WAVExtractor.cpp revision e52267adab4c89169aafd10a0fa2656ccb02c039 
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 "WAVExtractor"
19#include <utils/Log.h>
20
21#include "include/WAVExtractor.h"
22
23#include <media/stagefright/foundation/ADebug.h>
24#include <media/stagefright/DataSource.h>
25#include <media/stagefright/MediaBufferGroup.h>
26#include <media/stagefright/MediaDefs.h>
27#include <media/stagefright/MediaErrors.h>
28#include <media/stagefright/MediaSource.h>
29#include <media/stagefright/MetaData.h>
30#include <utils/String8.h>
31#include <cutils/bitops.h>
32
33#define CHANNEL_MASK_USE_CHANNEL_ORDER 0
34
35namespace android {
36
37enum {
38    WAVE_FORMAT_PCM        = 0x0001,
39    WAVE_FORMAT_ALAW       = 0x0006,
40    WAVE_FORMAT_MULAW      = 0x0007,
41    WAVE_FORMAT_MSGSM      = 0x0031,
42    WAVE_FORMAT_EXTENSIBLE = 0xFFFE
43};
44
45static const char* WAVEEXT_SUBFORMAT = "\x00\x00\x00\x00\x10\x00\x80\x00\x00\xAA\x00\x38\x9B\x71";
46
47
48static uint32_t U32_LE_AT(const uint8_t *ptr) {
49    return ptr[3] << 24 | ptr[2] << 16 | ptr[1] << 8 | ptr[0];
50}
51
52static uint16_t U16_LE_AT(const uint8_t *ptr) {
53    return ptr[1] << 8 | ptr[0];
54}
55
56struct WAVSource : public MediaSource {
57    WAVSource(
58            const sp<DataSource> &dataSource,
59            const sp<MetaData> &meta,
60            uint16_t waveFormat,
61            int32_t bitsPerSample,
62            off64_t offset, size_t size);
63
64    virtual status_t start(MetaData *params = NULL);
65    virtual status_t stop();
66    virtual sp<MetaData> getFormat();
67
68    virtual status_t read(
69            MediaBuffer **buffer, const ReadOptions *options = NULL);
70
71protected:
72    virtual ~WAVSource();
73
74private:
75    static const size_t kMaxFrameSize;
76
77    sp<DataSource> mDataSource;
78    sp<MetaData> mMeta;
79    uint16_t mWaveFormat;
80    int32_t mSampleRate;
81    int32_t mNumChannels;
82    int32_t mBitsPerSample;
83    off64_t mOffset;
84    size_t mSize;
85    bool mStarted;
86    MediaBufferGroup *mGroup;
87    off64_t mCurrentPos;
88
89    WAVSource(const WAVSource &);
90    WAVSource &operator=(const WAVSource &);
91};
92
93WAVExtractor::WAVExtractor(const sp<DataSource> &source)
94    : mDataSource(source),
95      mValidFormat(false),
96      mChannelMask(CHANNEL_MASK_USE_CHANNEL_ORDER) {
97    mInitCheck = init();
98}
99
100WAVExtractor::~WAVExtractor() {
101}
102
103sp<MetaData> WAVExtractor::getMetaData() {
104    sp<MetaData> meta = new MetaData;
105
106    if (mInitCheck != OK) {
107        return meta;
108    }
109
110    meta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_CONTAINER_WAV);
111
112    return meta;
113}
114
115size_t WAVExtractor::countTracks() {
116    return mInitCheck == OK ? 1 : 0;
117}
118
119sp<MediaSource> WAVExtractor::getTrack(size_t index) {
120    if (mInitCheck != OK || index > 0) {
121        return NULL;
122    }
123
124    return new WAVSource(
125            mDataSource, mTrackMeta,
126            mWaveFormat, mBitsPerSample, mDataOffset, mDataSize);
127}
128
129sp<MetaData> WAVExtractor::getTrackMetaData(
130        size_t index, uint32_t /* flags */) {
131    if (mInitCheck != OK || index > 0) {
132        return NULL;
133    }
134
135    return mTrackMeta;
136}
137
138status_t WAVExtractor::init() {
139    uint8_t header[12];
140    if (mDataSource->readAt(
141                0, header, sizeof(header)) < (ssize_t)sizeof(header)) {
142        return NO_INIT;
143    }
144
145    if (memcmp(header, "RIFF", 4) || memcmp(&header[8], "WAVE", 4)) {
146        return NO_INIT;
147    }
148
149    size_t totalSize = U32_LE_AT(&header[4]);
150
151    off64_t offset = 12;
152    size_t remainingSize = totalSize;
153    while (remainingSize >= 8) {
154        uint8_t chunkHeader[8];
155        if (mDataSource->readAt(offset, chunkHeader, 8) < 8) {
156            return NO_INIT;
157        }
158
159        remainingSize -= 8;
160        offset += 8;
161
162        uint32_t chunkSize = U32_LE_AT(&chunkHeader[4]);
163
164        if (chunkSize > remainingSize) {
165            return NO_INIT;
166        }
167
168        if (!memcmp(chunkHeader, "fmt ", 4)) {
169            if (chunkSize < 16) {
170                return NO_INIT;
171            }
172
173            uint8_t formatSpec[40];
174            if (mDataSource->readAt(offset, formatSpec, 2) < 2) {
175                return NO_INIT;
176            }
177
178            mWaveFormat = U16_LE_AT(formatSpec);
179            if (mWaveFormat != WAVE_FORMAT_PCM
180                    && mWaveFormat != WAVE_FORMAT_ALAW
181                    && mWaveFormat != WAVE_FORMAT_MULAW
182                    && mWaveFormat != WAVE_FORMAT_MSGSM
183                    && mWaveFormat != WAVE_FORMAT_EXTENSIBLE) {
184                return ERROR_UNSUPPORTED;
185            }
186
187            uint8_t fmtSize = 16;
188            if (mWaveFormat == WAVE_FORMAT_EXTENSIBLE) {
189                fmtSize = 40;
190            }
191            if (mDataSource->readAt(offset, formatSpec, fmtSize) < fmtSize) {
192                return NO_INIT;
193            }
194
195            mNumChannels = U16_LE_AT(&formatSpec[2]);
196            if (mWaveFormat != WAVE_FORMAT_EXTENSIBLE) {
197                if (mNumChannels != 1 && mNumChannels != 2) {
198                    ALOGW("More than 2 channels (%d) in non-WAVE_EXT, unknown channel mask",
199                            mNumChannels);
200                }
201            } else {
202                if (mNumChannels < 1 && mNumChannels > 8) {
203                    return ERROR_UNSUPPORTED;
204                }
205            }
206
207            mSampleRate = U32_LE_AT(&formatSpec[4]);
208
209            if (mSampleRate == 0) {
210                return ERROR_MALFORMED;
211            }
212
213            mBitsPerSample = U16_LE_AT(&formatSpec[14]);
214
215            if (mWaveFormat == WAVE_FORMAT_PCM
216                    || mWaveFormat == WAVE_FORMAT_EXTENSIBLE) {
217                if (mBitsPerSample != 8 && mBitsPerSample != 16
218                    && mBitsPerSample != 24) {
219                    return ERROR_UNSUPPORTED;
220                }
221            } else if (mWaveFormat == WAVE_FORMAT_MSGSM) {
222                if (mBitsPerSample != 0) {
223                    return ERROR_UNSUPPORTED;
224                }
225            } else {
226                CHECK(mWaveFormat == WAVE_FORMAT_MULAW
227                        || mWaveFormat == WAVE_FORMAT_ALAW);
228                if (mBitsPerSample != 8) {
229                    return ERROR_UNSUPPORTED;
230                }
231            }
232
233            if (mWaveFormat == WAVE_FORMAT_EXTENSIBLE) {
234                uint16_t validBitsPerSample = U16_LE_AT(&formatSpec[18]);
235                if (validBitsPerSample != mBitsPerSample) {
236                    if (validBitsPerSample != 0) {
237                        ALOGE("validBits(%d) != bitsPerSample(%d) are not supported",
238                                validBitsPerSample, mBitsPerSample);
239                        return ERROR_UNSUPPORTED;
240                    } else {
241                        // we only support valitBitsPerSample == bitsPerSample but some WAV_EXT
242                        // writers don't correctly set the valid bits value, and leave it at 0.
243                        ALOGW("WAVE_EXT has 0 valid bits per sample, ignoring");
244                    }
245                }
246
247                mChannelMask = U32_LE_AT(&formatSpec[20]);
248                ALOGV("numChannels=%d channelMask=0x%x", mNumChannels, mChannelMask);
249                if ((mChannelMask >> 18) != 0) {
250                    ALOGE("invalid channel mask 0x%x", mChannelMask);
251                    return ERROR_MALFORMED;
252                }
253
254                if ((mChannelMask != CHANNEL_MASK_USE_CHANNEL_ORDER)
255                        && (popcount(mChannelMask) != mNumChannels)) {
256                    ALOGE("invalid number of channels (%d) in channel mask (0x%x)",
257                            popcount(mChannelMask), mChannelMask);
258                    return ERROR_MALFORMED;
259                }
260
261                // In a WAVE_EXT header, the first two bytes of the GUID stored at byte 24 contain
262                // the sample format, using the same definitions as a regular WAV header
263                mWaveFormat = U16_LE_AT(&formatSpec[24]);
264                if (mWaveFormat != WAVE_FORMAT_PCM
265                        && mWaveFormat != WAVE_FORMAT_ALAW
266                        && mWaveFormat != WAVE_FORMAT_MULAW) {
267                    return ERROR_UNSUPPORTED;
268                }
269                if (memcmp(&formatSpec[26], WAVEEXT_SUBFORMAT, 14)) {
270                    ALOGE("unsupported GUID");
271                    return ERROR_UNSUPPORTED;
272                }
273            }
274
275            mValidFormat = true;
276        } else if (!memcmp(chunkHeader, "data", 4)) {
277            if (mValidFormat) {
278                mDataOffset = offset;
279                mDataSize = chunkSize;
280
281                mTrackMeta = new MetaData;
282
283                switch (mWaveFormat) {
284                    case WAVE_FORMAT_PCM:
285                        mTrackMeta->setCString(
286                                kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_RAW);
287                        break;
288                    case WAVE_FORMAT_ALAW:
289                        mTrackMeta->setCString(
290                                kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_G711_ALAW);
291                        break;
292                    case WAVE_FORMAT_MSGSM:
293                        mTrackMeta->setCString(
294                                kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_MSGSM);
295                        break;
296                    default:
297                        CHECK_EQ(mWaveFormat, (uint16_t)WAVE_FORMAT_MULAW);
298                        mTrackMeta->setCString(
299                                kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_G711_MLAW);
300                        break;
301                }
302
303                mTrackMeta->setInt32(kKeyChannelCount, mNumChannels);
304                mTrackMeta->setInt32(kKeyChannelMask, mChannelMask);
305                mTrackMeta->setInt32(kKeySampleRate, mSampleRate);
306
307                int64_t durationUs = 0;
308                if (mWaveFormat == WAVE_FORMAT_MSGSM) {
309                    // 65 bytes decode to 320 8kHz samples
310                    durationUs =
311                        1000000LL * (mDataSize / 65 * 320) / 8000;
312                } else {
313                    size_t bytesPerSample = mBitsPerSample >> 3;
314
315                    if (!bytesPerSample || !mNumChannels)
316                        return ERROR_MALFORMED;
317
318                    size_t num_samples = mDataSize / (mNumChannels * bytesPerSample);
319
320                    if (!mSampleRate)
321                        return ERROR_MALFORMED;
322
323                    durationUs =
324                        1000000LL * num_samples / mSampleRate;
325                }
326
327                mTrackMeta->setInt64(kKeyDuration, durationUs);
328
329                return OK;
330            }
331        }
332
333        offset += chunkSize;
334    }
335
336    return NO_INIT;
337}
338
339const size_t WAVSource::kMaxFrameSize = 32768;
340
341WAVSource::WAVSource(
342        const sp<DataSource> &dataSource,
343        const sp<MetaData> &meta,
344        uint16_t waveFormat,
345        int32_t bitsPerSample,
346        off64_t offset, size_t size)
347    : mDataSource(dataSource),
348      mMeta(meta),
349      mWaveFormat(waveFormat),
350      mSampleRate(0),
351      mNumChannels(0),
352      mBitsPerSample(bitsPerSample),
353      mOffset(offset),
354      mSize(size),
355      mStarted(false),
356      mGroup(NULL) {
357    CHECK(mMeta->findInt32(kKeySampleRate, &mSampleRate));
358    CHECK(mMeta->findInt32(kKeyChannelCount, &mNumChannels));
359
360    mMeta->setInt32(kKeyMaxInputSize, kMaxFrameSize);
361}
362
363WAVSource::~WAVSource() {
364    if (mStarted) {
365        stop();
366    }
367}
368
369status_t WAVSource::start(MetaData * /* params */) {
370    ALOGV("WAVSource::start");
371
372    CHECK(!mStarted);
373
374    mGroup = new MediaBufferGroup;
375    mGroup->add_buffer(new MediaBuffer(kMaxFrameSize));
376
377    if (mBitsPerSample == 8) {
378        // As a temporary buffer for 8->16 bit conversion.
379        mGroup->add_buffer(new MediaBuffer(kMaxFrameSize));
380    }
381
382    mCurrentPos = mOffset;
383
384    mStarted = true;
385
386    return OK;
387}
388
389status_t WAVSource::stop() {
390    ALOGV("WAVSource::stop");
391
392    CHECK(mStarted);
393
394    delete mGroup;
395    mGroup = NULL;
396
397    mStarted = false;
398
399    return OK;
400}
401
402sp<MetaData> WAVSource::getFormat() {
403    ALOGV("WAVSource::getFormat");
404
405    return mMeta;
406}
407
408status_t WAVSource::read(
409        MediaBuffer **out, const ReadOptions *options) {
410    *out = NULL;
411
412    int64_t seekTimeUs;
413    ReadOptions::SeekMode mode;
414    if (options != NULL && options->getSeekTo(&seekTimeUs, &mode)) {
415        int64_t pos = 0;
416
417        if (mWaveFormat == WAVE_FORMAT_MSGSM) {
418            // 65 bytes decode to 320 8kHz samples
419            int64_t samplenumber = (seekTimeUs * mSampleRate) / 1000000;
420            int64_t framenumber = samplenumber / 320;
421            pos = framenumber * 65;
422        } else {
423            pos = (seekTimeUs * mSampleRate) / 1000000 * mNumChannels * (mBitsPerSample >> 3);
424        }
425        if (pos > (off64_t)mSize) {
426            pos = mSize;
427        }
428        mCurrentPos = pos + mOffset;
429    }
430
431    MediaBuffer *buffer;
432    status_t err = mGroup->acquire_buffer(&buffer);
433    if (err != OK) {
434        return err;
435    }
436
437    // make sure that maxBytesToRead is multiple of 3, in 24-bit case
438    size_t maxBytesToRead =
439        mBitsPerSample == 8 ? kMaxFrameSize / 2 :
440        (mBitsPerSample == 24 ? 3*(kMaxFrameSize/3): kMaxFrameSize);
441
442    size_t maxBytesAvailable =
443        (mCurrentPos - mOffset >= (off64_t)mSize)
444            ? 0 : mSize - (mCurrentPos - mOffset);
445
446    if (maxBytesToRead > maxBytesAvailable) {
447        maxBytesToRead = maxBytesAvailable;
448    }
449
450    if (mWaveFormat == WAVE_FORMAT_MSGSM) {
451        // Microsoft packs 2 frames into 65 bytes, rather than using separate 33-byte frames,
452        // so read multiples of 65, and use smaller buffers to account for ~10:1 expansion ratio
453        if (maxBytesToRead > 1024) {
454            maxBytesToRead = 1024;
455        }
456        maxBytesToRead = (maxBytesToRead / 65) * 65;
457    } else {
458        // read only integral amounts of audio unit frames.
459        const size_t inputUnitFrameSize = mNumChannels * mBitsPerSample / 8;
460        maxBytesToRead -= maxBytesToRead % inputUnitFrameSize;
461    }
462
463    ssize_t n = mDataSource->readAt(
464            mCurrentPos, buffer->data(),
465            maxBytesToRead);
466
467    if (n <= 0) {
468        buffer->release();
469        buffer = NULL;
470
471        return ERROR_END_OF_STREAM;
472    }
473
474    buffer->set_range(0, n);
475
476    if (mWaveFormat == WAVE_FORMAT_PCM || mWaveFormat == WAVE_FORMAT_EXTENSIBLE) {
477        if (mBitsPerSample == 8) {
478            // Convert 8-bit unsigned samples to 16-bit signed.
479
480            MediaBuffer *tmp;
481            CHECK_EQ(mGroup->acquire_buffer(&tmp), (status_t)OK);
482
483            // The new buffer holds the sample number of samples, but each
484            // one is 2 bytes wide.
485            tmp->set_range(0, 2 * n);
486
487            int16_t *dst = (int16_t *)tmp->data();
488            const uint8_t *src = (const uint8_t *)buffer->data();
489            ssize_t numBytes = n;
490
491            while (numBytes-- > 0) {
492                *dst++ = ((int16_t)(*src) - 128) * 256;
493                ++src;
494            }
495
496            buffer->release();
497            buffer = tmp;
498        } else if (mBitsPerSample == 24) {
499            // Convert 24-bit signed samples to 16-bit signed.
500
501            const uint8_t *src =
502                (const uint8_t *)buffer->data() + buffer->range_offset();
503            int16_t *dst = (int16_t *)src;
504
505            size_t numSamples = buffer->range_length() / 3;
506            for (size_t i = 0; i < numSamples; ++i) {
507                int32_t x = (int32_t)(src[0] | src[1] << 8 | src[2] << 16);
508                x = (x << 8) >> 8;  // sign extension
509
510                x = x >> 8;
511                *dst++ = (int16_t)x;
512                src += 3;
513            }
514
515            buffer->set_range(buffer->range_offset(), 2 * numSamples);
516        }
517    }
518
519    int64_t timeStampUs = 0;
520
521    if (mWaveFormat == WAVE_FORMAT_MSGSM) {
522        timeStampUs = 1000000LL * (mCurrentPos - mOffset) * 320 / 65 / mSampleRate;
523    } else {
524        size_t bytesPerSample = mBitsPerSample >> 3;
525        timeStampUs = 1000000LL * (mCurrentPos - mOffset)
526                / (mNumChannels * bytesPerSample) / mSampleRate;
527    }
528
529    buffer->meta_data()->setInt64(kKeyTime, timeStampUs);
530
531    buffer->meta_data()->setInt32(kKeyIsSyncFrame, 1);
532    mCurrentPos += n;
533
534    *out = buffer;
535
536    return OK;
537}
538
539////////////////////////////////////////////////////////////////////////////////
540
541bool SniffWAV(
542        const sp<DataSource> &source, String8 *mimeType, float *confidence,
543        sp<AMessage> *) {
544    char header[12];
545    if (source->readAt(0, header, sizeof(header)) < (ssize_t)sizeof(header)) {
546        return false;
547    }
548
549    if (memcmp(header, "RIFF", 4) || memcmp(&header[8], "WAVE", 4)) {
550        return false;
551    }
552
553    sp<MediaExtractor> extractor = new WAVExtractor(source);
554    if (extractor->countTracks() == 0) {
555        return false;
556    }
557
558    *mimeType = MEDIA_MIMETYPE_CONTAINER_WAV;
559    *confidence = 0.3f;
560
561    return true;
562}
563
564}  // namespace android
565