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) support
12 * Copyright (c) 2006 Daniel Maas <dmaas@maasdigital.com>
13 *
14 * This file is part of FFmpeg.
15 *
16 * FFmpeg is free software; you can redistribute it and/or
17 * modify it under the terms of the GNU Lesser General Public
18 * License as published by the Free Software Foundation; either
19 * version 2.1 of the License, or (at your option) any later version.
20 *
21 * FFmpeg is distributed in the hope that it will be useful,
22 * but WITHOUT ANY WARRANTY; without even the implied warranty of
23 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
24 * Lesser General Public License for more details.
25 *
26 * You should have received a copy of the GNU Lesser General Public
27 * License along with FFmpeg; if not, write to the Free Software
28 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
29 */
30 #include <time.h>
31
32 #include "libavutil/time_internal.h"
33 #include "avformat.h"
34 #include "internal.h"
35 #include "libavcodec/dv_profile.h"
36 #include "libavcodec/dv.h"
37 #include "dv.h"
38 #include "libavutil/avassert.h"
39 #include "libavutil/fifo.h"
40 #include "libavutil/mathematics.h"
41 #include "libavutil/intreadwrite.h"
42 #include "libavutil/opt.h"
43 #include "libavutil/timecode.h"
44
45 #define MAX_AUDIO_FRAME_SIZE 192000 // 1 second of 48khz 32-bit audio
46
47 struct DVMuxContext {
48 AVClass *av_class;
49 const AVDVProfile* sys; /* current DV profile, e.g.: 525/60, 625/50 */
50 int n_ast; /* number of stereo audio streams (up to 2) */
51 AVStream *ast[4]; /* stereo audio streams */
52 AVFifoBuffer *audio_data[4]; /* FIFO for storing excessive amounts of PCM */
53 int frames; /* current frame number */
54 int64_t start_time; /* recording start time */
55 int has_audio; /* frame under construction has audio */
56 int has_video; /* frame under construction has video */
57 uint8_t frame_buf[DV_MAX_FRAME_SIZE]; /* frame under construction */
58 AVTimecode tc; /* timecode context */
59 };
60
61 static const int dv_aaux_packs_dist[12][9] = {
62 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
63 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
64 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
65 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
66 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
67 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
68 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
69 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
70 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
71 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
72 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
73 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
74 };
75
brktimegm(time_t secs,struct tm * tm)76 static void brktimegm(time_t secs, struct tm *tm)
77 {
78 tm = gmtime_r(&secs, tm);
79
80 tm->tm_year += 1900; /* unlike gmtime_r we store complete year here */
81 tm->tm_mon += 1; /* unlike gmtime_r tm_mon is from 1 to 12 */
82 }
83
dv_audio_frame_size(const AVDVProfile * sys,int frame,int sample_rate)84 static int dv_audio_frame_size(const AVDVProfile* sys, int frame, int sample_rate)
85 {
86 if ((sys->time_base.den == 25 || sys->time_base.den == 50) && sys->time_base.num == 1) {
87 if (sample_rate == 32000) return 1280;
88 else if (sample_rate == 44100) return 1764;
89 else return 1920;
90 }
91
92 av_assert0(sample_rate == 48000);
93
94 return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist) /
95 sizeof(sys->audio_samples_dist[0]))];
96 }
97
dv_write_pack(enum dv_pack_type pack_id,DVMuxContext * c,uint8_t * buf,int channel,int seq)98 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf, int channel, int seq)
99 {
100 struct tm tc;
101 time_t ct;
102 uint32_t timecode;
103 int audio_type = 0;
104
105 buf[0] = (uint8_t)pack_id;
106 switch (pack_id) {
107 case dv_timecode:
108 timecode = av_timecode_get_smpte_from_framenum(&c->tc, c->frames);
109 timecode |= 1<<23 | 1<<15 | 1<<7 | 1<<6; // biphase and binary group flags
110 AV_WB32(buf + 1, timecode);
111 break;
112 case dv_audio_source: /* AAUX source pack */
113 if (c->ast[channel]->codecpar->sample_rate == 44100) {
114 audio_type = 1;
115 } else if (c->ast[channel]->codecpar->sample_rate == 32000)
116 audio_type = 2;
117 buf[1] = (1 << 7) | /* locked mode -- SMPTE only supports locked mode */
118 (1 << 6) | /* reserved -- always 1 */
119 (dv_audio_frame_size(c->sys, c->frames, c->ast[channel]->codecpar->sample_rate) -
120 c->sys->audio_min_samples[audio_type]);
121 /* # of samples */
122 buf[2] = (0 << 7) | /* multi-stereo */
123 (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
124 (0 << 4) | /* pair bit: 0 -- one pair of channels */
125 (seq >= c->sys->difseg_size/2); /* audio mode (1st or 2nd channel) */
126 buf[3] = (1 << 7) | /* res */
127 (1 << 6) | /* multi-language flag */
128 (c->sys->dsf << 5) | /* system: 60fields/50fields */
129 (DV_PROFILE_IS_HD(c->sys) ? 0x3 : c->sys->video_stype ? 2 : 0); /* stype */
130 buf[4] = (1 << 7) | /* emphasis: 1 -- off */
131 (0 << 6) | /* emphasis time constant: 0 -- reserved */
132 (audio_type << 3) | /* frequency: 0 -- 48kHz, 1 -- 44,1kHz, 2 -- 32kHz */
133 0; /* quantization: 0 -- 16-bit linear, 1 -- 12-bit nonlinear */
134
135 break;
136 case dv_audio_control:
137 buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
138 (1 << 4) | /* input source: 1 -- digital input */
139 (3 << 2) | /* compression: 3 -- no information */
140 0; /* misc. info/SMPTE emphasis off */
141 buf[2] = (1 << 7) | /* recording start point: 1 -- no */
142 (1 << 6) | /* recording end point: 1 -- no */
143 (1 << 3) | /* recording mode: 1 -- original */
144 7;
145 buf[3] = (1 << 7) | /* direction: 1 -- forward */
146 (c->sys->pix_fmt == AV_PIX_FMT_YUV420P ? 0x20 : /* speed */
147 c->sys->ltc_divisor * 4);
148 buf[4] = (1 << 7) | /* reserved -- always 1 */
149 0x7f; /* genre category */
150 break;
151 case dv_audio_recdate:
152 case dv_video_recdate: /* VAUX recording date */
153 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
154 c->sys->time_base.den, AV_ROUND_DOWN);
155 brktimegm(ct, &tc);
156 buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
157 /* 0xff is very likely to be "unknown" */
158 buf[2] = (3 << 6) | /* reserved -- always 1 */
159 ((tc.tm_mday / 10) << 4) | /* Tens of day */
160 (tc.tm_mday % 10); /* Units of day */
161 buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
162 ((tc.tm_mon / 10) << 4) | /* Tens of month */
163 (tc.tm_mon % 10); /* Units of month */
164 buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
165 (tc.tm_year % 10); /* Units of year */
166 break;
167 case dv_audio_rectime: /* AAUX recording time */
168 case dv_video_rectime: /* VAUX recording time */
169 ct = c->start_time + av_rescale_rnd(c->frames, c->sys->time_base.num,
170 c->sys->time_base.den, AV_ROUND_DOWN);
171 brktimegm(ct, &tc);
172 buf[1] = (3 << 6) | /* reserved -- always 1 */
173 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
174 buf[2] = (1 << 7) | /* reserved -- always 1 */
175 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
176 (tc.tm_sec % 10); /* Units of seconds */
177 buf[3] = (1 << 7) | /* reserved -- always 1 */
178 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
179 (tc.tm_min % 10); /* Units of minutes */
180 buf[4] = (3 << 6) | /* reserved -- always 1 */
181 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
182 (tc.tm_hour % 10); /* Units of hours */
183 break;
184 default:
185 buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
186 }
187 return 5;
188 }
189
dv_inject_audio(DVMuxContext * c,int channel,uint8_t * frame_ptr)190 static void dv_inject_audio(DVMuxContext *c, int channel, uint8_t* frame_ptr)
191 {
192 int i, j, d, of, size;
193 size = 4 * dv_audio_frame_size(c->sys, c->frames, c->ast[channel]->codecpar->sample_rate);
194 frame_ptr += channel * c->sys->difseg_size * 150 * 80;
195 for (i = 0; i < c->sys->difseg_size; i++) {
196 frame_ptr += 6 * 80; /* skip DIF segment header */
197 for (j = 0; j < 9; j++) {
198 dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3], channel, i);
199 for (d = 8; d < 80; d+=2) {
200 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
201 if (of*2 >= size)
202 continue;
203
204 frame_ptr[d] = *av_fifo_peek2(c->audio_data[channel], of*2+1); // FIXME: maybe we have to admit
205 frame_ptr[d+1] = *av_fifo_peek2(c->audio_data[channel], of*2); // that DV is a big-endian PCM
206 }
207 frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
208 }
209 }
210 }
211
dv_inject_metadata(DVMuxContext * c,uint8_t * frame)212 static void dv_inject_metadata(DVMuxContext *c, uint8_t* frame)
213 {
214 int j, k;
215 uint8_t* buf;
216 int seq = 0;
217
218 for (buf = frame; buf < frame + c->sys->frame_size; buf += 150 * 80, seq++) {
219 /* DV subcode: 2nd and 3d DIFs */
220 for (j = 80; j < 80 * 3; j += 80) {
221 for (k = 6; k < 6 * 8; k += 8)
222 dv_write_pack(dv_timecode, c, &buf[j+k], 0, seq);
223
224 if (((long)(buf-frame)/(c->sys->frame_size/(c->sys->difseg_size*c->sys->n_difchan))%c->sys->difseg_size) > 5) { /* FIXME: is this really needed ? */
225 dv_write_pack(dv_video_recdate, c, &buf[j+14], 0, seq);
226 dv_write_pack(dv_video_rectime, c, &buf[j+22], 0, seq);
227 dv_write_pack(dv_video_recdate, c, &buf[j+38], 0, seq);
228 dv_write_pack(dv_video_rectime, c, &buf[j+46], 0, seq);
229 }
230 }
231
232 /* DV VAUX: 4th, 5th and 6th 3DIFs */
233 for (j = 80*3 + 3; j < 80*6; j += 80) {
234 dv_write_pack(dv_video_recdate, c, &buf[j+5* 2], 0, seq);
235 dv_write_pack(dv_video_rectime, c, &buf[j+5* 3], 0, seq);
236 dv_write_pack(dv_video_recdate, c, &buf[j+5*11], 0, seq);
237 dv_write_pack(dv_video_rectime, c, &buf[j+5*12], 0, seq);
238 }
239 }
240 }
241
242 /*
243 * The following 3 functions constitute our interface to the world
244 */
245
dv_assemble_frame(AVFormatContext * s,DVMuxContext * c,AVStream * st,uint8_t * data,int data_size,uint8_t ** frame)246 static int dv_assemble_frame(AVFormatContext *s,
247 DVMuxContext *c, AVStream* st,
248 uint8_t* data, int data_size, uint8_t** frame)
249 {
250 int i, reqasize;
251
252 *frame = &c->frame_buf[0];
253
254 switch (st->codecpar->codec_type) {
255 case AVMEDIA_TYPE_VIDEO:
256 /* FIXME: we have to have more sensible approach than this one */
257 if (c->has_video)
258 av_log(s, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
259 if (data_size != c->sys->frame_size) {
260 av_log(s, AV_LOG_ERROR, "Unexpected frame size, %d != %d\n",
261 data_size, c->sys->frame_size);
262 return AVERROR(ENOSYS);
263 }
264
265 memcpy(*frame, data, c->sys->frame_size);
266 c->has_video = 1;
267 break;
268 case AVMEDIA_TYPE_AUDIO:
269 for (i = 0; i < c->n_ast && st != c->ast[i]; i++);
270
271 /* FIXME: we have to have more sensible approach than this one */
272 if (av_fifo_size(c->audio_data[i]) + data_size >= 100*MAX_AUDIO_FRAME_SIZE)
273 av_log(s, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
274 av_fifo_generic_write(c->audio_data[i], data, data_size, NULL);
275
276 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames, st->codecpar->sample_rate);
277
278 /* Let us see if we've got enough audio for one DV frame. */
279 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
280
281 break;
282 default:
283 break;
284 }
285
286 /* Let us see if we have enough data to construct one DV frame. */
287 if (c->has_video == 1 && c->has_audio + 1 == 1 << c->n_ast) {
288 dv_inject_metadata(c, *frame);
289 c->has_audio = 0;
290 for (i=0; i < c->n_ast; i++) {
291 dv_inject_audio(c, i, *frame);
292 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames, c->ast[i]->codecpar->sample_rate);
293 av_fifo_drain(c->audio_data[i], reqasize);
294 c->has_audio |= ((reqasize <= av_fifo_size(c->audio_data[i])) << i);
295 }
296
297 c->has_video = 0;
298
299 c->frames++;
300
301 return c->sys->frame_size;
302 }
303
304 return 0;
305 }
306
dv_init_mux(AVFormatContext * s)307 static DVMuxContext* dv_init_mux(AVFormatContext* s)
308 {
309 DVMuxContext *c = s->priv_data;
310 AVStream *vst = NULL;
311 int i;
312
313 /* we support at most 1 video and 2 audio streams */
314 if (s->nb_streams > 5)
315 return NULL;
316
317 /* We have to sort out where audio and where video stream is */
318 for (i=0; i<s->nb_streams; i++) {
319 AVStream *st = s->streams[i];
320 switch (st->codecpar->codec_type) {
321 case AVMEDIA_TYPE_VIDEO:
322 if (vst) return NULL;
323 if (st->codecpar->codec_id != AV_CODEC_ID_DVVIDEO)
324 goto bail_out;
325 vst = st;
326 break;
327 case AVMEDIA_TYPE_AUDIO:
328 if (c->n_ast > 1) return NULL;
329 /* Some checks -- DV format is very picky about its incoming streams */
330 if(st->codecpar->codec_id != AV_CODEC_ID_PCM_S16LE ||
331 st->codecpar->channels != 2)
332 goto bail_out;
333 if (st->codecpar->sample_rate != 48000 &&
334 st->codecpar->sample_rate != 44100 &&
335 st->codecpar->sample_rate != 32000 )
336 goto bail_out;
337 c->ast[c->n_ast++] = st;
338 break;
339 default:
340 goto bail_out;
341 }
342 }
343
344 if (!vst)
345 goto bail_out;
346
347 c->sys = av_dv_codec_profile2(vst->codecpar->width, vst->codecpar->height,
348 vst->codecpar->format, vst->time_base);
349 if (!c->sys)
350 goto bail_out;
351
352 if ((c->sys->time_base.den != 25 && c->sys->time_base.den != 50) || c->sys->time_base.num != 1) {
353 if (c->ast[0] && c->ast[0]->codecpar->sample_rate != 48000)
354 goto bail_out;
355 if (c->ast[1] && c->ast[1]->codecpar->sample_rate != 48000)
356 goto bail_out;
357 }
358
359 if (((c->n_ast > 1) && (c->sys->n_difchan < 2)) ||
360 ((c->n_ast > 2) && (c->sys->n_difchan < 4))) {
361 /* only 2 stereo pairs allowed in 50Mbps mode */
362 goto bail_out;
363 }
364
365 /* Ok, everything seems to be in working order */
366 c->frames = 0;
367 c->has_audio = 0;
368 c->has_video = 0;
369 ff_parse_creation_time_metadata(s, &c->start_time, 1);
370
371 for (i=0; i < c->n_ast; i++) {
372 if (c->ast[i] && !(c->audio_data[i]=av_fifo_alloc_array(100, MAX_AUDIO_FRAME_SIZE))) {
373 goto bail_out;
374 }
375 }
376
377 return c;
378
379 bail_out:
380 return NULL;
381 }
382
dv_write_header(AVFormatContext * s)383 static int dv_write_header(AVFormatContext *s)
384 {
385 AVRational rate;
386 DVMuxContext *dvc = s->priv_data;
387 AVDictionaryEntry *tcr = av_dict_get(s->metadata, "timecode", NULL, 0);
388
389 if (!dv_init_mux(s)) {
390 av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
391 "Make sure that you supply exactly two streams:\n"
392 " video: 25fps or 29.97fps, audio: 2ch/48|44|32kHz/PCM\n"
393 " (50Mbps allows an optional second audio stream)\n");
394 return -1;
395 }
396 rate.num = dvc->sys->ltc_divisor;
397 rate.den = 1;
398 if (!tcr) { // no global timecode, look into the streams
399 int i;
400 for (i = 0; i < s->nb_streams; i++) {
401 tcr = av_dict_get(s->streams[i]->metadata, "timecode", NULL, 0);
402 if (tcr)
403 break;
404 }
405 }
406 if (tcr && av_timecode_init_from_string(&dvc->tc, rate, tcr->value, s) >= 0)
407 return 0;
408 return av_timecode_init(&dvc->tc, rate, 0, 0, s);
409 }
410
dv_write_packet(struct AVFormatContext * s,AVPacket * pkt)411 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
412 {
413 uint8_t* frame;
414 int fsize;
415
416 fsize = dv_assemble_frame(s, s->priv_data, s->streams[pkt->stream_index],
417 pkt->data, pkt->size, &frame);
418 if (fsize < 0) {
419 return fsize;
420 }
421 avio_write(s->pb, frame, fsize);
422 return 0;
423 }
424
425 /*
426 * We might end up with some extra A/V data without matching counterpart.
427 * E.g. video data without enough audio to write the complete frame.
428 * Currently we simply drop the last frame. I don't know whether this
429 * is the best strategy of all
430 */
dv_deinit(AVFormatContext * s)431 static void dv_deinit(AVFormatContext *s)
432 {
433 DVMuxContext *c = s->priv_data;
434
435 for (int i = 0; i < c->n_ast; i++)
436 av_fifo_freep(&c->audio_data[i]);
437 }
438
439 AVOutputFormat ff_dv_muxer = {
440 .name = "dv",
441 .long_name = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
442 .extensions = "dv",
443 .priv_data_size = sizeof(DVMuxContext),
444 .audio_codec = AV_CODEC_ID_PCM_S16LE,
445 .video_codec = AV_CODEC_ID_DVVIDEO,
446 .write_header = dv_write_header,
447 .write_packet = dv_write_packet,
448 .deinit = dv_deinit,
449 };
450