FFmpeg
 All Data Structures Namespaces Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
dv.c
Go to the documentation of this file.
1 /*
2  * General DV 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 FFmpeg.
16  *
17  * FFmpeg 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  * FFmpeg 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 FFmpeg; 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/dv_profile.h"
35 #include "libavcodec/dv.h"
37 #include "libavutil/intreadwrite.h"
38 #include "libavutil/mathematics.h"
39 #include "libavutil/timecode.h"
40 #include "dv.h"
41 #include "libavutil/avassert.h"
42 
44  const AVDVProfile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
49  uint8_t audio_buf[4][8192];
50  int ach;
51  int frames;
52  uint64_t abytes;
53 };
54 
55 static inline uint16_t dv_audio_12to16(uint16_t sample)
56 {
57  uint16_t shift, result;
58 
59  sample = (sample < 0x800) ? sample : sample | 0xf000;
60  shift = (sample & 0xf00) >> 8;
61 
62  if (shift < 0x2 || shift > 0xd) {
63  result = sample;
64  } else if (shift < 0x8) {
65  shift--;
66  result = (sample - (256 * shift)) << shift;
67  } else {
68  shift = 0xe - shift;
69  result = ((sample + ((256 * shift) + 1)) << shift) - 1;
70  }
71 
72  return result;
73 }
74 
76 {
77  int offs;
78  int c;
79 
80  for (c = 0; c < 10; c++) {
81  switch (t) {
82  case dv_audio_source:
83  if (c&1) offs = (80 * 6 + 80 * 16 * 0 + 3 + c*12000);
84  else offs = (80 * 6 + 80 * 16 * 3 + 3 + c*12000);
85  break;
86  case dv_audio_control:
87  if (c&1) offs = (80 * 6 + 80 * 16 * 1 + 3 + c*12000);
88  else offs = (80 * 6 + 80 * 16 * 4 + 3 + c*12000);
89  break;
90  case dv_video_control:
91  if (c&1) offs = (80 * 3 + 8 + c*12000);
92  else offs = (80 * 5 + 48 + 5 + c*12000);
93  break;
94  case dv_timecode:
95  offs = (80*1 + 3 + 3);
96  break;
97  default:
98  return NULL;
99  }
100  if (frame[offs] == t)
101  break;
102  }
103 
104  return frame[offs] == t ? &frame[offs] : NULL;
105 }
106 
107 static const int dv_audio_frequency[3] = {
108  48000, 44100, 32000,
109 };
110 
111 /*
112  * There's a couple of assumptions being made here:
113  * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
114  * We can pass them upwards when libavcodec will be ready to deal with them.
115  * 2. We don't do software emphasis.
116  * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
117  * are converted into 16bit linear ones.
118  */
119 static int dv_extract_audio(uint8_t *frame, uint8_t **ppcm,
120  const AVDVProfile *sys)
121 {
122  int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
123  uint16_t lc, rc;
124  const uint8_t *as_pack;
125  uint8_t *pcm, ipcm;
126 
127  as_pack = dv_extract_pack(frame, dv_audio_source);
128  if (!as_pack) /* No audio ? */
129  return 0;
130 
131  smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
132  freq = as_pack[4] >> 3 & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
133  quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
134 
135  if (quant > 1)
136  return -1; /* unsupported quantization */
137 
138  if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency))
139  return AVERROR_INVALIDDATA;
140 
141  size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
142  half_ch = sys->difseg_size / 2;
143 
144  /* We work with 720p frames split in half, thus even frames have
145  * channels 0,1 and odd 2,3. */
146  ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
147 
148  if (ipcm + sys->n_difchan > (quant == 1 ? 2 : 4)) {
149  av_log(NULL, AV_LOG_ERROR, "too many dv pcm frames\n");
150  return AVERROR_INVALIDDATA;
151  }
152 
153  /* for each DIF channel */
154  for (chan = 0; chan < sys->n_difchan; chan++) {
155  av_assert0(ipcm<4);
156  pcm = ppcm[ipcm++];
157  if (!pcm)
158  break;
159 
160  /* for each DIF segment */
161  for (i = 0; i < sys->difseg_size; i++) {
162  frame += 6 * 80; /* skip DIF segment header */
163  if (quant == 1 && i == half_ch) {
164  /* next stereo channel (12bit mode only) */
165  av_assert0(ipcm<4);
166  pcm = ppcm[ipcm++];
167  if (!pcm)
168  break;
169  }
170 
171  /* for each AV sequence */
172  for (j = 0; j < 9; j++) {
173  for (d = 8; d < 80; d += 2) {
174  if (quant == 0) { /* 16bit quantization */
175  of = sys->audio_shuffle[i][j] +
176  (d - 8) / 2 * sys->audio_stride;
177  if (of * 2 >= size)
178  continue;
179 
180  /* FIXME: maybe we have to admit that DV is a
181  * big-endian PCM */
182  pcm[of * 2] = frame[d + 1];
183  pcm[of * 2 + 1] = frame[d];
184 
185  if (pcm[of * 2 + 1] == 0x80 && pcm[of * 2] == 0x00)
186  pcm[of * 2 + 1] = 0;
187  } else { /* 12bit quantization */
188  lc = ((uint16_t)frame[d] << 4) |
189  ((uint16_t)frame[d + 2] >> 4);
190  rc = ((uint16_t)frame[d + 1] << 4) |
191  ((uint16_t)frame[d + 2] & 0x0f);
192  lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
193  rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
194 
195  of = sys->audio_shuffle[i % half_ch][j] +
196  (d - 8) / 3 * sys->audio_stride;
197  if (of * 2 >= size)
198  continue;
199 
200  /* FIXME: maybe we have to admit that DV is a
201  * big-endian PCM */
202  pcm[of * 2] = lc & 0xff;
203  pcm[of * 2 + 1] = lc >> 8;
204  of = sys->audio_shuffle[i % half_ch + half_ch][j] +
205  (d - 8) / 3 * sys->audio_stride;
206  /* FIXME: maybe we have to admit that DV is a
207  * big-endian PCM */
208  pcm[of * 2] = rc & 0xff;
209  pcm[of * 2 + 1] = rc >> 8;
210  ++d;
211  }
212  }
213 
214  frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
215  }
216  }
217  }
218 
219  return size;
220 }
221 
223 {
224  const uint8_t *as_pack;
225  int freq, stype, smpls, quant, i, ach;
226 
227  as_pack = dv_extract_pack(frame, dv_audio_source);
228  if (!as_pack || !c->sys) { /* No audio ? */
229  c->ach = 0;
230  return 0;
231  }
232 
233  smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
234  freq = as_pack[4] >> 3 & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
235  stype = as_pack[3] & 0x1f; /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
236  quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
237 
238  if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency)) {
240  "Unrecognized audio sample rate index (%d)\n", freq);
241  return 0;
242  }
243 
244  if (stype > 3) {
245  av_log(c->fctx, AV_LOG_ERROR, "stype %d is invalid\n", stype);
246  c->ach = 0;
247  return 0;
248  }
249 
250  /* note: ach counts PAIRS of channels (i.e. stereo channels) */
251  ach = ((int[4]) { 1, 0, 2, 4 })[stype];
252  if (ach == 1 && quant && freq == 2)
253  ach = 2;
254 
255  /* Dynamic handling of the audio streams in DV */
256  for (i = 0; i < ach; i++) {
257  if (!c->ast[i]) {
258  c->ast[i] = avformat_new_stream(c->fctx, NULL);
259  if (!c->ast[i])
260  break;
261  avpriv_set_pts_info(c->ast[i], 64, 1, 30000);
264 
265  av_init_packet(&c->audio_pkt[i]);
266  c->audio_pkt[i].size = 0;
267  c->audio_pkt[i].data = c->audio_buf[i];
268  c->audio_pkt[i].stream_index = c->ast[i]->index;
270  }
271  c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
272  c->ast[i]->codec->channels = 2;
274  c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
275  c->ast[i]->start_time = 0;
276  }
277  c->ach = i;
278 
279  return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
280 }
281 
283 {
284  const uint8_t *vsc_pack;
285  AVCodecContext *avctx;
286  int apt, is16_9;
287  int size = 0;
288 
289  if (c->sys) {
290  avctx = c->vst->codec;
291 
293  c->sys->time_base.den);
295 
296  /* finding out SAR is a little bit messy */
297  vsc_pack = dv_extract_pack(frame, dv_video_control);
298  apt = frame[4] & 0x07;
299  is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
300  (!apt && (vsc_pack[2] & 0x07) == 0x07)));
301  c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
302  avctx->bit_rate = av_rescale_q(c->sys->frame_size,
303  (AVRational) { 8, 1 },
304  c->sys->time_base);
305  size = c->sys->frame_size;
306  }
307  return size;
308 }
309 
311 {
312  const uint8_t *tc_pack;
313 
314  // For PAL systems, drop frame bit is replaced by an arbitrary
315  // bit so its value should not be considered. Drop frame timecode
316  // is only relevant for NTSC systems.
317  int prevent_df = c->sys->ltc_divisor == 25 || c->sys->ltc_divisor == 50;
318 
319  tc_pack = dv_extract_pack(frame, dv_timecode);
320  if (!tc_pack)
321  return 0;
322  av_timecode_make_smpte_tc_string(tc, AV_RB32(tc_pack + 1), prevent_df);
323  return 1;
324 }
325 
326 /* The following 3 functions constitute our interface to the world */
327 
329 {
330  DVDemuxContext *c;
331 
332  c = av_mallocz(sizeof(DVDemuxContext));
333  if (!c)
334  return NULL;
335 
336  c->vst = avformat_new_stream(s, NULL);
337  if (!c->vst) {
338  av_free(c);
339  return NULL;
340  }
341 
342  c->fctx = s;
345  c->vst->codec->bit_rate = 25000000;
346  c->vst->start_time = 0;
347 
348  return c;
349 }
350 
352 {
353  int size = -1;
354  int i;
355 
356  for (i = 0; i < c->ach; i++) {
357  if (c->ast[i] && c->audio_pkt[i].size) {
358  *pkt = c->audio_pkt[i];
359  c->audio_pkt[i].size = 0;
360  size = pkt->size;
361  break;
362  }
363  }
364 
365  return size;
366 }
367 
369  uint8_t *buf, int buf_size, int64_t pos)
370 {
371  int size, i;
372  uint8_t *ppcm[5] = { 0 };
373 
374  if (buf_size < DV_PROFILE_BYTES ||
375  !(c->sys = av_dv_frame_profile(c->sys, buf, buf_size)) ||
376  buf_size < c->sys->frame_size) {
377  return -1; /* Broken frame, or not enough data */
378  }
379 
380  /* Queueing audio packet */
381  /* FIXME: in case of no audio/bad audio we have to do something */
382  size = dv_extract_audio_info(c, buf);
383  for (i = 0; i < c->ach; i++) {
384  c->audio_pkt[i].pos = pos;
385  c->audio_pkt[i].size = size;
386  c->audio_pkt[i].pts = c->abytes * 30000 * 8 /
387  c->ast[i]->codec->bit_rate;
388  ppcm[i] = c->audio_buf[i];
389  }
390  if (c->ach)
391  dv_extract_audio(buf, ppcm, c->sys);
392 
393  /* We work with 720p frames split in half, thus even frames have
394  * channels 0,1 and odd 2,3. */
395  if (c->sys->height == 720) {
396  if (buf[1] & 0x0C) {
397  c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
398  } else {
399  c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
400  c->abytes += size;
401  }
402  } else {
403  c->abytes += size;
404  }
405 
406  /* Now it's time to return video packet */
407  size = dv_extract_video_info(c, buf);
408  av_init_packet(pkt);
409  pkt->data = buf;
410  pkt->pos = pos;
411  pkt->size = size;
412  pkt->flags |= AV_PKT_FLAG_KEY;
413  pkt->stream_index = c->vst->index;
414  pkt->pts = c->frames;
415 
416  c->frames++;
417 
418  return size;
419 }
420 
422  int64_t timestamp, int flags)
423 {
424  // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
425  const AVDVProfile *sys = av_dv_codec_profile(c->vst->codec->width, c->vst->codec->height,
426  c->vst->codec->pix_fmt);
427  int64_t offset;
428  int64_t size = avio_size(s->pb) - s->data_offset;
429  int64_t max_offset = ((size - 1) / sys->frame_size) * sys->frame_size;
430 
431  offset = sys->frame_size * timestamp;
432 
433  if (size >= 0 && offset > max_offset)
434  offset = max_offset;
435  else if (offset < 0)
436  offset = 0;
437 
438  return offset + s->data_offset;
439 }
440 
441 void ff_dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
442 {
443  c->frames = frame_offset;
444  if (c->ach) {
445  if (c->sys) {
446  c->abytes = av_rescale_q(c->frames, c->sys->time_base,
447  (AVRational) { 8, c->ast[0]->codec->bit_rate });
448  } else
449  av_log(c->fctx, AV_LOG_ERROR, "cannot adjust audio bytes\n");
450  }
451  c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
452  c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
453 }
454 
455 /************************************************************
456  * Implementation of the easiest DV storage of all -- raw DV.
457  ************************************************************/
458 
459 typedef struct RawDVContext {
462 } RawDVContext;
463 
465  int ret;
466  char timecode[AV_TIMECODE_STR_SIZE];
467  int64_t pos = avio_tell(s->pb);
468 
469  // Read 3 DIF blocks: Header block and 2 Subcode blocks.
470  int partial_frame_size = 3 * 80;
471  uint8_t *partial_frame = av_mallocz(sizeof(*partial_frame) *
472  partial_frame_size);
473 
474  RawDVContext *c = s->priv_data;
475  ret = avio_read(s->pb, partial_frame, partial_frame_size);
476  if (ret < 0)
477  goto finish;
478 
479  if (ret < partial_frame_size) {
480  ret = -1;
481  goto finish;
482  }
483 
484  ret = dv_extract_timecode(c->dv_demux, partial_frame, timecode);
485  if (ret)
486  av_dict_set(&s->metadata, "timecode", timecode, 0);
487  else
488  av_log(s, AV_LOG_ERROR, "Detected timecode is invalid\n");
489 
490 finish:
491  av_free(partial_frame);
492  avio_seek(s->pb, pos, SEEK_SET);
493  return ret;
494 }
495 
497 {
498  unsigned state, marker_pos = 0;
499  RawDVContext *c = s->priv_data;
500 
502  if (!c->dv_demux)
503  return -1;
504 
505  state = avio_rb32(s->pb);
506  while ((state & 0xffffff7f) != 0x1f07003f) {
507  if (avio_feof(s->pb)) {
508  av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
509  return -1;
510  }
511  if (state == 0x003f0700 || state == 0xff3f0700)
512  marker_pos = avio_tell(s->pb);
513  if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
514  avio_seek(s->pb, -163, SEEK_CUR);
515  state = avio_rb32(s->pb);
516  break;
517  }
518  state = (state << 8) | avio_r8(s->pb);
519  }
520  AV_WB32(c->buf, state);
521 
522  if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) != DV_PROFILE_BYTES - 4 ||
523  avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
524  return AVERROR(EIO);
525 
527  c->buf,
529  if (!c->dv_demux->sys) {
530  av_log(s, AV_LOG_ERROR,
531  "Can't determine profile of DV input stream.\n");
532  return -1;
533  }
534 
536  (AVRational) { 8, 1 },
537  c->dv_demux->sys->time_base);
538 
539  if (s->pb->seekable)
541 
542  return 0;
543 }
544 
546 {
547  int size;
548  RawDVContext *c = s->priv_data;
549 
550  size = avpriv_dv_get_packet(c->dv_demux, pkt);
551 
552  if (size < 0) {
553  int64_t pos = avio_tell(s->pb);
554  if (!c->dv_demux->sys)
555  return AVERROR(EIO);
556  size = c->dv_demux->sys->frame_size;
557  if (avio_read(s->pb, c->buf, size) <= 0)
558  return AVERROR(EIO);
559 
560  size = avpriv_dv_produce_packet(c->dv_demux, pkt, c->buf, size, pos);
561  }
562 
563  return size;
564 }
565 
566 static int dv_read_seek(AVFormatContext *s, int stream_index,
567  int64_t timestamp, int flags)
568 {
569  RawDVContext *r = s->priv_data;
570  DVDemuxContext *c = r->dv_demux;
571  int64_t offset = dv_frame_offset(s, c, timestamp, flags);
572 
573  if (avio_seek(s->pb, offset, SEEK_SET) < 0)
574  return -1;
575 
576  ff_dv_offset_reset(c, offset / c->sys->frame_size);
577  return 0;
578 }
579 
581 {
582  RawDVContext *c = s->priv_data;
583  av_free(c->dv_demux);
584  return 0;
585 }
586 
587 static int dv_probe(AVProbeData *p)
588 {
589  unsigned marker_pos = 0;
590  int i;
591  int matches = 0;
592  int firstmatch = 0;
593  int secondary_matches = 0;
594 
595  if (p->buf_size < 5)
596  return 0;
597 
598  for (i = 0; i < p->buf_size-4; i++) {
599  unsigned state = AV_RB32(p->buf+i);
600  if ((state & 0x0007f840) == 0x00070000) {
601  // any section header, also with seq/chan num != 0,
602  // should appear around every 12000 bytes, at least 10 per frame
603  if ((state & 0xff07ff7f) == 0x1f07003f) {
604  secondary_matches++;
605  if ((state & 0xffffff7f) == 0x1f07003f) {
606  matches++;
607  if (!i)
608  firstmatch = 1;
609  }
610  }
611  if (state == 0x003f0700 || state == 0xff3f0700)
612  marker_pos = i;
613  if (state == 0xff3f0701 && i - marker_pos == 80)
614  matches++;
615  }
616  }
617 
618  if (matches && p->buf_size / matches < 1024 * 1024) {
619  if (matches > 4 || firstmatch ||
620  (secondary_matches >= 10 &&
621  p->buf_size / secondary_matches < 24000))
622  // not max to avoid dv in mov to match
623  return AVPROBE_SCORE_MAX * 3 / 4;
624  return AVPROBE_SCORE_MAX / 4;
625  }
626  return 0;
627 }
628 
630  .name = "dv",
631  .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
632  .priv_data_size = sizeof(RawDVContext),
633  .read_probe = dv_probe,
638  .extensions = "dv,dif",
639 };