importer/wave_imp.c

   1 /*****************************************************************************
   2  * wave_imp.c
   3  *****************************************************************************
   4  * Copyright (C) 2014 L-SMASH project
   5  *
   6  * Authors: Yusuke Nakamura <muken.the.vfrmaniac@gmail.com>
   7  *
   8  * Permission to use, copy, modify, and/or distribute this software for any
   9  * purpose with or without fee is hereby granted, provided that the above
  10  * copyright notice and this permission notice appear in all copies.
  11  *
  12  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  13  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  14  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
  15  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  16  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
  17  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
  18  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  19  *****************************************************************************/
  20
  21 /* This file is available under an ISC license. */
  22
  23 #include "common/internal.h" /* must be placed first */
  24
  25 #include <string.h>
  26
  27 #define LSMASH_IMPORTER_INTERNAL
  28 #include "importer.h"
  29
  30 /*********************************************************************************
  31     Waveform Audio File Format (WAVE) importer
  32
  33     References
  34         Multimedia Programming Interface and Data Specifications 1.0
  35         New Multimedia Data Types and Data Techniques April 15, 1994 Revision: 3.0
  36         Multiple channel audio data and WAVE files March 7, 2007
  37         Microsoft Windows SDK MMReg.h
  38 **********************************************************************************/
  39 #define WAVE_MIN_FILESIZE 45
  40
  41 #define WAVE_FORMAT_TYPE_ID_PCM        0x0001   /* WAVE_FORMAT_PCM */
  42 #define WAVE_FORMAT_TYPE_ID_EXTENSIBLE 0xFFFE   /* WAVE_FORMAT_EXTENSIBLE */
  43
  44 #define PASS_GUID( _0, _1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11, _12, _13, _14, _15 ) \
  45                  { _0, _1, _2, _3, _4, _5, _6, _7, _8, _9, _10, _11, _12, _13, _14, _15 }
  46 #define DEFINE_WAVEFORMAT_EXTENSIBLE_SUBTYPE_GUID( name, value ) \
  47         static const uint8_t name[16] = value
  48
  49 /* KSDATAFORMAT_SUBTYPE_PCM := 00000001-0000-0010-8000-00aa00389b71 */
  50 DEFINE_WAVEFORMAT_EXTENSIBLE_SUBTYPE_GUID
  51 (
  52     WAVEFORMAT_EXTENSIBLE_SUBTYPE_GUID_PCM,
  53     PASS_GUID( 0x01, 0x00, 0x00, 0x00, 0x00, 0x00, 0x10, 0x00,
  54                0x80, 0x00, 0x00, 0xaa, 0x00, 0x38, 0x9b, 0x71 )
  55 );
  56
  57 typedef struct
  58 {
  59     uint16_t wFormatTag;
  60     uint16_t nChannels;
  61     uint32_t nSamplesPerSec;
  62     uint32_t nAvgBytesPerSec;
  63     uint16_t nBlockAlign;
  64     uint16_t wBitsPerSample;
  65     uint16_t cbSize;
  66 } waveformat_extended_t;
  67
  68 typedef struct
  69 {
  70     waveformat_extended_t wfx;
  71     union
  72     {
  73         uint16_t wValidBitsPerSample;
  74         uint16_t wSamplesPerBlock;
  75         uint16_t wReserved;
  76     } Samples;
  77     uint32_t dwChannelMask;
  78     uint8_t guid[16];
  79 } waveformat_extensible_t;
  80
  81 typedef struct
  82 {
  83     uint32_t number_of_samples;
  84     uint32_t au_length;
  85     uint32_t au_number;
  86     waveformat_extensible_t fmt;
  87 } wave_importer_t;
  88
  89 static void remove_wave_importer( wave_importer_t *wave_imp )
  90 {
  91     if( !wave_imp )
  92         return;
  93     lsmash_free( wave_imp );
  94 }
  95
  96 static wave_importer_t *create_wave_importer( importer_t *importer )
  97 {
  98     return (wave_importer_t *)lsmash_malloc_zero( sizeof(wave_importer_t) );
  99 }
 100
 101 static void wave_importer_cleanup( importer_t *importer )
 102 {
 103     debug_if( importer && importer->info )
 104         remove_wave_importer( importer->info );
 105 }
 106
 107 static int wave_importer_get_accessunit( importer_t *importer, uint32_t track_number, lsmash_sample_t **p_sample )
 108 {
 109     if( !importer->info )
 110         return LSMASH_ERR_NAMELESS;
 111     if( track_number != 1 )
 112         return LSMASH_ERR_FUNCTION_PARAM;
 113     lsmash_audio_summary_t *summary = (lsmash_audio_summary_t *)lsmash_get_entry_data( importer->summaries, track_number );
 114     if( !summary )
 115         return LSMASH_ERR_NAMELESS;
 116     wave_importer_t *wave_imp = (wave_importer_t *)importer->info;
 117     importer_status current_status = importer->status;
 118     if( current_status == IMPORTER_ERROR )
 119         return LSMASH_ERR_NAMELESS;
 120     if( current_status == IMPORTER_EOF )
 121         return IMPORTER_EOF;
 122     if( wave_imp->number_of_samples / summary->samples_in_frame > wave_imp->au_number )
 123         wave_imp->au_length = summary->bytes_per_frame;
 124     else
 125     {
 126         wave_imp->au_length = wave_imp->fmt.wfx.nBlockAlign * (wave_imp->number_of_samples % summary->samples_in_frame);
 127         importer->status = IMPORTER_EOF;
 128         if( wave_imp->au_length == 0 )
 129             return IMPORTER_EOF;
 130     }
 131     lsmash_sample_t *sample = lsmash_create_sample( wave_imp->au_length );
 132     if( !sample )
 133         return LSMASH_ERR_MEMORY_ALLOC;
 134     *p_sample = sample;
 135     if( lsmash_bs_get_bytes_ex( importer->bs, wave_imp->au_length, sample->data ) != wave_imp->au_length )
 136     {
 137         importer->status = IMPORTER_ERROR;
 138         return LSMASH_ERR_INVALID_DATA;
 139     }
 140     sample->length        = wave_imp->au_length;
 141     sample->dts           = wave_imp->au_number ++ * summary->samples_in_frame;
 142     sample->cts           = sample->dts;
 143     sample->prop.ra_flags = ISOM_SAMPLE_RANDOM_ACCESS_FLAG_SYNC;
 144     return current_status;
 145 }
 146
 147 static inline int wave_fmt_subtype_cmp( const waveformat_extensible_t *fmt, const uint8_t guid[16] )
 148 {
 149     return memcmp( fmt->guid, guid, 16 );
 150 }
 151
 152 static int wave_parse_fmt_chunk( wave_importer_t *wave_imp, lsmash_bs_t *bs )
 153 {
 154     waveformat_extensible_t *fmt = &wave_imp->fmt;
 155     waveformat_extended_t   *wfx = &fmt->wfx;
 156     wfx->wFormatTag      = lsmash_bs_get_le16( bs );
 157     wfx->nChannels       = lsmash_bs_get_le16( bs );
 158     wfx->nSamplesPerSec  = lsmash_bs_get_le32( bs );
 159     wfx->nAvgBytesPerSec = lsmash_bs_get_le32( bs );
 160     wfx->nBlockAlign     = lsmash_bs_get_le16( bs );
 161     wfx->wBitsPerSample  = lsmash_bs_get_le16( bs );
 162     switch( wfx->wFormatTag )
 163     {
 164         case WAVE_FORMAT_TYPE_ID_PCM :
 165             return 0;
 166         case WAVE_FORMAT_TYPE_ID_EXTENSIBLE :
 167             wfx->cbSize = lsmash_bs_get_le16( bs );
 168             if( wfx->cbSize < 22 )
 169                 return LSMASH_ERR_INVALID_DATA;
 170             fmt->Samples.wValidBitsPerSample = lsmash_bs_get_le16( bs );
 171             fmt->dwChannelMask               = lsmash_bs_get_le32( bs );
 172             if( lsmash_bs_get_bytes_ex( bs, 16, fmt->guid ) != 16 )
 173                 return LSMASH_ERR_NAMELESS;
 174             /* We support only PCM audio currently. */
 175             if( wave_fmt_subtype_cmp( fmt, WAVEFORMAT_EXTENSIBLE_SUBTYPE_GUID_PCM ) )
 176                 return LSMASH_ERR_INVALID_DATA;
 177             return 0;
 178         default :
 179             return LSMASH_ERR_NAMELESS;
 180     }
 181 }
 182
 183 static lsmash_audio_summary_t *wave_create_summary( waveformat_extensible_t *fmt )
 184 {
 185     lsmash_audio_summary_t *summary = (lsmash_audio_summary_t *)lsmash_create_summary( LSMASH_SUMMARY_TYPE_AUDIO );
 186     if( !summary )
 187         return NULL;
 188     waveformat_extended_t *wfx = &fmt->wfx;
 189     summary->sample_type      = QT_CODEC_TYPE_LPCM_AUDIO;
 190     summary->aot              = MP4A_AUDIO_OBJECT_TYPE_NULL;
 191     summary->frequency        = wfx->nSamplesPerSec;
 192     summary->channels         = wfx->nChannels;
 193     summary->sample_size      = wfx->wFormatTag == WAVE_FORMAT_TYPE_ID_EXTENSIBLE
 194                               ? fmt->Samples.wValidBitsPerSample
 195                               : wfx->wBitsPerSample;
 196     summary->samples_in_frame = 1000;   /* arbitrary */
 197     summary->sbr_mode         = MP4A_AAC_SBR_NOT_SPECIFIED;
 198     summary->bytes_per_frame  = wfx->nBlockAlign * summary->samples_in_frame;
 199     summary->max_au_length    = summary->bytes_per_frame;
 200     lsmash_codec_specific_t *cs = lsmash_create_codec_specific_data( LSMASH_CODEC_SPECIFIC_DATA_TYPE_QT_AUDIO_FORMAT_SPECIFIC_FLAGS,
 201                                                                      LSMASH_CODEC_SPECIFIC_FORMAT_STRUCTURED );
 202     if( !cs )
 203         goto fail;
 204     lsmash_qt_audio_format_specific_flags_t *lpcm = (lsmash_qt_audio_format_specific_flags_t *)cs->data.structured;
 205     if( (summary->sample_size & 7) == 0 )
 206         lpcm->format_flags |= QT_AUDIO_FORMAT_FLAG_PACKED;
 207     else
 208         lpcm->format_flags |= QT_AUDIO_FORMAT_FLAG_ALIGNED_HIGH;
 209     if( summary->sample_size > 8 )
 210         lpcm->format_flags |= QT_AUDIO_FORMAT_FLAG_SIGNED_INTEGER;
 211     if( lsmash_add_entry( &summary->opaque->list, cs ) < 0 )
 212     {
 213         lsmash_destroy_codec_specific_data( cs );
 214         goto fail;
 215     }
 216     if( wfx->wFormatTag == WAVE_FORMAT_TYPE_ID_EXTENSIBLE || wfx->nChannels > 2 )
 217     {
 218         cs = lsmash_create_codec_specific_data( LSMASH_CODEC_SPECIFIC_DATA_TYPE_QT_AUDIO_CHANNEL_LAYOUT,
 219                                                 LSMASH_CODEC_SPECIFIC_FORMAT_STRUCTURED );
 220         if( !cs )
 221             goto fail;
 222         lsmash_qt_audio_channel_layout_t *layout = (lsmash_qt_audio_channel_layout_t *)cs->data.structured;
 223         if( wfx->wFormatTag == WAVE_FORMAT_TYPE_ID_EXTENSIBLE )
 224         {
 225             layout->channelLayoutTag = QT_CHANNEL_LAYOUT_USE_CHANNEL_BITMAP;
 226             layout->channelBitmap    = fmt->dwChannelMask;
 227         }
 228         else
 229         {
 230             layout->channelLayoutTag = QT_CHANNEL_LAYOUT_UNKNOWN | wfx->nChannels;
 231             layout->channelBitmap    = 0;
 232         }
 233         if( lsmash_add_entry( &summary->opaque->list, cs ) < 0 )
 234         {
 235             lsmash_destroy_codec_specific_data( cs );
 236             goto fail;
 237         }
 238     }
 239     return summary;
 240 fail:
 241     lsmash_cleanup_summary( (lsmash_summary_t *)summary );
 242     return NULL;
 243 }
 244
 245 static int wave_importer_probe( importer_t *importer )
 246 {
 247     wave_importer_t *wave_imp = create_wave_importer( importer );
 248     if( !wave_imp )
 249         return LSMASH_ERR_MEMORY_ALLOC;
 250     int err = 0;
 251     uint32_t filesize;
 252     lsmash_bs_t *bs = importer->bs;
 253     if( lsmash_bs_get_be32( bs ) != LSMASH_4CC( 'R', 'I', 'F', 'F' )
 254      || ((filesize = lsmash_bs_get_le32( bs ) + 8) < WAVE_MIN_FILESIZE && filesize > 8)
 255      || lsmash_bs_get_be32( bs ) != LSMASH_4CC( 'W', 'A', 'V', 'E' ) )
 256     {
 257         err = LSMASH_ERR_INVALID_DATA;
 258         goto fail;
 259     }
 260     int fmt_chunk_present  = 0;
 261     int data_chunk_present = 0;
 262     while( !bs->eob && !(fmt_chunk_present && data_chunk_present) )
 263     {
 264         uint32_t ckID   = lsmash_bs_get_be32( bs );
 265         uint32_t ckSize = lsmash_bs_get_le32( bs );
 266         lsmash_bs_reset_counter( bs );
 267         switch( ckID )
 268         {
 269             case LSMASH_4CC( 'f', 'm', 't', ' ' ) :
 270                 if( ckSize < 16 )
 271                 {
 272                     err = LSMASH_ERR_INVALID_DATA;
 273                     goto fail;
 274                 }
 275                 if( (err = wave_parse_fmt_chunk( wave_imp, bs )) < 0 )
 276                     goto fail;
 277                 fmt_chunk_present = 1;
 278                 break;
 279             case LSMASH_4CC( 'd', 'a', 't', 'a' ) :
 280                 if( !fmt_chunk_present )
 281                 {
 282                     /* The 'fmt ' chunk must be present before the 'data' chunk. */
 283                     err = LSMASH_ERR_INVALID_DATA;
 284                     goto fail;
 285                 }
 286                 wave_imp->number_of_samples = ckSize / wave_imp->fmt.wfx.nBlockAlign;
 287                 data_chunk_present = 1;
 288                 break;
 289             default :
 290                 break;
 291         }
 292         if( !data_chunk_present )
 293         {
 294             /* Skip the rest of this chunk.
 295              * Note that ckData is word-aligned even if ckSize is an odd number. */
 296             uint32_t skip_size = ckSize;
 297             if( skip_size & 1 )
 298                 skip_size++;
 299             if( skip_size > lsmash_bs_count( bs ) )
 300             {
 301                 skip_size -= lsmash_bs_count( bs );
 302                 lsmash_bs_read_seek( bs, skip_size, SEEK_CUR );
 303             }
 304         }
 305     }
 306     if( !(fmt_chunk_present && data_chunk_present) )
 307     {
 308         err = LSMASH_ERR_INVALID_DATA;
 309         goto fail;
 310     }
 311     lsmash_audio_summary_t *summary = wave_create_summary( &wave_imp->fmt );
 312     if( !summary )
 313     {
 314         err = LSMASH_ERR_NAMELESS;
 315         goto fail;
 316     }
 317     if( (err = lsmash_add_entry( importer->summaries, summary )) < 0 )
 318     {
 319         lsmash_cleanup_summary( (lsmash_summary_t *)summary );
 320         goto fail;
 321     }
 322     importer->info   = wave_imp;
 323     importer->status = IMPORTER_OK;
 324     return 0;
 325 fail:
 326     remove_wave_importer( wave_imp );
 327     importer->info = NULL;
 328     return err;
 329 }
 330
 331 static uint32_t wave_importer_get_last_delta( importer_t *importer, uint32_t track_number )
 332 {
 333     debug_if( !importer || !importer->info )
 334         return 0;
 335     wave_importer_t *wave_imp = (wave_importer_t *)importer->info;
 336     if( !wave_imp || track_number != 1 || importer->status != IMPORTER_EOF )
 337         return 0;
 338     lsmash_audio_summary_t *summary = (lsmash_audio_summary_t *)lsmash_get_entry_data( importer->summaries, track_number );
 339     if( !summary )
 340         return 0;
 341     return wave_imp->number_of_samples / summary->samples_in_frame >= wave_imp->au_number
 342          ? summary->samples_in_frame
 343          : (wave_imp->number_of_samples % summary->samples_in_frame);
 344 }
 345
 346 const importer_functions wave_importer =
 347 {
 348     { "WAVE", offsetof( importer_t, log_level ) },
 349     1,
 350     wave_importer_probe,
 351     wave_importer_get_accessunit,
 352     wave_importer_get_last_delta,
 353     wave_importer_cleanup
 354 };