Android 4.0 MutliMedia 流程分析

对于应用层来说，只需要知道如果使用framework层提供的MediaPlayer和AudioManager就可以完成简单的音视频的播放。

但是，了解Media framework中一系列的实现方法和流程对于以后的扩展或者fix bug都是很有帮助的。

Step 1. new MediaPlayer对象

         |- Java 层

      public MediaPlayer() {        Looper looper;        if ((looper = Looper.myLooper()) != null) {            mEventHandler = new EventHandler(this, looper);        } else if ((looper = Looper.getMainLooper()) != null) {            mEventHandler = new EventHandler(this, looper);        } else {            mEventHandler = null;        }        /* Native setup requires a weak reference to our object.         * It's easier to create it here than in C++.         */        native_setup(new WeakReference<MediaPlayer>(this));      }

          这个构造函数主要做了两个事情。第一，new了一个EventHandler去处理Looper中事件；第二，调用JNI的native_setup，并将自己用为参数传了下去.

           从MediaPlayer中这段代码，我们知道之后就开始调用JNI。

    static {        System.loadLibrary("media_jni");        native_init();    }

 |- JNI层   (frameworks/base/media/jni/android_media_MediaPlayer.cpp)

android_media_MediaPlayer_native_setup(JNIEnv *env, jobject thiz, jobject weak_this){    LOGV("native_setup");    sp<MediaPlayer> mp = new MediaPlayer();  // new 了一个native MediaPlayer.    if (mp == NULL) {        jniThrowException(env, "java/lang/RuntimeException", "Out of memory");        return;    }    // create new listener and give it to MediaPlayer    sp<JNIMediaPlayerListener> listener = new JNIMediaPlayerListener(env, thiz, weak_this);    mp->setListener(listener);    // Stow our new C++ MediaPlayer in an opaque field in the Java object.    setMediaPlayer(env, thiz, mp);}

         native_setup中new了一个native层的MediaPlayer，然后给MediaPlayer加一个listener作为回调函数，暂时不知道这个是用来干嘛的，可以继续往下看。

       |- Native (frameworks/base/media/java/android/media/MediaPlayer.cpp)

MediaPlayer::MediaPlayer(){    LOGV("constructor");    mListener = NULL;    mCookie = NULL;    mDuration = -1;    mStreamType = AUDIO_STREAM_MUSIC;    mCurrentPosition = -1;    mSeekPosition = -1;    mCurrentState = MEDIA_PLAYER_IDLE;    mPrepareSync = false;    mPrepareStatus = NO_ERROR;    mLoop = false;    mLeftVolume = mRightVolume = 1.0;    mVideoWidth = mVideoHeight = 0;    mLockThreadId = 0;    mAudioSessionId = AudioSystem::newAudioSessionId();    AudioSystem::acquireAudioSessionId(mAudioSessionId);    mSendLevel = 0;}

       这个构造函数中只做了基本的初始化，都设置为默认值，其中有个mAudioSessionId = AudioSystem::newAudioSessionId(),其实就是调用AudioFlinger::newAudioSessionId(), 并且用AudioSessionRef去保存这个SessionId，应该是用来作为一个唯一标志符的。之后就会调用acquireAduioSeesionId到AudioFlinger中去查找，如果发现这个SeesionId之前就有了，那就把对应的AudioSessionRef的引用值+1，然后找不到说明之前不存在，就新建一个AudioSessionRef，然后放到mAudioSessionRefs中去保管。

status_t MediaPlayer::setListener(const sp<MediaPlayerListener>& listener){    LOGV("setListener");    Mutex::Autolock _l(mLock);    mListener = listener;    return NO_ERROR;}

       把JNI传下来的listener赋值给mListener作为一个观察者。

       这样MediaPlayer的构造算是完成了，继续往下走。

Step 2. MediaPlayer.setDataSource    

|- Java层 （in  frameworks/base/media/java/android/media/ ）

public void setDataSource(Context context, Uri uri)public void setDataSource(String path)...

       MediaPlayer.java提供了三种setDataSource，可以使用content Uri or file-path or http/rtsp URL.

|- JNI层   ---- JNI （frameworks/base/media/jni）

static voidandroid_media_MediaPlayer_setDataSource(JNIEnv *env, jobject thiz, jstring path){    android_media_MediaPlayer_setDataSourceAndHeaders(env, thiz, path, NULL, NULL);}

static voidandroid_media_MediaPlayer_setDataSourceAndHeaders(        JNIEnv *env, jobject thiz, jstring path,        jobjectArray keys, jobjectArray values) {    sp<MediaPlayer> mp = getMediaPlayer(env, thiz);    ......    status_t opStatus =        mp->setDataSource(   //进入到                pathStr,                headersVector.size() > 0? &headersVector : NULL);}

|- Native层  

status_t MediaPlayer::setDataSource(        const char *url, const KeyedVector<String8, String8> *headers){    LOGV("setDataSource(%s)", url);    status_t err = BAD_VALUE;    if (url != NULL) {        const sp<IMediaPlayerService>& service(getMediaPlayerService());  //通过getMediaPlayerService()获得MediaPlayerServie的IBinder对象，赋值给service        if (service != 0) {            sp<IMediaPlayer> player(service->create(getpid(), this, mAudioSessionId));            if (NO_ERROR != player->setDataSource(url, headers)) {                player.clear();            }            err = attachNewPlayer(player);        }    }    return err;}

     |- sp<IMediaPlayer> player(service->create(getpid(), this, mAudioSessionId));  通过Binder机制去调用MediaPlayerService的Create方法，在Create中会new一个Client出来返回给IMediaPlayer给客户端的MediaPlayer使用，并且会把这个Client保存在服务端的mClients中。

         在Client的构造函数中使用刚才提到的audioSessionId，“mAudioSessionId = audioSessionId;”

sp<IMediaPlayer> MediaPlayerService::create(pid_t pid, const sp<IMediaPlayerClient>& client,        int audioSessionId){    int32_t connId = android_atomic_inc(&mNextConnId);    sp<Client> c = new Client(            this, pid, connId, client, audioSessionId,            IPCThreadState::self()->getCallingUid());    LOGV("Create new client(%d) from pid %d, uid %d, ", connId, pid,         IPCThreadState::self()->getCallingUid());    wp<Client> w = c;    {        Mutex::Autolock lock(mLock);        mClients.add(w);    }    return c;}

     等Client创建完成并且返回给MediaPlayer客户端后，通过调用player->setDataSouce(url, source)去间接调用Client::setDataSource()

status_t MediaPlayerService::Client::setDataSource(        const char *url, const KeyedVector<String8, String8> *headers){    LOGV("setDataSource(%s)", url);    ......    if (strncmp(url, "content://", 10) == 0) {        // get a filedescriptor for the content Uri and        // pass it to the setDataSource(fd) method        String16 url16(url);        int fd = android::openContentProviderFile(url16);        if (fd < 0)        {            LOGE("Couldn't open fd for %s", url);            return UNKNOWN_ERROR;        }        setDataSource(fd, 0, 0x7fffffffffLL); // this sets mStatus        close(fd);        return mStatus;    } else {        player_type playerType = getPlayerType(url);        LOGV("player type = %d", playerType);        // create the right type of player        sp<MediaPlayerBase> p = createPlayer(playerType);        if (p == NULL) return NO_INIT;        if (!p->hardwareOutput()) {            mAudioOutput = new AudioOutput(mAudioSessionId);            static_cast<MediaPlayerInterface*>(p.get())->setAudioSink(mAudioOutput);        }        // now set data source        LOGV(" setDataSource");        mStatus = p->setDataSource(url, headers);        if (mStatus == NO_ERROR) {            mPlayer = p;        } else {            LOGE("  error: %d", mStatus);        }        return mStatus;    }}

       在这个setDataSource中通过getPlayerTyep去根据url的类型判断要创建哪种Player，然后调用creatPlayer, 并把返回值付给成员变量mPlayer,也就说这个mPlayer才是真正的用于工作的一个Player。

sp<MediaPlayerBase> MediaPlayerService::Client::createPlayer(player_type playerType){    // determine if we have the right player type    sp<MediaPlayerBase> p = mPlayer;    if ((p != NULL) && (p->playerType() != playerType)) {        LOGV("delete player");        p.clear();    }    if (p == NULL) {        p = android::createPlayer(playerType, this, notify); //去调用父类的createPlayer    }    if (p != NULL) {        p->setUID(mUID);    }    return p;}static sp<MediaPlayerBase> createPlayer(player_type playerType, void* cookie,        notify_callback_f notifyFunc){    sp<MediaPlayerBase> p;    switch (playerType) {        case SONIVOX_PLAYER:            LOGV(" create MidiFile");            p = new MidiFile();            break;        case STAGEFRIGHT_PLAYER:            LOGV(" create StagefrightPlayer");            p = new StagefrightPlayer;  //大多数情况下都是创建StagefrighetPlayer            break;        case NU_PLAYER:            LOGV(" create NuPlayer");            p = new NuPlayerDriver;            break;        case TEST_PLAYER:            LOGV("Create Test Player stub");            p = new TestPlayerStub();            break;        default:            LOGE("Unknown player type: %d", playerType);            return NULL;    }    if (p != NULL) {        if (p->initCheck() == NO_ERROR) {            p->setNotifyCallback(cookie, notifyFunc);        } else {            p.clear();        }    }    if (p == NULL) {        LOGE("Failed to create player object");    }    return p;}

看到StagefrightPlayer的构造函数，里面new了一个AwesomePlayer,然后把自己作为一个监听器注入到AwesomePlayer中，感觉StagefirghtPlayer更像是一个装饰类一样。实际工作的还是AwesomePlayer.

StagefrightPlayer::StagefrightPlayer()    : mPlayer(new AwesomePlayer) {    LOGV("StagefrightPlayer");    mPlayer->setListener(this);}

         到这边，整个的createPlayer算是创建完了，之后在Client::setDataSource还要做两件事，看看是否有AudioOutput了，如果没有就创建一个，并放到新创建出来的mPlayer的AudioSink中。这里又用到了AudioSessionId。注意AudioOutput继承AudioSink。

        if (!p->hardwareOutput()) {            mAudioOutput = new AudioOutput(mAudioSessionId);            static_cast<MediaPlayerInterface*>(p.get())->setAudioSink(mAudioOutput);        }        // now set data source        LOGV(" setDataSource");                mStatus = p->setDataSource(url, headers);  //调用AwesomePlayer的setDataSource()

    之后才去调用StagefrightPlayer的setDataSource,实际上就等于调用AwesomePlayer的setDataSource(). 接下来我们看看AwesomPlayer的setDataSource。

status_t AwesomePlayer::setDataSource(        const char *uri, const KeyedVector<String8, String8> *headers) {    Mutex::Autolock autoLock(mLock);    return setDataSource_l(uri, headers);}status_t AwesomePlayer::setDataSource_l(        const char *uri, const KeyedVector<String8, String8> *headers) {    reset_l();    mUri = uri;    ...    // The actual work will be done during preparation in the call to    // ::finishSetDataSource_l to avoid blocking the calling thread in    // setDataSource for any significant time.    {        Mutex::Autolock autoLock(mStatsLock);        mStats.mFd = -1;            mStats.mURI = mUri;    }    return OK;}

 在AwesomePlayer中用mStats来保存播放文件的一些参数，在SetDataSource中只是将这些参数放到mStats中并未做任何处理。

到这里整个SetDataSource的动作就算完成了，里面用了Binder机制去实现了C/S构架。MediaPlayer是跟MediaPlayerService相对应的，在MediaPlayerService中创建一个Client然后返回给Client端MediaPlayer中的player。而在Client中又有个真正用于干实事的mPlayer,也就是StageFrightPlayer或者其他的player。里面用了好多设计模式，代理，工厂，装饰...

Step 3. MediaPlayer.prepare()

|- Java层

public native void prepare() throws IOException, IllegalStateException;public native void prepareAsync() throws IllegalStateException;

一般都会使用prepareAsync去异步完成prepare的动作，防止出现ANR。当prepare完成后可以通过OnPreparedListener来进行prepare之后的工作。

|- JNI层

status_t MediaPlayer::prepareAsync_l(){    if ( (mPlayer != 0) && ( mCurrentState & ( MEDIA_PLAYER_INITIALIZED | MEDIA_PLAYER_STOPPED) ) ) {        mPlayer->setAudioStreamType(mStreamType);        mCurrentState = MEDIA_PLAYER_PREPARING;        return mPlayer->prepareAsync();    }    LOGE("prepareAsync called in state %d", mCurrentState);    return INVALID_OPERATION;}

只是简单的判断了下现在的状态，直接就调用Service端的prepareAsync(), 然后直接就走到了AwesomePlayer的prepareAsync_l()。在prepareAsync_l()中并未真正的做处理，而是通过发送一个Event到TimedEventQueue中进行事件调度，在new一个AwesomeEvent的时候会把回调函数也一起放进去，所以我们只要跟踪对应的回调函数就可以了，就是onPrepareAsyncEvent().

status_t MediaPlayerService::Client::prepareAsync(){    ...    status_t ret = p->prepareAsync();    ...    return ret;}

status_t AwesomePlayer::prepareAsync_l() {    ...    if (!mQueueStarted) {        mQueue.start();        mQueueStarted = true;    }    modifyFlags(PREPARING, SET);    mAsyncPrepareEvent = new AwesomeEvent(            this, &AwesomePlayer::onPrepareAsyncEvent);    mQueue.postEvent(mAsyncPrepareEvent);    return OK;}

      在onPrepareAsyncEvent中，会先去判断mUri的size，如果>0也就是我们一开始调用的setDataSource(uri,...)，跟着走到fininshSetDataSource_l();

void AwesomePlayer::onPrepareAsyncEvent() {    Mutex::Autolock autoLock(mLock);    ... ...    if (mUri.size() > 0) {        status_t err = finishSetDataSource_l();        ... ...    }    if (mVideoTrack != NULL && mVideoSource == NULL) {        status_t err = initVideoDecoder();        if (err != OK) {            abortPrepare(err);            return;        }    }    if (mAudioTrack != NULL && mAudioSource == NULL) {        status_t err = initAudioDecoder();        if (err != OK) {            abortPrepare(err);            return;        }    }    modifyFlags(PREPARING_CONNECTED, SET);    if (isStreamingHTTP()) {        postBufferingEvent_l();    } else {        finishAsyncPrepare_l();    }}

finishSetDataSource_l()看上去还蛮复杂的，做的事情也蛮多的。创建了DataSource，Extractor。我们这边先撇开了流媒体，首先通过DataSource::CreateFromURI创建出一个对应Uri的FileDataSource赋值给dataSource. 然后根据DataSource去创建MediaExtractor。MediaExtractor::Create会根据dataSource的类型，返回出对应的Extractor.比如MPEG4Extractor(source)，MP3Extractor(source, meta)，并且把dataSource作为自己的成员。这样Extractor就可以直接操作dataSource。在最后又调用了setDataSource_l(const sp<MediaExtractor> &extractor) .

status_t AwesomePlayer::finishSetDataSource_l() {    sp<DataSource> dataSource;    ... ...    AString sniffedMIME;    if (!strncasecmp("http://", mUri.string(), 7)            || !strncasecmp("https://", mUri.string(), 8)            || isWidevineStreaming) {      ...//主要对流媒体    } else {        dataSource = DataSource::CreateFromURI(mUri.string(), &mUriHeaders);  //返回一个FileSource主要用来对文件进行操作。单例模式？    }    ....    sp<MediaExtractor> extractor;    if (isWidevineStreaming) {       ...    } else {        extractor = MediaExtractor::Create(                dataSource, sniffedMIME.empty() ? NULL : sniffedMIME.c_str());        if (extractor == NULL) {            return UNKNOWN_ERROR;        }    }    dataSource->getDrmInfo(mDecryptHandle, &mDrmManagerClient);    if (mDecryptHandle != NULL) {        CHECK(mDrmManagerClient);        if (RightsStatus::RIGHTS_VALID != mDecryptHandle->status) {            notifyListener_l(MEDIA_ERROR, MEDIA_ERROR_UNKNOWN, ERROR_DRM_NO_LICENSE);        }    }    status_t err = setDataSource_l(extractor);     ...    return OK;}

在setDataSource_l(extractor)中主要判断extractor中的track数，是否有video和audio，如果有分别调用setVideoSource(extractor->getTrack(i));和setAudioSource。extractor->getTrack(i)会返回一个对应的MediaSource。如果是MP3就返回一个MP3Source。MediaSource也有dataSource的一个指针。在setVideSource所做的工作就是mVideoTrack = source; 所以MediaSource就对应于AwesomePlayer中的mVideoTrack。同理mAudioTrack也会有一个MediaSource.

status_t AwesomePlayer::setDataSource_l(const sp<MediaExtractor> &extractor) {    // Attempt to approximate overall stream bitrate by summing all    // tracks' individual bitrates, if not all of them advertise bitrate,    // we have to fail.    int64_t totalBitRate = 0;    for (size_t i = 0; i < extractor->countTracks(); ++i) {        sp<MetaData> meta = extractor->getTrackMetaData(i);        int32_t bitrate;        if (!meta->findInt32(kKeyBitRate, &bitrate)) {            const char *mime;            CHECK(meta->findCString(kKeyMIMEType, &mime));            LOGV("track of type '%s' does not publish bitrate", mime);            totalBitRate = -1;            break;        }        totalBitRate += bitrate;    }    mBitrate = totalBitRate;    LOGV("mBitrate = %lld bits/sec", mBitrate);    bool haveAudio = false;    bool haveVideo = false;    for (size_t i = 0; i < extractor->countTracks(); ++i) {        sp<MetaData> meta = extractor->getTrackMetaData(i);        const char *_mime;        CHECK(meta->findCString(kKeyMIMEType, &_mime));        String8 mime = String8(_mime);        if (!haveVideo && !strncasecmp(mime.string(), "video/", 6)) {            setVideoSource(extractor->getTrack(i));            haveVideo = true;  //是否有Video            // Set the presentation/display size            int32_t displayWidth, displayHeight;            bool success = meta->findInt32(kKeyDisplayWidth, &displayWidth);            if (success) {                success = meta->findInt32(kKeyDisplayHeight, &displayHeight);            }            if (success) {                mDisplayWidth = displayWidth;                mDisplayHeight = displayHeight;            }            {                Mutex::Autolock autoLock(mStatsLock);                mStats.mVideoTrackIndex = mStats.mTracks.size();                mStats.mTracks.push();                TrackStat *stat =                    &mStats.mTracks.editItemAt(mStats.mVideoTrackIndex);                stat->mMIME = mime.string();            }        } else if (!haveAudio && !strncasecmp(mime.string(), "audio/", 6)) {            setAudioSource(extractor->getTrack(i));            haveAudio = true; //是否有Audio            {                Mutex::Autolock autoLock(mStatsLock);                mStats.mAudioTrackIndex = mStats.mTracks.size();                mStats.mTracks.push();                TrackStat *stat =                    &mStats.mTracks.editItemAt(mStats.mAudioTrackIndex);                stat->mMIME = mime.string();            }            if (!strcasecmp(mime.string(), MEDIA_MIMETYPE_AUDIO_VORBIS)) {                // Only do this for vorbis audio, none of the other audio                // formats even support this ringtone specific hack and                // retrieving the metadata on some extractors may turn out                // to be very expensive.                sp<MetaData> fileMeta = extractor->getMetaData();                int32_t loop;                if (fileMeta != NULL                        && fileMeta->findInt32(kKeyAutoLoop, &loop) && loop != 0) {                    modifyFlags(AUTO_LOOPING, SET);                }            }        } else if (!strcasecmp(mime.string(), MEDIA_MIMETYPE_TEXT_3GPP)) {            addTextSource(extractor->getTrack(i));        }    }    if (!haveAudio && !haveVideo) {        return UNKNOWN_ERROR;    }    mExtractorFlags = extractor->flags();    return OK;}

总算这步骤完成之后，我们的prepare也就完成了。

Step 4: MediaPlayer.start()
|- Java 层

 简单的调用JNI的方法

    public  void start() throws IllegalStateException {        stayAwake(true);        mIsSuspend = false;        broadcastIntent(PLAYING_STATUS_CHANGED_ACTION, true);        _start();    }    private native void _start() throws IllegalStateException;

|- JNI 层

调用native层的start();

static voidandroid_media_MediaPlayer_start(JNIEnv *env, jobject thiz){    LOGV("start");    sp<MediaPlayer> mp = getMediaPlayer(env, thiz);    ... ...    process_media_player_call( env, thiz, mp->start(), NULL, NULL );}

|- Native 层

在Native层主要是setLooping,setVolume，然后直接start(); 省略中间的调用，到达了AwesomePlayer::play_l() 

status_t MediaPlayer::start(){    LOGV("start");    Mutex::Autolock _l(mLock);    if (mCurrentState & MEDIA_PLAYER_STARTED)        return NO_ERROR;    if ( (mPlayer != 0) && ( mCurrentState & ( MEDIA_PLAYER_PREPARED |                    MEDIA_PLAYER_PLAYBACK_COMPLETE | MEDIA_PLAYER_PAUSED ) ) ) {        mPlayer->setLooping(mLoop);        mPlayer->setVolume(mLeftVolume, mRightVolume);        mPlayer->setAuxEffectSendLevel(mSendLevel);        mCurrentState = MEDIA_PLAYER_STARTED;        status_t ret = mPlayer->start();        ...    }    LOGE("start called in state %d", mCurrentState);    return INVALID_OPERATION;}

在play_l(), 如果只有Audio，则先去创建AudioPlayer，mTimeSource=mAudioPlayer，然后就直接startAudioPlayer_l；如果还有Video，则postVideoEvent_l(); 我们假设只有Audio，之后video,后面可以慢慢分析。

status_t AwesomePlayer::play_l() {    modifyFlags(SEEK_PREVIEW, CLEAR);    if (mAudioSource != NULL) {        if (mAudioPlayer == NULL) {            if (mAudioSink != NULL) {                mAudioPlayer = new AudioPlayer(mAudioSink, this);                mAudioPlayer->setSource(mAudioSource);                mTimeSource = mAudioPlayer;                // If there was a seek request before we ever started,                // honor the request now.                // Make sure to do this before starting the audio player                // to avoid a race condition.                seekAudioIfNecessary_l();            }        }        CHECK(!(mFlags & AUDIO_RUNNING));        if (mVideoSource == NULL) {            // We don't want to post an error notification at this point,            // the error returned from MediaPlayer::start() will suffice.            status_t err = startAudioPlayer_l(                    false /* sendErrorNotification */);        }    }    if (mTimeSource == NULL && mAudioPlayer == NULL) {        mTimeSource = &mSystemTimeSource;    }    if (mVideoSource != NULL) {        // Kick off video playback        postVideoEvent_l();        if (mAudioSource != NULL && mVideoSource != NULL) {            postVideoLagEvent_l();        }    }    if (mFlags & AT_EOS) {        // Legacy behaviour, if a stream finishes playing and then        // is started again, we play from the start...        seekTo_l(0);    }    uint32_t params = IMediaPlayerService::kBatteryDataCodecStarted        | IMediaPlayerService::kBatteryDataTrackDecoder;    if ((mAudioSource != NULL) && (mAudioSource != mAudioTrack)) {        params |= IMediaPlayerService::kBatteryDataTrackAudio;    }    if (mVideoSource != NULL) {        params |= IMediaPlayerService::kBatteryDataTrackVideo;    }    addBatteryData(params);    if (mVideoSource != NULL) {        if (mMDClient == NULL) {            mMDClient = new MultiDisplayClient();        }        int wcom = 0;        if (mNativeWindow != NULL)            mNativeWindow->query(mNativeWindow.get(), NATIVE_WINDOW_QUEUES_TO_WINDOW_COMPOSER, &wcom);        /*         * 0 means the buffers do not go directly to the window compositor;         * 1 means the ANativeWindow DOES send queued buffers         * directly to the window compositor;         */        if (wcom == 1) {            sp<MetaData> meta = NULL;            int32_t displayW, displayH, frameRate;            displayW = displayH = frameRate = 0;            MDSVideoInfo info;            memset(&info, 0, sizeof(MDSVideoInfo));            info.isplaying = true;            info.isprotected = (mDecryptHandle != NULL);            bool success = false;            if (mVideoTrack != NULL)                meta = mVideoTrack->getFormat();            if (meta != NULL) {                success = meta->findInt32(kKeyFrameRate, &frameRate);                if (!success)                    frameRate = 0;            }            if (mVideoSource != NULL)                meta = mVideoSource->getFormat();            if (meta != NULL) {                success = meta->findInt32(kKeyWidth, &displayW);                if (!success)                    displayW = 0;                success = meta->findInt32(kKeyHeight, &displayH);                if (!success)                    displayH = 0;            }            info.frameRate = frameRate;            info.displayW  = displayW;            info.displayH  = displayH;            mMDClient->updateVideoInfo(&info);        }    }    return OK;}

那我们接下来看看startAudioPlayer_l，startAudioPlayer_l中直接调用 mAudioPlayer->start(true /* sourceAlreadyStarted */);里面总过做了四件事情：

1. mSource->start();

2. mSource->read(&mFirstBuffer, &options);   // 从MP3Source中读取一段Buffer，然后返回给mFirstBuffer

3. mAudioSink->open()  //如果有AudioSink，会创建一个new AudioTrack。创建就要联系到AudioFlinger了。在创建AudioTrack的时候，AudioPlayer会把自己的AudioSinkCallback传递进去，这个callback就是用进行数据交换的。在AudioSinkCallback中会调用AudioPlayer::fillBuffer()，在fillBuffer中其实就是调用了AudioSource的read方法。

4. mAudioSink->start();  即调用mTrack->start();  //关于Audio System相关的可以参考网上一些人的总结，AudioTrack, AudioFlinger, AudioSystem, etc

status_t AudioPlayer::start(bool sourceAlreadyStarted) {    CHECK(!mStarted);    CHECK(mSource != NULL);    status_t err;    if (!sourceAlreadyStarted) {        err = mSource->start();        if (err != OK) {            return err;        }    }    // We allow an optional INFO_FORMAT_CHANGED at the very beginning    // of playback, if there is one, getFormat below will retrieve the    // updated format, if there isn't, we'll stash away the valid buffer    // of data to be used on the first audio callback.    CHECK(mFirstBuffer == NULL);    MediaSource::ReadOptions options;    if (mSeeking) {        options.setSeekTo(mSeekTimeUs);        mSeeking = false;    }    mFirstBufferResult = mSource->read(&mFirstBuffer, &options);    if (mFirstBufferResult == INFO_FORMAT_CHANGED) {        LOGV("INFO_FORMAT_CHANGED!!!");        CHECK(mFirstBuffer == NULL);        mFirstBufferResult = OK;        mIsFirstBuffer = false;    } else {        mIsFirstBuffer = true;    }    sp<MetaData> format = mSource->getFormat();    const char *mime;    bool success = format->findCString(kKeyMIMEType, &mime);    CHECK(success);    CHECK(!strcasecmp(mime, MEDIA_MIMETYPE_AUDIO_RAW));    success = format->findInt32(kKeySampleRate, &mSampleRate);    CHECK(success);    int32_t numChannels;    success = format->findInt32(kKeyChannelCount, &numChannels);    CHECK(success);    mChannels = numChannels;    if (mAudioSink.get() != NULL) {        status_t err = mAudioSink->open(                mSampleRate, numChannels, AUDIO_FORMAT_PCM_16_BIT,                DEFAULT_AUDIOSINK_BUFFERCOUNT,                &AudioPlayer::AudioSinkCallback, this);        if (err != OK) {            if (mFirstBuffer != NULL) {                mFirstBuffer->release();                mFirstBuffer = NULL;            }            if (!sourceAlreadyStarted) {                mSource->stop();            }            return err;        }        mLatencyUs = (int64_t)mAudioSink->latency() * 1000;        mFrameSize = mAudioSink->frameSize();        mAudioSink->start();    } else {        mAudioTrack = new AudioTrack(                AUDIO_STREAM_MUSIC, mSampleRate, AUDIO_FORMAT_PCM_16_BIT,                (numChannels == 2)                    ? AUDIO_CHANNEL_OUT_STEREO                    : AUDIO_CHANNEL_OUT_MONO,                0, 0, &AudioCallback, this, 0);        if ((err = mAudioTrack->initCheck()) != OK) {            delete mAudioTrack;            mAudioTrack = NULL;            if (mFirstBuffer != NULL) {                mFirstBuffer->release();                mFirstBuffer = NULL;            }            if (!sourceAlreadyStarted) {                mSource->stop();            }            return err;        }        mLatencyUs = (int64_t)mAudioTrack->latency() * 1000;        mFrameSize = mAudioTrack->frameSize();        mAudioTrack->start();    }    {        Mutex::Autolock autoLock(mLock);        mIsClockAdjustmentOn = true;    }    mStarted = true;    return OK;}

参考

    http://blog.csdn.net/siobhan/article/details/7179694

    AudioTrack分析 ：http://blog.csdn.net/Innost/article/details/6125779

    AudioFlinger分析：http://blog.csdn.net/innost/article/details/6142812

                                      http://blog.csdn.net/DroidPhone/article/details/5951999

    AudioPolicyServer和AudioPolicyManager分析：http://blog.csdn.net/DroidPhone/article/details/5949280

   AudioTrack 如何与AudioFlinger交换音频数据    http://blog.csdn.net/DroidPhone/article/details/5941344

    Android 4.1 Audio系统的变化： http://www.cnblogs.com/innost/archive/2012/07/16/2593305.html