| 1 | /* |
| 2 | * MP3 demuxer |
| 3 | * Copyright (c) 2003 Fabrice Bellard |
| 4 | * |
| 5 | * This file is part of FFmpeg. |
| 6 | * |
| 7 | * FFmpeg is free software; you can redistribute it and/or |
| 8 | * modify it under the terms of the GNU Lesser General Public |
| 9 | * License as published by the Free Software Foundation; either |
| 10 | * version 2.1 of the License, or (at your option) any later version. |
| 11 | * |
| 12 | * FFmpeg is distributed in the hope that it will be useful, |
| 13 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 14 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 15 | * Lesser General Public License for more details. |
| 16 | * |
| 17 | * You should have received a copy of the GNU Lesser General Public |
| 18 | * License along with FFmpeg; if not, write to the Free Software |
| 19 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA |
| 20 | */ |
| 21 | |
| 22 | #include "libavutil/opt.h" |
| 23 | #include "libavutil/avstring.h" |
| 24 | #include "libavutil/intreadwrite.h" |
| 25 | #include "libavutil/crc.h" |
| 26 | #include "libavutil/dict.h" |
| 27 | #include "libavutil/mathematics.h" |
| 28 | #include "avformat.h" |
| 29 | #include "internal.h" |
| 30 | #include "avio_internal.h" |
| 31 | #include "id3v2.h" |
| 32 | #include "id3v1.h" |
| 33 | #include "replaygain.h" |
| 34 | |
| 35 | #include "libavcodec/mpegaudiodecheader.h" |
| 36 | |
| 37 | #define XING_FLAG_FRAMES 0x01 |
| 38 | #define XING_FLAG_SIZE 0x02 |
| 39 | #define XING_FLAG_TOC 0x04 |
| 40 | #define XING_FLAC_QSCALE 0x08 |
| 41 | |
| 42 | #define XING_TOC_COUNT 100 |
| 43 | |
| 44 | typedef struct { |
| 45 | AVClass *class; |
| 46 | int64_t filesize; |
| 47 | int xing_toc; |
| 48 | int start_pad; |
| 49 | int end_pad; |
| 50 | int usetoc; |
| 51 | unsigned frames; /* Total number of frames in file */ |
| 52 | unsigned header_filesize; /* Total number of bytes in the stream */ |
| 53 | int is_cbr; |
| 54 | } MP3DecContext; |
| 55 | |
| 56 | /* mp3 read */ |
| 57 | |
| 58 | static int mp3_read_probe(AVProbeData *p) |
| 59 | { |
| 60 | int max_frames, first_frames = 0; |
| 61 | int fsize, frames, sample_rate; |
| 62 | uint32_t header; |
| 63 | const uint8_t *buf, *buf0, *buf2, *end; |
| 64 | AVCodecContext avctx; |
| 65 | |
| 66 | buf0 = p->buf; |
| 67 | end = p->buf + p->buf_size - sizeof(uint32_t); |
| 68 | while(buf0 < end && !*buf0) |
| 69 | buf0++; |
| 70 | |
| 71 | max_frames = 0; |
| 72 | buf = buf0; |
| 73 | |
| 74 | for(; buf < end; buf= buf2+1) { |
| 75 | buf2 = buf; |
| 76 | if(ff_mpa_check_header(AV_RB32(buf2))) |
| 77 | continue; |
| 78 | |
| 79 | for(frames = 0; buf2 < end; frames++) { |
| 80 | header = AV_RB32(buf2); |
| 81 | fsize = avpriv_mpa_decode_header(&avctx, header, &sample_rate, &sample_rate, &sample_rate, &sample_rate); |
| 82 | if(fsize < 0) |
| 83 | break; |
| 84 | buf2 += fsize; |
| 85 | } |
| 86 | max_frames = FFMAX(max_frames, frames); |
| 87 | if(buf == buf0) |
| 88 | first_frames= frames; |
| 89 | } |
| 90 | // keep this in sync with ac3 probe, both need to avoid |
| 91 | // issues with MPEG-files! |
| 92 | if (first_frames>=4) return AVPROBE_SCORE_EXTENSION + 1; |
| 93 | else if(max_frames>200)return AVPROBE_SCORE_EXTENSION; |
| 94 | else if(max_frames>=4 && max_frames >= p->buf_size/10000) return AVPROBE_SCORE_EXTENSION / 2; |
| 95 | else if(ff_id3v2_match(buf0, ID3v2_DEFAULT_MAGIC) && 2*ff_id3v2_tag_len(buf0) >= p->buf_size) |
| 96 | return p->buf_size < PROBE_BUF_MAX ? AVPROBE_SCORE_EXTENSION / 4 : AVPROBE_SCORE_EXTENSION - 2; |
| 97 | else if(max_frames>=1 && max_frames >= p->buf_size/10000) return 1; |
| 98 | else return 0; |
| 99 | //mpegps_mp3_unrecognized_format.mpg has max_frames=3 |
| 100 | } |
| 101 | |
| 102 | static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration) |
| 103 | { |
| 104 | int i; |
| 105 | MP3DecContext *mp3 = s->priv_data; |
| 106 | int fill_index = mp3->usetoc && duration > 0; |
| 107 | |
| 108 | if (!filesize && |
| 109 | !(filesize = avio_size(s->pb))) { |
| 110 | av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n"); |
| 111 | fill_index = 0; |
| 112 | } |
| 113 | |
| 114 | for (i = 0; i < XING_TOC_COUNT; i++) { |
| 115 | uint8_t b = avio_r8(s->pb); |
| 116 | if (fill_index) |
| 117 | av_add_index_entry(s->streams[0], |
| 118 | av_rescale(b, filesize, 256), |
| 119 | av_rescale(i, duration, XING_TOC_COUNT), |
| 120 | 0, 0, AVINDEX_KEYFRAME); |
| 121 | } |
| 122 | if (fill_index) |
| 123 | mp3->xing_toc = 1; |
| 124 | } |
| 125 | |
| 126 | static void mp3_parse_info_tag(AVFormatContext *s, AVStream *st, |
| 127 | MPADecodeHeader *c, uint32_t spf) |
| 128 | { |
| 129 | #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1)) |
| 130 | #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m))) |
| 131 | |
| 132 | uint16_t crc; |
| 133 | uint32_t v; |
| 134 | |
| 135 | char version[10]; |
| 136 | |
| 137 | uint32_t peak = 0; |
| 138 | int32_t r_gain = INT32_MIN, a_gain = INT32_MIN; |
| 139 | |
| 140 | MP3DecContext *mp3 = s->priv_data; |
| 141 | static const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}}; |
| 142 | uint64_t fsize = avio_size(s->pb); |
| 143 | |
| 144 | /* Check for Xing / Info tag */ |
| 145 | avio_skip(s->pb, xing_offtbl[c->lsf == 1][c->nb_channels == 1]); |
| 146 | v = avio_rb32(s->pb); |
| 147 | mp3->is_cbr = v == MKBETAG('I', 'n', 'f', 'o'); |
| 148 | if (v != MKBETAG('X', 'i', 'n', 'g') && !mp3->is_cbr) |
| 149 | return; |
| 150 | |
| 151 | v = avio_rb32(s->pb); |
| 152 | if (v & XING_FLAG_FRAMES) |
| 153 | mp3->frames = avio_rb32(s->pb); |
| 154 | if (v & XING_FLAG_SIZE) |
| 155 | mp3->header_filesize = avio_rb32(s->pb); |
| 156 | if (fsize && mp3->header_filesize) { |
| 157 | uint64_t min, delta; |
| 158 | min = FFMIN(fsize, mp3->header_filesize); |
| 159 | delta = FFMAX(fsize, mp3->header_filesize) - min; |
| 160 | if (fsize > mp3->header_filesize && delta > min >> 4) { |
| 161 | mp3->frames = 0; |
| 162 | } else if (delta > min >> 4) { |
| 163 | av_log(s, AV_LOG_WARNING, |
| 164 | "filesize and duration do not match (growing file?)\n"); |
| 165 | } |
| 166 | } |
| 167 | if (v & XING_FLAG_TOC) |
| 168 | read_xing_toc(s, mp3->header_filesize, av_rescale_q(mp3->frames, |
| 169 | (AVRational){spf, c->sample_rate}, |
| 170 | st->time_base)); |
| 171 | /* VBR quality */ |
| 172 | if (v & XING_FLAC_QSCALE) |
| 173 | avio_rb32(s->pb); |
| 174 | |
| 175 | /* Encoder short version string */ |
| 176 | memset(version, 0, sizeof(version)); |
| 177 | avio_read(s->pb, version, 9); |
| 178 | |
| 179 | /* Info Tag revision + VBR method */ |
| 180 | avio_r8(s->pb); |
| 181 | |
| 182 | /* Lowpass filter value */ |
| 183 | avio_r8(s->pb); |
| 184 | |
| 185 | /* ReplayGain peak */ |
| 186 | v = avio_rb32(s->pb); |
| 187 | peak = av_rescale(v, 100000, 1 << 23); |
| 188 | |
| 189 | /* Radio ReplayGain */ |
| 190 | v = avio_rb16(s->pb); |
| 191 | |
| 192 | if (MIDDLE_BITS(v, 13, 15) == 1) { |
| 193 | r_gain = MIDDLE_BITS(v, 0, 8) * 10000; |
| 194 | |
| 195 | if (v & (1 << 9)) |
| 196 | r_gain *= -1; |
| 197 | } |
| 198 | |
| 199 | /* Audiophile ReplayGain */ |
| 200 | v = avio_rb16(s->pb); |
| 201 | |
| 202 | if (MIDDLE_BITS(v, 13, 15) == 2) { |
| 203 | a_gain = MIDDLE_BITS(v, 0, 8) * 10000; |
| 204 | |
| 205 | if (v & (1 << 9)) |
| 206 | a_gain *= -1; |
| 207 | } |
| 208 | |
| 209 | /* Encoding flags + ATH Type */ |
| 210 | avio_r8(s->pb); |
| 211 | |
| 212 | /* if ABR {specified bitrate} else {minimal bitrate} */ |
| 213 | avio_r8(s->pb); |
| 214 | |
| 215 | /* Encoder delays */ |
| 216 | v= avio_rb24(s->pb); |
| 217 | if(AV_RB32(version) == MKBETAG('L', 'A', 'M', 'E') |
| 218 | || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'f') |
| 219 | || AV_RB32(version) == MKBETAG('L', 'a', 'v', 'c') |
| 220 | ) { |
| 221 | |
| 222 | mp3->start_pad = v>>12; |
| 223 | mp3-> end_pad = v&4095; |
| 224 | st->skip_samples = mp3->start_pad + 528 + 1; |
| 225 | if (mp3->frames) { |
| 226 | st->first_discard_sample = -mp3->end_pad + 528 + 1 + mp3->frames * (int64_t)spf; |
| 227 | st->last_discard_sample = mp3->frames * (int64_t)spf; |
| 228 | } |
| 229 | if (!st->start_time) |
| 230 | st->start_time = av_rescale_q(st->skip_samples, |
| 231 | (AVRational){1, c->sample_rate}, |
| 232 | st->time_base); |
| 233 | av_log(s, AV_LOG_DEBUG, "pad %d %d\n", mp3->start_pad, mp3-> end_pad); |
| 234 | } |
| 235 | |
| 236 | /* Misc */ |
| 237 | avio_r8(s->pb); |
| 238 | |
| 239 | /* MP3 gain */ |
| 240 | avio_r8(s->pb); |
| 241 | |
| 242 | /* Preset and surround info */ |
| 243 | avio_rb16(s->pb); |
| 244 | |
| 245 | /* Music length */ |
| 246 | avio_rb32(s->pb); |
| 247 | |
| 248 | /* Music CRC */ |
| 249 | avio_rb16(s->pb); |
| 250 | |
| 251 | /* Info Tag CRC */ |
| 252 | crc = ffio_get_checksum(s->pb); |
| 253 | v = avio_rb16(s->pb); |
| 254 | |
| 255 | if (v == crc) { |
| 256 | ff_replaygain_export_raw(st, r_gain, peak, a_gain, 0); |
| 257 | av_dict_set(&st->metadata, "encoder", version, 0); |
| 258 | } |
| 259 | } |
| 260 | |
| 261 | static void mp3_parse_vbri_tag(AVFormatContext *s, AVStream *st, int64_t base) |
| 262 | { |
| 263 | uint32_t v; |
| 264 | MP3DecContext *mp3 = s->priv_data; |
| 265 | |
| 266 | /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */ |
| 267 | avio_seek(s->pb, base + 4 + 32, SEEK_SET); |
| 268 | v = avio_rb32(s->pb); |
| 269 | if (v == MKBETAG('V', 'B', 'R', 'I')) { |
| 270 | /* Check tag version */ |
| 271 | if (avio_rb16(s->pb) == 1) { |
| 272 | /* skip delay and quality */ |
| 273 | avio_skip(s->pb, 4); |
| 274 | mp3->header_filesize = avio_rb32(s->pb); |
| 275 | mp3->frames = avio_rb32(s->pb); |
| 276 | } |
| 277 | } |
| 278 | } |
| 279 | |
| 280 | /** |
| 281 | * Try to find Xing/Info/VBRI tags and compute duration from info therein |
| 282 | */ |
| 283 | static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base) |
| 284 | { |
| 285 | uint32_t v, spf; |
| 286 | MPADecodeHeader c; |
| 287 | int vbrtag_size = 0; |
| 288 | MP3DecContext *mp3 = s->priv_data; |
| 289 | |
| 290 | ffio_init_checksum(s->pb, ff_crcA001_update, 0); |
| 291 | |
| 292 | v = avio_rb32(s->pb); |
| 293 | if(ff_mpa_check_header(v) < 0) |
| 294 | return -1; |
| 295 | |
| 296 | if (avpriv_mpegaudio_decode_header(&c, v) == 0) |
| 297 | vbrtag_size = c.frame_size; |
| 298 | if(c.layer != 3) |
| 299 | return -1; |
| 300 | |
| 301 | spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */ |
| 302 | |
| 303 | mp3->frames = 0; |
| 304 | mp3->header_filesize = 0; |
| 305 | |
| 306 | mp3_parse_info_tag(s, st, &c, spf); |
| 307 | mp3_parse_vbri_tag(s, st, base); |
| 308 | |
| 309 | if (!mp3->frames && !mp3->header_filesize) |
| 310 | return -1; |
| 311 | |
| 312 | /* Skip the vbr tag frame */ |
| 313 | avio_seek(s->pb, base + vbrtag_size, SEEK_SET); |
| 314 | |
| 315 | if (mp3->frames) |
| 316 | st->duration = av_rescale_q(mp3->frames, (AVRational){spf, c.sample_rate}, |
| 317 | st->time_base); |
| 318 | if (mp3->header_filesize && mp3->frames && !mp3->is_cbr) |
| 319 | st->codec->bit_rate = av_rescale(mp3->header_filesize, 8 * c.sample_rate, mp3->frames * (int64_t)spf); |
| 320 | |
| 321 | return 0; |
| 322 | } |
| 323 | |
| 324 | static int mp3_read_header(AVFormatContext *s) |
| 325 | { |
| 326 | MP3DecContext *mp3 = s->priv_data; |
| 327 | AVStream *st; |
| 328 | int64_t off; |
| 329 | int ret; |
| 330 | |
| 331 | st = avformat_new_stream(s, NULL); |
| 332 | if (!st) |
| 333 | return AVERROR(ENOMEM); |
| 334 | |
| 335 | st->codec->codec_type = AVMEDIA_TYPE_AUDIO; |
| 336 | st->codec->codec_id = AV_CODEC_ID_MP3; |
| 337 | st->need_parsing = AVSTREAM_PARSE_FULL_RAW; |
| 338 | st->start_time = 0; |
| 339 | |
| 340 | // lcm of all mp3 sample rates |
| 341 | avpriv_set_pts_info(st, 64, 1, 14112000); |
| 342 | |
| 343 | s->pb->maxsize = -1; |
| 344 | off = avio_tell(s->pb); |
| 345 | |
| 346 | if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX)) |
| 347 | ff_id3v1_read(s); |
| 348 | |
| 349 | if(s->pb->seekable) |
| 350 | mp3->filesize = avio_size(s->pb); |
| 351 | |
| 352 | if (mp3_parse_vbr_tags(s, st, off) < 0) |
| 353 | avio_seek(s->pb, off, SEEK_SET); |
| 354 | |
| 355 | ret = ff_replaygain_export(st, s->metadata); |
| 356 | if (ret < 0) |
| 357 | return ret; |
| 358 | |
| 359 | /* the parameters will be extracted from the compressed bitstream */ |
| 360 | return 0; |
| 361 | } |
| 362 | |
| 363 | #define MP3_PACKET_SIZE 1024 |
| 364 | |
| 365 | static int mp3_read_packet(AVFormatContext *s, AVPacket *pkt) |
| 366 | { |
| 367 | MP3DecContext *mp3 = s->priv_data; |
| 368 | int ret, size; |
| 369 | int64_t pos; |
| 370 | |
| 371 | size= MP3_PACKET_SIZE; |
| 372 | pos = avio_tell(s->pb); |
| 373 | if(mp3->filesize > ID3v1_TAG_SIZE && pos < mp3->filesize) |
| 374 | size= FFMIN(size, mp3->filesize - pos); |
| 375 | |
| 376 | ret= av_get_packet(s->pb, pkt, size); |
| 377 | if (ret <= 0) { |
| 378 | if(ret<0) |
| 379 | return ret; |
| 380 | return AVERROR_EOF; |
| 381 | } |
| 382 | |
| 383 | pkt->flags &= ~AV_PKT_FLAG_CORRUPT; |
| 384 | pkt->stream_index = 0; |
| 385 | |
| 386 | if (ret >= ID3v1_TAG_SIZE && |
| 387 | memcmp(&pkt->data[ret - ID3v1_TAG_SIZE], "TAG", 3) == 0) |
| 388 | ret -= ID3v1_TAG_SIZE; |
| 389 | |
| 390 | /* note: we need to modify the packet size here to handle the last |
| 391 | packet */ |
| 392 | pkt->size = ret; |
| 393 | return ret; |
| 394 | } |
| 395 | |
| 396 | static int check(AVFormatContext *s, int64_t pos) |
| 397 | { |
| 398 | int64_t ret = avio_seek(s->pb, pos, SEEK_SET); |
| 399 | unsigned header; |
| 400 | MPADecodeHeader sd; |
| 401 | if (ret < 0) |
| 402 | return ret; |
| 403 | header = avio_rb32(s->pb); |
| 404 | if (ff_mpa_check_header(header) < 0) |
| 405 | return -1; |
| 406 | if (avpriv_mpegaudio_decode_header(&sd, header) == 1) |
| 407 | return -1; |
| 408 | return sd.frame_size; |
| 409 | } |
| 410 | |
| 411 | static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp, |
| 412 | int flags) |
| 413 | { |
| 414 | MP3DecContext *mp3 = s->priv_data; |
| 415 | AVIndexEntry *ie, ie1; |
| 416 | AVStream *st = s->streams[0]; |
| 417 | int64_t ret = av_index_search_timestamp(st, timestamp, flags); |
| 418 | int i, j; |
| 419 | int dir = (flags&AVSEEK_FLAG_BACKWARD) ? -1 : 1; |
| 420 | int64_t best_pos; |
| 421 | int best_score; |
| 422 | |
| 423 | if ( mp3->is_cbr |
| 424 | && st->duration > 0 |
| 425 | && mp3->header_filesize > s->data_offset |
| 426 | && mp3->frames) { |
| 427 | int64_t filesize = avio_size(s->pb); |
| 428 | int64_t duration; |
| 429 | if (filesize <= s->data_offset) |
| 430 | filesize = mp3->header_filesize; |
| 431 | filesize -= s->data_offset; |
| 432 | duration = av_rescale(st->duration, filesize, mp3->header_filesize - s->data_offset); |
| 433 | ie = &ie1; |
| 434 | timestamp = av_clip64(timestamp, 0, duration); |
| 435 | ie->timestamp = timestamp; |
| 436 | ie->pos = av_rescale(timestamp, filesize, duration) + s->data_offset; |
| 437 | } else if (mp3->xing_toc) { |
| 438 | if (ret < 0) |
| 439 | return ret; |
| 440 | |
| 441 | ie = &st->index_entries[ret]; |
| 442 | } else { |
| 443 | st->skip_samples = timestamp <= 0 ? mp3->start_pad + 528 + 1 : 0; |
| 444 | |
| 445 | return -1; |
| 446 | } |
| 447 | |
| 448 | avio_seek(s->pb, FFMAX(ie->pos - 4096, 0), SEEK_SET); |
| 449 | ret = avio_seek(s->pb, ie->pos, SEEK_SET); |
| 450 | if (ret < 0) |
| 451 | return ret; |
| 452 | |
| 453 | #define MIN_VALID 3 |
| 454 | best_pos = ie->pos; |
| 455 | best_score = 999; |
| 456 | for(i=0; i<4096; i++) { |
| 457 | int64_t pos = ie->pos + (dir > 0 ? i - 1024 : -i); |
| 458 | int64_t candidate = -1; |
| 459 | int score = 999; |
| 460 | |
| 461 | if (pos < 0) |
| 462 | continue; |
| 463 | |
| 464 | for(j=0; j<MIN_VALID; j++) { |
| 465 | ret = check(s, pos); |
| 466 | if(ret < 0) |
| 467 | break; |
| 468 | if ((ie->pos - pos)*dir <= 0 && abs(MIN_VALID/2-j) < score) { |
| 469 | candidate = pos; |
| 470 | score = abs(MIN_VALID/2-j); |
| 471 | } |
| 472 | pos += ret; |
| 473 | } |
| 474 | if (best_score > score && j == MIN_VALID) { |
| 475 | best_pos = candidate; |
| 476 | best_score = score; |
| 477 | if(score == 0) |
| 478 | break; |
| 479 | } |
| 480 | } |
| 481 | |
| 482 | ret = avio_seek(s->pb, best_pos, SEEK_SET); |
| 483 | if (ret < 0) |
| 484 | return ret; |
| 485 | ff_update_cur_dts(s, st, ie->timestamp); |
| 486 | st->skip_samples = ie->timestamp <= 0 ? mp3->start_pad + 528 + 1 : 0; |
| 487 | return 0; |
| 488 | } |
| 489 | |
| 490 | static const AVOption options[] = { |
| 491 | { "usetoc", "use table of contents", offsetof(MP3DecContext, usetoc), AV_OPT_TYPE_INT, {.i64 = -1}, -1, 1, AV_OPT_FLAG_DECODING_PARAM}, |
| 492 | { NULL }, |
| 493 | }; |
| 494 | |
| 495 | static const AVClass demuxer_class = { |
| 496 | .class_name = "mp3", |
| 497 | .item_name = av_default_item_name, |
| 498 | .option = options, |
| 499 | .version = LIBAVUTIL_VERSION_INT, |
| 500 | .category = AV_CLASS_CATEGORY_DEMUXER, |
| 501 | }; |
| 502 | |
| 503 | AVInputFormat ff_mp3_demuxer = { |
| 504 | .name = "mp3", |
| 505 | .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"), |
| 506 | .read_probe = mp3_read_probe, |
| 507 | .read_header = mp3_read_header, |
| 508 | .read_packet = mp3_read_packet, |
| 509 | .read_seek = mp3_seek, |
| 510 | .priv_data_size = sizeof(MP3DecContext), |
| 511 | .flags = AVFMT_GENERIC_INDEX, |
| 512 | .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */ |
| 513 | .priv_class = &demuxer_class, |
| 514 | }; |