Support playback of incorrectly cut DV files where the headers do not
[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 "libavutil/intreadwrite.h"
35 #include "dv.h"
36
37 struct DVDemuxContext {
38 const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
39 AVFormatContext* fctx;
40 AVStream* vst;
41 AVStream* ast[4];
42 AVPacket audio_pkt[4];
43 uint8_t audio_buf[4][8192];
44 int ach;
45 int frames;
46 uint64_t abytes;
47 };
48
49 static inline uint16_t dv_audio_12to16(uint16_t sample)
50 {
51 uint16_t shift, result;
52
53 sample = (sample < 0x800) ? sample : sample | 0xf000;
54 shift = (sample & 0xf00) >> 8;
55
56 if (shift < 0x2 || shift > 0xd) {
57 result = sample;
58 } else if (shift < 0x8) {
59 shift--;
60 result = (sample - (256 * shift)) << shift;
61 } else {
62 shift = 0xe - shift;
63 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
64 }
65
66 return result;
67 }
68
69 /*
70 * This is the dumbest implementation of all -- it simply looks at
71 * a fixed offset and if pack isn't there -- fails. We might want
72 * to have a fallback mechanism for complete search of missing packs.
73 */
74 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
75 {
76 int offs;
77
78 switch (t) {
79 case dv_audio_source:
80 offs = (80*6 + 80*16*3 + 3);
81 break;
82 case dv_audio_control:
83 offs = (80*6 + 80*16*4 + 3);
84 break;
85 case dv_video_control:
86 offs = (80*5 + 48 + 5);
87 break;
88 default:
89 return NULL;
90 }
91
92 return frame[offs] == t ? &frame[offs] : NULL;
93 }
94
95 /*
96 * There's a couple of assumptions being made here:
97 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
98 * We can pass them upwards when ffmpeg will be ready to deal with them.
99 * 2. We don't do software emphasis.
100 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
101 * are converted into 16bit linear ones.
102 */
103 static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4],
104 const DVprofile *sys)
105 {
106 int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
107 uint16_t lc, rc;
108 const uint8_t* as_pack;
109 uint8_t *pcm, ipcm;
110
111 as_pack = dv_extract_pack(frame, dv_audio_source);
112 if (!as_pack) /* No audio ? */
113 return 0;
114
115 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
116 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
117 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
118
119 if (quant > 1)
120 return -1; /* unsupported quantization */
121
122 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
123 half_ch = sys->difseg_size / 2;
124
125 /* We work with 720p frames split in half, thus even frames have
126 * channels 0,1 and odd 2,3. */
127 ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
128 pcm = ppcm[ipcm++];
129
130 /* for each DIF channel */
131 for (chan = 0; chan < sys->n_difchan; chan++) {
132 /* for each DIF segment */
133 for (i = 0; i < sys->difseg_size; i++) {
134 frame += 6 * 80; /* skip DIF segment header */
135 if (quant == 1 && i == half_ch) {
136 /* next stereo channel (12bit mode only) */
137 pcm = ppcm[ipcm++];
138 if (!pcm)
139 break;
140 }
141
142 /* for each AV sequence */
143 for (j = 0; j < 9; j++) {
144 for (d = 8; d < 80; d += 2) {
145 if (quant == 0) { /* 16bit quantization */
146 of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride;
147 if (of*2 >= size)
148 continue;
149
150 pcm[of*2] = frame[d+1]; // FIXME: maybe we have to admit
151 pcm[of*2+1] = frame[d]; // that DV is a big-endian PCM
152 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
153 pcm[of*2+1] = 0;
154 } else { /* 12bit quantization */
155 lc = ((uint16_t)frame[d] << 4) |
156 ((uint16_t)frame[d+2] >> 4);
157 rc = ((uint16_t)frame[d+1] << 4) |
158 ((uint16_t)frame[d+2] & 0x0f);
159 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
160 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
161
162 of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride;
163 if (of*2 >= size)
164 continue;
165
166 pcm[of*2] = lc & 0xff; // FIXME: maybe we have to admit
167 pcm[of*2+1] = lc >> 8; // that DV is a big-endian PCM
168 of = sys->audio_shuffle[i%half_ch+half_ch][j] +
169 (d - 8) / 3 * sys->audio_stride;
170 pcm[of*2] = rc & 0xff; // FIXME: maybe we have to admit
171 pcm[of*2+1] = rc >> 8; // that DV is a big-endian PCM
172 ++d;
173 }
174 }
175
176 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
177 }
178 }
179
180 /* next stereo channel (50Mbps and 100Mbps only) */
181 pcm = ppcm[ipcm++];
182 if (!pcm)
183 break;
184 }
185
186 return size;
187 }
188
189 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
190 {
191 const uint8_t* as_pack;
192 int freq, stype, smpls, quant, i, ach;
193
194 as_pack = dv_extract_pack(frame, dv_audio_source);
195 if (!as_pack || !c->sys) { /* No audio ? */
196 c->ach = 0;
197 return 0;
198 }
199
200 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
201 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
202 stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
203 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
204
205 /* note: ach counts PAIRS of channels (i.e. stereo channels) */
206 ach = ((int[4]){ 1, 0, 2, 4})[stype];
207 if (ach == 1 && quant && freq == 2)
208 ach = 2;
209
210 /* Dynamic handling of the audio streams in DV */
211 for (i = 0; i < ach; i++) {
212 if (!c->ast[i]) {
213 c->ast[i] = av_new_stream(c->fctx, 0);
214 if (!c->ast[i])
215 break;
216 av_set_pts_info(c->ast[i], 64, 1, 30000);
217 c->ast[i]->codec->codec_type = CODEC_TYPE_AUDIO;
218 c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE;
219
220 av_init_packet(&c->audio_pkt[i]);
221 c->audio_pkt[i].size = 0;
222 c->audio_pkt[i].data = c->audio_buf[i];
223 c->audio_pkt[i].stream_index = c->ast[i]->index;
224 c->audio_pkt[i].flags |= PKT_FLAG_KEY;
225 }
226 c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
227 c->ast[i]->codec->channels = 2;
228 c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
229 c->ast[i]->start_time = 0;
230 }
231 c->ach = i;
232
233 return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
234 }
235
236 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
237 {
238 const uint8_t* vsc_pack;
239 AVCodecContext* avctx;
240 int apt, is16_9;
241 int size = 0;
242
243 if (c->sys) {
244 avctx = c->vst->codec;
245
246 av_set_pts_info(c->vst, 64, c->sys->time_base.num,
247 c->sys->time_base.den);
248 avctx->time_base= c->sys->time_base;
249 if (!avctx->width){
250 avctx->width = c->sys->width;
251 avctx->height = c->sys->height;
252 }
253 avctx->pix_fmt = c->sys->pix_fmt;
254
255 /* finding out SAR is a little bit messy */
256 vsc_pack = dv_extract_pack(frame, dv_video_control);
257 apt = frame[4] & 0x07;
258 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
259 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
260 c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
261 avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1},
262 c->sys->time_base);
263 size = c->sys->frame_size;
264 }
265 return size;
266 }
267
268 /*
269 * The following 3 functions constitute our interface to the world
270 */
271
272 DVDemuxContext* dv_init_demux(AVFormatContext *s)
273 {
274 DVDemuxContext *c;
275
276 c = av_mallocz(sizeof(DVDemuxContext));
277 if (!c)
278 return NULL;
279
280 c->vst = av_new_stream(s, 0);
281 if (!c->vst) {
282 av_free(c);
283 return NULL;
284 }
285
286 c->sys = NULL;
287 c->fctx = s;
288 memset(c->ast, 0, sizeof(c->ast));
289 c->ach = 0;
290 c->frames = 0;
291 c->abytes = 0;
292
293 c->vst->codec->codec_type = CODEC_TYPE_VIDEO;
294 c->vst->codec->codec_id = CODEC_ID_DVVIDEO;
295 c->vst->codec->bit_rate = 25000000;
296 c->vst->start_time = 0;
297
298 return c;
299 }
300
301 int dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
302 {
303 int size = -1;
304 int i;
305
306 for (i = 0; i < c->ach; i++) {
307 if (c->ast[i] && c->audio_pkt[i].size) {
308 *pkt = c->audio_pkt[i];
309 c->audio_pkt[i].size = 0;
310 size = pkt->size;
311 break;
312 }
313 }
314
315 return size;
316 }
317
318 int dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
319 uint8_t* buf, int buf_size)
320 {
321 int size, i;
322 uint8_t *ppcm[4] = {0};
323
324 if (buf_size < DV_PROFILE_BYTES ||
325 !(c->sys = dv_frame_profile(buf)) ||
326 buf_size < c->sys->frame_size) {
327 return -1; /* Broken frame, or not enough data */
328 }
329
330 /* Queueing audio packet */
331 /* FIXME: in case of no audio/bad audio we have to do something */
332 size = dv_extract_audio_info(c, buf);
333 for (i = 0; i < c->ach; i++) {
334 c->audio_pkt[i].size = size;
335 c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate;
336 ppcm[i] = c->audio_buf[i];
337 }
338 dv_extract_audio(buf, ppcm, c->sys);
339 c->abytes += size;
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 }
349
350 /* Now it's time to return video packet */
351 size = dv_extract_video_info(c, buf);
352 av_init_packet(pkt);
353 pkt->data = buf;
354 pkt->size = size;
355 pkt->flags |= PKT_FLAG_KEY;
356 pkt->stream_index = c->vst->id;
357 pkt->pts = c->frames;
358
359 c->frames++;
360
361 return size;
362 }
363
364 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
365 int64_t timestamp, int flags)
366 {
367 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
368 const DVprofile* sys = dv_codec_profile(c->vst->codec);
369 int64_t offset;
370 int64_t size = url_fsize(s->pb);
371 int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
372
373 offset = sys->frame_size * timestamp;
374
375 if (offset > max_offset) offset = max_offset;
376 else if (offset < 0) offset = 0;
377
378 return offset;
379 }
380
381 void dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
382 {
383 c->frames= frame_offset;
384 if (c->ach)
385 c->abytes= av_rescale_q(c->frames, c->sys->time_base,
386 (AVRational){8, c->ast[0]->codec->bit_rate});
387 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
388 c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
389 }
390
391 /************************************************************
392 * Implementation of the easiest DV storage of all -- raw DV.
393 ************************************************************/
394
395 typedef struct RawDVContext {
396 DVDemuxContext* dv_demux;
397 uint8_t buf[DV_MAX_FRAME_SIZE];
398 } RawDVContext;
399
400 static int dv_read_header(AVFormatContext *s,
401 AVFormatParameters *ap)
402 {
403 unsigned state;
404 RawDVContext *c = s->priv_data;
405
406 c->dv_demux = dv_init_demux(s);
407 if (!c->dv_demux)
408 return -1;
409
410 state = get_be32(s->pb);
411 while ((state & 0xffffff7f) != 0x1f07003f) {
412 if (url_feof(s->pb)) {
413 av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
414 return -1;
415 }
416 state = (state << 8) | get_byte(s->pb);
417 }
418 AV_WB32(c->buf, state);
419
420 if (get_buffer(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) <= 0 ||
421 url_fseek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
422 return AVERROR(EIO);
423
424 c->dv_demux->sys = dv_frame_profile(c->buf);
425 if (!c->dv_demux->sys) {
426 av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n");
427 return -1;
428 }
429
430 s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size, (AVRational){8,1},
431 c->dv_demux->sys->time_base);
432
433 return 0;
434 }
435
436
437 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
438 {
439 int size;
440 RawDVContext *c = s->priv_data;
441
442 size = dv_get_packet(c->dv_demux, pkt);
443
444 if (size < 0) {
445 if (!c->dv_demux->sys)
446 return AVERROR(EIO);
447 size = c->dv_demux->sys->frame_size;
448 if (get_buffer(s->pb, c->buf, size) <= 0)
449 return AVERROR(EIO);
450
451 size = dv_produce_packet(c->dv_demux, pkt, c->buf, size);
452 }
453
454 return size;
455 }
456
457 static int dv_read_seek(AVFormatContext *s, int stream_index,
458 int64_t timestamp, int flags)
459 {
460 RawDVContext *r = s->priv_data;
461 DVDemuxContext *c = r->dv_demux;
462 int64_t offset = dv_frame_offset(s, c, timestamp, flags);
463
464 dv_offset_reset(c, offset / c->sys->frame_size);
465
466 offset = url_fseek(s->pb, offset, SEEK_SET);
467 return (offset < 0) ? offset : 0;
468 }
469
470 static int dv_read_close(AVFormatContext *s)
471 {
472 RawDVContext *c = s->priv_data;
473 av_free(c->dv_demux);
474 return 0;
475 }
476
477 #if CONFIG_DV_DEMUXER
478 AVInputFormat dv_demuxer = {
479 "dv",
480 NULL_IF_CONFIG_SMALL("DV video format"),
481 sizeof(RawDVContext),
482 NULL,
483 dv_read_header,
484 dv_read_packet,
485 dv_read_close,
486 dv_read_seek,
487 .extensions = "dv,dif",
488 };
489 #endif