lavc: export DV profile API used by muxer/demuxer as public
[libav.git] / libavformat / dv.c
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 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/dv_profile.h"
35 #include "libavcodec/dv.h"
36 #include "libavutil/channel_layout.h"
37 #include "libavutil/intreadwrite.h"
38 #include "libavutil/mathematics.h"
39 #include "dv.h"
40
41 struct DVDemuxContext {
42 const AVDVProfile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
43 AVFormatContext* fctx;
44 AVStream* vst;
45 AVStream* ast[4];
46 AVPacket audio_pkt[4];
47 uint8_t audio_buf[4][8192];
48 int ach;
49 int frames;
50 uint64_t abytes;
51 };
52
53 static inline uint16_t dv_audio_12to16(uint16_t sample)
54 {
55 uint16_t shift, result;
56
57 sample = (sample < 0x800) ? sample : sample | 0xf000;
58 shift = (sample & 0xf00) >> 8;
59
60 if (shift < 0x2 || shift > 0xd) {
61 result = sample;
62 } else if (shift < 0x8) {
63 shift--;
64 result = (sample - (256 * shift)) << shift;
65 } else {
66 shift = 0xe - shift;
67 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
68 }
69
70 return result;
71 }
72
73 /*
74 * This is the dumbest implementation of all -- it simply looks at
75 * a fixed offset and if pack isn't there -- fails. We might want
76 * to have a fallback mechanism for complete search of missing packs.
77 */
78 static const uint8_t *dv_extract_pack(uint8_t *frame, enum dv_pack_type t)
79 {
80 int offs;
81
82 switch (t) {
83 case dv_audio_source:
84 offs = (80 * 6 + 80 * 16 * 3 + 3);
85 break;
86 case dv_audio_control:
87 offs = (80 * 6 + 80 * 16 * 4 + 3);
88 break;
89 case dv_video_control:
90 offs = (80 * 5 + 48 + 5);
91 break;
92 default:
93 return NULL;
94 }
95
96 return frame[offs] == t ? &frame[offs] : NULL;
97 }
98
99 static const int dv_audio_frequency[3] = {
100 48000, 44100, 32000,
101 };
102
103 /*
104 * There's a couple of assumptions being made here:
105 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
106 * We can pass them upwards when libavcodec will be ready to deal with them.
107 * 2. We don't do software emphasis.
108 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
109 * are converted into 16bit linear ones.
110 */
111 static int dv_extract_audio(uint8_t *frame, uint8_t **ppcm,
112 const AVDVProfile *sys)
113 {
114 int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
115 uint16_t lc, rc;
116 const uint8_t *as_pack;
117 uint8_t *pcm, ipcm;
118
119 as_pack = dv_extract_pack(frame, dv_audio_source);
120 if (!as_pack) /* No audio ? */
121 return 0;
122
123 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
124 freq = as_pack[4] >> 3 & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
125 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
126
127 if (quant > 1)
128 return -1; /* unsupported quantization */
129
130 if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency))
131 return AVERROR_INVALIDDATA;
132
133 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
134 half_ch = sys->difseg_size / 2;
135
136 /* We work with 720p frames split in half, thus even frames have
137 * channels 0,1 and odd 2,3. */
138 ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
139
140 /* for each DIF channel */
141 for (chan = 0; chan < sys->n_difchan; chan++) {
142 /* next stereo channel (50Mbps and 100Mbps only) */
143 pcm = ppcm[ipcm++];
144 if (!pcm)
145 break;
146
147 /* for each DIF segment */
148 for (i = 0; i < sys->difseg_size; i++) {
149 frame += 6 * 80; /* skip DIF segment header */
150 if (quant == 1 && i == half_ch) {
151 /* next stereo channel (12bit mode only) */
152 pcm = ppcm[ipcm++];
153 if (!pcm)
154 break;
155 }
156
157 /* for each AV sequence */
158 for (j = 0; j < 9; j++) {
159 for (d = 8; d < 80; d += 2) {
160 if (quant == 0) { /* 16bit quantization */
161 of = sys->audio_shuffle[i][j] +
162 (d - 8) / 2 * sys->audio_stride;
163 if (of * 2 >= size)
164 continue;
165
166 /* FIXME: maybe we have to admit that DV is a
167 * big-endian PCM */
168 pcm[of * 2] = frame[d + 1];
169 pcm[of * 2 + 1] = frame[d];
170
171 if (pcm[of * 2 + 1] == 0x80 && pcm[of * 2] == 0x00)
172 pcm[of * 2 + 1] = 0;
173 } else { /* 12bit quantization */
174 lc = ((uint16_t)frame[d] << 4) |
175 ((uint16_t)frame[d + 2] >> 4);
176 rc = ((uint16_t)frame[d + 1] << 4) |
177 ((uint16_t)frame[d + 2] & 0x0f);
178 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
179 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
180
181 of = sys->audio_shuffle[i % half_ch][j] +
182 (d - 8) / 3 * sys->audio_stride;
183 if (of * 2 >= size)
184 continue;
185
186 /* FIXME: maybe we have to admit that DV is a
187 * big-endian PCM */
188 pcm[of * 2] = lc & 0xff;
189 pcm[of * 2 + 1] = lc >> 8;
190 of = sys->audio_shuffle[i % half_ch + half_ch][j] +
191 (d - 8) / 3 * sys->audio_stride;
192 /* FIXME: maybe we have to admit that DV is a
193 * big-endian PCM */
194 pcm[of * 2] = rc & 0xff;
195 pcm[of * 2 + 1] = rc >> 8;
196 ++d;
197 }
198 }
199
200 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
201 }
202 }
203 }
204
205 return size;
206 }
207
208 static int dv_extract_audio_info(DVDemuxContext *c, uint8_t *frame)
209 {
210 const uint8_t *as_pack;
211 int freq, stype, smpls, quant, i, ach;
212
213 as_pack = dv_extract_pack(frame, dv_audio_source);
214 if (!as_pack || !c->sys) { /* No audio ? */
215 c->ach = 0;
216 return 0;
217 }
218
219 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
220 freq = as_pack[4] >> 3 & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
221 stype = as_pack[3] & 0x1f; /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
222 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
223
224 if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency)) {
225 av_log(c->fctx, AV_LOG_ERROR,
226 "Unrecognized audio sample rate index (%d)\n", freq);
227 return 0;
228 }
229
230 if (stype > 3) {
231 av_log(c->fctx, AV_LOG_ERROR, "stype %d is invalid\n", stype);
232 c->ach = 0;
233 return 0;
234 }
235
236 /* note: ach counts PAIRS of channels (i.e. stereo channels) */
237 ach = ((int[4]) { 1, 0, 2, 4 })[stype];
238 if (ach == 1 && quant && freq == 2)
239 ach = 2;
240
241 /* Dynamic handling of the audio streams in DV */
242 for (i = 0; i < ach; i++) {
243 if (!c->ast[i]) {
244 c->ast[i] = avformat_new_stream(c->fctx, NULL);
245 if (!c->ast[i])
246 break;
247 avpriv_set_pts_info(c->ast[i], 64, 1, 30000);
248 c->ast[i]->codec->codec_type = AVMEDIA_TYPE_AUDIO;
249 c->ast[i]->codec->codec_id = AV_CODEC_ID_PCM_S16LE;
250
251 av_init_packet(&c->audio_pkt[i]);
252 c->audio_pkt[i].size = 0;
253 c->audio_pkt[i].data = c->audio_buf[i];
254 c->audio_pkt[i].stream_index = c->ast[i]->index;
255 c->audio_pkt[i].flags |= AV_PKT_FLAG_KEY;
256 }
257 c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
258 c->ast[i]->codec->channels = 2;
259 c->ast[i]->codec->channel_layout = AV_CH_LAYOUT_STEREO;
260 c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
261 c->ast[i]->start_time = 0;
262 }
263 c->ach = i;
264
265 return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
266 }
267
268 static int dv_extract_video_info(DVDemuxContext *c, uint8_t *frame)
269 {
270 const uint8_t *vsc_pack;
271 AVCodecContext *avctx;
272 int apt, is16_9;
273 int size = 0;
274
275 if (c->sys) {
276 avctx = c->vst->codec;
277
278 avpriv_set_pts_info(c->vst, 64, c->sys->time_base.num,
279 c->sys->time_base.den);
280 c->vst->avg_frame_rate = av_inv_q(c->vst->time_base);
281 if (!avctx->width) {
282 avctx->width = c->sys->width;
283 avctx->height = c->sys->height;
284 }
285 avctx->pix_fmt = c->sys->pix_fmt;
286
287 /* finding out SAR is a little bit messy */
288 vsc_pack = dv_extract_pack(frame, dv_video_control);
289 apt = frame[4] & 0x07;
290 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
291 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
292 c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
293 avctx->bit_rate = av_rescale_q(c->sys->frame_size,
294 (AVRational) { 8, 1 },
295 c->sys->time_base);
296 size = c->sys->frame_size;
297 }
298 return size;
299 }
300
301 /* The following 3 functions constitute our interface to the world */
302
303 DVDemuxContext *avpriv_dv_init_demux(AVFormatContext *s)
304 {
305 DVDemuxContext *c;
306
307 c = av_mallocz(sizeof(DVDemuxContext));
308 if (!c)
309 return NULL;
310
311 c->vst = avformat_new_stream(s, NULL);
312 if (!c->vst) {
313 av_free(c);
314 return NULL;
315 }
316
317 c->fctx = s;
318 c->vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
319 c->vst->codec->codec_id = AV_CODEC_ID_DVVIDEO;
320 c->vst->codec->bit_rate = 25000000;
321 c->vst->start_time = 0;
322
323 return c;
324 }
325
326 int avpriv_dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
327 {
328 int size = -1;
329 int i;
330
331 for (i = 0; i < c->ach; i++) {
332 if (c->ast[i] && c->audio_pkt[i].size) {
333 *pkt = c->audio_pkt[i];
334 c->audio_pkt[i].size = 0;
335 size = pkt->size;
336 break;
337 }
338 }
339
340 return size;
341 }
342
343 int avpriv_dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
344 uint8_t *buf, int buf_size)
345 {
346 int size, i;
347 uint8_t *ppcm[5] = { 0 };
348
349 if (buf_size < DV_PROFILE_BYTES ||
350 !(c->sys = av_dv_frame_profile(c->sys, buf, buf_size)) ||
351 buf_size < c->sys->frame_size) {
352 return -1; /* Broken frame, or not enough data */
353 }
354
355 /* Queueing audio packet */
356 /* FIXME: in case of no audio/bad audio we have to do something */
357 size = dv_extract_audio_info(c, buf);
358 for (i = 0; i < c->ach; i++) {
359 c->audio_pkt[i].size = size;
360 c->audio_pkt[i].pts = c->abytes * 30000 * 8 /
361 c->ast[i]->codec->bit_rate;
362 ppcm[i] = c->audio_buf[i];
363 }
364 if (c->ach)
365 dv_extract_audio(buf, ppcm, c->sys);
366
367 /* We work with 720p frames split in half, thus even frames have
368 * channels 0,1 and odd 2,3. */
369 if (c->sys->height == 720) {
370 if (buf[1] & 0x0C) {
371 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
372 } else {
373 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
374 c->abytes += size;
375 }
376 } else {
377 c->abytes += size;
378 }
379
380 /* Now it's time to return video packet */
381 size = dv_extract_video_info(c, buf);
382 av_init_packet(pkt);
383 pkt->data = buf;
384 pkt->size = size;
385 pkt->flags |= AV_PKT_FLAG_KEY;
386 pkt->stream_index = c->vst->index;
387 pkt->pts = c->frames;
388
389 c->frames++;
390
391 return size;
392 }
393
394 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
395 int64_t timestamp, int flags)
396 {
397 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
398 const AVDVProfile *sys = av_dv_codec_profile(c->vst->codec->width, c->vst->codec->height,
399 c->vst->codec->pix_fmt);
400 int64_t offset;
401 int64_t size = avio_size(s->pb) - s->data_offset;
402 int64_t max_offset = ((size - 1) / sys->frame_size) * sys->frame_size;
403
404 offset = sys->frame_size * timestamp;
405
406 if (size >= 0 && offset > max_offset)
407 offset = max_offset;
408 else if (offset < 0)
409 offset = 0;
410
411 return offset + s->data_offset;
412 }
413
414 void ff_dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
415 {
416 c->frames = frame_offset;
417 if (c->ach)
418 c->abytes = av_rescale_q(c->frames, c->sys->time_base,
419 (AVRational) { 8, c->ast[0]->codec->bit_rate });
420
421 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
422 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
423 }
424
425 /************************************************************
426 * Implementation of the easiest DV storage of all -- raw DV.
427 ************************************************************/
428
429 typedef struct RawDVContext {
430 DVDemuxContext *dv_demux;
431 uint8_t buf[DV_MAX_FRAME_SIZE];
432 } RawDVContext;
433
434 static int dv_read_header(AVFormatContext *s)
435 {
436 unsigned state, marker_pos = 0;
437 RawDVContext *c = s->priv_data;
438
439 c->dv_demux = avpriv_dv_init_demux(s);
440 if (!c->dv_demux)
441 return -1;
442
443 state = avio_rb32(s->pb);
444 while ((state & 0xffffff7f) != 0x1f07003f) {
445 if (s->pb->eof_reached) {
446 av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
447 return -1;
448 }
449 if (state == 0x003f0700 || state == 0xff3f0700)
450 marker_pos = avio_tell(s->pb);
451 if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
452 avio_seek(s->pb, -163, SEEK_CUR);
453 state = avio_rb32(s->pb);
454 break;
455 }
456 state = (state << 8) | avio_r8(s->pb);
457 }
458 AV_WB32(c->buf, state);
459
460 if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) <= 0 ||
461 avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
462 return AVERROR(EIO);
463
464 c->dv_demux->sys = av_dv_frame_profile(c->dv_demux->sys,
465 c->buf,
466 DV_PROFILE_BYTES);
467 if (!c->dv_demux->sys) {
468 av_log(s, AV_LOG_ERROR,
469 "Can't determine profile of DV input stream.\n");
470 return -1;
471 }
472
473 s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size,
474 (AVRational) { 8, 1 },
475 c->dv_demux->sys->time_base);
476
477 return 0;
478 }
479
480 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
481 {
482 int size;
483 RawDVContext *c = s->priv_data;
484
485 size = avpriv_dv_get_packet(c->dv_demux, pkt);
486
487 if (size < 0) {
488 if (!c->dv_demux->sys)
489 return AVERROR(EIO);
490 size = c->dv_demux->sys->frame_size;
491 if (avio_read(s->pb, c->buf, size) <= 0)
492 return AVERROR(EIO);
493
494 size = avpriv_dv_produce_packet(c->dv_demux, pkt, c->buf, size);
495 }
496
497 return size;
498 }
499
500 static int dv_read_seek(AVFormatContext *s, int stream_index,
501 int64_t timestamp, int flags)
502 {
503 RawDVContext *r = s->priv_data;
504 DVDemuxContext *c = r->dv_demux;
505 int64_t offset = dv_frame_offset(s, c, timestamp, flags);
506
507 if (avio_seek(s->pb, offset, SEEK_SET) < 0)
508 return -1;
509
510 ff_dv_offset_reset(c, offset / c->sys->frame_size);
511 return 0;
512 }
513
514 static int dv_read_close(AVFormatContext *s)
515 {
516 RawDVContext *c = s->priv_data;
517 av_free(c->dv_demux);
518 return 0;
519 }
520
521 static int dv_probe(AVProbeData *p)
522 {
523 unsigned state, marker_pos = 0;
524 int i;
525 int matches = 0;
526 int secondary_matches = 0;
527
528 if (p->buf_size < 5)
529 return 0;
530
531 state = AV_RB32(p->buf);
532 for (i = 4; i < p->buf_size; i++) {
533 if ((state & 0xffffff7f) == 0x1f07003f)
534 matches++;
535 // any section header, also with seq/chan num != 0,
536 // should appear around every 12000 bytes, at least 10 per frame
537 if ((state & 0xff07ff7f) == 0x1f07003f)
538 secondary_matches++;
539 if (state == 0x003f0700 || state == 0xff3f0700)
540 marker_pos = i;
541 if (state == 0xff3f0701 && i - marker_pos == 80)
542 matches++;
543 state = (state << 8) | p->buf[i];
544 }
545
546 if (matches && p->buf_size / matches < 1024 * 1024) {
547 if (matches > 4 ||
548 (secondary_matches >= 10 &&
549 p->buf_size / secondary_matches < 24000))
550 // not max to avoid dv in mov to match
551 return AVPROBE_SCORE_MAX * 3 / 4;
552 return AVPROBE_SCORE_MAX / 4;
553 }
554 return 0;
555 }
556
557 AVInputFormat ff_dv_demuxer = {
558 .name = "dv",
559 .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
560 .priv_data_size = sizeof(RawDVContext),
561 .read_probe = dv_probe,
562 .read_header = dv_read_header,
563 .read_packet = dv_read_packet,
564 .read_close = dv_read_close,
565 .read_seek = dv_read_seek,
566 .extensions = "dv,dif",
567 };