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/intfloat.h"
43 #include "libavutil/intreadwrite.h"
44 #include "libavutil/lzo.h"
45 #include "libavutil/mastering_display_metadata.h"
46 #include "libavutil/mathematics.h"
47 #include "libavutil/opt.h"
48 #include "libavutil/time_internal.h"
49 #include "libavutil/spherical.h"
50
51 #include "libavcodec/bytestream.h"
52 #include "libavcodec/flac.h"
53 #include "libavcodec/mpeg4audio.h"
54 #include "libavcodec/packet_internal.h"
55
56 #include "avformat.h"
57 #include "avio_internal.h"
58 #include "demux.h"
59 #include "dovi_isom.h"
60 #include "internal.h"
61 #include "isom.h"
62 #include "matroska.h"
63 #include "oggdec.h"
64 /* For ff_codec_get_id(). */
65 #include "riff.h"
66 #include "rmsipr.h"
67
68 #if CONFIG_BZLIB
69 #include <bzlib.h>
70 #endif
71 #if CONFIG_ZLIB
72 #include <zlib.h>
73 #endif
74
75 #include "qtpalette.h"
76
77 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
78 #define NEEDS_CHECKING 2 /* Indicates that some error checks
79 * still need to be performed */
80 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
81 * syntax level used for parsing ended. */
82 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
83 * of unkown, potentially damaged data is encountered,
84 * it is considered an error. */
85 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
86 * to this many bytes of unknown data for the
87 * SKIP_THRESHOLD check. */
88
89 typedef enum {
90 EBML_NONE,
91 EBML_UINT,
92 EBML_SINT,
93 EBML_FLOAT,
94 EBML_STR,
95 EBML_UTF8,
96 EBML_BIN,
97 EBML_NEST,
98 EBML_LEVEL1,
99 EBML_STOP,
100 EBML_TYPE_COUNT
101 } EbmlType;
102
103 typedef struct CountedElement {
104 union {
105 uint64_t u;
106 int64_t i;
107 double f;
108 char *s;
109 } el;
110 unsigned count;
111 } CountedElement;
112
113 typedef const struct EbmlSyntax {
114 uint32_t id;
115 uint8_t type;
116 uint8_t is_counted;
117 size_t list_elem_size;
118 size_t data_offset;
119 union {
120 int64_t i;
121 uint64_t u;
122 double f;
123 const char *s;
124 const struct EbmlSyntax *n;
125 } def;
126 } EbmlSyntax;
127
128 typedef struct EbmlList {
129 int nb_elem;
130 unsigned int alloc_elem_size;
131 void *elem;
132 } EbmlList;
133
134 typedef struct EbmlBin {
135 int size;
136 AVBufferRef *buf;
137 uint8_t *data;
138 int64_t pos;
139 } EbmlBin;
140
141 typedef struct Ebml {
142 uint64_t version;
143 uint64_t max_size;
144 uint64_t id_length;
145 char *doctype;
146 uint64_t doctype_version;
147 } Ebml;
148
149 typedef struct MatroskaTrackCompression {
150 uint64_t algo;
151 EbmlBin settings;
152 } MatroskaTrackCompression;
153
154 typedef struct MatroskaTrackEncryption {
155 uint64_t algo;
156 EbmlBin key_id;
157 } MatroskaTrackEncryption;
158
159 typedef struct MatroskaTrackEncoding {
160 uint64_t scope;
161 uint64_t type;
162 MatroskaTrackCompression compression;
163 MatroskaTrackEncryption encryption;
164 } MatroskaTrackEncoding;
165
166 typedef struct MatroskaMasteringMeta {
167 double r_x;
168 double r_y;
169 double g_x;
170 double g_y;
171 double b_x;
172 double b_y;
173 double white_x;
174 double white_y;
175 double max_luminance;
176 CountedElement min_luminance;
177 } MatroskaMasteringMeta;
178
179 typedef struct MatroskaTrackVideoColor {
180 uint64_t matrix_coefficients;
181 uint64_t bits_per_channel;
182 uint64_t chroma_sub_horz;
183 uint64_t chroma_sub_vert;
184 uint64_t cb_sub_horz;
185 uint64_t cb_sub_vert;
186 uint64_t chroma_siting_horz;
187 uint64_t chroma_siting_vert;
188 uint64_t range;
189 uint64_t transfer_characteristics;
190 uint64_t primaries;
191 uint64_t max_cll;
192 uint64_t max_fall;
193 MatroskaMasteringMeta mastering_meta;
194 } MatroskaTrackVideoColor;
195
196 typedef struct MatroskaTrackVideoProjection {
197 uint64_t type;
198 EbmlBin private;
199 double yaw;
200 double pitch;
201 double roll;
202 } MatroskaTrackVideoProjection;
203
204 typedef struct MatroskaTrackVideo {
205 double frame_rate;
206 uint64_t display_width;
207 uint64_t display_height;
208 uint64_t pixel_width;
209 uint64_t pixel_height;
210 EbmlBin color_space;
211 uint64_t display_unit;
212 uint64_t interlaced;
213 uint64_t field_order;
214 uint64_t stereo_mode;
215 uint64_t alpha_mode;
216 EbmlList color;
217 MatroskaTrackVideoProjection projection;
218 } MatroskaTrackVideo;
219
220 typedef struct MatroskaTrackAudio {
221 double samplerate;
222 double out_samplerate;
223 uint64_t bitdepth;
224 uint64_t channels;
225
226 /* real audio header (extracted from extradata) */
227 int coded_framesize;
228 int sub_packet_h;
229 int frame_size;
230 int sub_packet_size;
231 int sub_packet_cnt;
232 int pkt_cnt;
233 uint64_t buf_timecode;
234 uint8_t *buf;
235 } MatroskaTrackAudio;
236
237 typedef struct MatroskaTrackPlane {
238 uint64_t uid;
239 uint64_t type;
240 } MatroskaTrackPlane;
241
242 typedef struct MatroskaTrackOperation {
243 EbmlList combine_planes;
244 } MatroskaTrackOperation;
245
246 typedef struct MatroskaBlockAdditionMapping {
247 uint64_t value;
248 char *name;
249 uint64_t type;
250 EbmlBin extradata;
251 } MatroskaBlockAdditionMapping;
252
253 typedef struct MatroskaTrack {
254 uint64_t num;
255 uint64_t uid;
256 uint64_t type;
257 char *name;
258 char *codec_id;
259 EbmlBin codec_priv;
260 char *language;
261 double time_scale;
262 uint64_t default_duration;
263 uint64_t flag_default;
264 uint64_t flag_forced;
265 uint64_t flag_comment;
266 uint64_t flag_hearingimpaired;
267 uint64_t flag_visualimpaired;
268 uint64_t flag_textdescriptions;
269 CountedElement flag_original;
270 uint64_t seek_preroll;
271 MatroskaTrackVideo video;
272 MatroskaTrackAudio audio;
273 MatroskaTrackOperation operation;
274 EbmlList encodings;
275 uint64_t codec_delay;
276 uint64_t codec_delay_in_track_tb;
277
278 AVStream *stream;
279 int64_t end_timecode;
280 int ms_compat;
281 int needs_decoding;
282 uint64_t max_block_additional_id;
283 EbmlList block_addition_mappings;
284
285 uint32_t palette[AVPALETTE_COUNT];
286 int has_palette;
287 } MatroskaTrack;
288
289 typedef struct MatroskaAttachment {
290 uint64_t uid;
291 char *filename;
292 char *description;
293 char *mime;
294 EbmlBin bin;
295
296 AVStream *stream;
297 } MatroskaAttachment;
298
299 typedef struct MatroskaChapter {
300 uint64_t start;
301 uint64_t end;
302 uint64_t uid;
303 char *title;
304
305 AVChapter *chapter;
306 } MatroskaChapter;
307
308 typedef struct MatroskaIndexPos {
309 uint64_t track;
310 uint64_t pos;
311 } MatroskaIndexPos;
312
313 typedef struct MatroskaIndex {
314 uint64_t time;
315 EbmlList pos;
316 } MatroskaIndex;
317
318 typedef struct MatroskaTag {
319 char *name;
320 char *string;
321 char *lang;
322 uint64_t def;
323 EbmlList sub;
324 } MatroskaTag;
325
326 typedef struct MatroskaTagTarget {
327 char *type;
328 uint64_t typevalue;
329 uint64_t trackuid;
330 uint64_t chapteruid;
331 uint64_t attachuid;
332 } MatroskaTagTarget;
333
334 typedef struct MatroskaTags {
335 MatroskaTagTarget target;
336 EbmlList tag;
337 } MatroskaTags;
338
339 typedef struct MatroskaSeekhead {
340 uint64_t id;
341 uint64_t pos;
342 } MatroskaSeekhead;
343
344 typedef struct MatroskaLevel {
345 uint64_t start;
346 uint64_t length;
347 } MatroskaLevel;
348
349 typedef struct MatroskaBlock {
350 uint64_t duration;
351 CountedElement reference;
352 uint64_t non_simple;
353 EbmlBin bin;
354 uint64_t additional_id;
355 EbmlBin additional;
356 int64_t discard_padding;
357 } MatroskaBlock;
358
359 typedef struct MatroskaCluster {
360 MatroskaBlock block;
361 uint64_t timecode;
362 int64_t pos;
363 } MatroskaCluster;
364
365 typedef struct MatroskaLevel1Element {
366 int64_t pos;
367 uint32_t id;
368 int parsed;
369 } MatroskaLevel1Element;
370
371 typedef struct MatroskaDemuxContext {
372 const AVClass *class;
373 AVFormatContext *ctx;
374
375 /* EBML stuff */
376 MatroskaLevel levels[EBML_MAX_DEPTH];
377 int num_levels;
378 uint32_t current_id;
379 int64_t resync_pos;
380 int unknown_count;
381
382 uint64_t time_scale;
383 double duration;
384 char *title;
385 char *muxingapp;
386 EbmlBin date_utc;
387 EbmlList tracks;
388 EbmlList attachments;
389 EbmlList chapters;
390 EbmlList index;
391 EbmlList tags;
392 EbmlList seekhead;
393
394 /* byte position of the segment inside the stream */
395 int64_t segment_start;
396
397 /* This packet coincides with FFFormatContext.parse_pkt
398 * and is not owned by us. */
399 AVPacket *pkt;
400
401 /* the packet queue */
402 PacketList queue;
403
404 int done;
405
406 /* What to skip before effectively reading a packet. */
407 int skip_to_keyframe;
408 uint64_t skip_to_timecode;
409
410 /* File has a CUES element, but we defer parsing until it is needed. */
411 int cues_parsing_deferred;
412
413 /* Level1 elements and whether they were read yet */
414 MatroskaLevel1Element level1_elems[64];
415 int num_level1_elems;
416
417 MatroskaCluster current_cluster;
418
419 /* WebM DASH Manifest live flag */
420 int is_live;
421
422 /* Bandwidth value for WebM DASH Manifest */
423 int bandwidth;
424 } MatroskaDemuxContext;
425
426 #define CHILD_OF(parent) { .def = { .n = parent } }
427
428 // The following forward declarations need their size because
429 // a tentative definition with internal linkage must not be an
430 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
431 // Removing the sizes breaks MSVC.
432 static EbmlSyntax ebml_syntax[3], matroska_segment[9], matroska_track_video_color[15], matroska_track_video[19],
433 matroska_track[33], matroska_track_encoding[6], matroska_track_encodings[2],
434 matroska_track_combine_planes[2], matroska_track_operation[2], matroska_block_addition_mapping[5], matroska_tracks[2],
435 matroska_attachments[2], matroska_chapter_entry[9], matroska_chapter[6], matroska_chapters[2],
436 matroska_index_entry[3], matroska_index[2], matroska_tag[3], matroska_tags[2], matroska_seekhead[2],
437 matroska_blockadditions[2], matroska_blockgroup[8], matroska_cluster_parsing[8];
438
439 static EbmlSyntax ebml_header[] = {
440 { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
441 { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, 0, offsetof(Ebml, max_size), { .u = 8 } },
442 { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, 0, offsetof(Ebml, id_length), { .u = 4 } },
443 { EBML_ID_DOCTYPE, EBML_STR, 0, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
444 { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
445 { EBML_ID_EBMLVERSION, EBML_NONE },
446 { EBML_ID_DOCTYPEVERSION, EBML_NONE },
447 CHILD_OF(ebml_syntax)
448 };
449
450 static EbmlSyntax ebml_syntax[] = {
451 { EBML_ID_HEADER, EBML_NEST, 0, 0, 0, { .n = ebml_header } },
452 { MATROSKA_ID_SEGMENT, EBML_STOP },
453 { 0 }
454 };
455
456 static EbmlSyntax matroska_info[] = {
457 { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
458 { MATROSKA_ID_DURATION, EBML_FLOAT, 0, 0, offsetof(MatroskaDemuxContext, duration) },
459 { MATROSKA_ID_TITLE, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, title) },
460 { MATROSKA_ID_WRITINGAPP, EBML_NONE },
461 { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, 0, offsetof(MatroskaDemuxContext, muxingapp) },
462 { MATROSKA_ID_DATEUTC, EBML_BIN, 0, 0, offsetof(MatroskaDemuxContext, date_utc) },
463 { MATROSKA_ID_SEGMENTUID, EBML_NONE },
464 CHILD_OF(matroska_segment)
465 };
466
467 static EbmlSyntax matroska_mastering_meta[] = {
468 { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_x) },
469 { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, r_y) },
470 { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_x) },
471 { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, g_y) },
472 { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_x) },
473 { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, b_y) },
474 { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_x) },
475 { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, white_y) },
476 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 1, 0, offsetof(MatroskaMasteringMeta, min_luminance) },
477 { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, 0, offsetof(MatroskaMasteringMeta, max_luminance) },
478 CHILD_OF(matroska_track_video_color)
479 };
480
481 static EbmlSyntax matroska_track_video_color[] = {
482 { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
483 { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u = 0 } },
484 { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz) },
485 { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert) },
486 { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz) },
487 { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert) },
488 { MATROSKA_ID_VIDEOCOLORCHROMASITINGHORZ, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_horz), { .u = MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED } },
489 { MATROSKA_ID_VIDEOCOLORCHROMASITINGVERT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, chroma_siting_vert), { .u = MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED } },
490 { MATROSKA_ID_VIDEOCOLORRANGE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, range), { .u = AVCOL_RANGE_UNSPECIFIED } },
491 { MATROSKA_ID_VIDEOCOLORTRANSFERCHARACTERISTICS, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, transfer_characteristics), { .u = AVCOL_TRC_UNSPECIFIED } },
492 { MATROSKA_ID_VIDEOCOLORPRIMARIES, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, primaries), { .u = AVCOL_PRI_UNSPECIFIED } },
493 { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_cll) },
494 { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoColor, max_fall) },
495 { MATROSKA_ID_VIDEOCOLORMASTERINGMETA, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideoColor, mastering_meta), { .n = matroska_mastering_meta } },
496 CHILD_OF(matroska_track_video)
497 };
498
499 static EbmlSyntax matroska_track_video_projection[] = {
500 { MATROSKA_ID_VIDEOPROJECTIONTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideoProjection, type), { .u = MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR } },
501 { MATROSKA_ID_VIDEOPROJECTIONPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideoProjection, private) },
502 { MATROSKA_ID_VIDEOPROJECTIONPOSEYAW, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, yaw), { .f = 0.0 } },
503 { MATROSKA_ID_VIDEOPROJECTIONPOSEPITCH, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, pitch), { .f = 0.0 } },
504 { MATROSKA_ID_VIDEOPROJECTIONPOSEROLL, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideoProjection, roll), { .f = 0.0 } },
505 CHILD_OF(matroska_track_video)
506 };
507
508 static EbmlSyntax matroska_track_video[] = {
509 { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackVideo, frame_rate) },
510 { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
511 { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
512 { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_width) },
513 { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, pixel_height) },
514 { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, 0, offsetof(MatroskaTrackVideo, color_space) },
515 { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, alpha_mode), { .u = 0 } },
516 { MATROSKA_ID_VIDEOCOLOR, EBML_NEST, 0, sizeof(MatroskaTrackVideoColor), offsetof(MatroskaTrackVideo, color), { .n = matroska_track_video_color } },
517 { MATROSKA_ID_VIDEOPROJECTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackVideo, projection), { .n = matroska_track_video_projection } },
518 { MATROSKA_ID_VIDEOPIXELCROPB, EBML_NONE },
519 { MATROSKA_ID_VIDEOPIXELCROPT, EBML_NONE },
520 { MATROSKA_ID_VIDEOPIXELCROPL, EBML_NONE },
521 { MATROSKA_ID_VIDEOPIXELCROPR, EBML_NONE },
522 { MATROSKA_ID_VIDEODISPLAYUNIT, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, display_unit), { .u= MATROSKA_VIDEO_DISPLAYUNIT_PIXELS } },
523 { MATROSKA_ID_VIDEOFLAGINTERLACED, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, interlaced), { .u = MATROSKA_VIDEO_INTERLACE_FLAG_UNDETERMINED } },
524 { MATROSKA_ID_VIDEOFIELDORDER, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, field_order), { .u = MATROSKA_VIDEO_FIELDORDER_UNDETERMINED } },
525 { MATROSKA_ID_VIDEOSTEREOMODE, EBML_UINT, 0, 0, offsetof(MatroskaTrackVideo, stereo_mode), { .u = MATROSKA_VIDEO_STEREOMODE_TYPE_NB } },
526 { MATROSKA_ID_VIDEOASPECTRATIO, EBML_NONE },
527 CHILD_OF(matroska_track)
528 };
529
530 static EbmlSyntax matroska_track_audio[] = {
531 { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
532 { MATROSKA_ID_AUDIOOUTSAMPLINGFREQ, EBML_FLOAT, 0, 0, offsetof(MatroskaTrackAudio, out_samplerate) },
533 { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, bitdepth) },
534 { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
535 CHILD_OF(matroska_track)
536 };
537
538 static EbmlSyntax matroska_track_encoding_compression[] = {
539 { MATROSKA_ID_ENCODINGCOMPALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackCompression, algo), { .u = MATROSKA_TRACK_ENCODING_COMP_ZLIB } },
540 { MATROSKA_ID_ENCODINGCOMPSETTINGS, EBML_BIN, 0, 0, offsetof(MatroskaTrackCompression, settings) },
541 CHILD_OF(matroska_track_encoding)
542 };
543
544 static EbmlSyntax matroska_track_encoding_encryption[] = {
545 { MATROSKA_ID_ENCODINGENCALGO, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncryption,algo), {.u = 0} },
546 { MATROSKA_ID_ENCODINGENCKEYID, EBML_BIN, 0, 0, offsetof(MatroskaTrackEncryption,key_id) },
547 { MATROSKA_ID_ENCODINGENCAESSETTINGS, EBML_NONE },
548 { MATROSKA_ID_ENCODINGSIGALGO, EBML_NONE },
549 { MATROSKA_ID_ENCODINGSIGHASHALGO, EBML_NONE },
550 { MATROSKA_ID_ENCODINGSIGKEYID, EBML_NONE },
551 { MATROSKA_ID_ENCODINGSIGNATURE, EBML_NONE },
552 CHILD_OF(matroska_track_encoding)
553 };
554 static EbmlSyntax matroska_track_encoding[] = {
555 { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
556 { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
557 { MATROSKA_ID_ENCODINGCOMPRESSION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, compression), { .n = matroska_track_encoding_compression } },
558 { MATROSKA_ID_ENCODINGENCRYPTION, EBML_NEST, 0, 0, offsetof(MatroskaTrackEncoding, encryption), { .n = matroska_track_encoding_encryption } },
559 { MATROSKA_ID_ENCODINGORDER, EBML_NONE },
560 CHILD_OF(matroska_track_encodings)
561 };
562
563 static EbmlSyntax matroska_track_encodings[] = {
564 { MATROSKA_ID_TRACKCONTENTENCODING, EBML_NEST, 0, sizeof(MatroskaTrackEncoding), offsetof(MatroskaTrack, encodings), { .n = matroska_track_encoding } },
565 CHILD_OF(matroska_track)
566 };
567
568 static EbmlSyntax matroska_track_plane[] = {
569 { MATROSKA_ID_TRACKPLANEUID, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,uid) },
570 { MATROSKA_ID_TRACKPLANETYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrackPlane,type) },
571 CHILD_OF(matroska_track_combine_planes)
572 };
573
574 static EbmlSyntax matroska_track_combine_planes[] = {
575 { MATROSKA_ID_TRACKPLANE, EBML_NEST, 0, sizeof(MatroskaTrackPlane), offsetof(MatroskaTrackOperation,combine_planes), {.n = matroska_track_plane} },
576 CHILD_OF(matroska_track_operation)
577 };
578
579 static EbmlSyntax matroska_track_operation[] = {
580 { MATROSKA_ID_TRACKCOMBINEPLANES, EBML_NEST, 0, 0, 0, {.n = matroska_track_combine_planes} },
581 CHILD_OF(matroska_track)
582 };
583
584 static EbmlSyntax matroska_block_addition_mapping[] = {
585 { MATROSKA_ID_BLKADDIDVALUE, EBML_UINT, 0, 0, offsetof(MatroskaBlockAdditionMapping, value) },
586 { MATROSKA_ID_BLKADDIDNAME, EBML_STR, 0, 0, offsetof(MatroskaBlockAdditionMapping, name) },
587 { MATROSKA_ID_BLKADDIDTYPE, EBML_UINT, 0, 0, offsetof(MatroskaBlockAdditionMapping, type) },
588 { MATROSKA_ID_BLKADDIDEXTRADATA, EBML_BIN, 0, 0, offsetof(MatroskaBlockAdditionMapping, extradata) },
589 CHILD_OF(matroska_track)
590 };
591
592 static EbmlSyntax matroska_track[] = {
593 { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, 0, offsetof(MatroskaTrack, num) },
594 { MATROSKA_ID_TRACKNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTrack, name) },
595 { MATROSKA_ID_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, uid) },
596 { MATROSKA_ID_TRACKTYPE, EBML_UINT, 0, 0, offsetof(MatroskaTrack, type) },
597 { MATROSKA_ID_CODECID, EBML_STR, 0, 0, offsetof(MatroskaTrack, codec_id) },
598 { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, 0, offsetof(MatroskaTrack, codec_priv) },
599 { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, codec_delay), { .u = 0 } },
600 { MATROSKA_ID_TRACKLANGUAGE, EBML_STR, 0, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
601 { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, 0, offsetof(MatroskaTrack, default_duration) },
602 { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
603 { MATROSKA_ID_TRACKFLAGCOMMENTARY, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_comment), { .u = 0 } },
604 { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
605 { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
606 { MATROSKA_ID_TRACKFLAGHEARINGIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_hearingimpaired), { .u = 0 } },
607 { MATROSKA_ID_TRACKFLAGVISUALIMPAIRED, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_visualimpaired), { .u = 0 } },
608 { MATROSKA_ID_TRACKFLAGTEXTDESCRIPTIONS, EBML_UINT, 0, 0, offsetof(MatroskaTrack, flag_textdescriptions), { .u = 0 } },
609 { MATROSKA_ID_TRACKFLAGORIGINAL, EBML_UINT, 1, 0, offsetof(MatroskaTrack, flag_original), {.u = 0 } },
610 { MATROSKA_ID_TRACKVIDEO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, video), { .n = matroska_track_video } },
611 { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
612 { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
613 { MATROSKA_ID_TRACKCONTENTENCODINGS, EBML_NEST, 0, 0, 0, { .n = matroska_track_encodings } },
614 { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, 0, offsetof(MatroskaTrack, max_block_additional_id), { .u = 0 } },
615 { MATROSKA_ID_TRACKBLKADDMAPPING, EBML_NEST, 0, sizeof(MatroskaBlockAdditionMapping), offsetof(MatroskaTrack, block_addition_mappings), { .n = matroska_block_addition_mapping } },
616 { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, 0, offsetof(MatroskaTrack, seek_preroll), { .u = 0 } },
617 { MATROSKA_ID_TRACKFLAGENABLED, EBML_NONE },
618 { MATROSKA_ID_TRACKFLAGLACING, EBML_NONE },
619 { MATROSKA_ID_CODECNAME, EBML_NONE },
620 { MATROSKA_ID_CODECDECODEALL, EBML_NONE },
621 { MATROSKA_ID_CODECINFOURL, EBML_NONE },
622 { MATROSKA_ID_CODECDOWNLOADURL, EBML_NONE },
623 { MATROSKA_ID_TRACKMINCACHE, EBML_NONE },
624 { MATROSKA_ID_TRACKMAXCACHE, EBML_NONE },
625 CHILD_OF(matroska_tracks)
626 };
627
628 static EbmlSyntax matroska_tracks[] = {
629 { MATROSKA_ID_TRACKENTRY, EBML_NEST, 0, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
630 CHILD_OF(matroska_segment)
631 };
632
633 static EbmlSyntax matroska_attachment[] = {
634 { MATROSKA_ID_FILEUID, EBML_UINT, 0, 0, offsetof(MatroskaAttachment, uid) },
635 { MATROSKA_ID_FILENAME, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, filename) },
636 { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, 0, offsetof(MatroskaAttachment, mime) },
637 { MATROSKA_ID_FILEDATA, EBML_BIN, 0, 0, offsetof(MatroskaAttachment, bin) },
638 { MATROSKA_ID_FILEDESC, EBML_UTF8, 0, 0, offsetof(MatroskaAttachment, description) },
639 CHILD_OF(matroska_attachments)
640 };
641
642 static EbmlSyntax matroska_attachments[] = {
643 { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, 0, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
644 CHILD_OF(matroska_segment)
645 };
646
647 static EbmlSyntax matroska_chapter_display[] = {
648 { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaChapter, title) },
649 { MATROSKA_ID_CHAPLANG, EBML_NONE },
650 { MATROSKA_ID_CHAPCOUNTRY, EBML_NONE },
651 CHILD_OF(matroska_chapter_entry)
652 };
653
654 static EbmlSyntax matroska_chapter_entry[] = {
655 { MATROSKA_ID_CHAPTERTIMESTART, EBML_UINT, 0, 0, offsetof(MatroskaChapter, start), { .u = AV_NOPTS_VALUE } },
656 { MATROSKA_ID_CHAPTERTIMEEND, EBML_UINT, 0, 0, offsetof(MatroskaChapter, end), { .u = AV_NOPTS_VALUE } },
657 { MATROSKA_ID_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaChapter, uid) },
658 { MATROSKA_ID_CHAPTERDISPLAY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter_display } },
659 { MATROSKA_ID_CHAPTERFLAGHIDDEN, EBML_NONE },
660 { MATROSKA_ID_CHAPTERFLAGENABLED, EBML_NONE },
661 { MATROSKA_ID_CHAPTERPHYSEQUIV, EBML_NONE },
662 { MATROSKA_ID_CHAPTERATOM, EBML_NONE },
663 CHILD_OF(matroska_chapter)
664 };
665
666 static EbmlSyntax matroska_chapter[] = {
667 { MATROSKA_ID_CHAPTERATOM, EBML_NEST, 0, sizeof(MatroskaChapter), offsetof(MatroskaDemuxContext, chapters), { .n = matroska_chapter_entry } },
668 { MATROSKA_ID_EDITIONUID, EBML_NONE },
669 { MATROSKA_ID_EDITIONFLAGHIDDEN, EBML_NONE },
670 { MATROSKA_ID_EDITIONFLAGDEFAULT, EBML_NONE },
671 { MATROSKA_ID_EDITIONFLAGORDERED, EBML_NONE },
672 CHILD_OF(matroska_chapters)
673 };
674
675 static EbmlSyntax matroska_chapters[] = {
676 { MATROSKA_ID_EDITIONENTRY, EBML_NEST, 0, 0, 0, { .n = matroska_chapter } },
677 CHILD_OF(matroska_segment)
678 };
679
680 static EbmlSyntax matroska_index_pos[] = {
681 { MATROSKA_ID_CUETRACK, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, track) },
682 { MATROSKA_ID_CUECLUSTERPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaIndexPos, pos) },
683 { MATROSKA_ID_CUERELATIVEPOSITION,EBML_NONE },
684 { MATROSKA_ID_CUEDURATION, EBML_NONE },
685 { MATROSKA_ID_CUEBLOCKNUMBER, EBML_NONE },
686 CHILD_OF(matroska_index_entry)
687 };
688
689 static EbmlSyntax matroska_index_entry[] = {
690 { MATROSKA_ID_CUETIME, EBML_UINT, 0, 0, offsetof(MatroskaIndex, time) },
691 { MATROSKA_ID_CUETRACKPOSITION, EBML_NEST, 0, sizeof(MatroskaIndexPos), offsetof(MatroskaIndex, pos), { .n = matroska_index_pos } },
692 CHILD_OF(matroska_index)
693 };
694
695 static EbmlSyntax matroska_index[] = {
696 { MATROSKA_ID_POINTENTRY, EBML_NEST, 0, sizeof(MatroskaIndex), offsetof(MatroskaDemuxContext, index), { .n = matroska_index_entry } },
697 CHILD_OF(matroska_segment)
698 };
699
700 static EbmlSyntax matroska_simpletag[] = {
701 { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, 0, offsetof(MatroskaTag, name) },
702 { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, 0, offsetof(MatroskaTag, string) },
703 { MATROSKA_ID_TAGLANG, EBML_STR, 0, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
704 { MATROSKA_ID_TAGDEFAULT, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
705 { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, 0, offsetof(MatroskaTag, def) },
706 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
707 CHILD_OF(matroska_tag)
708 };
709
710 static EbmlSyntax matroska_tagtargets[] = {
711 { MATROSKA_ID_TAGTARGETS_TYPE, EBML_STR, 0, 0, offsetof(MatroskaTagTarget, type) },
712 { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
713 { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, trackuid), { .u = 0 } },
714 { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, chapteruid), { .u = 0 } },
715 { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, 0, offsetof(MatroskaTagTarget, attachuid), { .u = 0 } },
716 CHILD_OF(matroska_tag)
717 };
718
719 static EbmlSyntax matroska_tag[] = {
720 { MATROSKA_ID_SIMPLETAG, EBML_NEST, 0, sizeof(MatroskaTag), offsetof(MatroskaTags, tag), { .n = matroska_simpletag } },
721 { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
722 CHILD_OF(matroska_tags)
723 };
724
725 static EbmlSyntax matroska_tags[] = {
726 { MATROSKA_ID_TAG, EBML_NEST, 0, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
727 CHILD_OF(matroska_segment)
728 };
729
730 static EbmlSyntax matroska_seekhead_entry[] = {
731 { MATROSKA_ID_SEEKID, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, id) },
732 { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
733 CHILD_OF(matroska_seekhead)
734 };
735
736 static EbmlSyntax matroska_seekhead[] = {
737 { MATROSKA_ID_SEEKENTRY, EBML_NEST, 0, sizeof(MatroskaSeekhead), offsetof(MatroskaDemuxContext, seekhead), { .n = matroska_seekhead_entry } },
738 CHILD_OF(matroska_segment)
739 };
740
741 static EbmlSyntax matroska_segment[] = {
742 { MATROSKA_ID_CLUSTER, EBML_STOP },
743 { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, 0, { .n = matroska_info } },
744 { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tracks } },
745 { MATROSKA_ID_ATTACHMENTS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_attachments } },
746 { MATROSKA_ID_CHAPTERS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_chapters } },
747 { MATROSKA_ID_CUES, EBML_LEVEL1, 0, 0, 0, { .n = matroska_index } },
748 { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, 0, { .n = matroska_tags } },
749 { MATROSKA_ID_SEEKHEAD, EBML_LEVEL1, 0, 0, 0, { .n = matroska_seekhead } },
750 { 0 } /* We don't want to go back to level 0, so don't add the parent. */
751 };
752
753 static EbmlSyntax matroska_segments[] = {
754 { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, 0, { .n = matroska_segment } },
755 { 0 }
756 };
757
758 static EbmlSyntax matroska_blockmore[] = {
759 { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
760 { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, 0, offsetof(MatroskaBlock,additional) },
761 CHILD_OF(matroska_blockadditions)
762 };
763
764 static EbmlSyntax matroska_blockadditions[] = {
765 { MATROSKA_ID_BLOCKMORE, EBML_NEST, 0, 0, 0, {.n = matroska_blockmore} },
766 CHILD_OF(matroska_blockgroup)
767 };
768
769 static EbmlSyntax matroska_blockgroup[] = {
770 { MATROSKA_ID_BLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
771 { MATROSKA_ID_BLOCKADDITIONS, EBML_NEST, 0, 0, 0, { .n = matroska_blockadditions} },
772 { MATROSKA_ID_BLOCKDURATION, EBML_UINT, 0, 0, offsetof(MatroskaBlock, duration) },
773 { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, 0, offsetof(MatroskaBlock, discard_padding) },
774 { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 1, 0, offsetof(MatroskaBlock, reference) },
775 { MATROSKA_ID_CODECSTATE, EBML_NONE },
776 { 1, EBML_UINT, 0, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
777 CHILD_OF(matroska_cluster_parsing)
778 };
779
780 // The following array contains SimpleBlock and BlockGroup twice
781 // in order to reuse the other values for matroska_cluster_enter.
782 static EbmlSyntax matroska_cluster_parsing[] = {
783 { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, 0, offsetof(MatroskaBlock, bin) },
784 { MATROSKA_ID_BLOCKGROUP, EBML_NEST, 0, 0, 0, { .n = matroska_blockgroup } },
785 { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, 0, offsetof(MatroskaCluster, timecode) },
786 { MATROSKA_ID_SIMPLEBLOCK, EBML_STOP },
787 { MATROSKA_ID_BLOCKGROUP, EBML_STOP },
788 { MATROSKA_ID_CLUSTERPOSITION, EBML_NONE },
789 { MATROSKA_ID_CLUSTERPREVSIZE, EBML_NONE },
790 CHILD_OF(matroska_segment)
791 };
792
793 static EbmlSyntax matroska_cluster_enter[] = {
794 { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, 0, { .n = &matroska_cluster_parsing[2] } },
795 { 0 }
796 };
797 #undef CHILD_OF
798
799 static const CodecMime mkv_image_mime_tags[] = {
800 {"image/gif" , AV_CODEC_ID_GIF},
801 {"image/jpeg" , AV_CODEC_ID_MJPEG},
802 {"image/png" , AV_CODEC_ID_PNG},
803 {"image/tiff" , AV_CODEC_ID_TIFF},
804
805 {"" , AV_CODEC_ID_NONE}
806 };
807
808 static const CodecMime mkv_mime_tags[] = {
809 {"application/x-truetype-font", AV_CODEC_ID_TTF},
810 {"application/x-font" , AV_CODEC_ID_TTF},
811 {"application/vnd.ms-opentype", AV_CODEC_ID_OTF},
812 {"binary" , AV_CODEC_ID_BIN_DATA},
813
814 {"" , AV_CODEC_ID_NONE}
815 };
816
817 static const char *const matroska_doctypes[] = { "matroska", "webm" };
818
819 /*
820 * This function prepares the status for parsing of level 1 elements.
821 */
matroska_reset_status(MatroskaDemuxContext *matroska, uint32_t id, int64_t position)822 static int matroska_reset_status(MatroskaDemuxContext *matroska,
823 uint32_t id, int64_t position)
824 {
825 int64_t err = 0;
826 if (position >= 0) {
827 err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
828 if (err > 0)
829 err = 0;
830 } else
831 position = avio_tell(matroska->ctx->pb);
832
833 matroska->current_id = id;
834 matroska->num_levels = 1;
835 matroska->unknown_count = 0;
836 matroska->resync_pos = position;
837 if (id)
838 matroska->resync_pos -= (av_log2(id) + 7) / 8;
839
840 return err;
841 }
842
matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)843 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
844 {
845 AVIOContext *pb = matroska->ctx->pb;
846 uint32_t id;
847
848 /* Try to seek to the last position to resync from. If this doesn't work,
849 * we resync from the earliest position available: The start of the buffer. */
850 if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
851 av_log(matroska->ctx, AV_LOG_WARNING,
852 "Seek to desired resync point failed. Seeking to "
853 "earliest point available instead.\n");
854 avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
855 last_pos + 1), SEEK_SET);
856 }
857
858 id = avio_rb32(pb);
859
860 // try to find a toplevel element
861 while (!avio_feof(pb)) {
862 if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
863 id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
864 id == MATROSKA_ID_SEEKHEAD || id == MATROSKA_ID_ATTACHMENTS ||
865 id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
866 /* Prepare the context for parsing of a level 1 element. */
867 matroska_reset_status(matroska, id, -1);
868 /* Given that we are here means that an error has occurred,
869 * so treat the segment as unknown length in order not to
870 * discard valid data that happens to be beyond the designated
871 * end of the segment. */
872 matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
873 return 0;
874 }
875 id = (id << 8) | avio_r8(pb);
876 }
877
878 matroska->done = 1;
879 return pb->error ? pb->error : AVERROR_EOF;
880 }
881
882 /*
883 * Read: an "EBML number", which is defined as a variable-length
884 * array of bytes. The first byte indicates the length by giving a
885 * number of 0-bits followed by a one. The position of the first
886 * "one" bit inside the first byte indicates the length of this
887 * number.
888 * Returns: number of bytes read, < 0 on error
889 */
ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb, int max_size, uint64_t *number, int eof_forbidden)890 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
891 int max_size, uint64_t *number, int eof_forbidden)
892 {
893 int read, n = 1;
894 uint64_t total;
895 int64_t pos;
896
897 /* The first byte tells us the length in bytes - except when it is zero. */
898 total = avio_r8(pb);
899 if (pb->eof_reached)
900 goto err;
901
902 /* get the length of the EBML number */
903 read = 8 - ff_log2_tab[total];
904
905 if (!total || read > max_size) {
906 pos = avio_tell(pb) - 1;
907 if (!total) {
908 av_log(matroska->ctx, AV_LOG_ERROR,
909 "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
910 "of an EBML number\n", pos, pos);
911 } else {
912 av_log(matroska->ctx, AV_LOG_ERROR,
913 "Length %d indicated by an EBML number's first byte 0x%02x "
914 "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
915 read, (uint8_t) total, pos, pos, max_size);
916 }
917 return AVERROR_INVALIDDATA;
918 }
919
920 /* read out length */
921 total ^= 1 << ff_log2_tab[total];
922 while (n++ < read)
923 total = (total << 8) | avio_r8(pb);
924
925 if (pb->eof_reached) {
926 eof_forbidden = 1;
927 goto err;
928 }
929
930 *number = total;
931
932 return read;
933
934 err:
935 pos = avio_tell(pb);
936 if (pb->error) {
937 av_log(matroska->ctx, AV_LOG_ERROR,
938 "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
939 pos, pos);
940 return pb->error;
941 }
942 if (eof_forbidden) {
943 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
944 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
945 return AVERROR(EIO);
946 }
947 return AVERROR_EOF;
948 }
949
950 /**
951 * Read a EBML length value.
952 * This needs special handling for the "unknown length" case which has multiple
953 * encodings.
954 */
ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb, uint64_t *number)955 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
956 uint64_t *number)
957 {
958 int res = ebml_read_num(matroska, pb, 8, number, 1);
959 if (res > 0 && *number + 1 == 1ULL << (7 * res))
960 *number = EBML_UNKNOWN_LENGTH;
961 return res;
962 }
963
964 /*
965 * Read the next element as an unsigned int.
966 * Returns NEEDS_CHECKING unless size == 0.
967 */
ebml_read_uint(AVIOContext *pb, int size, uint64_t default_value, uint64_t *num)968 static int ebml_read_uint(AVIOContext *pb, int size,
969 uint64_t default_value, uint64_t *num)
970 {
971 int n = 0;
972
973 if (size == 0) {
974 *num = default_value;
975 return 0;
976 }
977 /* big-endian ordering; build up number */
978 *num = 0;
979 while (n++ < size)
980 *num = (*num << 8) | avio_r8(pb);
981
982 return NEEDS_CHECKING;
983 }
984
985 /*
986 * Read the next element as a signed int.
987 * Returns NEEDS_CHECKING unless size == 0.
988 */
ebml_read_sint(AVIOContext *pb, int size, int64_t default_value, int64_t *num)989 static int ebml_read_sint(AVIOContext *pb, int size,
990 int64_t default_value, int64_t *num)
991 {
992 int n = 1;
993
994 if (size == 0) {
995 *num = default_value;
996 return 0;
997 } else {
998 *num = sign_extend(avio_r8(pb), 8);
999
1000 /* big-endian ordering; build up number */
1001 while (n++ < size)
1002 *num = ((uint64_t)*num << 8) | avio_r8(pb);
1003 }
1004
1005 return NEEDS_CHECKING;
1006 }
1007
1008 /*
1009 * Read the next element as a float.
1010 * Returns 0 if size == 0, NEEDS_CHECKING or < 0 on obvious failure.
1011 */
ebml_read_float(AVIOContext *pb, int size, double default_value, double *num)1012 static int ebml_read_float(AVIOContext *pb, int size,
1013 double default_value, double *num)
1014 {
1015 if (size == 0) {
1016 *num = default_value;
1017 return 0;
1018 } else if (size == 4) {
1019 *num = av_int2float(avio_rb32(pb));
1020 } else if (size == 8) {
1021 *num = av_int2double(avio_rb64(pb));
1022 } else
1023 return AVERROR_INVALIDDATA;
1024
1025 return NEEDS_CHECKING;
1026 }
1027
1028 /*
1029 * Read the next element as an ASCII string.
1030 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1031 */
ebml_read_ascii(AVIOContext *pb, int size, const char *default_value, char **str)1032 static int ebml_read_ascii(AVIOContext *pb, int size,
1033 const char *default_value, char **str)
1034 {
1035 char *res;
1036 int ret;
1037
1038 if (size == 0 && default_value) {
1039 res = av_strdup(default_value);
1040 if (!res)
1041 return AVERROR(ENOMEM);
1042 } else {
1043 /* EBML strings are usually not 0-terminated, so we allocate one
1044 * byte more, read the string and NUL-terminate it ourselves. */
1045 if (!(res = av_malloc(size + 1)))
1046 return AVERROR(ENOMEM);
1047 if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
1048 av_free(res);
1049 return ret < 0 ? ret : NEEDS_CHECKING;
1050 }
1051 (res)[size] = '\0';
1052 }
1053 av_free(*str);
1054 *str = res;
1055
1056 return 0;
1057 }
1058
1059 /*
1060 * Read the next element as binary data.
1061 * 0 is success, < 0 or NEEDS_CHECKING is failure.
1062 */
ebml_read_binary(AVIOContext *pb, int length, int64_t pos, EbmlBin *bin)1063 static int ebml_read_binary(AVIOContext *pb, int length,
1064 int64_t pos, EbmlBin *bin)
1065 {
1066 int ret;
1067
1068 ret = av_buffer_realloc(&bin->buf, length + AV_INPUT_BUFFER_PADDING_SIZE);
1069 if (ret < 0)
1070 return ret;
1071 memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1072
1073 bin->data = bin->buf->data;
1074 bin->size = length;
1075 bin->pos = pos;
1076 if ((ret = avio_read(pb, bin->data, length)) != length) {
1077 av_buffer_unref(&bin->buf);
1078 bin->data = NULL;
1079 bin->size = 0;
1080 return ret < 0 ? ret : NEEDS_CHECKING;
1081 }
1082
1083 return 0;
1084 }
1085
1086 /*
1087 * Read the next element, but only the header. The contents
1088 * are supposed to be sub-elements which can be read separately.
1089 * 0 is success, < 0 is failure.
1090 */
ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length, int64_t pos)1091 static int ebml_read_master(MatroskaDemuxContext *matroska,
1092 uint64_t length, int64_t pos)
1093 {
1094 MatroskaLevel *level;
1095
1096 if (matroska->num_levels >= EBML_MAX_DEPTH) {
1097 av_log(matroska->ctx, AV_LOG_ERROR,
1098 "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1099 return AVERROR(ENOSYS);
1100 }
1101
1102 level = &matroska->levels[matroska->num_levels++];
1103 level->start = pos;
1104 level->length = length;
1105
1106 return 0;
1107 }
1108
1109 /*
1110 * Read a signed "EBML number"
1111 * Return: number of bytes processed, < 0 on error
1112 */
matroska_ebmlnum_sint(MatroskaDemuxContext *matroska, AVIOContext *pb, int64_t *num)1113 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1114 AVIOContext *pb, int64_t *num)
1115 {
1116 uint64_t unum;
1117 int res;
1118
1119 /* read as unsigned number first */
1120 if ((res = ebml_read_num(matroska, pb, 8, &unum, 1)) < 0)
1121 return res;
1122
1123 /* make signed (weird way) */
1124 *num = unum - ((1LL << (7 * res - 1)) - 1);
1125
1126 return res;
1127 }
1128
1129 static int ebml_parse(MatroskaDemuxContext *matroska,
1130 EbmlSyntax *syntax, void *data);
1131
ebml_parse_id(EbmlSyntax *syntax, uint32_t id)1132 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1133 {
1134 int i;
1135
1136 // Whoever touches this should be aware of the duplication
1137 // existing in matroska_cluster_parsing.
1138 for (i = 0; syntax[i].id; i++)
1139 if (id == syntax[i].id)
1140 break;
1141
1142 return &syntax[i];
1143 }
1144
ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax, void *data)1145 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1146 void *data)
1147 {
1148 int res;
1149
1150 if (data) {
1151 for (int i = 0; syntax[i].id; i++) {
1152 void *dst = (char *)data + syntax[i].data_offset;
1153 switch (syntax[i].type) {
1154 case EBML_UINT:
1155 *(uint64_t *)dst = syntax[i].def.u;
1156 break;
1157 case EBML_SINT:
1158 *(int64_t *) dst = syntax[i].def.i;
1159 break;
1160 case EBML_FLOAT:
1161 *(double *) dst = syntax[i].def.f;
1162 break;
1163 case EBML_STR:
1164 case EBML_UTF8:
1165 // the default may be NULL
1166 if (syntax[i].def.s) {
1167 *(char**)dst = av_strdup(syntax[i].def.s);
1168 if (!*(char**)dst)
1169 return AVERROR(ENOMEM);
1170 }
1171 break;
1172 }
1173 }
1174
1175 if (!matroska->levels[matroska->num_levels - 1].length) {
1176 matroska->num_levels--;
1177 return 0;
1178 }
1179 }
1180
1181 do {
1182 res = ebml_parse(matroska, syntax, data);
1183 } while (!res);
1184
1185 return res == LEVEL_ENDED ? 0 : res;
1186 }
1187
is_ebml_id_valid(uint32_t id)1188 static int is_ebml_id_valid(uint32_t id)
1189 {
1190 // Due to endian nonsense in Matroska, the highest byte with any bits set
1191 // will contain the leading length bit. This bit in turn identifies the
1192 // total byte length of the element by its position within the byte.
1193 unsigned int bits = av_log2(id);
1194 return id && (bits + 7) / 8 == (8 - bits % 8);
1195 }
1196
1197 /*
1198 * Allocate and return the entry for the level1 element with the given ID. If
1199 * an entry already exists, return the existing entry.
1200 */
matroska_find_level1_elem(MatroskaDemuxContext *matroska, uint32_t id, int64_t pos)1201 static MatroskaLevel1Element *matroska_find_level1_elem(MatroskaDemuxContext *matroska,
1202 uint32_t id, int64_t pos)
1203 {
1204 int i;
1205 MatroskaLevel1Element *elem;
1206
1207 if (!is_ebml_id_valid(id))
1208 return NULL;
1209
1210 // Some files link to all clusters; useless.
1211 if (id == MATROSKA_ID_CLUSTER)
1212 return NULL;
1213
1214 // There can be multiple SeekHeads and Tags.
1215 for (i = 0; i < matroska->num_level1_elems; i++) {
1216 if (matroska->level1_elems[i].id == id) {
1217 if (matroska->level1_elems[i].pos == pos ||
1218 id != MATROSKA_ID_SEEKHEAD && id != MATROSKA_ID_TAGS)
1219 return &matroska->level1_elems[i];
1220 }
1221 }
1222
1223 // Only a completely broken file would have more elements.
1224 if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1225 av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements.\n");
1226 return NULL;
1227 }
1228
1229 elem = &matroska->level1_elems[matroska->num_level1_elems++];
1230 *elem = (MatroskaLevel1Element){.id = id};
1231
1232 return elem;
1233 }
1234
ebml_parse(MatroskaDemuxContext *matroska, EbmlSyntax *syntax, void *data)1235 static int ebml_parse(MatroskaDemuxContext *matroska,
1236 EbmlSyntax *syntax, void *data)
1237 {
1238 static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1239 // Forbid unknown-length EBML_NONE elements.
1240 [EBML_NONE] = EBML_UNKNOWN_LENGTH - 1,
1241 [EBML_UINT] = 8,
1242 [EBML_SINT] = 8,
1243 [EBML_FLOAT] = 8,
1244 // max. 16 MB for strings
1245 [EBML_STR] = 0x1000000,
1246 [EBML_UTF8] = 0x1000000,
1247 // max. 256 MB for binary data
1248 [EBML_BIN] = 0x10000000,
1249 // no limits for anything else
1250 };
1251 AVIOContext *pb = matroska->ctx->pb;
1252 uint32_t id;
1253 uint64_t length;
1254 int64_t pos = avio_tell(pb), pos_alt;
1255 int res, update_pos = 1, level_check;
1256 MatroskaLevel1Element *level1_elem;
1257 MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1258
1259 if (!matroska->current_id) {
1260 uint64_t id;
1261 res = ebml_read_num(matroska, pb, 4, &id, 0);
1262 if (res < 0) {
1263 if (pb->eof_reached && res == AVERROR_EOF) {
1264 if (matroska->is_live)
1265 // in live mode, finish parsing if EOF is reached.
1266 return 1;
1267 if (level && pos == avio_tell(pb)) {
1268 if (level->length == EBML_UNKNOWN_LENGTH) {
1269 // Unknown-length levels automatically end at EOF.
1270 matroska->num_levels--;
1271 return LEVEL_ENDED;
1272 } else {
1273 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1274 "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1275 }
1276 }
1277 }
1278 return res;
1279 }
1280 matroska->current_id = id | 1 << 7 * res;
1281 pos_alt = pos + res;
1282 } else {
1283 pos_alt = pos;
1284 pos -= (av_log2(matroska->current_id) + 7) / 8;
1285 }
1286
1287 id = matroska->current_id;
1288
1289 syntax = ebml_parse_id(syntax, id);
1290 if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1291 if (level && level->length == EBML_UNKNOWN_LENGTH) {
1292 // Unknown-length levels end when an element from an upper level
1293 // in the hierarchy is encountered.
1294 while (syntax->def.n) {
1295 syntax = ebml_parse_id(syntax->def.n, id);
1296 if (syntax->id) {
1297 matroska->num_levels--;
1298 return LEVEL_ENDED;
1299 }
1300 };
1301 }
1302
1303 av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1304 "%"PRId64"\n", id, pos);
1305 update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1306 }
1307
1308 if (data) {
1309 data = (char *) data + syntax->data_offset;
1310 if (syntax->list_elem_size) {
1311 EbmlList *list = data;
1312 void *newelem;
1313
1314 if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1315 return AVERROR(ENOMEM);
1316 newelem = av_fast_realloc(list->elem,
1317 &list->alloc_elem_size,
1318 (list->nb_elem + 1) * syntax->list_elem_size);
1319 if (!newelem)
1320 return AVERROR(ENOMEM);
1321 list->elem = newelem;
1322 data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1323 memset(data, 0, syntax->list_elem_size);
1324 list->nb_elem++;
1325 }
1326 }
1327
1328 if (syntax->type != EBML_STOP) {
1329 matroska->current_id = 0;
1330 if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1331 return res;
1332
1333 pos_alt += res;
1334
1335 if (matroska->num_levels > 0) {
1336 if (length != EBML_UNKNOWN_LENGTH &&
1337 level->length != EBML_UNKNOWN_LENGTH) {
1338 uint64_t elem_end = pos_alt + length,
1339 level_end = level->start + level->length;
1340
1341 if (elem_end < level_end) {
1342 level_check = 0;
1343 } else if (elem_end == level_end) {
1344 level_check = LEVEL_ENDED;
1345 } else {
1346 av_log(matroska->ctx, AV_LOG_ERROR,
1347 "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1348 "containing master element ending at 0x%"PRIx64"\n",
1349 pos, elem_end, level_end);
1350 return AVERROR_INVALIDDATA;
1351 }
1352 } else if (length != EBML_UNKNOWN_LENGTH) {
1353 level_check = 0;
1354 } else if (level->length != EBML_UNKNOWN_LENGTH) {
1355 av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1356 "at 0x%"PRIx64" inside parent with finite size\n", pos);
1357 return AVERROR_INVALIDDATA;
1358 } else {
1359 level_check = 0;
1360 if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1361 || syntax->type == EBML_NEST)) {
1362 // According to the current specifications only clusters and
1363 // segments are allowed to be unknown-length. We also accept
1364 // other unknown-length master elements.
1365 av_log(matroska->ctx, AV_LOG_WARNING,
1366 "Found unknown-length element 0x%"PRIX32" other than "
1367 "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1368 "parsing will nevertheless be attempted.\n", id, pos);
1369 update_pos = -1;
1370 }
1371 }
1372 } else
1373 level_check = 0;
1374
1375 if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1376 if (length != EBML_UNKNOWN_LENGTH) {
1377 av_log(matroska->ctx, AV_LOG_ERROR,
1378 "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1379 "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1380 length, max_lengths[syntax->type], id, pos);
1381 } else if (syntax->type != EBML_NONE) {
1382 av_log(matroska->ctx, AV_LOG_ERROR,
1383 "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1384 "unknown length, yet the length of an element of its "
1385 "type must be known.\n", id, pos);
1386 } else {
1387 av_log(matroska->ctx, AV_LOG_ERROR,
1388 "Found unknown-length element with ID 0x%"PRIX32" at "
1389 "pos. 0x%"PRIx64" for which no syntax for parsing is "
1390 "available.\n", id, pos);
1391 }
1392 return AVERROR_INVALIDDATA;
1393 }
1394
1395 if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1396 // Loosing sync will likely manifest itself as encountering unknown
1397 // elements which are not reliably distinguishable from elements
1398 // belonging to future extensions of the format.
1399 // We use a heuristic to detect such situations: If the current
1400 // element is not expected at the current syntax level and there
1401 // were only a few unknown elements in a row, then the element is
1402 // skipped or considered defective based upon the length of the
1403 // current element (i.e. how much would be skipped); if there were
1404 // more than a few skipped elements in a row and skipping the current
1405 // element would lead us more than SKIP_THRESHOLD away from the last
1406 // known good position, then it is inferred that an error occurred.
1407 // The dependency on the number of unknown elements in a row exists
1408 // because the distance to the last known good position is
1409 // automatically big if the last parsed element was big.
1410 // In both cases, each unknown element is considered equivalent to
1411 // UNKNOWN_EQUIV of skipped bytes for the check.
1412 // The whole check is only done for non-seekable output, because
1413 // in this situation skipped data can't simply be rechecked later.
1414 // This is especially important when using unkown length elements
1415 // as the check for whether a child exceeds its containing master
1416 // element is not effective in this situation.
1417 if (update_pos) {
1418 matroska->unknown_count = 0;
1419 } else {
1420 int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1421
1422 if (matroska->unknown_count > 3)
1423 dist += pos_alt - matroska->resync_pos;
1424
1425 if (dist > SKIP_THRESHOLD) {
1426 av_log(matroska->ctx, AV_LOG_ERROR,
1427 "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1428 "length 0x%"PRIx64" considered as invalid data. Last "
1429 "known good position 0x%"PRIx64", %d unknown elements"
1430 " in a row\n", id, pos, length, matroska->resync_pos,
1431 matroska->unknown_count);
1432 return AVERROR_INVALIDDATA;
1433 }
1434 }
1435 }
1436
1437 if (update_pos > 0) {
1438 // We have found an element that is allowed at this place
1439 // in the hierarchy and it passed all checks, so treat the beginning
1440 // of the element as the "last known good" position.
1441 matroska->resync_pos = pos;
1442 }
1443
1444 if (!data && length != EBML_UNKNOWN_LENGTH)
1445 goto skip;
1446 }
1447
1448 switch (syntax->type) {
1449 case EBML_UINT:
1450 res = ebml_read_uint(pb, length, syntax->def.u, data);
1451 break;
1452 case EBML_SINT:
1453 res = ebml_read_sint(pb, length, syntax->def.i, data);
1454 break;
1455 case EBML_FLOAT:
1456 res = ebml_read_float(pb, length, syntax->def.f, data);
1457 break;
1458 case EBML_STR:
1459 case EBML_UTF8:
1460 res = ebml_read_ascii(pb, length, syntax->def.s, data);
1461 break;
1462 case EBML_BIN:
1463 res = ebml_read_binary(pb, length, pos_alt, data);
1464 break;
1465 case EBML_LEVEL1:
1466 case EBML_NEST:
1467 if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1468 return res;
1469 if (id == MATROSKA_ID_SEGMENT)
1470 matroska->segment_start = pos_alt;
1471 if (id == MATROSKA_ID_CUES)
1472 matroska->cues_parsing_deferred = 0;
1473 if (syntax->type == EBML_LEVEL1 &&
1474 (level1_elem = matroska_find_level1_elem(matroska, syntax->id, pos))) {
1475 if (!level1_elem->pos) {
1476 // Zero is not a valid position for a level 1 element.
1477 level1_elem->pos = pos;
1478 } else if (level1_elem->pos != pos)
1479 av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1480 level1_elem->parsed = 1;
1481 }
1482 if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1483 return res;
1484 break;
1485 case EBML_STOP:
1486 return 1;
1487 skip:
1488 default:
1489 if (length) {
1490 int64_t res2;
1491 if (ffio_limit(pb, length) != length) {
1492 // ffio_limit emits its own error message,
1493 // so we don't have to.
1494 return AVERROR(EIO);
1495 }
1496 if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1497 // avio_skip might take us past EOF. We check for this
1498 // by skipping only length - 1 bytes, reading a byte and
1499 // checking the error flags. This is done in order to check
1500 // that the element has been properly skipped even when
1501 // no filesize (that ffio_limit relies on) is available.
1502 avio_r8(pb);
1503 res = NEEDS_CHECKING;
1504 } else
1505 res = res2;
1506 } else
1507 res = 0;
1508 }
1509 if (res) {
1510 if (res == NEEDS_CHECKING) {
1511 if (pb->eof_reached) {
1512 if (pb->error)
1513 res = pb->error;
1514 else
1515 res = AVERROR_EOF;
1516 } else
1517 goto level_check;
1518 }
1519
1520 if (res == AVERROR_INVALIDDATA)
1521 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1522 else if (res == AVERROR(EIO))
1523 av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1524 else if (res == AVERROR_EOF) {
1525 av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1526 res = AVERROR(EIO);
1527 }
1528
1529 return res;
1530 }
1531
1532 level_check:
1533 if (syntax->is_counted && data) {
1534 CountedElement *elem = data;
1535 if (elem->count != UINT_MAX)
1536 elem->count++;
1537 }
1538
1539 if (level_check == LEVEL_ENDED && matroska->num_levels) {
1540 level = &matroska->levels[matroska->num_levels - 1];
1541 pos = avio_tell(pb);
1542
1543 // Given that pos >= level->start no check for
1544 // level->length != EBML_UNKNOWN_LENGTH is necessary.
1545 while (matroska->num_levels && pos == level->start + level->length) {
1546 matroska->num_levels--;
1547 level--;
1548 }
1549 }
1550
1551 return level_check;
1552 }
1553
ebml_free(EbmlSyntax *syntax, void *data)1554 static void ebml_free(EbmlSyntax *syntax, void *data)
1555 {
1556 int i, j;
1557 for (i = 0; syntax[i].id; i++) {
1558 void *data_off = (char *) data + syntax[i].data_offset;
1559 switch (syntax[i].type) {
1560 case EBML_STR:
1561 case EBML_UTF8:
1562 av_freep(data_off);
1563 break;
1564 case EBML_BIN:
1565 av_buffer_unref(&((EbmlBin *) data_off)->buf);
1566 break;
1567 case EBML_LEVEL1:
1568 case EBML_NEST:
1569 if (syntax[i].list_elem_size) {
1570 EbmlList *list = data_off;
1571 char *ptr = list->elem;
1572 for (j = 0; j < list->nb_elem;
1573 j++, ptr += syntax[i].list_elem_size)
1574 ebml_free(syntax[i].def.n, ptr);
1575 av_freep(&list->elem);
1576 list->nb_elem = 0;
1577 list->alloc_elem_size = 0;
1578 } else
1579 ebml_free(syntax[i].def.n, data_off);
1580 default:
1581 break;
1582 }
1583 }
1584 }
1585
1586 /*
1587 * Autodetecting...
1588 */
matroska_probe(const AVProbeData *p)1589 static int matroska_probe(const AVProbeData *p)
1590 {
1591 uint64_t total = 0;
1592 int len_mask = 0x80, size = 1, n = 1, i;
1593
1594 /* EBML header? */
1595 if (AV_RB32(p->buf) != EBML_ID_HEADER)
1596 return 0;
1597
1598 /* length of header */
1599 total = p->buf[4];
1600 while (size <= 8 && !(total & len_mask)) {
1601 size++;
1602 len_mask >>= 1;
1603 }
1604 if (size > 8)
1605 return 0;
1606 total &= (len_mask - 1);
1607 while (n < size)
1608 total = (total << 8) | p->buf[4 + n++];
1609
1610 if (total + 1 == 1ULL << (7 * size)){
1611 /* Unknown-length header - simply parse the whole buffer. */
1612 total = p->buf_size - 4 - size;
1613 } else {
1614 /* Does the probe data contain the whole header? */
1615 if (p->buf_size < 4 + size + total)
1616 return 0;
1617 }
1618
1619 /* The header should contain a known document type. For now,
1620 * we don't parse the whole header but simply check for the
1621 * availability of that array of characters inside the header.
1622 * Not fully fool-proof, but good enough. */
1623 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1624 size_t probelen = strlen(matroska_doctypes[i]);
1625 if (total < probelen)
1626 continue;
1627 for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1628 if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1629 return AVPROBE_SCORE_MAX;
1630 }
1631
1632 // probably valid EBML header but no recognized doctype
1633 return AVPROBE_SCORE_EXTENSION;
1634 }
1635
matroska_find_track_by_num(MatroskaDemuxContext *matroska, uint64_t num)1636 static MatroskaTrack *matroska_find_track_by_num(MatroskaDemuxContext *matroska,
1637 uint64_t num)
1638 {
1639 MatroskaTrack *tracks = matroska->tracks.elem;
1640 int i;
1641
1642 for (i = 0; i < matroska->tracks.nb_elem; i++)
1643 if (tracks[i].num == num)
1644 return &tracks[i];
1645
1646 av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1647 return NULL;
1648 }
1649
matroska_decode_buffer(uint8_t **buf, int *buf_size, MatroskaTrack *track)1650 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1651 MatroskaTrack *track)
1652 {
1653 MatroskaTrackEncoding *encodings = track->encodings.elem;
1654 uint8_t *data = *buf;
1655 int isize = *buf_size;
1656 uint8_t *pkt_data = NULL;
1657 uint8_t av_unused *newpktdata;
1658 int pkt_size = isize;
1659 int result = 0;
1660 #if CONFIG_LZO
1661 int olen;
1662 #endif
1663
1664 if (pkt_size >= 10000000U)
1665 return AVERROR_INVALIDDATA;
1666
1667 switch (encodings[0].compression.algo) {
1668 case MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP:
1669 {
1670 int header_size = encodings[0].compression.settings.size;
1671 uint8_t *header = encodings[0].compression.settings.data;
1672
1673 if (header_size && !header) {
1674 av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1675 return -1;
1676 }
1677
1678 if (!header_size)
1679 return 0;
1680
1681 pkt_size = isize + header_size;
1682 pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1683 if (!pkt_data)
1684 return AVERROR(ENOMEM);
1685
1686 memcpy(pkt_data, header, header_size);
1687 memcpy(pkt_data + header_size, data, isize);
1688 break;
1689 }
1690 #if CONFIG_LZO
1691 case MATROSKA_TRACK_ENCODING_COMP_LZO:
1692 do {
1693 int insize = isize;
1694 olen = pkt_size *= 3;
1695 newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1696 + AV_INPUT_BUFFER_PADDING_SIZE);
1697 if (!newpktdata) {
1698 result = AVERROR(ENOMEM);
1699 goto failed;
1700 }
1701 pkt_data = newpktdata;
1702 result = av_lzo1x_decode(pkt_data, &olen, data, &insize);
1703 } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1704 if (result) {
1705 result = AVERROR_INVALIDDATA;
1706 goto failed;
1707 }
1708 pkt_size -= olen;
1709 break;
1710 #endif
1711 #if CONFIG_ZLIB
1712 case MATROSKA_TRACK_ENCODING_COMP_ZLIB:
1713 {
1714 z_stream zstream = { 0 };
1715 if (!pkt_size || inflateInit(&zstream) != Z_OK)
1716 return -1;
1717 zstream.next_in = data;
1718 zstream.avail_in = isize;
1719 do {
1720 pkt_size *= 3;
1721 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1722 if (!newpktdata) {
1723 inflateEnd(&zstream);
1724 result = AVERROR(ENOMEM);
1725 goto failed;
1726 }
1727 pkt_data = newpktdata;
1728 zstream.avail_out = pkt_size - zstream.total_out;
1729 zstream.next_out = pkt_data + zstream.total_out;
1730 result = inflate(&zstream, Z_NO_FLUSH);
1731 } while (result == Z_OK && pkt_size < 10000000);
1732 pkt_size = zstream.total_out;
1733 inflateEnd(&zstream);
1734 if (result != Z_STREAM_END) {
1735 if (result == Z_MEM_ERROR)
1736 result = AVERROR(ENOMEM);
1737 else
1738 result = AVERROR_INVALIDDATA;
1739 goto failed;
1740 }
1741 break;
1742 }
1743 #endif
1744 #if CONFIG_BZLIB
1745 case MATROSKA_TRACK_ENCODING_COMP_BZLIB:
1746 {
1747 bz_stream bzstream = { 0 };
1748 if (!pkt_size || BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1749 return -1;
1750 bzstream.next_in = data;
1751 bzstream.avail_in = isize;
1752 do {
1753 pkt_size *= 3;
1754 newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1755 if (!newpktdata) {
1756 BZ2_bzDecompressEnd(&bzstream);
1757 result = AVERROR(ENOMEM);
1758 goto failed;
1759 }
1760 pkt_data = newpktdata;
1761 bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1762 bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1763 result = BZ2_bzDecompress(&bzstream);
1764 } while (result == BZ_OK && pkt_size < 10000000);
1765 pkt_size = bzstream.total_out_lo32;
1766 BZ2_bzDecompressEnd(&bzstream);
1767 if (result != BZ_STREAM_END) {
1768 if (result == BZ_MEM_ERROR)
1769 result = AVERROR(ENOMEM);
1770 else
1771 result = AVERROR_INVALIDDATA;
1772 goto failed;
1773 }
1774 break;
1775 }
1776 #endif
1777 default:
1778 return AVERROR_INVALIDDATA;
1779 }
1780
1781 memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1782
1783 *buf = pkt_data;
1784 *buf_size = pkt_size;
1785 return 0;
1786
1787 failed:
1788 av_free(pkt_data);
1789 return result;
1790 }
1791
matroska_convert_tag(AVFormatContext *s, EbmlList *list, AVDictionary **metadata, char *prefix)1792 static void matroska_convert_tag(AVFormatContext *s, EbmlList *list,
1793 AVDictionary **metadata, char *prefix)
1794 {
1795 MatroskaTag *tags = list->elem;
1796 char key[1024];
1797 int i;
1798
1799 for (i = 0; i < list->nb_elem; i++) {
1800 const char *lang = tags[i].lang &&
1801 strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1802
1803 if (!tags[i].name) {
1804 av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1805 continue;
1806 }
1807 if (prefix)
1808 snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1809 else
1810 av_strlcpy(key, tags[i].name, sizeof(key));
1811 if (tags[i].def || !lang) {
1812 av_dict_set(metadata, key, tags[i].string, 0);
1813 if (tags[i].sub.nb_elem)
1814 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1815 }
1816 if (lang) {
1817 av_strlcat(key, "-", sizeof(key));
1818 av_strlcat(key, lang, sizeof(key));
1819 av_dict_set(metadata, key, tags[i].string, 0);
1820 if (tags[i].sub.nb_elem)
1821 matroska_convert_tag(s, &tags[i].sub, metadata, key);
1822 }
1823 }
1824 ff_metadata_conv(metadata, NULL, ff_mkv_metadata_conv);
1825 }
1826
matroska_convert_tags(AVFormatContext *s)1827 static void matroska_convert_tags(AVFormatContext *s)
1828 {
1829 MatroskaDemuxContext *matroska = s->priv_data;
1830 MatroskaTags *tags = matroska->tags.elem;
1831 int i, j;
1832
1833 for (i = 0; i < matroska->tags.nb_elem; i++) {
1834 if (tags[i].target.attachuid) {
1835 MatroskaAttachment *attachment = matroska->attachments.elem;
1836 int found = 0;
1837 for (j = 0; j < matroska->attachments.nb_elem; j++) {
1838 if (attachment[j].uid == tags[i].target.attachuid &&
1839 attachment[j].stream) {
1840 matroska_convert_tag(s, &tags[i].tag,
1841 &attachment[j].stream->metadata, NULL);
1842 found = 1;
1843 }
1844 }
1845 if (!found) {
1846 av_log(s, AV_LOG_WARNING,
1847 "The tags at index %d refer to a "
1848 "non-existent attachment %"PRId64".\n",
1849 i, tags[i].target.attachuid);
1850 }
1851 } else if (tags[i].target.chapteruid) {
1852 MatroskaChapter *chapter = matroska->chapters.elem;
1853 int found = 0;
1854 for (j = 0; j < matroska->chapters.nb_elem; j++) {
1855 if (chapter[j].uid == tags[i].target.chapteruid &&
1856 chapter[j].chapter) {
1857 matroska_convert_tag(s, &tags[i].tag,
1858 &chapter[j].chapter->metadata, NULL);
1859 found = 1;
1860 }
1861 }
1862 if (!found) {
1863 av_log(s, AV_LOG_WARNING,
1864 "The tags at index %d refer to a non-existent chapter "
1865 "%"PRId64".\n",
1866 i, tags[i].target.chapteruid);
1867 }
1868 } else if (tags[i].target.trackuid) {
1869 MatroskaTrack *track = matroska->tracks.elem;
1870 int found = 0;
1871 for (j = 0; j < matroska->tracks.nb_elem; j++) {
1872 if (track[j].uid == tags[i].target.trackuid &&
1873 track[j].stream) {
1874 matroska_convert_tag(s, &tags[i].tag,
1875 &track[j].stream->metadata, NULL);
1876 found = 1;
1877 }
1878 }
1879 if (!found) {
1880 av_log(s, AV_LOG_WARNING,
1881 "The tags at index %d refer to a non-existent track "
1882 "%"PRId64".\n",
1883 i, tags[i].target.trackuid);
1884 }
1885 } else {
1886 matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1887 tags[i].target.type);
1888 }
1889 }
1890 }
1891
matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska, int64_t pos)1892 static int matroska_parse_seekhead_entry(MatroskaDemuxContext *matroska,
1893 int64_t pos)
1894 {
1895 uint32_t saved_id = matroska->current_id;
1896 int64_t before_pos = avio_tell(matroska->ctx->pb);
1897 int ret = 0;
1898 int ret2;
1899
1900 /* seek */
1901 if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1902 /* We don't want to lose our seekhead level, so we add
1903 * a dummy. This is a crude hack. */
1904 if (matroska->num_levels == EBML_MAX_DEPTH) {
1905 av_log(matroska->ctx, AV_LOG_INFO,
1906 "Max EBML element depth (%d) reached, "
1907 "cannot parse further.\n", EBML_MAX_DEPTH);
1908 ret = AVERROR_INVALIDDATA;
1909 } else {
1910 matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1911 matroska->num_levels++;
1912 matroska->current_id = 0;
1913
1914 ret = ebml_parse(matroska, matroska_segment, matroska);
1915 if (ret == LEVEL_ENDED) {
1916 /* This can only happen if the seek brought us beyond EOF. */
1917 ret = AVERROR_EOF;
1918 }
1919 }
1920 }
1921 /* Seek back - notice that in all instances where this is used
1922 * it is safe to set the level to 1. */
1923 ret2 = matroska_reset_status(matroska, saved_id, before_pos);
1924 if (ret >= 0)
1925 ret = ret2;
1926
1927 return ret;
1928 }
1929
matroska_execute_seekhead(MatroskaDemuxContext *matroska)1930 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1931 {
1932 EbmlList *seekhead_list = &matroska->seekhead;
1933 int i;
1934
1935 // we should not do any seeking in the streaming case
1936 if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1937 return;
1938
1939 for (i = 0; i < seekhead_list->nb_elem; i++) {
1940 MatroskaSeekhead *seekheads = seekhead_list->elem;
1941 uint32_t id = seekheads[i].id;
1942 int64_t pos = seekheads[i].pos + matroska->segment_start;
1943 MatroskaLevel1Element *elem;
1944
1945 if (id != seekheads[i].id || pos < matroska->segment_start)
1946 continue;
1947
1948 elem = matroska_find_level1_elem(matroska, id, pos);
1949 if (!elem || elem->parsed)
1950 continue;
1951
1952 elem->pos = pos;
1953
1954 // defer cues parsing until we actually need cue data.
1955 if (id == MATROSKA_ID_CUES)
1956 continue;
1957
1958 if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1959 // mark index as broken
1960 matroska->cues_parsing_deferred = -1;
1961 break;
1962 }
1963
1964 elem->parsed = 1;
1965 }
1966 }
1967
matroska_add_index_entries(MatroskaDemuxContext *matroska)1968 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1969 {
1970 EbmlList *index_list;
1971 MatroskaIndex *index;
1972 uint64_t index_scale = 1;
1973 int i, j;
1974
1975 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1976 return;
1977
1978 index_list = &matroska->index;
1979 index = index_list->elem;
1980 if (index_list->nb_elem < 2)
1981 return;
1982 if (index[1].time > 1E14 / matroska->time_scale) {
1983 av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1984 return;
1985 }
1986 for (i = 0; i < index_list->nb_elem; i++) {
1987 EbmlList *pos_list = &index[i].pos;
1988 MatroskaIndexPos *pos = pos_list->elem;
1989 for (j = 0; j < pos_list->nb_elem; j++) {
1990 MatroskaTrack *track = matroska_find_track_by_num(matroska,
1991 pos[j].track);
1992 if (track && track->stream)
1993 av_add_index_entry(track->stream,
1994 pos[j].pos + matroska->segment_start,
1995 index[i].time / index_scale, 0, 0,
1996 AVINDEX_KEYFRAME);
1997 }
1998 }
1999 }
2000
matroska_parse_cues(MatroskaDemuxContext *matroska)2001 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
2002 int i;
2003
2004 if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
2005 return;
2006
2007 for (i = 0; i < matroska->num_level1_elems; i++) {
2008 MatroskaLevel1Element *elem = &matroska->level1_elems[i];
2009 if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
2010 if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
2011 matroska->cues_parsing_deferred = -1;
2012 elem->parsed = 1;
2013 break;
2014 }
2015 }
2016
2017 matroska_add_index_entries(matroska);
2018 }
2019
matroska_aac_profile(char *codec_id)2020 static int matroska_aac_profile(char *codec_id)
2021 {
2022 static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
2023 int profile;
2024
2025 for (profile = 0; profile < FF_ARRAY_ELEMS(aac_profiles); profile++)
2026 if (strstr(codec_id, aac_profiles[profile]))
2027 break;
2028 return profile + 1;
2029 }
2030
matroska_aac_sri(int samplerate)2031 static int matroska_aac_sri(int samplerate)
2032 {
2033 int sri;
2034
2035 for (sri = 0; sri < FF_ARRAY_ELEMS(ff_mpeg4audio_sample_rates); sri++)
2036 if (ff_mpeg4audio_sample_rates[sri] == samplerate)
2037 break;
2038 return sri;
2039 }
2040
matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)2041 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
2042 {
2043 /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
2044 avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
2045 }
2046
matroska_parse_flac(AVFormatContext *s, MatroskaTrack *track, int *offset)2047 static int matroska_parse_flac(AVFormatContext *s,
2048 MatroskaTrack *track,
2049 int *offset)
2050 {
2051 AVStream *st = track->stream;
2052 uint8_t *p = track->codec_priv.data;
2053 int size = track->codec_priv.size;
2054
2055 if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
2056 av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
2057 track->codec_priv.size = 0;
2058 return 0;
2059 }
2060 *offset = 8;
2061 track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
2062
2063 p += track->codec_priv.size;
2064 size -= track->codec_priv.size;
2065
2066 /* parse the remaining metadata blocks if present */
2067 while (size >= 4) {
2068 int block_last, block_type, block_size;
2069
2070 flac_parse_block_header(p, &block_last, &block_type, &block_size);
2071
2072 p += 4;
2073 size -= 4;
2074 if (block_size > size)
2075 return 0;
2076
2077 /* check for the channel mask */
2078 if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2079 AVDictionary *dict = NULL;
2080 AVDictionaryEntry *chmask;
2081
2082 ff_vorbis_comment(s, &dict, p, block_size, 0);
2083 chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2084 if (chmask) {
2085 uint64_t mask = strtol(chmask->value, NULL, 0);
2086 if (!mask || mask & ~0x3ffffULL) {
2087 av_log(s, AV_LOG_WARNING,
2088 "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2089 } else
2090 av_channel_layout_from_mask(&st->codecpar->ch_layout, mask);
2091 }
2092 av_dict_free(&dict);
2093 }
2094
2095 p += block_size;
2096 size -= block_size;
2097 }
2098
2099 return 0;
2100 }
2101
mkv_field_order(MatroskaDemuxContext *matroska, uint64_t field_order)2102 static int mkv_field_order(MatroskaDemuxContext *matroska, uint64_t field_order)
2103 {
2104 int minor, micro, bttb = 0;
2105
2106 /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2107 * this function, and fixed in 57.52 */
2108 if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf57.%d.%d", &minor, µ) == 2)
2109 bttb = (minor >= 36 && minor <= 51 && micro >= 100);
2110
2111 switch (field_order) {
2112 case MATROSKA_VIDEO_FIELDORDER_PROGRESSIVE:
2113 return AV_FIELD_PROGRESSIVE;
2114 case MATROSKA_VIDEO_FIELDORDER_UNDETERMINED:
2115 return AV_FIELD_UNKNOWN;
2116 case MATROSKA_VIDEO_FIELDORDER_TT:
2117 return AV_FIELD_TT;
2118 case MATROSKA_VIDEO_FIELDORDER_BB:
2119 return AV_FIELD_BB;
2120 case MATROSKA_VIDEO_FIELDORDER_BT:
2121 return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2122 case MATROSKA_VIDEO_FIELDORDER_TB:
2123 return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2124 default:
2125 return AV_FIELD_UNKNOWN;
2126 }
2127 }
2128
mkv_stereo_mode_display_mul(int stereo_mode, int *h_width, int *h_height)2129 static void mkv_stereo_mode_display_mul(int stereo_mode,
2130 int *h_width, int *h_height)
2131 {
2132 switch (stereo_mode) {
2133 case MATROSKA_VIDEO_STEREOMODE_TYPE_MONO:
2134 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_RL:
2135 case MATROSKA_VIDEO_STEREOMODE_TYPE_CHECKERBOARD_LR:
2136 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_RL:
2137 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTH_EYES_BLOCK_LR:
2138 break;
2139 case MATROSKA_VIDEO_STEREOMODE_TYPE_RIGHT_LEFT:
2140 case MATROSKA_VIDEO_STEREOMODE_TYPE_LEFT_RIGHT:
2141 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_RL:
2142 case MATROSKA_VIDEO_STEREOMODE_TYPE_COL_INTERLEAVED_LR:
2143 *h_width = 2;
2144 break;
2145 case MATROSKA_VIDEO_STEREOMODE_TYPE_BOTTOM_TOP:
2146 case MATROSKA_VIDEO_STEREOMODE_TYPE_TOP_BOTTOM:
2147 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL:
2148 case MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_LR:
2149 *h_height = 2;
2150 break;
2151 }
2152 }
2153
mkv_parse_video_color(AVStream *st, const MatroskaTrack *track)2154 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2155 const MatroskaTrackVideoColor *color = track->video.color.elem;
2156 const MatroskaMasteringMeta *mastering_meta;
2157 int has_mastering_primaries, has_mastering_luminance;
2158
2159 if (!track->video.color.nb_elem)
2160 return 0;
2161
2162 mastering_meta = &color->mastering_meta;
2163 // Mastering primaries are CIE 1931 coords, and must be > 0.
2164 has_mastering_primaries =
2165 mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2166 mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2167 mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2168 mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2169 has_mastering_luminance = mastering_meta->max_luminance >
2170 mastering_meta->min_luminance.el.f &&
2171 mastering_meta->min_luminance.el.f >= 0 &&
2172 mastering_meta->min_luminance.count;
2173
2174 if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2175 st->codecpar->color_space = color->matrix_coefficients;
2176 if (color->primaries != AVCOL_PRI_RESERVED &&
2177 color->primaries != AVCOL_PRI_RESERVED0)
2178 st->codecpar->color_primaries = color->primaries;
2179 if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2180 color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2181 st->codecpar->color_trc = color->transfer_characteristics;
2182 if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2183 color->range <= AVCOL_RANGE_JPEG)
2184 st->codecpar->color_range = color->range;
2185 if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2186 color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2187 color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2188 color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2189 st->codecpar->chroma_location =
2190 avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2191 (color->chroma_siting_vert - 1) << 7);
2192 }
2193 if (color->max_cll && color->max_fall) {
2194 size_t size = 0;
2195 int ret;
2196 AVContentLightMetadata *metadata = av_content_light_metadata_alloc(&size);
2197 if (!metadata)
2198 return AVERROR(ENOMEM);
2199 ret = av_stream_add_side_data(st, AV_PKT_DATA_CONTENT_LIGHT_LEVEL,
2200 (uint8_t *)metadata, size);
2201 if (ret < 0) {
2202 av_freep(&metadata);
2203 return ret;
2204 }
2205 metadata->MaxCLL = color->max_cll;
2206 metadata->MaxFALL = color->max_fall;
2207 }
2208
2209 if (has_mastering_primaries || has_mastering_luminance) {
2210 AVMasteringDisplayMetadata *metadata =
2211 (AVMasteringDisplayMetadata*) av_stream_new_side_data(
2212 st, AV_PKT_DATA_MASTERING_DISPLAY_METADATA,
2213 sizeof(AVMasteringDisplayMetadata));
2214 if (!metadata) {
2215 return AVERROR(ENOMEM);
2216 }
2217 memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2218 if (has_mastering_primaries) {
2219 metadata->display_primaries[0][0] = av_d2q(mastering_meta->r_x, INT_MAX);
2220 metadata->display_primaries[0][1] = av_d2q(mastering_meta->r_y, INT_MAX);
2221 metadata->display_primaries[1][0] = av_d2q(mastering_meta->g_x, INT_MAX);
2222 metadata->display_primaries[1][1] = av_d2q(mastering_meta->g_y, INT_MAX);
2223 metadata->display_primaries[2][0] = av_d2q(mastering_meta->b_x, INT_MAX);
2224 metadata->display_primaries[2][1] = av_d2q(mastering_meta->b_y, INT_MAX);
2225 metadata->white_point[0] = av_d2q(mastering_meta->white_x, INT_MAX);
2226 metadata->white_point[1] = av_d2q(mastering_meta->white_y, INT_MAX);
2227 metadata->has_primaries = 1;
2228 }
2229 if (has_mastering_luminance) {
2230 metadata->max_luminance = av_d2q(mastering_meta->max_luminance, INT_MAX);
2231 metadata->min_luminance = av_d2q(mastering_meta->min_luminance.el.f, INT_MAX);
2232 metadata->has_luminance = 1;
2233 }
2234 }
2235 return 0;
2236 }
2237
mkv_create_display_matrix(AVStream *st, const MatroskaTrackVideoProjection *proj, void *logctx)2238 static int mkv_create_display_matrix(AVStream *st,
2239 const MatroskaTrackVideoProjection *proj,
2240 void *logctx)
2241 {
2242 double pitch = proj->pitch, yaw = proj->yaw, roll = proj->roll;
2243 int32_t *matrix;
2244 int hflip;
2245
2246 if (pitch == 0.0 && yaw == 0.0 && roll == 0.0)
2247 return 0;
2248
2249 /* Note: The following constants are exactly representable
2250 * as floating-point numbers. */
2251 if (pitch != 0.0 || (yaw != 0.0 && yaw != 180.0 && yaw != -180.0) ||
2252 isnan(roll)) {
2253 av_log(logctx, AV_LOG_WARNING, "Ignoring non-2D rectangular "
2254 "projection in stream %u (yaw %f, pitch %f, roll %f)\n",
2255 st->index, yaw, pitch, roll);
2256 return 0;
2257 }
2258 matrix = (int32_t*)av_stream_new_side_data(st, AV_PKT_DATA_DISPLAYMATRIX,
2259 9 * sizeof(*matrix));
2260 if (!matrix)
2261 return AVERROR(ENOMEM);
2262
2263 hflip = yaw != 0.0;
2264 /* ProjectionPoseRoll is in the counter-clockwise direction
2265 * whereas av_display_rotation_set() expects its argument
2266 * to be oriented clockwise, so we need to negate roll.
2267 * Furthermore, if hflip is set, we need to negate it again
2268 * to account for the fact that the Matroska specifications
2269 * require the yaw rotation to be applied first. */
2270 av_display_rotation_set(matrix, roll * (2 * hflip - 1));
2271 av_display_matrix_flip(matrix, hflip, 0);
2272
2273 return 0;
2274 }
2275
mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track, void *logctx)2276 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track,
2277 void *logctx)
2278 {
2279 AVSphericalMapping *spherical;
2280 const MatroskaTrackVideoProjection *mkv_projection = &track->video.projection;
2281 const uint8_t *priv_data = mkv_projection->private.data;
2282 enum AVSphericalProjection projection;
2283 size_t spherical_size;
2284 uint32_t l = 0, t = 0, r = 0, b = 0;
2285 uint32_t padding = 0;
2286 int ret;
2287
2288 if (mkv_projection->private.size && priv_data[0] != 0) {
2289 av_log(logctx, AV_LOG_WARNING, "Unknown spherical metadata\n");
2290 return 0;
2291 }
2292
2293 switch (track->video.projection.type) {
2294 case MATROSKA_VIDEO_PROJECTION_TYPE_RECTANGULAR:
2295 return mkv_create_display_matrix(st, mkv_projection, logctx);
2296 case MATROSKA_VIDEO_PROJECTION_TYPE_EQUIRECTANGULAR:
2297 if (track->video.projection.private.size == 20) {
2298 t = AV_RB32(priv_data + 4);
2299 b = AV_RB32(priv_data + 8);
2300 l = AV_RB32(priv_data + 12);
2301 r = AV_RB32(priv_data + 16);
2302
2303 if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2304 av_log(logctx, AV_LOG_ERROR,
2305 "Invalid bounding rectangle coordinates "
2306 "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2307 l, t, r, b);
2308 return AVERROR_INVALIDDATA;
2309 }
2310 } else if (track->video.projection.private.size != 0) {
2311 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2312 return AVERROR_INVALIDDATA;
2313 }
2314
2315 if (l || t || r || b)
2316 projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2317 else
2318 projection = AV_SPHERICAL_EQUIRECTANGULAR;
2319 break;
2320 case MATROSKA_VIDEO_PROJECTION_TYPE_CUBEMAP:
2321 if (track->video.projection.private.size < 4) {
2322 av_log(logctx, AV_LOG_ERROR, "Missing projection private properties\n");
2323 return AVERROR_INVALIDDATA;
2324 } else if (track->video.projection.private.size == 12) {
2325 uint32_t layout = AV_RB32(priv_data + 4);
2326 if (layout) {
2327 av_log(logctx, AV_LOG_WARNING,
2328 "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2329 return 0;
2330 }
2331 projection = AV_SPHERICAL_CUBEMAP;
2332 padding = AV_RB32(priv_data + 8);
2333 } else {
2334 av_log(logctx, AV_LOG_ERROR, "Unknown spherical metadata\n");
2335 return AVERROR_INVALIDDATA;
2336 }
2337 break;
2338 default:
2339 av_log(logctx, AV_LOG_WARNING,
2340 "Unknown spherical metadata type %"PRIu64"\n",
2341 track->video.projection.type);
2342 return 0;
2343 }
2344
2345 spherical = av_spherical_alloc(&spherical_size);
2346 if (!spherical)
2347 return AVERROR(ENOMEM);
2348
2349 spherical->projection = projection;
2350
2351 spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2352 spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2353 spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2354
2355 spherical->padding = padding;
2356
2357 spherical->bound_left = l;
2358 spherical->bound_top = t;
2359 spherical->bound_right = r;
2360 spherical->bound_bottom = b;
2361
2362 ret = av_stream_add_side_data(st, AV_PKT_DATA_SPHERICAL, (uint8_t *)spherical,
2363 spherical_size);
2364 if (ret < 0) {
2365 av_freep(&spherical);
2366 return ret;
2367 }
2368
2369 return 0;
2370 }
2371
mkv_parse_dvcc_dvvc(AVFormatContext *s, AVStream *st, const MatroskaTrack *track, EbmlBin *bin)2372 static int mkv_parse_dvcc_dvvc(AVFormatContext *s, AVStream *st, const MatroskaTrack *track,
2373 EbmlBin *bin)
2374 {
2375 return ff_isom_parse_dvcc_dvvc(s, st, bin->data, bin->size);
2376 }
2377
mkv_parse_block_addition_mappings(AVFormatContext *s, AVStream *st, const MatroskaTrack *track)2378 static int mkv_parse_block_addition_mappings(AVFormatContext *s, AVStream *st, const MatroskaTrack *track)
2379 {
2380 const EbmlList *mappings_list = &track->block_addition_mappings;
2381 MatroskaBlockAdditionMapping *mappings = mappings_list->elem;
2382 int ret;
2383
2384 for (int i = 0; i < mappings_list->nb_elem; i++) {
2385 MatroskaBlockAdditionMapping *mapping = &mappings[i];
2386
2387 switch (mapping->type) {
2388 case MKBETAG('d','v','c','C'):
2389 case MKBETAG('d','v','v','C'):
2390 if ((ret = mkv_parse_dvcc_dvvc(s, st, track, &mapping->extradata)) < 0)
2391 return ret;
2392
2393 break;
2394 default:
2395 av_log(s, AV_LOG_DEBUG,
2396 "Unknown block additional mapping type 0x%"PRIx64", value %"PRIu64", name \"%s\"\n",
2397 mapping->type, mapping->value, mapping->name ? mapping->name : "");
2398 }
2399 }
2400
2401 return 0;
2402 }
2403
get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)2404 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2405 {
2406 const AVCodecTag *codec_tags;
2407
2408 codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2409 ff_codec_movvideo_tags : ff_codec_movaudio_tags;
2410
2411 /* Normalize noncompliant private data that starts with the fourcc
2412 * by expanding/shifting the data by 4 bytes and storing the data
2413 * size at the start. */
2414 if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2415 int ret = av_buffer_realloc(&track->codec_priv.buf,
2416 track->codec_priv.size + 4 + AV_INPUT_BUFFER_PADDING_SIZE);
2417 if (ret < 0)
2418 return ret;
2419
2420 track->codec_priv.data = track->codec_priv.buf->data;
2421 memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2422 track->codec_priv.size += 4;
2423 AV_WB32(track->codec_priv.data, track->codec_priv.size);
2424 }
2425
2426 *fourcc = AV_RL32(track->codec_priv.data + 4);
2427 *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2428
2429 return 0;
2430 }
2431
matroska_parse_tracks(AVFormatContext *s)2432 static int matroska_parse_tracks(AVFormatContext *s)
2433 {
2434 MatroskaDemuxContext *matroska = s->priv_data;
2435 MatroskaTrack *tracks = matroska->tracks.elem;
2436 int i, j, ret;
2437 int k;
2438
2439 for (i = 0; i < matroska->tracks.nb_elem; i++) {
2440 MatroskaTrack *track = &tracks[i];
2441 enum AVCodecID codec_id = AV_CODEC_ID_NONE;
2442 EbmlList *encodings_list = &track->encodings;
2443 MatroskaTrackEncoding *encodings = encodings_list->elem;
2444 uint8_t *extradata = NULL;
2445 int extradata_size = 0;
2446 int extradata_offset = 0;
2447 uint32_t fourcc = 0;
2448 FFIOContext b;
2449 AVStream *st;
2450 FFStream *sti;
2451 char* key_id_base64 = NULL;
2452 int bit_depth = -1;
2453
2454 /* Apply some sanity checks. */
2455 if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2456 track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2457 track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2458 track->type != MATROSKA_TRACK_TYPE_METADATA) {
2459 av_log(matroska->ctx, AV_LOG_INFO,
2460 "Unknown or unsupported track type %"PRIu64"\n",
2461 track->type);
2462 continue;
2463 }
2464 if (!track->codec_id)
2465 continue;
2466
2467 if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
2468 || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
2469 || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2470 || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2471 ) {
2472 av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
2473 continue;
2474 }
2475
2476 if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2477 isnan(track->audio.samplerate)) {
2478 av_log(matroska->ctx, AV_LOG_WARNING,
2479 "Invalid sample rate %f, defaulting to 8000 instead.\n",
2480 track->audio.samplerate);
2481 track->audio.samplerate = 8000;
2482 }
2483
2484 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2485 if (!track->default_duration && track->video.frame_rate > 0) {
2486 double default_duration = 1000000000 / track->video.frame_rate;
2487 if (default_duration > UINT64_MAX || default_duration < 0) {
2488 av_log(matroska->ctx, AV_LOG_WARNING,
2489 "Invalid frame rate %e. Cannot calculate default duration.\n",
2490 track->video.frame_rate);
2491 } else {
2492 track->default_duration = default_duration;
2493 }
2494 }
2495 if (track->video.display_width == -1)
2496 track->video.display_width = track->video.pixel_width;
2497 if (track->video.display_height == -1)
2498 track->video.display_height = track->video.pixel_height;
2499 if (track->video.color_space.size == 4)
2500 fourcc = AV_RL32(track->video.color_space.data);
2501 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2502 if (!track->audio.out_samplerate)
2503 track->audio.out_samplerate = track->audio.samplerate;
2504 }
2505 if (encodings_list->nb_elem > 1) {
2506 av_log(matroska->ctx, AV_LOG_ERROR,
2507 "Multiple combined encodings not supported");
2508 } else if (encodings_list->nb_elem == 1) {
2509 if (encodings[0].type) {
2510 if (encodings[0].encryption.key_id.size > 0) {
2511 /* Save the encryption key id to be stored later as a
2512 metadata tag. */
2513 const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2514 key_id_base64 = av_malloc(b64_size);
2515 if (key_id_base64 == NULL)
2516 return AVERROR(ENOMEM);
2517
2518 av_base64_encode(key_id_base64, b64_size,
2519 encodings[0].encryption.key_id.data,
2520 encodings[0].encryption.key_id.size);
2521 } else {
2522 encodings[0].scope = 0;
2523 av_log(matroska->ctx, AV_LOG_ERROR,
2524 "Unsupported encoding type");
2525 }
2526 } else if (
2527 #if CONFIG_ZLIB
2528 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2529 #endif
2530 #if CONFIG_BZLIB
2531 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2532 #endif
2533 #if CONFIG_LZO
2534 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2535 #endif
2536 encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2537 encodings[0].scope = 0;
2538 av_log(matroska->ctx, AV_LOG_ERROR,
2539 "Unsupported encoding type");
2540 } else if (track->codec_priv.size && encodings[0].scope & 2) {
2541 uint8_t *codec_priv = track->codec_priv.data;
2542 int ret = matroska_decode_buffer(&track->codec_priv.data,
2543 &track->codec_priv.size,
2544 track);
2545 if (ret < 0) {
2546 track->codec_priv.data = NULL;
2547 track->codec_priv.size = 0;
2548 av_log(matroska->ctx, AV_LOG_ERROR,
2549 "Failed to decode codec private data\n");
2550 }
2551
2552 if (codec_priv != track->codec_priv.data) {
2553 av_buffer_unref(&track->codec_priv.buf);
2554 if (track->codec_priv.data) {
2555 track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2556 track->codec_priv.size + AV_INPUT_BUFFER_PADDING_SIZE,
2557 NULL, NULL, 0);
2558 if (!track->codec_priv.buf) {
2559 av_freep(&track->codec_priv.data);
2560 track->codec_priv.size = 0;
2561 return AVERROR(ENOMEM);
2562 }
2563 }
2564 }
2565 }
2566 }
2567 track->needs_decoding = encodings && !encodings[0].type &&
2568 encodings[0].scope & 1 &&
2569 (encodings[0].compression.algo !=
2570 MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP ||
2571 encodings[0].compression.settings.size);
2572
2573 for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2574 if (av_strstart(track->codec_id, ff_mkv_codec_tags[j].str, NULL)) {
2575 codec_id = ff_mkv_codec_tags[j].id;
2576 break;
2577 }
2578 }
2579
2580 st = track->stream = avformat_new_stream(s, NULL);
2581 if (!st) {
2582 av_free(key_id_base64);
2583 return AVERROR(ENOMEM);
2584 }
2585 sti = ffstream(st);
2586
2587 if (key_id_base64) {
2588 /* export encryption key id as base64 metadata tag */
2589 av_dict_set(&st->metadata, "enc_key_id", key_id_base64,
2590 AV_DICT_DONT_STRDUP_VAL);
2591 }
2592
2593 if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2594 track->codec_priv.size >= 40 &&
2595 track->codec_priv.data) {
2596 track->ms_compat = 1;
2597 bit_depth = AV_RL16(track->codec_priv.data + 14);
2598 fourcc = AV_RL32(track->codec_priv.data + 16);
2599 codec_id = ff_codec_get_id(ff_codec_bmp_tags,
2600 fourcc);
2601 if (!codec_id)
2602 codec_id = ff_codec_get_id(ff_codec_movvideo_tags,
2603 fourcc);
2604 extradata_offset = 40;
2605 } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2606 track->codec_priv.size >= 14 &&
2607 track->codec_priv.data) {
2608 int ret;
2609 ffio_init_context(&b, track->codec_priv.data,
2610 track->codec_priv.size,
2611 0, NULL, NULL, NULL, NULL);
2612 ret = ff_get_wav_header(s, &b.pub, st->codecpar,
2613 track->codec_priv.size, 0);
2614 if (ret < 0)
2615 return ret;
2616 codec_id = st->codecpar->codec_id;
2617 fourcc = st->codecpar->codec_tag;
2618 extradata_offset = FFMIN(track->codec_priv.size, 18);
2619 } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2620 /* Normally 36, but allow noncompliant private data */
2621 && (track->codec_priv.size >= 32)
2622 && (track->codec_priv.data)) {
2623 uint16_t sample_size;
2624 int ret = get_qt_codec(track, &fourcc, &codec_id);
2625 if (ret < 0)
2626 return ret;
2627 sample_size = AV_RB16(track->codec_priv.data + 26);
2628 if (fourcc == 0) {
2629 if (sample_size == 8) {
2630 fourcc = MKTAG('r','a','w',' ');
2631 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2632 } else if (sample_size == 16) {
2633 fourcc = MKTAG('t','w','o','s');
2634 codec_id = ff_codec_get_id(ff_codec_movaudio_tags, fourcc);
2635 }
2636 }
2637 if ((fourcc == MKTAG('t','w','o','s') ||
2638 fourcc == MKTAG('s','o','w','t')) &&
2639 sample_size == 8)
2640 codec_id = AV_CODEC_ID_PCM_S8;
2641 } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2642 (track->codec_priv.size >= 21) &&
2643 (track->codec_priv.data)) {
2644 int ret = get_qt_codec(track, &fourcc, &codec_id);
2645 if (ret < 0)
2646 return ret;
2647 if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2648 fourcc = MKTAG('S','V','Q','3');
2649 codec_id = ff_codec_get_id(ff_codec_movvideo_tags, fourcc);
2650 }
2651 if (codec_id == AV_CODEC_ID_NONE)
2652 av_log(matroska->ctx, AV_LOG_ERROR,
2653 "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2654 if (track->codec_priv.size >= 86) {
2655 bit_depth = AV_RB16(track->codec_priv.data + 82);
2656 ffio_init_context(&b, track->codec_priv.data,
2657 track->codec_priv.size,
2658 0, NULL, NULL, NULL, NULL);
2659 if (ff_get_qtpalette(codec_id, &b.pub, track->palette)) {
2660 bit_depth &= 0x1F;
2661 track->has_palette = 1;
2662 }
2663 }
2664 } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2665 switch (track->audio.bitdepth) {
2666 case 8:
2667 codec_id = AV_CODEC_ID_PCM_U8;
2668 break;
2669 case 24:
2670 codec_id = AV_CODEC_ID_PCM_S24BE;
2671 break;
2672 case 32:
2673 codec_id = AV_CODEC_ID_PCM_S32BE;
2674 break;
2675 }
2676 } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2677 switch (track->audio.bitdepth) {
2678 case 8:
2679 codec_id = AV_CODEC_ID_PCM_U8;
2680 break;
2681 case 24:
2682 codec_id = AV_CODEC_ID_PCM_S24LE;
2683 break;
2684 case 32:
2685 codec_id = AV_CODEC_ID_PCM_S32LE;
2686 break;
2687 }
2688 } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2689 track->audio.bitdepth == 64) {
2690 codec_id = AV_CODEC_ID_PCM_F64LE;
2691 } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2692 int profile = matroska_aac_profile(track->codec_id);
2693 int sri = matroska_aac_sri(track->audio.samplerate);
2694 extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2695 if (!extradata)
2696 return AVERROR(ENOMEM);
2697 extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2698 extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2699 if (strstr(track->codec_id, "SBR")) {
2700 sri = matroska_aac_sri(track->audio.out_samplerate);
2701 extradata[2] = 0x56;
2702 extradata[3] = 0xE5;
2703 extradata[4] = 0x80 | (sri << 3);
2704 extradata_size = 5;
2705 } else
2706 extradata_size = 2;
2707 } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2708 /* Only ALAC's magic cookie is stored in Matroska's track headers.
2709 * Create the "atom size", "tag", and "tag version" fields the
2710 * decoder expects manually. */
2711 extradata_size = 12 + track->codec_priv.size;
2712 extradata = av_mallocz(extradata_size +
2713 AV_INPUT_BUFFER_PADDING_SIZE);
2714 if (!extradata)
2715 return AVERROR(ENOMEM);
2716 AV_WB32(extradata, extradata_size);
2717 memcpy(&extradata[4], "alac", 4);
2718 AV_WB32(&extradata[8], 0);
2719 memcpy(&extradata[12], track->codec_priv.data,
2720 track->codec_priv.size);
2721 } else if (codec_id == AV_CODEC_ID_TTA) {
2722 uint8_t *ptr;
2723 if (track->audio.channels > UINT16_MAX ||
2724 track->audio.bitdepth > UINT16_MAX) {
2725 av_log(matroska->ctx, AV_LOG_WARNING,
2726 "Too large audio channel number %"PRIu64
2727 " or bitdepth %"PRIu64". Skipping track.\n",
2728 track->audio.channels, track->audio.bitdepth);
2729 if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2730 return AVERROR_INVALIDDATA;
2731 else
2732 continue;
2733 }
2734 if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2735 return AVERROR_INVALIDDATA;
2736 extradata_size = 22;
2737 extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2738 if (!extradata)
2739 return AVERROR(ENOMEM);
2740 ptr = extradata;
2741 bytestream_put_be32(&ptr, AV_RB32("TTA1"));
2742 bytestream_put_le16(&ptr, 1);
2743 bytestream_put_le16(&ptr, track->audio.channels);
2744 bytestream_put_le16(&ptr, track->audio.bitdepth);
2745 bytestream_put_le32(&ptr, track->audio.out_samplerate);
2746 bytestream_put_le32(&ptr, av_rescale(matroska->duration * matroska->time_scale,
2747 track->audio.out_samplerate,
2748 AV_TIME_BASE * 1000));
2749 } else if (codec_id == AV_CODEC_ID_RV10 ||
2750 codec_id == AV_CODEC_ID_RV20 ||
2751 codec_id == AV_CODEC_ID_RV30 ||
2752 codec_id == AV_CODEC_ID_RV40) {
2753 extradata_offset = 26;
2754 } else if (codec_id == AV_CODEC_ID_RA_144) {
2755 track->audio.out_samplerate = 8000;
2756 track->audio.channels = 1;
2757 } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2758 codec_id == AV_CODEC_ID_COOK ||
2759 codec_id == AV_CODEC_ID_ATRAC3 ||
2760 codec_id == AV_CODEC_ID_SIPR)
2761 && track->codec_priv.data) {
2762 const uint8_t *ptr = track->codec_priv.data;
2763 int flavor;
2764
2765 if (track->codec_priv.size < 46)
2766 return AVERROR_INVALIDDATA;
2767 ptr += 22;
2768 flavor = bytestream_get_be16(&ptr);
2769 track->audio.coded_framesize = bytestream_get_be32(&ptr);
2770 ptr += 12;
2771 track->audio.sub_packet_h = bytestream_get_be16(&ptr);
2772 track->audio.frame_size = bytestream_get_be16(&ptr);
2773 track->audio.sub_packet_size = bytestream_get_be16(&ptr);
2774 if (track->audio.coded_framesize <= 0 ||
2775 track->audio.sub_packet_h <= 0 ||
2776 track->audio.frame_size <= 0)
2777 return AVERROR_INVALIDDATA;
2778
2779 if (codec_id == AV_CODEC_ID_RA_288) {
2780 if (track->audio.sub_packet_h & 1 || 2 * track->audio.frame_size
2781 != (int64_t)track->audio.sub_packet_h * track->audio.coded_framesize)
2782 return AVERROR_INVALIDDATA;
2783 st->codecpar->block_align = track->audio.coded_framesize;
2784 track->codec_priv.size = 0;
2785 } else {
2786 if (codec_id == AV_CODEC_ID_SIPR) {
2787 static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2788 if (flavor > 3)
2789 return AVERROR_INVALIDDATA;
2790 track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2791 st->codecpar->bit_rate = sipr_bit_rate[flavor];
2792 } else if (track->audio.sub_packet_size <= 0 ||
2793 track->audio.frame_size % track->audio.sub_packet_size)
2794 return AVERROR_INVALIDDATA;
2795 st->codecpar->block_align = track->audio.sub_packet_size;
2796 extradata_offset = 78;
2797 }
2798 track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2799 track->audio.frame_size);
2800 if (!track->audio.buf)
2801 return AVERROR(ENOMEM);
2802 } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2803 ret = matroska_parse_flac(s, track, &extradata_offset);
2804 if (ret < 0)
2805 return ret;
2806 } else if (codec_id == AV_CODEC_ID_WAVPACK && track->codec_priv.size < 2) {
2807 av_log(matroska->ctx, AV_LOG_INFO, "Assuming WavPack version 4.10 "
2808 "in absence of valid CodecPrivate.\n");
2809 extradata_size = 2;
2810 extradata = av_mallocz(2 + AV_INPUT_BUFFER_PADDING_SIZE);
2811 if (!extradata)
2812 return AVERROR(ENOMEM);
2813 AV_WL16(extradata, 0x410);
2814 } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2815 fourcc = AV_RL32(track->codec_priv.data);
2816 } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2817 /* we don't need any value stored in CodecPrivate.
2818 make sure that it's not exported as extradata. */
2819 track->codec_priv.size = 0;
2820 }
2821 track->codec_priv.size -= extradata_offset;
2822
2823 if (codec_id == AV_CODEC_ID_NONE)
2824 av_log(matroska->ctx, AV_LOG_INFO,
2825 "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2826
2827 if (track->time_scale < 0.01) {
2828 av_log(matroska->ctx, AV_LOG_WARNING,
2829 "Track TimestampScale too small %f, assuming 1.0.\n",
2830 track->time_scale);
2831 track->time_scale = 1.0;
2832 }
2833 avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2834 1000 * 1000 * 1000); /* 64 bit pts in ns */
2835
2836 /* convert the delay from ns to the track timebase */
2837 track->codec_delay_in_track_tb = av_rescale_q(track->codec_delay,
2838 (AVRational){ 1, 1000000000 },
2839 st->time_base);
2840
2841 st->codecpar->codec_id = codec_id;
2842
2843 if (strcmp(track->language, "und"))
2844 av_dict_set(&st->metadata, "language", track->language, 0);
2845 av_dict_set(&st->metadata, "title", track->name, 0);
2846
2847 if (track->flag_default)
2848 st->disposition |= AV_DISPOSITION_DEFAULT;
2849 if (track->flag_forced)
2850 st->disposition |= AV_DISPOSITION_FORCED;
2851 if (track->flag_comment)
2852 st->disposition |= AV_DISPOSITION_COMMENT;
2853 if (track->flag_hearingimpaired)
2854 st->disposition |= AV_DISPOSITION_HEARING_IMPAIRED;
2855 if (track->flag_visualimpaired)
2856 st->disposition |= AV_DISPOSITION_VISUAL_IMPAIRED;
2857 if (track->flag_original.count > 0)
2858 st->disposition |= track->flag_original.el.u ? AV_DISPOSITION_ORIGINAL
2859 : AV_DISPOSITION_DUB;
2860
2861 if (!st->codecpar->extradata) {
2862 if (extradata) {
2863 st->codecpar->extradata = extradata;
2864 st->codecpar->extradata_size = extradata_size;
2865 } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2866 if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2867 return AVERROR(ENOMEM);
2868 memcpy(st->codecpar->extradata,
2869 track->codec_priv.data + extradata_offset,
2870 track->codec_priv.size);
2871 }
2872 }
2873
2874 if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2875 MatroskaTrackPlane *planes = track->operation.combine_planes.elem;
2876 int display_width_mul = 1;
2877 int display_height_mul = 1;
2878
2879 st->codecpar->codec_type = AVMEDIA_TYPE_VIDEO;
2880 st->codecpar->codec_tag = fourcc;
2881 if (bit_depth >= 0)
2882 st->codecpar->bits_per_coded_sample = bit_depth;
2883 st->codecpar->width = track->video.pixel_width;
2884 st->codecpar->height = track->video.pixel_height;
2885
2886 if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_INTERLACED)
2887 st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2888 else if (track->video.interlaced == MATROSKA_VIDEO_INTERLACE_FLAG_PROGRESSIVE)
2889 st->codecpar->field_order = AV_FIELD_PROGRESSIVE;
2890
2891 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2892 mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2893
2894 if (track->video.display_unit < MATROSKA_VIDEO_DISPLAYUNIT_UNKNOWN) {
2895 if (track->video.display_width && track->video.display_height &&
2896 st->codecpar->height < INT64_MAX / track->video.display_width / display_width_mul &&
2897 st->codecpar->width < INT64_MAX / track->video.display_height / display_height_mul)
2898 av_reduce(&st->sample_aspect_ratio.num,
2899 &st->sample_aspect_ratio.den,
2900 st->codecpar->height * track->video.display_width * display_width_mul,
2901 st->codecpar->width * track->video.display_height * display_height_mul,
2902 INT_MAX);
2903 }
2904 if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2905 sti->need_parsing = AVSTREAM_PARSE_HEADERS;
2906
2907 if (track->default_duration) {
2908 int div = track->default_duration <= INT64_MAX ? 1 : 2;
2909 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
2910 1000000000 / div, track->default_duration / div, 30000);
2911 #if FF_API_R_FRAME_RATE
2912 if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2913 && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2914 st->r_frame_rate = st->avg_frame_rate;
2915 #endif
2916 }
2917
2918 /* export stereo mode flag as metadata tag */
2919 if (track->video.stereo_mode && track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB)
2920 av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2921
2922 /* export alpha mode flag as metadata tag */
2923 if (track->video.alpha_mode)
2924 av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2925
2926 /* if we have virtual track, mark the real tracks */
2927 for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2928 char buf[32];
2929 if (planes[j].type >= MATROSKA_VIDEO_STEREO_PLANE_COUNT)
2930 continue;
2931 snprintf(buf, sizeof(buf), "%s_%d",
2932 ff_matroska_video_stereo_plane[planes[j].type], i);
2933 for (k=0; k < matroska->tracks.nb_elem; k++)
2934 if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2935 av_dict_set(&tracks[k].stream->metadata,
2936 "stereo_mode", buf, 0);
2937 break;
2938 }
2939 }
2940 // add stream level stereo3d side data if it is a supported format
2941 if (track->video.stereo_mode < MATROSKA_VIDEO_STEREOMODE_TYPE_NB &&
2942 track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2943 int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2944 if (ret < 0)
2945 return ret;
2946 }
2947
2948 ret = mkv_parse_video_color(st, track);
2949 if (ret < 0)
2950 return ret;
2951 ret = mkv_parse_video_projection(st, track, matroska->ctx);
2952 if (ret < 0)
2953 return ret;
2954 } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2955 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
2956 st->codecpar->codec_tag = fourcc;
2957 st->codecpar->sample_rate = track->audio.out_samplerate;
2958 // channel layout may be already set by codec private checks above
2959 if (!av_channel_layout_check(&st->codecpar->ch_layout)) {
2960 st->codecpar->ch_layout.order = AV_CHANNEL_ORDER_UNSPEC;
2961 st->codecpar->ch_layout.nb_channels = track->audio.channels;
2962 }
2963 if (!st->codecpar->bits_per_coded_sample)
2964 st->codecpar->bits_per_coded_sample = track->audio.bitdepth;
2965 if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2966 st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2967 st->codecpar->codec_id == AV_CODEC_ID_TRUEHD)
2968 sti->need_parsing = AVSTREAM_PARSE_FULL;
2969 else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2970 sti->need_parsing = AVSTREAM_PARSE_HEADERS;
2971 if (track->codec_delay > 0) {
2972 st->codecpar->initial_padding = av_rescale_q(track->codec_delay,
2973 (AVRational){1, 1000000000},
2974 (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2975 48000 : st->codecpar->sample_rate});
2976 }
2977 if (track->seek_preroll > 0) {
2978 st->codecpar->seek_preroll = av_rescale_q(track->seek_preroll,
2979 (AVRational){1, 1000000000},
2980 (AVRational){1, st->codecpar->sample_rate});
2981 }
2982 } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2983 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2984
2985 if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2986 st->disposition |= AV_DISPOSITION_CAPTIONS;
2987 } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2988 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2989 } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2990 st->disposition |= AV_DISPOSITION_METADATA;
2991 }
2992 } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2993 st->codecpar->codec_type = AVMEDIA_TYPE_SUBTITLE;
2994
2995 if (track->flag_textdescriptions)
2996 st->disposition |= AV_DISPOSITION_DESCRIPTIONS;
2997 }
2998
2999 ret = mkv_parse_block_addition_mappings(s, st, track);
3000 if (ret < 0)
3001 return ret;
3002 }
3003
3004 return 0;
3005 }
3006
matroska_read_header(AVFormatContext *s)3007 static int matroska_read_header(AVFormatContext *s)
3008 {
3009 FFFormatContext *const si = ffformatcontext(s);
3010 MatroskaDemuxContext *matroska = s->priv_data;
3011 EbmlList *attachments_list = &matroska->attachments;
3012 EbmlList *chapters_list = &matroska->chapters;
3013 MatroskaAttachment *attachments;
3014 MatroskaChapter *chapters;
3015 uint64_t max_start = 0;
3016 int64_t pos;
3017 Ebml ebml = { 0 };
3018 int i, j, res;
3019
3020 matroska->ctx = s;
3021 matroska->cues_parsing_deferred = 1;
3022
3023 /* First read the EBML header. */
3024 if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
3025 av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
3026 ebml_free(ebml_syntax, &ebml);
3027 return AVERROR_INVALIDDATA;
3028 }
3029 if (ebml.version > EBML_VERSION ||
3030 ebml.max_size > sizeof(uint64_t) ||
3031 ebml.id_length > sizeof(uint32_t) ||
3032 ebml.doctype_version > 3) {
3033 avpriv_report_missing_feature(matroska->ctx,
3034 "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
3035 ebml.version, ebml.doctype, ebml.doctype_version);
3036 ebml_free(ebml_syntax, &ebml);
3037 return AVERROR_PATCHWELCOME;
3038 } else if (ebml.doctype_version == 3) {
3039 av_log(matroska->ctx, AV_LOG_WARNING,
3040 "EBML header using unsupported features\n"
3041 "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
3042 ebml.version, ebml.doctype, ebml.doctype_version);
3043 }
3044 for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
3045 if (!strcmp(ebml.doctype, matroska_doctypes[i]))
3046 break;
3047 if (i >= FF_ARRAY_ELEMS(matroska_doctypes)) {
3048 av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
3049 if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
3050 ebml_free(ebml_syntax, &ebml);
3051 return AVERROR_INVALIDDATA;
3052 }
3053 }
3054 ebml_free(ebml_syntax, &ebml);
3055
3056 matroska->pkt = si->parse_pkt;
3057
3058 /* The next thing is a segment. */
3059 pos = avio_tell(matroska->ctx->pb);
3060 res = ebml_parse(matroska, matroska_segments, matroska);
3061 // Try resyncing until we find an EBML_STOP type element.
3062 while (res != 1) {
3063 res = matroska_resync(matroska, pos);
3064 if (res < 0)
3065 return res;
3066 pos = avio_tell(matroska->ctx->pb);
3067 res = ebml_parse(matroska, matroska_segment, matroska);
3068 if (res == AVERROR(EIO)) // EOF is translated to EIO, this exists the loop on EOF
3069 return res;
3070 }
3071 /* Set data_offset as it might be needed later by seek_frame_generic. */
3072 if (matroska->current_id == MATROSKA_ID_CLUSTER)
3073 si->data_offset = avio_tell(matroska->ctx->pb) - 4;
3074 matroska_execute_seekhead(matroska);
3075
3076 if (!matroska->time_scale)
3077 matroska->time_scale = 1000000;
3078 if (isnan(matroska->duration))
3079 matroska->duration = 0;
3080 if (matroska->duration)
3081 matroska->ctx->duration = matroska->duration * matroska->time_scale *
3082 1000 / AV_TIME_BASE;
3083 av_dict_set(&s->metadata, "title", matroska->title, 0);
3084 av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
3085
3086 if (matroska->date_utc.size == 8)
3087 matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
3088
3089 res = matroska_parse_tracks(s);
3090 if (res < 0)
3091 return res;
3092
3093 attachments = attachments_list->elem;
3094 for (j = 0; j < attachments_list->nb_elem; j++) {
3095 if (!(attachments[j].filename && attachments[j].mime &&
3096 attachments[j].bin.data && attachments[j].bin.size > 0)) {
3097 av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
3098 } else {
3099 AVStream *st = avformat_new_stream(s, NULL);
3100 if (!st)
3101 break;
3102 av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
3103 av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
3104 if (attachments[j].description)
3105 av_dict_set(&st->metadata, "title", attachments[j].description, 0);
3106 st->codecpar->codec_id = AV_CODEC_ID_NONE;
3107
3108 for (i = 0; mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3109 if (av_strstart(attachments[j].mime, mkv_image_mime_tags[i].str, NULL)) {
3110 st->codecpar->codec_id = mkv_image_mime_tags[i].id;
3111 break;
3112 }
3113 }
3114
3115 attachments[j].stream = st;
3116
3117 if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
3118 res = ff_add_attached_pic(s, st, NULL, &attachments[j].bin.buf, 0);
3119 if (res < 0)
3120 return res;
3121 } else {
3122 st->codecpar->codec_type = AVMEDIA_TYPE_ATTACHMENT;
3123 if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
3124 break;
3125 memcpy(st->codecpar->extradata, attachments[j].bin.data,
3126 attachments[j].bin.size);
3127
3128 for (i = 0; mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
3129 if (av_strstart(attachments[j].mime, mkv_mime_tags[i].str, NULL)) {
3130 st->codecpar->codec_id = mkv_mime_tags[i].id;
3131 break;
3132 }
3133 }
3134 }
3135 }
3136 }
3137
3138 chapters = chapters_list->elem;
3139 for (i = 0; i < chapters_list->nb_elem; i++)
3140 if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
3141 (max_start == 0 || chapters[i].start > max_start)) {
3142 chapters[i].chapter =
3143 avpriv_new_chapter(s, chapters[i].uid,
3144 (AVRational) { 1, 1000000000 },
3145 chapters[i].start, chapters[i].end,
3146 chapters[i].title);
3147 max_start = chapters[i].start;
3148 }
3149
3150 matroska_add_index_entries(matroska);
3151
3152 matroska_convert_tags(s);
3153
3154 return 0;
3155 }
3156
3157 /*
3158 * Put one packet in an application-supplied AVPacket struct.
3159 * Returns 0 on success or -1 on failure.
3160 */
matroska_deliver_packet(MatroskaDemuxContext *matroska, AVPacket *pkt)3161 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
3162 AVPacket *pkt)
3163 {
3164 if (matroska->queue.head) {
3165 MatroskaTrack *tracks = matroska->tracks.elem;
3166 MatroskaTrack *track;
3167
3168 avpriv_packet_list_get(&matroska->queue, pkt);
3169 track = &tracks[pkt->stream_index];
3170 if (track->has_palette) {
3171 uint8_t *pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
3172 if (!pal) {
3173 av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3174 } else {
3175 memcpy(pal, track->palette, AVPALETTE_SIZE);
3176 }
3177 track->has_palette = 0;
3178 }
3179 return 0;
3180 }
3181
3182 return -1;
3183 }
3184
3185 /*
3186 * Free all packets in our internal queue.
3187 */
matroska_clear_queue(MatroskaDemuxContext *matroska)3188 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3189 {
3190 avpriv_packet_list_free(&matroska->queue);
3191 }
3192
matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf, int size, int type, AVIOContext *pb, uint32_t lace_size[256], int *laces)3193 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3194 int size, int type, AVIOContext *pb,
3195 uint32_t lace_size[256], int *laces)
3196 {
3197 int n;
3198 uint8_t *data = *buf;
3199
3200 if (!type) {
3201 *laces = 1;
3202 lace_size[0] = size;
3203 return 0;
3204 }
3205
3206 if (size <= 0)
3207 return AVERROR_INVALIDDATA;
3208
3209 *laces = *data + 1;
3210 data += 1;
3211 size -= 1;
3212
3213 switch (type) {
3214 case 0x1: /* Xiph lacing */
3215 {
3216 uint8_t temp;
3217 uint32_t total = 0;
3218 for (n = 0; n < *laces - 1; n++) {
3219 lace_size[n] = 0;
3220
3221 do {
3222 if (size <= total)
3223 return AVERROR_INVALIDDATA;
3224 temp = *data;
3225 total += temp;
3226 lace_size[n] += temp;
3227 data += 1;
3228 size -= 1;
3229 } while (temp == 0xff);
3230 }
3231 if (size < total)
3232 return AVERROR_INVALIDDATA;
3233
3234 lace_size[n] = size - total;
3235 break;
3236 }
3237
3238 case 0x2: /* fixed-size lacing */
3239 if (size % (*laces))
3240 return AVERROR_INVALIDDATA;
3241 for (n = 0; n < *laces; n++)
3242 lace_size[n] = size / *laces;
3243 break;
3244
3245 case 0x3: /* EBML lacing */
3246 {
3247 uint64_t num;
3248 uint64_t total;
3249 int offset;
3250
3251 avio_skip(pb, 4);
3252
3253 n = ebml_read_num(matroska, pb, 8, &num, 1);
3254 if (n < 0)
3255 return n;
3256 if (num > INT_MAX)
3257 return AVERROR_INVALIDDATA;
3258
3259 total = lace_size[0] = num;
3260 offset = n;
3261 for (n = 1; n < *laces - 1; n++) {
3262 int64_t snum;
3263 int r;
3264 r = matroska_ebmlnum_sint(matroska, pb, &snum);
3265 if (r < 0)
3266 return r;
3267 if (lace_size[n - 1] + snum > (uint64_t)INT_MAX)
3268 return AVERROR_INVALIDDATA;
3269
3270 lace_size[n] = lace_size[n - 1] + snum;
3271 total += lace_size[n];
3272 offset += r;
3273 }
3274 data += offset;
3275 size -= offset;
3276 if (size < total)
3277 return AVERROR_INVALIDDATA;
3278
3279 lace_size[*laces - 1] = size - total;
3280 break;
3281 }
3282 }
3283
3284 *buf = data;
3285
3286 return 0;
3287 }
3288
matroska_parse_rm_audio(MatroskaDemuxContext *matroska, MatroskaTrack *track, AVStream *st, uint8_t *data, int size, uint64_t timecode, int64_t pos)3289 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3290 MatroskaTrack *track, AVStream *st,
3291 uint8_t *data, int size, uint64_t timecode,
3292 int64_t pos)
3293 {
3294 const int a = st->codecpar->block_align;
3295 const int sps = track->audio.sub_packet_size;
3296 const int cfs = track->audio.coded_framesize;
3297 const int h = track->audio.sub_packet_h;
3298 const int w = track->audio.frame_size;
3299 int y = track->audio.sub_packet_cnt;
3300 int x;
3301
3302 if (!track->audio.pkt_cnt) {
3303 if (track->audio.sub_packet_cnt == 0)
3304 track->audio.buf_timecode = timecode;
3305 if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3306 if (size < cfs * h / 2) {
3307 av_log(matroska->ctx, AV_LOG_ERROR,
3308 "Corrupt int4 RM-style audio packet size\n");
3309 return AVERROR_INVALIDDATA;
3310 }
3311 for (x = 0; x < h / 2; x++)
3312 memcpy(track->audio.buf + x * 2 * w + y * cfs,
3313 data + x * cfs, cfs);
3314 } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3315 if (size < w) {
3316 av_log(matroska->ctx, AV_LOG_ERROR,
3317 "Corrupt sipr RM-style audio packet size\n");
3318 return AVERROR_INVALIDDATA;
3319 }
3320 memcpy(track->audio.buf + y * w, data, w);
3321 } else {
3322 if (size < w) {
3323 av_log(matroska->ctx, AV_LOG_ERROR,
3324 "Corrupt generic RM-style audio packet size\n");
3325 return AVERROR_INVALIDDATA;
3326 }
3327 for (x = 0; x < w / sps; x++)
3328 memcpy(track->audio.buf +
3329 sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3330 data + x * sps, sps);
3331 }
3332
3333 if (++track->audio.sub_packet_cnt >= h) {
3334 if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3335 ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3336 track->audio.sub_packet_cnt = 0;
3337 track->audio.pkt_cnt = h * w / a;
3338 }
3339 }
3340
3341 while (track->audio.pkt_cnt) {
3342 int ret;
3343 AVPacket *pkt = matroska->pkt;
3344
3345 ret = av_new_packet(pkt, a);
3346 if (ret < 0) {
3347 return ret;
3348 }
3349 memcpy(pkt->data,
3350 track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3351 a);
3352 pkt->pts = track->audio.buf_timecode;
3353 track->audio.buf_timecode = AV_NOPTS_VALUE;
3354 pkt->pos = pos;
3355 pkt->stream_index = st->index;
3356 ret = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
3357 if (ret < 0) {
3358 av_packet_unref(pkt);
3359 return AVERROR(ENOMEM);
3360 }
3361 }
3362
3363 return 0;
3364 }
3365
3366 /* reconstruct full wavpack blocks from mangled matroska ones */
matroska_parse_wavpack(MatroskaTrack *track, uint8_t **data, int *size)3367 static int matroska_parse_wavpack(MatroskaTrack *track,
3368 uint8_t **data, int *size)
3369 {
3370 uint8_t *dst = NULL;
3371 uint8_t *src = *data;
3372 int dstlen = 0;
3373 int srclen = *size;
3374 uint32_t samples;
3375 uint16_t ver;
3376 int ret, offset = 0;
3377
3378 if (srclen < 12)
3379 return AVERROR_INVALIDDATA;
3380
3381 av_assert1(track->stream->codecpar->extradata_size >= 2);
3382 ver = AV_RL16(track->stream->codecpar->extradata);
3383
3384 samples = AV_RL32(src);
3385 src += 4;
3386 srclen -= 4;
3387
3388 while (srclen >= 8) {
3389 int multiblock;
3390 uint32_t blocksize;
3391 uint8_t *tmp;
3392
3393 uint32_t flags = AV_RL32(src);
3394 uint32_t crc = AV_RL32(src + 4);
3395 src += 8;
3396 srclen -= 8;
3397
3398 multiblock = (flags & 0x1800) != 0x1800;
3399 if (multiblock) {
3400 if (srclen < 4) {
3401 ret = AVERROR_INVALIDDATA;
3402 goto fail;
3403 }
3404 blocksize = AV_RL32(src);
3405 src += 4;
3406 srclen -= 4;
3407 } else
3408 blocksize = srclen;
3409
3410 if (blocksize > srclen) {
3411 ret = AVERROR_INVALIDDATA;
3412 goto fail;
3413 }
3414
3415 tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3416 if (!tmp) {
3417 ret = AVERROR(ENOMEM);
3418 goto fail;
3419 }
3420 dst = tmp;
3421 dstlen += blocksize + 32;
3422
3423 AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3424 AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3425 AV_WL16(dst + offset + 8, ver); // version
3426 AV_WL16(dst + offset + 10, 0); // track/index_no
3427 AV_WL32(dst + offset + 12, 0); // total samples
3428 AV_WL32(dst + offset + 16, 0); // block index
3429 AV_WL32(dst + offset + 20, samples); // number of samples
3430 AV_WL32(dst + offset + 24, flags); // flags
3431 AV_WL32(dst + offset + 28, crc); // crc
3432 memcpy(dst + offset + 32, src, blocksize); // block data
3433
3434 src += blocksize;
3435 srclen -= blocksize;
3436 offset += blocksize + 32;
3437 }
3438
3439 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3440
3441 *data = dst;
3442 *size = dstlen;
3443
3444 return 0;
3445
3446 fail:
3447 av_freep(&dst);
3448 return ret;
3449 }
3450
matroska_parse_prores(MatroskaTrack *track, uint8_t **data, int *size)3451 static int matroska_parse_prores(MatroskaTrack *track,
3452 uint8_t **data, int *size)
3453 {
3454 uint8_t *dst;
3455 int dstlen = *size + 8;
3456
3457 dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3458 if (!dst)
3459 return AVERROR(ENOMEM);
3460
3461 AV_WB32(dst, dstlen);
3462 AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3463 memcpy(dst + 8, *data, dstlen - 8);
3464 memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3465
3466 *data = dst;
3467 *size = dstlen;
3468
3469 return 0;
3470 }
3471
matroska_parse_webvtt(MatroskaDemuxContext *matroska, MatroskaTrack *track, AVStream *st, uint8_t *data, int data_len, uint64_t timecode, uint64_t duration, int64_t pos)3472 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3473 MatroskaTrack *track,
3474 AVStream *st,
3475 uint8_t *data, int data_len,
3476 uint64_t timecode,
3477 uint64_t duration,
3478 int64_t pos)
3479 {
3480 AVPacket *pkt = matroska->pkt;
3481 uint8_t *id, *settings, *text, *buf;
3482 int id_len, settings_len, text_len;
3483 uint8_t *p, *q;
3484 int err;
3485
3486 if (data_len <= 0)
3487 return AVERROR_INVALIDDATA;
3488
3489 p = data;
3490 q = data + data_len;
3491
3492 id = p;
3493 id_len = -1;
3494 while (p < q) {
3495 if (*p == '\r' || *p == '\n') {
3496 id_len = p - id;
3497 if (*p == '\r')
3498 p++;
3499 break;
3500 }
3501 p++;
3502 }
3503
3504 if (p >= q || *p != '\n')
3505 return AVERROR_INVALIDDATA;
3506 p++;
3507
3508 settings = p;
3509 settings_len = -1;
3510 while (p < q) {
3511 if (*p == '\r' || *p == '\n') {
3512 settings_len = p - settings;
3513 if (*p == '\r')
3514 p++;
3515 break;
3516 }
3517 p++;
3518 }
3519
3520 if (p >= q || *p != '\n')
3521 return AVERROR_INVALIDDATA;
3522 p++;
3523
3524 text = p;
3525 text_len = q - p;
3526 while (text_len > 0) {
3527 const int len = text_len - 1;
3528 const uint8_t c = p[len];
3529 if (c != '\r' && c != '\n')
3530 break;
3531 text_len = len;
3532 }
3533
3534 if (text_len <= 0)
3535 return AVERROR_INVALIDDATA;
3536
3537 err = av_new_packet(pkt, text_len);
3538 if (err < 0) {
3539 return err;
3540 }
3541
3542 memcpy(pkt->data, text, text_len);
3543
3544 if (id_len > 0) {
3545 buf = av_packet_new_side_data(pkt,
3546 AV_PKT_DATA_WEBVTT_IDENTIFIER,
3547 id_len);
3548 if (!buf) {
3549 av_packet_unref(pkt);
3550 return AVERROR(ENOMEM);
3551 }
3552 memcpy(buf, id, id_len);
3553 }
3554
3555 if (settings_len > 0) {
3556 buf = av_packet_new_side_data(pkt,
3557 AV_PKT_DATA_WEBVTT_SETTINGS,
3558 settings_len);
3559 if (!buf) {
3560 av_packet_unref(pkt);
3561 return AVERROR(ENOMEM);
3562 }
3563 memcpy(buf, settings, settings_len);
3564 }
3565
3566 // Do we need this for subtitles?
3567 // pkt->flags = AV_PKT_FLAG_KEY;
3568
3569 pkt->stream_index = st->index;
3570 pkt->pts = timecode;
3571
3572 // Do we need this for subtitles?
3573 // pkt->dts = timecode;
3574
3575 pkt->duration = duration;
3576 pkt->pos = pos;
3577
3578 err = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
3579 if (err < 0) {
3580 av_packet_unref(pkt);
3581 return AVERROR(ENOMEM);
3582 }
3583
3584 return 0;
3585 }
3586
matroska_parse_frame(MatroskaDemuxContext *matroska, MatroskaTrack *track, AVStream *st, AVBufferRef *buf, uint8_t *data, int pkt_size, uint64_t timecode, uint64_t lace_duration, int64_t pos, int is_keyframe, uint8_t *additional, uint64_t additional_id, int additional_size, int64_t discard_padding)3587 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3588 MatroskaTrack *track, AVStream *st,
3589 AVBufferRef *buf, uint8_t *data, int pkt_size,
3590 uint64_t timecode, uint64_t lace_duration,
3591 int64_t pos, int is_keyframe,
3592 uint8_t *additional, uint64_t additional_id, int additional_size,
3593 int64_t discard_padding)
3594 {
3595 uint8_t *pkt_data = data;
3596 int res = 0;
3597 AVPacket *pkt = matroska->pkt;
3598
3599 if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3600 res = matroska_parse_wavpack(track, &pkt_data, &pkt_size);
3601 if (res < 0) {
3602 av_log(matroska->ctx, AV_LOG_ERROR,
3603 "Error parsing a wavpack block.\n");
3604 goto fail;
3605 }
3606 if (!buf)
3607 av_freep(&data);
3608 buf = NULL;
3609 }
3610
3611 if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3612 AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3613 res = matroska_parse_prores(track, &pkt_data, &pkt_size);
3614 if (res < 0) {
3615 av_log(matroska->ctx, AV_LOG_ERROR,
3616 "Error parsing a prores block.\n");
3617 goto fail;
3618 }
3619 if (!buf)
3620 av_freep(&data);
3621 buf = NULL;
3622 }
3623
3624 if (!pkt_size && !additional_size)
3625 goto no_output;
3626
3627 if (!buf)
3628 pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3629 NULL, NULL, 0);
3630 else
3631 pkt->buf = av_buffer_ref(buf);
3632
3633 if (!pkt->buf) {
3634 res = AVERROR(ENOMEM);
3635 goto fail;
3636 }
3637
3638 pkt->data = pkt_data;
3639 pkt->size = pkt_size;
3640 pkt->flags = is_keyframe;
3641 pkt->stream_index = st->index;
3642
3643 if (additional_size > 0) {
3644 uint8_t *side_data = av_packet_new_side_data(pkt,
3645 AV_PKT_DATA_MATROSKA_BLOCKADDITIONAL,
3646 additional_size + 8);
3647 if (!side_data) {
3648 av_packet_unref(pkt);
3649 return AVERROR(ENOMEM);
3650 }
3651 AV_WB64(side_data, additional_id);
3652 memcpy(side_data + 8, additional, additional_size);
3653 }
3654
3655 if (discard_padding) {
3656 uint8_t *side_data = av_packet_new_side_data(pkt,
3657 AV_PKT_DATA_SKIP_SAMPLES,
3658 10);
3659 if (!side_data) {
3660 av_packet_unref(pkt);
3661 return AVERROR(ENOMEM);
3662 }
3663 discard_padding = av_rescale_q(discard_padding,
3664 (AVRational){1, 1000000000},
3665 (AVRational){1, st->codecpar->sample_rate});
3666 if (discard_padding > 0) {
3667 AV_WL32(side_data + 4, discard_padding);
3668 } else {
3669 AV_WL32(side_data, -discard_padding);
3670 }
3671 }
3672
3673 if (track->ms_compat)
3674 pkt->dts = timecode;
3675 else
3676 pkt->pts = timecode;
3677 pkt->pos = pos;
3678 pkt->duration = lace_duration;
3679
3680 res = avpriv_packet_list_put(&matroska->queue, pkt, NULL, 0);
3681 if (res < 0) {
3682 av_packet_unref(pkt);
3683 return AVERROR(ENOMEM);
3684 }
3685
3686 return 0;
3687
3688 no_output:
3689 fail:
3690 if (!buf)
3691 av_free(pkt_data);
3692 return res;
3693 }
3694
matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data, int size, int64_t pos, uint64_t cluster_time, uint64_t block_duration, int is_keyframe, uint8_t *additional, uint64_t additional_id, int additional_size, int64_t cluster_pos, int64_t discard_padding)3695 static int matroska_parse_block(MatroskaDemuxContext *matroska, AVBufferRef *buf, uint8_t *data,
3696 int size, int64_t pos, uint64_t cluster_time,
3697 uint64_t block_duration, int is_keyframe,
3698 uint8_t *additional, uint64_t additional_id, int additional_size,
3699 int64_t cluster_pos, int64_t discard_padding)
3700 {
3701 uint64_t timecode = AV_NOPTS_VALUE;
3702 MatroskaTrack *track;
3703 FFIOContext pb;
3704 int res = 0;
3705 AVStream *st;
3706 int16_t block_time;
3707 uint32_t lace_size[256];
3708 int n, flags, laces = 0;
3709 uint64_t num;
3710 int trust_default_duration;
3711
3712 av_assert1(buf);
3713
3714 ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
3715
3716 if ((n = ebml_read_num(matroska, &pb.pub, 8, &num, 1)) < 0)
3717 return n;
3718 data += n;
3719 size -= n;
3720
3721 track = matroska_find_track_by_num(matroska, num);
3722 if (!track || size < 3)
3723 return AVERROR_INVALIDDATA;
3724
3725 if (!(st = track->stream)) {
3726 av_log(matroska->ctx, AV_LOG_VERBOSE,
3727 "No stream associated to TrackNumber %"PRIu64". "
3728 "Ignoring Block with this TrackNumber.\n", num);
3729 return 0;
3730 }
3731
3732 if (st->discard >= AVDISCARD_ALL)
3733 return res;
3734 if (block_duration > INT64_MAX)
3735 block_duration = INT64_MAX;
3736
3737 block_time = sign_extend(AV_RB16(data), 16);
3738 data += 2;
3739 flags = *data++;
3740 size -= 3;
3741 if (is_keyframe == -1)
3742 is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3743
3744 if (cluster_time != (uint64_t) -1 &&
3745 (block_time >= 0 || cluster_time >= -block_time)) {
3746 uint64_t timecode_cluster_in_track_tb = (double) cluster_time / track->time_scale;
3747 timecode = timecode_cluster_in_track_tb + block_time - track->codec_delay_in_track_tb;
3748 if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3749 timecode < track->end_timecode)
3750 is_keyframe = 0; /* overlapping subtitles are not key frame */
3751 if (is_keyframe) {
3752 ff_reduce_index(matroska->ctx, st->index);
3753 av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3754 AVINDEX_KEYFRAME);
3755 }
3756 }
3757
3758 if (matroska->skip_to_keyframe &&
3759 track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3760 // Compare signed timecodes. Timecode may be negative due to codec delay
3761 // offset. We don't support timestamps greater than int64_t anyway - see
3762 // AVPacket's pts.
3763 if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3764 return res;
3765 if (is_keyframe)
3766 matroska->skip_to_keyframe = 0;
3767 else if (!ffstream(st)->skip_to_keyframe) {
3768 av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3769 matroska->skip_to_keyframe = 0;
3770 }
3771 }
3772
3773 res = matroska_parse_laces(matroska, &data, size, (flags & 0x06) >> 1,
3774 &pb.pub, lace_size, &laces);
3775 if (res < 0) {
3776 av_log(matroska->ctx, AV_LOG_ERROR, "Error parsing frame sizes.\n");
3777 return res;
3778 }
3779
3780 trust_default_duration = track->default_duration != 0;
3781 if (track->audio.samplerate == 8000 && trust_default_duration) {
3782 // If this is needed for more codecs, then add them here
3783 if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3784 if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3785 trust_default_duration = 0;
3786 }
3787 }
3788
3789 if (!block_duration && trust_default_duration)
3790 block_duration = track->default_duration * laces / matroska->time_scale;
3791
3792 if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3793 track->end_timecode =
3794 FFMAX(track->end_timecode, timecode + block_duration);
3795
3796 for (n = 0; n < laces; n++) {
3797 int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3798 uint8_t *out_data = data;
3799 int out_size = lace_size[n];
3800
3801 if (track->needs_decoding) {
3802 res = matroska_decode_buffer(&out_data, &out_size, track);
3803 if (res < 0)
3804 return res;
3805 /* Given that we are here means that out_data is no longer
3806 * owned by buf, so set it to NULL. This depends upon
3807 * zero-length header removal compression being ignored. */
3808 av_assert1(out_data != data);
3809 buf = NULL;
3810 }
3811
3812 if (track->audio.buf) {
3813 res = matroska_parse_rm_audio(matroska, track, st,
3814 out_data, out_size,
3815 timecode, pos);
3816 if (!buf)
3817 av_free(out_data);
3818 if (res)
3819 return res;
3820 } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3821 res = matroska_parse_webvtt(matroska, track, st,
3822 out_data, out_size,
3823 timecode, lace_duration,
3824 pos);
3825 if (!buf)
3826 av_free(out_data);
3827 if (res)
3828 return res;
3829 } else {
3830 res = matroska_parse_frame(matroska, track, st, buf, out_data,
3831 out_size, timecode, lace_duration,
3832 pos, !n ? is_keyframe : 0,
3833 additional, additional_id, additional_size,
3834 discard_padding);
3835 if (res)
3836 return res;
3837 }
3838
3839 if (timecode != AV_NOPTS_VALUE)
3840 timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3841 data += lace_size[n];
3842 }
3843
3844 return 0;
3845 }
3846
matroska_parse_cluster(MatroskaDemuxContext *matroska)3847 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3848 {
3849 MatroskaCluster *cluster = &matroska->current_cluster;
3850 MatroskaBlock *block = &cluster->block;
3851 int res;
3852
3853 av_assert0(matroska->num_levels <= 2);
3854
3855 if (matroska->num_levels == 1) {
3856 res = ebml_parse(matroska, matroska_segment, NULL);
3857
3858 if (res == 1) {
3859 /* Found a cluster: subtract the size of the ID already read. */
3860 cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3861
3862 res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3863 if (res < 0)
3864 return res;
3865 }
3866 }
3867
3868 if (matroska->num_levels == 2) {
3869 /* We are inside a cluster. */
3870 res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3871
3872 if (res >= 0 && block->bin.size > 0) {
3873 int is_keyframe = block->non_simple ? block->reference.count == 0 : -1;
3874 uint8_t* additional = block->additional.size > 0 ?
3875 block->additional.data : NULL;
3876
3877 res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3878 block->bin.size, block->bin.pos,
3879 cluster->timecode, block->duration,
3880 is_keyframe, additional, block->additional_id,
3881 block->additional.size, cluster->pos,
3882 block->discard_padding);
3883 }
3884
3885 ebml_free(matroska_blockgroup, block);
3886 memset(block, 0, sizeof(*block));
3887 } else if (!matroska->num_levels) {
3888 if (!avio_feof(matroska->ctx->pb)) {
3889 avio_r8(matroska->ctx->pb);
3890 if (!avio_feof(matroska->ctx->pb)) {
3891 av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3892 "end of segment.\n");
3893 return AVERROR_INVALIDDATA;
3894 }
3895 }
3896 matroska->done = 1;
3897 return AVERROR_EOF;
3898 }
3899
3900 return res;
3901 }
3902
matroska_read_packet(AVFormatContext *s, AVPacket *pkt)3903 static int matroska_read_packet(AVFormatContext *s, AVPacket *pkt)
3904 {
3905 MatroskaDemuxContext *matroska = s->priv_data;
3906 int ret = 0;
3907
3908 if (matroska->resync_pos == -1) {
3909 // This can only happen if generic seeking has been used.
3910 matroska->resync_pos = avio_tell(s->pb);
3911 }
3912
3913 while (matroska_deliver_packet(matroska, pkt)) {
3914 if (matroska->done)
3915 return (ret < 0) ? ret : AVERROR_EOF;
3916 if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3917 ret = matroska_resync(matroska, matroska->resync_pos);
3918 }
3919
3920 return 0;
3921 }
3922
matroska_read_seek(AVFormatContext *s, int stream_index, int64_t timestamp, int flags)3923 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3924 int64_t timestamp, int flags)
3925 {
3926 MatroskaDemuxContext *matroska = s->priv_data;
3927 MatroskaTrack *tracks = NULL;
3928 AVStream *st = s->streams[stream_index];
3929 FFStream *const sti = ffstream(st);
3930 int i, index;
3931
3932 /* Parse the CUES now since we need the index data to seek. */
3933 if (matroska->cues_parsing_deferred > 0) {
3934 matroska->cues_parsing_deferred = 0;
3935 matroska_parse_cues(matroska);
3936 }
3937
3938 if (!sti->nb_index_entries)
3939 goto err;
3940 timestamp = FFMAX(timestamp, sti->index_entries[0].timestamp);
3941
3942 if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 ||
3943 index == sti->nb_index_entries - 1) {
3944 matroska_reset_status(matroska, 0, sti->index_entries[sti->nb_index_entries - 1].pos);
3945 while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 ||
3946 index == sti->nb_index_entries - 1) {
3947 matroska_clear_queue(matroska);
3948 if (matroska_parse_cluster(matroska) < 0)
3949 break;
3950 }
3951 }
3952
3953 matroska_clear_queue(matroska);
3954 if (index < 0 || (matroska->cues_parsing_deferred < 0 &&
3955 index == sti->nb_index_entries - 1))
3956 goto err;
3957
3958 tracks = matroska->tracks.elem;
3959 for (i = 0; i < matroska->tracks.nb_elem; i++) {
3960 tracks[i].audio.pkt_cnt = 0;
3961 tracks[i].audio.sub_packet_cnt = 0;
3962 tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3963 tracks[i].end_timecode = 0;
3964 }
3965
3966 /* We seek to a level 1 element, so set the appropriate status. */
3967 matroska_reset_status(matroska, 0, sti->index_entries[index].pos);
3968 if (flags & AVSEEK_FLAG_ANY) {
3969 sti->skip_to_keyframe = 0;
3970 matroska->skip_to_timecode = timestamp;
3971 } else {
3972 sti->skip_to_keyframe = 1;
3973 matroska->skip_to_timecode = sti->index_entries[index].timestamp;
3974 }
3975 matroska->skip_to_keyframe = 1;
3976 matroska->done = 0;
3977 avpriv_update_cur_dts(s, st, sti->index_entries[index].timestamp);
3978 return 0;
3979 err:
3980 // slightly hackish but allows proper fallback to
3981 // the generic seeking code.
3982 matroska_reset_status(matroska, 0, -1);
3983 matroska->resync_pos = -1;
3984 matroska_clear_queue(matroska);
3985 sti->skip_to_keyframe =
3986 matroska->skip_to_keyframe = 0;
3987 matroska->done = 0;
3988 return -1;
3989 }
3990
matroska_read_close(AVFormatContext *s)3991 static int matroska_read_close(AVFormatContext *s)
3992 {
3993 MatroskaDemuxContext *matroska = s->priv_data;
3994 MatroskaTrack *tracks = matroska->tracks.elem;
3995 int n;
3996
3997 matroska_clear_queue(matroska);
3998
3999 for (n = 0; n < matroska->tracks.nb_elem; n++)
4000 if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
4001 av_freep(&tracks[n].audio.buf);
4002 ebml_free(matroska_segment, matroska);
4003
4004 return 0;
4005 }
4006
4007 #if CONFIG_WEBM_DASH_MANIFEST_DEMUXER
4008 typedef struct {
4009 int64_t start_time_ns;
4010 int64_t end_time_ns;
4011 int64_t start_offset;
4012 int64_t end_offset;
4013 } CueDesc;
4014
4015 /* This function searches all the Cues and returns the CueDesc corresponding to
4016 * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
4017 * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
4018 */
get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start)4019 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
4020 MatroskaDemuxContext *matroska = s->priv_data;
4021 FFStream *const sti = ffstream(s->streams[0]);
4022 AVIndexEntry *const index_entries = sti->index_entries;
4023 int nb_index_entries = sti->nb_index_entries;
4024 CueDesc cue_desc;
4025 int i;
4026
4027 if (ts >= (int64_t)(matroska->duration * matroska->time_scale))
4028 return (CueDesc) {-1, -1, -1, -1};
4029 for (i = 1; i < nb_index_entries; i++) {
4030 if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
4031 index_entries[i].timestamp * matroska->time_scale > ts) {
4032 break;
4033 }
4034 }
4035 --i;
4036 cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
4037 cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
4038 if (i != nb_index_entries - 1) {
4039 cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
4040 cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
4041 } else {
4042 cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
4043 // FIXME: this needs special handling for files where Cues appear
4044 // before Clusters. the current logic assumes Cues appear after
4045 // Clusters.
4046 cue_desc.end_offset = cues_start - matroska->segment_start;
4047 }
4048 return cue_desc;
4049 }
4050
webm_clusters_start_with_keyframe(AVFormatContext *s)4051 static int webm_clusters_start_with_keyframe(AVFormatContext *s)
4052 {
4053 MatroskaDemuxContext *matroska = s->priv_data;
4054 AVStream *const st = s->streams[0];
4055 FFStream *const sti = ffstream(st);
4056 uint32_t id = matroska->current_id;
4057 int64_t cluster_pos, before_pos;
4058 int index, rv = 1;
4059
4060 if (sti->nb_index_entries <= 0)
4061 return 0;
4062
4063 // seek to the first cluster using cues.
4064 index = av_index_search_timestamp(st, 0, 0);
4065 if (index < 0)
4066 return 0;
4067 cluster_pos = sti->index_entries[index].pos;
4068 before_pos = avio_tell(s->pb);
4069 while (1) {
4070 uint64_t cluster_id, cluster_length;
4071 int read;
4072 AVPacket *pkt;
4073 avio_seek(s->pb, cluster_pos, SEEK_SET);
4074 // read cluster id and length
4075 read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
4076 if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
4077 break;
4078 read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
4079 if (read < 0)
4080 break;
4081
4082 matroska_reset_status(matroska, 0, cluster_pos);
4083 matroska_clear_queue(matroska);
4084 if (matroska_parse_cluster(matroska) < 0 ||
4085 !matroska->queue.head) {
4086 break;
4087 }
4088 pkt = &matroska->queue.head->pkt;
4089 // 4 + read is the length of the cluster id and the cluster length field.
4090 cluster_pos += 4 + read + cluster_length;
4091 if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
4092 rv = 0;
4093 break;
4094 }
4095 }
4096
4097 /* Restore the status after matroska_read_header: */
4098 matroska_reset_status(matroska, id, before_pos);
4099
4100 return rv;
4101 }
4102
buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps, double min_buffer, double* buffer, double* sec_to_download, AVFormatContext *s, int64_t cues_start)4103 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
4104 double min_buffer, double* buffer,
4105 double* sec_to_download, AVFormatContext *s,
4106 int64_t cues_start)
4107 {
4108 double nano_seconds_per_second = 1000000000.0;
4109 double time_sec = time_ns / nano_seconds_per_second;
4110 int rv = 0;
4111 int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
4112 int64_t end_time_ns = time_ns + time_to_search_ns;
4113 double sec_downloaded = 0.0;
4114 CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
4115 if (desc_curr.start_time_ns == -1)
4116 return -1;
4117 *sec_to_download = 0.0;
4118
4119 // Check for non cue start time.
4120 if (time_ns > desc_curr.start_time_ns) {
4121 int64_t cue_nano = desc_curr.end_time_ns - time_ns;
4122 double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
4123 double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
4124 double timeToDownload = (cueBytes * 8.0) / bps;
4125
4126 sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
4127 *sec_to_download += timeToDownload;
4128
4129 // Check if the search ends within the first cue.
4130 if (desc_curr.end_time_ns >= end_time_ns) {
4131 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4132 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4133 sec_downloaded = percent_to_sub * sec_downloaded;
4134 *sec_to_download = percent_to_sub * *sec_to_download;
4135 }
4136
4137 if ((sec_downloaded + *buffer) <= min_buffer) {
4138 return 1;
4139 }
4140
4141 // Get the next Cue.
4142 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4143 }
4144
4145 while (desc_curr.start_time_ns != -1) {
4146 int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
4147 int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
4148 double desc_sec = desc_ns / nano_seconds_per_second;
4149 double bits = (desc_bytes * 8.0);
4150 double time_to_download = bits / bps;
4151
4152 sec_downloaded += desc_sec - time_to_download;
4153 *sec_to_download += time_to_download;
4154
4155 if (desc_curr.end_time_ns >= end_time_ns) {
4156 double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4157 double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4158 sec_downloaded = percent_to_sub * sec_downloaded;
4159 *sec_to_download = percent_to_sub * *sec_to_download;
4160
4161 if ((sec_downloaded + *buffer) <= min_buffer)
4162 rv = 1;
4163 break;
4164 }
4165
4166 if ((sec_downloaded + *buffer) <= min_buffer) {
4167 rv = 1;
4168 break;
4169 }
4170
4171 desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4172 }
4173 *buffer = *buffer + sec_downloaded;
4174 return rv;
4175 }
4176
4177 /* This function computes the bandwidth of the WebM file with the help of
4178 * buffer_size_after_time_downloaded() function. Both of these functions are
4179 * adapted from WebM Tools project and are adapted to work with FFmpeg's
4180 * Matroska parsing mechanism.
4181 *
4182 * Returns the bandwidth of the file on success; -1 on error.
4183 * */
webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)4184 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4185 {
4186 MatroskaDemuxContext *matroska = s->priv_data;
4187 AVStream *st = s->streams[0];
4188 FFStream *const sti = ffstream(st);
4189 double bandwidth = 0.0;
4190
4191 for (int i = 0; i < sti->nb_index_entries; i++) {
4192 int64_t prebuffer_ns = 1000000000;
4193 int64_t time_ns = sti->index_entries[i].timestamp * matroska->time_scale;
4194 double nano_seconds_per_second = 1000000000.0;
4195 int64_t prebuffered_ns;
4196 double prebuffer_bytes = 0.0;
4197 int64_t temp_prebuffer_ns = prebuffer_ns;
4198 int64_t pre_bytes, pre_ns;
4199 double pre_sec, prebuffer, bits_per_second;
4200 CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4201 // Start with the first Cue.
4202 CueDesc desc_end = desc_beg;
4203
4204 if (time_ns > INT64_MAX - prebuffer_ns)
4205 return -1;
4206 prebuffered_ns = time_ns + prebuffer_ns;
4207
4208 // Figure out how much data we have downloaded for the prebuffer. This will
4209 // be used later to adjust the bits per sample to try.
4210 while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4211 // Prebuffered the entire Cue.
4212 prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4213 temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4214 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4215 }
4216 if (desc_end.start_time_ns == -1) {
4217 // The prebuffer is larger than the duration.
4218 if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4219 return -1;
4220 bits_per_second = 0.0;
4221 } else {
4222 // The prebuffer ends in the last Cue. Estimate how much data was
4223 // prebuffered.
4224 pre_bytes = desc_end.end_offset - desc_end.start_offset;
4225 pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4226 if (pre_ns <= 0)
4227 return -1;
4228 pre_sec = pre_ns / nano_seconds_per_second;
4229 prebuffer_bytes +=
4230 pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4231
4232 prebuffer = prebuffer_ns / nano_seconds_per_second;
4233
4234 // Set this to 0.0 in case our prebuffer buffers the entire video.
4235 bits_per_second = 0.0;
4236 do {
4237 int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4238 int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4239 double desc_sec, calc_bits_per_second, percent, mod_bits_per_second;
4240 if (desc_bytes <= 0)
4241 return -1;
4242
4243 desc_sec = desc_ns / nano_seconds_per_second;
4244 calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4245
4246 // Drop the bps by the percentage of bytes buffered.
4247 percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4248 mod_bits_per_second = calc_bits_per_second * percent;
4249
4250 if (prebuffer < desc_sec) {
4251 double search_sec =
4252 (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4253
4254 // Add 1 so the bits per second should be a little bit greater than file
4255 // datarate.
4256 int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4257 const double min_buffer = 0.0;
4258 double buffer = prebuffer;
4259 double sec_to_download = 0.0;
4260
4261 int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4262 min_buffer, &buffer, &sec_to_download,
4263 s, cues_start);
4264 if (rv < 0) {
4265 return -1;
4266 } else if (rv == 0) {
4267 bits_per_second = (double)(bps);
4268 break;
4269 }
4270 }
4271
4272 desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4273 } while (desc_end.start_time_ns != -1);
4274 }
4275 if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4276 }
4277 return (int64_t)bandwidth;
4278 }
4279
webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)4280 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4281 {
4282 MatroskaDemuxContext *matroska = s->priv_data;
4283 EbmlList *seekhead_list = &matroska->seekhead;
4284 MatroskaSeekhead *seekhead = seekhead_list->elem;
4285 AVStream *const st = s->streams[0];
4286 FFStream *const sti = ffstream(st);
4287 AVBPrint bprint;
4288 char *buf;
4289 int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4290 int i;
4291 int ret;
4292
4293 // determine cues start and end positions
4294 for (i = 0; i < seekhead_list->nb_elem; i++)
4295 if (seekhead[i].id == MATROSKA_ID_CUES)
4296 break;
4297
4298 if (i >= seekhead_list->nb_elem) return -1;
4299
4300 before_pos = avio_tell(matroska->ctx->pb);
4301 cues_start = seekhead[i].pos + matroska->segment_start;
4302 if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4303 // cues_end is computed as cues_start + cues_length + length of the
4304 // Cues element ID (i.e. 4) + EBML length of the Cues element.
4305 // cues_end is inclusive and the above sum is reduced by 1.
4306 uint64_t cues_length, cues_id;
4307 int bytes_read;
4308 bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4309 if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4310 return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4311 bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4312 if (bytes_read < 0)
4313 return bytes_read;
4314 cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4315 }
4316 avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4317 if (cues_start == -1 || cues_end == -1) return -1;
4318
4319 // parse the cues
4320 matroska_parse_cues(matroska);
4321
4322 if (!sti->nb_index_entries)
4323 return AVERROR_INVALIDDATA;
4324
4325 // cues start
4326 av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4327
4328 // cues end
4329 av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4330
4331 // if the file has cues at the start, fix up the init range so that
4332 // it does not include it
4333 if (cues_start <= init_range)
4334 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4335
4336 // bandwidth
4337 bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4338 if (bandwidth < 0) return -1;
4339 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4340
4341 // check if all clusters start with key frames
4342 av_dict_set_int(&s->streams[0]->metadata, CLUSTER_KEYFRAME, webm_clusters_start_with_keyframe(s), 0);
4343
4344 // Store cue point timestamps as a comma separated list
4345 // for checking subsegment alignment in the muxer.
4346 av_bprint_init(&bprint, 0, AV_BPRINT_SIZE_UNLIMITED);
4347 for (int i = 0; i < sti->nb_index_entries; i++)
4348 av_bprintf(&bprint, "%" PRId64",", sti->index_entries[i].timestamp);
4349 if (!av_bprint_is_complete(&bprint)) {
4350 av_bprint_finalize(&bprint, NULL);
4351 return AVERROR(ENOMEM);
4352 }
4353 // Remove the trailing ','
4354 bprint.str[--bprint.len] = '\0';
4355 if ((ret = av_bprint_finalize(&bprint, &buf)) < 0)
4356 return ret;
4357 av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS,
4358 buf, AV_DICT_DONT_STRDUP_VAL);
4359
4360 return 0;
4361 }
4362
webm_dash_manifest_read_header(AVFormatContext *s)4363 static int webm_dash_manifest_read_header(AVFormatContext *s)
4364 {
4365 char *buf;
4366 int ret = matroska_read_header(s);
4367 int64_t init_range;
4368 MatroskaTrack *tracks;
4369 MatroskaDemuxContext *matroska = s->priv_data;
4370 if (ret) {
4371 av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4372 return -1;
4373 }
4374 if (!matroska->tracks.nb_elem || !s->nb_streams) {
4375 av_log(s, AV_LOG_ERROR, "No track found\n");
4376 return AVERROR_INVALIDDATA;
4377 }
4378
4379 if (!matroska->is_live) {
4380 buf = av_asprintf("%g", matroska->duration);
4381 if (!buf)
4382 return AVERROR(ENOMEM);
4383 av_dict_set(&s->streams[0]->metadata, DURATION,
4384 buf, AV_DICT_DONT_STRDUP_VAL);
4385
4386 // initialization range
4387 // 5 is the offset of Cluster ID.
4388 init_range = avio_tell(s->pb) - 5;
4389 av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4390 }
4391
4392 // basename of the file
4393 buf = strrchr(s->url, '/');
4394 av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4395
4396 // track number
4397 tracks = matroska->tracks.elem;
4398 av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4399
4400 // parse the cues and populate Cue related fields
4401 if (!matroska->is_live) {
4402 ret = webm_dash_manifest_cues(s, init_range);
4403 if (ret < 0) {
4404 av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4405 return ret;
4406 }
4407 }
4408
4409 // use the bandwidth from the command line if it was provided
4410 if (matroska->bandwidth > 0) {
4411 av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4412 matroska->bandwidth, 0);
4413 }
4414 return 0;
4415 }
4416
webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)4417 static int webm_dash_manifest_read_packet(AVFormatContext *s, AVPacket *pkt)
4418 {
4419 return AVERROR_EOF;
4420 }
4421
4422 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4423 static const AVOption options[] = {
4424 { "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 },
4425 { "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 },
4426 { NULL },
4427 };
4428
4429 static const AVClass webm_dash_class = {
4430 .class_name = "WebM DASH Manifest demuxer",
4431 .item_name = av_default_item_name,
4432 .option = options,
4433 .version = LIBAVUTIL_VERSION_INT,
4434 };
4435
4436 const AVInputFormat ff_webm_dash_manifest_demuxer = {
4437 .name = "webm_dash_manifest",
4438 .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4439 .priv_class = &webm_dash_class,
4440 .priv_data_size = sizeof(MatroskaDemuxContext),
4441 .flags_internal = FF_FMT_INIT_CLEANUP,
4442 .read_header = webm_dash_manifest_read_header,
4443 .read_packet = webm_dash_manifest_read_packet,
4444 .read_close = matroska_read_close,
4445 };
4446 #endif
4447
4448 const AVInputFormat ff_matroska_demuxer = {
4449 .name = "matroska,webm",
4450 .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4451 .extensions = "mkv,mk3d,mka,mks,webm",
4452 .priv_data_size = sizeof(MatroskaDemuxContext),
4453 .flags_internal = FF_FMT_INIT_CLEANUP,
4454 .read_probe = matroska_probe,
4455 .read_header = matroska_read_header,
4456 .read_packet = matroska_read_packet,
4457 .read_close = matroska_read_close,
4458 .read_seek = matroska_read_seek,
4459 .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4460 };
4461