caf: use int64_t for num_packets
[libav.git] / libavformat / cafdec.c
CommitLineData
5f14b565
PR
1/*
2 * Core Audio Format demuxer
3 * Copyright (c) 2007 Justin Ruggles
4 * Copyright (c) 2009 Peter Ross
5 *
2912e87a 6 * This file is part of Libav.
5f14b565 7 *
2912e87a 8 * Libav is free software; you can redistribute it and/or
5f14b565
PR
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.
12 *
2912e87a 13 * Libav is distributed in the hope that it will be useful,
5f14b565
PR
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.
17 *
18 * You should have received a copy of the GNU Lesser General Public
2912e87a 19 * License along with Libav; if not, write to the Free Software
5f14b565
PR
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23/**
ba87f080 24 * @file
5f14b565
PR
25 * Core Audio Format demuxer
26 */
27
28#include "avformat.h"
c3f9ebf7 29#include "internal.h"
5f14b565
PR
30#include "riff.h"
31#include "isom.h"
32#include "libavutil/intreadwrite.h"
3383a53e 33#include "libavutil/intfloat.h"
d2d67e42 34#include "libavutil/dict.h"
5f14b565
PR
35#include "caf.h"
36
37typedef struct {
38 int bytes_per_packet; ///< bytes in a packet, or 0 if variable
39 int frames_per_packet; ///< frames in a packet, or 0 if variable
40 int64_t num_bytes; ///< total number of bytes in stream
41
42 int64_t packet_cnt; ///< packet counter
43 int64_t frame_cnt; ///< frame counter
44
45 int64_t data_start; ///< data start position, in bytes
46 int64_t data_size; ///< raw data size, in bytes
47} CaffContext;
48
49static int probe(AVProbeData *p)
50{
51 if (AV_RB32(p->buf) == MKBETAG('c','a','f','f') && AV_RB16(&p->buf[4]) == 1)
52 return AVPROBE_SCORE_MAX;
53 return 0;
54}
55
56/** Read audio description chunk */
57static int read_desc_chunk(AVFormatContext *s)
58{
ae628ec1 59 AVIOContext *pb = s->pb;
5f14b565
PR
60 CaffContext *caf = s->priv_data;
61 AVStream *st;
62 int flags;
63
64 /* new audio stream */
3b3bbdd3 65 st = avformat_new_stream(s, NULL);
5f14b565 66 if (!st)
2874c81c 67 return AVERROR(ENOMEM);
5f14b565
PR
68
69 /* parse format description */
72415b2a 70 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
3383a53e 71 st->codec->sample_rate = av_int2double(avio_rb64(pb));
b7effd4e
AK
72 st->codec->codec_tag = avio_rb32(pb);
73 flags = avio_rb32(pb);
74 caf->bytes_per_packet = avio_rb32(pb);
5f14b565 75 st->codec->block_align = caf->bytes_per_packet;
b7effd4e
AK
76 caf->frames_per_packet = avio_rb32(pb);
77 st->codec->channels = avio_rb32(pb);
78 st->codec->bits_per_coded_sample = avio_rb32(pb);
5f14b565
PR
79
80 /* calculate bit rate for constant size packets */
81 if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
82 st->codec->bit_rate = (uint64_t)st->codec->sample_rate * (uint64_t)caf->bytes_per_packet * 8
83 / (uint64_t)caf->frames_per_packet;
84 } else {
85 st->codec->bit_rate = 0;
86 }
87
88 /* determine codec */
89 if (st->codec->codec_tag == MKBETAG('l','p','c','m'))
90 st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, (flags ^ 0x2) | 0x4);
91 else
92 st->codec->codec_id = ff_codec_get_id(ff_codec_caf_tags, st->codec->codec_tag);
93 return 0;
94}
95
96/** Read magic cookie chunk */
97static int read_kuki_chunk(AVFormatContext *s, int64_t size)
98{
ae628ec1 99 AVIOContext *pb = s->pb;
5f14b565
PR
100 AVStream *st = s->streams[0];
101
102 if (size < 0 || size > INT_MAX - FF_INPUT_BUFFER_PADDING_SIZE)
103 return -1;
104
105 if (st->codec->codec_id == CODEC_ID_AAC) {
106 /* The magic cookie format for AAC is an mp4 esds atom.
107 The lavc AAC decoder requires the data from the codec specific
108 description as extradata input. */
109 int strt, skip;
110 MOVAtom atom;
111
a2704c97 112 strt = avio_tell(pb);
5f14b565 113 ff_mov_read_esds(s, pb, atom);
a2704c97 114 skip = size - (avio_tell(pb) - strt);
5f14b565
PR
115 if (skip < 0 || !st->codec->extradata ||
116 st->codec->codec_id != CODEC_ID_AAC) {
117 av_log(s, AV_LOG_ERROR, "invalid AAC magic cookie\n");
118 return AVERROR_INVALIDDATA;
119 }
45a8a02a 120 avio_skip(pb, skip);
1e6f8d53
PR
121 } else if (st->codec->codec_id == CODEC_ID_ALAC) {
122#define ALAC_PREAMBLE 12
123#define ALAC_HEADER 36
b0b77b9c
JR
124#define ALAC_NEW_KUKI 24
125 uint8_t preamble[12];
126 if (size < ALAC_NEW_KUKI || size > ALAC_PREAMBLE + ALAC_HEADER) {
1e6f8d53 127 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
45a8a02a 128 avio_skip(pb, size);
1e6f8d53
PR
129 return AVERROR_INVALIDDATA;
130 }
b0b77b9c
JR
131 avio_read(pb, preamble, ALAC_PREAMBLE);
132
1e6f8d53
PR
133 st->codec->extradata = av_mallocz(ALAC_HEADER + FF_INPUT_BUFFER_PADDING_SIZE);
134 if (!st->codec->extradata)
135 return AVERROR(ENOMEM);
b0b77b9c
JR
136
137 /* For the old style cookie, we skip 12 bytes, then read 36 bytes.
138 * The new style cookie only contains the last 24 bytes of what was
139 * 36 bytes in the old style cookie, so we fabricate the first 12 bytes
140 * in that case to maintain compatibility. */
141 if (!memcmp(&preamble[4], "frmaalac", 8)) {
142 if (size < ALAC_PREAMBLE + ALAC_HEADER) {
143 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
144 av_freep(&st->codec->extradata);
145 return AVERROR_INVALIDDATA;
146 }
147 avio_read(pb, st->codec->extradata, ALAC_HEADER);
148 avio_skip(pb, size - ALAC_PREAMBLE - ALAC_HEADER);
149 } else {
150 AV_WB32(st->codec->extradata, 36);
151 memcpy(&st->codec->extradata[4], "alac", 4);
152 AV_WB32(&st->codec->extradata[8], 0);
153 memcpy(&st->codec->extradata[12], preamble, 12);
154 avio_read(pb, &st->codec->extradata[24], ALAC_NEW_KUKI - 12);
155 avio_skip(pb, size - ALAC_NEW_KUKI);
156 }
1e6f8d53 157 st->codec->extradata_size = ALAC_HEADER;
5f14b565
PR
158 } else {
159 st->codec->extradata = av_mallocz(size + FF_INPUT_BUFFER_PADDING_SIZE);
160 if (!st->codec->extradata)
161 return AVERROR(ENOMEM);
b7effd4e 162 avio_read(pb, st->codec->extradata, size);
5f14b565
PR
163 st->codec->extradata_size = size;
164 }
165
166 return 0;
167}
168
169/** Read packet table chunk */
170static int read_pakt_chunk(AVFormatContext *s, int64_t size)
171{
ae628ec1 172 AVIOContext *pb = s->pb;
5f14b565
PR
173 AVStream *st = s->streams[0];
174 CaffContext *caf = s->priv_data;
c0196a14
JR
175 int64_t pos = 0, ccount, num_packets;
176 int i;
5f14b565 177
a2704c97 178 ccount = avio_tell(pb);
5f14b565 179
b7effd4e 180 num_packets = avio_rb64(pb);
5f14b565
PR
181 if (num_packets < 0 || INT32_MAX / sizeof(AVIndexEntry) < num_packets)
182 return AVERROR_INVALIDDATA;
183
b7effd4e
AK
184 st->nb_frames = avio_rb64(pb); /* valid frames */
185 st->nb_frames += avio_rb32(pb); /* priming frames */
186 st->nb_frames += avio_rb32(pb); /* remainder frames */
5f14b565
PR
187
188 st->duration = 0;
189 for (i = 0; i < num_packets; i++) {
190 av_add_index_entry(s->streams[0], pos, st->duration, 0, 0, AVINDEX_KEYFRAME);
191 pos += caf->bytes_per_packet ? caf->bytes_per_packet : ff_mp4_read_descr_len(pb);
192 st->duration += caf->frames_per_packet ? caf->frames_per_packet : ff_mp4_read_descr_len(pb);
193 }
194
f5a9c35f 195 if (avio_tell(pb) - ccount > size) {
5f14b565 196 av_log(s, AV_LOG_ERROR, "error reading packet table\n");
f5a9c35f 197 return AVERROR_INVALIDDATA;
5f14b565 198 }
f5a9c35f 199 avio_skip(pb, ccount + size - avio_tell(pb));
5f14b565
PR
200
201 caf->num_bytes = pos;
202 return 0;
203}
204
205/** Read information chunk */
206static void read_info_chunk(AVFormatContext *s, int64_t size)
207{
ae628ec1 208 AVIOContext *pb = s->pb;
5f14b565 209 unsigned int i;
b7effd4e 210 unsigned int nb_entries = avio_rb32(pb);
5f14b565
PR
211 for (i = 0; i < nb_entries; i++) {
212 char key[32];
213 char value[1024];
e16ead07
AK
214 avio_get_str(pb, INT_MAX, key, sizeof(key));
215 avio_get_str(pb, INT_MAX, value, sizeof(value));
d2d67e42 216 av_dict_set(&s->metadata, key, value, 0);
5f14b565
PR
217 }
218}
219
6e9651d1 220static int read_header(AVFormatContext *s)
5f14b565 221{
ae628ec1 222 AVIOContext *pb = s->pb;
5f14b565
PR
223 CaffContext *caf = s->priv_data;
224 AVStream *st;
225 uint32_t tag = 0;
226 int found_data, ret;
227 int64_t size;
228
45a8a02a 229 avio_skip(pb, 8); /* magic, version, file flags */
5f14b565
PR
230
231 /* audio description chunk */
b7effd4e 232 if (avio_rb32(pb) != MKBETAG('d','e','s','c')) {
5f14b565
PR
233 av_log(s, AV_LOG_ERROR, "desc chunk not present\n");
234 return AVERROR_INVALIDDATA;
235 }
b7effd4e 236 size = avio_rb64(pb);
5f14b565
PR
237 if (size != 32)
238 return AVERROR_INVALIDDATA;
239
240 ret = read_desc_chunk(s);
241 if (ret)
242 return ret;
243 st = s->streams[0];
244
245 /* parse each chunk */
246 found_data = 0;
66e5b1df 247 while (!pb->eof_reached) {
5f14b565
PR
248
249 /* stop at data chunk if seeking is not supported or
250 data chunk size is unknown */
8978feda 251 if (found_data && (caf->data_size < 0 || !pb->seekable))
5f14b565
PR
252 break;
253
b7effd4e
AK
254 tag = avio_rb32(pb);
255 size = avio_rb64(pb);
66e5b1df 256 if (pb->eof_reached)
5f14b565
PR
257 break;
258
259 switch (tag) {
260 case MKBETAG('d','a','t','a'):
45a8a02a 261 avio_skip(pb, 4); /* edit count */
a2704c97 262 caf->data_start = avio_tell(pb);
5f14b565 263 caf->data_size = size < 0 ? -1 : size - 4;
8978feda 264 if (caf->data_size > 0 && pb->seekable)
45a8a02a 265 avio_skip(pb, caf->data_size);
5f14b565
PR
266 found_data = 1;
267 break;
268
269 /* magic cookie chunk */
270 case MKBETAG('k','u','k','i'):
271 if (read_kuki_chunk(s, size))
272 return AVERROR_INVALIDDATA;
273 break;
274
275 /* packet table chunk */
276 case MKBETAG('p','a','k','t'):
277 if (read_pakt_chunk(s, size))
278 return AVERROR_INVALIDDATA;
279 break;
280
281 case MKBETAG('i','n','f','o'):
282 read_info_chunk(s, size);
283 break;
284
285 default:
286#define _(x) ((x) >= ' ' ? (x) : ' ')
287 av_log(s, AV_LOG_WARNING, "skipping CAF chunk: %08X (%c%c%c%c)\n",
288 tag, _(tag>>24), _((tag>>16)&0xFF), _((tag>>8)&0xFF), _(tag&0xFF));
289#undef _
290 case MKBETAG('f','r','e','e'):
291 if (size < 0)
292 return AVERROR_INVALIDDATA;
45a8a02a 293 avio_skip(pb, size);
5f14b565
PR
294 break;
295 }
296 }
297
298 if (!found_data)
299 return AVERROR_INVALIDDATA;
300
301 if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
302 if (caf->data_size > 0)
303 st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet;
304 } else if (st->nb_index_entries) {
305 st->codec->bit_rate = st->codec->sample_rate * caf->data_size * 8 /
306 st->duration;
307 } else {
308 av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when "
309 "block size or frame size are variable.\n");
310 return AVERROR_INVALIDDATA;
311 }
5f14b565 312
c3f9ebf7 313 avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
5f14b565
PR
314 st->start_time = 0;
315
316 /* position the stream at the start of data */
317 if (caf->data_size >= 0)
6b4aa5da 318 avio_seek(pb, caf->data_start, SEEK_SET);
5f14b565
PR
319
320 return 0;
321}
322
323#define CAF_MAX_PKT_SIZE 4096
324
325static int read_packet(AVFormatContext *s, AVPacket *pkt)
326{
ae628ec1 327 AVIOContext *pb = s->pb;
5f14b565
PR
328 AVStream *st = s->streams[0];
329 CaffContext *caf = s->priv_data;
330 int res, pkt_size = 0, pkt_frames = 0;
331 int64_t left = CAF_MAX_PKT_SIZE;
332
66e5b1df 333 if (pb->eof_reached)
5ae092ee 334 return AVERROR(EIO);
5f14b565
PR
335
336 /* don't read past end of data chunk */
337 if (caf->data_size > 0) {
a2704c97 338 left = (caf->data_start + caf->data_size) - avio_tell(pb);
5f14b565 339 if (left <= 0)
5ae092ee 340 return AVERROR(EIO);
5f14b565
PR
341 }
342
343 pkt_frames = caf->frames_per_packet;
344 pkt_size = caf->bytes_per_packet;
345
346 if (pkt_size > 0 && pkt_frames == 1) {
347 pkt_size = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size;
348 pkt_size = FFMIN(pkt_size, left);
349 pkt_frames = pkt_size / caf->bytes_per_packet;
350 } else if (st->nb_index_entries) {
351 if (caf->packet_cnt < st->nb_index_entries - 1) {
352 pkt_size = st->index_entries[caf->packet_cnt + 1].pos - st->index_entries[caf->packet_cnt].pos;
353 pkt_frames = st->index_entries[caf->packet_cnt + 1].timestamp - st->index_entries[caf->packet_cnt].timestamp;
354 } else if (caf->packet_cnt == st->nb_index_entries - 1) {
355 pkt_size = caf->num_bytes - st->index_entries[caf->packet_cnt].pos;
356 pkt_frames = st->duration - st->index_entries[caf->packet_cnt].timestamp;
357 } else {
5ae092ee 358 return AVERROR(EIO);
5f14b565
PR
359 }
360 }
361
362 if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left)
5ae092ee 363 return AVERROR(EIO);
5f14b565
PR
364
365 res = av_get_packet(pb, pkt, pkt_size);
366 if (res < 0)
367 return res;
368
369 pkt->size = res;
370 pkt->stream_index = 0;
371 pkt->dts = pkt->pts = caf->frame_cnt;
372
373 caf->packet_cnt++;
374 caf->frame_cnt += pkt_frames;
375
376 return 0;
377}
378
379static int read_seek(AVFormatContext *s, int stream_index,
380 int64_t timestamp, int flags)
381{
382 AVStream *st = s->streams[0];
383 CaffContext *caf = s->priv_data;
af701d42 384 int64_t pos, packet_cnt, frame_cnt;
5f14b565
PR
385
386 timestamp = FFMAX(timestamp, 0);
387
388 if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
389 /* calculate new byte position based on target frame position */
390 pos = caf->bytes_per_packet * timestamp / caf->frames_per_packet;
391 if (caf->data_size > 0)
392 pos = FFMIN(pos, caf->data_size);
af701d42
JP
393 packet_cnt = pos / caf->bytes_per_packet;
394 frame_cnt = caf->frames_per_packet * packet_cnt;
5f14b565 395 } else if (st->nb_index_entries) {
af701d42
JP
396 packet_cnt = av_index_search_timestamp(st, timestamp, flags);
397 frame_cnt = st->index_entries[packet_cnt].timestamp;
398 pos = st->index_entries[packet_cnt].pos;
5f14b565
PR
399 } else {
400 return -1;
401 }
402
af701d42
JP
403 if (avio_seek(s->pb, pos + caf->data_start, SEEK_SET) < 0)
404 return -1;
405
406 caf->packet_cnt = packet_cnt;
407 caf->frame_cnt = frame_cnt;
408
5f14b565
PR
409 return 0;
410}
411
c6610a21 412AVInputFormat ff_caf_demuxer = {
dfc2c4d9
AK
413 .name = "caf",
414 .long_name = NULL_IF_CONFIG_SMALL("Apple Core Audio Format"),
415 .priv_data_size = sizeof(CaffContext),
416 .read_probe = probe,
417 .read_header = read_header,
418 .read_packet = read_packet,
419 .read_seek = read_seek,
67ffcb96 420 .codec_tag = (const AVCodecTag* const []){ ff_codec_caf_tags, 0 },
5f14b565 421};