1/*
2 * Copyright (C) 2009 The Android Open Source Project
3 *
4 * Licensed under the Apache License, Version 2.0 (the "License");
5 * you may not use this file except in compliance with the License.
6 * You may obtain a copy of the License at
7 *
8 *      http://www.apache.org/licenses/LICENSE-2.0
9 *
10 * Unless required by applicable law or agreed to in writing, software
11 * distributed under the License is distributed on an "AS IS" BASIS,
12 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
13 * See the License for the specific language governing permissions and
14 * limitations under the License.
15 */
16
17//#define LOG_NDEBUG 0
18#define LOG_TAG "MP3Extractor"
19#include <utils/Log.h>
20
21#include "include/MP3Extractor.h"
22
23#include "include/avc_utils.h"
24#include "include/ID3.h"
25#include "include/VBRISeeker.h"
26#include "include/XINGSeeker.h"
27
28#include <media/stagefright/foundation/ADebug.h>
29#include <media/stagefright/foundation/AMessage.h>
30#include <media/stagefright/DataSource.h>
31#include <media/stagefright/MediaBuffer.h>
32#include <media/stagefright/MediaBufferGroup.h>
33#include <media/stagefright/MediaDefs.h>
34#include <media/stagefright/MediaErrors.h>
35#include <media/stagefright/MediaSource.h>
36#include <media/stagefright/MetaData.h>
37#include <media/stagefright/Utils.h>
38#include <utils/String8.h>
39
40namespace android {
41
42// Everything must match except for
43// protection, bitrate, padding, private bits, mode, mode extension,
44// copyright bit, original bit and emphasis.
45// Yes ... there are things that must indeed match...
46static const uint32_t kMask = 0xfffe0c00;
47
48static bool Resync(
49        const sp<DataSource> &source, uint32_t match_header,
50        off64_t *inout_pos, off64_t *post_id3_pos, uint32_t *out_header) {
51    if (post_id3_pos != NULL) {
52        *post_id3_pos = 0;
53    }
54
55    if (*inout_pos == 0) {
56        // Skip an optional ID3 header if syncing at the very beginning
57        // of the datasource.
58
59        for (;;) {
60            uint8_t id3header[10];
61            if (source->readAt(*inout_pos, id3header, sizeof(id3header))
62                    < (ssize_t)sizeof(id3header)) {
63                // If we can't even read these 10 bytes, we might as well bail
64                // out, even if there _were_ 10 bytes of valid mp3 audio data...
65                return false;
66            }
67
68            if (memcmp("ID3", id3header, 3)) {
69                break;
70            }
71
72            // Skip the ID3v2 header.
73
74            size_t len =
75                ((id3header[6] & 0x7f) << 21)
76                | ((id3header[7] & 0x7f) << 14)
77                | ((id3header[8] & 0x7f) << 7)
78                | (id3header[9] & 0x7f);
79
80            len += 10;
81
82            *inout_pos += len;
83
84            ALOGV("skipped ID3 tag, new starting offset is %lld (0x%016llx)",
85                 *inout_pos, *inout_pos);
86        }
87
88        if (post_id3_pos != NULL) {
89            *post_id3_pos = *inout_pos;
90        }
91    }
92
93    off64_t pos = *inout_pos;
94    bool valid = false;
95
96    const size_t kMaxReadBytes = 1024;
97    const size_t kMaxBytesChecked = 128 * 1024;
98    uint8_t buf[kMaxReadBytes];
99    ssize_t bytesToRead = kMaxReadBytes;
100    ssize_t totalBytesRead = 0;
101    ssize_t remainingBytes = 0;
102    bool reachEOS = false;
103    uint8_t *tmp = buf;
104
105    do {
106        if (pos >= *inout_pos + kMaxBytesChecked) {
107            // Don't scan forever.
108            ALOGV("giving up at offset %lld", pos);
109            break;
110        }
111
112        if (remainingBytes < 4) {
113            if (reachEOS) {
114                break;
115            } else {
116                memcpy(buf, tmp, remainingBytes);
117                bytesToRead = kMaxReadBytes - remainingBytes;
118
119                /*
120                 * The next read position should start from the end of
121                 * the last buffer, and thus should include the remaining
122                 * bytes in the buffer.
123                 */
124                totalBytesRead = source->readAt(pos + remainingBytes,
125                                                buf + remainingBytes,
126                                                bytesToRead);
127                if (totalBytesRead <= 0) {
128                    break;
129                }
130                reachEOS = (totalBytesRead != bytesToRead);
131                totalBytesRead += remainingBytes;
132                remainingBytes = totalBytesRead;
133                tmp = buf;
134                continue;
135            }
136        }
137
138        uint32_t header = U32_AT(tmp);
139
140        if (match_header != 0 && (header & kMask) != (match_header & kMask)) {
141            ++pos;
142            ++tmp;
143            --remainingBytes;
144            continue;
145        }
146
147        size_t frame_size;
148        int sample_rate, num_channels, bitrate;
149        if (!GetMPEGAudioFrameSize(
150                    header, &frame_size,
151                    &sample_rate, &num_channels, &bitrate)) {
152            ++pos;
153            ++tmp;
154            --remainingBytes;
155            continue;
156        }
157
158        ALOGV("found possible 1st frame at %lld (header = 0x%08x)", pos, header);
159
160        // We found what looks like a valid frame,
161        // now find its successors.
162
163        off64_t test_pos = pos + frame_size;
164
165        valid = true;
166        for (int j = 0; j < 3; ++j) {
167            uint8_t tmp[4];
168            if (source->readAt(test_pos, tmp, 4) < 4) {
169                valid = false;
170                break;
171            }
172
173            uint32_t test_header = U32_AT(tmp);
174
175            ALOGV("subsequent header is %08x", test_header);
176
177            if ((test_header & kMask) != (header & kMask)) {
178                valid = false;
179                break;
180            }
181
182            size_t test_frame_size;
183            if (!GetMPEGAudioFrameSize(
184                        test_header, &test_frame_size)) {
185                valid = false;
186                break;
187            }
188
189            ALOGV("found subsequent frame #%d at %lld", j + 2, test_pos);
190
191            test_pos += test_frame_size;
192        }
193
194        if (valid) {
195            *inout_pos = pos;
196
197            if (out_header != NULL) {
198                *out_header = header;
199            }
200        } else {
201            ALOGV("no dice, no valid sequence of frames found.");
202        }
203
204        ++pos;
205        ++tmp;
206        --remainingBytes;
207    } while (!valid);
208
209    return valid;
210}
211
212class MP3Source : public MediaSource {
213public:
214    MP3Source(
215            const sp<MetaData> &meta, const sp<DataSource> &source,
216            off64_t first_frame_pos, uint32_t fixed_header,
217            const sp<MP3Seeker> &seeker);
218
219    virtual status_t start(MetaData *params = NULL);
220    virtual status_t stop();
221
222    virtual sp<MetaData> getFormat();
223
224    virtual status_t read(
225            MediaBuffer **buffer, const ReadOptions *options = NULL);
226
227protected:
228    virtual ~MP3Source();
229
230private:
231    static const size_t kMaxFrameSize;
232    sp<MetaData> mMeta;
233    sp<DataSource> mDataSource;
234    off64_t mFirstFramePos;
235    uint32_t mFixedHeader;
236    off64_t mCurrentPos;
237    int64_t mCurrentTimeUs;
238    bool mStarted;
239    sp<MP3Seeker> mSeeker;
240    MediaBufferGroup *mGroup;
241
242    int64_t mBasisTimeUs;
243    int64_t mSamplesRead;
244
245    MP3Source(const MP3Source &);
246    MP3Source &operator=(const MP3Source &);
247};
248
249MP3Extractor::MP3Extractor(
250        const sp<DataSource> &source, const sp<AMessage> &meta)
251    : mInitCheck(NO_INIT),
252      mDataSource(source),
253      mFirstFramePos(-1),
254      mFixedHeader(0) {
255    off64_t pos = 0;
256    off64_t post_id3_pos;
257    uint32_t header;
258    bool success;
259
260    int64_t meta_offset;
261    uint32_t meta_header;
262    int64_t meta_post_id3_offset;
263    if (meta != NULL
264            && meta->findInt64("offset", &meta_offset)
265            && meta->findInt32("header", (int32_t *)&meta_header)
266            && meta->findInt64("post-id3-offset", &meta_post_id3_offset)) {
267        // The sniffer has already done all the hard work for us, simply
268        // accept its judgement.
269        pos = (off64_t)meta_offset;
270        header = meta_header;
271        post_id3_pos = (off64_t)meta_post_id3_offset;
272
273        success = true;
274    } else {
275        success = Resync(mDataSource, 0, &pos, &post_id3_pos, &header);
276    }
277
278    if (!success) {
279        // mInitCheck will remain NO_INIT
280        return;
281    }
282
283    mFirstFramePos = pos;
284    mFixedHeader = header;
285
286    size_t frame_size;
287    int sample_rate;
288    int num_channels;
289    int bitrate;
290    GetMPEGAudioFrameSize(
291            header, &frame_size, &sample_rate, &num_channels, &bitrate);
292
293    unsigned layer = 4 - ((header >> 17) & 3);
294
295    mMeta = new MetaData;
296
297    switch (layer) {
298        case 1:
299            mMeta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG_LAYER_I);
300            break;
301        case 2:
302            mMeta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG_LAYER_II);
303            break;
304        case 3:
305            mMeta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG);
306            break;
307        default:
308            TRESPASS();
309    }
310
311    mMeta->setInt32(kKeySampleRate, sample_rate);
312    mMeta->setInt32(kKeyBitRate, bitrate * 1000);
313    mMeta->setInt32(kKeyChannelCount, num_channels);
314
315    sp<XINGSeeker> seeker = XINGSeeker::CreateFromSource(mDataSource, mFirstFramePos);
316
317    if (seeker == NULL) {
318        mSeeker = VBRISeeker::CreateFromSource(mDataSource, post_id3_pos);
319    } else {
320        mSeeker = seeker;
321        int encd = seeker->getEncoderDelay();
322        int encp = seeker->getEncoderPadding();
323        if (encd != 0 || encp != 0) {
324            mMeta->setInt32(kKeyEncoderDelay, encd);
325            mMeta->setInt32(kKeyEncoderPadding, encp);
326        }
327    }
328
329    if (mSeeker != NULL) {
330        // While it is safe to send the XING/VBRI frame to the decoder, this will
331        // result in an extra 1152 samples being output. The real first frame to
332        // decode is after the XING/VBRI frame, so skip there.
333        mFirstFramePos += frame_size;
334    }
335
336    int64_t durationUs;
337
338    if (mSeeker == NULL || !mSeeker->getDuration(&durationUs)) {
339        off64_t fileSize;
340        if (mDataSource->getSize(&fileSize) == OK) {
341            durationUs = 8000LL * (fileSize - mFirstFramePos) / bitrate;
342        } else {
343            durationUs = -1;
344        }
345    }
346
347    if (durationUs >= 0) {
348        mMeta->setInt64(kKeyDuration, durationUs);
349    }
350
351    mInitCheck = OK;
352
353    // get iTunes-style gapless info if present
354    ID3 id3(mDataSource);
355    if (id3.isValid()) {
356        ID3::Iterator *com = new ID3::Iterator(id3, "COM");
357        if (com->done()) {
358            delete com;
359            com = new ID3::Iterator(id3, "COMM");
360        }
361        while(!com->done()) {
362            String8 commentdesc;
363            String8 commentvalue;
364            com->getString(&commentdesc, &commentvalue);
365            const char * desc = commentdesc.string();
366            const char * value = commentvalue.string();
367
368            // first 3 characters are the language, which we don't care about
369            if(strlen(desc) > 3 && strcmp(desc + 3, "iTunSMPB") == 0) {
370
371                int32_t delay, padding;
372                if (sscanf(value, " %*x %x %x %*x", &delay, &padding) == 2) {
373                    mMeta->setInt32(kKeyEncoderDelay, delay);
374                    mMeta->setInt32(kKeyEncoderPadding, padding);
375                }
376                break;
377            }
378            com->next();
379        }
380        delete com;
381        com = NULL;
382    }
383}
384
385size_t MP3Extractor::countTracks() {
386    return mInitCheck != OK ? 0 : 1;
387}
388
389sp<MediaSource> MP3Extractor::getTrack(size_t index) {
390    if (mInitCheck != OK || index != 0) {
391        return NULL;
392    }
393
394    return new MP3Source(
395            mMeta, mDataSource, mFirstFramePos, mFixedHeader,
396            mSeeker);
397}
398
399sp<MetaData> MP3Extractor::getTrackMetaData(size_t index, uint32_t flags) {
400    if (mInitCheck != OK || index != 0) {
401        return NULL;
402    }
403
404    return mMeta;
405}
406
407////////////////////////////////////////////////////////////////////////////////
408
409// The theoretical maximum frame size for an MPEG audio stream should occur
410// while playing a Layer 2, MPEGv2.5 audio stream at 160kbps (with padding).
411// The size of this frame should be...
412// ((1152 samples/frame * 160000 bits/sec) /
413//  (8000 samples/sec * 8 bits/byte)) + 1 padding byte/frame = 2881 bytes/frame.
414// Set our max frame size to the nearest power of 2 above this size (aka, 4kB)
415const size_t MP3Source::kMaxFrameSize = (1 << 12); /* 4096 bytes */
416MP3Source::MP3Source(
417        const sp<MetaData> &meta, const sp<DataSource> &source,
418        off64_t first_frame_pos, uint32_t fixed_header,
419        const sp<MP3Seeker> &seeker)
420    : mMeta(meta),
421      mDataSource(source),
422      mFirstFramePos(first_frame_pos),
423      mFixedHeader(fixed_header),
424      mCurrentPos(0),
425      mCurrentTimeUs(0),
426      mStarted(false),
427      mSeeker(seeker),
428      mGroup(NULL),
429      mBasisTimeUs(0),
430      mSamplesRead(0) {
431}
432
433MP3Source::~MP3Source() {
434    if (mStarted) {
435        stop();
436    }
437}
438
439status_t MP3Source::start(MetaData *) {
440    CHECK(!mStarted);
441
442    mGroup = new MediaBufferGroup;
443
444    mGroup->add_buffer(new MediaBuffer(kMaxFrameSize));
445
446    mCurrentPos = mFirstFramePos;
447    mCurrentTimeUs = 0;
448
449    mBasisTimeUs = mCurrentTimeUs;
450    mSamplesRead = 0;
451
452    mStarted = true;
453
454    return OK;
455}
456
457status_t MP3Source::stop() {
458    CHECK(mStarted);
459
460    delete mGroup;
461    mGroup = NULL;
462
463    mStarted = false;
464
465    return OK;
466}
467
468sp<MetaData> MP3Source::getFormat() {
469    return mMeta;
470}
471
472status_t MP3Source::read(
473        MediaBuffer **out, const ReadOptions *options) {
474    *out = NULL;
475
476    int64_t seekTimeUs;
477    ReadOptions::SeekMode mode;
478    bool seekCBR = false;
479
480    if (options != NULL && options->getSeekTo(&seekTimeUs, &mode)) {
481        int64_t actualSeekTimeUs = seekTimeUs;
482        if (mSeeker == NULL
483                || !mSeeker->getOffsetForTime(&actualSeekTimeUs, &mCurrentPos)) {
484            int32_t bitrate;
485            if (!mMeta->findInt32(kKeyBitRate, &bitrate)) {
486                // bitrate is in bits/sec.
487                ALOGI("no bitrate");
488
489                return ERROR_UNSUPPORTED;
490            }
491
492            mCurrentTimeUs = seekTimeUs;
493            mCurrentPos = mFirstFramePos + seekTimeUs * bitrate / 8000000;
494            seekCBR = true;
495        } else {
496            mCurrentTimeUs = actualSeekTimeUs;
497        }
498
499        mBasisTimeUs = mCurrentTimeUs;
500        mSamplesRead = 0;
501    }
502
503    MediaBuffer *buffer;
504    status_t err = mGroup->acquire_buffer(&buffer);
505    if (err != OK) {
506        return err;
507    }
508
509    size_t frame_size;
510    int bitrate;
511    int num_samples;
512    int sample_rate;
513    for (;;) {
514        ssize_t n = mDataSource->readAt(mCurrentPos, buffer->data(), 4);
515        if (n < 4) {
516            buffer->release();
517            buffer = NULL;
518
519            return ERROR_END_OF_STREAM;
520        }
521
522        uint32_t header = U32_AT((const uint8_t *)buffer->data());
523
524        if ((header & kMask) == (mFixedHeader & kMask)
525            && GetMPEGAudioFrameSize(
526                header, &frame_size, &sample_rate, NULL,
527                &bitrate, &num_samples)) {
528
529            // re-calculate mCurrentTimeUs because we might have called Resync()
530            if (seekCBR) {
531                mCurrentTimeUs = (mCurrentPos - mFirstFramePos) * 8000 / bitrate;
532                mBasisTimeUs = mCurrentTimeUs;
533            }
534
535            break;
536        }
537
538        // Lost sync.
539        ALOGV("lost sync! header = 0x%08x, old header = 0x%08x\n", header, mFixedHeader);
540
541        off64_t pos = mCurrentPos;
542        if (!Resync(mDataSource, mFixedHeader, &pos, NULL, NULL)) {
543            ALOGE("Unable to resync. Signalling end of stream.");
544
545            buffer->release();
546            buffer = NULL;
547
548            return ERROR_END_OF_STREAM;
549        }
550
551        mCurrentPos = pos;
552
553        // Try again with the new position.
554    }
555
556    CHECK(frame_size <= buffer->size());
557
558    ssize_t n = mDataSource->readAt(mCurrentPos, buffer->data(), frame_size);
559    if (n < (ssize_t)frame_size) {
560        buffer->release();
561        buffer = NULL;
562
563        return ERROR_END_OF_STREAM;
564    }
565
566    buffer->set_range(0, frame_size);
567
568    buffer->meta_data()->setInt64(kKeyTime, mCurrentTimeUs);
569    buffer->meta_data()->setInt32(kKeyIsSyncFrame, 1);
570
571    mCurrentPos += frame_size;
572
573    mSamplesRead += num_samples;
574    mCurrentTimeUs = mBasisTimeUs + ((mSamplesRead * 1000000) / sample_rate);
575
576    *out = buffer;
577
578    return OK;
579}
580
581sp<MetaData> MP3Extractor::getMetaData() {
582    sp<MetaData> meta = new MetaData;
583
584    if (mInitCheck != OK) {
585        return meta;
586    }
587
588    meta->setCString(kKeyMIMEType, "audio/mpeg");
589
590    ID3 id3(mDataSource);
591
592    if (!id3.isValid()) {
593        return meta;
594    }
595
596    struct Map {
597        int key;
598        const char *tag1;
599        const char *tag2;
600    };
601    static const Map kMap[] = {
602        { kKeyAlbum, "TALB", "TAL" },
603        { kKeyArtist, "TPE1", "TP1" },
604        { kKeyAlbumArtist, "TPE2", "TP2" },
605        { kKeyComposer, "TCOM", "TCM" },
606        { kKeyGenre, "TCON", "TCO" },
607        { kKeyTitle, "TIT2", "TT2" },
608        { kKeyYear, "TYE", "TYER" },
609        { kKeyAuthor, "TXT", "TEXT" },
610        { kKeyCDTrackNumber, "TRK", "TRCK" },
611        { kKeyDiscNumber, "TPA", "TPOS" },
612        { kKeyCompilation, "TCP", "TCMP" },
613    };
614    static const size_t kNumMapEntries = sizeof(kMap) / sizeof(kMap[0]);
615
616    for (size_t i = 0; i < kNumMapEntries; ++i) {
617        ID3::Iterator *it = new ID3::Iterator(id3, kMap[i].tag1);
618        if (it->done()) {
619            delete it;
620            it = new ID3::Iterator(id3, kMap[i].tag2);
621        }
622
623        if (it->done()) {
624            delete it;
625            continue;
626        }
627
628        String8 s;
629        it->getString(&s);
630        delete it;
631
632        meta->setCString(kMap[i].key, s);
633    }
634
635    size_t dataSize;
636    String8 mime;
637    const void *data = id3.getAlbumArt(&dataSize, &mime);
638
639    if (data) {
640        meta->setData(kKeyAlbumArt, MetaData::TYPE_NONE, data, dataSize);
641        meta->setCString(kKeyAlbumArtMIME, mime.string());
642    }
643
644    return meta;
645}
646
647bool SniffMP3(
648        const sp<DataSource> &source, String8 *mimeType,
649        float *confidence, sp<AMessage> *meta) {
650    off64_t pos = 0;
651    off64_t post_id3_pos;
652    uint32_t header;
653    if (!Resync(source, 0, &pos, &post_id3_pos, &header)) {
654        return false;
655    }
656
657    *meta = new AMessage;
658    (*meta)->setInt64("offset", pos);
659    (*meta)->setInt32("header", header);
660    (*meta)->setInt64("post-id3-offset", post_id3_pos);
661
662    *mimeType = MEDIA_MIMETYPE_AUDIO_MPEG;
663    *confidence = 0.2f;
664
665    return true;
666}
667
668}  // namespace android
669