Details | Last modification | View Log | RSS feed
Rev | Author | Line No. | Line |
---|---|---|---|
6147 | serge | 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 |
||
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 | |||
51 | static 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 */ |
||
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->codec->codec_type = AVMEDIA_TYPE_AUDIO; |
||
73 | st->codec->sample_rate = av_int2double(avio_rb64(pb)); |
||
74 | st->codec->codec_tag = avio_rl32(pb); |
||
75 | flags = avio_rb32(pb); |
||
76 | caf->bytes_per_packet = avio_rb32(pb); |
||
77 | st->codec->block_align = caf->bytes_per_packet; |
||
78 | caf->frames_per_packet = avio_rb32(pb); |
||
79 | st->codec->channels = avio_rb32(pb); |
||
80 | st->codec->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->codec->bit_rate = (uint64_t)st->codec->sample_rate * (uint64_t)caf->bytes_per_packet * 8 |
||
85 | / (uint64_t)caf->frames_per_packet; |
||
86 | } else { |
||
87 | st->codec->bit_rate = 0; |
||
88 | } |
||
89 | |||
90 | /* determine codec */ |
||
91 | if (st->codec->codec_tag == MKTAG('l','p','c','m')) |
||
92 | st->codec->codec_id = ff_mov_get_lpcm_codec_id(st->codec->bits_per_coded_sample, (flags ^ 0x2) | 0x4); |
||
93 | else |
||
94 | st->codec->codec_id = ff_codec_get_id(ff_codec_caf_tags, st->codec->codec_tag); |
||
95 | return 0; |
||
96 | } |
||
97 | |||
98 | /** Read magic cookie chunk */ |
||
99 | static 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->codec->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->codec->extradata || |
||
117 | st->codec->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->codec->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->codec->extradata); |
||
138 | if (ff_alloc_extradata(st->codec, 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->codec->extradata); |
||
149 | return AVERROR_INVALIDDATA; |
||
150 | } |
||
151 | if (avio_read(pb, st->codec->extradata, ALAC_HEADER) != ALAC_HEADER) { |
||
152 | av_log(s, AV_LOG_ERROR, "failed to read kuki header\n"); |
||
153 | av_freep(&st->codec->extradata); |
||
154 | return AVERROR_INVALIDDATA; |
||
155 | } |
||
156 | avio_skip(pb, size - ALAC_PREAMBLE - ALAC_HEADER); |
||
157 | } else { |
||
158 | AV_WB32(st->codec->extradata, 36); |
||
159 | memcpy(&st->codec->extradata[4], "alac", 4); |
||
160 | AV_WB32(&st->codec->extradata[8], 0); |
||
161 | memcpy(&st->codec->extradata[12], preamble, 12); |
||
162 | if (avio_read(pb, &st->codec->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->codec->extradata); |
||
165 | return AVERROR_INVALIDDATA; |
||
166 | } |
||
167 | avio_skip(pb, size - ALAC_NEW_KUKI); |
||
168 | } |
||
169 | } else { |
||
170 | av_freep(&st->codec->extradata); |
||
171 | if (ff_get_extradata(st->codec, pb, size) < 0) |
||
172 | return AVERROR(ENOMEM); |
||
173 | } |
||
174 | |||
175 | return 0; |
||
176 | } |
||
177 | |||
178 | /** Read packet table chunk */ |
||
179 | static 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 */ |
||
215 | static 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 | |||
229 | static 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)) |
||
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) |
||
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 | #define _(x) ((x) >= ' ' ? (x) : ' ') |
||
302 | av_log(s, AV_LOG_WARNING, |
||
303 | "skipping CAF chunk: %08"PRIX32" (%c%c%c%c), size %"PRId64"\n", |
||
304 | tag, _(tag>>24), _((tag>>16)&0xFF), _((tag>>8)&0xFF), _(tag&0xFF), size); |
||
305 | #undef _ |
||
306 | case MKBETAG('f','r','e','e'): |
||
307 | if (size < 0) |
||
308 | return AVERROR_INVALIDDATA; |
||
309 | break; |
||
310 | } |
||
311 | |||
312 | if (size > 0) { |
||
313 | if (pos > INT64_MAX - size) |
||
314 | return AVERROR_INVALIDDATA; |
||
315 | avio_skip(pb, FFMAX(0, pos + size - avio_tell(pb))); |
||
316 | } |
||
317 | } |
||
318 | |||
319 | if (!found_data) |
||
320 | return AVERROR_INVALIDDATA; |
||
321 | |||
322 | if (caf->bytes_per_packet > 0 && caf->frames_per_packet > 0) { |
||
323 | if (caf->data_size > 0) |
||
324 | st->nb_frames = (caf->data_size / caf->bytes_per_packet) * caf->frames_per_packet; |
||
325 | } else if (st->nb_index_entries && st->duration > 0) { |
||
326 | st->codec->bit_rate = st->codec->sample_rate * caf->data_size * 8 / |
||
327 | st->duration; |
||
328 | } else { |
||
329 | av_log(s, AV_LOG_ERROR, "Missing packet table. It is required when " |
||
330 | "block size or frame size are variable.\n"); |
||
331 | return AVERROR_INVALIDDATA; |
||
332 | } |
||
333 | |||
334 | avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate); |
||
335 | st->start_time = 0; |
||
336 | |||
337 | /* position the stream at the start of data */ |
||
338 | if (caf->data_size >= 0) |
||
339 | avio_seek(pb, caf->data_start, SEEK_SET); |
||
340 | |||
341 | return 0; |
||
342 | } |
||
343 | |||
344 | #define CAF_MAX_PKT_SIZE 4096 |
||
345 | |||
346 | static int read_packet(AVFormatContext *s, AVPacket *pkt) |
||
347 | { |
||
348 | AVIOContext *pb = s->pb; |
||
349 | AVStream *st = s->streams[0]; |
||
350 | CafContext *caf = s->priv_data; |
||
351 | int res, pkt_size = 0, pkt_frames = 0; |
||
352 | int64_t left = CAF_MAX_PKT_SIZE; |
||
353 | |||
354 | if (avio_feof(pb)) |
||
355 | return AVERROR_EOF; |
||
356 | |||
357 | /* don't read past end of data chunk */ |
||
358 | if (caf->data_size > 0) { |
||
359 | left = (caf->data_start + caf->data_size) - avio_tell(pb); |
||
360 | if (!left) |
||
361 | return AVERROR_EOF; |
||
362 | if (left < 0) |
||
363 | return AVERROR(EIO); |
||
364 | } |
||
365 | |||
366 | pkt_frames = caf->frames_per_packet; |
||
367 | pkt_size = caf->bytes_per_packet; |
||
368 | |||
369 | if (pkt_size > 0 && pkt_frames == 1) { |
||
370 | pkt_size = (CAF_MAX_PKT_SIZE / pkt_size) * pkt_size; |
||
371 | pkt_size = FFMIN(pkt_size, left); |
||
372 | pkt_frames = pkt_size / caf->bytes_per_packet; |
||
373 | } else if (st->nb_index_entries) { |
||
374 | if (caf->packet_cnt < st->nb_index_entries - 1) { |
||
375 | pkt_size = st->index_entries[caf->packet_cnt + 1].pos - st->index_entries[caf->packet_cnt].pos; |
||
376 | pkt_frames = st->index_entries[caf->packet_cnt + 1].timestamp - st->index_entries[caf->packet_cnt].timestamp; |
||
377 | } else if (caf->packet_cnt == st->nb_index_entries - 1) { |
||
378 | pkt_size = caf->num_bytes - st->index_entries[caf->packet_cnt].pos; |
||
379 | pkt_frames = st->duration - st->index_entries[caf->packet_cnt].timestamp; |
||
380 | } else { |
||
381 | return AVERROR(EIO); |
||
382 | } |
||
383 | } |
||
384 | |||
385 | if (pkt_size == 0 || pkt_frames == 0 || pkt_size > left) |
||
386 | return AVERROR(EIO); |
||
387 | |||
388 | res = av_get_packet(pb, pkt, pkt_size); |
||
389 | if (res < 0) |
||
390 | return res; |
||
391 | |||
392 | pkt->size = res; |
||
393 | pkt->stream_index = 0; |
||
394 | pkt->dts = pkt->pts = caf->frame_cnt; |
||
395 | |||
396 | caf->packet_cnt++; |
||
397 | caf->frame_cnt += pkt_frames; |
||
398 | |||
399 | return 0; |
||
400 | } |
||
401 | |||
402 | static int read_seek(AVFormatContext *s, int stream_index, |
||
403 | int64_t timestamp, int flags) |
||
404 | { |
||
405 | AVStream *st = s->streams[0]; |
||
406 | CafContext *caf = s->priv_data; |
||
407 | int64_t pos, packet_cnt, frame_cnt; |
||
408 | |||
409 | timestamp = FFMAX(timestamp, 0); |
||
410 | |||
411 | if (caf->frames_per_packet > 0 && caf->bytes_per_packet > 0) { |
||
412 | /* calculate new byte position based on target frame position */ |
||
413 | pos = caf->bytes_per_packet * (timestamp / caf->frames_per_packet); |
||
414 | if (caf->data_size > 0) |
||
415 | pos = FFMIN(pos, caf->data_size); |
||
416 | packet_cnt = pos / caf->bytes_per_packet; |
||
417 | frame_cnt = caf->frames_per_packet * packet_cnt; |
||
418 | } else if (st->nb_index_entries) { |
||
419 | packet_cnt = av_index_search_timestamp(st, timestamp, flags); |
||
420 | frame_cnt = st->index_entries[packet_cnt].timestamp; |
||
421 | pos = st->index_entries[packet_cnt].pos; |
||
422 | } else { |
||
423 | return -1; |
||
424 | } |
||
425 | |||
426 | if (avio_seek(s->pb, pos + caf->data_start, SEEK_SET) < 0) |
||
427 | return -1; |
||
428 | |||
429 | caf->packet_cnt = packet_cnt; |
||
430 | caf->frame_cnt = frame_cnt; |
||
431 | |||
432 | return 0; |
||
433 | } |
||
434 | |||
435 | AVInputFormat ff_caf_demuxer = { |
||
436 | .name = "caf", |
||
437 | .long_name = NULL_IF_CONFIG_SMALL("Apple CAF (Core Audio Format)"), |
||
438 | .priv_data_size = sizeof(CafContext), |
||
439 | .read_probe = probe, |
||
440 | .read_header = read_header, |
||
441 | .read_packet = read_packet, |
||
442 | .read_seek = read_seek, |
||
443 | .codec_tag = (const AVCodecTag* const []){ ff_codec_caf_tags, 0 }, |
||
444 | };>>>>>>>>>>>>>>>>>>>>>>>> |