Replace ffmpeg references with more accurate libav* references.
[libav.git] / libavformat / dv.c
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 "libavcodec/dvdata.h"
34 #include "libavutil/intreadwrite.h"
35 #include "libavutil/mathematics.h"
36 #include "dv.h"
37
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
50 static inline uint16_t dv_audio_12to16(uint16_t sample)
51 {
52 uint16_t shift, result;
53
54 sample = (sample < 0x800) ? sample : sample | 0xf000;
55 shift = (sample & 0xf00) >> 8;
56
57 if (shift < 0x2 || shift > 0xd) {
58 result = sample;
59 } else if (shift < 0x8) {
60 shift--;
61 result = (sample - (256 * shift)) << shift;
62 } else {
63 shift = 0xe - shift;
64 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
65 }
66
67 return result;
68 }
69
70 /*
71 * This is the dumbest implementation of all -- it simply looks at
72 * a fixed offset and if pack isn't there -- fails. We might want
73 * to have a fallback mechanism for complete search of missing packs.
74 */
75 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
76 {
77 int offs;
78
79 switch (t) {
80 case dv_audio_source:
81 offs = (80*6 + 80*16*3 + 3);
82 break;
83 case dv_audio_control:
84 offs = (80*6 + 80*16*4 + 3);
85 break;
86 case dv_video_control:
87 offs = (80*5 + 48 + 5);
88 break;
89 default:
90 return NULL;
91 }
92
93 return frame[offs] == t ? &frame[offs] : NULL;
94 }
95
96 /*
97 * There's a couple of assumptions being made here:
98 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
99 * We can pass them upwards when libavcodec will be ready to deal with them.
100 * 2. We don't do software emphasis.
101 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
102 * are converted into 16bit linear ones.
103 */
104 static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4],
105 const DVprofile *sys)
106 {
107 int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
108 uint16_t lc, rc;
109 const uint8_t* as_pack;
110 uint8_t *pcm, ipcm;
111
112 as_pack = dv_extract_pack(frame, dv_audio_source);
113 if (!as_pack) /* No audio ? */
114 return 0;
115
116 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
117 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
118 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
119
120 if (quant > 1)
121 return -1; /* unsupported quantization */
122
123 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
124 half_ch = sys->difseg_size / 2;
125
126 /* We work with 720p frames split in half, thus even frames have
127 * channels 0,1 and odd 2,3. */
128 ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
129 pcm = ppcm[ipcm++];
130
131 /* for each DIF channel */
132 for (chan = 0; chan < sys->n_difchan; chan++) {
133 /* for each DIF segment */
134 for (i = 0; i < sys->difseg_size; i++) {
135 frame += 6 * 80; /* skip DIF segment header */
136 if (quant == 1 && i == half_ch) {
137 /* next stereo channel (12bit mode only) */
138 pcm = ppcm[ipcm++];
139 if (!pcm)
140 break;
141 }
142
143 /* for each AV sequence */
144 for (j = 0; j < 9; j++) {
145 for (d = 8; d < 80; d += 2) {
146 if (quant == 0) { /* 16bit quantization */
147 of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride;
148 if (of*2 >= size)
149 continue;
150
151 pcm[of*2] = frame[d+1]; // FIXME: maybe we have to admit
152 pcm[of*2+1] = frame[d]; // that DV is a big-endian PCM
153 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
154 pcm[of*2+1] = 0;
155 } else { /* 12bit quantization */
156 lc = ((uint16_t)frame[d] << 4) |
157 ((uint16_t)frame[d+2] >> 4);
158 rc = ((uint16_t)frame[d+1] << 4) |
159 ((uint16_t)frame[d+2] & 0x0f);
160 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
161 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
162
163 of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride;
164 if (of*2 >= size)
165 continue;
166
167 pcm[of*2] = lc & 0xff; // FIXME: maybe we have to admit
168 pcm[of*2+1] = lc >> 8; // that DV is a big-endian PCM
169 of = sys->audio_shuffle[i%half_ch+half_ch][j] +
170 (d - 8) / 3 * sys->audio_stride;
171 pcm[of*2] = rc & 0xff; // FIXME: maybe we have to admit
172 pcm[of*2+1] = rc >> 8; // that DV is a big-endian PCM
173 ++d;
174 }
175 }
176
177 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
178 }
179 }
180
181 /* next stereo channel (50Mbps and 100Mbps only) */
182 pcm = ppcm[ipcm++];
183 if (!pcm)
184 break;
185 }
186
187 return size;
188 }
189
190 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
191 {
192 const uint8_t* as_pack;
193 int freq, stype, smpls, quant, i, ach;
194
195 as_pack = dv_extract_pack(frame, dv_audio_source);
196 if (!as_pack || !c->sys) { /* No audio ? */
197 c->ach = 0;
198 return 0;
199 }
200
201 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
202 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
203 stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
204 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
205
206 /* note: ach counts PAIRS of channels (i.e. stereo channels) */
207 ach = ((int[4]){ 1, 0, 2, 4})[stype];
208 if (ach == 1 && quant && freq == 2)
209 ach = 2;
210
211 /* Dynamic handling of the audio streams in DV */
212 for (i = 0; i < ach; i++) {
213 if (!c->ast[i]) {
214 c->ast[i] = avformat_new_stream(c->fctx, NULL);
215 if (!c->ast[i])
216 break;
217 av_set_pts_info(c->ast[i], 64, 1, 30000);
218 c->ast[i]->codec->codec_type = AVMEDIA_TYPE_AUDIO;
219 c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE;
220
221 av_init_packet(&c->audio_pkt[i]);
222 c->audio_pkt[i].size = 0;
223 c->audio_pkt[i].data = c->audio_buf[i];
224 c->audio_pkt[i].stream_index = c->ast[i]->index;
225 c->audio_pkt[i].flags |= AV_PKT_FLAG_KEY;
226 }
227 c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
228 c->ast[i]->codec->channels = 2;
229 c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
230 c->ast[i]->start_time = 0;
231 }
232 c->ach = i;
233
234 return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
235 }
236
237 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
238 {
239 const uint8_t* vsc_pack;
240 AVCodecContext* avctx;
241 int apt, is16_9;
242 int size = 0;
243
244 if (c->sys) {
245 avctx = c->vst->codec;
246
247 av_set_pts_info(c->vst, 64, c->sys->time_base.num,
248 c->sys->time_base.den);
249 avctx->time_base= c->sys->time_base;
250 if (!avctx->width){
251 avctx->width = c->sys->width;
252 avctx->height = c->sys->height;
253 }
254 avctx->pix_fmt = c->sys->pix_fmt;
255
256 /* finding out SAR is a little bit messy */
257 vsc_pack = dv_extract_pack(frame, dv_video_control);
258 apt = frame[4] & 0x07;
259 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
260 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
261 c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
262 avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1},
263 c->sys->time_base);
264 size = c->sys->frame_size;
265 }
266 return size;
267 }
268
269 /*
270 * The following 3 functions constitute our interface to the world
271 */
272
273 DVDemuxContext* avpriv_dv_init_demux(AVFormatContext *s)
274 {
275 DVDemuxContext *c;
276
277 c = av_mallocz(sizeof(DVDemuxContext));
278 if (!c)
279 return NULL;
280
281 c->vst = avformat_new_stream(s, NULL);
282 if (!c->vst) {
283 av_free(c);
284 return NULL;
285 }
286
287 c->sys = NULL;
288 c->fctx = s;
289 memset(c->ast, 0, sizeof(c->ast));
290 c->ach = 0;
291 c->frames = 0;
292 c->abytes = 0;
293
294 c->vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
295 c->vst->codec->codec_id = CODEC_ID_DVVIDEO;
296 c->vst->codec->bit_rate = 25000000;
297 c->vst->start_time = 0;
298
299 return c;
300 }
301
302 int avpriv_dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
303 {
304 int size = -1;
305 int i;
306
307 for (i = 0; i < c->ach; i++) {
308 if (c->ast[i] && c->audio_pkt[i].size) {
309 *pkt = c->audio_pkt[i];
310 c->audio_pkt[i].size = 0;
311 size = pkt->size;
312 break;
313 }
314 }
315
316 return size;
317 }
318
319 int avpriv_dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
320 uint8_t* buf, int buf_size)
321 {
322 int size, i;
323 uint8_t *ppcm[4] = {0};
324
325 if (buf_size < DV_PROFILE_BYTES ||
326 !(c->sys = avpriv_dv_frame_profile(c->sys, buf, buf_size)) ||
327 buf_size < c->sys->frame_size) {
328 return -1; /* Broken frame, or not enough data */
329 }
330
331 /* Queueing audio packet */
332 /* FIXME: in case of no audio/bad audio we have to do something */
333 size = dv_extract_audio_info(c, buf);
334 for (i = 0; i < c->ach; i++) {
335 c->audio_pkt[i].size = size;
336 c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate;
337 ppcm[i] = c->audio_buf[i];
338 }
339 dv_extract_audio(buf, ppcm, c->sys);
340
341 /* We work with 720p frames split in half, thus even frames have
342 * channels 0,1 and odd 2,3. */
343 if (c->sys->height == 720) {
344 if (buf[1] & 0x0C) {
345 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
346 } else {
347 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
348 c->abytes += size;
349 }
350 } else {
351 c->abytes += size;
352 }
353
354 /* Now it's time to return video packet */
355 size = dv_extract_video_info(c, buf);
356 av_init_packet(pkt);
357 pkt->data = buf;
358 pkt->size = size;
359 pkt->flags |= AV_PKT_FLAG_KEY;
360 pkt->stream_index = c->vst->id;
361 pkt->pts = c->frames;
362
363 c->frames++;
364
365 return size;
366 }
367
368 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
369 int64_t timestamp, int flags)
370 {
371 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
372 const DVprofile* sys = avpriv_dv_codec_profile(c->vst->codec);
373 int64_t offset;
374 int64_t size = avio_size(s->pb) - s->data_offset;
375 int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
376
377 offset = sys->frame_size * timestamp;
378
379 if (size >= 0 && offset > max_offset) offset = max_offset;
380 else if (offset < 0) offset = 0;
381
382 return offset + s->data_offset;
383 }
384
385 void dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
386 {
387 c->frames= frame_offset;
388 if (c->ach)
389 c->abytes= av_rescale_q(c->frames, c->sys->time_base,
390 (AVRational){8, c->ast[0]->codec->bit_rate});
391 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
392 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
393 }
394
395 /************************************************************
396 * Implementation of the easiest DV storage of all -- raw DV.
397 ************************************************************/
398
399 typedef struct RawDVContext {
400 DVDemuxContext* dv_demux;
401 uint8_t buf[DV_MAX_FRAME_SIZE];
402 } RawDVContext;
403
404 static int dv_read_header(AVFormatContext *s,
405 AVFormatParameters *ap)
406 {
407 unsigned state, marker_pos = 0;
408 RawDVContext *c = s->priv_data;
409
410 c->dv_demux = avpriv_dv_init_demux(s);
411 if (!c->dv_demux)
412 return -1;
413
414 state = avio_rb32(s->pb);
415 while ((state & 0xffffff7f) != 0x1f07003f) {
416 if (s->pb->eof_reached) {
417 av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
418 return -1;
419 }
420 if (state == 0x003f0700 || state == 0xff3f0700)
421 marker_pos = avio_tell(s->pb);
422 if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
423 avio_seek(s->pb, -163, SEEK_CUR);
424 state = avio_rb32(s->pb);
425 break;
426 }
427 state = (state << 8) | avio_r8(s->pb);
428 }
429 AV_WB32(c->buf, state);
430
431 if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) <= 0 ||
432 avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
433 return AVERROR(EIO);
434
435 c->dv_demux->sys = avpriv_dv_frame_profile(c->dv_demux->sys, c->buf, DV_PROFILE_BYTES);
436 if (!c->dv_demux->sys) {
437 av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n");
438 return -1;
439 }
440
441 s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size, (AVRational){8,1},
442 c->dv_demux->sys->time_base);
443
444 return 0;
445 }
446
447
448 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
449 {
450 int size;
451 RawDVContext *c = s->priv_data;
452
453 size = avpriv_dv_get_packet(c->dv_demux, pkt);
454
455 if (size < 0) {
456 if (!c->dv_demux->sys)
457 return AVERROR(EIO);
458 size = c->dv_demux->sys->frame_size;
459 if (avio_read(s->pb, c->buf, size) <= 0)
460 return AVERROR(EIO);
461
462 size = avpriv_dv_produce_packet(c->dv_demux, pkt, c->buf, size);
463 }
464
465 return size;
466 }
467
468 static int dv_read_seek(AVFormatContext *s, int stream_index,
469 int64_t timestamp, int flags)
470 {
471 RawDVContext *r = s->priv_data;
472 DVDemuxContext *c = r->dv_demux;
473 int64_t offset = dv_frame_offset(s, c, timestamp, flags);
474
475 dv_offset_reset(c, offset / c->sys->frame_size);
476
477 offset = avio_seek(s->pb, offset, SEEK_SET);
478 return (offset < 0) ? offset : 0;
479 }
480
481 static int dv_read_close(AVFormatContext *s)
482 {
483 RawDVContext *c = s->priv_data;
484 av_free(c->dv_demux);
485 return 0;
486 }
487
488 static int dv_probe(AVProbeData *p)
489 {
490 unsigned state, marker_pos = 0;
491 int i;
492 int matches = 0;
493 int secondary_matches = 0;
494
495 if (p->buf_size < 5)
496 return 0;
497
498 state = AV_RB32(p->buf);
499 for (i = 4; i < p->buf_size; i++) {
500 if ((state & 0xffffff7f) == 0x1f07003f)
501 matches++;
502 // any section header, also with seq/chan num != 0,
503 // should appear around every 12000 bytes, at least 10 per frame
504 if ((state & 0xff07ff7f) == 0x1f07003f)
505 secondary_matches++;
506 if (state == 0x003f0700 || state == 0xff3f0700)
507 marker_pos = i;
508 if (state == 0xff3f0701 && i - marker_pos == 80)
509 matches++;
510 state = (state << 8) | p->buf[i];
511 }
512
513 if (matches && p->buf_size / matches < 1024*1024) {
514 if (matches > 4 || (secondary_matches >= 10 && p->buf_size / secondary_matches < 24000))
515 return AVPROBE_SCORE_MAX*3/4; // not max to avoid dv in mov to match
516 return AVPROBE_SCORE_MAX/4;
517 }
518 return 0;
519 }
520
521 #if CONFIG_DV_DEMUXER
522 AVInputFormat ff_dv_demuxer = {
523 .name = "dv",
524 .long_name = NULL_IF_CONFIG_SMALL("DV video format"),
525 .priv_data_size = sizeof(RawDVContext),
526 .read_probe = dv_probe,
527 .read_header = dv_read_header,
528 .read_packet = dv_read_packet,
529 .read_close = dv_read_close,
530 .read_seek = dv_read_seek,
531 .extensions = "dv,dif",
532 };
533 #endif