You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

540 lines
17KB

  1. /*
  2. * General DV muxer/demuxer
  3. * Copyright (c) 2003 Roman Shaposhnik
  4. *
  5. * Many thanks to Dan Dennedy <dan@dennedy.org> for providing wealth
  6. * of DV technical info.
  7. *
  8. * Raw DV format
  9. * Copyright (c) 2002 Fabrice Bellard
  10. *
  11. * 50 Mbps (DVCPRO50) and 100 Mbps (DVCPRO HD) support
  12. * Copyright (c) 2006 Daniel Maas <dmaas@maasdigital.com>
  13. * Funded by BBC Research & Development
  14. *
  15. * This file is part of Libav.
  16. *
  17. * Libav is free software; you can redistribute it and/or
  18. * modify it under the terms of the GNU Lesser General Public
  19. * License as published by the Free Software Foundation; either
  20. * version 2.1 of the License, or (at your option) any later version.
  21. *
  22. * Libav is distributed in the hope that it will be useful,
  23. * but WITHOUT ANY WARRANTY; without even the implied warranty of
  24. * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
  25. * Lesser General Public License for more details.
  26. *
  27. * You should have received a copy of the GNU Lesser General Public
  28. * License along with Libav; if not, write to the Free Software
  29. * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  30. */
  31. #include <time.h>
  32. #include "avformat.h"
  33. #include "internal.h"
  34. #include "libavcodec/dvdata.h"
  35. #include "libavutil/intreadwrite.h"
  36. #include "libavutil/mathematics.h"
  37. #include "dv.h"
  38. struct DVDemuxContext {
  39. const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
  40. AVFormatContext* fctx;
  41. AVStream* vst;
  42. AVStream* ast[4];
  43. AVPacket audio_pkt[4];
  44. uint8_t audio_buf[4][8192];
  45. int ach;
  46. int frames;
  47. uint64_t abytes;
  48. };
  49. static inline uint16_t dv_audio_12to16(uint16_t sample)
  50. {
  51. uint16_t shift, result;
  52. sample = (sample < 0x800) ? sample : sample | 0xf000;
  53. shift = (sample & 0xf00) >> 8;
  54. if (shift < 0x2 || shift > 0xd) {
  55. result = sample;
  56. } else if (shift < 0x8) {
  57. shift--;
  58. result = (sample - (256 * shift)) << shift;
  59. } else {
  60. shift = 0xe - shift;
  61. result = ((sample + ((256 * shift) + 1)) << shift) - 1;
  62. }
  63. return result;
  64. }
  65. /*
  66. * This is the dumbest implementation of all -- it simply looks at
  67. * a fixed offset and if pack isn't there -- fails. We might want
  68. * to have a fallback mechanism for complete search of missing packs.
  69. */
  70. static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
  71. {
  72. int offs;
  73. switch (t) {
  74. case dv_audio_source:
  75. offs = (80*6 + 80*16*3 + 3);
  76. break;
  77. case dv_audio_control:
  78. offs = (80*6 + 80*16*4 + 3);
  79. break;
  80. case dv_video_control:
  81. offs = (80*5 + 48 + 5);
  82. break;
  83. default:
  84. return NULL;
  85. }
  86. return frame[offs] == t ? &frame[offs] : NULL;
  87. }
  88. /*
  89. * There's a couple of assumptions being made here:
  90. * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
  91. * We can pass them upwards when libavcodec will be ready to deal with them.
  92. * 2. We don't do software emphasis.
  93. * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
  94. * are converted into 16bit linear ones.
  95. */
  96. static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4],
  97. const DVprofile *sys)
  98. {
  99. int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
  100. uint16_t lc, rc;
  101. const uint8_t* as_pack;
  102. uint8_t *pcm, ipcm;
  103. as_pack = dv_extract_pack(frame, dv_audio_source);
  104. if (!as_pack) /* No audio ? */
  105. return 0;
  106. smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
  107. freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
  108. quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
  109. if (quant > 1)
  110. return -1; /* unsupported quantization */
  111. size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
  112. half_ch = sys->difseg_size / 2;
  113. /* We work with 720p frames split in half, thus even frames have
  114. * channels 0,1 and odd 2,3. */
  115. ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
  116. /* for each DIF channel */
  117. for (chan = 0; chan < sys->n_difchan; chan++) {
  118. /* next stereo channel (50Mbps and 100Mbps only) */
  119. pcm = ppcm[ipcm++];
  120. if (!pcm)
  121. break;
  122. /* for each DIF segment */
  123. for (i = 0; i < sys->difseg_size; i++) {
  124. frame += 6 * 80; /* skip DIF segment header */
  125. if (quant == 1 && i == half_ch) {
  126. /* next stereo channel (12bit mode only) */
  127. pcm = ppcm[ipcm++];
  128. if (!pcm)
  129. break;
  130. }
  131. /* for each AV sequence */
  132. for (j = 0; j < 9; j++) {
  133. for (d = 8; d < 80; d += 2) {
  134. if (quant == 0) { /* 16bit quantization */
  135. of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride;
  136. if (of*2 >= size)
  137. continue;
  138. pcm[of*2] = frame[d+1]; // FIXME: maybe we have to admit
  139. pcm[of*2+1] = frame[d]; // that DV is a big-endian PCM
  140. if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
  141. pcm[of*2+1] = 0;
  142. } else { /* 12bit quantization */
  143. lc = ((uint16_t)frame[d] << 4) |
  144. ((uint16_t)frame[d+2] >> 4);
  145. rc = ((uint16_t)frame[d+1] << 4) |
  146. ((uint16_t)frame[d+2] & 0x0f);
  147. lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
  148. rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
  149. of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride;
  150. if (of*2 >= size)
  151. continue;
  152. pcm[of*2] = lc & 0xff; // FIXME: maybe we have to admit
  153. pcm[of*2+1] = lc >> 8; // that DV is a big-endian PCM
  154. of = sys->audio_shuffle[i%half_ch+half_ch][j] +
  155. (d - 8) / 3 * sys->audio_stride;
  156. pcm[of*2] = rc & 0xff; // FIXME: maybe we have to admit
  157. pcm[of*2+1] = rc >> 8; // that DV is a big-endian PCM
  158. ++d;
  159. }
  160. }
  161. frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
  162. }
  163. }
  164. }
  165. return size;
  166. }
  167. static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
  168. {
  169. const uint8_t* as_pack;
  170. int freq, stype, smpls, quant, i, ach;
  171. as_pack = dv_extract_pack(frame, dv_audio_source);
  172. if (!as_pack || !c->sys) { /* No audio ? */
  173. c->ach = 0;
  174. return 0;
  175. }
  176. smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
  177. freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
  178. stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
  179. quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
  180. if (stype > 3) {
  181. av_log(c->fctx, AV_LOG_ERROR, "stype %d is invalid\n", stype);
  182. c->ach = 0;
  183. return 0;
  184. }
  185. /* note: ach counts PAIRS of channels (i.e. stereo channels) */
  186. ach = ((int[4]){ 1, 0, 2, 4})[stype];
  187. if (ach == 1 && quant && freq == 2)
  188. ach = 2;
  189. /* Dynamic handling of the audio streams in DV */
  190. for (i = 0; i < ach; i++) {
  191. if (!c->ast[i]) {
  192. c->ast[i] = avformat_new_stream(c->fctx, NULL);
  193. if (!c->ast[i])
  194. break;
  195. avpriv_set_pts_info(c->ast[i], 64, 1, 30000);
  196. c->ast[i]->codec->codec_type = AVMEDIA_TYPE_AUDIO;
  197. c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE;
  198. av_init_packet(&c->audio_pkt[i]);
  199. c->audio_pkt[i].size = 0;
  200. c->audio_pkt[i].data = c->audio_buf[i];
  201. c->audio_pkt[i].stream_index = c->ast[i]->index;
  202. c->audio_pkt[i].flags |= AV_PKT_FLAG_KEY;
  203. }
  204. c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
  205. c->ast[i]->codec->channels = 2;
  206. c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
  207. c->ast[i]->start_time = 0;
  208. }
  209. c->ach = i;
  210. return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
  211. }
  212. static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
  213. {
  214. const uint8_t* vsc_pack;
  215. AVCodecContext* avctx;
  216. int apt, is16_9;
  217. int size = 0;
  218. if (c->sys) {
  219. avctx = c->vst->codec;
  220. avpriv_set_pts_info(c->vst, 64, c->sys->time_base.num,
  221. c->sys->time_base.den);
  222. avctx->time_base= c->sys->time_base;
  223. if (!avctx->width){
  224. avctx->width = c->sys->width;
  225. avctx->height = c->sys->height;
  226. }
  227. avctx->pix_fmt = c->sys->pix_fmt;
  228. /* finding out SAR is a little bit messy */
  229. vsc_pack = dv_extract_pack(frame, dv_video_control);
  230. apt = frame[4] & 0x07;
  231. is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
  232. (!apt && (vsc_pack[2] & 0x07) == 0x07)));
  233. c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
  234. avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1},
  235. c->sys->time_base);
  236. size = c->sys->frame_size;
  237. }
  238. return size;
  239. }
  240. /*
  241. * The following 3 functions constitute our interface to the world
  242. */
  243. DVDemuxContext* avpriv_dv_init_demux(AVFormatContext *s)
  244. {
  245. DVDemuxContext *c;
  246. c = av_mallocz(sizeof(DVDemuxContext));
  247. if (!c)
  248. return NULL;
  249. c->vst = avformat_new_stream(s, NULL);
  250. if (!c->vst) {
  251. av_free(c);
  252. return NULL;
  253. }
  254. c->sys = NULL;
  255. c->fctx = s;
  256. memset(c->ast, 0, sizeof(c->ast));
  257. c->ach = 0;
  258. c->frames = 0;
  259. c->abytes = 0;
  260. c->vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
  261. c->vst->codec->codec_id = CODEC_ID_DVVIDEO;
  262. c->vst->codec->bit_rate = 25000000;
  263. c->vst->start_time = 0;
  264. return c;
  265. }
  266. int avpriv_dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
  267. {
  268. int size = -1;
  269. int i;
  270. for (i = 0; i < c->ach; i++) {
  271. if (c->ast[i] && c->audio_pkt[i].size) {
  272. *pkt = c->audio_pkt[i];
  273. c->audio_pkt[i].size = 0;
  274. size = pkt->size;
  275. break;
  276. }
  277. }
  278. return size;
  279. }
  280. int avpriv_dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
  281. uint8_t* buf, int buf_size)
  282. {
  283. int size, i;
  284. uint8_t *ppcm[4] = {0};
  285. if (buf_size < DV_PROFILE_BYTES ||
  286. !(c->sys = avpriv_dv_frame_profile(c->sys, buf, buf_size)) ||
  287. buf_size < c->sys->frame_size) {
  288. return -1; /* Broken frame, or not enough data */
  289. }
  290. /* Queueing audio packet */
  291. /* FIXME: in case of no audio/bad audio we have to do something */
  292. size = dv_extract_audio_info(c, buf);
  293. for (i = 0; i < c->ach; i++) {
  294. c->audio_pkt[i].size = size;
  295. c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate;
  296. ppcm[i] = c->audio_buf[i];
  297. }
  298. if (c->ach)
  299. dv_extract_audio(buf, ppcm, c->sys);
  300. /* We work with 720p frames split in half, thus even frames have
  301. * channels 0,1 and odd 2,3. */
  302. if (c->sys->height == 720) {
  303. if (buf[1] & 0x0C) {
  304. c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
  305. } else {
  306. c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
  307. c->abytes += size;
  308. }
  309. } else {
  310. c->abytes += size;
  311. }
  312. /* Now it's time to return video packet */
  313. size = dv_extract_video_info(c, buf);
  314. av_init_packet(pkt);
  315. pkt->data = buf;
  316. pkt->size = size;
  317. pkt->flags |= AV_PKT_FLAG_KEY;
  318. pkt->stream_index = c->vst->id;
  319. pkt->pts = c->frames;
  320. c->frames++;
  321. return size;
  322. }
  323. static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
  324. int64_t timestamp, int flags)
  325. {
  326. // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
  327. const DVprofile* sys = avpriv_dv_codec_profile(c->vst->codec);
  328. int64_t offset;
  329. int64_t size = avio_size(s->pb) - s->data_offset;
  330. int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
  331. offset = sys->frame_size * timestamp;
  332. if (size >= 0 && offset > max_offset) offset = max_offset;
  333. else if (offset < 0) offset = 0;
  334. return offset + s->data_offset;
  335. }
  336. void dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
  337. {
  338. c->frames= frame_offset;
  339. if (c->ach)
  340. c->abytes= av_rescale_q(c->frames, c->sys->time_base,
  341. (AVRational){8, c->ast[0]->codec->bit_rate});
  342. c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
  343. c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
  344. }
  345. /************************************************************
  346. * Implementation of the easiest DV storage of all -- raw DV.
  347. ************************************************************/
  348. typedef struct RawDVContext {
  349. DVDemuxContext* dv_demux;
  350. uint8_t buf[DV_MAX_FRAME_SIZE];
  351. } RawDVContext;
  352. static int dv_read_header(AVFormatContext *s)
  353. {
  354. unsigned state, marker_pos = 0;
  355. RawDVContext *c = s->priv_data;
  356. c->dv_demux = avpriv_dv_init_demux(s);
  357. if (!c->dv_demux)
  358. return -1;
  359. state = avio_rb32(s->pb);
  360. while ((state & 0xffffff7f) != 0x1f07003f) {
  361. if (s->pb->eof_reached) {
  362. av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
  363. return -1;
  364. }
  365. if (state == 0x003f0700 || state == 0xff3f0700)
  366. marker_pos = avio_tell(s->pb);
  367. if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
  368. avio_seek(s->pb, -163, SEEK_CUR);
  369. state = avio_rb32(s->pb);
  370. break;
  371. }
  372. state = (state << 8) | avio_r8(s->pb);
  373. }
  374. AV_WB32(c->buf, state);
  375. if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) <= 0 ||
  376. avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
  377. return AVERROR(EIO);
  378. c->dv_demux->sys = avpriv_dv_frame_profile(c->dv_demux->sys, c->buf, DV_PROFILE_BYTES);
  379. if (!c->dv_demux->sys) {
  380. av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n");
  381. return -1;
  382. }
  383. s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size, (AVRational){8,1},
  384. c->dv_demux->sys->time_base);
  385. return 0;
  386. }
  387. static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
  388. {
  389. int size;
  390. RawDVContext *c = s->priv_data;
  391. size = avpriv_dv_get_packet(c->dv_demux, pkt);
  392. if (size < 0) {
  393. if (!c->dv_demux->sys)
  394. return AVERROR(EIO);
  395. size = c->dv_demux->sys->frame_size;
  396. if (avio_read(s->pb, c->buf, size) <= 0)
  397. return AVERROR(EIO);
  398. size = avpriv_dv_produce_packet(c->dv_demux, pkt, c->buf, size);
  399. }
  400. return size;
  401. }
  402. static int dv_read_seek(AVFormatContext *s, int stream_index,
  403. int64_t timestamp, int flags)
  404. {
  405. RawDVContext *r = s->priv_data;
  406. DVDemuxContext *c = r->dv_demux;
  407. int64_t offset = dv_frame_offset(s, c, timestamp, flags);
  408. dv_offset_reset(c, offset / c->sys->frame_size);
  409. offset = avio_seek(s->pb, offset, SEEK_SET);
  410. return (offset < 0) ? offset : 0;
  411. }
  412. static int dv_read_close(AVFormatContext *s)
  413. {
  414. RawDVContext *c = s->priv_data;
  415. av_free(c->dv_demux);
  416. return 0;
  417. }
  418. static int dv_probe(AVProbeData *p)
  419. {
  420. unsigned state, marker_pos = 0;
  421. int i;
  422. int matches = 0;
  423. int secondary_matches = 0;
  424. if (p->buf_size < 5)
  425. return 0;
  426. state = AV_RB32(p->buf);
  427. for (i = 4; i < p->buf_size; i++) {
  428. if ((state & 0xffffff7f) == 0x1f07003f)
  429. matches++;
  430. // any section header, also with seq/chan num != 0,
  431. // should appear around every 12000 bytes, at least 10 per frame
  432. if ((state & 0xff07ff7f) == 0x1f07003f)
  433. secondary_matches++;
  434. if (state == 0x003f0700 || state == 0xff3f0700)
  435. marker_pos = i;
  436. if (state == 0xff3f0701 && i - marker_pos == 80)
  437. matches++;
  438. state = (state << 8) | p->buf[i];
  439. }
  440. if (matches && p->buf_size / matches < 1024*1024) {
  441. if (matches > 4 || (secondary_matches >= 10 && p->buf_size / secondary_matches < 24000))
  442. return AVPROBE_SCORE_MAX*3/4; // not max to avoid dv in mov to match
  443. return AVPROBE_SCORE_MAX/4;
  444. }
  445. return 0;
  446. }
  447. #if CONFIG_DV_DEMUXER
  448. AVInputFormat ff_dv_demuxer = {
  449. .name = "dv",
  450. .long_name = NULL_IF_CONFIG_SMALL("DV video format"),
  451. .priv_data_size = sizeof(RawDVContext),
  452. .read_probe = dv_probe,
  453. .read_header = dv_read_header,
  454. .read_packet = dv_read_packet,
  455. .read_close = dv_read_close,
  456. .read_seek = dv_read_seek,
  457. .extensions = "dv,dif",
  458. };
  459. #endif