FFmpeg
matroskadec.c
Go to the documentation of this file.
1 /*
2  * Matroska file demuxer
3  * Copyright (c) 2003-2008 The FFmpeg Project
4  *
5  * This file is part of FFmpeg.
6  *
7  * FFmpeg is free software; you can redistribute it and/or
8  * modify it under the terms of the GNU Lesser General Public
9  * License as published by the Free Software Foundation; either
10  * version 2.1 of the License, or (at your option) any later version.
11  *
12  * FFmpeg is distributed in the hope that it will be useful,
13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15  * Lesser General Public License for more details.
16  *
17  * You should have received a copy of the GNU Lesser General Public
18  * License along with FFmpeg; if not, write to the Free Software
19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20  */
21 
22 /**
23  * @file
24  * Matroska file demuxer
25  * @author Ronald Bultje <rbultje@ronald.bitfreak.net>
26  * @author with a little help from Moritz Bunkus <moritz@bunkus.org>
27  * @author totally reworked by Aurelien Jacobs <aurel@gnuage.org>
28  * @see specs available on the Matroska project page: http://www.matroska.org/
29  */
30 
31 #include "config.h"
32 
33 #include <inttypes.h>
34 #include <stdio.h>
35 
36 #include "libavutil/avstring.h"
37 #include "libavutil/base64.h"
38 #include "libavutil/dict.h"
39 #include "libavutil/intfloat.h"
40 #include "libavutil/intreadwrite.h"
41 #include "libavutil/lzo.h"
43 #include "libavutil/mathematics.h"
44 #include "libavutil/opt.h"
46 #include "libavutil/spherical.h"
47 
48 #include "libavcodec/bytestream.h"
49 #include "libavcodec/flac.h"
50 #include "libavcodec/mpeg4audio.h"
51 
52 #include "avformat.h"
53 #include "avio_internal.h"
54 #include "internal.h"
55 #include "isom.h"
56 #include "matroska.h"
57 #include "oggdec.h"
58 /* For ff_codec_get_id(). */
59 #include "riff.h"
60 #include "rmsipr.h"
61 
62 #if CONFIG_BZLIB
63 #include <bzlib.h>
64 #endif
65 #if CONFIG_ZLIB
66 #include <zlib.h>
67 #endif
68 
69 #include "qtpalette.h"
70 
71 #define EBML_UNKNOWN_LENGTH UINT64_MAX /* EBML unknown length, in uint64_t */
72 #define NEEDS_CHECKING 2 /* Indicates that some error checks
73  * still need to be performed */
74 #define LEVEL_ENDED 3 /* return value of ebml_parse when the
75  * syntax level used for parsing ended. */
76 #define SKIP_THRESHOLD 1024 * 1024 /* In non-seekable mode, if more than SKIP_THRESHOLD
77  * of unkown, potentially damaged data is encountered,
78  * it is considered an error. */
79 #define UNKNOWN_EQUIV 50 * 1024 /* An unknown element is considered equivalent
80  * to this many bytes of unknown data for the
81  * SKIP_THRESHOLD check. */
82 
83 typedef enum {
89  EBML_UTF8,
90  EBML_BIN,
96 
97 typedef const struct EbmlSyntax {
98  uint32_t id;
101  size_t data_offset;
102  union {
103  int64_t i;
104  uint64_t u;
105  double f;
106  const char *s;
107  const struct EbmlSyntax *n;
108  } def;
109 } EbmlSyntax;
110 
111 typedef struct EbmlList {
112  int nb_elem;
113  unsigned int alloc_elem_size;
114  void *elem;
116 
117 typedef struct EbmlBin {
118  int size;
121  int64_t pos;
123 
124 typedef struct Ebml {
125  uint64_t version;
126  uint64_t max_size;
127  uint64_t id_length;
128  char *doctype;
129  uint64_t doctype_version;
130 } Ebml;
131 
132 typedef struct MatroskaTrackCompression {
133  uint64_t algo;
136 
137 typedef struct MatroskaTrackEncryption {
138  uint64_t algo;
141 
142 typedef struct MatroskaTrackEncoding {
143  uint64_t scope;
144  uint64_t type;
148 
149 typedef struct MatroskaMasteringMeta {
150  double r_x;
151  double r_y;
152  double g_x;
153  double g_y;
154  double b_x;
155  double b_y;
156  double white_x;
157  double white_y;
161 
162 typedef struct MatroskaTrackVideoColor {
165  uint64_t chroma_sub_horz;
166  uint64_t chroma_sub_vert;
167  uint64_t cb_sub_horz;
168  uint64_t cb_sub_vert;
171  uint64_t range;
172  uint64_t transfer_characteristics;
173  uint64_t primaries;
174  uint64_t max_cll;
175  uint64_t max_fall;
178 
179 typedef struct MatroskaTrackVideoProjection {
180  uint64_t type;
181  EbmlBin private;
182  double yaw;
183  double pitch;
184  double roll;
186 
187 typedef struct MatroskaTrackVideo {
188  double frame_rate;
189  uint64_t display_width;
190  uint64_t display_height;
191  uint64_t pixel_width;
192  uint64_t pixel_height;
194  uint64_t display_unit;
195  uint64_t interlaced;
196  uint64_t field_order;
197  uint64_t stereo_mode;
198  uint64_t alpha_mode;
202 
203 typedef struct MatroskaTrackAudio {
204  double samplerate;
206  uint64_t bitdepth;
207  uint64_t channels;
208 
209  /* real audio header (extracted from extradata) */
212  int frame_size;
213  int sub_packet_size;
215  int pkt_cnt;
216  uint64_t buf_timecode;
217  uint8_t *buf;
219 
220 typedef struct MatroskaTrackPlane {
221  uint64_t uid;
222  uint64_t type;
224 
225 typedef struct MatroskaTrackOperation {
228 
229 typedef struct MatroskaTrack {
230  uint64_t num;
231  uint64_t uid;
232  uint64_t type;
233  char *name;
234  char *codec_id;
236  char *language;
237  double time_scale;
239  uint64_t flag_default;
240  uint64_t flag_forced;
241  uint64_t seek_preroll;
246  uint64_t codec_delay;
247  uint64_t codec_delay_in_track_tb;
248 
250  int64_t end_timecode;
251  int ms_compat;
253 
257 
258 typedef struct MatroskaAttachment {
259  uint64_t uid;
260  char *filename;
261  char *mime;
263 
266 
267 typedef struct MatroskaChapter {
268  uint64_t start;
269  uint64_t end;
270  uint64_t uid;
271  char *title;
272 
275 
276 typedef struct MatroskaIndexPos {
277  uint64_t track;
278  uint64_t pos;
280 
281 typedef struct MatroskaIndex {
282  uint64_t time;
285 
286 typedef struct MatroskaTag {
287  char *name;
288  char *string;
289  char *lang;
290  uint64_t def;
293 
294 typedef struct MatroskaTagTarget {
295  char *type;
296  uint64_t typevalue;
297  uint64_t trackuid;
298  uint64_t chapteruid;
299  uint64_t attachuid;
301 
302 typedef struct MatroskaTags {
304  EbmlList tag;
305 } MatroskaTags;
306 
307 typedef struct MatroskaSeekhead {
308  uint64_t id;
309  uint64_t pos;
311 
312 typedef struct MatroskaLevel {
313  uint64_t start;
314  uint64_t length;
316 
317 typedef struct MatroskaBlock {
318  uint64_t duration;
319  int64_t reference;
320  uint64_t non_simple;
322  uint64_t additional_id;
325 } MatroskaBlock;
326 
327 typedef struct MatroskaCluster {
329  uint64_t timecode;
330  int64_t pos;
332 
333 typedef struct MatroskaLevel1Element {
334  int64_t pos;
335  uint32_t id;
336  int parsed;
338 
339 typedef struct MatroskaDemuxContext {
340  const AVClass *class;
342 
343  /* EBML stuff */
346  uint32_t current_id;
347  int64_t resync_pos;
349 
350  uint64_t time_scale;
351  double duration;
352  char *title;
353  char *muxingapp;
358  EbmlList index;
359  EbmlList tags;
361 
362  /* byte position of the segment inside the stream */
363  int64_t segment_start;
364 
365  /* the packet queue */
368 
369  int done;
370 
371  /* What to skip before effectively reading a packet. */
372  int skip_to_keyframe;
374 
375  /* File has a CUES element, but we defer parsing until it is needed. */
377 
378  /* Level1 elements and whether they were read yet */
380  int num_level1_elems;
381 
383 
384  /* WebM DASH Manifest live flag */
385  int is_live;
386 
387  /* Bandwidth value for WebM DASH Manifest */
388  int bandwidth;
390 
391 #define CHILD_OF(parent) { .def = { .n = parent } }
392 
393 // The following forward declarations need their size because
394 // a tentative definition with internal linkage must not be an
395 // incomplete type (6.7.2 in C90, 6.9.2 in C99).
396 // Removing the sizes breaks MSVC.
403 
404 static EbmlSyntax ebml_header[] = {
405  { EBML_ID_EBMLREADVERSION, EBML_UINT, 0, offsetof(Ebml, version), { .u = EBML_VERSION } },
406  { EBML_ID_EBMLMAXSIZELENGTH, EBML_UINT, 0, offsetof(Ebml, max_size), { .u = 8 } },
407  { EBML_ID_EBMLMAXIDLENGTH, EBML_UINT, 0, offsetof(Ebml, id_length), { .u = 4 } },
408  { EBML_ID_DOCTYPE, EBML_STR, 0, offsetof(Ebml, doctype), { .s = "(none)" } },
409  { EBML_ID_DOCTYPEREADVERSION, EBML_UINT, 0, offsetof(Ebml, doctype_version), { .u = 1 } },
413 };
414 
416  { EBML_ID_HEADER, EBML_NEST, 0, 0, { .n = ebml_header } },
418  { 0 }
419 };
420 
421 static EbmlSyntax matroska_info[] = {
422  { MATROSKA_ID_TIMECODESCALE, EBML_UINT, 0, offsetof(MatroskaDemuxContext, time_scale), { .u = 1000000 } },
424  { MATROSKA_ID_TITLE, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, title) },
426  { MATROSKA_ID_MUXINGAPP, EBML_UTF8, 0, offsetof(MatroskaDemuxContext, muxingapp) },
427  { MATROSKA_ID_DATEUTC, EBML_BIN, 0, offsetof(MatroskaDemuxContext, date_utc) },
430 };
431 
433  { MATROSKA_ID_VIDEOCOLOR_RX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_x), { .f=-1 } },
434  { MATROSKA_ID_VIDEOCOLOR_RY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, r_y), { .f=-1 } },
435  { MATROSKA_ID_VIDEOCOLOR_GX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_x), { .f=-1 } },
436  { MATROSKA_ID_VIDEOCOLOR_GY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, g_y), { .f=-1 } },
437  { MATROSKA_ID_VIDEOCOLOR_BX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_x), { .f=-1 } },
438  { MATROSKA_ID_VIDEOCOLOR_BY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, b_y), { .f=-1 } },
439  { MATROSKA_ID_VIDEOCOLOR_WHITEX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_x), { .f=-1 } },
440  { MATROSKA_ID_VIDEOCOLOR_WHITEY, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, white_y), { .f=-1 } },
441  { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMIN, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, min_luminance), { .f=-1 } },
442  { MATROSKA_ID_VIDEOCOLOR_LUMINANCEMAX, EBML_FLOAT, 0, offsetof(MatroskaMasteringMeta, max_luminance), { .f=-1 } },
444 };
445 
447  { MATROSKA_ID_VIDEOCOLORMATRIXCOEFF, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, matrix_coefficients), { .u = AVCOL_SPC_UNSPECIFIED } },
448  { MATROSKA_ID_VIDEOCOLORBITSPERCHANNEL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, bits_per_channel), { .u=0 } },
449  { MATROSKA_ID_VIDEOCOLORCHROMASUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_horz), { .u=0 } },
450  { MATROSKA_ID_VIDEOCOLORCHROMASUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, chroma_sub_vert), { .u=0 } },
451  { MATROSKA_ID_VIDEOCOLORCBSUBHORZ, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_horz), { .u=0 } },
452  { MATROSKA_ID_VIDEOCOLORCBSUBVERT, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, cb_sub_vert), { .u=0 } },
458  { MATROSKA_ID_VIDEOCOLORMAXCLL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_cll), { .u=0 } },
459  { MATROSKA_ID_VIDEOCOLORMAXFALL, EBML_UINT, 0, offsetof(MatroskaTrackVideoColor, max_fall), { .u=0 } },
462 };
463 
471 };
472 
473 static EbmlSyntax matroska_track_video[] = {
474  { MATROSKA_ID_VIDEOFRAMERATE, EBML_FLOAT, 0, offsetof(MatroskaTrackVideo, frame_rate) },
475  { MATROSKA_ID_VIDEODISPLAYWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_width), { .u=-1 } },
476  { MATROSKA_ID_VIDEODISPLAYHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, display_height), { .u=-1 } },
477  { MATROSKA_ID_VIDEOPIXELWIDTH, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_width) },
478  { MATROSKA_ID_VIDEOPIXELHEIGHT, EBML_UINT, 0, offsetof(MatroskaTrackVideo, pixel_height) },
479  { MATROSKA_ID_VIDEOCOLORSPACE, EBML_BIN, 0, offsetof(MatroskaTrackVideo, color_space) },
480  { MATROSKA_ID_VIDEOALPHAMODE, EBML_UINT, 0, offsetof(MatroskaTrackVideo, alpha_mode) },
493 };
494 
495 static EbmlSyntax matroska_track_audio[] = {
496  { MATROSKA_ID_AUDIOSAMPLINGFREQ, EBML_FLOAT, 0, offsetof(MatroskaTrackAudio, samplerate), { .f = 8000.0 } },
498  { MATROSKA_ID_AUDIOBITDEPTH, EBML_UINT, 0, offsetof(MatroskaTrackAudio, bitdepth) },
499  { MATROSKA_ID_AUDIOCHANNELS, EBML_UINT, 0, offsetof(MatroskaTrackAudio, channels), { .u = 1 } },
501 };
502 
507 };
508 
518 };
520  { MATROSKA_ID_ENCODINGSCOPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, scope), { .u = 1 } },
521  { MATROSKA_ID_ENCODINGTYPE, EBML_UINT, 0, offsetof(MatroskaTrackEncoding, type), { .u = 0 } },
526 };
527 
531 };
532 
533 static EbmlSyntax matroska_track_plane[] = {
537 };
538 
542 };
543 
547 };
548 
549 static EbmlSyntax matroska_track[] = {
550  { MATROSKA_ID_TRACKNUMBER, EBML_UINT, 0, offsetof(MatroskaTrack, num) },
552  { MATROSKA_ID_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTrack, uid) },
555  { MATROSKA_ID_CODECPRIVATE, EBML_BIN, 0, offsetof(MatroskaTrack, codec_priv) },
556  { MATROSKA_ID_CODECDELAY, EBML_UINT, 0, offsetof(MatroskaTrack, codec_delay) },
557  { MATROSKA_ID_TRACKLANGUAGE, EBML_UTF8, 0, offsetof(MatroskaTrack, language), { .s = "eng" } },
558  { MATROSKA_ID_TRACKDEFAULTDURATION, EBML_UINT, 0, offsetof(MatroskaTrack, default_duration) },
559  { MATROSKA_ID_TRACKTIMECODESCALE, EBML_FLOAT, 0, offsetof(MatroskaTrack, time_scale), { .f = 1.0 } },
560  { MATROSKA_ID_TRACKFLAGDEFAULT, EBML_UINT, 0, offsetof(MatroskaTrack, flag_default), { .u = 1 } },
561  { MATROSKA_ID_TRACKFLAGFORCED, EBML_UINT, 0, offsetof(MatroskaTrack, flag_forced), { .u = 0 } },
563  { MATROSKA_ID_TRACKAUDIO, EBML_NEST, 0, offsetof(MatroskaTrack, audio), { .n = matroska_track_audio } },
564  { MATROSKA_ID_TRACKOPERATION, EBML_NEST, 0, offsetof(MatroskaTrack, operation), { .n = matroska_track_operation } },
566  { MATROSKA_ID_TRACKMAXBLKADDID, EBML_UINT, 0, offsetof(MatroskaTrack, max_block_additional_id) },
567  { MATROSKA_ID_SEEKPREROLL, EBML_UINT, 0, offsetof(MatroskaTrack, seek_preroll) },
577 };
578 
579 static EbmlSyntax matroska_tracks[] = {
580  { MATROSKA_ID_TRACKENTRY, EBML_NEST, sizeof(MatroskaTrack), offsetof(MatroskaDemuxContext, tracks), { .n = matroska_track } },
582 };
583 
584 static EbmlSyntax matroska_attachment[] = {
586  { MATROSKA_ID_FILENAME, EBML_UTF8, 0, offsetof(MatroskaAttachment, filename) },
587  { MATROSKA_ID_FILEMIMETYPE, EBML_STR, 0, offsetof(MatroskaAttachment, mime) },
588  { MATROSKA_ID_FILEDATA, EBML_BIN, 0, offsetof(MatroskaAttachment, bin) },
591 };
592 
593 static EbmlSyntax matroska_attachments[] = {
594  { MATROSKA_ID_ATTACHEDFILE, EBML_NEST, sizeof(MatroskaAttachment), offsetof(MatroskaDemuxContext, attachments), { .n = matroska_attachment } },
596 };
597 
599  { MATROSKA_ID_CHAPSTRING, EBML_UTF8, 0, offsetof(MatroskaChapter, title) },
603 };
604 
615 };
616 
617 static EbmlSyntax matroska_chapter[] = {
624 };
625 
626 static EbmlSyntax matroska_chapters[] = {
629 };
630 
631 static EbmlSyntax matroska_index_pos[] = {
632  { MATROSKA_ID_CUETRACK, EBML_UINT, 0, offsetof(MatroskaIndexPos, track) },
638 };
639 
640 static EbmlSyntax matroska_index_entry[] = {
641  { MATROSKA_ID_CUETIME, EBML_UINT, 0, offsetof(MatroskaIndex, time) },
644 };
645 
646 static EbmlSyntax matroska_index[] = {
649 };
650 
651 static EbmlSyntax matroska_simpletag[] = {
652  { MATROSKA_ID_TAGNAME, EBML_UTF8, 0, offsetof(MatroskaTag, name) },
653  { MATROSKA_ID_TAGSTRING, EBML_UTF8, 0, offsetof(MatroskaTag, string) },
654  { MATROSKA_ID_TAGLANG, EBML_STR, 0, offsetof(MatroskaTag, lang), { .s = "und" } },
656  { MATROSKA_ID_TAGDEFAULT_BUG, EBML_UINT, 0, offsetof(MatroskaTag, def) },
657  { MATROSKA_ID_SIMPLETAG, EBML_NEST, sizeof(MatroskaTag), offsetof(MatroskaTag, sub), { .n = matroska_simpletag } },
659 };
660 
661 static EbmlSyntax matroska_tagtargets[] = {
663  { MATROSKA_ID_TAGTARGETS_TYPEVALUE, EBML_UINT, 0, offsetof(MatroskaTagTarget, typevalue), { .u = 50 } },
664  { MATROSKA_ID_TAGTARGETS_TRACKUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, trackuid) },
665  { MATROSKA_ID_TAGTARGETS_CHAPTERUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, chapteruid) },
666  { MATROSKA_ID_TAGTARGETS_ATTACHUID, EBML_UINT, 0, offsetof(MatroskaTagTarget, attachuid) },
668 };
669 
670 static EbmlSyntax matroska_tag[] = {
672  { MATROSKA_ID_TAGTARGETS, EBML_NEST, 0, offsetof(MatroskaTags, target), { .n = matroska_tagtargets } },
674 };
675 
676 static EbmlSyntax matroska_tags[] = {
677  { MATROSKA_ID_TAG, EBML_NEST, sizeof(MatroskaTags), offsetof(MatroskaDemuxContext, tags), { .n = matroska_tag } },
679 };
680 
682  { MATROSKA_ID_SEEKID, EBML_UINT, 0, offsetof(MatroskaSeekhead, id) },
683  { MATROSKA_ID_SEEKPOSITION, EBML_UINT, 0, offsetof(MatroskaSeekhead, pos), { .u = -1 } },
685 };
686 
687 static EbmlSyntax matroska_seekhead[] = {
690 };
691 
692 static EbmlSyntax matroska_segment[] = {
694  { MATROSKA_ID_INFO, EBML_LEVEL1, 0, 0, { .n = matroska_info } },
695  { MATROSKA_ID_TRACKS, EBML_LEVEL1, 0, 0, { .n = matroska_tracks } },
699  { MATROSKA_ID_TAGS, EBML_LEVEL1, 0, 0, { .n = matroska_tags } },
701  { 0 } /* We don't want to go back to level 0, so don't add the parent. */
702 };
703 
704 static EbmlSyntax matroska_segments[] = {
705  { MATROSKA_ID_SEGMENT, EBML_NEST, 0, 0, { .n = matroska_segment } },
706  { 0 }
707 };
708 
709 static EbmlSyntax matroska_blockmore[] = {
710  { MATROSKA_ID_BLOCKADDID, EBML_UINT, 0, offsetof(MatroskaBlock,additional_id), { .u = 1 } },
711  { MATROSKA_ID_BLOCKADDITIONAL, EBML_BIN, 0, offsetof(MatroskaBlock,additional) },
713 };
714 
718 };
719 
720 static EbmlSyntax matroska_blockgroup[] = {
721  { MATROSKA_ID_BLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
724  { MATROSKA_ID_DISCARDPADDING, EBML_SINT, 0, offsetof(MatroskaBlock, discard_padding) },
725  { MATROSKA_ID_BLOCKREFERENCE, EBML_SINT, 0, offsetof(MatroskaBlock, reference), { .i = INT64_MIN } },
727  { 1, EBML_UINT, 0, offsetof(MatroskaBlock, non_simple), { .u = 1 } },
729 };
730 
731 // The following array contains SimpleBlock and BlockGroup twice
732 // in order to reuse the other values for matroska_cluster_enter.
734  { MATROSKA_ID_SIMPLEBLOCK, EBML_BIN, 0, offsetof(MatroskaBlock, bin) },
736  { MATROSKA_ID_CLUSTERTIMECODE, EBML_UINT, 0, offsetof(MatroskaCluster, timecode) },
742 };
743 
745  { MATROSKA_ID_CLUSTER, EBML_NEST, 0, 0, { .n = &matroska_cluster_parsing[2] } },
746  { 0 }
747 };
748 #undef CHILD_OF
749 
750 static const char *const matroska_doctypes[] = { "matroska", "webm" };
751 
753 
754 /*
755  * This function prepares the status for parsing of level 1 elements.
756  */
757 static int matroska_reset_status(MatroskaDemuxContext *matroska,
758  uint32_t id, int64_t position)
759 {
760  int64_t err = 0;
761  if (position >= 0) {
762  err = avio_seek(matroska->ctx->pb, position, SEEK_SET);
763  if (err > 0)
764  err = 0;
765  } else
766  position = avio_tell(matroska->ctx->pb);
767 
768  matroska->current_id = id;
769  matroska->num_levels = 1;
770  matroska->unknown_count = 0;
771  matroska->resync_pos = position;
772  if (id)
773  matroska->resync_pos -= (av_log2(id) + 7) / 8;
774 
775  return err;
776 }
777 
778 static int matroska_resync(MatroskaDemuxContext *matroska, int64_t last_pos)
779 {
780  AVIOContext *pb = matroska->ctx->pb;
781  uint32_t id;
782 
783  /* Try to seek to the last position to resync from. If this doesn't work,
784  * we resync from the earliest position available: The start of the buffer. */
785  if (last_pos < avio_tell(pb) && avio_seek(pb, last_pos + 1, SEEK_SET) < 0) {
786  av_log(matroska->ctx, AV_LOG_WARNING,
787  "Seek to desired resync point failed. Seeking to "
788  "earliest point available instead.\n");
789  avio_seek(pb, FFMAX(avio_tell(pb) + (pb->buffer - pb->buf_ptr),
790  last_pos + 1), SEEK_SET);
791  }
792 
793  id = avio_rb32(pb);
794 
795  // try to find a toplevel element
796  while (!avio_feof(pb)) {
797  if (id == MATROSKA_ID_INFO || id == MATROSKA_ID_TRACKS ||
798  id == MATROSKA_ID_CUES || id == MATROSKA_ID_TAGS ||
800  id == MATROSKA_ID_CLUSTER || id == MATROSKA_ID_CHAPTERS) {
801  /* Prepare the context for parsing of a level 1 element. */
802  matroska_reset_status(matroska, id, -1);
803  /* Given that we are here means that an error has occurred,
804  * so treat the segment as unknown length in order not to
805  * discard valid data that happens to be beyond the designated
806  * end of the segment. */
807  matroska->levels[0].length = EBML_UNKNOWN_LENGTH;
808  return 0;
809  }
810  id = (id << 8) | avio_r8(pb);
811  }
812 
813  matroska->done = 1;
814  return pb->error ? pb->error : AVERROR_EOF;
815 }
816 
817 /*
818  * Read: an "EBML number", which is defined as a variable-length
819  * array of bytes. The first byte indicates the length by giving a
820  * number of 0-bits followed by a one. The position of the first
821  * "one" bit inside the first byte indicates the length of this
822  * number.
823  * Returns: number of bytes read, < 0 on error
824  */
825 static int ebml_read_num(MatroskaDemuxContext *matroska, AVIOContext *pb,
826  int max_size, uint64_t *number, int eof_forbidden)
827 {
828  int read, n = 1;
829  uint64_t total;
830  int64_t pos;
831 
832  /* The first byte tells us the length in bytes - except when it is zero. */
833  total = avio_r8(pb);
834  if (pb->eof_reached)
835  goto err;
836 
837  /* get the length of the EBML number */
838  read = 8 - ff_log2_tab[total];
839 
840  if (!total || read > max_size) {
841  pos = avio_tell(pb) - 1;
842  if (!total) {
843  av_log(matroska->ctx, AV_LOG_ERROR,
844  "0x00 at pos %"PRId64" (0x%"PRIx64") invalid as first byte "
845  "of an EBML number\n", pos, pos);
846  } else {
847  av_log(matroska->ctx, AV_LOG_ERROR,
848  "Length %d indicated by an EBML number's first byte 0x%02x "
849  "at pos %"PRId64" (0x%"PRIx64") exceeds max length %d.\n",
850  read, (uint8_t) total, pos, pos, max_size);
851  }
852  return AVERROR_INVALIDDATA;
853  }
854 
855  /* read out length */
856  total ^= 1 << ff_log2_tab[total];
857  while (n++ < read)
858  total = (total << 8) | avio_r8(pb);
859 
860  if (pb->eof_reached) {
861  eof_forbidden = 1;
862  goto err;
863  }
864 
865  *number = total;
866 
867  return read;
868 
869 err:
870  pos = avio_tell(pb);
871  if (pb->error) {
872  av_log(matroska->ctx, AV_LOG_ERROR,
873  "Read error at pos. %"PRIu64" (0x%"PRIx64")\n",
874  pos, pos);
875  return pb->error;
876  }
877  if (eof_forbidden) {
878  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
879  "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
880  return AVERROR(EIO);
881  }
882  return AVERROR_EOF;
883 }
884 
885 /**
886  * Read a EBML length value.
887  * This needs special handling for the "unknown length" case which has multiple
888  * encodings.
889  */
890 static int ebml_read_length(MatroskaDemuxContext *matroska, AVIOContext *pb,
891  uint64_t *number)
892 {
893  int res = ebml_read_num(matroska, pb, 8, number, 1);
894  if (res > 0 && *number + 1 == 1ULL << (7 * res))
895  *number = EBML_UNKNOWN_LENGTH;
896  return res;
897 }
898 
899 /*
900  * Read the next element as an unsigned int.
901  * Returns NEEDS_CHECKING.
902  */
903 static int ebml_read_uint(AVIOContext *pb, int size, uint64_t *num)
904 {
905  int n = 0;
906 
907  /* big-endian ordering; build up number */
908  *num = 0;
909  while (n++ < size)
910  *num = (*num << 8) | avio_r8(pb);
911 
912  return NEEDS_CHECKING;
913 }
914 
915 /*
916  * Read the next element as a signed int.
917  * Returns NEEDS_CHECKING.
918  */
919 static int ebml_read_sint(AVIOContext *pb, int size, int64_t *num)
920 {
921  int n = 1;
922 
923  if (size == 0) {
924  *num = 0;
925  } else {
926  *num = sign_extend(avio_r8(pb), 8);
927 
928  /* big-endian ordering; build up number */
929  while (n++ < size)
930  *num = ((uint64_t)*num << 8) | avio_r8(pb);
931  }
932 
933  return NEEDS_CHECKING;
934 }
935 
936 /*
937  * Read the next element as a float.
938  * Returns NEEDS_CHECKING or < 0 on obvious failure.
939  */
940 static int ebml_read_float(AVIOContext *pb, int size, double *num)
941 {
942  if (size == 0)
943  *num = 0;
944  else if (size == 4)
945  *num = av_int2float(avio_rb32(pb));
946  else if (size == 8)
947  *num = av_int2double(avio_rb64(pb));
948  else
949  return AVERROR_INVALIDDATA;
950 
951  return NEEDS_CHECKING;
952 }
953 
954 /*
955  * Read the next element as an ASCII string.
956  * 0 is success, < 0 or NEEDS_CHECKING is failure.
957  */
958 static int ebml_read_ascii(AVIOContext *pb, int size, char **str)
959 {
960  char *res;
961  int ret;
962 
963  /* EBML strings are usually not 0-terminated, so we allocate one
964  * byte more, read the string and NULL-terminate it ourselves. */
965  if (!(res = av_malloc(size + 1)))
966  return AVERROR(ENOMEM);
967  if ((ret = avio_read(pb, (uint8_t *) res, size)) != size) {
968  av_free(res);
969  return ret < 0 ? ret : NEEDS_CHECKING;
970  }
971  (res)[size] = '\0';
972  av_free(*str);
973  *str = res;
974 
975  return 0;
976 }
977 
978 /*
979  * Read the next element as binary data.
980  * 0 is success, < 0 or NEEDS_CHECKING is failure.
981  */
982 static int ebml_read_binary(AVIOContext *pb, int length,
983  int64_t pos, EbmlBin *bin)
984 {
985  int ret;
986 
988  if (ret < 0)
989  return ret;
990  memset(bin->buf->data + length, 0, AV_INPUT_BUFFER_PADDING_SIZE);
991 
992  bin->data = bin->buf->data;
993  bin->size = length;
994  bin->pos = pos;
995  if ((ret = avio_read(pb, bin->data, length)) != length) {
996  av_buffer_unref(&bin->buf);
997  bin->data = NULL;
998  bin->size = 0;
999  return ret < 0 ? ret : NEEDS_CHECKING;
1000  }
1001 
1002  return 0;
1003 }
1005 /*
1006  * Read the next element, but only the header. The contents
1007  * are supposed to be sub-elements which can be read separately.
1008  * 0 is success, < 0 is failure.
1009  */
1010 static int ebml_read_master(MatroskaDemuxContext *matroska,
1011  uint64_t length, int64_t pos)
1012 {
1014 
1015  if (matroska->num_levels >= EBML_MAX_DEPTH) {
1016  av_log(matroska->ctx, AV_LOG_ERROR,
1017  "File moves beyond max. allowed depth (%d)\n", EBML_MAX_DEPTH);
1018  return AVERROR(ENOSYS);
1019  }
1020 
1021  level = &matroska->levels[matroska->num_levels++];
1022  level->start = pos;
1023  level->length = length;
1024 
1025  return 0;
1027 
1028 /*
1029  * Read signed/unsigned "EBML" numbers.
1030  * Return: number of bytes processed, < 0 on error
1031  */
1032 static int matroska_ebmlnum_uint(MatroskaDemuxContext *matroska,
1033  uint8_t *data, uint32_t size, uint64_t *num)
1034 {
1035  AVIOContext pb;
1036  ffio_init_context(&pb, data, size, 0, NULL, NULL, NULL, NULL);
1037  return ebml_read_num(matroska, &pb, FFMIN(size, 8), num, 1);
1038 }
1039 
1040 /*
1041  * Same as above, but signed.
1042  */
1043 static int matroska_ebmlnum_sint(MatroskaDemuxContext *matroska,
1044  uint8_t *data, uint32_t size, int64_t *num)
1045 {
1046  uint64_t unum;
1047  int res;
1048 
1049  /* read as unsigned number first */
1050  if ((res = matroska_ebmlnum_uint(matroska, data, size, &unum)) < 0)
1051  return res;
1052 
1053  /* make signed (weird way) */
1054  *num = unum - ((1LL << (7 * res - 1)) - 1);
1055 
1056  return res;
1057 }
1058 
1059 static int ebml_parse(MatroskaDemuxContext *matroska,
1060  EbmlSyntax *syntax, void *data);
1061 
1062 static EbmlSyntax *ebml_parse_id(EbmlSyntax *syntax, uint32_t id)
1063 {
1064  int i;
1065 
1066  // Whoever touches this should be aware of the duplication
1067  // existing in matroska_cluster_parsing.
1068  for (i = 0; syntax[i].id; i++)
1069  if (id == syntax[i].id)
1070  break;
1071 
1072  return &syntax[i];
1073 }
1074 
1075 static int ebml_parse_nest(MatroskaDemuxContext *matroska, EbmlSyntax *syntax,
1076  void *data)
1077 {
1078  int res;
1079 
1080  if (data) {
1081  for (int i = 0; syntax[i].id; i++)
1082  switch (syntax[i].type) {
1083  case EBML_UINT:
1084  *(uint64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.u;
1085  break;
1086  case EBML_SINT:
1087  *(int64_t *) ((char *) data + syntax[i].data_offset) = syntax[i].def.i;
1088  break;
1089  case EBML_FLOAT:
1090  *(double *) ((char *) data + syntax[i].data_offset) = syntax[i].def.f;
1091  break;
1092  case EBML_STR:
1093  case EBML_UTF8:
1094  // the default may be NULL
1095  if (syntax[i].def.s) {
1096  uint8_t **dst = (uint8_t **) ((uint8_t *) data + syntax[i].data_offset);
1097  *dst = av_strdup(syntax[i].def.s);
1098  if (!*dst)
1099  return AVERROR(ENOMEM);
1100  }
1101  break;
1102  }
1103 
1104  if (!matroska->levels[matroska->num_levels - 1].length) {
1105  matroska->num_levels--;
1106  return 0;
1107  }
1108  }
1109 
1110  do {
1111  res = ebml_parse(matroska, syntax, data);
1112  } while (!res);
1113 
1114  return res == LEVEL_ENDED ? 0 : res;
1115 }
1116 
1117 static int is_ebml_id_valid(uint32_t id)
1118 {
1119  // Due to endian nonsense in Matroska, the highest byte with any bits set
1120  // will contain the leading length bit. This bit in turn identifies the
1121  // total byte length of the element by its position within the byte.
1122  unsigned int bits = av_log2(id);
1123  return id && (bits + 7) / 8 == (8 - bits % 8);
1125 
1126 /*
1127  * Allocate and return the entry for the level1 element with the given ID. If
1128  * an entry already exists, return the existing entry.
1129  */
1131  uint32_t id)
1132 {
1133  int i;
1134  MatroskaLevel1Element *elem;
1135 
1136  if (!is_ebml_id_valid(id))
1137  return NULL;
1138 
1139  // Some files link to all clusters; useless.
1140  if (id == MATROSKA_ID_CLUSTER)
1141  return NULL;
1142 
1143  // There can be multiple seekheads.
1144  if (id != MATROSKA_ID_SEEKHEAD) {
1145  for (i = 0; i < matroska->num_level1_elems; i++) {
1146  if (matroska->level1_elems[i].id == id)
1147  return &matroska->level1_elems[i];
1148  }
1149  }
1150 
1151  // Only a completely broken file would have more elements.
1152  // It also provides a low-effort way to escape from circular seekheads
1153  // (every iteration will add a level1 entry).
1154  if (matroska->num_level1_elems >= FF_ARRAY_ELEMS(matroska->level1_elems)) {
1155  av_log(matroska->ctx, AV_LOG_ERROR, "Too many level1 elements or circular seekheads.\n");
1156  return NULL;
1157  }
1158 
1159  elem = &matroska->level1_elems[matroska->num_level1_elems++];
1160  *elem = (MatroskaLevel1Element){.id = id};
1161 
1162  return elem;
1163 }
1164 
1165 static int ebml_parse(MatroskaDemuxContext *matroska,
1166  EbmlSyntax *syntax, void *data)
1167 {
1168  static const uint64_t max_lengths[EBML_TYPE_COUNT] = {
1169  // Forbid unknown-length EBML_NONE elements.
1171  [EBML_UINT] = 8,
1172  [EBML_SINT] = 8,
1173  [EBML_FLOAT] = 8,
1174  // max. 16 MB for strings
1175  [EBML_STR] = 0x1000000,
1176  [EBML_UTF8] = 0x1000000,
1177  // max. 256 MB for binary data
1178  [EBML_BIN] = 0x10000000,
1179  // no limits for anything else
1180  };
1181  AVIOContext *pb = matroska->ctx->pb;
1182  uint32_t id;
1183  uint64_t length;
1184  int64_t pos = avio_tell(pb), pos_alt;
1185  int res, update_pos = 1, level_check;
1186  MatroskaLevel1Element *level1_elem;
1187  MatroskaLevel *level = matroska->num_levels ? &matroska->levels[matroska->num_levels - 1] : NULL;
1188 
1189  if (!matroska->current_id) {
1190  uint64_t id;
1191  res = ebml_read_num(matroska, pb, 4, &id, 0);
1192  if (res < 0) {
1193  if (pb->eof_reached && res == AVERROR_EOF) {
1194  if (matroska->is_live)
1195  // in live mode, finish parsing if EOF is reached.
1196  return 1;
1197  if (level && pos == avio_tell(pb)) {
1198  if (level->length == EBML_UNKNOWN_LENGTH) {
1199  // Unknown-length levels automatically end at EOF.
1200  matroska->num_levels--;
1201  return LEVEL_ENDED;
1202  } else {
1203  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely "
1204  "at pos. %"PRIu64" (0x%"PRIx64")\n", pos, pos);
1205  }
1206  }
1207  }
1208  return res;
1209  }
1210  matroska->current_id = id | 1 << 7 * res;
1211  pos_alt = pos + res;
1212  } else {
1213  pos_alt = pos;
1214  pos -= (av_log2(matroska->current_id) + 7) / 8;
1215  }
1216 
1217  id = matroska->current_id;
1218 
1219  syntax = ebml_parse_id(syntax, id);
1220  if (!syntax->id && id != EBML_ID_VOID && id != EBML_ID_CRC32) {
1221  if (level && level->length == EBML_UNKNOWN_LENGTH) {
1222  // Unknown-length levels end when an element from an upper level
1223  // in the hierarchy is encountered.
1224  while (syntax->def.n) {
1225  syntax = ebml_parse_id(syntax->def.n, id);
1226  if (syntax->id) {
1227  matroska->num_levels--;
1228  return LEVEL_ENDED;
1229  }
1230  };
1231  }
1232 
1233  av_log(matroska->ctx, AV_LOG_DEBUG, "Unknown entry 0x%"PRIX32" at pos. "
1234  "%"PRId64"\n", id, pos);
1235  update_pos = 0; /* Don't update resync_pos as an error might have happened. */
1236  }
1237 
1238  if (data) {
1239  data = (char *) data + syntax->data_offset;
1240  if (syntax->list_elem_size) {
1241  EbmlList *list = data;
1242  void *newelem;
1243 
1244  if ((unsigned)list->nb_elem + 1 >= UINT_MAX / syntax->list_elem_size)
1245  return AVERROR(ENOMEM);
1246  newelem = av_fast_realloc(list->elem,
1247  &list->alloc_elem_size,
1248  (list->nb_elem + 1) * syntax->list_elem_size);
1249  if (!newelem)
1250  return AVERROR(ENOMEM);
1251  list->elem = newelem;
1252  data = (char *) list->elem + list->nb_elem * syntax->list_elem_size;
1253  memset(data, 0, syntax->list_elem_size);
1254  list->nb_elem++;
1255  }
1256  }
1257 
1258  if (syntax->type != EBML_STOP) {
1259  matroska->current_id = 0;
1260  if ((res = ebml_read_length(matroska, pb, &length)) < 0)
1261  return res;
1262 
1263  pos_alt += res;
1264 
1265  if (matroska->num_levels > 0) {
1266  if (length != EBML_UNKNOWN_LENGTH &&
1267  level->length != EBML_UNKNOWN_LENGTH) {
1268  uint64_t elem_end = pos_alt + length,
1269  level_end = level->start + level->length;
1270 
1271  if (elem_end < level_end) {
1272  level_check = 0;
1273  } else if (elem_end == level_end) {
1274  level_check = LEVEL_ENDED;
1275  } else {
1276  av_log(matroska->ctx, AV_LOG_ERROR,
1277  "Element at 0x%"PRIx64" ending at 0x%"PRIx64" exceeds "
1278  "containing master element ending at 0x%"PRIx64"\n",
1279  pos, elem_end, level_end);
1280  return AVERROR_INVALIDDATA;
1281  }
1282  } else if (length != EBML_UNKNOWN_LENGTH) {
1283  level_check = 0;
1284  } else if (level->length != EBML_UNKNOWN_LENGTH) {
1285  av_log(matroska->ctx, AV_LOG_ERROR, "Unknown-sized element "
1286  "at 0x%"PRIx64" inside parent with finite size\n", pos);
1287  return AVERROR_INVALIDDATA;
1288  } else {
1289  level_check = 0;
1290  if (id != MATROSKA_ID_CLUSTER && (syntax->type == EBML_LEVEL1
1291  || syntax->type == EBML_NEST)) {
1292  // According to the current specifications only clusters and
1293  // segments are allowed to be unknown-length. We also accept
1294  // other unknown-length master elements.
1295  av_log(matroska->ctx, AV_LOG_WARNING,
1296  "Found unknown-length element 0x%"PRIX32" other than "
1297  "a cluster at 0x%"PRIx64". Spec-incompliant, but "
1298  "parsing will nevertheless be attempted.\n", id, pos);
1299  update_pos = -1;
1300  }
1301  }
1302  } else
1303  level_check = 0;
1304 
1305  if (max_lengths[syntax->type] && length > max_lengths[syntax->type]) {
1306  if (length != EBML_UNKNOWN_LENGTH) {
1307  av_log(matroska->ctx, AV_LOG_ERROR,
1308  "Invalid length 0x%"PRIx64" > 0x%"PRIx64" for element "
1309  "with ID 0x%"PRIX32" at 0x%"PRIx64"\n",
1310  length, max_lengths[syntax->type], id, pos);
1311  } else if (syntax->type != EBML_NONE) {
1312  av_log(matroska->ctx, AV_LOG_ERROR,
1313  "Element with ID 0x%"PRIX32" at pos. 0x%"PRIx64" has "
1314  "unknown length, yet the length of an element of its "
1315  "type must be known.\n", id, pos);
1316  } else {
1317  av_log(matroska->ctx, AV_LOG_ERROR,
1318  "Found unknown-length element with ID 0x%"PRIX32" at "
1319  "pos. 0x%"PRIx64" for which no syntax for parsing is "
1320  "available.\n", id, pos);
1321  }
1322  return AVERROR_INVALIDDATA;
1323  }
1324 
1325  if (!(pb->seekable & AVIO_SEEKABLE_NORMAL)) {
1326  // Loosing sync will likely manifest itself as encountering unknown
1327  // elements which are not reliably distinguishable from elements
1328  // belonging to future extensions of the format.
1329  // We use a heuristic to detect such situations: If the current
1330  // element is not expected at the current syntax level and there
1331  // were only a few unknown elements in a row, then the element is
1332  // skipped or considered defective based upon the length of the
1333  // current element (i.e. how much would be skipped); if there were
1334  // more than a few skipped elements in a row and skipping the current
1335  // element would lead us more than SKIP_THRESHOLD away from the last
1336  // known good position, then it is inferred that an error occurred.
1337  // The dependency on the number of unknown elements in a row exists
1338  // because the distance to the last known good position is
1339  // automatically big if the last parsed element was big.
1340  // In both cases, each unknown element is considered equivalent to
1341  // UNKNOWN_EQUIV of skipped bytes for the check.
1342  // The whole check is only done for non-seekable output, because
1343  // in this situation skipped data can't simply be rechecked later.
1344  // This is especially important when using unkown length elements
1345  // as the check for whether a child exceeds its containing master
1346  // element is not effective in this situation.
1347  if (update_pos) {
1348  matroska->unknown_count = 0;
1349  } else {
1350  int64_t dist = length + UNKNOWN_EQUIV * matroska->unknown_count++;
1351 
1352  if (matroska->unknown_count > 3)
1353  dist += pos_alt - matroska->resync_pos;
1354 
1355  if (dist > SKIP_THRESHOLD) {
1356  av_log(matroska->ctx, AV_LOG_ERROR,
1357  "Unknown element %"PRIX32" at pos. 0x%"PRIx64" with "
1358  "length 0x%"PRIx64" considered as invalid data. Last "
1359  "known good position 0x%"PRIx64", %d unknown elements"
1360  " in a row\n", id, pos, length, matroska->resync_pos,
1361  matroska->unknown_count);
1362  return AVERROR_INVALIDDATA;
1363  }
1364  }
1365  }
1366 
1367  if (update_pos > 0) {
1368  // We have found an element that is allowed at this place
1369  // in the hierarchy and it passed all checks, so treat the beginning
1370  // of the element as the "last known good" position.
1371  matroska->resync_pos = pos;
1372  }
1373 
1374  if (!data && length != EBML_UNKNOWN_LENGTH)
1375  goto skip;
1376  }
1377 
1378  switch (syntax->type) {
1379  case EBML_UINT:
1380  res = ebml_read_uint(pb, length, data);
1381  break;
1382  case EBML_SINT:
1383  res = ebml_read_sint(pb, length, data);
1384  break;
1385  case EBML_FLOAT:
1386  res = ebml_read_float(pb, length, data);
1387  break;
1388  case EBML_STR:
1389  case EBML_UTF8:
1390  res = ebml_read_ascii(pb, length, data);
1391  break;
1392  case EBML_BIN:
1393  res = ebml_read_binary(pb, length, pos_alt, data);
1394  break;
1395  case EBML_LEVEL1:
1396  case EBML_NEST:
1397  if ((res = ebml_read_master(matroska, length, pos_alt)) < 0)
1398  return res;
1399  if (id == MATROSKA_ID_SEGMENT)
1400  matroska->segment_start = pos_alt;
1401  if (id == MATROSKA_ID_CUES)
1402  matroska->cues_parsing_deferred = 0;
1403  if (syntax->type == EBML_LEVEL1 &&
1404  (level1_elem = matroska_find_level1_elem(matroska, syntax->id))) {
1405  if (!level1_elem->pos) {
1406  // Zero is not a valid position for a level 1 element.
1407  level1_elem->pos = pos;
1408  } else if (level1_elem->pos != pos)
1409  av_log(matroska->ctx, AV_LOG_ERROR, "Duplicate element\n");
1410  level1_elem->parsed = 1;
1411  }
1412  if (res = ebml_parse_nest(matroska, syntax->def.n, data))
1413  return res;
1414  break;
1415  case EBML_STOP:
1416  return 1;
1417  skip:
1418  default:
1419  if (length) {
1420  int64_t res2;
1421  if (ffio_limit(pb, length) != length) {
1422  // ffio_limit emits its own error message,
1423  // so we don't have to.
1424  return AVERROR(EIO);
1425  }
1426  if ((res2 = avio_skip(pb, length - 1)) >= 0) {
1427  // avio_skip might take us past EOF. We check for this
1428  // by skipping only length - 1 bytes, reading a byte and
1429  // checking the error flags. This is done in order to check
1430  // that the element has been properly skipped even when
1431  // no filesize (that ffio_limit relies on) is available.
1432  avio_r8(pb);
1433  res = NEEDS_CHECKING;
1434  } else
1435  res = res2;
1436  } else
1437  res = 0;
1438  }
1439  if (res) {
1440  if (res == NEEDS_CHECKING) {
1441  if (pb->eof_reached) {
1442  if (pb->error)
1443  res = pb->error;
1444  else
1445  res = AVERROR_EOF;
1446  } else
1447  goto level_check;
1448  }
1449 
1450  if (res == AVERROR_INVALIDDATA)
1451  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid element\n");
1452  else if (res == AVERROR(EIO))
1453  av_log(matroska->ctx, AV_LOG_ERROR, "Read error\n");
1454  else if (res == AVERROR_EOF) {
1455  av_log(matroska->ctx, AV_LOG_ERROR, "File ended prematurely\n");
1456  res = AVERROR(EIO);
1457  }
1458 
1459  return res;
1460  }
1461 
1462 level_check:
1463  if (level_check == LEVEL_ENDED && matroska->num_levels) {
1464  level = &matroska->levels[matroska->num_levels - 1];
1465  pos = avio_tell(pb);
1466 
1467  // Given that pos >= level->start no check for
1468  // level->length != EBML_UNKNOWN_LENGTH is necessary.
1469  while (matroska->num_levels && pos == level->start + level->length) {
1470  matroska->num_levels--;
1471  level--;
1472  }
1473  }
1474 
1475  return level_check;
1476 }
1477 
1478 static void ebml_free(EbmlSyntax *syntax, void *data)
1479 {
1480  int i, j;
1481  for (i = 0; syntax[i].id; i++) {
1482  void *data_off = (char *) data + syntax[i].data_offset;
1483  switch (syntax[i].type) {
1484  case EBML_STR:
1485  case EBML_UTF8:
1486  av_freep(data_off);
1487  break;
1488  case EBML_BIN:
1489  av_buffer_unref(&((EbmlBin *) data_off)->buf);
1490  break;
1491  case EBML_LEVEL1:
1492  case EBML_NEST:
1493  if (syntax[i].list_elem_size) {
1494  EbmlList *list = data_off;
1495  char *ptr = list->elem;
1496  for (j = 0; j < list->nb_elem;
1497  j++, ptr += syntax[i].list_elem_size)
1498  ebml_free(syntax[i].def.n, ptr);
1499  av_freep(&list->elem);
1500  list->nb_elem = 0;
1501  list->alloc_elem_size = 0;
1502  } else
1503  ebml_free(syntax[i].def.n, data_off);
1504  default:
1505  break;
1506  }
1507  }
1508 }
1509 
1510 /*
1511  * Autodetecting...
1512  */
1513 static int matroska_probe(const AVProbeData *p)
1514 {
1515  uint64_t total = 0;
1516  int len_mask = 0x80, size = 1, n = 1, i;
1517 
1518  /* EBML header? */
1519  if (AV_RB32(p->buf) != EBML_ID_HEADER)
1520  return 0;
1521 
1522  /* length of header */
1523  total = p->buf[4];
1524  while (size <= 8 && !(total & len_mask)) {
1525  size++;
1526  len_mask >>= 1;
1527  }
1528  if (size > 8)
1529  return 0;
1530  total &= (len_mask - 1);
1531  while (n < size)
1532  total = (total << 8) | p->buf[4 + n++];
1533 
1534  if (total + 1 == 1ULL << (7 * size)){
1535  /* Unknown-length header - simply parse the whole buffer. */
1536  total = p->buf_size - 4 - size;
1537  } else {
1538  /* Does the probe data contain the whole header? */
1539  if (p->buf_size < 4 + size + total)
1540  return 0;
1541  }
1542 
1543  /* The header should contain a known document type. For now,
1544  * we don't parse the whole header but simply check for the
1545  * availability of that array of characters inside the header.
1546  * Not fully fool-proof, but good enough. */
1547  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++) {
1548  size_t probelen = strlen(matroska_doctypes[i]);
1549  if (total < probelen)
1550  continue;
1551  for (n = 4 + size; n <= 4 + size + total - probelen; n++)
1552  if (!memcmp(p->buf + n, matroska_doctypes[i], probelen))
1553  return AVPROBE_SCORE_MAX;
1554  }
1555 
1556  // probably valid EBML header but no recognized doctype
1557  return AVPROBE_SCORE_EXTENSION;
1558 }
1559 
1561  uint64_t num)
1562 {
1563  MatroskaTrack *tracks = matroska->tracks.elem;
1564  int i;
1565 
1566  for (i = 0; i < matroska->tracks.nb_elem; i++)
1567  if (tracks[i].num == num)
1568  return &tracks[i];
1569 
1570  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid track number %"PRIu64"\n", num);
1571  return NULL;
1572 }
1573 
1574 static int matroska_decode_buffer(uint8_t **buf, int *buf_size,
1575  MatroskaTrack *track)
1576 {
1577  MatroskaTrackEncoding *encodings = track->encodings.elem;
1578  uint8_t *data = *buf;
1579  int isize = *buf_size;
1580  uint8_t *pkt_data = NULL;
1581  uint8_t av_unused *newpktdata;
1582  int pkt_size = isize;
1583  int result = 0;
1584  int olen;
1585 
1586  if (pkt_size >= 10000000U)
1587  return AVERROR_INVALIDDATA;
1588 
1589  switch (encodings[0].compression.algo) {
1591  {
1592  int header_size = encodings[0].compression.settings.size;
1593  uint8_t *header = encodings[0].compression.settings.data;
1594 
1595  if (header_size && !header) {
1596  av_log(NULL, AV_LOG_ERROR, "Compression size but no data in headerstrip\n");
1597  return -1;
1598  }
1599 
1600  if (!header_size)
1601  return 0;
1602 
1603  pkt_size = isize + header_size;
1604  pkt_data = av_malloc(pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1605  if (!pkt_data)
1606  return AVERROR(ENOMEM);
1607 
1608  memcpy(pkt_data, header, header_size);
1609  memcpy(pkt_data + header_size, data, isize);
1610  break;
1611  }
1612 #if CONFIG_LZO
1614  do {
1615  olen = pkt_size *= 3;
1616  newpktdata = av_realloc(pkt_data, pkt_size + AV_LZO_OUTPUT_PADDING
1618  if (!newpktdata) {
1619  result = AVERROR(ENOMEM);
1620  goto failed;
1621  }
1622  pkt_data = newpktdata;
1623  result = av_lzo1x_decode(pkt_data, &olen, data, &isize);
1624  } while (result == AV_LZO_OUTPUT_FULL && pkt_size < 10000000);
1625  if (result) {
1627  goto failed;
1628  }
1629  pkt_size -= olen;
1630  break;
1631 #endif
1632 #if CONFIG_ZLIB
1634  {
1635  z_stream zstream = { 0 };
1636  if (!pkt_size || inflateInit(&zstream) != Z_OK)
1637  return -1;
1638  zstream.next_in = data;
1639  zstream.avail_in = isize;
1640  do {
1641  pkt_size *= 3;
1642  newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1643  if (!newpktdata) {
1644  inflateEnd(&zstream);
1645  result = AVERROR(ENOMEM);
1646  goto failed;
1647  }
1648  pkt_data = newpktdata;
1649  zstream.avail_out = pkt_size - zstream.total_out;
1650  zstream.next_out = pkt_data + zstream.total_out;
1651  result = inflate(&zstream, Z_NO_FLUSH);
1652  } while (result == Z_OK && pkt_size < 10000000);
1653  pkt_size = zstream.total_out;
1654  inflateEnd(&zstream);
1655  if (result != Z_STREAM_END) {
1656  if (result == Z_MEM_ERROR)
1657  result = AVERROR(ENOMEM);
1658  else
1660  goto failed;
1661  }
1662  break;
1663  }
1664 #endif
1665 #if CONFIG_BZLIB
1667  {
1668  bz_stream bzstream = { 0 };
1669  if (!pkt_size || BZ2_bzDecompressInit(&bzstream, 0, 0) != BZ_OK)
1670  return -1;
1671  bzstream.next_in = data;
1672  bzstream.avail_in = isize;
1673  do {
1674  pkt_size *= 3;
1675  newpktdata = av_realloc(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE);
1676  if (!newpktdata) {
1677  BZ2_bzDecompressEnd(&bzstream);
1678  result = AVERROR(ENOMEM);
1679  goto failed;
1680  }
1681  pkt_data = newpktdata;
1682  bzstream.avail_out = pkt_size - bzstream.total_out_lo32;
1683  bzstream.next_out = pkt_data + bzstream.total_out_lo32;
1684  result = BZ2_bzDecompress(&bzstream);
1685  } while (result == BZ_OK && pkt_size < 10000000);
1686  pkt_size = bzstream.total_out_lo32;
1687  BZ2_bzDecompressEnd(&bzstream);
1688  if (result != BZ_STREAM_END) {
1689  if (result == BZ_MEM_ERROR)
1690  result = AVERROR(ENOMEM);
1691  else
1693  goto failed;
1694  }
1695  break;
1696  }
1697 #endif
1698  default:
1699  return AVERROR_INVALIDDATA;
1700  }
1701 
1702  memset(pkt_data + pkt_size, 0, AV_INPUT_BUFFER_PADDING_SIZE);
1703 
1704  *buf = pkt_data;
1705  *buf_size = pkt_size;
1706  return 0;
1708 failed:
1709  av_free(pkt_data);
1710  return result;
1711 }
1712 
1714  AVDictionary **metadata, char *prefix)
1715 {
1716  MatroskaTag *tags = list->elem;
1717  char key[1024];
1718  int i;
1719 
1720  for (i = 0; i < list->nb_elem; i++) {
1721  const char *lang = tags[i].lang &&
1722  strcmp(tags[i].lang, "und") ? tags[i].lang : NULL;
1723 
1724  if (!tags[i].name) {
1725  av_log(s, AV_LOG_WARNING, "Skipping invalid tag with no TagName.\n");
1726  continue;
1727  }
1728  if (prefix)
1729  snprintf(key, sizeof(key), "%s/%s", prefix, tags[i].name);
1730  else
1731  av_strlcpy(key, tags[i].name, sizeof(key));
1732  if (tags[i].def || !lang) {
1733  av_dict_set(metadata, key, tags[i].string, 0);
1734  if (tags[i].sub.nb_elem)
1735  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1736  }
1737  if (lang) {
1738  av_strlcat(key, "-", sizeof(key));
1739  av_strlcat(key, lang, sizeof(key));
1740  av_dict_set(metadata, key, tags[i].string, 0);
1741  if (tags[i].sub.nb_elem)
1742  matroska_convert_tag(s, &tags[i].sub, metadata, key);
1743  }
1744  }
1746 }
1747 
1749 {
1750  MatroskaDemuxContext *matroska = s->priv_data;
1751  MatroskaTags *tags = matroska->tags.elem;
1752  int i, j;
1753 
1754  for (i = 0; i < matroska->tags.nb_elem; i++) {
1755  if (tags[i].target.attachuid) {
1756  MatroskaAttachment *attachment = matroska->attachments.elem;
1757  int found = 0;
1758  for (j = 0; j < matroska->attachments.nb_elem; j++) {
1759  if (attachment[j].uid == tags[i].target.attachuid &&
1760  attachment[j].stream) {
1761  matroska_convert_tag(s, &tags[i].tag,
1762  &attachment[j].stream->metadata, NULL);
1763  found = 1;
1764  }
1765  }
1766  if (!found) {
1768  "The tags at index %d refer to a "
1769  "non-existent attachment %"PRId64".\n",
1770  i, tags[i].target.attachuid);
1771  }
1772  } else if (tags[i].target.chapteruid) {
1773  MatroskaChapter *chapter = matroska->chapters.elem;
1774  int found = 0;
1775  for (j = 0; j < matroska->chapters.nb_elem; j++) {
1776  if (chapter[j].uid == tags[i].target.chapteruid &&
1777  chapter[j].chapter) {
1778  matroska_convert_tag(s, &tags[i].tag,
1779  &chapter[j].chapter->metadata, NULL);
1780  found = 1;
1781  }
1782  }
1783  if (!found) {
1785  "The tags at index %d refer to a non-existent chapter "
1786  "%"PRId64".\n",
1787  i, tags[i].target.chapteruid);
1788  }
1789  } else if (tags[i].target.trackuid) {
1790  MatroskaTrack *track = matroska->tracks.elem;
1791  int found = 0;
1792  for (j = 0; j < matroska->tracks.nb_elem; j++) {
1793  if (track[j].uid == tags[i].target.trackuid &&
1794  track[j].stream) {
1795  matroska_convert_tag(s, &tags[i].tag,
1796  &track[j].stream->metadata, NULL);
1797  found = 1;
1798  }
1799  }
1800  if (!found) {
1802  "The tags at index %d refer to a non-existent track "
1803  "%"PRId64".\n",
1804  i, tags[i].target.trackuid);
1805  }
1806  } else {
1807  matroska_convert_tag(s, &tags[i].tag, &s->metadata,
1808  tags[i].target.type);
1809  }
1810  }
1811 }
1812 
1814  int64_t pos)
1815 {
1816  uint32_t saved_id = matroska->current_id;
1817  int64_t before_pos = avio_tell(matroska->ctx->pb);
1818  int ret = 0;
1819  int ret2;
1820 
1821  /* seek */
1822  if (avio_seek(matroska->ctx->pb, pos, SEEK_SET) == pos) {
1823  /* We don't want to lose our seekhead level, so we add
1824  * a dummy. This is a crude hack. */
1825  if (matroska->num_levels == EBML_MAX_DEPTH) {
1826  av_log(matroska->ctx, AV_LOG_INFO,
1827  "Max EBML element depth (%d) reached, "
1828  "cannot parse further.\n", EBML_MAX_DEPTH);
1830  } else {
1831  matroska->levels[matroska->num_levels] = (MatroskaLevel) { 0, EBML_UNKNOWN_LENGTH };
1832  matroska->num_levels++;
1833  matroska->current_id = 0;
1834 
1835  ret = ebml_parse(matroska, matroska_segment, matroska);
1836  if (ret == LEVEL_ENDED) {
1837  /* This can only happen if the seek brought us beyond EOF. */
1838  ret = AVERROR_EOF;
1839  }
1840  }
1841  }
1842  /* Seek back - notice that in all instances where this is used
1843  * it is safe to set the level to 1. */
1844  ret2 = matroska_reset_status(matroska, saved_id, before_pos);
1845  if (ret >= 0)
1846  ret = ret2;
1847 
1848  return ret;
1849 }
1850 
1851 static void matroska_execute_seekhead(MatroskaDemuxContext *matroska)
1852 {
1853  EbmlList *seekhead_list = &matroska->seekhead;
1854  int i;
1855 
1856  // we should not do any seeking in the streaming case
1857  if (!(matroska->ctx->pb->seekable & AVIO_SEEKABLE_NORMAL))
1858  return;
1859 
1860  for (i = 0; i < seekhead_list->nb_elem; i++) {
1861  MatroskaSeekhead *seekheads = seekhead_list->elem;
1862  uint32_t id = seekheads[i].id;
1863  int64_t pos = seekheads[i].pos + matroska->segment_start;
1864  MatroskaLevel1Element *elem;
1865 
1866  if (id != seekheads[i].id || pos < matroska->segment_start)
1867  continue;
1868 
1869  elem = matroska_find_level1_elem(matroska, id);
1870  if (!elem || elem->parsed)
1871  continue;
1872 
1873  elem->pos = pos;
1874 
1875  // defer cues parsing until we actually need cue data.
1876  if (id == MATROSKA_ID_CUES)
1877  continue;
1878 
1879  if (matroska_parse_seekhead_entry(matroska, pos) < 0) {
1880  // mark index as broken
1881  matroska->cues_parsing_deferred = -1;
1882  break;
1883  }
1884 
1885  elem->parsed = 1;
1886  }
1887 }
1888 
1889 static void matroska_add_index_entries(MatroskaDemuxContext *matroska)
1890 {
1891  EbmlList *index_list;
1893  uint64_t index_scale = 1;
1894  int i, j;
1895 
1896  if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1897  return;
1898 
1899  index_list = &matroska->index;
1900  index = index_list->elem;
1901  if (index_list->nb_elem < 2)
1902  return;
1903  if (index[1].time > 1E14 / matroska->time_scale) {
1904  av_log(matroska->ctx, AV_LOG_WARNING, "Dropping apparently-broken index.\n");
1905  return;
1906  }
1907  for (i = 0; i < index_list->nb_elem; i++) {
1908  EbmlList *pos_list = &index[i].pos;
1909  MatroskaIndexPos *pos = pos_list->elem;
1910  for (j = 0; j < pos_list->nb_elem; j++) {
1911  MatroskaTrack *track = matroska_find_track_by_num(matroska,
1912  pos[j].track);
1913  if (track && track->stream)
1914  av_add_index_entry(track->stream,
1915  pos[j].pos + matroska->segment_start,
1916  index[i].time / index_scale, 0, 0,
1918  }
1919  }
1920 }
1921 
1922 static void matroska_parse_cues(MatroskaDemuxContext *matroska) {
1923  int i;
1924 
1925  if (matroska->ctx->flags & AVFMT_FLAG_IGNIDX)
1926  return;
1927 
1928  for (i = 0; i < matroska->num_level1_elems; i++) {
1929  MatroskaLevel1Element *elem = &matroska->level1_elems[i];
1930  if (elem->id == MATROSKA_ID_CUES && !elem->parsed) {
1931  if (matroska_parse_seekhead_entry(matroska, elem->pos) < 0)
1932  matroska->cues_parsing_deferred = -1;
1933  elem->parsed = 1;
1934  break;
1935  }
1936  }
1937 
1938  matroska_add_index_entries(matroska);
1939 }
1940 
1941 static int matroska_aac_profile(char *codec_id)
1942 {
1943  static const char *const aac_profiles[] = { "MAIN", "LC", "SSR" };
1944  int profile;
1945 
1947  if (strstr(codec_id, aac_profiles[profile]))
1948  break;
1949  return profile + 1;
1950 }
1951 
1952 static int matroska_aac_sri(int samplerate)
1953 {
1954  int sri;
1955 
1956  for (sri = 0; sri < FF_ARRAY_ELEMS(avpriv_mpeg4audio_sample_rates); sri++)
1957  if (avpriv_mpeg4audio_sample_rates[sri] == samplerate)
1958  break;
1959  return sri;
1960 }
1961 
1962 static void matroska_metadata_creation_time(AVDictionary **metadata, int64_t date_utc)
1963 {
1964  /* Convert to seconds and adjust by number of seconds between 2001-01-01 and Epoch */
1965  avpriv_dict_set_timestamp(metadata, "creation_time", date_utc / 1000 + 978307200000000LL);
1966 }
1967 
1969  MatroskaTrack *track,
1970  int *offset)
1971 {
1972  AVStream *st = track->stream;
1973  uint8_t *p = track->codec_priv.data;
1974  int size = track->codec_priv.size;
1975 
1976  if (size < 8 + FLAC_STREAMINFO_SIZE || p[4] & 0x7f) {
1977  av_log(s, AV_LOG_WARNING, "Invalid FLAC private data\n");
1978  track->codec_priv.size = 0;
1979  return 0;
1980  }
1981  *offset = 8;
1982  track->codec_priv.size = 8 + FLAC_STREAMINFO_SIZE;
1983 
1984  p += track->codec_priv.size;
1985  size -= track->codec_priv.size;
1986 
1987  /* parse the remaining metadata blocks if present */
1988  while (size >= 4) {
1989  int block_last, block_type, block_size;
1990 
1991  flac_parse_block_header(p, &block_last, &block_type, &block_size);
1992 
1993  p += 4;
1994  size -= 4;
1995  if (block_size > size)
1996  return 0;
1997 
1998  /* check for the channel mask */
1999  if (block_type == FLAC_METADATA_TYPE_VORBIS_COMMENT) {
2000  AVDictionary *dict = NULL;
2001  AVDictionaryEntry *chmask;
2002 
2003  ff_vorbis_comment(s, &dict, p, block_size, 0);
2004  chmask = av_dict_get(dict, "WAVEFORMATEXTENSIBLE_CHANNEL_MASK", NULL, 0);
2005  if (chmask) {
2006  uint64_t mask = strtol(chmask->value, NULL, 0);
2007  if (!mask || mask & ~0x3ffffULL) {
2009  "Invalid value of WAVEFORMATEXTENSIBLE_CHANNEL_MASK\n");
2010  } else
2011  st->codecpar->channel_layout = mask;
2012  }
2013  av_dict_free(&dict);
2014  }
2015 
2016  p += block_size;
2017  size -= block_size;
2018  }
2019 
2020  return 0;
2021 }
2022 
2023 static int mkv_field_order(MatroskaDemuxContext *matroska, int64_t field_order)
2024 {
2025  int major, minor, micro, bttb = 0;
2026 
2027  /* workaround a bug in our Matroska muxer, introduced in version 57.36 alongside
2028  * this function, and fixed in 57.52 */
2029  if (matroska->muxingapp && sscanf(matroska->muxingapp, "Lavf%d.%d.%d", &major, &minor, &micro) == 3)
2030  bttb = (major == 57 && minor >= 36 && minor <= 51 && micro >= 100);
2031 
2032  switch (field_order) {
2034  return AV_FIELD_PROGRESSIVE;
2036  return AV_FIELD_UNKNOWN;
2038  return AV_FIELD_TT;
2040  return AV_FIELD_BB;
2042  return bttb ? AV_FIELD_TB : AV_FIELD_BT;
2044  return bttb ? AV_FIELD_BT : AV_FIELD_TB;
2045  default:
2046  return AV_FIELD_UNKNOWN;
2047  }
2048 }
2049 
2050 static void mkv_stereo_mode_display_mul(int stereo_mode,
2051  int *h_width, int *h_height)
2052 {
2053  switch (stereo_mode) {
2059  break;
2064  *h_width = 2;
2065  break;
2070  *h_height = 2;
2071  break;
2072  }
2073 }
2074 
2075 static int mkv_parse_video_color(AVStream *st, const MatroskaTrack *track) {
2076  const MatroskaTrackVideoColor *color = track->video.color.elem;
2077  const MatroskaMasteringMeta *mastering_meta;
2078  int has_mastering_primaries, has_mastering_luminance;
2079 
2080  if (!track->video.color.nb_elem)
2081  return 0;
2082 
2083  mastering_meta = &color->mastering_meta;
2084  // Mastering primaries are CIE 1931 coords, and must be > 0.
2085  has_mastering_primaries =
2086  mastering_meta->r_x > 0 && mastering_meta->r_y > 0 &&
2087  mastering_meta->g_x > 0 && mastering_meta->g_y > 0 &&
2088  mastering_meta->b_x > 0 && mastering_meta->b_y > 0 &&
2089  mastering_meta->white_x > 0 && mastering_meta->white_y > 0;
2090  has_mastering_luminance = mastering_meta->max_luminance > 0;
2091 
2092  if (color->matrix_coefficients != AVCOL_SPC_RESERVED)
2093  st->codecpar->color_space = color->matrix_coefficients;
2094  if (color->primaries != AVCOL_PRI_RESERVED &&
2095  color->primaries != AVCOL_PRI_RESERVED0)
2096  st->codecpar->color_primaries = color->primaries;
2097  if (color->transfer_characteristics != AVCOL_TRC_RESERVED &&
2098  color->transfer_characteristics != AVCOL_TRC_RESERVED0)
2099  st->codecpar->color_trc = color->transfer_characteristics;
2100  if (color->range != AVCOL_RANGE_UNSPECIFIED &&
2101  color->range <= AVCOL_RANGE_JPEG)
2102  st->codecpar->color_range = color->range;
2103  if (color->chroma_siting_horz != MATROSKA_COLOUR_CHROMASITINGHORZ_UNDETERMINED &&
2104  color->chroma_siting_vert != MATROSKA_COLOUR_CHROMASITINGVERT_UNDETERMINED &&
2105  color->chroma_siting_horz < MATROSKA_COLOUR_CHROMASITINGHORZ_NB &&
2106  color->chroma_siting_vert < MATROSKA_COLOUR_CHROMASITINGVERT_NB) {
2107  st->codecpar->chroma_location =
2108  avcodec_chroma_pos_to_enum((color->chroma_siting_horz - 1) << 7,
2109  (color->chroma_siting_vert - 1) << 7);
2110  }
2111  if (color->max_cll && color->max_fall) {
2112  size_t size = 0;
2113  int ret;
2115  if (!metadata)
2116  return AVERROR(ENOMEM);
2118  (uint8_t *)metadata, size);
2119  if (ret < 0) {
2120  av_freep(&metadata);
2121  return ret;
2122  }
2123  metadata->MaxCLL = color->max_cll;
2124  metadata->MaxFALL = color->max_fall;
2125  }
2126 
2127  if (has_mastering_primaries || has_mastering_luminance) {
2128  // Use similar rationals as other standards.
2129  const int chroma_den = 50000;
2130  const int luma_den = 10000;
2131  AVMasteringDisplayMetadata *metadata =
2134  sizeof(AVMasteringDisplayMetadata));
2135  if (!metadata) {
2136  return AVERROR(ENOMEM);
2137  }
2138  memset(metadata, 0, sizeof(AVMasteringDisplayMetadata));
2139  if (has_mastering_primaries) {
2140  metadata->display_primaries[0][0] = av_make_q(
2141  round(mastering_meta->r_x * chroma_den), chroma_den);
2142  metadata->display_primaries[0][1] = av_make_q(
2143  round(mastering_meta->r_y * chroma_den), chroma_den);
2144  metadata->display_primaries[1][0] = av_make_q(
2145  round(mastering_meta->g_x * chroma_den), chroma_den);
2146  metadata->display_primaries[1][1] = av_make_q(
2147  round(mastering_meta->g_y * chroma_den), chroma_den);
2148  metadata->display_primaries[2][0] = av_make_q(
2149  round(mastering_meta->b_x * chroma_den), chroma_den);
2150  metadata->display_primaries[2][1] = av_make_q(
2151  round(mastering_meta->b_y * chroma_den), chroma_den);
2152  metadata->white_point[0] = av_make_q(
2153  round(mastering_meta->white_x * chroma_den), chroma_den);
2154  metadata->white_point[1] = av_make_q(
2155  round(mastering_meta->white_y * chroma_den), chroma_den);
2156  metadata->has_primaries = 1;
2157  }
2158  if (has_mastering_luminance) {
2159  metadata->max_luminance = av_make_q(
2160  round(mastering_meta->max_luminance * luma_den), luma_den);
2161  metadata->min_luminance = av_make_q(
2162  round(mastering_meta->min_luminance * luma_den), luma_den);
2163  metadata->has_luminance = 1;
2164  }
2165  }
2166  return 0;
2167 }
2168 
2169 static int mkv_parse_video_projection(AVStream *st, const MatroskaTrack *track) {
2170  AVSphericalMapping *spherical;
2171  enum AVSphericalProjection projection;
2172  size_t spherical_size;
2173  uint32_t l = 0, t = 0, r = 0, b = 0;
2174  uint32_t padding = 0;
2175  int ret;
2176  GetByteContext gb;
2177 
2179  track->video.projection.private.size);
2180 
2181  if (bytestream2_get_byte(&gb) != 0) {
2182  av_log(NULL, AV_LOG_WARNING, "Unknown spherical metadata\n");
2183  return 0;
2184  }
2185 
2186  bytestream2_skip(&gb, 3); // flags
2187 
2188  switch (track->video.projection.type) {
2190  if (track->video.projection.private.size == 20) {
2191  t = bytestream2_get_be32(&gb);
2192  b = bytestream2_get_be32(&gb);
2193  l = bytestream2_get_be32(&gb);
2194  r = bytestream2_get_be32(&gb);
2195 
2196  if (b >= UINT_MAX - t || r >= UINT_MAX - l) {
2198  "Invalid bounding rectangle coordinates "
2199  "%"PRIu32",%"PRIu32",%"PRIu32",%"PRIu32"\n",
2200  l, t, r, b);
2201  return AVERROR_INVALIDDATA;
2202  }
2203  } else if (track->video.projection.private.size != 0) {
2204  av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2205  return AVERROR_INVALIDDATA;
2206  }
2207 
2208  if (l || t || r || b)
2209  projection = AV_SPHERICAL_EQUIRECTANGULAR_TILE;
2210  else
2211  projection = AV_SPHERICAL_EQUIRECTANGULAR;
2212  break;
2214  if (track->video.projection.private.size < 4) {
2215  av_log(NULL, AV_LOG_ERROR, "Missing projection private properties\n");
2216  return AVERROR_INVALIDDATA;
2217  } else if (track->video.projection.private.size == 12) {
2218  uint32_t layout = bytestream2_get_be32(&gb);
2219  if (layout) {
2221  "Unknown spherical cubemap layout %"PRIu32"\n", layout);
2222  return 0;
2223  }
2224  projection = AV_SPHERICAL_CUBEMAP;
2225  padding = bytestream2_get_be32(&gb);
2226  } else {
2227  av_log(NULL, AV_LOG_ERROR, "Unknown spherical metadata\n");
2228  return AVERROR_INVALIDDATA;
2229  }
2230  break;
2232  /* No Spherical metadata */
2233  return 0;
2234  default:
2236  "Unknown spherical metadata type %"PRIu64"\n",
2237  track->video.projection.type);
2238  return 0;
2239  }
2240 
2241  spherical = av_spherical_alloc(&spherical_size);
2242  if (!spherical)
2243  return AVERROR(ENOMEM);
2244 
2245  spherical->projection = projection;
2246 
2247  spherical->yaw = (int32_t) (track->video.projection.yaw * (1 << 16));
2248  spherical->pitch = (int32_t) (track->video.projection.pitch * (1 << 16));
2249  spherical->roll = (int32_t) (track->video.projection.roll * (1 << 16));
2250 
2251  spherical->padding = padding;
2252 
2253  spherical->bound_left = l;
2254  spherical->bound_top = t;
2255  spherical->bound_right = r;
2256  spherical->bound_bottom = b;
2257 
2259  spherical_size);
2260  if (ret < 0) {
2261  av_freep(&spherical);
2262  return ret;
2263  }
2264 
2265  return 0;
2266 }
2267 
2268 static int get_qt_codec(MatroskaTrack *track, uint32_t *fourcc, enum AVCodecID *codec_id)
2269 {
2270  const AVCodecTag *codec_tags;
2271 
2272  codec_tags = track->type == MATROSKA_TRACK_TYPE_VIDEO ?
2274 
2275  /* Normalize noncompliant private data that starts with the fourcc
2276  * by expanding/shifting the data by 4 bytes and storing the data
2277  * size at the start. */
2278  if (ff_codec_get_id(codec_tags, AV_RL32(track->codec_priv.data))) {
2279  int ret = av_buffer_realloc(&track->codec_priv.buf,
2281  if (ret < 0)
2282  return ret;
2283 
2284  track->codec_priv.data = track->codec_priv.buf->data;
2285  memmove(track->codec_priv.data + 4, track->codec_priv.data, track->codec_priv.size);
2286  track->codec_priv.size += 4;
2287  AV_WB32(track->codec_priv.data, track->codec_priv.size);
2288  }
2289 
2290  *fourcc = AV_RL32(track->codec_priv.data + 4);
2291  *codec_id = ff_codec_get_id(codec_tags, *fourcc);
2292 
2293  return 0;
2294 }
2295 
2297 {
2298  MatroskaDemuxContext *matroska = s->priv_data;
2299  MatroskaTrack *tracks = matroska->tracks.elem;
2300  AVStream *st;
2301  int i, j, ret;
2302  int k;
2303 
2304  for (i = 0; i < matroska->tracks.nb_elem; i++) {
2305  MatroskaTrack *track = &tracks[i];
2307  EbmlList *encodings_list = &track->encodings;
2308  MatroskaTrackEncoding *encodings = encodings_list->elem;
2309  uint8_t *extradata = NULL;
2310  int extradata_size = 0;
2311  int extradata_offset = 0;
2312  uint32_t fourcc = 0;
2313  AVIOContext b;
2314  char* key_id_base64 = NULL;
2315  int bit_depth = -1;
2316 
2317  /* Apply some sanity checks. */
2318  if (track->type != MATROSKA_TRACK_TYPE_VIDEO &&
2319  track->type != MATROSKA_TRACK_TYPE_AUDIO &&
2320  track->type != MATROSKA_TRACK_TYPE_SUBTITLE &&
2321  track->type != MATROSKA_TRACK_TYPE_METADATA) {
2322  av_log(matroska->ctx, AV_LOG_INFO,
2323  "Unknown or unsupported track type %"PRIu64"\n",
2324  track->type);
2325  continue;
2326  }
2327  if (!track->codec_id)
2328  continue;
2329 
2330  if ( track->type == MATROSKA_TRACK_TYPE_AUDIO && track->codec_id[0] != 'A'
2331  || track->type == MATROSKA_TRACK_TYPE_VIDEO && track->codec_id[0] != 'V'
2332  || track->type == MATROSKA_TRACK_TYPE_SUBTITLE && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2333  || track->type == MATROSKA_TRACK_TYPE_METADATA && track->codec_id[0] != 'D' && track->codec_id[0] != 'S'
2334  ) {
2335  av_log(matroska->ctx, AV_LOG_INFO, "Inconsistent track type\n");
2336  continue;
2337  }
2338 
2339  if (track->audio.samplerate < 0 || track->audio.samplerate > INT_MAX ||
2340  isnan(track->audio.samplerate)) {
2341  av_log(matroska->ctx, AV_LOG_WARNING,
2342  "Invalid sample rate %f, defaulting to 8000 instead.\n",
2343  track->audio.samplerate);
2344  track->audio.samplerate = 8000;
2345  }
2346 
2347  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2348  if (!track->default_duration && track->video.frame_rate > 0) {
2349  double default_duration = 1000000000 / track->video.frame_rate;
2350  if (default_duration > UINT64_MAX || default_duration < 0) {
2351  av_log(matroska->ctx, AV_LOG_WARNING,
2352  "Invalid frame rate %e. Cannot calculate default duration.\n",
2353  track->video.frame_rate);
2354  } else {
2355  track->default_duration = default_duration;
2356  }
2357  }
2358  if (track->video.display_width == -1)
2359  track->video.display_width = track->video.pixel_width;
2360  if (track->video.display_height == -1)
2361  track->video.display_height = track->video.pixel_height;
2362  if (track->video.color_space.size == 4)
2363  fourcc = AV_RL32(track->video.color_space.data);
2364  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2365  if (!track->audio.out_samplerate)
2366  track->audio.out_samplerate = track->audio.samplerate;
2367  }
2368  if (encodings_list->nb_elem > 1) {
2369  av_log(matroska->ctx, AV_LOG_ERROR,
2370  "Multiple combined encodings not supported");
2371  } else if (encodings_list->nb_elem == 1) {
2372  if (encodings[0].type) {
2373  if (encodings[0].encryption.key_id.size > 0) {
2374  /* Save the encryption key id to be stored later as a
2375  metadata tag. */
2376  const int b64_size = AV_BASE64_SIZE(encodings[0].encryption.key_id.size);
2377  key_id_base64 = av_malloc(b64_size);
2378  if (key_id_base64 == NULL)
2379  return AVERROR(ENOMEM);
2380 
2381  av_base64_encode(key_id_base64, b64_size,
2382  encodings[0].encryption.key_id.data,
2383  encodings[0].encryption.key_id.size);
2384  } else {
2385  encodings[0].scope = 0;
2386  av_log(matroska->ctx, AV_LOG_ERROR,
2387  "Unsupported encoding type");
2388  }
2389  } else if (
2390 #if CONFIG_ZLIB
2391  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_ZLIB &&
2392 #endif
2393 #if CONFIG_BZLIB
2394  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_BZLIB &&
2395 #endif
2396 #if CONFIG_LZO
2397  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_LZO &&
2398 #endif
2399  encodings[0].compression.algo != MATROSKA_TRACK_ENCODING_COMP_HEADERSTRIP) {
2400  encodings[0].scope = 0;
2401  av_log(matroska->ctx, AV_LOG_ERROR,
2402  "Unsupported encoding type");
2403  } else if (track->codec_priv.size && encodings[0].scope & 2) {
2404  uint8_t *codec_priv = track->codec_priv.data;
2405  int ret = matroska_decode_buffer(&track->codec_priv.data,
2406  &track->codec_priv.size,
2407  track);
2408  if (ret < 0) {
2409  track->codec_priv.data = NULL;
2410  track->codec_priv.size = 0;
2411  av_log(matroska->ctx, AV_LOG_ERROR,
2412  "Failed to decode codec private data\n");
2413  }
2414 
2415  if (codec_priv != track->codec_priv.data) {
2416  av_buffer_unref(&track->codec_priv.buf);
2417  if (track->codec_priv.data) {
2418  track->codec_priv.buf = av_buffer_create(track->codec_priv.data,
2420  NULL, NULL, 0);
2421  if (!track->codec_priv.buf) {
2422  av_freep(&track->codec_priv.data);
2423  track->codec_priv.size = 0;
2424  return AVERROR(ENOMEM);
2425  }
2426  }
2427  }
2428  }
2429  }
2430 
2431  for (j = 0; ff_mkv_codec_tags[j].id != AV_CODEC_ID_NONE; j++) {
2432  if (!strncmp(ff_mkv_codec_tags[j].str, track->codec_id,
2433  strlen(ff_mkv_codec_tags[j].str))) {
2435  break;
2436  }
2437  }
2438 
2439  st = track->stream = avformat_new_stream(s, NULL);
2440  if (!st) {
2441  av_free(key_id_base64);
2442  return AVERROR(ENOMEM);
2443  }
2444 
2445  if (key_id_base64) {
2446  /* export encryption key id as base64 metadata tag */
2447  av_dict_set(&st->metadata, "enc_key_id", key_id_base64, 0);
2448  av_freep(&key_id_base64);
2449  }
2450 
2451  if (!strcmp(track->codec_id, "V_MS/VFW/FOURCC") &&
2452  track->codec_priv.size >= 40 &&
2453  track->codec_priv.data) {
2454  track->ms_compat = 1;
2455  bit_depth = AV_RL16(track->codec_priv.data + 14);
2456  fourcc = AV_RL32(track->codec_priv.data + 16);
2458  fourcc);
2459  if (!codec_id)
2461  fourcc);
2462  extradata_offset = 40;
2463  } else if (!strcmp(track->codec_id, "A_MS/ACM") &&
2464  track->codec_priv.size >= 14 &&
2465  track->codec_priv.data) {
2466  int ret;
2468  track->codec_priv.size,
2469  0, NULL, NULL, NULL, NULL);
2470  ret = ff_get_wav_header(s, &b, st->codecpar, track->codec_priv.size, 0);
2471  if (ret < 0)
2472  return ret;
2473  codec_id = st->codecpar->codec_id;
2474  fourcc = st->codecpar->codec_tag;
2475  extradata_offset = FFMIN(track->codec_priv.size, 18);
2476  } else if (!strcmp(track->codec_id, "A_QUICKTIME")
2477  /* Normally 36, but allow noncompliant private data */
2478  && (track->codec_priv.size >= 32)
2479  && (track->codec_priv.data)) {
2480  uint16_t sample_size;
2481  int ret = get_qt_codec(track, &fourcc, &codec_id);
2482  if (ret < 0)
2483  return ret;
2484  sample_size = AV_RB16(track->codec_priv.data + 26);
2485  if (fourcc == 0) {
2486  if (sample_size == 8) {
2487  fourcc = MKTAG('r','a','w',' ');
2489  } else if (sample_size == 16) {
2490  fourcc = MKTAG('t','w','o','s');
2492  }
2493  }
2494  if ((fourcc == MKTAG('t','w','o','s') ||
2495  fourcc == MKTAG('s','o','w','t')) &&
2496  sample_size == 8)
2498  } else if (!strcmp(track->codec_id, "V_QUICKTIME") &&
2499  (track->codec_priv.size >= 21) &&
2500  (track->codec_priv.data)) {
2501  int ret = get_qt_codec(track, &fourcc, &codec_id);
2502  if (ret < 0)
2503  return ret;
2504  if (codec_id == AV_CODEC_ID_NONE && AV_RL32(track->codec_priv.data+4) == AV_RL32("SMI ")) {
2505  fourcc = MKTAG('S','V','Q','3');
2507  }
2508  if (codec_id == AV_CODEC_ID_NONE)
2509  av_log(matroska->ctx, AV_LOG_ERROR,
2510  "mov FourCC not found %s.\n", av_fourcc2str(fourcc));
2511  if (track->codec_priv.size >= 86) {
2512  bit_depth = AV_RB16(track->codec_priv.data + 82);
2514  track->codec_priv.size,
2515  0, NULL, NULL, NULL, NULL);
2516  if (ff_get_qtpalette(codec_id, &b, track->palette)) {
2517  bit_depth &= 0x1F;
2518  track->has_palette = 1;
2519  }
2520  }
2521  } else if (codec_id == AV_CODEC_ID_PCM_S16BE) {
2522  switch (track->audio.bitdepth) {
2523  case 8:
2525  break;
2526  case 24:
2528  break;
2529  case 32:
2531  break;
2532  }
2533  } else if (codec_id == AV_CODEC_ID_PCM_S16LE) {
2534  switch (track->audio.bitdepth) {
2535  case 8:
2537  break;
2538  case 24:
2540  break;
2541  case 32:
2543  break;
2544  }
2545  } else if (codec_id == AV_CODEC_ID_PCM_F32LE &&
2546  track->audio.bitdepth == 64) {
2548  } else if (codec_id == AV_CODEC_ID_AAC && !track->codec_priv.size) {
2549  int profile = matroska_aac_profile(track->codec_id);
2550  int sri = matroska_aac_sri(track->audio.samplerate);
2551  extradata = av_mallocz(5 + AV_INPUT_BUFFER_PADDING_SIZE);
2552  if (!extradata)
2553  return AVERROR(ENOMEM);
2554  extradata[0] = (profile << 3) | ((sri & 0x0E) >> 1);
2555  extradata[1] = ((sri & 0x01) << 7) | (track->audio.channels << 3);
2556  if (strstr(track->codec_id, "SBR")) {
2557  sri = matroska_aac_sri(track->audio.out_samplerate);
2558  extradata[2] = 0x56;
2559  extradata[3] = 0xE5;
2560  extradata[4] = 0x80 | (sri << 3);
2561  extradata_size = 5;
2562  } else
2563  extradata_size = 2;
2564  } else if (codec_id == AV_CODEC_ID_ALAC && track->codec_priv.size && track->codec_priv.size < INT_MAX - 12 - AV_INPUT_BUFFER_PADDING_SIZE) {
2565  /* Only ALAC's magic cookie is stored in Matroska's track headers.
2566  * Create the "atom size", "tag", and "tag version" fields the
2567  * decoder expects manually. */
2568  extradata_size = 12 + track->codec_priv.size;
2569  extradata = av_mallocz(extradata_size +
2571  if (!extradata)
2572  return AVERROR(ENOMEM);
2573  AV_WB32(extradata, extradata_size);
2574  memcpy(&extradata[4], "alac", 4);
2575  AV_WB32(&extradata[8], 0);
2576  memcpy(&extradata[12], track->codec_priv.data,
2577  track->codec_priv.size);
2578  } else if (codec_id == AV_CODEC_ID_TTA) {
2579  if (track->audio.channels > UINT16_MAX ||
2580  track->audio.bitdepth > UINT16_MAX) {
2581  av_log(matroska->ctx, AV_LOG_WARNING,
2582  "Too large audio channel number %"PRIu64
2583  " or bitdepth %"PRIu64". Skipping track.\n",
2584  track->audio.channels, track->audio.bitdepth);
2585  if (matroska->ctx->error_recognition & AV_EF_EXPLODE)
2586  return AVERROR_INVALIDDATA;
2587  else
2588  continue;
2589  }
2590  if (track->audio.out_samplerate < 0 || track->audio.out_samplerate > INT_MAX)
2591  return AVERROR_INVALIDDATA;
2592  extradata_size = 30;
2593  extradata = av_mallocz(extradata_size + AV_INPUT_BUFFER_PADDING_SIZE);
2594  if (!extradata)
2595  return AVERROR(ENOMEM);
2596  ffio_init_context(&b, extradata, extradata_size, 1,
2597  NULL, NULL, NULL, NULL);
2598  avio_write(&b, "TTA1", 4);
2599  avio_wl16(&b, 1);
2600  avio_wl16(&b, track->audio.channels);
2601  avio_wl16(&b, track->audio.bitdepth);
2602  avio_wl32(&b, track->audio.out_samplerate);
2603  avio_wl32(&b, av_rescale((matroska->duration * matroska->time_scale),
2604  track->audio.out_samplerate,
2605  AV_TIME_BASE * 1000));
2606  } else if (codec_id == AV_CODEC_ID_RV10 ||
2610  extradata_offset = 26;
2611  } else if (codec_id == AV_CODEC_ID_RA_144) {
2612  track->audio.out_samplerate = 8000;
2613  track->audio.channels = 1;
2614  } else if ((codec_id == AV_CODEC_ID_RA_288 ||
2618  && track->codec_priv.data) {
2619  int flavor;
2620 
2622  track->codec_priv.size,
2623  0, NULL, NULL, NULL, NULL);
2624  avio_skip(&b, 22);
2625  flavor = avio_rb16(&b);
2626  track->audio.coded_framesize = avio_rb32(&b);
2627  avio_skip(&b, 12);
2628  track->audio.sub_packet_h = avio_rb16(&b);
2629  track->audio.frame_size = avio_rb16(&b);
2630  track->audio.sub_packet_size = avio_rb16(&b);
2631  if (flavor < 0 ||
2632  track->audio.coded_framesize <= 0 ||
2633  track->audio.sub_packet_h <= 0 ||
2634  track->audio.frame_size <= 0 ||
2635  track->audio.sub_packet_size <= 0 && codec_id != AV_CODEC_ID_SIPR)
2636  return AVERROR_INVALIDDATA;
2637  track->audio.buf = av_malloc_array(track->audio.sub_packet_h,
2638  track->audio.frame_size);
2639  if (!track->audio.buf)
2640  return AVERROR(ENOMEM);
2641  if (codec_id == AV_CODEC_ID_RA_288) {
2642  st->codecpar->block_align = track->audio.coded_framesize;
2643  track->codec_priv.size = 0;
2644  } else {
2645  if (codec_id == AV_CODEC_ID_SIPR && flavor < 4) {
2646  static const int sipr_bit_rate[4] = { 6504, 8496, 5000, 16000 };
2647  track->audio.sub_packet_size = ff_sipr_subpk_size[flavor];
2648  st->codecpar->bit_rate = sipr_bit_rate[flavor];
2649  }
2650  st->codecpar->block_align = track->audio.sub_packet_size;
2651  extradata_offset = 78;
2652  }
2653  } else if (codec_id == AV_CODEC_ID_FLAC && track->codec_priv.size) {
2654  ret = matroska_parse_flac(s, track, &extradata_offset);
2655  if (ret < 0)
2656  return ret;
2657  } else if (codec_id == AV_CODEC_ID_PRORES && track->codec_priv.size == 4) {
2658  fourcc = AV_RL32(track->codec_priv.data);
2659  } else if (codec_id == AV_CODEC_ID_VP9 && track->codec_priv.size) {
2660  /* we don't need any value stored in CodecPrivate.
2661  make sure that it's not exported as extradata. */
2662  track->codec_priv.size = 0;
2663  } else if (codec_id == AV_CODEC_ID_AV1 && track->codec_priv.size) {
2664  /* For now, propagate only the OBUs, if any. Once libavcodec is
2665  updated to handle isobmff style extradata this can be removed. */
2666  extradata_offset = 4;
2667  }
2668  track->codec_priv.size -= extradata_offset;
2669 
2670  if (codec_id == AV_CODEC_ID_NONE)
2671  av_log(matroska->ctx, AV_LOG_INFO,
2672  "Unknown/unsupported AVCodecID %s.\n", track->codec_id);
2673 
2674  if (track->time_scale < 0.01)
2675  track->time_scale = 1.0;
2676  avpriv_set_pts_info(st, 64, matroska->time_scale * track->time_scale,
2677  1000 * 1000 * 1000); /* 64 bit pts in ns */
2678 
2679  /* convert the delay from ns to the track timebase */
2681  (AVRational){ 1, 1000000000 },
2682  st->time_base);
2683 
2684  st->codecpar->codec_id = codec_id;
2685 
2686  if (strcmp(track->language, "und"))
2687  av_dict_set(&st->metadata, "language", track->language, 0);
2688  av_dict_set(&st->metadata, "title", track->name, 0);
2689 
2690  if (track->flag_default)
2692  if (track->flag_forced)
2694 
2695  if (!st->codecpar->extradata) {
2696  if (extradata) {
2697  st->codecpar->extradata = extradata;
2698  st->codecpar->extradata_size = extradata_size;
2699  } else if (track->codec_priv.data && track->codec_priv.size > 0) {
2700  if (ff_alloc_extradata(st->codecpar, track->codec_priv.size))
2701  return AVERROR(ENOMEM);
2702  memcpy(st->codecpar->extradata,
2703  track->codec_priv.data + extradata_offset,
2704  track->codec_priv.size);
2705  }
2706  }
2707 
2708  if (track->type == MATROSKA_TRACK_TYPE_VIDEO) {
2710  int display_width_mul = 1;
2711  int display_height_mul = 1;
2712 
2714  st->codecpar->codec_tag = fourcc;
2715  if (bit_depth >= 0)
2717  st->codecpar->width = track->video.pixel_width;
2718  st->codecpar->height = track->video.pixel_height;
2719 
2721  st->codecpar->field_order = mkv_field_order(matroska, track->video.field_order);
2724 
2726  mkv_stereo_mode_display_mul(track->video.stereo_mode, &display_width_mul, &display_height_mul);
2727 
2730  &st->sample_aspect_ratio.den,
2731  st->codecpar->height * track->video.display_width * display_width_mul,
2732  st->codecpar->width * track->video.display_height * display_height_mul,
2733  255);
2734  }
2735  if (st->codecpar->codec_id != AV_CODEC_ID_HEVC)
2737 
2738  if (track->default_duration) {
2739  int div = track->default_duration <= INT64_MAX ? 1 : 2;
2741  1000000000 / div, track->default_duration / div, 30000);
2742 #if FF_API_R_FRAME_RATE
2743  if ( st->avg_frame_rate.num < st->avg_frame_rate.den * 1000LL
2744  && st->avg_frame_rate.num > st->avg_frame_rate.den * 5LL)
2745  st->r_frame_rate = st->avg_frame_rate;
2746 #endif
2747  }
2748 
2749  /* export stereo mode flag as metadata tag */
2751  av_dict_set(&st->metadata, "stereo_mode", ff_matroska_video_stereo_mode[track->video.stereo_mode], 0);
2752 
2753  /* export alpha mode flag as metadata tag */
2754  if (track->video.alpha_mode)
2755  av_dict_set(&st->metadata, "alpha_mode", "1", 0);
2756 
2757  /* if we have virtual track, mark the real tracks */
2758  for (j=0; j < track->operation.combine_planes.nb_elem; j++) {
2759  char buf[32];
2761  continue;
2762  snprintf(buf, sizeof(buf), "%s_%d",
2764  for (k=0; k < matroska->tracks.nb_elem; k++)
2765  if (planes[j].uid == tracks[k].uid && tracks[k].stream) {
2766  av_dict_set(&tracks[k].stream->metadata,
2767  "stereo_mode", buf, 0);
2768  break;
2769  }
2770  }
2771  // add stream level stereo3d side data if it is a supported format
2773  track->video.stereo_mode != 10 && track->video.stereo_mode != 12) {
2774  int ret = ff_mkv_stereo3d_conv(st, track->video.stereo_mode);
2775  if (ret < 0)
2776  return ret;
2777  }
2778 
2779  ret = mkv_parse_video_color(st, track);
2780  if (ret < 0)
2781  return ret;
2782  ret = mkv_parse_video_projection(st, track);
2783  if (ret < 0)
2784  return ret;
2785  } else if (track->type == MATROSKA_TRACK_TYPE_AUDIO) {
2787  st->codecpar->codec_tag = fourcc;
2788  st->codecpar->sample_rate = track->audio.out_samplerate;
2789  st->codecpar->channels = track->audio.channels;
2790  if (!st->codecpar->bits_per_coded_sample)
2792  if (st->codecpar->codec_id == AV_CODEC_ID_MP3 ||
2793  st->codecpar->codec_id == AV_CODEC_ID_MLP ||
2796  else if (st->codecpar->codec_id != AV_CODEC_ID_AAC)
2798  if (track->codec_delay > 0) {
2800  (AVRational){1, 1000000000},
2801  (AVRational){1, st->codecpar->codec_id == AV_CODEC_ID_OPUS ?
2802  48000 : st->codecpar->sample_rate});
2803  }
2804  if (track->seek_preroll > 0) {
2806  (AVRational){1, 1000000000},
2807  (AVRational){1, st->codecpar->sample_rate});
2808  }
2809  } else if (codec_id == AV_CODEC_ID_WEBVTT) {
2811 
2812  if (!strcmp(track->codec_id, "D_WEBVTT/CAPTIONS")) {
2814  } else if (!strcmp(track->codec_id, "D_WEBVTT/DESCRIPTIONS")) {
2816  } else if (!strcmp(track->codec_id, "D_WEBVTT/METADATA")) {
2818  }
2819  } else if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE) {
2821  }
2822  }
2823 
2824  return 0;
2825 }
2826 
2828 {
2829  MatroskaDemuxContext *matroska = s->priv_data;
2830  EbmlList *attachments_list = &matroska->attachments;
2831  EbmlList *chapters_list = &matroska->chapters;
2832  MatroskaAttachment *attachments;
2833  MatroskaChapter *chapters;
2834  uint64_t max_start = 0;
2835  int64_t pos;
2836  Ebml ebml = { 0 };
2837  int i, j, res;
2838 
2839  matroska->ctx = s;
2840  matroska->cues_parsing_deferred = 1;
2841 
2842  /* First read the EBML header. */
2843  if (ebml_parse(matroska, ebml_syntax, &ebml) || !ebml.doctype) {
2844  av_log(matroska->ctx, AV_LOG_ERROR, "EBML header parsing failed\n");
2845  ebml_free(ebml_syntax, &ebml);
2846  return AVERROR_INVALIDDATA;
2847  }
2848  if (ebml.version > EBML_VERSION ||
2849  ebml.max_size > sizeof(uint64_t) ||
2850  ebml.id_length > sizeof(uint32_t) ||
2851  ebml.doctype_version > 3) {
2853  "EBML version %"PRIu64", doctype %s, doc version %"PRIu64,
2854  ebml.version, ebml.doctype, ebml.doctype_version);
2855  ebml_free(ebml_syntax, &ebml);
2856  return AVERROR_PATCHWELCOME;
2857  } else if (ebml.doctype_version == 3) {
2858  av_log(matroska->ctx, AV_LOG_WARNING,
2859  "EBML header using unsupported features\n"
2860  "(EBML version %"PRIu64", doctype %s, doc version %"PRIu64")\n",
2861  ebml.version, ebml.doctype, ebml.doctype_version);
2862  }
2863  for (i = 0; i < FF_ARRAY_ELEMS(matroska_doctypes); i++)
2864  if (!strcmp(ebml.doctype, matroska_doctypes[i]))
2865  break;
2867  av_log(s, AV_LOG_WARNING, "Unknown EBML doctype '%s'\n", ebml.doctype);
2868  if (matroska->ctx->error_recognition & AV_EF_EXPLODE) {
2869  ebml_free(ebml_syntax, &ebml);
2870  return AVERROR_INVALIDDATA;
2871  }
2872  }
2873  ebml_free(ebml_syntax, &ebml);
2874 
2875  /* The next thing is a segment. */
2876  pos = avio_tell(matroska->ctx->pb);
2877  res = ebml_parse(matroska, matroska_segments, matroska);
2878  // Try resyncing until we find an EBML_STOP type element.
2879  while (res != 1) {
2880  res = matroska_resync(matroska, pos);
2881  if (res < 0)
2882  goto fail;
2883  pos = avio_tell(matroska->ctx->pb);
2884  res = ebml_parse(matroska, matroska_segment, matroska);
2885  if (res == AVERROR(EIO)) // EOF is translated to EIO, this exists the loop on EOF
2886  goto fail;
2887  }
2888  /* Set data_offset as it might be needed later by seek_frame_generic. */
2889  if (matroska->current_id == MATROSKA_ID_CLUSTER)
2890  s->internal->data_offset = avio_tell(matroska->ctx->pb) - 4;
2891  matroska_execute_seekhead(matroska);
2892 
2893  if (!matroska->time_scale)
2894  matroska->time_scale = 1000000;
2895  if (isnan(matroska->duration))
2896  matroska->duration = 0;
2897  if (matroska->duration)
2898  matroska->ctx->duration = matroska->duration * matroska->time_scale *
2899  1000 / AV_TIME_BASE;
2900  av_dict_set(&s->metadata, "title", matroska->title, 0);
2901  av_dict_set(&s->metadata, "encoder", matroska->muxingapp, 0);
2902 
2903  if (matroska->date_utc.size == 8)
2904  matroska_metadata_creation_time(&s->metadata, AV_RB64(matroska->date_utc.data));
2905 
2906  res = matroska_parse_tracks(s);
2907  if (res < 0)
2908  goto fail;
2909 
2910  attachments = attachments_list->elem;
2911  for (j = 0; j < attachments_list->nb_elem; j++) {
2912  if (!(attachments[j].filename && attachments[j].mime &&
2913  attachments[j].bin.data && attachments[j].bin.size > 0)) {
2914  av_log(matroska->ctx, AV_LOG_ERROR, "incomplete attachment\n");
2915  } else {
2917  if (!st)
2918  break;
2919  av_dict_set(&st->metadata, "filename", attachments[j].filename, 0);
2920  av_dict_set(&st->metadata, "mimetype", attachments[j].mime, 0);
2922 
2923  for (i = 0; ff_mkv_image_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2924  if (!strncmp(ff_mkv_image_mime_tags[i].str, attachments[j].mime,
2925  strlen(ff_mkv_image_mime_tags[i].str))) {
2927  break;
2928  }
2929  }
2930 
2931  attachments[j].stream = st;
2932 
2933  if (st->codecpar->codec_id != AV_CODEC_ID_NONE) {
2934  AVPacket *pkt = &st->attached_pic;
2935 
2938 
2940  pkt->buf = av_buffer_ref(attachments[j].bin.buf);
2941  if (!pkt->buf)
2942  return AVERROR(ENOMEM);
2943  pkt->data = attachments[j].bin.data;
2944  pkt->size = attachments[j].bin.size;
2945  pkt->stream_index = st->index;
2947  } else {
2949  if (ff_alloc_extradata(st->codecpar, attachments[j].bin.size))
2950  break;
2951  memcpy(st->codecpar->extradata, attachments[j].bin.data,
2952  attachments[j].bin.size);
2953 
2954  for (i = 0; ff_mkv_mime_tags[i].id != AV_CODEC_ID_NONE; i++) {
2955  if (!strncmp(ff_mkv_mime_tags[i].str, attachments[j].mime,
2956  strlen(ff_mkv_mime_tags[i].str))) {
2958  break;
2959  }
2960  }
2961  }
2962  }
2963  }
2964 
2965  chapters = chapters_list->elem;
2966  for (i = 0; i < chapters_list->nb_elem; i++)
2967  if (chapters[i].start != AV_NOPTS_VALUE && chapters[i].uid &&
2968  (max_start == 0 || chapters[i].start > max_start)) {
2969  chapters[i].chapter =
2970  avpriv_new_chapter(s, chapters[i].uid,
2971  (AVRational) { 1, 1000000000 },
2972  chapters[i].start, chapters[i].end,
2973  chapters[i].title);
2974  if (chapters[i].chapter) {
2975  av_dict_set(&chapters[i].chapter->metadata,
2976  "title", chapters[i].title, 0);
2977  }
2978  max_start = chapters[i].start;
2979  }
2980 
2981  matroska_add_index_entries(matroska);
2982 
2984 
2985  return 0;
2986 fail:
2988  return res;
2990 
2991 /*
2992  * Put one packet in an application-supplied AVPacket struct.
2993  * Returns 0 on success or -1 on failure.
2994  */
2995 static int matroska_deliver_packet(MatroskaDemuxContext *matroska,
2996  AVPacket *pkt)
2997 {
2998  if (matroska->queue) {
2999  MatroskaTrack *tracks = matroska->tracks.elem;
3000  MatroskaTrack *track;
3001 
3002  ff_packet_list_get(&matroska->queue, &matroska->queue_end, pkt);
3003  track = &tracks[pkt->stream_index];
3004  if (track->has_palette) {
3006  if (!pal) {
3007  av_log(matroska->ctx, AV_LOG_ERROR, "Cannot append palette to packet\n");
3008  } else {
3009  memcpy(pal, track->palette, AVPALETTE_SIZE);
3010  }
3011  track->has_palette = 0;
3012  }
3013  return 0;
3014  }
3015 
3016  return -1;
3017 }
3018 
3019 /*
3020  * Free all packets in our internal queue.
3021  */
3022 static void matroska_clear_queue(MatroskaDemuxContext *matroska)
3023 {
3024  ff_packet_list_free(&matroska->queue, &matroska->queue_end);
3025 }
3026 
3027 static int matroska_parse_laces(MatroskaDemuxContext *matroska, uint8_t **buf,
3028  int *buf_size, int type,
3029  uint32_t **lace_buf, int *laces)
3030 {
3031  int res = 0, n, size = *buf_size;
3032  uint8_t *data = *buf;
3033  uint32_t *lace_size;
3034 
3035  if (!type) {
3036  *laces = 1;
3037  *lace_buf = av_malloc(sizeof(**lace_buf));
3038  if (!*lace_buf)
3039  return AVERROR(ENOMEM);
3040 
3041  *lace_buf[0] = size;
3042  return 0;
3043  }
3044 
3045  if (size <= 0)
3046  return AVERROR_INVALIDDATA;
3047 
3048  *laces = *data + 1;
3049  data += 1;
3050  size -= 1;
3051  lace_size = av_malloc_array(*laces, sizeof(*lace_size));
3052  if (!lace_size)
3053  return AVERROR(ENOMEM);
3054 
3055  switch (type) {
3056  case 0x1: /* Xiph lacing */
3057  {
3058  uint8_t temp;
3059  uint32_t total = 0;
3060  for (n = 0; res == 0 && n < *laces - 1; n++) {
3061  lace_size[n] = 0;
3062 
3063  while (1) {
3064  if (size <= total) {
3065  res = AVERROR_INVALIDDATA;
3066  break;
3067  }
3068  temp = *data;
3069  total += temp;
3070  lace_size[n] += temp;
3071  data += 1;
3072  size -= 1;
3073  if (temp != 0xff)
3074  break;
3075  }
3076  }
3077  if (size < total) {
3078  res = AVERROR_INVALIDDATA;
3079  break;
3080  }
3081 
3082  lace_size[n] = size - total;
3083  break;
3084  }
3085 
3086  case 0x2: /* fixed-size lacing */
3087  if (size % (*laces)) {
3088  res = AVERROR_INVALIDDATA;
3089  break;
3090  }
3091  for (n = 0; n < *laces; n++)
3092  lace_size[n] = size / *laces;
3093  break;
3094 
3095  case 0x3: /* EBML lacing */
3096  {
3097  uint64_t num;
3098  uint64_t total;
3099  n = matroska_ebmlnum_uint(matroska, data, size, &num);
3100  if (n < 0 || num > INT_MAX) {
3101  av_log(matroska->ctx, AV_LOG_INFO,
3102  "EBML block data error\n");
3103  res = n<0 ? n : AVERROR_INVALIDDATA;
3104  break;
3105  }
3106  data += n;
3107  size -= n;
3108  total = lace_size[0] = num;
3109  for (n = 1; res == 0 && n < *laces - 1; n++) {
3110  int64_t snum;
3111  int r;
3112  r = matroska_ebmlnum_sint(matroska, data, size, &snum);
3113  if (r < 0 || lace_size[n - 1] + snum > (uint64_t)INT_MAX) {
3114  av_log(matroska->ctx, AV_LOG_INFO,
3115  "EBML block data error\n");
3116  res = r<0 ? r : AVERROR_INVALIDDATA;
3117  break;
3118  }
3119  data += r;
3120  size -= r;
3121  lace_size[n] = lace_size[n - 1] + snum;
3122  total += lace_size[n];
3123  }
3124  if (size < total) {
3125  res = AVERROR_INVALIDDATA;
3126  break;
3127  }
3128  lace_size[*laces - 1] = size - total;
3129  break;
3130  }
3131  }
3132 
3133  *buf = data;
3134  *lace_buf = lace_size;
3135  *buf_size = size;
3136 
3137  return res;
3138 }
3139 
3140 static int matroska_parse_rm_audio(MatroskaDemuxContext *matroska,
3141  MatroskaTrack *track, AVStream *st,
3142  uint8_t *data, int size, uint64_t timecode,
3143  int64_t pos)
3144 {
3145  int a = st->codecpar->block_align;
3146  int sps = track->audio.sub_packet_size;
3147  int cfs = track->audio.coded_framesize;
3148  int h = track->audio.sub_packet_h;
3149  int y = track->audio.sub_packet_cnt;
3150  int w = track->audio.frame_size;
3151  int x;
3152 
3153  if (!track->audio.pkt_cnt) {
3154  if (track->audio.sub_packet_cnt == 0)
3155  track->audio.buf_timecode = timecode;
3156  if (st->codecpar->codec_id == AV_CODEC_ID_RA_288) {
3157  if (size < cfs * h / 2) {
3158  av_log(matroska->ctx, AV_LOG_ERROR,
3159  "Corrupt int4 RM-style audio packet size\n");
3160  return AVERROR_INVALIDDATA;
3161  }
3162  for (x = 0; x < h / 2; x++)
3163  memcpy(track->audio.buf + x * 2 * w + y * cfs,
3164  data + x * cfs, cfs);
3165  } else if (st->codecpar->codec_id == AV_CODEC_ID_SIPR) {
3166  if (size < w) {
3167  av_log(matroska->ctx, AV_LOG_ERROR,
3168  "Corrupt sipr RM-style audio packet size\n");
3169  return AVERROR_INVALIDDATA;
3170  }
3171  memcpy(track->audio.buf + y * w, data, w);
3172  } else {
3173  if (size < sps * w / sps || h<=0 || w%sps) {
3174  av_log(matroska->ctx, AV_LOG_ERROR,
3175  "Corrupt generic RM-style audio packet size\n");
3176  return AVERROR_INVALIDDATA;
3177  }
3178  for (x = 0; x < w / sps; x++)
3179  memcpy(track->audio.buf +
3180  sps * (h * x + ((h + 1) / 2) * (y & 1) + (y >> 1)),
3181  data + x * sps, sps);
3182  }
3183 
3184  if (++track->audio.sub_packet_cnt >= h) {
3185  if (st->codecpar->codec_id == AV_CODEC_ID_SIPR)
3186  ff_rm_reorder_sipr_data(track->audio.buf, h, w);
3187  track->audio.sub_packet_cnt = 0;
3188  track->audio.pkt_cnt = h * w / a;
3189  }
3190  }
3191 
3192  while (track->audio.pkt_cnt) {
3193  int ret;
3194  AVPacket pktl, *pkt = &pktl;
3195 
3196  ret = av_new_packet(pkt, a);
3197  if (ret < 0) {
3198  return ret;
3199  }
3200  memcpy(pkt->data,
3201  track->audio.buf + a * (h * w / a - track->audio.pkt_cnt--),
3202  a);
3203  pkt->pts = track->audio.buf_timecode;
3205  pkt->pos = pos;
3206  pkt->stream_index = st->index;
3207  ret = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3208  if (ret < 0) {
3210  return AVERROR(ENOMEM);
3211  }
3212  }
3213 
3214  return 0;
3215 }
3216 
3217 /* reconstruct full wavpack blocks from mangled matroska ones */
3218 static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src,
3219  uint8_t **pdst, int *size)
3220 {
3221  uint8_t *dst = NULL;
3222  int dstlen = 0;
3223  int srclen = *size;
3224  uint32_t samples;
3225  uint16_t ver;
3226  int ret, offset = 0;
3227 
3228  if (srclen < 12 || track->stream->codecpar->extradata_size < 2)
3229  return AVERROR_INVALIDDATA;
3230 
3231  ver = AV_RL16(track->stream->codecpar->extradata);
3232 
3233  samples = AV_RL32(src);
3234  src += 4;
3235  srclen -= 4;
3236 
3237  while (srclen >= 8) {
3238  int multiblock;
3239  uint32_t blocksize;
3240  uint8_t *tmp;
3241 
3242  uint32_t flags = AV_RL32(src);
3243  uint32_t crc = AV_RL32(src + 4);
3244  src += 8;
3245  srclen -= 8;
3246 
3247  multiblock = (flags & 0x1800) != 0x1800;
3248  if (multiblock) {
3249  if (srclen < 4) {
3251  goto fail;
3252  }
3253  blocksize = AV_RL32(src);
3254  src += 4;
3255  srclen -= 4;
3256  } else
3257  blocksize = srclen;
3258 
3259  if (blocksize > srclen) {
3261  goto fail;
3262  }
3263 
3264  tmp = av_realloc(dst, dstlen + blocksize + 32 + AV_INPUT_BUFFER_PADDING_SIZE);
3265  if (!tmp) {
3266  ret = AVERROR(ENOMEM);
3267  goto fail;
3268  }
3269  dst = tmp;
3270  dstlen += blocksize + 32;
3271 
3272  AV_WL32(dst + offset, MKTAG('w', 'v', 'p', 'k')); // tag
3273  AV_WL32(dst + offset + 4, blocksize + 24); // blocksize - 8
3274  AV_WL16(dst + offset + 8, ver); // version
3275  AV_WL16(dst + offset + 10, 0); // track/index_no
3276  AV_WL32(dst + offset + 12, 0); // total samples
3277  AV_WL32(dst + offset + 16, 0); // block index
3278  AV_WL32(dst + offset + 20, samples); // number of samples
3279  AV_WL32(dst + offset + 24, flags); // flags
3280  AV_WL32(dst + offset + 28, crc); // crc
3281  memcpy(dst + offset + 32, src, blocksize); // block data
3282 
3283  src += blocksize;
3284  srclen -= blocksize;
3285  offset += blocksize + 32;
3286  }
3287 
3288  memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3289 
3290  *pdst = dst;
3291  *size = dstlen;
3292 
3293  return 0;
3295 fail:
3296  av_freep(&dst);
3297  return ret;
3298 }
3299 
3300 static int matroska_parse_prores(MatroskaTrack *track, uint8_t *src,
3301  uint8_t **pdst, int *size)
3302 {
3303  uint8_t *dst;
3304  int dstlen = *size + 8;
3305 
3306  dst = av_malloc(dstlen + AV_INPUT_BUFFER_PADDING_SIZE);
3307  if (!dst)
3308  return AVERROR(ENOMEM);
3309 
3310  AV_WB32(dst, dstlen);
3311  AV_WB32(dst + 4, MKBETAG('i', 'c', 'p', 'f'));
3312  memcpy(dst + 8, src, dstlen - 8);
3313  memset(dst + dstlen, 0, AV_INPUT_BUFFER_PADDING_SIZE);
3314 
3315  *pdst = dst;
3316  *size = dstlen;
3317 
3318  return 0;
3319 }
3320 
3321 static int matroska_parse_webvtt(MatroskaDemuxContext *matroska,
3322  MatroskaTrack *track,
3323  AVStream *st,
3324  uint8_t *data, int data_len,
3325  uint64_t timecode,
3326  uint64_t duration,
3327  int64_t pos)
3328 {
3329  AVPacket pktl, *pkt = &pktl;
3330  uint8_t *id, *settings, *text, *buf;
3331  int id_len, settings_len, text_len;
3332  uint8_t *p, *q;
3333  int err;
3334 
3335  if (data_len <= 0)
3336  return AVERROR_INVALIDDATA;
3337 
3338  p = data;
3339  q = data + data_len;
3340 
3341  id = p;
3342  id_len = -1;
3343  while (p < q) {
3344  if (*p == '\r' || *p == '\n') {
3345  id_len = p - id;
3346  if (*p == '\r')
3347  p++;
3348  break;
3349  }
3350  p++;
3351  }
3352 
3353  if (p >= q || *p != '\n')
3354  return AVERROR_INVALIDDATA;
3355  p++;
3356 
3357  settings = p;
3358  settings_len = -1;
3359  while (p < q) {
3360  if (*p == '\r' || *p == '\n') {
3361  settings_len = p - settings;
3362  if (*p == '\r')
3363  p++;
3364  break;
3365  }
3366  p++;
3367  }
3368 
3369  if (p >= q || *p != '\n')
3370  return AVERROR_INVALIDDATA;
3371  p++;
3372 
3373  text = p;
3374  text_len = q - p;
3375  while (text_len > 0) {
3376  const int len = text_len - 1;
3377  const uint8_t c = p[len];
3378  if (c != '\r' && c != '\n')
3379  break;
3380  text_len = len;
3381  }
3382 
3383  if (text_len <= 0)
3384  return AVERROR_INVALIDDATA;
3385 
3386  err = av_new_packet(pkt, text_len);
3387  if (err < 0) {
3388  return err;
3389  }
3390 
3391  memcpy(pkt->data, text, text_len);
3392 
3393  if (id_len > 0) {
3396  id_len);
3397  if (!buf) {
3399  return AVERROR(ENOMEM);
3400  }
3401  memcpy(buf, id, id_len);
3402  }
3403 
3404  if (settings_len > 0) {
3407  settings_len);
3408  if (!buf) {
3410  return AVERROR(ENOMEM);
3411  }
3412  memcpy(buf, settings, settings_len);
3413  }
3414 
3415  // Do we need this for subtitles?
3416  // pkt->flags = AV_PKT_FLAG_KEY;
3417 
3418  pkt->stream_index = st->index;
3419  pkt->pts = timecode;
3420 
3421  // Do we need this for subtitles?
3422  // pkt->dts = timecode;
3423 
3424  pkt->duration = duration;
3425  pkt->pos = pos;
3426 
3427  err = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3428  if (err < 0) {
3430  return AVERROR(ENOMEM);
3431  }
3432 
3433  return 0;
3434 }
3435 
3436 static int matroska_parse_frame(MatroskaDemuxContext *matroska,
3437  MatroskaTrack *track, AVStream *st,
3438  AVBufferRef *buf, uint8_t *data, int pkt_size,
3439  uint64_t timecode, uint64_t lace_duration,
3440  int64_t pos, int is_keyframe,
3441  uint8_t *additional, uint64_t additional_id, int additional_size,
3442  int64_t discard_padding)
3443 {
3444  MatroskaTrackEncoding *encodings = track->encodings.elem;
3445  uint8_t *pkt_data = data;
3446  int res = 0;
3447  AVPacket pktl, *pkt = &pktl;
3448 
3449  if (encodings && !encodings->type && encodings->scope & 1) {
3450  res = matroska_decode_buffer(&pkt_data, &pkt_size, track);
3451  if (res < 0)
3452  return res;
3453  }
3454 
3455  if (st->codecpar->codec_id == AV_CODEC_ID_WAVPACK) {
3456  uint8_t *wv_data;
3457  res = matroska_parse_wavpack(track, pkt_data, &wv_data, &pkt_size);
3458  if (res < 0) {
3459  av_log(matroska->ctx, AV_LOG_ERROR,
3460  "Error parsing a wavpack block.\n");
3461  goto fail;
3462  }
3463  if (pkt_data != data)
3464  av_freep(&pkt_data);
3465  pkt_data = wv_data;
3466  }
3467 
3468  if (st->codecpar->codec_id == AV_CODEC_ID_PRORES &&
3469  AV_RB32(pkt_data + 4) != MKBETAG('i', 'c', 'p', 'f')) {
3470  uint8_t *pr_data;
3471  res = matroska_parse_prores(track, pkt_data, &pr_data, &pkt_size);
3472  if (res < 0) {
3473  av_log(matroska->ctx, AV_LOG_ERROR,
3474  "Error parsing a prores block.\n");
3475  goto fail;
3476  }
3477  if (pkt_data != data)
3478  av_freep(&pkt_data);
3479  pkt_data = pr_data;
3480  }
3481 
3482  if (!pkt_size && !additional_size)
3483  goto no_output;
3484 
3486  if (pkt_data != data)
3487  pkt->buf = av_buffer_create(pkt_data, pkt_size + AV_INPUT_BUFFER_PADDING_SIZE,
3488  NULL, NULL, 0);
3489  else
3490  pkt->buf = av_buffer_ref(buf);
3491 
3492  if (!pkt->buf) {
3493  res = AVERROR(ENOMEM);
3494  goto fail;
3495  }
3496 
3497  pkt->data = pkt_data;
3498  pkt->size = pkt_size;
3499  pkt->flags = is_keyframe;
3500  pkt->stream_index = st->index;
3501 
3502  if (additional_size > 0) {
3503  uint8_t *side_data = av_packet_new_side_data(pkt,
3505  additional_size + 8);
3506  if (!side_data) {
3508  return AVERROR(ENOMEM);
3509  }
3510  AV_WB64(side_data, additional_id);
3511  memcpy(side_data + 8, additional, additional_size);
3512  }
3513 
3514  if (discard_padding) {
3515  uint8_t *side_data = av_packet_new_side_data(pkt,
3517  10);
3518  if (!side_data) {
3520  return AVERROR(ENOMEM);
3521  }
3522  discard_padding = av_rescale_q(discard_padding,
3523  (AVRational){1, 1000000000},
3524  (AVRational){1, st->codecpar->sample_rate});
3525  if (discard_padding > 0) {
3526  AV_WL32(side_data + 4, discard_padding);
3527  } else {
3528  AV_WL32(side_data, -discard_padding);
3529  }
3530  }
3531 
3532  if (track->ms_compat)
3533  pkt->dts = timecode;
3534  else
3535  pkt->pts = timecode;
3536  pkt->pos = pos;
3537  pkt->duration = lace_duration;
3538 
3539 #if FF_API_CONVERGENCE_DURATION
3541  if (st->codecpar->codec_id == AV_CODEC_ID_SUBRIP) {
3542  pkt->convergence_duration = lace_duration;
3543  }
3545 #endif
3546 
3547  res = ff_packet_list_put(&matroska->queue, &matroska->queue_end, pkt, 0);
3548  if (res < 0) {
3550  return AVERROR(ENOMEM);
3551  }
3552 
3553  return 0;
3554 
3555 no_output:
3557  if (pkt_data != data)
3558  av_freep(&pkt_data);
3559  return res;
3560 }
3561 
3563  int size, int64_t pos, uint64_t cluster_time,
3564  uint64_t block_duration, int is_keyframe,
3565  uint8_t *additional, uint64_t additional_id, int additional_size,
3566  int64_t cluster_pos, int64_t discard_padding)
3567 {
3568  uint64_t timecode = AV_NOPTS_VALUE;
3569  MatroskaTrack *track;
3570  int res = 0;
3571  AVStream *st;
3572  int16_t block_time;
3573  uint32_t *lace_size = NULL;
3574  int n, flags, laces = 0;
3575  uint64_t num;
3576  int trust_default_duration = 1;
3577 
3578  if ((n = matroska_ebmlnum_uint(matroska, data, size, &num)) < 0) {
3579  return n;
3580  }
3581  data += n;
3582  size -= n;
3583 
3584  track = matroska_find_track_by_num(matroska, num);
3585  if (!track || size < 3)
3586  return AVERROR_INVALIDDATA;
3587 
3588  if (!(st = track->stream)) {
3589  av_log(matroska->ctx, AV_LOG_VERBOSE,
3590  "No stream associated to TrackNumber %"PRIu64". "
3591  "Ignoring Block with this TrackNumber.\n", num);
3592  return 0;
3593  }
3594 
3595  if (st->discard >= AVDISCARD_ALL)
3596  return res;
3597  if (block_duration > INT64_MAX)
3598  block_duration = INT64_MAX;
3599 
3600  block_time = sign_extend(AV_RB16(data), 16);
3601  data += 2;
3602  flags = *data++;
3603  size -= 3;
3604  if (is_keyframe == -1)
3605  is_keyframe = flags & 0x80 ? AV_PKT_FLAG_KEY : 0;
3606 
3607  if (cluster_time != (uint64_t) -1 &&
3608  (block_time >= 0 || cluster_time >= -block_time)) {
3609  timecode = cluster_time + block_time - track->codec_delay_in_track_tb;
3610  if (track->type == MATROSKA_TRACK_TYPE_SUBTITLE &&
3611  timecode < track->end_timecode)
3612  is_keyframe = 0; /* overlapping subtitles are not key frame */
3613  if (is_keyframe) {
3614  ff_reduce_index(matroska->ctx, st->index);
3615  av_add_index_entry(st, cluster_pos, timecode, 0, 0,
3617  }
3618  }
3619 
3620  if (matroska->skip_to_keyframe &&
3621  track->type != MATROSKA_TRACK_TYPE_SUBTITLE) {
3622  // Compare signed timecodes. Timecode may be negative due to codec delay
3623  // offset. We don't support timestamps greater than int64_t anyway - see
3624  // AVPacket's pts.
3625  if ((int64_t)timecode < (int64_t)matroska->skip_to_timecode)
3626  return res;
3627  if (is_keyframe)
3628  matroska->skip_to_keyframe = 0;
3629  else if (!st->skip_to_keyframe) {
3630  av_log(matroska->ctx, AV_LOG_ERROR, "File is broken, keyframes not correctly marked!\n");
3631  matroska->skip_to_keyframe = 0;
3632  }
3633  }
3634 
3635  res = matroska_parse_laces(matroska, &data, &size, (flags & 0x06) >> 1,
3636  &lace_size, &laces);
3637 
3638  if (res)
3639  goto end;
3640 
3641  if (track->audio.samplerate == 8000) {
3642  // If this is needed for more codecs, then add them here
3643  if (st->codecpar->codec_id == AV_CODEC_ID_AC3) {
3644  if (track->audio.samplerate != st->codecpar->sample_rate || !st->codecpar->frame_size)
3645  trust_default_duration = 0;
3646  }
3647  }
3648 
3649  if (!block_duration && trust_default_duration)
3650  block_duration = track->default_duration * laces / matroska->time_scale;
3651 
3652  if (cluster_time != (uint64_t)-1 && (block_time >= 0 || cluster_time >= -block_time))
3653  track->end_timecode =
3654  FFMAX(track->end_timecode, timecode + block_duration);
3655 
3656  for (n = 0; n < laces; n++) {
3657  int64_t lace_duration = block_duration*(n+1) / laces - block_duration*n / laces;
3658 
3659  if (lace_size[n] > size) {
3660  av_log(matroska->ctx, AV_LOG_ERROR, "Invalid packet size\n");
3661  break;
3662  }
3663 
3664  if ((st->codecpar->codec_id == AV_CODEC_ID_RA_288 ||
3668  st->codecpar->block_align && track->audio.sub_packet_size) {
3669  res = matroska_parse_rm_audio(matroska, track, st, data,
3670  lace_size[n],
3671  timecode, pos);
3672  if (res)
3673  goto end;
3674 
3675  } else if (st->codecpar->codec_id == AV_CODEC_ID_WEBVTT) {
3676  res = matroska_parse_webvtt(matroska, track, st,
3677  data, lace_size[n],
3678  timecode, lace_duration,
3679  pos);
3680  if (res)
3681  goto end;
3682  } else {
3683  res = matroska_parse_frame(matroska, track, st, buf, data, lace_size[n],
3684  timecode, lace_duration, pos,
3685  !n ? is_keyframe : 0,
3686  additional, additional_id, additional_size,
3687  discard_padding);
3688  if (res)
3689  goto end;
3690  }
3691 
3692  if (timecode != AV_NOPTS_VALUE)
3693  timecode = lace_duration ? timecode + lace_duration : AV_NOPTS_VALUE;
3694  data += lace_size[n];
3695  size -= lace_size[n];
3696  }
3698 end:
3699  av_free(lace_size);
3700  return res;
3701 }
3702 
3703 static int matroska_parse_cluster(MatroskaDemuxContext *matroska)
3704 {
3705  MatroskaCluster *cluster = &matroska->current_cluster;
3706  MatroskaBlock *block = &cluster->block;
3707  int res;
3708 
3709  av_assert0(matroska->num_levels <= 2);
3710 
3711  if (matroska->num_levels == 1) {
3712  res = ebml_parse(matroska, matroska_segment, NULL);
3713 
3714  if (res == 1) {
3715  /* Found a cluster: subtract the size of the ID already read. */
3716  cluster->pos = avio_tell(matroska->ctx->pb) - 4;
3717 
3718  res = ebml_parse(matroska, matroska_cluster_enter, cluster);
3719  if (res < 0)
3720  return res;
3721  }
3722  }
3723 
3724  if (matroska->num_levels == 2) {
3725  /* We are inside a cluster. */
3726  res = ebml_parse(matroska, matroska_cluster_parsing, cluster);
3727 
3728  if (res >= 0 && block->bin.size > 0) {
3729  int is_keyframe = block->non_simple ? block->reference == INT64_MIN : -1;
3730  uint8_t* additional = block->additional.size > 0 ?
3731  block->additional.data : NULL;
3732 
3733  res = matroska_parse_block(matroska, block->bin.buf, block->bin.data,
3734  block->bin.size, block->bin.pos,
3735  cluster->timecode, block->duration,
3736  is_keyframe, additional, block->additional_id,
3737  block->additional.size, cluster->pos,
3738  block->discard_padding);
3739  }
3740 
3742  memset(block, 0, sizeof(*block));
3743  } else if (!matroska->num_levels) {
3744  if (!avio_feof(matroska->ctx->pb)) {
3745  avio_r8(matroska->ctx->pb);
3746  if (!avio_feof(matroska->ctx->pb)) {
3747  av_log(matroska->ctx, AV_LOG_WARNING, "File extends beyond "
3748  "end of segment.\n");
3749  return AVERROR_INVALIDDATA;
3750  }
3751  }
3752  matroska->done = 1;
3753  return AVERROR_EOF;
3754  }
3755 
3756  return res;
3757 }
3758 
3760 {
3761  MatroskaDemuxContext *matroska = s->priv_data;
3762  int ret = 0;
3763 
3764  if (matroska->resync_pos == -1) {
3765  // This can only happen if generic seeking has been used.
3766  matroska->resync_pos = avio_tell(s->pb);
3767  }
3768 
3769  while (matroska_deliver_packet(matroska, pkt)) {
3770  if (matroska->done)
3771  return (ret < 0) ? ret : AVERROR_EOF;
3772  if (matroska_parse_cluster(matroska) < 0 && !matroska->done)
3773  ret = matroska_resync(matroska, matroska->resync_pos);
3774  }
3775 
3776  return 0;
3777 }
3778 
3779 static int matroska_read_seek(AVFormatContext *s, int stream_index,
3780  int64_t timestamp, int flags)
3781 {
3782  MatroskaDemuxContext *matroska = s->priv_data;
3783  MatroskaTrack *tracks = NULL;
3784  AVStream *st = s->streams[stream_index];
3785  int i, index;
3786 
3787  /* Parse the CUES now since we need the index data to seek. */
3788  if (matroska->cues_parsing_deferred > 0) {
3789  matroska->cues_parsing_deferred = 0;
3790  matroska_parse_cues(matroska);
3791  }
3792 
3793  if (!st->nb_index_entries)
3794  goto err;
3795  timestamp = FFMAX(timestamp, st->index_entries[0].timestamp);
3796 
3797  if ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3798  matroska_reset_status(matroska, 0, st->index_entries[st->nb_index_entries - 1].pos);
3799  while ((index = av_index_search_timestamp(st, timestamp, flags)) < 0 || index == st->nb_index_entries - 1) {
3800  matroska_clear_queue(matroska);
3801  if (matroska_parse_cluster(matroska) < 0)
3802  break;
3803  }
3804  }
3805 
3806  matroska_clear_queue(matroska);
3807  if (index < 0 || (matroska->cues_parsing_deferred < 0 && index == st->nb_index_entries - 1))
3808  goto err;
3809 
3810  tracks = matroska->tracks.elem;
3811  for (i = 0; i < matroska->tracks.nb_elem; i++) {
3812  tracks[i].audio.pkt_cnt = 0;
3813  tracks[i].audio.sub_packet_cnt = 0;
3814  tracks[i].audio.buf_timecode = AV_NOPTS_VALUE;
3815  tracks[i].end_timecode = 0;
3816  }
3817 
3818  /* We seek to a level 1 element, so set the appropriate status. */
3819  matroska_reset_status(matroska, 0, st->index_entries[index].pos);
3820  if (flags & AVSEEK_FLAG_ANY) {
3821  st->skip_to_keyframe = 0;
3822  matroska->skip_to_timecode = timestamp;
3823  } else {
3824  st->skip_to_keyframe = 1;
3825  matroska->skip_to_timecode = st->index_entries[index].timestamp;
3826  }
3827  matroska->skip_to_keyframe = 1;
3828  matroska->done = 0;
3830  return 0;
3831 err:
3832  // slightly hackish but allows proper fallback to
3833  // the generic seeking code.
3834  matroska_reset_status(matroska, 0, -1);
3835  matroska->resync_pos = -1;
3836  matroska_clear_queue(matroska);
3838  matroska->skip_to_keyframe = 0;
3839  matroska->done = 0;
3840  return -1;
3841 }
3842 
3844 {
3845  MatroskaDemuxContext *matroska = s->priv_data;
3846  MatroskaTrack *tracks = matroska->tracks.elem;
3847  int n;
3848 
3849  matroska_clear_queue(matroska);
3850 
3851  for (n = 0; n < matroska->tracks.nb_elem; n++)
3852  if (tracks[n].type == MATROSKA_TRACK_TYPE_AUDIO)
3853  av_freep(&tracks[n].audio.buf);
3856  return 0;
3858 
3859 typedef struct {
3860  int64_t start_time_ns;
3861  int64_t end_time_ns;
3862  int64_t start_offset;
3863  int64_t end_offset;
3865 
3866 /* This function searches all the Cues and returns the CueDesc corresponding to
3867  * the timestamp ts. Returned CueDesc will be such that start_time_ns <= ts <
3868  * end_time_ns. All 4 fields will be set to -1 if ts >= file's duration.
3869  */
3870 static CueDesc get_cue_desc(AVFormatContext *s, int64_t ts, int64_t cues_start) {
3871  MatroskaDemuxContext *matroska = s->priv_data;
3872  CueDesc cue_desc;
3873  int i;
3874  int nb_index_entries = s->streams[0]->nb_index_entries;
3875  AVIndexEntry *index_entries = s->streams[0]->index_entries;
3876 
3877  if (ts >= (int64_t)(matroska->duration * matroska->time_scale))
3878  return (CueDesc) {-1, -1, -1, -1};
3879  for (i = 1; i < nb_index_entries; i++) {
3880  if (index_entries[i - 1].timestamp * matroska->time_scale <= ts &&
3881  index_entries[i].timestamp * matroska->time_scale > ts) {
3882  break;
3883  }
3884  }
3885  --i;
3886  cue_desc.start_time_ns = index_entries[i].timestamp * matroska->time_scale;
3887  cue_desc.start_offset = index_entries[i].pos - matroska->segment_start;
3888  if (i != nb_index_entries - 1) {
3889  cue_desc.end_time_ns = index_entries[i + 1].timestamp * matroska->time_scale;
3890  cue_desc.end_offset = index_entries[i + 1].pos - matroska->segment_start;
3891  } else {
3892  cue_desc.end_time_ns = matroska->duration * matroska->time_scale;
3893  // FIXME: this needs special handling for files where Cues appear
3894  // before Clusters. the current logic assumes Cues appear after
3895  // Clusters.
3896  cue_desc.end_offset = cues_start - matroska->segment_start;
3897  }
3898  return cue_desc;
3899 }
3900 
3902 {
3903  MatroskaDemuxContext *matroska = s->priv_data;
3904  uint32_t id = matroska->current_id;
3905  int64_t cluster_pos, before_pos;
3906  int index, rv = 1;
3907  if (s->streams[0]->nb_index_entries <= 0) return 0;
3908  // seek to the first cluster using cues.
3909  index = av_index_search_timestamp(s->streams[0], 0, 0);
3910  if (index < 0) return 0;
3911  cluster_pos = s->streams[0]->index_entries[index].pos;
3912  before_pos = avio_tell(s->pb);
3913  while (1) {
3914  uint64_t cluster_id, cluster_length;
3915  int read;
3916  AVPacket *pkt;
3917  avio_seek(s->pb, cluster_pos, SEEK_SET);
3918  // read cluster id and length
3919  read = ebml_read_num(matroska, matroska->ctx->pb, 4, &cluster_id, 1);
3920  if (read < 0 || cluster_id != 0xF43B675) // done with all clusters
3921  break;
3922  read = ebml_read_length(matroska, matroska->ctx->pb, &cluster_length);
3923  if (read < 0)
3924  break;
3925 
3926  matroska_reset_status(matroska, 0, cluster_pos);
3927  matroska_clear_queue(matroska);
3928  if (matroska_parse_cluster(matroska) < 0 ||
3929  !matroska->queue) {
3930  break;
3931  }
3932  pkt = &matroska->queue->pkt;
3933  // 4 + read is the length of the cluster id and the cluster length field.
3934  cluster_pos += 4 + read + cluster_length;
3935  if (!(pkt->flags & AV_PKT_FLAG_KEY)) {
3936  rv = 0;
3937  break;
3938  }
3939  }
3940 
3941  /* Restore the status after matroska_read_header: */
3942  matroska_reset_status(matroska, id, before_pos);
3943 
3944  return rv;
3945 }
3946 
3947 static int buffer_size_after_time_downloaded(int64_t time_ns, double search_sec, int64_t bps,
3948  double min_buffer, double* buffer,
3949  double* sec_to_download, AVFormatContext *s,
3950  int64_t cues_start)
3951 {
3952  double nano_seconds_per_second = 1000000000.0;
3953  double time_sec = time_ns / nano_seconds_per_second;
3954  int rv = 0;
3955  int64_t time_to_search_ns = (int64_t)(search_sec * nano_seconds_per_second);
3956  int64_t end_time_ns = time_ns + time_to_search_ns;
3957  double sec_downloaded = 0.0;
3958  CueDesc desc_curr = get_cue_desc(s, time_ns, cues_start);
3959  if (desc_curr.start_time_ns == -1)
3960  return -1;
3961  *sec_to_download = 0.0;
3962 
3963  // Check for non cue start time.
3964  if (time_ns > desc_curr.start_time_ns) {
3965  int64_t cue_nano = desc_curr.end_time_ns - time_ns;
3966  double percent = (double)(cue_nano) / (desc_curr.end_time_ns - desc_curr.start_time_ns);
3967  double cueBytes = (desc_curr.end_offset - desc_curr.start_offset) * percent;
3968  double timeToDownload = (cueBytes * 8.0) / bps;
3969 
3970  sec_downloaded += (cue_nano / nano_seconds_per_second) - timeToDownload;
3971  *sec_to_download += timeToDownload;
3972 
3973  // Check if the search ends within the first cue.
3974  if (desc_curr.end_time_ns >= end_time_ns) {
3975  double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
3976  double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
3977  sec_downloaded = percent_to_sub * sec_downloaded;
3978  *sec_to_download = percent_to_sub * *sec_to_download;
3979  }
3980 
3981  if ((sec_downloaded + *buffer) <= min_buffer) {
3982  return 1;
3983  }
3984 
3985  // Get the next Cue.
3986  desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
3987  }
3988 
3989  while (desc_curr.start_time_ns != -1) {
3990  int64_t desc_bytes = desc_curr.end_offset - desc_curr.start_offset;
3991  int64_t desc_ns = desc_curr.end_time_ns - desc_curr.start_time_ns;
3992  double desc_sec = desc_ns / nano_seconds_per_second;
3993  double bits = (desc_bytes * 8.0);
3994  double time_to_download = bits / bps;
3995 
3996  sec_downloaded += desc_sec - time_to_download;
3997  *sec_to_download += time_to_download;
3998 
3999  if (desc_curr.end_time_ns >= end_time_ns) {
4000  double desc_end_time_sec = desc_curr.end_time_ns / nano_seconds_per_second;
4001  double percent_to_sub = search_sec / (desc_end_time_sec - time_sec);
4002  sec_downloaded = percent_to_sub * sec_downloaded;
4003  *sec_to_download = percent_to_sub * *sec_to_download;
4004 
4005  if ((sec_downloaded + *buffer) <= min_buffer)
4006  rv = 1;
4007  break;
4008  }
4009 
4010  if ((sec_downloaded + *buffer) <= min_buffer) {
4011  rv = 1;
4012  break;
4013  }
4014 
4015  desc_curr = get_cue_desc(s, desc_curr.end_time_ns, cues_start);
4016  }
4017  *buffer = *buffer + sec_downloaded;
4018  return rv;
4019 }
4020 
4021 /* This function computes the bandwidth of the WebM file with the help of
4022  * buffer_size_after_time_downloaded() function. Both of these functions are
4023  * adapted from WebM Tools project and are adapted to work with FFmpeg's
4024  * Matroska parsing mechanism.
4025  *
4026  * Returns the bandwidth of the file on success; -1 on error.
4027  * */
4028 static int64_t webm_dash_manifest_compute_bandwidth(AVFormatContext *s, int64_t cues_start)
4029 {
4030  MatroskaDemuxContext *matroska = s->priv_data;
4031  AVStream *st = s->streams[0];
4032  double bandwidth = 0.0;
4033  int i;
4034 
4035  for (i = 0; i < st->nb_index_entries; i++) {
4036  int64_t prebuffer_ns = 1000000000;
4037  int64_t time_ns = st->index_entries[i].timestamp * matroska->time_scale;
4038  double nano_seconds_per_second = 1000000000.0;
4039  int64_t prebuffered_ns = time_ns + prebuffer_ns;
4040  double prebuffer_bytes = 0.0;
4041  int64_t temp_prebuffer_ns = prebuffer_ns;
4042  int64_t pre_bytes, pre_ns;
4043  double pre_sec, prebuffer, bits_per_second;
4044  CueDesc desc_beg = get_cue_desc(s, time_ns, cues_start);
4045 
4046  // Start with the first Cue.
4047  CueDesc desc_end = desc_beg;
4048 
4049  // Figure out how much data we have downloaded for the prebuffer. This will
4050  // be used later to adjust the bits per sample to try.
4051  while (desc_end.start_time_ns != -1 && desc_end.end_time_ns < prebuffered_ns) {
4052  // Prebuffered the entire Cue.
4053  prebuffer_bytes += desc_end.end_offset - desc_end.start_offset;
4054  temp_prebuffer_ns -= desc_end.end_time_ns - desc_end.start_time_ns;
4055  desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4056  }
4057  if (desc_end.start_time_ns == -1) {
4058  // The prebuffer is larger than the duration.
4059  if (matroska->duration * matroska->time_scale >= prebuffered_ns)
4060  return -1;
4061  bits_per_second = 0.0;
4062  } else {
4063  // The prebuffer ends in the last Cue. Estimate how much data was
4064  // prebuffered.
4065  pre_bytes = desc_end.end_offset - desc_end.start_offset;
4066  pre_ns = desc_end.end_time_ns - desc_end.start_time_ns;
4067  if (pre_ns <= 0)
4068  return -1;
4069  pre_sec = pre_ns / nano_seconds_per_second;
4070  prebuffer_bytes +=
4071  pre_bytes * ((temp_prebuffer_ns / nano_seconds_per_second) / pre_sec);
4072 
4073  prebuffer = prebuffer_ns / nano_seconds_per_second;
4074 
4075  // Set this to 0.0 in case our prebuffer buffers the entire video.
4076  bits_per_second = 0.0;
4077  do {
4078  int64_t desc_bytes = desc_end.end_offset - desc_beg.start_offset;
4079  int64_t desc_ns = desc_end.end_time_ns - desc_beg.start_time_ns;
4080  double desc_sec, calc_bits_per_second, percent, mod_bits_per_second;
4081  if (desc_bytes <= 0)
4082  return -1;
4083 
4084  desc_sec = desc_ns / nano_seconds_per_second;
4085  calc_bits_per_second = (desc_bytes * 8) / desc_sec;
4086 
4087  // Drop the bps by the percentage of bytes buffered.
4088  percent = (desc_bytes - prebuffer_bytes) / desc_bytes;
4089  mod_bits_per_second = calc_bits_per_second * percent;
4090 
4091  if (prebuffer < desc_sec) {
4092  double search_sec =
4093  (double)(matroska->duration * matroska->time_scale) / nano_seconds_per_second;
4094 
4095  // Add 1 so the bits per second should be a little bit greater than file
4096  // datarate.
4097  int64_t bps = (int64_t)(mod_bits_per_second) + 1;
4098  const double min_buffer = 0.0;
4099  double buffer = prebuffer;
4100  double sec_to_download = 0.0;
4101 
4102  int rv = buffer_size_after_time_downloaded(prebuffered_ns, search_sec, bps,
4103  min_buffer, &buffer, &sec_to_download,
4104  s, cues_start);
4105  if (rv < 0) {
4106  return -1;
4107  } else if (rv == 0) {
4108  bits_per_second = (double)(bps);
4109  break;
4110  }
4111  }
4112 
4113  desc_end = get_cue_desc(s, desc_end.end_time_ns, cues_start);
4114  } while (desc_end.start_time_ns != -1);
4115  }
4116  if (bandwidth < bits_per_second) bandwidth = bits_per_second;
4117  }
4118  return (int64_t)bandwidth;
4119 }
4120 
4121 static int webm_dash_manifest_cues(AVFormatContext *s, int64_t init_range)
4122 {
4123  MatroskaDemuxContext *matroska = s->priv_data;
4124  EbmlList *seekhead_list = &matroska->seekhead;
4125  MatroskaSeekhead *seekhead = seekhead_list->elem;
4126  char *buf;
4127  int64_t cues_start = -1, cues_end = -1, before_pos, bandwidth;
4128  int i;
4129  int end = 0;
4130 
4131  // determine cues start and end positions
4132  for (i = 0; i < seekhead_list->nb_elem; i++)
4133  if (seekhead[i].id == MATROSKA_ID_CUES)
4134  break;
4135 
4136  if (i >= seekhead_list->nb_elem) return -1;
4137 
4138  before_pos = avio_tell(matroska->ctx->pb);
4139  cues_start = seekhead[i].pos + matroska->segment_start;
4140  if (avio_seek(matroska->ctx->pb, cues_start, SEEK_SET) == cues_start) {
4141  // cues_end is computed as cues_start + cues_length + length of the
4142  // Cues element ID (i.e. 4) + EBML length of the Cues element.
4143  // cues_end is inclusive and the above sum is reduced by 1.
4144  uint64_t cues_length, cues_id;
4145  int bytes_read;
4146  bytes_read = ebml_read_num (matroska, matroska->ctx->pb, 4, &cues_id, 1);
4147  if (bytes_read < 0 || cues_id != (MATROSKA_ID_CUES & 0xfffffff))
4148  return bytes_read < 0 ? bytes_read : AVERROR_INVALIDDATA;
4149  bytes_read = ebml_read_length(matroska, matroska->ctx->pb, &cues_length);
4150  if (bytes_read < 0)
4151  return bytes_read;
4152  cues_end = cues_start + 4 + bytes_read + cues_length - 1;
4153  }
4154  avio_seek(matroska->ctx->pb, before_pos, SEEK_SET);
4155  if (cues_start == -1 || cues_end == -1) return -1;
4156 
4157  // parse the cues
4158  matroska_parse_cues(matroska);
4159 
4160  // cues start
4161  av_dict_set_int(&s->streams[0]->metadata, CUES_START, cues_start, 0);
4162 
4163  // cues end
4164  av_dict_set_int(&s->streams[0]->metadata, CUES_END, cues_end, 0);
4165 
4166  // if the file has cues at the start, fix up the init range so that
4167  // it does not include it
4168  if (cues_start <= init_range)
4169  av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, cues_start - 1, 0);
4170 
4171  // bandwidth
4172  bandwidth = webm_dash_manifest_compute_bandwidth(s, cues_start);
4173  if (bandwidth < 0) return -1;
4174  av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH, bandwidth, 0);
4175 
4176  // check if all clusters start with key frames
4178 
4179  // store cue point timestamps as a comma separated list for checking subsegment alignment in
4180  // the muxer. assumes that each timestamp cannot be more than 20 characters long.
4181  buf = av_malloc_array(s->streams[0]->nb_index_entries, 20);
4182  if (!buf) return -1;
4183  strcpy(buf, "");
4184  for (i = 0; i < s->streams[0]->nb_index_entries; i++) {
4185  int ret = snprintf(buf + end, 20,
4186  "%" PRId64"%s", s->streams[0]->index_entries[i].timestamp,
4187  i != s->streams[0]->nb_index_entries - 1 ? "," : "");
4188  if (ret <= 0 || (ret == 20 && i == s->streams[0]->nb_index_entries - 1)) {
4189  av_log(s, AV_LOG_ERROR, "timestamp too long.\n");
4190  av_free(buf);
4191  return AVERROR_INVALIDDATA;
4192  }
4193  end += ret;
4194  }
4195  av_dict_set(&s->streams[0]->metadata, CUE_TIMESTAMPS, buf, 0);
4196  av_free(buf);
4197 
4198  return 0;
4199 }
4200 
4202 {
4203  char *buf;
4204  int ret = matroska_read_header(s);
4205  int64_t init_range;
4206  MatroskaTrack *tracks;
4207  MatroskaDemuxContext *matroska = s->priv_data;
4208  if (ret) {
4209  av_log(s, AV_LOG_ERROR, "Failed to read file headers\n");
4210  return -1;
4211  }
4212  if (!matroska->tracks.nb_elem || !s->nb_streams) {
4213  av_log(s, AV_LOG_ERROR, "No track found\n");
4215  goto fail;
4216  }
4217 
4218  if (!matroska->is_live) {
4219  buf = av_asprintf("%g", matroska->duration);
4220  if (!buf) {
4221  ret = AVERROR(ENOMEM);
4222  goto fail;
4223  }
4224  av_dict_set(&s->streams[0]->metadata, DURATION, buf, 0);
4225  av_free(buf);
4226 
4227  // initialization range
4228  // 5 is the offset of Cluster ID.
4229  init_range = avio_tell(s->pb) - 5;
4230  av_dict_set_int(&s->streams[0]->metadata, INITIALIZATION_RANGE, init_range, 0);
4231  }
4232 
4233  // basename of the file
4234  buf = strrchr(s->url, '/');
4235  av_dict_set(&s->streams[0]->metadata, FILENAME, buf ? ++buf : s->url, 0);
4236 
4237  // track number
4238  tracks = matroska->tracks.elem;
4239  av_dict_set_int(&s->streams[0]->metadata, TRACK_NUMBER, tracks[0].num, 0);
4240 
4241  // parse the cues and populate Cue related fields
4242  if (!matroska->is_live) {
4243  ret = webm_dash_manifest_cues(s, init_range);
4244  if (ret < 0) {
4245  av_log(s, AV_LOG_ERROR, "Error parsing Cues\n");
4246  goto fail;
4247  }
4248  }
4249 
4250  // use the bandwidth from the command line if it was provided
4251  if (matroska->bandwidth > 0) {
4252  av_dict_set_int(&s->streams[0]->metadata, BANDWIDTH,
4253  matroska->bandwidth, 0);
4254  }
4255  return 0;
4256 fail:
4258  return ret;
4259 }
4262 {
4263  return AVERROR_EOF;
4264 }
4265 
4266 #define OFFSET(x) offsetof(MatroskaDemuxContext, x)
4267 static const AVOption options[] = {
4268  { "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 },
4269  { "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 },
4270  { NULL },
4271 };
4272 
4273 static const AVClass webm_dash_class = {
4274  .class_name = "WebM DASH Manifest demuxer",
4275  .item_name = av_default_item_name,
4276  .option = options,
4277  .version = LIBAVUTIL_VERSION_INT,
4278 };
4279 
4281  .name = "matroska,webm",
4282  .long_name = NULL_IF_CONFIG_SMALL("Matroska / WebM"),
4283  .extensions = "mkv,mk3d,mka,mks",
4284  .priv_data_size = sizeof(MatroskaDemuxContext),
4290  .mime_type = "audio/webm,audio/x-matroska,video/webm,video/x-matroska"
4291 };
4292 
4294  .name = "webm_dash_manifest",
4295  .long_name = NULL_IF_CONFIG_SMALL("WebM DASH Manifest"),
4296  .priv_data_size = sizeof(MatroskaDemuxContext),
4300  .priv_class = &webm_dash_class,
4301 };
AVCOL_PRI_RESERVED
@ AVCOL_PRI_RESERVED
Definition: pixfmt.h:447
MatroskaCluster::timecode
uint64_t timecode
Definition: matroskadec.c:323
AVStream::index_entries
AVIndexEntry * index_entries
Only used if the format does not support seeking natively.
Definition: avformat.h:1099
MATROSKA_ID_ENCODINGENCRYPTION
#define MATROSKA_ID_ENCODINGENCRYPTION
Definition: matroska.h:178
MatroskaDemuxContext::segment_start
int64_t segment_start
Definition: matroskadec.c:357
AV_CODEC_ID_PCM_S16LE
@ AV_CODEC_ID_PCM_S16LE
Definition: avcodec.h:463
MATROSKA_ID_TAGTARGETS_ATTACHUID
#define MATROSKA_ID_TAGTARGETS_ATTACHUID
Definition: matroska.h:214
MATROSKA_ID_CHAPCOUNTRY
#define MATROSKA_ID_CHAPCOUNTRY
Definition: matroska.h:257
AVMasteringDisplayMetadata::has_primaries
int has_primaries
Flag indicating whether the display primaries (and white point) are set.
Definition: mastering_display_metadata.h:62
ebml_read_master
static int ebml_read_master(MatroskaDemuxContext *matroska, uint64_t length, int64_t pos)
Definition: matroskadec.c:1004
ebml_read_binary
static int ebml_read_binary(AVIOContext *pb, int length, int64_t pos, EbmlBin *bin)
Definition: matroskadec.c:976
av_packet_unref
void av_packet_unref(AVPacket *pkt)
Wipe the packet.
Definition: avpacket.c:599
MATROSKA_ID_VIDEODISPLAYUNIT
#define MATROSKA_ID_VIDEODISPLAYUNIT
Definition: matroska.h:120
FF_ENABLE_DEPRECATION_WARNINGS
#define FF_ENABLE_DEPRECATION_WARNINGS
Definition: internal.h:85
MATROSKA_ID_CODECPRIVATE
#define MATROSKA_ID_CODECPRIVATE
Definition: matroska.h:89
MATROSKA_ID_TRACKNUMBER
#define MATROSKA_ID_TRACKNUMBER
Definition: matroska.h:78
MatroskaTag
Definition: matroskadec.c:280
ff_packet_list_get
int ff_packet_list_get(AVPacketList **head, AVPacketList **tail, AVPacket *pkt)
Remove the oldest AVPacket in the list and return it.
Definition: utils.c:1563
AVMEDIA_TYPE_SUBTITLE
@ AVMEDIA_TYPE_SUBTITLE
Definition: avutil.h:204
bit_depth
static void bit_depth(AudioStatsContext *s, uint64_t mask, uint64_t imask, AVRational *depth)
Definition: af_astats.c:226
MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL
@ MATROSKA_VIDEO_STEREOMODE_TYPE_ROW_INTERLEAVED_RL
Definition: matroska.h:307
AV_LOG_WARNING
#define AV_LOG_WARNING
Something somehow does not look correct.
Definition: log.h:182
MATROSKA_ID_CHAPSTRING
#define MATROSKA_ID_CHAPSTRING
Definition: matroska.h:255
matroska_parse_wavpack
static int matroska_parse_wavpack(MatroskaTrack *track, uint8_t *src, uint8_t **pdst, int *size)
Definition: matroskadec.c:3212
MATROSKA_ID_ENCODINGSIGHASHALGO
#define MATROSKA_ID_ENCODINGSIGHASHALGO
Definition: matroska.h:183
MatroskaTrack::codec_priv
EbmlBin codec_priv
Definition: matroskadec.c:229
MATROSKA_ID_TAGTARGETS
#define MATROSKA_ID_TAGTARGETS
Definition: matroska.h:209
AVMasteringDisplayMetadata::max_luminance
AVRational max_luminance
Max luminance of mastering display (cd/m^2).
Definition: mastering_display_metadata.h:57
AVSphericalProjection
AVSphericalProjection
Projection of the video surface(s) on a sphere.
Definition: spherical.h:51
AVCodecParameters::extradata
uint8_t * extradata
Extra binary data needed for initializing the decoder, codec-dependent.
Definition: avcodec.h:3971
EbmlSyntax::id
uin