Modify the Ogg/Speex demuxer and the libspeex decoder so that they always treat
[FFMpeg-mirror/lagarith.git] / libavformat / mov.c
blob74698e05479cec2a4a648e7fca03af58df95ddca
1 /*
2 * MOV demuxer
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 FFmpeg.
8 * FFmpeg 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 * FFmpeg 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 FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 #include <limits.h>
25 //#define DEBUG
26 //#define DEBUG_METADATA
27 //#define MOV_EXPORT_ALL_METADATA
29 #include "libavutil/intreadwrite.h"
30 #include "libavutil/avstring.h"
31 #include "avformat.h"
32 #include "riff.h"
33 #include "isom.h"
34 #include "libavcodec/mpeg4audio.h"
35 #include "libavcodec/mpegaudiodata.h"
36 #include "libavcodec/get_bits.h"
38 #if CONFIG_ZLIB
39 #include <zlib.h>
40 #endif
43 * First version by Francois Revol revol@free.fr
44 * Seek function by Gael Chardon gael.dev@4now.net
46 * Features and limitations:
47 * - reads most of the QT files I have (at least the structure),
48 * Sample QuickTime files with mp3 audio can be found at: http://www.3ivx.com/showcase.html
49 * - the code is quite ugly... maybe I won't do it recursive next time :-)
51 * Funny I didn't know about http://sourceforge.net/projects/qt-ffmpeg/
52 * when coding this :) (it's a writer anyway)
54 * Reference documents:
55 * http://www.geocities.com/xhelmboyx/quicktime/formats/qtm-layout.txt
56 * Apple:
57 * http://developer.apple.com/documentation/QuickTime/QTFF/
58 * http://developer.apple.com/documentation/QuickTime/QTFF/qtff.pdf
59 * QuickTime is a trademark of Apple (AFAIK :))
62 #include "qtpalette.h"
65 #undef NDEBUG
66 #include <assert.h>
68 /* XXX: it's the first time I make a recursive parser I think... sorry if it's ugly :P */
70 /* those functions parse an atom */
71 /* return code:
72 0: continue to parse next atom
73 <0: error occurred, exit
75 /* links atom IDs to parse functions */
76 typedef struct MOVParseTableEntry {
77 uint32_t type;
78 int (*parse)(MOVContext *ctx, ByteIOContext *pb, MOVAtom atom);
79 } MOVParseTableEntry;
81 static const MOVParseTableEntry mov_default_parse_table[];
83 static int mov_metadata_trkn(MOVContext *c, ByteIOContext *pb, unsigned len)
85 char buf[16];
87 get_be16(pb); // unknown
88 snprintf(buf, sizeof(buf), "%d", get_be16(pb));
89 av_metadata_set(&c->fc->metadata, "track", buf);
91 get_be16(pb); // total tracks
93 return 0;
96 static int mov_read_udta_string(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
98 #ifdef MOV_EXPORT_ALL_METADATA
99 char tmp_key[5];
100 #endif
101 char str[1024], key2[16], language[4] = {0};
102 const char *key = NULL;
103 uint16_t str_size;
104 int (*parse)(MOVContext*, ByteIOContext*, unsigned) = NULL;
106 switch (atom.type) {
107 case MKTAG(0xa9,'n','a','m'): key = "title"; break;
108 case MKTAG(0xa9,'a','u','t'):
109 case MKTAG(0xa9,'A','R','T'):
110 case MKTAG(0xa9,'w','r','t'): key = "author"; break;
111 case MKTAG(0xa9,'c','p','y'): key = "copyright"; break;
112 case MKTAG(0xa9,'c','m','t'):
113 case MKTAG(0xa9,'i','n','f'): key = "comment"; break;
114 case MKTAG(0xa9,'a','l','b'): key = "album"; break;
115 case MKTAG(0xa9,'d','a','y'): key = "year"; break;
116 case MKTAG(0xa9,'g','e','n'): key = "genre"; break;
117 case MKTAG(0xa9,'t','o','o'):
118 case MKTAG(0xa9,'e','n','c'): key = "muxer"; break;
119 case MKTAG( 't','r','k','n'): key = "track";
120 parse = mov_metadata_trkn; break;
123 if (c->itunes_metadata && atom.size > 8) {
124 int data_size = get_be32(pb);
125 int tag = get_le32(pb);
126 if (tag == MKTAG('d','a','t','a')) {
127 get_be32(pb); // type
128 get_be32(pb); // unknown
129 str_size = data_size - 16;
130 atom.size -= 16;
131 } else return 0;
132 } else if (atom.size > 4 && key && !c->itunes_metadata) {
133 str_size = get_be16(pb); // string length
134 ff_mov_lang_to_iso639(get_be16(pb), language);
135 atom.size -= 4;
136 } else
137 str_size = atom.size;
139 #ifdef MOV_EXPORT_ALL_METADATA
140 if (!key) {
141 snprintf(tmp_key, 5, "%.4s", (char*)&atom.type);
142 key = tmp_key;
144 #endif
146 if (!key)
147 return 0;
148 if (atom.size < 0)
149 return -1;
151 str_size = FFMIN3(sizeof(str)-1, str_size, atom.size);
153 if (parse)
154 parse(c, pb, str_size);
155 else {
156 get_buffer(pb, str, str_size);
157 str[str_size] = 0;
158 av_metadata_set(&c->fc->metadata, key, str);
159 if (*language && strcmp(language, "und")) {
160 snprintf(key2, sizeof(key2), "%s-%s", key, language);
161 av_metadata_set(&c->fc->metadata, key2, str);
164 #ifdef DEBUG_METADATA
165 av_log(c->fc, AV_LOG_DEBUG, "lang \"%3s\" ", language);
166 av_log(c->fc, AV_LOG_DEBUG, "tag \"%s\" value \"%s\" atom \"%.4s\" %d %lld\n",
167 key, str, (char*)&atom.type, str_size, atom.size);
168 #endif
170 return 0;
173 static int mov_read_default(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
175 int64_t total_size = 0;
176 MOVAtom a;
177 int i;
178 int err = 0;
180 a.offset = atom.offset;
182 if (atom.size < 0)
183 atom.size = INT64_MAX;
184 while(((total_size + 8) < atom.size) && !url_feof(pb) && !err) {
185 int (*parse)(MOVContext*, ByteIOContext*, MOVAtom) = NULL;
186 a.size = atom.size;
187 a.type=0;
188 if(atom.size >= 8) {
189 a.size = get_be32(pb);
190 a.type = get_le32(pb);
192 total_size += 8;
193 a.offset += 8;
194 dprintf(c->fc, "type: %08x %.4s sz: %"PRIx64" %"PRIx64" %"PRIx64"\n",
195 a.type, (char*)&a.type, a.size, atom.size, total_size);
196 if (a.size == 1) { /* 64 bit extended size */
197 a.size = get_be64(pb) - 8;
198 a.offset += 8;
199 total_size += 8;
201 if (a.size == 0) {
202 a.size = atom.size - total_size;
203 if (a.size <= 8)
204 break;
206 a.size -= 8;
207 if(a.size < 0)
208 break;
209 a.size = FFMIN(a.size, atom.size - total_size);
211 for (i = 0; mov_default_parse_table[i].type; i++)
212 if (mov_default_parse_table[i].type == a.type) {
213 parse = mov_default_parse_table[i].parse;
214 break;
217 // container is user data
218 if (!parse && (atom.type == MKTAG('u','d','t','a') ||
219 atom.type == MKTAG('i','l','s','t')))
220 parse = mov_read_udta_string;
222 if (!parse) { /* skip leaf atoms data */
223 url_fskip(pb, a.size);
224 } else {
225 int64_t start_pos = url_ftell(pb);
226 int64_t left;
227 err = parse(c, pb, a);
228 if (url_is_streamed(pb) && c->found_moov && c->found_mdat)
229 break;
230 left = a.size - url_ftell(pb) + start_pos;
231 if (left > 0) /* skip garbage at atom end */
232 url_fskip(pb, left);
235 a.offset += a.size;
236 total_size += a.size;
239 if (!err && total_size < atom.size && atom.size < 0x7ffff)
240 url_fskip(pb, atom.size - total_size);
242 return err;
245 static int mov_read_dref(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
247 AVStream *st;
248 MOVStreamContext *sc;
249 int entries, i, j;
251 if (c->fc->nb_streams < 1)
252 return 0;
253 st = c->fc->streams[c->fc->nb_streams-1];
254 sc = st->priv_data;
256 get_be32(pb); // version + flags
257 entries = get_be32(pb);
258 if (entries >= UINT_MAX / sizeof(*sc->drefs))
259 return -1;
260 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
261 if (!sc->drefs)
262 return AVERROR(ENOMEM);
263 sc->drefs_count = entries;
265 for (i = 0; i < sc->drefs_count; i++) {
266 MOVDref *dref = &sc->drefs[i];
267 uint32_t size = get_be32(pb);
268 int64_t next = url_ftell(pb) + size - 4;
270 dref->type = get_le32(pb);
271 get_be32(pb); // version + flags
272 dprintf(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
274 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
275 /* macintosh alias record */
276 uint16_t volume_len, len;
277 char volume[28];
278 int16_t type;
280 url_fskip(pb, 10);
282 volume_len = get_byte(pb);
283 volume_len = FFMIN(volume_len, 27);
284 get_buffer(pb, volume, 27);
285 volume[volume_len] = 0;
286 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", volume, volume_len);
288 url_fskip(pb, 112);
290 for (type = 0; type != -1 && url_ftell(pb) < next; ) {
291 type = get_be16(pb);
292 len = get_be16(pb);
293 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
294 if (len&1)
295 len += 1;
296 if (type == 2) { // absolute path
297 av_free(dref->path);
298 dref->path = av_mallocz(len+1);
299 if (!dref->path)
300 return AVERROR(ENOMEM);
301 get_buffer(pb, dref->path, len);
302 if (len > volume_len && !strncmp(dref->path, volume, volume_len)) {
303 len -= volume_len;
304 memmove(dref->path, dref->path+volume_len, len);
305 dref->path[len] = 0;
307 for (j = 0; j < len; j++)
308 if (dref->path[j] == ':')
309 dref->path[j] = '/';
310 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
311 } else
312 url_fskip(pb, len);
315 url_fseek(pb, next, SEEK_SET);
317 return 0;
320 static int mov_read_hdlr(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
322 AVStream *st;
323 uint32_t type;
324 uint32_t ctype;
326 if (c->fc->nb_streams < 1) // meta before first trak
327 return 0;
329 st = c->fc->streams[c->fc->nb_streams-1];
331 get_byte(pb); /* version */
332 get_be24(pb); /* flags */
334 /* component type */
335 ctype = get_le32(pb);
336 type = get_le32(pb); /* component subtype */
338 dprintf(c->fc, "ctype= %.4s (0x%08x)\n", (char*)&ctype, ctype);
339 dprintf(c->fc, "stype= %.4s\n", (char*)&type);
341 if (type == MKTAG('v','i','d','e'))
342 st->codec->codec_type = CODEC_TYPE_VIDEO;
343 else if(type == MKTAG('s','o','u','n'))
344 st->codec->codec_type = CODEC_TYPE_AUDIO;
345 else if(type == MKTAG('m','1','a',' '))
346 st->codec->codec_id = CODEC_ID_MP2;
347 else if(type == MKTAG('s','u','b','p'))
348 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
350 get_be32(pb); /* component manufacture */
351 get_be32(pb); /* component flags */
352 get_be32(pb); /* component flags mask */
354 if(atom.size <= 24)
355 return 0; /* nothing left to read */
357 url_fskip(pb, atom.size - (url_ftell(pb) - atom.offset));
358 return 0;
361 static int mp4_read_descr_len(ByteIOContext *pb)
363 int len = 0;
364 int count = 4;
365 while (count--) {
366 int c = get_byte(pb);
367 len = (len << 7) | (c & 0x7f);
368 if (!(c & 0x80))
369 break;
371 return len;
374 static int mp4_read_descr(MOVContext *c, ByteIOContext *pb, int *tag)
376 int len;
377 *tag = get_byte(pb);
378 len = mp4_read_descr_len(pb);
379 dprintf(c->fc, "MPEG4 description: tag=0x%02x len=%d\n", *tag, len);
380 return len;
383 #define MP4ESDescrTag 0x03
384 #define MP4DecConfigDescrTag 0x04
385 #define MP4DecSpecificDescrTag 0x05
387 static const AVCodecTag mp4_audio_types[] = {
388 { CODEC_ID_MP3ON4, AOT_PS }, /* old mp3on4 draft */
389 { CODEC_ID_MP3ON4, AOT_L1 }, /* layer 1 */
390 { CODEC_ID_MP3ON4, AOT_L2 }, /* layer 2 */
391 { CODEC_ID_MP3ON4, AOT_L3 }, /* layer 3 */
392 { CODEC_ID_MP4ALS, AOT_ALS }, /* MPEG-4 ALS */
393 { CODEC_ID_NONE, AOT_NULL },
396 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
398 AVStream *st;
399 int tag, len;
401 if (c->fc->nb_streams < 1)
402 return 0;
403 st = c->fc->streams[c->fc->nb_streams-1];
405 get_be32(pb); /* version + flags */
406 len = mp4_read_descr(c, pb, &tag);
407 if (tag == MP4ESDescrTag) {
408 get_be16(pb); /* ID */
409 get_byte(pb); /* priority */
410 } else
411 get_be16(pb); /* ID */
413 len = mp4_read_descr(c, pb, &tag);
414 if (tag == MP4DecConfigDescrTag) {
415 int object_type_id = get_byte(pb);
416 get_byte(pb); /* stream type */
417 get_be24(pb); /* buffer size db */
418 get_be32(pb); /* max bitrate */
419 get_be32(pb); /* avg bitrate */
421 st->codec->codec_id= ff_codec_get_id(ff_mp4_obj_type, object_type_id);
422 dprintf(c->fc, "esds object type id %d\n", object_type_id);
423 len = mp4_read_descr(c, pb, &tag);
424 if (tag == MP4DecSpecificDescrTag) {
425 dprintf(c->fc, "Specific MPEG4 header len=%d\n", len);
426 if((uint64_t)len > (1<<30))
427 return -1;
428 st->codec->extradata = av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
429 if (!st->codec->extradata)
430 return AVERROR(ENOMEM);
431 get_buffer(pb, st->codec->extradata, len);
432 st->codec->extradata_size = len;
433 if (st->codec->codec_id == CODEC_ID_AAC) {
434 MPEG4AudioConfig cfg;
435 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
436 st->codec->extradata_size);
437 if (cfg.chan_config > 7)
438 return -1;
439 st->codec->channels = ff_mpeg4audio_channels[cfg.chan_config];
440 if (cfg.object_type == 29 && cfg.sampling_index < 3) // old mp3on4
441 st->codec->sample_rate = ff_mpa_freq_tab[cfg.sampling_index];
442 else
443 st->codec->sample_rate = cfg.sample_rate; // ext sample rate ?
444 dprintf(c->fc, "mp4a config channels %d obj %d ext obj %d "
445 "sample rate %d ext sample rate %d\n", st->codec->channels,
446 cfg.object_type, cfg.ext_object_type,
447 cfg.sample_rate, cfg.ext_sample_rate);
448 if (!(st->codec->codec_id = ff_codec_get_id(mp4_audio_types,
449 cfg.object_type)))
450 st->codec->codec_id = CODEC_ID_AAC;
454 return 0;
457 static int mov_read_pasp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
459 const int num = get_be32(pb);
460 const int den = get_be32(pb);
461 AVStream *st;
463 if (c->fc->nb_streams < 1)
464 return 0;
465 st = c->fc->streams[c->fc->nb_streams-1];
467 if (den != 0) {
468 if ((st->sample_aspect_ratio.den != 1 || st->sample_aspect_ratio.num) && // default
469 (den != st->sample_aspect_ratio.den || num != st->sample_aspect_ratio.num))
470 av_log(c->fc, AV_LOG_WARNING,
471 "sample aspect ratio already set to %d:%d, overriding by 'pasp' atom\n",
472 st->sample_aspect_ratio.num, st->sample_aspect_ratio.den);
473 st->sample_aspect_ratio.num = num;
474 st->sample_aspect_ratio.den = den;
476 return 0;
479 /* this atom contains actual media data */
480 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
482 if(atom.size == 0) /* wrong one (MP4) */
483 return 0;
484 c->found_mdat=1;
485 return 0; /* now go for moov */
488 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
490 uint32_t type = get_le32(pb);
492 if (type != MKTAG('q','t',' ',' '))
493 c->isom = 1;
494 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
495 get_be32(pb); /* minor version */
496 url_fskip(pb, atom.size - 8);
497 return 0;
500 /* this atom should contain all header atoms */
501 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
503 if (mov_read_default(c, pb, atom) < 0)
504 return -1;
505 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
506 /* so we don't parse the whole file if over a network */
507 c->found_moov=1;
508 return 0; /* now go for mdat */
511 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
513 c->fragment.moof_offset = url_ftell(pb) - 8;
514 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
515 return mov_read_default(c, pb, atom);
518 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
520 AVStream *st;
521 MOVStreamContext *sc;
522 int version;
523 char language[4] = {0};
524 unsigned lang;
526 if (c->fc->nb_streams < 1)
527 return 0;
528 st = c->fc->streams[c->fc->nb_streams-1];
529 sc = st->priv_data;
531 version = get_byte(pb);
532 if (version > 1)
533 return -1; /* unsupported */
535 get_be24(pb); /* flags */
536 if (version == 1) {
537 get_be64(pb);
538 get_be64(pb);
539 } else {
540 get_be32(pb); /* creation time */
541 get_be32(pb); /* modification time */
544 sc->time_scale = get_be32(pb);
545 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
547 lang = get_be16(pb); /* language */
548 if (ff_mov_lang_to_iso639(lang, language))
549 av_metadata_set(&st->metadata, "language", language);
550 get_be16(pb); /* quality */
552 return 0;
555 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
557 int version = get_byte(pb); /* version */
558 get_be24(pb); /* flags */
560 if (version == 1) {
561 get_be64(pb);
562 get_be64(pb);
563 } else {
564 get_be32(pb); /* creation time */
565 get_be32(pb); /* modification time */
567 c->time_scale = get_be32(pb); /* time scale */
569 dprintf(c->fc, "time scale = %i\n", c->time_scale);
571 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
572 get_be32(pb); /* preferred scale */
574 get_be16(pb); /* preferred volume */
576 url_fskip(pb, 10); /* reserved */
578 url_fskip(pb, 36); /* display matrix */
580 get_be32(pb); /* preview time */
581 get_be32(pb); /* preview duration */
582 get_be32(pb); /* poster time */
583 get_be32(pb); /* selection time */
584 get_be32(pb); /* selection duration */
585 get_be32(pb); /* current time */
586 get_be32(pb); /* next track ID */
588 return 0;
591 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
593 AVStream *st;
595 if (c->fc->nb_streams < 1)
596 return 0;
597 st = c->fc->streams[c->fc->nb_streams-1];
599 if((uint64_t)atom.size > (1<<30))
600 return -1;
602 // currently SVQ3 decoder expect full STSD header - so let's fake it
603 // this should be fixed and just SMI header should be passed
604 av_free(st->codec->extradata);
605 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
606 if (!st->codec->extradata)
607 return AVERROR(ENOMEM);
608 st->codec->extradata_size = 0x5a + atom.size;
609 memcpy(st->codec->extradata, "SVQ3", 4); // fake
610 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
611 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
612 return 0;
615 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
617 AVStream *st;
618 int little_endian;
620 if (c->fc->nb_streams < 1)
621 return 0;
622 st = c->fc->streams[c->fc->nb_streams-1];
624 little_endian = get_be16(pb);
625 dprintf(c->fc, "enda %d\n", little_endian);
626 if (little_endian == 1) {
627 switch (st->codec->codec_id) {
628 case CODEC_ID_PCM_S24BE:
629 st->codec->codec_id = CODEC_ID_PCM_S24LE;
630 break;
631 case CODEC_ID_PCM_S32BE:
632 st->codec->codec_id = CODEC_ID_PCM_S32LE;
633 break;
634 case CODEC_ID_PCM_F32BE:
635 st->codec->codec_id = CODEC_ID_PCM_F32LE;
636 break;
637 case CODEC_ID_PCM_F64BE:
638 st->codec->codec_id = CODEC_ID_PCM_F64LE;
639 break;
640 default:
641 break;
644 return 0;
647 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
648 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
650 AVStream *st;
651 uint64_t size;
652 uint8_t *buf;
654 if (c->fc->nb_streams < 1) // will happen with jp2 files
655 return 0;
656 st= c->fc->streams[c->fc->nb_streams-1];
657 size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
658 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
659 return -1;
660 buf= av_realloc(st->codec->extradata, size);
661 if(!buf)
662 return -1;
663 st->codec->extradata= buf;
664 buf+= st->codec->extradata_size;
665 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
666 AV_WB32( buf , atom.size + 8);
667 AV_WL32( buf + 4, atom.type);
668 get_buffer(pb, buf + 8, atom.size);
669 return 0;
672 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
674 AVStream *st;
676 if (c->fc->nb_streams < 1)
677 return 0;
678 st = c->fc->streams[c->fc->nb_streams-1];
680 if((uint64_t)atom.size > (1<<30))
681 return -1;
683 if (st->codec->codec_id == CODEC_ID_QDM2) {
684 // pass all frma atom to codec, needed at least for QDM2
685 av_free(st->codec->extradata);
686 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
687 if (!st->codec->extradata)
688 return AVERROR(ENOMEM);
689 st->codec->extradata_size = atom.size;
690 get_buffer(pb, st->codec->extradata, atom.size);
691 } else if (atom.size > 8) { /* to read frma, esds atoms */
692 if (mov_read_default(c, pb, atom) < 0)
693 return -1;
694 } else
695 url_fskip(pb, atom.size);
696 return 0;
700 * This function reads atom content and puts data in extradata without tag
701 * nor size unlike mov_read_extradata.
703 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
705 AVStream *st;
707 if (c->fc->nb_streams < 1)
708 return 0;
709 st = c->fc->streams[c->fc->nb_streams-1];
711 if((uint64_t)atom.size > (1<<30))
712 return -1;
714 av_free(st->codec->extradata);
715 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
716 if (!st->codec->extradata)
717 return AVERROR(ENOMEM);
718 st->codec->extradata_size = atom.size;
719 get_buffer(pb, st->codec->extradata, atom.size);
720 return 0;
723 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
725 AVStream *st;
726 MOVStreamContext *sc;
727 unsigned int i, entries;
729 if (c->fc->nb_streams < 1)
730 return 0;
731 st = c->fc->streams[c->fc->nb_streams-1];
732 sc = st->priv_data;
734 get_byte(pb); /* version */
735 get_be24(pb); /* flags */
737 entries = get_be32(pb);
739 if(entries >= UINT_MAX/sizeof(int64_t))
740 return -1;
742 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
743 if (!sc->chunk_offsets)
744 return AVERROR(ENOMEM);
745 sc->chunk_count = entries;
747 if (atom.type == MKTAG('s','t','c','o'))
748 for(i=0; i<entries; i++)
749 sc->chunk_offsets[i] = get_be32(pb);
750 else if (atom.type == MKTAG('c','o','6','4'))
751 for(i=0; i<entries; i++)
752 sc->chunk_offsets[i] = get_be64(pb);
753 else
754 return -1;
756 return 0;
760 * Compute codec id for 'lpcm' tag.
761 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
763 static enum CodecID mov_get_lpcm_codec_id(int bps, int flags)
765 if (flags & 1) { // floating point
766 if (flags & 2) { // big endian
767 if (bps == 32) return CODEC_ID_PCM_F32BE;
768 else if (bps == 64) return CODEC_ID_PCM_F64BE;
769 } else {
770 if (bps == 32) return CODEC_ID_PCM_F32LE;
771 else if (bps == 64) return CODEC_ID_PCM_F64LE;
773 } else {
774 if (flags & 2) {
775 if (bps == 8)
776 // signed integer
777 if (flags & 4) return CODEC_ID_PCM_S8;
778 else return CODEC_ID_PCM_U8;
779 else if (bps == 16) return CODEC_ID_PCM_S16BE;
780 else if (bps == 24) return CODEC_ID_PCM_S24BE;
781 else if (bps == 32) return CODEC_ID_PCM_S32BE;
782 } else {
783 if (bps == 8)
784 if (flags & 4) return CODEC_ID_PCM_S8;
785 else return CODEC_ID_PCM_U8;
786 else if (bps == 16) return CODEC_ID_PCM_S16LE;
787 else if (bps == 24) return CODEC_ID_PCM_S24LE;
788 else if (bps == 32) return CODEC_ID_PCM_S32LE;
791 return CODEC_ID_NONE;
794 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
796 AVStream *st;
797 MOVStreamContext *sc;
798 int j, entries, pseudo_stream_id;
800 if (c->fc->nb_streams < 1)
801 return 0;
802 st = c->fc->streams[c->fc->nb_streams-1];
803 sc = st->priv_data;
805 get_byte(pb); /* version */
806 get_be24(pb); /* flags */
808 entries = get_be32(pb);
810 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
811 //Parsing Sample description table
812 enum CodecID id;
813 int dref_id = 1;
814 MOVAtom a = { 0, 0, 0 };
815 int64_t start_pos = url_ftell(pb);
816 int size = get_be32(pb); /* size */
817 uint32_t format = get_le32(pb); /* data format */
819 if (size >= 16) {
820 get_be32(pb); /* reserved */
821 get_be16(pb); /* reserved */
822 dref_id = get_be16(pb);
825 if (st->codec->codec_tag &&
826 st->codec->codec_tag != format &&
827 (c->fc->video_codec_id ? ff_codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
828 : st->codec->codec_tag != MKTAG('j','p','e','g'))
830 /* Multiple fourcc, we skip JPEG. This is not correct, we should
831 * export it as a separate AVStream but this needs a few changes
832 * in the MOV demuxer, patch welcome. */
833 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
834 url_fskip(pb, size - (url_ftell(pb) - start_pos));
835 continue;
837 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
838 sc->dref_id= dref_id;
840 st->codec->codec_tag = format;
841 id = ff_codec_get_id(codec_movaudio_tags, format);
842 if (id<=0 && ((format&0xFFFF) == 'm'+('s'<<8) || (format&0xFFFF) == 'T'+('S'<<8)))
843 id = ff_codec_get_id(ff_codec_wav_tags, bswap_32(format)&0xFFFF);
845 if (st->codec->codec_type != CODEC_TYPE_VIDEO && id > 0) {
846 st->codec->codec_type = CODEC_TYPE_AUDIO;
847 } else if (st->codec->codec_type != CODEC_TYPE_AUDIO && /* do not overwrite codec type */
848 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
849 id = ff_codec_get_id(codec_movvideo_tags, format);
850 if (id <= 0)
851 id = ff_codec_get_id(ff_codec_bmp_tags, format);
852 if (id > 0)
853 st->codec->codec_type = CODEC_TYPE_VIDEO;
854 else if(st->codec->codec_type == CODEC_TYPE_DATA){
855 id = ff_codec_get_id(ff_codec_movsubtitle_tags, format);
856 if(id > 0)
857 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
861 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
862 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
863 (format >> 24) & 0xff, st->codec->codec_type);
865 if(st->codec->codec_type==CODEC_TYPE_VIDEO) {
866 uint8_t codec_name[32];
867 unsigned int color_depth;
868 int color_greyscale;
870 st->codec->codec_id = id;
871 get_be16(pb); /* version */
872 get_be16(pb); /* revision level */
873 get_be32(pb); /* vendor */
874 get_be32(pb); /* temporal quality */
875 get_be32(pb); /* spatial quality */
877 st->codec->width = get_be16(pb); /* width */
878 st->codec->height = get_be16(pb); /* height */
880 get_be32(pb); /* horiz resolution */
881 get_be32(pb); /* vert resolution */
882 get_be32(pb); /* data size, always 0 */
883 get_be16(pb); /* frames per samples */
885 get_buffer(pb, codec_name, 32); /* codec name, pascal string */
886 if (codec_name[0] <= 31) {
887 memcpy(st->codec->codec_name, &codec_name[1],codec_name[0]);
888 st->codec->codec_name[codec_name[0]] = 0;
891 st->codec->bits_per_coded_sample = get_be16(pb); /* depth */
892 st->codec->color_table_id = get_be16(pb); /* colortable id */
893 dprintf(c->fc, "depth %d, ctab id %d\n",
894 st->codec->bits_per_coded_sample, st->codec->color_table_id);
895 /* figure out the palette situation */
896 color_depth = st->codec->bits_per_coded_sample & 0x1F;
897 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
899 /* if the depth is 2, 4, or 8 bpp, file is palettized */
900 if ((color_depth == 2) || (color_depth == 4) ||
901 (color_depth == 8)) {
902 /* for palette traversal */
903 unsigned int color_start, color_count, color_end;
904 unsigned char r, g, b;
906 st->codec->palctrl = av_malloc(sizeof(*st->codec->palctrl));
907 if (color_greyscale) {
908 int color_index, color_dec;
909 /* compute the greyscale palette */
910 st->codec->bits_per_coded_sample = color_depth;
911 color_count = 1 << color_depth;
912 color_index = 255;
913 color_dec = 256 / (color_count - 1);
914 for (j = 0; j < color_count; j++) {
915 r = g = b = color_index;
916 st->codec->palctrl->palette[j] =
917 (r << 16) | (g << 8) | (b);
918 color_index -= color_dec;
919 if (color_index < 0)
920 color_index = 0;
922 } else if (st->codec->color_table_id) {
923 const uint8_t *color_table;
924 /* if flag bit 3 is set, use the default palette */
925 color_count = 1 << color_depth;
926 if (color_depth == 2)
927 color_table = ff_qt_default_palette_4;
928 else if (color_depth == 4)
929 color_table = ff_qt_default_palette_16;
930 else
931 color_table = ff_qt_default_palette_256;
933 for (j = 0; j < color_count; j++) {
934 r = color_table[j * 3 + 0];
935 g = color_table[j * 3 + 1];
936 b = color_table[j * 3 + 2];
937 st->codec->palctrl->palette[j] =
938 (r << 16) | (g << 8) | (b);
940 } else {
941 /* load the palette from the file */
942 color_start = get_be32(pb);
943 color_count = get_be16(pb);
944 color_end = get_be16(pb);
945 if ((color_start <= 255) &&
946 (color_end <= 255)) {
947 for (j = color_start; j <= color_end; j++) {
948 /* each R, G, or B component is 16 bits;
949 * only use the top 8 bits; skip alpha bytes
950 * up front */
951 get_byte(pb);
952 get_byte(pb);
953 r = get_byte(pb);
954 get_byte(pb);
955 g = get_byte(pb);
956 get_byte(pb);
957 b = get_byte(pb);
958 get_byte(pb);
959 st->codec->palctrl->palette[j] =
960 (r << 16) | (g << 8) | (b);
964 st->codec->palctrl->palette_changed = 1;
966 } else if(st->codec->codec_type==CODEC_TYPE_AUDIO) {
967 int bits_per_sample, flags;
968 uint16_t version = get_be16(pb);
970 st->codec->codec_id = id;
971 get_be16(pb); /* revision level */
972 get_be32(pb); /* vendor */
974 st->codec->channels = get_be16(pb); /* channel count */
975 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
976 st->codec->bits_per_coded_sample = get_be16(pb); /* sample size */
978 sc->audio_cid = get_be16(pb);
979 get_be16(pb); /* packet size = 0 */
981 st->codec->sample_rate = ((get_be32(pb) >> 16));
983 //Read QT version 1 fields. In version 0 these do not exist.
984 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
985 if(!c->isom) {
986 if(version==1) {
987 sc->samples_per_frame = get_be32(pb);
988 get_be32(pb); /* bytes per packet */
989 sc->bytes_per_frame = get_be32(pb);
990 get_be32(pb); /* bytes per sample */
991 } else if(version==2) {
992 get_be32(pb); /* sizeof struct only */
993 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
994 st->codec->channels = get_be32(pb);
995 get_be32(pb); /* always 0x7F000000 */
996 st->codec->bits_per_coded_sample = get_be32(pb); /* bits per channel if sound is uncompressed */
997 flags = get_be32(pb); /* lcpm format specific flag */
998 sc->bytes_per_frame = get_be32(pb); /* bytes per audio packet if constant */
999 sc->samples_per_frame = get_be32(pb); /* lpcm frames per audio packet if constant */
1000 if (format == MKTAG('l','p','c','m'))
1001 st->codec->codec_id = mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, flags);
1005 switch (st->codec->codec_id) {
1006 case CODEC_ID_PCM_S8:
1007 case CODEC_ID_PCM_U8:
1008 if (st->codec->bits_per_coded_sample == 16)
1009 st->codec->codec_id = CODEC_ID_PCM_S16BE;
1010 break;
1011 case CODEC_ID_PCM_S16LE:
1012 case CODEC_ID_PCM_S16BE:
1013 if (st->codec->bits_per_coded_sample == 8)
1014 st->codec->codec_id = CODEC_ID_PCM_S8;
1015 else if (st->codec->bits_per_coded_sample == 24)
1016 st->codec->codec_id =
1017 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
1018 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
1019 break;
1020 /* set values for old format before stsd version 1 appeared */
1021 case CODEC_ID_MACE3:
1022 sc->samples_per_frame = 6;
1023 sc->bytes_per_frame = 2*st->codec->channels;
1024 break;
1025 case CODEC_ID_MACE6:
1026 sc->samples_per_frame = 6;
1027 sc->bytes_per_frame = 1*st->codec->channels;
1028 break;
1029 case CODEC_ID_ADPCM_IMA_QT:
1030 sc->samples_per_frame = 64;
1031 sc->bytes_per_frame = 34*st->codec->channels;
1032 break;
1033 case CODEC_ID_GSM:
1034 sc->samples_per_frame = 160;
1035 sc->bytes_per_frame = 33;
1036 break;
1037 default:
1038 break;
1041 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
1042 if (bits_per_sample) {
1043 st->codec->bits_per_coded_sample = bits_per_sample;
1044 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
1046 } else if(st->codec->codec_type==CODEC_TYPE_SUBTITLE){
1047 // ttxt stsd contains display flags, justification, background
1048 // color, fonts, and default styles, so fake an atom to read it
1049 MOVAtom fake_atom = { .size = size - (url_ftell(pb) - start_pos) };
1050 if (format != AV_RL32("mp4s")) // mp4s contains a regular esds atom
1051 mov_read_glbl(c, pb, fake_atom);
1052 st->codec->codec_id= id;
1053 st->codec->width = sc->width;
1054 st->codec->height = sc->height;
1055 } else {
1056 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
1057 url_fskip(pb, size - (url_ftell(pb) - start_pos));
1059 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
1060 a.size = size - (url_ftell(pb) - start_pos);
1061 if (a.size > 8) {
1062 if (mov_read_default(c, pb, a) < 0)
1063 return -1;
1064 } else if (a.size > 0)
1065 url_fskip(pb, a.size);
1068 if(st->codec->codec_type==CODEC_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
1069 st->codec->sample_rate= sc->time_scale;
1071 /* special codec parameters handling */
1072 switch (st->codec->codec_id) {
1073 #if CONFIG_DV_DEMUXER
1074 case CODEC_ID_DVAUDIO:
1075 c->dv_fctx = avformat_alloc_context();
1076 c->dv_demux = dv_init_demux(c->dv_fctx);
1077 if (!c->dv_demux) {
1078 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
1079 return -1;
1081 sc->dv_audio_container = 1;
1082 st->codec->codec_id = CODEC_ID_PCM_S16LE;
1083 break;
1084 #endif
1085 /* no ifdef since parameters are always those */
1086 case CODEC_ID_QCELP:
1087 // force sample rate for qcelp when not stored in mov
1088 if (st->codec->codec_tag != MKTAG('Q','c','l','p'))
1089 st->codec->sample_rate = 8000;
1090 st->codec->frame_size= 160;
1091 st->codec->channels= 1; /* really needed */
1092 break;
1093 case CODEC_ID_AMR_NB:
1094 case CODEC_ID_AMR_WB:
1095 st->codec->frame_size= sc->samples_per_frame;
1096 st->codec->channels= 1; /* really needed */
1097 /* force sample rate for amr, stsd in 3gp does not store sample rate */
1098 if (st->codec->codec_id == CODEC_ID_AMR_NB)
1099 st->codec->sample_rate = 8000;
1100 else if (st->codec->codec_id == CODEC_ID_AMR_WB)
1101 st->codec->sample_rate = 16000;
1102 break;
1103 case CODEC_ID_MP2:
1104 case CODEC_ID_MP3:
1105 st->codec->codec_type = CODEC_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
1106 st->need_parsing = AVSTREAM_PARSE_FULL;
1107 break;
1108 case CODEC_ID_GSM:
1109 case CODEC_ID_ADPCM_MS:
1110 case CODEC_ID_ADPCM_IMA_WAV:
1111 st->codec->block_align = sc->bytes_per_frame;
1112 break;
1113 case CODEC_ID_ALAC:
1114 if (st->codec->extradata_size == 36) {
1115 st->codec->frame_size = AV_RB32(st->codec->extradata+12);
1116 st->codec->channels = AV_RB8 (st->codec->extradata+21);
1118 break;
1119 default:
1120 break;
1123 return 0;
1126 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1128 AVStream *st;
1129 MOVStreamContext *sc;
1130 unsigned int i, entries;
1132 if (c->fc->nb_streams < 1)
1133 return 0;
1134 st = c->fc->streams[c->fc->nb_streams-1];
1135 sc = st->priv_data;
1137 get_byte(pb); /* version */
1138 get_be24(pb); /* flags */
1140 entries = get_be32(pb);
1142 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1144 if(entries >= UINT_MAX / sizeof(*sc->stsc_data))
1145 return -1;
1146 sc->stsc_data = av_malloc(entries * sizeof(*sc->stsc_data));
1147 if (!sc->stsc_data)
1148 return AVERROR(ENOMEM);
1149 sc->stsc_count = entries;
1151 for(i=0; i<entries; i++) {
1152 sc->stsc_data[i].first = get_be32(pb);
1153 sc->stsc_data[i].count = get_be32(pb);
1154 sc->stsc_data[i].id = get_be32(pb);
1156 return 0;
1159 static int mov_read_stps(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1161 AVStream *st;
1162 MOVStreamContext *sc;
1163 unsigned i, entries;
1165 if (c->fc->nb_streams < 1)
1166 return 0;
1167 st = c->fc->streams[c->fc->nb_streams-1];
1168 sc = st->priv_data;
1170 get_be32(pb); // version + flags
1172 entries = get_be32(pb);
1173 if (entries >= UINT_MAX / sizeof(*sc->stps_data))
1174 return -1;
1175 sc->stps_data = av_malloc(entries * sizeof(*sc->stps_data));
1176 if (!sc->stps_data)
1177 return AVERROR(ENOMEM);
1178 sc->stps_count = entries;
1180 for (i = 0; i < entries; i++) {
1181 sc->stps_data[i] = get_be32(pb);
1182 //dprintf(c->fc, "stps %d\n", sc->stps_data[i]);
1185 return 0;
1188 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1190 AVStream *st;
1191 MOVStreamContext *sc;
1192 unsigned int i, entries;
1194 if (c->fc->nb_streams < 1)
1195 return 0;
1196 st = c->fc->streams[c->fc->nb_streams-1];
1197 sc = st->priv_data;
1199 get_byte(pb); /* version */
1200 get_be24(pb); /* flags */
1202 entries = get_be32(pb);
1204 dprintf(c->fc, "keyframe_count = %d\n", entries);
1206 if(entries >= UINT_MAX / sizeof(int))
1207 return -1;
1208 sc->keyframes = av_malloc(entries * sizeof(int));
1209 if (!sc->keyframes)
1210 return AVERROR(ENOMEM);
1211 sc->keyframe_count = entries;
1213 for(i=0; i<entries; i++) {
1214 sc->keyframes[i] = get_be32(pb);
1215 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1217 return 0;
1220 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1222 AVStream *st;
1223 MOVStreamContext *sc;
1224 unsigned int i, entries, sample_size, field_size, num_bytes;
1225 GetBitContext gb;
1226 unsigned char* buf;
1228 if (c->fc->nb_streams < 1)
1229 return 0;
1230 st = c->fc->streams[c->fc->nb_streams-1];
1231 sc = st->priv_data;
1233 get_byte(pb); /* version */
1234 get_be24(pb); /* flags */
1236 if (atom.type == MKTAG('s','t','s','z')) {
1237 sample_size = get_be32(pb);
1238 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1239 sc->sample_size = sample_size;
1240 field_size = 32;
1241 } else {
1242 sample_size = 0;
1243 get_be24(pb); /* reserved */
1244 field_size = get_byte(pb);
1246 entries = get_be32(pb);
1248 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, entries);
1250 sc->sample_count = entries;
1251 if (sample_size)
1252 return 0;
1254 if (field_size != 4 && field_size != 8 && field_size != 16 && field_size != 32) {
1255 av_log(c->fc, AV_LOG_ERROR, "Invalid sample field size %d\n", field_size);
1256 return -1;
1259 if(entries >= UINT_MAX / sizeof(int))
1260 return -1;
1261 sc->sample_sizes = av_malloc(entries * sizeof(int));
1262 if (!sc->sample_sizes)
1263 return AVERROR(ENOMEM);
1265 num_bytes = (entries*field_size+4)>>3;
1267 buf = av_malloc(num_bytes+FF_INPUT_BUFFER_PADDING_SIZE);
1268 if (!buf) {
1269 av_freep(&sc->sample_sizes);
1270 return AVERROR(ENOMEM);
1273 if (get_buffer(pb, buf, num_bytes) < num_bytes) {
1274 av_freep(&sc->sample_sizes);
1275 av_free(buf);
1276 return -1;
1279 init_get_bits(&gb, buf, 8*num_bytes);
1281 for(i=0; i<entries; i++)
1282 sc->sample_sizes[i] = get_bits_long(&gb, field_size);
1284 av_free(buf);
1285 return 0;
1288 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1290 AVStream *st;
1291 MOVStreamContext *sc;
1292 unsigned int i, entries;
1293 int64_t duration=0;
1294 int64_t total_sample_count=0;
1296 if (c->fc->nb_streams < 1)
1297 return 0;
1298 st = c->fc->streams[c->fc->nb_streams-1];
1299 sc = st->priv_data;
1301 get_byte(pb); /* version */
1302 get_be24(pb); /* flags */
1303 entries = get_be32(pb);
1305 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1307 if(entries >= UINT_MAX / sizeof(*sc->stts_data))
1308 return -1;
1309 sc->stts_data = av_malloc(entries * sizeof(*sc->stts_data));
1310 if (!sc->stts_data)
1311 return AVERROR(ENOMEM);
1312 sc->stts_count = entries;
1314 for(i=0; i<entries; i++) {
1315 int sample_duration;
1316 int sample_count;
1318 sample_count=get_be32(pb);
1319 sample_duration = get_be32(pb);
1320 sc->stts_data[i].count= sample_count;
1321 sc->stts_data[i].duration= sample_duration;
1323 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1325 duration+=(int64_t)sample_duration*sample_count;
1326 total_sample_count+=sample_count;
1329 st->nb_frames= total_sample_count;
1330 if(duration)
1331 st->duration= duration;
1332 return 0;
1335 static int mov_read_cslg(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1337 AVStream *st;
1338 MOVStreamContext *sc;
1340 if (c->fc->nb_streams < 1)
1341 return 0;
1342 st = c->fc->streams[c->fc->nb_streams-1];
1343 sc = st->priv_data;
1345 get_be32(pb); // version + flags
1347 sc->dts_shift = get_be32(pb);
1348 dprintf(c->fc, "dts shift %d\n", sc->dts_shift);
1350 get_be32(pb); // least dts to pts delta
1351 get_be32(pb); // greatest dts to pts delta
1352 get_be32(pb); // pts start
1353 get_be32(pb); // pts end
1355 return 0;
1358 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1360 AVStream *st;
1361 MOVStreamContext *sc;
1362 unsigned int i, entries;
1364 if (c->fc->nb_streams < 1)
1365 return 0;
1366 st = c->fc->streams[c->fc->nb_streams-1];
1367 sc = st->priv_data;
1369 get_byte(pb); /* version */
1370 get_be24(pb); /* flags */
1371 entries = get_be32(pb);
1373 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1375 if(entries >= UINT_MAX / sizeof(*sc->ctts_data))
1376 return -1;
1377 sc->ctts_data = av_malloc(entries * sizeof(*sc->ctts_data));
1378 if (!sc->ctts_data)
1379 return AVERROR(ENOMEM);
1380 sc->ctts_count = entries;
1382 for(i=0; i<entries; i++) {
1383 int count =get_be32(pb);
1384 int duration =get_be32(pb);
1386 sc->ctts_data[i].count = count;
1387 sc->ctts_data[i].duration= duration;
1389 return 0;
1392 static void mov_build_index(MOVContext *mov, AVStream *st)
1394 MOVStreamContext *sc = st->priv_data;
1395 int64_t current_offset;
1396 int64_t current_dts = 0;
1397 unsigned int stts_index = 0;
1398 unsigned int stsc_index = 0;
1399 unsigned int stss_index = 0;
1400 unsigned int stps_index = 0;
1401 unsigned int i, j;
1403 /* adjust first dts according to edit list */
1404 if (sc->time_offset) {
1405 int rescaled = sc->time_offset < 0 ? av_rescale(sc->time_offset, sc->time_scale, mov->time_scale) : sc->time_offset;
1406 current_dts = -rescaled;
1407 if (sc->ctts_data && sc->ctts_data[0].duration / sc->stts_data[0].duration > 16) {
1408 /* more than 16 frames delay, dts are likely wrong
1409 this happens with files created by iMovie */
1410 sc->wrong_dts = 1;
1411 st->codec->has_b_frames = 1;
1415 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1416 if (!(st->codec->codec_type == CODEC_TYPE_AUDIO &&
1417 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1418 unsigned int current_sample = 0;
1419 unsigned int stts_sample = 0;
1420 unsigned int sample_size;
1421 unsigned int distance = 0;
1422 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1424 current_dts -= sc->dts_shift;
1426 st->nb_frames = sc->sample_count;
1427 for (i = 0; i < sc->chunk_count; i++) {
1428 current_offset = sc->chunk_offsets[i];
1429 if (stsc_index + 1 < sc->stsc_count &&
1430 i + 1 == sc->stsc_data[stsc_index + 1].first)
1431 stsc_index++;
1432 for (j = 0; j < sc->stsc_data[stsc_index].count; j++) {
1433 int keyframe = 0;
1434 if (current_sample >= sc->sample_count) {
1435 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1436 return;
1439 if (!sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index]) {
1440 keyframe = 1;
1441 if (stss_index + 1 < sc->keyframe_count)
1442 stss_index++;
1443 } else if (sc->stps_count && current_sample+key_off == sc->stps_data[stps_index]) {
1444 keyframe = 1;
1445 if (stps_index + 1 < sc->stps_count)
1446 stps_index++;
1448 if (keyframe)
1449 distance = 0;
1450 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1451 if(sc->pseudo_stream_id == -1 ||
1452 sc->stsc_data[stsc_index].id - 1 == sc->pseudo_stream_id) {
1453 av_add_index_entry(st, current_offset, current_dts, sample_size, distance,
1454 keyframe ? AVINDEX_KEYFRAME : 0);
1455 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1456 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1457 current_offset, current_dts, sample_size, distance, keyframe);
1460 current_offset += sample_size;
1461 current_dts += sc->stts_data[stts_index].duration;
1462 distance++;
1463 stts_sample++;
1464 current_sample++;
1465 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1466 stts_sample = 0;
1467 stts_index++;
1471 } else {
1472 for (i = 0; i < sc->chunk_count; i++) {
1473 unsigned chunk_samples;
1475 current_offset = sc->chunk_offsets[i];
1476 if (stsc_index + 1 < sc->stsc_count &&
1477 i + 1 == sc->stsc_data[stsc_index + 1].first)
1478 stsc_index++;
1479 chunk_samples = sc->stsc_data[stsc_index].count;
1481 if (sc->samples_per_frame && chunk_samples % sc->samples_per_frame) {
1482 av_log(mov->fc, AV_LOG_ERROR, "error unaligned chunk\n");
1483 return;
1486 while (chunk_samples > 0) {
1487 unsigned size, samples;
1489 if (sc->samples_per_frame >= 160) { // gsm
1490 samples = sc->samples_per_frame;
1491 size = sc->bytes_per_frame;
1492 } else {
1493 if (sc->samples_per_frame > 1) {
1494 samples = FFMIN((1024 / sc->samples_per_frame)*
1495 sc->samples_per_frame, chunk_samples);
1496 size = (samples / sc->samples_per_frame) * sc->bytes_per_frame;
1497 } else {
1498 samples = FFMIN(1024, chunk_samples);
1499 size = samples * sc->sample_size;
1503 av_add_index_entry(st, current_offset, current_dts, size, 0, AVINDEX_KEYFRAME);
1504 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1505 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1506 size, samples);
1508 current_offset += size;
1509 current_dts += samples;
1510 chunk_samples -= samples;
1516 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1518 AVStream *st;
1519 MOVStreamContext *sc;
1520 int ret;
1522 st = av_new_stream(c->fc, c->fc->nb_streams);
1523 if (!st) return AVERROR(ENOMEM);
1524 sc = av_mallocz(sizeof(MOVStreamContext));
1525 if (!sc) return AVERROR(ENOMEM);
1527 st->priv_data = sc;
1528 st->codec->codec_type = CODEC_TYPE_DATA;
1529 sc->ffindex = st->index;
1531 if ((ret = mov_read_default(c, pb, atom)) < 0)
1532 return ret;
1534 /* sanity checks */
1535 if (sc->chunk_count && (!sc->stts_count || !sc->stsc_count ||
1536 (!sc->sample_size && !sc->sample_count))) {
1537 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1538 st->index);
1539 return 0;
1542 if (!sc->time_scale)
1543 sc->time_scale = c->time_scale;
1545 av_set_pts_info(st, 64, 1, sc->time_scale);
1547 if (st->codec->codec_type == CODEC_TYPE_AUDIO &&
1548 !st->codec->frame_size && sc->stts_count == 1) {
1549 st->codec->frame_size = av_rescale(sc->stts_data[0].duration,
1550 st->codec->sample_rate, sc->time_scale);
1551 dprintf(c->fc, "frame size %d\n", st->codec->frame_size);
1554 mov_build_index(c, st);
1556 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1557 if (url_fopen(&sc->pb, sc->drefs[sc->dref_id-1].path, URL_RDONLY) < 0)
1558 av_log(c->fc, AV_LOG_ERROR, "stream %d, error opening file %s: %s\n",
1559 st->index, sc->drefs[sc->dref_id-1].path, strerror(errno));
1560 } else
1561 sc->pb = c->fc->pb;
1563 switch (st->codec->codec_id) {
1564 #if CONFIG_H261_DECODER
1565 case CODEC_ID_H261:
1566 #endif
1567 #if CONFIG_H263_DECODER
1568 case CODEC_ID_H263:
1569 #endif
1570 #if CONFIG_H264_DECODER
1571 case CODEC_ID_H264:
1572 #endif
1573 #if CONFIG_MPEG4_DECODER
1574 case CODEC_ID_MPEG4:
1575 #endif
1576 st->codec->width = 0; /* let decoder init width/height */
1577 st->codec->height= 0;
1578 break;
1581 /* Do not need those anymore. */
1582 av_freep(&sc->chunk_offsets);
1583 av_freep(&sc->stsc_data);
1584 av_freep(&sc->sample_sizes);
1585 av_freep(&sc->keyframes);
1586 av_freep(&sc->stts_data);
1587 av_freep(&sc->stps_data);
1589 return 0;
1592 static int mov_read_ilst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1594 int ret;
1595 c->itunes_metadata = 1;
1596 ret = mov_read_default(c, pb, atom);
1597 c->itunes_metadata = 0;
1598 return ret;
1601 static int mov_read_meta(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1603 while (atom.size > 8) {
1604 uint32_t tag = get_le32(pb);
1605 atom.size -= 4;
1606 if (tag == MKTAG('h','d','l','r')) {
1607 url_fseek(pb, -8, SEEK_CUR);
1608 atom.size += 8;
1609 return mov_read_default(c, pb, atom);
1612 return 0;
1615 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1617 int i;
1618 int width;
1619 int height;
1620 int64_t disp_transform[2];
1621 int display_matrix[3][2];
1622 AVStream *st;
1623 MOVStreamContext *sc;
1624 int version;
1626 if (c->fc->nb_streams < 1)
1627 return 0;
1628 st = c->fc->streams[c->fc->nb_streams-1];
1629 sc = st->priv_data;
1631 version = get_byte(pb);
1632 get_be24(pb); /* flags */
1634 MOV_TRACK_ENABLED 0x0001
1635 MOV_TRACK_IN_MOVIE 0x0002
1636 MOV_TRACK_IN_PREVIEW 0x0004
1637 MOV_TRACK_IN_POSTER 0x0008
1640 if (version == 1) {
1641 get_be64(pb);
1642 get_be64(pb);
1643 } else {
1644 get_be32(pb); /* creation time */
1645 get_be32(pb); /* modification time */
1647 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1648 get_be32(pb); /* reserved */
1650 /* highlevel (considering edits) duration in movie timebase */
1651 (version == 1) ? get_be64(pb) : get_be32(pb);
1652 get_be32(pb); /* reserved */
1653 get_be32(pb); /* reserved */
1655 get_be16(pb); /* layer */
1656 get_be16(pb); /* alternate group */
1657 get_be16(pb); /* volume */
1658 get_be16(pb); /* reserved */
1660 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1661 // they're kept in fixed point format through all calculations
1662 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1663 for (i = 0; i < 3; i++) {
1664 display_matrix[i][0] = get_be32(pb); // 16.16 fixed point
1665 display_matrix[i][1] = get_be32(pb); // 16.16 fixed point
1666 get_be32(pb); // 2.30 fixed point (not used)
1669 width = get_be32(pb); // 16.16 fixed point track width
1670 height = get_be32(pb); // 16.16 fixed point track height
1671 sc->width = width >> 16;
1672 sc->height = height >> 16;
1674 // transform the display width/height according to the matrix
1675 // skip this if the display matrix is the default identity matrix
1676 // or if it is rotating the picture, ex iPhone 3GS
1677 // to keep the same scale, use [width height 1<<16]
1678 if (width && height &&
1679 ((display_matrix[0][0] != 65536 ||
1680 display_matrix[1][1] != 65536) &&
1681 !display_matrix[0][1] &&
1682 !display_matrix[1][0] &&
1683 !display_matrix[2][0] && !display_matrix[2][1])) {
1684 for (i = 0; i < 2; i++)
1685 disp_transform[i] =
1686 (int64_t) width * display_matrix[0][i] +
1687 (int64_t) height * display_matrix[1][i] +
1688 ((int64_t) display_matrix[2][i] << 16);
1690 //sample aspect ratio is new width/height divided by old width/height
1691 st->sample_aspect_ratio = av_d2q(
1692 ((double) disp_transform[0] * height) /
1693 ((double) disp_transform[1] * width), INT_MAX);
1695 return 0;
1698 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1700 MOVFragment *frag = &c->fragment;
1701 MOVTrackExt *trex = NULL;
1702 int flags, track_id, i;
1704 get_byte(pb); /* version */
1705 flags = get_be24(pb);
1707 track_id = get_be32(pb);
1708 if (!track_id)
1709 return -1;
1710 frag->track_id = track_id;
1711 for (i = 0; i < c->trex_count; i++)
1712 if (c->trex_data[i].track_id == frag->track_id) {
1713 trex = &c->trex_data[i];
1714 break;
1716 if (!trex) {
1717 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
1718 return -1;
1721 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
1722 else frag->base_data_offset = frag->moof_offset;
1723 if (flags & 0x02) frag->stsd_id = get_be32(pb);
1724 else frag->stsd_id = trex->stsd_id;
1726 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
1727 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
1728 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
1729 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
1730 return 0;
1733 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1735 MOVTrackExt *trex;
1737 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
1738 return -1;
1739 trex = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
1740 if (!trex)
1741 return AVERROR(ENOMEM);
1742 c->trex_data = trex;
1743 trex = &c->trex_data[c->trex_count++];
1744 get_byte(pb); /* version */
1745 get_be24(pb); /* flags */
1746 trex->track_id = get_be32(pb);
1747 trex->stsd_id = get_be32(pb);
1748 trex->duration = get_be32(pb);
1749 trex->size = get_be32(pb);
1750 trex->flags = get_be32(pb);
1751 return 0;
1754 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1756 MOVFragment *frag = &c->fragment;
1757 AVStream *st = NULL;
1758 MOVStreamContext *sc;
1759 uint64_t offset;
1760 int64_t dts;
1761 int data_offset = 0;
1762 unsigned entries, first_sample_flags = frag->flags;
1763 int flags, distance, i;
1765 for (i = 0; i < c->fc->nb_streams; i++) {
1766 if (c->fc->streams[i]->id == frag->track_id) {
1767 st = c->fc->streams[i];
1768 break;
1771 if (!st) {
1772 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding track id %d\n", frag->track_id);
1773 return -1;
1775 sc = st->priv_data;
1776 if (sc->pseudo_stream_id+1 != frag->stsd_id)
1777 return 0;
1778 get_byte(pb); /* version */
1779 flags = get_be24(pb);
1780 entries = get_be32(pb);
1781 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
1782 if (flags & 0x001) data_offset = get_be32(pb);
1783 if (flags & 0x004) first_sample_flags = get_be32(pb);
1784 if (flags & 0x800) {
1785 MOVStts *ctts_data;
1786 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
1787 return -1;
1788 ctts_data = av_realloc(sc->ctts_data,
1789 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
1790 if (!ctts_data)
1791 return AVERROR(ENOMEM);
1792 sc->ctts_data = ctts_data;
1794 dts = st->duration;
1795 offset = frag->base_data_offset + data_offset;
1796 distance = 0;
1797 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
1798 for (i = 0; i < entries; i++) {
1799 unsigned sample_size = frag->size;
1800 int sample_flags = i ? frag->flags : first_sample_flags;
1801 unsigned sample_duration = frag->duration;
1802 int keyframe;
1804 if (flags & 0x100) sample_duration = get_be32(pb);
1805 if (flags & 0x200) sample_size = get_be32(pb);
1806 if (flags & 0x400) sample_flags = get_be32(pb);
1807 if (flags & 0x800) {
1808 sc->ctts_data[sc->ctts_count].count = 1;
1809 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
1810 sc->ctts_count++;
1812 if ((keyframe = st->codec->codec_type == CODEC_TYPE_AUDIO ||
1813 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
1814 distance = 0;
1815 av_add_index_entry(st, offset, dts, sample_size, distance,
1816 keyframe ? AVINDEX_KEYFRAME : 0);
1817 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1818 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
1819 offset, dts, sample_size, distance, keyframe);
1820 distance++;
1821 dts += sample_duration;
1822 offset += sample_size;
1824 frag->moof_offset = offset;
1825 st->duration = dts;
1826 return 0;
1829 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
1830 /* like the files created with Adobe Premiere 5.0, for samples see */
1831 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
1832 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1834 int err;
1836 if (atom.size < 8)
1837 return 0; /* continue */
1838 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
1839 url_fskip(pb, atom.size - 4);
1840 return 0;
1842 atom.type = get_le32(pb);
1843 atom.offset += 8;
1844 atom.size -= 8;
1845 if (atom.type != MKTAG('m','d','a','t')) {
1846 url_fskip(pb, atom.size);
1847 return 0;
1849 err = mov_read_mdat(c, pb, atom);
1850 return err;
1853 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1855 #if CONFIG_ZLIB
1856 ByteIOContext ctx;
1857 uint8_t *cmov_data;
1858 uint8_t *moov_data; /* uncompressed data */
1859 long cmov_len, moov_len;
1860 int ret = -1;
1862 get_be32(pb); /* dcom atom */
1863 if (get_le32(pb) != MKTAG('d','c','o','m'))
1864 return -1;
1865 if (get_le32(pb) != MKTAG('z','l','i','b')) {
1866 av_log(c->fc, AV_LOG_ERROR, "unknown compression for cmov atom !");
1867 return -1;
1869 get_be32(pb); /* cmvd atom */
1870 if (get_le32(pb) != MKTAG('c','m','v','d'))
1871 return -1;
1872 moov_len = get_be32(pb); /* uncompressed size */
1873 cmov_len = atom.size - 6 * 4;
1875 cmov_data = av_malloc(cmov_len);
1876 if (!cmov_data)
1877 return AVERROR(ENOMEM);
1878 moov_data = av_malloc(moov_len);
1879 if (!moov_data) {
1880 av_free(cmov_data);
1881 return AVERROR(ENOMEM);
1883 get_buffer(pb, cmov_data, cmov_len);
1884 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
1885 goto free_and_return;
1886 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
1887 goto free_and_return;
1888 atom.type = MKTAG('m','o','o','v');
1889 atom.offset = 0;
1890 atom.size = moov_len;
1891 #ifdef DEBUG
1892 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
1893 #endif
1894 ret = mov_read_default(c, &ctx, atom);
1895 free_and_return:
1896 av_free(moov_data);
1897 av_free(cmov_data);
1898 return ret;
1899 #else
1900 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
1901 return -1;
1902 #endif
1905 /* edit list atom */
1906 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1908 MOVStreamContext *sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
1909 int i, edit_count;
1911 get_byte(pb); /* version */
1912 get_be24(pb); /* flags */
1913 edit_count = get_be32(pb); /* entries */
1915 if((uint64_t)edit_count*12+8 > atom.size)
1916 return -1;
1918 for(i=0; i<edit_count; i++){
1919 int time;
1920 int duration = get_be32(pb); /* Track duration */
1921 time = get_be32(pb); /* Media time */
1922 get_be32(pb); /* Media rate */
1923 if (i == 0 && time >= -1) {
1924 sc->time_offset = time != -1 ? time : -duration;
1928 if(edit_count > 1)
1929 av_log(c->fc, AV_LOG_WARNING, "multiple edit list entries, "
1930 "a/v desync might occur, patch welcome\n");
1932 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, edit_count);
1933 return 0;
1936 static const MOVParseTableEntry mov_default_parse_table[] = {
1937 { MKTAG('a','v','s','s'), mov_read_extradata },
1938 { MKTAG('c','o','6','4'), mov_read_stco },
1939 { MKTAG('c','s','l','g'), mov_read_cslg },
1940 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
1941 { MKTAG('d','i','n','f'), mov_read_default },
1942 { MKTAG('d','r','e','f'), mov_read_dref },
1943 { MKTAG('e','d','t','s'), mov_read_default },
1944 { MKTAG('e','l','s','t'), mov_read_elst },
1945 { MKTAG('e','n','d','a'), mov_read_enda },
1946 { MKTAG('f','i','e','l'), mov_read_extradata },
1947 { MKTAG('f','t','y','p'), mov_read_ftyp },
1948 { MKTAG('g','l','b','l'), mov_read_glbl },
1949 { MKTAG('h','d','l','r'), mov_read_hdlr },
1950 { MKTAG('i','l','s','t'), mov_read_ilst },
1951 { MKTAG('j','p','2','h'), mov_read_extradata },
1952 { MKTAG('m','d','a','t'), mov_read_mdat },
1953 { MKTAG('m','d','h','d'), mov_read_mdhd },
1954 { MKTAG('m','d','i','a'), mov_read_default },
1955 { MKTAG('m','e','t','a'), mov_read_meta },
1956 { MKTAG('m','i','n','f'), mov_read_default },
1957 { MKTAG('m','o','o','f'), mov_read_moof },
1958 { MKTAG('m','o','o','v'), mov_read_moov },
1959 { MKTAG('m','v','e','x'), mov_read_default },
1960 { MKTAG('m','v','h','d'), mov_read_mvhd },
1961 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
1962 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
1963 { MKTAG('a','v','c','C'), mov_read_glbl },
1964 { MKTAG('p','a','s','p'), mov_read_pasp },
1965 { MKTAG('s','t','b','l'), mov_read_default },
1966 { MKTAG('s','t','c','o'), mov_read_stco },
1967 { MKTAG('s','t','p','s'), mov_read_stps },
1968 { MKTAG('s','t','s','c'), mov_read_stsc },
1969 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
1970 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
1971 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
1972 { MKTAG('s','t','t','s'), mov_read_stts },
1973 { MKTAG('s','t','z','2'), mov_read_stsz }, /* compact sample size */
1974 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
1975 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
1976 { MKTAG('t','r','a','k'), mov_read_trak },
1977 { MKTAG('t','r','a','f'), mov_read_default },
1978 { MKTAG('t','r','e','x'), mov_read_trex },
1979 { MKTAG('t','r','u','n'), mov_read_trun },
1980 { MKTAG('u','d','t','a'), mov_read_default },
1981 { MKTAG('w','a','v','e'), mov_read_wave },
1982 { MKTAG('e','s','d','s'), mov_read_esds },
1983 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
1984 { MKTAG('c','m','o','v'), mov_read_cmov },
1985 { 0, NULL }
1988 static int mov_probe(AVProbeData *p)
1990 unsigned int offset;
1991 uint32_t tag;
1992 int score = 0;
1994 /* check file header */
1995 offset = 0;
1996 for(;;) {
1997 /* ignore invalid offset */
1998 if ((offset + 8) > (unsigned int)p->buf_size)
1999 return score;
2000 tag = AV_RL32(p->buf + offset + 4);
2001 switch(tag) {
2002 /* check for obvious tags */
2003 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
2004 case MKTAG('m','o','o','v'):
2005 case MKTAG('m','d','a','t'):
2006 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
2007 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
2008 case MKTAG('f','t','y','p'):
2009 return AVPROBE_SCORE_MAX;
2010 /* those are more common words, so rate then a bit less */
2011 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
2012 case MKTAG('w','i','d','e'):
2013 case MKTAG('f','r','e','e'):
2014 case MKTAG('j','u','n','k'):
2015 case MKTAG('p','i','c','t'):
2016 return AVPROBE_SCORE_MAX - 5;
2017 case MKTAG(0x82,0x82,0x7f,0x7d):
2018 case MKTAG('s','k','i','p'):
2019 case MKTAG('u','u','i','d'):
2020 case MKTAG('p','r','f','l'):
2021 offset = AV_RB32(p->buf+offset) + offset;
2022 /* if we only find those cause probedata is too small at least rate them */
2023 score = AVPROBE_SCORE_MAX - 50;
2024 break;
2025 default:
2026 /* unrecognized tag */
2027 return score;
2030 return score;
2033 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
2035 MOVContext *mov = s->priv_data;
2036 ByteIOContext *pb = s->pb;
2037 int err;
2038 MOVAtom atom = { 0, 0, 0 };
2040 mov->fc = s;
2041 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
2042 if(!url_is_streamed(pb))
2043 atom.size = url_fsize(pb);
2044 else
2045 atom.size = INT64_MAX;
2047 /* check MOV header */
2048 if ((err = mov_read_default(mov, pb, atom)) < 0) {
2049 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
2050 return err;
2052 if (!mov->found_moov) {
2053 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
2054 return -1;
2056 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
2058 return 0;
2061 static AVIndexEntry *mov_find_next_sample(AVFormatContext *s, AVStream **st)
2063 AVIndexEntry *sample = NULL;
2064 int64_t best_dts = INT64_MAX;
2065 int i;
2066 for (i = 0; i < s->nb_streams; i++) {
2067 AVStream *avst = s->streams[i];
2068 MOVStreamContext *msc = avst->priv_data;
2069 if (msc->pb && msc->current_sample < avst->nb_index_entries) {
2070 AVIndexEntry *current_sample = &avst->index_entries[msc->current_sample];
2071 int64_t dts = av_rescale(current_sample->timestamp, AV_TIME_BASE, msc->time_scale);
2072 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
2073 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
2074 (!url_is_streamed(s->pb) &&
2075 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
2076 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
2077 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
2078 sample = current_sample;
2079 best_dts = dts;
2080 *st = avst;
2084 return sample;
2087 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
2089 MOVContext *mov = s->priv_data;
2090 MOVStreamContext *sc;
2091 AVIndexEntry *sample;
2092 AVStream *st = NULL;
2093 int ret;
2094 retry:
2095 sample = mov_find_next_sample(s, &st);
2096 if (!sample) {
2097 mov->found_mdat = 0;
2098 if (!url_is_streamed(s->pb) ||
2099 mov_read_default(mov, s->pb, (MOVAtom){ 0, 0, INT64_MAX }) < 0 ||
2100 url_feof(s->pb))
2101 return AVERROR_EOF;
2102 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
2103 goto retry;
2105 sc = st->priv_data;
2106 /* must be done just before reading, to avoid infinite loop on sample */
2107 sc->current_sample++;
2109 if (st->discard != AVDISCARD_ALL) {
2110 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
2111 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
2112 sc->ffindex, sample->pos);
2113 return -1;
2115 ret = av_get_packet(sc->pb, pkt, sample->size);
2116 if (ret < 0)
2117 return ret;
2118 #if CONFIG_DV_DEMUXER
2119 if (mov->dv_demux && sc->dv_audio_container) {
2120 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
2121 av_free(pkt->data);
2122 pkt->size = 0;
2123 ret = dv_get_packet(mov->dv_demux, pkt);
2124 if (ret < 0)
2125 return ret;
2127 #endif
2130 pkt->stream_index = sc->ffindex;
2131 pkt->dts = sample->timestamp;
2132 if (sc->ctts_data) {
2133 pkt->pts = pkt->dts + sc->dts_shift + sc->ctts_data[sc->ctts_index].duration;
2134 /* update ctts context */
2135 sc->ctts_sample++;
2136 if (sc->ctts_index < sc->ctts_count &&
2137 sc->ctts_data[sc->ctts_index].count == sc->ctts_sample) {
2138 sc->ctts_index++;
2139 sc->ctts_sample = 0;
2141 if (sc->wrong_dts)
2142 pkt->dts = AV_NOPTS_VALUE;
2143 } else {
2144 int64_t next_dts = (sc->current_sample < st->nb_index_entries) ?
2145 st->index_entries[sc->current_sample].timestamp : st->duration;
2146 pkt->duration = next_dts - pkt->dts;
2147 pkt->pts = pkt->dts;
2149 if (st->discard == AVDISCARD_ALL)
2150 goto retry;
2151 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? PKT_FLAG_KEY : 0;
2152 pkt->pos = sample->pos;
2153 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
2154 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
2155 return 0;
2158 static int mov_seek_stream(AVStream *st, int64_t timestamp, int flags)
2160 MOVStreamContext *sc = st->priv_data;
2161 int sample, time_sample;
2162 int i;
2164 sample = av_index_search_timestamp(st, timestamp, flags);
2165 dprintf(st->codec, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
2166 if (sample < 0) /* not sure what to do */
2167 return -1;
2168 sc->current_sample = sample;
2169 dprintf(st->codec, "stream %d, found sample %d\n", st->index, sc->current_sample);
2170 /* adjust ctts index */
2171 if (sc->ctts_data) {
2172 time_sample = 0;
2173 for (i = 0; i < sc->ctts_count; i++) {
2174 int next = time_sample + sc->ctts_data[i].count;
2175 if (next > sc->current_sample) {
2176 sc->ctts_index = i;
2177 sc->ctts_sample = sc->current_sample - time_sample;
2178 break;
2180 time_sample = next;
2183 return sample;
2186 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
2188 AVStream *st;
2189 int64_t seek_timestamp, timestamp;
2190 int sample;
2191 int i;
2193 if (stream_index >= s->nb_streams)
2194 return -1;
2195 if (sample_time < 0)
2196 sample_time = 0;
2198 st = s->streams[stream_index];
2199 sample = mov_seek_stream(st, sample_time, flags);
2200 if (sample < 0)
2201 return -1;
2203 /* adjust seek timestamp to found sample timestamp */
2204 seek_timestamp = st->index_entries[sample].timestamp;
2206 for (i = 0; i < s->nb_streams; i++) {
2207 st = s->streams[i];
2208 if (stream_index == i)
2209 continue;
2211 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
2212 mov_seek_stream(st, timestamp, flags);
2214 return 0;
2217 static int mov_read_close(AVFormatContext *s)
2219 MOVContext *mov = s->priv_data;
2220 int i, j;
2222 for (i = 0; i < s->nb_streams; i++) {
2223 AVStream *st = s->streams[i];
2224 MOVStreamContext *sc = st->priv_data;
2226 av_freep(&sc->ctts_data);
2227 for (j = 0; j < sc->drefs_count; j++)
2228 av_freep(&sc->drefs[j].path);
2229 av_freep(&sc->drefs);
2230 if (sc->pb && sc->pb != s->pb)
2231 url_fclose(sc->pb);
2233 av_freep(&st->codec->palctrl);
2236 if (mov->dv_demux) {
2237 for(i = 0; i < mov->dv_fctx->nb_streams; i++) {
2238 av_freep(&mov->dv_fctx->streams[i]->codec);
2239 av_freep(&mov->dv_fctx->streams[i]);
2241 av_freep(&mov->dv_fctx);
2242 av_freep(&mov->dv_demux);
2245 av_freep(&mov->trex_data);
2247 return 0;
2250 AVInputFormat mov_demuxer = {
2251 "mov,mp4,m4a,3gp,3g2,mj2",
2252 NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),
2253 sizeof(MOVContext),
2254 mov_probe,
2255 mov_read_header,
2256 mov_read_packet,
2257 mov_read_close,
2258 mov_read_seek,