AACExtractor.cpp revision d411b4ca2945cd8974a3a78199fce94646950128 
1/*
2 * Copyright (C) 2011 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 "AACExtractor"
19#include <utils/Log.h>
20
21#include "include/AACExtractor.h"
22#include "include/avc_utils.h"
23
24#include <media/stagefright/foundation/ABuffer.h>
25#include <media/stagefright/foundation/AMessage.h>
26#include <media/stagefright/foundation/ADebug.h>
27#include <media/stagefright/DataSource.h>
28#include <media/stagefright/MediaBufferGroup.h>
29#include <media/stagefright/MediaDefs.h>
30#include <media/stagefright/MediaErrors.h>
31#include <media/stagefright/MediaSource.h>
32#include <media/stagefright/MetaData.h>
33#include <utils/String8.h>
34
35namespace android {
36
37class AACSource : public MediaSource {
38public:
39    AACSource(const sp<DataSource> &source,
40              const sp<MetaData> &meta,
41              const Vector<uint64_t> &offset_vector,
42              int64_t frame_duration_us);
43
44    virtual status_t start(MetaData *params = NULL);
45    virtual status_t stop();
46
47    virtual sp<MetaData> getFormat();
48
49    virtual status_t read(
50            MediaBuffer **buffer, const ReadOptions *options = NULL);
51
52protected:
53    virtual ~AACSource();
54
55private:
56    static const size_t kMaxFrameSize;
57    sp<DataSource> mDataSource;
58    sp<MetaData> mMeta;
59
60    off64_t mOffset;
61    int64_t mCurrentTimeUs;
62    bool mStarted;
63    MediaBufferGroup *mGroup;
64
65    Vector<uint64_t> mOffsetVector;
66    int64_t mFrameDurationUs;
67
68    AACSource(const AACSource &);
69    AACSource &operator=(const AACSource &);
70};
71
72////////////////////////////////////////////////////////////////////////////////
73
74// Returns the sample rate based on the sampling frequency index
75uint32_t get_sample_rate(const uint8_t sf_index)
76{
77    static const uint32_t sample_rates[] =
78    {
79        96000, 88200, 64000, 48000, 44100, 32000,
80        24000, 22050, 16000, 12000, 11025, 8000
81    };
82
83    if (sf_index < sizeof(sample_rates) / sizeof(sample_rates[0])) {
84        return sample_rates[sf_index];
85    }
86
87    return 0;
88}
89
90// Returns the frame length in bytes as described in an ADTS header starting at the given offset,
91//     or 0 if the size can't be read due to an error in the header or a read failure.
92// The returned value is the AAC frame size with the ADTS header length (regardless of
93//     the presence of the CRC).
94// If headerSize is non-NULL, it will be used to return the size of the header of this ADTS frame.
95static size_t getAdtsFrameLength(const sp<DataSource> &source, off64_t offset, size_t* headerSize) {
96
97    const size_t kAdtsHeaderLengthNoCrc = 7;
98    const size_t kAdtsHeaderLengthWithCrc = 9;
99
100    size_t frameSize = 0;
101
102    uint8_t syncword[2];
103    if (source->readAt(offset, &syncword, 2) != 2) {
104        return 0;
105    }
106    if ((syncword[0] != 0xff) || ((syncword[1] & 0xf6) != 0xf0)) {
107        return 0;
108    }
109
110    uint8_t protectionAbsent;
111    if (source->readAt(offset + 1, &protectionAbsent, 1) < 1) {
112        return 0;
113    }
114    protectionAbsent &= 0x1;
115
116    uint8_t header[3];
117    if (source->readAt(offset + 3, &header, 3) < 3) {
118        return 0;
119    }
120
121    frameSize = (header[0] & 0x3) << 11 | header[1] << 3 | header[2] >> 5;
122
123    // protectionAbsent is 0 if there is CRC
124    size_t headSize = protectionAbsent ? kAdtsHeaderLengthNoCrc : kAdtsHeaderLengthWithCrc;
125    if (headSize > frameSize) {
126        return 0;
127    }
128    if (headerSize != NULL) {
129        *headerSize = headSize;
130    }
131
132    return frameSize;
133}
134
135AACExtractor::AACExtractor(
136        const sp<DataSource> &source, const sp<AMessage> &_meta)
137    : mDataSource(source),
138      mInitCheck(NO_INIT),
139      mFrameDurationUs(0) {
140    sp<AMessage> meta = _meta;
141
142    if (meta == NULL) {
143        String8 mimeType;
144        float confidence;
145        sp<AMessage> _meta;
146
147        if (!SniffAAC(mDataSource, &mimeType, &confidence, &meta)) {
148            return;
149        }
150    }
151
152    int64_t offset;
153    CHECK(meta->findInt64("offset", &offset));
154
155    uint8_t profile, sf_index, channel, header[2];
156    if (mDataSource->readAt(offset + 2, &header, 2) < 2) {
157        return;
158    }
159
160    profile = (header[0] >> 6) & 0x3;
161    sf_index = (header[0] >> 2) & 0xf;
162    uint32_t sr = get_sample_rate(sf_index);
163    if (sr == 0) {
164        return;
165    }
166    channel = (header[0] & 0x1) << 2 | (header[1] >> 6);
167
168    mMeta = MakeAACCodecSpecificData(profile, sf_index, channel);
169
170    off64_t streamSize, numFrames = 0;
171    size_t frameSize = 0;
172    int64_t duration = 0;
173
174    if (mDataSource->getSize(&streamSize) == OK) {
175         while (offset < streamSize) {
176            if ((frameSize = getAdtsFrameLength(source, offset, NULL)) == 0) {
177                return;
178            }
179
180            mOffsetVector.push(offset);
181
182            offset += frameSize;
183            numFrames ++;
184        }
185
186        // Round up and get the duration
187        mFrameDurationUs = (1024 * 1000000ll + (sr - 1)) / sr;
188        duration = numFrames * mFrameDurationUs;
189        mMeta->setInt64(kKeyDuration, duration);
190    }
191
192    mInitCheck = OK;
193}
194
195AACExtractor::~AACExtractor() {
196}
197
198sp<MetaData> AACExtractor::getMetaData() {
199    sp<MetaData> meta = new MetaData;
200
201    if (mInitCheck != OK) {
202        return meta;
203    }
204
205    meta->setCString(kKeyMIMEType, MEDIA_MIMETYPE_AUDIO_AAC_ADTS);
206
207    return meta;
208}
209
210size_t AACExtractor::countTracks() {
211    return mInitCheck == OK ? 1 : 0;
212}
213
214sp<MediaSource> AACExtractor::getTrack(size_t index) {
215    if (mInitCheck != OK || index != 0) {
216        return NULL;
217    }
218
219    return new AACSource(mDataSource, mMeta, mOffsetVector, mFrameDurationUs);
220}
221
222sp<MetaData> AACExtractor::getTrackMetaData(size_t index, uint32_t /* flags */) {
223    if (mInitCheck != OK || index != 0) {
224        return NULL;
225    }
226
227    return mMeta;
228}
229
230////////////////////////////////////////////////////////////////////////////////
231
232// 8192 = 2^13, 13bit AAC frame size (in bytes)
233const size_t AACSource::kMaxFrameSize = 8192;
234
235AACSource::AACSource(
236        const sp<DataSource> &source, const sp<MetaData> &meta,
237        const Vector<uint64_t> &offset_vector,
238        int64_t frame_duration_us)
239    : mDataSource(source),
240      mMeta(meta),
241      mOffset(0),
242      mCurrentTimeUs(0),
243      mStarted(false),
244      mGroup(NULL),
245      mOffsetVector(offset_vector),
246      mFrameDurationUs(frame_duration_us) {
247}
248
249AACSource::~AACSource() {
250    if (mStarted) {
251        stop();
252    }
253}
254
255status_t AACSource::start(MetaData * /* params */) {
256    CHECK(!mStarted);
257
258    if (mOffsetVector.empty()) {
259        mOffset = 0;
260    } else {
261        mOffset = mOffsetVector.itemAt(0);
262    }
263
264    mCurrentTimeUs = 0;
265    mGroup = new MediaBufferGroup;
266    mGroup->add_buffer(new MediaBuffer(kMaxFrameSize));
267    mStarted = true;
268
269    return OK;
270}
271
272status_t AACSource::stop() {
273    CHECK(mStarted);
274
275    delete mGroup;
276    mGroup = NULL;
277
278    mStarted = false;
279    return OK;
280}
281
282sp<MetaData> AACSource::getFormat() {
283    return mMeta;
284}
285
286status_t AACSource::read(
287        MediaBuffer **out, const ReadOptions *options) {
288    *out = NULL;
289
290    int64_t seekTimeUs;
291    ReadOptions::SeekMode mode;
292    if (options && options->getSeekTo(&seekTimeUs, &mode)) {
293        if (mFrameDurationUs > 0) {
294            int64_t seekFrame = seekTimeUs / mFrameDurationUs;
295            mCurrentTimeUs = seekFrame * mFrameDurationUs;
296
297            mOffset = mOffsetVector.itemAt(seekFrame);
298        }
299    }
300
301    size_t frameSize, frameSizeWithoutHeader, headerSize;
302    if ((frameSize = getAdtsFrameLength(mDataSource, mOffset, &headerSize)) == 0) {
303        return ERROR_END_OF_STREAM;
304    }
305
306    MediaBuffer *buffer;
307    status_t err = mGroup->acquire_buffer(&buffer);
308    if (err != OK) {
309        return err;
310    }
311
312    frameSizeWithoutHeader = frameSize - headerSize;
313    if (mDataSource->readAt(mOffset + headerSize, buffer->data(),
314                frameSizeWithoutHeader) != (ssize_t)frameSizeWithoutHeader) {
315        buffer->release();
316        buffer = NULL;
317
318        return ERROR_IO;
319    }
320
321    buffer->set_range(0, frameSizeWithoutHeader);
322    buffer->meta_data()->setInt64(kKeyTime, mCurrentTimeUs);
323    buffer->meta_data()->setInt32(kKeyIsSyncFrame, 1);
324
325    mOffset += frameSize;
326    mCurrentTimeUs += mFrameDurationUs;
327
328    *out = buffer;
329    return OK;
330}
331
332////////////////////////////////////////////////////////////////////////////////
333
334bool SniffAAC(
335        const sp<DataSource> &source, String8 *mimeType, float *confidence,
336        sp<AMessage> *meta) {
337    off64_t pos = 0;
338
339    for (;;) {
340        uint8_t id3header[10];
341        if (source->readAt(pos, id3header, sizeof(id3header))
342                < (ssize_t)sizeof(id3header)) {
343            return false;
344        }
345
346        if (memcmp("ID3", id3header, 3)) {
347            break;
348        }
349
350        // Skip the ID3v2 header.
351
352        size_t len =
353            ((id3header[6] & 0x7f) << 21)
354            | ((id3header[7] & 0x7f) << 14)
355            | ((id3header[8] & 0x7f) << 7)
356            | (id3header[9] & 0x7f);
357
358        len += 10;
359
360        pos += len;
361
362        ALOGV("skipped ID3 tag, new starting offset is %lld (0x%016llx)",
363             pos, pos);
364    }
365
366    uint8_t header[2];
367
368    if (source->readAt(pos, &header, 2) != 2) {
369        return false;
370    }
371
372    // ADTS syncword
373    if ((header[0] == 0xff) && ((header[1] & 0xf6) == 0xf0)) {
374        *mimeType = MEDIA_MIMETYPE_AUDIO_AAC_ADTS;
375        *confidence = 0.2;
376
377        *meta = new AMessage;
378        (*meta)->setInt64("offset", pos);
379
380        return true;
381    }
382
383    return false;
384}
385
386}  // namespace android
387