1 /*
2 * MOV demuxer
3 * Copyright (c) 2001 Fabrice Bellard
4 * Copyright (c) 2009 Baptiste Coudurier <baptiste dot coudurier at gmail dot com>
5 *
6 * first version by Francois Revol <revol@free.fr>
7 * seek function by Gael Chardon <gael.dev@4now.net>
8 *
9 * This file is part of FFmpeg.
10 *
11 * FFmpeg is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public
13 * License as published by the Free Software Foundation; either
14 * version 2.1 of the License, or (at your option) any later version.
15 *
16 * FFmpeg is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
20 *
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with FFmpeg; if not, write to the Free Software
23 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
24 */
25
26 #include <inttypes.h>
27 #include <limits.h>
28 #include <stdint.h>
29
30 //#define MOV_EXPORT_ALL_METADATA
31
32 #include "libavutil/attributes.h"
33 #include "libavutil/channel_layout.h"
34 #include "libavutil/display.h"
35 #include "libavutil/intreadwrite.h"
36 #include "libavutil/intfloat.h"
37 #include "libavutil/mathematics.h"
38 #include "libavutil/avstring.h"
39 #include "libavutil/dict.h"
40 #include "libavutil/opt.h"
41 #include "libavutil/timecode.h"
42 #include "libavcodec/ac3tab.h"
43 #include "avformat.h"
44 #include "internal.h"
45 #include "avio_internal.h"
46 #include "riff.h"
47 #include "isom.h"
48 #include "libavcodec/get_bits.h"
49 #include "id3v1.h"
50 #include "mov_chan.h"
51 #include "replaygain.h"
52
53 #if CONFIG_ZLIB
54 #include <zlib.h>
55 #endif
56
57 #include "qtpalette.h"
58
59
60 #undef NDEBUG
61 #include <assert.h>
62
63 /* those functions parse an atom */
64 /* links atom IDs to parse functions */
65 typedef struct MOVParseTableEntry {
66 uint32_t type;
67 int (*parse)(MOVContext *ctx, AVIOContext *pb, MOVAtom atom);
68 } MOVParseTableEntry;
69
70 static int mov_read_default(MOVContext *c, AVIOContext *pb, MOVAtom atom);
71
mov_metadata_track_or_disc_number(MOVContext * c,AVIOContext * pb,unsigned len,const char * key)72 static int mov_metadata_track_or_disc_number(MOVContext *c, AVIOContext *pb,
73 unsigned len, const char *key)
74 {
75 char buf[16];
76
77 short current, total = 0;
78 avio_rb16(pb); // unknown
79 current = avio_rb16(pb);
80 if (len >= 6)
81 total = avio_rb16(pb);
82 if (!total)
83 snprintf(buf, sizeof(buf), "%d", current);
84 else
85 snprintf(buf, sizeof(buf), "%d/%d", current, total);
86 c->fc->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
87 av_dict_set(&c->fc->metadata, key, buf, 0);
88
89 return 0;
90 }
91
mov_metadata_int8_bypass_padding(MOVContext * c,AVIOContext * pb,unsigned len,const char * key)92 static int mov_metadata_int8_bypass_padding(MOVContext *c, AVIOContext *pb,
93 unsigned len, const char *key)
94 {
95 /* bypass padding bytes */
96 avio_r8(pb);
97 avio_r8(pb);
98 avio_r8(pb);
99
100 c->fc->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
101 av_dict_set_int(&c->fc->metadata, key, avio_r8(pb), 0);
102
103 return 0;
104 }
105
mov_metadata_int8_no_padding(MOVContext * c,AVIOContext * pb,unsigned len,const char * key)106 static int mov_metadata_int8_no_padding(MOVContext *c, AVIOContext *pb,
107 unsigned len, const char *key)
108 {
109 c->fc->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
110 av_dict_set_int(&c->fc->metadata, key, avio_r8(pb), 0);
111
112 return 0;
113 }
114
mov_metadata_gnre(MOVContext * c,AVIOContext * pb,unsigned len,const char * key)115 static int mov_metadata_gnre(MOVContext *c, AVIOContext *pb,
116 unsigned len, const char *key)
117 {
118 short genre;
119
120 avio_r8(pb); // unknown
121
122 genre = avio_r8(pb);
123 if (genre < 1 || genre > ID3v1_GENRE_MAX)
124 return 0;
125 c->fc->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
126 av_dict_set(&c->fc->metadata, key, ff_id3v1_genre_str[genre-1], 0);
127
128 return 0;
129 }
130
131 static const uint32_t mac_to_unicode[128] = {
132 0x00C4,0x00C5,0x00C7,0x00C9,0x00D1,0x00D6,0x00DC,0x00E1,
133 0x00E0,0x00E2,0x00E4,0x00E3,0x00E5,0x00E7,0x00E9,0x00E8,
134 0x00EA,0x00EB,0x00ED,0x00EC,0x00EE,0x00EF,0x00F1,0x00F3,
135 0x00F2,0x00F4,0x00F6,0x00F5,0x00FA,0x00F9,0x00FB,0x00FC,
136 0x2020,0x00B0,0x00A2,0x00A3,0x00A7,0x2022,0x00B6,0x00DF,
137 0x00AE,0x00A9,0x2122,0x00B4,0x00A8,0x2260,0x00C6,0x00D8,
138 0x221E,0x00B1,0x2264,0x2265,0x00A5,0x00B5,0x2202,0x2211,
139 0x220F,0x03C0,0x222B,0x00AA,0x00BA,0x03A9,0x00E6,0x00F8,
140 0x00BF,0x00A1,0x00AC,0x221A,0x0192,0x2248,0x2206,0x00AB,
141 0x00BB,0x2026,0x00A0,0x00C0,0x00C3,0x00D5,0x0152,0x0153,
142 0x2013,0x2014,0x201C,0x201D,0x2018,0x2019,0x00F7,0x25CA,
143 0x00FF,0x0178,0x2044,0x20AC,0x2039,0x203A,0xFB01,0xFB02,
144 0x2021,0x00B7,0x201A,0x201E,0x2030,0x00C2,0x00CA,0x00C1,
145 0x00CB,0x00C8,0x00CD,0x00CE,0x00CF,0x00CC,0x00D3,0x00D4,
146 0xF8FF,0x00D2,0x00DA,0x00DB,0x00D9,0x0131,0x02C6,0x02DC,
147 0x00AF,0x02D8,0x02D9,0x02DA,0x00B8,0x02DD,0x02DB,0x02C7,
148 };
149
mov_read_mac_string(MOVContext * c,AVIOContext * pb,int len,char * dst,int dstlen)150 static int mov_read_mac_string(MOVContext *c, AVIOContext *pb, int len,
151 char *dst, int dstlen)
152 {
153 char *p = dst;
154 char *end = dst+dstlen-1;
155 int i;
156
157 for (i = 0; i < len; i++) {
158 uint8_t t, c = avio_r8(pb);
159 if (c < 0x80 && p < end)
160 *p++ = c;
161 else if (p < end)
162 PUT_UTF8(mac_to_unicode[c-0x80], t, if (p < end) *p++ = t;);
163 }
164 *p = 0;
165 return p - dst;
166 }
167
mov_read_covr(MOVContext * c,AVIOContext * pb,int type,int len)168 static int mov_read_covr(MOVContext *c, AVIOContext *pb, int type, int len)
169 {
170 AVPacket pkt;
171 AVStream *st;
172 MOVStreamContext *sc;
173 enum AVCodecID id;
174 int ret;
175
176 switch (type) {
177 case 0xd: id = AV_CODEC_ID_MJPEG; break;
178 case 0xe: id = AV_CODEC_ID_PNG; break;
179 case 0x1b: id = AV_CODEC_ID_BMP; break;
180 default:
181 av_log(c->fc, AV_LOG_WARNING, "Unknown cover type: 0x%x.\n", type);
182 avio_skip(pb, len);
183 return 0;
184 }
185
186 st = avformat_new_stream(c->fc, NULL);
187 if (!st)
188 return AVERROR(ENOMEM);
189 sc = av_mallocz(sizeof(*sc));
190 if (!sc)
191 return AVERROR(ENOMEM);
192 st->priv_data = sc;
193
194 ret = av_get_packet(pb, &pkt, len);
195 if (ret < 0)
196 return ret;
197
198 st->disposition |= AV_DISPOSITION_ATTACHED_PIC;
199
200 st->attached_pic = pkt;
201 st->attached_pic.stream_index = st->index;
202 st->attached_pic.flags |= AV_PKT_FLAG_KEY;
203
204 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
205 st->codec->codec_id = id;
206
207 return 0;
208 }
209
mov_metadata_raw(MOVContext * c,AVIOContext * pb,unsigned len,const char * key)210 static int mov_metadata_raw(MOVContext *c, AVIOContext *pb,
211 unsigned len, const char *key)
212 {
213 char *value = av_malloc(len + 1);
214 if (!value)
215 return AVERROR(ENOMEM);
216 avio_read(pb, value, len);
217 value[len] = 0;
218 return av_dict_set(&c->fc->metadata, key, value, AV_DICT_DONT_STRDUP_VAL);
219 }
220
mov_metadata_loci(MOVContext * c,AVIOContext * pb,unsigned len)221 static int mov_metadata_loci(MOVContext *c, AVIOContext *pb, unsigned len)
222 {
223 char language[4] = { 0 };
224 char buf[100];
225 uint16_t langcode = 0;
226 double longitude, latitude;
227 const char *key = "location";
228
229 if (len < 4 + 2 + 1 + 1 + 4 + 4 + 4)
230 return AVERROR_INVALIDDATA;
231
232 avio_skip(pb, 4); // version+flags
233 langcode = avio_rb16(pb);
234 ff_mov_lang_to_iso639(langcode, language);
235 len -= 6;
236
237 len -= avio_get_str(pb, len, buf, sizeof(buf)); // place name
238 if (len < 1)
239 return AVERROR_INVALIDDATA;
240 avio_skip(pb, 1); // role
241 len -= 1;
242
243 if (len < 14)
244 return AVERROR_INVALIDDATA;
245 longitude = ((int32_t) avio_rb32(pb)) / (float) (1 << 16);
246 latitude = ((int32_t) avio_rb32(pb)) / (float) (1 << 16);
247
248 // Try to output in the same format as the ?xyz field
249 snprintf(buf, sizeof(buf), "%+08.4f%+09.4f/", latitude, longitude);
250 if (*language && strcmp(language, "und")) {
251 char key2[16];
252 snprintf(key2, sizeof(key2), "%s-%s", key, language);
253 av_dict_set(&c->fc->metadata, key2, buf, 0);
254 }
255 c->fc->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
256 return av_dict_set(&c->fc->metadata, key, buf, 0);
257 }
258
mov_read_udta_string(MOVContext * c,AVIOContext * pb,MOVAtom atom)259 static int mov_read_udta_string(MOVContext *c, AVIOContext *pb, MOVAtom atom)
260 {
261 #ifdef MOV_EXPORT_ALL_METADATA
262 char tmp_key[5];
263 #endif
264 char str[1024], key2[16], language[4] = {0};
265 const char *key = NULL;
266 uint16_t langcode = 0;
267 uint32_t data_type = 0, str_size;
268 int (*parse)(MOVContext*, AVIOContext*, unsigned, const char*) = NULL;
269
270 switch (atom.type) {
271 case MKTAG(0xa9,'n','a','m'): key = "title"; break;
272 case MKTAG(0xa9,'a','u','t'):
273 case MKTAG(0xa9,'A','R','T'): key = "artist"; break;
274 case MKTAG( 'a','A','R','T'): key = "album_artist"; break;
275 case MKTAG(0xa9,'w','r','t'): key = "composer"; break;
276 case MKTAG( 'c','p','i','l'): key = "compilation";
277 parse = mov_metadata_int8_no_padding; break;
278 case MKTAG( 'c','p','r','t'):
279 case MKTAG(0xa9,'c','p','y'): key = "copyright"; break;
280 case MKTAG(0xa9,'g','r','p'): key = "grouping"; break;
281 case MKTAG(0xa9,'l','y','r'): key = "lyrics"; break;
282 case MKTAG(0xa9,'c','m','t'):
283 case MKTAG(0xa9,'i','n','f'): key = "comment"; break;
284 case MKTAG(0xa9,'a','l','b'): key = "album"; break;
285 case MKTAG(0xa9,'d','a','y'): key = "date"; break;
286 case MKTAG(0xa9,'g','e','n'): key = "genre"; break;
287 case MKTAG( 'g','n','r','e'): key = "genre";
288 parse = mov_metadata_gnre; break;
289 case MKTAG(0xa9,'t','o','o'):
290 case MKTAG(0xa9,'s','w','r'): key = "encoder"; break;
291 case MKTAG(0xa9,'e','n','c'): key = "encoder"; break;
292 case MKTAG(0xa9,'m','a','k'): key = "make"; break;
293 case MKTAG(0xa9,'m','o','d'): key = "model"; break;
294 case MKTAG(0xa9,'x','y','z'): key = "location"; break;
295 case MKTAG( 'd','e','s','c'): key = "description";break;
296 case MKTAG( 'l','d','e','s'): key = "synopsis"; break;
297 case MKTAG( 't','v','s','h'): key = "show"; break;
298 case MKTAG( 't','v','e','n'): key = "episode_id";break;
299 case MKTAG( 't','v','n','n'): key = "network"; break;
300 case MKTAG( 't','r','k','n'): key = "track";
301 parse = mov_metadata_track_or_disc_number; break;
302 case MKTAG( 'd','i','s','k'): key = "disc";
303 parse = mov_metadata_track_or_disc_number; break;
304 case MKTAG( 't','v','e','s'): key = "episode_sort";
305 parse = mov_metadata_int8_bypass_padding; break;
306 case MKTAG( 't','v','s','n'): key = "season_number";
307 parse = mov_metadata_int8_bypass_padding; break;
308 case MKTAG( 's','t','i','k'): key = "media_type";
309 parse = mov_metadata_int8_no_padding; break;
310 case MKTAG( 'h','d','v','d'): key = "hd_video";
311 parse = mov_metadata_int8_no_padding; break;
312 case MKTAG( 'p','g','a','p'): key = "gapless_playback";
313 parse = mov_metadata_int8_no_padding; break;
314 case MKTAG( '@','P','R','M'):
315 return mov_metadata_raw(c, pb, atom.size, "premiere_version");
316 case MKTAG( '@','P','R','Q'):
317 return mov_metadata_raw(c, pb, atom.size, "quicktime_version");
318 case MKTAG( 'l','o','c','i'):
319 return mov_metadata_loci(c, pb, atom.size);
320 case MKTAG( 'X','M','P','_'):
321 return mov_metadata_raw(c, pb, atom.size, "xmp");
322 }
323
324 if (c->itunes_metadata && atom.size > 8) {
325 int data_size = avio_rb32(pb);
326 int tag = avio_rl32(pb);
327 if (tag == MKTAG('d','a','t','a')) {
328 data_type = avio_rb32(pb); // type
329 avio_rb32(pb); // unknown
330 str_size = data_size - 16;
331 atom.size -= 16;
332
333 if (atom.type == MKTAG('c', 'o', 'v', 'r')) {
334 int ret = mov_read_covr(c, pb, data_type, str_size);
335 if (ret < 0) {
336 av_log(c->fc, AV_LOG_ERROR, "Error parsing cover art.\n");
337 }
338 return ret;
339 }
340 } else return 0;
341 } else if (atom.size > 4 && key && !c->itunes_metadata) {
342 str_size = avio_rb16(pb); // string length
343 langcode = avio_rb16(pb);
344 ff_mov_lang_to_iso639(langcode, language);
345 atom.size -= 4;
346 } else
347 str_size = atom.size;
348
349 #ifdef MOV_EXPORT_ALL_METADATA
350 if (!key) {
351 snprintf(tmp_key, 5, "%.4s", (char*)&atom.type);
352 key = tmp_key;
353 }
354 #endif
355
356 if (!key)
357 return 0;
358 if (atom.size < 0)
359 return AVERROR_INVALIDDATA;
360
361 str_size = FFMIN3(sizeof(str)-1, str_size, atom.size);
362
363 if (parse)
364 parse(c, pb, str_size, key);
365 else {
366 if (data_type == 3 || (data_type == 0 && (langcode < 0x400 || langcode == 0x7fff))) { // MAC Encoded
367 mov_read_mac_string(c, pb, str_size, str, sizeof(str));
368 } else {
369 int ret = avio_read(pb, str, str_size);
370 if (ret != str_size)
371 return ret < 0 ? ret : AVERROR_INVALIDDATA;
372 str[str_size] = 0;
373 }
374 c->fc->event_flags |= AVFMT_EVENT_FLAG_METADATA_UPDATED;
375 av_dict_set(&c->fc->metadata, key, str, 0);
376 if (*language && strcmp(language, "und")) {
377 snprintf(key2, sizeof(key2), "%s-%s", key, language);
378 av_dict_set(&c->fc->metadata, key2, str, 0);
379 }
380 }
381 av_dlog(c->fc, "lang \"%3s\" ", language);
382 av_dlog(c->fc, "tag \"%s\" value \"%s\" atom \"%.4s\" %d %"PRId64"\n",
383 key, str, (char*)&atom.type, str_size, atom.size);
384
385 return 0;
386 }
387
mov_read_chpl(MOVContext * c,AVIOContext * pb,MOVAtom atom)388 static int mov_read_chpl(MOVContext *c, AVIOContext *pb, MOVAtom atom)
389 {
390 int64_t start;
391 int i, nb_chapters, str_len, version;
392 char str[256+1];
393
394 if ((atom.size -= 5) < 0)
395 return 0;
396
397 version = avio_r8(pb);
398 avio_rb24(pb);
399 if (version)
400 avio_rb32(pb); // ???
401 nb_chapters = avio_r8(pb);
402
403 for (i = 0; i < nb_chapters; i++) {
404
405 if (atom.size < 9)
406 return 0;
407
408 start = avio_rb64(pb);
409 str_len = avio_r8(pb);
410
411 if ((atom.size -= 9+str_len) < 0)
412 return 0;
413
414 avio_read(pb, str, str_len);
415 str[str_len] = 0;
416 avpriv_new_chapter(c->fc, i, (AVRational){1,10000000}, start, AV_NOPTS_VALUE, str);
417 }
418 return 0;
419 }
420
421 #define MIN_DATA_ENTRY_BOX_SIZE 12
mov_read_dref(MOVContext * c,AVIOContext * pb,MOVAtom atom)422 static int mov_read_dref(MOVContext *c, AVIOContext *pb, MOVAtom atom)
423 {
424 AVStream *st;
425 MOVStreamContext *sc;
426 int entries, i, j;
427
428 if (c->fc->nb_streams < 1)
429 return 0;
430 st = c->fc->streams[c->fc->nb_streams-1];
431 sc = st->priv_data;
432
433 avio_rb32(pb); // version + flags
434 entries = avio_rb32(pb);
435 if (entries > (atom.size - 1) / MIN_DATA_ENTRY_BOX_SIZE + 1 ||
436 entries >= UINT_MAX / sizeof(*sc->drefs))
437 return AVERROR_INVALIDDATA;
438 av_free(sc->drefs);
439 sc->drefs_count = 0;
440 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
441 if (!sc->drefs)
442 return AVERROR(ENOMEM);
443 sc->drefs_count = entries;
444
445 for (i = 0; i < sc->drefs_count; i++) {
446 MOVDref *dref = &sc->drefs[i];
447 uint32_t size = avio_rb32(pb);
448 int64_t next = avio_tell(pb) + size - 4;
449
450 if (size < 12)
451 return AVERROR_INVALIDDATA;
452
453 dref->type = avio_rl32(pb);
454 avio_rb32(pb); // version + flags
455 av_dlog(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
456
457 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
458 /* macintosh alias record */
459 uint16_t volume_len, len;
460 int16_t type;
461
462 avio_skip(pb, 10);
463
464 volume_len = avio_r8(pb);
465 volume_len = FFMIN(volume_len, 27);
466 avio_read(pb, dref->volume, 27);
467 dref->volume[volume_len] = 0;
468 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", dref->volume, volume_len);
469
470 avio_skip(pb, 12);
471
472 len = avio_r8(pb);
473 len = FFMIN(len, 63);
474 avio_read(pb, dref->filename, 63);
475 dref->filename[len] = 0;
476 av_log(c->fc, AV_LOG_DEBUG, "filename %s, len %d\n", dref->filename, len);
477
478 avio_skip(pb, 16);
479
480 /* read next level up_from_alias/down_to_target */
481 dref->nlvl_from = avio_rb16(pb);
482 dref->nlvl_to = avio_rb16(pb);
483 av_log(c->fc, AV_LOG_DEBUG, "nlvl from %d, nlvl to %d\n",
484 dref->nlvl_from, dref->nlvl_to);
485
486 avio_skip(pb, 16);
487
488 for (type = 0; type != -1 && avio_tell(pb) < next; ) {
489 if(avio_feof(pb))
490 return AVERROR_EOF;
491 type = avio_rb16(pb);
492 len = avio_rb16(pb);
493 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
494 if (len&1)
495 len += 1;
496 if (type == 2) { // absolute path
497 av_free(dref->path);
498 dref->path = av_mallocz(len+1);
499 if (!dref->path)
500 return AVERROR(ENOMEM);
501 avio_read(pb, dref->path, len);
502 if (len > volume_len && !strncmp(dref->path, dref->volume, volume_len)) {
503 len -= volume_len;
504 memmove(dref->path, dref->path+volume_len, len);
505 dref->path[len] = 0;
506 }
507 for (j = 0; j < len; j++)
508 if (dref->path[j] == ':')
509 dref->path[j] = '/';
510 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
511 } else if (type == 0) { // directory name
512 av_free(dref->dir);
513 dref->dir = av_malloc(len+1);
514 if (!dref->dir)
515 return AVERROR(ENOMEM);
516 if (avio_read(pb, dref->dir, len) != len)
517 return AVERROR_INVALIDDATA;
518 dref->dir[len] = 0;
519 for (j = 0; j < len; j++)
520 if (dref->dir[j] == ':')
521 dref->dir[j] = '/';
522 av_log(c->fc, AV_LOG_DEBUG, "dir %s\n", dref->dir);
523 } else
524 avio_skip(pb, len);
525 }
526 }
527 avio_seek(pb, next, SEEK_SET);
528 }
529 return 0;
530 }
531
mov_read_hdlr(MOVContext * c,AVIOContext * pb,MOVAtom atom)532 static int mov_read_hdlr(MOVContext *c, AVIOContext *pb, MOVAtom atom)
533 {
534 AVStream *st;
535 uint32_t type;
536 uint32_t av_unused ctype;
537 int title_size;
538 char *title_str;
539
540 if (c->fc->nb_streams < 1) // meta before first trak
541 return 0;
542
543 st = c->fc->streams[c->fc->nb_streams-1];
544
545 avio_r8(pb); /* version */
546 avio_rb24(pb); /* flags */
547
548 /* component type */
549 ctype = avio_rl32(pb);
550 type = avio_rl32(pb); /* component subtype */
551
552 av_dlog(c->fc, "ctype= %.4s (0x%08x)\n", (char*)&ctype, ctype);
553 av_dlog(c->fc, "stype= %.4s\n", (char*)&type);
554
555 if (type == MKTAG('v','i','d','e'))
556 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
557 else if (type == MKTAG('s','o','u','n'))
558 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
559 else if (type == MKTAG('m','1','a',' '))
560 st->codec->codec_id = AV_CODEC_ID_MP2;
561 else if ((type == MKTAG('s','u','b','p')) || (type == MKTAG('c','l','c','p')))
562 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
563
564 avio_rb32(pb); /* component manufacture */
565 avio_rb32(pb); /* component flags */
566 avio_rb32(pb); /* component flags mask */
567
568 title_size = atom.size - 24;
569 if (title_size > 0) {
570 title_str = av_malloc(title_size + 1); /* Add null terminator */
571 if (!title_str)
572 return AVERROR(ENOMEM);
573 avio_read(pb, title_str, title_size);
574 title_str[title_size] = 0;
575 if (title_str[0])
576 av_dict_set(&st->metadata, "handler_name", title_str +
577 (!c->isom && title_str[0] == title_size - 1), 0);
578 av_freep(&title_str);
579 }
580
581 return 0;
582 }
583
ff_mov_read_esds(AVFormatContext * fc,AVIOContext * pb)584 int ff_mov_read_esds(AVFormatContext *fc, AVIOContext *pb)
585 {
586 AVStream *st;
587 int tag;
588
589 if (fc->nb_streams < 1)
590 return 0;
591 st = fc->streams[fc->nb_streams-1];
592
593 avio_rb32(pb); /* version + flags */
594 ff_mp4_read_descr(fc, pb, &tag);
595 if (tag == MP4ESDescrTag) {
596 ff_mp4_parse_es_descr(pb, NULL);
597 } else
598 avio_rb16(pb); /* ID */
599
600 ff_mp4_read_descr(fc, pb, &tag);
601 if (tag == MP4DecConfigDescrTag)
602 ff_mp4_read_dec_config_descr(fc, st, pb);
603 return 0;
604 }
605
mov_read_esds(MOVContext * c,AVIOContext * pb,MOVAtom atom)606 static int mov_read_esds(MOVContext *c, AVIOContext *pb, MOVAtom atom)
607 {
608 return ff_mov_read_esds(c->fc, pb);
609 }
610
mov_read_dac3(MOVContext * c,AVIOContext * pb,MOVAtom atom)611 static int mov_read_dac3(MOVContext *c, AVIOContext *pb, MOVAtom atom)
612 {
613 AVStream *st;
614 int ac3info, acmod, lfeon, bsmod;
615
616 if (c->fc->nb_streams < 1)
617 return 0;
618 st = c->fc->streams[c->fc->nb_streams-1];
619
620 ac3info = avio_rb24(pb);
621 bsmod = (ac3info >> 14) & 0x7;
622 acmod = (ac3info >> 11) & 0x7;
623 lfeon = (ac3info >> 10) & 0x1;
624
625 st->codec->channels = ((int[]){2,1,2,3,3,4,4,5})[acmod] + lfeon;
626 st->codec->channel_layout = avpriv_ac3_channel_layout_tab[acmod];
627 if (lfeon)
628 st->codec->channel_layout |= AV_CH_LOW_FREQUENCY;
629 st->codec->audio_service_type = bsmod;
630 if (st->codec->channels > 1 && bsmod == 0x7)
631 st->codec->audio_service_type = AV_AUDIO_SERVICE_TYPE_KARAOKE;
632
633 return 0;
634 }
635
mov_read_dec3(MOVContext * c,AVIOContext * pb,MOVAtom atom)636 static int mov_read_dec3(MOVContext *c, AVIOContext *pb, MOVAtom atom)
637 {
638 AVStream *st;
639 int eac3info, acmod, lfeon, bsmod;
640
641 if (c->fc->nb_streams < 1)
642 return 0;
643 st = c->fc->streams[c->fc->nb_streams-1];
644
645 /* No need to parse fields for additional independent substreams and its
646 * associated dependent substreams since libavcodec's E-AC-3 decoder
647 * does not support them yet. */
648 avio_rb16(pb); /* data_rate and num_ind_sub */
649 eac3info = avio_rb24(pb);
650 bsmod = (eac3info >> 12) & 0x1f;
651 acmod = (eac3info >> 9) & 0x7;
652 lfeon = (eac3info >> 8) & 0x1;
653 st->codec->channel_layout = avpriv_ac3_channel_layout_tab[acmod];
654 if (lfeon)
655 st->codec->channel_layout |= AV_CH_LOW_FREQUENCY;
656 st->codec->channels = av_get_channel_layout_nb_channels(st->codec->channel_layout);
657 st->codec->audio_service_type = bsmod;
658 if (st->codec->channels > 1 && bsmod == 0x7)
659 st->codec->audio_service_type = AV_AUDIO_SERVICE_TYPE_KARAOKE;
660
661 return 0;
662 }
663
mov_read_chan(MOVContext * c,AVIOContext * pb,MOVAtom atom)664 static int mov_read_chan(MOVContext *c, AVIOContext *pb, MOVAtom atom)
665 {
666 AVStream *st;
667
668 if (c->fc->nb_streams < 1)
669 return 0;
670 st = c->fc->streams[c->fc->nb_streams-1];
671
672 if (atom.size < 16)
673 return 0;
674
675 /* skip version and flags */
676 avio_skip(pb, 4);
677
678 ff_mov_read_chan(c->fc, pb, st, atom.size - 4);
679
680 return 0;
681 }
682
mov_read_wfex(MOVContext * c,AVIOContext * pb,MOVAtom atom)683 static int mov_read_wfex(MOVContext *c, AVIOContext *pb, MOVAtom atom)
684 {
685 AVStream *st;
686
687 if (c->fc->nb_streams < 1)
688 return 0;
689 st = c->fc->streams[c->fc->nb_streams-1];
690
691 if (ff_get_wav_header(pb, st->codec, atom.size) < 0) {
692 av_log(c->fc, AV_LOG_WARNING, "get_wav_header failed\n");
693 }
694
695 return 0;
696 }
697
mov_read_pasp(MOVContext * c,AVIOContext * pb,MOVAtom atom)698 static int mov_read_pasp(MOVContext *c, AVIOContext *pb, MOVAtom atom)
699 {
700 const int num = avio_rb32(pb);
701 const int den = avio_rb32(pb);
702 AVStream *st;
703
704 if (c->fc->nb_streams < 1)
705 return 0;
706 st = c->fc->streams[c->fc->nb_streams-1];
707
708 if ((st->sample_aspect_ratio.den != 1 || st->sample_aspect_ratio.num) && // default
709 (den != st->sample_aspect_ratio.den || num != st->sample_aspect_ratio.num)) {
710 av_log(c->fc, AV_LOG_WARNING,
711 "sample aspect ratio already set to %d:%d, ignoring 'pasp' atom (%d:%d)\n",
712 st->sample_aspect_ratio.num, st->sample_aspect_ratio.den,
713 num, den);
714 } else if (den != 0) {
715 st->sample_aspect_ratio.num = num;
716 st->sample_aspect_ratio.den = den;
717 }
718 return 0;
719 }
720
721 /* this atom contains actual media data */
mov_read_mdat(MOVContext * c,AVIOContext * pb,MOVAtom atom)722 static int mov_read_mdat(MOVContext *c, AVIOContext *pb, MOVAtom atom)
723 {
724 if (atom.size == 0) /* wrong one (MP4) */
725 return 0;
726 c->found_mdat=1;
727 return 0; /* now go for moov */
728 }
729
730 /* read major brand, minor version and compatible brands and store them as metadata */
mov_read_ftyp(MOVContext * c,AVIOContext * pb,MOVAtom atom)731 static int mov_read_ftyp(MOVContext *c, AVIOContext *pb, MOVAtom atom)
732 {
733 uint32_t minor_ver;
734 int comp_brand_size;
735 char* comp_brands_str;
736 uint8_t type[5] = {0};
737
738 avio_read(pb, type, 4);
739 if (strcmp(type, "qt "))
740 c->isom = 1;
741 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
742 av_dict_set(&c->fc->metadata, "major_brand", type, 0);
743 minor_ver = avio_rb32(pb); /* minor version */
744 av_dict_set_int(&c->fc->metadata, "minor_version", minor_ver, 0);
745
746 comp_brand_size = atom.size - 8;
747 if (comp_brand_size < 0)
748 return AVERROR_INVALIDDATA;
749 comp_brands_str = av_malloc(comp_brand_size + 1); /* Add null terminator */
750 if (!comp_brands_str)
751 return AVERROR(ENOMEM);
752 avio_read(pb, comp_brands_str, comp_brand_size);
753 comp_brands_str[comp_brand_size] = 0;
754 av_dict_set(&c->fc->metadata, "compatible_brands", comp_brands_str, 0);
755 av_freep(&comp_brands_str);
756
757 return 0;
758 }
759
760 /* this atom should contain all header atoms */
mov_read_moov(MOVContext * c,AVIOContext * pb,MOVAtom atom)761 static int mov_read_moov(MOVContext *c, AVIOContext *pb, MOVAtom atom)
762 {
763 int ret;
764
765 if (c->found_moov) {
766 av_log(c->fc, AV_LOG_WARNING, "Found duplicated MOOV Atom. Skipped it\n");
767 avio_skip(pb, atom.size);
768 return 0;
769 }
770
771 if ((ret = mov_read_default(c, pb, atom)) < 0)
772 return ret;
773 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
774 /* so we don't parse the whole file if over a network */
775 c->found_moov=1;
776 return 0; /* now go for mdat */
777 }
778
mov_read_moof(MOVContext * c,AVIOContext * pb,MOVAtom atom)779 static int mov_read_moof(MOVContext *c, AVIOContext *pb, MOVAtom atom)
780 {
781 c->fragment.moof_offset = c->fragment.implicit_offset = avio_tell(pb) - 8;
782 av_dlog(c->fc, "moof offset %"PRIx64"\n", c->fragment.moof_offset);
783 return mov_read_default(c, pb, atom);
784 }
785
mov_metadata_creation_time(AVDictionary ** metadata,int64_t time)786 static void mov_metadata_creation_time(AVDictionary **metadata, int64_t time)
787 {
788 char buffer[32];
789 if (time) {
790 struct tm *ptm;
791 time_t timet;
792 if(time >= 2082844800)
793 time -= 2082844800; /* seconds between 1904-01-01 and Epoch */
794 timet = time;
795 ptm = gmtime(&timet);
796 if (!ptm) return;
797 strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm);
798 av_dict_set(metadata, "creation_time", buffer, 0);
799 }
800 }
801
mov_read_mdhd(MOVContext * c,AVIOContext * pb,MOVAtom atom)802 static int mov_read_mdhd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
803 {
804 AVStream *st;
805 MOVStreamContext *sc;
806 int version;
807 char language[4] = {0};
808 unsigned lang;
809 int64_t creation_time;
810
811 if (c->fc->nb_streams < 1)
812 return 0;
813 st = c->fc->streams[c->fc->nb_streams-1];
814 sc = st->priv_data;
815
816 if (sc->time_scale) {
817 av_log(c->fc, AV_LOG_ERROR, "Multiple mdhd?\n");
818 return AVERROR_INVALIDDATA;
819 }
820
821 version = avio_r8(pb);
822 if (version > 1) {
823 avpriv_request_sample(c->fc, "Version %d", version);
824 return AVERROR_PATCHWELCOME;
825 }
826 avio_rb24(pb); /* flags */
827 if (version == 1) {
828 creation_time = avio_rb64(pb);
829 avio_rb64(pb);
830 } else {
831 creation_time = avio_rb32(pb);
832 avio_rb32(pb); /* modification time */
833 }
834 mov_metadata_creation_time(&st->metadata, creation_time);
835
836 sc->time_scale = avio_rb32(pb);
837 st->duration = (version == 1) ? avio_rb64(pb) : avio_rb32(pb); /* duration */
838
839 lang = avio_rb16(pb); /* language */
840 if (ff_mov_lang_to_iso639(lang, language))
841 av_dict_set(&st->metadata, "language", language, 0);
842 avio_rb16(pb); /* quality */
843
844 return 0;
845 }
846
mov_read_mvhd(MOVContext * c,AVIOContext * pb,MOVAtom atom)847 static int mov_read_mvhd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
848 {
849 int64_t creation_time;
850 int version = avio_r8(pb); /* version */
851 avio_rb24(pb); /* flags */
852
853 if (version == 1) {
854 creation_time = avio_rb64(pb);
855 avio_rb64(pb);
856 } else {
857 creation_time = avio_rb32(pb);
858 avio_rb32(pb); /* modification time */
859 }
860 mov_metadata_creation_time(&c->fc->metadata, creation_time);
861 c->time_scale = avio_rb32(pb); /* time scale */
862
863 av_dlog(c->fc, "time scale = %i\n", c->time_scale);
864
865 c->duration = (version == 1) ? avio_rb64(pb) : avio_rb32(pb); /* duration */
866 // set the AVCodecContext duration because the duration of individual tracks
867 // may be inaccurate
868 if (c->time_scale > 0 && !c->trex_data)
869 c->fc->duration = av_rescale(c->duration, AV_TIME_BASE, c->time_scale);
870 avio_rb32(pb); /* preferred scale */
871
872 avio_rb16(pb); /* preferred volume */
873
874 avio_skip(pb, 10); /* reserved */
875
876 avio_skip(pb, 36); /* display matrix */
877
878 avio_rb32(pb); /* preview time */
879 avio_rb32(pb); /* preview duration */
880 avio_rb32(pb); /* poster time */
881 avio_rb32(pb); /* selection time */
882 avio_rb32(pb); /* selection duration */
883 avio_rb32(pb); /* current time */
884 avio_rb32(pb); /* next track ID */
885 return 0;
886 }
887
mov_read_enda(MOVContext * c,AVIOContext * pb,MOVAtom atom)888 static int mov_read_enda(MOVContext *c, AVIOContext *pb, MOVAtom atom)
889 {
890 AVStream *st;
891 int little_endian;
892
893 if (c->fc->nb_streams < 1)
894 return 0;
895 st = c->fc->streams[c->fc->nb_streams-1];
896
897 little_endian = avio_rb16(pb) & 0xFF;
898 av_dlog(c->fc, "enda %d\n", little_endian);
899 if (little_endian == 1) {
900 switch (st->codec->codec_id) {
901 case AV_CODEC_ID_PCM_S24BE:
902 st->codec->codec_id = AV_CODEC_ID_PCM_S24LE;
903 break;
904 case AV_CODEC_ID_PCM_S32BE:
905 st->codec->codec_id = AV_CODEC_ID_PCM_S32LE;
906 break;
907 case AV_CODEC_ID_PCM_F32BE:
908 st->codec->codec_id = AV_CODEC_ID_PCM_F32LE;
909 break;
910 case AV_CODEC_ID_PCM_F64BE:
911 st->codec->codec_id = AV_CODEC_ID_PCM_F64LE;
912 break;
913 default:
914 break;
915 }
916 }
917 return 0;
918 }
919
mov_read_fiel(MOVContext * c,AVIOContext * pb,MOVAtom atom)920 static int mov_read_fiel(MOVContext *c, AVIOContext *pb, MOVAtom atom)
921 {
922 AVStream *st;
923 unsigned mov_field_order;
924 enum AVFieldOrder decoded_field_order = AV_FIELD_UNKNOWN;
925
926 if (c->fc->nb_streams < 1) // will happen with jp2 files
927 return 0;
928 st = c->fc->streams[c->fc->nb_streams-1];
929 if (atom.size < 2)
930 return AVERROR_INVALIDDATA;
931 mov_field_order = avio_rb16(pb);
932 if ((mov_field_order & 0xFF00) == 0x0100)
933 decoded_field_order = AV_FIELD_PROGRESSIVE;
934 else if ((mov_field_order & 0xFF00) == 0x0200) {
935 switch (mov_field_order & 0xFF) {
936 case 0x01: decoded_field_order = AV_FIELD_TT;
937 break;
938 case 0x06: decoded_field_order = AV_FIELD_BB;
939 break;
940 case 0x09: decoded_field_order = AV_FIELD_TB;
941 break;
942 case 0x0E: decoded_field_order = AV_FIELD_BT;
943 break;
944 }
945 }
946 if (decoded_field_order == AV_FIELD_UNKNOWN && mov_field_order) {
947 av_log(NULL, AV_LOG_ERROR, "Unknown MOV field order 0x%04x\n", mov_field_order);
948 }
949 st->codec->field_order = decoded_field_order;
950
951 return 0;
952 }
953
954 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
mov_read_extradata(MOVContext * c,AVIOContext * pb,MOVAtom atom,enum AVCodecID codec_id)955 static int mov_read_extradata(MOVContext *c, AVIOContext *pb, MOVAtom atom,
956 enum AVCodecID codec_id)
957 {
958 AVStream *st;
959 uint64_t size;
960 uint8_t *buf;
961 int err;
962
963 if (c->fc->nb_streams < 1) // will happen with jp2 files
964 return 0;
965 st= c->fc->streams[c->fc->nb_streams-1];
966
967 if (st->codec->codec_id != codec_id)
968 return 0; /* unexpected codec_id - don't mess with extradata */
969
970 size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
971 if (size > INT_MAX || (uint64_t)atom.size > INT_MAX)
972 return AVERROR_INVALIDDATA;
973 if ((err = av_reallocp(&st->codec->extradata, size)) < 0) {
974 st->codec->extradata_size = 0;
975 return err;
976 }
977 buf = st->codec->extradata + st->codec->extradata_size;
978 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
979 AV_WB32( buf , atom.size + 8);
980 AV_WL32( buf + 4, atom.type);
981 err = avio_read(pb, buf + 8, atom.size);
982 if (err < 0) {
983 return err;
984 } else if (err < atom.size) {
985 av_log(c->fc, AV_LOG_WARNING, "truncated extradata\n");
986 st->codec->extradata_size -= atom.size - err;
987 }
988 memset(buf + 8 + err, 0, FF_INPUT_BUFFER_PADDING_SIZE);
989 return 0;
990 }
991
992 /* wrapper functions for reading ALAC/AVS/MJPEG/MJPEG2000 extradata atoms only for those codecs */
mov_read_alac(MOVContext * c,AVIOContext * pb,MOVAtom atom)993 static int mov_read_alac(MOVContext *c, AVIOContext *pb, MOVAtom atom)
994 {
995 return mov_read_extradata(c, pb, atom, AV_CODEC_ID_ALAC);
996 }
997
mov_read_avss(MOVContext * c,AVIOContext * pb,MOVAtom atom)998 static int mov_read_avss(MOVContext *c, AVIOContext *pb, MOVAtom atom)
999 {
1000 return mov_read_extradata(c, pb, atom, AV_CODEC_ID_AVS);
1001 }
1002
mov_read_jp2h(MOVContext * c,AVIOContext * pb,MOVAtom atom)1003 static int mov_read_jp2h(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1004 {
1005 return mov_read_extradata(c, pb, atom, AV_CODEC_ID_JPEG2000);
1006 }
1007
mov_read_avid(MOVContext * c,AVIOContext * pb,MOVAtom atom)1008 static int mov_read_avid(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1009 {
1010 int ret = mov_read_extradata(c, pb, atom, AV_CODEC_ID_AVUI);
1011 if(ret == 0)
1012 ret = mov_read_extradata(c, pb, atom, AV_CODEC_ID_DNXHD);
1013 return ret;
1014 }
1015
mov_read_targa_y216(MOVContext * c,AVIOContext * pb,MOVAtom atom)1016 static int mov_read_targa_y216(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1017 {
1018 int ret = mov_read_extradata(c, pb, atom, AV_CODEC_ID_TARGA_Y216);
1019
1020 if (!ret && c->fc->nb_streams >= 1) {
1021 AVCodecContext *avctx = c->fc->streams[c->fc->nb_streams-1]->codec;
1022 if (avctx->extradata_size >= 40) {
1023 avctx->height = AV_RB16(&avctx->extradata[36]);
1024 avctx->width = AV_RB16(&avctx->extradata[38]);
1025 }
1026 }
1027 return ret;
1028 }
1029
mov_read_ares(MOVContext * c,AVIOContext * pb,MOVAtom atom)1030 static int mov_read_ares(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1031 {
1032 if (c->fc->nb_streams >= 1) {
1033 AVCodecContext *codec = c->fc->streams[c->fc->nb_streams-1]->codec;
1034 if (codec->codec_tag == MKTAG('A', 'V', 'i', 'n') &&
1035 codec->codec_id == AV_CODEC_ID_H264 &&
1036 atom.size > 11) {
1037 avio_skip(pb, 10);
1038 /* For AVID AVCI50, force width of 1440 to be able to select the correct SPS and PPS */
1039 if (avio_rb16(pb) == 0xd4d)
1040 codec->width = 1440;
1041 return 0;
1042 }
1043 }
1044
1045 return mov_read_avid(c, pb, atom);
1046 }
1047
mov_read_svq3(MOVContext * c,AVIOContext * pb,MOVAtom atom)1048 static int mov_read_svq3(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1049 {
1050 return mov_read_extradata(c, pb, atom, AV_CODEC_ID_SVQ3);
1051 }
1052
mov_read_wave(MOVContext * c,AVIOContext * pb,MOVAtom atom)1053 static int mov_read_wave(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1054 {
1055 AVStream *st;
1056
1057 if (c->fc->nb_streams < 1)
1058 return 0;
1059 st = c->fc->streams[c->fc->nb_streams-1];
1060
1061 if ((uint64_t)atom.size > (1<<30))
1062 return AVERROR_INVALIDDATA;
1063
1064 if (st->codec->codec_id == AV_CODEC_ID_QDM2 ||
1065 st->codec->codec_id == AV_CODEC_ID_QDMC ||
1066 st->codec->codec_id == AV_CODEC_ID_SPEEX) {
1067 // pass all frma atom to codec, needed at least for QDMC and QDM2
1068 av_free(st->codec->extradata);
1069 if (ff_get_extradata(st->codec, pb, atom.size) < 0)
1070 return AVERROR(ENOMEM);
1071 } else if (atom.size > 8) { /* to read frma, esds atoms */
1072 int ret;
1073 if ((ret = mov_read_default(c, pb, atom)) < 0)
1074 return ret;
1075 } else
1076 avio_skip(pb, atom.size);
1077 return 0;
1078 }
1079
1080 /**
1081 * This function reads atom content and puts data in extradata without tag
1082 * nor size unlike mov_read_extradata.
1083 */
mov_read_glbl(MOVContext * c,AVIOContext * pb,MOVAtom atom)1084 static int mov_read_glbl(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1085 {
1086 AVStream *st;
1087
1088 if (c->fc->nb_streams < 1)
1089 return 0;
1090 st = c->fc->streams[c->fc->nb_streams-1];
1091
1092 if ((uint64_t)atom.size > (1<<30))
1093 return AVERROR_INVALIDDATA;
1094
1095 if (atom.size >= 10) {
1096 // Broken files created by legacy versions of libavformat will
1097 // wrap a whole fiel atom inside of a glbl atom.
1098 unsigned size = avio_rb32(pb);
1099 unsigned type = avio_rl32(pb);
1100 avio_seek(pb, -8, SEEK_CUR);
1101 if (type == MKTAG('f','i','e','l') && size == atom.size)
1102 return mov_read_default(c, pb, atom);
1103 }
1104 av_free(st->codec->extradata);
1105 if (ff_get_extradata(st->codec, pb, atom.size) < 0)
1106 return AVERROR(ENOMEM);
1107
1108 return 0;
1109 }
1110
mov_read_dvc1(MOVContext * c,AVIOContext * pb,MOVAtom atom)1111 static int mov_read_dvc1(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1112 {
1113 AVStream *st;
1114 uint8_t profile_level;
1115 int ret;
1116
1117 if (c->fc->nb_streams < 1)
1118 return 0;
1119 st = c->fc->streams[c->fc->nb_streams-1];
1120
1121 if (atom.size >= (1<<28) || atom.size < 7)
1122 return AVERROR_INVALIDDATA;
1123
1124 profile_level = avio_r8(pb);
1125 if ((profile_level & 0xf0) != 0xc0)
1126 return 0;
1127
1128 avio_seek(pb, 6, SEEK_CUR);
1129 av_free(st->codec->extradata);
1130 if ((ret = ff_get_extradata(st->codec, pb, atom.size - 7)) < 0)
1131 return ret;
1132
1133 return 0;
1134 }
1135
1136 /**
1137 * An strf atom is a BITMAPINFOHEADER struct. This struct is 40 bytes itself,
1138 * but can have extradata appended at the end after the 40 bytes belonging
1139 * to the struct.
1140 */
mov_read_strf(MOVContext * c,AVIOContext * pb,MOVAtom atom)1141 static int mov_read_strf(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1142 {
1143 AVStream *st;
1144
1145 if (c->fc->nb_streams < 1)
1146 return 0;
1147 if (atom.size <= 40)
1148 return 0;
1149 st = c->fc->streams[c->fc->nb_streams-1];
1150
1151 if ((uint64_t)atom.size > (1<<30))
1152 return AVERROR_INVALIDDATA;
1153
1154 avio_skip(pb, 40);
1155 av_free(st->codec->extradata);
1156 if (ff_get_extradata(st->codec, pb, atom.size - 40) < 0)
1157 return AVERROR(ENOMEM);
1158 return 0;
1159 }
1160
mov_read_stco(MOVContext * c,AVIOContext * pb,MOVAtom atom)1161 static int mov_read_stco(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1162 {
1163 AVStream *st;
1164 MOVStreamContext *sc;
1165 unsigned int i, entries;
1166
1167 if (c->fc->nb_streams < 1)
1168 return 0;
1169 st = c->fc->streams[c->fc->nb_streams-1];
1170 sc = st->priv_data;
1171
1172 avio_r8(pb); /* version */
1173 avio_rb24(pb); /* flags */
1174
1175 entries = avio_rb32(pb);
1176
1177 if (!entries)
1178 return 0;
1179 if (entries >= UINT_MAX/sizeof(int64_t))
1180 return AVERROR_INVALIDDATA;
1181
1182 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
1183 if (!sc->chunk_offsets)
1184 return AVERROR(ENOMEM);
1185 sc->chunk_count = entries;
1186
1187 if (atom.type == MKTAG('s','t','c','o'))
1188 for (i = 0; i < entries && !pb->eof_reached; i++)
1189 sc->chunk_offsets[i] = avio_rb32(pb);
1190 else if (atom.type == MKTAG('c','o','6','4'))
1191 for (i = 0; i < entries && !pb->eof_reached; i++)
1192 sc->chunk_offsets[i] = avio_rb64(pb);
1193 else
1194 return AVERROR_INVALIDDATA;
1195
1196 sc->chunk_count = i;
1197
1198 if (pb->eof_reached)
1199 return AVERROR_EOF;
1200
1201 return 0;
1202 }
1203
1204 /**
1205 * Compute codec id for 'lpcm' tag.
1206 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
1207 */
ff_mov_get_lpcm_codec_id(int bps,int flags)1208 enum AVCodecID ff_mov_get_lpcm_codec_id(int bps, int flags)
1209 {
1210 /* lpcm flags:
1211 * 0x1 = float
1212 * 0x2 = big-endian
1213 * 0x4 = signed
1214 */
1215 return ff_get_pcm_codec_id(bps, flags & 1, flags & 2, flags & 4 ? -1 : 0);
1216 }
1217
mov_codec_id(AVStream * st,uint32_t format)1218 static int mov_codec_id(AVStream *st, uint32_t format)
1219 {
1220 int id = ff_codec_get_id(ff_codec_movaudio_tags, format);
1221
1222 if (id <= 0 &&
1223 ((format & 0xFFFF) == 'm' + ('s' << 8) ||
1224 (format & 0xFFFF) == 'T' + ('S' << 8)))
1225 id = ff_codec_get_id(ff_codec_wav_tags, av_bswap32(format) & 0xFFFF);
1226
1227 if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO && id > 0) {
1228 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1229 } else if (st->codec->codec_type != AVMEDIA_TYPE_AUDIO &&
1230 /* skip old asf mpeg4 tag */
1231 format && format != MKTAG('m','p','4','s')) {
1232 id = ff_codec_get_id(ff_codec_movvideo_tags, format);
1233 if (id <= 0)
1234 id = ff_codec_get_id(ff_codec_bmp_tags, format);
1235 if (id > 0)
1236 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1237 else if (st->codec->codec_type == AVMEDIA_TYPE_DATA ||
1238 (st->codec->codec_type == AVMEDIA_TYPE_SUBTITLE &&
1239 st->codec->codec_id == AV_CODEC_ID_NONE)) {
1240 id = ff_codec_get_id(ff_codec_movsubtitle_tags, format);
1241 if (id > 0)
1242 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1243 }
1244 }
1245
1246 st->codec->codec_tag = format;
1247
1248 return id;
1249 }
1250
mov_parse_stsd_video(MOVContext * c,AVIOContext * pb,AVStream * st,MOVStreamContext * sc)1251 static void mov_parse_stsd_video(MOVContext *c, AVIOContext *pb,
1252 AVStream *st, MOVStreamContext *sc)
1253 {
1254 uint8_t codec_name[32];
1255 unsigned int color_depth, len, j;
1256 int color_greyscale;
1257 int color_table_id;
1258
1259 avio_rb16(pb); /* version */
1260 avio_rb16(pb); /* revision level */
1261 avio_rb32(pb); /* vendor */
1262 avio_rb32(pb); /* temporal quality */
1263 avio_rb32(pb); /* spatial quality */
1264
1265 st->codec->width = avio_rb16(pb); /* width */
1266 st->codec->height = avio_rb16(pb); /* height */
1267
1268 avio_rb32(pb); /* horiz resolution */
1269 avio_rb32(pb); /* vert resolution */
1270 avio_rb32(pb); /* data size, always 0 */
1271 avio_rb16(pb); /* frames per samples */
1272
1273 len = avio_r8(pb); /* codec name, pascal string */
1274 if (len > 31)
1275 len = 31;
1276 mov_read_mac_string(c, pb, len, codec_name, sizeof(codec_name));
1277 if (len < 31)
1278 avio_skip(pb, 31 - len);
1279
1280 if (codec_name[0])
1281 av_dict_set(&st->metadata, "encoder", codec_name, 0);
1282
1283 /* codec_tag YV12 triggers an UV swap in rawdec.c */
1284 if (!memcmp(codec_name, "Planar Y'CbCr 8-bit 4:2:0", 25)) {
1285 st->codec->codec_tag = MKTAG('I', '4', '2', '0');
1286 st->codec->width &= ~1;
1287 st->codec->height &= ~1;
1288 }
1289 /* Flash Media Server uses tag H263 with Sorenson Spark */
1290 if (st->codec->codec_tag == MKTAG('H','2','6','3') &&
1291 !memcmp(codec_name, "Sorenson H263", 13))
1292 st->codec->codec_id = AV_CODEC_ID_FLV1;
1293
1294 st->codec->bits_per_coded_sample = avio_rb16(pb); /* depth */
1295 color_table_id = avio_rb16(pb); /* colortable id */
1296 av_dlog(c->fc, "depth %d, ctab id %d\n",
1297 st->codec->bits_per_coded_sample, color_table_id);
1298 /* figure out the palette situation */
1299 color_depth = st->codec->bits_per_coded_sample & 0x1F;
1300 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
1301 /* Do not create a greyscale palette for cinepak */
1302 if (color_greyscale && st->codec->codec_id == AV_CODEC_ID_CINEPAK)
1303 return;
1304
1305 /* if the depth is 2, 4, or 8 bpp, file is palettized */
1306 if ((color_depth == 2) || (color_depth == 4) || (color_depth == 8)) {
1307 /* for palette traversal */
1308 unsigned int color_start, color_count, color_end;
1309 unsigned char a, r, g, b;
1310
1311 if (color_greyscale) {
1312 int color_index, color_dec;
1313 /* compute the greyscale palette */
1314 st->codec->bits_per_coded_sample = color_depth;
1315 color_count = 1 << color_depth;
1316 color_index = 255;
1317 color_dec = 256 / (color_count - 1);
1318 for (j = 0; j < color_count; j++) {
1319 r = g = b = color_index;
1320 sc->palette[j] = (0xFFU << 24) | (r << 16) | (g << 8) | (b);
1321 color_index -= color_dec;
1322 if (color_index < 0)
1323 color_index = 0;
1324 }
1325 } else if (color_table_id) {
1326 const uint8_t *color_table;
1327 /* if flag bit 3 is set, use the default palette */
1328 color_count = 1 << color_depth;
1329 if (color_depth == 2)
1330 color_table = ff_qt_default_palette_4;
1331 else if (color_depth == 4)
1332 color_table = ff_qt_default_palette_16;
1333 else
1334 color_table = ff_qt_default_palette_256;
1335
1336 for (j = 0; j < color_count; j++) {
1337 r = color_table[j * 3 + 0];
1338 g = color_table[j * 3 + 1];
1339 b = color_table[j * 3 + 2];
1340 sc->palette[j] = (0xFFU << 24) | (r << 16) | (g << 8) | (b);
1341 }
1342 } else {
1343 /* load the palette from the file */
1344 color_start = avio_rb32(pb);
1345 color_count = avio_rb16(pb);
1346 color_end = avio_rb16(pb);
1347 if ((color_start <= 255) && (color_end <= 255)) {
1348 for (j = color_start; j <= color_end; j++) {
1349 /* each A, R, G, or B component is 16 bits;
1350 * only use the top 8 bits */
1351 a = avio_r8(pb);
1352 avio_r8(pb);
1353 r = avio_r8(pb);
1354 avio_r8(pb);
1355 g = avio_r8(pb);
1356 avio_r8(pb);
1357 b = avio_r8(pb);
1358 avio_r8(pb);
1359 sc->palette[j] = (a << 24 ) | (r << 16) | (g << 8) | (b);
1360 }
1361 }
1362 }
1363 sc->has_palette = 1;
1364 }
1365 }
1366
mov_parse_stsd_audio(MOVContext * c,AVIOContext * pb,AVStream * st,MOVStreamContext * sc)1367 static void mov_parse_stsd_audio(MOVContext *c, AVIOContext *pb,
1368 AVStream *st, MOVStreamContext *sc)
1369 {
1370 int bits_per_sample, flags;
1371 uint16_t version = avio_rb16(pb);
1372 AVDictionaryEntry *compatible_brands = av_dict_get(c->fc->metadata, "compatible_brands", NULL, AV_DICT_MATCH_CASE);
1373
1374 avio_rb16(pb); /* revision level */
1375 avio_rb32(pb); /* vendor */
1376
1377 st->codec->channels = avio_rb16(pb); /* channel count */
1378 st->codec->bits_per_coded_sample = avio_rb16(pb); /* sample size */
1379 av_dlog(c->fc, "audio channels %d\n", st->codec->channels);
1380
1381 sc->audio_cid = avio_rb16(pb);
1382 avio_rb16(pb); /* packet size = 0 */
1383
1384 st->codec->sample_rate = ((avio_rb32(pb) >> 16));
1385
1386 // Read QT version 1 fields. In version 0 these do not exist.
1387 av_dlog(c->fc, "version =%d, isom =%d\n", version, c->isom);
1388 if (!c->isom ||
1389 (compatible_brands && strstr(compatible_brands->value, "qt "))) {
1390
1391 if (version == 1) {
1392 sc->samples_per_frame = avio_rb32(pb);
1393 avio_rb32(pb); /* bytes per packet */
1394 sc->bytes_per_frame = avio_rb32(pb);
1395 avio_rb32(pb); /* bytes per sample */
1396 } else if (version == 2) {
1397 avio_rb32(pb); /* sizeof struct only */
1398 st->codec->sample_rate = av_int2double(avio_rb64(pb));
1399 st->codec->channels = avio_rb32(pb);
1400 avio_rb32(pb); /* always 0x7F000000 */
1401 st->codec->bits_per_coded_sample = avio_rb32(pb);
1402
1403 flags = avio_rb32(pb); /* lpcm format specific flag */
1404 sc->bytes_per_frame = avio_rb32(pb);
1405 sc->samples_per_frame = avio_rb32(pb);
1406 if (st->codec->codec_tag == MKTAG('l','p','c','m'))
1407 st->codec->codec_id =
1408 ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample,
1409 flags);
1410 }
1411 }
1412
1413 switch (st->codec->codec_id) {
1414 case AV_CODEC_ID_PCM_S8:
1415 case AV_CODEC_ID_PCM_U8:
1416 if (st->codec->bits_per_coded_sample == 16)
1417 st->codec->codec_id = AV_CODEC_ID_PCM_S16BE;
1418 break;
1419 case AV_CODEC_ID_PCM_S16LE:
1420 case AV_CODEC_ID_PCM_S16BE:
1421 if (st->codec->bits_per_coded_sample == 8)
1422 st->codec->codec_id = AV_CODEC_ID_PCM_S8;
1423 else if (st->codec->bits_per_coded_sample == 24)
1424 st->codec->codec_id =
1425 st->codec->codec_id == AV_CODEC_ID_PCM_S16BE ?
1426 AV_CODEC_ID_PCM_S24BE : AV_CODEC_ID_PCM_S24LE;
1427 break;
1428 /* set values for old format before stsd version 1 appeared */
1429 case AV_CODEC_ID_MACE3:
1430 sc->samples_per_frame = 6;
1431 sc->bytes_per_frame = 2 * st->codec->channels;
1432 break;
1433 case AV_CODEC_ID_MACE6:
1434 sc->samples_per_frame = 6;
1435 sc->bytes_per_frame = 1 * st->codec->channels;
1436 break;
1437 case AV_CODEC_ID_ADPCM_IMA_QT:
1438 sc->samples_per_frame = 64;
1439 sc->bytes_per_frame = 34 * st->codec->channels;
1440 break;
1441 case AV_CODEC_ID_GSM:
1442 sc->samples_per_frame = 160;
1443 sc->bytes_per_frame = 33;
1444 break;
1445 default:
1446 break;
1447 }
1448
1449 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
1450 if (bits_per_sample) {
1451 st->codec->bits_per_coded_sample = bits_per_sample;
1452 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
1453 }
1454 }
1455
mov_parse_stsd_subtitle(MOVContext * c,AVIOContext * pb,AVStream * st,MOVStreamContext * sc,int size)1456 static void mov_parse_stsd_subtitle(MOVContext *c, AVIOContext *pb,
1457 AVStream *st, MOVStreamContext *sc,
1458 int size)
1459 {
1460 // ttxt stsd contains display flags, justification, background
1461 // color, fonts, and default styles, so fake an atom to read it
1462 MOVAtom fake_atom = { .size = size };
1463 // mp4s contains a regular esds atom
1464 if (st->codec->codec_tag != AV_RL32("mp4s"))
1465 mov_read_glbl(c, pb, fake_atom);
1466 st->codec->width = sc->width;
1467 st->codec->height = sc->height;
1468 }
1469
yuv_to_rgba(uint32_t ycbcr)1470 static uint32_t yuv_to_rgba(uint32_t ycbcr)
1471 {
1472 uint8_t r, g, b;
1473 int y, cb, cr;
1474
1475 y = (ycbcr >> 16) & 0xFF;
1476 cr = (ycbcr >> 8) & 0xFF;
1477 cb = ycbcr & 0xFF;
1478
1479 b = av_clip_uint8(1.164 * (y - 16) + 2.018 * (cb - 128));
1480 g = av_clip_uint8(1.164 * (y - 16) - 0.813 * (cr - 128) - 0.391 * (cb - 128));
1481 r = av_clip_uint8(1.164 * (y - 16) + 1.596 * (cr - 128));
1482
1483 return (r << 16) | (g << 8) | b;
1484 }
1485
mov_rewrite_dvd_sub_extradata(AVStream * st)1486 static int mov_rewrite_dvd_sub_extradata(AVStream *st)
1487 {
1488 char buf[256] = {0};
1489 uint8_t *src = st->codec->extradata;
1490 int i;
1491
1492 if (st->codec->extradata_size != 64)
1493 return 0;
1494
1495 if (st->codec->width > 0 && st->codec->height > 0)
1496 snprintf(buf, sizeof(buf), "size: %dx%d\n",
1497 st->codec->width, st->codec->height);
1498 av_strlcat(buf, "palette: ", sizeof(buf));
1499
1500 for (i = 0; i < 16; i++) {
1501 uint32_t yuv = AV_RB32(src + i * 4);
1502 uint32_t rgba = yuv_to_rgba(yuv);
1503
1504 av_strlcatf(buf, sizeof(buf), "%06"PRIx32"%s", rgba, i != 15 ? ", " : "");
1505 }
1506
1507 if (av_strlcat(buf, "\n", sizeof(buf)) >= sizeof(buf))
1508 return 0;
1509
1510 av_freep(&st->codec->extradata);
1511 st->codec->extradata_size = 0;
1512 st->codec->extradata = av_mallocz(strlen(buf) + FF_INPUT_BUFFER_PADDING_SIZE);
1513 if (!st->codec->extradata)
1514 return AVERROR(ENOMEM);
1515 st->codec->extradata_size = strlen(buf);
1516 memcpy(st->codec->extradata, buf, st->codec->extradata_size);
1517
1518 return 0;
1519 }
1520
mov_parse_stsd_data(MOVContext * c,AVIOContext * pb,AVStream * st,MOVStreamContext * sc,int size)1521 static int mov_parse_stsd_data(MOVContext *c, AVIOContext *pb,
1522 AVStream *st, MOVStreamContext *sc,
1523 int size)
1524 {
1525 if (st->codec->codec_tag == MKTAG('t','m','c','d')) {
1526 if (ff_get_extradata(st->codec, pb, size) < 0)
1527 return AVERROR(ENOMEM);
1528 if (size > 16) {
1529 MOVStreamContext *tmcd_ctx = st->priv_data;
1530 int val;
1531 val = AV_RB32(st->codec->extradata + 4);
1532 tmcd_ctx->tmcd_flags = val;
1533 if (val & 1)
1534 st->codec->flags2 |= CODEC_FLAG2_DROP_FRAME_TIMECODE;
1535 st->codec->time_base.den = st->codec->extradata[16]; /* number of frame */
1536 st->codec->time_base.num = 1;
1537 if (size > 30) {
1538 uint32_t len = AV_RB32(st->codec->extradata + 18); /* name atom length */
1539 uint32_t format = AV_RB32(st->codec->extradata + 22);
1540 if (format == AV_RB32("name") && (int64_t)size >= (int64_t)len + 18) {
1541 uint16_t str_size = AV_RB16(st->codec->extradata + 26); /* string length */
1542 if (str_size > 0 && size >= (int)str_size + 26) {
1543 char *reel_name = av_malloc(str_size + 1);
1544 if (!reel_name)
1545 return AVERROR(ENOMEM);
1546 memcpy(reel_name, st->codec->extradata + 30, str_size);
1547 reel_name[str_size] = 0; /* Add null terminator */
1548 /* don't add reel_name if emtpy string */
1549 if (*reel_name == 0) {
1550 av_free(reel_name);
1551 } else {
1552 av_dict_set(&st->metadata, "reel_name", reel_name, AV_DICT_DONT_STRDUP_VAL);
1553 }
1554 }
1555 }
1556 }
1557 }
1558 } else {
1559 /* other codec type, just skip (rtp, mp4s ...) */
1560 avio_skip(pb, size);
1561 }
1562 return 0;
1563 }
1564
mov_finalize_stsd_codec(MOVContext * c,AVIOContext * pb,AVStream * st,MOVStreamContext * sc)1565 static int mov_finalize_stsd_codec(MOVContext *c, AVIOContext *pb,
1566 AVStream *st, MOVStreamContext *sc)
1567 {
1568 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO &&
1569 !st->codec->sample_rate && sc->time_scale > 1)
1570 st->codec->sample_rate = sc->time_scale;
1571
1572 /* special codec parameters handling */
1573 switch (st->codec->codec_id) {
1574 #if CONFIG_DV_DEMUXER
1575 case AV_CODEC_ID_DVAUDIO:
1576 c->dv_fctx = avformat_alloc_context();
1577 c->dv_demux = avpriv_dv_init_demux(c->dv_fctx);
1578 if (!c->dv_demux) {
1579 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
1580 return AVERROR(ENOMEM);
1581 }
1582 sc->dv_audio_container = 1;
1583 st->codec->codec_id = AV_CODEC_ID_PCM_S16LE;
1584 break;
1585 #endif
1586 /* no ifdef since parameters are always those */
1587 case AV_CODEC_ID_QCELP:
1588 st->codec->channels = 1;
1589 // force sample rate for qcelp when not stored in mov
1590 if (st->codec->codec_tag != MKTAG('Q','c','l','p'))
1591 st->codec->sample_rate = 8000;
1592 // FIXME: Why is the following needed for some files?
1593 sc->samples_per_frame = 160;
1594 if (!sc->bytes_per_frame)
1595 sc->bytes_per_frame = 35;
1596 break;
1597 case AV_CODEC_ID_AMR_NB:
1598 st->codec->channels = 1;
1599 /* force sample rate for amr, stsd in 3gp does not store sample rate */
1600 st->codec->sample_rate = 8000;
1601 break;
1602 case AV_CODEC_ID_AMR_WB:
1603 st->codec->channels = 1;
1604 st->codec->sample_rate = 16000;
1605 break;
1606 case AV_CODEC_ID_MP2:
1607 case AV_CODEC_ID_MP3:
1608 /* force type after stsd for m1a hdlr */
1609 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1610 st->need_parsing = AVSTREAM_PARSE_FULL;
1611 break;
1612 case AV_CODEC_ID_GSM:
1613 case AV_CODEC_ID_ADPCM_MS:
1614 case AV_CODEC_ID_ADPCM_IMA_WAV:
1615 case AV_CODEC_ID_ILBC:
1616 case AV_CODEC_ID_MACE3:
1617 case AV_CODEC_ID_MACE6:
1618 case AV_CODEC_ID_QDM2:
1619 st->codec->block_align = sc->bytes_per_frame;
1620 break;
1621 case AV_CODEC_ID_ALAC:
1622 if (st->codec->extradata_size == 36) {
1623 st->codec->channels = AV_RB8 (st->codec->extradata + 21);
1624 st->codec->sample_rate = AV_RB32(st->codec->extradata + 32);
1625 }
1626 break;
1627 case AV_CODEC_ID_AC3:
1628 case AV_CODEC_ID_MPEG1VIDEO:
1629 case AV_CODEC_ID_VC1:
1630 st->need_parsing = AVSTREAM_PARSE_FULL;
1631 break;
1632 default:
1633 break;
1634 }
1635 return 0;
1636 }
1637
mov_skip_multiple_stsd(MOVContext * c,AVIOContext * pb,int codec_tag,int format,int64_t size)1638 static int mov_skip_multiple_stsd(MOVContext *c, AVIOContext *pb,
1639 int codec_tag, int format,
1640 int64_t size)
1641 {
1642 int video_codec_id = ff_codec_get_id(ff_codec_movvideo_tags, format);
1643
1644 if (codec_tag &&
1645 (codec_tag != format &&
1646 (c->fc->video_codec_id ? video_codec_id != c->fc->video_codec_id
1647 : codec_tag != MKTAG('j','p','e','g')))) {
1648 /* Multiple fourcc, we skip JPEG. This is not correct, we should
1649 * export it as a separate AVStream but this needs a few changes
1650 * in the MOV demuxer, patch welcome. */
1651
1652 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
1653 avio_skip(pb, size);
1654 return 1;
1655 }
1656 if ( codec_tag == AV_RL32("avc1") ||
1657 codec_tag == AV_RL32("hvc1") ||
1658 codec_tag == AV_RL32("hev1")
1659 )
1660 av_log(c->fc, AV_LOG_WARNING, "Concatenated H.264 or H.265 might not play correctly.\n");
1661
1662 return 0;
1663 }
1664
ff_mov_read_stsd_entries(MOVContext * c,AVIOContext * pb,int entries)1665 int ff_mov_read_stsd_entries(MOVContext *c, AVIOContext *pb, int entries)
1666 {
1667 AVStream *st;
1668 MOVStreamContext *sc;
1669 int pseudo_stream_id;
1670
1671 if (c->fc->nb_streams < 1)
1672 return 0;
1673 st = c->fc->streams[c->fc->nb_streams-1];
1674 sc = st->priv_data;
1675
1676 for (pseudo_stream_id = 0;
1677 pseudo_stream_id < entries && !pb->eof_reached;
1678 pseudo_stream_id++) {
1679 //Parsing Sample description table
1680 enum AVCodecID id;
1681 int ret, dref_id = 1;
1682 MOVAtom a = { AV_RL32("stsd") };
1683 int64_t start_pos = avio_tell(pb);
1684 int64_t size = avio_rb32(pb); /* size */
1685 uint32_t format = avio_rl32(pb); /* data format */
1686
1687 if (size >= 16) {
1688 avio_rb32(pb); /* reserved */
1689 avio_rb16(pb); /* reserved */
1690 dref_id = avio_rb16(pb);
1691 }else if (size <= 7){
1692 av_log(c->fc, AV_LOG_ERROR, "invalid size %"PRId64" in stsd\n", size);
1693 return AVERROR_INVALIDDATA;
1694 }
1695
1696 if (mov_skip_multiple_stsd(c, pb, st->codec->codec_tag, format,
1697 size - (avio_tell(pb) - start_pos)))
1698 continue;
1699
1700 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
1701 sc->dref_id= dref_id;
1702
1703 id = mov_codec_id(st, format);
1704
1705 av_dlog(c->fc, "size=%"PRId64" 4CC= %c%c%c%c codec_type=%d\n", size,
1706 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
1707 (format >> 24) & 0xff, st->codec->codec_type);
1708
1709 if (st->codec->codec_type==AVMEDIA_TYPE_VIDEO) {
1710 st->codec->codec_id = id;
1711 mov_parse_stsd_video(c, pb, st, sc);
1712 } else if (st->codec->codec_type==AVMEDIA_TYPE_AUDIO) {
1713 st->codec->codec_id = id;
1714 mov_parse_stsd_audio(c, pb, st, sc);
1715 } else if (st->codec->codec_type==AVMEDIA_TYPE_SUBTITLE){
1716 st->codec->codec_id = id;
1717 mov_parse_stsd_subtitle(c, pb, st, sc,
1718 size - (avio_tell(pb) - start_pos));
1719 } else {
1720 ret = mov_parse_stsd_data(c, pb, st, sc,
1721 size - (avio_tell(pb) - start_pos));
1722 if (ret < 0)
1723 return ret;
1724 }
1725 /* this will read extra atoms at the end (wave, alac, damr, avcC, hvcC, SMI ...) */
1726 a.size = size - (avio_tell(pb) - start_pos);
1727 if (a.size > 8) {
1728 if ((ret = mov_read_default(c, pb, a)) < 0)
1729 return ret;
1730 } else if (a.size > 0)
1731 avio_skip(pb, a.size);
1732 }
1733
1734 if (pb->eof_reached)
1735 return AVERROR_EOF;
1736
1737 return mov_finalize_stsd_codec(c, pb, st, sc);
1738 }
1739
mov_read_stsd(MOVContext * c,AVIOContext * pb,MOVAtom atom)1740 static int mov_read_stsd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1741 {
1742 int entries;
1743
1744 avio_r8(pb); /* version */
1745 avio_rb24(pb); /* flags */
1746 entries = avio_rb32(pb);
1747
1748 return ff_mov_read_stsd_entries(c, pb, entries);
1749 }
1750
mov_read_stsc(MOVContext * c,AVIOContext * pb,MOVAtom atom)1751 static int mov_read_stsc(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1752 {
1753 AVStream *st;
1754 MOVStreamContext *sc;
1755 unsigned int i, entries;
1756
1757 if (c->fc->nb_streams < 1)
1758 return 0;
1759 st = c->fc->streams[c->fc->nb_streams-1];
1760 sc = st->priv_data;
1761
1762 avio_r8(pb); /* version */
1763 avio_rb24(pb); /* flags */
1764
1765 entries = avio_rb32(pb);
1766
1767 av_dlog(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1768
1769 if (!entries)
1770 return 0;
1771 if (entries >= UINT_MAX / sizeof(*sc->stsc_data))
1772 return AVERROR_INVALIDDATA;
1773 sc->stsc_data = av_malloc(entries * sizeof(*sc->stsc_data));
1774 if (!sc->stsc_data)
1775 return AVERROR(ENOMEM);
1776
1777 for (i = 0; i < entries && !pb->eof_reached; i++) {
1778 sc->stsc_data[i].first = avio_rb32(pb);
1779 sc->stsc_data[i].count = avio_rb32(pb);
1780 sc->stsc_data[i].id = avio_rb32(pb);
1781 }
1782
1783 sc->stsc_count = i;
1784
1785 if (pb->eof_reached)
1786 return AVERROR_EOF;
1787
1788 return 0;
1789 }
1790
mov_read_stps(MOVContext * c,AVIOContext * pb,MOVAtom atom)1791 static int mov_read_stps(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1792 {
1793 AVStream *st;
1794 MOVStreamContext *sc;
1795 unsigned i, entries;
1796
1797 if (c->fc->nb_streams < 1)
1798 return 0;
1799 st = c->fc->streams[c->fc->nb_streams-1];
1800 sc = st->priv_data;
1801
1802 avio_rb32(pb); // version + flags
1803
1804 entries = avio_rb32(pb);
1805 if (entries >= UINT_MAX / sizeof(*sc->stps_data))
1806 return AVERROR_INVALIDDATA;
1807 sc->stps_data = av_malloc(entries * sizeof(*sc->stps_data));
1808 if (!sc->stps_data)
1809 return AVERROR(ENOMEM);
1810
1811 for (i = 0; i < entries && !pb->eof_reached; i++) {
1812 sc->stps_data[i] = avio_rb32(pb);
1813 //av_dlog(c->fc, "stps %d\n", sc->stps_data[i]);
1814 }
1815
1816 sc->stps_count = i;
1817
1818 if (pb->eof_reached)
1819 return AVERROR_EOF;
1820
1821 return 0;
1822 }
1823
mov_read_stss(MOVContext * c,AVIOContext * pb,MOVAtom atom)1824 static int mov_read_stss(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1825 {
1826 AVStream *st;
1827 MOVStreamContext *sc;
1828 unsigned int i, entries;
1829
1830 if (c->fc->nb_streams < 1)
1831 return 0;
1832 st = c->fc->streams[c->fc->nb_streams-1];
1833 sc = st->priv_data;
1834
1835 avio_r8(pb); /* version */
1836 avio_rb24(pb); /* flags */
1837
1838 entries = avio_rb32(pb);
1839
1840 av_dlog(c->fc, "keyframe_count = %d\n", entries);
1841
1842 if (!entries)
1843 {
1844 sc->keyframe_absent = 1;
1845 if (!st->need_parsing && st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
1846 st->need_parsing = AVSTREAM_PARSE_HEADERS;
1847 return 0;
1848 }
1849 if (entries >= UINT_MAX / sizeof(int))
1850 return AVERROR_INVALIDDATA;
1851 sc->keyframes = av_malloc(entries * sizeof(int));
1852 if (!sc->keyframes)
1853 return AVERROR(ENOMEM);
1854
1855 for (i = 0; i < entries && !pb->eof_reached; i++) {
1856 sc->keyframes[i] = avio_rb32(pb);
1857 //av_dlog(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1858 }
1859
1860 sc->keyframe_count = i;
1861
1862 if (pb->eof_reached)
1863 return AVERROR_EOF;
1864
1865 return 0;
1866 }
1867
mov_read_stsz(MOVContext * c,AVIOContext * pb,MOVAtom atom)1868 static int mov_read_stsz(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1869 {
1870 AVStream *st;
1871 MOVStreamContext *sc;
1872 unsigned int i, entries, sample_size, field_size, num_bytes;
1873 GetBitContext gb;
1874 unsigned char* buf;
1875
1876 if (c->fc->nb_streams < 1)
1877 return 0;
1878 st = c->fc->streams[c->fc->nb_streams-1];
1879 sc = st->priv_data;
1880
1881 avio_r8(pb); /* version */
1882 avio_rb24(pb); /* flags */
1883
1884 if (atom.type == MKTAG('s','t','s','z')) {
1885 sample_size = avio_rb32(pb);
1886 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1887 sc->sample_size = sample_size;
1888 sc->stsz_sample_size = sample_size;
1889 field_size = 32;
1890 } else {
1891 sample_size = 0;
1892 avio_rb24(pb); /* reserved */
1893 field_size = avio_r8(pb);
1894 }
1895 entries = avio_rb32(pb);
1896
1897 av_dlog(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, entries);
1898
1899 sc->sample_count = entries;
1900 if (sample_size)
1901 return 0;
1902
1903 if (field_size != 4 && field_size != 8 && field_size != 16 && field_size != 32) {
1904 av_log(c->fc, AV_LOG_ERROR, "Invalid sample field size %d\n", field_size);
1905 return AVERROR_INVALIDDATA;
1906 }
1907
1908 if (!entries)
1909 return 0;
1910 if (entries >= UINT_MAX / sizeof(int) || entries >= (UINT_MAX - 4) / field_size)
1911 return AVERROR_INVALIDDATA;
1912 sc->sample_sizes = av_malloc(entries * sizeof(int));
1913 if (!sc->sample_sizes)
1914 return AVERROR(ENOMEM);
1915
1916 num_bytes = (entries*field_size+4)>>3;
1917
1918 buf = av_malloc(num_bytes+FF_INPUT_BUFFER_PADDING_SIZE);
1919 if (!buf) {
1920 av_freep(&sc->sample_sizes);
1921 return AVERROR(ENOMEM);
1922 }
1923
1924 if (avio_read(pb, buf, num_bytes) < num_bytes) {
1925 av_freep(&sc->sample_sizes);
1926 av_free(buf);
1927 return AVERROR_INVALIDDATA;
1928 }
1929
1930 init_get_bits(&gb, buf, 8*num_bytes);
1931
1932 for (i = 0; i < entries && !pb->eof_reached; i++) {
1933 sc->sample_sizes[i] = get_bits_long(&gb, field_size);
1934 sc->data_size += sc->sample_sizes[i];
1935 }
1936
1937 sc->sample_count = i;
1938
1939 if (pb->eof_reached)
1940 return AVERROR_EOF;
1941
1942 av_free(buf);
1943 return 0;
1944 }
1945
mov_read_stts(MOVContext * c,AVIOContext * pb,MOVAtom atom)1946 static int mov_read_stts(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1947 {
1948 AVStream *st;
1949 MOVStreamContext *sc;
1950 unsigned int i, entries;
1951 int64_t duration=0;
1952 int64_t total_sample_count=0;
1953
1954 if (c->fc->nb_streams < 1)
1955 return 0;
1956 st = c->fc->streams[c->fc->nb_streams-1];
1957 sc = st->priv_data;
1958
1959 avio_r8(pb); /* version */
1960 avio_rb24(pb); /* flags */
1961 entries = avio_rb32(pb);
1962
1963 av_dlog(c->fc, "track[%i].stts.entries = %i\n",
1964 c->fc->nb_streams-1, entries);
1965
1966 if (entries >= UINT_MAX / sizeof(*sc->stts_data))
1967 return -1;
1968
1969 av_free(sc->stts_data);
1970 sc->stts_data = av_malloc(entries * sizeof(*sc->stts_data));
1971 if (!sc->stts_data)
1972 return AVERROR(ENOMEM);
1973
1974 for (i = 0; i < entries && !pb->eof_reached; i++) {
1975 int sample_duration;
1976 int sample_count;
1977
1978 sample_count=avio_rb32(pb);
1979 sample_duration = avio_rb32(pb);
1980
1981 /* sample_duration < 0 is invalid based on the spec */
1982 if (sample_duration < 0) {
1983 av_log(c->fc, AV_LOG_ERROR, "Invalid SampleDelta %d in STTS, at %d st:%d\n",
1984 sample_duration, i, c->fc->nb_streams-1);
1985 sample_duration = 1;
1986 }
1987 if (sample_count < 0) {
1988 av_log(c->fc, AV_LOG_ERROR, "Invalid sample_count=%d\n", sample_count);
1989 return AVERROR_INVALIDDATA;
1990 }
1991 sc->stts_data[i].count= sample_count;
1992 sc->stts_data[i].duration= sample_duration;
1993
1994 av_dlog(c->fc, "sample_count=%d, sample_duration=%d\n",
1995 sample_count, sample_duration);
1996
1997 if ( i+1 == entries
1998 && i
1999 && sample_count == 1
2000 && total_sample_count > 100
2001 && sample_duration/10 > duration / total_sample_count)
2002 sample_duration = duration / total_sample_count;
2003 duration+=(int64_t)sample_duration*sample_count;
2004 total_sample_count+=sample_count;
2005 }
2006
2007 sc->stts_count = i;
2008
2009 sc->duration_for_fps += duration;
2010 sc->nb_frames_for_fps += total_sample_count;
2011
2012 if (pb->eof_reached)
2013 return AVERROR_EOF;
2014
2015 st->nb_frames= total_sample_count;
2016 if (duration)
2017 st->duration= duration;
2018 sc->track_end = duration;
2019 return 0;
2020 }
2021
mov_update_dts_shift(MOVStreamContext * sc,int duration)2022 static void mov_update_dts_shift(MOVStreamContext *sc, int duration)
2023 {
2024 if (duration < 0) {
2025 sc->dts_shift = FFMAX(sc->dts_shift, -duration);
2026 }
2027 }
2028
mov_read_ctts(MOVContext * c,AVIOContext * pb,MOVAtom atom)2029 static int mov_read_ctts(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2030 {
2031 AVStream *st;
2032 MOVStreamContext *sc;
2033 unsigned int i, entries;
2034
2035 if (c->fc->nb_streams < 1)
2036 return 0;
2037 st = c->fc->streams[c->fc->nb_streams-1];
2038 sc = st->priv_data;
2039
2040 avio_r8(pb); /* version */
2041 avio_rb24(pb); /* flags */
2042 entries = avio_rb32(pb);
2043
2044 av_dlog(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
2045
2046 if (!entries)
2047 return 0;
2048 if (entries >= UINT_MAX / sizeof(*sc->ctts_data))
2049 return AVERROR_INVALIDDATA;
2050 sc->ctts_data = av_malloc(entries * sizeof(*sc->ctts_data));
2051 if (!sc->ctts_data)
2052 return AVERROR(ENOMEM);
2053
2054 for (i = 0; i < entries && !pb->eof_reached; i++) {
2055 int count =avio_rb32(pb);
2056 int duration =avio_rb32(pb);
2057
2058 sc->ctts_data[i].count = count;
2059 sc->ctts_data[i].duration= duration;
2060
2061 av_dlog(c->fc, "count=%d, duration=%d\n",
2062 count, duration);
2063
2064 if (FFABS(duration) > (1<<28) && i+2<entries) {
2065 av_log(c->fc, AV_LOG_WARNING, "CTTS invalid\n");
2066 av_freep(&sc->ctts_data);
2067 sc->ctts_count = 0;
2068 return 0;
2069 }
2070
2071 if (i+2<entries)
2072 mov_update_dts_shift(sc, duration);
2073 }
2074
2075 sc->ctts_count = i;
2076
2077 if (pb->eof_reached)
2078 return AVERROR_EOF;
2079
2080 av_dlog(c->fc, "dts shift %d\n", sc->dts_shift);
2081
2082 return 0;
2083 }
2084
mov_read_sbgp(MOVContext * c,AVIOContext * pb,MOVAtom atom)2085 static int mov_read_sbgp(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2086 {
2087 AVStream *st;
2088 MOVStreamContext *sc;
2089 unsigned int i, entries;
2090 uint8_t version;
2091 uint32_t grouping_type;
2092
2093 if (c->fc->nb_streams < 1)
2094 return 0;
2095 st = c->fc->streams[c->fc->nb_streams-1];
2096 sc = st->priv_data;
2097
2098 version = avio_r8(pb); /* version */
2099 avio_rb24(pb); /* flags */
2100 grouping_type = avio_rl32(pb);
2101 if (grouping_type != MKTAG( 'r','a','p',' '))
2102 return 0; /* only support 'rap ' grouping */
2103 if (version == 1)
2104 avio_rb32(pb); /* grouping_type_parameter */
2105
2106 entries = avio_rb32(pb);
2107 if (!entries)
2108 return 0;
2109 if (entries >= UINT_MAX / sizeof(*sc->rap_group))
2110 return AVERROR_INVALIDDATA;
2111 sc->rap_group = av_malloc(entries * sizeof(*sc->rap_group));
2112 if (!sc->rap_group)
2113 return AVERROR(ENOMEM);
2114
2115 for (i = 0; i < entries && !pb->eof_reached; i++) {
2116 sc->rap_group[i].count = avio_rb32(pb); /* sample_count */
2117 sc->rap_group[i].index = avio_rb32(pb); /* group_description_index */
2118 }
2119
2120 sc->rap_group_count = i;
2121
2122 return pb->eof_reached ? AVERROR_EOF : 0;
2123 }
2124
mov_build_index(MOVContext * mov,AVStream * st)2125 static void mov_build_index(MOVContext *mov, AVStream *st)
2126 {
2127 MOVStreamContext *sc = st->priv_data;
2128 int64_t current_offset;
2129 int64_t current_dts = 0;
2130 unsigned int stts_index = 0;
2131 unsigned int stsc_index = 0;
2132 unsigned int stss_index = 0;
2133 unsigned int stps_index = 0;
2134 unsigned int i, j;
2135 uint64_t stream_size = 0;
2136
2137 /* adjust first dts according to edit list */
2138 if ((sc->empty_duration || sc->start_time) && mov->time_scale > 0) {
2139 if (sc->empty_duration)
2140 sc->empty_duration = av_rescale(sc->empty_duration, sc->time_scale, mov->time_scale);
2141 sc->time_offset = sc->start_time - sc->empty_duration;
2142 current_dts = -sc->time_offset;
2143 if (sc->ctts_count>0 && sc->stts_count>0 &&
2144 sc->ctts_data[0].duration / FFMAX(sc->stts_data[0].duration, 1) > 16) {
2145 /* more than 16 frames delay, dts are likely wrong
2146 this happens with files created by iMovie */
2147 sc->wrong_dts = 1;
2148 st->codec->has_b_frames = 1;
2149 }
2150 }
2151
2152 /* only use old uncompressed audio chunk demuxing when stts specifies it */
2153 if (!(st->codec->codec_type == AVMEDIA_TYPE_AUDIO &&
2154 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
2155 unsigned int current_sample = 0;
2156 unsigned int stts_sample = 0;
2157 unsigned int sample_size;
2158 unsigned int distance = 0;
2159 unsigned int rap_group_index = 0;
2160 unsigned int rap_group_sample = 0;
2161 int rap_group_present = sc->rap_group_count && sc->rap_group;
2162 int key_off = (sc->keyframe_count && sc->keyframes[0] > 0) || (sc->stps_count && sc->stps_data[0] > 0);
2163
2164 current_dts -= sc->dts_shift;
2165
2166 if (!sc->sample_count || st->nb_index_entries)
2167 return;
2168 if (sc->sample_count >= UINT_MAX / sizeof(*st->index_entries) - st->nb_index_entries)
2169 return;
2170 if (av_reallocp_array(&st->index_entries,
2171 st->nb_index_entries + sc->sample_count,
2172 sizeof(*st->index_entries)) < 0) {
2173 st->nb_index_entries = 0;
2174 return;
2175 }
2176 st->index_entries_allocated_size = (st->nb_index_entries + sc->sample_count) * sizeof(*st->index_entries);
2177
2178 for (i = 0; i < sc->chunk_count; i++) {
2179 int64_t next_offset = i+1 < sc->chunk_count ? sc->chunk_offsets[i+1] : INT64_MAX;
2180 current_offset = sc->chunk_offsets[i];
2181 while (stsc_index + 1 < sc->stsc_count &&
2182 i + 1 == sc->stsc_data[stsc_index + 1].first)
2183 stsc_index++;
2184
2185 if (next_offset > current_offset && sc->sample_size>0 && sc->sample_size < sc->stsz_sample_size &&
2186 sc->stsc_data[stsc_index].count * (int64_t)sc->stsz_sample_size > next_offset - current_offset) {
2187 av_log(mov->fc, AV_LOG_WARNING, "STSZ sample size %d invalid (too large), ignoring\n", sc->stsz_sample_size);
2188 sc->stsz_sample_size = sc->sample_size;
2189 }
2190 if (sc->stsz_sample_size>0 && sc->stsz_sample_size < sc->sample_size) {
2191 av_log(mov->fc, AV_LOG_WARNING, "STSZ sample size %d invalid (too small), ignoring\n", sc->stsz_sample_size);
2192 sc->stsz_sample_size = sc->sample_size;
2193 }
2194
2195 for (j = 0; j < sc->stsc_data[stsc_index].count; j++) {
2196 int keyframe = 0;
2197 if (current_sample >= sc->sample_count) {
2198 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
2199 return;
2200 }
2201
2202 if (!sc->keyframe_absent && (!sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index])) {
2203 keyframe = 1;
2204 if (stss_index + 1 < sc->keyframe_count)
2205 stss_index++;
2206 } else if (sc->stps_count && current_sample+key_off == sc->stps_data[stps_index]) {
2207 keyframe = 1;
2208 if (stps_index + 1 < sc->stps_count)
2209 stps_index++;
2210 }
2211 if (rap_group_present && rap_group_index < sc->rap_group_count) {
2212 if (sc->rap_group[rap_group_index].index > 0)
2213 keyframe = 1;
2214 if (++rap_group_sample == sc->rap_group[rap_group_index].count) {
2215 rap_group_sample = 0;
2216 rap_group_index++;
2217 }
2218 }
2219 if (sc->keyframe_absent
2220 && !sc->stps_count
2221 && !rap_group_present
2222 && (st->codec->codec_type == AVMEDIA_TYPE_AUDIO || (i==0 && j==0)))
2223 keyframe = 1;
2224 if (keyframe)
2225 distance = 0;
2226 sample_size = sc->stsz_sample_size > 0 ? sc->stsz_sample_size : sc->sample_sizes[current_sample];
2227 if (sc->pseudo_stream_id == -1 ||
2228 sc->stsc_data[stsc_index].id - 1 == sc->pseudo_stream_id) {
2229 AVIndexEntry *e = &st->index_entries[st->nb_index_entries++];
2230 e->pos = current_offset;
2231 e->timestamp = current_dts;
2232 e->size = sample_size;
2233 e->min_distance = distance;
2234 e->flags = keyframe ? AVINDEX_KEYFRAME : 0;
2235 av_dlog(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
2236 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
2237 current_offset, current_dts, sample_size, distance, keyframe);
2238 if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO && st->nb_index_entries < 100)
2239 ff_rfps_add_frame(mov->fc, st, current_dts);
2240 }
2241
2242 current_offset += sample_size;
2243 stream_size += sample_size;
2244 current_dts += sc->stts_data[stts_index].duration;
2245 distance++;
2246 stts_sample++;
2247 current_sample++;
2248 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
2249 stts_sample = 0;
2250 stts_index++;
2251 }
2252 }
2253 }
2254 if (st->duration > 0)
2255 st->codec->bit_rate = stream_size*8*sc->time_scale/st->duration;
2256 } else {
2257 unsigned chunk_samples, total = 0;
2258
2259 // compute total chunk count
2260 for (i = 0; i < sc->stsc_count; i++) {
2261 unsigned count, chunk_count;
2262
2263 chunk_samples = sc->stsc_data[i].count;
2264 if (i != sc->stsc_count - 1 &&
2265 sc->samples_per_frame && chunk_samples % sc->samples_per_frame) {
2266 av_log(mov->fc, AV_LOG_ERROR, "error unaligned chunk\n");
2267 return;
2268 }
2269
2270 if (sc->samples_per_frame >= 160) { // gsm
2271 count = chunk_samples / sc->samples_per_frame;
2272 } else if (sc->samples_per_frame > 1) {
2273 unsigned samples = (1024/sc->samples_per_frame)*sc->samples_per_frame;
2274 count = (chunk_samples+samples-1) / samples;
2275 } else {
2276 count = (chunk_samples+1023) / 1024;
2277 }
2278
2279 if (i < sc->stsc_count - 1)
2280 chunk_count = sc->stsc_data[i+1].first - sc->stsc_data[i].first;
2281 else
2282 chunk_count = sc->chunk_count - (sc->stsc_data[i].first - 1);
2283 total += chunk_count * count;
2284 }
2285
2286 av_dlog(mov->fc, "chunk count %d\n", total);
2287 if (total >= UINT_MAX / sizeof(*st->index_entries) - st->nb_index_entries)
2288 return;
2289 if (av_reallocp_array(&st->index_entries,
2290 st->nb_index_entries + total,
2291 sizeof(*st->index_entries)) < 0) {
2292 st->nb_index_entries = 0;
2293 return;
2294 }
2295 st->index_entries_allocated_size = (st->nb_index_entries + total) * sizeof(*st->index_entries);
2296
2297 // populate index
2298 for (i = 0; i < sc->chunk_count; i++) {
2299 current_offset = sc->chunk_offsets[i];
2300 if (stsc_index + 1 < sc->stsc_count &&
2301 i + 1 == sc->stsc_data[stsc_index + 1].first)
2302 stsc_index++;
2303 chunk_samples = sc->stsc_data[stsc_index].count;
2304
2305 while (chunk_samples > 0) {
2306 AVIndexEntry *e;
2307 unsigned size, samples;
2308
2309 if (sc->samples_per_frame >= 160) { // gsm
2310 samples = sc->samples_per_frame;
2311 size = sc->bytes_per_frame;
2312 } else {
2313 if (sc->samples_per_frame > 1) {
2314 samples = FFMIN((1024 / sc->samples_per_frame)*
2315 sc->samples_per_frame, chunk_samples);
2316 size = (samples / sc->samples_per_frame) * sc->bytes_per_frame;
2317 } else {
2318 samples = FFMIN(1024, chunk_samples);
2319 size = samples * sc->sample_size;
2320 }
2321 }
2322
2323 if (st->nb_index_entries >= total) {
2324 av_log(mov->fc, AV_LOG_ERROR, "wrong chunk count %d\n", total);
2325 return;
2326 }
2327 e = &st->index_entries[st->nb_index_entries++];
2328 e->pos = current_offset;
2329 e->timestamp = current_dts;
2330 e->size = size;
2331 e->min_distance = 0;
2332 e->flags = AVINDEX_KEYFRAME;
2333 av_dlog(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
2334 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
2335 size, samples);
2336
2337 current_offset += size;
2338 current_dts += samples;
2339 chunk_samples -= samples;
2340 }
2341 }
2342 }
2343 }
2344
mov_open_dref(AVIOContext ** pb,const char * src,MOVDref * ref,AVIOInterruptCB * int_cb,int use_absolute_path,AVFormatContext * fc)2345 static int mov_open_dref(AVIOContext **pb, const char *src, MOVDref *ref,
2346 AVIOInterruptCB *int_cb, int use_absolute_path, AVFormatContext *fc)
2347 {
2348 /* try relative path, we do not try the absolute because it can leak information about our
2349 system to an attacker */
2350 if (ref->nlvl_to > 0 && ref->nlvl_from > 0) {
2351 char filename[1024];
2352 const char *src_path;
2353 int i, l;
2354
2355 /* find a source dir */
2356 src_path = strrchr(src, '/');
2357 if (src_path)
2358 src_path++;
2359 else
2360 src_path = src;
2361
2362 /* find a next level down to target */
2363 for (i = 0, l = strlen(ref->path) - 1; l >= 0; l--)
2364 if (ref->path[l] == '/') {
2365 if (i == ref->nlvl_to - 1)
2366 break;
2367 else
2368 i++;
2369 }
2370
2371 /* compose filename if next level down to target was found */
2372 if (i == ref->nlvl_to - 1 && src_path - src < sizeof(filename)) {
2373 memcpy(filename, src, src_path - src);
2374 filename[src_path - src] = 0;
2375
2376 for (i = 1; i < ref->nlvl_from; i++)
2377 av_strlcat(filename, "../", 1024);
2378
2379 av_strlcat(filename, ref->path + l + 1, 1024);
2380
2381 if (!avio_open2(pb, filename, AVIO_FLAG_READ, int_cb, NULL))
2382 return 0;
2383 }
2384 } else if (use_absolute_path) {
2385 av_log(fc, AV_LOG_WARNING, "Using absolute path on user request, "
2386 "this is a possible security issue\n");
2387 if (!avio_open2(pb, ref->path, AVIO_FLAG_READ, int_cb, NULL))
2388 return 0;
2389 }
2390
2391 return AVERROR(ENOENT);
2392 }
2393
fix_timescale(MOVContext * c,MOVStreamContext * sc)2394 static void fix_timescale(MOVContext *c, MOVStreamContext *sc)
2395 {
2396 if (sc->time_scale <= 0) {
2397 av_log(c->fc, AV_LOG_WARNING, "stream %d, timescale not set\n", sc->ffindex);
2398 sc->time_scale = c->time_scale;
2399 if (sc->time_scale <= 0)
2400 sc->time_scale = 1;
2401 }
2402 }
2403
mov_read_trak(MOVContext * c,AVIOContext * pb,MOVAtom atom)2404 static int mov_read_trak(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2405 {
2406 AVStream *st;
2407 MOVStreamContext *sc;
2408 int ret;
2409
2410 st = avformat_new_stream(c->fc, NULL);
2411 if (!st) return AVERROR(ENOMEM);
2412 st->id = c->fc->nb_streams;
2413 sc = av_mallocz(sizeof(MOVStreamContext));
2414 if (!sc) return AVERROR(ENOMEM);
2415
2416 st->priv_data = sc;
2417 st->codec->codec_type = AVMEDIA_TYPE_DATA;
2418 sc->ffindex = st->index;
2419
2420 if ((ret = mov_read_default(c, pb, atom)) < 0)
2421 return ret;
2422
2423 /* sanity checks */
2424 if (sc->chunk_count && (!sc->stts_count || !sc->stsc_count ||
2425 (!sc->sample_size && !sc->sample_count))) {
2426 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
2427 st->index);
2428 return 0;
2429 }
2430
2431 fix_timescale(c, sc);
2432
2433 avpriv_set_pts_info(st, 64, 1, sc->time_scale);
2434
2435 mov_build_index(c, st);
2436
2437 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
2438 MOVDref *dref = &sc->drefs[sc->dref_id - 1];
2439 if (mov_open_dref(&sc->pb, c->fc->filename, dref, &c->fc->interrupt_callback,
2440 c->use_absolute_path, c->fc) < 0)
2441 av_log(c->fc, AV_LOG_ERROR,
2442 "stream %d, error opening alias: path='%s', dir='%s', "
2443 "filename='%s', volume='%s', nlvl_from=%d, nlvl_to=%d\n",
2444 st->index, dref->path, dref->dir, dref->filename,
2445 dref->volume, dref->nlvl_from, dref->nlvl_to);
2446 } else {
2447 sc->pb = c->fc->pb;
2448 sc->pb_is_copied = 1;
2449 }
2450
2451 if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
2452 if (!st->sample_aspect_ratio.num &&
2453 (st->codec->width != sc->width || st->codec->height != sc->height)) {
2454 st->sample_aspect_ratio = av_d2q(((double)st->codec->height * sc->width) /
2455 ((double)st->codec->width * sc->height), INT_MAX);
2456 }
2457
2458 #if FF_API_R_FRAME_RATE
2459 if (sc->stts_count == 1 || (sc->stts_count == 2 && sc->stts_data[1].count == 1))
2460 av_reduce(&st->r_frame_rate.num, &st->r_frame_rate.den,
2461 sc->time_scale, sc->stts_data[0].duration, INT_MAX);
2462 #endif
2463 }
2464
2465 // done for ai5q, ai52, ai55, ai1q, ai12 and ai15.
2466 if (!st->codec->extradata_size && st->codec->codec_id == AV_CODEC_ID_H264 &&
2467 TAG_IS_AVCI(st->codec->codec_tag)) {
2468 ret = ff_generate_avci_extradata(st);
2469 if (ret < 0)
2470 return ret;
2471 }
2472
2473 switch (st->codec->codec_id) {
2474 #if CONFIG_H261_DECODER
2475 case AV_CODEC_ID_H261:
2476 #endif
2477 #if CONFIG_H263_DECODER
2478 case AV_CODEC_ID_H263:
2479 #endif
2480 #if CONFIG_MPEG4_DECODER
2481 case AV_CODEC_ID_MPEG4:
2482 #endif
2483 st->codec->width = 0; /* let decoder init width/height */
2484 st->codec->height= 0;
2485 break;
2486 }
2487
2488 /* Do not need those anymore. */
2489 av_freep(&sc->chunk_offsets);
2490 av_freep(&sc->stsc_data);
2491 av_freep(&sc->sample_sizes);
2492 av_freep(&sc->keyframes);
2493 av_freep(&sc->stts_data);
2494 av_freep(&sc->stps_data);
2495 av_freep(&sc->rap_group);
2496
2497 return 0;
2498 }
2499
mov_read_ilst(MOVContext * c,AVIOContext * pb,MOVAtom atom)2500 static int mov_read_ilst(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2501 {
2502 int ret;
2503 c->itunes_metadata = 1;
2504 ret = mov_read_default(c, pb, atom);
2505 c->itunes_metadata = 0;
2506 return ret;
2507 }
2508
mov_read_custom_2plus(MOVContext * c,AVIOContext * pb,int size)2509 static int mov_read_custom_2plus(MOVContext *c, AVIOContext *pb, int size)
2510 {
2511 int64_t end = avio_tell(pb) + size;
2512 uint8_t *key = NULL, *val = NULL;
2513 int i;
2514 AVStream *st;
2515 MOVStreamContext *sc;
2516
2517 if (c->fc->nb_streams < 1)
2518 return 0;
2519 st = c->fc->streams[c->fc->nb_streams-1];
2520 sc = st->priv_data;
2521
2522 for (i = 0; i < 2; i++) {
2523 uint8_t **p;
2524 uint32_t len, tag;
2525
2526 if (end - avio_tell(pb) <= 12)
2527 break;
2528
2529 len = avio_rb32(pb);
2530 tag = avio_rl32(pb);
2531 avio_skip(pb, 4); // flags
2532
2533 if (len < 12 || len - 12 > end - avio_tell(pb))
2534 break;
2535 len -= 12;
2536
2537 if (tag == MKTAG('n', 'a', 'm', 'e'))
2538 p = &key;
2539 else if (tag == MKTAG('d', 'a', 't', 'a') && len > 4) {
2540 avio_skip(pb, 4);
2541 len -= 4;
2542 p = &val;
2543 } else
2544 break;
2545
2546 *p = av_malloc(len + 1);
2547 if (!*p)
2548 break;
2549 avio_read(pb, *p, len);
2550 (*p)[len] = 0;
2551 }
2552
2553 if (key && val) {
2554 if (strcmp(key, "iTunSMPB") == 0) {
2555 int priming, remainder, samples;
2556 if(sscanf(val, "%*X %X %X %X", &priming, &remainder, &samples) == 3){
2557 if(priming>0 && priming<16384)
2558 sc->start_pad = priming;
2559 }
2560 }
2561 if (strcmp(key, "cdec") != 0) {
2562 av_dict_set(&c->fc->metadata, key, val,
2563 AV_DICT_DONT_STRDUP_KEY | AV_DICT_DONT_STRDUP_VAL);
2564 key = val = NULL;
2565 }
2566 }
2567
2568 avio_seek(pb, end, SEEK_SET);
2569 av_freep(&key);
2570 av_freep(&val);
2571 return 0;
2572 }
2573
mov_read_custom(MOVContext * c,AVIOContext * pb,MOVAtom atom)2574 static int mov_read_custom(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2575 {
2576 int64_t end = avio_tell(pb) + atom.size;
2577 uint32_t tag, len;
2578
2579 if (atom.size < 8)
2580 goto fail;
2581
2582 len = avio_rb32(pb);
2583 tag = avio_rl32(pb);
2584
2585 if (len > atom.size)
2586 goto fail;
2587
2588 if (tag == MKTAG('m', 'e', 'a', 'n') && len > 12) {
2589 uint8_t domain[128];
2590 int domain_len;
2591
2592 avio_skip(pb, 4); // flags
2593 len -= 12;
2594
2595 domain_len = avio_get_str(pb, len, domain, sizeof(domain));
2596 avio_skip(pb, len - domain_len);
2597 return mov_read_custom_2plus(c, pb, end - avio_tell(pb));
2598 }
2599
2600 fail:
2601 av_log(c->fc, AV_LOG_VERBOSE,
2602 "Unhandled or malformed custom metadata of size %"PRId64"\n", atom.size);
2603 return 0;
2604 }
2605
mov_read_meta(MOVContext * c,AVIOContext * pb,MOVAtom atom)2606 static int mov_read_meta(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2607 {
2608 while (atom.size > 8) {
2609 uint32_t tag = avio_rl32(pb);
2610 atom.size -= 4;
2611 if (tag == MKTAG('h','d','l','r')) {
2612 avio_seek(pb, -8, SEEK_CUR);
2613 atom.size += 8;
2614 return mov_read_default(c, pb, atom);
2615 }
2616 }
2617 return 0;
2618 }
2619
mov_read_tkhd(MOVContext * c,AVIOContext * pb,MOVAtom atom)2620 static int mov_read_tkhd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2621 {
2622 int i;
2623 int width;
2624 int height;
2625 int64_t disp_transform[2];
2626 int display_matrix[3][3];
2627 AVStream *st;
2628 MOVStreamContext *sc;
2629 int version;
2630 int flags;
2631
2632 if (c->fc->nb_streams < 1)
2633 return 0;
2634 st = c->fc->streams[c->fc->nb_streams-1];
2635 sc = st->priv_data;
2636
2637 version = avio_r8(pb);
2638 flags = avio_rb24(pb);
2639 st->disposition |= (flags & MOV_TKHD_FLAG_ENABLED) ? AV_DISPOSITION_DEFAULT : 0;
2640
2641 if (version == 1) {
2642 avio_rb64(pb);
2643 avio_rb64(pb);
2644 } else {
2645 avio_rb32(pb); /* creation time */
2646 avio_rb32(pb); /* modification time */
2647 }
2648 st->id = (int)avio_rb32(pb); /* track id (NOT 0 !)*/
2649 avio_rb32(pb); /* reserved */
2650
2651 /* highlevel (considering edits) duration in movie timebase */
2652 (version == 1) ? avio_rb64(pb) : avio_rb32(pb);
2653 avio_rb32(pb); /* reserved */
2654 avio_rb32(pb); /* reserved */
2655
2656 avio_rb16(pb); /* layer */
2657 avio_rb16(pb); /* alternate group */
2658 avio_rb16(pb); /* volume */
2659 avio_rb16(pb); /* reserved */
2660
2661 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
2662 // they're kept in fixed point format through all calculations
2663 // save u,v,z to store the whole matrix in the AV_PKT_DATA_DISPLAYMATRIX
2664 // side data, but the scale factor is not needed to calculate aspect ratio
2665 for (i = 0; i < 3; i++) {
2666 display_matrix[i][0] = avio_rb32(pb); // 16.16 fixed point
2667 display_matrix[i][1] = avio_rb32(pb); // 16.16 fixed point
2668 display_matrix[i][2] = avio_rb32(pb); // 2.30 fixed point
2669 }
2670
2671 width = avio_rb32(pb); // 16.16 fixed point track width
2672 height = avio_rb32(pb); // 16.16 fixed point track height
2673 sc->width = width >> 16;
2674 sc->height = height >> 16;
2675
2676 // save the matrix and add rotate metadata when it is not the default
2677 // identity
2678 if (display_matrix[0][0] != (1 << 16) ||
2679 display_matrix[1][1] != (1 << 16) ||
2680 display_matrix[2][2] != (1 << 30) ||
2681 display_matrix[0][1] || display_matrix[0][2] ||
2682 display_matrix[1][0] || display_matrix[1][2] ||
2683 display_matrix[2][0] || display_matrix[2][1]) {
2684 int i, j;
2685 double rotate;
2686
2687 av_freep(&sc->display_matrix);
2688 sc->display_matrix = av_malloc(sizeof(int32_t) * 9);
2689 if (!sc->display_matrix)
2690 return AVERROR(ENOMEM);
2691
2692 for (i = 0; i < 3; i++)
2693 for (j = 0; j < 3; j++)
2694 sc->display_matrix[i * 3 + j] = display_matrix[j][i];
2695
2696 rotate = av_display_rotation_get(sc->display_matrix);
2697 if (!isnan(rotate)) {
2698 char rotate_buf[64];
2699 rotate = -rotate;
2700 if (rotate < 0) // for backward compatibility
2701 rotate += 360;
2702 snprintf(rotate_buf, sizeof(rotate_buf), "%g", rotate);
2703 av_dict_set(&st->metadata, "rotate", rotate_buf, 0);
2704 }
2705 }
2706
2707 // transform the display width/height according to the matrix
2708 // skip this if the display matrix is the default identity matrix
2709 // or if it is rotating the picture, ex iPhone 3GS
2710 // to keep the same scale, use [width height 1<<16]
2711 if (width && height &&
2712 ((display_matrix[0][0] != 65536 ||
2713 display_matrix[1][1] != 65536) &&
2714 !display_matrix[0][1] &&
2715 !display_matrix[1][0] &&
2716 !display_matrix[2][0] && !display_matrix[2][1])) {
2717 for (i = 0; i < 2; i++)
2718 disp_transform[i] =
2719 (int64_t) width * display_matrix[0][i] +
2720 (int64_t) height * display_matrix[1][i] +
2721 ((int64_t) display_matrix[2][i] << 16);
2722
2723 //sample aspect ratio is new width/height divided by old width/height
2724 st->sample_aspect_ratio = av_d2q(
2725 ((double) disp_transform[0] * height) /
2726 ((double) disp_transform[1] * width), INT_MAX);
2727 }
2728 return 0;
2729 }
2730
mov_read_tfhd(MOVContext * c,AVIOContext * pb,MOVAtom atom)2731 static int mov_read_tfhd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2732 {
2733 MOVFragment *frag = &c->fragment;
2734 MOVTrackExt *trex = NULL;
2735 int flags, track_id, i;
2736
2737 avio_r8(pb); /* version */
2738 flags = avio_rb24(pb);
2739
2740 track_id = avio_rb32(pb);
2741 if (!track_id)
2742 return AVERROR_INVALIDDATA;
2743 frag->track_id = track_id;
2744 for (i = 0; i < c->trex_count; i++)
2745 if (c->trex_data[i].track_id == frag->track_id) {
2746 trex = &c->trex_data[i];
2747 break;
2748 }
2749 if (!trex) {
2750 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
2751 return AVERROR_INVALIDDATA;
2752 }
2753
2754 frag->base_data_offset = flags & MOV_TFHD_BASE_DATA_OFFSET ?
2755 avio_rb64(pb) : flags & MOV_TFHD_DEFAULT_BASE_IS_MOOF ?
2756 frag->moof_offset : frag->implicit_offset;
2757 frag->stsd_id = flags & MOV_TFHD_STSD_ID ? avio_rb32(pb) : trex->stsd_id;
2758
2759 frag->duration = flags & MOV_TFHD_DEFAULT_DURATION ?
2760 avio_rb32(pb) : trex->duration;
2761 frag->size = flags & MOV_TFHD_DEFAULT_SIZE ?
2762 avio_rb32(pb) : trex->size;
2763 frag->flags = flags & MOV_TFHD_DEFAULT_FLAGS ?
2764 avio_rb32(pb) : trex->flags;
2765 av_dlog(c->fc, "frag flags 0x%x\n", frag->flags);
2766 return 0;
2767 }
2768
mov_read_chap(MOVContext * c,AVIOContext * pb,MOVAtom atom)2769 static int mov_read_chap(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2770 {
2771 c->chapter_track = avio_rb32(pb);
2772 return 0;
2773 }
2774
mov_read_trex(MOVContext * c,AVIOContext * pb,MOVAtom atom)2775 static int mov_read_trex(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2776 {
2777 MOVTrackExt *trex;
2778 int err;
2779
2780 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
2781 return AVERROR_INVALIDDATA;
2782 if ((err = av_reallocp_array(&c->trex_data, c->trex_count + 1,
2783 sizeof(*c->trex_data))) < 0) {
2784 c->trex_count = 0;
2785 return err;
2786 }
2787
2788 c->fc->duration = AV_NOPTS_VALUE; // the duration from mvhd is not representing the whole file when fragments are used.
2789
2790 trex = &c->trex_data[c->trex_count++];
2791 avio_r8(pb); /* version */
2792 avio_rb24(pb); /* flags */
2793 trex->track_id = avio_rb32(pb);
2794 trex->stsd_id = avio_rb32(pb);
2795 trex->duration = avio_rb32(pb);
2796 trex->size = avio_rb32(pb);
2797 trex->flags = avio_rb32(pb);
2798 return 0;
2799 }
2800
mov_read_trun(MOVContext * c,AVIOContext * pb,MOVAtom atom)2801 static int mov_read_trun(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2802 {
2803 MOVFragment *frag = &c->fragment;
2804 AVStream *st = NULL;
2805 MOVStreamContext *sc;
2806 MOVStts *ctts_data;
2807 uint64_t offset;
2808 int64_t dts;
2809 int data_offset = 0;
2810 unsigned entries, first_sample_flags = frag->flags;
2811 int flags, distance, i, found_keyframe = 0, err;
2812
2813 for (i = 0; i < c->fc->nb_streams; i++) {
2814 if (c->fc->streams[i]->id == frag->track_id) {
2815 st = c->fc->streams[i];
2816 break;
2817 }
2818 }
2819 if (!st) {
2820 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding track id %d\n", frag->track_id);
2821 return AVERROR_INVALIDDATA;
2822 }
2823 sc = st->priv_data;
2824 if (sc->pseudo_stream_id+1 != frag->stsd_id && sc->pseudo_stream_id != -1)
2825 return 0;
2826 avio_r8(pb); /* version */
2827 flags = avio_rb24(pb);
2828 entries = avio_rb32(pb);
2829 av_dlog(c->fc, "flags 0x%x entries %d\n", flags, entries);
2830
2831 /* Always assume the presence of composition time offsets.
2832 * Without this assumption, for instance, we cannot deal with a track in fragmented movies that meet the following.
2833 * 1) in the initial movie, there are no samples.
2834 * 2) in the first movie fragment, there is only one sample without composition time offset.
2835 * 3) in the subsequent movie fragments, there are samples with composition time offset. */
2836 if (!sc->ctts_count && sc->sample_count)
2837 {
2838 /* Complement ctts table if moov atom doesn't have ctts atom. */
2839 ctts_data = av_realloc(NULL, sizeof(*sc->ctts_data));
2840 if (!ctts_data)
2841 return AVERROR(ENOMEM);
2842 sc->ctts_data = ctts_data;
2843 sc->ctts_data[sc->ctts_count].count = sc->sample_count;
2844 sc->ctts_data[sc->ctts_count].duration = 0;
2845 sc->ctts_count++;
2846 }
2847 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
2848 return AVERROR_INVALIDDATA;
2849 if ((err = av_reallocp_array(&sc->ctts_data, entries + sc->ctts_count,
2850 sizeof(*sc->ctts_data))) < 0) {
2851 sc->ctts_count = 0;
2852 return err;
2853 }
2854 if (flags & MOV_TRUN_DATA_OFFSET) data_offset = avio_rb32(pb);
2855 if (flags & MOV_TRUN_FIRST_SAMPLE_FLAGS) first_sample_flags = avio_rb32(pb);
2856 dts = sc->track_end - sc->time_offset;
2857 offset = frag->base_data_offset + data_offset;
2858 distance = 0;
2859 av_dlog(c->fc, "first sample flags 0x%x\n", first_sample_flags);
2860 for (i = 0; i < entries && !pb->eof_reached; i++) {
2861 unsigned sample_size = frag->size;
2862 int sample_flags = i ? frag->flags : first_sample_flags;
2863 unsigned sample_duration = frag->duration;
2864 int keyframe = 0;
2865
2866 if (flags & MOV_TRUN_SAMPLE_DURATION) sample_duration = avio_rb32(pb);
2867 if (flags & MOV_TRUN_SAMPLE_SIZE) sample_size = avio_rb32(pb);
2868 if (flags & MOV_TRUN_SAMPLE_FLAGS) sample_flags = avio_rb32(pb);
2869 sc->ctts_data[sc->ctts_count].count = 1;
2870 sc->ctts_data[sc->ctts_count].duration = (flags & MOV_TRUN_SAMPLE_CTS) ?
2871 avio_rb32(pb) : 0;
2872 mov_update_dts_shift(sc, sc->ctts_data[sc->ctts_count].duration);
2873 sc->ctts_count++;
2874 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
2875 keyframe = 1;
2876 else if (!found_keyframe)
2877 keyframe = found_keyframe =
2878 !(sample_flags & (MOV_FRAG_SAMPLE_FLAG_IS_NON_SYNC |
2879 MOV_FRAG_SAMPLE_FLAG_DEPENDS_YES));
2880 if (keyframe)
2881 distance = 0;
2882 av_add_index_entry(st, offset, dts, sample_size, distance,
2883 keyframe ? AVINDEX_KEYFRAME : 0);
2884 av_dlog(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
2885 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
2886 offset, dts, sample_size, distance, keyframe);
2887 distance++;
2888 dts += sample_duration;
2889 offset += sample_size;
2890 sc->data_size += sample_size;
2891 sc->duration_for_fps += sample_duration;
2892 sc->nb_frames_for_fps ++;
2893 }
2894
2895 if (pb->eof_reached)
2896 return AVERROR_EOF;
2897
2898 frag->implicit_offset = offset;
2899 st->duration = sc->track_end = dts + sc->time_offset;
2900 return 0;
2901 }
2902
2903 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
2904 /* like the files created with Adobe Premiere 5.0, for samples see */
2905 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
mov_read_wide(MOVContext * c,AVIOContext * pb,MOVAtom atom)2906 static int mov_read_wide(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2907 {
2908 int err;
2909
2910 if (atom.size < 8)
2911 return 0; /* continue */
2912 if (avio_rb32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
2913 avio_skip(pb, atom.size - 4);
2914 return 0;
2915 }
2916 atom.type = avio_rl32(pb);
2917 atom.size -= 8;
2918 if (atom.type != MKTAG('m','d','a','t')) {
2919 avio_skip(pb, atom.size);
2920 return 0;
2921 }
2922 err = mov_read_mdat(c, pb, atom);
2923 return err;
2924 }
2925
mov_read_cmov(MOVContext * c,AVIOContext * pb,MOVAtom atom)2926 static int mov_read_cmov(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2927 {
2928 #if CONFIG_ZLIB
2929 AVIOContext ctx;
2930 uint8_t *cmov_data;
2931 uint8_t *moov_data; /* uncompressed data */
2932 long cmov_len, moov_len;
2933 int ret = -1;
2934
2935 avio_rb32(pb); /* dcom atom */
2936 if (avio_rl32(pb) != MKTAG('d','c','o','m'))
2937 return AVERROR_INVALIDDATA;
2938 if (avio_rl32(pb) != MKTAG('z','l','i','b')) {
2939 av_log(c->fc, AV_LOG_ERROR, "unknown compression for cmov atom !\n");
2940 return AVERROR_INVALIDDATA;
2941 }
2942 avio_rb32(pb); /* cmvd atom */
2943 if (avio_rl32(pb) != MKTAG('c','m','v','d'))
2944 return AVERROR_INVALIDDATA;
2945 moov_len = avio_rb32(pb); /* uncompressed size */
2946 cmov_len = atom.size - 6 * 4;
2947
2948 cmov_data = av_malloc(cmov_len);
2949 if (!cmov_data)
2950 return AVERROR(ENOMEM);
2951 moov_data = av_malloc(moov_len);
2952 if (!moov_data) {
2953 av_free(cmov_data);
2954 return AVERROR(ENOMEM);
2955 }
2956 avio_read(pb, cmov_data, cmov_len);
2957 if (uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
2958 goto free_and_return;
2959 if (ffio_init_context(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
2960 goto free_and_return;
2961 atom.type = MKTAG('m','o','o','v');
2962 atom.size = moov_len;
2963 ret = mov_read_default(c, &ctx, atom);
2964 free_and_return:
2965 av_free(moov_data);
2966 av_free(cmov_data);
2967 return ret;
2968 #else
2969 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
2970 return AVERROR(ENOSYS);
2971 #endif
2972 }
2973
2974 /* edit list atom */
mov_read_elst(MOVContext * c,AVIOContext * pb,MOVAtom atom)2975 static int mov_read_elst(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2976 {
2977 MOVStreamContext *sc;
2978 int i, edit_count, version, edit_start_index = 0;
2979 int unsupported = 0;
2980
2981 if (c->fc->nb_streams < 1 || c->ignore_editlist)
2982 return 0;
2983 sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
2984
2985 version = avio_r8(pb); /* version */
2986 avio_rb24(pb); /* flags */
2987 edit_count = avio_rb32(pb); /* entries */
2988
2989 if ((uint64_t)edit_count*12+8 > atom.size)
2990 return AVERROR_INVALIDDATA;
2991
2992 av_dlog(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, edit_count);
2993 for (i=0; i<edit_count; i++){
2994 int64_t time;
2995 int64_t duration;
2996 int rate;
2997 if (version == 1) {
2998 duration = avio_rb64(pb);
2999 time = avio_rb64(pb);
3000 } else {
3001 duration = avio_rb32(pb); /* segment duration */
3002 time = (int32_t)avio_rb32(pb); /* media time */
3003 }
3004 rate = avio_rb32(pb);
3005 if (i == 0 && time == -1) {
3006 sc->empty_duration = duration;
3007 edit_start_index = 1;
3008 } else if (i == edit_start_index && time >= 0)
3009 sc->start_time = time;
3010 else
3011 unsupported = 1;
3012
3013 av_dlog(c->fc, "duration=%"PRId64" time=%"PRId64" rate=%f\n",
3014 duration, time, rate / 65536.0);
3015 }
3016
3017 if (unsupported)
3018 av_log(c->fc, AV_LOG_WARNING, "multiple edit list entries, "
3019 "a/v desync might occur, patch welcome\n");
3020
3021 return 0;
3022 }
3023
mov_read_tmcd(MOVContext * c,AVIOContext * pb,MOVAtom atom)3024 static int mov_read_tmcd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
3025 {
3026 MOVStreamContext *sc;
3027
3028 if (c->fc->nb_streams < 1)
3029 return AVERROR_INVALIDDATA;
3030 sc = c->fc->streams[c->fc->nb_streams - 1]->priv_data;
3031 sc->timecode_track = avio_rb32(pb);
3032 return 0;
3033 }
3034
mov_read_uuid(MOVContext * c,AVIOContext * pb,MOVAtom atom)3035 static int mov_read_uuid(MOVContext *c, AVIOContext *pb, MOVAtom atom)
3036 {
3037 int ret;
3038 uint8_t uuid[16];
3039 static const uint8_t uuid_isml_manifest[] = {
3040 0xa5, 0xd4, 0x0b, 0x30, 0xe8, 0x14, 0x11, 0xdd,
3041 0xba, 0x2f, 0x08, 0x00, 0x20, 0x0c, 0x9a, 0x66
3042 };
3043
3044 if (atom.size < sizeof(uuid) || atom.size == INT64_MAX)
3045 return AVERROR_INVALIDDATA;
3046
3047 ret = avio_read(pb, uuid, sizeof(uuid));
3048 if (ret < 0) {
3049 return ret;
3050 } else if (ret != sizeof(uuid)) {
3051 return AVERROR_INVALIDDATA;
3052 }
3053 if (!memcmp(uuid, uuid_isml_manifest, sizeof(uuid))) {
3054 uint8_t *buffer, *ptr;
3055 char *endptr;
3056 size_t len = atom.size - sizeof(uuid);
3057
3058 if (len < 4) {
3059 return AVERROR_INVALIDDATA;
3060 }
3061 ret = avio_skip(pb, 4); // zeroes
3062 len -= 4;
3063
3064 buffer = av_mallocz(len + 1);
3065 if (!buffer) {
3066 return AVERROR(ENOMEM);
3067 }
3068 ret = avio_read(pb, buffer, len);
3069 if (ret < 0) {
3070 av_free(buffer);
3071 return ret;
3072 } else if (ret != len) {
3073 av_free(buffer);
3074 return AVERROR_INVALIDDATA;
3075 }
3076
3077 ptr = buffer;
3078 while ((ptr = av_stristr(ptr, "systemBitrate=\""))) {
3079 ptr += sizeof("systemBitrate=\"") - 1;
3080 c->bitrates_count++;
3081 c->bitrates = av_realloc_f(c->bitrates, c->bitrates_count, sizeof(*c->bitrates));
3082 if (!c->bitrates) {
3083 c->bitrates_count = 0;
3084 av_free(buffer);
3085 return AVERROR(ENOMEM);
3086 }
3087 errno = 0;
3088 ret = strtol(ptr, &endptr, 10);
3089 if (ret < 0 || errno || *endptr != '"') {
3090 c->bitrates[c->bitrates_count - 1] = 0;
3091 } else {
3092 c->bitrates[c->bitrates_count - 1] = ret;
3093 }
3094 }
3095
3096 av_free(buffer);
3097 }
3098 return 0;
3099 }
3100
3101 static const MOVParseTableEntry mov_default_parse_table[] = {
3102 { MKTAG('A','C','L','R'), mov_read_avid },
3103 { MKTAG('A','P','R','G'), mov_read_avid },
3104 { MKTAG('A','A','L','P'), mov_read_avid },
3105 { MKTAG('A','R','E','S'), mov_read_ares },
3106 { MKTAG('a','v','s','s'), mov_read_avss },
3107 { MKTAG('c','h','p','l'), mov_read_chpl },
3108 { MKTAG('c','o','6','4'), mov_read_stco },
3109 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
3110 { MKTAG('d','i','n','f'), mov_read_default },
3111 { MKTAG('d','r','e','f'), mov_read_dref },
3112 { MKTAG('e','d','t','s'), mov_read_default },
3113 { MKTAG('e','l','s','t'), mov_read_elst },
3114 { MKTAG('e','n','d','a'), mov_read_enda },
3115 { MKTAG('f','i','e','l'), mov_read_fiel },
3116 { MKTAG('f','t','y','p'), mov_read_ftyp },
3117 { MKTAG('g','l','b','l'), mov_read_glbl },
3118 { MKTAG('h','d','l','r'), mov_read_hdlr },
3119 { MKTAG('i','l','s','t'), mov_read_ilst },
3120 { MKTAG('j','p','2','h'), mov_read_jp2h },
3121 { MKTAG('m','d','a','t'), mov_read_mdat },
3122 { MKTAG('m','d','h','d'), mov_read_mdhd },
3123 { MKTAG('m','d','i','a'), mov_read_default },
3124 { MKTAG('m','e','t','a'), mov_read_meta },
3125 { MKTAG('m','i','n','f'), mov_read_default },
3126 { MKTAG('m','o','o','f'), mov_read_moof },
3127 { MKTAG('m','o','o','v'), mov_read_moov },
3128 { MKTAG('m','v','e','x'), mov_read_default },
3129 { MKTAG('m','v','h','d'), mov_read_mvhd },
3130 { MKTAG('S','M','I',' '), mov_read_svq3 },
3131 { MKTAG('a','l','a','c'), mov_read_alac }, /* alac specific atom */
3132 { MKTAG('a','v','c','C'), mov_read_glbl },
3133 { MKTAG('p','a','s','p'), mov_read_pasp },
3134 { MKTAG('s','t','b','l'), mov_read_default },
3135 { MKTAG('s','t','c','o'), mov_read_stco },
3136 { MKTAG('s','t','p','s'), mov_read_stps },
3137 { MKTAG('s','t','r','f'), mov_read_strf },
3138 { MKTAG('s','t','s','c'), mov_read_stsc },
3139 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
3140 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
3141 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
3142 { MKTAG('s','t','t','s'), mov_read_stts },
3143 { MKTAG('s','t','z','2'), mov_read_stsz }, /* compact sample size */
3144 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
3145 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
3146 { MKTAG('t','r','a','k'), mov_read_trak },
3147 { MKTAG('t','r','a','f'), mov_read_default },
3148 { MKTAG('t','r','e','f'), mov_read_default },
3149 { MKTAG('t','m','c','d'), mov_read_tmcd },
3150 { MKTAG('c','h','a','p'), mov_read_chap },
3151 { MKTAG('t','r','e','x'), mov_read_trex },
3152 { MKTAG('t','r','u','n'), mov_read_trun },
3153 { MKTAG('u','d','t','a'), mov_read_default },
3154 { MKTAG('w','a','v','e'), mov_read_wave },
3155 { MKTAG('e','s','d','s'), mov_read_esds },
3156 { MKTAG('d','a','c','3'), mov_read_dac3 }, /* AC-3 info */
3157 { MKTAG('d','e','c','3'), mov_read_dec3 }, /* EAC-3 info */
3158 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
3159 { MKTAG('w','f','e','x'), mov_read_wfex },
3160 { MKTAG('c','m','o','v'), mov_read_cmov },
3161 { MKTAG('c','h','a','n'), mov_read_chan }, /* channel layout */
3162 { MKTAG('d','v','c','1'), mov_read_dvc1 },
3163 { MKTAG('s','b','g','p'), mov_read_sbgp },
3164 { MKTAG('h','v','c','C'), mov_read_glbl },
3165 { MKTAG('u','u','i','d'), mov_read_uuid },
3166 { MKTAG('C','i','n', 0x8e), mov_read_targa_y216 },
3167 { MKTAG('-','-','-','-'), mov_read_custom },
3168 { 0, NULL }
3169 };
3170
mov_read_default(MOVContext * c,AVIOContext * pb,MOVAtom atom)3171 static int mov_read_default(MOVContext *c, AVIOContext *pb, MOVAtom atom)
3172 {
3173 int64_t total_size = 0;
3174 MOVAtom a;
3175 int i;
3176
3177 if (atom.size < 0)
3178 atom.size = INT64_MAX;
3179 while (total_size + 8 <= atom.size && !avio_feof(pb)) {
3180 int (*parse)(MOVContext*, AVIOContext*, MOVAtom) = NULL;
3181 a.size = atom.size;
3182 a.type=0;
3183 if (atom.size >= 8) {
3184 a.size = avio_rb32(pb);
3185 a.type = avio_rl32(pb);
3186 if (a.type == MKTAG('f','r','e','e') &&
3187 a.size >= 8 &&
3188 c->moov_retry) {
3189 uint8_t buf[8];
3190 uint32_t *type = (uint32_t *)buf + 1;
3191 avio_read(pb, buf, 8);
3192 avio_seek(pb, -8, SEEK_CUR);
3193 if (*type == MKTAG('m','v','h','d') ||
3194 *type == MKTAG('c','m','o','v')) {
3195 av_log(c->fc, AV_LOG_ERROR, "Detected moov in a free atom.\n");
3196 a.type = MKTAG('m','o','o','v');
3197 }
3198 }
3199 if (atom.type != MKTAG('r','o','o','t') &&
3200 atom.type != MKTAG('m','o','o','v'))
3201 {
3202 if (a.type == MKTAG('t','r','a','k') || a.type == MKTAG('m','d','a','t'))
3203 {
3204 av_log(c->fc, AV_LOG_ERROR, "Broken file, trak/mdat not at top-level\n");
3205 avio_skip(pb, -8);
3206 return 0;
3207 }
3208 }
3209 total_size += 8;
3210 if (a.size == 1) { /* 64 bit extended size */
3211 a.size = avio_rb64(pb) - 8;
3212 total_size += 8;
3213 }
3214 }
3215 av_dlog(c->fc, "type: %08x '%.4s' parent:'%.4s' sz: %"PRId64" %"PRId64" %"PRId64"\n",
3216 a.type, (char*)&a.type, (char*)&atom.type, a.size, total_size, atom.size);
3217 if (a.size == 0) {
3218 a.size = atom.size - total_size + 8;
3219 }
3220 a.size -= 8;
3221 if (a.size < 0)
3222 break;
3223 a.size = FFMIN(a.size, atom.size - total_size);
3224
3225 for (i = 0; mov_default_parse_table[i].type; i++)
3226 if (mov_default_parse_table[i].type == a.type) {
3227 parse = mov_default_parse_table[i].parse;
3228 break;
3229 }
3230
3231 // container is user data
3232 if (!parse && (atom.type == MKTAG('u','d','t','a') ||
3233 atom.type == MKTAG('i','l','s','t')))
3234 parse = mov_read_udta_string;
3235
3236 if (!parse) { /* skip leaf atoms data */
3237 avio_skip(pb, a.size);
3238 } else {
3239 int64_t start_pos = avio_tell(pb);
3240 int64_t left;
3241 int err = parse(c, pb, a);
3242 if (err < 0)
3243 return err;
3244 if (c->found_moov && c->found_mdat &&
3245 ((!pb->seekable || c->fc->flags & AVFMT_FLAG_IGNIDX) ||
3246 start_pos + a.size == avio_size(pb))) {
3247 if (!pb->seekable || c->fc->flags & AVFMT_FLAG_IGNIDX)
3248 c->next_root_atom = start_pos + a.size;
3249 return 0;
3250 }
3251 left = a.size - avio_tell(pb) + start_pos;
3252 if (left > 0) /* skip garbage at atom end */
3253 avio_skip(pb, left);
3254 else if (left < 0) {
3255 av_log(c->fc, AV_LOG_WARNING,
3256 "overread end of atom '%.4s' by %"PRId64" bytes\n",
3257 (char*)&a.type, -left);
3258 avio_seek(pb, left, SEEK_CUR);
3259 }
3260 }
3261
3262 total_size += a.size;
3263 }
3264
3265 if (total_size < atom.size && atom.size < 0x7ffff)
3266 avio_skip(pb, atom.size - total_size);
3267
3268 return 0;
3269 }
3270
mov_probe(AVProbeData * p)3271 static int mov_probe(AVProbeData *p)
3272 {
3273 int64_t offset;
3274 uint32_t tag;
3275 int score = 0;
3276 int moov_offset = -1;
3277
3278 /* check file header */
3279 offset = 0;
3280 for (;;) {
3281 /* ignore invalid offset */
3282 if ((offset + 8) > (unsigned int)p->buf_size)
3283 break;
3284 tag = AV_RL32(p->buf + offset + 4);
3285 switch(tag) {
3286 /* check for obvious tags */
3287 case MKTAG('m','o','o','v'):
3288 moov_offset = offset + 4;
3289 case MKTAG('m','d','a','t'):
3290 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
3291 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
3292 case MKTAG('f','t','y','p'):
3293 if (AV_RB32(p->buf+offset) < 8 &&
3294 (AV_RB32(p->buf+offset) != 1 ||
3295 offset + 12 > (unsigned int)p->buf_size ||
3296 AV_RB64(p->buf+offset + 8) == 0)) {
3297 score = FFMAX(score, AVPROBE_SCORE_EXTENSION);
3298 } else if (tag == MKTAG('f','t','y','p') &&
3299 AV_RL32(p->buf + offset + 8) == MKTAG('j','p','2',' ')) {
3300 score = FFMAX(score, 5);
3301 } else {
3302 score = AVPROBE_SCORE_MAX;
3303 }
3304 offset = FFMAX(4, AV_RB32(p->buf+offset)) + offset;
3305 break;
3306 /* those are more common words, so rate then a bit less */
3307 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
3308 case MKTAG('w','i','d','e'):
3309 case MKTAG('f','r','e','e'):
3310 case MKTAG('j','u','n','k'):
3311 case MKTAG('p','i','c','t'):
3312 score = FFMAX(score, AVPROBE_SCORE_MAX - 5);
3313 offset = FFMAX(4, AV_RB32(p->buf+offset)) + offset;
3314 break;
3315 case MKTAG(0x82,0x82,0x7f,0x7d):
3316 case MKTAG('s','k','i','p'):
3317 case MKTAG('u','u','i','d'):
3318 case MKTAG('p','r','f','l'):
3319 /* if we only find those cause probedata is too small at least rate them */
3320 score = FFMAX(score, AVPROBE_SCORE_EXTENSION);
3321 offset = FFMAX(4, AV_RB32(p->buf+offset)) + offset;
3322 break;
3323 default:
3324 offset = FFMAX(4, AV_RB32(p->buf+offset)) + offset;
3325 }
3326 }
3327 if(score > AVPROBE_SCORE_MAX - 50 && moov_offset != -1) {
3328 /* moov atom in the header - we should make sure that this is not a
3329 * MOV-packed MPEG-PS */
3330 offset = moov_offset;
3331
3332 while(offset < (p->buf_size - 16)){ /* Sufficient space */
3333 /* We found an actual hdlr atom */
3334 if(AV_RL32(p->buf + offset ) == MKTAG('h','d','l','r') &&
3335 AV_RL32(p->buf + offset + 8) == MKTAG('m','h','l','r') &&
3336 AV_RL32(p->buf + offset + 12) == MKTAG('M','P','E','G')){
3337 av_log(NULL, AV_LOG_WARNING, "Found media data tag MPEG indicating this is a MOV-packed MPEG-PS.\n");
3338 /* We found a media handler reference atom describing an
3339 * MPEG-PS-in-MOV, return a
3340 * low score to force expanding the probe window until
3341 * mpegps_probe finds what it needs */
3342 return 5;
3343 }else
3344 /* Keep looking */
3345 offset+=2;
3346 }
3347 }
3348
3349 return score;
3350 }
3351
3352 // must be done after parsing all trak because there's no order requirement
mov_read_chapters(AVFormatContext * s)3353 static void mov_read_chapters(AVFormatContext *s)
3354 {
3355 MOVContext *mov = s->priv_data;
3356 AVStream *st = NULL;
3357 MOVStreamContext *sc;
3358 int64_t cur_pos;
3359 int i;
3360
3361 for (i = 0; i < s->nb_streams; i++)
3362 if (s->streams[i]->id == mov->chapter_track) {
3363 st = s->streams[i];
3364 break;
3365 }
3366 if (!st) {
3367 av_log(s, AV_LOG_ERROR, "Referenced QT chapter track not found\n");
3368 return;
3369 }
3370
3371 st->discard = AVDISCARD_ALL;
3372 sc = st->priv_data;
3373 cur_pos = avio_tell(sc->pb);
3374
3375 for (i = 0; i < st->nb_index_entries; i++) {
3376 AVIndexEntry *sample = &st->index_entries[i];
3377 int64_t end = i+1 < st->nb_index_entries ? st->index_entries[i+1].timestamp : st->duration;
3378 uint8_t *title;
3379 uint16_t ch;
3380 int len, title_len;
3381
3382 if (end < sample->timestamp) {
3383 av_log(s, AV_LOG_WARNING, "ignoring stream duration which is shorter than chapters\n");
3384 end = AV_NOPTS_VALUE;
3385 }
3386
3387 if (avio_seek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
3388 av_log(s, AV_LOG_ERROR, "Chapter %d not found in file\n", i);
3389 goto finish;
3390 }
3391
3392 // the first two bytes are the length of the title
3393 len = avio_rb16(sc->pb);
3394 if (len > sample->size-2)
3395 continue;
3396 title_len = 2*len + 1;
3397 if (!(title = av_mallocz(title_len)))
3398 goto finish;
3399
3400 // The samples could theoretically be in any encoding if there's an encd
3401 // atom following, but in practice are only utf-8 or utf-16, distinguished
3402 // instead by the presence of a BOM
3403 if (!len) {
3404 title[0] = 0;
3405 } else {
3406 ch = avio_rb16(sc->pb);
3407 if (ch == 0xfeff)
3408 avio_get_str16be(sc->pb, len, title, title_len);
3409 else if (ch == 0xfffe)
3410 avio_get_str16le(sc->pb, len, title, title_len);
3411 else {
3412 AV_WB16(title, ch);
3413 if (len == 1 || len == 2)
3414 title[len] = 0;
3415 else
3416 avio_get_str(sc->pb, INT_MAX, title + 2, len - 1);
3417 }
3418 }
3419
3420 avpriv_new_chapter(s, i, st->time_base, sample->timestamp, end, title);
3421 av_freep(&title);
3422 }
3423 finish:
3424 avio_seek(sc->pb, cur_pos, SEEK_SET);
3425 }
3426
parse_timecode_in_framenum_format(AVFormatContext * s,AVStream * st,uint32_t value,int flags)3427 static int parse_timecode_in_framenum_format(AVFormatContext *s, AVStream *st,
3428 uint32_t value, int flags)
3429 {
3430 AVTimecode tc;
3431 char buf[AV_TIMECODE_STR_SIZE];
3432 AVRational rate = {st->codec->time_base.den,
3433 st->codec->time_base.num};
3434 int ret = av_timecode_init(&tc, rate, flags, 0, s);
3435 if (ret < 0)
3436 return ret;
3437 av_dict_set(&st->metadata, "timecode",
3438 av_timecode_make_string(&tc, buf, value), 0);
3439 return 0;
3440 }
3441
mov_read_timecode_track(AVFormatContext * s,AVStream * st)3442 static int mov_read_timecode_track(AVFormatContext *s, AVStream *st)
3443 {
3444 MOVStreamContext *sc = st->priv_data;
3445 int flags = 0;
3446 int64_t cur_pos = avio_tell(sc->pb);
3447 uint32_t value;
3448
3449 if (!st->nb_index_entries)
3450 return -1;
3451
3452 avio_seek(sc->pb, st->index_entries->pos, SEEK_SET);
3453 value = avio_rb32(s->pb);
3454
3455 if (sc->tmcd_flags & 0x0001) flags |= AV_TIMECODE_FLAG_DROPFRAME;
3456 if (sc->tmcd_flags & 0x0002) flags |= AV_TIMECODE_FLAG_24HOURSMAX;
3457 if (sc->tmcd_flags & 0x0004) flags |= AV_TIMECODE_FLAG_ALLOWNEGATIVE;
3458
3459 /* Assume Counter flag is set to 1 in tmcd track (even though it is likely
3460 * not the case) and thus assume "frame number format" instead of QT one.
3461 * No sample with tmcd track can be found with a QT timecode at the moment,
3462 * despite what the tmcd track "suggests" (Counter flag set to 0 means QT
3463 * format). */
3464 parse_timecode_in_framenum_format(s, st, value, flags);
3465
3466 avio_seek(sc->pb, cur_pos, SEEK_SET);
3467 return 0;
3468 }
3469
mov_read_close(AVFormatContext * s)3470 static int mov_read_close(AVFormatContext *s)
3471 {
3472 MOVContext *mov = s->priv_data;
3473 int i, j;
3474
3475 for (i = 0; i < s->nb_streams; i++) {
3476 AVStream *st = s->streams[i];
3477 MOVStreamContext *sc = st->priv_data;
3478
3479 av_freep(&sc->ctts_data);
3480 for (j = 0; j < sc->drefs_count; j++) {
3481 av_freep(&sc->drefs[j].path);
3482 av_freep(&sc->drefs[j].dir);
3483 }
3484 av_freep(&sc->drefs);
3485
3486 sc->drefs_count = 0;
3487
3488 if (!sc->pb_is_copied)
3489 avio_close(sc->pb);
3490
3491 sc->pb = NULL;
3492 av_freep(&sc->chunk_offsets);
3493 av_freep(&sc->stsc_data);
3494 av_freep(&sc->sample_sizes);
3495 av_freep(&sc->keyframes);
3496 av_freep(&sc->stts_data);
3497 av_freep(&sc->stps_data);
3498 av_freep(&sc->rap_group);
3499 av_freep(&sc->display_matrix);
3500 }
3501
3502 if (mov->dv_demux) {
3503 avformat_free_context(mov->dv_fctx);
3504 mov->dv_fctx = NULL;
3505 }
3506
3507 av_freep(&mov->trex_data);
3508 av_freep(&mov->bitrates);
3509
3510 return 0;
3511 }
3512
tmcd_is_referenced(AVFormatContext * s,int tmcd_id)3513 static int tmcd_is_referenced(AVFormatContext *s, int tmcd_id)
3514 {
3515 int i;
3516
3517 for (i = 0; i < s->nb_streams; i++) {
3518 AVStream *st = s->streams[i];
3519 MOVStreamContext *sc = st->priv_data;
3520
3521 if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO &&
3522 sc->timecode_track == tmcd_id)
3523 return 1;
3524 }
3525 return 0;
3526 }
3527
3528 /* look for a tmcd track not referenced by any video track, and export it globally */
export_orphan_timecode(AVFormatContext * s)3529 static void export_orphan_timecode(AVFormatContext *s)
3530 {
3531 int i;
3532
3533 for (i = 0; i < s->nb_streams; i++) {
3534 AVStream *st = s->streams[i];
3535
3536 if (st->codec->codec_tag == MKTAG('t','m','c','d') &&
3537 !tmcd_is_referenced(s, i + 1)) {
3538 AVDictionaryEntry *tcr = av_dict_get(st->metadata, "timecode", NULL, 0);
3539 if (tcr) {
3540 av_dict_set(&s->metadata, "timecode", tcr->value, 0);
3541 break;
3542 }
3543 }
3544 }
3545 }
3546
mov_read_header(AVFormatContext * s)3547 static int mov_read_header(AVFormatContext *s)
3548 {
3549 MOVContext *mov = s->priv_data;
3550 AVIOContext *pb = s->pb;
3551 int j, err;
3552 MOVAtom atom = { AV_RL32("root") };
3553 int i;
3554
3555 mov->fc = s;
3556 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
3557 if (pb->seekable)
3558 atom.size = avio_size(pb);
3559 else
3560 atom.size = INT64_MAX;
3561
3562 /* check MOV header */
3563 do {
3564 if (mov->moov_retry)
3565 avio_seek(pb, 0, SEEK_SET);
3566 if ((err = mov_read_default(mov, pb, atom)) < 0) {
3567 av_log(s, AV_LOG_ERROR, "error reading header\n");
3568 mov_read_close(s);
3569 return err;
3570 }
3571 } while (pb->seekable && !mov->found_moov && !mov->moov_retry++);
3572 if (!mov->found_moov) {
3573 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
3574 mov_read_close(s);
3575 return AVERROR_INVALIDDATA;
3576 }
3577 av_dlog(mov->fc, "on_parse_exit_offset=%"PRId64"\n", avio_tell(pb));
3578
3579 if (pb->seekable) {
3580 if (mov->chapter_track > 0)
3581 mov_read_chapters(s);
3582 for (i = 0; i < s->nb_streams; i++)
3583 if (s->streams[i]->codec->codec_tag == AV_RL32("tmcd"))
3584 mov_read_timecode_track(s, s->streams[i]);
3585 }
3586
3587 /* copy timecode metadata from tmcd tracks to the related video streams */
3588 for (i = 0; i < s->nb_streams; i++) {
3589 AVStream *st = s->streams[i];
3590 MOVStreamContext *sc = st->priv_data;
3591 if (sc->timecode_track > 0) {
3592 AVDictionaryEntry *tcr;
3593 int tmcd_st_id = -1;
3594
3595 for (j = 0; j < s->nb_streams; j++)
3596 if (s->streams[j]->id == sc->timecode_track)
3597 tmcd_st_id = j;
3598
3599 if (tmcd_st_id < 0 || tmcd_st_id == i)
3600 continue;
3601 tcr = av_dict_get(s->streams[tmcd_st_id]->metadata, "timecode", NULL, 0);
3602 if (tcr)
3603 av_dict_set(&st->metadata, "timecode", tcr->value, 0);
3604 }
3605 }
3606 export_orphan_timecode(s);
3607
3608 for (i = 0; i < s->nb_streams; i++) {
3609 AVStream *st = s->streams[i];
3610 MOVStreamContext *sc = st->priv_data;
3611 fix_timescale(mov, sc);
3612 if(st->codec->codec_type == AVMEDIA_TYPE_AUDIO && st->codec->codec_id == AV_CODEC_ID_AAC) {
3613 st->skip_samples = sc->start_pad;
3614 }
3615 if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO && sc->nb_frames_for_fps > 0 && sc->duration_for_fps > 0)
3616 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
3617 sc->time_scale*(int64_t)sc->nb_frames_for_fps, sc->duration_for_fps, INT_MAX);
3618 if (st->codec->codec_type == AVMEDIA_TYPE_SUBTITLE) {
3619 if (st->codec->width <= 0 && st->codec->height <= 0) {
3620 st->codec->width = sc->width;
3621 st->codec->height = sc->height;
3622 }
3623 if (st->codec->codec_id == AV_CODEC_ID_DVD_SUBTITLE) {
3624 if ((err = mov_rewrite_dvd_sub_extradata(st)) < 0)
3625 return err;
3626 }
3627 }
3628 }
3629
3630 if (mov->trex_data) {
3631 for (i = 0; i < s->nb_streams; i++) {
3632 AVStream *st = s->streams[i];
3633 MOVStreamContext *sc = st->priv_data;
3634 if (st->duration > 0)
3635 st->codec->bit_rate = sc->data_size * 8 * sc->time_scale / st->duration;
3636 }
3637 }
3638
3639 for (i = 0; i < mov->bitrates_count && i < s->nb_streams; i++) {
3640 if (mov->bitrates[i]) {
3641 s->streams[i]->codec->bit_rate = mov->bitrates[i];
3642 }
3643 }
3644
3645 ff_rfps_calculate(s);
3646
3647 for (i = 0; i < s->nb_streams; i++) {
3648 AVStream *st = s->streams[i];
3649 MOVStreamContext *sc = st->priv_data;
3650
3651 switch (st->codec->codec_type) {
3652 case AVMEDIA_TYPE_AUDIO:
3653 err = ff_replaygain_export(st, s->metadata);
3654 if (err < 0) {
3655 mov_read_close(s);
3656 return err;
3657 }
3658 break;
3659 case AVMEDIA_TYPE_VIDEO:
3660 if (sc->display_matrix) {
3661 AVPacketSideData *sd, *tmp;
3662
3663 tmp = av_realloc_array(st->side_data,
3664 st->nb_side_data + 1, sizeof(*tmp));
3665 if (!tmp)
3666 return AVERROR(ENOMEM);
3667
3668 st->side_data = tmp;
3669 st->nb_side_data++;
3670
3671 sd = &st->side_data[st->nb_side_data - 1];
3672 sd->type = AV_PKT_DATA_DISPLAYMATRIX;
3673 sd->size = sizeof(int32_t) * 9;
3674 sd->data = (uint8_t*)sc->display_matrix;
3675 sc->display_matrix = NULL;
3676 }
3677 break;
3678 }
3679 }
3680
3681 return 0;
3682 }
3683
mov_find_next_sample(AVFormatContext * s,AVStream ** st)3684 static AVIndexEntry *mov_find_next_sample(AVFormatContext *s, AVStream **st)
3685 {
3686 AVIndexEntry *sample = NULL;
3687 int64_t best_dts = INT64_MAX;
3688 int i;
3689 for (i = 0; i < s->nb_streams; i++) {
3690 AVStream *avst = s->streams[i];
3691 MOVStreamContext *msc = avst->priv_data;
3692 if (msc->pb && msc->current_sample < avst->nb_index_entries) {
3693 AVIndexEntry *current_sample = &avst->index_entries[msc->current_sample];
3694 int64_t dts = av_rescale(current_sample->timestamp, AV_TIME_BASE, msc->time_scale);
3695 av_dlog(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
3696 if (!sample || (!s->pb->seekable && current_sample->pos < sample->pos) ||
3697 (s->pb->seekable &&
3698 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
3699 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
3700 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
3701 sample = current_sample;
3702 best_dts = dts;
3703 *st = avst;
3704 }
3705 }
3706 }
3707 return sample;
3708 }
3709
mov_read_packet(AVFormatContext * s,AVPacket * pkt)3710 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
3711 {
3712 MOVContext *mov = s->priv_data;
3713 MOVStreamContext *sc;
3714 AVIndexEntry *sample;
3715 AVStream *st = NULL;
3716 int ret;
3717 mov->fc = s;
3718 retry:
3719 sample = mov_find_next_sample(s, &st);
3720 if (!sample) {
3721 mov->found_mdat = 0;
3722 if (!mov->next_root_atom)
3723 return AVERROR_EOF;
3724 avio_seek(s->pb, mov->next_root_atom, SEEK_SET);
3725 mov->next_root_atom = 0;
3726 if (mov_read_default(mov, s->pb, (MOVAtom){ AV_RL32("root"), INT64_MAX }) < 0 ||
3727 avio_feof(s->pb))
3728 return AVERROR_EOF;
3729 av_dlog(s, "read fragments, offset 0x%"PRIx64"\n", avio_tell(s->pb));
3730 goto retry;
3731 }
3732 sc = st->priv_data;
3733 /* must be done just before reading, to avoid infinite loop on sample */
3734 sc->current_sample++;
3735
3736 if (mov->next_root_atom) {
3737 sample->pos = FFMIN(sample->pos, mov->next_root_atom);
3738 sample->size = FFMIN(sample->size, (mov->next_root_atom - sample->pos));
3739 }
3740
3741 if (st->discard != AVDISCARD_ALL) {
3742 if (avio_seek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
3743 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
3744 sc->ffindex, sample->pos);
3745 return AVERROR_INVALIDDATA;
3746 }
3747 ret = av_get_packet(sc->pb, pkt, sample->size);
3748 if (ret < 0)
3749 return ret;
3750 if (sc->has_palette) {
3751 uint8_t *pal;
3752
3753 pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
3754 if (!pal) {
3755 av_log(mov->fc, AV_LOG_ERROR, "Cannot append palette to packet\n");
3756 } else {
3757 memcpy(pal, sc->palette, AVPALETTE_SIZE);
3758 sc->has_palette = 0;
3759 }
3760 }
3761 #if CONFIG_DV_DEMUXER
3762 if (mov->dv_demux && sc->dv_audio_container) {
3763 avpriv_dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size, pkt->pos);
3764 av_free(pkt->data);
3765 pkt->size = 0;
3766 ret = avpriv_dv_get_packet(mov->dv_demux, pkt);
3767 if (ret < 0)
3768 return ret;
3769 }
3770 #endif
3771 }
3772
3773 pkt->stream_index = sc->ffindex;
3774 pkt->dts = sample->timestamp;
3775 if (sc->ctts_data && sc->ctts_index < sc->ctts_count) {
3776 pkt->pts = pkt->dts + sc->dts_shift + sc->ctts_data[sc->ctts_index].duration;
3777 /* update ctts context */
3778 sc->ctts_sample++;
3779 if (sc->ctts_index < sc->ctts_count &&
3780 sc->ctts_data[sc->ctts_index].count == sc->ctts_sample) {
3781 sc->ctts_index++;
3782 sc->ctts_sample = 0;
3783 }
3784 if (sc->wrong_dts)
3785 pkt->dts = AV_NOPTS_VALUE;
3786 } else {
3787 int64_t next_dts = (sc->current_sample < st->nb_index_entries) ?
3788 st->index_entries[sc->current_sample].timestamp : st->duration;
3789 pkt->duration = next_dts - pkt->dts;
3790 pkt->pts = pkt->dts;
3791 }
3792 if (st->discard == AVDISCARD_ALL)
3793 goto retry;
3794 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? AV_PKT_FLAG_KEY : 0;
3795 pkt->pos = sample->pos;
3796 av_dlog(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
3797 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
3798 return 0;
3799 }
3800
mov_seek_stream(AVFormatContext * s,AVStream * st,int64_t timestamp,int flags)3801 static int mov_seek_stream(AVFormatContext *s, AVStream *st, int64_t timestamp, int flags)
3802 {
3803 MOVStreamContext *sc = st->priv_data;
3804 int sample, time_sample;
3805 int i;
3806
3807 sample = av_index_search_timestamp(st, timestamp, flags);
3808 av_dlog(s, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
3809 if (sample < 0 && st->nb_index_entries && timestamp < st->index_entries[0].timestamp)
3810 sample = 0;
3811 if (sample < 0) /* not sure what to do */
3812 return AVERROR_INVALIDDATA;
3813 sc->current_sample = sample;
3814 av_dlog(s, "stream %d, found sample %d\n", st->index, sc->current_sample);
3815 /* adjust ctts index */
3816 if (sc->ctts_data) {
3817 time_sample = 0;
3818 for (i = 0; i < sc->ctts_count; i++) {
3819 int next = time_sample + sc->ctts_data[i].count;
3820 if (next > sc->current_sample) {
3821 sc->ctts_index = i;
3822 sc->ctts_sample = sc->current_sample - time_sample;
3823 break;
3824 }
3825 time_sample = next;
3826 }
3827 }
3828 return sample;
3829 }
3830
mov_read_seek(AVFormatContext * s,int stream_index,int64_t sample_time,int flags)3831 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
3832 {
3833 AVStream *st;
3834 int64_t seek_timestamp, timestamp;
3835 int sample;
3836 int i;
3837
3838 if (stream_index >= s->nb_streams)
3839 return AVERROR_INVALIDDATA;
3840
3841 st = s->streams[stream_index];
3842 sample = mov_seek_stream(s, st, sample_time, flags);
3843 if (sample < 0)
3844 return sample;
3845
3846 /* adjust seek timestamp to found sample timestamp */
3847 seek_timestamp = st->index_entries[sample].timestamp;
3848
3849 for (i = 0; i < s->nb_streams; i++) {
3850 MOVStreamContext *sc = s->streams[i]->priv_data;
3851 st = s->streams[i];
3852 st->skip_samples = (sample_time <= 0) ? sc->start_pad : 0;
3853
3854 if (stream_index == i)
3855 continue;
3856
3857 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
3858 mov_seek_stream(s, st, timestamp, flags);
3859 }
3860 return 0;
3861 }
3862
3863 static const AVOption options[] = {
3864 {"use_absolute_path",
3865 "allow using absolute path when opening alias, this is a possible security issue",
3866 offsetof(MOVContext, use_absolute_path), FF_OPT_TYPE_INT, {.i64 = 0},
3867 0, 1, AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_DECODING_PARAM},
3868 {"ignore_editlist", "", offsetof(MOVContext, ignore_editlist), FF_OPT_TYPE_INT, {.i64 = 0},
3869 0, 1, AV_OPT_FLAG_VIDEO_PARAM|AV_OPT_FLAG_DECODING_PARAM},
3870 {NULL}
3871 };
3872
3873 static const AVClass mov_class = {
3874 .class_name = "mov,mp4,m4a,3gp,3g2,mj2",
3875 .item_name = av_default_item_name,
3876 .option = options,
3877 .version = LIBAVUTIL_VERSION_INT,
3878 };
3879
3880 AVInputFormat ff_mov_demuxer = {
3881 .name = "mov,mp4,m4a,3gp,3g2,mj2",
3882 .long_name = NULL_IF_CONFIG_SMALL("QuickTime / MOV"),
3883 .priv_data_size = sizeof(MOVContext),
3884 .extensions = "mov,mp4,m4a,3gp,3g2,mj2",
3885 .read_probe = mov_probe,
3886 .read_header = mov_read_header,
3887 .read_packet = mov_read_packet,
3888 .read_close = mov_read_close,
3889 .read_seek = mov_read_seek,
3890 .priv_class = &mov_class,
3891 .flags = AVFMT_NO_BYTE_SEEK,
3892 };
3893