libavcodec/flacdec.c

   1 /*
   2  * FLAC (Free Lossless Audio Codec) decoder
   3  * Copyright (c) 2003 Alex Beregszaszi
   4  *
   5  * This file is part of Libav.
   6  *
   7  * Libav is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * Libav is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with Libav; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 /**
  23  * @file
  24  * FLAC (Free Lossless Audio Codec) decoder
  25  * @author Alex Beregszaszi
  26  * @see http://flac.sourceforge.net/
  27  *
  28  * This decoder can be used in 1 of 2 ways: Either raw FLAC data can be fed
  29  * through, starting from the initial 'fLaC' signature; or by passing the
  30  * 34-byte streaminfo structure through avctx->extradata[_size] followed
  31  * by data starting with the 0xFFF8 marker.
  32  */
  33
  34 #include <limits.h>
  35
  36 #include "libavutil/channel_layout.h"
  37 #include "libavutil/crc.h"
  38 #include "avcodec.h"
  39 #include "internal.h"
  40 #include "get_bits.h"
  41 #include "bytestream.h"
  42 #include "golomb.h"
  43 #include "flac.h"
  44 #include "flacdata.h"
  45 #include "flacdsp.h"
  46
  47 #undef NDEBUG
  48 #include <assert.h>
  49
  50 typedef struct FLACContext {
  51     FLACSTREAMINFO
  52
  53     AVCodecContext *avctx;                  ///< parent AVCodecContext
  54     GetBitContext gb;                       ///< GetBitContext initialized to start at the current frame
  55
  56     int blocksize;                          ///< number of samples in the current frame
  57     int sample_shift;                       ///< shift required to make output samples 16-bit or 32-bit
  58     int ch_mode;                            ///< channel decorrelation type in the current frame
  59     int got_streaminfo;                     ///< indicates if the STREAMINFO has been read
  60
  61     int32_t *decoded[FLAC_MAX_CHANNELS];    ///< decoded samples
  62     uint8_t *decoded_buffer;
  63     unsigned int decoded_buffer_size;
  64
  65     FLACDSPContext dsp;
  66 } FLACContext;
  67
  68 static int allocate_buffers(FLACContext *s);
  69
  70 static void flac_set_bps(FLACContext *s)
  71 {
  72     enum AVSampleFormat req = s->avctx->request_sample_fmt;
  73     int need32 = s->bps > 16;
  74     int want32 = av_get_bytes_per_sample(req) > 2;
  75     int planar = av_sample_fmt_is_planar(req);
  76
  77     if (need32 || want32) {
  78         if (planar)
  79             s->avctx->sample_fmt = AV_SAMPLE_FMT_S32P;
  80         else
  81             s->avctx->sample_fmt = AV_SAMPLE_FMT_S32;
  82         s->sample_shift = 32 - s->bps;
  83     } else {
  84         if (planar)
  85             s->avctx->sample_fmt = AV_SAMPLE_FMT_S16P;
  86         else
  87             s->avctx->sample_fmt = AV_SAMPLE_FMT_S16;
  88         s->sample_shift = 16 - s->bps;
  89     }
  90 }
  91
  92 static av_cold int flac_decode_init(AVCodecContext *avctx)
  93 {
  94     enum FLACExtradataFormat format;
  95     uint8_t *streaminfo;
  96     int ret;
  97     FLACContext *s = avctx->priv_data;
  98     s->avctx = avctx;
  99
 100     /* for now, the raw FLAC header is allowed to be passed to the decoder as
 101        frame data instead of extradata. */
 102     if (!avctx->extradata)
 103         return 0;
 104
 105     if (!avpriv_flac_is_extradata_valid(avctx, &format, &streaminfo))
 106         return -1;
 107
 108     /* initialize based on the demuxer-supplied streamdata header */
 109     avpriv_flac_parse_streaminfo(avctx, (FLACStreaminfo *)s, streaminfo);
 110     ret = allocate_buffers(s);
 111     if (ret < 0)
 112         return ret;
 113     flac_set_bps(s);
 114     ff_flacdsp_init(&s->dsp, avctx->sample_fmt, s->bps);
 115     s->got_streaminfo = 1;
 116
 117     return 0;
 118 }
 119
 120 static void dump_headers(AVCodecContext *avctx, FLACStreaminfo *s)
 121 {
 122     av_log(avctx, AV_LOG_DEBUG, "  Max Blocksize: %d\n", s->max_blocksize);
 123     av_log(avctx, AV_LOG_DEBUG, "  Max Framesize: %d\n", s->max_framesize);
 124     av_log(avctx, AV_LOG_DEBUG, "  Samplerate: %d\n", s->samplerate);
 125     av_log(avctx, AV_LOG_DEBUG, "  Channels: %d\n", s->channels);
 126     av_log(avctx, AV_LOG_DEBUG, "  Bits: %d\n", s->bps);
 127 }
 128
 129 static int allocate_buffers(FLACContext *s)
 130 {
 131     int buf_size;
 132
 133     assert(s->max_blocksize);
 134
 135     buf_size = av_samples_get_buffer_size(NULL, s->channels, s->max_blocksize,
 136                                           AV_SAMPLE_FMT_S32P, 0);
 137     if (buf_size < 0)
 138         return buf_size;
 139
 140     av_fast_malloc(&s->decoded_buffer, &s->decoded_buffer_size, buf_size);
 141     if (!s->decoded_buffer)
 142         return AVERROR(ENOMEM);
 143
 144     return av_samples_fill_arrays((uint8_t **)s->decoded, NULL,
 145                                   s->decoded_buffer, s->channels,
 146                                   s->max_blocksize, AV_SAMPLE_FMT_S32P, 0);
 147 }
 148
 149 /**
 150  * Parse the STREAMINFO from an inline header.
 151  * @param s the flac decoding context
 152  * @param buf input buffer, starting with the "fLaC" marker
 153  * @param buf_size buffer size
 154  * @return non-zero if metadata is invalid
 155  */
 156 static int parse_streaminfo(FLACContext *s, const uint8_t *buf, int buf_size)
 157 {
 158     int metadata_type, metadata_size, ret;
 159
 160     if (buf_size < FLAC_STREAMINFO_SIZE+8) {
 161         /* need more data */
 162         return 0;
 163     }
 164     avpriv_flac_parse_block_header(&buf[4], NULL, &metadata_type, &metadata_size);
 165     if (metadata_type != FLAC_METADATA_TYPE_STREAMINFO ||
 166         metadata_size != FLAC_STREAMINFO_SIZE) {
 167         return AVERROR_INVALIDDATA;
 168     }
 169     avpriv_flac_parse_streaminfo(s->avctx, (FLACStreaminfo *)s, &buf[8]);
 170     ret = allocate_buffers(s);
 171     if (ret < 0)
 172         return ret;
 173     flac_set_bps(s);
 174     ff_flacdsp_init(&s->dsp, s->avctx->sample_fmt, s->bps);
 175     s->got_streaminfo = 1;
 176
 177     return 0;
 178 }
 179
 180 /**
 181  * Determine the size of an inline header.
 182  * @param buf input buffer, starting with the "fLaC" marker
 183  * @param buf_size buffer size
 184  * @return number of bytes in the header, or 0 if more data is needed
 185  */
 186 static int get_metadata_size(const uint8_t *buf, int buf_size)
 187 {
 188     int metadata_last, metadata_size;
 189     const uint8_t *buf_end = buf + buf_size;
 190
 191     buf += 4;
 192     do {
 193         if (buf_end - buf < 4)
 194             return 0;
 195         avpriv_flac_parse_block_header(buf, &metadata_last, NULL, &metadata_size);
 196         buf += 4;
 197         if (buf_end - buf < metadata_size) {
 198             /* need more data in order to read the complete header */
 199             return 0;
 200         }
 201         buf += metadata_size;
 202     } while (!metadata_last);
 203
 204     return buf_size - (buf_end - buf);
 205 }
 206
 207 static int decode_residuals(FLACContext *s, int32_t *decoded, int pred_order)
 208 {
 209     int i, tmp, partition, method_type, rice_order;
 210     int rice_bits, rice_esc;
 211     int samples;
 212
 213     method_type = get_bits(&s->gb, 2);
 214     if (method_type > 1) {
 215         av_log(s->avctx, AV_LOG_ERROR, "illegal residual coding method %d\n",
 216                method_type);
 217         return -1;
 218     }
 219
 220     rice_order = get_bits(&s->gb, 4);
 221
 222     samples= s->blocksize >> rice_order;
 223     if (pred_order > samples) {
 224         av_log(s->avctx, AV_LOG_ERROR, "invalid predictor order: %i > %i\n",
 225                pred_order, samples);
 226         return -1;
 227     }
 228
 229     rice_bits = 4 + method_type;
 230     rice_esc  = (1 << rice_bits) - 1;
 231
 232     decoded += pred_order;
 233     i= pred_order;
 234     for (partition = 0; partition < (1 << rice_order); partition++) {
 235         tmp = get_bits(&s->gb, rice_bits);
 236         if (tmp == rice_esc) {
 237             tmp = get_bits(&s->gb, 5);
 238             for (; i < samples; i++)
 239                 *decoded++ = get_sbits_long(&s->gb, tmp);
 240         } else {
 241             for (; i < samples; i++) {
 242                 *decoded++ = get_sr_golomb_flac(&s->gb, tmp, INT_MAX, 0);
 243             }
 244         }
 245         i= 0;
 246     }
 247
 248     return 0;
 249 }
 250
 251 static int decode_subframe_fixed(FLACContext *s, int32_t *decoded,
 252                                  int pred_order, int bps)
 253 {
 254     const int blocksize = s->blocksize;
 255     int a, b, c, d, i;
 256
 257     /* warm up samples */
 258     for (i = 0; i < pred_order; i++) {
 259         decoded[i] = get_sbits_long(&s->gb, bps);
 260     }
 261
 262     if (decode_residuals(s, decoded, pred_order) < 0)
 263         return -1;
 264
 265     if (pred_order > 0)
 266         a = decoded[pred_order-1];
 267     if (pred_order > 1)
 268         b = a - decoded[pred_order-2];
 269     if (pred_order > 2)
 270         c = b - decoded[pred_order-2] + decoded[pred_order-3];
 271     if (pred_order > 3)
 272         d = c - decoded[pred_order-2] + 2*decoded[pred_order-3] - decoded[pred_order-4];
 273
 274     switch (pred_order) {
 275     case 0:
 276         break;
 277     case 1:
 278         for (i = pred_order; i < blocksize; i++)
 279             decoded[i] = a += decoded[i];
 280         break;
 281     case 2:
 282         for (i = pred_order; i < blocksize; i++)
 283             decoded[i] = a += b += decoded[i];
 284         break;
 285     case 3:
 286         for (i = pred_order; i < blocksize; i++)
 287             decoded[i] = a += b += c += decoded[i];
 288         break;
 289     case 4:
 290         for (i = pred_order; i < blocksize; i++)
 291             decoded[i] = a += b += c += d += decoded[i];
 292         break;
 293     default:
 294         av_log(s->avctx, AV_LOG_ERROR, "illegal pred order %d\n", pred_order);
 295         return -1;
 296     }
 297
 298     return 0;
 299 }
 300
 301 static int decode_subframe_lpc(FLACContext *s, int32_t *decoded, int pred_order,
 302                                int bps)
 303 {
 304     int i;
 305     int coeff_prec, qlevel;
 306     int coeffs[32];
 307
 308     /* warm up samples */
 309     for (i = 0; i < pred_order; i++) {
 310         decoded[i] = get_sbits_long(&s->gb, bps);
 311     }
 312
 313     coeff_prec = get_bits(&s->gb, 4) + 1;
 314     if (coeff_prec == 16) {
 315         av_log(s->avctx, AV_LOG_ERROR, "invalid coeff precision\n");
 316         return -1;
 317     }
 318     qlevel = get_sbits(&s->gb, 5);
 319     if (qlevel < 0) {
 320         av_log(s->avctx, AV_LOG_ERROR, "qlevel %d not supported, maybe buggy stream\n",
 321                qlevel);
 322         return -1;
 323     }
 324
 325     for (i = 0; i < pred_order; i++) {
 326         coeffs[pred_order - i - 1] = get_sbits(&s->gb, coeff_prec);
 327     }
 328
 329     if (decode_residuals(s, decoded, pred_order) < 0)
 330         return -1;
 331
 332     s->dsp.lpc(decoded, coeffs, pred_order, qlevel, s->blocksize);
 333
 334     return 0;
 335 }
 336
 337 static inline int decode_subframe(FLACContext *s, int channel)
 338 {
 339     int32_t *decoded = s->decoded[channel];
 340     int type, wasted = 0;
 341     int bps = s->bps;
 342     int i, tmp;
 343
 344     if (channel == 0) {
 345         if (s->ch_mode == FLAC_CHMODE_RIGHT_SIDE)
 346             bps++;
 347     } else {
 348         if (s->ch_mode == FLAC_CHMODE_LEFT_SIDE || s->ch_mode == FLAC_CHMODE_MID_SIDE)
 349             bps++;
 350     }
 351
 352     if (get_bits1(&s->gb)) {
 353         av_log(s->avctx, AV_LOG_ERROR, "invalid subframe padding\n");
 354         return -1;
 355     }
 356     type = get_bits(&s->gb, 6);
 357
 358     if (get_bits1(&s->gb)) {
 359         int left = get_bits_left(&s->gb);
 360         wasted = 1;
 361         if ( left < 0 ||
 362             (left < bps && !show_bits_long(&s->gb, left)) ||
 363                            !show_bits_long(&s->gb, bps)) {
 364             av_log(s->avctx, AV_LOG_ERROR,
 365                    "Invalid number of wasted bits > available bits (%d) - left=%d\n",
 366                    bps, left);
 367             return AVERROR_INVALIDDATA;
 368         }
 369         while (!get_bits1(&s->gb))
 370             wasted++;
 371         bps -= wasted;
 372     }
 373     if (bps > 32) {
 374         av_log_missing_feature(s->avctx, "Decorrelated bit depth > 32", 0);
 375         return AVERROR_PATCHWELCOME;
 376     }
 377
 378 //FIXME use av_log2 for types
 379     if (type == 0) {
 380         tmp = get_sbits_long(&s->gb, bps);
 381         for (i = 0; i < s->blocksize; i++)
 382             decoded[i] = tmp;
 383     } else if (type == 1) {
 384         for (i = 0; i < s->blocksize; i++)
 385             decoded[i] = get_sbits_long(&s->gb, bps);
 386     } else if ((type >= 8) && (type <= 12)) {
 387         if (decode_subframe_fixed(s, decoded, type & ~0x8, bps) < 0)
 388             return -1;
 389     } else if (type >= 32) {
 390         if (decode_subframe_lpc(s, decoded, (type & ~0x20)+1, bps) < 0)
 391             return -1;
 392     } else {
 393         av_log(s->avctx, AV_LOG_ERROR, "invalid coding type\n");
 394         return -1;
 395     }
 396
 397     if (wasted) {
 398         int i;
 399         for (i = 0; i < s->blocksize; i++)
 400             decoded[i] <<= wasted;
 401     }
 402
 403     return 0;
 404 }
 405
 406 static int decode_frame(FLACContext *s)
 407 {
 408     int i, ret;
 409     GetBitContext *gb = &s->gb;
 410     FLACFrameInfo fi;
 411
 412     if (ff_flac_decode_frame_header(s->avctx, gb, &fi, 0)) {
 413         av_log(s->avctx, AV_LOG_ERROR, "invalid frame header\n");
 414         return -1;
 415     }
 416
 417     if (s->channels && fi.channels != s->channels && s->got_streaminfo) {
 418         s->channels = s->avctx->channels = fi.channels;
 419         ff_flac_set_channel_layout(s->avctx);
 420         ret = allocate_buffers(s);
 421         if (ret < 0)
 422             return ret;
 423     }
 424     s->channels = s->avctx->channels = fi.channels;
 425     if (!s->avctx->channel_layout)
 426         ff_flac_set_channel_layout(s->avctx);
 427     s->ch_mode = fi.ch_mode;
 428
 429     if (!s->bps && !fi.bps) {
 430         av_log(s->avctx, AV_LOG_ERROR, "bps not found in STREAMINFO or frame header\n");
 431         return -1;
 432     }
 433     if (!fi.bps) {
 434         fi.bps = s->bps;
 435     } else if (s->bps && fi.bps != s->bps) {
 436         av_log(s->avctx, AV_LOG_ERROR, "switching bps mid-stream is not "
 437                                        "supported\n");
 438         return -1;
 439     }
 440
 441     if (!s->bps) {
 442         s->bps = s->avctx->bits_per_raw_sample = fi.bps;
 443         flac_set_bps(s);
 444     }
 445
 446     if (!s->max_blocksize)
 447         s->max_blocksize = FLAC_MAX_BLOCKSIZE;
 448     if (fi.blocksize > s->max_blocksize) {
 449         av_log(s->avctx, AV_LOG_ERROR, "blocksize %d > %d\n", fi.blocksize,
 450                s->max_blocksize);
 451         return -1;
 452     }
 453     s->blocksize = fi.blocksize;
 454
 455     if (!s->samplerate && !fi.samplerate) {
 456         av_log(s->avctx, AV_LOG_ERROR, "sample rate not found in STREAMINFO"
 457                                         " or frame header\n");
 458         return -1;
 459     }
 460     if (fi.samplerate == 0)
 461         fi.samplerate = s->samplerate;
 462     s->samplerate = s->avctx->sample_rate = fi.samplerate;
 463
 464     if (!s->got_streaminfo) {
 465         ret = allocate_buffers(s);
 466         if (ret < 0)
 467             return ret;
 468         ff_flacdsp_init(&s->dsp, s->avctx->sample_fmt, s->bps);
 469         s->got_streaminfo = 1;
 470         dump_headers(s->avctx, (FLACStreaminfo *)s);
 471     }
 472
 473 //    dump_headers(s->avctx, (FLACStreaminfo *)s);
 474
 475     /* subframes */
 476     for (i = 0; i < s->channels; i++) {
 477         if (decode_subframe(s, i) < 0)
 478             return -1;
 479     }
 480
 481     align_get_bits(gb);
 482
 483     /* frame footer */
 484     skip_bits(gb, 16); /* data crc */
 485
 486     return 0;
 487 }
 488
 489 static int flac_decode_frame(AVCodecContext *avctx, void *data,
 490                              int *got_frame_ptr, AVPacket *avpkt)
 491 {
 492     AVFrame *frame     = data;
 493     const uint8_t *buf = avpkt->data;
 494     int buf_size = avpkt->size;
 495     FLACContext *s = avctx->priv_data;
 496     int bytes_read = 0;
 497     int ret;
 498
 499     *got_frame_ptr = 0;
 500
 501     if (s->max_framesize == 0) {
 502         s->max_framesize =
 503             ff_flac_get_max_frame_size(s->max_blocksize ? s->max_blocksize : FLAC_MAX_BLOCKSIZE,
 504                                        FLAC_MAX_CHANNELS, 32);
 505     }
 506
 507     /* check that there is at least the smallest decodable amount of data.
 508        this amount corresponds to the smallest valid FLAC frame possible.
 509        FF F8 69 02 00 00 9A 00 00 34 46 */
 510     if (buf_size < FLAC_MIN_FRAME_SIZE)
 511         return buf_size;
 512
 513     /* check for inline header */
 514     if (AV_RB32(buf) == MKBETAG('f','L','a','C')) {
 515         if (!s->got_streaminfo && parse_streaminfo(s, buf, buf_size)) {
 516             av_log(s->avctx, AV_LOG_ERROR, "invalid header\n");
 517             return -1;
 518         }
 519         return get_metadata_size(buf, buf_size);
 520     }
 521
 522     /* decode frame */
 523     init_get_bits(&s->gb, buf, buf_size*8);
 524     if (decode_frame(s) < 0) {
 525         av_log(s->avctx, AV_LOG_ERROR, "decode_frame() failed\n");
 526         return -1;
 527     }
 528     bytes_read = (get_bits_count(&s->gb)+7)/8;
 529
 530     /* get output buffer */
 531     frame->nb_samples = s->blocksize;
 532     if ((ret = ff_get_buffer(avctx, frame, 0)) < 0) {
 533         av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
 534         return ret;
 535     }
 536
 537     s->dsp.decorrelate[s->ch_mode](frame->data, s->decoded, s->channels,
 538                                    s->blocksize, s->sample_shift);
 539
 540     if (bytes_read > buf_size) {
 541         av_log(s->avctx, AV_LOG_ERROR, "overread: %d\n", bytes_read - buf_size);
 542         return -1;
 543     }
 544     if (bytes_read < buf_size) {
 545         av_log(s->avctx, AV_LOG_DEBUG, "underread: %d orig size: %d\n",
 546                buf_size - bytes_read, buf_size);
 547     }
 548
 549     *got_frame_ptr = 1;
 550
 551     return bytes_read;
 552 }
 553
 554 static av_cold int flac_decode_close(AVCodecContext *avctx)
 555 {
 556     FLACContext *s = avctx->priv_data;
 557
 558     av_freep(&s->decoded_buffer);
 559
 560     return 0;
 561 }
 562
 563 AVCodec ff_flac_decoder = {
 564     .name           = "flac",
 565     .type           = AVMEDIA_TYPE_AUDIO,
 566     .id             = AV_CODEC_ID_FLAC,
 567     .priv_data_size = sizeof(FLACContext),
 568     .init           = flac_decode_init,
 569     .close          = flac_decode_close,
 570     .decode         = flac_decode_frame,
 571     .capabilities   = CODEC_CAP_DR1,
 572     .long_name      = NULL_IF_CONFIG_SMALL("FLAC (Free Lossless Audio Codec)"),
 573     .sample_fmts    = (const enum AVSampleFormat[]) { AV_SAMPLE_FMT_S16,
 574                                                       AV_SAMPLE_FMT_S16P,
 575                                                       AV_SAMPLE_FMT_S32,
 576                                                       AV_SAMPLE_FMT_S32P,
 577                                                       -1 },
 578 };