3 * Copyright (c) 2003 Fabrice Bellard
5 * This file is part of FFmpeg.
7 * FFmpeg 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.
12 * FFmpeg 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.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with FFmpeg; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 #include "libavutil/opt.h"
23 #include "libavutil/avstring.h"
24 #include "libavutil/intreadwrite.h"
25 #include "libavutil/crc.h"
26 #include "libavutil/dict.h"
27 #include "libavutil/mathematics.h"
30 #include "avio_internal.h"
33 #include "replaygain.h"
35 #include "libavcodec/mpegaudiodecheader.h"
37 #define XING_FLAG_FRAMES 0x01
38 #define XING_FLAG_SIZE 0x02
39 #define XING_FLAG_TOC 0x04
40 #define XING_FLAC_QSCALE 0x08
42 #define XING_TOC_COUNT 100
51 unsigned frames
; /* Total number of frames in file */
52 unsigned header_filesize
; /* Total number of bytes in the stream */
58 static int mp3_read_probe(AVProbeData
*p
)
60 int max_frames
, first_frames
= 0;
61 int fsize
, frames
, sample_rate
;
63 const uint8_t *buf
, *buf0
, *buf2
, *end
;
67 end
= p
->buf
+ p
->buf_size
- sizeof(uint32_t);
68 while(buf0
< end
&& !*buf0
)
74 for(; buf
< end
; buf
= buf2
+1) {
76 if(ff_mpa_check_header(AV_RB32(buf2
)))
79 for(frames
= 0; buf2
< end
; frames
++) {
80 header
= AV_RB32(buf2
);
81 fsize
= avpriv_mpa_decode_header(&avctx
, header
, &sample_rate
, &sample_rate
, &sample_rate
, &sample_rate
);
86 max_frames
= FFMAX(max_frames
, frames
);
90 // keep this in sync with ac3 probe, both need to avoid
91 // issues with MPEG-files!
92 if (first_frames
>=4) return AVPROBE_SCORE_EXTENSION
+ 1;
93 else if(max_frames
>200)return AVPROBE_SCORE_EXTENSION
;
94 else if(max_frames
>=4 && max_frames
>= p
->buf_size
/10000) return AVPROBE_SCORE_EXTENSION
/ 2;
95 else if(ff_id3v2_match(buf0
, ID3v2_DEFAULT_MAGIC
) && 2*ff_id3v2_tag_len(buf0
) >= p
->buf_size
)
96 return p
->buf_size
< PROBE_BUF_MAX
? AVPROBE_SCORE_EXTENSION
/ 4 : AVPROBE_SCORE_EXTENSION
- 2;
97 else if(max_frames
>=1 && max_frames
>= p
->buf_size
/10000) return 1;
99 //mpegps_mp3_unrecognized_format.mpg has max_frames=3
102 static void read_xing_toc(AVFormatContext
*s
, int64_t filesize
, int64_t duration
)
105 MP3DecContext
*mp3
= s
->priv_data
;
106 int fill_index
= mp3
->usetoc
&& duration
> 0;
109 !(filesize
= avio_size(s
->pb
))) {
110 av_log(s
, AV_LOG_WARNING
, "Cannot determine file size, skipping TOC table.\n");
114 for (i
= 0; i
< XING_TOC_COUNT
; i
++) {
115 uint8_t b
= avio_r8(s
->pb
);
117 av_add_index_entry(s
->streams
[0],
118 av_rescale(b
, filesize
, 256),
119 av_rescale(i
, duration
, XING_TOC_COUNT
),
120 0, 0, AVINDEX_KEYFRAME
);
126 static void mp3_parse_info_tag(AVFormatContext
*s
, AVStream
*st
,
127 MPADecodeHeader
*c
, uint32_t spf
)
129 #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1))
130 #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m)))
138 int32_t r_gain
= INT32_MIN
, a_gain
= INT32_MIN
;
140 MP3DecContext
*mp3
= s
->priv_data
;
141 static const int64_t xing_offtbl
[2][2] = {{32, 17}, {17,9}};
142 uint64_t fsize
= avio_size(s
->pb
);
144 /* Check for Xing / Info tag */
145 avio_skip(s
->pb
, xing_offtbl
[c
->lsf
== 1][c
->nb_channels
== 1]);
146 v
= avio_rb32(s
->pb
);
147 mp3
->is_cbr
= v
== MKBETAG('I', 'n', 'f', 'o');
148 if (v
!= MKBETAG('X', 'i', 'n', 'g') && !mp3
->is_cbr
)
151 v
= avio_rb32(s
->pb
);
152 if (v
& XING_FLAG_FRAMES
)
153 mp3
->frames
= avio_rb32(s
->pb
);
154 if (v
& XING_FLAG_SIZE
)
155 mp3
->header_filesize
= avio_rb32(s
->pb
);
156 if (fsize
&& mp3
->header_filesize
) {
158 min
= FFMIN(fsize
, mp3
->header_filesize
);
159 delta
= FFMAX(fsize
, mp3
->header_filesize
) - min
;
160 if (fsize
> mp3
->header_filesize
&& delta
> min
>> 4) {
162 } else if (delta
> min
>> 4) {
163 av_log(s
, AV_LOG_WARNING
,
164 "filesize and duration do not match (growing file?)\n");
167 if (v
& XING_FLAG_TOC
)
168 read_xing_toc(s
, mp3
->header_filesize
, av_rescale_q(mp3
->frames
,
169 (AVRational
){spf
, c
->sample_rate
},
172 if (v
& XING_FLAC_QSCALE
)
175 /* Encoder short version string */
176 memset(version
, 0, sizeof(version
));
177 avio_read(s
->pb
, version
, 9);
179 /* Info Tag revision + VBR method */
182 /* Lowpass filter value */
185 /* ReplayGain peak */
186 v
= avio_rb32(s
->pb
);
187 peak
= av_rescale(v
, 100000, 1 << 23);
189 /* Radio ReplayGain */
190 v
= avio_rb16(s
->pb
);
192 if (MIDDLE_BITS(v
, 13, 15) == 1) {
193 r_gain
= MIDDLE_BITS(v
, 0, 8) * 10000;
199 /* Audiophile ReplayGain */
200 v
= avio_rb16(s
->pb
);
202 if (MIDDLE_BITS(v
, 13, 15) == 2) {
203 a_gain
= MIDDLE_BITS(v
, 0, 8) * 10000;
209 /* Encoding flags + ATH Type */
212 /* if ABR {specified bitrate} else {minimal bitrate} */
217 if(AV_RB32(version
) == MKBETAG('L', 'A', 'M', 'E')
218 || AV_RB32(version
) == MKBETAG('L', 'a', 'v', 'f')
219 || AV_RB32(version
) == MKBETAG('L', 'a', 'v', 'c')
222 mp3
->start_pad
= v
>>12;
223 mp3
-> end_pad
= v
&4095;
224 st
->skip_samples
= mp3
->start_pad
+ 528 + 1;
226 st
->first_discard_sample
= -mp3
->end_pad
+ 528 + 1 + mp3
->frames
* (int64_t)spf
;
227 st
->last_discard_sample
= mp3
->frames
* (int64_t)spf
;
230 st
->start_time
= av_rescale_q(st
->skip_samples
,
231 (AVRational
){1, c
->sample_rate
},
233 av_log(s
, AV_LOG_DEBUG
, "pad %d %d\n", mp3
->start_pad
, mp3
-> end_pad
);
242 /* Preset and surround info */
252 crc
= ffio_get_checksum(s
->pb
);
253 v
= avio_rb16(s
->pb
);
256 ff_replaygain_export_raw(st
, r_gain
, peak
, a_gain
, 0);
257 av_dict_set(&st
->metadata
, "encoder", version
, 0);
261 static void mp3_parse_vbri_tag(AVFormatContext
*s
, AVStream
*st
, int64_t base
)
264 MP3DecContext
*mp3
= s
->priv_data
;
266 /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
267 avio_seek(s
->pb
, base
+ 4 + 32, SEEK_SET
);
268 v
= avio_rb32(s
->pb
);
269 if (v
== MKBETAG('V', 'B', 'R', 'I')) {
270 /* Check tag version */
271 if (avio_rb16(s
->pb
) == 1) {
272 /* skip delay and quality */
274 mp3
->header_filesize
= avio_rb32(s
->pb
);
275 mp3
->frames
= avio_rb32(s
->pb
);
281 * Try to find Xing/Info/VBRI tags and compute duration from info therein
283 static int mp3_parse_vbr_tags(AVFormatContext
*s
, AVStream
*st
, int64_t base
)
288 MP3DecContext
*mp3
= s
->priv_data
;
290 ffio_init_checksum(s
->pb
, ff_crcA001_update
, 0);
292 v
= avio_rb32(s
->pb
);
293 if(ff_mpa_check_header(v
) < 0)
296 if (avpriv_mpegaudio_decode_header(&c
, v
) == 0)
297 vbrtag_size
= c
.frame_size
;
301 spf
= c
.lsf
? 576 : 1152; /* Samples per frame, layer 3 */
304 mp3
->header_filesize
= 0;
306 mp3_parse_info_tag(s
, st
, &c
, spf
);
307 mp3_parse_vbri_tag(s
, st
, base
);
309 if (!mp3
->frames
&& !mp3
->header_filesize
)
312 /* Skip the vbr tag frame */
313 avio_seek(s
->pb
, base
+ vbrtag_size
, SEEK_SET
);
316 st
->duration
= av_rescale_q(mp3
->frames
, (AVRational
){spf
, c
.sample_rate
},
318 if (mp3
->header_filesize
&& mp3
->frames
&& !mp3
->is_cbr
)
319 st
->codec
->bit_rate
= av_rescale(mp3
->header_filesize
, 8 * c
.sample_rate
, mp3
->frames
* (int64_t)spf
);
324 static int mp3_read_header(AVFormatContext
*s
)
326 MP3DecContext
*mp3
= s
->priv_data
;
331 st
= avformat_new_stream(s
, NULL
);
333 return AVERROR(ENOMEM
);
335 st
->codec
->codec_type
= AVMEDIA_TYPE_AUDIO
;
336 st
->codec
->codec_id
= AV_CODEC_ID_MP3
;
337 st
->need_parsing
= AVSTREAM_PARSE_FULL_RAW
;
340 // lcm of all mp3 sample rates
341 avpriv_set_pts_info(st
, 64, 1, 14112000);
344 off
= avio_tell(s
->pb
);
346 if (!av_dict_get(s
->metadata
, "", NULL
, AV_DICT_IGNORE_SUFFIX
))
350 mp3
->filesize
= avio_size(s
->pb
);
352 if (mp3_parse_vbr_tags(s
, st
, off
) < 0)
353 avio_seek(s
->pb
, off
, SEEK_SET
);
355 ret
= ff_replaygain_export(st
, s
->metadata
);
359 /* the parameters will be extracted from the compressed bitstream */
363 #define MP3_PACKET_SIZE 1024
365 static int mp3_read_packet(AVFormatContext
*s
, AVPacket
*pkt
)
367 MP3DecContext
*mp3
= s
->priv_data
;
371 size
= MP3_PACKET_SIZE
;
372 pos
= avio_tell(s
->pb
);
373 if(mp3
->filesize
> ID3v1_TAG_SIZE
&& pos
< mp3
->filesize
)
374 size
= FFMIN(size
, mp3
->filesize
- pos
);
376 ret
= av_get_packet(s
->pb
, pkt
, size
);
383 pkt
->flags
&= ~AV_PKT_FLAG_CORRUPT
;
384 pkt
->stream_index
= 0;
386 if (ret
>= ID3v1_TAG_SIZE
&&
387 memcmp(&pkt
->data
[ret
- ID3v1_TAG_SIZE
], "TAG", 3) == 0)
388 ret
-= ID3v1_TAG_SIZE
;
390 /* note: we need to modify the packet size here to handle the last
396 static int check(AVFormatContext
*s
, int64_t pos
)
398 int64_t ret
= avio_seek(s
->pb
, pos
, SEEK_SET
);
403 header
= avio_rb32(s
->pb
);
404 if (ff_mpa_check_header(header
) < 0)
406 if (avpriv_mpegaudio_decode_header(&sd
, header
) == 1)
408 return sd
.frame_size
;
411 static int mp3_seek(AVFormatContext
*s
, int stream_index
, int64_t timestamp
,
414 MP3DecContext
*mp3
= s
->priv_data
;
415 AVIndexEntry
*ie
, ie1
;
416 AVStream
*st
= s
->streams
[0];
417 int64_t ret
= av_index_search_timestamp(st
, timestamp
, flags
);
419 int dir
= (flags
&AVSEEK_FLAG_BACKWARD
) ? -1 : 1;
425 && mp3
->header_filesize
> s
->data_offset
427 int64_t filesize
= avio_size(s
->pb
);
429 if (filesize
<= s
->data_offset
)
430 filesize
= mp3
->header_filesize
;
431 filesize
-= s
->data_offset
;
432 duration
= av_rescale(st
->duration
, filesize
, mp3
->header_filesize
- s
->data_offset
);
434 timestamp
= av_clip64(timestamp
, 0, duration
);
435 ie
->timestamp
= timestamp
;
436 ie
->pos
= av_rescale(timestamp
, filesize
, duration
) + s
->data_offset
;
437 } else if (mp3
->xing_toc
) {
441 ie
= &st
->index_entries
[ret
];
443 st
->skip_samples
= timestamp
<= 0 ? mp3
->start_pad
+ 528 + 1 : 0;
448 avio_seek(s
->pb
, FFMAX(ie
->pos
- 4096, 0), SEEK_SET
);
449 ret
= avio_seek(s
->pb
, ie
->pos
, SEEK_SET
);
456 for(i
=0; i
<4096; i
++) {
457 int64_t pos
= ie
->pos
+ (dir
> 0 ? i
- 1024 : -i
);
458 int64_t candidate
= -1;
464 for(j
=0; j
<MIN_VALID
; j
++) {
468 if ((ie
->pos
- pos
)*dir
<= 0 && abs(MIN_VALID
/2-j
) < score
) {
470 score
= abs(MIN_VALID
/2-j
);
474 if (best_score
> score
&& j
== MIN_VALID
) {
475 best_pos
= candidate
;
482 ret
= avio_seek(s
->pb
, best_pos
, SEEK_SET
);
485 ff_update_cur_dts(s
, st
, ie
->timestamp
);
486 st
->skip_samples
= ie
->timestamp
<= 0 ? mp3
->start_pad
+ 528 + 1 : 0;
490 static const AVOption options
[] = {
491 { "usetoc", "use table of contents", offsetof(MP3DecContext
, usetoc
), AV_OPT_TYPE_INT
, {.i64
= -1}, -1, 1, AV_OPT_FLAG_DECODING_PARAM
},
495 static const AVClass demuxer_class
= {
497 .item_name
= av_default_item_name
,
499 .version
= LIBAVUTIL_VERSION_INT
,
500 .category
= AV_CLASS_CATEGORY_DEMUXER
,
503 AVInputFormat ff_mp3_demuxer
= {
505 .long_name
= NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
506 .read_probe
= mp3_read_probe
,
507 .read_header
= mp3_read_header
,
508 .read_packet
= mp3_read_packet
,
509 .read_seek
= mp3_seek
,
510 .priv_data_size
= sizeof(MP3DecContext
),
511 .flags
= AVFMT_GENERIC_INDEX
,
512 .extensions
= "mp2,mp3,m2a,mpa", /* XXX: use probe */
513 .priv_class
= &demuxer_class
,