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 channels 0,1 and odd 2,3 */
 125     ipcm = (sys->height == 720 && !(frame[1]&0x0C))?2:0;
 126     pcm = ppcm[ipcm++];
 127
 128     /* for each DIF channel */
 129     for (chan = 0; chan < sys->n_difchan; chan++) {
 130         /* for each DIF segment */
 131         for (i = 0; i < sys->difseg_size; i++) {
 132             frame += 6 * 80; /* skip DIF segment header */
 133             if (quant == 1 && i == half_ch) {
 134                 /* next stereo channel (12bit mode only) */
 135                 pcm = ppcm[ipcm++];
 136                 if (!pcm)
 137                     break;
 138             }
 139
 140             /* for each AV sequence */
 141             for (j = 0; j < 9; j++) {
 142                 for (d = 8; d < 80; d += 2) {
 143                     if (quant == 0) {  /* 16bit quantization */
 144                         of = sys->audio_shuffle[i][j] + (d - 8)/2 * sys->audio_stride;
 145                         if (of*2 >= size)
 146                             continue;
 147
 148                         pcm[of*2]   = frame[d+1]; // FIXME: maybe we have to admit
 149                         pcm[of*2+1] = frame[d];   //        that DV is a big-endian PCM
 150                         if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
 151                             pcm[of*2+1] = 0;
 152                     } else {           /* 12bit quantization */
 153                         lc = ((uint16_t)frame[d] << 4) |
 154                              ((uint16_t)frame[d+2] >> 4);
 155                         rc = ((uint16_t)frame[d+1] << 4) |
 156                              ((uint16_t)frame[d+2] & 0x0f);
 157                         lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
 158                         rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
 159
 160                         of = sys->audio_shuffle[i%half_ch][j] + (d - 8)/3 * sys->audio_stride;
 161                         if (of*2 >= size)
 162                             continue;
 163
 164                         pcm[of*2]   = lc & 0xff; // FIXME: maybe we have to admit
 165                         pcm[of*2+1] = lc >> 8;   //        that DV is a big-endian PCM
 166                         of = sys->audio_shuffle[i%half_ch+half_ch][j] +
 167                             (d - 8)/3 * sys->audio_stride;
 168                         pcm[of*2]   = rc & 0xff; // FIXME: maybe we have to admit
 169                         pcm[of*2+1] = rc >> 8;   //        that DV is a big-endian PCM
 170                         ++d;
 171                     }
 172                 }
 173
 174                 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
 175             }
 176         }
 177
 178         /* next stereo channel (50Mbps and 100Mbps only) */
 179         pcm = ppcm[ipcm++];
 180         if (!pcm)
 181             break;
 182     }
 183
 184     return size;
 185 }
 186
 187 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
 188 {
 189     const uint8_t* as_pack;
 190     int freq, stype, smpls, quant, i, ach;
 191
 192     as_pack = dv_extract_pack(frame, dv_audio_source);
 193     if (!as_pack || !c->sys) {    /* No audio ? */
 194         c->ach = 0;
 195         return 0;
 196     }
 197
 198     smpls = as_pack[1] & 0x3f; /* samples in this frame - min. samples */
 199     freq = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
 200     stype = (as_pack[3] & 0x1f); /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
 201     quant = as_pack[4] & 0x07; /* 0 - 16bit linear, 1 - 12bit nonlinear */
 202
 203     /* note: ach counts PAIRS of channels (i.e. stereo channels) */
 204     ach = ((int[4]){  1,  0,  2,  4})[stype];
 205     if (ach == 1 && quant && freq == 2)
 206         ach = 2;
 207
 208     /* Dynamic handling of the audio streams in DV */
 209     for (i=0; i<ach; i++) {
 210        if (!c->ast[i]) {
 211            c->ast[i] = av_new_stream(c->fctx, 0);
 212            if (!c->ast[i])
 213                break;
 214            av_set_pts_info(c->ast[i], 64, 1, 30000);
 215            c->ast[i]->codec->codec_type = CODEC_TYPE_AUDIO;
 216            c->ast[i]->codec->codec_id = CODEC_ID_PCM_S16LE;
 217
 218            av_init_packet(&c->audio_pkt[i]);
 219            c->audio_pkt[i].size     = 0;
 220            c->audio_pkt[i].data     = c->audio_buf[i];
 221            c->audio_pkt[i].stream_index = c->ast[i]->index;
 222            c->audio_pkt[i].flags |= PKT_FLAG_KEY;
 223        }
 224        c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
 225        c->ast[i]->codec->channels = 2;
 226        c->ast[i]->codec->bit_rate = 2 * dv_audio_frequency[freq] * 16;
 227        c->ast[i]->start_time = 0;
 228     }
 229     c->ach = i;
 230
 231     return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
 232 }
 233
 234 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
 235 {
 236     const uint8_t* vsc_pack;
 237     AVCodecContext* avctx;
 238     int apt, is16_9;
 239     int size = 0;
 240
 241     if (c->sys) {
 242         avctx = c->vst->codec;
 243
 244         av_set_pts_info(c->vst, 64, c->sys->time_base.num, c->sys->time_base.den);
 245         avctx->time_base= c->sys->time_base;
 246         if(!avctx->width){
 247             avctx->width = c->sys->width;
 248             avctx->height = c->sys->height;
 249         }
 250         avctx->pix_fmt = c->sys->pix_fmt;
 251
 252         /* finding out SAR is a little bit messy */
 253         vsc_pack = dv_extract_pack(frame, dv_video_control);
 254         apt = frame[4] & 0x07;
 255         is16_9 = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
 256                                (!apt && (vsc_pack[2] & 0x07) == 0x07)));
 257         c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
 258         avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1},
 259                                        c->sys->time_base);
 260         size = c->sys->frame_size;
 261     }
 262     return size;
 263 }
 264
 265 /*
 266  * The following 3 functions constitute our interface to the world
 267  */
 268
 269 DVDemuxContext* dv_init_demux(AVFormatContext *s)
 270 {
 271     DVDemuxContext *c;
 272
 273     c = av_mallocz(sizeof(DVDemuxContext));
 274     if (!c)
 275         return NULL;
 276
 277     c->vst = av_new_stream(s, 0);
 278     if (!c->vst) {
 279         av_free(c);
 280         return NULL;
 281     }
 282
 283     c->sys = NULL;
 284     c->fctx = s;
 285     memset(c->ast, 0, sizeof(c->ast));
 286     c->ach = 0;
 287     c->frames = 0;
 288     c->abytes = 0;
 289
 290     c->vst->codec->codec_type = CODEC_TYPE_VIDEO;
 291     c->vst->codec->codec_id = CODEC_ID_DVVIDEO;
 292     c->vst->codec->bit_rate = 25000000;
 293     c->vst->start_time = 0;
 294
 295     return c;
 296 }
 297
 298 int dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
 299 {
 300     int size = -1;
 301     int i;
 302
 303     for (i=0; i<c->ach; i++) {
 304        if (c->ast[i] && c->audio_pkt[i].size) {
 305            *pkt = c->audio_pkt[i];
 306            c->audio_pkt[i].size = 0;
 307            size = pkt->size;
 308            break;
 309        }
 310     }
 311
 312     return size;
 313 }
 314
 315 int dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
 316                       uint8_t* buf, int buf_size)
 317 {
 318     int size, i;
 319     uint8_t *ppcm[4] = {0};
 320
 321     if (buf_size < DV_PROFILE_BYTES ||
 322         !(c->sys = dv_frame_profile(buf)) ||
 323         buf_size < c->sys->frame_size) {
 324           return -1;   /* Broken frame, or not enough data */
 325     }
 326
 327     /* Queueing audio packet */
 328     /* FIXME: in case of no audio/bad audio we have to do something */
 329     size = dv_extract_audio_info(c, buf);
 330     for (i=0; i<c->ach; i++) {
 331        c->audio_pkt[i].size = size;
 332        c->audio_pkt[i].pts  = c->abytes * 30000*8 / c->ast[i]->codec->bit_rate;
 333        ppcm[i] = c->audio_buf[i];
 334     }
 335     dv_extract_audio(buf, ppcm, c->sys);
 336     c->abytes += size;
 337
 338     /* We work with 720p frames split in half, thus even frames have channels 0,1 and odd 2,3 */
 339     if (c->sys->height == 720) {
 340         if (buf[1]&0x0C)
 341             c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
 342         else
 343             c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
 344     }
 345
 346     /* Now it's time to return video packet */
 347     size = dv_extract_video_info(c, buf);
 348     av_init_packet(pkt);
 349     pkt->data     = buf;
 350     pkt->size     = size;
 351     pkt->flags   |= PKT_FLAG_KEY;
 352     pkt->stream_index = c->vst->id;
 353     pkt->pts      = c->frames;
 354
 355     c->frames++;
 356
 357     return size;
 358 }
 359
 360 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
 361                               int64_t timestamp, int flags)
 362 {
 363     // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
 364     const DVprofile* sys = dv_codec_profile(c->vst->codec);
 365     int64_t offset;
 366     int64_t size = url_fsize(s->pb);
 367     int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
 368
 369     offset = sys->frame_size * timestamp;
 370
 371     if (offset > max_offset) offset = max_offset;
 372     else if (offset < 0) offset = 0;
 373
 374     return offset;
 375 }
 376
 377 void dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
 378 {
 379     c->frames= frame_offset;
 380     if (c->ach)
 381         c->abytes= av_rescale_q(c->frames, c->sys->time_base,
 382                                 (AVRational){8, c->ast[0]->codec->bit_rate});
 383     c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
 384     c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
 385 }
 386
 387 /************************************************************
 388  * Implementation of the easiest DV storage of all -- raw DV.
 389  ************************************************************/
 390
 391 typedef struct RawDVContext {
 392     DVDemuxContext* dv_demux;
 393     uint8_t         buf[DV_MAX_FRAME_SIZE];
 394 } RawDVContext;
 395
 396 static int dv_read_header(AVFormatContext *s,
 397                           AVFormatParameters *ap)
 398 {
 399     RawDVContext *c = s->priv_data;
 400
 401     c->dv_demux = dv_init_demux(s);
 402     if (!c->dv_demux)
 403         return -1;
 404
 405     if (get_buffer(s->pb, c->buf, DV_PROFILE_BYTES) <= 0 ||
 406         url_fseek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
 407         return AVERROR(EIO);
 408
 409     c->dv_demux->sys = dv_frame_profile(c->buf);
 410     if (!c->dv_demux->sys) {
 411         av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n");
 412         return -1;
 413     }
 414
 415     s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size, (AVRational){8,1},
 416                                c->dv_demux->sys->time_base);
 417
 418     return 0;
 419 }
 420
 421
 422 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
 423 {
 424     int size;
 425     RawDVContext *c = s->priv_data;
 426
 427     size = dv_get_packet(c->dv_demux, pkt);
 428
 429     if (size < 0) {
 430         size = c->dv_demux->sys->frame_size;
 431         if (get_buffer(s->pb, c->buf, size) <= 0)
 432             return AVERROR(EIO);
 433
 434         size = dv_produce_packet(c->dv_demux, pkt, c->buf, size);
 435     }
 436
 437     return size;
 438 }
 439
 440 static int dv_read_seek(AVFormatContext *s, int stream_index,
 441                        int64_t timestamp, int flags)
 442 {
 443     RawDVContext *r = s->priv_data;
 444     DVDemuxContext *c = r->dv_demux;
 445     int64_t offset= dv_frame_offset(s, c, timestamp, flags);
 446
 447     dv_offset_reset(c, offset / c->sys->frame_size);
 448
 449     offset = url_fseek(s->pb, offset, SEEK_SET);
 450     return (offset < 0)?offset:0;
 451 }
 452
 453 static int dv_read_close(AVFormatContext *s)
 454 {
 455     RawDVContext *c = s->priv_data;
 456     av_free(c->dv_demux);
 457     return 0;
 458 }
 459
 460 #ifdef CONFIG_DV_DEMUXER
 461 AVInputFormat dv_demuxer = {
 462     "dv",
 463     NULL_IF_CONFIG_SMALL("DV video format"),
 464     sizeof(RawDVContext),
 465     NULL,
 466     dv_read_header,
 467     dv_read_packet,
 468     dv_read_close,
 469     dv_read_seek,
 470     .extensions = "dv,dif",
 471 };
 472 #endif