FFmpeg
 All Data Structures Files Functions Variables Typedefs Enumerations Enumerator Macros Groups Pages
dv.c
Go to the documentation of this file.
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 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/dvdata.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 DVprofile* 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 
75 /*
76  * This is the dumbest implementation of all -- it simply looks at
77  * a fixed offset and if pack isn't there -- fails. We might want
78  * to have a fallback mechanism for complete search of missing packs.
79  */
81 {
82  int offs;
83 
84  switch (t) {
85  case dv_audio_source:
86  offs = (80*6 + 80*16*3 + 3);
87  break;
88  case dv_audio_control:
89  offs = (80*6 + 80*16*4 + 3);
90  break;
91  case dv_video_control:
92  offs = (80*5 + 48 + 5);
93  break;
94  case dv_timecode:
95  offs = (80*1 + 3 + 3);
96  break;
97  default:
98  return NULL;
99  }
100 
101  return frame[offs] == t ? &frame[offs] : NULL;
102 }
103 
104 static const int dv_audio_frequency[3] = {
105  48000, 44100, 32000,
106 };
107 
108 /*
109  * There's a couple of assumptions being made here:
110  * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
111  * We can pass them upwards when libavcodec will be ready to deal with them.
112  * 2. We don't do software emphasis.
113  * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
114  * are converted into 16bit linear ones.
115  */
116 static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4],
117  const DVprofile *sys)
118 {
119  int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
120  uint16_t lc, rc;
121  const uint8_t* as_pack;
122  uint8_t *pcm, ipcm;
123 
124  as_pack = dv_extract_pack(frame, dv_audio_source);
125  if (!as_pack) /* No audio ? */
126  return 0;
127 
128  smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
129  freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
130  quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
131 
132  if (quant > 1)
133  return -1; /* unsupported quantization */
134 
135  if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency))
136  return AVERROR_INVALIDDATA;
137 
138  size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
139  half_ch = sys->difseg_size / 2;
140 
141  /* We work with 720p frames split in half, thus even frames have
142  * channels 0,1 and odd 2,3. */
143  ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
144 
145  if (ipcm + sys->n_difchan > (quant == 1 ? 2 : 4)) {
146  av_log(NULL, AV_LOG_ERROR, "too many dv pcm frames\n");
147  return AVERROR_INVALIDDATA;
148  }
149 
150  /* for each DIF channel */
151  for (chan = 0; chan < sys->n_difchan; chan++) {
152  av_assert0(ipcm<4);
153  pcm = ppcm[ipcm++];
154  if (!pcm)
155  break;
156 
157  /* for each DIF segment */
158  for (i = 0; i < sys->difseg_size; i++) {
159  frame += 6 * 80; /* skip DIF segment header */
160  if (quant == 1 && i == half_ch) {
161  /* next stereo channel (12bit mode only) */
162  av_assert0(ipcm<4);
163  pcm = ppcm[ipcm++];
164  if (!pcm)
165  break;
166  }
167 
168  /* for each AV sequence */
169  for (j = 0; j < 9; j++) {
170  for (d = 8; d < 80; d += 2) {
171  if (quant == 0) { /* 16bit quantization */
172  of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride;
173  if (of*2 >= size)
174  continue;
175 
176  pcm[of*2] = frame[d+1]; // FIXME: maybe we have to admit
177  pcm[of*2+1] = frame[d]; // that DV is a big-endian PCM
178  if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
179  pcm[of*2+1] = 0;
180  } else { /* 12bit quantization */
181  lc = ((uint16_t)frame[d] << 4) |
182  ((uint16_t)frame[d+2] >> 4);
183  rc = ((uint16_t)frame[d+1] << 4) |
184  ((uint16_t)frame[d+2] & 0x0f);
185  lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
186  rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
187 
188  of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride;
189  if (of*2 >= size)
190  continue;
191 
192  pcm[of*2] = lc & 0xff; // FIXME: maybe we have to admit
193  pcm[of*2+1] = lc >> 8; // that DV is a big-endian PCM
194  of = sys->audio_shuffle[i%half_ch+half_ch][j] +
195  (d - 8) / 3 * sys->audio_stride;
196  pcm[of*2] = rc & 0xff; // FIXME: maybe we have to admit
197  pcm[of*2+1] = rc >> 8; // that DV is a big-endian PCM
198  ++d;
199  }
200  }
201 
202  frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
203  }
204  }
205  }
206 
207  return size;
208 }
209 
211 {
212  const uint8_t* as_pack;
213  int freq, stype, smpls, quant, i, ach;
214 
215  as_pack = dv_extract_pack(frame, dv_audio_source);
216  if (!as_pack || !c->sys) { /* No audio ? */
217  c->ach = 0;
218  return 0;
219  }
220 
221  smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
222  freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
223  stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
224  quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
225 
226  if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency)) {
228  "Unrecognized audio sample rate index (%d)\n", freq);
229  return 0;
230  }
231 
232  if (stype > 3) {
233  av_log(c->fctx, AV_LOG_ERROR, "stype %d is invalid\n", stype);
234  c->ach = 0;
235  return 0;
236  }
237 
238  /* note: ach counts PAIRS of channels (i.e. stereo channels) */
239  ach = ((int[4]){ 1, 0, 2, 4})[stype];
240  if (ach == 1 && quant && freq == 2)
241  ach = 2;
242 
243  /* Dynamic handling of the audio streams in DV */
244  for (i = 0; i < ach; i++) {
245  if (!c->ast[i]) {
246  c->ast[i] = avformat_new_stream(c->fctx, NULL);
247  if (!c->ast[i])
248  break;
249  avpriv_set_pts_info(c->ast[i], 64, 1, 30000);
252 
253  av_init_packet(&c->audio_pkt[i]);
254  c->audio_pkt[i].size = 0;
255  c->audio_pkt[i].data = c->audio_buf[i];
256  c->audio_pkt[i].stream_index = c->ast[i]->index;
258  }
259  c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
260  c->ast[i]->codec->channels = 2;
262  c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
263  c->ast[i]->start_time = 0;
264  }
265  c->ach = i;
266 
267  return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
268 }
269 
271 {
272  const uint8_t* vsc_pack;
273  AVCodecContext* avctx;
274  int apt, is16_9;
275  int size = 0;
276 
277  if (c->sys) {
278  avctx = c->vst->codec;
279 
281  c->sys->time_base.den);
282  avctx->time_base= c->sys->time_base;
283 
284  /* finding out SAR is a little bit messy */
285  vsc_pack = dv_extract_pack(frame, dv_video_control);
286  apt = frame[4] & 0x07;
287  is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
288  (!apt && (vsc_pack[2] & 0x07) == 0x07)));
289  c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
290  avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1},
291  c->sys->time_base);
292  size = c->sys->frame_size;
293  }
294  return size;
295 }
296 
298 {
299  const uint8_t *tc_pack;
300 
301  // For PAL systems, drop frame bit is replaced by an arbitrary
302  // bit so its value should not be considered. Drop frame timecode
303  // is only relevant for NTSC systems.
304  int prevent_df = c->sys->ltc_divisor == 25 || c->sys->ltc_divisor == 50;
305 
306  tc_pack = dv_extract_pack(frame, dv_timecode);
307  if (!tc_pack)
308  return 0;
309  av_timecode_make_smpte_tc_string(tc, AV_RB32(tc_pack + 1), prevent_df);
310  return 1;
311 }
312 
313 /*
314  * The following 3 functions constitute our interface to the world
315  */
316 
318 {
319  DVDemuxContext *c;
320 
321  c = av_mallocz(sizeof(DVDemuxContext));
322  if (!c)
323  return NULL;
324 
325  c->vst = avformat_new_stream(s, NULL);
326  if (!c->vst) {
327  av_free(c);
328  return NULL;
329  }
330 
331  c->fctx = s;
334  c->vst->codec->bit_rate = 25000000;
335  c->vst->start_time = 0;
336 
337  return c;
338 }
339 
341 {
342  int size = -1;
343  int i;
344 
345  for (i = 0; i < c->ach; i++) {
346  if (c->ast[i] && c->audio_pkt[i].size) {
347  *pkt = c->audio_pkt[i];
348  c->audio_pkt[i].size = 0;
349  size = pkt->size;
350  break;
351  }
352  }
353 
354  return size;
355 }
356 
358  uint8_t* buf, int buf_size, int64_t pos)
359 {
360  int size, i;
361  uint8_t *ppcm[4] = {0};
362 
363  if (buf_size < DV_PROFILE_BYTES ||
364  !(c->sys = avpriv_dv_frame_profile(c->sys, buf, buf_size)) ||
365  buf_size < c->sys->frame_size) {
366  return -1; /* Broken frame, or not enough data */
367  }
368 
369  /* Queueing audio packet */
370  /* FIXME: in case of no audio/bad audio we have to do something */
371  size = dv_extract_audio_info(c, buf);
372  for (i = 0; i < c->ach; i++) {
373  c->audio_pkt[i].pos = pos;
374  c->audio_pkt[i].size = size;
375  c->audio_pkt[i].pts = c->abytes * 30000 * 8 / c->ast[i]->codec->bit_rate;
376  ppcm[i] = c->audio_buf[i];
377  }
378  if (c->ach)
379  dv_extract_audio(buf, ppcm, c->sys);
380 
381  /* We work with 720p frames split in half, thus even frames have
382  * channels 0,1 and odd 2,3. */
383  if (c->sys->height == 720) {
384  if (buf[1] & 0x0C) {
385  c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
386  } else {
387  c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
388  c->abytes += size;
389  }
390  } else {
391  c->abytes += size;
392  }
393 
394  /* Now it's time to return video packet */
395  size = dv_extract_video_info(c, buf);
396  av_init_packet(pkt);
397  pkt->data = buf;
398  pkt->pos = pos;
399  pkt->size = size;
400  pkt->flags |= AV_PKT_FLAG_KEY;
401  pkt->stream_index = c->vst->index;
402  pkt->pts = c->frames;
403 
404  c->frames++;
405 
406  return size;
407 }
408 
410  int64_t timestamp, int flags)
411 {
412  // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
413  const DVprofile* sys = avpriv_dv_codec_profile(c->vst->codec);
414  int64_t offset;
415  int64_t size = avio_size(s->pb) - s->data_offset;
416  int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
417 
418  offset = sys->frame_size * timestamp;
419 
420  if (size >= 0 && offset > max_offset) offset = max_offset;
421  else if (offset < 0) offset = 0;
422 
423  return offset + s->data_offset;
424 }
425 
426 void ff_dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
427 {
428  c->frames= frame_offset;
429  if (c->ach) {
430  if (c->sys) {
431  c->abytes= av_rescale_q(c->frames, c->sys->time_base,
432  (AVRational){8, c->ast[0]->codec->bit_rate});
433  }else
434  av_log(c->fctx, AV_LOG_ERROR, "cannot adjust audio bytes\n");
435  }
436  c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
437  c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
438 }
439 
440 /************************************************************
441  * Implementation of the easiest DV storage of all -- raw DV.
442  ************************************************************/
443 
444 typedef struct RawDVContext {
447 } RawDVContext;
448 
450  int ret;
451  char timecode[AV_TIMECODE_STR_SIZE];
452  int64_t pos = avio_tell(s->pb);
453 
454  // Read 3 DIF blocks: Header block and 2 Subcode blocks.
455  int partial_frame_size = 3 * 80;
456  uint8_t *partial_frame = av_mallocz(sizeof(*partial_frame) *
457  partial_frame_size);
458 
459  RawDVContext *c = s->priv_data;
460  ret = avio_read(s->pb, partial_frame, partial_frame_size);
461  if (ret < 0)
462  goto finish;
463 
464  if (ret < partial_frame_size) {
465  ret = -1;
466  goto finish;
467  }
468 
469  ret = dv_extract_timecode(c->dv_demux, partial_frame, timecode);
470  if (ret)
471  av_dict_set(&s->metadata, "timecode", timecode, 0);
472  else
473  av_log(s, AV_LOG_ERROR, "Detected timecode is invalid\n");
474 
475 finish:
476  av_free(partial_frame);
477  avio_seek(s->pb, pos, SEEK_SET);
478  return ret;
479 }
480 
482 {
483  unsigned state, marker_pos = 0;
484  RawDVContext *c = s->priv_data;
485 
487  if (!c->dv_demux)
488  return -1;
489 
490  state = avio_rb32(s->pb);
491  while ((state & 0xffffff7f) != 0x1f07003f) {
492  if (url_feof(s->pb)) {
493  av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
494  return -1;
495  }
496  if (state == 0x003f0700 || state == 0xff3f0700)
497  marker_pos = avio_tell(s->pb);
498  if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
499  avio_seek(s->pb, -163, SEEK_CUR);
500  state = avio_rb32(s->pb);
501  break;
502  }
503  state = (state << 8) | avio_r8(s->pb);
504  }
505  AV_WB32(c->buf, state);
506 
507  if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) != DV_PROFILE_BYTES - 4 ||
508  avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
509  return AVERROR(EIO);
510 
512  if (!c->dv_demux->sys) {
513  av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n");
514  return -1;
515  }
516 
518  c->dv_demux->sys->time_base);
519 
520  if (s->pb->seekable)
521  dv_read_timecode(s);
522 
523  return 0;
524 }
525 
526 
528 {
529  int size;
530  RawDVContext *c = s->priv_data;
531 
532  size = avpriv_dv_get_packet(c->dv_demux, pkt);
533 
534  if (size < 0) {
535  int64_t pos = avio_tell(s->pb);
536  if (!c->dv_demux->sys)
537  return AVERROR(EIO);
538  size = c->dv_demux->sys->frame_size;
539  if (avio_read(s->pb, c->buf, size) <= 0)
540  return AVERROR(EIO);
541 
542  size = avpriv_dv_produce_packet(c->dv_demux, pkt, c->buf, size, pos);
543  }
544 
545  return size;
546 }
547 
548 static int dv_read_seek(AVFormatContext *s, int stream_index,
549  int64_t timestamp, int flags)
550 {
551  RawDVContext *r = s->priv_data;
552  DVDemuxContext *c = r->dv_demux;
553  int64_t offset = dv_frame_offset(s, c, timestamp, flags);
554 
555  if (avio_seek(s->pb, offset, SEEK_SET) < 0)
556  return -1;
557 
558  ff_dv_offset_reset(c, offset / c->sys->frame_size);
559  return 0;
560 }
561 
563 {
564  RawDVContext *c = s->priv_data;
565  av_free(c->dv_demux);
566  return 0;
567 }
568 
569 static int dv_probe(AVProbeData *p)
570 {
571  unsigned state, marker_pos = 0;
572  int i;
573  int matches = 0;
574  int secondary_matches = 0;
575 
576  if (p->buf_size < 5)
577  return 0;
578 
579  state = AV_RB32(p->buf);
580  for (i = 4; i < p->buf_size; i++) {
581  if ((state & 0xffffff7f) == 0x1f07003f)
582  matches++;
583  // any section header, also with seq/chan num != 0,
584  // should appear around every 12000 bytes, at least 10 per frame
585  if ((state & 0xff07ff7f) == 0x1f07003f)
586  secondary_matches++;
587  if (state == 0x003f0700 || state == 0xff3f0700)
588  marker_pos = i;
589  if (state == 0xff3f0701 && i - marker_pos == 80)
590  matches++;
591  state = (state << 8) | p->buf[i];
592  }
593 
594  if (matches && p->buf_size / matches < 1024*1024) {
595  if (matches > 4 || (secondary_matches >= 10 && p->buf_size / secondary_matches < 24000))
596  return AVPROBE_SCORE_MAX*3/4; // not max to avoid dv in mov to match
597  return AVPROBE_SCORE_MAX/4;
598  }
599  return 0;
600 }
601 
602 #if CONFIG_DV_DEMUXER
603 AVInputFormat ff_dv_demuxer = {
604  .name = "dv",
605  .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
606  .priv_data_size = sizeof(RawDVContext),
607  .read_probe = dv_probe,
612  .extensions = "dv,dif",
613 };
614 #endif