changing AVCodecContext codec -> *codec in AVStream so additions to AVCodecContext...
[libav.git] / libavformat / dv.c
1 /*
2 * General DV muxer/demuxer
3 * Copyright (c) 2003 Roman Shaposhnick
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 * This library is free software; you can redistribute it and/or
12 * modify it under the terms of the GNU Lesser General Public
13 * License as published by the Free Software Foundation; either
14 * version 2 of the License, or (at your option) any later version.
15 *
16 * This library is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * Lesser General Public License for more details.
20 *
21 * You should have received a copy of the GNU Lesser General Public
22 * License along with this library; if not, write to the Free Software
23 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
24 */
25 #include <time.h>
26 #include "avformat.h"
27 #include "dvdata.h"
28 #include "dv.h"
29
30 struct DVDemuxContext {
31 AVFormatContext* fctx;
32 AVStream* vst;
33 AVStream* ast[2];
34 AVPacket audio_pkt[2];
35 uint8_t audio_buf[2][8192];
36 int ach;
37 int frames;
38 uint64_t abytes;
39 };
40
41 struct DVMuxContext {
42 const DVprofile* sys; /* Current DV profile. E.g.: 525/60, 625/50 */
43 uint8_t frame_buf[144000]; /* frame under contruction */
44 FifoBuffer audio_data; /* Fifo for storing excessive amounts of PCM */
45 int frames; /* Number of a current frame */
46 time_t start_time; /* Start time of recording */
47 uint8_t aspect; /* Aspect ID 0 - 4:3, 7 - 16:9 */
48 int has_audio; /* frame under contruction has audio */
49 int has_video; /* frame under contruction has video */
50 };
51
52 enum dv_section_type {
53 dv_sect_header = 0x1f,
54 dv_sect_subcode = 0x3f,
55 dv_sect_vaux = 0x56,
56 dv_sect_audio = 0x76,
57 dv_sect_video = 0x96,
58 };
59
60 enum dv_pack_type {
61 dv_header525 = 0x3f, /* see dv_write_pack for important details on */
62 dv_header625 = 0xbf, /* these two packs */
63 dv_timecode = 0x13,
64 dv_audio_source = 0x50,
65 dv_audio_control = 0x51,
66 dv_audio_recdate = 0x52,
67 dv_audio_rectime = 0x53,
68 dv_video_source = 0x60,
69 dv_video_control = 0x61,
70 dv_viedo_recdate = 0x62,
71 dv_video_rectime = 0x63,
72 dv_unknown_pack = 0xff,
73 };
74
75
76
77 /*
78 * The reason why the following three big ugly looking tables are
79 * here is my lack of DV spec IEC 61834. The tables were basically
80 * constructed to make code that places packs in SSYB, VAUX and
81 * AAUX blocks very simple and table-driven. They conform to the
82 * SMPTE 314M and the output of my personal DV camcorder, neither
83 * of which is sufficient for a reliable DV stream producing. Thus
84 * while code is still in development I'll be gathering input from
85 * people with different DV equipment and modifying the tables to
86 * accommodate all the quirks. Later on, if possible, some of them
87 * will be folded into smaller tables and/or switch-if logic. For
88 * now, my only excuse is -- they don't eat up that much of a space.
89 */
90
91 static const int dv_ssyb_packs_dist[12][6] = {
92 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
93 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
94 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
95 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
96 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
97 { 0x13, 0x13, 0x13, 0x13, 0x13, 0x13 },
98 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
99 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
100 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
101 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
102 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
103 { 0x13, 0x62, 0x63, 0x13, 0x62, 0x63 },
104 };
105
106 static const int dv_vaux_packs_dist[12][15] = {
107 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
108 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
109 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
110 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
111 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
112 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
113 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
114 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
115 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
116 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
117 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
118 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
119 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
120 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
121 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
122 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
123 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
124 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
125 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
126 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
127 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
128 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
129 { 0x60, 0x61, 0x62, 0x63, 0xff, 0xff, 0xff, 0xff, 0xff,
130 0x60, 0x61, 0x62, 0x63, 0xff, 0xff },
131 };
132
133 static const int dv_aaux_packs_dist[12][9] = {
134 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
135 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
136 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
137 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
138 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
139 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
140 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
141 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
142 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
143 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
144 { 0xff, 0xff, 0xff, 0x50, 0x51, 0x52, 0x53, 0xff, 0xff },
145 { 0x50, 0x51, 0x52, 0x53, 0xff, 0xff, 0xff, 0xff, 0xff },
146 };
147
148 static inline uint16_t dv_audio_12to16(uint16_t sample)
149 {
150 uint16_t shift, result;
151
152 sample = (sample < 0x800) ? sample : sample | 0xf000;
153 shift = (sample & 0xf00) >> 8;
154
155 if (shift < 0x2 || shift > 0xd) {
156 result = sample;
157 } else if (shift < 0x8) {
158 shift--;
159 result = (sample - (256 * shift)) << shift;
160 } else {
161 shift = 0xe - shift;
162 result = ((sample + ((256 * shift) + 1)) << shift) - 1;
163 }
164
165 return result;
166 }
167
168 static int dv_audio_frame_size(const DVprofile* sys, int frame)
169 {
170 return sys->audio_samples_dist[frame % (sizeof(sys->audio_samples_dist)/
171 sizeof(sys->audio_samples_dist[0]))];
172 }
173
174 static int dv_write_pack(enum dv_pack_type pack_id, DVMuxContext *c, uint8_t* buf)
175 {
176 struct tm tc;
177 time_t ct;
178 int ltc_frame;
179
180 buf[0] = (uint8_t)pack_id;
181 switch (pack_id) {
182 case dv_header525: /* I can't imagine why these two weren't defined as real */
183 case dv_header625: /* packs in SMPTE314M -- they definitely look like ones */
184 buf[1] = 0xf8 | /* reserved -- always 1 */
185 (0 & 0x07); /* APT: Track application ID */
186 buf[2] = (0 << 7) | /* TF1: audio data is 0 - valid; 1 - invalid */
187 (0x0f << 3) | /* reserved -- always 1 */
188 (0 & 0x07); /* AP1: Audio application ID */
189 buf[3] = (0 << 7) | /* TF2: video data is 0 - valid; 1 - invalid */
190 (0x0f << 3) | /* reserved -- always 1 */
191 (0 & 0x07); /* AP2: Video application ID */
192 buf[4] = (0 << 7) | /* TF3: subcode(SSYB) is 0 - valid; 1 - invalid */
193 (0x0f << 3) | /* reserved -- always 1 */
194 (0 & 0x07); /* AP3: Subcode application ID */
195 break;
196 case dv_timecode:
197 ct = (time_t)(c->frames / ((float)c->sys->frame_rate /
198 (float)c->sys->frame_rate_base));
199 brktimegm(ct, &tc);
200 /*
201 * LTC drop-frame frame counter drops two frames (0 and 1) every
202 * minute, unless it is exactly divisible by 10
203 */
204 ltc_frame = (c->frames + 2*ct/60 - 2*ct/600) % c->sys->ltc_divisor;
205 buf[1] = (0 << 7) | /* Color fame: 0 - unsync; 1 - sync mode */
206 (1 << 6) | /* Drop frame timecode: 0 - nondrop; 1 - drop */
207 ((ltc_frame / 10) << 4) | /* Tens of frames */
208 (ltc_frame % 10); /* Units of frames */
209 buf[2] = (1 << 7) | /* Biphase mark polarity correction: 0 - even; 1 - odd */
210 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
211 (tc.tm_sec % 10); /* Units of seconds */
212 buf[3] = (1 << 7) | /* Binary group flag BGF0 */
213 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
214 (tc.tm_min % 10); /* Units of minutes */
215 buf[4] = (1 << 7) | /* Binary group flag BGF2 */
216 (1 << 6) | /* Binary group flag BGF1 */
217 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
218 (tc.tm_hour % 10); /* Units of hours */
219 break;
220 case dv_audio_source: /* AAUX source pack */
221 buf[1] = (0 << 7) | /* locked mode */
222 (1 << 6) | /* reserved -- always 1 */
223 (dv_audio_frame_size(c->sys, c->frames) -
224 c->sys->audio_min_samples[0]);
225 /* # of samples */
226 buf[2] = (0 << 7) | /* multi-stereo */
227 (0 << 5) | /* #of audio channels per block: 0 -- 1 channel */
228 (0 << 4) | /* pair bit: 0 -- one pair of channels */
229 0; /* audio mode */
230 buf[3] = (1 << 7) | /* res */
231 (1 << 6) | /* multi-language flag */
232 (c->sys->dsf << 5) | /* system: 60fields/50fields */
233 0; /* definition: 0 -- SD (525/625) */
234 buf[4] = (1 << 7) | /* emphasis: 1 -- off */
235 (0 << 6) | /* emphasis time constant: 0 -- reserved */
236 (0 << 3) | /* frequency: 0 -- 48Khz, 1 -- 44,1Khz, 2 -- 32Khz */
237 0; /* quantization: 0 -- 16bit linear, 1 -- 12bit nonlinear */
238 break;
239 case dv_audio_control:
240 buf[1] = (0 << 6) | /* copy protection: 0 -- unrestricted */
241 (1 << 4) | /* input source: 1 -- digital input */
242 (3 << 2) | /* compression: 3 -- no information */
243 0; /* misc. info/SMPTE emphasis off */
244 buf[2] = (1 << 7) | /* recording start point: 1 -- no */
245 (1 << 6) | /* recording end point: 1 -- no */
246 (1 << 3) | /* recording mode: 1 -- original */
247 7;
248 buf[3] = (1 << 7) | /* direction: 1 -- forward */
249 0x20; /* speed */
250 buf[4] = (1 << 7) | /* reserved -- always 1 */
251 0x7f; /* genre category */
252 break;
253 case dv_audio_recdate:
254 case dv_viedo_recdate: /* VAUX recording date */
255 ct = c->start_time + (time_t)(c->frames /
256 ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
257 brktimegm(ct, &tc);
258 buf[1] = 0xff; /* ds, tm, tens of time zone, units of time zone */
259 /* 0xff is very likely to be "unknown" */
260 buf[2] = (3 << 6) | /* reserved -- always 1 */
261 ((tc.tm_mday / 10) << 4) | /* Tens of day */
262 (tc.tm_mday % 10); /* Units of day */
263 buf[3] = /* we set high 4 bits to 0, shouldn't we set them to week? */
264 ((tc.tm_mon / 10) << 4) | /* Tens of month */
265 (tc.tm_mon % 10); /* Units of month */
266 buf[4] = (((tc.tm_year % 100) / 10) << 4) | /* Tens of year */
267 (tc.tm_year % 10); /* Units of year */
268 break;
269 case dv_audio_rectime: /* AAUX recording time */
270 case dv_video_rectime: /* VAUX recording time */
271 ct = c->start_time + (time_t)(c->frames /
272 ((float)c->sys->frame_rate / (float)c->sys->frame_rate_base));
273 brktimegm(ct, &tc);
274 buf[1] = (3 << 6) | /* reserved -- always 1 */
275 0x3f; /* tens of frame, units of frame: 0x3f - "unknown" ? */
276 buf[2] = (1 << 7) | /* reserved -- always 1 */
277 ((tc.tm_sec / 10) << 4) | /* Tens of seconds */
278 (tc.tm_sec % 10); /* Units of seconds */
279 buf[3] = (1 << 7) | /* reserved -- always 1 */
280 ((tc.tm_min / 10) << 4) | /* Tens of minutes */
281 (tc.tm_min % 10); /* Units of minutes */
282 buf[4] = (3 << 6) | /* reserved -- always 1 */
283 ((tc.tm_hour / 10) << 4) | /* Tens of hours */
284 (tc.tm_hour % 10); /* Units of hours */
285 break;
286 case dv_video_source:
287 buf[1] = 0xff; /* reserved -- always 1 */
288 buf[2] = (1 << 7) | /* B/W: 0 - b/w, 1 - color */
289 (1 << 6) | /* following CLF is valid - 0, invalid - 1 */
290 (3 << 4) | /* CLF: color frames id (see ITU-R BT.470-4) */
291 0xf; /* reserved -- always 1 */
292 buf[3] = (3 << 6) | /* reserved -- always 1 */
293 (c->sys->dsf << 5) | /* system: 60fields/50fields */
294 0; /* signal type video compression */
295 buf[4] = 0xff; /* VISC: 0xff -- no information */
296 break;
297 case dv_video_control:
298 buf[1] = (0 << 6) | /* Copy generation management (CGMS) 0 -- free */
299 0x3f; /* reserved -- always 1 */
300 buf[2] = 0xc8 | /* reserved -- always b11001xxx */
301 c->aspect;
302 buf[3] = (1 << 7) | /* Frame/field flag 1 -- frame, 0 -- field */
303 (1 << 6) | /* First/second field flag 0 -- field 2, 1 -- field 1 */
304 (1 << 5) | /* Frame change flag 0 -- same picture as before, 1 -- different */
305 (1 << 4) | /* 1 - interlaced, 0 - noninterlaced */
306 0xc; /* reserved -- always b1100 */
307 buf[4] = 0xff; /* reserved -- always 1 */
308 break;
309 default:
310 buf[1] = buf[2] = buf[3] = buf[4] = 0xff;
311 }
312 return 5;
313 }
314
315 static inline int dv_write_dif_id(enum dv_section_type t, uint8_t seq_num,
316 uint8_t dif_num, uint8_t* buf)
317 {
318 buf[0] = (uint8_t)t; /* Section type */
319 buf[1] = (seq_num<<4) | /* DIF seq number 0-9 for 525/60; 0-11 for 625/50 */
320 (0 << 3) | /* FSC: for 50Mb/s 0 - first channel; 1 - second */
321 7; /* reserved -- always 1 */
322 buf[2] = dif_num; /* DIF block number Video: 0-134, Audio: 0-8 */
323 return 3;
324 }
325
326 static inline int dv_write_ssyb_id(uint8_t syb_num, uint8_t fr, uint8_t* buf)
327 {
328 if (syb_num == 0 || syb_num == 6) {
329 buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
330 (0<<4) | /* AP3 (Subcode application ID) */
331 0x0f; /* reserved -- always 1 */
332 }
333 else if (syb_num == 11) {
334 buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
335 0x7f; /* reserved -- always 1 */
336 }
337 else {
338 buf[0] = (fr<<7) | /* FR ID 1 - first half of each channel; 0 - second */
339 (0<<4) | /* APT (Track application ID) */
340 0x0f; /* reserved -- always 1 */
341 }
342 buf[1] = 0xf0 | /* reserved -- always 1 */
343 (syb_num & 0x0f); /* SSYB number 0 - 11 */
344 buf[2] = 0xff; /* reserved -- always 1 */
345 return 3;
346 }
347
348 static void dv_format_frame(DVMuxContext *c, uint8_t* buf)
349 {
350 int i, j, k;
351
352 for (i = 0; i < c->sys->difseg_size; i++) {
353 memset(buf, 0xff, 80 * 6); /* First 6 DIF blocks are for control data */
354
355 /* DV header: 1DIF */
356 buf += dv_write_dif_id(dv_sect_header, i, 0, buf);
357 buf += dv_write_pack((c->sys->dsf ? dv_header625 : dv_header525), c, buf);
358 buf += 72; /* unused bytes */
359
360 /* DV subcode: 2DIFs */
361 for (j = 0; j < 2; j++) {
362 buf += dv_write_dif_id( dv_sect_subcode, i, j, buf);
363 for (k = 0; k < 6; k++) {
364 buf += dv_write_ssyb_id(k, (i < c->sys->difseg_size/2), buf);
365 buf += dv_write_pack(dv_ssyb_packs_dist[i][k], c, buf);
366 }
367 buf += 29; /* unused bytes */
368 }
369
370 /* DV VAUX: 3DIFs */
371 for (j = 0; j < 3; j++) {
372 buf += dv_write_dif_id(dv_sect_vaux, i, j, buf);
373 for (k = 0; k < 15 ; k++)
374 buf += dv_write_pack(dv_vaux_packs_dist[i][k], c, buf);
375 buf += 2; /* unused bytes */
376 }
377
378 /* DV Audio/Video: 135 Video DIFs + 9 Audio DIFs */
379 for (j = 0; j < 135; j++) {
380 if (j%15 == 0) {
381 memset(buf, 0xff, 80);
382 buf += dv_write_dif_id(dv_sect_audio, i, j/15, buf);
383 buf += 77; /* audio control & shuffled PCM audio */
384 }
385 buf += dv_write_dif_id(dv_sect_video, i, j, buf);
386 buf += 77; /* 1 video macro block: 1 bytes control
387 4 * 14 bytes Y 8x8 data
388 10 bytes Cr 8x8 data
389 10 bytes Cb 8x8 data */
390 }
391 }
392 }
393
394 static void dv_inject_audio(DVMuxContext *c, const uint8_t* pcm, uint8_t* frame_ptr)
395 {
396 int i, j, d, of, size;
397 size = 4 * dv_audio_frame_size(c->sys, c->frames);
398 for (i = 0; i < c->sys->difseg_size; i++) {
399 frame_ptr += 6 * 80; /* skip DIF segment header */
400 for (j = 0; j < 9; j++) {
401 dv_write_pack(dv_aaux_packs_dist[i][j], c, &frame_ptr[3]);
402 for (d = 8; d < 80; d+=2) {
403 of = c->sys->audio_shuffle[i][j] + (d - 8)/2 * c->sys->audio_stride;
404 if (of*2 >= size)
405 continue;
406
407 frame_ptr[d] = pcm[of*2+1]; // FIXME: may be we have to admit
408 frame_ptr[d+1] = pcm[of*2]; // that DV is a big endian PCM
409 }
410 frame_ptr += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
411 }
412 }
413 }
414
415 static void dv_inject_video(DVMuxContext *c, const uint8_t* video_data, uint8_t* frame_ptr)
416 {
417 int i, j;
418 int ptr = 0;
419
420 for (i = 0; i < c->sys->difseg_size; i++) {
421 ptr += 6 * 80; /* skip DIF segment header */
422 for (j = 0; j < 135; j++) {
423 if (j%15 == 0)
424 ptr += 80; /* skip Audio DIF */
425 ptr += 3;
426 memcpy(frame_ptr + ptr, video_data + ptr, 77);
427 ptr += 77;
428 }
429 }
430 }
431
432 /*
433 * This is the dumbest implementation of all -- it simply looks at
434 * a fixed offset and if pack isn't there -- fails. We might want
435 * to have a fallback mechanism for complete search of missing packs.
436 */
437 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
438 {
439 int offs;
440
441 switch (t) {
442 case dv_audio_source:
443 offs = (80*6 + 80*16*3 + 3);
444 break;
445 case dv_audio_control:
446 offs = (80*6 + 80*16*4 + 3);
447 break;
448 case dv_video_control:
449 offs = (80*5 + 48 + 5);
450 break;
451 default:
452 return NULL;
453 }
454
455 return (frame[offs] == t ? &frame[offs] : NULL);
456 }
457
458 /*
459 * There's a couple of assumptions being made here:
460 * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
461 * We can pass them upwards when ffmpeg will be ready to deal with them.
462 * 2. We don't do software emphasis.
463 * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
464 * are converted into 16bit linear ones.
465 */
466 static int dv_extract_audio(uint8_t* frame, uint8_t* pcm, uint8_t* pcm2)
467 {
468 int size, i, j, d, of, smpls, freq, quant, half_ch;
469 uint16_t lc, rc;
470 const DVprofile* sys;
471 const uint8_t* as_pack;
472
473 as_pack = dv_extract_pack(frame, dv_audio_source);
474 if (!as_pack) /* No audio ? */
475 return 0;
476
477 sys = dv_frame_profile(frame);
478 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
479 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
480 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
481
482 if (quant > 1)
483 return -1; /* Unsupported quantization */
484
485 size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
486 half_ch = sys->difseg_size/2;
487
488 /* for each DIF segment */
489 for (i = 0; i < sys->difseg_size; i++) {
490 frame += 6 * 80; /* skip DIF segment header */
491 if (quant == 1 && i == half_ch) {
492 if (!pcm2)
493 break;
494 else
495 pcm = pcm2;
496 }
497
498 for (j = 0; j < 9; j++) {
499 for (d = 8; d < 80; d += 2) {
500 if (quant == 0) { /* 16bit quantization */
501 of = sys->audio_shuffle[i][j] + (d - 8)/2 * sys->audio_stride;
502 if (of*2 >= size)
503 continue;
504
505 pcm[of*2] = frame[d+1]; // FIXME: may be we have to admit
506 pcm[of*2+1] = frame[d]; // that DV is a big endian PCM
507 if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
508 pcm[of*2+1] = 0;
509 } else { /* 12bit quantization */
510 lc = ((uint16_t)frame[d] << 4) |
511 ((uint16_t)frame[d+2] >> 4);
512 rc = ((uint16_t)frame[d+1] << 4) |
513 ((uint16_t)frame[d+2] & 0x0f);
514 lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
515 rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
516
517 of = sys->audio_shuffle[i%half_ch][j] + (d - 8)/3 * sys->audio_stride;
518 if (of*2 >= size)
519 continue;
520
521 pcm[of*2] = lc & 0xff; // FIXME: may be we have to admit
522 pcm[of*2+1] = lc >> 8; // that DV is a big endian PCM
523 of = sys->audio_shuffle[i%half_ch+half_ch][j] +
524 (d - 8)/3 * sys->audio_stride;
525 pcm[of*2] = rc & 0xff; // FIXME: may be we have to admit
526 pcm[of*2+1] = rc >> 8; // that DV is a big endian PCM
527 ++d;
528 }
529 }
530
531 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
532 }
533 }
534
535 return size;
536 }
537
538 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
539 {
540 const uint8_t* as_pack;
541 const DVprofile* sys;
542 int freq, smpls, quant, i, ach;
543
544 sys = dv_frame_profile(frame);
545 as_pack = dv_extract_pack(frame, dv_audio_source);
546 if (!as_pack || !sys) { /* No audio ? */
547 c->ach = 0;
548 return 0;
549 }
550
551 smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
552 freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
553 quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
554 ach = (quant && freq == 2) ? 2 : 1;
555
556 /* Dynamic handling of the audio streams in DV */
557 for (i=0; i<ach; i++) {
558 if (!c->ast[i]) {
559 c->ast[i] = av_new_stream(c->fctx, 0);
560 if (!c->ast[i])
561 break;
562 av_set_pts_info(c->ast[i], 64, 1, 30000);
563 c->ast[i]->codec->codec_type = CODEC_TYPE_AUDIO;
564 c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE;
565
566 av_init_packet(&c->audio_pkt[i]);
567 c->audio_pkt[i].size = 0;
568 c->audio_pkt[i].data = c->audio_buf[i];
569 c->audio_pkt[i].stream_index = c->ast[i]->index;
570 c->audio_pkt[i].flags |= PKT_FLAG_KEY;
571 }
572 c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
573 c->ast[i]->codec->channels = 2;
574 c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
575 c->ast[i]->start_time = 0;
576 }
577 c->ach = i;
578
579 return (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
580 }
581
582 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
583 {
584 const DVprofile* sys;
585 const uint8_t* vsc_pack;
586 AVCodecContext* avctx;
587 int apt, is16_9;
588 int size = 0;
589
590 sys = dv_frame_profile(frame);
591 if (sys) {
592 avctx = c->vst->codec;
593
594 av_set_pts_info(c->vst, 64, sys->frame_rate_base, sys->frame_rate);
595 avctx->time_base= (AVRational){sys->frame_rate_base, sys->frame_rate};
596 avctx->width = sys->width;
597 avctx->height = sys->height;
598 avctx->pix_fmt = sys->pix_fmt;
599
600 /* finding out SAR is a little bit messy */
601 vsc_pack = dv_extract_pack(frame, dv_video_control);
602 apt = frame[4] & 0x07;
603 is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
604 (!apt && (vsc_pack[2] & 0x07) == 0x07)));
605 avctx->sample_aspect_ratio = sys->sar[is16_9];
606
607 size = sys->frame_size;
608 }
609 return size;
610 }
611
612 /*
613 * The following 6 functions constitute our interface to the world
614 */
615
616 int dv_assemble_frame(DVMuxContext *c, AVStream* st,
617 const uint8_t* data, int data_size, uint8_t** frame)
618 {
619 uint8_t pcm[8192];
620 int fsize, reqasize;
621
622 *frame = &c->frame_buf[0];
623 if (c->has_audio && c->has_video) { /* must be a stale frame */
624 dv_format_frame(c, *frame);
625 c->frames++;
626 if (c->has_audio > 0)
627 c->has_audio = 0;
628 c->has_video = 0;
629 }
630
631 if (st->codec->codec_type == CODEC_TYPE_VIDEO) {
632 /* FIXME: we have to have more sensible approach than this one */
633 if (c->has_video)
634 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient audio data or severe sync problem.\n", c->frames);
635
636 dv_inject_video(c, data, *frame);
637 c->has_video = 1;
638 data_size = 0;
639 if (c->has_audio < 0)
640 goto out;
641 }
642
643 reqasize = 4 * dv_audio_frame_size(c->sys, c->frames);
644 fsize = fifo_size(&c->audio_data, c->audio_data.rptr);
645 if (st->codec->codec_type == CODEC_TYPE_AUDIO || (c->has_video && fsize >= reqasize)) {
646 if (fsize + data_size >= reqasize && !c->has_audio) {
647 if (fsize >= reqasize) {
648 fifo_read(&c->audio_data, &pcm[0], reqasize, &c->audio_data.rptr);
649 } else {
650 fifo_read(&c->audio_data, &pcm[0], fsize, &c->audio_data.rptr);
651 memcpy(&pcm[fsize], &data[0], reqasize - fsize);
652 data += reqasize - fsize;
653 data_size -= reqasize - fsize;
654 }
655 dv_inject_audio(c, &pcm[0], *frame);
656 c->has_audio = 1;
657 }
658
659 /* FIXME: we have to have more sensible approach than this one */
660 if (fifo_size(&c->audio_data, c->audio_data.rptr) + data_size >= 100*AVCODEC_MAX_AUDIO_FRAME_SIZE)
661 av_log(st->codec, AV_LOG_ERROR, "Can't process DV frame #%d. Insufficient video data or severe sync problem.\n", c->frames);
662 fifo_write(&c->audio_data, (uint8_t *)data, data_size, &c->audio_data.wptr);
663 }
664
665 out:
666 return (c->has_audio && c->has_video) ? c->sys->frame_size : 0;
667 }
668
669 DVMuxContext* dv_init_mux(AVFormatContext* s)
670 {
671 DVMuxContext *c;
672 AVStream *vst = NULL;
673 AVStream *ast = NULL;
674 int i;
675
676 if (s->nb_streams > 2)
677 return NULL;
678
679 c = av_mallocz(sizeof(DVMuxContext));
680 if (!c)
681 return NULL;
682
683 /* We have to sort out where audio and where video stream is */
684 for (i=0; i<s->nb_streams; i++) {
685 switch (s->streams[i]->codec->codec_type) {
686 case CODEC_TYPE_VIDEO:
687 vst = s->streams[i];
688 break;
689 case CODEC_TYPE_AUDIO:
690 ast = s->streams[i];
691 break;
692 default:
693 goto bail_out;
694 }
695 }
696
697 /* Some checks -- DV format is very picky about its incoming streams */
698 if (!vst || vst->codec->codec_id != CODEC_ID_DVVIDEO)
699 goto bail_out;
700 if (ast && (ast->codec->codec_id != CODEC_ID_PCM_S16LE ||
701 ast->codec->sample_rate != 48000 ||
702 ast->codec->channels != 2))
703 goto bail_out;
704 c->sys = dv_codec_profile(vst->codec);
705 if (!c->sys)
706 goto bail_out;
707
708 /* Ok, everything seems to be in working order */
709 c->frames = 0;
710 c->has_audio = ast ? 0 : -1;
711 c->has_video = 0;
712 c->start_time = (time_t)s->timestamp;
713 c->aspect = 0; /* 4:3 is the default */
714 if ((int)(av_q2d(vst->codec->sample_aspect_ratio) * vst->codec->width / vst->codec->height * 10) == 17) /* 16:9 */
715 c->aspect = 0x07;
716
717 if (ast && fifo_init(&c->audio_data, 100*AVCODEC_MAX_AUDIO_FRAME_SIZE) < 0)
718 goto bail_out;
719
720 dv_format_frame(c, &c->frame_buf[0]);
721
722 return c;
723
724 bail_out:
725 av_free(c);
726 return NULL;
727 }
728
729 void dv_delete_mux(DVMuxContext *c)
730 {
731 fifo_free(&c->audio_data);
732 }
733
734 DVDemuxContext* dv_init_demux(AVFormatContext *s)
735 {
736 DVDemuxContext *c;
737
738 c = av_mallocz(sizeof(DVDemuxContext));
739 if (!c)
740 return NULL;
741
742 c->vst = av_new_stream(s, 0);
743 if (!c->vst) {
744 av_free(c);
745 return NULL;
746 }
747
748 c->fctx = s;
749 c->ast[0] = c->ast[1] = NULL;
750 c->ach = 0;
751 c->frames = 0;
752 c->abytes = 0;
753
754 c->vst->codec->codec_type = CODEC_TYPE_VIDEO;
755 c->vst->codec->codec_id = CODEC_ID_DVVIDEO;
756 c->vst->codec->bit_rate = 25000000;
757 c->vst->start_time = 0;
758
759 return c;
760 }
761
762 int dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
763 {
764 int size = -1;
765 int i;
766
767 for (i=0; i<c->ach; i++) {
768 if (c->ast[i] && c->audio_pkt[i].size) {
769 *pkt = c->audio_pkt[i];
770 c->audio_pkt[i].size = 0;
771 size = pkt->size;
772 break;
773 }
774 }
775
776 return size;
777 }
778
779 int dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
780 uint8_t* buf, int buf_size)
781 {
782 int size, i;
783 const DVprofile* sys = dv_frame_profile(buf);
784
785 if (buf_size < 4 || buf_size < sys->frame_size)
786 return -1; /* Broken frame, or not enough data */
787
788 /* Queueing audio packet */
789 /* FIXME: in case of no audio/bad audio we have to do something */
790 size = dv_extract_audio_info(c, buf);
791 for (i=0; i<c->ach; i++) {
792 c->audio_pkt[i].size = size;
793 c->audio_pkt[i].pts = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate;
794 }
795 dv_extract_audio(buf, c->audio_buf[0], c->audio_buf[1]);
796 c->abytes += size;
797
798 /* Now it's time to return video packet */
799 size = dv_extract_video_info(c, buf);
800 av_init_packet(pkt);
801 pkt->data = buf;
802 pkt->size = size;
803 pkt->flags |= PKT_FLAG_KEY;
804 pkt->stream_index = c->vst->id;
805 pkt->pts = c->frames;
806
807 c->frames++;
808
809 return size;
810 }
811
812 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
813 int64_t timestamp, int flags)
814 {
815 // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
816 const DVprofile* sys = dv_codec_profile(c->vst->codec);
817 int64_t offset;
818 int64_t size = url_fsize(&s->pb);
819 int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
820
821 offset = sys->frame_size * timestamp;
822
823 if (offset > max_offset) offset = max_offset;
824 else if (offset < 0) offset = 0;
825
826 return offset;
827 }
828
829 void dv_flush_audio_packets(DVDemuxContext *c)
830 {
831 c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
832 }
833
834 /************************************************************
835 * Implementation of the easiest DV storage of all -- raw DV.
836 ************************************************************/
837
838 typedef struct RawDVContext {
839 uint8_t buf[144000];
840 DVDemuxContext* dv_demux;
841 } RawDVContext;
842
843 static int dv_read_header(AVFormatContext *s,
844 AVFormatParameters *ap)
845 {
846 RawDVContext *c = s->priv_data;
847 const DVprofile* sys;
848 c->dv_demux = dv_init_demux(s);
849 if (!c->dv_demux)
850 return -1;
851
852 if (get_buffer(&s->pb, c->buf, 4) <= 0 || url_fseek(&s->pb, -4, SEEK_CUR) < 0)
853 return AVERROR_IO;
854
855 sys = dv_frame_profile(c->buf);
856 s->bit_rate = av_rescale(sys->frame_size * 8, sys->frame_rate, sys->frame_rate_base);
857
858 return 0;
859 }
860
861
862 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
863 {
864 int size;
865 RawDVContext *c = s->priv_data;
866
867 size = dv_get_packet(c->dv_demux, pkt);
868
869 if (size < 0) {
870 if (get_buffer(&s->pb, c->buf, 4) <= 0)
871 return AVERROR_IO;
872
873 size = dv_frame_profile(c->buf)->frame_size;
874 if (get_buffer(&s->pb, c->buf + 4, size - 4) <= 0)
875 return AVERROR_IO;
876
877 size = dv_produce_packet(c->dv_demux, pkt, c->buf, size);
878 }
879
880 return size;
881 }
882
883 static int dv_read_seek(AVFormatContext *s, int stream_index,
884 int64_t timestamp, int flags)
885 {
886 RawDVContext *r = s->priv_data;
887 DVDemuxContext *c = r->dv_demux;
888 int64_t offset= dv_frame_offset(s, c, timestamp, flags);
889 const DVprofile* sys = dv_codec_profile(c->vst->codec);
890
891 c->frames= offset / sys->frame_size;
892 if (c->ach)
893 c->abytes= av_rescale(c->frames,
894 c->ast[0]->codec->bit_rate * (int64_t)sys->frame_rate_base,
895 8*sys->frame_rate);
896
897 dv_flush_audio_packets(c);
898 return url_fseek(&s->pb, offset, SEEK_SET);
899 }
900
901 static int dv_read_close(AVFormatContext *s)
902 {
903 RawDVContext *c = s->priv_data;
904 av_free(c->dv_demux);
905 return 0;
906 }
907
908 static int dv_write_header(AVFormatContext *s)
909 {
910 s->priv_data = dv_init_mux(s);
911 if (!s->priv_data) {
912 av_log(s, AV_LOG_ERROR, "Can't initialize DV format!\n"
913 "Make sure that you supply exactly two streams:\n"
914 " video: 25fps or 29.97fps, audio: 2ch/48Khz/PCM\n");
915 return -1;
916 }
917 return 0;
918 }
919
920 static int dv_write_packet(struct AVFormatContext *s, AVPacket *pkt)
921 {
922 uint8_t* frame;
923 int fsize;
924
925 fsize = dv_assemble_frame((DVMuxContext *)s->priv_data, s->streams[pkt->stream_index],
926 pkt->data, pkt->size, &frame);
927 if (fsize > 0) {
928 put_buffer(&s->pb, frame, fsize);
929 put_flush_packet(&s->pb);
930 }
931 return 0;
932 }
933
934 /*
935 * We might end up with some extra A/V data without matching counterpart.
936 * E.g. video data without enough audio to write the complete frame.
937 * Currently we simply drop the last frame. I don't know whether this
938 * is the best strategy of all
939 */
940 static int dv_write_trailer(struct AVFormatContext *s)
941 {
942 dv_delete_mux((DVMuxContext *)s->priv_data);
943 return 0;
944 }
945
946 static AVInputFormat dv_iformat = {
947 "dv",
948 "DV video format",
949 sizeof(RawDVContext),
950 NULL,
951 dv_read_header,
952 dv_read_packet,
953 dv_read_close,
954 dv_read_seek,
955 .extensions = "dv,dif",
956 };
957
958 static AVOutputFormat dv_oformat = {
959 "dv",
960 "DV video format",
961 NULL,
962 "dv",
963 sizeof(DVMuxContext),
964 CODEC_ID_PCM_S16LE,
965 CODEC_ID_DVVIDEO,
966 dv_write_header,
967 dv_write_packet,
968 dv_write_trailer,
969 };
970
971 int ff_dv_init(void)
972 {
973 av_register_input_format(&dv_iformat);
974 av_register_output_format(&dv_oformat);
975 return 0;
976 }