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) 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 #include "avformat.h"
  32 #include "dvdata.h"
  33 #include "dv.h"
  34
  35 struct DVDemuxContext {
  36     const DVprofile*  sys;    /* Current DV profile. E.g.: 525/60, 625/50 */
  37     AVFormatContext* fctx;
  38     AVStream*        vst;
  39     AVStream*        ast[2];
  40     AVPacket         audio_pkt[2];
  41     uint8_t          audio_buf[2][8192];
  42     int              ach;
  43     int              frames;
  44     uint64_t         abytes;
  45 };
  46
  47 static inline uint16_t dv_audio_12to16(uint16_t sample)
  48 {
  49     uint16_t shift, result;
  50
  51     sample = (sample < 0x800) ? sample : sample | 0xf000;
  52     shift = (sample & 0xf00) >> 8;
  53
  54     if (shift < 0x2 || shift > 0xd) {
  55         result = sample;
  56     } else if (shift < 0x8) {
  57         shift--;
  58         result = (sample - (256 * shift)) << shift;
  59     } else {
  60         shift = 0xe - shift;
  61         result = ((sample + ((256 * shift) + 1)) << shift) - 1;
  62     }
  63
  64     return result;
  65 }
  66
  67 /*
  68  * This is the dumbest implementation of all -- it simply looks at
  69  * a fixed offset and if pack isn't there -- fails. We might want
  70  * to have a fallback mechanism for complete search of missing packs.
  71  */
  72 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
  73 {
  74     int offs;
  75
  76     switch (t) {
  77     case dv_audio_source:
  78           offs = (80*6 + 80*16*3 + 3);
  79           break;
  80     case dv_audio_control:
  81           offs = (80*6 + 80*16*4 + 3);
  82           break;
  83     case dv_video_control:
  84           offs = (80*5 + 48 + 5);
  85           break;
  86     default:
  87           return NULL;
  88     }
  89
  90     return (frame[offs] == t ? &frame[offs] : NULL);
  91 }
  92
  93 /*
  94  * There's a couple of assumptions being made here:
  95  * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
  96  *    We can pass them upwards when ffmpeg will be ready to deal with them.
  97  * 2. We don't do software emphasis.
  98  * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
  99  *    are converted into 16bit linear ones.
 100  */
 101 static int dv_extract_audio(uint8_t* frame, uint8_t* pcm, uint8_t* pcm2,
 102                             const DVprofile *sys)
 103 {
 104     int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
 105     uint16_t lc, rc;
 106     const uint8_t* as_pack;
 107
 108     as_pack = dv_extract_pack(frame, dv_audio_source);
 109     if (!as_pack)    /* No audio ? */
 110         return 0;
 111
 112     smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
 113     freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
 114     quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
 115
 116     if (quant > 1)
 117         return -1; /* Unsupported quantization */
 118
 119     size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
 120     half_ch = sys->difseg_size/2;
 121
 122     /* for each DIF channel */
 123     for (chan = 0; chan < sys->n_difchan; chan++) {
 124         /* for each DIF segment */
 125         for (i = 0; i < sys->difseg_size; i++) {
 126             frame += 6 * 80; /* skip DIF segment header */
 127             if (quant == 1 && i == half_ch) {
 128                 /* next stereo channel (12bit mode only) */
 129                 if (!pcm2)
 130                     break;
 131                 else
 132                     pcm = pcm2;
 133             }
 134
 135             /* for each AV sequence */
 136             for (j = 0; j < 9; j++) {
 137                 for (d = 8; d < 80; d += 2) {
 138                     if (quant == 0) {  /* 16bit quantization */
 139                         of = sys->audio_shuffle[i][j] + (d - 8)/2 * sys->audio_stride;
 140                         if (of*2 >= size)
 141                             continue;
 142
 143                         pcm[of*2] = frame[d+1]; // FIXME: may be we have to admit
 144                         pcm[of*2+1] = frame[d]; //        that DV is a big endian PCM
 145                         if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
 146                             pcm[of*2+1] = 0;
 147                     } else {           /* 12bit quantization */
 148                         lc = ((uint16_t)frame[d] << 4) |
 149                              ((uint16_t)frame[d+2] >> 4);
 150                         rc = ((uint16_t)frame[d+1] << 4) |
 151                              ((uint16_t)frame[d+2] & 0x0f);
 152                         lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
 153                         rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
 154
 155                         of = sys->audio_shuffle[i%half_ch][j] + (d - 8)/3 * sys->audio_stride;
 156                         if (of*2 >= size)
 157                             continue;
 158
 159                         pcm[of*2] = lc & 0xff; // FIXME: may be we have to admit
 160                         pcm[of*2+1] = lc >> 8; //        that DV is a big endian PCM
 161                         of = sys->audio_shuffle[i%half_ch+half_ch][j] +
 162                             (d - 8)/3 * sys->audio_stride;
 163                         pcm[of*2] = rc & 0xff; // FIXME: may be we have to admit
 164                         pcm[of*2+1] = rc >> 8; //        that DV is a big endian PCM
 165                         ++d;
 166                     }
 167                 }
 168
 169                 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
 170             }
 171         }
 172
 173         /* next stereo channel (50Mbps only) */
 174         if(!pcm2)
 175             break;
 176         pcm = pcm2;
 177     }
 178
 179     return size;
 180 }
 181
 182 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
 183 {
 184     const uint8_t* as_pack;
 185     int freq, stype, smpls, quant, i, ach;
 186
 187     as_pack = dv_extract_pack(frame, dv_audio_source);
 188     if (!as_pack || !c->sys) {    /* No audio ? */
 189         c->ach = 0;
 190         return 0;
 191     }
 192
 193     smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
 194     freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48KHz, 1 - 44,1kHz, 2 - 32 kHz */
 195     stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH */
 196     quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
 197
 198     /* note: ach counts PAIRS of channels (i.e. stereo channels) */
 199     ach = (stype == 2 || (quant && (freq == 2))) ? 2 : 1;
 200
 201     /* Dynamic handling of the audio streams in DV */
 202     for (i=0; i<ach; i++) {
 203        if (!c->ast[i]) {
 204            c->ast[i] = av_new_stream(c->fctx, 0);
 205            if (!c->ast[i])
 206                break;
 207            av_set_pts_info(c->ast[i], 64, 1, 30000);
 208            c->ast[i]->codec->codec_type = CODEC_TYPE_AUDIO;
 209            c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE;
 210
 211            av_init_packet(&c->audio_pkt[i]);
 212            c->audio_pkt[i].size     = 0;
 213            c->audio_pkt[i].data     = c->audio_buf[i];
 214            c->audio_pkt[i].stream_index = c->ast[i]->index;
 215            c->audio_pkt[i].flags |= PKT_FLAG_KEY;
 216        }
 217        c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
 218        c->ast[i]->codec->channels = 2;
 219        c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
 220        c->ast[i]->start_time = 0;
 221     }
 222     c->ach = i;
 223
 224     return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
 225 }
 226
 227 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
 228 {
 229     const uint8_t* vsc_pack;
 230     AVCodecContext* avctx;
 231     int apt, is16_9;
 232     int size = 0;
 233
 234     if (c->sys) {
 235         avctx = c->vst->codec;
 236
 237         av_set_pts_info(c->vst, 64, c->sys->frame_rate_base, c->sys->frame_rate);
 238         avctx->time_base= (AVRational){c->sys->frame_rate_base, c->sys->frame_rate};
 239         if(!avctx->width){
 240             avctx->width = c->sys->width;
 241             avctx->height = c->sys->height;
 242         }
 243         avctx->pix_fmt = c->sys->pix_fmt;
 244
 245         /* finding out SAR is a little bit messy */
 246         vsc_pack = dv_extract_pack(frame, dv_video_control);
 247         apt = frame[4] & 0x07;
 248         is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
 249                                (!apt && (vsc_pack[2] & 0x07) == 0x07)));
 250         avctx->sample_aspect_ratio = c->sys->sar[is16_9];
 251         avctx->bit_rate = av_rescale(c->sys->frame_size * 8,
 252                                      c->sys->frame_rate,
 253                                      c->sys->frame_rate_base);
 254         size = c->sys->frame_size;
 255     }
 256     return size;
 257 }
 258
 259 /*
 260  * The following 3 functions constitute our interface to the world
 261  */
 262
 263 DVDemuxContext* dv_init_demux(AVFormatContext *s)
 264 {
 265     DVDemuxContext *c;
 266
 267     c = av_mallocz(sizeof(DVDemuxContext));
 268     if (!c)
 269         return NULL;
 270
 271     c->vst = av_new_stream(s, 0);
 272     if (!c->vst) {
 273         av_free(c);
 274         return NULL;
 275     }
 276
 277     c->sys = NULL;
 278     c->fctx = s;
 279     c->ast[0] = c->ast[1] = NULL;
 280     c->ach = 0;
 281     c->frames = 0;
 282     c->abytes = 0;
 283
 284     c->vst->codec->codec_type = CODEC_TYPE_VIDEO;
 285     c->vst->codec->codec_id = CODEC_ID_DVVIDEO;
 286     c->vst->codec->bit_rate = 25000000;
 287     c->vst->start_time = 0;
 288
 289     return c;
 290 }
 291
 292 int dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
 293 {
 294     int size = -1;
 295     int i;
 296
 297     for (i=0; i<c->ach; i++) {
 298        if (c->ast[i] && c->audio_pkt[i].size) {
 299            *pkt = c->audio_pkt[i];
 300            c->audio_pkt[i].size = 0;
 301            size = pkt->size;
 302            break;
 303        }
 304     }
 305
 306     return size;
 307 }
 308
 309 int dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
 310                       uint8_t* buf, int buf_size)
 311 {
 312     int size, i;
 313
 314     if (buf_size < DV_PROFILE_BYTES ||
 315         !(c->sys = dv_frame_profile(buf)) ||
 316         buf_size < c->sys->frame_size) {
 317           return -1;   /* Broken frame, or not enough data */
 318     }
 319
 320     /* Queueing audio packet */
 321     /* FIXME: in case of no audio/bad audio we have to do something */
 322     size = dv_extract_audio_info(c, buf);
 323     for (i=0; i<c->ach; i++) {
 324        c->audio_pkt[i].size = size;
 325        c->audio_pkt[i].pts  = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate;
 326     }
 327     dv_extract_audio(buf, c->audio_buf[0], c->audio_buf[1], c->sys);
 328     c->abytes += size;
 329
 330     /* Now it's time to return video packet */
 331     size = dv_extract_video_info(c, buf);
 332     av_init_packet(pkt);
 333     pkt->data     = buf;
 334     pkt->size     = size;
 335     pkt->flags   |= PKT_FLAG_KEY;
 336     pkt->stream_index = c->vst->id;
 337     pkt->pts      = c->frames;
 338
 339     c->frames++;
 340
 341     return size;
 342 }
 343
 344 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
 345                               int64_t timestamp, int flags)
 346 {
 347     // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
 348     const DVprofile* sys = dv_codec_profile(c->vst->codec);
 349     int64_t offset;
 350     int64_t size = url_fsize(s->pb);
 351     int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
 352
 353     offset = sys->frame_size * timestamp;
 354
 355     if (offset > max_offset) offset = max_offset;
 356     else if (offset < 0) offset = 0;
 357
 358     return offset;
 359 }
 360
 361 void dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
 362 {
 363     c->frames= frame_offset;
 364     if (c->ach)
 365         c->abytes= av_rescale(c->frames,
 366                           c->ast[0]->codec->bit_rate * (int64_t)c->sys->frame_rate_base,
 367                           8*c->sys->frame_rate);
 368     c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
 369 }
 370
 371 /************************************************************
 372  * Implementation of the easiest DV storage of all -- raw DV.
 373  ************************************************************/
 374
 375 typedef struct RawDVContext {
 376     DVDemuxContext* dv_demux;
 377     uint8_t         buf[DV_MAX_FRAME_SIZE];
 378 } RawDVContext;
 379
 380 static int dv_read_header(AVFormatContext *s,
 381                           AVFormatParameters *ap)
 382 {
 383     RawDVContext *c = s->priv_data;
 384
 385     c->dv_demux = dv_init_demux(s);
 386     if (!c->dv_demux)
 387         return -1;
 388
 389     if (get_buffer(s->pb, c->buf, DV_PROFILE_BYTES) <= 0 ||
 390         url_fseek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
 391         return AVERROR(EIO);
 392
 393     c->dv_demux->sys = dv_frame_profile(c->buf);
 394     s->bit_rate = av_rescale(c->dv_demux->sys->frame_size * 8,
 395                              c->dv_demux->sys->frame_rate,
 396                              c->dv_demux->sys->frame_rate_base);
 397
 398     return 0;
 399 }
 400
 401
 402 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
 403 {
 404     int size;
 405     RawDVContext *c = s->priv_data;
 406
 407     size = dv_get_packet(c->dv_demux, pkt);
 408
 409     if (size < 0) {
 410         size = c->dv_demux->sys->frame_size;
 411         if (get_buffer(s->pb, c->buf, size) <= 0)
 412             return AVERROR(EIO);
 413
 414         size = dv_produce_packet(c->dv_demux, pkt, c->buf, size);
 415     }
 416
 417     return size;
 418 }
 419
 420 static int dv_read_seek(AVFormatContext *s, int stream_index,
 421                        int64_t timestamp, int flags)
 422 {
 423     RawDVContext *r = s->priv_data;
 424     DVDemuxContext *c = r->dv_demux;
 425     int64_t offset= dv_frame_offset(s, c, timestamp, flags);
 426
 427     dv_offset_reset(c, offset / c->sys->frame_size);
 428
 429     offset = url_fseek(s->pb, offset, SEEK_SET);
 430     return (offset < 0)?offset:0;
 431 }
 432
 433 static int dv_read_close(AVFormatContext *s)
 434 {
 435     RawDVContext *c = s->priv_data;
 436     av_free(c->dv_demux);
 437     return 0;
 438 }
 439
 440 #ifdef CONFIG_DV_DEMUXER
 441 AVInputFormat dv_demuxer = {
 442     "dv",
 443     "DV video format",
 444     sizeof(RawDVContext),
 445     NULL,
 446     dv_read_header,
 447     dv_read_packet,
 448     dv_read_close,
 449     dv_read_seek,
 450     .extensions = "dv,dif",
 451 };
 452 #endif