FFmpeg: libavformat/wav.c Source File

00001 /*
00002  * WAV muxer and demuxer
00003  * Copyright (c) 2001, 2002 Fabrice Bellard
00004  *
00005  * Sony Wave64 demuxer
00006  * RF64 demuxer
00007  * Copyright (c) 2009 Daniel Verkamp
00008  *
00009  * This file is part of FFmpeg.
00010  *
00011  * FFmpeg is free software; you can redistribute it and/or
00012  * modify it under the terms of the GNU Lesser General Public
00013  * License as published by the Free Software Foundation; either
00014  * version 2.1 of the License, or (at your option) any later version.
00015  *
00016  * FFmpeg is distributed in the hope that it will be useful,
00017  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00018  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00019  * Lesser General Public License for more details.
00020  *
00021  * You should have received a copy of the GNU Lesser General Public
00022  * License along with FFmpeg; if not, write to the Free Software
00023  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00024  */
00025 #include "avformat.h"
00026 #include "avio_internal.h"
00027 #include "pcm.h"
00028 #include "riff.h"
00029 #include "metadata.h"
00030 
00031 typedef struct {
00032     int64_t data;
00033     int64_t data_end;
00034     int64_t minpts;
00035     int64_t maxpts;
00036     int last_duration;
00037     int w64;
00038 } WAVContext;
00039 
00040 #if CONFIG_WAV_MUXER
00041 static int wav_write_header(AVFormatContext *s)
00042 {
00043     WAVContext *wav = s->priv_data;
00044     AVIOContext *pb = s->pb;
00045     int64_t fmt, fact;
00046 
00047     ffio_wfourcc(pb, "RIFF");
00048     avio_wl32(pb, 0); /* file length */
00049     ffio_wfourcc(pb, "WAVE");
00050 
00051     /* format header */
00052     fmt = ff_start_tag(pb, "fmt ");
00053     if (ff_put_wav_header(pb, s->streams[0]->codec) < 0) {
00054         av_log(s, AV_LOG_ERROR, "%s codec not supported in WAVE format\n",
00055                s->streams[0]->codec->codec ? s->streams[0]->codec->codec->name : "NONE");
00056         return -1;
00057     }
00058     ff_end_tag(pb, fmt);
00059 
00060     if (s->streams[0]->codec->codec_tag != 0x01 /* hence for all other than PCM */
00061         && s->pb->seekable) {
00062         fact = ff_start_tag(pb, "fact");
00063         avio_wl32(pb, 0);
00064         ff_end_tag(pb, fact);
00065     }
00066 
00067     av_set_pts_info(s->streams[0], 64, 1, s->streams[0]->codec->sample_rate);
00068     wav->maxpts = wav->last_duration = 0;
00069     wav->minpts = INT64_MAX;
00070 
00071     /* data header */
00072     wav->data = ff_start_tag(pb, "data");
00073 
00074     avio_flush(pb);
00075 
00076     return 0;
00077 }
00078 
00079 static int wav_write_packet(AVFormatContext *s, AVPacket *pkt)
00080 {
00081     AVIOContext *pb  = s->pb;
00082     WAVContext    *wav = s->priv_data;
00083     avio_write(pb, pkt->data, pkt->size);
00084     if(pkt->pts != AV_NOPTS_VALUE) {
00085         wav->minpts        = FFMIN(wav->minpts, pkt->pts);
00086         wav->maxpts        = FFMAX(wav->maxpts, pkt->pts);
00087         wav->last_duration = pkt->duration;
00088     } else
00089         av_log(s, AV_LOG_ERROR, "wav_write_packet: NOPTS\n");
00090     return 0;
00091 }
00092 
00093 static int wav_write_trailer(AVFormatContext *s)
00094 {
00095     AVIOContext *pb  = s->pb;
00096     WAVContext    *wav = s->priv_data;
00097     int64_t file_size;
00098 
00099     avio_flush(pb);
00100 
00101     if (s->pb->seekable) {
00102         ff_end_tag(pb, wav->data);
00103 
00104         /* update file size */
00105         file_size = avio_tell(pb);
00106         avio_seek(pb, 4, SEEK_SET);
00107         avio_wl32(pb, (uint32_t)(file_size - 8));
00108         avio_seek(pb, file_size, SEEK_SET);
00109 
00110         avio_flush(pb);
00111 
00112         if(s->streams[0]->codec->codec_tag != 0x01) {
00113             /* Update num_samps in fact chunk */
00114             int number_of_samples;
00115             number_of_samples = av_rescale(wav->maxpts - wav->minpts + wav->last_duration,
00116                                            s->streams[0]->codec->sample_rate * (int64_t)s->streams[0]->time_base.num,
00117                                            s->streams[0]->time_base.den);
00118             avio_seek(pb, wav->data-12, SEEK_SET);
00119             avio_wl32(pb, number_of_samples);
00120             avio_seek(pb, file_size, SEEK_SET);
00121             avio_flush(pb);
00122         }
00123     }
00124     return 0;
00125 }
00126 
00127 AVOutputFormat ff_wav_muxer = {
00128     "wav",
00129     NULL_IF_CONFIG_SMALL("WAV format"),
00130     "audio/x-wav",
00131     "wav",
00132     sizeof(WAVContext),
00133     CODEC_ID_PCM_S16LE,
00134     CODEC_ID_NONE,
00135     wav_write_header,
00136     wav_write_packet,
00137     wav_write_trailer,
00138     .codec_tag= (const AVCodecTag* const []){ff_codec_wav_tags, 0},
00139 };
00140 #endif /* CONFIG_WAV_MUXER */
00141 
00142 
00143 #if CONFIG_WAV_DEMUXER
00144 
00145 static int64_t next_tag(AVIOContext *pb, unsigned int *tag)
00146 {
00147     *tag = avio_rl32(pb);
00148     return avio_rl32(pb);
00149 }
00150 
00151 /* return the size of the found tag */
00152 static int64_t find_tag(AVIOContext *pb, uint32_t tag1)
00153 {
00154     unsigned int tag;
00155     int64_t size;
00156 
00157     for (;;) {
00158         if (url_feof(pb))
00159             return -1;
00160         size = next_tag(pb, &tag);
00161         if (tag == tag1)
00162             break;
00163         avio_skip(pb, size);
00164     }
00165     return size;
00166 }
00167 
00168 static int wav_probe(AVProbeData *p)
00169 {
00170     /* check file header */
00171     if (p->buf_size <= 32)
00172         return 0;
00173     if (!memcmp(p->buf + 8, "WAVE", 4)) {
00174         if (!memcmp(p->buf, "RIFF", 4))
00175             /*
00176               Since ACT demuxer has standard WAV header at top of it's own,
00177               returning score is decreased to avoid probe conflict
00178               between ACT and WAV.
00179             */
00180             return AVPROBE_SCORE_MAX - 1;
00181         else if (!memcmp(p->buf,      "RF64", 4) &&
00182                  !memcmp(p->buf + 12, "ds64", 4))
00183             return AVPROBE_SCORE_MAX;
00184     }
00185     return 0;
00186 }
00187 
00188 static int wav_parse_fmt_tag(AVFormatContext *s, int64_t size, AVStream **st)
00189 {
00190     AVIOContext *pb = s->pb;
00191     int ret;
00192 
00193     /* parse fmt header */
00194     *st = av_new_stream(s, 0);
00195     if (!*st)
00196         return AVERROR(ENOMEM);
00197 
00198     ret = ff_get_wav_header(pb, (*st)->codec, size);
00199     if (ret < 0)
00200         return ret;
00201     (*st)->need_parsing = AVSTREAM_PARSE_FULL;
00202 
00203     av_set_pts_info(*st, 64, 1, (*st)->codec->sample_rate);
00204 
00205     return 0;
00206 }
00207 
00208 static inline int wav_parse_bext_string(AVFormatContext *s, const char *key, int length)
00209 {
00210     char temp[257];
00211     int ret;
00212 
00213     if ((ret = avio_read(s->pb, temp, length)) < 0)
00214         return ret;
00215 
00216     temp[length] = 0;
00217 
00218     if (strlen(temp))
00219         return av_dict_set(&s->metadata, key, temp, 0);
00220 
00221     return 0;
00222 }
00223 
00224 static int wav_parse_bext_tag(AVFormatContext *s, int64_t size)
00225 {
00226     char temp[131], *coding_history;
00227     int ret, x;
00228     uint64_t time_reference;
00229     int64_t umid_parts[8], umid_mask = 0;
00230 
00231     if ((ret = wav_parse_bext_string(s, "description", 256)) < 0 ||
00232         (ret = wav_parse_bext_string(s, "originator", 32)) < 0 ||
00233         (ret = wav_parse_bext_string(s, "originator_reference", 32)) < 0 ||
00234         (ret = wav_parse_bext_string(s, "origination_date", 10)) < 0 ||
00235         (ret = wav_parse_bext_string(s, "origination_time", 8)) < 0)
00236         return ret;
00237 
00238     time_reference = avio_rl64(s->pb);
00239     snprintf(temp, sizeof(temp), "%"PRIu64, time_reference);
00240     if ((ret = av_dict_set(&s->metadata, "time_reference", temp, 0)) < 0)
00241         return ret;
00242 
00243     /* check if version is >= 1, in which case an UMID may be present */
00244     if (avio_rl16(s->pb) >= 1) {
00245         for (x = 0; x < 8; x++)
00246             umid_mask |= umid_parts[x] = avio_rb64(s->pb);
00247 
00248         if (umid_mask) {
00249             /* the string formatting below is per SMPTE 330M-2004 Annex C */
00250             if (umid_parts[4] == 0 && umid_parts[5] == 0 && umid_parts[6] == 0 && umid_parts[7] == 0) {
00251                 /* basic UMID */
00252                 snprintf(temp, sizeof(temp), "0x%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64,
00253                          umid_parts[0], umid_parts[1], umid_parts[2], umid_parts[3]);
00254             } else {
00255                 /* extended UMID */
00256                 snprintf(temp, sizeof(temp), "0x%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64
00257                                              "0x%016"PRIX64"%016"PRIX64"%016"PRIX64"%016"PRIX64,
00258                          umid_parts[0], umid_parts[1], umid_parts[2], umid_parts[3],
00259                          umid_parts[4], umid_parts[5], umid_parts[6], umid_parts[7]);
00260             }
00261 
00262             if ((ret = av_dict_set(&s->metadata, "umid", temp, 0)) < 0)
00263                 return ret;
00264         }
00265 
00266         avio_skip(s->pb, 190);
00267     } else
00268         avio_skip(s->pb, 254);
00269 
00270     if (size > 602) {
00271         /* CodingHistory present */
00272         size -= 602;
00273 
00274         if (!(coding_history = av_malloc(size+1)))
00275             return AVERROR(ENOMEM);
00276 
00277         if ((ret = avio_read(s->pb, coding_history, size)) < 0)
00278             return ret;
00279 
00280         coding_history[size] = 0;
00281         if ((ret = av_dict_set(&s->metadata, "coding_history", coding_history,
00282                                AV_METADATA_DONT_STRDUP_VAL)) < 0)
00283             return ret;
00284     }
00285 
00286     return 0;
00287 }
00288 
00289 static const AVMetadataConv wav_metadata_conv[] = {
00290     {"description",      "comment"      },
00291     {"originator",       "encoded_by"   },
00292     {"origination_date", "date"         },
00293     {"origination_time", "creation_time"},
00294     {0},
00295 };
00296 
00297 /* wav input */
00298 static int wav_read_header(AVFormatContext *s,
00299                            AVFormatParameters *ap)
00300 {
00301     int64_t size, av_uninit(data_size);
00302     int64_t sample_count=0;
00303     int rf64;
00304     unsigned int tag;
00305     AVIOContext *pb = s->pb;
00306     AVStream *st;
00307     WAVContext *wav = s->priv_data;
00308     int ret, got_fmt = 0;
00309     int64_t next_tag_ofs, data_ofs = -1;
00310 
00311     /* check RIFF header */
00312     tag = avio_rl32(pb);
00313 
00314     rf64 = tag == MKTAG('R', 'F', '6', '4');
00315     if (!rf64 && tag != MKTAG('R', 'I', 'F', 'F'))
00316         return -1;
00317     avio_rl32(pb); /* file size */
00318     tag = avio_rl32(pb);
00319     if (tag != MKTAG('W', 'A', 'V', 'E'))
00320         return -1;
00321 
00322     if (rf64) {
00323         if (avio_rl32(pb) != MKTAG('d', 's', '6', '4'))
00324             return -1;
00325         size = avio_rl32(pb);
00326         if (size < 24)
00327             return -1;
00328         avio_rl64(pb); /* RIFF size */
00329         data_size = avio_rl64(pb);
00330         sample_count = avio_rl64(pb);
00331         if (data_size < 0 || sample_count < 0) {
00332             av_log(s, AV_LOG_ERROR, "negative data_size and/or sample_count in "
00333                    "ds64: data_size = %"PRId64", sample_count = %"PRId64"\n",
00334                    data_size, sample_count);
00335             return AVERROR_INVALIDDATA;
00336         }
00337         avio_skip(pb, size - 24); /* skip rest of ds64 chunk */
00338     }
00339 
00340     for (;;) {
00341         size = next_tag(pb, &tag);
00342         next_tag_ofs = avio_tell(pb) + size;
00343 
00344         if (url_feof(pb))
00345             break;
00346 
00347         switch (tag) {
00348         case MKTAG('f', 'm', 't', ' '):
00349             /* only parse the first 'fmt ' tag found */
00350             if (!got_fmt && (ret = wav_parse_fmt_tag(s, size, &st) < 0)) {
00351                 return ret;
00352             } else if (got_fmt)
00353                 av_log(s, AV_LOG_WARNING, "found more than one 'fmt ' tag\n");
00354 
00355             got_fmt = 1;
00356             break;
00357         case MKTAG('d', 'a', 't', 'a'):
00358             if (!got_fmt) {
00359                 av_log(s, AV_LOG_ERROR, "found no 'fmt ' tag before the 'data' tag\n");
00360                 return AVERROR_INVALIDDATA;
00361             }
00362 
00363             if (rf64) {
00364                 next_tag_ofs = wav->data_end = avio_tell(pb) + data_size;
00365             } else {
00366                 data_size = size;
00367                 next_tag_ofs = wav->data_end = size ? next_tag_ofs : INT64_MAX;
00368             }
00369 
00370             data_ofs = avio_tell(pb);
00371 
00372             /* don't look for footer metadata if we can't seek or if we don't
00373              * know where the data tag ends
00374              */
00375             if (!pb->seekable || (!rf64 && !size))
00376                 goto break_loop;
00377             break;
00378         case MKTAG('f','a','c','t'):
00379             if(!sample_count)
00380                 sample_count = avio_rl32(pb);
00381             break;
00382         case MKTAG('b','e','x','t'):
00383             if ((ret = wav_parse_bext_tag(s, size)) < 0)
00384                 return ret;
00385             break;
00386         }
00387 
00388         /* seek to next tag unless we know that we'll run into EOF */
00389         if ((avio_size(pb) > 0 && next_tag_ofs >= avio_size(pb)) ||
00390             avio_seek(pb, next_tag_ofs, SEEK_SET) < 0) {
00391             break;
00392         }
00393     }
00394 break_loop:
00395     if (data_ofs < 0) {
00396         av_log(s, AV_LOG_ERROR, "no 'data' tag found\n");
00397         return AVERROR_INVALIDDATA;
00398     }
00399 
00400     avio_seek(pb, data_ofs, SEEK_SET);
00401 
00402     if (!sample_count && st->codec->channels && av_get_bits_per_sample(st->codec->codec_id))
00403         sample_count = (data_size<<3) / (st->codec->channels * (uint64_t)av_get_bits_per_sample(st->codec->codec_id));
00404     if (sample_count)
00405         st->duration = sample_count;
00406 
00407     ff_metadata_conv_ctx(s, NULL, wav_metadata_conv);
00408 
00409     return 0;
00410 }
00411 
00415 static int64_t find_guid(AVIOContext *pb, const uint8_t guid1[16])
00416 {
00417     uint8_t guid[16];
00418     int64_t size;
00419 
00420     while (!url_feof(pb)) {
00421         avio_read(pb, guid, 16);
00422         size = avio_rl64(pb);
00423         if (size <= 24)
00424             return -1;
00425         if (!memcmp(guid, guid1, 16))
00426             return size;
00427         avio_skip(pb, FFALIGN(size, INT64_C(8)) - 24);
00428     }
00429     return -1;
00430 }
00431 
00432 static const uint8_t guid_data[16] = { 'd', 'a', 't', 'a',
00433     0xF3, 0xAC, 0xD3, 0x11, 0x8C, 0xD1, 0x00, 0xC0, 0x4F, 0x8E, 0xDB, 0x8A };
00434 
00435 #define MAX_SIZE 4096
00436 
00437 static int wav_read_packet(AVFormatContext *s,
00438                            AVPacket *pkt)
00439 {
00440     int ret, size;
00441     int64_t left;
00442     AVStream *st;
00443     WAVContext *wav = s->priv_data;
00444 
00445     st = s->streams[0];
00446 
00447     left = wav->data_end - avio_tell(s->pb);
00448     if (left <= 0){
00449         if (CONFIG_W64_DEMUXER && wav->w64)
00450             left = find_guid(s->pb, guid_data) - 24;
00451         else
00452             left = find_tag(s->pb, MKTAG('d', 'a', 't', 'a'));
00453         if (left < 0)
00454             return AVERROR_EOF;
00455         wav->data_end= avio_tell(s->pb) + left;
00456     }
00457 
00458     size = MAX_SIZE;
00459     if (st->codec->block_align > 1) {
00460         if (size < st->codec->block_align)
00461             size = st->codec->block_align;
00462         size = (size / st->codec->block_align) * st->codec->block_align;
00463     }
00464     size = FFMIN(size, left);
00465     ret  = av_get_packet(s->pb, pkt, size);
00466     if (ret < 0)
00467         return ret;
00468     pkt->stream_index = 0;
00469 
00470     return ret;
00471 }
00472 
00473 static int wav_read_seek(AVFormatContext *s,
00474                          int stream_index, int64_t timestamp, int flags)
00475 {
00476     AVStream *st;
00477 
00478     st = s->streams[0];
00479     switch (st->codec->codec_id) {
00480     case CODEC_ID_MP2:
00481     case CODEC_ID_MP3:
00482     case CODEC_ID_AC3:
00483     case CODEC_ID_DTS:
00484         /* use generic seeking with dynamically generated indexes */
00485         return -1;
00486     default:
00487         break;
00488     }
00489     return pcm_read_seek(s, stream_index, timestamp, flags);
00490 }
00491 
00492 AVInputFormat ff_wav_demuxer = {
00493     "wav",
00494     NULL_IF_CONFIG_SMALL("WAV format"),
00495     sizeof(WAVContext),
00496     wav_probe,
00497     wav_read_header,
00498     wav_read_packet,
00499     NULL,
00500     wav_read_seek,
00501     .flags= AVFMT_GENERIC_INDEX,
00502     .codec_tag= (const AVCodecTag* const []){ff_codec_wav_tags, 0},
00503 };
00504 #endif /* CONFIG_WAV_DEMUXER */
00505 
00506 
00507 #if CONFIG_W64_DEMUXER
00508 static const uint8_t guid_riff[16] = { 'r', 'i', 'f', 'f',
00509     0x2E, 0x91, 0xCF, 0x11, 0xA5, 0xD6, 0x28, 0xDB, 0x04, 0xC1, 0x00, 0x00 };
00510 
00511 static const uint8_t guid_wave[16] = { 'w', 'a', 'v', 'e',
00512     0xF3, 0xAC, 0xD3, 0x11, 0x8C, 0xD1, 0x00, 0xC0, 0x4F, 0x8E, 0xDB, 0x8A };
00513 
00514 static const uint8_t guid_fmt [16] = { 'f', 'm', 't', ' ',
00515     0xF3, 0xAC, 0xD3, 0x11, 0x8C, 0xD1, 0x00, 0xC0, 0x4F, 0x8E, 0xDB, 0x8A };
00516 
00517 static int w64_probe(AVProbeData *p)
00518 {
00519     if (p->buf_size <= 40)
00520         return 0;
00521     if (!memcmp(p->buf,      guid_riff, 16) &&
00522         !memcmp(p->buf + 24, guid_wave, 16))
00523         return AVPROBE_SCORE_MAX;
00524     else
00525         return 0;
00526 }
00527 
00528 static int w64_read_header(AVFormatContext *s, AVFormatParameters *ap)
00529 {
00530     int64_t size;
00531     AVIOContext *pb  = s->pb;
00532     WAVContext    *wav = s->priv_data;
00533     AVStream *st;
00534     uint8_t guid[16];
00535     int ret;
00536 
00537     avio_read(pb, guid, 16);
00538     if (memcmp(guid, guid_riff, 16))
00539         return -1;
00540 
00541     if (avio_rl64(pb) < 16 + 8 + 16 + 8 + 16 + 8) /* riff + wave + fmt + sizes */
00542         return -1;
00543 
00544     avio_read(pb, guid, 16);
00545     if (memcmp(guid, guid_wave, 16)) {
00546         av_log(s, AV_LOG_ERROR, "could not find wave guid\n");
00547         return -1;
00548     }
00549 
00550     size = find_guid(pb, guid_fmt);
00551     if (size < 0) {
00552         av_log(s, AV_LOG_ERROR, "could not find fmt guid\n");
00553         return -1;
00554     }
00555 
00556     st = av_new_stream(s, 0);
00557     if (!st)
00558         return AVERROR(ENOMEM);
00559 
00560     /* subtract chunk header size - normal wav file doesn't count it */
00561     ret = ff_get_wav_header(pb, st->codec, size - 24);
00562     if (ret < 0)
00563         return ret;
00564     avio_skip(pb, FFALIGN(size, INT64_C(8)) - size);
00565 
00566     st->need_parsing = AVSTREAM_PARSE_FULL;
00567 
00568     av_set_pts_info(st, 64, 1, st->codec->sample_rate);
00569 
00570     size = find_guid(pb, guid_data);
00571     if (size < 0) {
00572         av_log(s, AV_LOG_ERROR, "could not find data guid\n");
00573         return -1;
00574     }
00575     wav->data_end = avio_tell(pb) + size - 24;
00576     wav->w64      = 1;
00577 
00578     return 0;
00579 }
00580 
00581 AVInputFormat ff_w64_demuxer = {
00582     "w64",
00583     NULL_IF_CONFIG_SMALL("Sony Wave64 format"),
00584     sizeof(WAVContext),
00585     w64_probe,
00586     w64_read_header,
00587     wav_read_packet,
00588     NULL,
00589     wav_read_seek,
00590     .flags = AVFMT_GENERIC_INDEX,
00591     .codec_tag = (const AVCodecTag* const []){ff_codec_wav_tags, 0},
00592 };
00593 #endif /* CONFIG_W64_DEMUXER */