/libavformat/oggparsevorbis.c
C | 505 lines | 372 code | 76 blank | 57 comment | 112 complexity | 9dc9306ac26ba4700d8c5cca70bf3be4 MD5 | raw file
1/* 2 * Copyright (C) 2005 Michael Ahlberg, Måns Rullgård 3 * 4 * Permission is hereby granted, free of charge, to any person 5 * obtaining a copy of this software and associated documentation 6 * files (the "Software"), to deal in the Software without 7 * restriction, including without limitation the rights to use, copy, 8 * modify, merge, publish, distribute, sublicense, and/or sell copies 9 * of the Software, and to permit persons to whom the Software is 10 * furnished to do so, subject to the following conditions: 11 * 12 * The above copyright notice and this permission notice shall be 13 * included in all copies or substantial portions of the Software. 14 * 15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, 16 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF 17 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND 18 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT 19 * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, 20 * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, 21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER 22 * DEALINGS IN THE SOFTWARE. 23 */ 24 25#include <stdlib.h> 26 27#include "libavutil/avstring.h" 28#include "libavutil/base64.h" 29#include "libavutil/bswap.h" 30#include "libavutil/dict.h" 31 32#include "libavcodec/bytestream.h" 33#include "libavcodec/vorbis_parser.h" 34 35#include "avformat.h" 36#include "flac_picture.h" 37#include "internal.h" 38#include "oggdec.h" 39#include "vorbiscomment.h" 40#include "replaygain.h" 41 42static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val) 43{ 44 int i, cnum, h, m, s, ms, keylen = strlen(key); 45 AVChapter *chapter = NULL; 46 47 if (keylen < 9 || av_strncasecmp(key, "CHAPTER", 7) || sscanf(key+7, "%03d", &cnum) != 1) 48 return 0; 49 50 if (keylen <= 10) { 51 if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4) 52 return 0; 53 54 avpriv_new_chapter(as, cnum, (AVRational) { 1, 1000 }, 55 ms + 1000 * (s + 60 * (m + 60 * h)), 56 AV_NOPTS_VALUE, NULL); 57 av_free(val); 58 } else if (!av_strcasecmp(key + keylen - 4, "NAME")) { 59 for (i = 0; i < as->nb_chapters; i++) 60 if (as->chapters[i]->id == cnum) { 61 chapter = as->chapters[i]; 62 break; 63 } 64 if (!chapter) 65 return 0; 66 67 av_dict_set(&chapter->metadata, "title", val, AV_DICT_DONT_STRDUP_VAL); 68 } else 69 return 0; 70 71 av_free(key); 72 return 1; 73} 74 75int ff_vorbis_stream_comment(AVFormatContext *as, AVStream *st, 76 const uint8_t *buf, int size) 77{ 78 int updates = ff_vorbis_comment(as, &st->metadata, buf, size, 1); 79 80 if (updates > 0) { 81 st->event_flags |= AVSTREAM_EVENT_FLAG_METADATA_UPDATED; 82 } 83 84 return updates; 85} 86 87int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m, 88 const uint8_t *buf, int size, 89 int parse_picture) 90{ 91 const uint8_t *p = buf; 92 const uint8_t *end = buf + size; 93 int updates = 0; 94 unsigned n; 95 int s; 96 97 /* must have vendor_length and user_comment_list_length */ 98 if (size < 8) 99 return AVERROR_INVALIDDATA; 100 101 s = bytestream_get_le32(&p); 102 103 if (end - p - 4 < s || s < 0) 104 return AVERROR_INVALIDDATA; 105 106 p += s; 107 108 n = bytestream_get_le32(&p); 109 110 while (end - p >= 4 && n > 0) { 111 const char *t, *v; 112 int tl, vl; 113 114 s = bytestream_get_le32(&p); 115 116 if (end - p < s || s < 0) 117 break; 118 119 t = p; 120 p += s; 121 n--; 122 123 v = memchr(t, '=', s); 124 if (!v) 125 continue; 126 127 tl = v - t; 128 vl = s - tl - 1; 129 v++; 130 131 if (tl && vl) { 132 char *tt, *ct; 133 134 tt = av_malloc(tl + 1); 135 ct = av_malloc(vl + 1); 136 if (!tt || !ct) { 137 av_freep(&tt); 138 av_freep(&ct); 139 return AVERROR(ENOMEM); 140 } 141 142 memcpy(tt, t, tl); 143 tt[tl] = 0; 144 145 memcpy(ct, v, vl); 146 ct[vl] = 0; 147 148 /* The format in which the pictures are stored is the FLAC format. 149 * Xiph says: "The binary FLAC picture structure is base64 encoded 150 * and placed within a VorbisComment with the tag name 151 * 'METADATA_BLOCK_PICTURE'. This is the preferred and 152 * recommended way of embedding cover art within VorbisComments." 153 */ 154 if (!av_strcasecmp(tt, "METADATA_BLOCK_PICTURE") && parse_picture) { 155 int ret, len = AV_BASE64_DECODE_SIZE(vl); 156 char *pict = av_malloc(len); 157 158 if (!pict) { 159 av_log(as, AV_LOG_WARNING, "out-of-memory error. Skipping cover art block.\n"); 160 av_freep(&tt); 161 av_freep(&ct); 162 continue; 163 } 164 ret = av_base64_decode(pict, ct, len); 165 av_freep(&tt); 166 av_freep(&ct); 167 if (ret > 0) 168 ret = ff_flac_parse_picture(as, pict, ret); 169 av_freep(&pict); 170 if (ret < 0) { 171 av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n"); 172 continue; 173 } 174 } else if (!ogm_chapter(as, tt, ct)) { 175 updates++; 176 if (av_dict_get(*m, tt, NULL, 0)) { 177 av_dict_set(m, tt, ";", AV_DICT_APPEND); 178 } 179 av_dict_set(m, tt, ct, 180 AV_DICT_DONT_STRDUP_KEY | AV_DICT_DONT_STRDUP_VAL | 181 AV_DICT_APPEND); 182 } 183 } 184 } 185 186 if (p != end) 187 av_log(as, AV_LOG_INFO, 188 "%"PTRDIFF_SPECIFIER" bytes of comment header remain\n", end - p); 189 if (n > 0) 190 av_log(as, AV_LOG_INFO, 191 "truncated comment header, %i comments not found\n", n); 192 193 ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv); 194 195 return updates; 196} 197 198/* 199 * Parse the vorbis header 200 * 201 * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec 202 * [vorbis_version] = read 32 bits as unsigned integer | Not used 203 * [audio_channels] = read 8 bit integer as unsigned | Used 204 * [audio_sample_rate] = read 32 bits as unsigned integer | Used 205 * [bitrate_maximum] = read 32 bits as signed integer | Not used yet 206 * [bitrate_nominal] = read 32 bits as signed integer | Not used yet 207 * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate 208 * [blocksize_0] = read 4 bits as unsigned integer | Not Used 209 * [blocksize_1] = read 4 bits as unsigned integer | Not Used 210 * [framing_flag] = read one bit | Not Used 211 */ 212 213struct oggvorbis_private { 214 unsigned int len[3]; 215 unsigned char *packet[3]; 216 AVVorbisParseContext *vp; 217 int64_t final_pts; 218 int final_duration; 219}; 220 221static int fixup_vorbis_headers(AVFormatContext *as, 222 struct oggvorbis_private *priv, 223 uint8_t **buf) 224{ 225 int i, offset, len, err; 226 int buf_len; 227 unsigned char *ptr; 228 229 len = priv->len[0] + priv->len[1] + priv->len[2]; 230 buf_len = len + len / 255 + 64; 231 232 if (*buf) 233 return AVERROR_INVALIDDATA; 234 235 ptr = *buf = av_realloc(NULL, buf_len); 236 if (!ptr) 237 return AVERROR(ENOMEM); 238 memset(*buf, '\0', buf_len); 239 240 ptr[0] = 2; 241 offset = 1; 242 offset += av_xiphlacing(&ptr[offset], priv->len[0]); 243 offset += av_xiphlacing(&ptr[offset], priv->len[1]); 244 for (i = 0; i < 3; i++) { 245 memcpy(&ptr[offset], priv->packet[i], priv->len[i]); 246 offset += priv->len[i]; 247 av_freep(&priv->packet[i]); 248 } 249 if ((err = av_reallocp(buf, offset + AV_INPUT_BUFFER_PADDING_SIZE)) < 0) 250 return err; 251 return offset; 252} 253 254static void vorbis_cleanup(AVFormatContext *s, int idx) 255{ 256 struct ogg *ogg = s->priv_data; 257 struct ogg_stream *os = ogg->streams + idx; 258 struct oggvorbis_private *priv = os->private; 259 int i; 260 if (os->private) { 261 av_vorbis_parse_free(&priv->vp); 262 for (i = 0; i < 3; i++) 263 av_freep(&priv->packet[i]); 264 } 265} 266 267static int vorbis_update_metadata(AVFormatContext *s, int idx) 268{ 269 struct ogg *ogg = s->priv_data; 270 struct ogg_stream *os = ogg->streams + idx; 271 AVStream *st = s->streams[idx]; 272 int ret; 273 274 if (os->psize <= 8) 275 return 0; 276 277 /* New metadata packet; release old data. */ 278 av_dict_free(&st->metadata); 279 ret = ff_vorbis_stream_comment(s, st, os->buf + os->pstart + 7, 280 os->psize - 8); 281 if (ret < 0) 282 return ret; 283 284 /* Update the metadata if possible. */ 285 av_freep(&os->new_metadata); 286 if (st->metadata) { 287 os->new_metadata = av_packet_pack_dictionary(st->metadata, &os->new_metadata_size); 288 /* Send an empty dictionary to indicate that metadata has been cleared. */ 289 } else { 290 os->new_metadata = av_malloc(1); 291 os->new_metadata_size = 0; 292 } 293 294 return ret; 295} 296 297static int vorbis_header(AVFormatContext *s, int idx) 298{ 299 struct ogg *ogg = s->priv_data; 300 AVStream *st = s->streams[idx]; 301 struct ogg_stream *os = ogg->streams + idx; 302 struct oggvorbis_private *priv; 303 int pkt_type = os->buf[os->pstart]; 304 305 if (!os->private) { 306 os->private = av_mallocz(sizeof(struct oggvorbis_private)); 307 if (!os->private) 308 return AVERROR(ENOMEM); 309 } 310 311 priv = os->private; 312 313 if (!(pkt_type & 1)) 314 return priv->vp ? 0 : AVERROR_INVALIDDATA; 315 316 if (os->psize < 1 || pkt_type > 5) 317 return AVERROR_INVALIDDATA; 318 319 if (priv->packet[pkt_type >> 1]) 320 return AVERROR_INVALIDDATA; 321 if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1]) 322 return priv->vp ? 0 : AVERROR_INVALIDDATA; 323 324 priv->len[pkt_type >> 1] = os->psize; 325 priv->packet[pkt_type >> 1] = av_mallocz(os->psize); 326 if (!priv->packet[pkt_type >> 1]) 327 return AVERROR(ENOMEM); 328 memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize); 329 if (os->buf[os->pstart] == 1) { 330 const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */ 331 unsigned blocksize, bs0, bs1; 332 int srate; 333 int channels; 334 335 if (os->psize != 30) 336 return AVERROR_INVALIDDATA; 337 338 if (bytestream_get_le32(&p) != 0) /* vorbis_version */ 339 return AVERROR_INVALIDDATA; 340 341 channels = bytestream_get_byte(&p); 342 if (st->codecpar->channels && channels != st->codecpar->channels) { 343 av_log(s, AV_LOG_ERROR, "Channel change is not supported\n"); 344 return AVERROR_PATCHWELCOME; 345 } 346 st->codecpar->channels = channels; 347 srate = bytestream_get_le32(&p); 348 p += 4; // skip maximum bitrate 349 st->codecpar->bit_rate = bytestream_get_le32(&p); // nominal bitrate 350 p += 4; // skip minimum bitrate 351 352 blocksize = bytestream_get_byte(&p); 353 bs0 = blocksize & 15; 354 bs1 = blocksize >> 4; 355 356 if (bs0 > bs1) 357 return AVERROR_INVALIDDATA; 358 if (bs0 < 6 || bs1 > 13) 359 return AVERROR_INVALIDDATA; 360 361 if (bytestream_get_byte(&p) != 1) /* framing_flag */ 362 return AVERROR_INVALIDDATA; 363 364 st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO; 365 st->codecpar->codec_id = AV_CODEC_ID_VORBIS; 366 367 if (srate > 0) { 368 st->codecpar->sample_rate = srate; 369 avpriv_set_pts_info(st, 64, 1, srate); 370 } 371 } else if (os->buf[os->pstart] == 3) { 372 if (vorbis_update_metadata(s, idx) >= 0 && priv->len[1] > 10) { 373 unsigned new_len; 374 375 int ret = ff_replaygain_export(st, st->metadata); 376 if (ret < 0) 377 return ret; 378 379 // drop all metadata we parsed and which is not required by libvorbis 380 new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1; 381 if (new_len >= 16 && new_len < os->psize) { 382 AV_WL32(priv->packet[1] + new_len - 5, 0); 383 priv->packet[1][new_len - 1] = 1; 384 priv->len[1] = new_len; 385 } 386 } 387 } else { 388 int ret = fixup_vorbis_headers(s, priv, &st->codecpar->extradata); 389 if (ret < 0) { 390 st->codecpar->extradata_size = 0; 391 return ret; 392 } 393 st->codecpar->extradata_size = ret; 394 395 priv->vp = av_vorbis_parse_init(st->codecpar->extradata, st->codecpar->extradata_size); 396 if (!priv->vp) { 397 av_freep(&st->codecpar->extradata); 398 st->codecpar->extradata_size = 0; 399 return AVERROR_UNKNOWN; 400 } 401 } 402 403 return 1; 404} 405 406static int vorbis_packet(AVFormatContext *s, int idx) 407{ 408 struct ogg *ogg = s->priv_data; 409 struct ogg_stream *os = ogg->streams + idx; 410 struct oggvorbis_private *priv = os->private; 411 int duration, flags = 0; 412 413 if (!priv->vp) 414 return AVERROR_INVALIDDATA; 415 416 /* first packet handling 417 * here we parse the duration of each packet in the first page and compare 418 * the total duration to the page granule to find the encoder delay and 419 * set the first timestamp */ 420 if ((!os->lastpts || os->lastpts == AV_NOPTS_VALUE) && !(os->flags & OGG_FLAG_EOS) && (int64_t)os->granule>=0) { 421 int seg, d; 422 uint8_t *last_pkt = os->buf + os->pstart; 423 uint8_t *next_pkt = last_pkt; 424 425 av_vorbis_parse_reset(priv->vp); 426 duration = 0; 427 seg = os->segp; 428 d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags); 429 if (d < 0) { 430 os->pflags |= AV_PKT_FLAG_CORRUPT; 431 return 0; 432 } else if (flags & VORBIS_FLAG_COMMENT) { 433 vorbis_update_metadata(s, idx); 434 flags = 0; 435 } 436 duration += d; 437 last_pkt = next_pkt = next_pkt + os->psize; 438 for (; seg < os->nsegs; seg++) { 439 if (os->segments[seg] < 255) { 440 int d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags); 441 if (d < 0) { 442 duration = os->granule; 443 break; 444 } else if (flags & VORBIS_FLAG_COMMENT) { 445 vorbis_update_metadata(s, idx); 446 flags = 0; 447 } 448 duration += d; 449 last_pkt = next_pkt + os->segments[seg]; 450 } 451 next_pkt += os->segments[seg]; 452 } 453 os->lastpts = 454 os->lastdts = os->granule - duration; 455 456 if (!os->granule && duration) //hack to deal with broken files (Ticket3710) 457 os->lastpts = os->lastdts = AV_NOPTS_VALUE; 458 459 if (s->streams[idx]->start_time == AV_NOPTS_VALUE) { 460 s->streams[idx]->start_time = FFMAX(os->lastpts, 0); 461 if (s->streams[idx]->duration != AV_NOPTS_VALUE) 462 s->streams[idx]->duration -= s->streams[idx]->start_time; 463 } 464 priv->final_pts = AV_NOPTS_VALUE; 465 av_vorbis_parse_reset(priv->vp); 466 } 467 468 /* parse packet duration */ 469 if (os->psize > 0) { 470 duration = av_vorbis_parse_frame_flags(priv->vp, os->buf + os->pstart, 1, &flags); 471 if (duration < 0) { 472 os->pflags |= AV_PKT_FLAG_CORRUPT; 473 return 0; 474 } else if (flags & VORBIS_FLAG_COMMENT) { 475 vorbis_update_metadata(s, idx); 476 flags = 0; 477 } 478 os->pduration = duration; 479 } 480 481 /* final packet handling 482 * here we save the pts of the first packet in the final page, sum up all 483 * packet durations in the final page except for the last one, and compare 484 * to the page granule to find the duration of the final packet */ 485 if (os->flags & OGG_FLAG_EOS) { 486 if (os->lastpts != AV_NOPTS_VALUE) { 487 priv->final_pts = os->lastpts; 488 priv->final_duration = 0; 489 } 490 if (os->segp == os->nsegs) 491 os->pduration = os->granule - priv->final_pts - priv->final_duration; 492 priv->final_duration += os->pduration; 493 } 494 495 return 0; 496} 497 498const struct ogg_codec ff_vorbis_codec = { 499 .magic = "\001vorbis", 500 .magicsize = 7, 501 .header = vorbis_header, 502 .packet = vorbis_packet, 503 .cleanup = vorbis_cleanup, 504 .nb_header = 3, 505};