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
39 typedef 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
probe(const AVProbeData * p)51 static int probe(const 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 */
read_desc_chunk(AVFormatContext * s)59 static 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_clipd(av_int2double(avio_rb64(pb)), 0, INT_MAX);
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 if (caf->bytes_per_packet < 0 || caf->frames_per_packet < 0)
83 return AVERROR_INVALIDDATA;
84
85 /* calculate bit rate for constant size packets */
86 if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
87 st->codecpar->bit_rate = (uint64_t)st->codecpar->sample_rate * (uint64_t)caf->bytes_per_packet * 8
88 / (uint64_t)caf->frames_per_packet;
89 } else {
90 st->codecpar->bit_rate = 0;
91 }
92
93 /* determine codec */
94 if (st->codecpar->codec_tag == MKTAG('l','p','c','m'))
95 st->codecpar->codec_id = ff_mov_get_lpcm_codec_id(st->codecpar->bits_per_coded_sample, (flags ^ 0x2) | 0x4);
96 else
97 st->codecpar->codec_id = ff_codec_get_id(ff_codec_caf_tags, st->codecpar->codec_tag);
98 return 0;
99 }
100
101 /** Read magic cookie chunk */
read_kuki_chunk(AVFormatContext * s,int64_t size)102 static int read_kuki_chunk(AVFormatContext *s, int64_t size)
103 {
104 AVIOContext *pb = s->pb;
105 AVStream *st = s->streams[0];
106 int ret;
107
108 if (size < 0 || size > INT_MAX - AV_INPUT_BUFFER_PADDING_SIZE)
109 return -1;
110
111 if (st->codecpar->codec_id == AV_CODEC_ID_AAC) {
112 /* The magic cookie format for AAC is an mp4 esds atom.
113 The lavc AAC decoder requires the data from the codec specific
114 description as extradata input. */
115 int strt, skip;
116
117 strt = avio_tell(pb);
118 ff_mov_read_esds(s, pb);
119 skip = size - (avio_tell(pb) - strt);
120 if (skip < 0 || !st->codecpar->extradata ||
121 st->codecpar->codec_id != AV_CODEC_ID_AAC) {
122 av_log(s, AV_LOG_ERROR, "invalid AAC magic cookie\n");
123 return AVERROR_INVALIDDATA;
124 }
125 avio_skip(pb, skip);
126 } else if (st->codecpar->codec_id == AV_CODEC_ID_ALAC) {
127 #define ALAC_PREAMBLE 12
128 #define ALAC_HEADER 36
129 #define ALAC_NEW_KUKI 24
130 uint8_t preamble[12];
131 if (size < ALAC_NEW_KUKI) {
132 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
133 avio_skip(pb, size);
134 return AVERROR_INVALIDDATA;
135 }
136 if (avio_read(pb, preamble, ALAC_PREAMBLE) != ALAC_PREAMBLE) {
137 av_log(s, AV_LOG_ERROR, "failed to read preamble\n");
138 return AVERROR_INVALIDDATA;
139 }
140
141 if ((ret = ff_alloc_extradata(st->codecpar, ALAC_HEADER)) < 0)
142 return ret;
143
144 /* For the old style cookie, we skip 12 bytes, then read 36 bytes.
145 * The new style cookie only contains the last 24 bytes of what was
146 * 36 bytes in the old style cookie, so we fabricate the first 12 bytes
147 * in that case to maintain compatibility. */
148 if (!memcmp(&preamble[4], "frmaalac", 8)) {
149 if (size < ALAC_PREAMBLE + ALAC_HEADER) {
150 av_log(s, AV_LOG_ERROR, "invalid ALAC magic cookie\n");
151 av_freep(&st->codecpar->extradata);
152 return AVERROR_INVALIDDATA;
153 }
154 if (avio_read(pb, st->codecpar->extradata, ALAC_HEADER) != ALAC_HEADER) {
155 av_log(s, AV_LOG_ERROR, "failed to read kuki header\n");
156 av_freep(&st->codecpar->extradata);
157 return AVERROR_INVALIDDATA;
158 }
159 avio_skip(pb, size - ALAC_PREAMBLE - ALAC_HEADER);
160 } else {
161 AV_WB32(st->codecpar->extradata, 36);
162 memcpy(&st->codecpar->extradata[4], "alac", 4);
163 AV_WB32(&st->codecpar->extradata[8], 0);
164 memcpy(&st->codecpar->extradata[12], preamble, 12);
165 if (avio_read(pb, &st->codecpar->extradata[24], ALAC_NEW_KUKI - 12) != ALAC_NEW_KUKI - 12) {
166 av_log(s, AV_LOG_ERROR, "failed to read new kuki header\n");
167 av_freep(&st->codecpar->extradata);
168 return AVERROR_INVALIDDATA;
169 }
170 avio_skip(pb, size - ALAC_NEW_KUKI);
171 }
172 } else if (st->codecpar->codec_id == AV_CODEC_ID_OPUS) {
173 // The data layout for Opus is currently unknown, so we do not export
174 // extradata at all. Multichannel streams are not supported.
175 if (st->codecpar->channels > 2) {
176 avpriv_request_sample(s, "multichannel Opus in CAF");
177 return AVERROR_PATCHWELCOME;
178 }
179 avio_skip(pb, size);
180 } else if ((ret = ff_get_extradata(s, st->codecpar, pb, size)) < 0) {
181 return ret;
182 }
183
184 return 0;
185 }
186
187 /** Read packet table chunk */
read_pakt_chunk(AVFormatContext * s,int64_t size)188 static int read_pakt_chunk(AVFormatContext *s, int64_t size)
189 {
190 AVIOContext *pb = s->pb;
191 AVStream *st = s->streams[0];
192 CafContext *caf = s->priv_data;
193 int64_t pos = 0, ccount, num_packets;
194 int i;
195 int ret;
196
197 ccount = avio_tell(pb);
198
199 num_packets = avio_rb64(pb);
200 if (num_packets < 0 || INT32_MAX / sizeof(AVIndexEntry) < num_packets)
201 return AVERROR_INVALIDDATA;
202
203 st->nb_frames = avio_rb64(pb); /* valid frames */
204 st->nb_frames += avio_rb32(pb); /* priming frames */
205 st->nb_frames += avio_rb32(pb); /* remainder frames */
206
207 if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
208 st->duration = caf->frames_per_packet * num_packets;
209 pos = caf-> bytes_per_packet * num_packets;
210 } else {
211 st->duration = 0;
212 for (i = 0; i < num_packets; i++) {
213 if (avio_feof(pb))
214 return AVERROR_INVALIDDATA;
215 ret = av_add_index_entry(s->streams[0], pos, st->duration, 0, 0, AVINDEX_KEYFRAME);
216 if (ret < 0)
217 return ret;
218 pos += caf->bytes_per_packet ? caf->bytes_per_packet : ff_mp4_read_descr_len(pb);
219 st->duration += caf->frames_per_packet ? caf->frames_per_packet : ff_mp4_read_descr_len(pb);
220 }
221 }
222
223 if (avio_tell(pb) - ccount > size) {
224 av_log(s, AV_LOG_ERROR, "error reading packet table\n");
225 return AVERROR_INVALIDDATA;
226 }
227 avio_skip(pb, ccount + size - avio_tell(pb));
228
229 caf->num_bytes = pos;
230 return 0;
231 }
232
233 /** Read information chunk */
read_info_chunk(AVFormatContext * s,int64_t size)234 static void read_info_chunk(AVFormatContext *s, int64_t size)
235 {
236 AVIOContext *pb = s->pb;
237 unsigned int i;
238 unsigned int nb_entries = avio_rb32(pb);
239 for (i = 0; i < nb_entries && !avio_feof(pb); i++) {
240 char key[32];
241 char value[1024];
242 avio_get_str(pb, INT_MAX, key, sizeof(key));
243 avio_get_str(pb, INT_MAX, value, sizeof(value));
244 av_dict_set(&s->metadata, key, value, 0);
245 }
246 }
247
read_header(AVFormatContext * s)248 static int read_header(AVFormatContext *s)
249 {
250 AVIOContext *pb = s->pb;
251 CafContext *caf = s->priv_data;
252 AVStream *st;
253 uint32_t tag = 0;
254 int found_data, ret;
255 int64_t size, pos;
256
257 avio_skip(pb, 8); /* magic, version, file flags */
258
259 /* audio description chunk */
260 if (avio_rb32(pb) != MKBETAG('d','e','s','c')) {
261 av_log(s, AV_LOG_ERROR, "desc chunk not present\n");
262 return AVERROR_INVALIDDATA;
263 }
264 size = avio_rb64(pb);
265 if (size != 32)
266 return AVERROR_INVALIDDATA;
267
268 ret = read_desc_chunk(s);
269 if (ret)
270 return ret;
271 st = s->streams[0];
272
273 /* parse each chunk */
274 found_data = 0;
275 while (!avio_feof(pb)) {
276
277 /* stop at data chunk if seeking is not supported or
278 data chunk size is unknown */
279 if (found_data && (caf->data_size < 0 || !(pb->seekable & AVIO_SEEKABLE_NORMAL)))
280 break;
281
282 tag = avio_rb32(pb);
283 size = avio_rb64(pb);
284 pos = avio_tell(pb);
285 if (avio_feof(pb))
286 break;
287
288 switch (tag) {
289 case MKBETAG('d','a','t','a'):
290 avio_skip(pb, 4); /* edit count */
291 caf->data_start = avio_tell(pb);
292 caf->data_size = size < 0 ? -1 : size - 4;
293 if (caf->data_size > 0 && (pb->seekable & AVIO_SEEKABLE_NORMAL))
294 avio_skip(pb, caf->data_size);
295 found_data = 1;
296 break;
297
298 case MKBETAG('c','h','a','n'):
299 if ((ret = ff_mov_read_chan(s, s->pb, st, size)) < 0)
300 return ret;
301 break;
302
303 /* magic cookie chunk */
304 case MKBETAG('k','u','k','i'):
305 if (read_kuki_chunk(s, size))
306 return AVERROR_INVALIDDATA;
307 break;
308
309 /* packet table chunk */
310 case MKBETAG('p','a','k','t'):
311 if (read_pakt_chunk(s, size))
312 return AVERROR_INVALIDDATA;
313 break;
314
315 case MKBETAG('i','n','f','o'):
316 read_info_chunk(s, size);
317 break;
318
319 default:
320 av_log(s, AV_LOG_WARNING,
321 "skipping CAF chunk: %08"PRIX32" (%s), size %"PRId64"\n",
322 tag, av_fourcc2str(av_bswap32(tag)), size);
323 case MKBETAG('f','r','e','e'):
324 if (size < 0 && found_data)
325 goto found_data;
326 if (size < 0)
327 return AVERROR_INVALIDDATA;
328 break;
329 }
330
331 if (size > 0) {
332 if (pos > INT64_MAX - size)
333 return AVERROR_INVALIDDATA;
334 avio_skip(pb, FFMAX(0, pos + size - avio_tell(pb)));
335 }
336 }
337
338 if (!found_data)
339 return AVERROR_INVALIDDATA;
340
341 found_data:
342 if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) {
343 if (caf->data_size > 0)
344 st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet;
345 } else if (st->nb_index_entries && st->duration > 0) {
346 if (st->codecpar->sample_rate && caf->data_size / st->duration > INT64_MAX / st->codecpar->sample_rate / 8) {
347 av_log(s, AV_LOG_ERROR, "Overflow during bit rate calculation %d * 8 * %"PRId64"\n",
348 st->codecpar->sample_rate, caf->data_size / st->duration);
349 return AVERROR_INVALIDDATA;
350 }
351 st->codecpar->bit_rate = st->codecpar->sample_rate * 8LL *
352 (caf->data_size / st->duration);
353 } else {
354 av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when "
355 "block size or frame size are variable.\n");
356 return AVERROR_INVALIDDATA;
357 }
358
359 avpriv_set_pts_info(st, 64, 1, st->codecpar->sample_rate);
360 st->start_time = 0;
361
362 /* position the stream at the start of data */
363 if (caf->data_size >= 0)
364 avio_seek(pb, caf->data_start, SEEK_SET);
365
366 return 0;
367 }
368
369 #define CAF_MAX_PKT_SIZE 4096
370
read_packet(AVFormatContext * s,AVPacket * pkt)371 static int read_packet(AVFormatContext *s, AVPacket *pkt)
372 {
373 AVIOContext *pb = s->pb;
374 AVStream *st = s->streams[0];
375 CafContext *caf = s->priv_data;
376 int res, pkt_size = 0, pkt_frames = 0;
377 int64_t left = CAF_MAX_PKT_SIZE;
378
379 if (avio_feof(pb))
380 return AVERROR_EOF;
381
382 /* don't read past end of data chunk */
383 if (caf->data_size > 0) {
384 left = (caf->data_start + caf->data_size) - avio_tell(pb);
385 if (!left)
386 return AVERROR_EOF;
387 if (left < 0)
388 return AVERROR(EIO);
389 }
390
391 pkt_frames = caf->frames_per_packet;
392 pkt_size = caf->bytes_per_packet;
393
394 if (pkt_size > 0 && pkt_frames == 1) {
395 pkt_size = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size;
396 pkt_size = FFMIN(pkt_size, left);
397 pkt_frames = pkt_size / caf->bytes_per_packet;
398 } else if (st->nb_index_entries) {
399 if (caf->packet_cnt < st->nb_index_entries - 1) {
400 pkt_size = st->index_entries[caf->packet_cnt + 1].pos - st->index_entries[caf->packet_cnt].pos;
401 pkt_frames = st->index_entries[caf->packet_cnt + 1].timestamp - st->index_entries[caf->packet_cnt].timestamp;
402 } else if (caf->packet_cnt == st->nb_index_entries - 1) {
403 pkt_size = caf->num_bytes - st->index_entries[caf->packet_cnt].pos;
404 pkt_frames = st->duration - st->index_entries[caf->packet_cnt].timestamp;
405 } else {
406 return AVERROR(EIO);
407 }
408 }
409
410 if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left)
411 return AVERROR(EIO);
412
413 res = av_get_packet(pb, pkt, pkt_size);
414 if (res < 0)
415 return res;
416
417 pkt->size = res;
418 pkt->stream_index = 0;
419 pkt->dts = pkt->pts = caf->frame_cnt;
420
421 caf->packet_cnt++;
422 caf->frame_cnt += pkt_frames;
423
424 return 0;
425 }
426
read_seek(AVFormatContext * s,int stream_index,int64_t timestamp,int flags)427 static int read_seek(AVFormatContext *s, int stream_index,
428 int64_t timestamp, int flags)
429 {
430 AVStream *st = s->streams[0];
431 CafContext *caf = s->priv_data;
432 int64_t pos, packet_cnt, frame_cnt;
433
434 timestamp = FFMAX(timestamp, 0);
435
436 if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) {
437 /* calculate new byte position based on target frame position */
438 pos = caf->bytes_per_packet * (timestamp / caf->frames_per_packet);
439 if (caf->data_size > 0)
440 pos = FFMIN(pos, caf->data_size);
441 packet_cnt = pos / caf->bytes_per_packet;
442 frame_cnt = caf->frames_per_packet * packet_cnt;
443 } else if (st->nb_index_entries) {
444 packet_cnt = av_index_search_timestamp(st, timestamp, flags);
445 frame_cnt = st->index_entries[packet_cnt].timestamp;
446 pos = st->index_entries[packet_cnt].pos;
447 } else {
448 return -1;
449 }
450
451 if (avio_seek(s->pb, pos + caf->data_start, SEEK_SET) < 0)
452 return -1;
453
454 caf->packet_cnt = packet_cnt;
455 caf->frame_cnt = frame_cnt;
456
457 return 0;
458 }
459
460 AVInputFormat ff_caf_demuxer = {
461 .name = "caf",
462 .long_name = NULL_IF_CONFIG_SMALL("Apple CAF (Core Audio Format)"),
463 .priv_data_size = sizeof(CafContext),
464 .read_probe = probe,
465 .read_header = read_header,
466 .read_packet = read_packet,
467 .read_seek = read_seek,
468 .codec_tag = ff_caf_codec_tags_list,
469 };
470