fa529f68d2783ffa41a81c711fd3fccda21f7ae7
[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 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 "libavcodec/dvdata.h"
34 #include "dv.h"
35
36 struct DVDemuxContext {
37 const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
38 AVFormatContext* fctx;
39 AVStream* vst;
40 AVStream* ast[4];
41 AVPacket audio_pkt[4];
42 uint8_t audio_buf[4][8192];
43 int ach;
44 int frames;
45 uint64_t abytes;
46 };
47
48 static inline uint16_t dv_audio_12to16(uint16_t sample)
49 {
50 uint16_t shift, result;
51
52 sample = (sample < 0x800) ? sample : sample | 0xf000;
53 shift = (sample & 0xf00) >> 8;
54
55 if (shift < 0x2 || shift > 0xd) {
56 result = sample;
57 } else if (shift < 0x8) {
58 shift--;
59 result = (sample - (256 * shift)) << shift;
60 } else {
61 shift = 0xe - shift;
62 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
63 }
64
65 return result;
66 }
67
68 /*
69 * This is the dumbest implementation of all -- it simply looks at
70 * a fixed offset and if pack isn't there -- fails. We might want
71 * to have a fallback mechanism for complete search of missing packs.
72 */
73 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
74 {
75 int offs;
76
77 switch (t) {
78 case dv_audio_source:
79 offs = (80*6 + 80*16*3 + 3);
80 break;
81 case dv_audio_control:
82 offs = (80*6 + 80*16*4 + 3);
83 break;
84 case dv_video_control:
85 offs = (80*5 + 48 + 5);
86 break;
87 default:
88 return NULL;
89 }
90
91 return frame[offs] == t ? &frame[offs] : NULL;
92 }
93
94 /*
95 * There's a couple of assumptions being made here:
96 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
97 * We can pass them upwards when ffmpeg will be ready to deal with them.
98 * 2. We don't do software emphasis.
99 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
100 * are converted into 16bit linear ones.
101 */
102 static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4],
103 const DVprofile *sys)
104 {
105 int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
106 uint16_t lc, rc;
107 const uint8_t* as_pack;
108 uint8_t *pcm, ipcm;
109
110 as_pack = dv_extract_pack(frame, dv_audio_source);
111 if (!as_pack) /* No audio ? */
112 return 0;
113
114 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
115 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
116 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
117
118 if (quant > 1)
119 return -1; /* unsupported quantization */
120
121 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
122 half_ch = sys->difseg_size / 2;
123
124 /* We work with 720p frames split in half, thus even frames have
125 * channels 0,1 and odd 2,3. */
126 ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
127 pcm = ppcm[ipcm++];
128
129 /* for each DIF channel */
130 for (chan = 0; chan < sys->n_difchan; chan++) {
131 /* for each DIF segment */
132 for (i = 0; i < sys->difseg_size; i++) {
133 frame += 6 * 80; /* skip DIF segment header */
134 if (quant == 1 && i == half_ch) {
135 /* next stereo channel (12bit mode only) */
136 pcm = ppcm[ipcm++];
137 if (!pcm)
138 break;
139 }
140
141 /* for each AV sequence */
142 for (j = 0; j < 9; j++) {
143 for (d = 8; d < 80; d += 2) {
144 if (quant == 0) { /* 16bit quantization */
145 of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride;
146 if (of*2 >= size)
147 continue;
148
149 pcm[of*2] = frame[d+1]; // FIXME: maybe we have to admit
150 pcm[of*2+1] = frame[d]; // that DV is a big-endian PCM
151 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
152 pcm[of*2+1] = 0;
153 } else { /* 12bit quantization */
154 lc = ((uint16_t)frame[d] << 4) |
155 ((uint16_t)frame[d+2] >> 4);
156 rc = ((uint16_t)frame[d+1] << 4) |
157 ((uint16_t)frame[d+2] & 0x0f);
158 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
159 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
160
161 of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride;
162 if (of*2 >= size)
163 continue;
164
165 pcm[of*2] = lc & 0xff; // FIXME: maybe we have to admit
166 pcm[of*2+1] = lc >> 8; // that DV is a big-endian PCM
167 of = sys->audio_shuffle[i%half_ch+half_ch][j] +
168 (d - 8) / 3 * sys->audio_stride;
169 pcm[of*2] = rc & 0xff; // FIXME: maybe we have to admit
170 pcm[of*2+1] = rc >> 8; // that DV is a big-endian PCM
171 ++d;
172 }
173 }
174
175 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
176 }
177 }
178
179 /* next stereo channel (50Mbps and 100Mbps only) */
180 pcm = ppcm[ipcm++];
181 if (!pcm)
182 break;
183 }
184
185 return size;
186 }
187
188 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
189 {
190 const uint8_t* as_pack;
191 int freq, stype, smpls, quant, i, ach;
192
193 as_pack = dv_extract_pack(frame, dv_audio_source);
194 if (!as_pack || !c->sys) { /* No audio ? */
195 c->ach = 0;
196 return 0;
197 }
198
199 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
200 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
201 stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
202 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
203
204 /* note: ach counts PAIRS of channels (i.e. stereo channels) */
205 ach = ((int[4]){ 1, 0, 2, 4})[stype];
206 if (ach == 1 && quant && freq == 2)
207 ach = 2;
208
209 /* Dynamic handling of the audio streams in DV */
210 for (i = 0; i < ach; i++) {
211 if (!c->ast[i]) {
212 c->ast[i] = av_new_stream(c->fctx, 0);
213 if (!c->ast[i])
214 break;
215 av_set_pts_info(c->ast[i], 64, 1, 30000);
216 c->ast[i]->codec->codec_type = CODEC_TYPE_AUDIO;
217 c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE;
218
219 av_init_packet(&c->audio_pkt[i]);
220 c->audio_pkt[i].size = 0;
221 c->audio_pkt[i].data = c->audio_buf[i];
222 c->audio_pkt[i].stream_index = c->ast[i]->index;
223 c->audio_pkt[i].flags |= PKT_FLAG_KEY;
224 }
225 c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
226 c->ast[i]->codec->channels = 2;
227 c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
228 c->ast[i]->start_time = 0;
229 }
230 c->ach = i;
231
232 return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
233 }
234
235 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
236 {
237 const uint8_t* vsc_pack;
238 AVCodecContext* avctx;
239 int apt, is16_9;
240 int size = 0;
241
242 if (c->sys) {
243 avctx = c->vst->codec;
244
245 av_set_pts_info(c->vst, 64, c->sys->time_base.num,
246 c->sys->time_base.den);
247 avctx->time_base= c->sys->time_base;
248 if (!avctx->width){
249 avctx->width = c->sys->width;
250 avctx->height = c->sys->height;
251 }
252 avctx->pix_fmt = c->sys->pix_fmt;
253
254 /* finding out SAR is a little bit messy */
255 vsc_pack = dv_extract_pack(frame, dv_video_control);
256 apt = frame[4] & 0x07;
257 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
258 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
259 c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
260 avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1},
261 c->sys->time_base);
262 size = c->sys->frame_size;
263 }
264 return size;
265 }
266
267 /*
268 * The following 3 functions constitute our interface to the world
269 */
270
271 DVDemuxContext* dv_init_demux(AVFormatContext *s)
272 {
273 DVDemuxContext *c;
274
275 c = av_mallocz(sizeof(DVDemuxContext));
276 if (!c)
277 return NULL;
278
279 c->vst = av_new_stream(s, 0);
280 if (!c->vst) {
281 av_free(c);
282 return NULL;
283 }
284
285 c->sys = NULL;
286 c->fctx = s;
287 memset(c->ast, 0, sizeof(c->ast));
288 c->ach = 0;
289 c->frames = 0;
290 c->abytes = 0;
291
292 c->vst->codec->codec_type = CODEC_TYPE_VIDEO;
293 c->vst->codec->codec_id = CODEC_ID_DVVIDEO;
294 c->vst->codec->bit_rate = 25000000;
295 c->vst->start_time = 0;
296
297 return c;
298 }
299
300 int dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
301 {
302 int size = -1;
303 int i;
304
305 for (i = 0; i < c->ach; i++) {
306 if (c->ast[i] && c->audio_pkt[i].size) {
307 *pkt = c->audio_pkt[i];
308 c->audio_pkt[i].size = 0;
309 size = pkt->size;
310 break;
311 }
312 }
313
314 return size;
315 }
316
317 int dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
318 uint8_t* buf, int buf_size)
319 {
320 int size, i;
321 uint8_t *ppcm[4] = {0};
322
323 if (buf_size < DV_PROFILE_BYTES ||
324 !(c->sys = dv_frame_profile(buf)) ||
325 buf_size < c->sys->frame_size) {
326 return -1; /* Broken frame, or not enough data */
327 }
328
329 /* Queueing audio packet */
330 /* FIXME: in case of no audio/bad audio we have to do something */
331 size = dv_extract_audio_info(c, buf);
332 for (i = 0; i < c->ach; i++) {
333 c->audio_pkt[i].size = size;
334 c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate;
335 ppcm[i] = c->audio_buf[i];
336 }
337 dv_extract_audio(buf, ppcm, c->sys);
338 c->abytes += size;
339
340 /* We work with 720p frames split in half, thus even frames have
341 * channels 0,1 and odd 2,3. */
342 if (c->sys->height == 720) {
343 if (buf[1] & 0x0C)
344 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
345 else
346 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
347 }
348
349 /* Now it's time to return video packet */
350 size = dv_extract_video_info(c, buf);
351 av_init_packet(pkt);
352 pkt->data = buf;
353 pkt->size = size;
354 pkt->flags |= PKT_FLAG_KEY;
355 pkt->stream_index = c->vst->id;
356 pkt->pts = c->frames;
357
358 c->frames++;
359
360 return size;
361 }
362
363 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
364 int64_t timestamp, int flags)
365 {
366 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
367 const DVprofile* sys = dv_codec_profile(c->vst->codec);
368 int64_t offset;
369 int64_t size = url_fsize(s->pb);
370 int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
371
372 offset = sys->frame_size * timestamp;
373
374 if (offset > max_offset) offset = max_offset;
375 else if (offset < 0) offset = 0;
376
377 return offset;
378 }
379
380 void dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
381 {
382 c->frames= frame_offset;
383 if (c->ach)
384 c->abytes= av_rescale_q(c->frames, c->sys->time_base,
385 (AVRational){8, c->ast[0]->codec->bit_rate});
386 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
387 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
388 }
389
390 /************************************************************
391 * Implementation of the easiest DV storage of all -- raw DV.
392 ************************************************************/
393
394 typedef struct RawDVContext {
395 DVDemuxContext* dv_demux;
396 uint8_t buf[DV_MAX_FRAME_SIZE];
397 } RawDVContext;
398
399 static int dv_read_header(AVFormatContext *s,
400 AVFormatParameters *ap)
401 {
402 RawDVContext *c = s->priv_data;
403
404 c->dv_demux = dv_init_demux(s);
405 if (!c->dv_demux)
406 return -1;
407
408 if (get_buffer(s->pb, c->buf, DV_PROFILE_BYTES) <= 0 ||
409 url_fseek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
410 return AVERROR(EIO);
411
412 c->dv_demux->sys = dv_frame_profile(c->buf);
413 if (!c->dv_demux->sys) {
414 av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n");
415 return -1;
416 }
417
418 s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size, (AVRational){8,1},
419 c->dv_demux->sys->time_base);
420
421 return 0;
422 }
423
424
425 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
426 {
427 int size;
428 RawDVContext *c = s->priv_data;
429
430 size = dv_get_packet(c->dv_demux, pkt);
431
432 if (size < 0) {
433 if (!c->dv_demux->sys)
434 return AVERROR(EIO);
435 size = c->dv_demux->sys->frame_size;
436 if (get_buffer(s->pb, c->buf, size) <= 0)
437 return AVERROR(EIO);
438
439 size = dv_produce_packet(c->dv_demux, pkt, c->buf, size);
440 }
441
442 return size;
443 }
444
445 static int dv_read_seek(AVFormatContext *s, int stream_index,
446 int64_t timestamp, int flags)
447 {
448 RawDVContext *r = s->priv_data;
449 DVDemuxContext *c = r->dv_demux;
450 int64_t offset = dv_frame_offset(s, c, timestamp, flags);
451
452 dv_offset_reset(c, offset / c->sys->frame_size);
453
454 offset = url_fseek(s->pb, offset, SEEK_SET);
455 return (offset < 0) ? offset : 0;
456 }
457
458 static int dv_read_close(AVFormatContext *s)
459 {
460 RawDVContext *c = s->priv_data;
461 av_free(c->dv_demux);
462 return 0;
463 }
464
465 #if CONFIG_DV_DEMUXER
466 AVInputFormat dv_demuxer = {
467 "dv",
468 NULL_IF_CONFIG_SMALL("DV video format"),
469 sizeof(RawDVContext),
470 NULL,
471 dv_read_header,
472 dv_read_packet,
473 dv_read_close,
474 dv_read_seek,
475 .extensions = "dv,dif",
476 };
477 #endif