1/*
2 * MP3 muxer
3 * Copyright (c) 2003 Fabrice Bellard
4 *
5 * This file is part of FFmpeg.
6 *
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.
11 *
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.
16 *
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
20 */
21
22#include "avformat.h"
23#include "avio_internal.h"
24#include "id3v1.h"
25#include "id3v2.h"
26#include "rawenc.h"
27#include "libavutil/avstring.h"
28#include "libavcodec/mpegaudio.h"
29#include "libavcodec/mpegaudiodata.h"
30#include "libavcodec/mpegaudiodecheader.h"
31#include "libavutil/intreadwrite.h"
32#include "libavutil/opt.h"
33#include "libavutil/dict.h"
34#include "libavutil/avassert.h"
35
36static int id3v1_set_string(AVFormatContext *s, const char *key,
37                            uint8_t *buf, int buf_size)
38{
39    AVDictionaryEntry *tag;
40    if ((tag = av_dict_get(s->metadata, key, NULL, 0)))
41        av_strlcpy(buf, tag->value, buf_size);
42    return !!tag;
43}
44
45static int id3v1_create_tag(AVFormatContext *s, uint8_t *buf)
46{
47    AVDictionaryEntry *tag;
48    int i, count = 0;
49
50    memset(buf, 0, ID3v1_TAG_SIZE); /* fail safe */
51    buf[0] = 'T';
52    buf[1] = 'A';
53    buf[2] = 'G';
54    /* we knowingly overspecify each tag length by one byte to compensate for the mandatory null byte added by av_strlcpy */
55    count += id3v1_set_string(s, "TIT2",    buf +  3, 30 + 1);       //title
56    count += id3v1_set_string(s, "TPE1",    buf + 33, 30 + 1);       //author|artist
57    count += id3v1_set_string(s, "TALB",    buf + 63, 30 + 1);       //album
58    count += id3v1_set_string(s, "TDRL",    buf + 93,  4 + 1);       //date
59    count += id3v1_set_string(s, "comment", buf + 97, 30 + 1);
60    if ((tag = av_dict_get(s->metadata, "TRCK", NULL, 0))) { //track
61        buf[125] = 0;
62        buf[126] = atoi(tag->value);
63        count++;
64    }
65    buf[127] = 0xFF; /* default to unknown genre */
66    if ((tag = av_dict_get(s->metadata, "TCON", NULL, 0))) { //genre
67        for(i = 0; i <= ID3v1_GENRE_MAX; i++) {
68            if (!av_strcasecmp(tag->value, ff_id3v1_genre_str[i])) {
69                buf[127] = i;
70                count++;
71                break;
72            }
73        }
74    }
75    return count;
76}
77
78#define XING_NUM_BAGS 400
79#define XING_TOC_SIZE 100
80// maximum size of the xing frame: offset/Xing/flags/frames/size/TOC
81#define XING_MAX_SIZE (32 + 4 + 4 + 4 + 4 + XING_TOC_SIZE)
82
83typedef struct MP3Context {
84    const AVClass *class;
85    ID3v2EncContext id3;
86    int id3v2_version;
87    int write_id3v1;
88    int write_xing;
89
90    /* xing header */
91    int64_t xing_offset;
92    int32_t frames;
93    int32_t size;
94    uint32_t want;
95    uint32_t seen;
96    uint32_t pos;
97    uint64_t bag[XING_NUM_BAGS];
98    int initial_bitrate;
99    int has_variable_bitrate;
100
101    /* index of the audio stream */
102    int audio_stream_idx;
103    /* number of attached pictures we still need to write */
104    int pics_to_write;
105
106    /* audio packets are queued here until we get all the attached pictures */
107    AVPacketList *queue, *queue_end;
108} MP3Context;
109
110static const uint8_t xing_offtbl[2][2] = {{32, 17}, {17, 9}};
111
112/*
113 * Write an empty XING header and initialize respective data.
114 */
115static int mp3_write_xing(AVFormatContext *s)
116{
117    MP3Context       *mp3 = s->priv_data;
118    AVCodecContext *codec = s->streams[mp3->audio_stream_idx]->codec;
119    int32_t          header;
120    MPADecodeHeader  mpah;
121    int srate_idx, i, channels;
122    int bitrate_idx;
123    int best_bitrate_idx = -1;
124    int best_bitrate_error = INT_MAX;
125    int xing_offset;
126    int ver = 0;
127    int bytes_needed;
128    const char *vendor = (s->flags & AVFMT_FLAG_BITEXACT) ? "Lavf" : LIBAVFORMAT_IDENT;
129
130    if (!s->pb->seekable || !mp3->write_xing)
131        return 0;
132
133    for (i = 0; i < FF_ARRAY_ELEMS(avpriv_mpa_freq_tab); i++) {
134        const uint16_t base_freq = avpriv_mpa_freq_tab[i];
135
136        if      (codec->sample_rate == base_freq)     ver = 0x3; // MPEG 1
137        else if (codec->sample_rate == base_freq / 2) ver = 0x2; // MPEG 2
138        else if (codec->sample_rate == base_freq / 4) ver = 0x0; // MPEG 2.5
139        else continue;
140
141        srate_idx = i;
142        break;
143    }
144    if (i == FF_ARRAY_ELEMS(avpriv_mpa_freq_tab)) {
145        av_log(s, AV_LOG_WARNING, "Unsupported sample rate, not writing Xing header.\n");
146        return -1;
147    }
148
149    switch (codec->channels) {
150    case 1:  channels = MPA_MONO;                                          break;
151    case 2:  channels = MPA_STEREO;                                        break;
152    default: av_log(s, AV_LOG_WARNING, "Unsupported number of channels, "
153                    "not writing Xing header.\n");
154             return -1;
155    }
156
157    /* dummy MPEG audio header */
158    header  =  0xffU                                 << 24; // sync
159    header |= (0x7 << 5 | ver << 3 | 0x1 << 1 | 0x1) << 16; // sync/audio-version/layer 3/no crc*/
160    header |= (srate_idx << 2) << 8;
161    header |= channels << 6;
162
163    for (bitrate_idx = 1; bitrate_idx < 15; bitrate_idx++) {
164        int bit_rate = 1000 * avpriv_mpa_bitrate_tab[ver != 3][3 - 1][bitrate_idx];
165        int error    = FFABS(bit_rate - codec->bit_rate);
166
167        if (error < best_bitrate_error) {
168            best_bitrate_error = error;
169            best_bitrate_idx   = bitrate_idx;
170        }
171    }
172    av_assert0(best_bitrate_idx >= 0);
173
174    for (bitrate_idx = best_bitrate_idx; ; bitrate_idx++) {
175        int32_t mask = bitrate_idx << (4 + 8);
176        if (15 == bitrate_idx)
177            return -1;
178        header |= mask;
179
180        avpriv_mpegaudio_decode_header(&mpah, header);
181        xing_offset=xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1];
182        bytes_needed = 4              // header
183               + xing_offset
184               + 4              // xing tag
185               + 4              // frames/size/toc flags
186               + 4              // frames
187               + 4              // size
188               + XING_TOC_SIZE   // toc
189               + 24
190               ;
191
192        if (bytes_needed <= mpah.frame_size)
193            break;
194
195        header &= ~mask;
196    }
197
198    avio_wb32(s->pb, header);
199
200    ffio_fill(s->pb, 0, xing_offset);
201    mp3->xing_offset = avio_tell(s->pb);
202    ffio_wfourcc(s->pb, "Xing");
203    avio_wb32(s->pb, 0x01 | 0x02 | 0x04);  // frames / size / TOC
204
205    mp3->size = mpah.frame_size;
206    mp3->want=1;
207    mp3->seen=0;
208    mp3->pos=0;
209
210    avio_wb32(s->pb, 0);  // frames
211    avio_wb32(s->pb, 0);  // size
212
213    // toc
214    for (i = 0; i < XING_TOC_SIZE; ++i)
215        avio_w8(s->pb, (uint8_t)(255 * i / XING_TOC_SIZE));
216
217    for (i = 0; i < strlen(vendor); ++i)
218        avio_w8(s->pb, vendor[i]);
219    for (; i < 21; ++i)
220        avio_w8(s->pb, 0);
221    avio_wb24(s->pb, FFMAX(codec->delay - 528 - 1, 0)<<12);
222
223    ffio_fill(s->pb, 0, mpah.frame_size - bytes_needed);
224
225    return 0;
226}
227
228/*
229 * Add a frame to XING data.
230 * Following lame's "VbrTag.c".
231 */
232static void mp3_xing_add_frame(MP3Context *mp3, AVPacket *pkt)
233{
234    int i;
235
236    mp3->frames++;
237    mp3->seen++;
238    mp3->size += pkt->size;
239
240    if (mp3->want == mp3->seen) {
241        mp3->bag[mp3->pos] = mp3->size;
242
243        if (XING_NUM_BAGS == ++mp3->pos) {
244            /* shrink table to half size by throwing away each second bag. */
245            for (i = 1; i < XING_NUM_BAGS; i += 2)
246                mp3->bag[i >> 1] = mp3->bag[i];
247
248            /* double wanted amount per bag. */
249            mp3->want *= 2;
250            /* adjust current position to half of table size. */
251            mp3->pos = XING_NUM_BAGS / 2;
252        }
253
254        mp3->seen = 0;
255    }
256}
257
258static int mp3_write_audio_packet(AVFormatContext *s, AVPacket *pkt)
259{
260    MP3Context  *mp3 = s->priv_data;
261
262    if (pkt->data && pkt->size >= 4) {
263        MPADecodeHeader mpah;
264        int av_unused base;
265        uint32_t h;
266
267        h = AV_RB32(pkt->data);
268        if (ff_mpa_check_header(h) == 0) {
269            avpriv_mpegaudio_decode_header(&mpah, h);
270            if (!mp3->initial_bitrate)
271                mp3->initial_bitrate = mpah.bit_rate;
272            if ((mpah.bit_rate == 0) || (mp3->initial_bitrate != mpah.bit_rate))
273                mp3->has_variable_bitrate = 1;
274        } else {
275            av_log(s, AV_LOG_WARNING, "Audio packet of size %d (starting with %08X...) "
276                   "is invalid, writing it anyway.\n", pkt->size, h);
277        }
278
279#ifdef FILTER_VBR_HEADERS
280        /* filter out XING and INFO headers. */
281        base = 4 + xing_offtbl[mpah.lsf == 1][mpah.nb_channels == 1];
282
283        if (base + 4 <= pkt->size) {
284            uint32_t v = AV_RB32(pkt->data + base);
285
286            if (MKBETAG('X','i','n','g') == v || MKBETAG('I','n','f','o') == v)
287                return 0;
288        }
289
290        /* filter out VBRI headers. */
291        base = 4 + 32;
292
293        if (base + 4 <= pkt->size && MKBETAG('V','B','R','I') == AV_RB32(pkt->data + base))
294            return 0;
295#endif
296
297        if (mp3->xing_offset)
298            mp3_xing_add_frame(mp3, pkt);
299    }
300
301    return ff_raw_write_packet(s, pkt);
302}
303
304static int mp3_queue_flush(AVFormatContext *s)
305{
306    MP3Context *mp3 = s->priv_data;
307    AVPacketList *pktl;
308    int ret = 0, write = 1;
309
310    ff_id3v2_finish(&mp3->id3, s->pb, s->metadata_header_padding);
311    mp3_write_xing(s);
312
313    while ((pktl = mp3->queue)) {
314        if (write && (ret = mp3_write_audio_packet(s, &pktl->pkt)) < 0)
315            write = 0;
316        av_free_packet(&pktl->pkt);
317        mp3->queue = pktl->next;
318        av_freep(&pktl);
319    }
320    mp3->queue_end = NULL;
321    return ret;
322}
323
324static void mp3_update_xing(AVFormatContext *s)
325{
326    MP3Context  *mp3 = s->priv_data;
327    int i;
328
329    /* replace "Xing" identification string with "Info" for CBR files. */
330    if (!mp3->has_variable_bitrate) {
331        avio_seek(s->pb, mp3->xing_offset, SEEK_SET);
332        ffio_wfourcc(s->pb, "Info");
333    }
334
335    avio_seek(s->pb, mp3->xing_offset + 8, SEEK_SET);
336    avio_wb32(s->pb, mp3->frames);
337    avio_wb32(s->pb, mp3->size);
338
339    avio_w8(s->pb, 0);  // first toc entry has to be zero.
340
341    for (i = 1; i < XING_TOC_SIZE; ++i) {
342        int j = i * mp3->pos / XING_TOC_SIZE;
343        int seek_point = 256LL * mp3->bag[j] / mp3->size;
344        avio_w8(s->pb, FFMIN(seek_point, 255));
345    }
346
347    avio_seek(s->pb, 0, SEEK_END);
348}
349
350static int mp3_write_trailer(struct AVFormatContext *s)
351{
352    uint8_t buf[ID3v1_TAG_SIZE];
353    MP3Context *mp3 = s->priv_data;
354
355    if (mp3->pics_to_write) {
356        av_log(s, AV_LOG_WARNING, "No packets were sent for some of the "
357               "attached pictures.\n");
358        mp3_queue_flush(s);
359    }
360
361    /* write the id3v1 tag */
362    if (mp3->write_id3v1 && id3v1_create_tag(s, buf) > 0) {
363        avio_write(s->pb, buf, ID3v1_TAG_SIZE);
364    }
365
366    if (mp3->xing_offset)
367        mp3_update_xing(s);
368
369    return 0;
370}
371
372static int query_codec(enum AVCodecID id, int std_compliance)
373{
374    const CodecMime *cm= ff_id3v2_mime_tags;
375    while(cm->id != AV_CODEC_ID_NONE) {
376        if(id == cm->id)
377            return MKTAG('A', 'P', 'I', 'C');
378        cm++;
379    }
380    return -1;
381}
382
383#if CONFIG_MP2_MUXER
384AVOutputFormat ff_mp2_muxer = {
385    .name              = "mp2",
386    .long_name         = NULL_IF_CONFIG_SMALL("MP2 (MPEG audio layer 2)"),
387    .mime_type         = "audio/mpeg",
388    .extensions        = "mp2,m2a,mpa",
389    .audio_codec       = AV_CODEC_ID_MP2,
390    .video_codec       = AV_CODEC_ID_NONE,
391    .write_packet      = ff_raw_write_packet,
392    .flags             = AVFMT_NOTIMESTAMPS,
393};
394#endif
395
396#if CONFIG_MP3_MUXER
397
398static const AVOption options[] = {
399    { "id3v2_version", "Select ID3v2 version to write. Currently 3 and 4 are supported.",
400      offsetof(MP3Context, id3v2_version), AV_OPT_TYPE_INT, {.i64 = 4}, 0, 4, AV_OPT_FLAG_ENCODING_PARAM},
401    { "write_id3v1", "Enable ID3v1 writing. ID3v1 tags are written in UTF-8 which may not be supported by most software.",
402      offsetof(MP3Context, write_id3v1), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
403    { "write_xing",  "Write the Xing header containing file duration.",
404      offsetof(MP3Context, write_xing),  AV_OPT_TYPE_INT, {.i64 = 1}, 0, 1, AV_OPT_FLAG_ENCODING_PARAM},
405    { NULL },
406};
407
408static const AVClass mp3_muxer_class = {
409    .class_name     = "MP3 muxer",
410    .item_name      = av_default_item_name,
411    .option         = options,
412    .version        = LIBAVUTIL_VERSION_INT,
413};
414
415static int mp3_write_packet(AVFormatContext *s, AVPacket *pkt)
416{
417    MP3Context *mp3 = s->priv_data;
418
419    if (pkt->stream_index == mp3->audio_stream_idx) {
420        if (mp3->pics_to_write) {
421            /* buffer audio packets until we get all the pictures */
422            AVPacketList *pktl = av_mallocz(sizeof(*pktl));
423            int ret;
424            if (!pktl)
425                return AVERROR(ENOMEM);
426
427            ret = av_copy_packet(&pktl->pkt, pkt);
428            if (ret < 0) {
429                av_freep(&pktl);
430                return ret;
431            }
432
433            if (mp3->queue_end)
434                mp3->queue_end->next = pktl;
435            else
436                mp3->queue = pktl;
437            mp3->queue_end = pktl;
438        } else
439            return mp3_write_audio_packet(s, pkt);
440    } else {
441        int ret;
442
443        /* warn only once for each stream */
444        if (s->streams[pkt->stream_index]->nb_frames == 1) {
445            av_log(s, AV_LOG_WARNING, "Got more than one picture in stream %d,"
446                   " ignoring.\n", pkt->stream_index);
447        }
448        if (!mp3->pics_to_write || s->streams[pkt->stream_index]->nb_frames >= 1)
449            return 0;
450
451        if ((ret = ff_id3v2_write_apic(s, &mp3->id3, pkt)) < 0)
452            return ret;
453        mp3->pics_to_write--;
454
455        /* flush the buffered audio packets */
456        if (!mp3->pics_to_write &&
457            (ret = mp3_queue_flush(s)) < 0)
458            return ret;
459    }
460
461    return 0;
462}
463
464/**
465 * Write an ID3v2 header at beginning of stream
466 */
467
468static int mp3_write_header(struct AVFormatContext *s)
469{
470    MP3Context  *mp3 = s->priv_data;
471    int ret, i;
472
473    if (mp3->id3v2_version      &&
474        mp3->id3v2_version != 3 &&
475        mp3->id3v2_version != 4) {
476        av_log(s, AV_LOG_ERROR, "Invalid ID3v2 version requested: %d. Only "
477               "3, 4 or 0 (disabled) are allowed.\n", mp3->id3v2_version);
478        return AVERROR(EINVAL);
479    }
480
481    /* check the streams -- we want exactly one audio and arbitrary number of
482     * video (attached pictures) */
483    mp3->audio_stream_idx = -1;
484    for (i = 0; i < s->nb_streams; i++) {
485        AVStream *st = s->streams[i];
486        if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
487            if (mp3->audio_stream_idx >= 0 || st->codec->codec_id != AV_CODEC_ID_MP3) {
488                av_log(s, AV_LOG_ERROR, "Invalid audio stream. Exactly one MP3 "
489                       "audio stream is required.\n");
490                return AVERROR(EINVAL);
491            }
492            mp3->audio_stream_idx = i;
493        } else if (st->codec->codec_type != AVMEDIA_TYPE_VIDEO) {
494            av_log(s, AV_LOG_ERROR, "Only audio streams and pictures are allowed in MP3.\n");
495            return AVERROR(EINVAL);
496        }
497    }
498    if (mp3->audio_stream_idx < 0) {
499        av_log(s, AV_LOG_ERROR, "No audio stream present.\n");
500        return AVERROR(EINVAL);
501    }
502    mp3->pics_to_write = s->nb_streams - 1;
503
504    if (mp3->pics_to_write && !mp3->id3v2_version) {
505        av_log(s, AV_LOG_ERROR, "Attached pictures were requested, but the "
506               "ID3v2 header is disabled.\n");
507        return AVERROR(EINVAL);
508    }
509
510    if (mp3->id3v2_version) {
511        ff_id3v2_start(&mp3->id3, s->pb, mp3->id3v2_version, ID3v2_DEFAULT_MAGIC);
512        ret = ff_id3v2_write_metadata(s, &mp3->id3);
513        if (ret < 0)
514            return ret;
515    }
516
517    if (!mp3->pics_to_write) {
518        if (mp3->id3v2_version)
519            ff_id3v2_finish(&mp3->id3, s->pb, s->metadata_header_padding);
520        mp3_write_xing(s);
521    }
522
523    return 0;
524}
525
526AVOutputFormat ff_mp3_muxer = {
527    .name              = "mp3",
528    .long_name         = NULL_IF_CONFIG_SMALL("MP3 (MPEG audio layer 3)"),
529    .mime_type         = "audio/mpeg",
530    .extensions        = "mp3",
531    .priv_data_size    = sizeof(MP3Context),
532    .audio_codec       = AV_CODEC_ID_MP3,
533    .video_codec       = AV_CODEC_ID_PNG,
534    .write_header      = mp3_write_header,
535    .write_packet      = mp3_write_packet,
536    .write_trailer     = mp3_write_trailer,
537    .query_codec       = query_codec,
538    .flags             = AVFMT_NOTIMESTAMPS,
539    .priv_class        = &mp3_muxer_class,
540};
541#endif
542