3 * Copyright (c) 2001 Fabrice Bellard
4 * Copyright (c) 2009 Baptiste Coudurier <baptiste dot coudurier at gmail dot com>
6 * This file is part of Libav.
8 * Libav is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
13 * Libav is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with Libav; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
26 //#define MOV_EXPORT_ALL_METADATA
28 #include "libavutil/intreadwrite.h"
29 #include "libavutil/intfloat_readwrite.h"
30 #include "libavutil/mathematics.h"
31 #include "libavutil/avstring.h"
32 #include "libavutil/dict.h"
34 #include "avio_internal.h"
37 #include "libavcodec/get_bits.h"
45 * First version by Francois Revol revol@free.fr
46 * Seek function by Gael Chardon gael.dev@4now.net
48 * Features and limitations:
49 * - reads most of the QT files I have (at least the structure),
50 * Sample QuickTime files with mp3 audio can be found at: http://www.3ivx.com/showcase.html
51 * - the code is quite ugly... maybe I won't do it recursive next time :-)
53 * Funny I didn't know about http://sourceforge.net/projects/qt-ffmpeg/
54 * when coding this :) (it's a writer anyway)
56 * Reference documents:
57 * http://www.geocities.com/xhelmboyx/quicktime/formats/qtm-layout.txt
59 * http://developer.apple.com/documentation/QuickTime/QTFF/
60 * http://developer.apple.com/documentation/QuickTime/QTFF/qtff.pdf
61 * QuickTime is a trademark of Apple (AFAIK :))
64 #include "qtpalette.h"
70 /* XXX: it's the first time I make a recursive parser I think... sorry if it's ugly :P */
72 /* those functions parse an atom */
74 0: continue to parse next atom
75 <0: error occurred, exit
77 /* links atom IDs to parse functions */
78 typedef struct MOVParseTableEntry {
80 int (*parse)(MOVContext *ctx, AVIOContext *pb, MOVAtom atom);
83 static const MOVParseTableEntry mov_default_parse_table[];
85 static int mov_metadata_track_or_disc_number(MOVContext *c, AVIOContext *pb,
86 unsigned len, const char *key)
91 avio_rb16(pb); // unknown
92 current = avio_rb16(pb);
93 total = avio_rb16(pb);
95 snprintf(buf, sizeof(buf), "%d", current);
97 snprintf(buf, sizeof(buf), "%d/%d", current, total);
98 av_dict_set(&c->fc->metadata, key, buf, 0);
103 static int mov_metadata_int8_bypass_padding(MOVContext *c, AVIOContext *pb,
104 unsigned len, const char *key)
108 /* bypass padding bytes */
113 snprintf(buf, sizeof(buf), "%hu", avio_r8(pb));
114 av_dict_set(&c->fc->metadata, key, buf, 0);
119 static int mov_metadata_int8_no_padding(MOVContext *c, AVIOContext *pb,
120 unsigned len, const char *key)
124 snprintf(buf, sizeof(buf), "%hu", avio_r8(pb));
125 av_dict_set(&c->fc->metadata, key, buf, 0);
130 static int mov_metadata_gnre(MOVContext *c, AVIOContext *pb,
131 unsigned len, const char *key)
136 avio_r8(pb); // unknown
139 if (genre < 1 || genre > ID3v1_GENRE_MAX)
141 snprintf(buf, sizeof(buf), "%s", ff_id3v1_genre_str[genre-1]);
142 av_dict_set(&c->fc->metadata, key, buf, 0);
147 static const uint32_t mac_to_unicode[128] = {
148 0x00C4,0x00C5,0x00C7,0x00C9,0x00D1,0x00D6,0x00DC,0x00E1,
149 0x00E0,0x00E2,0x00E4,0x00E3,0x00E5,0x00E7,0x00E9,0x00E8,
150 0x00EA,0x00EB,0x00ED,0x00EC,0x00EE,0x00EF,0x00F1,0x00F3,
151 0x00F2,0x00F4,0x00F6,0x00F5,0x00FA,0x00F9,0x00FB,0x00FC,
152 0x2020,0x00B0,0x00A2,0x00A3,0x00A7,0x2022,0x00B6,0x00DF,
153 0x00AE,0x00A9,0x2122,0x00B4,0x00A8,0x2260,0x00C6,0x00D8,
154 0x221E,0x00B1,0x2264,0x2265,0x00A5,0x00B5,0x2202,0x2211,
155 0x220F,0x03C0,0x222B,0x00AA,0x00BA,0x03A9,0x00E6,0x00F8,
156 0x00BF,0x00A1,0x00AC,0x221A,0x0192,0x2248,0x2206,0x00AB,
157 0x00BB,0x2026,0x00A0,0x00C0,0x00C3,0x00D5,0x0152,0x0153,
158 0x2013,0x2014,0x201C,0x201D,0x2018,0x2019,0x00F7,0x25CA,
159 0x00FF,0x0178,0x2044,0x20AC,0x2039,0x203A,0xFB01,0xFB02,
160 0x2021,0x00B7,0x201A,0x201E,0x2030,0x00C2,0x00CA,0x00C1,
161 0x00CB,0x00C8,0x00CD,0x00CE,0x00CF,0x00CC,0x00D3,0x00D4,
162 0xF8FF,0x00D2,0x00DA,0x00DB,0x00D9,0x0131,0x02C6,0x02DC,
163 0x00AF,0x02D8,0x02D9,0x02DA,0x00B8,0x02DD,0x02DB,0x02C7,
166 static int mov_read_mac_string(MOVContext *c, AVIOContext *pb, int len,
167 char *dst, int dstlen)
170 char *end = dst+dstlen-1;
173 for (i = 0; i < len; i++) {
174 uint8_t t, c = avio_r8(pb);
175 if (c < 0x80 && p < end)
178 PUT_UTF8(mac_to_unicode[c-0x80], t, if (p < end) *p++ = t;);
184 static int mov_read_udta_string(MOVContext *c, AVIOContext *pb, MOVAtom atom)
186 #ifdef MOV_EXPORT_ALL_METADATA
189 char str[1024], key2[16], language[4] = {0};
190 const char *key = NULL;
191 uint16_t str_size, langcode = 0;
192 uint32_t data_type = 0;
193 int (*parse)(MOVContext*, AVIOContext*, unsigned, const char*) = NULL;
196 case MKTAG(0xa9,'n','a','m'): key = "title"; break;
197 case MKTAG(0xa9,'a','u','t'):
198 case MKTAG(0xa9,'A','R','T'): key = "artist"; break;
199 case MKTAG( 'a','A','R','T'): key = "album_artist"; break;
200 case MKTAG(0xa9,'w','r','t'): key = "composer"; break;
201 case MKTAG( 'c','p','r','t'):
202 case MKTAG(0xa9,'c','p','y'): key = "copyright"; break;
203 case MKTAG(0xa9,'c','m','t'):
204 case MKTAG(0xa9,'i','n','f'): key = "comment"; break;
205 case MKTAG(0xa9,'a','l','b'): key = "album"; break;
206 case MKTAG(0xa9,'d','a','y'): key = "date"; break;
207 case MKTAG(0xa9,'g','e','n'): key = "genre"; break;
208 case MKTAG( 'g','n','r','e'): key = "genre";
209 parse = mov_metadata_gnre; break;
210 case MKTAG(0xa9,'t','o','o'):
211 case MKTAG(0xa9,'s','w','r'): key = "encoder"; break;
212 case MKTAG(0xa9,'e','n','c'): key = "encoder"; break;
213 case MKTAG( 'd','e','s','c'): key = "description";break;
214 case MKTAG( 'l','d','e','s'): key = "synopsis"; break;
215 case MKTAG( 't','v','s','h'): key = "show"; break;
216 case MKTAG( 't','v','e','n'): key = "episode_id";break;
217 case MKTAG( 't','v','n','n'): key = "network"; break;
218 case MKTAG( 't','r','k','n'): key = "track";
219 parse = mov_metadata_track_or_disc_number; break;
220 case MKTAG( 'd','i','s','k'): key = "disc";
221 parse = mov_metadata_track_or_disc_number; break;
222 case MKTAG( 't','v','e','s'): key = "episode_sort";
223 parse = mov_metadata_int8_bypass_padding; break;
224 case MKTAG( 't','v','s','n'): key = "season_number";
225 parse = mov_metadata_int8_bypass_padding; break;
226 case MKTAG( 's','t','i','k'): key = "media_type";
227 parse = mov_metadata_int8_no_padding; break;
228 case MKTAG( 'h','d','v','d'): key = "hd_video";
229 parse = mov_metadata_int8_no_padding; break;
230 case MKTAG( 'p','g','a','p'): key = "gapless_playback";
231 parse = mov_metadata_int8_no_padding; break;
234 if (c->itunes_metadata && atom.size > 8) {
235 int data_size = avio_rb32(pb);
236 int tag = avio_rl32(pb);
237 if (tag == MKTAG('d','a','t','a')) {
238 data_type = avio_rb32(pb); // type
239 avio_rb32(pb); // unknown
240 str_size = data_size - 16;
243 } else if (atom.size > 4 && key && !c->itunes_metadata) {
244 str_size = avio_rb16(pb); // string length
245 langcode = avio_rb16(pb);
246 ff_mov_lang_to_iso639(langcode, language);
249 str_size = atom.size;
251 #ifdef MOV_EXPORT_ALL_METADATA
253 snprintf(tmp_key, 5, "%.4s", (char*)&atom.type);
263 str_size = FFMIN3(sizeof(str)-1, str_size, atom.size);
266 parse(c, pb, str_size, key);
268 if (data_type == 3 || (data_type == 0 && langcode < 0x800)) { // MAC Encoded
269 mov_read_mac_string(c, pb, str_size, str, sizeof(str));
271 avio_read(pb, str, str_size);
274 av_dict_set(&c->fc->metadata, key, str, 0);
275 if (*language && strcmp(language, "und")) {
276 snprintf(key2, sizeof(key2), "%s-%s", key, language);
277 av_dict_set(&c->fc->metadata, key2, str, 0);
280 av_dlog(c->fc, "lang \"%3s\" ", language);
281 av_dlog(c->fc, "tag \"%s\" value \"%s\" atom \"%.4s\" %d %"PRId64"\n",
282 key, str, (char*)&atom.type, str_size, atom.size);
287 static int mov_read_chpl(MOVContext *c, AVIOContext *pb, MOVAtom atom)
290 int i, nb_chapters, str_len, version;
293 if ((atom.size -= 5) < 0)
296 version = avio_r8(pb);
299 avio_rb32(pb); // ???
300 nb_chapters = avio_r8(pb);
302 for (i = 0; i < nb_chapters; i++) {
306 start = avio_rb64(pb);
307 str_len = avio_r8(pb);
309 if ((atom.size -= 9+str_len) < 0)
312 avio_read(pb, str, str_len);
314 avpriv_new_chapter(c->fc, i, (AVRational){1,10000000}, start, AV_NOPTS_VALUE, str);
319 static int mov_read_default(MOVContext *c, AVIOContext *pb, MOVAtom atom)
321 int64_t total_size = 0;
326 atom.size = INT64_MAX;
327 while (total_size + 8 < atom.size && !pb->eof_reached) {
328 int (*parse)(MOVContext*, AVIOContext*, MOVAtom) = NULL;
331 if (atom.size >= 8) {
332 a.size = avio_rb32(pb);
333 a.type = avio_rl32(pb);
335 av_dlog(c->fc, "type: %08x '%.4s' parent:'%.4s' sz: %"PRId64" %"PRId64" %"PRId64"\n",
336 a.type, (char*)&a.type, (char*)&atom.type, a.size, total_size, atom.size);
338 if (a.size == 1) { /* 64 bit extended size */
339 a.size = avio_rb64(pb) - 8;
343 a.size = atom.size - total_size;
350 a.size = FFMIN(a.size, atom.size - total_size);
352 for (i = 0; mov_default_parse_table[i].type; i++)
353 if (mov_default_parse_table[i].type == a.type) {
354 parse = mov_default_parse_table[i].parse;
358 // container is user data
359 if (!parse && (atom.type == MKTAG('u','d','t','a') ||
360 atom.type == MKTAG('i','l','s','t')))
361 parse = mov_read_udta_string;
363 if (!parse) { /* skip leaf atoms data */
364 avio_skip(pb, a.size);
366 int64_t start_pos = avio_tell(pb);
368 int err = parse(c, pb, a);
371 if (c->found_moov && c->found_mdat &&
372 (!pb->seekable || start_pos + a.size == avio_size(pb)))
374 left = a.size - avio_tell(pb) + start_pos;
375 if (left > 0) /* skip garbage at atom end */
379 total_size += a.size;
382 if (total_size < atom.size && atom.size < 0x7ffff)
383 avio_skip(pb, atom.size - total_size);
388 static int mov_read_dref(MOVContext *c, AVIOContext *pb, MOVAtom atom)
391 MOVStreamContext *sc;
394 if (c->fc->nb_streams < 1)
396 st = c->fc->streams[c->fc->nb_streams-1];
399 avio_rb32(pb); // version + flags
400 entries = avio_rb32(pb);
401 if (entries >= UINT_MAX / sizeof(*sc->drefs))
403 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
405 return AVERROR(ENOMEM);
406 sc->drefs_count = entries;
408 for (i = 0; i < sc->drefs_count; i++) {
409 MOVDref *dref = &sc->drefs[i];
410 uint32_t size = avio_rb32(pb);
411 int64_t next = avio_tell(pb) + size - 4;
416 dref->type = avio_rl32(pb);
417 avio_rb32(pb); // version + flags
418 av_dlog(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
420 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
421 /* macintosh alias record */
422 uint16_t volume_len, len;
427 volume_len = avio_r8(pb);
428 volume_len = FFMIN(volume_len, 27);
429 avio_read(pb, dref->volume, 27);
430 dref->volume[volume_len] = 0;
431 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", dref->volume, volume_len);
436 len = FFMIN(len, 63);
437 avio_read(pb, dref->filename, 63);
438 dref->filename[len] = 0;
439 av_log(c->fc, AV_LOG_DEBUG, "filename %s, len %d\n", dref->filename, len);
443 /* read next level up_from_alias/down_to_target */
444 dref->nlvl_from = avio_rb16(pb);
445 dref->nlvl_to = avio_rb16(pb);
446 av_log(c->fc, AV_LOG_DEBUG, "nlvl from %d, nlvl to %d\n",
447 dref->nlvl_from, dref->nlvl_to);
451 for (type = 0; type != -1 && avio_tell(pb) < next; ) {
452 type = avio_rb16(pb);
454 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
457 if (type == 2) { // absolute path
459 dref->path = av_mallocz(len+1);
461 return AVERROR(ENOMEM);
462 avio_read(pb, dref->path, len);
463 if (len > volume_len && !strncmp(dref->path, dref->volume, volume_len)) {
465 memmove(dref->path, dref->path+volume_len, len);
468 for (j = 0; j < len; j++)
469 if (dref->path[j] == ':')
471 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
472 } else if (type == 0) { // directory name
474 dref->dir = av_malloc(len+1);
476 return AVERROR(ENOMEM);
477 avio_read(pb, dref->dir, len);
479 for (j = 0; j < len; j++)
480 if (dref->dir[j] == ':')
482 av_log(c->fc, AV_LOG_DEBUG, "dir %s\n", dref->dir);
487 avio_seek(pb, next, SEEK_SET);
492 static int mov_read_hdlr(MOVContext *c, AVIOContext *pb, MOVAtom atom)
496 uint32_t av_unused ctype;
498 if (c->fc->nb_streams < 1) // meta before first trak
501 st = c->fc->streams[c->fc->nb_streams-1];
503 avio_r8(pb); /* version */
504 avio_rb24(pb); /* flags */
507 ctype = avio_rl32(pb);
508 type = avio_rl32(pb); /* component subtype */
510 av_dlog(c->fc, "ctype= %.4s (0x%08x)\n", (char*)&ctype, ctype);
511 av_dlog(c->fc, "stype= %.4s\n", (char*)&type);
513 if (type == MKTAG('v','i','d','e'))
514 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
515 else if (type == MKTAG('s','o','u','n'))
516 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
517 else if (type == MKTAG('m','1','a',' '))
518 st->codec->codec_id = CODEC_ID_MP2;
519 else if ((type == MKTAG('s','u','b','p')) || (type == MKTAG('c','l','c','p')))
520 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
522 avio_rb32(pb); /* component manufacture */
523 avio_rb32(pb); /* component flags */
524 avio_rb32(pb); /* component flags mask */
529 int ff_mov_read_esds(AVFormatContext *fc, AVIOContext *pb, MOVAtom atom)
534 if (fc->nb_streams < 1)
536 st = fc->streams[fc->nb_streams-1];
538 avio_rb32(pb); /* version + flags */
539 ff_mp4_read_descr(fc, pb, &tag);
540 if (tag == MP4ESDescrTag) {
541 ff_mp4_parse_es_descr(pb, NULL);
543 avio_rb16(pb); /* ID */
545 ff_mp4_read_descr(fc, pb, &tag);
546 if (tag == MP4DecConfigDescrTag)
547 ff_mp4_read_dec_config_descr(fc, st, pb);
551 static int mov_read_esds(MOVContext *c, AVIOContext *pb, MOVAtom atom)
553 return ff_mov_read_esds(c->fc, pb, atom);
556 static int mov_read_dac3(MOVContext *c, AVIOContext *pb, MOVAtom atom)
559 int ac3info, acmod, lfeon, bsmod;
561 if (c->fc->nb_streams < 1)
563 st = c->fc->streams[c->fc->nb_streams-1];
565 ac3info = avio_rb24(pb);
566 bsmod = (ac3info >> 14) & 0x7;
567 acmod = (ac3info >> 11) & 0x7;
568 lfeon = (ac3info >> 10) & 0x1;
569 st->codec->channels = ((int[]){2,1,2,3,3,4,4,5})[acmod] + lfeon;
570 st->codec->audio_service_type = bsmod;
571 if (st->codec->channels > 1 && bsmod == 0x7)
572 st->codec->audio_service_type = AV_AUDIO_SERVICE_TYPE_KARAOKE;
577 static int mov_read_wfex(MOVContext *c, AVIOContext *pb, MOVAtom atom)
581 if (c->fc->nb_streams < 1)
583 st = c->fc->streams[c->fc->nb_streams-1];
585 ff_get_wav_header(pb, st->codec, atom.size);
590 static int mov_read_pasp(MOVContext *c, AVIOContext *pb, MOVAtom atom)
592 const int num = avio_rb32(pb);
593 const int den = avio_rb32(pb);
596 if (c->fc->nb_streams < 1)
598 st = c->fc->streams[c->fc->nb_streams-1];
600 if ((st->sample_aspect_ratio.den != 1 || st->sample_aspect_ratio.num) && // default
601 (den != st->sample_aspect_ratio.den || num != st->sample_aspect_ratio.num)) {
602 av_log(c->fc, AV_LOG_WARNING,
603 "sample aspect ratio already set to %d:%d, ignoring 'pasp' atom (%d:%d)\n",
604 st->sample_aspect_ratio.num, st->sample_aspect_ratio.den,
606 } else if (den != 0) {
607 st->sample_aspect_ratio.num = num;
608 st->sample_aspect_ratio.den = den;
613 /* this atom contains actual media data */
614 static int mov_read_mdat(MOVContext *c, AVIOContext *pb, MOVAtom atom)
616 if (atom.size == 0) /* wrong one (MP4) */
619 return 0; /* now go for moov */
622 /* read major brand, minor version and compatible brands and store them as metadata */
623 static int mov_read_ftyp(MOVContext *c, AVIOContext *pb, MOVAtom atom)
627 char minor_ver_str[11]; /* 32 bit integer -> 10 digits + null */
628 char* comp_brands_str;
629 uint8_t type[5] = {0};
631 avio_read(pb, type, 4);
632 if (strcmp(type, "qt "))
634 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
635 av_dict_set(&c->fc->metadata, "major_brand", type, 0);
636 minor_ver = avio_rb32(pb); /* minor version */
637 snprintf(minor_ver_str, sizeof(minor_ver_str), "%d", minor_ver);
638 av_dict_set(&c->fc->metadata, "minor_version", minor_ver_str, 0);
640 comp_brand_size = atom.size - 8;
641 if (comp_brand_size < 0)
643 comp_brands_str = av_malloc(comp_brand_size + 1); /* Add null terminator */
644 if (!comp_brands_str)
645 return AVERROR(ENOMEM);
646 avio_read(pb, comp_brands_str, comp_brand_size);
647 comp_brands_str[comp_brand_size] = 0;
648 av_dict_set(&c->fc->metadata, "compatible_brands", comp_brands_str, 0);
649 av_freep(&comp_brands_str);
654 /* this atom should contain all header atoms */
655 static int mov_read_moov(MOVContext *c, AVIOContext *pb, MOVAtom atom)
657 if (mov_read_default(c, pb, atom) < 0)
659 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
660 /* so we don't parse the whole file if over a network */
662 return 0; /* now go for mdat */
665 static int mov_read_moof(MOVContext *c, AVIOContext *pb, MOVAtom atom)
667 c->fragment.moof_offset = avio_tell(pb) - 8;
668 av_dlog(c->fc, "moof offset %"PRIx64"\n", c->fragment.moof_offset);
669 return mov_read_default(c, pb, atom);
672 static void mov_metadata_creation_time(AVDictionary **metadata, time_t time)
677 time -= 2082844800; /* seconds between 1904-01-01 and Epoch */
680 strftime(buffer, sizeof(buffer), "%Y-%m-%d %H:%M:%S", ptm);
681 av_dict_set(metadata, "creation_time", buffer, 0);
685 static int mov_read_mdhd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
688 MOVStreamContext *sc;
690 char language[4] = {0};
692 time_t creation_time;
694 if (c->fc->nb_streams < 1)
696 st = c->fc->streams[c->fc->nb_streams-1];
699 version = avio_r8(pb);
701 return -1; /* unsupported */
703 avio_rb24(pb); /* flags */
705 creation_time = avio_rb64(pb);
708 creation_time = avio_rb32(pb);
709 avio_rb32(pb); /* modification time */
711 mov_metadata_creation_time(&st->metadata, creation_time);
713 sc->time_scale = avio_rb32(pb);
714 st->duration = (version == 1) ? avio_rb64(pb) : avio_rb32(pb); /* duration */
716 lang = avio_rb16(pb); /* language */
717 if (ff_mov_lang_to_iso639(lang, language))
718 av_dict_set(&st->metadata, "language", language, 0);
719 avio_rb16(pb); /* quality */
724 static int mov_read_mvhd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
726 time_t creation_time;
727 int version = avio_r8(pb); /* version */
728 avio_rb24(pb); /* flags */
731 creation_time = avio_rb64(pb);
734 creation_time = avio_rb32(pb);
735 avio_rb32(pb); /* modification time */
737 mov_metadata_creation_time(&c->fc->metadata, creation_time);
738 c->time_scale = avio_rb32(pb); /* time scale */
740 av_dlog(c->fc, "time scale = %i\n", c->time_scale);
742 c->duration = (version == 1) ? avio_rb64(pb) : avio_rb32(pb); /* duration */
743 avio_rb32(pb); /* preferred scale */
745 avio_rb16(pb); /* preferred volume */
747 avio_skip(pb, 10); /* reserved */
749 avio_skip(pb, 36); /* display matrix */
751 avio_rb32(pb); /* preview time */
752 avio_rb32(pb); /* preview duration */
753 avio_rb32(pb); /* poster time */
754 avio_rb32(pb); /* selection time */
755 avio_rb32(pb); /* selection duration */
756 avio_rb32(pb); /* current time */
757 avio_rb32(pb); /* next track ID */
762 static int mov_read_smi(MOVContext *c, AVIOContext *pb, MOVAtom atom)
766 if (c->fc->nb_streams < 1)
768 st = c->fc->streams[c->fc->nb_streams-1];
770 if ((uint64_t)atom.size > (1<<30))
773 // currently SVQ3 decoder expect full STSD header - so let's fake it
774 // this should be fixed and just SMI header should be passed
775 av_free(st->codec->extradata);
776 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
777 if (!st->codec->extradata)
778 return AVERROR(ENOMEM);
779 st->codec->extradata_size = 0x5a + atom.size;
780 memcpy(st->codec->extradata, "SVQ3", 4); // fake
781 avio_read(pb, st->codec->extradata + 0x5a, atom.size);
782 av_dlog(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
786 static int mov_read_enda(MOVContext *c, AVIOContext *pb, MOVAtom atom)
791 if (c->fc->nb_streams < 1)
793 st = c->fc->streams[c->fc->nb_streams-1];
795 little_endian = avio_rb16(pb);
796 av_dlog(c->fc, "enda %d\n", little_endian);
797 if (little_endian == 1) {
798 switch (st->codec->codec_id) {
799 case CODEC_ID_PCM_S24BE:
800 st->codec->codec_id = CODEC_ID_PCM_S24LE;
802 case CODEC_ID_PCM_S32BE:
803 st->codec->codec_id = CODEC_ID_PCM_S32LE;
805 case CODEC_ID_PCM_F32BE:
806 st->codec->codec_id = CODEC_ID_PCM_F32LE;
808 case CODEC_ID_PCM_F64BE:
809 st->codec->codec_id = CODEC_ID_PCM_F64LE;
818 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
819 static int mov_read_extradata(MOVContext *c, AVIOContext *pb, MOVAtom atom)
825 if (c->fc->nb_streams < 1) // will happen with jp2 files
827 st= c->fc->streams[c->fc->nb_streams-1];
828 size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
829 if (size > INT_MAX || (uint64_t)atom.size > INT_MAX)
831 buf= av_realloc(st->codec->extradata, size);
834 st->codec->extradata= buf;
835 buf+= st->codec->extradata_size;
836 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
837 AV_WB32( buf , atom.size + 8);
838 AV_WL32( buf + 4, atom.type);
839 avio_read(pb, buf + 8, atom.size);
843 static int mov_read_wave(MOVContext *c, AVIOContext *pb, MOVAtom atom)
847 if (c->fc->nb_streams < 1)
849 st = c->fc->streams[c->fc->nb_streams-1];
851 if ((uint64_t)atom.size > (1<<30))
854 if (st->codec->codec_id == CODEC_ID_QDM2 || st->codec->codec_id == CODEC_ID_QDMC) {
855 // pass all frma atom to codec, needed at least for QDMC and QDM2
856 av_free(st->codec->extradata);
857 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
858 if (!st->codec->extradata)
859 return AVERROR(ENOMEM);
860 st->codec->extradata_size = atom.size;
861 avio_read(pb, st->codec->extradata, atom.size);
862 } else if (atom.size > 8) { /* to read frma, esds atoms */
863 if (mov_read_default(c, pb, atom) < 0)
866 avio_skip(pb, atom.size);
871 * This function reads atom content and puts data in extradata without tag
872 * nor size unlike mov_read_extradata.
874 static int mov_read_glbl(MOVContext *c, AVIOContext *pb, MOVAtom atom)
878 if (c->fc->nb_streams < 1)
880 st = c->fc->streams[c->fc->nb_streams-1];
882 if ((uint64_t)atom.size > (1<<30))
885 av_free(st->codec->extradata);
886 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
887 if (!st->codec->extradata)
888 return AVERROR(ENOMEM);
889 st->codec->extradata_size = atom.size;
890 avio_read(pb, st->codec->extradata, atom.size);
895 * An strf atom is a BITMAPINFOHEADER struct. This struct is 40 bytes itself,
896 * but can have extradata appended at the end after the 40 bytes belonging
899 static int mov_read_strf(MOVContext *c, AVIOContext *pb, MOVAtom atom)
903 if (c->fc->nb_streams < 1)
907 st = c->fc->streams[c->fc->nb_streams-1];
909 if ((uint64_t)atom.size > (1<<30))
912 av_free(st->codec->extradata);
913 st->codec->extradata = av_mallocz(atom.size - 40 + FF_INPUT_BUFFER_PADDING_SIZE);
914 if (!st->codec->extradata)
915 return AVERROR(ENOMEM);
916 st->codec->extradata_size = atom.size - 40;
918 avio_read(pb, st->codec->extradata, atom.size - 40);
922 static int mov_read_stco(MOVContext *c, AVIOContext *pb, MOVAtom atom)
925 MOVStreamContext *sc;
926 unsigned int i, entries;
928 if (c->fc->nb_streams < 1)
930 st = c->fc->streams[c->fc->nb_streams-1];
933 avio_r8(pb); /* version */
934 avio_rb24(pb); /* flags */
936 entries = avio_rb32(pb);
938 if (entries >= UINT_MAX/sizeof(int64_t))
941 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
942 if (!sc->chunk_offsets)
943 return AVERROR(ENOMEM);
944 sc->chunk_count = entries;
946 if (atom.type == MKTAG('s','t','c','o'))
947 for (i=0; i<entries; i++)
948 sc->chunk_offsets[i] = avio_rb32(pb);
949 else if (atom.type == MKTAG('c','o','6','4'))
950 for (i=0; i<entries; i++)
951 sc->chunk_offsets[i] = avio_rb64(pb);
959 * Compute codec id for 'lpcm' tag.
960 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
962 enum CodecID ff_mov_get_lpcm_codec_id(int bps, int flags)
964 if (flags & 1) { // floating point
965 if (flags & 2) { // big endian
966 if (bps == 32) return CODEC_ID_PCM_F32BE;
967 else if (bps == 64) return CODEC_ID_PCM_F64BE;
969 if (bps == 32) return CODEC_ID_PCM_F32LE;
970 else if (bps == 64) return CODEC_ID_PCM_F64LE;
976 if (flags & 4) return CODEC_ID_PCM_S8;
977 else return CODEC_ID_PCM_U8;
978 else if (bps == 16) return CODEC_ID_PCM_S16BE;
979 else if (bps == 24) return CODEC_ID_PCM_S24BE;
980 else if (bps == 32) return CODEC_ID_PCM_S32BE;
983 if (flags & 4) return CODEC_ID_PCM_S8;
984 else return CODEC_ID_PCM_U8;
985 else if (bps == 16) return CODEC_ID_PCM_S16LE;
986 else if (bps == 24) return CODEC_ID_PCM_S24LE;
987 else if (bps == 32) return CODEC_ID_PCM_S32LE;
990 return CODEC_ID_NONE;
993 int ff_mov_read_stsd_entries(MOVContext *c, AVIOContext *pb, int entries)
996 MOVStreamContext *sc;
997 int j, pseudo_stream_id;
999 if (c->fc->nb_streams < 1)
1001 st = c->fc->streams[c->fc->nb_streams-1];
1004 for (pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
1005 //Parsing Sample description table
1008 MOVAtom a = { AV_RL32("stsd") };
1009 int64_t start_pos = avio_tell(pb);
1010 int size = avio_rb32(pb); /* size */
1011 uint32_t format = avio_rl32(pb); /* data format */
1014 avio_rb32(pb); /* reserved */
1015 avio_rb16(pb); /* reserved */
1016 dref_id = avio_rb16(pb);
1019 if (st->codec->codec_tag &&
1020 st->codec->codec_tag != format &&
1021 (c->fc->video_codec_id ? ff_codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
1022 : st->codec->codec_tag != MKTAG('j','p','e','g'))
1024 /* Multiple fourcc, we skip JPEG. This is not correct, we should
1025 * export it as a separate AVStream but this needs a few changes
1026 * in the MOV demuxer, patch welcome. */
1028 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
1029 avio_skip(pb, size - (avio_tell(pb) - start_pos));
1032 /* we cannot demux concatenated h264 streams because of different extradata */
1033 if (st->codec->codec_tag && st->codec->codec_tag == AV_RL32("avc1"))
1035 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
1036 sc->dref_id= dref_id;
1038 st->codec->codec_tag = format;
1039 id = ff_codec_get_id(codec_movaudio_tags, format);
1040 if (id<=0 && ((format&0xFFFF) == 'm'+('s'<<8) || (format&0xFFFF) == 'T'+('S'<<8)))
1041 id = ff_codec_get_id(ff_codec_wav_tags, av_bswap32(format)&0xFFFF);
1043 if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO && id > 0) {
1044 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
1045 } else if (st->codec->codec_type != AVMEDIA_TYPE_AUDIO && /* do not overwrite codec type */
1046 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
1047 id = ff_codec_get_id(codec_movvideo_tags, format);
1049 id = ff_codec_get_id(ff_codec_bmp_tags, format);
1051 st->codec->codec_type = AVMEDIA_TYPE_VIDEO;
1052 else if (st->codec->codec_type == AVMEDIA_TYPE_DATA){
1053 id = ff_codec_get_id(ff_codec_movsubtitle_tags, format);
1055 st->codec->codec_type = AVMEDIA_TYPE_SUBTITLE;
1059 av_dlog(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
1060 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
1061 (format >> 24) & 0xff, st->codec->codec_type);
1063 if (st->codec->codec_type==AVMEDIA_TYPE_VIDEO) {
1064 unsigned int color_depth, len;
1065 int color_greyscale;
1067 st->codec->codec_id = id;
1068 avio_rb16(pb); /* version */
1069 avio_rb16(pb); /* revision level */
1070 avio_rb32(pb); /* vendor */
1071 avio_rb32(pb); /* temporal quality */
1072 avio_rb32(pb); /* spatial quality */
1074 st->codec->width = avio_rb16(pb); /* width */
1075 st->codec->height = avio_rb16(pb); /* height */
1077 avio_rb32(pb); /* horiz resolution */
1078 avio_rb32(pb); /* vert resolution */
1079 avio_rb32(pb); /* data size, always 0 */
1080 avio_rb16(pb); /* frames per samples */
1082 len = avio_r8(pb); /* codec name, pascal string */
1085 mov_read_mac_string(c, pb, len, st->codec->codec_name, 32);
1087 avio_skip(pb, 31 - len);
1088 /* codec_tag YV12 triggers an UV swap in rawdec.c */
1089 if (!memcmp(st->codec->codec_name, "Planar Y'CbCr 8-bit 4:2:0", 25))
1090 st->codec->codec_tag=MKTAG('I', '4', '2', '0');
1092 st->codec->bits_per_coded_sample = avio_rb16(pb); /* depth */
1093 st->codec->color_table_id = avio_rb16(pb); /* colortable id */
1094 av_dlog(c->fc, "depth %d, ctab id %d\n",
1095 st->codec->bits_per_coded_sample, st->codec->color_table_id);
1096 /* figure out the palette situation */
1097 color_depth = st->codec->bits_per_coded_sample & 0x1F;
1098 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
1100 /* if the depth is 2, 4, or 8 bpp, file is palettized */
1101 if ((color_depth == 2) || (color_depth == 4) ||
1102 (color_depth == 8)) {
1103 /* for palette traversal */
1104 unsigned int color_start, color_count, color_end;
1105 unsigned char r, g, b;
1107 if (color_greyscale) {
1108 int color_index, color_dec;
1109 /* compute the greyscale palette */
1110 st->codec->bits_per_coded_sample = color_depth;
1111 color_count = 1 << color_depth;
1113 color_dec = 256 / (color_count - 1);
1114 for (j = 0; j < color_count; j++) {
1115 r = g = b = color_index;
1117 (r << 16) | (g << 8) | (b);
1118 color_index -= color_dec;
1119 if (color_index < 0)
1122 } else if (st->codec->color_table_id) {
1123 const uint8_t *color_table;
1124 /* if flag bit 3 is set, use the default palette */
1125 color_count = 1 << color_depth;
1126 if (color_depth == 2)
1127 color_table = ff_qt_default_palette_4;
1128 else if (color_depth == 4)
1129 color_table = ff_qt_default_palette_16;
1131 color_table = ff_qt_default_palette_256;
1133 for (j = 0; j < color_count; j++) {
1134 r = color_table[j * 3 + 0];
1135 g = color_table[j * 3 + 1];
1136 b = color_table[j * 3 + 2];
1138 (r << 16) | (g << 8) | (b);
1141 /* load the palette from the file */
1142 color_start = avio_rb32(pb);
1143 color_count = avio_rb16(pb);
1144 color_end = avio_rb16(pb);
1145 if ((color_start <= 255) &&
1146 (color_end <= 255)) {
1147 for (j = color_start; j <= color_end; j++) {
1148 /* each R, G, or B component is 16 bits;
1149 * only use the top 8 bits; skip alpha bytes
1160 (r << 16) | (g << 8) | (b);
1164 sc->has_palette = 1;
1166 } else if (st->codec->codec_type==AVMEDIA_TYPE_AUDIO) {
1167 int bits_per_sample, flags;
1168 uint16_t version = avio_rb16(pb);
1170 st->codec->codec_id = id;
1171 avio_rb16(pb); /* revision level */
1172 avio_rb32(pb); /* vendor */
1174 st->codec->channels = avio_rb16(pb); /* channel count */
1175 av_dlog(c->fc, "audio channels %d\n", st->codec->channels);
1176 st->codec->bits_per_coded_sample = avio_rb16(pb); /* sample size */
1178 sc->audio_cid = avio_rb16(pb);
1179 avio_rb16(pb); /* packet size = 0 */
1181 st->codec->sample_rate = ((avio_rb32(pb) >> 16));
1183 //Read QT version 1 fields. In version 0 these do not exist.
1184 av_dlog(c->fc, "version =%d, isom =%d\n",version,c->isom);
1187 sc->samples_per_frame = avio_rb32(pb);
1188 avio_rb32(pb); /* bytes per packet */
1189 sc->bytes_per_frame = avio_rb32(pb);
1190 avio_rb32(pb); /* bytes per sample */
1191 } else if (version==2) {
1192 avio_rb32(pb); /* sizeof struct only */
1193 st->codec->sample_rate = av_int2dbl(avio_rb64(pb)); /* float 64 */
1194 st->codec->channels = avio_rb32(pb);
1195 avio_rb32(pb); /* always 0x7F000000 */
1196 st->codec->bits_per_coded_sample = avio_rb32(pb); /* bits per channel if sound is uncompressed */
1197 flags = avio_rb32(pb); /* lpcm format specific flag */
1198 sc->bytes_per_frame = avio_rb32(pb); /* bytes per audio packet if constant */
1199 sc->samples_per_frame = avio_rb32(pb); /* lpcm frames per audio packet if constant */
1200 if (format == MKTAG('l','p','c','m'))
1201 st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, flags);
1205 switch (st->codec->codec_id) {
1206 case CODEC_ID_PCM_S8:
1207 case CODEC_ID_PCM_U8:
1208 if (st->codec->bits_per_coded_sample == 16)
1209 st->codec->codec_id = CODEC_ID_PCM_S16BE;
1211 case CODEC_ID_PCM_S16LE:
1212 case CODEC_ID_PCM_S16BE:
1213 if (st->codec->bits_per_coded_sample == 8)
1214 st->codec->codec_id = CODEC_ID_PCM_S8;
1215 else if (st->codec->bits_per_coded_sample == 24)
1216 st->codec->codec_id =
1217 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
1218 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
1220 /* set values for old format before stsd version 1 appeared */
1221 case CODEC_ID_MACE3:
1222 sc->samples_per_frame = 6;
1223 sc->bytes_per_frame = 2*st->codec->channels;
1225 case CODEC_ID_MACE6:
1226 sc->samples_per_frame = 6;
1227 sc->bytes_per_frame = 1*st->codec->channels;
1229 case CODEC_ID_ADPCM_IMA_QT:
1230 sc->samples_per_frame = 64;
1231 sc->bytes_per_frame = 34*st->codec->channels;
1234 sc->samples_per_frame = 160;
1235 sc->bytes_per_frame = 33;
1241 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
1242 if (bits_per_sample) {
1243 st->codec->bits_per_coded_sample = bits_per_sample;
1244 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
1246 } else if (st->codec->codec_type==AVMEDIA_TYPE_SUBTITLE){
1247 // ttxt stsd contains display flags, justification, background
1248 // color, fonts, and default styles, so fake an atom to read it
1249 MOVAtom fake_atom = { .size = size - (avio_tell(pb) - start_pos) };
1250 if (format != AV_RL32("mp4s")) // mp4s contains a regular esds atom
1251 mov_read_glbl(c, pb, fake_atom);
1252 st->codec->codec_id= id;
1253 st->codec->width = sc->width;
1254 st->codec->height = sc->height;
1256 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
1257 avio_skip(pb, size - (avio_tell(pb) - start_pos));
1259 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
1260 a.size = size - (avio_tell(pb) - start_pos);
1262 if (mov_read_default(c, pb, a) < 0)
1264 } else if (a.size > 0)
1265 avio_skip(pb, a.size);
1268 if (st->codec->codec_type==AVMEDIA_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
1269 st->codec->sample_rate= sc->time_scale;
1271 /* special codec parameters handling */
1272 switch (st->codec->codec_id) {
1273 #if CONFIG_DV_DEMUXER
1274 case CODEC_ID_DVAUDIO:
1275 c->dv_fctx = avformat_alloc_context();
1276 c->dv_demux = avpriv_dv_init_demux(c->dv_fctx);
1278 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
1281 sc->dv_audio_container = 1;
1282 st->codec->codec_id = CODEC_ID_PCM_S16LE;
1285 /* no ifdef since parameters are always those */
1286 case CODEC_ID_QCELP:
1287 // force sample rate for qcelp when not stored in mov
1288 if (st->codec->codec_tag != MKTAG('Q','c','l','p'))
1289 st->codec->sample_rate = 8000;
1290 st->codec->frame_size= 160;
1291 st->codec->channels= 1; /* really needed */
1293 case CODEC_ID_AMR_NB:
1294 case CODEC_ID_AMR_WB:
1295 st->codec->frame_size= sc->samples_per_frame;
1296 st->codec->channels= 1; /* really needed */
1297 /* force sample rate for amr, stsd in 3gp does not store sample rate */
1298 if (st->codec->codec_id == CODEC_ID_AMR_NB)
1299 st->codec->sample_rate = 8000;
1300 else if (st->codec->codec_id == CODEC_ID_AMR_WB)
1301 st->codec->sample_rate = 16000;
1305 st->codec->codec_type = AVMEDIA_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
1306 st->need_parsing = AVSTREAM_PARSE_FULL;
1309 case CODEC_ID_ADPCM_MS:
1310 case CODEC_ID_ADPCM_IMA_WAV:
1311 st->codec->frame_size = sc->samples_per_frame;
1312 st->codec->block_align = sc->bytes_per_frame;
1315 if (st->codec->extradata_size == 36) {
1316 st->codec->frame_size = AV_RB32(st->codec->extradata+12);
1317 st->codec->channels = AV_RB8 (st->codec->extradata+21);
1318 st->codec->sample_rate = AV_RB32(st->codec->extradata+32);
1328 static int mov_read_stsd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1332 avio_r8(pb); /* version */
1333 avio_rb24(pb); /* flags */
1334 entries = avio_rb32(pb);
1336 return ff_mov_read_stsd_entries(c, pb, entries);
1339 static int mov_read_stsc(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1342 MOVStreamContext *sc;
1343 unsigned int i, entries;
1345 if (c->fc->nb_streams < 1)
1347 st = c->fc->streams[c->fc->nb_streams-1];
1350 avio_r8(pb); /* version */
1351 avio_rb24(pb); /* flags */
1353 entries = avio_rb32(pb);
1355 av_dlog(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1357 if (entries >= UINT_MAX / sizeof(*sc->stsc_data))
1359 sc->stsc_data = av_malloc(entries * sizeof(*sc->stsc_data));
1361 return AVERROR(ENOMEM);
1362 sc->stsc_count = entries;
1364 for (i=0; i<entries; i++) {
1365 sc->stsc_data[i].first = avio_rb32(pb);
1366 sc->stsc_data[i].count = avio_rb32(pb);
1367 sc->stsc_data[i].id = avio_rb32(pb);
1372 static int mov_read_stps(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1375 MOVStreamContext *sc;
1376 unsigned i, entries;
1378 if (c->fc->nb_streams < 1)
1380 st = c->fc->streams[c->fc->nb_streams-1];
1383 avio_rb32(pb); // version + flags
1385 entries = avio_rb32(pb);
1386 if (entries >= UINT_MAX / sizeof(*sc->stps_data))
1388 sc->stps_data = av_malloc(entries * sizeof(*sc->stps_data));
1390 return AVERROR(ENOMEM);
1391 sc->stps_count = entries;
1393 for (i = 0; i < entries; i++) {
1394 sc->stps_data[i] = avio_rb32(pb);
1395 //av_dlog(c->fc, "stps %d\n", sc->stps_data[i]);
1401 static int mov_read_stss(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1404 MOVStreamContext *sc;
1405 unsigned int i, entries;
1407 if (c->fc->nb_streams < 1)
1409 st = c->fc->streams[c->fc->nb_streams-1];
1412 avio_r8(pb); /* version */
1413 avio_rb24(pb); /* flags */
1415 entries = avio_rb32(pb);
1417 av_dlog(c->fc, "keyframe_count = %d\n", entries);
1419 if (entries >= UINT_MAX / sizeof(int))
1421 sc->keyframes = av_malloc(entries * sizeof(int));
1423 return AVERROR(ENOMEM);
1424 sc->keyframe_count = entries;
1426 for (i=0; i<entries; i++) {
1427 sc->keyframes[i] = avio_rb32(pb);
1428 //av_dlog(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1433 static int mov_read_stsz(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1436 MOVStreamContext *sc;
1437 unsigned int i, entries, sample_size, field_size, num_bytes;
1441 if (c->fc->nb_streams < 1)
1443 st = c->fc->streams[c->fc->nb_streams-1];
1446 avio_r8(pb); /* version */
1447 avio_rb24(pb); /* flags */
1449 if (atom.type == MKTAG('s','t','s','z')) {
1450 sample_size = avio_rb32(pb);
1451 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1452 sc->sample_size = sample_size;
1456 avio_rb24(pb); /* reserved */
1457 field_size = avio_r8(pb);
1459 entries = avio_rb32(pb);
1461 av_dlog(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, entries);
1463 sc->sample_count = entries;
1467 if (field_size != 4 && field_size != 8 && field_size != 16 && field_size != 32) {
1468 av_log(c->fc, AV_LOG_ERROR, "Invalid sample field size %d\n", field_size);
1472 if (entries >= UINT_MAX / sizeof(int) || entries >= (UINT_MAX - 4) / field_size)
1474 sc->sample_sizes = av_malloc(entries * sizeof(int));
1475 if (!sc->sample_sizes)
1476 return AVERROR(ENOMEM);
1478 num_bytes = (entries*field_size+4)>>3;
1480 buf = av_malloc(num_bytes+FF_INPUT_BUFFER_PADDING_SIZE);
1482 av_freep(&sc->sample_sizes);
1483 return AVERROR(ENOMEM);
1486 if (avio_read(pb, buf, num_bytes) < num_bytes) {
1487 av_freep(&sc->sample_sizes);
1492 init_get_bits(&gb, buf, 8*num_bytes);
1494 for (i=0; i<entries; i++)
1495 sc->sample_sizes[i] = get_bits_long(&gb, field_size);
1501 static int mov_read_stts(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1504 MOVStreamContext *sc;
1505 unsigned int i, entries;
1507 int64_t total_sample_count=0;
1509 if (c->fc->nb_streams < 1)
1511 st = c->fc->streams[c->fc->nb_streams-1];
1514 avio_r8(pb); /* version */
1515 avio_rb24(pb); /* flags */
1516 entries = avio_rb32(pb);
1518 av_dlog(c->fc, "track[%i].stts.entries = %i\n",
1519 c->fc->nb_streams-1, entries);
1521 if (!entries || entries >= UINT_MAX / sizeof(*sc->stts_data))
1522 return AVERROR(EINVAL);
1524 sc->stts_data = av_malloc(entries * sizeof(*sc->stts_data));
1526 return AVERROR(ENOMEM);
1528 sc->stts_count = entries;
1530 for (i=0; i<entries; i++) {
1531 int sample_duration;
1534 sample_count=avio_rb32(pb);
1535 sample_duration = avio_rb32(pb);
1536 sc->stts_data[i].count= sample_count;
1537 sc->stts_data[i].duration= sample_duration;
1539 av_dlog(c->fc, "sample_count=%d, sample_duration=%d\n",
1540 sample_count, sample_duration);
1542 duration+=(int64_t)sample_duration*sample_count;
1543 total_sample_count+=sample_count;
1546 st->nb_frames= total_sample_count;
1548 st->duration= duration;
1552 static int mov_read_ctts(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1555 MOVStreamContext *sc;
1556 unsigned int i, entries;
1558 if (c->fc->nb_streams < 1)
1560 st = c->fc->streams[c->fc->nb_streams-1];
1563 avio_r8(pb); /* version */
1564 avio_rb24(pb); /* flags */
1565 entries = avio_rb32(pb);
1567 av_dlog(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1569 if (entries >= UINT_MAX / sizeof(*sc->ctts_data))
1571 sc->ctts_data = av_malloc(entries * sizeof(*sc->ctts_data));
1573 return AVERROR(ENOMEM);
1574 sc->ctts_count = entries;
1576 for (i=0; i<entries; i++) {
1577 int count =avio_rb32(pb);
1578 int duration =avio_rb32(pb);
1580 sc->ctts_data[i].count = count;
1581 sc->ctts_data[i].duration= duration;
1583 sc->dts_shift = FFMAX(sc->dts_shift, -duration);
1586 av_dlog(c->fc, "dts shift %d\n", sc->dts_shift);
1591 static void mov_build_index(MOVContext *mov, AVStream *st)
1593 MOVStreamContext *sc = st->priv_data;
1594 int64_t current_offset;
1595 int64_t current_dts = 0;
1596 unsigned int stts_index = 0;
1597 unsigned int stsc_index = 0;
1598 unsigned int stss_index = 0;
1599 unsigned int stps_index = 0;
1601 uint64_t stream_size = 0;
1603 /* adjust first dts according to edit list */
1604 if (sc->time_offset && mov->time_scale > 0) {
1605 if (sc->time_offset < 0)
1606 sc->time_offset = av_rescale(sc->time_offset, sc->time_scale, mov->time_scale);
1607 current_dts = -sc->time_offset;
1608 if (sc->ctts_data && sc->stts_data &&
1609 sc->ctts_data[0].duration / sc->stts_data[0].duration > 16) {
1610 /* more than 16 frames delay, dts are likely wrong
1611 this happens with files created by iMovie */
1613 st->codec->has_b_frames = 1;
1617 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1618 if (!(st->codec->codec_type == AVMEDIA_TYPE_AUDIO &&
1619 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1620 unsigned int current_sample = 0;
1621 unsigned int stts_sample = 0;
1622 unsigned int sample_size;
1623 unsigned int distance = 0;
1624 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1626 current_dts -= sc->dts_shift;
1628 if (sc->sample_count >= UINT_MAX / sizeof(*st->index_entries))
1630 st->index_entries = av_malloc(sc->sample_count*sizeof(*st->index_entries));
1631 if (!st->index_entries)
1633 st->index_entries_allocated_size = sc->sample_count*sizeof(*st->index_entries);
1635 for (i = 0; i < sc->chunk_count; i++) {
1636 current_offset = sc->chunk_offsets[i];
1637 while (stsc_index + 1 < sc->stsc_count &&
1638 i + 1 == sc->stsc_data[stsc_index + 1].first)
1640 for (j = 0; j < sc->stsc_data[stsc_index].count; j++) {
1642 if (current_sample >= sc->sample_count) {
1643 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1647 if (!sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index]) {
1649 if (stss_index + 1 < sc->keyframe_count)
1651 } else if (sc->stps_count && current_sample+key_off == sc->stps_data[stps_index]) {
1653 if (stps_index + 1 < sc->stps_count)
1658 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1659 if (sc->pseudo_stream_id == -1 ||
1660 sc->stsc_data[stsc_index].id - 1 == sc->pseudo_stream_id) {
1661 AVIndexEntry *e = &st->index_entries[st->nb_index_entries++];
1662 e->pos = current_offset;
1663 e->timestamp = current_dts;
1664 e->size = sample_size;
1665 e->min_distance = distance;
1666 e->flags = keyframe ? AVINDEX_KEYFRAME : 0;
1667 av_dlog(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1668 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1669 current_offset, current_dts, sample_size, distance, keyframe);
1672 current_offset += sample_size;
1673 stream_size += sample_size;
1674 current_dts += sc->stts_data[stts_index].duration;
1678 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1684 if (st->duration > 0)
1685 st->codec->bit_rate = stream_size*8*sc->time_scale/st->duration;
1687 unsigned chunk_samples, total = 0;
1689 // compute total chunk count
1690 for (i = 0; i < sc->stsc_count; i++) {
1691 unsigned count, chunk_count;
1693 chunk_samples = sc->stsc_data[i].count;
1694 if (sc->samples_per_frame && chunk_samples % sc->samples_per_frame) {
1695 av_log(mov->fc, AV_LOG_ERROR, "error unaligned chunk\n");
1699 if (sc->samples_per_frame >= 160) { // gsm
1700 count = chunk_samples / sc->samples_per_frame;
1701 } else if (sc->samples_per_frame > 1) {
1702 unsigned samples = (1024/sc->samples_per_frame)*sc->samples_per_frame;
1703 count = (chunk_samples+samples-1) / samples;
1705 count = (chunk_samples+1023) / 1024;
1708 if (i < sc->stsc_count - 1)
1709 chunk_count = sc->stsc_data[i+1].first - sc->stsc_data[i].first;
1711 chunk_count = sc->chunk_count - (sc->stsc_data[i].first - 1);
1712 total += chunk_count * count;
1715 av_dlog(mov->fc, "chunk count %d\n", total);
1716 if (total >= UINT_MAX / sizeof(*st->index_entries))
1718 st->index_entries = av_malloc(total*sizeof(*st->index_entries));
1719 if (!st->index_entries)
1721 st->index_entries_allocated_size = total*sizeof(*st->index_entries);
1724 for (i = 0; i < sc->chunk_count; i++) {
1725 current_offset = sc->chunk_offsets[i];
1726 if (stsc_index + 1 < sc->stsc_count &&
1727 i + 1 == sc->stsc_data[stsc_index + 1].first)
1729 chunk_samples = sc->stsc_data[stsc_index].count;
1731 while (chunk_samples > 0) {
1733 unsigned size, samples;
1735 if (sc->samples_per_frame >= 160) { // gsm
1736 samples = sc->samples_per_frame;
1737 size = sc->bytes_per_frame;
1739 if (sc->samples_per_frame > 1) {
1740 samples = FFMIN((1024 / sc->samples_per_frame)*
1741 sc->samples_per_frame, chunk_samples);
1742 size = (samples / sc->samples_per_frame) * sc->bytes_per_frame;
1744 samples = FFMIN(1024, chunk_samples);
1745 size = samples * sc->sample_size;
1749 if (st->nb_index_entries >= total) {
1750 av_log(mov->fc, AV_LOG_ERROR, "wrong chunk count %d\n", total);
1753 e = &st->index_entries[st->nb_index_entries++];
1754 e->pos = current_offset;
1755 e->timestamp = current_dts;
1757 e->min_distance = 0;
1758 e->flags = AVINDEX_KEYFRAME;
1759 av_dlog(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1760 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1763 current_offset += size;
1764 current_dts += samples;
1765 chunk_samples -= samples;
1771 static int mov_open_dref(AVIOContext **pb, char *src, MOVDref *ref)
1773 /* try relative path, we do not try the absolute because it can leak information about our
1774 system to an attacker */
1775 if (ref->nlvl_to > 0 && ref->nlvl_from > 0) {
1776 char filename[1024];
1780 /* find a source dir */
1781 src_path = strrchr(src, '/');
1787 /* find a next level down to target */
1788 for (i = 0, l = strlen(ref->path) - 1; l >= 0; l--)
1789 if (ref->path[l] == '/') {
1790 if (i == ref->nlvl_to - 1)
1796 /* compose filename if next level down to target was found */
1797 if (i == ref->nlvl_to - 1 && src_path - src < sizeof(filename)) {
1798 memcpy(filename, src, src_path - src);
1799 filename[src_path - src] = 0;
1801 for (i = 1; i < ref->nlvl_from; i++)
1802 av_strlcat(filename, "../", 1024);
1804 av_strlcat(filename, ref->path + l + 1, 1024);
1806 if (!avio_open(pb, filename, AVIO_FLAG_READ))
1811 return AVERROR(ENOENT);
1814 static int mov_read_trak(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1817 MOVStreamContext *sc;
1820 st = avformat_new_stream(c->fc, NULL);
1821 if (!st) return AVERROR(ENOMEM);
1822 st->id = c->fc->nb_streams;
1823 sc = av_mallocz(sizeof(MOVStreamContext));
1824 if (!sc) return AVERROR(ENOMEM);
1827 st->codec->codec_type = AVMEDIA_TYPE_DATA;
1828 sc->ffindex = st->index;
1830 if ((ret = mov_read_default(c, pb, atom)) < 0)
1834 if (sc->chunk_count && (!sc->stts_count || !sc->stsc_count ||
1835 (!sc->sample_size && !sc->sample_count))) {
1836 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1841 if (sc->time_scale <= 0) {
1842 av_log(c->fc, AV_LOG_WARNING, "stream %d, timescale not set\n", st->index);
1843 sc->time_scale = c->time_scale;
1844 if (sc->time_scale <= 0)
1848 av_set_pts_info(st, 64, 1, sc->time_scale);
1850 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO &&
1851 !st->codec->frame_size && sc->stts_count == 1) {
1852 st->codec->frame_size = av_rescale(sc->stts_data[0].duration,
1853 st->codec->sample_rate, sc->time_scale);
1854 av_dlog(c->fc, "frame size %d\n", st->codec->frame_size);
1857 mov_build_index(c, st);
1859 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1860 MOVDref *dref = &sc->drefs[sc->dref_id - 1];
1861 if (mov_open_dref(&sc->pb, c->fc->filename, dref) < 0)
1862 av_log(c->fc, AV_LOG_ERROR,
1863 "stream %d, error opening alias: path='%s', dir='%s', "
1864 "filename='%s', volume='%s', nlvl_from=%d, nlvl_to=%d\n",
1865 st->index, dref->path, dref->dir, dref->filename,
1866 dref->volume, dref->nlvl_from, dref->nlvl_to);
1870 if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
1871 if (!st->sample_aspect_ratio.num &&
1872 (st->codec->width != sc->width || st->codec->height != sc->height)) {
1873 st->sample_aspect_ratio = av_d2q(((double)st->codec->height * sc->width) /
1874 ((double)st->codec->width * sc->height), INT_MAX);
1877 av_reduce(&st->avg_frame_rate.num, &st->avg_frame_rate.den,
1878 sc->time_scale*st->nb_frames, st->duration, INT_MAX);
1880 if (sc->stts_count == 1 || (sc->stts_count == 2 && sc->stts_data[1].count == 1))
1881 av_reduce(&st->r_frame_rate.num, &st->r_frame_rate.den,
1882 sc->time_scale, sc->stts_data[0].duration, INT_MAX);
1885 switch (st->codec->codec_id) {
1886 #if CONFIG_H261_DECODER
1889 #if CONFIG_H263_DECODER
1892 #if CONFIG_H264_DECODER
1895 #if CONFIG_MPEG4_DECODER
1896 case CODEC_ID_MPEG4:
1898 st->codec->width = 0; /* let decoder init width/height */
1899 st->codec->height= 0;
1903 /* Do not need those anymore. */
1904 av_freep(&sc->chunk_offsets);
1905 av_freep(&sc->stsc_data);
1906 av_freep(&sc->sample_sizes);
1907 av_freep(&sc->keyframes);
1908 av_freep(&sc->stts_data);
1909 av_freep(&sc->stps_data);
1914 static int mov_read_ilst(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1917 c->itunes_metadata = 1;
1918 ret = mov_read_default(c, pb, atom);
1919 c->itunes_metadata = 0;
1923 static int mov_read_meta(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1925 while (atom.size > 8) {
1926 uint32_t tag = avio_rl32(pb);
1928 if (tag == MKTAG('h','d','l','r')) {
1929 avio_seek(pb, -8, SEEK_CUR);
1931 return mov_read_default(c, pb, atom);
1937 static int mov_read_tkhd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
1942 int64_t disp_transform[2];
1943 int display_matrix[3][2];
1945 MOVStreamContext *sc;
1948 if (c->fc->nb_streams < 1)
1950 st = c->fc->streams[c->fc->nb_streams-1];
1953 version = avio_r8(pb);
1954 avio_rb24(pb); /* flags */
1956 MOV_TRACK_ENABLED 0x0001
1957 MOV_TRACK_IN_MOVIE 0x0002
1958 MOV_TRACK_IN_PREVIEW 0x0004
1959 MOV_TRACK_IN_POSTER 0x0008
1966 avio_rb32(pb); /* creation time */
1967 avio_rb32(pb); /* modification time */
1969 st->id = (int)avio_rb32(pb); /* track id (NOT 0 !)*/
1970 avio_rb32(pb); /* reserved */
1972 /* highlevel (considering edits) duration in movie timebase */
1973 (version == 1) ? avio_rb64(pb) : avio_rb32(pb);
1974 avio_rb32(pb); /* reserved */
1975 avio_rb32(pb); /* reserved */
1977 avio_rb16(pb); /* layer */
1978 avio_rb16(pb); /* alternate group */
1979 avio_rb16(pb); /* volume */
1980 avio_rb16(pb); /* reserved */
1982 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1983 // they're kept in fixed point format through all calculations
1984 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1985 for (i = 0; i < 3; i++) {
1986 display_matrix[i][0] = avio_rb32(pb); // 16.16 fixed point
1987 display_matrix[i][1] = avio_rb32(pb); // 16.16 fixed point
1988 avio_rb32(pb); // 2.30 fixed point (not used)
1991 width = avio_rb32(pb); // 16.16 fixed point track width
1992 height = avio_rb32(pb); // 16.16 fixed point track height
1993 sc->width = width >> 16;
1994 sc->height = height >> 16;
1996 // transform the display width/height according to the matrix
1997 // skip this if the display matrix is the default identity matrix
1998 // or if it is rotating the picture, ex iPhone 3GS
1999 // to keep the same scale, use [width height 1<<16]
2000 if (width && height &&
2001 ((display_matrix[0][0] != 65536 ||
2002 display_matrix[1][1] != 65536) &&
2003 !display_matrix[0][1] &&
2004 !display_matrix[1][0] &&
2005 !display_matrix[2][0] && !display_matrix[2][1])) {
2006 for (i = 0; i < 2; i++)
2008 (int64_t) width * display_matrix[0][i] +
2009 (int64_t) height * display_matrix[1][i] +
2010 ((int64_t) display_matrix[2][i] << 16);
2012 //sample aspect ratio is new width/height divided by old width/height
2013 st->sample_aspect_ratio = av_d2q(
2014 ((double) disp_transform[0] * height) /
2015 ((double) disp_transform[1] * width), INT_MAX);
2020 static int mov_read_tfhd(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2022 MOVFragment *frag = &c->fragment;
2023 MOVTrackExt *trex = NULL;
2024 int flags, track_id, i;
2026 avio_r8(pb); /* version */
2027 flags = avio_rb24(pb);
2029 track_id = avio_rb32(pb);
2032 frag->track_id = track_id;
2033 for (i = 0; i < c->trex_count; i++)
2034 if (c->trex_data[i].track_id == frag->track_id) {
2035 trex = &c->trex_data[i];
2039 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
2043 if (flags & 0x01) frag->base_data_offset = avio_rb64(pb);
2044 else frag->base_data_offset = frag->moof_offset;
2045 if (flags & 0x02) frag->stsd_id = avio_rb32(pb);
2046 else frag->stsd_id = trex->stsd_id;
2048 frag->duration = flags & 0x08 ? avio_rb32(pb) : trex->duration;
2049 frag->size = flags & 0x10 ? avio_rb32(pb) : trex->size;
2050 frag->flags = flags & 0x20 ? avio_rb32(pb) : trex->flags;
2051 av_dlog(c->fc, "frag flags 0x%x\n", frag->flags);
2055 static int mov_read_chap(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2057 c->chapter_track = avio_rb32(pb);
2061 static int mov_read_trex(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2065 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
2067 trex = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
2069 return AVERROR(ENOMEM);
2070 c->trex_data = trex;
2071 trex = &c->trex_data[c->trex_count++];
2072 avio_r8(pb); /* version */
2073 avio_rb24(pb); /* flags */
2074 trex->track_id = avio_rb32(pb);
2075 trex->stsd_id = avio_rb32(pb);
2076 trex->duration = avio_rb32(pb);
2077 trex->size = avio_rb32(pb);
2078 trex->flags = avio_rb32(pb);
2082 static int mov_read_trun(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2084 MOVFragment *frag = &c->fragment;
2085 AVStream *st = NULL;
2086 MOVStreamContext *sc;
2090 int data_offset = 0;
2091 unsigned entries, first_sample_flags = frag->flags;
2092 int flags, distance, i;
2094 for (i = 0; i < c->fc->nb_streams; i++) {
2095 if (c->fc->streams[i]->id == frag->track_id) {
2096 st = c->fc->streams[i];
2101 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding track id %d\n", frag->track_id);
2105 if (sc->pseudo_stream_id+1 != frag->stsd_id)
2107 avio_r8(pb); /* version */
2108 flags = avio_rb24(pb);
2109 entries = avio_rb32(pb);
2110 av_dlog(c->fc, "flags 0x%x entries %d\n", flags, entries);
2112 /* Always assume the presence of composition time offsets.
2113 * Without this assumption, for instance, we cannot deal with a track in fragmented movies that meet the following.
2114 * 1) in the initial movie, there are no samples.
2115 * 2) in the first movie fragment, there is only one sample without composition time offset.
2116 * 3) in the subsequent movie fragments, there are samples with composition time offset. */
2117 if (!sc->ctts_count && sc->sample_count)
2119 /* Complement ctts table if moov atom doesn't have ctts atom. */
2120 ctts_data = av_malloc(sizeof(*sc->ctts_data));
2122 return AVERROR(ENOMEM);
2123 sc->ctts_data = ctts_data;
2124 sc->ctts_data[sc->ctts_count].count = sc->sample_count;
2125 sc->ctts_data[sc->ctts_count].duration = 0;
2128 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
2130 ctts_data = av_realloc(sc->ctts_data,
2131 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
2133 return AVERROR(ENOMEM);
2134 sc->ctts_data = ctts_data;
2136 if (flags & 0x001) data_offset = avio_rb32(pb);
2137 if (flags & 0x004) first_sample_flags = avio_rb32(pb);
2138 dts = st->duration - sc->time_offset;
2139 offset = frag->base_data_offset + data_offset;
2141 av_dlog(c->fc, "first sample flags 0x%x\n", first_sample_flags);
2142 for (i = 0; i < entries; i++) {
2143 unsigned sample_size = frag->size;
2144 int sample_flags = i ? frag->flags : first_sample_flags;
2145 unsigned sample_duration = frag->duration;
2148 if (flags & 0x100) sample_duration = avio_rb32(pb);
2149 if (flags & 0x200) sample_size = avio_rb32(pb);
2150 if (flags & 0x400) sample_flags = avio_rb32(pb);
2151 sc->ctts_data[sc->ctts_count].count = 1;
2152 sc->ctts_data[sc->ctts_count].duration = (flags & 0x800) ? avio_rb32(pb) : 0;
2154 if ((keyframe = st->codec->codec_type == AVMEDIA_TYPE_AUDIO ||
2155 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
2157 av_add_index_entry(st, offset, dts, sample_size, distance,
2158 keyframe ? AVINDEX_KEYFRAME : 0);
2159 av_dlog(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
2160 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
2161 offset, dts, sample_size, distance, keyframe);
2163 dts += sample_duration;
2164 offset += sample_size;
2166 frag->moof_offset = offset;
2167 st->duration = dts + sc->time_offset;
2171 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
2172 /* like the files created with Adobe Premiere 5.0, for samples see */
2173 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
2174 static int mov_read_wide(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2179 return 0; /* continue */
2180 if (avio_rb32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
2181 avio_skip(pb, atom.size - 4);
2184 atom.type = avio_rl32(pb);
2186 if (atom.type != MKTAG('m','d','a','t')) {
2187 avio_skip(pb, atom.size);
2190 err = mov_read_mdat(c, pb, atom);
2194 static int mov_read_cmov(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2199 uint8_t *moov_data; /* uncompressed data */
2200 long cmov_len, moov_len;
2203 avio_rb32(pb); /* dcom atom */
2204 if (avio_rl32(pb) != MKTAG('d','c','o','m'))
2206 if (avio_rl32(pb) != MKTAG('z','l','i','b')) {
2207 av_log(c->fc, AV_LOG_ERROR, "unknown compression for cmov atom !");
2210 avio_rb32(pb); /* cmvd atom */
2211 if (avio_rl32(pb) != MKTAG('c','m','v','d'))
2213 moov_len = avio_rb32(pb); /* uncompressed size */
2214 cmov_len = atom.size - 6 * 4;
2216 cmov_data = av_malloc(cmov_len);
2218 return AVERROR(ENOMEM);
2219 moov_data = av_malloc(moov_len);
2222 return AVERROR(ENOMEM);
2224 avio_read(pb, cmov_data, cmov_len);
2225 if (uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
2226 goto free_and_return;
2227 if (ffio_init_context(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
2228 goto free_and_return;
2229 atom.type = MKTAG('m','o','o','v');
2230 atom.size = moov_len;
2231 ret = mov_read_default(c, &ctx, atom);
2237 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
2242 /* edit list atom */
2243 static int mov_read_elst(MOVContext *c, AVIOContext *pb, MOVAtom atom)
2245 MOVStreamContext *sc;
2246 int i, edit_count, version;
2248 if (c->fc->nb_streams < 1)
2250 sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
2252 version = avio_r8(pb); /* version */
2253 avio_rb24(pb); /* flags */
2254 edit_count = avio_rb32(pb); /* entries */
2256 if ((uint64_t)edit_count*12+8 > atom.size)
2259 for (i=0; i<edit_count; i++){
2263 duration = avio_rb64(pb);
2264 time = avio_rb64(pb);
2266 duration = avio_rb32(pb); /* segment duration */
2267 time = (int32_t)avio_rb32(pb); /* media time */
2269 avio_rb32(pb); /* Media rate */
2270 if (i == 0 && time >= -1) {
2271 sc->time_offset = time != -1 ? time : -duration;
2276 av_log(c->fc, AV_LOG_WARNING, "multiple edit list entries, "
2277 "a/v desync might occur, patch welcome\n");
2279 av_dlog(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, edit_count);
2283 static const MOVParseTableEntry mov_default_parse_table[] = {
2284 { MKTAG('a','v','s','s'), mov_read_extradata },
2285 { MKTAG('c','h','p','l'), mov_read_chpl },
2286 { MKTAG('c','o','6','4'), mov_read_stco },
2287 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
2288 { MKTAG('d','i','n','f'), mov_read_default },
2289 { MKTAG('d','r','e','f'), mov_read_dref },
2290 { MKTAG('e','d','t','s'), mov_read_default },
2291 { MKTAG('e','l','s','t'), mov_read_elst },
2292 { MKTAG('e','n','d','a'), mov_read_enda },
2293 { MKTAG('f','i','e','l'), mov_read_extradata },
2294 { MKTAG('f','t','y','p'), mov_read_ftyp },
2295 { MKTAG('g','l','b','l'), mov_read_glbl },
2296 { MKTAG('h','d','l','r'), mov_read_hdlr },
2297 { MKTAG('i','l','s','t'), mov_read_ilst },
2298 { MKTAG('j','p','2','h'), mov_read_extradata },
2299 { MKTAG('m','d','a','t'), mov_read_mdat },
2300 { MKTAG('m','d','h','d'), mov_read_mdhd },
2301 { MKTAG('m','d','i','a'), mov_read_default },
2302 { MKTAG('m','e','t','a'), mov_read_meta },
2303 { MKTAG('m','i','n','f'), mov_read_default },
2304 { MKTAG('m','o','o','f'), mov_read_moof },
2305 { MKTAG('m','o','o','v'), mov_read_moov },
2306 { MKTAG('m','v','e','x'), mov_read_default },
2307 { MKTAG('m','v','h','d'), mov_read_mvhd },
2308 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
2309 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
2310 { MKTAG('a','v','c','C'), mov_read_glbl },
2311 { MKTAG('p','a','s','p'), mov_read_pasp },
2312 { MKTAG('s','t','b','l'), mov_read_default },
2313 { MKTAG('s','t','c','o'), mov_read_stco },
2314 { MKTAG('s','t','p','s'), mov_read_stps },
2315 { MKTAG('s','t','r','f'), mov_read_strf },
2316 { MKTAG('s','t','s','c'), mov_read_stsc },
2317 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
2318 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
2319 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
2320 { MKTAG('s','t','t','s'), mov_read_stts },
2321 { MKTAG('s','t','z','2'), mov_read_stsz }, /* compact sample size */
2322 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
2323 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
2324 { MKTAG('t','r','a','k'), mov_read_trak },
2325 { MKTAG('t','r','a','f'), mov_read_default },
2326 { MKTAG('t','r','e','f'), mov_read_default },
2327 { MKTAG('c','h','a','p'), mov_read_chap },
2328 { MKTAG('t','r','e','x'), mov_read_trex },
2329 { MKTAG('t','r','u','n'), mov_read_trun },
2330 { MKTAG('u','d','t','a'), mov_read_default },
2331 { MKTAG('w','a','v','e'), mov_read_wave },
2332 { MKTAG('e','s','d','s'), mov_read_esds },
2333 { MKTAG('d','a','c','3'), mov_read_dac3 }, /* AC-3 info */
2334 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
2335 { MKTAG('w','f','e','x'), mov_read_wfex },
2336 { MKTAG('c','m','o','v'), mov_read_cmov },
2340 static int mov_probe(AVProbeData *p)
2342 unsigned int offset;
2346 /* check file header */
2349 /* ignore invalid offset */
2350 if ((offset + 8) > (unsigned int)p->buf_size)
2352 tag = AV_RL32(p->buf + offset + 4);
2354 /* check for obvious tags */
2355 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
2356 case MKTAG('m','o','o','v'):
2357 case MKTAG('m','d','a','t'):
2358 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
2359 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
2360 case MKTAG('f','t','y','p'):
2361 return AVPROBE_SCORE_MAX;
2362 /* those are more common words, so rate then a bit less */
2363 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
2364 case MKTAG('w','i','d','e'):
2365 case MKTAG('f','r','e','e'):
2366 case MKTAG('j','u','n','k'):
2367 case MKTAG('p','i','c','t'):
2368 return AVPROBE_SCORE_MAX - 5;
2369 case MKTAG(0x82,0x82,0x7f,0x7d):
2370 case MKTAG('s','k','i','p'):
2371 case MKTAG('u','u','i','d'):
2372 case MKTAG('p','r','f','l'):
2373 offset = AV_RB32(p->buf+offset) + offset;
2374 /* if we only find those cause probedata is too small at least rate them */
2375 score = AVPROBE_SCORE_MAX - 50;
2378 /* unrecognized tag */
2384 // must be done after parsing all trak because there's no order requirement
2385 static void mov_read_chapters(AVFormatContext *s)
2387 MOVContext *mov = s->priv_data;
2388 AVStream *st = NULL;
2389 MOVStreamContext *sc;
2393 for (i = 0; i < s->nb_streams; i++)
2394 if (s->streams[i]->id == mov->chapter_track) {
2399 av_log(s, AV_LOG_ERROR, "Referenced QT chapter track not found\n");
2403 st->discard = AVDISCARD_ALL;
2405 cur_pos = avio_tell(sc->pb);
2407 for (i = 0; i < st->nb_index_entries; i++) {
2408 AVIndexEntry *sample = &st->index_entries[i];
2409 int64_t end = i+1 < st->nb_index_entries ? st->index_entries[i+1].timestamp : st->duration;
2414 if (avio_seek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
2415 av_log(s, AV_LOG_ERROR, "Chapter %d not found in file\n", i);
2419 // the first two bytes are the length of the title
2420 len = avio_rb16(sc->pb);
2421 if (len > sample->size-2)
2423 title_len = 2*len + 1;
2424 if (!(title = av_mallocz(title_len)))
2427 // The samples could theoretically be in any encoding if there's an encd
2428 // atom following, but in practice are only utf-8 or utf-16, distinguished
2429 // instead by the presence of a BOM
2433 ch = avio_rb16(sc->pb);
2435 avio_get_str16be(sc->pb, len, title, title_len);
2436 else if (ch == 0xfffe)
2437 avio_get_str16le(sc->pb, len, title, title_len);
2440 if (len == 1 || len == 2)
2443 avio_get_str(sc->pb, len - 2, title + 2, title_len - 2);
2447 avpriv_new_chapter(s, i, st->time_base, sample->timestamp, end, title);
2451 avio_seek(sc->pb, cur_pos, SEEK_SET);
2454 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
2456 MOVContext *mov = s->priv_data;
2457 AVIOContext *pb = s->pb;
2459 MOVAtom atom = { AV_RL32("root") };
2462 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
2464 atom.size = avio_size(pb);
2466 atom.size = INT64_MAX;
2468 /* check MOV header */
2469 if ((err = mov_read_default(mov, pb, atom)) < 0) {
2470 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
2473 if (!mov->found_moov) {
2474 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
2477 av_dlog(mov->fc, "on_parse_exit_offset=%"PRId64"\n", avio_tell(pb));
2479 if (pb->seekable && mov->chapter_track > 0)
2480 mov_read_chapters(s);
2485 static AVIndexEntry *mov_find_next_sample(AVFormatContext *s, AVStream **st)
2487 AVIndexEntry *sample = NULL;
2488 int64_t best_dts = INT64_MAX;
2490 for (i = 0; i < s->nb_streams; i++) {
2491 AVStream *avst = s->streams[i];
2492 MOVStreamContext *msc = avst->priv_data;
2493 if (msc->pb && msc->current_sample < avst->nb_index_entries) {
2494 AVIndexEntry *current_sample = &avst->index_entries[msc->current_sample];
2495 int64_t dts = av_rescale(current_sample->timestamp, AV_TIME_BASE, msc->time_scale);
2496 av_dlog(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
2497 if (!sample || (!s->pb->seekable && current_sample->pos < sample->pos) ||
2499 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
2500 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
2501 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
2502 sample = current_sample;
2511 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
2513 MOVContext *mov = s->priv_data;
2514 MOVStreamContext *sc;
2515 AVIndexEntry *sample;
2516 AVStream *st = NULL;
2519 sample = mov_find_next_sample(s, &st);
2521 mov->found_mdat = 0;
2522 if (s->pb->seekable||
2523 mov_read_default(mov, s->pb, (MOVAtom){ AV_RL32("root"), INT64_MAX }) < 0 ||
2526 av_dlog(s, "read fragments, offset 0x%"PRIx64"\n", avio_tell(s->pb));
2530 /* must be done just before reading, to avoid infinite loop on sample */
2531 sc->current_sample++;
2533 if (st->discard != AVDISCARD_ALL) {
2534 if (avio_seek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
2535 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
2536 sc->ffindex, sample->pos);
2539 ret = av_get_packet(sc->pb, pkt, sample->size);
2542 if (sc->has_palette) {
2545 pal = av_packet_new_side_data(pkt, AV_PKT_DATA_PALETTE, AVPALETTE_SIZE);
2547 av_log(mov->fc, AV_LOG_ERROR, "Cannot append palette to packet\n");
2549 memcpy(pal, sc->palette, AVPALETTE_SIZE);
2550 sc->has_palette = 0;
2553 #if CONFIG_DV_DEMUXER
2554 if (mov->dv_demux && sc->dv_audio_container) {
2555 avpriv_dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
2558 ret = avpriv_dv_get_packet(mov->dv_demux, pkt);
2565 pkt->stream_index = sc->ffindex;
2566 pkt->dts = sample->timestamp;
2567 if (sc->ctts_data) {
2568 pkt->pts = pkt->dts + sc->dts_shift + sc->ctts_data[sc->ctts_index].duration;
2569 /* update ctts context */
2571 if (sc->ctts_index < sc->ctts_count &&
2572 sc->ctts_data[sc->ctts_index].count == sc->ctts_sample) {
2574 sc->ctts_sample = 0;
2577 pkt->dts = AV_NOPTS_VALUE;
2579 int64_t next_dts = (sc->current_sample < st->nb_index_entries) ?
2580 st->index_entries[sc->current_sample].timestamp : st->duration;
2581 pkt->duration = next_dts - pkt->dts;
2582 pkt->pts = pkt->dts;
2584 if (st->discard == AVDISCARD_ALL)
2586 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? AV_PKT_FLAG_KEY : 0;
2587 pkt->pos = sample->pos;
2588 av_dlog(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
2589 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
2593 static int mov_seek_stream(AVFormatContext *s, AVStream *st, int64_t timestamp, int flags)
2595 MOVStreamContext *sc = st->priv_data;
2596 int sample, time_sample;
2599 sample = av_index_search_timestamp(st, timestamp, flags);
2600 av_dlog(s, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
2601 if (sample < 0 && st->nb_index_entries && timestamp < st->index_entries[0].timestamp)
2603 if (sample < 0) /* not sure what to do */
2605 sc->current_sample = sample;
2606 av_dlog(s, "stream %d, found sample %d\n", st->index, sc->current_sample);
2607 /* adjust ctts index */
2608 if (sc->ctts_data) {
2610 for (i = 0; i < sc->ctts_count; i++) {
2611 int next = time_sample + sc->ctts_data[i].count;
2612 if (next > sc->current_sample) {
2614 sc->ctts_sample = sc->current_sample - time_sample;
2623 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
2626 int64_t seek_timestamp, timestamp;
2630 if (stream_index >= s->nb_streams)
2632 if (sample_time < 0)
2635 st = s->streams[stream_index];
2636 sample = mov_seek_stream(s, st, sample_time, flags);
2640 /* adjust seek timestamp to found sample timestamp */
2641 seek_timestamp = st->index_entries[sample].timestamp;
2643 for (i = 0; i < s->nb_streams; i++) {
2645 if (stream_index == i)
2648 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
2649 mov_seek_stream(s, st, timestamp, flags);
2654 static int mov_read_close(AVFormatContext *s)
2656 MOVContext *mov = s->priv_data;
2659 for (i = 0; i < s->nb_streams; i++) {
2660 AVStream *st = s->streams[i];
2661 MOVStreamContext *sc = st->priv_data;
2663 av_freep(&sc->ctts_data);
2664 for (j = 0; j < sc->drefs_count; j++) {
2665 av_freep(&sc->drefs[j].path);
2666 av_freep(&sc->drefs[j].dir);
2668 av_freep(&sc->drefs);
2669 if (sc->pb && sc->pb != s->pb)
2673 if (mov->dv_demux) {
2674 for (i = 0; i < mov->dv_fctx->nb_streams; i++) {
2675 av_freep(&mov->dv_fctx->streams[i]->codec);
2676 av_freep(&mov->dv_fctx->streams[i]);
2678 av_freep(&mov->dv_fctx);
2679 av_freep(&mov->dv_demux);
2682 av_freep(&mov->trex_data);
2687 AVInputFormat ff_mov_demuxer = {
2688 .name = "mov,mp4,m4a,3gp,3g2,mj2",
2689 .long_name = NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),
2690 .priv_data_size = sizeof(MOVContext),
2691 .read_probe = mov_probe,
2692 .read_header = mov_read_header,
2693 .read_packet = mov_read_packet,
2694 .read_close = mov_read_close,
2695 .read_seek = mov_read_seek,