oggenc.c 16 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516
  1. /*
  2. * Ogg muxer
  3. * Copyright (c) 2007 Baptiste Coudurier <baptiste dot coudurier at free dot fr>
  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. #include "libavutil/crc.h"
  22. #include "libavutil/random_seed.h"
  23. #include "libavcodec/xiph.h"
  24. #include "libavcodec/bytestream.h"
  25. #include "libavcodec/flac.h"
  26. #include "avformat.h"
  27. #include "internal.h"
  28. #include "vorbiscomment.h"
  29. #define MAX_PAGE_SIZE 65025
  30. typedef struct {
  31. int64_t granule;
  32. int stream_index;
  33. uint8_t flags;
  34. uint8_t segments_count;
  35. uint8_t segments[255];
  36. uint8_t data[MAX_PAGE_SIZE];
  37. uint16_t size;
  38. } OGGPage;
  39. typedef struct {
  40. unsigned page_counter;
  41. uint8_t *header[3];
  42. int header_len[3];
  43. /** for theora granule */
  44. int kfgshift;
  45. int64_t last_kf_pts;
  46. int vrev;
  47. int eos;
  48. unsigned page_count; ///< number of page buffered
  49. OGGPage page; ///< current page
  50. unsigned serial_num; ///< serial number
  51. int64_t last_granule; ///< last packet granule
  52. } OGGStreamContext;
  53. typedef struct OGGPageList {
  54. OGGPage page;
  55. struct OGGPageList *next;
  56. } OGGPageList;
  57. typedef struct {
  58. OGGPageList *page_list;
  59. } OGGContext;
  60. static void ogg_update_checksum(AVFormatContext *s, ByteIOContext *pb, int64_t crc_offset)
  61. {
  62. int64_t pos = url_ftell(pb);
  63. uint32_t checksum = get_checksum(pb);
  64. url_fseek(pb, crc_offset, SEEK_SET);
  65. put_be32(pb, checksum);
  66. url_fseek(pb, pos, SEEK_SET);
  67. }
  68. static int ogg_write_page(AVFormatContext *s, OGGPage *page, int extra_flags)
  69. {
  70. OGGStreamContext *oggstream = s->streams[page->stream_index]->priv_data;
  71. ByteIOContext *pb;
  72. int64_t crc_offset;
  73. int ret, size;
  74. uint8_t *buf;
  75. ret = url_open_dyn_buf(&pb);
  76. if (ret < 0)
  77. return ret;
  78. init_checksum(pb, ff_crc04C11DB7_update, 0);
  79. put_tag(pb, "OggS");
  80. put_byte(pb, 0);
  81. put_byte(pb, page->flags | extra_flags);
  82. put_le64(pb, page->granule);
  83. put_le32(pb, oggstream->serial_num);
  84. put_le32(pb, oggstream->page_counter++);
  85. crc_offset = url_ftell(pb);
  86. put_le32(pb, 0); // crc
  87. put_byte(pb, page->segments_count);
  88. put_buffer(pb, page->segments, page->segments_count);
  89. put_buffer(pb, page->data, page->size);
  90. ogg_update_checksum(s, pb, crc_offset);
  91. put_flush_packet(pb);
  92. size = url_close_dyn_buf(pb, &buf);
  93. if (size < 0)
  94. return size;
  95. put_buffer(s->pb, buf, size);
  96. put_flush_packet(s->pb);
  97. av_free(buf);
  98. oggstream->page_count--;
  99. return 0;
  100. }
  101. static int64_t ogg_granule_to_timestamp(OGGStreamContext *oggstream, int64_t granule)
  102. {
  103. if (oggstream->kfgshift)
  104. return (granule>>oggstream->kfgshift) +
  105. (granule & ((1<<oggstream->kfgshift)-1));
  106. else
  107. return granule;
  108. }
  109. static int ogg_compare_granule(AVFormatContext *s, OGGPage *next, OGGPage *page)
  110. {
  111. AVStream *st2 = s->streams[next->stream_index];
  112. AVStream *st = s->streams[page->stream_index];
  113. int64_t next_granule, cur_granule;
  114. if (next->granule == -1 || page->granule == -1)
  115. return 0;
  116. next_granule = av_rescale_q(ogg_granule_to_timestamp(st2->priv_data, next->granule),
  117. st2->time_base, AV_TIME_BASE_Q);
  118. cur_granule = av_rescale_q(ogg_granule_to_timestamp(st->priv_data, page->granule),
  119. st ->time_base, AV_TIME_BASE_Q);
  120. return next_granule > cur_granule;
  121. }
  122. static int ogg_reset_cur_page(OGGStreamContext *oggstream)
  123. {
  124. oggstream->page.granule = -1;
  125. oggstream->page.flags = 0;
  126. oggstream->page.segments_count = 0;
  127. oggstream->page.size = 0;
  128. return 0;
  129. }
  130. static int ogg_buffer_page(AVFormatContext *s, OGGStreamContext *oggstream)
  131. {
  132. OGGContext *ogg = s->priv_data;
  133. OGGPageList **p = &ogg->page_list;
  134. OGGPageList *l = av_mallocz(sizeof(*l));
  135. if (!l)
  136. return AVERROR(ENOMEM);
  137. l->page = oggstream->page;
  138. oggstream->page_count++;
  139. ogg_reset_cur_page(oggstream);
  140. while (*p) {
  141. if (ogg_compare_granule(s, &(*p)->page, &l->page))
  142. break;
  143. p = &(*p)->next;
  144. }
  145. l->next = *p;
  146. *p = l;
  147. return 0;
  148. }
  149. static int ogg_buffer_data(AVFormatContext *s, AVStream *st,
  150. uint8_t *data, unsigned size, int64_t granule)
  151. {
  152. OGGStreamContext *oggstream = st->priv_data;
  153. int total_segments = size / 255 + 1;
  154. uint8_t *p = data;
  155. int i, segments, len, flush = 0;
  156. // Handles VFR by flushing page because this frame needs to have a timestamp
  157. if (st->codec->codec_id == CODEC_ID_THEORA &&
  158. ogg_granule_to_timestamp(oggstream, granule) >
  159. ogg_granule_to_timestamp(oggstream, oggstream->last_granule) + 1) {
  160. if (oggstream->page.granule != -1)
  161. ogg_buffer_page(s, oggstream);
  162. flush = 1;
  163. }
  164. for (i = 0; i < total_segments; ) {
  165. OGGPage *page = &oggstream->page;
  166. segments = FFMIN(total_segments - i, 255 - page->segments_count);
  167. if (i && !page->segments_count)
  168. page->flags |= 1; // continued packet
  169. memset(page->segments+page->segments_count, 255, segments - 1);
  170. page->segments_count += segments - 1;
  171. len = FFMIN(size, segments*255);
  172. page->segments[page->segments_count++] = len - (segments-1)*255;
  173. memcpy(page->data+page->size, p, len);
  174. p += len;
  175. size -= len;
  176. i += segments;
  177. page->size += len;
  178. if (i == total_segments)
  179. page->granule = granule;
  180. if (page->segments_count == 255) {
  181. ogg_buffer_page(s, oggstream);
  182. }
  183. }
  184. if (flush && oggstream->page.granule != -1)
  185. ogg_buffer_page(s, oggstream);
  186. return 0;
  187. }
  188. static uint8_t *ogg_write_vorbiscomment(int offset, int bitexact,
  189. int *header_len, AVMetadata **m, int framing_bit)
  190. {
  191. const char *vendor = bitexact ? "ffmpeg" : LIBAVFORMAT_IDENT;
  192. int size;
  193. uint8_t *p, *p0;
  194. unsigned int count;
  195. ff_metadata_conv(m, ff_vorbiscomment_metadata_conv, NULL);
  196. size = offset + ff_vorbiscomment_length(*m, vendor, &count) + framing_bit;
  197. p = av_mallocz(size);
  198. if (!p)
  199. return NULL;
  200. p0 = p;
  201. p += offset;
  202. ff_vorbiscomment_write(&p, m, vendor, count);
  203. if (framing_bit)
  204. bytestream_put_byte(&p, 1);
  205. *header_len = size;
  206. return p0;
  207. }
  208. static int ogg_build_flac_headers(AVCodecContext *avctx,
  209. OGGStreamContext *oggstream, int bitexact,
  210. AVMetadata **m)
  211. {
  212. enum FLACExtradataFormat format;
  213. uint8_t *streaminfo;
  214. uint8_t *p;
  215. if (!ff_flac_is_extradata_valid(avctx, &format, &streaminfo))
  216. return -1;
  217. // first packet: STREAMINFO
  218. oggstream->header_len[0] = 51;
  219. oggstream->header[0] = av_mallocz(51); // per ogg flac specs
  220. p = oggstream->header[0];
  221. if (!p)
  222. return AVERROR(ENOMEM);
  223. bytestream_put_byte(&p, 0x7F);
  224. bytestream_put_buffer(&p, "FLAC", 4);
  225. bytestream_put_byte(&p, 1); // major version
  226. bytestream_put_byte(&p, 0); // minor version
  227. bytestream_put_be16(&p, 1); // headers packets without this one
  228. bytestream_put_buffer(&p, "fLaC", 4);
  229. bytestream_put_byte(&p, 0x00); // streaminfo
  230. bytestream_put_be24(&p, 34);
  231. bytestream_put_buffer(&p, streaminfo, FLAC_STREAMINFO_SIZE);
  232. // second packet: VorbisComment
  233. p = ogg_write_vorbiscomment(4, bitexact, &oggstream->header_len[1], m, 0);
  234. if (!p)
  235. return AVERROR(ENOMEM);
  236. oggstream->header[1] = p;
  237. bytestream_put_byte(&p, 0x84); // last metadata block and vorbis comment
  238. bytestream_put_be24(&p, oggstream->header_len[1] - 4);
  239. return 0;
  240. }
  241. #define SPEEX_HEADER_SIZE 80
  242. static int ogg_build_speex_headers(AVCodecContext *avctx,
  243. OGGStreamContext *oggstream, int bitexact,
  244. AVMetadata **m)
  245. {
  246. uint8_t *p;
  247. if (avctx->extradata_size < SPEEX_HEADER_SIZE)
  248. return -1;
  249. // first packet: Speex header
  250. p = av_mallocz(SPEEX_HEADER_SIZE);
  251. if (!p)
  252. return AVERROR(ENOMEM);
  253. oggstream->header[0] = p;
  254. oggstream->header_len[0] = SPEEX_HEADER_SIZE;
  255. bytestream_put_buffer(&p, avctx->extradata, SPEEX_HEADER_SIZE);
  256. AV_WL32(&oggstream->header[0][68], 0); // set extra_headers to 0
  257. // second packet: VorbisComment
  258. p = ogg_write_vorbiscomment(0, bitexact, &oggstream->header_len[1], m, 0);
  259. if (!p)
  260. return AVERROR(ENOMEM);
  261. oggstream->header[1] = p;
  262. return 0;
  263. }
  264. static int ogg_write_header(AVFormatContext *s)
  265. {
  266. OGGStreamContext *oggstream;
  267. int i, j;
  268. for (i = 0; i < s->nb_streams; i++) {
  269. AVStream *st = s->streams[i];
  270. unsigned serial_num = i;
  271. if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO)
  272. av_set_pts_info(st, 64, 1, st->codec->sample_rate);
  273. else if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO)
  274. av_set_pts_info(st, 64, st->codec->time_base.num, st->codec->time_base.den);
  275. if (st->codec->codec_id != CODEC_ID_VORBIS &&
  276. st->codec->codec_id != CODEC_ID_THEORA &&
  277. st->codec->codec_id != CODEC_ID_SPEEX &&
  278. st->codec->codec_id != CODEC_ID_FLAC) {
  279. av_log(s, AV_LOG_ERROR, "Unsupported codec id in stream %d\n", i);
  280. return -1;
  281. }
  282. if (!st->codec->extradata || !st->codec->extradata_size) {
  283. av_log(s, AV_LOG_ERROR, "No extradata present\n");
  284. return -1;
  285. }
  286. oggstream = av_mallocz(sizeof(*oggstream));
  287. oggstream->page.stream_index = i;
  288. if (!(st->codec->flags & CODEC_FLAG_BITEXACT))
  289. do {
  290. serial_num = av_get_random_seed();
  291. for (j = 0; j < i; j++) {
  292. OGGStreamContext *sc = s->streams[j]->priv_data;
  293. if (serial_num == sc->serial_num)
  294. break;
  295. }
  296. } while (j < i);
  297. oggstream->serial_num = serial_num;
  298. st->priv_data = oggstream;
  299. if (st->codec->codec_id == CODEC_ID_FLAC) {
  300. int err = ogg_build_flac_headers(st->codec, oggstream,
  301. st->codec->flags & CODEC_FLAG_BITEXACT,
  302. &s->metadata);
  303. if (err) {
  304. av_log(s, AV_LOG_ERROR, "Error writing FLAC headers\n");
  305. av_freep(&st->priv_data);
  306. return err;
  307. }
  308. } else if (st->codec->codec_id == CODEC_ID_SPEEX) {
  309. int err = ogg_build_speex_headers(st->codec, oggstream,
  310. st->codec->flags & CODEC_FLAG_BITEXACT,
  311. &s->metadata);
  312. if (err) {
  313. av_log(s, AV_LOG_ERROR, "Error writing Speex headers\n");
  314. av_freep(&st->priv_data);
  315. return err;
  316. }
  317. } else {
  318. uint8_t *p;
  319. const char *cstr = st->codec->codec_id == CODEC_ID_VORBIS ? "vorbis" : "theora";
  320. int header_type = st->codec->codec_id == CODEC_ID_VORBIS ? 3 : 0x81;
  321. int framing_bit = st->codec->codec_id == CODEC_ID_VORBIS ? 1 : 0;
  322. if (ff_split_xiph_headers(st->codec->extradata, st->codec->extradata_size,
  323. st->codec->codec_id == CODEC_ID_VORBIS ? 30 : 42,
  324. oggstream->header, oggstream->header_len) < 0) {
  325. av_log(s, AV_LOG_ERROR, "Extradata corrupted\n");
  326. av_freep(&st->priv_data);
  327. return -1;
  328. }
  329. p = ogg_write_vorbiscomment(7, st->codec->flags & CODEC_FLAG_BITEXACT,
  330. &oggstream->header_len[1], &s->metadata,
  331. framing_bit);
  332. if (!p)
  333. return AVERROR(ENOMEM);
  334. oggstream->header[1] = p;
  335. bytestream_put_byte(&p, header_type);
  336. bytestream_put_buffer(&p, cstr, 6);
  337. if (st->codec->codec_id == CODEC_ID_THEORA) {
  338. /** KFGSHIFT is the width of the less significant section of the granule position
  339. The less significant section is the frame count since the last keyframe */
  340. oggstream->kfgshift = ((oggstream->header[0][40]&3)<<3)|(oggstream->header[0][41]>>5);
  341. oggstream->vrev = oggstream->header[0][9];
  342. av_log(s, AV_LOG_DEBUG, "theora kfgshift %d, vrev %d\n",
  343. oggstream->kfgshift, oggstream->vrev);
  344. }
  345. }
  346. }
  347. for (j = 0; j < s->nb_streams; j++) {
  348. OGGStreamContext *oggstream = s->streams[j]->priv_data;
  349. ogg_buffer_data(s, s->streams[j], oggstream->header[0],
  350. oggstream->header_len[0], 0);
  351. oggstream->page.flags |= 2; // bos
  352. ogg_buffer_page(s, oggstream);
  353. }
  354. for (j = 0; j < s->nb_streams; j++) {
  355. AVStream *st = s->streams[j];
  356. OGGStreamContext *oggstream = st->priv_data;
  357. for (i = 1; i < 3; i++) {
  358. if (oggstream && oggstream->header_len[i])
  359. ogg_buffer_data(s, st, oggstream->header[i],
  360. oggstream->header_len[i], 0);
  361. }
  362. ogg_buffer_page(s, oggstream);
  363. }
  364. return 0;
  365. }
  366. static void ogg_write_pages(AVFormatContext *s, int flush)
  367. {
  368. OGGContext *ogg = s->priv_data;
  369. OGGPageList *next, *p;
  370. if (!ogg->page_list)
  371. return;
  372. for (p = ogg->page_list; p; ) {
  373. OGGStreamContext *oggstream =
  374. s->streams[p->page.stream_index]->priv_data;
  375. if (oggstream->page_count < 2 && !flush)
  376. break;
  377. ogg_write_page(s, &p->page,
  378. flush && oggstream->page_count == 1 ? 4 : 0); // eos
  379. next = p->next;
  380. av_freep(&p);
  381. p = next;
  382. }
  383. ogg->page_list = p;
  384. }
  385. static int ogg_write_packet(AVFormatContext *s, AVPacket *pkt)
  386. {
  387. AVStream *st = s->streams[pkt->stream_index];
  388. OGGStreamContext *oggstream = st->priv_data;
  389. int ret;
  390. int64_t granule;
  391. if (st->codec->codec_id == CODEC_ID_THEORA) {
  392. int64_t pts = oggstream->vrev < 1 ? pkt->pts : pkt->pts + pkt->duration;
  393. int pframe_count;
  394. if (pkt->flags & AV_PKT_FLAG_KEY)
  395. oggstream->last_kf_pts = pts;
  396. pframe_count = pts - oggstream->last_kf_pts;
  397. // prevent frame count from overflow if key frame flag is not set
  398. if (pframe_count >= (1<<oggstream->kfgshift)) {
  399. oggstream->last_kf_pts += pframe_count;
  400. pframe_count = 0;
  401. }
  402. granule = (oggstream->last_kf_pts<<oggstream->kfgshift) | pframe_count;
  403. } else
  404. granule = pkt->pts + pkt->duration;
  405. ret = ogg_buffer_data(s, st, pkt->data, pkt->size, granule);
  406. if (ret < 0)
  407. return ret;
  408. ogg_write_pages(s, 0);
  409. oggstream->last_granule = granule;
  410. return 0;
  411. }
  412. static int ogg_write_trailer(AVFormatContext *s)
  413. {
  414. int i;
  415. /* flush current page */
  416. for (i = 0; i < s->nb_streams; i++)
  417. ogg_buffer_page(s, s->streams[i]->priv_data);
  418. ogg_write_pages(s, 1);
  419. for (i = 0; i < s->nb_streams; i++) {
  420. AVStream *st = s->streams[i];
  421. OGGStreamContext *oggstream = st->priv_data;
  422. if (st->codec->codec_id == CODEC_ID_FLAC ||
  423. st->codec->codec_id == CODEC_ID_SPEEX) {
  424. av_free(oggstream->header[0]);
  425. av_free(oggstream->header[1]);
  426. }
  427. av_freep(&st->priv_data);
  428. }
  429. return 0;
  430. }
  431. AVOutputFormat ff_ogg_muxer = {
  432. "ogg",
  433. NULL_IF_CONFIG_SMALL("Ogg"),
  434. "application/ogg",
  435. "ogg,ogv,spx",
  436. sizeof(OGGContext),
  437. CODEC_ID_FLAC,
  438. CODEC_ID_THEORA,
  439. ogg_write_header,
  440. ogg_write_packet,
  441. ogg_write_trailer,
  442. };