3 * Copyright (c) 2003 Fabrice Bellard
5 * This file is part of Libav.
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.
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.
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
22 #include "libavutil/avstring.h"
23 #include "libavutil/intreadwrite.h"
24 #include "libavutil/crc.h"
25 #include "libavutil/dict.h"
26 #include "libavutil/mathematics.h"
29 #include "avio_internal.h"
32 #include "replaygain.h"
34 #include "libavcodec/avcodec.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
44 typedef struct MP3DecContext {
46 unsigned frames; /* Total number of frames in file */
47 unsigned size; /* Total number of bytes in the stream */
53 static int mp3_read_probe(AVProbeData *p)
55 int max_frames, first_frames = 0;
56 int fsize, frames, sample_rate;
58 uint8_t *buf, *buf0, *buf2, *end;
59 AVCodecContext *avctx = avcodec_alloc_context3(NULL);
62 return AVERROR(ENOMEM);
65 end = p->buf + p->buf_size - sizeof(uint32_t);
66 while(buf0 < end && !*buf0)
72 for(; buf < end; buf= buf2+1) {
75 for(frames = 0; buf2 < end; frames++) {
76 header = AV_RB32(buf2);
77 fsize = avpriv_mpa_decode_header(avctx, header, &sample_rate,
78 &sample_rate, &sample_rate,
84 max_frames = FFMAX(max_frames, frames);
88 avcodec_free_context(&avctx);
89 // keep this in sync with ac3 probe, both need to avoid
90 // issues with MPEG-files!
91 if (first_frames >= 10)
92 return AVPROBE_SCORE_EXTENSION + 5;
93 if (first_frames >= 4)
94 return AVPROBE_SCORE_EXTENSION + 1;
98 unsigned int code = -1;
100 #define VIDEO_ID 0x000001e0
101 #define AUDIO_ID 0x000001c0
102 /* do a search for mpegps headers to be able to properly bias
103 * towards mpegps if we detect this stream as both. */
104 for (i = 0; i<p->buf_size; i++) {
105 code = (code << 8) + p->buf[i];
106 if ((code & 0xffffff00) == 0x100) {
107 if ((code & 0x1f0) == VIDEO_ID) pes++;
108 else if((code & 0x1e0) == AUDIO_ID) pes++;
113 max_frames = (max_frames + pes - 1) / pes;
115 if (max_frames > 500) return AVPROBE_SCORE_EXTENSION;
116 else if (max_frames >= 4) return AVPROBE_SCORE_EXTENSION / 2;
117 else if (max_frames >= 1) return 1;
119 //mpegps_mp3_unrecognized_format.mpg has max_frames=3
122 static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration)
125 MP3DecContext *mp3 = s->priv_data;
128 !(filesize = avio_size(s->pb))) {
129 av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n");
133 for (i = 0; i < XING_TOC_COUNT; i++) {
134 uint8_t b = avio_r8(s->pb);
136 av_add_index_entry(s->streams[0],
137 av_rescale(b, filesize, 256),
138 av_rescale(i, duration, XING_TOC_COUNT),
139 0, 0, AVINDEX_KEYFRAME);
144 static void mp3_parse_info_tag(AVFormatContext *s, AVStream *st,
145 MPADecodeHeader *c, uint32_t spf)
147 #define LAST_BITS(k, n) ((k) & ((1 << (n)) - 1))
148 #define MIDDLE_BITS(k, m, n) LAST_BITS((k) >> (m), ((n) - (m)))
156 int32_t r_gain = INT32_MIN, a_gain = INT32_MIN;
158 MP3DecContext *mp3 = s->priv_data;
159 const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}};
161 /* Check for Xing / Info tag */
162 avio_skip(s->pb, xing_offtbl[c->lsf == 1][c->nb_channels == 1]);
163 v = avio_rb32(s->pb);
164 mp3->is_cbr = v == MKBETAG('I', 'n', 'f', 'o');
165 if (v != MKBETAG('X', 'i', 'n', 'g') && !mp3->is_cbr)
168 v = avio_rb32(s->pb);
169 if (v & XING_FLAG_FRAMES)
170 mp3->frames = avio_rb32(s->pb);
171 if (v & XING_FLAG_SIZE)
172 mp3->size = avio_rb32(s->pb);
173 if (v & XING_FLAG_TOC && mp3->frames)
174 read_xing_toc(s, mp3->size, av_rescale_q(mp3->frames,
175 (AVRational){spf, c->sample_rate},
179 if (v & XING_FLAC_QSCALE)
182 /* Encoder short version string */
183 memset(version, 0, sizeof(version));
184 avio_read(s->pb, version, 9);
186 /* Info Tag revision + VBR method */
189 /* Lowpass filter value */
192 /* ReplayGain peak */
193 v = avio_rb32(s->pb);
194 peak = av_rescale(v, 100000, 1 << 23);
196 /* Radio ReplayGain */
197 v = avio_rb16(s->pb);
199 if (MIDDLE_BITS(v, 13, 15) == 1) {
200 r_gain = MIDDLE_BITS(v, 0, 8) * 10000;
206 /* Audiophile ReplayGain */
207 v = avio_rb16(s->pb);
209 if (MIDDLE_BITS(v, 13, 15) == 2) {
210 a_gain = MIDDLE_BITS(v, 0, 8) * 10000;
216 /* Encoding flags + ATH Type */
219 /* if ABR {specified bitrate} else {minimal bitrate} */
231 /* Preset and surround info */
241 crc = ffio_get_checksum(s->pb);
242 v = avio_rb16(s->pb);
245 ff_replaygain_export_raw(st, r_gain, peak, a_gain, 0);
246 av_dict_set(&st->metadata, "encoder", version, 0);
250 static void mp3_parse_vbri_tag(AVFormatContext *s, AVStream *st, int64_t base)
253 MP3DecContext *mp3 = s->priv_data;
255 /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
256 avio_seek(s->pb, base + 4 + 32, SEEK_SET);
257 v = avio_rb32(s->pb);
258 if (v == MKBETAG('V', 'B', 'R', 'I')) {
259 /* Check tag version */
260 if (avio_rb16(s->pb) == 1) {
261 /* skip delay and quality */
263 mp3->size = avio_rb32(s->pb);
264 mp3->frames = avio_rb32(s->pb);
270 * Try to find Xing/Info/VBRI tags and compute duration from info therein
272 static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
277 MP3DecContext *mp3 = s->priv_data;
279 ffio_init_checksum(s->pb, ff_crcA001_update, 0);
281 v = avio_rb32(s->pb);
282 if(ff_mpa_check_header(v) < 0)
285 if (avpriv_mpegaudio_decode_header(&c, v) == 0)
286 vbrtag_size = c.frame_size;
290 spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
295 mp3_parse_info_tag(s, st, &c, spf);
296 mp3_parse_vbri_tag(s, st, base);
298 if (!mp3->frames && !mp3->size)
301 /* Skip the vbr tag frame */
302 avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
305 st->duration = av_rescale_q(mp3->frames, (AVRational){spf, c.sample_rate},
307 if (mp3->size && mp3->frames && !mp3->is_cbr)
308 st->codec->bit_rate = av_rescale(mp3->size, 8 * c.sample_rate, mp3->frames * (int64_t)spf);
313 static int mp3_read_header(AVFormatContext *s)
319 st = avformat_new_stream(s, NULL);
321 return AVERROR(ENOMEM);
323 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
324 st->codec->codec_id = AV_CODEC_ID_MP3;
325 st->need_parsing = AVSTREAM_PARSE_FULL;
328 // lcm of all mp3 sample rates
329 avpriv_set_pts_info(st, 64, 1, 14112000);
331 off = avio_tell(s->pb);
333 if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
336 if (mp3_parse_vbr_tags(s, st, off) < 0)
337 avio_seek(s->pb, off, SEEK_SET);
339 ret = ff_replaygain_export(st, s->metadata);
343 /* the parameters will be extracted from the compressed bitstream */
347 #define MP3_PACKET_SIZE 1024
349 static int mp3_read_packet(AVFormatContext *s, AVPacket *pkt)
353 ret = av_get_packet(s->pb, pkt, MP3_PACKET_SIZE);
357 pkt->stream_index = 0;
359 if (ret > ID3v1_TAG_SIZE &&
360 memcmp(&pkt->data[ret - ID3v1_TAG_SIZE], "TAG", 3) == 0)
361 ret -= ID3v1_TAG_SIZE;
363 /* note: we need to modify the packet size here to handle the last
369 #define SEEK_PACKETS 4
370 #define SEEK_WINDOW (SEEK_PACKETS * MP3_PACKET_SIZE)
372 /* The toc entry can position to the wrong byte offset, try to pick
373 * the closest frame by probing the data in a window of 4 packets.
376 static int check(AVIOContext *pb, int64_t pos, int64_t *out_pos)
378 MPADecodeHeader mh = { 0 };
384 for (i = 0; i < SEEK_PACKETS; i++) {
385 off = avio_seek(pb, pos + mh.frame_size, SEEK_SET);
389 header = avio_rb32(pb);
391 if (ff_mpa_check_header(header) < 0 ||
392 avpriv_mpegaudio_decode_header(&mh, header))
400 static int reposition(AVFormatContext *s, int64_t pos)
402 int ret, best_valid = -1;
403 int64_t p, best_pos = -1;
405 for (p = FFMAX(pos - SEEK_WINDOW / 2, 0); p < pos + SEEK_WINDOW / 2; p++) {
406 int64_t out_pos[SEEK_PACKETS];
407 ret = check(s->pb, p, out_pos);
409 if (best_valid < ret) {
411 for (i = 0; i < ret; i++) {
412 if (llabs(best_pos - pos) > llabs(out_pos[i] - pos)) {
413 best_pos = out_pos[i];
417 if (best_pos == pos && best_valid == SEEK_PACKETS)
423 return AVERROR(ENOSYS);
425 p = avio_seek(s->pb, best_pos, SEEK_SET);
432 static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
435 MP3DecContext *mp3 = s->priv_data;
437 AVStream *st = s->streams[0];
438 int64_t ret = av_index_search_timestamp(st, timestamp, flags);
441 return AVERROR(ENOSYS);
446 ie = &st->index_entries[ret];
448 ret = reposition(s, ie->pos);
452 ff_update_cur_dts(s, st, ie->timestamp);
457 AVInputFormat ff_mp3_demuxer = {
459 .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
460 .read_probe = mp3_read_probe,
461 .read_header = mp3_read_header,
462 .read_packet = mp3_read_packet,
463 .read_seek = mp3_seek,
464 .priv_data_size = sizeof(MP3DecContext),
465 .flags = AVFMT_GENERIC_INDEX,
466 .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */