/libavformat/oggparsevorbis.c

http://github.com/FFmpeg/FFmpeg · 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. #include <stdlib.h>
  25. #include "libavutil/avstring.h"
  26. #include "libavutil/base64.h"
  27. #include "libavutil/bswap.h"
  28. #include "libavutil/dict.h"
  29. #include "libavcodec/bytestream.h"
  30. #include "libavcodec/vorbis_parser.h"
  31. #include "avformat.h"
  32. #include "flac_picture.h"
  33. #include "internal.h"
  34. #include "oggdec.h"
  35. #include "vorbiscomment.h"
  36. #include "replaygain.h"
  37. static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
  38. {
  39. int i, cnum, h, m, s, ms, keylen = strlen(key);
  40. AVChapter *chapter = NULL;
  41. if (keylen < 9 || av_strncasecmp(key, "CHAPTER", 7) || sscanf(key+7, "%03d", &cnum) != 1)
  42. return 0;
  43. if (keylen <= 10) {
  44. if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
  45. return 0;
  46. avpriv_new_chapter(as, cnum, (AVRational) { 1, 1000 },
  47. ms + 1000 * (s + 60 * (m + 60 * h)),
  48. AV_NOPTS_VALUE, NULL);
  49. av_free(val);
  50. } else if (!av_strcasecmp(key + keylen - 4, "NAME")) {
  51. for (i = 0; i < as->nb_chapters; i++)
  52. if (as->chapters[i]->id == cnum) {
  53. chapter = as->chapters[i];
  54. break;
  55. }
  56. if (!chapter)
  57. return 0;
  58. av_dict_set(&chapter->metadata, "title", val, AV_DICT_DONT_STRDUP_VAL);
  59. } else
  60. return 0;
  61. av_free(key);
  62. return 1;
  63. }
  64. int ff_vorbis_stream_comment(AVFormatContext *as, AVStream *st,
  65. const uint8_t *buf, int size)
  66. {
  67. int updates = ff_vorbis_comment(as, &st->metadata, buf, size, 1);
  68. if (updates > 0) {
  69. st->event_flags |= AVSTREAM_EVENT_FLAG_METADATA_UPDATED;
  70. }
  71. return updates;
  72. }
  73. int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m,
  74. const uint8_t *buf, int size,
  75. int parse_picture)
  76. {
  77. const uint8_t *p = buf;
  78. const uint8_t *end = buf + size;
  79. int updates = 0;
  80. unsigned n;
  81. int s;
  82. /* must have vendor_length and user_comment_list_length */
  83. if (size < 8)
  84. return AVERROR_INVALIDDATA;
  85. s = bytestream_get_le32(&p);
  86. if (end - p - 4 < s || s < 0)
  87. return AVERROR_INVALIDDATA;
  88. p += s;
  89. n = bytestream_get_le32(&p);
  90. while (end - p >= 4 && n > 0) {
  91. const char *t, *v;
  92. int tl, vl;
  93. s = bytestream_get_le32(&p);
  94. if (end - p < s || s < 0)
  95. break;
  96. t = p;
  97. p += s;
  98. n--;
  99. v = memchr(t, '=', s);
  100. if (!v)
  101. continue;
  102. tl = v - t;
  103. vl = s - tl - 1;
  104. v++;
  105. if (tl && vl) {
  106. char *tt, *ct;
  107. tt = av_malloc(tl + 1);
  108. ct = av_malloc(vl + 1);
  109. if (!tt || !ct) {
  110. av_freep(&tt);
  111. av_freep(&ct);
  112. return AVERROR(ENOMEM);
  113. }
  114. memcpy(tt, t, tl);
  115. tt[tl] = 0;
  116. memcpy(ct, v, vl);
  117. ct[vl] = 0;
  118. /* The format in which the pictures are stored is the FLAC format.
  119. * Xiph says: "The binary FLAC picture structure is base64 encoded
  120. * and placed within a VorbisComment with the tag name
  121. * 'METADATA_BLOCK_PICTURE'. This is the preferred and
  122. * recommended way of embedding cover art within VorbisComments."
  123. */
  124. if (!av_strcasecmp(tt, "METADATA_BLOCK_PICTURE") && parse_picture) {
  125. int ret, len = AV_BASE64_DECODE_SIZE(vl);
  126. char *pict = av_malloc(len);
  127. if (!pict) {
  128. av_log(as, AV_LOG_WARNING, "out-of-memory error. Skipping cover art block.\n");
  129. av_freep(&tt);
  130. av_freep(&ct);
  131. continue;
  132. }
  133. ret = av_base64_decode(pict, ct, len);
  134. av_freep(&tt);
  135. av_freep(&ct);
  136. if (ret > 0)
  137. ret = ff_flac_parse_picture(as, pict, ret);
  138. av_freep(&pict);
  139. if (ret < 0) {
  140. av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n");
  141. continue;
  142. }
  143. } else if (!ogm_chapter(as, tt, ct)) {
  144. updates++;
  145. if (av_dict_get(*m, tt, NULL, 0)) {
  146. av_dict_set(m, tt, ";", AV_DICT_APPEND);
  147. }
  148. av_dict_set(m, tt, ct,
  149. AV_DICT_DONT_STRDUP_KEY | AV_DICT_DONT_STRDUP_VAL |
  150. AV_DICT_APPEND);
  151. }
  152. }
  153. }
  154. if (p != end)
  155. av_log(as, AV_LOG_INFO,
  156. "%"PTRDIFF_SPECIFIER" bytes of comment header remain\n", end - p);
  157. if (n > 0)
  158. av_log(as, AV_LOG_INFO,
  159. "truncated comment header, %i comments not found\n", n);
  160. ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
  161. return updates;
  162. }
  163. /*
  164. * Parse the vorbis header
  165. *
  166. * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
  167. * [vorbis_version] = read 32 bits as unsigned integer | Not used
  168. * [audio_channels] = read 8 bit integer as unsigned | Used
  169. * [audio_sample_rate] = read 32 bits as unsigned integer | Used
  170. * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
  171. * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
  172. * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
  173. * [blocksize_0] = read 4 bits as unsigned integer | Not Used
  174. * [blocksize_1] = read 4 bits as unsigned integer | Not Used
  175. * [framing_flag] = read one bit | Not Used
  176. */
  177. struct oggvorbis_private {
  178. unsigned int len[3];
  179. unsigned char *packet[3];
  180. AVVorbisParseContext *vp;
  181. int64_t final_pts;
  182. int final_duration;
  183. };
  184. static int fixup_vorbis_headers(AVFormatContext *as,
  185. struct oggvorbis_private *priv,
  186. uint8_t **buf)
  187. {
  188. int i, offset, len, err;
  189. int buf_len;
  190. unsigned char *ptr;
  191. len = priv->len[0] + priv->len[1] + priv->len[2];
  192. buf_len = len + len / 255 + 64;
  193. if (*buf)
  194. return AVERROR_INVALIDDATA;
  195. ptr = *buf = av_realloc(NULL, buf_len);
  196. if (!ptr)
  197. return AVERROR(ENOMEM);
  198. memset(*buf, '\0', buf_len);
  199. ptr[0] = 2;
  200. offset = 1;
  201. offset += av_xiphlacing(&ptr[offset], priv->len[0]);
  202. offset += av_xiphlacing(&ptr[offset], priv->len[1]);
  203. for (i = 0; i < 3; i++) {
  204. memcpy(&ptr[offset], priv->packet[i], priv->len[i]);
  205. offset += priv->len[i];
  206. av_freep(&priv->packet[i]);
  207. }
  208. if ((err = av_reallocp(buf, offset + AV_INPUT_BUFFER_PADDING_SIZE)) < 0)
  209. return err;
  210. return offset;
  211. }
  212. static void vorbis_cleanup(AVFormatContext *s, int idx)
  213. {
  214. struct ogg *ogg = s->priv_data;
  215. struct ogg_stream *os = ogg->streams + idx;
  216. struct oggvorbis_private *priv = os->private;
  217. int i;
  218. if (os->private) {
  219. av_vorbis_parse_free(&priv->vp);
  220. for (i = 0; i < 3; i++)
  221. av_freep(&priv->packet[i]);
  222. }
  223. }
  224. static int vorbis_update_metadata(AVFormatContext *s, int idx)
  225. {
  226. struct ogg *ogg = s->priv_data;
  227. struct ogg_stream *os = ogg->streams + idx;
  228. AVStream *st = s->streams[idx];
  229. int ret;
  230. if (os->psize <= 8)
  231. return 0;
  232. /* New metadata packet; release old data. */
  233. av_dict_free(&st->metadata);
  234. ret = ff_vorbis_stream_comment(s, st, os->buf + os->pstart + 7,
  235. os->psize - 8);
  236. if (ret < 0)
  237. return ret;
  238. /* Update the metadata if possible. */
  239. av_freep(&os->new_metadata);
  240. if (st->metadata) {
  241. os->new_metadata = av_packet_pack_dictionary(st->metadata, &os->new_metadata_size);
  242. /* Send an empty dictionary to indicate that metadata has been cleared. */
  243. } else {
  244. os->new_metadata = av_malloc(1);
  245. os->new_metadata_size = 0;
  246. }
  247. return ret;
  248. }
  249. static int vorbis_header(AVFormatContext *s, int idx)
  250. {
  251. struct ogg *ogg = s->priv_data;
  252. AVStream *st = s->streams[idx];
  253. struct ogg_stream *os = ogg->streams + idx;
  254. struct oggvorbis_private *priv;
  255. int pkt_type = os->buf[os->pstart];
  256. if (!os->private) {
  257. os->private = av_mallocz(sizeof(struct oggvorbis_private));
  258. if (!os->private)
  259. return AVERROR(ENOMEM);
  260. }
  261. priv = os->private;
  262. if (!(pkt_type & 1))
  263. return priv->vp ? 0 : AVERROR_INVALIDDATA;
  264. if (os->psize < 1 || pkt_type > 5)
  265. return AVERROR_INVALIDDATA;
  266. if (priv->packet[pkt_type >> 1])
  267. return AVERROR_INVALIDDATA;
  268. if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
  269. return priv->vp ? 0 : AVERROR_INVALIDDATA;
  270. priv->len[pkt_type >> 1] = os->psize;
  271. priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
  272. if (!priv->packet[pkt_type >> 1])
  273. return AVERROR(ENOMEM);
  274. memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
  275. if (os->buf[os->pstart] == 1) {
  276. const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
  277. unsigned blocksize, bs0, bs1;
  278. int srate;
  279. int channels;
  280. if (os->psize != 30)
  281. return AVERROR_INVALIDDATA;
  282. if (bytestream_get_le32(&p) != 0) /* vorbis_version */
  283. return AVERROR_INVALIDDATA;
  284. channels = bytestream_get_byte(&p);
  285. if (st->codecpar->channels && channels != st->codecpar->channels) {
  286. av_log(s, AV_LOG_ERROR, "Channel change is not supported\n");
  287. return AVERROR_PATCHWELCOME;
  288. }
  289. st->codecpar->channels = channels;
  290. srate = bytestream_get_le32(&p);
  291. p += 4; // skip maximum bitrate
  292. st->codecpar->bit_rate = bytestream_get_le32(&p); // nominal bitrate
  293. p += 4; // skip minimum bitrate
  294. blocksize = bytestream_get_byte(&p);
  295. bs0 = blocksize & 15;
  296. bs1 = blocksize >> 4;
  297. if (bs0 > bs1)
  298. return AVERROR_INVALIDDATA;
  299. if (bs0 < 6 || bs1 > 13)
  300. return AVERROR_INVALIDDATA;
  301. if (bytestream_get_byte(&p) != 1) /* framing_flag */
  302. return AVERROR_INVALIDDATA;
  303. st->codecpar->codec_type = AVMEDIA_TYPE_AUDIO;
  304. st->codecpar->codec_id = AV_CODEC_ID_VORBIS;
  305. if (srate > 0) {
  306. st->codecpar->sample_rate = srate;
  307. avpriv_set_pts_info(st, 64, 1, srate);
  308. }
  309. } else if (os->buf[os->pstart] == 3) {
  310. if (vorbis_update_metadata(s, idx) >= 0 && priv->len[1] > 10) {
  311. unsigned new_len;
  312. int ret = ff_replaygain_export(st, st->metadata);
  313. if (ret < 0)
  314. return ret;
  315. // drop all metadata we parsed and which is not required by libvorbis
  316. new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
  317. if (new_len >= 16 && new_len < os->psize) {
  318. AV_WL32(priv->packet[1] + new_len - 5, 0);
  319. priv->packet[1][new_len - 1] = 1;
  320. priv->len[1] = new_len;
  321. }
  322. }
  323. } else {
  324. int ret = fixup_vorbis_headers(s, priv, &st->codecpar->extradata);
  325. if (ret < 0) {
  326. st->codecpar->extradata_size = 0;
  327. return ret;
  328. }
  329. st->codecpar->extradata_size = ret;
  330. priv->vp = av_vorbis_parse_init(st->codecpar->extradata, st->codecpar->extradata_size);
  331. if (!priv->vp) {
  332. av_freep(&st->codecpar->extradata);
  333. st->codecpar->extradata_size = 0;
  334. return AVERROR_UNKNOWN;
  335. }
  336. }
  337. return 1;
  338. }
  339. static int vorbis_packet(AVFormatContext *s, int idx)
  340. {
  341. struct ogg *ogg = s->priv_data;
  342. struct ogg_stream *os = ogg->streams + idx;
  343. struct oggvorbis_private *priv = os->private;
  344. int duration, flags = 0;
  345. if (!priv->vp)
  346. return AVERROR_INVALIDDATA;
  347. /* first packet handling
  348. * here we parse the duration of each packet in the first page and compare
  349. * the total duration to the page granule to find the encoder delay and
  350. * set the first timestamp */
  351. if ((!os->lastpts || os->lastpts == AV_NOPTS_VALUE) && !(os->flags & OGG_FLAG_EOS) && (int64_t)os->granule>=0) {
  352. int seg, d;
  353. uint8_t *last_pkt = os->buf + os->pstart;
  354. uint8_t *next_pkt = last_pkt;
  355. av_vorbis_parse_reset(priv->vp);
  356. duration = 0;
  357. seg = os->segp;
  358. d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags);
  359. if (d < 0) {
  360. os->pflags |= AV_PKT_FLAG_CORRUPT;
  361. return 0;
  362. } else if (flags & VORBIS_FLAG_COMMENT) {
  363. vorbis_update_metadata(s, idx);
  364. flags = 0;
  365. }
  366. duration += d;
  367. last_pkt = next_pkt = next_pkt + os->psize;
  368. for (; seg < os->nsegs; seg++) {
  369. if (os->segments[seg] < 255) {
  370. int d = av_vorbis_parse_frame_flags(priv->vp, last_pkt, 1, &flags);
  371. if (d < 0) {
  372. duration = os->granule;
  373. break;
  374. } else if (flags & VORBIS_FLAG_COMMENT) {
  375. vorbis_update_metadata(s, idx);
  376. flags = 0;
  377. }
  378. duration += d;
  379. last_pkt = next_pkt + os->segments[seg];
  380. }
  381. next_pkt += os->segments[seg];
  382. }
  383. os->lastpts =
  384. os->lastdts = os->granule - duration;
  385. if (!os->granule && duration) //hack to deal with broken files (Ticket3710)
  386. os->lastpts = os->lastdts = AV_NOPTS_VALUE;
  387. if (s->streams[idx]->start_time == AV_NOPTS_VALUE) {
  388. s->streams[idx]->start_time = FFMAX(os->lastpts, 0);
  389. if (s->streams[idx]->duration != AV_NOPTS_VALUE)
  390. s->streams[idx]->duration -= s->streams[idx]->start_time;
  391. }
  392. priv->final_pts = AV_NOPTS_VALUE;
  393. av_vorbis_parse_reset(priv->vp);
  394. }
  395. /* parse packet duration */
  396. if (os->psize > 0) {
  397. duration = av_vorbis_parse_frame_flags(priv->vp, os->buf + os->pstart, 1, &flags);
  398. if (duration < 0) {
  399. os->pflags |= AV_PKT_FLAG_CORRUPT;
  400. return 0;
  401. } else if (flags & VORBIS_FLAG_COMMENT) {
  402. vorbis_update_metadata(s, idx);
  403. flags = 0;
  404. }
  405. os->pduration = duration;
  406. }
  407. /* final packet handling
  408. * here we save the pts of the first packet in the final page, sum up all
  409. * packet durations in the final page except for the last one, and compare
  410. * to the page granule to find the duration of the final packet */
  411. if (os->flags & OGG_FLAG_EOS) {
  412. if (os->lastpts != AV_NOPTS_VALUE) {
  413. priv->final_pts = os->lastpts;
  414. priv->final_duration = 0;
  415. }
  416. if (os->segp == os->nsegs)
  417. os->pduration = os->granule - priv->final_pts - priv->final_duration;
  418. priv->final_duration += os->pduration;
  419. }
  420. return 0;
  421. }
  422. const struct ogg_codec ff_vorbis_codec = {
  423. .magic = "\001vorbis",
  424. .magicsize = 7,
  425. .header = vorbis_header,
  426. .packet = vorbis_packet,
  427. .cleanup = vorbis_cleanup,
  428. .nb_header = 3,
  429. };