8f9f50b959e2c053fa27aa03884352fa7bf83820
[libav.git] / libavformat / mov.c
1 /*
2 * MOV demuxer
3 * Copyright (c) 2001 Fabrice Bellard
4 *
5 * This file is part of FFmpeg.
6 *
7 * FFmpeg is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
11 *
12 * FFmpeg is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
20 */
21
22 #include <limits.h>
23
24 //#define DEBUG
25
26 #include "libavutil/intreadwrite.h"
27 #include "avformat.h"
28 #include "riff.h"
29 #include "isom.h"
30 #include "dv.h"
31 #include "libavcodec/mpeg4audio.h"
32 #include "libavcodec/mpegaudiodata.h"
33
34 #if CONFIG_ZLIB
35 #include <zlib.h>
36 #endif
37
38 /*
39 * First version by Francois Revol revol@free.fr
40 * Seek function by Gael Chardon gael.dev@4now.net
41 *
42 * Features and limitations:
43 * - reads most of the QT files I have (at least the structure),
44 * Sample QuickTime files with mp3 audio can be found at: http://www.3ivx.com/showcase.html
45 * - the code is quite ugly... maybe I won't do it recursive next time :-)
46 *
47 * Funny I didn't know about http://sourceforge.net/projects/qt-ffmpeg/
48 * when coding this :) (it's a writer anyway)
49 *
50 * Reference documents:
51 * http://www.geocities.com/xhelmboyx/quicktime/formats/qtm-layout.txt
52 * Apple:
53 * http://developer.apple.com/documentation/QuickTime/QTFF/
54 * http://developer.apple.com/documentation/QuickTime/QTFF/qtff.pdf
55 * QuickTime is a trademark of Apple (AFAIK :))
56 */
57
58 #include "qtpalette.h"
59
60
61 #undef NDEBUG
62 #include <assert.h>
63
64 /* the QuickTime file format is quite convoluted...
65 * it has lots of index tables, each indexing something in another one...
66 * Here we just use what is needed to read the chunks
67 */
68
69 typedef struct {
70 int first;
71 int count;
72 int id;
73 } MOVStsc;
74
75 typedef struct {
76 uint32_t type;
77 char *path;
78 } MOVDref;
79
80 typedef struct {
81 uint32_t type;
82 int64_t offset;
83 int64_t size; /* total size (excluding the size and type fields) */
84 } MOVAtom;
85
86 struct MOVParseTableEntry;
87
88 typedef struct {
89 unsigned track_id;
90 uint64_t base_data_offset;
91 uint64_t moof_offset;
92 unsigned stsd_id;
93 unsigned duration;
94 unsigned size;
95 unsigned flags;
96 } MOVFragment;
97
98 typedef struct {
99 unsigned track_id;
100 unsigned stsd_id;
101 unsigned duration;
102 unsigned size;
103 unsigned flags;
104 } MOVTrackExt;
105
106 typedef struct MOVStreamContext {
107 ByteIOContext *pb;
108 int ffindex; /* the ffmpeg stream id */
109 int next_chunk;
110 unsigned int chunk_count;
111 int64_t *chunk_offsets;
112 unsigned int stts_count;
113 MOVStts *stts_data;
114 unsigned int ctts_count;
115 MOVStts *ctts_data;
116 unsigned int sample_to_chunk_sz;
117 MOVStsc *sample_to_chunk;
118 int ctts_index;
119 int ctts_sample;
120 unsigned int sample_size;
121 unsigned int sample_count;
122 int *sample_sizes;
123 unsigned int keyframe_count;
124 int *keyframes;
125 int time_scale;
126 int time_rate;
127 int time_offset; ///< time offset of the first edit list entry
128 int current_sample;
129 unsigned int bytes_per_frame;
130 unsigned int samples_per_frame;
131 int dv_audio_container;
132 int pseudo_stream_id; ///< -1 means demux all ids
133 int16_t audio_cid; ///< stsd audio compression id
134 unsigned drefs_count;
135 MOVDref *drefs;
136 int dref_id;
137 int wrong_dts; ///< dts are wrong due to negative ctts
138 int width; ///< tkhd width
139 int height; ///< tkhd height
140 } MOVStreamContext;
141
142 typedef struct MOVContext {
143 AVFormatContext *fc;
144 int time_scale;
145 int64_t duration; /* duration of the longest track */
146 int found_moov; /* when both 'moov' and 'mdat' sections has been found */
147 int found_mdat; /* we suppose we have enough data to read the file */
148 AVPaletteControl palette_control;
149 DVDemuxContext *dv_demux;
150 AVFormatContext *dv_fctx;
151 int isom; /* 1 if file is ISO Media (mp4/3gp) */
152 MOVFragment fragment; ///< current fragment in moof atom
153 MOVTrackExt *trex_data;
154 unsigned trex_count;
155 int itunes_metadata; ///< metadata are itunes style
156 } MOVContext;
157
158
159 /* XXX: it's the first time I make a recursive parser I think... sorry if it's ugly :P */
160
161 /* those functions parse an atom */
162 /* return code:
163 0: continue to parse next atom
164 <0: error occurred, exit
165 */
166 /* links atom IDs to parse functions */
167 typedef struct MOVParseTableEntry {
168 uint32_t type;
169 int (*parse)(MOVContext *ctx, ByteIOContext *pb, MOVAtom atom);
170 } MOVParseTableEntry;
171
172 static const MOVParseTableEntry mov_default_parse_table[];
173
174 static int mov_read_default(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
175 {
176 int64_t total_size = 0;
177 MOVAtom a;
178 int i;
179 int err = 0;
180
181 a.offset = atom.offset;
182
183 if (atom.size < 0)
184 atom.size = INT64_MAX;
185 while(((total_size + 8) < atom.size) && !url_feof(pb) && !err) {
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);
191 }
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;
200 }
201 if (a.size == 0) {
202 a.size = atom.size - total_size;
203 if (a.size <= 8)
204 break;
205 }
206 a.size -= 8;
207 if(a.size < 0)
208 break;
209 a.size = FFMIN(a.size, atom.size - total_size);
210
211 for (i = 0; mov_default_parse_table[i].type != 0
212 && mov_default_parse_table[i].type != a.type; i++)
213 /* empty */;
214
215 if (mov_default_parse_table[i].type == 0) { /* skip leaf atoms data */
216 url_fskip(pb, a.size);
217 } else {
218 int64_t start_pos = url_ftell(pb);
219 int64_t left;
220 err = mov_default_parse_table[i].parse(c, pb, a);
221 if (url_is_streamed(pb) && c->found_moov && c->found_mdat)
222 break;
223 left = a.size - url_ftell(pb) + start_pos;
224 if (left > 0) /* skip garbage at atom end */
225 url_fskip(pb, left);
226 }
227
228 a.offset += a.size;
229 total_size += a.size;
230 }
231
232 if (!err && total_size < atom.size && atom.size < 0x7ffff)
233 url_fskip(pb, atom.size - total_size);
234
235 return err;
236 }
237
238 static int mov_read_dref(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
239 {
240 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
241 MOVStreamContext *sc = st->priv_data;
242 int entries, i, j;
243
244 get_be32(pb); // version + flags
245 entries = get_be32(pb);
246 if (entries >= UINT_MAX / sizeof(*sc->drefs))
247 return -1;
248 sc->drefs_count = entries;
249 sc->drefs = av_mallocz(entries * sizeof(*sc->drefs));
250
251 for (i = 0; i < sc->drefs_count; i++) {
252 MOVDref *dref = &sc->drefs[i];
253 uint32_t size = get_be32(pb);
254 int64_t next = url_ftell(pb) + size - 4;
255
256 dref->type = get_le32(pb);
257 get_be32(pb); // version + flags
258 dprintf(c->fc, "type %.4s size %d\n", (char*)&dref->type, size);
259
260 if (dref->type == MKTAG('a','l','i','s') && size > 150) {
261 /* macintosh alias record */
262 uint16_t volume_len, len;
263 char volume[28];
264 int16_t type;
265
266 url_fskip(pb, 10);
267
268 volume_len = get_byte(pb);
269 volume_len = FFMIN(volume_len, 27);
270 get_buffer(pb, volume, 27);
271 volume[volume_len] = 0;
272 av_log(c->fc, AV_LOG_DEBUG, "volume %s, len %d\n", volume, volume_len);
273
274 url_fskip(pb, 112);
275
276 for (type = 0; type != -1 && url_ftell(pb) < next; ) {
277 type = get_be16(pb);
278 len = get_be16(pb);
279 av_log(c->fc, AV_LOG_DEBUG, "type %d, len %d\n", type, len);
280 if (len&1)
281 len += 1;
282 if (type == 2) { // absolute path
283 av_free(dref->path);
284 dref->path = av_mallocz(len+1);
285 if (!dref->path)
286 return AVERROR(ENOMEM);
287 get_buffer(pb, dref->path, len);
288 if (len > volume_len && !strncmp(dref->path, volume, volume_len)) {
289 len -= volume_len;
290 memmove(dref->path, dref->path+volume_len, len);
291 dref->path[len] = 0;
292 }
293 for (j = 0; j < len; j++)
294 if (dref->path[j] == ':')
295 dref->path[j] = '/';
296 av_log(c->fc, AV_LOG_DEBUG, "path %s\n", dref->path);
297 } else
298 url_fskip(pb, len);
299 }
300 }
301 url_fseek(pb, next, SEEK_SET);
302 }
303 return 0;
304 }
305
306 static int mov_read_hdlr(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
307 {
308 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
309 uint32_t type;
310 uint32_t ctype;
311
312 get_byte(pb); /* version */
313 get_be24(pb); /* flags */
314
315 /* component type */
316 ctype = get_le32(pb);
317 type = get_le32(pb); /* component subtype */
318
319 dprintf(c->fc, "ctype= %c%c%c%c (0x%08x)\n", *((char *)&ctype), ((char *)&ctype)[1],
320 ((char *)&ctype)[2], ((char *)&ctype)[3], (int) ctype);
321 dprintf(c->fc, "stype= %c%c%c%c\n",
322 *((char *)&type), ((char *)&type)[1], ((char *)&type)[2], ((char *)&type)[3]);
323 if(!ctype)
324 c->isom = 1;
325 if (type == MKTAG('v','i','d','e'))
326 st->codec->codec_type = CODEC_TYPE_VIDEO;
327 else if(type == MKTAG('s','o','u','n'))
328 st->codec->codec_type = CODEC_TYPE_AUDIO;
329 else if(type == MKTAG('m','1','a',' '))
330 st->codec->codec_id = CODEC_ID_MP2;
331 else if(type == MKTAG('s','u','b','p')) {
332 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
333 }
334 get_be32(pb); /* component manufacture */
335 get_be32(pb); /* component flags */
336 get_be32(pb); /* component flags mask */
337
338 if(atom.size <= 24)
339 return 0; /* nothing left to read */
340
341 url_fskip(pb, atom.size - (url_ftell(pb) - atom.offset));
342 return 0;
343 }
344
345 static int mp4_read_descr_len(ByteIOContext *pb)
346 {
347 int len = 0;
348 int count = 4;
349 while (count--) {
350 int c = get_byte(pb);
351 len = (len << 7) | (c & 0x7f);
352 if (!(c & 0x80))
353 break;
354 }
355 return len;
356 }
357
358 static int mp4_read_descr(MOVContext *c, ByteIOContext *pb, int *tag)
359 {
360 int len;
361 *tag = get_byte(pb);
362 len = mp4_read_descr_len(pb);
363 dprintf(c->fc, "MPEG4 description: tag=0x%02x len=%d\n", *tag, len);
364 return len;
365 }
366
367 #define MP4ESDescrTag 0x03
368 #define MP4DecConfigDescrTag 0x04
369 #define MP4DecSpecificDescrTag 0x05
370
371 static const AVCodecTag mp4_audio_types[] = {
372 { CODEC_ID_MP3ON4, 29 }, /* old mp3on4 draft */
373 { CODEC_ID_MP3ON4, 32 }, /* layer 1 */
374 { CODEC_ID_MP3ON4, 33 }, /* layer 2 */
375 { CODEC_ID_MP3ON4, 34 }, /* layer 3 */
376 { CODEC_ID_NONE, 0 },
377 };
378
379 static int mov_read_esds(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
380 {
381 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
382 int tag, len;
383
384 get_be32(pb); /* version + flags */
385 len = mp4_read_descr(c, pb, &tag);
386 if (tag == MP4ESDescrTag) {
387 get_be16(pb); /* ID */
388 get_byte(pb); /* priority */
389 } else
390 get_be16(pb); /* ID */
391
392 len = mp4_read_descr(c, pb, &tag);
393 if (tag == MP4DecConfigDescrTag) {
394 int object_type_id = get_byte(pb);
395 get_byte(pb); /* stream type */
396 get_be24(pb); /* buffer size db */
397 get_be32(pb); /* max bitrate */
398 get_be32(pb); /* avg bitrate */
399
400 st->codec->codec_id= codec_get_id(ff_mp4_obj_type, object_type_id);
401 dprintf(c->fc, "esds object type id %d\n", object_type_id);
402 len = mp4_read_descr(c, pb, &tag);
403 if (tag == MP4DecSpecificDescrTag) {
404 dprintf(c->fc, "Specific MPEG4 header len=%d\n", len);
405 if((uint64_t)len > (1<<30))
406 return -1;
407 st->codec->extradata = av_mallocz(len + FF_INPUT_BUFFER_PADDING_SIZE);
408 if (!st->codec->extradata)
409 return AVERROR(ENOMEM);
410 get_buffer(pb, st->codec->extradata, len);
411 st->codec->extradata_size = len;
412 if (st->codec->codec_id == CODEC_ID_AAC) {
413 MPEG4AudioConfig cfg;
414 ff_mpeg4audio_get_config(&cfg, st->codec->extradata,
415 st->codec->extradata_size);
416 if (cfg.chan_config > 7)
417 return -1;
418 st->codec->channels = ff_mpeg4audio_channels[cfg.chan_config];
419 if (cfg.object_type == 29 && cfg.sampling_index < 3) // old mp3on4
420 st->codec->sample_rate = ff_mpa_freq_tab[cfg.sampling_index];
421 else
422 st->codec->sample_rate = cfg.sample_rate; // ext sample rate ?
423 dprintf(c->fc, "mp4a config channels %d obj %d ext obj %d "
424 "sample rate %d ext sample rate %d\n", st->codec->channels,
425 cfg.object_type, cfg.ext_object_type,
426 cfg.sample_rate, cfg.ext_sample_rate);
427 if (!(st->codec->codec_id = codec_get_id(mp4_audio_types,
428 cfg.object_type)))
429 st->codec->codec_id = CODEC_ID_AAC;
430 }
431 }
432 }
433 return 0;
434 }
435
436 static int mov_read_pasp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
437 {
438 const int num = get_be32(pb);
439 const int den = get_be32(pb);
440 AVStream * const st = c->fc->streams[c->fc->nb_streams-1];
441 if (den != 0) {
442 if ((st->sample_aspect_ratio.den != 1 || st->sample_aspect_ratio.num) && // default
443 (den != st->sample_aspect_ratio.den || num != st->sample_aspect_ratio.num))
444 av_log(c->fc, AV_LOG_WARNING,
445 "sample aspect ratio already set to %d:%d, overriding by 'pasp' atom\n",
446 st->sample_aspect_ratio.num, st->sample_aspect_ratio.den);
447 st->sample_aspect_ratio.num = num;
448 st->sample_aspect_ratio.den = den;
449 }
450 return 0;
451 }
452
453 /* this atom contains actual media data */
454 static int mov_read_mdat(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
455 {
456 if(atom.size == 0) /* wrong one (MP4) */
457 return 0;
458 c->found_mdat=1;
459 return 0; /* now go for moov */
460 }
461
462 static int mov_read_ftyp(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
463 {
464 uint32_t type = get_le32(pb);
465
466 if (type != MKTAG('q','t',' ',' '))
467 c->isom = 1;
468 av_log(c->fc, AV_LOG_DEBUG, "ISO: File Type Major Brand: %.4s\n",(char *)&type);
469 get_be32(pb); /* minor version */
470 url_fskip(pb, atom.size - 8);
471 return 0;
472 }
473
474 /* this atom should contain all header atoms */
475 static int mov_read_moov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
476 {
477 if (mov_read_default(c, pb, atom) < 0)
478 return -1;
479 /* we parsed the 'moov' atom, we can terminate the parsing as soon as we find the 'mdat' */
480 /* so we don't parse the whole file if over a network */
481 c->found_moov=1;
482 return 0; /* now go for mdat */
483 }
484
485 static int mov_read_moof(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
486 {
487 c->fragment.moof_offset = url_ftell(pb) - 8;
488 dprintf(c->fc, "moof offset %llx\n", c->fragment.moof_offset);
489 return mov_read_default(c, pb, atom);
490 }
491
492 static int mov_read_mdhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
493 {
494 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
495 MOVStreamContext *sc = st->priv_data;
496 int version = get_byte(pb);
497 unsigned lang;
498
499 if (version > 1)
500 return -1; /* unsupported */
501
502 get_be24(pb); /* flags */
503 if (version == 1) {
504 get_be64(pb);
505 get_be64(pb);
506 } else {
507 get_be32(pb); /* creation time */
508 get_be32(pb); /* modification time */
509 }
510
511 sc->time_scale = get_be32(pb);
512 st->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
513
514 lang = get_be16(pb); /* language */
515 ff_mov_lang_to_iso639(lang, st->language);
516 get_be16(pb); /* quality */
517
518 return 0;
519 }
520
521 static int mov_read_mvhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
522 {
523 int version = get_byte(pb); /* version */
524 get_be24(pb); /* flags */
525
526 if (version == 1) {
527 get_be64(pb);
528 get_be64(pb);
529 } else {
530 get_be32(pb); /* creation time */
531 get_be32(pb); /* modification time */
532 }
533 c->time_scale = get_be32(pb); /* time scale */
534
535 dprintf(c->fc, "time scale = %i\n", c->time_scale);
536
537 c->duration = (version == 1) ? get_be64(pb) : get_be32(pb); /* duration */
538 get_be32(pb); /* preferred scale */
539
540 get_be16(pb); /* preferred volume */
541
542 url_fskip(pb, 10); /* reserved */
543
544 url_fskip(pb, 36); /* display matrix */
545
546 get_be32(pb); /* preview time */
547 get_be32(pb); /* preview duration */
548 get_be32(pb); /* poster time */
549 get_be32(pb); /* selection time */
550 get_be32(pb); /* selection duration */
551 get_be32(pb); /* current time */
552 get_be32(pb); /* next track ID */
553
554 return 0;
555 }
556
557 static int mov_read_smi(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
558 {
559 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
560
561 if((uint64_t)atom.size > (1<<30))
562 return -1;
563
564 // currently SVQ3 decoder expect full STSD header - so let's fake it
565 // this should be fixed and just SMI header should be passed
566 av_free(st->codec->extradata);
567 st->codec->extradata = av_mallocz(atom.size + 0x5a + FF_INPUT_BUFFER_PADDING_SIZE);
568 if (!st->codec->extradata)
569 return AVERROR(ENOMEM);
570 st->codec->extradata_size = 0x5a + atom.size;
571 memcpy(st->codec->extradata, "SVQ3", 4); // fake
572 get_buffer(pb, st->codec->extradata + 0x5a, atom.size);
573 dprintf(c->fc, "Reading SMI %"PRId64" %s\n", atom.size, st->codec->extradata + 0x5a);
574 return 0;
575 }
576
577 static int mov_read_enda(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
578 {
579 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
580 int little_endian = get_be16(pb);
581
582 dprintf(c->fc, "enda %d\n", little_endian);
583 if (little_endian == 1) {
584 switch (st->codec->codec_id) {
585 case CODEC_ID_PCM_S24BE:
586 st->codec->codec_id = CODEC_ID_PCM_S24LE;
587 break;
588 case CODEC_ID_PCM_S32BE:
589 st->codec->codec_id = CODEC_ID_PCM_S32LE;
590 break;
591 case CODEC_ID_PCM_F32BE:
592 st->codec->codec_id = CODEC_ID_PCM_F32LE;
593 break;
594 case CODEC_ID_PCM_F64BE:
595 st->codec->codec_id = CODEC_ID_PCM_F64LE;
596 break;
597 default:
598 break;
599 }
600 }
601 return 0;
602 }
603
604 /* FIXME modify qdm2/svq3/h264 decoders to take full atom as extradata */
605 static int mov_read_extradata(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
606 {
607 AVStream *st;
608 uint64_t size;
609 uint8_t *buf;
610
611 if (c->fc->nb_streams < 1) // will happen with jp2 files
612 return 0;
613 st= c->fc->streams[c->fc->nb_streams-1];
614 size= (uint64_t)st->codec->extradata_size + atom.size + 8 + FF_INPUT_BUFFER_PADDING_SIZE;
615 if(size > INT_MAX || (uint64_t)atom.size > INT_MAX)
616 return -1;
617 buf= av_realloc(st->codec->extradata, size);
618 if(!buf)
619 return -1;
620 st->codec->extradata= buf;
621 buf+= st->codec->extradata_size;
622 st->codec->extradata_size= size - FF_INPUT_BUFFER_PADDING_SIZE;
623 AV_WB32( buf , atom.size + 8);
624 AV_WL32( buf + 4, atom.type);
625 get_buffer(pb, buf + 8, atom.size);
626 return 0;
627 }
628
629 static int mov_read_wave(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
630 {
631 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
632
633 if((uint64_t)atom.size > (1<<30))
634 return -1;
635
636 if (st->codec->codec_id == CODEC_ID_QDM2) {
637 // pass all frma atom to codec, needed at least for QDM2
638 av_free(st->codec->extradata);
639 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
640 if (!st->codec->extradata)
641 return AVERROR(ENOMEM);
642 st->codec->extradata_size = atom.size;
643 get_buffer(pb, st->codec->extradata, atom.size);
644 } else if (atom.size > 8) { /* to read frma, esds atoms */
645 if (mov_read_default(c, pb, atom) < 0)
646 return -1;
647 } else
648 url_fskip(pb, atom.size);
649 return 0;
650 }
651
652 /**
653 * This function reads atom content and puts data in extradata without tag
654 * nor size unlike mov_read_extradata.
655 */
656 static int mov_read_glbl(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
657 {
658 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
659
660 if((uint64_t)atom.size > (1<<30))
661 return -1;
662
663 av_free(st->codec->extradata);
664 st->codec->extradata = av_mallocz(atom.size + FF_INPUT_BUFFER_PADDING_SIZE);
665 if (!st->codec->extradata)
666 return AVERROR(ENOMEM);
667 st->codec->extradata_size = atom.size;
668 get_buffer(pb, st->codec->extradata, atom.size);
669 return 0;
670 }
671
672 static int mov_read_stco(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
673 {
674 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
675 MOVStreamContext *sc = st->priv_data;
676 unsigned int i, entries;
677
678 get_byte(pb); /* version */
679 get_be24(pb); /* flags */
680
681 entries = get_be32(pb);
682
683 if(entries >= UINT_MAX/sizeof(int64_t))
684 return -1;
685
686 sc->chunk_count = entries;
687 sc->chunk_offsets = av_malloc(entries * sizeof(int64_t));
688 if (!sc->chunk_offsets)
689 return -1;
690 if (atom.type == MKTAG('s','t','c','o'))
691 for(i=0; i<entries; i++)
692 sc->chunk_offsets[i] = get_be32(pb);
693 else if (atom.type == MKTAG('c','o','6','4'))
694 for(i=0; i<entries; i++)
695 sc->chunk_offsets[i] = get_be64(pb);
696 else
697 return -1;
698
699 return 0;
700 }
701
702 /**
703 * Compute codec id for 'lpcm' tag.
704 * See CoreAudioTypes and AudioStreamBasicDescription at Apple.
705 */
706 static enum CodecID mov_get_lpcm_codec_id(int bps, int flags)
707 {
708 if (flags & 1) { // floating point
709 if (flags & 2) { // big endian
710 if (bps == 32) return CODEC_ID_PCM_F32BE;
711 else if (bps == 64) return CODEC_ID_PCM_F64BE;
712 } else {
713 if (bps == 32) return CODEC_ID_PCM_F32LE;
714 else if (bps == 64) return CODEC_ID_PCM_F64LE;
715 }
716 } else {
717 if (flags & 2) {
718 if (bps == 8)
719 // signed integer
720 if (flags & 4) return CODEC_ID_PCM_S8;
721 else return CODEC_ID_PCM_U8;
722 else if (bps == 16) return CODEC_ID_PCM_S16BE;
723 else if (bps == 24) return CODEC_ID_PCM_S24BE;
724 else if (bps == 32) return CODEC_ID_PCM_S32BE;
725 } else {
726 if (bps == 8)
727 if (flags & 4) return CODEC_ID_PCM_S8;
728 else return CODEC_ID_PCM_U8;
729 else if (bps == 16) return CODEC_ID_PCM_S16LE;
730 else if (bps == 24) return CODEC_ID_PCM_S24LE;
731 else if (bps == 32) return CODEC_ID_PCM_S32LE;
732 }
733 }
734 return CODEC_ID_NONE;
735 }
736
737 static int mov_read_stsd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
738 {
739 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
740 MOVStreamContext *sc = st->priv_data;
741 int j, entries, pseudo_stream_id;
742
743 get_byte(pb); /* version */
744 get_be24(pb); /* flags */
745
746 entries = get_be32(pb);
747
748 for(pseudo_stream_id=0; pseudo_stream_id<entries; pseudo_stream_id++) {
749 //Parsing Sample description table
750 enum CodecID id;
751 int dref_id;
752 MOVAtom a = { 0, 0, 0 };
753 int64_t start_pos = url_ftell(pb);
754 int size = get_be32(pb); /* size */
755 uint32_t format = get_le32(pb); /* data format */
756
757 get_be32(pb); /* reserved */
758 get_be16(pb); /* reserved */
759 dref_id = get_be16(pb);
760
761 if (st->codec->codec_tag &&
762 st->codec->codec_tag != format &&
763 (c->fc->video_codec_id ? codec_get_id(codec_movvideo_tags, format) != c->fc->video_codec_id
764 : st->codec->codec_tag != MKTAG('j','p','e','g'))
765 ){
766 /* Multiple fourcc, we skip JPEG. This is not correct, we should
767 * export it as a separate AVStream but this needs a few changes
768 * in the MOV demuxer, patch welcome. */
769 av_log(c->fc, AV_LOG_WARNING, "multiple fourcc not supported\n");
770 url_fskip(pb, size - (url_ftell(pb) - start_pos));
771 continue;
772 }
773 sc->pseudo_stream_id = st->codec->codec_tag ? -1 : pseudo_stream_id;
774 sc->dref_id= dref_id;
775
776 st->codec->codec_tag = format;
777 id = codec_get_id(codec_movaudio_tags, format);
778 if (id<=0 && (format&0xFFFF) == 'm'+('s'<<8))
779 id = codec_get_id(codec_wav_tags, bswap_32(format)&0xFFFF);
780
781 if (st->codec->codec_type != CODEC_TYPE_VIDEO && id > 0) {
782 st->codec->codec_type = CODEC_TYPE_AUDIO;
783 } else if (st->codec->codec_type != CODEC_TYPE_AUDIO && /* do not overwrite codec type */
784 format && format != MKTAG('m','p','4','s')) { /* skip old asf mpeg4 tag */
785 id = codec_get_id(codec_movvideo_tags, format);
786 if (id <= 0)
787 id = codec_get_id(codec_bmp_tags, format);
788 if (id > 0)
789 st->codec->codec_type = CODEC_TYPE_VIDEO;
790 else if(st->codec->codec_type == CODEC_TYPE_DATA){
791 id = codec_get_id(ff_codec_movsubtitle_tags, format);
792 if(id > 0)
793 st->codec->codec_type = CODEC_TYPE_SUBTITLE;
794 }
795 }
796
797 dprintf(c->fc, "size=%d 4CC= %c%c%c%c codec_type=%d\n", size,
798 (format >> 0) & 0xff, (format >> 8) & 0xff, (format >> 16) & 0xff,
799 (format >> 24) & 0xff, st->codec->codec_type);
800
801 if(st->codec->codec_type==CODEC_TYPE_VIDEO) {
802 uint8_t codec_name[32];
803 unsigned int color_depth;
804 int color_greyscale;
805
806 st->codec->codec_id = id;
807 get_be16(pb); /* version */
808 get_be16(pb); /* revision level */
809 get_be32(pb); /* vendor */
810 get_be32(pb); /* temporal quality */
811 get_be32(pb); /* spatial quality */
812
813 st->codec->width = get_be16(pb); /* width */
814 st->codec->height = get_be16(pb); /* height */
815
816 get_be32(pb); /* horiz resolution */
817 get_be32(pb); /* vert resolution */
818 get_be32(pb); /* data size, always 0 */
819 get_be16(pb); /* frames per samples */
820
821 get_buffer(pb, codec_name, 32); /* codec name, pascal string */
822 if (codec_name[0] <= 31) {
823 memcpy(st->codec->codec_name, &codec_name[1],codec_name[0]);
824 st->codec->codec_name[codec_name[0]] = 0;
825 }
826
827 st->codec->bits_per_coded_sample = get_be16(pb); /* depth */
828 st->codec->color_table_id = get_be16(pb); /* colortable id */
829 dprintf(c->fc, "depth %d, ctab id %d\n",
830 st->codec->bits_per_coded_sample, st->codec->color_table_id);
831 /* figure out the palette situation */
832 color_depth = st->codec->bits_per_coded_sample & 0x1F;
833 color_greyscale = st->codec->bits_per_coded_sample & 0x20;
834
835 /* if the depth is 2, 4, or 8 bpp, file is palettized */
836 if ((color_depth == 2) || (color_depth == 4) ||
837 (color_depth == 8)) {
838 /* for palette traversal */
839 unsigned int color_start, color_count, color_end;
840 unsigned char r, g, b;
841
842 if (color_greyscale) {
843 int color_index, color_dec;
844 /* compute the greyscale palette */
845 st->codec->bits_per_coded_sample = color_depth;
846 color_count = 1 << color_depth;
847 color_index = 255;
848 color_dec = 256 / (color_count - 1);
849 for (j = 0; j < color_count; j++) {
850 r = g = b = color_index;
851 c->palette_control.palette[j] =
852 (r << 16) | (g << 8) | (b);
853 color_index -= color_dec;
854 if (color_index < 0)
855 color_index = 0;
856 }
857 } else if (st->codec->color_table_id) {
858 const uint8_t *color_table;
859 /* if flag bit 3 is set, use the default palette */
860 color_count = 1 << color_depth;
861 if (color_depth == 2)
862 color_table = ff_qt_default_palette_4;
863 else if (color_depth == 4)
864 color_table = ff_qt_default_palette_16;
865 else
866 color_table = ff_qt_default_palette_256;
867
868 for (j = 0; j < color_count; j++) {
869 r = color_table[j * 4 + 0];
870 g = color_table[j * 4 + 1];
871 b = color_table[j * 4 + 2];
872 c->palette_control.palette[j] =
873 (r << 16) | (g << 8) | (b);
874 }
875 } else {
876 /* load the palette from the file */
877 color_start = get_be32(pb);
878 color_count = get_be16(pb);
879 color_end = get_be16(pb);
880 if ((color_start <= 255) &&
881 (color_end <= 255)) {
882 for (j = color_start; j <= color_end; j++) {
883 /* each R, G, or B component is 16 bits;
884 * only use the top 8 bits; skip alpha bytes
885 * up front */
886 get_byte(pb);
887 get_byte(pb);
888 r = get_byte(pb);
889 get_byte(pb);
890 g = get_byte(pb);
891 get_byte(pb);
892 b = get_byte(pb);
893 get_byte(pb);
894 c->palette_control.palette[j] =
895 (r << 16) | (g << 8) | (b);
896 }
897 }
898 }
899 st->codec->palctrl = &c->palette_control;
900 st->codec->palctrl->palette_changed = 1;
901 } else
902 st->codec->palctrl = NULL;
903 } else if(st->codec->codec_type==CODEC_TYPE_AUDIO) {
904 int bits_per_sample, flags;
905 uint16_t version = get_be16(pb);
906
907 st->codec->codec_id = id;
908 get_be16(pb); /* revision level */
909 get_be32(pb); /* vendor */
910
911 st->codec->channels = get_be16(pb); /* channel count */
912 dprintf(c->fc, "audio channels %d\n", st->codec->channels);
913 st->codec->bits_per_coded_sample = get_be16(pb); /* sample size */
914
915 sc->audio_cid = get_be16(pb);
916 get_be16(pb); /* packet size = 0 */
917
918 st->codec->sample_rate = ((get_be32(pb) >> 16));
919
920 //Read QT version 1 fields. In version 0 these do not exist.
921 dprintf(c->fc, "version =%d, isom =%d\n",version,c->isom);
922 if(!c->isom) {
923 if(version==1) {
924 sc->samples_per_frame = get_be32(pb);
925 get_be32(pb); /* bytes per packet */
926 sc->bytes_per_frame = get_be32(pb);
927 get_be32(pb); /* bytes per sample */
928 } else if(version==2) {
929 get_be32(pb); /* sizeof struct only */
930 st->codec->sample_rate = av_int2dbl(get_be64(pb)); /* float 64 */
931 st->codec->channels = get_be32(pb);
932 get_be32(pb); /* always 0x7F000000 */
933 st->codec->bits_per_coded_sample = get_be32(pb); /* bits per channel if sound is uncompressed */
934 flags = get_be32(pb); /* lcpm format specific flag */
935 sc->bytes_per_frame = get_be32(pb); /* bytes per audio packet if constant */
936 sc->samples_per_frame = get_be32(pb); /* lpcm frames per audio packet if constant */
937 if (format == MKTAG('l','p','c','m'))
938 st->codec->codec_id = mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, flags);
939 }
940 }
941
942 switch (st->codec->codec_id) {
943 case CODEC_ID_PCM_S8:
944 case CODEC_ID_PCM_U8:
945 if (st->codec->bits_per_coded_sample == 16)
946 st->codec->codec_id = CODEC_ID_PCM_S16BE;
947 break;
948 case CODEC_ID_PCM_S16LE:
949 case CODEC_ID_PCM_S16BE:
950 if (st->codec->bits_per_coded_sample == 8)
951 st->codec->codec_id = CODEC_ID_PCM_S8;
952 else if (st->codec->bits_per_coded_sample == 24)
953 st->codec->codec_id =
954 st->codec->codec_id == CODEC_ID_PCM_S16BE ?
955 CODEC_ID_PCM_S24BE : CODEC_ID_PCM_S24LE;
956 break;
957 /* set values for old format before stsd version 1 appeared */
958 case CODEC_ID_MACE3:
959 sc->samples_per_frame = 6;
960 sc->bytes_per_frame = 2*st->codec->channels;
961 break;
962 case CODEC_ID_MACE6:
963 sc->samples_per_frame = 6;
964 sc->bytes_per_frame = 1*st->codec->channels;
965 break;
966 case CODEC_ID_ADPCM_IMA_QT:
967 sc->samples_per_frame = 64;
968 sc->bytes_per_frame = 34*st->codec->channels;
969 break;
970 case CODEC_ID_GSM:
971 sc->samples_per_frame = 160;
972 sc->bytes_per_frame = 33;
973 break;
974 default:
975 break;
976 }
977
978 bits_per_sample = av_get_bits_per_sample(st->codec->codec_id);
979 if (bits_per_sample) {
980 st->codec->bits_per_coded_sample = bits_per_sample;
981 sc->sample_size = (bits_per_sample >> 3) * st->codec->channels;
982 }
983 } else if(st->codec->codec_type==CODEC_TYPE_SUBTITLE){
984 // ttxt stsd contains display flags, justification, background
985 // color, fonts, and default styles, so fake an atom to read it
986 MOVAtom fake_atom = { .size = size - (url_ftell(pb) - start_pos) };
987 mov_read_glbl(c, pb, fake_atom);
988 st->codec->codec_id= id;
989 st->codec->width = sc->width;
990 st->codec->height = sc->height;
991 } else {
992 /* other codec type, just skip (rtp, mp4s, tmcd ...) */
993 url_fskip(pb, size - (url_ftell(pb) - start_pos));
994 }
995 /* this will read extra atoms at the end (wave, alac, damr, avcC, SMI ...) */
996 a.size = size - (url_ftell(pb) - start_pos);
997 if (a.size > 8) {
998 if (mov_read_default(c, pb, a) < 0)
999 return -1;
1000 } else if (a.size > 0)
1001 url_fskip(pb, a.size);
1002 }
1003
1004 if(st->codec->codec_type==CODEC_TYPE_AUDIO && st->codec->sample_rate==0 && sc->time_scale>1)
1005 st->codec->sample_rate= sc->time_scale;
1006
1007 /* special codec parameters handling */
1008 switch (st->codec->codec_id) {
1009 #if CONFIG_DV_DEMUXER
1010 case CODEC_ID_DVAUDIO:
1011 c->dv_fctx = avformat_alloc_context();
1012 c->dv_demux = dv_init_demux(c->dv_fctx);
1013 if (!c->dv_demux) {
1014 av_log(c->fc, AV_LOG_ERROR, "dv demux context init error\n");
1015 return -1;
1016 }
1017 sc->dv_audio_container = 1;
1018 st->codec->codec_id = CODEC_ID_PCM_S16LE;
1019 break;
1020 #endif
1021 /* no ifdef since parameters are always those */
1022 case CODEC_ID_QCELP:
1023 st->codec->frame_size= 160;
1024 st->codec->channels= 1; /* really needed */
1025 break;
1026 case CODEC_ID_AMR_NB:
1027 case CODEC_ID_AMR_WB:
1028 st->codec->frame_size= sc->samples_per_frame;
1029 st->codec->channels= 1; /* really needed */
1030 /* force sample rate for amr, stsd in 3gp does not store sample rate */
1031 if (st->codec->codec_id == CODEC_ID_AMR_NB)
1032 st->codec->sample_rate = 8000;
1033 else if (st->codec->codec_id == CODEC_ID_AMR_WB)
1034 st->codec->sample_rate = 16000;
1035 break;
1036 case CODEC_ID_MP2:
1037 case CODEC_ID_MP3:
1038 st->codec->codec_type = CODEC_TYPE_AUDIO; /* force type after stsd for m1a hdlr */
1039 st->need_parsing = AVSTREAM_PARSE_FULL;
1040 break;
1041 case CODEC_ID_GSM:
1042 case CODEC_ID_ADPCM_MS:
1043 case CODEC_ID_ADPCM_IMA_WAV:
1044 st->codec->block_align = sc->bytes_per_frame;
1045 break;
1046 case CODEC_ID_ALAC:
1047 if (st->codec->extradata_size == 36) {
1048 st->codec->frame_size = AV_RB32(st->codec->extradata+12);
1049 st->codec->channels = AV_RB8 (st->codec->extradata+21);
1050 }
1051 break;
1052 default:
1053 break;
1054 }
1055
1056 return 0;
1057 }
1058
1059 static int mov_read_stsc(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1060 {
1061 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1062 MOVStreamContext *sc = st->priv_data;
1063 unsigned int i, entries;
1064
1065 get_byte(pb); /* version */
1066 get_be24(pb); /* flags */
1067
1068 entries = get_be32(pb);
1069
1070 if(entries >= UINT_MAX / sizeof(*sc->sample_to_chunk))
1071 return -1;
1072
1073 dprintf(c->fc, "track[%i].stsc.entries = %i\n", c->fc->nb_streams-1, entries);
1074
1075 sc->sample_to_chunk_sz = entries;
1076 sc->sample_to_chunk = av_malloc(entries * sizeof(*sc->sample_to_chunk));
1077 if (!sc->sample_to_chunk)
1078 return -1;
1079 for(i=0; i<entries; i++) {
1080 sc->sample_to_chunk[i].first = get_be32(pb);
1081 sc->sample_to_chunk[i].count = get_be32(pb);
1082 sc->sample_to_chunk[i].id = get_be32(pb);
1083 }
1084 return 0;
1085 }
1086
1087 static int mov_read_stss(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1088 {
1089 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1090 MOVStreamContext *sc = st->priv_data;
1091 unsigned int i, entries;
1092
1093 get_byte(pb); /* version */
1094 get_be24(pb); /* flags */
1095
1096 entries = get_be32(pb);
1097
1098 if(entries >= UINT_MAX / sizeof(int))
1099 return -1;
1100
1101 sc->keyframe_count = entries;
1102
1103 dprintf(c->fc, "keyframe_count = %d\n", sc->keyframe_count);
1104
1105 sc->keyframes = av_malloc(entries * sizeof(int));
1106 if (!sc->keyframes)
1107 return -1;
1108 for(i=0; i<entries; i++) {
1109 sc->keyframes[i] = get_be32(pb);
1110 //dprintf(c->fc, "keyframes[]=%d\n", sc->keyframes[i]);
1111 }
1112 return 0;
1113 }
1114
1115 static int mov_read_stsz(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1116 {
1117 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1118 MOVStreamContext *sc = st->priv_data;
1119 unsigned int i, entries, sample_size;
1120
1121 get_byte(pb); /* version */
1122 get_be24(pb); /* flags */
1123
1124 sample_size = get_be32(pb);
1125 if (!sc->sample_size) /* do not overwrite value computed in stsd */
1126 sc->sample_size = sample_size;
1127 entries = get_be32(pb);
1128 if(entries >= UINT_MAX / sizeof(int))
1129 return -1;
1130
1131 sc->sample_count = entries;
1132 if (sample_size)
1133 return 0;
1134
1135 dprintf(c->fc, "sample_size = %d sample_count = %d\n", sc->sample_size, sc->sample_count);
1136
1137 sc->sample_sizes = av_malloc(entries * sizeof(int));
1138 if (!sc->sample_sizes)
1139 return -1;
1140 for(i=0; i<entries; i++)
1141 sc->sample_sizes[i] = get_be32(pb);
1142 return 0;
1143 }
1144
1145 static int mov_read_stts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1146 {
1147 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1148 MOVStreamContext *sc = st->priv_data;
1149 unsigned int i, entries;
1150 int64_t duration=0;
1151 int64_t total_sample_count=0;
1152
1153 get_byte(pb); /* version */
1154 get_be24(pb); /* flags */
1155 entries = get_be32(pb);
1156 if(entries >= UINT_MAX / sizeof(*sc->stts_data))
1157 return -1;
1158
1159 sc->stts_count = entries;
1160 sc->stts_data = av_malloc(entries * sizeof(*sc->stts_data));
1161 if (!sc->stts_data)
1162 return -1;
1163 dprintf(c->fc, "track[%i].stts.entries = %i\n", c->fc->nb_streams-1, entries);
1164
1165 sc->time_rate=0;
1166
1167 for(i=0; i<entries; i++) {
1168 int sample_duration;
1169 int sample_count;
1170
1171 sample_count=get_be32(pb);
1172 sample_duration = get_be32(pb);
1173 sc->stts_data[i].count= sample_count;
1174 sc->stts_data[i].duration= sample_duration;
1175
1176 sc->time_rate= av_gcd(sc->time_rate, sample_duration);
1177
1178 dprintf(c->fc, "sample_count=%d, sample_duration=%d\n",sample_count,sample_duration);
1179
1180 duration+=(int64_t)sample_duration*sample_count;
1181 total_sample_count+=sample_count;
1182 }
1183
1184 st->nb_frames= total_sample_count;
1185 if(duration)
1186 st->duration= duration;
1187 return 0;
1188 }
1189
1190 static int mov_read_ctts(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1191 {
1192 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1193 MOVStreamContext *sc = st->priv_data;
1194 unsigned int i, entries;
1195
1196 get_byte(pb); /* version */
1197 get_be24(pb); /* flags */
1198 entries = get_be32(pb);
1199 if(entries >= UINT_MAX / sizeof(*sc->ctts_data))
1200 return -1;
1201
1202 sc->ctts_count = entries;
1203 sc->ctts_data = av_malloc(entries * sizeof(*sc->ctts_data));
1204 if (!sc->ctts_data)
1205 return -1;
1206 dprintf(c->fc, "track[%i].ctts.entries = %i\n", c->fc->nb_streams-1, entries);
1207
1208 for(i=0; i<entries; i++) {
1209 int count =get_be32(pb);
1210 int duration =get_be32(pb);
1211
1212 if (duration < 0) {
1213 sc->wrong_dts = 1;
1214 st->codec->has_b_frames = 1;
1215 }
1216 sc->ctts_data[i].count = count;
1217 sc->ctts_data[i].duration= duration;
1218
1219 sc->time_rate= av_gcd(sc->time_rate, FFABS(duration));
1220 }
1221 return 0;
1222 }
1223
1224 static void mov_build_index(MOVContext *mov, AVStream *st)
1225 {
1226 MOVStreamContext *sc = st->priv_data;
1227 int64_t current_offset;
1228 int64_t current_dts = 0;
1229 unsigned int stts_index = 0;
1230 unsigned int stsc_index = 0;
1231 unsigned int stss_index = 0;
1232 unsigned int i, j;
1233
1234 /* adjust first dts according to edit list */
1235 if (sc->time_offset) {
1236 assert(sc->time_offset % sc->time_rate == 0);
1237 current_dts = - (sc->time_offset / sc->time_rate);
1238 }
1239
1240 /* only use old uncompressed audio chunk demuxing when stts specifies it */
1241 if (!(st->codec->codec_type == CODEC_TYPE_AUDIO &&
1242 sc->stts_count == 1 && sc->stts_data[0].duration == 1)) {
1243 unsigned int current_sample = 0;
1244 unsigned int stts_sample = 0;
1245 unsigned int keyframe, sample_size;
1246 unsigned int distance = 0;
1247 int key_off = sc->keyframes && sc->keyframes[0] == 1;
1248
1249 st->nb_frames = sc->sample_count;
1250 for (i = 0; i < sc->chunk_count; i++) {
1251 current_offset = sc->chunk_offsets[i];
1252 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1253 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1254 stsc_index++;
1255 for (j = 0; j < sc->sample_to_chunk[stsc_index].count; j++) {
1256 if (current_sample >= sc->sample_count) {
1257 av_log(mov->fc, AV_LOG_ERROR, "wrong sample count\n");
1258 goto out;
1259 }
1260 keyframe = !sc->keyframe_count || current_sample+key_off == sc->keyframes[stss_index];
1261 if (keyframe) {
1262 distance = 0;
1263 if (stss_index + 1 < sc->keyframe_count)
1264 stss_index++;
1265 }
1266 sample_size = sc->sample_size > 0 ? sc->sample_size : sc->sample_sizes[current_sample];
1267 if(sc->pseudo_stream_id == -1 ||
1268 sc->sample_to_chunk[stsc_index].id - 1 == sc->pseudo_stream_id) {
1269 av_add_index_entry(st, current_offset, current_dts, sample_size, distance,
1270 keyframe ? AVINDEX_KEYFRAME : 0);
1271 dprintf(mov->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1272 "size %d, distance %d, keyframe %d\n", st->index, current_sample,
1273 current_offset, current_dts, sample_size, distance, keyframe);
1274 }
1275 current_offset += sample_size;
1276 assert(sc->stts_data[stts_index].duration % sc->time_rate == 0);
1277 current_dts += sc->stts_data[stts_index].duration / sc->time_rate;
1278 distance++;
1279 stts_sample++;
1280 current_sample++;
1281 if (stts_index + 1 < sc->stts_count && stts_sample == sc->stts_data[stts_index].count) {
1282 stts_sample = 0;
1283 stts_index++;
1284 }
1285 }
1286 }
1287 } else { /* read whole chunk */
1288 unsigned int chunk_samples, chunk_size, chunk_duration;
1289 unsigned int frames = 1;
1290 for (i = 0; i < sc->chunk_count; i++) {
1291 current_offset = sc->chunk_offsets[i];
1292 if (stsc_index + 1 < sc->sample_to_chunk_sz &&
1293 i + 1 == sc->sample_to_chunk[stsc_index + 1].first)
1294 stsc_index++;
1295 chunk_samples = sc->sample_to_chunk[stsc_index].count;
1296 /* get chunk size, beware of alaw/ulaw/mace */
1297 if (sc->samples_per_frame > 0 &&
1298 (chunk_samples * sc->bytes_per_frame % sc->samples_per_frame == 0)) {
1299 if (sc->samples_per_frame < 160)
1300 chunk_size = chunk_samples * sc->bytes_per_frame / sc->samples_per_frame;
1301 else {
1302 chunk_size = sc->bytes_per_frame;
1303 frames = chunk_samples / sc->samples_per_frame;
1304 chunk_samples = sc->samples_per_frame;
1305 }
1306 } else
1307 chunk_size = chunk_samples * sc->sample_size;
1308 for (j = 0; j < frames; j++) {
1309 av_add_index_entry(st, current_offset, current_dts, chunk_size, 0, AVINDEX_KEYFRAME);
1310 /* get chunk duration */
1311 chunk_duration = 0;
1312 while (chunk_samples > 0) {
1313 if (chunk_samples < sc->stts_data[stts_index].count) {
1314 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1315 sc->stts_data[stts_index].count -= chunk_samples;
1316 break;
1317 } else {
1318 chunk_duration += sc->stts_data[stts_index].duration * chunk_samples;
1319 chunk_samples -= sc->stts_data[stts_index].count;
1320 if (stts_index + 1 < sc->stts_count)
1321 stts_index++;
1322 }
1323 }
1324 current_offset += sc->bytes_per_frame;
1325 dprintf(mov->fc, "AVIndex stream %d, chunk %d, offset %"PRIx64", dts %"PRId64", "
1326 "size %d, duration %d\n", st->index, i, current_offset, current_dts,
1327 chunk_size, chunk_duration);
1328 assert(chunk_duration % sc->time_rate == 0);
1329 current_dts += chunk_duration / sc->time_rate;
1330 }
1331 }
1332 }
1333 out:
1334 /* adjust sample count to avindex entries */
1335 sc->sample_count = st->nb_index_entries;
1336 }
1337
1338 static int mov_read_trak(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1339 {
1340 AVStream *st;
1341 MOVStreamContext *sc;
1342 int ret;
1343
1344 st = av_new_stream(c->fc, c->fc->nb_streams);
1345 if (!st) return AVERROR(ENOMEM);
1346 sc = av_mallocz(sizeof(MOVStreamContext));
1347 if (!sc) return AVERROR(ENOMEM);
1348
1349 st->priv_data = sc;
1350 st->codec->codec_type = CODEC_TYPE_DATA;
1351 st->start_time = 0; /* XXX: check */
1352 sc->ffindex = st->index;
1353
1354 if ((ret = mov_read_default(c, pb, atom)) < 0)
1355 return ret;
1356
1357 /* sanity checks */
1358 if(sc->chunk_count && (!sc->stts_count || !sc->sample_to_chunk_sz ||
1359 (!sc->sample_size && !sc->sample_count))){
1360 av_log(c->fc, AV_LOG_ERROR, "stream %d, missing mandatory atoms, broken header\n",
1361 st->index);
1362 sc->sample_count = 0; //ignore track
1363 return 0;
1364 }
1365 if(!sc->time_rate)
1366 sc->time_rate=1;
1367 if(!sc->time_scale)
1368 sc->time_scale= c->time_scale;
1369 av_set_pts_info(st, 64, sc->time_rate, sc->time_scale);
1370
1371 if (st->codec->codec_type == CODEC_TYPE_AUDIO &&
1372 !st->codec->frame_size && sc->stts_count == 1)
1373 st->codec->frame_size = av_rescale(sc->time_rate, st->codec->sample_rate, sc->time_scale);
1374
1375 if(st->duration != AV_NOPTS_VALUE){
1376 assert(st->duration % sc->time_rate == 0);
1377 st->duration /= sc->time_rate;
1378 }
1379
1380 mov_build_index(c, st);
1381
1382 if (sc->dref_id-1 < sc->drefs_count && sc->drefs[sc->dref_id-1].path) {
1383 if (url_fopen(&sc->pb, sc->drefs[sc->dref_id-1].path, URL_RDONLY) < 0)
1384 av_log(c->fc, AV_LOG_ERROR, "stream %d, error opening file %s: %s\n",
1385 st->index, sc->drefs[sc->dref_id-1].path, strerror(errno));
1386 } else
1387 sc->pb = c->fc->pb;
1388
1389 switch (st->codec->codec_id) {
1390 #if CONFIG_H261_DECODER
1391 case CODEC_ID_H261:
1392 #endif
1393 #if CONFIG_H263_DECODER
1394 case CODEC_ID_H263:
1395 #endif
1396 #if CONFIG_MPEG4_DECODER
1397 case CODEC_ID_MPEG4:
1398 #endif
1399 st->codec->width= 0; /* let decoder init width/height */
1400 st->codec->height= 0;
1401 break;
1402 }
1403
1404 /* Do not need those anymore. */
1405 av_freep(&sc->chunk_offsets);
1406 av_freep(&sc->sample_to_chunk);
1407 av_freep(&sc->sample_sizes);
1408 av_freep(&sc->keyframes);
1409 av_freep(&sc->stts_data);
1410
1411 return 0;
1412 }
1413
1414 static int mov_read_ilst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1415 {
1416 int ret;
1417 c->itunes_metadata = 1;
1418 ret = mov_read_default(c, pb, atom);
1419 c->itunes_metadata = 0;
1420 return ret;
1421 }
1422
1423 static int mov_read_meta(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1424 {
1425 url_fskip(pb, 4); // version + flags
1426 atom.size -= 4;
1427 return mov_read_default(c, pb, atom);
1428 }
1429
1430 static int mov_read_trkn(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1431 {
1432 get_be32(pb); // type
1433 get_be32(pb); // unknown
1434 c->fc->track = get_be32(pb);
1435 dprintf(c->fc, "%.4s %d\n", (char*)&atom.type, c->fc->track);
1436 return 0;
1437 }
1438
1439 static int mov_read_udta_string(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1440 {
1441 char *str = NULL;
1442 int size;
1443 uint16_t str_size;
1444
1445 if (c->itunes_metadata) {
1446 int data_size = get_be32(pb);
1447 int tag = get_le32(pb);
1448 if (tag == MKTAG('d','a','t','a')) {
1449 get_be32(pb); // type
1450 get_be32(pb); // unknown
1451 str_size = data_size - 16;
1452 atom.size -= 16;
1453 } else return 0;
1454 } else {
1455 str_size = get_be16(pb); // string length
1456 get_be16(pb); // language
1457 atom.size -= 4;
1458 }
1459 switch (atom.type) {
1460 case MKTAG(0xa9,'n','a','m'):
1461 str = c->fc->title; size = sizeof(c->fc->title); break;
1462 case MKTAG(0xa9,'A','R','T'):
1463 case MKTAG(0xa9,'w','r','t'):
1464 str = c->fc->author; size = sizeof(c->fc->author); break;
1465 case MKTAG(0xa9,'c','p','y'):
1466 str = c->fc->copyright; size = sizeof(c->fc->copyright); break;
1467 case MKTAG(0xa9,'c','m','t'):
1468 case MKTAG(0xa9,'i','n','f'):
1469 str = c->fc->comment; size = sizeof(c->fc->comment); break;
1470 case MKTAG(0xa9,'a','l','b'):
1471 str = c->fc->album; size = sizeof(c->fc->album); break;
1472 }
1473 if (!str)
1474 return 0;
1475 if (atom.size < 0)
1476 return -1;
1477
1478 get_buffer(pb, str, FFMIN3(size, str_size, atom.size));
1479 dprintf(c->fc, "%.4s %s %d %lld\n", (char*)&atom.type, str, str_size, atom.size);
1480 return 0;
1481 }
1482
1483 static int mov_read_tkhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1484 {
1485 int i;
1486 int width;
1487 int height;
1488 int64_t disp_transform[2];
1489 int display_matrix[3][2];
1490 AVStream *st = c->fc->streams[c->fc->nb_streams-1];
1491 MOVStreamContext *sc = st->priv_data;
1492 int version = get_byte(pb);
1493
1494 get_be24(pb); /* flags */
1495 /*
1496 MOV_TRACK_ENABLED 0x0001
1497 MOV_TRACK_IN_MOVIE 0x0002
1498 MOV_TRACK_IN_PREVIEW 0x0004
1499 MOV_TRACK_IN_POSTER 0x0008
1500 */
1501
1502 if (version == 1) {
1503 get_be64(pb);
1504 get_be64(pb);
1505 } else {
1506 get_be32(pb); /* creation time */
1507 get_be32(pb); /* modification time */
1508 }
1509 st->id = (int)get_be32(pb); /* track id (NOT 0 !)*/
1510 get_be32(pb); /* reserved */
1511 st->start_time = 0; /* check */
1512 /* highlevel (considering edits) duration in movie timebase */
1513 (version == 1) ? get_be64(pb) : get_be32(pb);
1514 get_be32(pb); /* reserved */
1515 get_be32(pb); /* reserved */
1516
1517 get_be16(pb); /* layer */
1518 get_be16(pb); /* alternate group */
1519 get_be16(pb); /* volume */
1520 get_be16(pb); /* reserved */
1521
1522 //read in the display matrix (outlined in ISO 14496-12, Section 6.2.2)
1523 // they're kept in fixed point format through all calculations
1524 // ignore u,v,z b/c we don't need the scale factor to calc aspect ratio
1525 for (i = 0; i < 3; i++) {
1526 display_matrix[i][0] = get_be32(pb); // 16.16 fixed point
1527 display_matrix[i][1] = get_be32(pb); // 16.16 fixed point
1528 get_be32(pb); // 2.30 fixed point (not used)
1529 }
1530
1531 width = get_be32(pb); // 16.16 fixed point track width
1532 height = get_be32(pb); // 16.16 fixed point track height
1533 sc->width = width >> 16;
1534 sc->height = height >> 16;
1535
1536 //transform the display width/height according to the matrix
1537 // skip this if the display matrix is the default identity matrix
1538 // to keep the same scale, use [width height 1<<16]
1539 if (width && height &&
1540 (display_matrix[0][0] != 65536 || display_matrix[0][1] ||
1541 display_matrix[1][0] || display_matrix[1][1] != 65536 ||
1542 display_matrix[2][0] || display_matrix[2][1])) {
1543 for (i = 0; i < 2; i++)
1544 disp_transform[i] =
1545 (int64_t) width * display_matrix[0][i] +
1546 (int64_t) height * display_matrix[1][i] +
1547 ((int64_t) display_matrix[2][i] << 16);
1548
1549 //sample aspect ratio is new width/height divided by old width/height
1550 st->sample_aspect_ratio = av_d2q(
1551 ((double) disp_transform[0] * height) /
1552 ((double) disp_transform[1] * width), INT_MAX);
1553 }
1554 return 0;
1555 }
1556
1557 static int mov_read_tfhd(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1558 {
1559 MOVFragment *frag = &c->fragment;
1560 MOVTrackExt *trex = NULL;
1561 int flags, track_id, i;
1562
1563 get_byte(pb); /* version */
1564 flags = get_be24(pb);
1565
1566 track_id = get_be32(pb);
1567 if (!track_id || track_id > c->fc->nb_streams)
1568 return -1;
1569 frag->track_id = track_id;
1570 for (i = 0; i < c->trex_count; i++)
1571 if (c->trex_data[i].track_id == frag->track_id) {
1572 trex = &c->trex_data[i];
1573 break;
1574 }
1575 if (!trex) {
1576 av_log(c->fc, AV_LOG_ERROR, "could not find corresponding trex\n");
1577 return -1;
1578 }
1579
1580 if (flags & 0x01) frag->base_data_offset = get_be64(pb);
1581 else frag->base_data_offset = frag->moof_offset;
1582 if (flags & 0x02) frag->stsd_id = get_be32(pb);
1583 else frag->stsd_id = trex->stsd_id;
1584
1585 frag->duration = flags & 0x08 ? get_be32(pb) : trex->duration;
1586 frag->size = flags & 0x10 ? get_be32(pb) : trex->size;
1587 frag->flags = flags & 0x20 ? get_be32(pb) : trex->flags;
1588 dprintf(c->fc, "frag flags 0x%x\n", frag->flags);
1589 return 0;
1590 }
1591
1592 static int mov_read_trex(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1593 {
1594 MOVTrackExt *trex;
1595
1596 if ((uint64_t)c->trex_count+1 >= UINT_MAX / sizeof(*c->trex_data))
1597 return -1;
1598 c->trex_data = av_realloc(c->trex_data, (c->trex_count+1)*sizeof(*c->trex_data));
1599 if (!c->trex_data)
1600 return AVERROR(ENOMEM);
1601 trex = &c->trex_data[c->trex_count++];
1602 get_byte(pb); /* version */
1603 get_be24(pb); /* flags */
1604 trex->track_id = get_be32(pb);
1605 trex->stsd_id = get_be32(pb);
1606 trex->duration = get_be32(pb);
1607 trex->size = get_be32(pb);
1608 trex->flags = get_be32(pb);
1609 return 0;
1610 }
1611
1612 static int mov_read_trun(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1613 {
1614 MOVFragment *frag = &c->fragment;
1615 AVStream *st;
1616 MOVStreamContext *sc;
1617 uint64_t offset;
1618 int64_t dts;
1619 int data_offset = 0;
1620 unsigned entries, first_sample_flags = frag->flags;
1621 int flags, distance, i;
1622
1623 if (!frag->track_id || frag->track_id > c->fc->nb_streams)
1624 return -1;
1625 st = c->fc->streams[frag->track_id-1];
1626 sc = st->priv_data;
1627 if (sc->pseudo_stream_id+1 != frag->stsd_id)
1628 return 0;
1629 get_byte(pb); /* version */
1630 flags = get_be24(pb);
1631 entries = get_be32(pb);
1632 dprintf(c->fc, "flags 0x%x entries %d\n", flags, entries);
1633 if (flags & 0x001) data_offset = get_be32(pb);
1634 if (flags & 0x004) first_sample_flags = get_be32(pb);
1635 if (flags & 0x800) {
1636 if ((uint64_t)entries+sc->ctts_count >= UINT_MAX/sizeof(*sc->ctts_data))
1637 return -1;
1638 sc->ctts_data = av_realloc(sc->ctts_data,
1639 (entries+sc->ctts_count)*sizeof(*sc->ctts_data));
1640 if (!sc->ctts_data)
1641 return AVERROR(ENOMEM);
1642 }
1643 dts = st->duration;
1644 offset = frag->base_data_offset + data_offset;
1645 distance = 0;
1646 dprintf(c->fc, "first sample flags 0x%x\n", first_sample_flags);
1647 for (i = 0; i < entries; i++) {
1648 unsigned sample_size = frag->size;
1649 int sample_flags = i ? frag->flags : first_sample_flags;
1650 unsigned sample_duration = frag->duration;
1651 int keyframe;
1652
1653 if (flags & 0x100) sample_duration = get_be32(pb);
1654 if (flags & 0x200) sample_size = get_be32(pb);
1655 if (flags & 0x400) sample_flags = get_be32(pb);
1656 if (flags & 0x800) {
1657 sc->ctts_data[sc->ctts_count].count = 1;
1658 sc->ctts_data[sc->ctts_count].duration = get_be32(pb);
1659 sc->ctts_count++;
1660 }
1661 if ((keyframe = st->codec->codec_type == CODEC_TYPE_AUDIO ||
1662 (flags & 0x004 && !i && !sample_flags) || sample_flags & 0x2000000))
1663 distance = 0;
1664 av_add_index_entry(st, offset, dts, sample_size, distance,
1665 keyframe ? AVINDEX_KEYFRAME : 0);
1666 dprintf(c->fc, "AVIndex stream %d, sample %d, offset %"PRIx64", dts %"PRId64", "
1667 "size %d, distance %d, keyframe %d\n", st->index, sc->sample_count+i,
1668 offset, dts, sample_size, distance, keyframe);
1669 distance++;
1670 assert(sample_duration % sc->time_rate == 0);
1671 dts += sample_duration / sc->time_rate;
1672 offset += sample_size;
1673 }
1674 frag->moof_offset = offset;
1675 sc->sample_count = st->nb_index_entries;
1676 st->duration = dts;
1677 return 0;
1678 }
1679
1680 /* this atom should be null (from specs), but some buggy files put the 'moov' atom inside it... */
1681 /* like the files created with Adobe Premiere 5.0, for samples see */
1682 /* http://graphics.tudelft.nl/~wouter/publications/soundtests/ */
1683 static int mov_read_wide(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1684 {
1685 int err;
1686
1687 if (atom.size < 8)
1688 return 0; /* continue */
1689 if (get_be32(pb) != 0) { /* 0 sized mdat atom... use the 'wide' atom size */
1690 url_fskip(pb, atom.size - 4);
1691 return 0;
1692 }
1693 atom.type = get_le32(pb);
1694 atom.offset += 8;
1695 atom.size -= 8;
1696 if (atom.type != MKTAG('m','d','a','t')) {
1697 url_fskip(pb, atom.size);
1698 return 0;
1699 }
1700 err = mov_read_mdat(c, pb, atom);
1701 return err;
1702 }
1703
1704 static int mov_read_cmov(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1705 {
1706 #if CONFIG_ZLIB
1707 ByteIOContext ctx;
1708 uint8_t *cmov_data;
1709 uint8_t *moov_data; /* uncompressed data */
1710 long cmov_len, moov_len;
1711 int ret = -1;
1712
1713 get_be32(pb); /* dcom atom */
1714 if (get_le32(pb) != MKTAG('d','c','o','m'))
1715 return -1;
1716 if (get_le32(pb) != MKTAG('z','l','i','b')) {
1717 av_log(NULL, AV_LOG_ERROR, "unknown compression for cmov atom !");
1718 return -1;
1719 }
1720 get_be32(pb); /* cmvd atom */
1721 if (get_le32(pb) != MKTAG('c','m','v','d'))
1722 return -1;
1723 moov_len = get_be32(pb); /* uncompressed size */
1724 cmov_len = atom.size - 6 * 4;
1725
1726 cmov_data = av_malloc(cmov_len);
1727 if (!cmov_data)
1728 return -1;
1729 moov_data = av_malloc(moov_len);
1730 if (!moov_data) {
1731 av_free(cmov_data);
1732 return -1;
1733 }
1734 get_buffer(pb, cmov_data, cmov_len);
1735 if(uncompress (moov_data, (uLongf *) &moov_len, (const Bytef *)cmov_data, cmov_len) != Z_OK)
1736 goto free_and_return;
1737 if(init_put_byte(&ctx, moov_data, moov_len, 0, NULL, NULL, NULL, NULL) != 0)
1738 goto free_and_return;
1739 atom.type = MKTAG('m','o','o','v');
1740 atom.offset = 0;
1741 atom.size = moov_len;
1742 #ifdef DEBUG
1743 // { int fd = open("/tmp/uncompheader.mov", O_WRONLY | O_CREAT); write(fd, moov_data, moov_len); close(fd); }
1744 #endif
1745 ret = mov_read_default(c, &ctx, atom);
1746 free_and_return:
1747 av_free(moov_data);
1748 av_free(cmov_data);
1749 return ret;
1750 #else
1751 av_log(c->fc, AV_LOG_ERROR, "this file requires zlib support compiled in\n");
1752 return -1;
1753 #endif
1754 }
1755
1756 /* edit list atom */
1757 static int mov_read_elst(MOVContext *c, ByteIOContext *pb, MOVAtom atom)
1758 {
1759 MOVStreamContext *sc = c->fc->streams[c->fc->nb_streams-1]->priv_data;
1760 int i, edit_count;
1761
1762 get_byte(pb); /* version */
1763 get_be24(pb); /* flags */
1764 edit_count = get_be32(pb); /* entries */
1765
1766 for(i=0; i<edit_count; i++){
1767 int time;
1768 get_be32(pb); /* Track duration */
1769 time = get_be32(pb); /* Media time */
1770 get_be32(pb); /* Media rate */
1771 if (i == 0 && time != -1)
1772 sc->time_offset = time;
1773 }
1774
1775 if(edit_count > 1)
1776 av_log(c->fc, AV_LOG_WARNING, "multiple edit list entries, "
1777 "a/v desync might occur, patch welcome\n");
1778
1779 dprintf(c->fc, "track[%i].edit_count = %i\n", c->fc->nb_streams-1, sc->edit_count);
1780 return 0;
1781 }
1782
1783 static const MOVParseTableEntry mov_default_parse_table[] = {
1784 { MKTAG('a','v','s','s'), mov_read_extradata },
1785 { MKTAG('c','o','6','4'), mov_read_stco },
1786 { MKTAG('c','t','t','s'), mov_read_ctts }, /* composition time to sample */
1787 { MKTAG('d','i','n','f'), mov_read_default },
1788 { MKTAG('d','r','e','f'), mov_read_dref },
1789 { MKTAG('e','d','t','s'), mov_read_default },
1790 { MKTAG('e','l','s','t'), mov_read_elst },
1791 { MKTAG('e','n','d','a'), mov_read_enda },
1792 { MKTAG('f','i','e','l'), mov_read_extradata },
1793 { MKTAG('f','t','y','p'), mov_read_ftyp },
1794 { MKTAG('g','l','b','l'), mov_read_glbl },
1795 { MKTAG('h','d','l','r'), mov_read_hdlr },
1796 { MKTAG('i','l','s','t'), mov_read_ilst },
1797 { MKTAG('j','p','2','h'), mov_read_extradata },
1798 { MKTAG('m','d','a','t'), mov_read_mdat },
1799 { MKTAG('m','d','h','d'), mov_read_mdhd },
1800 { MKTAG('m','d','i','a'), mov_read_default },
1801 { MKTAG('m','e','t','a'), mov_read_meta },
1802 { MKTAG('m','i','n','f'), mov_read_default },
1803 { MKTAG('m','o','o','f'), mov_read_moof },
1804 { MKTAG('m','o','o','v'), mov_read_moov },
1805 { MKTAG('m','v','e','x'), mov_read_default },
1806 { MKTAG('m','v','h','d'), mov_read_mvhd },
1807 { MKTAG('S','M','I',' '), mov_read_smi }, /* Sorenson extension ??? */
1808 { MKTAG('a','l','a','c'), mov_read_extradata }, /* alac specific atom */
1809 { MKTAG('a','v','c','C'), mov_read_glbl },
1810 { MKTAG('p','a','s','p'), mov_read_pasp },
1811 { MKTAG('s','t','b','l'), mov_read_default },
1812 { MKTAG('s','t','c','o'), mov_read_stco },
1813 { MKTAG('s','t','s','c'), mov_read_stsc },
1814 { MKTAG('s','t','s','d'), mov_read_stsd }, /* sample description */
1815 { MKTAG('s','t','s','s'), mov_read_stss }, /* sync sample */
1816 { MKTAG('s','t','s','z'), mov_read_stsz }, /* sample size */
1817 { MKTAG('s','t','t','s'), mov_read_stts },
1818 { MKTAG('t','k','h','d'), mov_read_tkhd }, /* track header */
1819 { MKTAG('t','f','h','d'), mov_read_tfhd }, /* track fragment header */
1820 { MKTAG('t','r','a','k'), mov_read_trak },
1821 { MKTAG('t','r','a','f'), mov_read_default },
1822 { MKTAG('t','r','e','x'), mov_read_trex },
1823 { MKTAG('t','r','k','n'), mov_read_trkn },
1824 { MKTAG('t','r','u','n'), mov_read_trun },
1825 { MKTAG('u','d','t','a'), mov_read_default },
1826 { MKTAG('w','a','v','e'), mov_read_wave },
1827 { MKTAG('e','s','d','s'), mov_read_esds },
1828 { MKTAG('w','i','d','e'), mov_read_wide }, /* place holder */
1829 { MKTAG('c','m','o','v'), mov_read_cmov },
1830 { MKTAG(0xa9,'n','a','m'), mov_read_udta_string },
1831 { MKTAG(0xa9,'w','r','t'), mov_read_udta_string },
1832 { MKTAG(0xa9,'c','p','y'), mov_read_udta_string },
1833 { MKTAG(0xa9,'i','n','f'), mov_read_udta_string },
1834 { MKTAG(0xa9,'i','n','f'), mov_read_udta_string },
1835 { MKTAG(0xa9,'A','R','T'), mov_read_udta_string },
1836 { MKTAG(0xa9,'a','l','b'), mov_read_udta_string },
1837 { MKTAG(0xa9,'c','m','t'), mov_read_udta_string },
1838 { 0, NULL }
1839 };
1840
1841 static int mov_probe(AVProbeData *p)
1842 {
1843 unsigned int offset;
1844 uint32_t tag;
1845 int score = 0;
1846
1847 /* check file header */
1848 offset = 0;
1849 for(;;) {
1850 /* ignore invalid offset */
1851 if ((offset + 8) > (unsigned int)p->buf_size)
1852 return score;
1853 tag = AV_RL32(p->buf + offset + 4);
1854 switch(tag) {
1855 /* check for obvious tags */
1856 case MKTAG('j','P',' ',' '): /* jpeg 2000 signature */
1857 case MKTAG('m','o','o','v'):
1858 case MKTAG('m','d','a','t'):
1859 case MKTAG('p','n','o','t'): /* detect movs with preview pics like ew.mov and april.mov */
1860 case MKTAG('u','d','t','a'): /* Packet Video PVAuthor adds this and a lot of more junk */
1861 case MKTAG('f','t','y','p'):
1862 return AVPROBE_SCORE_MAX;
1863 /* those are more common words, so rate then a bit less */
1864 case MKTAG('e','d','i','w'): /* xdcam files have reverted first tags */
1865 case MKTAG('w','i','d','e'):
1866 case MKTAG('f','r','e','e'):
1867 case MKTAG('j','u','n','k'):
1868 case MKTAG('p','i','c','t'):
1869 return AVPROBE_SCORE_MAX - 5;
1870 case MKTAG(0x82,0x82,0x7f,0x7d):
1871 case MKTAG('s','k','i','p'):
1872 case MKTAG('u','u','i','d'):
1873 case MKTAG('p','r','f','l'):
1874 offset = AV_RB32(p->buf+offset) + offset;
1875 /* if we only find those cause probedata is too small at least rate them */
1876 score = AVPROBE_SCORE_MAX - 50;
1877 break;
1878 default:
1879 /* unrecognized tag */
1880 return score;
1881 }
1882 }
1883 return score;
1884 }
1885
1886 static int mov_read_header(AVFormatContext *s, AVFormatParameters *ap)
1887 {
1888 MOVContext *mov = s->priv_data;
1889 ByteIOContext *pb = s->pb;
1890 int err;
1891 MOVAtom atom = { 0, 0, 0 };
1892
1893 mov->fc = s;
1894 /* .mov and .mp4 aren't streamable anyway (only progressive download if moov is before mdat) */
1895 if(!url_is_streamed(pb))
1896 atom.size = url_fsize(pb);
1897 else
1898 atom.size = INT64_MAX;
1899
1900 /* check MOV header */
1901 if ((err = mov_read_default(mov, pb, atom)) < 0) {
1902 av_log(s, AV_LOG_ERROR, "error reading header: %d\n", err);
1903 return err;
1904 }
1905 if (!mov->found_moov) {
1906 av_log(s, AV_LOG_ERROR, "moov atom not found\n");
1907 return -1;
1908 }
1909 dprintf(mov->fc, "on_parse_exit_offset=%lld\n", url_ftell(pb));
1910
1911 return 0;
1912 }
1913
1914 static int mov_read_packet(AVFormatContext *s, AVPacket *pkt)
1915 {
1916 MOVContext *mov = s->priv_data;
1917 MOVStreamContext *sc = 0;
1918 AVIndexEntry *sample = 0;
1919 int64_t best_dts = INT64_MAX;
1920 int i;
1921 retry:
1922 for (i = 0; i < s->nb_streams; i++) {
1923 AVStream *st = s->streams[i];
1924 MOVStreamContext *msc = st->priv_data;
1925 if (st->discard != AVDISCARD_ALL && msc->pb && msc->current_sample < msc->sample_count) {
1926 AVIndexEntry *current_sample = &st->index_entries[msc->current_sample];
1927 int64_t dts = av_rescale(current_sample->timestamp * (int64_t)msc->time_rate,
1928 AV_TIME_BASE, msc->time_scale);
1929 dprintf(s, "stream %d, sample %d, dts %"PRId64"\n", i, msc->current_sample, dts);
1930 if (!sample || (url_is_streamed(s->pb) && current_sample->pos < sample->pos) ||
1931 (!url_is_streamed(s->pb) &&
1932 ((msc->pb != s->pb && dts < best_dts) || (msc->pb == s->pb &&
1933 ((FFABS(best_dts - dts) <= AV_TIME_BASE && current_sample->pos < sample->pos) ||
1934 (FFABS(best_dts - dts) > AV_TIME_BASE && dts < best_dts)))))) {
1935 sample = current_sample;
1936 best_dts = dts;
1937 sc = msc;
1938 }
1939 }
1940 }
1941 if (!sample) {
1942 mov->found_mdat = 0;
1943 if (!url_is_streamed(s->pb) ||
1944 mov_read_default(mov, s->pb, (MOVAtom){ 0, 0, INT64_MAX }) < 0 ||
1945 url_feof(s->pb))
1946 return -1;
1947 dprintf(s, "read fragments, offset 0x%llx\n", url_ftell(s->pb));
1948 goto retry;
1949 }
1950 /* must be done just before reading, to avoid infinite loop on sample */
1951 sc->current_sample++;
1952 if (url_fseek(sc->pb, sample->pos, SEEK_SET) != sample->pos) {
1953 av_log(mov->fc, AV_LOG_ERROR, "stream %d, offset 0x%"PRIx64": partial file\n",
1954 sc->ffindex, sample->pos);
1955 return -1;
1956 }
1957 av_get_packet(sc->pb, pkt, sample->size);
1958 #if CONFIG_DV_DEMUXER
1959 if (mov->dv_demux && sc->dv_audio_container) {
1960 dv_produce_packet(mov->dv_demux, pkt, pkt->data, pkt->size);
1961 av_free(pkt->data);
1962 pkt->size = 0;
1963 if (dv_get_packet(mov->dv_demux, pkt) < 0)
1964 return -1;
1965 }
1966 #endif
1967 pkt->stream_index = sc->ffindex;
1968 pkt->dts = sample->timestamp;
1969 if (sc->ctts_data) {
1970 assert(sc->ctts_data[sc->ctts_index].duration % sc->time_rate == 0);
1971 pkt->pts = pkt->dts + sc->ctts_data[sc->ctts_index].duration / sc->time_rate;
1972 /* update ctts context */
1973 sc->ctts_sample++;
1974 if (sc->ctts_index < sc->ctts_count &&
1975 sc->ctts_data[sc->ctts_index].count == sc->ctts_sample) {
1976 sc->ctts_index++;
1977 sc->ctts_sample = 0;
1978 }
1979 if (sc->wrong_dts)
1980 pkt->dts = AV_NOPTS_VALUE;
1981 } else {
1982 AVStream *st = s->streams[sc->ffindex];
1983 int64_t next_dts = (sc->current_sample < sc->sample_count) ?
1984 st->index_entries[sc->current_sample].timestamp : st->duration;
1985 pkt->duration = next_dts - pkt->dts;
1986 pkt->pts = pkt->dts;
1987 }
1988 pkt->flags |= sample->flags & AVINDEX_KEYFRAME ? PKT_FLAG_KEY : 0;
1989 pkt->pos = sample->pos;
1990 dprintf(s, "stream %d, pts %"PRId64", dts %"PRId64", pos 0x%"PRIx64", duration %d\n",
1991 pkt->stream_index, pkt->pts, pkt->dts, pkt->pos, pkt->duration);
1992 return 0;
1993 }
1994
1995 static int mov_seek_stream(AVStream *st, int64_t timestamp, int flags)
1996 {
1997 MOVStreamContext *sc = st->priv_data;
1998 int sample, time_sample;
1999 int i;
2000
2001 sample = av_index_search_timestamp(st, timestamp, flags);
2002 dprintf(st->codec, "stream %d, timestamp %"PRId64", sample %d\n", st->index, timestamp, sample);
2003 if (sample < 0) /* not sure what to do */
2004 return -1;
2005 sc->current_sample = sample;
2006 dprintf(st->codec, "stream %d, found sample %d\n", st->index, sc->current_sample);
2007 /* adjust ctts index */
2008 if (sc->ctts_data) {
2009 time_sample = 0;
2010 for (i = 0; i < sc->ctts_count; i++) {
2011 int next = time_sample + sc->ctts_data[i].count;
2012 if (next > sc->current_sample) {
2013 sc->ctts_index = i;
2014 sc->ctts_sample = sc->current_sample - time_sample;
2015 break;
2016 }
2017 time_sample = next;
2018 }
2019 }
2020 return sample;
2021 }
2022
2023 static int mov_read_seek(AVFormatContext *s, int stream_index, int64_t sample_time, int flags)
2024 {
2025 AVStream *st;
2026 int64_t seek_timestamp, timestamp;
2027 int sample;
2028 int i;
2029
2030 if (stream_index >= s->nb_streams)
2031 return -1;
2032 if (sample_time < 0)
2033 sample_time = 0;
2034
2035 st = s->streams[stream_index];
2036 sample = mov_seek_stream(st, sample_time, flags);
2037 if (sample < 0)
2038 return -1;
2039
2040 /* adjust seek timestamp to found sample timestamp */
2041 seek_timestamp = st->index_entries[sample].timestamp;
2042
2043 for (i = 0; i < s->nb_streams; i++) {
2044 st = s->streams[i];
2045 if (stream_index == i || st->discard == AVDISCARD_ALL)
2046 continue;
2047
2048 timestamp = av_rescale_q(seek_timestamp, s->streams[stream_index]->time_base, st->time_base);
2049 mov_seek_stream(st, timestamp, flags);
2050 }
2051 return 0;
2052 }
2053
2054 static int mov_read_close(AVFormatContext *s)
2055 {
2056 int i, j;
2057 MOVContext *mov = s->priv_data;
2058 for(i=0; i<s->nb_streams; i++) {
2059 MOVStreamContext *sc = s->streams[i]->priv_data;
2060 av_freep(&sc->ctts_data);
2061 for (j=0; j<sc->drefs_count; j++)
2062 av_freep(&sc->drefs[j].path);
2063 av_freep(&sc->drefs);
2064 if (sc->pb && sc->pb != s->pb)
2065 url_fclose(sc->pb);
2066 }
2067 if(mov->dv_demux){
2068 for(i=0; i<mov->dv_fctx->nb_streams; i++){
2069 av_freep(&mov->dv_fctx->streams[i]->codec);
2070 av_freep(&mov->dv_fctx->streams[i]);
2071 }
2072 av_freep(&mov->dv_fctx);
2073 av_freep(&mov->dv_demux);
2074 }
2075 av_freep(&mov->trex_data);
2076 return 0;
2077 }
2078
2079 AVInputFormat mov_demuxer = {
2080 "mov,mp4,m4a,3gp,3g2,mj2",
2081 NULL_IF_CONFIG_SMALL("QuickTime/MPEG-4/Motion JPEG 2000 format"),
2082 sizeof(MOVContext),
2083 mov_probe,
2084 mov_read_header,
2085 mov_read_packet,
2086 mov_read_close,
2087 mov_read_seek,
2088 };