oma: Report a timestamp
[libav.git] / libavformat / omadec.c
CommitLineData
b339f0c6
BL
1/*
2 * Sony OpenMG (OMA) demuxer
3 *
23d0fdcf 4 * Copyright (c) 2008, 2013 Maxim Poliakovski
b339f0c6 5 * 2008 Benjamin Larsson
61f4387d 6 * 2011 David Goldwich
b339f0c6 7 *
2912e87a 8 * This file is part of Libav.
b339f0c6 9 *
2912e87a 10 * Libav is free software; you can redistribute it and/or
b339f0c6
BL
11 * modify it under the terms of the GNU Lesser General Public
12 * License as published by the Free Software Foundation; either
13 * version 2.1 of the License, or (at your option) any later version.
14 *
2912e87a 15 * Libav is distributed in the hope that it will be useful,
b339f0c6
BL
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * Lesser General Public License for more details.
19 *
20 * You should have received a copy of the GNU Lesser General Public
2912e87a 21 * License along with Libav; if not, write to the Free Software
b339f0c6
BL
22 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 */
24
25/**
ba87f080 26 * @file
b339f0c6
BL
27 * This is a demuxer for Sony OpenMG Music files
28 *
29 * Known file extensions: ".oma", "aa3"
30 * The format of such files consists of three parts:
6afd0ee3
MK
31 * - "ea3" header carrying overall info and metadata. Except for starting with
32 * "ea" instead of "ID", it's an ID3v2 header.
b339f0c6
BL
33 * - "EA3" header is a Sony-specific header containing information about
34 * the OpenMG file: codec type (usually ATRAC, can also be MP3 or WMA),
35 * codec specific info (packet size, sample rate, channels and so on)
36 * and DRM related info (file encryption, content id).
37 * - Sound data organized in packets follow the EA3 header
38 * (can be encrypted using the Sony DRM!).
39 *
e6f0bb65 40 * Supported decoders: ATRAC3, ATRAC3+, MP3, LPCM
b339f0c6
BL
41 */
42
d92024f1
DB
43#include <inttypes.h>
44
c9759eb4 45#include "libavutil/channel_layout.h"
b339f0c6 46#include "avformat.h"
c3f9ebf7 47#include "internal.h"
b339f0c6 48#include "libavutil/intreadwrite.h"
61f4387d 49#include "libavutil/des.h"
e352520e 50#include "libavutil/mathematics.h"
ee20b332 51#include "oma.h"
e94204df 52#include "pcm.h"
6afd0ee3 53#include "id3v2.h"
b339f0c6 54
b339f0c6 55
61f4387d
DG
56static const uint64_t leaf_table[] = {
57 0xd79e8283acea4620, 0x7a9762f445afd0d8,
58 0x354d60a60b8c79f1, 0x584e1cde00b07aee,
59 0x1573cd93da7df623, 0x47f98d79620dd535
60};
61
62typedef struct OMAContext {
63 uint64_t content_start;
64 int encrypted;
65 uint16_t k_size;
66 uint16_t e_size;
67 uint16_t i_size;
68 uint16_t s_size;
69 uint32_t rid;
70 uint8_t r_val[24];
71 uint8_t n_val[24];
72 uint8_t m_val[8];
73 uint8_t s_val[8];
74 uint8_t sm_val[8];
75 uint8_t e_val[8];
76 uint8_t iv[8];
77 struct AVDES av_des;
78} OMAContext;
79
c1401054
LB
80static void hex_log(AVFormatContext *s, int level,
81 const char *name, const uint8_t *value, int len)
61f4387d
DG
82{
83 char buf[33];
84 len = FFMIN(len, 16);
85 if (av_log_get_level() < level)
86 return;
87 ff_data_to_hex(buf, value, len, 1);
c1401054 88 buf[len << 1] = '\0';
61f4387d
DG
89 av_log(s, level, "%s: %s\n", name, buf);
90}
91
c1401054
LB
92static int kset(AVFormatContext *s, const uint8_t *r_val, const uint8_t *n_val,
93 int len)
61f4387d
DG
94{
95 OMAContext *oc = s->priv_data;
96
97 if (!r_val && !n_val)
98 return -1;
99
100 len = FFMIN(len, 16);
101
102 /* use first 64 bits in the third round again */
103 if (r_val) {
104 if (r_val != oc->r_val) {
105 memset(oc->r_val, 0, 24);
106 memcpy(oc->r_val, r_val, len);
107 }
108 memcpy(&oc->r_val[16], r_val, 8);
109 }
110 if (n_val) {
111 if (n_val != oc->n_val) {
112 memset(oc->n_val, 0, 24);
113 memcpy(oc->n_val, n_val, len);
114 }
115 memcpy(&oc->n_val[16], n_val, 8);
116 }
117
118 return 0;
119}
120
9d0b45ad
LB
121#define OMA_RPROBE_M_VAL 48 + 1
122
123static int rprobe(AVFormatContext *s, uint8_t *enc_header, unsigned size,
124 const uint8_t *r_val)
61f4387d
DG
125{
126 OMAContext *oc = s->priv_data;
127 unsigned int pos;
128 struct AVDES av_des;
129
9d0b45ad
LB
130 if (!enc_header || !r_val ||
131 size < OMA_ENC_HEADER_SIZE + oc->k_size + oc->e_size + oc->i_size ||
132 size < OMA_RPROBE_M_VAL)
61f4387d
DG
133 return -1;
134
135 /* m_val */
136 av_des_init(&av_des, r_val, 192, 1);
137 av_des_crypt(&av_des, oc->m_val, &enc_header[48], 1, NULL, 1);
138
139 /* s_val */
140 av_des_init(&av_des, oc->m_val, 64, 0);
141 av_des_crypt(&av_des, oc->s_val, NULL, 1, NULL, 0);
142
143 /* sm_val */
144 pos = OMA_ENC_HEADER_SIZE + oc->k_size + oc->e_size;
145 av_des_init(&av_des, oc->s_val, 64, 0);
146 av_des_mac(&av_des, oc->sm_val, &enc_header[pos], (oc->i_size >> 3));
147
148 pos += oc->i_size;
149
150 return memcmp(&enc_header[pos], oc->sm_val, 8) ? -1 : 0;
151}
152
9d0b45ad 153static int nprobe(AVFormatContext *s, uint8_t *enc_header, unsigned size,
c1401054 154 const uint8_t *n_val)
61f4387d
DG
155{
156 OMAContext *oc = s->priv_data;
9d0b45ad
LB
157 uint64_t pos;
158 uint32_t taglen, datalen;
61f4387d
DG
159 struct AVDES av_des;
160
9d0b45ad
LB
161 if (!enc_header || !n_val ||
162 size < OMA_ENC_HEADER_SIZE + oc->k_size + 4)
61f4387d
DG
163 return -1;
164
165 pos = OMA_ENC_HEADER_SIZE + oc->k_size;
166 if (!memcmp(&enc_header[pos], "EKB ", 4))
167 pos += 32;
168
9d0b45ad
LB
169 if (size < pos + 44)
170 return -1;
171
61f4387d
DG
172 if (AV_RB32(&enc_header[pos]) != oc->rid)
173 av_log(s, AV_LOG_DEBUG, "Mismatching RID\n");
174
c1401054
LB
175 taglen = AV_RB32(&enc_header[pos + 32]);
176 datalen = AV_RB32(&enc_header[pos + 36]) >> 4;
61f4387d 177
342c43d1
MS
178 pos += 44;
179 if (size - pos < taglen)
180 return -1;
181
182 pos += taglen;
61f4387d 183
9d0b45ad
LB
184 if (datalen << 4 > size - pos)
185 return -1;
186
61f4387d
DG
187 av_des_init(&av_des, n_val, 192, 1);
188 while (datalen-- > 0) {
189 av_des_crypt(&av_des, oc->r_val, &enc_header[pos], 2, NULL, 1);
190 kset(s, oc->r_val, NULL, 16);
9d0b45ad 191 if (!rprobe(s, enc_header, size, oc->r_val))
61f4387d
DG
192 return 0;
193 pos += 16;
194 }
195
196 return -1;
197}
198
199static int decrypt_init(AVFormatContext *s, ID3v2ExtraMeta *em, uint8_t *header)
200{
201 OMAContext *oc = s->priv_data;
202 ID3v2ExtraMetaGEOB *geob = NULL;
203 uint8_t *gdata;
204
205 oc->encrypted = 1;
206 av_log(s, AV_LOG_INFO, "File is encrypted\n");
207
208 /* find GEOB metadata */
209 while (em) {
210 if (!strcmp(em->tag, "GEOB") &&
211 (geob = em->data) &&
e9607007
DG
212 (!strcmp(geob->description, "OMG_LSI") ||
213 !strcmp(geob->description, "OMG_BKLSI"))) {
61f4387d
DG
214 break;
215 }
216 em = em->next;
217 }
218 if (!em) {
219 av_log(s, AV_LOG_ERROR, "No encryption header found\n");
6f98508e 220 return AVERROR_INVALIDDATA;
61f4387d
DG
221 }
222
223 if (geob->datasize < 64) {
c1401054 224 av_log(s, AV_LOG_ERROR,
d92024f1 225 "Invalid GEOB data size: %"PRIu32"\n", geob->datasize);
6f98508e 226 return AVERROR_INVALIDDATA;
61f4387d
DG
227 }
228
229 gdata = geob->data;
230
231 if (AV_RB16(gdata) != 1)
232 av_log(s, AV_LOG_WARNING, "Unknown version in encryption header\n");
233
234 oc->k_size = AV_RB16(&gdata[2]);
235 oc->e_size = AV_RB16(&gdata[4]);
236 oc->i_size = AV_RB16(&gdata[6]);
237 oc->s_size = AV_RB16(&gdata[8]);
238
239 if (memcmp(&gdata[OMA_ENC_HEADER_SIZE], "KEYRING ", 12)) {
240 av_log(s, AV_LOG_ERROR, "Invalid encryption header\n");
6f98508e 241 return AVERROR_INVALIDDATA;
61f4387d 242 }
1c736bed
MN
243 if (OMA_ENC_HEADER_SIZE + oc->k_size + oc->e_size + oc->i_size + 8 > geob->datasize ||
244 OMA_ENC_HEADER_SIZE + 48 > geob->datasize) {
245 av_log(s, AV_LOG_ERROR, "Too little GEOB data\n");
246 return AVERROR_INVALIDDATA;
247 }
61f4387d 248 oc->rid = AV_RB32(&gdata[OMA_ENC_HEADER_SIZE + 28]);
d92024f1 249 av_log(s, AV_LOG_DEBUG, "RID: %.8"PRIx32"\n", oc->rid);
61f4387d
DG
250
251 memcpy(oc->iv, &header[0x58], 8);
252 hex_log(s, AV_LOG_DEBUG, "IV", oc->iv, 8);
253
c1401054
LB
254 hex_log(s, AV_LOG_DEBUG, "CBC-MAC",
255 &gdata[OMA_ENC_HEADER_SIZE + oc->k_size + oc->e_size + oc->i_size],
256 8);
61f4387d
DG
257
258 if (s->keylen > 0) {
259 kset(s, s->key, s->key, s->keylen);
260 }
261 if (!memcmp(oc->r_val, (const uint8_t[8]){0}, 8) ||
9d0b45ad 262 rprobe(s, gdata, geob->datasize, oc->r_val) < 0 &&
72ec043a 263 nprobe(s, gdata, geob->datasize, oc->n_val) < 0) {
61f4387d 264 int i;
934cd18a 265 for (i = 0; i < FF_ARRAY_ELEMS(leaf_table); i += 2) {
61f4387d 266 uint8_t buf[16];
c1401054
LB
267 AV_WL64(buf, leaf_table[i]);
268 AV_WL64(&buf[8], leaf_table[i + 1]);
61f4387d 269 kset(s, buf, buf, 16);
9d0b45ad 270 if (!rprobe(s, gdata, geob->datasize, oc->r_val) ||
c1401054 271 !nprobe(s, gdata, geob->datasize, oc->n_val))
61f4387d
DG
272 break;
273 }
97f50e92 274 if (i >= FF_ARRAY_ELEMS(leaf_table)) {
61f4387d 275 av_log(s, AV_LOG_ERROR, "Invalid key\n");
6f98508e 276 return AVERROR_INVALIDDATA;
61f4387d
DG
277 }
278 }
279
280 /* e_val */
281 av_des_init(&oc->av_des, oc->m_val, 64, 0);
c1401054
LB
282 av_des_crypt(&oc->av_des, oc->e_val,
283 &gdata[OMA_ENC_HEADER_SIZE + 40], 1, NULL, 0);
61f4387d
DG
284 hex_log(s, AV_LOG_DEBUG, "EK", oc->e_val, 8);
285
286 /* init e_val */
287 av_des_init(&oc->av_des, oc->e_val, 64, 1);
288
289 return 0;
290}
6afd0ee3 291
6e9651d1 292static int oma_read_header(AVFormatContext *s)
b339f0c6 293{
6afd0ee3 294 int ret, framesize, jsflag, samplerate;
23d0fdcf 295 uint32_t codec_params, channel_id;
b339f0c6
BL
296 int16_t eid;
297 uint8_t buf[EA3_HEADER_SIZE];
298 uint8_t *edata;
299 AVStream *st;
61f4387d
DG
300 ID3v2ExtraMeta *extra_meta = NULL;
301 OMAContext *oc = s->priv_data;
b339f0c6 302
393fd0d8 303 ff_id3v2_read(s, ID3v2_EA3_MAGIC, &extra_meta);
b7effd4e 304 ret = avio_read(s->pb, buf, EA3_HEADER_SIZE);
4c1e56ed
MR
305 if (ret < EA3_HEADER_SIZE)
306 return -1;
b339f0c6 307
c1401054
LB
308 if (memcmp(buf, ((const uint8_t[]){'E', 'A', '3'}), 3) ||
309 buf[4] != 0 || buf[5] != EA3_HEADER_SIZE) {
b339f0c6 310 av_log(s, AV_LOG_ERROR, "Couldn't find the EA3 header !\n");
6f98508e 311 return AVERROR_INVALIDDATA;
b339f0c6
BL
312 }
313
61f4387d
DG
314 oc->content_start = avio_tell(s->pb);
315
316 /* encrypted file */
b339f0c6 317 eid = AV_RB16(&buf[6]);
61f4387d
DG
318 if (eid != -1 && eid != -128 && decrypt_init(s, extra_meta, buf) < 0) {
319 ff_id3v2_free_extra_meta(&extra_meta);
b339f0c6
BL
320 return -1;
321 }
322
61f4387d
DG
323 ff_id3v2_free_extra_meta(&extra_meta);
324
b339f0c6 325 codec_params = AV_RB24(&buf[33]);
b339f0c6 326
3b3bbdd3 327 st = avformat_new_stream(s, NULL);
b339f0c6
BL
328 if (!st)
329 return AVERROR(ENOMEM);
330
de8c8181 331 st->start_time = 0;
c1401054
LB
332 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
333 st->codec->codec_tag = buf[32];
334 st->codec->codec_id = ff_codec_get_id(ff_oma_codec_tags,
335 st->codec->codec_tag);
de8c8181 336
b339f0c6 337 switch (buf[32]) {
c1401054
LB
338 case OMA_CODECID_ATRAC3:
339 samplerate = ff_oma_srate_tab[(codec_params >> 13) & 7] * 100;
340 if (!samplerate) {
341 av_log(s, AV_LOG_ERROR, "Unsupported sample rate\n");
342 return AVERROR_INVALIDDATA;
343 }
344 if (samplerate != 44100)
345 avpriv_request_sample(s, "Sample rate %d", samplerate);
346
347 framesize = (codec_params & 0x3FF) * 8;
348
349 /* get stereo coding mode, 1 for joint-stereo */
350 jsflag = (codec_params >> 17) & 1;
351
352 st->codec->channels = 2;
353 st->codec->channel_layout = AV_CH_LAYOUT_STEREO;
354 st->codec->sample_rate = samplerate;
355 st->codec->bit_rate = st->codec->sample_rate * framesize * 8 / 1024;
356
7df9e693 357 /* fake the ATRAC3 extradata
c1401054
LB
358 * (wav format, makes stream copy to wav work) */
359 st->codec->extradata_size = 14;
360 edata = av_mallocz(14 + FF_INPUT_BUFFER_PADDING_SIZE);
361 if (!edata)
362 return AVERROR(ENOMEM);
363
364 st->codec->extradata = edata;
365 AV_WL16(&edata[0], 1); // always 1
366 AV_WL32(&edata[2], samplerate); // samples rate
367 AV_WL16(&edata[6], jsflag); // coding mode
368 AV_WL16(&edata[8], jsflag); // coding mode
369 AV_WL16(&edata[10], 1); // always 1
370 // AV_WL16(&edata[12], 0); // always 0
371
372 avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
373 break;
374 case OMA_CODECID_ATRAC3P:
23d0fdcf
MP
375 channel_id = (codec_params >> 10) & 7;
376 if (!channel_id) {
377 av_log(s, AV_LOG_ERROR,
d92024f1 378 "Invalid ATRAC-X channel id: %"PRIu32"\n", channel_id);
23d0fdcf
MP
379 return AVERROR_INVALIDDATA;
380 }
381 st->codec->channel_layout = ff_oma_chid_to_native_layout[channel_id - 1];
382 st->codec->channels = ff_oma_chid_to_num_channels[channel_id - 1];
c1401054
LB
383 framesize = ((codec_params & 0x3FF) * 8) + 8;
384 samplerate = ff_oma_srate_tab[(codec_params >> 13) & 7] * 100;
385 if (!samplerate) {
386 av_log(s, AV_LOG_ERROR, "Unsupported sample rate\n");
387 return AVERROR_INVALIDDATA;
388 }
389 st->codec->sample_rate = samplerate;
487b5410 390 st->codec->bit_rate = samplerate * framesize * 8 / 2048;
c1401054 391 avpriv_set_pts_info(st, 64, 1, samplerate);
c1401054
LB
392 break;
393 case OMA_CODECID_MP3:
394 st->need_parsing = AVSTREAM_PARSE_FULL;
395 framesize = 1024;
396 break;
397 case OMA_CODECID_LPCM:
398 /* PCM 44.1 kHz 16 bit stereo big-endian */
399 st->codec->channels = 2;
400 st->codec->channel_layout = AV_CH_LAYOUT_STEREO;
401 st->codec->sample_rate = 44100;
402 framesize = 1024;
403 /* bit rate = sample rate x PCM block align (= 4) x 8 */
404 st->codec->bit_rate = st->codec->sample_rate * 32;
405 st->codec->bits_per_coded_sample =
406 av_get_bits_per_sample(st->codec->codec_id);
407 avpriv_set_pts_info(st, 64, 1, st->codec->sample_rate);
408 break;
409 default:
410 av_log(s, AV_LOG_ERROR, "Unsupported codec %d!\n", buf[32]);
6f98508e 411 return AVERROR(ENOSYS);
b339f0c6
BL
412 }
413
b339f0c6 414 st->codec->block_align = framesize;
b339f0c6
BL
415
416 return 0;
417}
418
419
420static int oma_read_packet(AVFormatContext *s, AVPacket *pkt)
421{
e352520e
LB
422 OMAContext *oc = s->priv_data;
423 AVStream *st = s->streams[0];
424 int packet_size = st->codec->block_align;
425 int byte_rate = st->codec->bit_rate >> 3;
426 int64_t pos = avio_tell(s->pb);
427 int ret = av_get_packet(s->pb, pkt, packet_size);
b339f0c6 428
2219e27b
LB
429 if (ret < packet_size)
430 pkt->flags |= AV_PKT_FLAG_CORRUPT;
431
db9aee6c
LB
432 if (ret < 0)
433 return ret;
434 if (!ret)
435 return AVERROR_EOF;
b339f0c6 436
61f4387d
DG
437 pkt->stream_index = 0;
438
e352520e
LB
439 if (pos > 0) {
440 pkt->pts =
441 pkt->dts = av_rescale(pos, st->time_base.den,
442 byte_rate * (int64_t)st->time_base.num);
443 }
444
61f4387d 445 if (oc->encrypted) {
c1401054
LB
446 /* previous unencrypted block saved in IV for
447 * the next packet (CBC mode) */
2219e27b
LB
448 if (ret == packet_size)
449 av_des_crypt(&oc->av_des, pkt->data, pkt->data,
450 (packet_size >> 3), oc->iv, 1);
451 else
452 memset(oc->iv, 0, 8);
61f4387d
DG
453 }
454
b339f0c6
BL
455 return ret;
456}
457
458static int oma_read_probe(AVProbeData *p)
459{
0a7fef39 460 const uint8_t *buf = p->buf;
6afd0ee3
MK
461 unsigned tag_len = 0;
462
0a7fef39
DG
463 if (p->buf_size >= ID3v2_HEADER_SIZE && ff_id3v2_match(buf, ID3v2_EA3_MAGIC))
464 tag_len = ff_id3v2_tag_len(buf);
8ae5eb75
DG
465
466 /* This check cannot overflow as tag_len has at most 28 bits */
467 if (p->buf_size < tag_len + 5)
468 /* EA3 header comes late, might be outside of the probe buffer */
0a7fef39 469 return tag_len ? AVPROBE_SCORE_EXTENSION : 0;
8ae5eb75 470
6afd0ee3
MK
471 buf += tag_len;
472
473 if (!memcmp(buf, "EA3", 3) && !buf[4] && buf[5] == EA3_HEADER_SIZE)
b339f0c6
BL
474 return AVPROBE_SCORE_MAX;
475 else
476 return 0;
477}
478
c1401054
LB
479static int oma_read_seek(struct AVFormatContext *s,
480 int stream_index, int64_t timestamp, int flags)
61f4387d
DG
481{
482 OMAContext *oc = s->priv_data;
4f03a77e
LB
483 int err = ff_pcm_read_seek(s, stream_index, timestamp, flags);
484
485 if (!oc->encrypted)
486 return err;
487
488 /* readjust IV for CBC */
489 if (err || avio_tell(s->pb) < oc->content_start)
490 goto wipe;
491 if ((err = avio_seek(s->pb, -8, SEEK_CUR)) < 0)
492 goto wipe;
493 if ((err = avio_read(s->pb, oc->iv, 8)) < 8) {
494 if (err >= 0)
495 err = AVERROR_EOF;
496 goto wipe;
61f4387d
DG
497 }
498
499 return 0;
4f03a77e
LB
500wipe:
501 memset(oc->iv, 0, 8);
502 return err;
61f4387d 503}
b339f0c6 504
c6610a21 505AVInputFormat ff_oma_demuxer = {
dfc2c4d9
AK
506 .name = "oma",
507 .long_name = NULL_IF_CONFIG_SMALL("Sony OpenMG audio"),
61f4387d 508 .priv_data_size = sizeof(OMAContext),
dfc2c4d9
AK
509 .read_probe = oma_read_probe,
510 .read_header = oma_read_header,
511 .read_packet = oma_read_packet,
61f4387d
DG
512 .read_seek = oma_read_seek,
513 .flags = AVFMT_GENERIC_INDEX,
514 .extensions = "oma,omg,aa3",
ee20b332 515 .codec_tag = (const AVCodecTag* const []){ff_oma_codec_tags, 0},
b339f0c6 516};