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 Libav.
  16  *
  17  * Libav 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  * Libav 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 Libav; 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 "internal.h"
  34 #include "libavcodec/dv_profile.h"
  35 #include "libavcodec/dvdata.h"
  36 #include "libavutil/channel_layout.h"
  37 #include "libavutil/intreadwrite.h"
  38 #include "libavutil/mathematics.h"
  39 #include "dv.h"
  40
  41 struct DVDemuxContext {
  42     const DVprofile*  sys;    /* Current DV profile. E.g.: 525/60, 625/50 */
  43     AVFormatContext*  fctx;
  44     AVStream*         vst;
  45     AVStream*         ast[4];
  46     AVPacket          audio_pkt[4];
  47     uint8_t           audio_buf[4][8192];
  48     int               ach;
  49     int               frames;
  50     uint64_t          abytes;
  51 };
  52
  53 static inline uint16_t dv_audio_12to16(uint16_t sample)
  54 {
  55     uint16_t shift, result;
  56
  57     sample = (sample < 0x800) ? sample : sample | 0xf000;
  58     shift  = (sample & 0xf00) >> 8;
  59
  60     if (shift < 0x2 || shift > 0xd) {
  61         result = sample;
  62     } else if (shift < 0x8) {
  63         shift--;
  64         result = (sample - (256 * shift)) << shift;
  65     } else {
  66         shift = 0xe - shift;
  67         result = ((sample + ((256 * shift) + 1)) << shift) - 1;
  68     }
  69
  70     return result;
  71 }
  72
  73 /*
  74  * This is the dumbest implementation of all -- it simply looks at
  75  * a fixed offset and if pack isn't there -- fails. We might want
  76  * to have a fallback mechanism for complete search of missing packs.
  77  */
  78 static const uint8_t* dv_extract_pack(uint8_t* frame, enum dv_pack_type t)
  79 {
  80     int offs;
  81
  82     switch (t) {
  83     case dv_audio_source:
  84         offs = (80*6 + 80*16*3 + 3);
  85         break;
  86     case dv_audio_control:
  87         offs = (80*6 + 80*16*4 + 3);
  88         break;
  89     case dv_video_control:
  90         offs = (80*5 + 48 + 5);
  91         break;
  92     default:
  93         return NULL;
  94     }
  95
  96     return frame[offs] == t ? &frame[offs] : NULL;
  97 }
  98
  99 static const int dv_audio_frequency[3] = {
 100     48000, 44100, 32000,
 101 };
 102
 103 /*
 104  * There's a couple of assumptions being made here:
 105  * 1. By default we silence erroneous (0x8000/16bit 0x800/12bit) audio samples.
 106  *    We can pass them upwards when libavcodec will be ready to deal with them.
 107  * 2. We don't do software emphasis.
 108  * 3. Audio is always returned as 16bit linear samples: 12bit nonlinear samples
 109  *    are converted into 16bit linear ones.
 110  */
 111 static int dv_extract_audio(uint8_t* frame, uint8_t* ppcm[4],
 112                             const DVprofile *sys)
 113 {
 114     int size, chan, i, j, d, of, smpls, freq, quant, half_ch;
 115     uint16_t lc, rc;
 116     const uint8_t* as_pack;
 117     uint8_t *pcm, ipcm;
 118
 119     as_pack = dv_extract_pack(frame, dv_audio_source);
 120     if (!as_pack)    /* No audio ? */
 121         return 0;
 122
 123     smpls =  as_pack[1] & 0x3f;       /* samples in this frame - min. samples */
 124     freq  = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
 125     quant =  as_pack[4] & 0x07;       /* 0 - 16bit linear, 1 - 12bit nonlinear */
 126
 127     if (quant > 1)
 128         return -1; /* unsupported quantization */
 129
 130     if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency))
 131         return AVERROR_INVALIDDATA;
 132
 133     size = (sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */
 134     half_ch = sys->difseg_size / 2;
 135
 136     /* We work with 720p frames split in half, thus even frames have
 137      * channels 0,1 and odd 2,3. */
 138     ipcm = (sys->height == 720 && !(frame[1] & 0x0C)) ? 2 : 0;
 139
 140     /* for each DIF channel */
 141     for (chan = 0; chan < sys->n_difchan; chan++) {
 142         /* next stereo channel (50Mbps and 100Mbps only) */
 143         pcm = ppcm[ipcm++];
 144         if (!pcm)
 145             break;
 146
 147         /* for each DIF segment */
 148         for (i = 0; i < sys->difseg_size; i++) {
 149             frame += 6 * 80; /* skip DIF segment header */
 150             if (quant == 1 && i == half_ch) {
 151                 /* next stereo channel (12bit mode only) */
 152                 pcm = ppcm[ipcm++];
 153                 if (!pcm)
 154                     break;
 155             }
 156
 157             /* for each AV sequence */
 158             for (j = 0; j < 9; j++) {
 159                 for (d = 8; d < 80; d += 2) {
 160                     if (quant == 0) {  /* 16bit quantization */
 161                         of = sys->audio_shuffle[i][j] + (d - 8) / 2 * sys->audio_stride;
 162                         if (of*2 >= size)
 163                             continue;
 164
 165                         pcm[of*2]   = frame[d+1]; // FIXME: maybe we have to admit
 166                         pcm[of*2+1] = frame[d];   //        that DV is a big-endian PCM
 167                         if (pcm[of*2+1] == 0x80 && pcm[of*2] == 0x00)
 168                             pcm[of*2+1] = 0;
 169                     } else {           /* 12bit quantization */
 170                         lc = ((uint16_t)frame[d]   << 4) |
 171                              ((uint16_t)frame[d+2] >> 4);
 172                         rc = ((uint16_t)frame[d+1] << 4) |
 173                              ((uint16_t)frame[d+2] & 0x0f);
 174                         lc = (lc == 0x800 ? 0 : dv_audio_12to16(lc));
 175                         rc = (rc == 0x800 ? 0 : dv_audio_12to16(rc));
 176
 177                         of = sys->audio_shuffle[i%half_ch][j] + (d - 8) / 3 * sys->audio_stride;
 178                         if (of*2 >= size)
 179                             continue;
 180
 181                         pcm[of*2]   = lc & 0xff; // FIXME: maybe we have to admit
 182                         pcm[of*2+1] = lc >> 8;   //        that DV is a big-endian PCM
 183                         of = sys->audio_shuffle[i%half_ch+half_ch][j] +
 184                             (d - 8) / 3 * sys->audio_stride;
 185                         pcm[of*2]   = rc & 0xff; // FIXME: maybe we have to admit
 186                         pcm[of*2+1] = rc >> 8;   //        that DV is a big-endian PCM
 187                         ++d;
 188                     }
 189                 }
 190
 191                 frame += 16 * 80; /* 15 Video DIFs + 1 Audio DIF */
 192             }
 193         }
 194     }
 195
 196     return size;
 197 }
 198
 199 static int dv_extract_audio_info(DVDemuxContext* c, uint8_t* frame)
 200 {
 201     const uint8_t* as_pack;
 202     int freq, stype, smpls, quant, i, ach;
 203
 204     as_pack = dv_extract_pack(frame, dv_audio_source);
 205     if (!as_pack || !c->sys) {    /* No audio ? */
 206         c->ach = 0;
 207         return 0;
 208     }
 209
 210     smpls =  as_pack[1] & 0x3f;       /* samples in this frame - min. samples */
 211     freq  = (as_pack[4] >> 3) & 0x07; /* 0 - 48kHz, 1 - 44,1kHz, 2 - 32kHz */
 212     stype = (as_pack[3] & 0x1f);      /* 0 - 2CH, 2 - 4CH, 3 - 8CH */
 213     quant =  as_pack[4] & 0x07;       /* 0 - 16bit linear, 1 - 12bit nonlinear */
 214
 215     if (freq >= FF_ARRAY_ELEMS(dv_audio_frequency)) {
 216         av_log(c->fctx, AV_LOG_ERROR,
 217                "Unrecognized audio sample rate index (%d)\n", freq);
 218         return 0;
 219     }
 220
 221     if (stype > 3) {
 222         av_log(c->fctx, AV_LOG_ERROR, "stype %d is invalid\n", stype);
 223         c->ach = 0;
 224         return 0;
 225     }
 226
 227     /* note: ach counts PAIRS of channels (i.e. stereo channels) */
 228     ach = ((int[4]){  1,  0,  2,  4})[stype];
 229     if (ach == 1 && quant && freq == 2)
 230         ach = 2;
 231
 232     /* Dynamic handling of the audio streams in DV */
 233     for (i = 0; i < ach; i++) {
 234         if (!c->ast[i]) {
 235             c->ast[i] = avformat_new_stream(c->fctx, NULL);
 236             if (!c->ast[i])
 237                 break;
 238             avpriv_set_pts_info(c->ast[i], 64, 1, 30000);
 239             c->ast[i]->codec->codec_type = AVMEDIA_TYPE_AUDIO;
 240             c->ast[i]->codec->codec_id   = AV_CODEC_ID_PCM_S16LE;
 241
 242             av_init_packet(&c->audio_pkt[i]);
 243             c->audio_pkt[i].size         = 0;
 244             c->audio_pkt[i].data         = c->audio_buf[i];
 245             c->audio_pkt[i].stream_index = c->ast[i]->index;
 246             c->audio_pkt[i].flags       |= AV_PKT_FLAG_KEY;
 247         }
 248         c->ast[i]->codec->sample_rate = dv_audio_frequency[freq];
 249         c->ast[i]->codec->channels    = 2;
 250         c->ast[i]->codec->channel_layout = AV_CH_LAYOUT_STEREO;
 251         c->ast[i]->codec->bit_rate    = 2 * dv_audio_frequency[freq] * 16;
 252         c->ast[i]->start_time         = 0;
 253     }
 254     c->ach = i;
 255
 256     return (c->sys->audio_min_samples[freq] + smpls) * 4; /* 2ch, 2bytes */;
 257 }
 258
 259 static int dv_extract_video_info(DVDemuxContext *c, uint8_t* frame)
 260 {
 261     const uint8_t* vsc_pack;
 262     AVCodecContext* avctx;
 263     int apt, is16_9;
 264     int size = 0;
 265
 266     if (c->sys) {
 267         avctx = c->vst->codec;
 268
 269         avpriv_set_pts_info(c->vst, 64, c->sys->time_base.num,
 270                             c->sys->time_base.den);
 271         avctx->time_base= c->sys->time_base;
 272         if (!avctx->width){
 273             avctx->width = c->sys->width;
 274             avctx->height = c->sys->height;
 275         }
 276         avctx->pix_fmt = c->sys->pix_fmt;
 277
 278         /* finding out SAR is a little bit messy */
 279         vsc_pack = dv_extract_pack(frame, dv_video_control);
 280         apt      = frame[4] & 0x07;
 281         is16_9   = (vsc_pack && ((vsc_pack[2] & 0x07) == 0x02 ||
 282                                  (!apt && (vsc_pack[2] & 0x07) == 0x07)));
 283         c->vst->sample_aspect_ratio = c->sys->sar[is16_9];
 284         avctx->bit_rate = av_rescale_q(c->sys->frame_size, (AVRational){8,1},
 285                                        c->sys->time_base);
 286         size = c->sys->frame_size;
 287     }
 288     return size;
 289 }
 290
 291 /*
 292  * The following 3 functions constitute our interface to the world
 293  */
 294
 295 DVDemuxContext* avpriv_dv_init_demux(AVFormatContext *s)
 296 {
 297     DVDemuxContext *c;
 298
 299     c = av_mallocz(sizeof(DVDemuxContext));
 300     if (!c)
 301         return NULL;
 302
 303     c->vst = avformat_new_stream(s, NULL);
 304     if (!c->vst) {
 305         av_free(c);
 306         return NULL;
 307     }
 308
 309     c->fctx                   = s;
 310     c->vst->codec->codec_type = AVMEDIA_TYPE_VIDEO;
 311     c->vst->codec->codec_id   = AV_CODEC_ID_DVVIDEO;
 312     c->vst->codec->bit_rate   = 25000000;
 313     c->vst->start_time        = 0;
 314
 315     return c;
 316 }
 317
 318 int avpriv_dv_get_packet(DVDemuxContext *c, AVPacket *pkt)
 319 {
 320     int size = -1;
 321     int i;
 322
 323     for (i = 0; i < c->ach; i++) {
 324         if (c->ast[i] && c->audio_pkt[i].size) {
 325             *pkt = c->audio_pkt[i];
 326             c->audio_pkt[i].size = 0;
 327             size = pkt->size;
 328             break;
 329         }
 330     }
 331
 332     return size;
 333 }
 334
 335 int avpriv_dv_produce_packet(DVDemuxContext *c, AVPacket *pkt,
 336                       uint8_t* buf, int buf_size)
 337 {
 338     int size, i;
 339     uint8_t *ppcm[4] = {0};
 340
 341     if (buf_size < DV_PROFILE_BYTES ||
 342         !(c->sys = avpriv_dv_frame_profile(c->sys, buf, buf_size)) ||
 343         buf_size < c->sys->frame_size) {
 344         return -1;   /* Broken frame, or not enough data */
 345     }
 346
 347     /* Queueing audio packet */
 348     /* FIXME: in case of no audio/bad audio we have to do something */
 349     size = dv_extract_audio_info(c, buf);
 350     for (i = 0; i < c->ach; i++) {
 351         c->audio_pkt[i].size = size;
 352         c->audio_pkt[i].pts  = c->abytes * 30000 * 8 / c->ast[i]->codec->bit_rate;
 353         ppcm[i] = c->audio_buf[i];
 354     }
 355     if (c->ach)
 356         dv_extract_audio(buf, ppcm, c->sys);
 357
 358     /* We work with 720p frames split in half, thus even frames have
 359      * channels 0,1 and odd 2,3. */
 360     if (c->sys->height == 720) {
 361         if (buf[1] & 0x0C) {
 362             c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
 363         } else {
 364             c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
 365             c->abytes += size;
 366         }
 367     } else {
 368         c->abytes += size;
 369     }
 370
 371     /* Now it's time to return video packet */
 372     size = dv_extract_video_info(c, buf);
 373     av_init_packet(pkt);
 374     pkt->data         = buf;
 375     pkt->size         = size;
 376     pkt->flags       |= AV_PKT_FLAG_KEY;
 377     pkt->stream_index = c->vst->index;
 378     pkt->pts          = c->frames;
 379
 380     c->frames++;
 381
 382     return size;
 383 }
 384
 385 static int64_t dv_frame_offset(AVFormatContext *s, DVDemuxContext *c,
 386                               int64_t timestamp, int flags)
 387 {
 388     // FIXME: sys may be wrong if last dv_read_packet() failed (buffer is junk)
 389     const DVprofile* sys = avpriv_dv_codec_profile(c->vst->codec);
 390     int64_t offset;
 391     int64_t size = avio_size(s->pb) - s->data_offset;
 392     int64_t max_offset = ((size-1) / sys->frame_size) * sys->frame_size;
 393
 394     offset = sys->frame_size * timestamp;
 395
 396     if (size >= 0 && offset > max_offset) offset = max_offset;
 397     else if (offset < 0) offset = 0;
 398
 399     return offset + s->data_offset;
 400 }
 401
 402 void ff_dv_offset_reset(DVDemuxContext *c, int64_t frame_offset)
 403 {
 404     c->frames= frame_offset;
 405     if (c->ach)
 406         c->abytes= av_rescale_q(c->frames, c->sys->time_base,
 407                                 (AVRational){8, c->ast[0]->codec->bit_rate});
 408     c->audio_pkt[0].size = c->audio_pkt[1].size = 0;
 409     c->audio_pkt[2].size = c->audio_pkt[3].size = 0;
 410 }
 411
 412 /************************************************************
 413  * Implementation of the easiest DV storage of all -- raw DV.
 414  ************************************************************/
 415
 416 typedef struct RawDVContext {
 417     DVDemuxContext* dv_demux;
 418     uint8_t         buf[DV_MAX_FRAME_SIZE];
 419 } RawDVContext;
 420
 421 static int dv_read_header(AVFormatContext *s)
 422 {
 423     unsigned state, marker_pos = 0;
 424     RawDVContext *c = s->priv_data;
 425
 426     c->dv_demux = avpriv_dv_init_demux(s);
 427     if (!c->dv_demux)
 428         return -1;
 429
 430     state = avio_rb32(s->pb);
 431     while ((state & 0xffffff7f) != 0x1f07003f) {
 432         if (s->pb->eof_reached) {
 433             av_log(s, AV_LOG_ERROR, "Cannot find DV header.\n");
 434             return -1;
 435         }
 436         if (state == 0x003f0700 || state == 0xff3f0700)
 437             marker_pos = avio_tell(s->pb);
 438         if (state == 0xff3f0701 && avio_tell(s->pb) - marker_pos == 80) {
 439             avio_seek(s->pb, -163, SEEK_CUR);
 440             state = avio_rb32(s->pb);
 441             break;
 442         }
 443         state = (state << 8) | avio_r8(s->pb);
 444     }
 445     AV_WB32(c->buf, state);
 446
 447     if (avio_read(s->pb, c->buf + 4, DV_PROFILE_BYTES - 4) <= 0 ||
 448         avio_seek(s->pb, -DV_PROFILE_BYTES, SEEK_CUR) < 0)
 449         return AVERROR(EIO);
 450
 451     c->dv_demux->sys = avpriv_dv_frame_profile(c->dv_demux->sys, c->buf, DV_PROFILE_BYTES);
 452     if (!c->dv_demux->sys) {
 453         av_log(s, AV_LOG_ERROR, "Can't determine profile of DV input stream.\n");
 454         return -1;
 455     }
 456
 457     s->bit_rate = av_rescale_q(c->dv_demux->sys->frame_size, (AVRational){8,1},
 458                                c->dv_demux->sys->time_base);
 459
 460     return 0;
 461 }
 462
 463
 464 static int dv_read_packet(AVFormatContext *s, AVPacket *pkt)
 465 {
 466     int size;
 467     RawDVContext *c = s->priv_data;
 468
 469     size = avpriv_dv_get_packet(c->dv_demux, pkt);
 470
 471     if (size < 0) {
 472         if (!c->dv_demux->sys)
 473             return AVERROR(EIO);
 474         size = c->dv_demux->sys->frame_size;
 475         if (avio_read(s->pb, c->buf, size) <= 0)
 476             return AVERROR(EIO);
 477
 478         size = avpriv_dv_produce_packet(c->dv_demux, pkt, c->buf, size);
 479     }
 480
 481     return size;
 482 }
 483
 484 static int dv_read_seek(AVFormatContext *s, int stream_index,
 485                        int64_t timestamp, int flags)
 486 {
 487     RawDVContext *r   = s->priv_data;
 488     DVDemuxContext *c = r->dv_demux;
 489     int64_t offset    = dv_frame_offset(s, c, timestamp, flags);
 490
 491     if (avio_seek(s->pb, offset, SEEK_SET) < 0)
 492         return -1;
 493
 494     ff_dv_offset_reset(c, offset / c->sys->frame_size);
 495     return 0;
 496 }
 497
 498 static int dv_read_close(AVFormatContext *s)
 499 {
 500     RawDVContext *c = s->priv_data;
 501     av_free(c->dv_demux);
 502     return 0;
 503 }
 504
 505 static int dv_probe(AVProbeData *p)
 506 {
 507     unsigned state, marker_pos = 0;
 508     int i;
 509     int matches = 0;
 510     int secondary_matches = 0;
 511
 512     if (p->buf_size < 5)
 513         return 0;
 514
 515     state = AV_RB32(p->buf);
 516     for (i = 4; i < p->buf_size; i++) {
 517         if ((state & 0xffffff7f) == 0x1f07003f)
 518             matches++;
 519         // any section header, also with seq/chan num != 0,
 520         // should appear around every 12000 bytes, at least 10 per frame
 521         if ((state & 0xff07ff7f) == 0x1f07003f)
 522             secondary_matches++;
 523         if (state == 0x003f0700 || state == 0xff3f0700)
 524             marker_pos = i;
 525         if (state == 0xff3f0701 && i - marker_pos == 80)
 526             matches++;
 527         state = (state << 8) | p->buf[i];
 528     }
 529
 530     if (matches && p->buf_size / matches < 1024*1024) {
 531         if (matches > 4 || (secondary_matches >= 10 && p->buf_size / secondary_matches < 24000))
 532             return AVPROBE_SCORE_MAX*3/4; // not max to avoid dv in mov to match
 533         return AVPROBE_SCORE_MAX/4;
 534     }
 535     return 0;
 536 }
 537
 538 #if CONFIG_DV_DEMUXER
 539 AVInputFormat ff_dv_demuxer = {
 540     .name           = "dv",
 541     .long_name      = NULL_IF_CONFIG_SMALL("DV (Digital Video)"),
 542     .priv_data_size = sizeof(RawDVContext),
 543     .read_probe     = dv_probe,
 544     .read_header    = dv_read_header,
 545     .read_packet    = dv_read_packet,
 546     .read_close     = dv_read_close,
 547     .read_seek      = dv_read_seek,
 548     .extensions     = "dv,dif",
 549 };
 550 #endif