2 * Copyright (C) 2003-2005 M. Bakker, Nero AG, http://www.nero.com
3 * FAAD2 - Freeware Advanced Audio (AAC) Decoder including SBR decoding
12 #include "portable_io.h"
17 * The three states of the mp4 parser. The parser only loads the audio specific
18 * values and tables when it is in the second state.
20 enum audio_track_state {
21 /** We haven't encountered an mp4a atom so far. */
23 /** We have seen an mp4a atom but no subsequent trak atom yet. */
25 /** A trak atom was seen *after* the mp4a atom. */
30 /* determines which atoms we still need to parse. */
31 enum audio_track_state state;
34 uint16_t channel_count;
38 uint32_t stsz_sample_size;
39 uint32_t stsz_sample_count;
43 uint32_t stts_entry_count;
44 uint32_t *stts_sample_count;
47 uint32_t stsc_entry_count;
48 uint32_t *stsc_first_chunk;
49 uint32_t *stsc_samples_per_chunk;
52 uint32_t stco_entry_count;
53 uint32_t *stco_chunk_offset;
61 const struct mp4_callback *cb;
73 struct mp4_track track;
74 struct mp4_metadata meta;
78 * Returns -E_MP4_READ, 0, or 1 on errors/EOF/success. Partial reads followed
79 * by EOF or read errors are treated as errors.
81 static int read_data(struct mp4 *f, void *data, size_t size)
84 ssize_t ret = f->cb->read(f->cb->user_data, data, size);
85 if (ret < 0 && errno == EINTR)
87 /* regard EAGAIN as an error as reads should be blocking. */
89 return ret < 0? -E_MP4_READ : 0;
95 static int read_int64(struct mp4 *f, uint64_t *result)
98 int ret = read_data(f, data, 8);
101 *result = read_u64_be(data);
105 static int read_int32(struct mp4 *f, uint32_t *result)
108 int ret = read_data(f, data, 4);
111 *result = read_u32_be(data);
115 static int read_int16(struct mp4 *f, uint16_t *result)
118 int ret = read_data(f, data, 2);
121 *result = read_u16_be(data);
126 ATOM_ITEM(MOOV, 'm', 'o', 'o', 'v') \
127 ATOM_ITEM(TRAK, 't', 'r', 'a', 'k') \
128 ATOM_ITEM(MDIA, 'm', 'd', 'i', 'a') \
129 ATOM_ITEM(MINF, 'm', 'i', 'n', 'f') \
130 ATOM_ITEM(STBL, 's', 't', 'b', 'l') \
131 ATOM_ITEM(UDTA, 'u', 'd', 't', 'a') \
132 ATOM_ITEM(ILST, 'i', 'l', 's', 't') /* iTunes Metadata list */ \
133 ATOM_ITEM(ARTIST, 0xa9, 'A', 'R', 'T') \
134 ATOM_ITEM(TITLE, 0xa9, 'n', 'a', 'm') \
135 ATOM_ITEM(ALBUM, 0xa9, 'a', 'l', 'b') \
136 ATOM_ITEM(DATE, 0xa9, 'd', 'a', 'y') \
137 ATOM_ITEM(COMMENT, 0xa9, 'c', 'm', 't') \
138 ATOM_ITEM(MDHD, 'm', 'd', 'h', 'd') /* track header */ \
139 ATOM_ITEM(STSD, 's', 't', 's', 'd') /* sample description box */ \
140 ATOM_ITEM(STTS, 's', 't', 't', 's') /* time to sample box */ \
141 ATOM_ITEM(STSZ, 's', 't', 's', 'z') /* sample size box */ \
142 ATOM_ITEM(STCO, 's', 't', 'c', 'o') /* chunk offset box */ \
143 ATOM_ITEM(STSC, 's', 't', 's', 'c') /* sample to chunk box */ \
144 ATOM_ITEM(MP4A, 'm', 'p', '4', 'a') \
145 ATOM_ITEM(META, 'm', 'e', 't', 'a') /* iTunes Metadata box */ \
146 ATOM_ITEM(DATA, 'd', 'a', 't', 'a') /* iTunes Metadata data box */ \
148 #define ATOM_ITEM(_name, a, b, c, d) ATOM_ ## _name,
149 enum atom {ATOM_ITEMS};
152 static uint8_t atom_name_to_type(uint8_t *p)
154 #define ATOM_VALUE(a, b, c, d) ((a << 24) + (b << 16) + (c << 8) + d)
155 #define ATOM_ITEM(_name, a, b, c, d) \
156 {.name = # _name, .val = ATOM_VALUE(a, b, c, d)},
157 static const struct {
160 } atom_table[] = {ATOM_ITEMS};
162 uint32_t val = read_u32_be(p);
164 for (uint8_t n = 0; n < ARRAY_SIZE(atom_table); n++)
165 if (val == atom_table[n].val)
170 /* read atom header, atom size is returned with header included. */
171 static int atom_read_header(struct mp4 *f, uint8_t *atom_type,
172 uint8_t *header_size, uint64_t *atom_size)
176 uint8_t atom_header[8];
178 ret = read_data(f, atom_header, 8);
181 size = read_u32_be(atom_header);
182 if (size == 1) { /* 64 bit atom size */
185 ret = read_int64(f, atom_size);
193 *atom_type = atom_name_to_type(atom_header + 4);
197 static off_t get_position(const struct mp4 *f)
199 return f->cb->seek(f->cb->user_data, 0, SEEK_CUR);
202 static void set_position(struct mp4 *f, off_t position)
204 f->cb->seek(f->cb->user_data, position, SEEK_SET);
207 static void skip_bytes(struct mp4 *f, off_t num_skip)
209 f->cb->seek(f->cb->user_data, num_skip, SEEK_CUR);
212 static int read_stsz(struct mp4 *f)
215 struct mp4_track *t = &f->track;
217 if (t->state != ATS_SEEN_MP4A || t->stsz_table)
219 skip_bytes(f, 4); /* version (1), flags (3) */
220 ret = read_int32(f, &t->stsz_sample_size);
223 ret = read_int32(f, &t->stsz_sample_count);
226 if (t->stsz_sample_size != 0)
228 t->stsz_table = para_malloc(t->stsz_sample_count * sizeof(int32_t));
229 for (uint32_t n = 0; n < t->stsz_sample_count; n++) {
230 ret = read_int32(f, &t->stsz_table[n]);
237 static int read_stts(struct mp4 *f)
240 struct mp4_track *t = &f->track;
242 if (t->state != ATS_SEEN_MP4A || t->stts_sample_count)
244 skip_bytes(f, 4); /* version (1), flags (3) */
245 ret = read_int32(f, &t->stts_entry_count);
248 t->stts_sample_count = para_malloc(t->stts_entry_count
250 for (uint32_t n = 0; n < t->stts_entry_count; n++) {
251 ret = read_int32(f, &t->stts_sample_count[n]);
254 skip_bytes(f, 4); /* sample delta */
259 static int read_stsc(struct mp4 *f)
262 struct mp4_track *t = &f->track;
264 if (t->state != ATS_SEEN_MP4A)
266 if (t->stsc_first_chunk || t->stsc_samples_per_chunk)
268 skip_bytes(f, 4); /* version (1), flags (3) */
269 ret = read_int32(f, &t->stsc_entry_count);
272 t->stsc_first_chunk = para_malloc(t->stsc_entry_count * sizeof(int32_t));
273 t->stsc_samples_per_chunk = para_malloc(t->stsc_entry_count
275 for (uint32_t n = 0; n < t->stsc_entry_count; n++) {
276 ret = read_int32(f, &t->stsc_first_chunk[n]);
279 ret = read_int32(f, &t->stsc_samples_per_chunk[n]);
282 skip_bytes(f, 4); /* sample desc index */
287 static int read_stco(struct mp4 *f)
290 struct mp4_track *t = &f->track;
292 if (t->state != ATS_SEEN_MP4A || t->stco_chunk_offset)
294 skip_bytes(f, 4); /* version (1), flags (3) */
295 ret = read_int32(f, &t->stco_entry_count);
298 t->stco_chunk_offset = para_malloc(t->stco_entry_count
300 for (uint32_t n = 0; n < t->stco_entry_count; n++) {
301 ret = read_int32(f, &t->stco_chunk_offset[n]);
308 static int read_stsd(struct mp4 *f)
311 uint32_t entry_count;
313 if (f->track.state != ATS_INITIAL)
315 skip_bytes(f, 4); /* version (1), flags (3) */
316 ret = read_int32(f, &entry_count);
319 for (uint32_t n = 0; n < entry_count; n++) {
320 uint64_t skip = get_position(f);
322 uint8_t atom_type = 0;
323 ret = atom_read_header(f, &atom_type, NULL, &size);
327 if (atom_type == ATOM_MP4A) {
328 f->track.state = ATS_SEEN_MP4A;
329 /* reserved (6), data reference index (2), reserved (8) */
331 ret = read_int16(f, &f->track.channel_count);
335 ret = read_int16(f, &f->track.sample_rate);
339 set_position(f, skip);
344 static const char *get_metadata_name(uint8_t atom_type)
347 case ATOM_TITLE: return "title";
348 case ATOM_ARTIST: return "artist";
349 case ATOM_ALBUM: return "album";
350 case ATOM_DATE: return "date";
351 case ATOM_COMMENT: return "comment";
352 default: return "unknown";
356 static int parse_tag(struct mp4 *f, uint8_t parent, int32_t size)
359 uint64_t subsize, sumsize;
368 set_position(f, destpos), sumsize += subsize
371 uint8_t header_size = 0;
372 ret = atom_read_header(f, &atom_type, &header_size, &subsize);
375 destpos = get_position(f) + subsize - header_size;
376 if (atom_type != ATOM_DATA)
378 skip_bytes(f, 8); /* version (1), flags (3), reserved (4) */
379 ret = -E_MP4_CORRUPT;
380 if (subsize < header_size + 8 || subsize > UINT_MAX)
382 len = subsize - (header_size + 8);
384 value = para_malloc(len + 1);
385 ret = read_data(f, value, len);
391 return -E_MP4_CORRUPT;
392 f->meta.tags = para_realloc(f->meta.tags, (f->meta.count + 1)
393 * sizeof(struct mp4_tag));
394 tag = f->meta.tags + f->meta.count;
395 tag->item = para_strdup(get_metadata_name(parent));
404 static int read_mdhd(struct mp4 *f)
408 struct mp4_track *t = &f->track;
410 if (t->state != ATS_INITIAL)
412 ret = read_int32(f, &version);
416 skip_bytes(f, 16); /* creation time (8), modification time (8) */
417 ret = read_int32(f, &t->time_scale);
420 ret = read_int64(f, &t->duration);
423 } else { /* version == 0 */
426 skip_bytes(f, 8); /* creation time (4), modification time (4) */
427 ret = read_int32(f, &t->time_scale);
430 ret = read_int32(f, &temp);
433 t->duration = (temp == (uint32_t) (-1))?
434 (uint64_t) (-1) : (uint64_t) (temp);
440 static int read_ilst(struct mp4 *f, int32_t size)
443 uint64_t sumsize = 0;
445 while (sumsize < size) {
447 uint64_t subsize, destpos;
448 uint8_t header_size = 0;
449 ret = atom_read_header(f, &atom_type, &header_size, &subsize);
452 destpos = get_position(f) + subsize - header_size;
459 ret = parse_tag(f, atom_type, subsize - header_size);
463 set_position(f, destpos);
469 static int read_meta(struct mp4 *f, uint64_t size)
472 uint64_t subsize, sumsize = 0;
474 uint8_t header_size = 0;
476 skip_bytes(f, 4); /* version (1), flags (3) */
477 while (sumsize < (size - (header_size + 4))) {
478 ret = atom_read_header(f, &atom_type, &header_size, &subsize);
481 if (subsize <= header_size + 4)
483 if (atom_type == ATOM_ILST) {
484 f->ilst_offset = get_position(f) - header_size;
485 f->ilst_size = subsize;
486 ret = read_ilst(f, subsize - (header_size + 4));
490 set_position(f, get_position(f) + subsize - header_size);
496 static bool need_atom(uint8_t atom_type, bool meta_only)
498 /* these are needed in any case */
509 /* meta-only opens don't need anything else */
512 /* these are only required for regular opens */
524 /* parse atoms that are sub atoms of other atoms */
525 static int parse_sub_atoms(struct mp4 *f, uint64_t total_size, bool meta_only)
528 uint64_t dest, size, end = get_position(f) + total_size;
530 for (dest = get_position(f); dest < end; set_position(f, dest)) {
531 uint8_t header_size, atom_type;
532 ret = atom_read_header(f, &atom_type, &header_size, &size);
536 return -E_MP4_CORRUPT;
537 dest = get_position(f) + size - header_size;
538 if (atom_type == ATOM_TRAK && f->track.state == ATS_SEEN_MP4A) {
539 f->track.state = ATS_TRACK_CHANGE;
542 if (atom_type == ATOM_UDTA) {
543 f->udta_offset = get_position(f) - header_size;
546 if (!need_atom(atom_type, meta_only))
549 case ATOM_STSZ: ret = read_stsz(f); break;
550 case ATOM_STTS: ret = read_stts(f); break;
551 case ATOM_STSC: ret = read_stsc(f); break;
552 case ATOM_STCO: ret = read_stco(f); break;
553 case ATOM_STSD: ret = read_stsd(f); break;
554 case ATOM_MDHD: ret = read_mdhd(f); break;
556 f->meta_offset = get_position(f) - header_size;
558 ret = read_meta(f, size);
561 ret = parse_sub_atoms(f, size - header_size, meta_only);
569 void mp4_close(struct mp4 *f)
571 free(f->track.stsz_table);
572 free(f->track.stts_sample_count);
573 free(f->track.stsc_first_chunk);
574 free(f->track.stsc_samples_per_chunk);
575 free(f->track.stco_chunk_offset);
576 for (uint32_t n = 0; n < f->meta.count; n++) {
577 free(f->meta.tags[n].item);
578 free(f->meta.tags[n].value);
584 static int open_file(const struct mp4_callback *cb, bool meta_only, struct mp4 **result)
588 uint8_t atom_type, header_size;
589 struct mp4 *f = para_calloc(sizeof(*f));
592 while ((ret = atom_read_header(f, &atom_type, &header_size, &size)) > 0) {
593 f->last_atom = atom_type;
594 if (atom_type != ATOM_MOOV || size <= header_size) { /* skip */
595 set_position(f, get_position(f) + size - header_size);
598 f->moov_offset = get_position(f) - header_size;
600 ret = parse_sub_atoms(f, size - header_size, meta_only);
607 if (f->track.channel_count == 0)
609 ret = -E_MP4_BAD_SAMPLERATE;
610 if (f->track.sample_rate == 0)
612 ret = -E_MP4_MISSING_ATOM;
613 if (f->udta_size == 0 || f->meta_size == 0 || f->ilst_size == 0)
623 int mp4_open(const struct mp4_callback *cb, struct mp4 **result)
629 ret = open_file(cb, false, &f);
632 ret = -E_MP4_BAD_SAMPLE_COUNT;
633 if (f->track.stsz_sample_count == 0)
635 ret = -E_MP4_CORRUPT;
636 if (f->track.time_scale == 0)
645 static int32_t chunk_of_sample(const struct mp4 *f, int32_t sample,
648 const struct mp4_track *t = &f->track;
649 uint32_t *fc = t->stsc_first_chunk, *spc = t->stsc_samples_per_chunk;
650 uint32_t chunk1, chunk1samples, n, total, k;
652 for (k = 1, total = 0; k < t->stsc_entry_count; k++, total += n) {
653 n = (fc[k] - fc[k - 1]) * spc[k - 1]; /* number of samples */
654 if (sample < total + n)
658 chunk1samples = spc[k - 1];
659 if (chunk1samples != 0)
660 *chunk = (sample - total) / chunk1samples + chunk1;
663 return total + (*chunk - chunk1) * chunk1samples;
667 * Return the number of milliseconds of the audio track.
669 * \param f As returned by \ref mp4_open_read(), must not be NULL.
671 uint64_t mp4_get_duration(const struct mp4 *f)
673 const struct mp4_track *t = &f->track;
675 return t->duration * 1000 / t->time_scale;
678 int mp4_set_sample_position(struct mp4 *f, uint32_t sample)
680 const struct mp4_track *t = &f->track;
681 int32_t offset, chunk, chunk_sample;
682 uint32_t n, srs; /* sample range size */
684 if (sample >= t->stsz_sample_count)
685 return -ERRNO_TO_PARA_ERROR(EINVAL);
686 chunk_sample = chunk_of_sample(f, sample, &chunk);
687 if (t->stsz_sample_size > 0)
688 srs = (sample - chunk_sample) * t->stsz_sample_size;
690 for (srs = 0, n = chunk_sample; n < sample; n++)
691 srs += t->stsz_table[n];
693 if (t->stco_entry_count > 0 && chunk > t->stco_entry_count)
694 offset = t->stco_chunk_offset[t->stco_entry_count - 1];
695 else if (t->stco_entry_count > 0)
696 offset = t->stco_chunk_offset[chunk - 1];
699 set_position(f, offset + srs);
703 int mp4_get_sample_size(const struct mp4 *f, uint32_t sample, uint32_t *result)
705 const struct mp4_track *t = &f->track;
707 if (sample >= t->stsz_sample_count)
708 return -ERRNO_TO_PARA_ERROR(EINVAL);
709 if (t->stsz_sample_size != 0)
710 *result = t->stsz_sample_size;
712 *result = t->stsz_table[sample];
716 uint16_t mp4_get_sample_rate(const struct mp4 *f)
718 return f->track.sample_rate;
721 uint16_t mp4_get_channel_count(const struct mp4 *f)
723 return f->track.channel_count;
726 uint32_t mp4_num_samples(const struct mp4 *f)
728 return f->track.stsz_sample_count;
731 int mp4_open_meta(const struct mp4_callback *cb, struct mp4 **result)
734 int ret = open_file(cb, true, &f);
743 * Return the metadata of an mp4 file.
745 * \param f As returned by either \ref mp4_open_read() or \ref mp4_open_meta().
747 * The caller is allowed to add, delete or modify the entries of the returned
748 * structure in order to pass the modified version to \ref mp4_meta_update().
750 struct mp4_metadata *mp4_get_meta(struct mp4 *f)
755 /** Total length of an on-disk metadata tag. */
756 #define TAG_LEN(_len) (24 + (_len))
757 static void create_ilst(const struct mp4_metadata *meta, uint8_t *out)
759 for (unsigned n = 0; n < meta->count; n++) {
760 struct mp4_tag *tag = meta->tags + n;
761 unsigned len = strlen(tag->value);
762 const char *atom_name;
764 if (!strcasecmp(tag->item, "title"))
765 atom_name = "\xA9" "nam";
766 else if (!strcasecmp(tag->item, "artist"))
767 atom_name = "\xA9" "ART";
768 else if (!strcasecmp(tag->item, "album"))
769 atom_name = "\xA9" "alb";
770 else if (!strcasecmp(tag->item, "date"))
771 atom_name = "\xA9" "day";
772 else if (!strcasecmp(tag->item, "comment"))
773 atom_name = "\xA9" "cmt";
776 write_u32_be(out, TAG_LEN(len));
777 memcpy(out + 4, atom_name, 4);
778 write_u32_be(out + 8, 8 /* data atom header */
779 + 8 /* flags + reserved */
781 memcpy(out + 12, "data", 4);
782 write_u32_be(out + 16, 1); /* flags */
783 write_u32_be(out + 20, 0); /* reserved */
784 memcpy(out + 24, tag->value, len);
789 static void *modify_moov(struct mp4 *f, uint32_t *out_size)
792 uint64_t total_base = f->moov_offset + 8;
793 uint32_t total_size = f->moov_size - 8;
794 uint32_t new_ilst_size = 0;
800 for (unsigned n = 0; n < f->meta.count; n++)
801 new_ilst_size += TAG_LEN(strlen(f->meta.tags[n].value));
802 size_delta = new_ilst_size - (f->ilst_size - 8);
803 *out_size = total_size + size_delta;
804 out_buffer = para_malloc(*out_size);
806 set_position(f, total_base);
807 ret = read_data(f, p_out, f->udta_offset - total_base);
810 p_out += f->udta_offset - total_base;
811 ret = read_int32(f, &tmp);
814 write_u32_be(p_out, tmp + size_delta);
816 ret = read_data(f, p_out, 4);
820 ret = read_data(f, p_out, f->meta_offset - f->udta_offset - 8);
823 p_out += f->meta_offset - f->udta_offset - 8;
824 ret = read_int32(f, &tmp);
827 write_u32_be(p_out, tmp + size_delta);
829 ret = read_data(f, p_out, 4);
833 ret = read_data(f, p_out, f->ilst_offset - f->meta_offset - 8);
836 p_out += f->ilst_offset - f->meta_offset - 8;
837 ret = read_int32(f, &tmp);
840 write_u32_be(p_out, tmp + size_delta);
842 ret = read_data(f, p_out, 4);
846 create_ilst(&f->meta, p_out);
847 p_out += new_ilst_size;
848 set_position(f, f->ilst_offset + f->ilst_size);
849 ret = read_data(f, p_out, total_size - (f->ilst_offset - total_base)
856 static int write_data(struct mp4 *f, void *data, size_t size)
859 ssize_t ret = f->cb->write(f->cb->user_data, data, size);
863 return -ERRNO_TO_PARA_ERROR(errno);
870 int mp4_update_meta(struct mp4 *f)
873 uint32_t new_moov_size;
874 uint8_t buf[8] = "----moov";
878 new_moov_data = modify_moov(f, &new_moov_size);
879 if (!new_moov_data ) {
883 if (f->last_atom != ATOM_MOOV) {
884 set_position(f, f->moov_offset + 4);
885 ret = write_data(f, "free", 4); /* rename old moov to free */
888 /* write new moov atom at EOF */
889 f->cb->seek(f->cb->user_data, 0, SEEK_END);
890 } else /* overwrite old moov atom */
891 set_position(f, f->moov_offset);
892 write_u32_be(buf, new_moov_size + 8);
893 ret = write_data(f, buf, sizeof(buf));
896 ret = write_data(f, new_moov_data, new_moov_size);
899 ret = f->cb->truncate(f->cb->user_data);
901 ret = -ERRNO_TO_PARA_ERROR(errno);
908 * Return the value of the given tag item.
910 * \param f Must not be NULL.
911 * \param item "artist", "title", "album", "comment", or "date".
913 * \return The function always returns NULL if the given item is not in the
914 * above list. Otherwise, if the file does not contain a tag for the given
915 * item, the function also returns NULL. Otherwise a copy of the tag value is
916 * returned and the caller should free this memory when it is no longer needed.
918 char *mp4_get_tag_value(const struct mp4 *f, const char *item)
920 for (unsigned n = 0; n < f->meta.count; n++)
921 if (!strcasecmp(f->meta.tags[n].item, item))
922 return para_strdup(f->meta.tags[n].value);