MP3Extractor.cpp revision 2944eca607304a095ea43ba2b8f0b9de61249f9f 
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            LOGV("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            LOGV("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        LOGV("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            LOGV("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            LOGV("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            LOGV("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    sp<MetaData> mMeta;
232    sp<DataSource> mDataSource;
233    off64_t mFirstFramePos;
234    uint32_t mFixedHeader;
235    off64_t mCurrentPos;
236    int64_t mCurrentTimeUs;
237    bool mStarted;
238    sp<MP3Seeker> mSeeker;
239    MediaBufferGroup *mGroup;
240
241    int64_t mBasisTimeUs;
242    int64_t mSamplesRead;
243
244    MP3Source(const MP3Source &);
245    MP3Source &operator=(const MP3Source &);
246};
247
248MP3Extractor::MP3Extractor(
249        const sp<DataSource> &source, const sp<AMessage> &meta)
250    : mInitCheck(NO_INIT),
251      mDataSource(source),
252      mFirstFramePos(-1),
253      mFixedHeader(0) {
254    off64_t pos = 0;
255    off64_t post_id3_pos;
256    uint32_t header;
257    bool success;
258
259    int64_t meta_offset;
260    uint32_t meta_header;
261    int64_t meta_post_id3_offset;
262    if (meta != NULL
263            && meta->findInt64("offset", &meta_offset)
264            && meta->findInt32("header", (int32_t *)&meta_header)
265            && meta->findInt64("post-id3-offset", &meta_post_id3_offset)) {
266        // The sniffer has already done all the hard work for us, simply
267        // accept its judgement.
268        pos = (off64_t)meta_offset;
269        header = meta_header;
270        post_id3_pos = (off64_t)meta_post_id3_offset;
271
272        success = true;
273    } else {
274        success = Resync(mDataSource, 0, &pos, &post_id3_pos, &header);
275    }
276
277    if (!success) {
278        // mInitCheck will remain NO_INIT
279        return;
280    }
281
282    mFirstFramePos = pos;
283    mFixedHeader = header;
284
285    size_t frame_size;
286    int sample_rate;
287    int num_channels;
288    int bitrate;
289    GetMPEGAudioFrameSize(
290            header, &frame_size, &sample_rate, &num_channels, &bitrate);
291
292    unsigned layer = 4 - ((header >> 17) & 3);
293
294    mMeta = new MetaData;
295
296    switch (layer) {
297        case 1:
298            mMeta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG_LAYER_I);
299            break;
300        case 2:
301            mMeta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG_LAYER_II);
302            break;
303        case 3:
304            mMeta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MPEG);
305            break;
306        default:
307            TRESPASS();
308    }
309
310    mMeta->setInt32(kKeySampleRate, sample_rate);
311    mMeta->setInt32(kKeyBitRate, bitrate * 1000);
312    mMeta->setInt32(kKeyChannelCount, num_channels);
313
314    mSeeker = XINGSeeker::CreateFromSource(mDataSource, mFirstFramePos);
315
316    if (mSeeker == NULL) {
317        mSeeker = VBRISeeker::CreateFromSource(mDataSource, post_id3_pos);
318    }
319
320    int64_t durationUs;
321
322    if (mSeeker == NULL || !mSeeker->getDuration(&durationUs)) {
323        off64_t fileSize;
324        if (mDataSource->getSize(&fileSize) == OK) {
325            durationUs = 8000LL * (fileSize - mFirstFramePos) / bitrate;
326        } else {
327            durationUs = -1;
328        }
329    }
330
331    if (durationUs >= 0) {
332        mMeta->setInt64(kKeyDuration, durationUs);
333    }
334
335    mInitCheck = OK;
336}
337
338size_t MP3Extractor::countTracks() {
339    return mInitCheck != OK ? 0 : 1;
340}
341
342sp<MediaSource> MP3Extractor::getTrack(size_t index) {
343    if (mInitCheck != OK || index != 0) {
344        return NULL;
345    }
346
347    return new MP3Source(
348            mMeta, mDataSource, mFirstFramePos, mFixedHeader,
349            mSeeker);
350}
351
352sp<MetaData> MP3Extractor::getTrackMetaData(size_t index, uint32_t flags) {
353    if (mInitCheck != OK || index != 0) {
354        return NULL;
355    }
356
357    return mMeta;
358}
359
360////////////////////////////////////////////////////////////////////////////////
361
362MP3Source::MP3Source(
363        const sp<MetaData> &meta, const sp<DataSource> &source,
364        off64_t first_frame_pos, uint32_t fixed_header,
365        const sp<MP3Seeker> &seeker)
366    : mMeta(meta),
367      mDataSource(source),
368      mFirstFramePos(first_frame_pos),
369      mFixedHeader(fixed_header),
370      mCurrentPos(0),
371      mCurrentTimeUs(0),
372      mStarted(false),
373      mSeeker(seeker),
374      mGroup(NULL),
375      mBasisTimeUs(0),
376      mSamplesRead(0) {
377}
378
379MP3Source::~MP3Source() {
380    if (mStarted) {
381        stop();
382    }
383}
384
385status_t MP3Source::start(MetaData *) {
386    CHECK(!mStarted);
387
388    mGroup = new MediaBufferGroup;
389
390    const size_t kMaxFrameSize = 32768;
391    mGroup->add_buffer(new MediaBuffer(kMaxFrameSize));
392
393    mCurrentPos = mFirstFramePos;
394    mCurrentTimeUs = 0;
395
396    mBasisTimeUs = mCurrentTimeUs;
397    mSamplesRead = 0;
398
399    mStarted = true;
400
401    return OK;
402}
403
404status_t MP3Source::stop() {
405    CHECK(mStarted);
406
407    delete mGroup;
408    mGroup = NULL;
409
410    mStarted = false;
411
412    return OK;
413}
414
415sp<MetaData> MP3Source::getFormat() {
416    return mMeta;
417}
418
419status_t MP3Source::read(
420        MediaBuffer **out, const ReadOptions *options) {
421    *out = NULL;
422
423    int64_t seekTimeUs;
424    ReadOptions::SeekMode mode;
425    bool seekCBR = false;
426
427    if (options != NULL && options->getSeekTo(&seekTimeUs, &mode)) {
428        int64_t actualSeekTimeUs = seekTimeUs;
429        if (mSeeker == NULL
430                || !mSeeker->getOffsetForTime(&actualSeekTimeUs, &mCurrentPos)) {
431            int32_t bitrate;
432            if (!mMeta->findInt32(kKeyBitRate, &bitrate)) {
433                // bitrate is in bits/sec.
434                LOGI("no bitrate");
435
436                return ERROR_UNSUPPORTED;
437            }
438
439            mCurrentTimeUs = seekTimeUs;
440            mCurrentPos = mFirstFramePos + seekTimeUs * bitrate / 8000000;
441            seekCBR = true;
442        } else {
443            mCurrentTimeUs = actualSeekTimeUs;
444        }
445
446        mBasisTimeUs = mCurrentTimeUs;
447        mSamplesRead = 0;
448    }
449
450    MediaBuffer *buffer;
451    status_t err = mGroup->acquire_buffer(&buffer);
452    if (err != OK) {
453        return err;
454    }
455
456    size_t frame_size;
457    int bitrate;
458    int num_samples;
459    int sample_rate;
460    for (;;) {
461        ssize_t n = mDataSource->readAt(mCurrentPos, buffer->data(), 4);
462        if (n < 4) {
463            buffer->release();
464            buffer = NULL;
465
466            return ERROR_END_OF_STREAM;
467        }
468
469        uint32_t header = U32_AT((const uint8_t *)buffer->data());
470
471        if ((header & kMask) == (mFixedHeader & kMask)
472            && GetMPEGAudioFrameSize(
473                header, &frame_size, &sample_rate, NULL,
474                &bitrate, &num_samples)) {
475
476            // re-calculate mCurrentTimeUs because we might have called Resync()
477            if (seekCBR) {
478                mCurrentTimeUs = (mCurrentPos - mFirstFramePos) * 8000 / bitrate;
479                mBasisTimeUs = mCurrentTimeUs;
480            }
481
482            break;
483        }
484
485        // Lost sync.
486        LOGV("lost sync! header = 0x%08x, old header = 0x%08x\n", header, mFixedHeader);
487
488        off64_t pos = mCurrentPos;
489        if (!Resync(mDataSource, mFixedHeader, &pos, NULL, NULL)) {
490            LOGE("Unable to resync. Signalling end of stream.");
491
492            buffer->release();
493            buffer = NULL;
494
495            return ERROR_END_OF_STREAM;
496        }
497
498        mCurrentPos = pos;
499
500        // Try again with the new position.
501    }
502
503    CHECK(frame_size <= buffer->size());
504
505    ssize_t n = mDataSource->readAt(mCurrentPos, buffer->data(), frame_size);
506    if (n < (ssize_t)frame_size) {
507        buffer->release();
508        buffer = NULL;
509
510        return ERROR_END_OF_STREAM;
511    }
512
513    buffer->set_range(0, frame_size);
514
515    buffer->meta_data()->setInt64(kKeyTime, mCurrentTimeUs);
516    buffer->meta_data()->setInt32(kKeyIsSyncFrame, 1);
517
518    mCurrentPos += frame_size;
519
520    mSamplesRead += num_samples;
521    mCurrentTimeUs = mBasisTimeUs + ((mSamplesRead * 1000000) / sample_rate);
522
523    *out = buffer;
524
525    return OK;
526}
527
528sp<MetaData> MP3Extractor::getMetaData() {
529    sp<MetaData> meta = new MetaData;
530
531    if (mInitCheck != OK) {
532        return meta;
533    }
534
535    meta->setCString(kKeyMIMEType, "audio/mpeg");
536
537    ID3 id3(mDataSource);
538
539    if (!id3.isValid()) {
540        return meta;
541    }
542
543    struct Map {
544        int key;
545        const char *tag1;
546        const char *tag2;
547    };
548    static const Map kMap[] = {
549        { kKeyAlbum, "TALB", "TAL" },
550        { kKeyArtist, "TPE1", "TP1" },
551        { kKeyAlbumArtist, "TPE2", "TP2" },
552        { kKeyComposer, "TCOM", "TCM" },
553        { kKeyGenre, "TCON", "TCO" },
554        { kKeyTitle, "TIT2", "TT2" },
555        { kKeyYear, "TYE", "TYER" },
556        { kKeyAuthor, "TXT", "TEXT" },
557        { kKeyCDTrackNumber, "TRK", "TRCK" },
558        { kKeyDiscNumber, "TPA", "TPOS" },
559        { kKeyCompilation, "TCP", "TCMP" },
560    };
561    static const size_t kNumMapEntries = sizeof(kMap) / sizeof(kMap[0]);
562
563    for (size_t i = 0; i < kNumMapEntries; ++i) {
564        ID3::Iterator *it = new ID3::Iterator(id3, kMap[i].tag1);
565        if (it->done()) {
566            delete it;
567            it = new ID3::Iterator(id3, kMap[i].tag2);
568        }
569
570        if (it->done()) {
571            delete it;
572            continue;
573        }
574
575        String8 s;
576        it->getString(&s);
577        delete it;
578
579        meta->setCString(kMap[i].key, s);
580    }
581
582    size_t dataSize;
583    String8 mime;
584    const void *data = id3.getAlbumArt(&dataSize, &mime);
585
586    if (data) {
587        meta->setData(kKeyAlbumArt, MetaData::TYPE_NONE, data, dataSize);
588        meta->setCString(kKeyAlbumArtMIME, mime.string());
589    }
590
591    return meta;
592}
593
594bool SniffMP3(
595        const sp<DataSource> &source, String8 *mimeType,
596        float *confidence, sp<AMessage> *meta) {
597    off64_t pos = 0;
598    off64_t post_id3_pos;
599    uint32_t header;
600    if (!Resync(source, 0, &pos, &post_id3_pos, &header)) {
601        return false;
602    }
603
604    *meta = new AMessage;
605    (*meta)->setInt64("offset", pos);
606    (*meta)->setInt32("header", header);
607    (*meta)->setInt64("post-id3-offset", post_id3_pos);
608
609    *mimeType = MEDIA_MIMETYPE_AUDIO_MPEG;
610    *confidence = 0.2f;
611
612    return true;
613}
614
615}  // namespace android
616