, 0);
2615 if (ret < 0)
2616 return ret;
2617 *extradata_offset = FFMIN(track->codec_priv.size, 18);
2618 return 0;
2619 } else if (!strcmp(track->codec_id, "A_QUICKTIME") &&
2620 /* Normally 36, but allow noncompliant private data */
2621 track->codec_priv.size >= 32) {
2622 enum AVCodecID codec_id;
2623 uint32_t fourcc;
2624 uint16_t sample_size;
2625
2626 ret = get_qt_codec(track, &fourcc, &codec_id);
2627 if (ret < 0)
2628 return ret;
2629 sample_size = AV_RB16(track->codec_priv.data + 26);
2630 if (fourcc == 0) {
2631 if (sample_size == 8) {
2632 fourcc = MKTAG('r','a','w',' ');
2633 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2634 } else if (sample_size == 16) {
2635 fourcc = MKTAG('t','w','o','s');
2636 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2637 }
2638 }
2639 if ((fourcc == MKTAG('t','w','o','s') ||
2640 fourcc == MKTAG('s','o','w','t')) && sample_size == 8)
2641 codec_id = AV_CODEC_ID_PCM_S8;
2642 par->codec_id = codec_id;
2643 par->codec_tag = fourcc;
2644 return 0;
2645 }
2646
2647 switch (par->codec_id) {
2648 case AV_CODEC_ID_PCM_S16BE:
2649 switch (track->audio.bitdepth) {
2650 case 8:
2651 par->codec_id = AV_CODEC_ID_PCM_U8;
2652 break;
2653 case 24:
2654 par->codec_id = AV_CODEC_ID_PCM_S24BE;
2655 break;
2656 case 32:
2657 par->codec_id = AV_CODEC_ID_PCM_S32BE;
2658 break;
2659 }
2660 break;
2661 case AV_CODEC_ID_PCM_S16LE:
2662 switch (track->audio.bitdepth) {
2663 case 8:
2664 par->codec_id = AV_CODEC_ID_PCM_U8;
2665 break;
2666 case 24:
2667 par->codec_id = AV_CODEC_ID_PCM_S24LE;
2668 break;
2669 case 32:
2670 par->codec_id = AV_CODEC_ID_PCM_S32LE;
2671 break;
2672 }
2673 break;
2674 case AV_CODEC_ID_PCM_F32LE:
2675 if (track->audio.bitdepth == 64)
2676 par->codec_id = AV_CODEC_ID_PCM_F64LE;
2677 break;
2678 case AV_CODEC_ID_AAC:
2679 if (!track->codec_priv.size) {
2680 int profile = matroska_aac_profile(track->codec_id);
2681 int sri = matroska_aac_sri(track->audio.samplerate);
2682
2683 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2684 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2685 if (strstr(track->codec_id, "SBR")) {
2686 sri = matroska_aac_sri(track->audio.out_samplerate);
2687 extradata[2] = 0x56;
2688 extradata[3] = 0xE5;
2689 extradata[4] = 0x80 | (sri << 3);
2690 extradata_size = 5;
2691 } else
2692 extradata_size = 2;
2693 }
2694 break;
2695 case AV_CODEC_ID_ALAC:
2696 if (track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2697 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2698 * Create the "atom size", "tag", and "tag version" fields the
2699 * decoder expects manually. */
2700 ret = ff_alloc_extradata(par, 12 + track->codec_priv.size);
2701 if (ret < 0)
2702 return ret;
2703 AV_WB32(par->extradata, par->extradata_size);
2704 AV_WB32(&par->extradata[4], MKBETAG('a', 'l', 'a', 'c'));
2705 AV_WB32(&par->extradata[8], 0);
2706 memcpy(&par->extradata[12], track->codec_priv.data,
2707 track->codec_priv.size);
2708 }
2709 break;
2710 case AV_CODEC_ID_TTA:
2711 {
2712 uint8_t *ptr;
2713 if (track->audio.channels > UINT16_MAX ||
2714 track->audio.bitdepth > UINT16_MAX) {
2715 av_log(matroska->ctx, AV_LOG_WARNING,
2716 "Too large audio channel number %"PRIu64
2717 " or bitdepth %"PRIu64". Skipping track.\n",
2718 track->audio.channels, track->audio.bitdepth);
2719 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2720 return AVERROR_INVALIDDATA;
2721 else
2722 return SKIP_TRACK;
2723 }
2724 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2725 return AVERROR_INVALIDDATA;
2726 extradata_size = TTA_EXTRADATA_SIZE;
2727 ptr = extradata;
2728 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2729 bytestream_put_le16(&ptr, 1);
2730 bytestream_put_le16(&ptr, track->audio.channels);
2731 bytestream_put_le16(&ptr, track->audio.bitdepth);
2732 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2733 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2734 track->audio.out_samplerate,
2735 AV_TIME_BASE * 1000));
2736 break;
2737 }
2738 case AV_CODEC_ID_RA_144:
2739 track->audio.out_samplerate = 8000;
2740 track->audio.channels = 1;
2741 break;
2742 case AV_CODEC_ID_RA_288:
2743 case AV_CODEC_ID_COOK:
2744 case AV_CODEC_ID_ATRAC3:
2745 case AV_CODEC_ID_SIPR:
2746 {
2747 const uint8_t *ptr = track->codec_priv.data;
2748 int flavor;
2749
2750 if (!track->codec_priv.size)
2751 break;
2752
2753 if (track->codec_priv.size < 46)
2754 return AVERROR_INVALIDDATA;
2755 ptr += 22;
2756 flavor = bytestream_get_be16(&ptr);
2757 track->audio.coded_framesize = bytestream_get_be32(&ptr);
2758 ptr += 12;
2759 track->audio.sub_packet_h = bytestream_get_be16(&ptr);
2760 track->audio.frame_size = bytestream_get_be16(&ptr);
2761 track->audio.sub_packet_size = bytestream_get_be16(&ptr);
2762 if (track->audio.coded_framesize <= 0 ||
2763 track->audio.sub_packet_h <= 0 ||
2764 track->audio.frame_size <= 0)
2765 return AVERROR_INVALIDDATA;
2766
2767 if (par->codec_id == AV_CODEC_ID_RA_288) {
2768 if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2769 != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2770 return AVERROR_INVALIDDATA;
2771 par->block_align = track->audio.coded_framesize;
2772 track->codec_priv.size = 0;
2773 } else {
2774 if (par->codec_id == AV_CODEC_ID_SIPR) {
2775 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2776 if (flavor > 3)
2777 return AVERROR_INVALIDDATA;
2778 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2779 par->bit_rate = sipr_bit_rate[flavor];
2780 } else if (track->audio.sub_packet_size <= 0 ||
2781 track->audio.frame_size % track->audio.sub_packet_size)
2782 return AVERROR_INVALIDDATA;
2783 par->block_align = track->audio.sub_packet_size;
2784 *extradata_offset = 78;
2785 }
2786 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2787 track->audio.frame_size);
2788 if (!track->audio.buf)
2789 return AVERROR(ENOMEM);
2790 break;
2791 }
2792 case AV_CODEC_ID_ATRAC1:
2793 /* ATRAC1 uses a constant frame size.
2794 * Typical ATRAC1 streams are either mono or stereo.
2795 * At most, ATRAC1 was used to store 8 channels of audio. */
2796 if (track->audio.channels > 8)
2797 return AVERROR_INVALIDDATA;
2798 par->block_align = track->audio.channels * 212;
2799 break;
2800 case AV_CODEC_ID_FLAC:
2801 if (track->codec_priv.size) {
2802 ret = matroska_parse_flac(s, track, extradata_offset);
2803 if (ret < 0)
2804 return ret;
2805 }
2806 break;
2807 case AV_CODEC_ID_WAVPACK:
2808 if (track->codec_priv.size < 2) {
2809 av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2810 "in absence of valid CodecPrivate.\n");
2811 extradata_size = WAVPACK_EXTRADATA_SIZE;
2812 AV_WL16(extradata, 0x410);
2813 }
2814 break;
2815 }
2816
2817 if (extradata_size > 0) {
2818 ret = ff_alloc_extradata(par, extradata_size);
2819 if (ret < 0)
2820 return ret;
2821 memcpy(par->extradata, extradata, extradata_size);
2822 }
2823
2824 return 0;
2825 }
2826
2827 /* Performs the generic part of parsing an audio track. */
2828 static int mka_parse_audio(MatroskaTrack *track, AVStream *st,
2829 AVCodecParameters *par,
2830 const MatroskaDemuxContext *matroska,
2831 AVFormatContext *s, int *extradata_offset)
2832 {
2833 FFStream *const sti = ffstream(st);
2834 int ret;
2835
2836 ret = mka_parse_audio_codec(track, par, matroska,
2837 s, extradata_offset);
2838 if (ret)
2839 return ret;
2840
2841 par->codec_type = AVMEDIA_TYPE_AUDIO;
2842 par->sample_rate = track->audio.out_samplerate;
2843 // channel layout may be already set by codec private checks above
2844 if (!av_channel_layout_check(&par->ch_layout)) {
2845 if (track->audio.channels > INT32_MAX)
2846 return AVERROR_PATCHWELCOME;
2847 par->ch_layout.order = AV_CHANNEL_ORDER_UNSPEC;
2848 par->ch_layout.nb_channels = track->audio.channels;
2849 }
2850 if (!par->bits_per_coded_sample)
2851 par->bits_per_coded_sample = track->audio.bitdepth;
2852 if (par->codec_id == AV_CODEC_ID_MP3 ||
2853 par->codec_id == AV_CODEC_ID_MLP ||
2854 par->codec_id == AV_CODEC_ID_TRUEHD)
2855 sti->need_parsing = AVSTREAM_PARSE_FULL;
2856 else if (par->codec_id != AV_CODEC_ID_AAC)
2857 sti->need_parsing = AVSTREAM_PARSE_HEADERS;
2858 if (track->codec_delay > 0) {
2859 par->initial_padding = av_rescale_q(track->codec_delay,
2860 (AVRational){1, 1000000000},
2861 (AVRational){1, par->codec_id == AV_CODEC_ID_OPUS ?
2862 48000 : par->sample_rate});
2863 }
2864 if (track->seek_preroll > 0) {
2865 par->seek_preroll = av_rescale_q(track->seek_preroll,
2866 (AVRational){1, 1000000000},
2867 (AVRational){1, par->sample_rate});
2868 }
2869
2870 return 0;
2871 }
2872
2873 /* Performs the codec-specific part of parsing a video track. */
2874 static int mkv_parse_video_codec(MatroskaTrack *track, AVCodecParameters *par,
2875 const MatroskaDemuxContext *matroska,
2876 int *extradata_offset)
2877 {
2878 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2879 track->codec_priv.size >= 40) {
2880 uint32_t size = AV_RL32A(track->codec_priv.data);
2881 // VFW extradata is padded to an even length, yet
2882 // the size field contains the real length.
2883 if (size & 1 && size == track->codec_priv.size - 1)
2884 --track->codec_priv.size;
2885 track->ms_compat = 1;
2886 par->bits_per_coded_sample = AV_RL16(track->codec_priv.data + 14);
2887 par->codec_tag = AV_RL32(track->codec_priv.data + 16);
2888 par->codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2889 par->codec_tag);
2890 if (!par->codec_id)
2891 par->codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2892 par->codec_tag);
2893 *extradata_offset = 40;
2894 return 0;
2895 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2896 track->codec_priv.size >= 21) {
2897 enum AVCodecID codec_id;
2898 uint32_t fourcc;
2899 int ret = get_qt_codec(track, &fourcc, &codec_id);
2900 if (ret < 0)
2901 return ret;
2902 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2903 fourcc = MKTAG('S','V','Q','3');
2904 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2905 }
2906 par->codec_id = codec_id;
2907 if (codec_id == AV_CODEC_ID_NONE)
2908 av_log(matroska->ctx, AV_LOG_ERROR,
2909 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2910 if (track->codec_priv.size >= 86) {
2911 FFIOContext b;
2912 unsigned bit_depth = AV_RB16(track->codec_priv.data + 82);
2913 ffio_init_read_context(&b, track->codec_priv.data,
2914 track->codec_priv.size);
2915 if (ff_get_qtpalette(codec_id, &b.pub, track->palette)) {
2916 bit_depth &= 0x1F;
2917 track->has_palette = 1;
2918 }
2919 par->bits_per_coded_sample = bit_depth;
2920 }
2921 par->codec_tag = fourcc;
2922 return 0;
2923 }
2924
2925 switch (par->codec_id) {
2926 case AV_CODEC_ID_RV10:
2927 case AV_CODEC_ID_RV20:
2928 case AV_CODEC_ID_RV30:
2929 case AV_CODEC_ID_RV40:
2930 *extradata_offset = 26;
2931 break;
2932 case AV_CODEC_ID_PRORES:
2933 if (track->codec_priv.size == 4)
2934 par->codec_tag = AV_RL32(track->codec_priv.data);
2935 break;
2936 case AV_CODEC_ID_VP9:
2937 /* we don't need any value stored in CodecPrivate.
2938 * make sure that it's not exported as extradata. */
2939 track->codec_priv.size = 0;
2940 break;
2941 }
2942
2943 return 0;
2944 }
2945
2946 /* Performs the generic part of parsing a video track. */
2947 static int mkv_parse_video(MatroskaTrack *track, AVStream *st,
2948 AVCodecParameters *par,
2949 const MatroskaDemuxContext *matroska,
2950 int *extradata_offset)
2951 {
2952 FFStream *const sti = ffstream(st);
2953 MatroskaTrackPlane *planes;
2954 int display_width_mul = 1;
2955 int display_height_mul = 1;
2956 int ret;
2957
2958 if (track->video.color_space.size == 4)
2959 par->codec_tag = AV_RL32(track->video.color_space.data);
2960
2961 ret = mkv_parse_video_codec(track, par, matroska,
2962 extradata_offset);
2963 if (ret < 0)
2964 return ret;
2965
2966 par->codec_type = AVMEDIA_TYPE_VIDEO;
2967 par->width = track->video.pixel_width;
2968 par->height = track->video.pixel_height;
2969
2970 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2971 par->field_order = mkv_field_order(matroska, track->video.field_order);
2972 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2973 par->field_order = AV_FIELD_PROGRESSIVE;
2974
2975 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2976 mkv_stereo_mode_display_mul(track->video.stereo_mode,
2977 &display_width_mul, &display_height_mul);
2978
2979 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2980 if (track->video.display_width && track->video.display_height &&
2981 track->video.display_width != -1 && track->video.display_height != -1 &&
2982 track->video.cropped_height < INT64_MAX / track->video.display_width / display_width_mul &&
2983 track->video.cropped_width < INT64_MAX / track->video.display_height / display_height_mul)
2984 av_reduce(&st->sample_aspect_ratio.num,
2985 &st->sample_aspect_ratio.den,
2986 track->video.cropped_height * track->video.display_width * display_width_mul,
2987 track->video.cropped_width * track->video.display_height * display_height_mul,
2988 INT_MAX);
2989 }
2990 if (track->video.cropped_width != track->video.pixel_width ||
2991 track->video.cropped_height != track->video.pixel_height) {
2992 uint8_t *cropping;
2993 AVPacketSideData *sd = av_packet_side_data_new(&st->codecpar->coded_side_data,
2994 &st->codecpar->nb_coded_side_data,
2995 AV_PKT_DATA_FRAME_CROPPING,
2996 sizeof(uint32_t) * 4, 0);
2997 if (!sd)
2998 return AVERROR(ENOMEM);
2999
3000 cropping = sd->data;
3001 bytestream_put_le32(&cropping, track->video.pixel_cropt);
3002 bytestream_put_le32(&cropping, track->video.pixel_cropb);
3003 bytestream_put_le32(&cropping, track->video.pixel_cropl);
3004 bytestream_put_le32(&cropping, track->video.pixel_cropr);
3005 }
3006 if (par->codec_id != AV_CODEC_ID_HEVC)
3007 sti->need_parsing = AVSTREAM_PARSE_HEADERS;
3008
3009 if (track->default_duration) {
3010 int div = track->default_duration <= INT64_MAX ? 1 : 2;
3011 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
3012 1000000000 / div, track->default_duration / div, 30000);
3013 #if FF_API_R_FRAME_RATE
3014 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
3015 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
3016 st->r_frame_rate = st->avg_frame_rate;
3017 #endif
3018 }
3019
3020 /* export stereo mode flag as metadata tag */
3021 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
3022 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
3023
3024 /* export alpha mode flag as metadata tag */
3025 if (track->video.alpha_mode)
3026 av_dict_set_int(&st->metadata, "alpha_mode", 1, 0);
3027
3028 /* if we have virtual track, mark the real tracks */
3029 planes = track->operation.combine_planes.elem;
3030 for (int j = 0; j < track->operation.combine_planes.nb_elem; j++) {
3031 MatroskaTrack *tracks = matroska->tracks.elem;
3032 char buf[32];
3033 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
3034 continue;
3035 snprintf(buf, sizeof(buf), "%s_%d",
3036 matroska_video_stereo_plane[planes[j].type], st->index);
3037 for (int k = 0; k < matroska->tracks.nb_elem; k++)
3038 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
3039 av_dict_set(&tracks[k].stream->metadata,
3040 "stereo_mode", buf, 0);
3041 break;
3042 }
3043 }
3044 // add stream level stereo3d side data if it is a supported format
3045 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
3046 track->video.stereo_mode != MATROSKA_VIDEO_STEREOMODE_TYPE_ANAGLYPH_CYAN_RED &&
3047 track->video.stereo_mode != MATROSKA_VIDEO_STEREOMODE_TYPE_ANAGLYPH_GREEN_MAG) {
3048 int ret = mkv_stereo3d_conv(st, track->video.stereo_mode);
3049 if (ret < 0)
3050 return ret;
3051 }
3052
3053 ret = mkv_parse_video_color(st, track);
3054 if (ret < 0)
3055 return ret;
3056 ret = mkv_parse_video_projection(st, track, matroska->ctx);
3057 if (ret < 0)
3058 return ret;
3059
3060 return 0;
3061 }
3062
3063 /* Performs the codec-specific part of parsing a subtitle track. */
3064 static int mkv_parse_subtitle_codec(MatroskaTrack *track, AVStream *st,
3065 AVCodecParameters *par,
3066 const MatroskaDemuxContext *matroska)
3067 {
3068 switch (par->codec_id) {
3069 case AV_CODEC_ID_ARIB_CAPTION:
3070 if (track->codec_priv.size == 3) {
3071 int component_tag = track->codec_priv.data[0];
3072 int data_component_id = AV_RB16(track->codec_priv.data + 1);
3073
3074 switch (data_component_id) {
3075 case 0x0008:
3076 // [0x30..0x37] are component tags utilized for
3077 // non-mobile captioning service ("profile A").
3078 if (component_tag >= 0x30 && component_tag <= 0x37) {
3079 par->profile = AV_PROFILE_ARIB_PROFILE_A;
3080 }
3081 break;
3082 case 0x0012:
3083 // component tag 0x87 signifies a mobile/partial reception
3084 // (1seg) captioning service ("profile C").
3085 if (component_tag == 0x87) {
3086 par->profile = AV_PROFILE_ARIB_PROFILE_C;
3087 }
3088 break;
3089 default:
3090 break;
3091 }
3092
3093 if (par->profile == AV_PROFILE_UNKNOWN)
3094 av_log(matroska->ctx, AV_LOG_WARNING,
3095 "Unknown ARIB caption profile utilized: %02x / %04x\n",
3096 component_tag, data_component_id);
3097
3098 track->codec_priv.size = 0;
3099 }
3100 break;
3101 case AV_CODEC_ID_WEBVTT:
3102 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
3103 st->disposition |= AV_DISPOSITION_CAPTIONS;
3104 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
3105 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
3106 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
3107 st->disposition |= AV_DISPOSITION_METADATA;
3108 }
3109 break;
3110 }
3111
3112 return 0;
3113 }
3114
3115 static int matroska_parse_tracks(AVFormatContext *s)
3116 {
3117 MatroskaDemuxContext *matroska = s->priv_data;
3118 MatroskaTrack *tracks = matroska->tracks.elem;
3119 int i, j, ret;
3120
3121 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3122 MatroskaTrack *track = &tracks[i];
3123 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
3124 AVCodecParameters *par;
3125 MatroskaTrackType type;
3126 int extradata_offset = 0;
3127 AVStream *st;
3128 char* key_id_base64 = NULL;
3129
3130 /* Apply some sanity checks. */
3131 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
3132 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
3133 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
3134 track->type != MATROSKA_TRACK_TYPE_METADATA) {
3135 av_log(matroska->ctx, AV_LOG_INFO,
3136 "Unknown or unsupported track type %"PRIu64"\n",
3137 track->type);
3138 continue;
3139 }
3140 if (!track->codec_id)
3141 continue;
3142
3143 if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
3144 || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
3145 || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
3146 || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
3147 ) {
3148 av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
3149 continue;
3150 }
3151
3152 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
3153 isnan(track->audio.samplerate)) {
3154 av_log(matroska->ctx, AV_LOG_WARNING,
3155 "Invalid sample rate %f, defaulting to 8000 instead.\n",
3156 track->audio.samplerate);
3157 track->audio.samplerate = 8000;
3158 }
3159
3160 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
3161 if (!track->default_duration && track->video.frame_rate > 0) {
3162 double default_duration = 1000000000 / track->video.frame_rate;
3163 if (default_duration > UINT64_MAX || default_duration < 0) {
3164 av_log(matroska->ctx, AV_LOG_WARNING,
3165 "Invalid frame rate %e. Cannot calculate default duration.\n",
3166 track->video.frame_rate);
3167 } else {
3168 track->default_duration = default_duration;
3169 }
3170 }
3171 int has_dimensions = track->video.pixel_width || track->video.pixel_height;
3172 if ((matroska->ctx->strict_std_compliance >= FF_COMPLIANCE_STRICT &&
3173 (!track->video.pixel_width || !track->video.pixel_height)) ||
3174 (track->video.pixel_cropl >= INT_MAX - track->video.pixel_cropr ||
3175 track->video.pixel_cropt >= INT_MAX - track->video.pixel_cropb ||
3176 (track->video.pixel_cropl + track->video.pixel_cropr) >= track->video.pixel_width + !has_dimensions ||
3177 (track->video.pixel_cropt + track->video.pixel_cropb) >= track->video.pixel_height + !has_dimensions)) {
3178 av_log(matroska->ctx, AV_LOG_ERROR,
3179 "Invalid coded dimensions %"PRId64"x%"PRId64" [%"PRId64", %"PRId64", %"PRId64", %"PRId64"].\n",
3180 track->video.pixel_width, track->video.pixel_height,
3181 track->video.pixel_cropl, track->video.pixel_cropr,
3182 track->video.pixel_cropt, track->video.pixel_cropb);
3183 return AVERROR_INVALIDDATA;
3184 }
3185 track->video.cropped_width = track->video.pixel_width -
3186 track->video.pixel_cropl - track->video.pixel_cropr;
3187 track->video.cropped_height = track->video.pixel_height -
3188 track->video.pixel_cropt - track->video.pixel_cropb;
3189 if (track->video.display_unit == MATROSKA_VIDEO_DISPLAYUNIT_PIXELS) {
3190 if (track->video.display_width == -1)
3191 track->video.display_width = track->video.cropped_width;
3192 if (track->video.display_height == -1)
3193 track->video.display_height = track->video.cropped_height;
3194 }
3195 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
3196 if (!track->audio.out_samplerate)
3197 track->audio.out_samplerate = track->audio.samplerate;
3198 }
3199 ret = matroska_parse_content_encodings(track->encodings.elem,
3200 track->encodings.nb_elem,
3201 track, &key_id_base64, matroska->ctx);
3202 if (ret < 0)
3203 return ret;
3204
3205 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
3206 if (av_strstart(track->codec_id, ff_mkv_codec_tags[j].str, NULL)) {
3207 codec_id = ff_mkv_codec_tags[j].id;
3208 break;
3209 }
3210 }
3211
3212 st = track->stream = avformat_new_stream(s, NULL);
3213 if (!st) {
3214 av_free(key_id_base64);
3215 return AVERROR(ENOMEM);
3216 }
3217 par = st->codecpar;
3218
3219 par->codec_id = codec_id;
3220
3221 if (track->flag_default)
3222 st->disposition |= AV_DISPOSITION_DEFAULT;
3223 if (track->flag_forced)
3224 st->disposition |= AV_DISPOSITION_FORCED;
3225 if (track->flag_comment)
3226 st->disposition |= AV_DISPOSITION_COMMENT;
3227 if (track->flag_hearingimpaired)
3228 st->disposition |= AV_DISPOSITION_HEARING_IMPAIRED;
3229 if (track->flag_visualimpaired)
3230 st->disposition |= AV_DISPOSITION_VISUAL_IMPAIRED;
3231 if (track->flag_original.count > 0)
3232 st->disposition |= track->flag_original.el.u ? AV_DISPOSITION_ORIGINAL
3233 : AV_DISPOSITION_DUB;
3234
3235 if (key_id_base64) {
3236 /* export encryption key id as base64 metadata tag */
3237 av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
3238 AV_DICT_DONT_STRDUP_VAL);
3239 }
3240
3241 if (strcmp(track->language, "und"))
3242 av_dict_set(&st->metadata, "language", track->language, 0);
3243 av_dict_set(&st->metadata, "title", track->name, 0);
3244
3245 if (track->time_scale < 0.01) {
3246 av_log(matroska->ctx, AV_LOG_WARNING,
3247 "Track TimestampScale too small %f, assuming 1.0.\n",
3248 track->time_scale);
3249 track->time_scale = 1.0;
3250 }
3251
3252 if (matroska->time_scale * track->time_scale > UINT_MAX)
3253 return AVERROR_INVALIDDATA;
3254
3255 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
3256 1000 * 1000 * 1000); /* 64 bit pts in ns */
3257
3258 /* convert the delay from ns to the track timebase */
3259 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
3260 (AVRational){ 1, 1000000000 },
3261 st->time_base);
3262
3263 type = track->type;
3264 if (par->codec_id == AV_CODEC_ID_WEBVTT)
3265 type = MATROSKA_TRACK_TYPE_SUBTITLE;
3266 switch (type) {
3267 case MATROSKA_TRACK_TYPE_AUDIO:
3268 ret = mka_parse_audio(track, st, par, matroska,
3269 s, &extradata_offset);
3270 if (ret < 0)
3271 return ret;
3272 if (ret == SKIP_TRACK)
3273 continue;
3274 break;
3275 case MATROSKA_TRACK_TYPE_VIDEO:
3276 ret = mkv_parse_video(track, st, par, matroska, &extradata_offset);
3277 if (ret < 0)
3278 return ret;
3279 break;
3280 case MATROSKA_TRACK_TYPE_SUBTITLE:
3281 ret = mkv_parse_subtitle_codec(track, st, par, matroska);
3282 if (ret < 0)
3283 return ret;
3284 par->codec_type = AVMEDIA_TYPE_SUBTITLE;
3285
3286 if (track->flag_textdescriptions)
3287 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
3288 break;
3289 }
3290
3291 if (par->codec_id == AV_CODEC_ID_NONE)
3292 av_log(matroska->ctx, AV_LOG_INFO,
3293 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
3294
3295 if (!par->extradata && track->codec_priv.size > extradata_offset) {
3296 const uint8_t *src = track->codec_priv.data + extradata_offset;
3297 unsigned extra_size = track->codec_priv.size - extradata_offset;
3298 ret = ff_alloc_extradata(par, extra_size);
3299 if (ret < 0)
3300 return ret;
3301 memcpy(par->extradata, src, extra_size);
3302 }
3303
3304 ret = mkv_parse_block_addition_mappings(s, st, track);
3305 if (ret < 0)
3306 return ret;
3307 }
3308
3309 return 0;
3310 }
3311
3312 static int matroska_read_header(AVFormatContext *s)
3313 {
3314 FFFormatContext *const si = ffformatcontext(s);
3315 MatroskaDemuxContext *matroska = s->priv_data;
3316 EbmlList *attachments_list = &matroska->attachments;
3317 EbmlList *chapters_list = &matroska->chapters;
3318 MatroskaAttachment *attachments;
3319 MatroskaChapter *chapters;
3320 uint64_t max_start = 0;
3321 int64_t pos;
3322 Ebml ebml = { 0 };
3323 int i, j, res;
3324
3325 matroska->ctx = s;
3326 matroska->cues_parsing_deferred = 1;
3327
3328 /* First read the EBML header. */
3329 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
3330 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
3331 ebml_free(ebml_syntax, &ebml);
3332 return AVERROR_INVALIDDATA;
3333 }
3334 if (ebml.version > EBML_VERSION ||
3335 ebml.max_size > sizeof(uint64_t) ||
3336 ebml.id_length > sizeof(uint32_t) ||
3337 ebml.doctype_version > 3) {
3338 avpriv_report_missing_feature(matroska->ctx,
3339 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
3340 ebml.version, ebml.doctype, ebml.doctype_version);
3341 ebml_free(ebml_syntax, &ebml);
3342 return AVERROR_PATCHWELCOME;
3343 } else if (ebml.doctype_version == 3) {
3344 av_log(matroska->ctx, AV_LOG_WARNING,
3345 "EBML header using unsupported features\n"
3346 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
3347 ebml.version, ebml.doctype, ebml.doctype_version);
3348 }
3349 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
3350 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
3351 break;
3352 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
3353 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
3354 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
3355 ebml_free(ebml_syntax, &ebml);
3356 return AVERROR_INVALIDDATA;
3357 }
3358 }
3359 matroska->is_webm = !strcmp(ebml.doctype, "webm");
3360
3361 ebml_free(ebml_syntax, &ebml);
3362
3363 matroska->pkt = si->parse_pkt;
3364
3365 /* The next thing is a segment. */
3366 pos = avio_tell(matroska->ctx->pb);
3367 res = ebml_parse(matroska, matroska_segments, matroska);
3368 // Try resyncing until we find an EBML_STOP type element.
3369 while (res != 1) {
3370 res = matroska_resync(matroska, pos);
3371 if (res < 0)
3372 return res;
3373 pos = avio_tell(matroska->ctx->pb);
3374 res = ebml_parse(matroska, matroska_segment, matroska);
3375 if (res == AVERROR(EIO)) // EOF is translated to EIO, this exists the loop on EOF
3376 return res;
3377 }
3378 /* Set data_offset as it might be needed later by seek_frame_generic. */
3379 if (matroska->current_id == MATROSKA_ID_CLUSTER)
3380 si->data_offset = avio_tell(matroska->ctx->pb) - 4;
3381 matroska_execute_seekhead(matroska);
3382
3383 if (!matroska->time_scale)
3384 matroska->time_scale = 1000000;
3385 if (isnan(matroska->duration))
3386 matroska->duration = 0;
3387 if (matroska->duration)
3388 matroska->ctx->duration = matroska->duration * matroska->time_scale *
3389 1000 / AV_TIME_BASE;
3390 av_dict_set(&s->metadata, "title", matroska->title, 0);
3391 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
3392
3393 if (matroska->date_utc.size == 8)
3394 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
3395
3396 res = matroska_parse_tracks(s);
3397 if (res < 0)
3398 return res;
3399
3400 attachments = attachments_list->elem;
3401 for (j = 0; j < attachments_list->nb_elem; j++) {
3402 if (!(attachments[j].filename && attachments[j].mime &&
3403 attachments[j].bin.data && attachments[j].bin.size > 0)) {
3404 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
3405 } else {
3406 AVStream *st = avformat_new_stream(s, NULL);
3407 if (!st)
3408 break;
3409 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
3410 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
3411 if (attachments[j].description)
3412 av_dict_set(&st->metadata, "title", attachments[j].description, 0);
3413 st->codecpar->codec_id = AV_CODEC_ID_NONE;
3414
3415 for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3416 if (av_strstart(attachments[j].mime, mkv_image_mime_tags[i].str, NULL)) {
3417 st->codecpar->codec_id = mkv_image_mime_tags[i].id;
3418 break;
3419 }
3420 }
3421
3422 attachments[j].stream = st;
3423
3424 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
3425 res = ff_add_attached_pic(s, st, NULL, &attachments[j].bin.buf, 0);
3426 if (res < 0)
3427 return res;
3428 } else {
3429 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
3430 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
3431 break;
3432 memcpy(st->codecpar->extradata, attachments[j].bin.data,
3433 attachments[j].bin.size);
3434
3435 for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3436 if (av_strstart(attachments[j].mime, mkv_mime_tags[i].str, NULL)) {
3437 st->codecpar->codec_id = mkv_mime_tags[i].id;
3438 break;
3439 }
3440 }
3441 }
3442 }
3443 }
3444
3445 chapters = chapters_list->elem;
3446 for (i = 0; i < chapters_list->nb_elem; i++)
3447 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
3448 (max_start == 0 || chapters[i].start > max_start)) {
3449 chapters[i].chapter =
3450 avpriv_new_chapter(s, chapters[i].uid,
3451 (AVRational) { 1, 1000000000 },
3452 chapters[i].start, chapters[i].end,
3453 chapters[i].title);
3454 max_start = chapters[i].start;
3455 }
3456
3457 matroska_add_index_entries(matroska);
3458
3459 matroska_convert_tags(s);
3460
3461 return 0;
3462 }
3463
3464 /*
3465 * Put one packet in an application-supplied AVPacket struct.
3466 * Returns 0 on success or -1 on failure.
3467 */
3468 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3469 AVPacket *pkt)
3470 {
3471 if (matroska->queue.head) {
3472 MatroskaTrack *tracks = matroska->tracks.elem;
3473 MatroskaTrack *track;
3474
3475 avpriv_packet_list_get(&matroska->queue, pkt);
3476 track = &tracks[pkt->stream_index];
3477 if (track->has_palette) {
3478 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
3479 if (!pal) {
3480 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3481 } else {
3482 memcpy(pal, track->palette, AVPALETTE_SIZE);
3483 }
3484 track->has_palette = 0;
3485 }
3486 return 0;
3487 }
3488
3489 return -1;
3490 }
3491
3492 /*
3493 * Free all packets in our internal queue.
3494 */
3495 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3496 {
3497 avpriv_packet_list_free(&matroska->queue);
3498 }
3499
3500 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3501 int size, int type, AVIOContext *pb,
3502 uint32_t lace_size[256], int *laces)
3503 {
3504 int n;
3505 uint8_t *data = *buf;
3506
3507 if (!type) {
3508 *laces = 1;
3509 lace_size[0] = size;
3510 return 0;
3511 }
3512
3513 if (size <= 0)
3514 return AVERROR_INVALIDDATA;
3515
3516 *laces = *data + 1;
3517 data += 1;
3518 size -= 1;
3519
3520 switch (type) {
3521 case 0x1: /* Xiph lacing */
3522 {
3523 uint8_t temp;
3524 uint32_t total = 0;
3525 for (n = 0; n < *laces - 1; n++) {
3526 lace_size[n] = 0;
3527
3528 do {
3529 if (size <= total)
3530 return AVERROR_INVALIDDATA;
3531 temp = *data;
3532 total += temp;
3533 lace_size[n] += temp;
3534 data += 1;
3535 size -= 1;
3536 } while (temp == 0xff);
3537 }
3538 if (size < total)
3539 return AVERROR_INVALIDDATA;
3540
3541 lace_size[n] = size - total;
3542 break;
3543 }
3544
3545 case 0x2: /* fixed-size lacing */
3546 if (size % (*laces))
3547 return AVERROR_INVALIDDATA;
3548 for (n = 0; n < *laces; n++)
3549 lace_size[n] = size / *laces;
3550 break;
3551
3552 case 0x3: /* EBML lacing */
3553 {
3554 uint64_t num;
3555 uint64_t total;
3556 int offset;
3557
3558 avio_skip(pb, 4);
3559
3560 n = ebml_read_num(matroska, pb, 8, &num, 1);
3561 if (n < 0)
3562 return n;
3563 if (num > INT_MAX)
3564 return AVERROR_INVALIDDATA;
3565
3566 total = lace_size[0] = num;
3567 offset = n;
3568 for (n = 1; n < *laces - 1; n++) {
3569 int64_t snum;
3570 int r;
3571 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3572 if (r < 0)
3573 return r;
3574 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3575 return AVERROR_INVALIDDATA;
3576
3577 lace_size[n] = lace_size[n - 1] + snum;
3578 total += lace_size[n];
3579 offset += r;
3580 }
3581 data += offset;
3582 size -= offset;
3583 if (size < total)
3584 return AVERROR_INVALIDDATA;
3585
3586 lace_size[*laces - 1] = size - total;
3587 break;
3588 }
3589 }
3590
3591 *buf = data;
3592
3593 return 0;
3594 }
3595
3596 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3597 MatroskaTrack *track, AVStream *st,
3598 uint8_t *data, int size, uint64_t timecode,
3599 int64_t pos)
3600 {
3601 const int a = st->codecpar->block_align;
3602 const int sps = track->audio.sub_packet_size;
3603 const int cfs = track->audio.coded_framesize;
3604 const int h = track->audio.sub_packet_h;
3605 const int w = track->audio.frame_size;
3606 int y = track->audio.sub_packet_cnt;
3607 int x;
3608
3609 if (!track->audio.pkt_cnt) {
3610 if (track->audio.sub_packet_cnt == 0)
3611 track->audio.buf_timecode = timecode;
3612 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3613 if (size < cfs * h / 2) {
3614 av_log(matroska->ctx, AV_LOG_ERROR,
3615 "Corrupt int4 RM-style audio packet size\n");
3616 return AVERROR_INVALIDDATA;
3617 }
3618 for (x = 0; x < h / 2; x++)
3619 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3620 data + x * cfs, cfs);
3621 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3622 if (size < w) {
3623 av_log(matroska->ctx, AV_LOG_ERROR,
3624 "Corrupt sipr RM-style audio packet size\n");
3625 return AVERROR_INVALIDDATA;
3626 }
3627 memcpy(track->audio.buf + y * w, data, w);
3628 } else {
3629 if (size < w) {
3630 av_log(matroska->ctx, AV_LOG_ERROR,
3631 "Corrupt generic RM-style audio packet size\n");
3632 return AVERROR_INVALIDDATA;
3633 }
3634 for (x = 0; x < w / sps; x++)
3635 memcpy(track->audio.buf +
3636 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3637 data + x * sps, sps);
3638 }
3639
3640 if (++track->audio.sub_packet_cnt >= h) {
3641 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3642 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3643 track->audio.sub_packet_cnt = 0;
3644 track->audio.pkt_cnt = h * w / a;
3645 }
3646 }
3647
3648 while (track->audio.pkt_cnt) {
3649 int ret;
3650 AVPacket *pkt = matroska->pkt;
3651
3652 ret = av_new_packet(pkt, a);
3653 if (ret < 0) {
3654 return ret;
3655 }
3656 memcpy(pkt->data,
3657 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3658 a);
3659 pkt->pts = track->audio.buf_timecode;
3660 track->audio.buf_timecode = AV_NOPTS_VALUE;
3661 pkt->pos = pos;
3662 pkt->stream_index = st->index;
3663 ret = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
3664 if (ret < 0) {
3665 av_packet_unref(pkt);
3666 return AVERROR(ENOMEM);
3667 }
3668 }
3669
3670 return 0;
3671 }
3672
3673 /* reconstruct full wavpack blocks from mangled matroska ones */
3674 static int matroska_parse_wavpack(MatroskaTrack *track,
3675 uint8_t **data, int *size)
3676 {
3677 uint8_t *dst = NULL;
3678 uint8_t *src = *data;
3679 int dstlen = 0;
3680 int srclen = *size;
3681 uint32_t samples;
3682 uint16_t ver;
3683 int ret, offset = 0;
3684
3685 if (srclen < 12)
3686 return AVERROR_INVALIDDATA;
3687
3688 av_assert1(track->stream->codecpar->extradata_size >= 2);
3689 ver = AV_RL16(track->stream->codecpar->extradata);
3690
3691 samples = AV_RL32(src);
3692 src += 4;
3693 srclen -= 4;
3694
3695 while (srclen >= 8) {
3696 int multiblock;
3697 uint32_t blocksize;
3698 uint8_t *tmp;
3699
3700 uint32_t flags = AV_RL32(src);
3701 uint32_t crc = AV_RL32(src + 4);
3702 src += 8;
3703 srclen -= 8;
3704
3705 multiblock = (flags & 0x1800) != 0x1800;
3706 if (multiblock) {
3707 if (srclen < 4) {
3708 ret = AVERROR_INVALIDDATA;
3709 goto fail;
3710 }
3711 blocksize = AV_RL32(src);
3712 src += 4;
3713 srclen -= 4;
3714 } else
3715 blocksize = srclen;
3716
3717 if (blocksize > srclen) {
3718 ret = AVERROR_INVALIDDATA;
3719 goto fail;
3720 }
3721
3722 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3723 if (!tmp) {
3724 ret = AVERROR(ENOMEM);
3725 goto fail;
3726 }
3727 dst = tmp;
3728 dstlen += blocksize + 32;
3729
3730 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3731 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3732 AV_WL16(dst + offset + 8, ver); // version
3733 AV_WL16(dst + offset + 10, 0); // track/index_no
3734 AV_WL32(dst + offset + 12, 0); // total samples
3735 AV_WL32(dst + offset + 16, 0); // block index
3736 AV_WL32(dst + offset + 20, samples); // number of samples
3737 AV_WL32(dst + offset + 24, flags); // flags
3738 AV_WL32(dst + offset + 28, crc); // crc
3739 memcpy(dst + offset + 32, src, blocksize); // block data
3740
3741 src += blocksize;
3742 srclen -= blocksize;
3743 offset += blocksize + 32;
3744 }
3745
3746 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3747
3748 *data = dst;
3749 *size = dstlen;
3750
3751 return 0;
3752
3753 fail:
3754 av_freep(&dst);
3755 return ret;
3756 }
3757
3758 static int matroska_parse_prores(MatroskaTrack *track,
3759 uint8_t **data, int *size)
3760 {
3761 uint8_t *dst;
3762 int dstlen = *size + 8;
3763
3764 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3765 if (!dst)
3766 return AVERROR(ENOMEM);
3767
3768 AV_WB32(dst, dstlen);
3769 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3770 memcpy(dst + 8, *data, dstlen - 8);
3771 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3772
3773 *data = dst;
3774 *size = dstlen;
3775
3776 return 0;
3777 }
3778
3779 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3780 MatroskaTrack *track,
3781 AVStream *st,
3782 uint8_t *data, int data_len,
3783 uint64_t timecode,
3784 uint64_t duration,
3785 int64_t pos)
3786 {
3787 AVPacket *pkt = matroska->pkt;
3788 uint8_t *id, *settings, *text, *buf;
3789 int id_len, settings_len, text_len;
3790 uint8_t *p, *q;
3791 int err;
3792
3793 if (data_len <= 0)
3794 return AVERROR_INVALIDDATA;
3795
3796 p = data;
3797 q = data + data_len;
3798
3799 id = p;
3800 id_len = -1;
3801 while (p < q) {
3802 if (*p == '\r' || *p == '\n') {
3803 id_len = p - id;
3804 if (*p == '\r')
3805 p++;
3806 break;
3807 }
3808 p++;
3809 }
3810
3811 if (p >= q || *p != '\n')
3812 return AVERROR_INVALIDDATA;
3813 p++;
3814
3815 settings = p;
3816 settings_len = -1;
3817 while (p < q) {
3818 if (*p == '\r' || *p == '\n') {
3819 settings_len = p - settings;
3820 if (*p == '\r')
3821 p++;
3822 break;
3823 }
3824 p++;
3825 }
3826
3827 if (p >= q || *p != '\n')
3828 return AVERROR_INVALIDDATA;
3829 p++;
3830
3831 text = p;
3832 text_len = q - p;
3833 while (text_len > 0) {
3834 const int len = text_len - 1;
3835 const uint8_t c = p[len];
3836 if (c != '\r' && c != '\n')
3837 break;
3838 text_len = len;
3839 }
3840
3841 err = av_new_packet(pkt, text_len);
3842 if (err < 0) {
3843 return err;
3844 }
3845
3846 memcpy(pkt->data, text, text_len);
3847
3848 if (id_len > 0) {
3849 buf = av_packet_new_side_data(pkt,
3850 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3851 id_len);
3852 if (!buf) {
3853 av_packet_unref(pkt);
3854 return AVERROR(ENOMEM);
3855 }
3856 memcpy(buf, id, id_len);
3857 }
3858
3859 if (settings_len > 0) {
3860 buf = av_packet_new_side_data(pkt,
3861 AV_PKT_DATA_WEBVTT_SETTINGS,
3862 settings_len);
3863 if (!buf) {
3864 av_packet_unref(pkt);
3865 return AVERROR(ENOMEM);
3866 }
3867 memcpy(buf, settings, settings_len);
3868 }
3869
3870 // Do we need this for subtitles?
3871 // pkt->flags = AV_PKT_FLAG_KEY;
3872
3873 pkt->stream_index = st->index;
3874 pkt->pts = timecode;
3875
3876 // Do we need this for subtitles?
3877 // pkt->dts = timecode;
3878
3879 pkt->duration = duration;
3880 pkt->pos = pos;
3881
3882 err = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
3883 if (err < 0) {
3884 av_packet_unref(pkt);
3885 return AVERROR(ENOMEM);
3886 }
3887
3888 return 0;
3889 }
3890
3891 static int matroska_parse_block_additional(MatroskaDemuxContext *matroska,
3892 MatroskaTrack *track, AVPacket *pkt,
3893 const uint8_t *data, int size, uint64_t id)
3894 {
3895 const EbmlList *mappings_list = &track->block_addition_mappings;
3896 MatroskaBlockAdditionMapping *mappings = mappings_list->elem, *mapping = NULL;
3897 uint8_t *side_data;
3898 int res;
3899
3900 if (!matroska->is_webm && track->max_block_additional_id && id > track->max_block_additional_id) {
3901 int strict = matroska->ctx->strict_std_compliance >= FF_COMPLIANCE_STRICT;
3902 av_log(matroska->ctx, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
3903 "BlockAddID %"PRIu64" is higher than the reported MaxBlockAdditionID %"PRIu64" "
3904 "for Track with TrackNumber %"PRIu64"\n", id, track->max_block_additional_id,
3905 track->num);
3906 if (strict)
3907 return AVERROR_INVALIDDATA;
3908 }
3909
3910 for (int i = 0; i < mappings_list->nb_elem; i++) {
3911 if (id != mappings[i].value)
3912 continue;
3913 mapping = &mappings[i];
3914 break;
3915 }
3916
3917 if (id != 1 && !matroska->is_webm && !mapping) {
3918 av_log(matroska->ctx, AV_LOG_WARNING, "BlockAddID %"PRIu64" has no mapping. Skipping\n", id);
3919 return 0;
3920 }
3921
3922 if (mapping && mapping->type)
3923 id = mapping->type;
3924
3925 switch (id) {
3926 case MATROSKA_BLOCK_ADD_ID_TYPE_ITU_T_T35: {
3927 GetByteContext bc;
3928 int country_code, provider_code;
3929 int provider_oriented_code, application_identifier;
3930 size_t hdrplus_size;
3931 AVDynamicHDRPlus *hdrplus;
3932
3933 if (size < 6)
3934 break; //ignore
3935
3936 bytestream2_init(&bc, data, size);
3937
3938 /* ITU-T T.35 metadata */
3939 country_code = bytestream2_get_byteu(&bc);
3940 provider_code = bytestream2_get_be16u(&bc);
3941
3942 if (country_code != ITU_T_T35_COUNTRY_CODE_US ||
3943 provider_code != ITU_T_T35_PROVIDER_CODE_SAMSUNG)
3944 break; // ignore
3945
3946 provider_oriented_code = bytestream2_get_be16u(&bc);
3947 application_identifier = bytestream2_get_byteu(&bc);
3948
3949 if (provider_oriented_code != 1 || application_identifier != 4)
3950 break; // ignore
3951
3952 hdrplus = av_dynamic_hdr_plus_alloc(&hdrplus_size);
3953 if (!hdrplus)
3954 return AVERROR(ENOMEM);
3955
3956 if ((res = av_dynamic_hdr_plus_from_t35(hdrplus, bc.buffer,
3957 bytestream2_get_bytes_left(&bc))) < 0 ||
3958 (res = av_packet_add_side_data(pkt, AV_PKT_DATA_DYNAMIC_HDR10_PLUS,
3959 (uint8_t *)hdrplus, hdrplus_size)) < 0) {
3960 av_free(hdrplus);
3961 return res;
3962 }
3963
3964 return 0;
3965 }
3966 default:
3967 break;
3968 }
3969
3970 side_data = av_packet_new_side_data(pkt, AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3971 size + (size_t)8);
3972 if (!side_data)
3973 return AVERROR(ENOMEM);
3974
3975 AV_WB64(side_data, id);
3976 memcpy(side_data + 8, data, size);
3977
3978 return 0;
3979 }
3980
3981 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3982 MatroskaTrack *track, AVStream *st,
3983 AVBufferRef *buf, uint8_t *data, int pkt_size,
3984 uint64_t timecode, uint64_t lace_duration,
3985 int64_t pos, int is_keyframe,
3986 MatroskaBlockMore *blockmore, int nb_blockmore,
3987 int64_t discard_padding)
3988 {
3989 uint8_t *pkt_data = data;
3990 int res = 0;
3991 AVPacket *pkt = matroska->pkt;
3992
3993 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3994 res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3995 if (res < 0) {
3996 av_log(matroska->ctx, AV_LOG_ERROR,
3997 "Error parsing a wavpack block.\n");
3998 goto fail;
3999 }
4000 if (!buf)
4001 av_freep(&data);
4002 buf = NULL;
4003 }
4004
4005 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
4006 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
4007 res = matroska_parse_prores(track, &pkt_data, &pkt_size);
4008 if (res < 0) {
4009 av_log(matroska->ctx, AV_LOG_ERROR,
4010 "Error parsing a prores block.\n");
4011 goto fail;
4012 }
4013 if (!buf)
4014 av_freep(&data);
4015 buf = NULL;
4016 }
4017
4018 if (!pkt_size && !nb_blockmore)
4019 goto no_output;
4020
4021 if (!matroska->is_webm && nb_blockmore && !track->max_block_additional_id) {
4022 int strict = matroska->ctx->strict_std_compliance >= FF_COMPLIANCE_STRICT;
4023 av_log(matroska->ctx, strict ? AV_LOG_ERROR : AV_LOG_WARNING,
4024 "Unexpected BlockAdditions found in a Block from Track with TrackNumber %"PRIu64" "
4025 "where MaxBlockAdditionID is 0\n", track->num);
4026 if (strict) {
4027 res = AVERROR_INVALIDDATA;
4028 goto fail;
4029 }
4030 }
4031
4032 if (!buf)
4033 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
4034 NULL, NULL, 0);
4035 else
4036 pkt->buf = av_buffer_ref(buf);
4037
4038 if (!pkt->buf) {
4039 res = AVERROR(ENOMEM);
4040 goto fail;
4041 }
4042
4043 pkt->data = pkt_data;
4044 pkt->size = pkt_size;
4045 pkt->flags = is_keyframe;
4046 pkt->stream_index = st->index;
4047
4048 for (int i = 0; i < nb_blockmore; i++) {
4049 MatroskaBlockMore *more = &blockmore[i];
4050
4051 if (!more->additional.size)
4052 continue;
4053
4054 res = matroska_parse_block_additional(matroska, track, pkt, more->additional.data,
4055 more->additional.size, more->additional_id);
4056 if (res < 0) {
4057 av_packet_unref(pkt);
4058 return res;
4059 }
4060 }
4061
4062 if (discard_padding) {
4063 uint8_t *side_data = av_packet_new_side_data(pkt,
4064 AV_PKT_DATA_SKIP_SAMPLES,
4065 10);
4066 if (!side_data) {
4067 av_packet_unref(pkt);
4068 return AVERROR(ENOMEM);
4069 }
4070 discard_padding = av_rescale_q(discard_padding,
4071 (AVRational){1, 1000000000},
4072 (AVRational){1, st->codecpar->sample_rate});
4073 if (discard_padding > 0) {
4074 AV_WL32A(side_data + 4, discard_padding);
4075 } else {
4076 AV_WL32A(side_data, -discard_padding);
4077 }
4078 }
4079
4080 if (track->ms_compat)
4081 pkt->dts = timecode;
4082 else
4083 pkt->pts = timecode;
4084 pkt->pos = pos;
4085 pkt->duration = lace_duration;
4086
4087 res = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
4088 if (res < 0) {
4089 av_packet_unref(pkt);
4090 return AVERROR(ENOMEM);
4091 }
4092
4093 return 0;
4094
4095 no_output:
4096 fail:
4097 if (!buf)
4098 av_free(pkt_data);
4099 return res;
4100 }
4101
4102 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
4103 int size, int64_t pos, uint64_t cluster_time,
4104 uint64_t block_duration, int is_keyframe,
4105 MatroskaBlockMore *blockmore, int nb_blockmore,
4106 int64_t cluster_pos, int64_t discard_padding)
4107 {
4108 uint64_t timecode = AV_NOPTS_VALUE;
4109 MatroskaTrack *track;
4110 FFIOContext pb;
4111 int res = 0;
4112 AVStream *st;
4113 int16_t block_time;
4114 uint32_t lace_size[256];
4115 int n, flags, laces = 0;
4116 uint64_t num;
4117 int trust_default_duration;
4118
4119 av_assert1(buf);
4120
4121 ffio_init_read_context(&pb, data, size);
4122
4123 if ((n = ebml_read_num(matroska, &pb.pub, 8, &num, 1)) < 0)
4124 return n;
4125 data += n;
4126 size -= n;
4127
4128 track = matroska_find_track_by_num(matroska, num);
4129 if (!track || size < 3)
4130 return AVERROR_INVALIDDATA;
4131
4132 if (!(st = track->stream)) {
4133 av_log(matroska->ctx, AV_LOG_VERBOSE,
4134 "No stream associated to TrackNumber %"PRIu64". "
4135 "Ignoring Block with this TrackNumber.\n", num);
4136 return 0;
4137 }
4138
4139 if (st->discard >= AVDISCARD_ALL)
4140 return res;
4141 if (block_duration > INT64_MAX)
4142 block_duration = INT64_MAX;
4143
4144 block_time = sign_extend(AV_RB16(data), 16);
4145 data += 2;
4146 flags = *data++;
4147 size -= 3;
4148 if (is_keyframe == -1)
4149 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
4150
4151 if (cluster_time != (uint64_t) -1 &&
4152 (block_time >= 0 || cluster_time >= -block_time)) {
4153 uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
4154 timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
4155 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
4156 timecode < track->end_timecode)
4157 is_keyframe = 0; /* overlapping subtitles are not key frame */
4158 if (is_keyframe) {
4159 ff_reduce_index(matroska->ctx, st->index);
4160 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
4161 AVINDEX_KEYFRAME);
4162 }
4163 }
4164
4165 if (matroska->skip_to_keyframe &&
4166 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
4167 // Compare signed timecodes. Timecode may be negative due to codec delay
4168 // offset. We don't support timestamps greater than int64_t anyway - see
4169 // AVPacket's pts.
4170 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
4171 return res;
4172 if (is_keyframe)
4173 matroska->skip_to_keyframe = 0;
4174 else if (!ffstream(st)->skip_to_keyframe) {
4175 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
4176 matroska->skip_to_keyframe = 0;
4177 }
4178 }
4179
4180 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
4181 &pb.pub, lace_size, &laces);
4182 if (res < 0) {
4183 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
4184 return res;
4185 }
4186
4187 trust_default_duration = track->default_duration != 0;
4188 if (track->audio.samplerate == 8000 && trust_default_duration) {
4189 // If this is needed for more codecs, then add them here
4190 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
4191 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
4192 trust_default_duration = 0;
4193 }
4194 }
4195
4196 if (!block_duration && trust_default_duration)
4197 block_duration = track->default_duration * laces / matroska->time_scale;
4198
4199 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
4200 track->end_timecode =
4201 FFMAX(track->end_timecode, timecode + block_duration);
4202
4203 for (n = 0; n < laces; n++) {
4204 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
4205 uint8_t *out_data = data;
4206 int out_size = lace_size[n];
4207
4208 if (track->needs_decoding) {
4209 res = matroska_decode_buffer(&out_data, &out_size, track);
4210 if (res < 0)
4211 return res;
4212 /* Given that we are here means that out_data is no longer
4213 * owned by buf, so set it to NULL. This depends upon
4214 * zero-length header removal compression being ignored. */
4215 av_assert1(out_data != data);
4216 buf = NULL;
4217 }
4218
4219 if (track->audio.buf) {
4220 res = matroska_parse_rm_audio(matroska, track, st,
4221 out_data, out_size,
4222 timecode, pos);
4223 if (!buf)
4224 av_free(out_data);
4225 if (res)
4226 return res;
4227 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
4228 res = matroska_parse_webvtt(matroska, track, st,
4229 out_data, out_size,
4230 timecode, lace_duration,
4231 pos);
4232 if (!buf)
4233 av_free(out_data);
4234 if (res)
4235 return res;
4236 } else {
4237 res = matroska_parse_frame(matroska, track, st, buf, out_data,
4238 out_size, timecode, lace_duration,
4239 pos, is_keyframe,
4240 blockmore, nb_blockmore,
4241 discard_padding);
4242 if (res)
4243 return res;
4244 }
4245
4246 if (timecode != AV_NOPTS_VALUE)
4247 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
4248 data += lace_size[n];
4249 }
4250
4251 return 0;
4252 }
4253
4254 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
4255 {
4256 MatroskaCluster *cluster = &matroska->current_cluster;
4257 MatroskaBlock *block = &cluster->block;
4258 int res;
4259
4260 av_assert0(matroska->num_levels <= 2U);
4261
4262 if (matroska->num_levels == 1) {
4263 res = ebml_parse(matroska, matroska_segment, NULL);
4264
4265 if (res == 1) {
4266 /* Found a cluster: subtract the size of the ID already read. */
4267 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
4268
4269 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
4270 if (res < 0)
4271 return res;
4272 }
4273 }
4274
4275 if (matroska->num_levels == 2) {
4276 /* We are inside a cluster. */
4277 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
4278
4279 if (res >= 0 && block->bin.size > 0) {
4280 int is_keyframe = block->non_simple ? block->reference.count == 0 : -1;
4281
4282 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
4283 block->bin.size, block->bin.pos,
4284 cluster->timecode, block->duration,
4285 is_keyframe, block->blockmore.elem,
4286 block->blockmore.nb_elem, cluster->pos,
4287 block->discard_padding);
4288 }
4289
4290 ebml_free(matroska_blockgroup, block);
4291 memset(block, 0, sizeof(*block));
4292 } else if (!matroska->num_levels) {
4293 if (!avio_feof(matroska->ctx->pb)) {
4294 avio_r8(matroska->ctx->pb);
4295 if (!avio_feof(matroska->ctx->pb)) {
4296 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
4297 "end of segment.\n");
4298 return AVERROR_INVALIDDATA;
4299 }
4300 }
4301 matroska->done = 1;
4302 return AVERROR_EOF;
4303 }
4304
4305 return res;
4306 }
4307
4308 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
4309 {
4310 MatroskaDemuxContext *matroska = s->priv_data;
4311 int ret = 0;
4312
4313 if (matroska->resync_pos == -1) {
4314 // This can only happen if generic seeking has been used.
4315 matroska->resync_pos = avio_tell(s->pb);
4316 }
4317
4318 while (matroska_deliver_packet(matroska, pkt)) {
4319 if (matroska->done)
4320 return (ret < 0) ? ret : AVERROR_EOF;
4321 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
4322 ret = matroska_resync(matroska, matroska->resync_pos);
4323 }
4324
4325 return 0;
4326 }
4327
4328 static int matroska_read_seek(AVFormatContext *s, int stream_index,
4329 int64_t timestamp, int flags)
4330 {
4331 MatroskaDemuxContext *matroska = s->priv_data;
4332 MatroskaTrack *tracks = NULL;
4333 AVStream *st = s->streams[stream_index];
4334 FFStream *const sti = ffstream(st);
4335 int i, index;
4336
4337 /* Parse the CUES now since we need the index data to seek. */
4338 if (matroska->cues_parsing_deferred > 0) {
4339 matroska->cues_parsing_deferred = 0;
4340 matroska_parse_cues(matroska);
4341 }
4342
4343 if (!sti->nb_index_entries)
4344 goto err;
4345 timestamp = FFMAX(timestamp, sti->index_entries[0].timestamp);
4346
4347 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 ||
4348 index == sti->nb_index_entries - 1) {
4349 matroska_reset_status(matroska, 0, sti->index_entries[sti->nb_index_entries - 1].pos);
4350 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 ||
4351 index == sti->nb_index_entries - 1) {
4352 matroska_clear_queue(matroska);
4353 if (matroska_parse_cluster(matroska) < 0)
4354 break;
4355 }
4356 }
4357
4358 matroska_clear_queue(matroska);
4359 if (index < 0 || (matroska->cues_parsing_deferred < 0 &&
4360 index == sti->nb_index_entries - 1))
4361 goto err;
4362
4363 tracks = matroska->tracks.elem;
4364 for (i = 0; i < matroska->tracks.nb_elem; i++) {
4365 tracks[i].audio.pkt_cnt = 0;
4366 tracks[i].audio.sub_packet_cnt = 0;
4367 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
4368 tracks[i].end_timecode = 0;
4369 }
4370
4371 /* We seek to a level 1 element, so set the appropriate status. */
4372 matroska_reset_status(matroska, 0, sti->index_entries[index].pos);
4373 if (flags & AVSEEK_FLAG_ANY) {
4374 sti->skip_to_keyframe = 0;
4375 matroska->skip_to_timecode = timestamp;
4376 } else {
4377 sti->skip_to_keyframe = 1;
4378 matroska->skip_to_timecode = sti->index_entries[index].timestamp;
4379 }
4380 matroska->skip_to_keyframe = 1;
4381 matroska->done = 0;
4382 avpriv_update_cur_dts(s, st, sti->index_entries[index].timestamp);
4383 return 0;
4384 err:
4385 // slightly hackish but allows proper fallback to
4386 // the generic seeking code.
4387 matroska_reset_status(matroska, 0, -1);
4388 matroska->resync_pos = -1;
4389 matroska_clear_queue(matroska);
4390 sti->skip_to_keyframe =
4391 matroska->skip_to_keyframe = 0;
4392 matroska->done = 0;
4393 return -1;
4394 }
4395
4396 static int matroska_read_close(AVFormatContext *s)
4397 {
4398 MatroskaDemuxContext *matroska = s->priv_data;
4399 MatroskaTrack *tracks = matroska->tracks.elem;
4400 int n;
4401
4402 matroska_clear_queue(matroska);
4403
4404 for (n = 0; n < matroska->tracks.nb_elem; n++)
4405 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
4406 av_freep(&tracks[n].audio.buf);
4407 ebml_free(matroska_segment, matroska);
4408
4409 return 0;
4410 }
4411
4412 #if CONFIG_WEBM_DASH_MANIFEST_DEMUXER
4413 typedef struct {
4414 int64_t start_time_ns;
4415 int64_t end_time_ns;
4416 int64_t start_offset;
4417 int64_t end_offset;
4418 } CueDesc;
4419
4420 /* This function searches all the Cues and returns the CueDesc corresponding to
4421 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
4422 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration or
4423 * if an error occurred.
4424 */
4425 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
4426 MatroskaDemuxContext *matroska = s->priv_data;
4427 FFStream *const sti = ffstream(s->streams[0]);
4428 AVIndexEntry *const index_entries = sti->index_entries;
4429 int nb_index_entries = sti->nb_index_entries;
4430 CueDesc cue_desc;
4431 int i;
4432
4433 if (ts >= (int64_t)(matroska->duration * matroska->time_scale))
4434 return (CueDesc) {-1, -1, -1, -1};
4435 for (i = 1; i < nb_index_entries; i++) {
4436 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
4437 index_entries[i].timestamp * matroska->time_scale > ts) {
4438 break;
4439 }
4440 }
4441 --i;
4442 if (index_entries[i].timestamp > matroska->duration)
4443 return (CueDesc) {-1, -1, -1, -1};
4444 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
4445 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
4446 if (i != nb_index_entries - 1) {
4447 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
4448 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
4449 } else {
4450 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
4451 // FIXME: this needs special handling for files where Cues appear
4452 // before Clusters. the current logic assumes Cues appear after
4453 // Clusters.
4454 cue_desc.end_offset = cues_start - matroska->segment_start;
4455 }
4456 return cue_desc;
4457 }
4458
4459 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
4460 {
4461 MatroskaDemuxContext *matroska = s->priv_data;
4462 AVStream *const st = s->streams[0];
4463 FFStream *const sti = ffstream(st);
4464 uint32_t id = matroska->current_id;
4465 int64_t cluster_pos, before_pos;
4466 int index, rv = 1;
4467
4468 if (sti->nb_index_entries <= 0)
4469 return 0;
4470
4471 // seek to the first cluster using cues.
4472 index = av_index_search_timestamp(st, 0, 0);
4473 if (index < 0)
4474 return 0;
4475 cluster_pos = sti->index_entries[index].pos;
4476 before_pos = avio_tell(s->pb);
4477 while (1) {
4478 uint64_t cluster_id, cluster_length;
4479 int read;
4480 AVPacket *pkt;
4481 avio_seek(s->pb, cluster_pos, SEEK_SET);
4482 // read cluster id and length
4483 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
4484 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
4485 break;
4486 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
4487 if (read < 0)
4488 break;
4489
4490 matroska_reset_status(matroska, 0, cluster_pos);
4491 matroska_clear_queue(matroska);
4492 if (matroska_parse_cluster(matroska) < 0 ||
4493 !matroska->queue.head) {
4494 break;
4495 }
4496 pkt = &matroska->queue.head->pkt;
4497 // 4 + read is the length of the cluster id and the cluster length field.
4498 cluster_pos += 4 + read + cluster_length;
4499 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
4500 rv = 0;
4501 break;
4502 }
4503 }
4504
4505 /* Restore the status after matroska_read_header: */
4506 matroska_reset_status(matroska, id, before_pos);
4507
4508 return rv;
4509 }
4510
4511 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
4512 double min_buffer, double* buffer,
4513 double* sec_to_download, AVFormatContext *s,
4514 int64_t cues_start)
4515 {
4516 double nano_seconds_per_second = 1000000000.0;
4517 double time_sec = time_ns / nano_seconds_per_second;
4518 int rv = 0;
4519 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
4520 int64_t end_time_ns = time_ns + time_to_search_ns;
4521 double sec_downloaded = 0.0;
4522 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
4523 if (desc_curr.start_time_ns == -1)
4524 return -1;
4525 *sec_to_download = 0.0;
4526
4527 // Check for non cue start time.
4528 if (time_ns > desc_curr.start_time_ns) {
4529 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
4530 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
4531 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
4532 double timeToDownload = (cueBytes * 8.0) / bps;
4533
4534 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
4535 *sec_to_download += timeToDownload;
4536
4537 // Check if the search ends within the first cue.
4538 if (desc_curr.end_time_ns >= end_time_ns) {
4539 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4540 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4541 sec_downloaded = percent_to_sub * sec_downloaded;
4542 *sec_to_download = percent_to_sub * *sec_to_download;
4543 }
4544
4545 if ((sec_downloaded + *buffer) <= min_buffer) {
4546 return 1;
4547 }
4548
4549 // Get the next Cue.
4550 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4551 }
4552
4553 while (desc_curr.start_time_ns != -1) {
4554 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
4555 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
4556 double desc_sec = desc_ns / nano_seconds_per_second;
4557 double bits = (desc_bytes * 8.0);
4558 double time_to_download = bits / bps;
4559
4560 sec_downloaded += desc_sec - time_to_download;
4561 *sec_to_download += time_to_download;
4562
4563 if (desc_curr.end_time_ns >= end_time_ns) {
4564 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4565 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4566 sec_downloaded = percent_to_sub * sec_downloaded;
4567 *sec_to_download = percent_to_sub * *sec_to_download;
4568
4569 if ((sec_downloaded + *buffer) <= min_buffer)
4570 rv = 1;
4571 break;
4572 }
4573
4574 if ((sec_downloaded + *buffer) <= min_buffer) {
4575 rv = 1;
4576 break;
4577 }
4578
4579 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4580 }
4581 *buffer = *buffer + sec_downloaded;
4582 return rv;
4583 }
4584
4585 /* This function computes the bandwidth of the WebM file with the help of
4586 * buffer_size_after_time_downloaded() function. Both of these functions are
4587 * adapted from WebM Tools project and are adapted to work with FFmpeg's
4588 * Matroska parsing mechanism.
4589 *
4590 * Returns the bandwidth of the file on success; -1 on error.
4591 * */
4592 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4593 {
4594 MatroskaDemuxContext *matroska = s->priv_data;
4595 AVStream *st = s->streams[0];
4596 FFStream *const sti = ffstream(st);
4597 double bandwidth = 0.0;
4598
4599 for (int i = 0; i < sti->nb_index_entries; i++) {
4600 int64_t prebuffer_ns = 1000000000;
4601 int64_t time_ns = sti->index_entries[i].timestamp * matroska->time_scale;
4602 double nano_seconds_per_second = 1000000000.0;
4603 int64_t prebuffered_ns;
4604 double prebuffer_bytes = 0.0;
4605 int64_t temp_prebuffer_ns = prebuffer_ns;
4606 int64_t pre_bytes, pre_ns;
4607 double pre_sec, prebuffer, bits_per_second;
4608 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4609 // Start with the first Cue.
4610 CueDesc desc_end = desc_beg;
4611
4612 if (time_ns > INT64_MAX - prebuffer_ns)
4613 return -1;
4614 prebuffered_ns = time_ns + prebuffer_ns;
4615
4616 // Figure out how much data we have downloaded for the prebuffer. This will
4617 // be used later to adjust the bits per sample to try.
4618 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4619 // Prebuffered the entire Cue.
4620 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4621 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4622 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4623 }
4624 if (desc_end.start_time_ns == -1) {
4625 // The prebuffer is larger than the duration.
4626 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4627 return -1;
4628 bits_per_second = 0.0;
4629 } else {
4630 // The prebuffer ends in the last Cue. Estimate how much data was
4631 // prebuffered.
4632 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4633 if (desc_end.end_time_ns <= desc_end.start_time_ns ||
4634 desc_end.end_time_ns - (uint64_t)desc_end.start_time_ns > INT64_MAX)
4635 return -1;
4636 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4637 pre_sec = pre_ns / nano_seconds_per_second;
4638 prebuffer_bytes +=
4639 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4640
4641 prebuffer = prebuffer_ns / nano_seconds_per_second;
4642
4643 // Set this to 0.0 in case our prebuffer buffers the entire video.
4644 bits_per_second = 0.0;
4645 do {
4646 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4647 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4648 double desc_sec, calc_bits_per_second, percent, mod_bits_per_second;
4649 if (desc_bytes <= 0 || desc_bytes > INT64_MAX/8)
4650 return -1;
4651
4652 desc_sec = desc_ns / nano_seconds_per_second;
4653 calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4654
4655 // Drop the bps by the percentage of bytes buffered.
4656 percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4657 mod_bits_per_second = calc_bits_per_second * percent;
4658
4659 if (prebuffer < desc_sec) {
4660 double search_sec =
4661 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4662
4663 // Add 1 so the bits per second should be a little bit greater than file
4664 // datarate.
4665 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4666 const double min_buffer = 0.0;
4667 double buffer = prebuffer;
4668 double sec_to_download = 0.0;
4669
4670 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4671 min_buffer, &buffer, &sec_to_download,
4672 s, cues_start);
4673 if (rv < 0) {
4674 return -1;
4675 } else if (rv == 0) {
4676 bits_per_second = (double)(bps);
4677 break;
4678 }
4679 }
4680
4681 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4682 } while (desc_end.start_time_ns != -1);
4683 }
4684 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4685 }
4686 return (int64_t)bandwidth;
4687 }
4688
4689 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4690 {
4691 MatroskaDemuxContext *matroska = s->priv_data;
4692 EbmlList *seekhead_list = &matroska->seekhead;
4693 MatroskaSeekhead *seekhead = seekhead_list->elem;
4694 AVStream *const st = s->streams[0];
4695 FFStream *const sti = ffstream(st);
4696 AVBPrint bprint;
4697 char *buf;
4698 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4699 int i;
4700 int ret;
4701
4702 // determine cues start and end positions
4703 for (i = 0; i < seekhead_list->nb_elem; i++)
4704 if (seekhead[i].id == MATROSKA_ID_CUES)
4705 break;
4706
4707 if (i >= seekhead_list->nb_elem) return -1;
4708
4709 before_pos = avio_tell(matroska->ctx->pb);
4710 cues_start = seekhead[i].pos + matroska->segment_start;
4711 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4712 // cues_end is computed as cues_start + cues_length + length of the
4713 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4714 // cues_end is inclusive and the above sum is reduced by 1.
4715 uint64_t cues_length, cues_id;
4716 int bytes_read;
4717 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4718 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4719 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4720 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4721 if (bytes_read < 0)
4722 return bytes_read;
4723 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4724 }
4725 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4726 if (cues_start == -1 || cues_end == -1) return -1;
4727
4728 // parse the cues
4729 matroska_parse_cues(matroska);
4730
4731 if (!sti->nb_index_entries)
4732 return AVERROR_INVALIDDATA;
4733
4734 // cues start
4735 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4736
4737 // cues end
4738 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4739
4740 // if the file has cues at the start, fix up the init range so that
4741 // it does not include it
4742 if (cues_start <= init_range)
4743 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4744
4745 // bandwidth
4746 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4747 if (bandwidth < 0) return -1;
4748 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4749
4750 // check if all clusters start with key frames
4751 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4752
4753 // Store cue point timestamps as a comma separated list
4754 // for checking subsegment alignment in the muxer.
4755 av_bprint_init(&bprint, 0, AV_BPRINT_SIZE_UNLIMITED);
4756 for (int i = 0; i < sti->nb_index_entries; i++)
4757 av_bprintf(&bprint, "%" PRId64",", sti->index_entries[i].timestamp);
4758 if (!av_bprint_is_complete(&bprint)) {
4759 av_bprint_finalize(&bprint, NULL);
4760 return AVERROR(ENOMEM);
4761 }
4762 // Remove the trailing ','
4763 bprint.str[--bprint.len] = '\0';
4764 if ((ret = av_bprint_finalize(&bprint, &buf)) < 0)
4765 return ret;
4766 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4767 buf, AV_DICT_DONT_STRDUP_VAL);
4768
4769 return 0;
4770 }
4771
4772 static int webm_dash_manifest_read_header(AVFormatContext *s)
4773 {
4774 char *buf;
4775 int ret = matroska_read_header(s);
4776 int64_t init_range;
4777 MatroskaTrack *tracks;
4778 MatroskaDemuxContext *matroska = s->priv_data;
4779 if (ret) {
4780 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4781 return -1;
4782 }
4783 if (!matroska->tracks.nb_elem || !s->nb_streams) {
4784 av_log(s, AV_LOG_ERROR, "No track found\n");
4785 return AVERROR_INVALIDDATA;
4786 }
4787
4788 if (!matroska->is_live) {
4789 buf = av_asprintf("%g", matroska->duration);
4790 if (!buf)
4791 return AVERROR(ENOMEM);
4792 av_dict_set(&s->streams[0]->metadata, DURATION,
4793 buf, AV_DICT_DONT_STRDUP_VAL);
4794
4795 // initialization range
4796 // 5 is the offset of Cluster ID.
4797 init_range = avio_tell(s->pb) - 5;
4798 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4799 }
4800
4801 // basename of the file
4802 buf = strrchr(s->url, '/');
4803 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4804
4805 // track number
4806 tracks = matroska->tracks.elem;
4807 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4808
4809 // parse the cues and populate Cue related fields
4810 if (!matroska->is_live) {
4811 ret = webm_dash_manifest_cues(s, init_range);
4812 if (ret < 0) {
4813 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4814 return ret;
4815 }
4816 }
4817
4818 // use the bandwidth from the command line if it was provided
4819 if (matroska->bandwidth > 0) {
4820 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4821 matroska->bandwidth, 0);
4822 }
4823 return 0;
4824 }
4825
4826 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4827 {
4828 return AVERROR_EOF;
4829 }
4830
4831 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4832 static const AVOption options[] = {
4833 { "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 },
4834 { "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 },
4835 { NULL },
4836 };
4837
4838 static const AVClass webm_dash_class = {
4839 .class_name = "WebM DASH Manifest demuxer",
4840 .item_name = av_default_item_name,
4841 .option = options,
4842 .version = LIBAVUTIL_VERSION_INT,
4843 };
4844
4845 const FFInputFormat ff_webm_dash_manifest_demuxer = {
4846 .p.name = "webm_dash_manifest",
4847 .p.long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4848 .p.priv_class = &webm_dash_class,
4849 .priv_data_size = sizeof(MatroskaDemuxContext),
4850 .flags_internal = FF_INFMT_FLAG_INIT_CLEANUP,
4851 .read_header = webm_dash_manifest_read_header,
4852 .read_packet = webm_dash_manifest_read_packet,
4853 .read_close = matroska_read_close,
4854 };
4855 #endif
4856
4857 const FFInputFormat ff_matroska_demuxer = {
4858 .p.name = "matroska,webm",
4859 .p.long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4860 .p.extensions = "mkv,mk3d,mka,mks,webm",
4861 .p.mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska",
4862 .priv_data_size = sizeof(MatroskaDemuxContext),
4863 .flags_internal = FF_INFMT_FLAG_INIT_CLEANUP,
4864 .read_probe = matroska_probe,
4865 .read_header = matroska_read_header,
4866 .read_packet = matroska_read_packet,
4867 .read_close = matroska_read_close,
4868 .read_seek = matroska_read_seek,
4869 };
FFmpeg git repo