summaryrefslogtreecommitdiff
path: root/libavformat/cafdec.c (plain)
blob: 19939d3a859fa66f62aed9f7cac1cd6cbc36b82d
1/*
2 * Core Audio Format demuxer
3 * Copyright (c) 2007 Justin Ruggles
4 * Copyright (c) 2009 Peter Ross
5 *
6 * This file is part of FFmpeg.
7 *
8 * FFmpeg is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * FFmpeg is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with FFmpeg; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23/**
24 * @file
25 * Core Audio Format demuxer
26 */
27
28#include <inttypes.h>
29
30#include "avformat.h"
31#include "internal.h"
32#include "isom.h"
33#include "mov_chan.h"
34#include "libavutil/intreadwrite.h"
35#include "libavutil/intfloat.h"
36#include "libavutil/dict.h"
37#include "caf.h"
38
39typedef struct CafContext {
40 int bytes_per_packet; ///< bytes in a packet, or 0 if variable
41 int frames_per_packet; ///< frames in a packet, or 0 if variable
42 int64_t num_bytes; ///< total number of bytes in stream
43
44 int64_t packet_cnt; ///< packet counter
45 int64_t frame_cnt; ///< frame counter
46
47 int64_t data_start; ///< data start position, in bytes
48 int64_t data_size; ///< raw data size, in bytes
49} CafContext;
50
51static int probe(AVProbeData *p)
52{
53 if (AV_RB32(p->buf) == MKBETAG('c','a','f','f') && AV_RB16(&p->buf[4]) == 1)
54 return AVPROBE_SCORE_MAX;
55 return 0;
56}
57
58/** Read audio description chunk */
59static int read_desc_chunk(AVFormatContext *s)
60{
61 AVIOContext *pb = s->pb;
62 CafContext *caf = s->priv_data;
63 AVStream *st;
64 int flags;
65
66 /* new audio stream */
67 st = avformat_new_stream(s, NULL);
68 if (!st)
69 return AVERROR(ENOMEM);
70
71 /* parse format description */
72 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
73 st->codecpar->sample_rate = av_int2double(avio_rb64(pb));
74 st->codecpar->codec_tag = avio_rl32(pb);
75 flags = avio_rb32(pb);
76 caf->bytes_per_packet = avio_rb32(pb);
77 st->codecpar->block_align = caf->bytes_per_packet;
78 caf->frames_per_packet = avio_rb32(pb);
79 st->codecpar->channels = avio_rb32(pb);
80 st->codecpar->bits_per_coded_sample = avio_rb32(pb);
81
82 /* calculate bit rate for constant size packets */
83 if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
84 st->codecpar->bit_rate = (uint64_t)st->codecpar->sample_rate * (uint64_t)caf->bytes_per_packet * 8
85 / (uint64_t)caf->frames_per_packet;
86 } else {
87 st->codecpar->bit_rate = 0;
88 }
89
90 /* determine codec */
91 if (st->codecpar->codec_tag == MKTAG('l','p','c','m'))
92 st->codecpar->codec_id = ff_mov_get_lpcm_codec_id(st->codecpar->bits_per_coded_sample, (flags ^ 0x2) | 0x4);
93 else
94 st->codecpar->codec_id = ff_codec_get_id(ff_codec_caf_tags, st->codecpar->codec_tag);
95 return 0;
96}
97
98/** Read magic cookie chunk */
99static int read_kuki_chunk(AVFormatContext *s, int64_t size)
100{
101 AVIOContext *pb = s->pb;
102 AVStream *st = s->streams[0];
103
104 if (size < 0 || size > INT_MAX - AV_INPUT_BUFFER_PADDING_SIZE)
105 return -1;
106
107 if (st->codecpar->codec_id == AV_CODEC_ID_AAC) {
108 /* The magic cookie format for AAC is an mp4 esds atom.
109 The lavc AAC decoder requires the data from the codec specific
110 description as extradata input. */
111 int strt, skip;
112
113 strt = avio_tell(pb);
114 ff_mov_read_esds(s, pb);
115 skip = size - (avio_tell(pb) - strt);
116 if (skip < 0 || !st->codecpar->extradata ||
117 st->codecpar->codec_id != AV_CODEC_ID_AAC) {
118 av_log(s, AV_LOG_ERROR, "invalid AAC magic cookie\n");
119 return AVERROR_INVALIDDATA;
120 }
121 avio_skip(pb, skip);
122 } else if (st->codecpar->codec_id == AV_CODEC_ID_ALAC) {
123#define ALAC_PREAMBLE 12
124#define ALAC_HEADER 36
125#define ALAC_NEW_KUKI 24
126 uint8_t preamble[12];
127 if (size < ALAC_NEW_KUKI) {
128 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
129 avio_skip(pb, size);
130 return AVERROR_INVALIDDATA;
131 }
132 if (avio_read(pb, preamble, ALAC_PREAMBLE) != ALAC_PREAMBLE) {
133 av_log(s, AV_LOG_ERROR, "failed to read preamble\n");
134 return AVERROR_INVALIDDATA;
135 }
136
137 av_freep(&st->codecpar->extradata);
138 if (ff_alloc_extradata(st->codecpar, ALAC_HEADER))
139 return AVERROR(ENOMEM);
140
141 /* For the old style cookie, we skip 12 bytes, then read 36 bytes.
142 * The new style cookie only contains the last 24 bytes of what was
143 * 36 bytes in the old style cookie, so we fabricate the first 12 bytes
144 * in that case to maintain compatibility. */
145 if (!memcmp(&preamble[4], "frmaalac", 8)) {
146 if (size < ALAC_PREAMBLE + ALAC_HEADER) {
147 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
148 av_freep(&st->codecpar->extradata);
149 return AVERROR_INVALIDDATA;
150 }
151 if (avio_read(pb, st->codecpar->extradata, ALAC_HEADER) != ALAC_HEADER) {
152 av_log(s, AV_LOG_ERROR, "failed to read kuki header\n");
153 av_freep(&st->codecpar->extradata);
154 return AVERROR_INVALIDDATA;
155 }
156 avio_skip(pb, size - ALAC_PREAMBLE - ALAC_HEADER);
157 } else {
158 AV_WB32(st->codecpar->extradata, 36);
159 memcpy(&st->codecpar->extradata[4], "alac", 4);
160 AV_WB32(&st->codecpar->extradata[8], 0);
161 memcpy(&st->codecpar->extradata[12], preamble, 12);
162 if (avio_read(pb, &st->codecpar->extradata[24], ALAC_NEW_KUKI - 12) != ALAC_NEW_KUKI - 12) {
163 av_log(s, AV_LOG_ERROR, "failed to read new kuki header\n");
164 av_freep(&st->codecpar->extradata);
165 return AVERROR_INVALIDDATA;
166 }
167 avio_skip(pb, size - ALAC_NEW_KUKI);
168 }
169 } else {
170 av_freep(&st->codecpar->extradata);
171 if (ff_get_extradata(s, st->codecpar, pb, size) < 0)
172 return AVERROR(ENOMEM);
173 }
174
175 return 0;
176}
177
178/** Read packet table chunk */
179static int read_pakt_chunk(AVFormatContext *s, int64_t size)
180{
181 AVIOContext *pb = s->pb;
182 AVStream *st = s->streams[0];
183 CafContext *caf = s->priv_data;
184 int64_t pos = 0, ccount, num_packets;
185 int i;
186
187 ccount = avio_tell(pb);
188
189 num_packets = avio_rb64(pb);
190 if (num_packets < 0 || INT32_MAX / sizeof(AVIndexEntry) < num_packets)
191 return AVERROR_INVALIDDATA;
192
193 st->nb_frames = avio_rb64(pb); /* valid frames */
194 st->nb_frames += avio_rb32(pb); /* priming frames */
195 st->nb_frames += avio_rb32(pb); /* remainder frames */
196
197 st->duration = 0;
198 for (i = 0; i < num_packets; i++) {
199 av_add_index_entry(s->streams[0], pos, st->duration, 0, 0, AVINDEX_KEYFRAME);
200 pos += caf->bytes_per_packet ? caf->bytes_per_packet : ff_mp4_read_descr_len(pb);
201 st->duration += caf->frames_per_packet ? caf->frames_per_packet : ff_mp4_read_descr_len(pb);
202 }
203
204 if (avio_tell(pb) - ccount > size) {
205 av_log(s, AV_LOG_ERROR, "error reading packet table\n");
206 return AVERROR_INVALIDDATA;
207 }
208 avio_skip(pb, ccount + size - avio_tell(pb));
209
210 caf->num_bytes = pos;
211 return 0;
212}
213
214/** Read information chunk */
215static void read_info_chunk(AVFormatContext *s, int64_t size)
216{
217 AVIOContext *pb = s->pb;
218 unsigned int i;
219 unsigned int nb_entries = avio_rb32(pb);
220 for (i = 0; i < nb_entries && !avio_feof(pb); i++) {
221 char key[32];
222 char value[1024];
223 avio_get_str(pb, INT_MAX, key, sizeof(key));
224 avio_get_str(pb, INT_MAX, value, sizeof(value));
225 av_dict_set(&s->metadata, key, value, 0);
226 }
227}
228
229static int read_header(AVFormatContext *s)
230{
231 AVIOContext *pb = s->pb;
232 CafContext *caf = s->priv_data;
233 AVStream *st;
234 uint32_t tag = 0;
235 int found_data, ret;
236 int64_t size, pos;
237
238 avio_skip(pb, 8); /* magic, version, file flags */
239
240 /* audio description chunk */
241 if (avio_rb32(pb) != MKBETAG('d','e','s','c')) {
242 av_log(s, AV_LOG_ERROR, "desc chunk not present\n");
243 return AVERROR_INVALIDDATA;
244 }
245 size = avio_rb64(pb);
246 if (size != 32)
247 return AVERROR_INVALIDDATA;
248
249 ret = read_desc_chunk(s);
250 if (ret)
251 return ret;
252 st = s->streams[0];
253
254 /* parse each chunk */
255 found_data = 0;
256 while (!avio_feof(pb)) {
257
258 /* stop at data chunk if seeking is not supported or
259 data chunk size is unknown */
260 if (found_data && (caf->data_size < 0 || !(pb->seekable & AVIO_SEEKABLE_NORMAL)))
261 break;
262
263 tag = avio_rb32(pb);
264 size = avio_rb64(pb);
265 pos = avio_tell(pb);
266 if (avio_feof(pb))
267 break;
268
269 switch (tag) {
270 case MKBETAG('d','a','t','a'):
271 avio_skip(pb, 4); /* edit count */
272 caf->data_start = avio_tell(pb);
273 caf->data_size = size < 0 ? -1 : size - 4;
274 if (caf->data_size > 0 && (pb->seekable & AVIO_SEEKABLE_NORMAL))
275 avio_skip(pb, caf->data_size);
276 found_data = 1;
277 break;
278
279 case MKBETAG('c','h','a','n'):
280 if ((ret = ff_mov_read_chan(s, s->pb, st, size)) < 0)
281 return ret;
282 break;
283
284 /* magic cookie chunk */
285 case MKBETAG('k','u','k','i'):
286 if (read_kuki_chunk(s, size))
287 return AVERROR_INVALIDDATA;
288 break;
289
290 /* packet table chunk */
291 case MKBETAG('p','a','k','t'):
292 if (read_pakt_chunk(s, size))
293 return AVERROR_INVALIDDATA;
294 break;
295
296 case MKBETAG('i','n','f','o'):
297 read_info_chunk(s, size);
298 break;
299
300 default:
301 av_log(s, AV_LOG_WARNING,
302 "skipping CAF chunk: %08"PRIX32" (%s), size %"PRId64"\n",
303 tag, av_fourcc2str(av_bswap32(tag)), size);
304 case MKBETAG('f','r','e','e'):
305 if (size < 0)
306 return AVERROR_INVALIDDATA;
307 break;
308 }
309
310 if (size > 0) {
311 if (pos > INT64_MAX - size)
312 return AVERROR_INVALIDDATA;
313 avio_skip(pb, FFMAX(0, pos + size - avio_tell(pb)));
314 }
315 }
316
317 if (!found_data)
318 return AVERROR_INVALIDDATA;
319
320 if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
321 if (caf->data_size > 0)
322 st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet;
323 } else if (st->nb_index_entries && st->duration > 0) {
324 if (st->codecpar->sample_rate && caf->data_size / st->duration > INT64_MAX / st->codecpar->sample_rate / 8) {
325 av_log(s, AV_LOG_ERROR, "Overflow during bit rate calculation %d * 8 * %"PRId64"\n",
326 st->codecpar->sample_rate, caf->data_size / st->duration);
327 return AVERROR_INVALIDDATA;
328 }
329 st->codecpar->bit_rate = st->codecpar->sample_rate * 8LL *
330 (caf->data_size / st->duration);
331 } else {
332 av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when "
333 "block size or frame size are variable.\n");
334 return AVERROR_INVALIDDATA;
335 }
336
337 avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate);
338 st->start_time = 0;
339
340 /* position the stream at the start of data */
341 if (caf->data_size >= 0)
342 avio_seek(pb, caf->data_start, SEEK_SET);
343
344 return 0;
345}
346
347#define CAF_MAX_PKT_SIZE 4096
348
349static int read_packet(AVFormatContext *s, AVPacket *pkt)
350{
351 AVIOContext *pb = s->pb;
352 AVStream *st = s->streams[0];
353 CafContext *caf = s->priv_data;
354 int res, pkt_size = 0, pkt_frames = 0;
355 int64_t left = CAF_MAX_PKT_SIZE;
356
357 if (avio_feof(pb))
358 return AVERROR_EOF;
359
360 /* don't read past end of data chunk */
361 if (caf->data_size > 0) {
362 left = (caf->data_start + caf->data_size) - avio_tell(pb);
363 if (!left)
364 return AVERROR_EOF;
365 if (left < 0)
366 return AVERROR(EIO);
367 }
368
369 pkt_frames = caf->frames_per_packet;
370 pkt_size = caf->bytes_per_packet;
371
372 if (pkt_size > 0 && pkt_frames == 1) {
373 pkt_size = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size;
374 pkt_size = FFMIN(pkt_size, left);
375 pkt_frames = pkt_size / caf->bytes_per_packet;
376 } else if (st->nb_index_entries) {
377 if (caf->packet_cnt < st->nb_index_entries - 1) {
378 pkt_size = st->index_entries[caf->packet_cnt + 1].pos - st->index_entries[caf->packet_cnt].pos;
379 pkt_frames = st->index_entries[caf->packet_cnt + 1].timestamp - st->index_entries[caf->packet_cnt].timestamp;
380 } else if (caf->packet_cnt == st->nb_index_entries - 1) {
381 pkt_size = caf->num_bytes - st->index_entries[caf->packet_cnt].pos;
382 pkt_frames = st->duration - st->index_entries[caf->packet_cnt].timestamp;
383 } else {
384 return AVERROR(EIO);
385 }
386 }
387
388 if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left)
389 return AVERROR(EIO);
390
391 res = av_get_packet(pb, pkt, pkt_size);
392 if (res < 0)
393 return res;
394
395 pkt->size = res;
396 pkt->stream_index = 0;
397 pkt->dts = pkt->pts = caf->frame_cnt;
398
399 caf->packet_cnt++;
400 caf->frame_cnt += pkt_frames;
401
402 return 0;
403}
404
405static int read_seek(AVFormatContext *s, int stream_index,
406 int64_t timestamp, int flags)
407{
408 AVStream *st = s->streams[0];
409 CafContext *caf = s->priv_data;
410 int64_t pos, packet_cnt, frame_cnt;
411
412 timestamp = FFMAX(timestamp, 0);
413
414 if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
415 /* calculate new byte position based on target frame position */
416 pos = caf->bytes_per_packet * (timestamp / caf->frames_per_packet);
417 if (caf->data_size > 0)
418 pos = FFMIN(pos, caf->data_size);
419 packet_cnt = pos / caf->bytes_per_packet;
420 frame_cnt = caf->frames_per_packet * packet_cnt;
421 } else if (st->nb_index_entries) {
422 packet_cnt = av_index_search_timestamp(st, timestamp, flags);
423 frame_cnt = st->index_entries[packet_cnt].timestamp;
424 pos = st->index_entries[packet_cnt].pos;
425 } else {
426 return -1;
427 }
428
429 if (avio_seek(s->pb, pos + caf->data_start, SEEK_SET) < 0)
430 return -1;
431
432 caf->packet_cnt = packet_cnt;
433 caf->frame_cnt = frame_cnt;
434
435 return 0;
436}
437
438AVInputFormat ff_caf_demuxer = {
439 .name = "caf",
440 .long_name = NULL_IF_CONFIG_SMALL("Apple CAF (Core Audio Format)"),
441 .priv_data_size = sizeof(CafContext),
442 .read_probe = probe,
443 .read_header = read_header,
444 .read_packet = read_packet,
445 .read_seek = read_seek,
446 .codec_tag = (const AVCodecTag* const []){ ff_codec_caf_tags, 0 },
447};
448
generated by cgit at 2024-05-17 04:45:12 (GMT)