前言
上一篇我们简单说了AudioFocus如何使用，那么今天就从源码角度看一下AudioFocus的实现原理。
正文
先说下requestAudioFocus，源码如下：
public int requestAudioFocus(@NonNull AudioFocusRequest focusRequest) {
return requestAudioFocus(focusRequest, null /* no AudioPolicy*/);
}
对外的api也是比较简单，这里说下这两个参数，第一个参数，最早传入的是streamtype，到8.0才改用AudioFocusRequest，随着功能的越来越丰富，谷歌特意封装了一个类来处理申请音源的信息，第二个参数 AudioPolicy，这里传的null，从字面理解audio策略，它是一个SystemApi，也就是一般应用是无法使用的，但它主要是允许我们把一些焦点的优先级的处理通过AudioPolicy拿到外部来处理而不是用Android自带的处理逻辑，是不是听着就有点小激动。
先继续分析
public int requestAudioFocus(@NonNull AudioFocusRequest afr, @Nullable AudioPolicy ap) {
if (afr == null) {
throw new NullPointerException("Illegal null AudioFocusRequest");
}
// this can only be checked now, not during the creation of the AudioFocusRequest instance
// 我们如果设置afr的locksFocus为ture时，需要指定ap，否则就抛异常，因为目前的版本lockfocus与audiopolicy是一起使用的因此这里做了这个判断
if (afr.locksFocus() && ap == null) {
throw new IllegalArgumentException(
"Illegal null audio policy when locking audio focus");
}
//（1）注册申请焦点info
registerAudioFocusRequest(afr);
final IAudioService service = getService();
final int status;
int sdk;
try {
sdk = getContext().getApplicationInfo().targetSdkVersion;
} catch (NullPointerException e) {
// some tests don't have a Context
sdk = Build.VERSION.SDK_INT;
}
// （2）clientId 就是把我们传入的AudioFocusRequest里的listener转成了String
final String clientId = getIdForAudioFocusListener(afr.getOnAudioFocusChangeListener());
final BlockingFocusResultReceiver focusReceiver;
synchronized (mFocusRequestsLock) {
try {
// TODO status contains result and generation counter for ext policy
//（3） 调用到了AudioService中requestAudioFocus
status = service.requestAudioFocus(afr.getAudioAttributes(),
afr.getFocusGain(), mICallBack,
mAudioFocusDispatcher,
clientId,
getContext().getOpPackageName() /* package name */, afr.getFlags(),
ap != null ? ap.cb() : null,
sdk);
} catch (RemoteException e) {
throw e.rethrowFromSystemServer();
}
// 上面说过了，目前Audio支持外部处理AudioFocus，这里判断如果不使用外部处理，则直接返回Android原生的申请焦点的结果。
if (status != AudioManager.AUDIOFOCUS_REQUEST_WAITING_FOR_EXT_POLICY) {
// default path with no external focus policy
return status;
}
// 如果map为null，则new 1个size为1的map，这个主要也是处理外部音频焦点策略的
if (mFocusRequestsAwaitingResult == null) {
mFocusRequestsAwaitingResult =
new HashMap<String, BlockingFocusResultReceiver>(1);
}
//（4）关于音频焦点拿到外部处理这块不是一句两句说的明白的，后续在分析
focusReceiver = new BlockingFocusResultReceiver(clientId);
mFocusRequestsAwaitingResult.put(clientId, focusReceiver);
}
// 等待200ms
focusReceiver.waitForResult(EXT_FOCUS_POLICY_TIMEOUT_MS);
if (DEBUG && !focusReceiver.receivedResult()) {
Log.e(TAG, "requestAudio response from ext policy timed out, denying request");
}
synchronized (mFocusRequestsLock) {
mFocusRequestsAwaitingResult.remove(clientId);
}
// 返回外部音频焦点处理结果
return focusReceiver.requestResult();
}
以上代码逻辑比较长，简单说下几个步骤
（1）参数的校验
（2）调用AudioService的requestAudioFocus
（3）判断是否有外部的audiopolicy，如果没有直接返回申请结果，如果有则等外外部音频焦点处理结果
这里在顺便说下 registerAudioFocusRequest(afr)
public void registerAudioFocusRequest(@NonNull AudioFocusRequest afr) {
final Handler h = afr.getOnAudioFocusChangeListenerHandler();
final FocusRequestInfo fri = new FocusRequestInfo(afr, (h == null) ? null :
new ServiceEventHandlerDelegate(h).getHandler());
final String key = getIdForAudioFocusListener(afr.getOnAudioFocusChangeListener());
mAudioFocusIdListenerMap.put(key, fri);
}
为什么单独说下这块呢，因为我们还记得上一篇我们在写AudioFocus的demo的时初始化AudioFocusReques的时候setOnAudioFocusChangeListener(mListener, mHandler)我们传入一个mHandler，这里用到了，那么它的作用是什么呢？其实决定的就是mListener的回调线程，如果不传mHandler那么默认回调到我们使用AudioManager申请焦点的这个线程，如果使用了mHander则回调到mHandler这个线程。建议不要规避开主线程。
那么我们继续分析AudioService
public int requestAudioFocus(AudioAttributes aa, int durationHint, IBinder cb,
IAudioFocusDispatcher fd, String clientId, String callingPackageName, int flags,
IAudioPolicyCallback pcb, int sdk) {
// permission checks
if ((flags & AudioManager.AUDIOFOCUS_FLAG_LOCK) == AudioManager.AUDIOFOCUS_FLAG_LOCK) {
if (AudioSystem.IN_VOICE_COMM_FOCUS_ID.equals(clientId)) {
if (PackageManager.PERMISSION_GRANTED != mContext.checkCallingOrSelfPermission(
android.Manifest.permission.MODIFY_PHONE_STATE)) {
Log.e(TAG, "Invalid permission to (un)lock audio focus", new Exception());
return AudioManager.AUDIOFOCUS_REQUEST_FAILED;
}
} else {
// only a registered audio policy can be used to lock focus
synchronized (mAudioPolicies) {
if (!mAudioPolicies.containsKey(pcb.asBinder())) {
Log.e(TAG, "Invalid unregistered AudioPolicy to (un)lock audio focus");
return AudioManager.AUDIOFOCUS_REQUEST_FAILED;
}
}
}
}
if (callingPackageName == null || clientId == null || aa == null) {
Log.e(TAG, "Invalid null parameter to request audio focus");
return AudioManager.AUDIOFOCUS_REQUEST_FAILED;
}
return mMediaFocusControl.requestAudioFocus(aa, durationHint, cb, fd,
clientId, callingPackageName, flags, sdk,
forceFocusDuckingForAccessibility(aa, durationHint, Binder.getCallingUid()));
}
大体也是分了三步
1.flag的检查，我们刚分析AudioManager的requestAudioFocus时已经看到了flag与audiopolicy的绑定判断，只不过到service又判断了一次，并详细说明了为何之前要绑定判断的原因，因为flag是lock的时候，如果不给电话用只能使用外部audiopolicy，所以之前要做与的判断。如果flag是lock状态并且给电话使用的话，那么我们就要使用requestAudioFocusForCall来申请。
2.参数判空的检查
3.调用了MediaFocusControl的requestAudioFocus
protected int requestAudioFocus(@NonNull AudioAttributes aa, int focusChangeHint, IBinder cb,
IAudioFocusDispatcher fd, @NonNull String clientId, @NonNull String callingPackageName,
int flags, int sdk, boolean forceDuck) {
mEventLogger.log((new AudioEventLogger.StringEvent(
"requestAudioFocus() from uid/pid " + Binder.getCallingUid()
+ "/" + Binder.getCallingPid()
+ " clientId=" + clientId + " callingPack=" + callingPackageName
+ " req=" + focusChangeHint
+ " flags=0x" + Integer.toHexString(flags)
+ " sdk=" + sdk))
.printLog(TAG));
// we need a valid binder callback for clients
if (!cb.pingBinder()) {
Log.e(TAG, " AudioFocus DOA client for requestAudioFocus(), aborting.");
return AudioManager.AUDIOFOCUS_REQUEST_FAILED;
}
if (mAppOps.noteOp(AppOpsManager.OP_TAKE_AUDIO_FOCUS, Binder.getCallingUid(),
callingPackageName) != AppOpsManager.MODE_ALLOWED) {
return AudioManager.AUDIOFOCUS_REQUEST_FAILED;
}
synchronized(mAudioFocusLock) {
//其实申请成功的焦点都放到栈里维护起来了，那么这个焦点栈最多可存100个音频焦点的信息
if (mFocusStack.size() > MAX_STACK_SIZE) {
Log.e(TAG, "Max AudioFocus stack size reached, failing requestAudioFocus()");
return AudioManager.AUDIOFOCUS_REQUEST_FAILED;
}
boolean enteringRingOrCall = !mRingOrCallActive
& (AudioSystem.IN_VOICE_COMM_FOCUS_ID.compareTo(clientId) == 0);
if (enteringRingOrCall) { mRingOrCallActive = true; }
final AudioFocusInfo afiForExtPolicy;
//假设我们不使用外部audiopolicy，只是用android原生的焦点策略，这个过
if (mFocusPolicy != null) {
// construct AudioFocusInfo as it will be communicated to audio focus policy
afiForExtPolicy = new AudioFocusInfo(aa, Binder.getCallingUid(),
clientId, callingPackageName, focusChangeHint, 0 /*lossReceived*/,
flags, sdk);
} else {
afiForExtPolicy = null;
}
// handle delayed focus
boolean focusGrantDelayed = false;
// 如果当前是电话，我们申请的又不是一个可以delay的焦点，那么直接返回申请失败，如果申请的是delay的，则更新delay的flag为true
if (!canReassignAudioFocus()) {
if ((flags & AudioManager.AUDIOFOCUS_FLAG_DELAY_OK) == 0) {
return AudioManager.AUDIOFOCUS_REQUEST_FAILED;
} else {
// request has AUDIOFOCUS_FLAG_DELAY_OK: focus can't be
// granted right now, so the requester will be inserted in the focus stack
// to receive focus later
focusGrantDelayed = true;
}
}
// external focus policy?
if (mFocusPolicy != null) {
if (notifyExtFocusPolicyFocusRequest_syncAf(afiForExtPolicy, fd, cb)) {
// stop handling focus request here as it is handled by external audio
// focus policy (return code will be handled in AudioManager)
return AudioManager.AUDIOFOCUS_REQUEST_WAITING_FOR_EXT_POLICY;
} else {
// an error occured, client already dead, bail early
return AudioManager.AUDIOFOCUS_REQUEST_FAILED;
}
}
// handle the potential premature death of the new holder of the focus
// (premature death == death before abandoning focus)
// Register for client death notification
AudioFocusDeathHandler afdh = new AudioFocusDeathHandler(cb);
try {
cb.linkToDeath(afdh, 0);
} catch (RemoteException e) {
// client has already died!
Log.w(TAG, "AudioFocus requestAudioFocus() could not link to "+cb+" binder death");
return AudioManager.AUDIOFOCUS_REQUEST_FAILED;
}
// 如果申请的音频焦点已经在栈顶存在，则直接返回成功
if (!mFocusStack.empty() && mFocusStack.peek().hasSameClient(clientId)) {
// if focus is already owned by this client and the reason for acquiring the focus
// hasn't changed, don't do anything
final FocusRequester fr = mFocusStack.peek();
if (fr.getGainRequest() == focusChangeHint && fr.getGrantFlags() == flags) {
// unlink death handler so it can be gc'ed.
// linkToDeath() creates a JNI global reference preventing collection.
cb.unlinkToDeath(afdh, 0);
notifyExtPolicyFocusGrant_syncAf(fr.toAudioFocusInfo(),
AudioManager.AUDIOFOCUS_REQUEST_GRANTED);
return AudioManager.AUDIOFOCUS_REQUEST_GRANTED;
}
// the reason for the audio focus request has changed: remove the current top of
// stack and respond as if we had a new focus owner
// 如果焦点栈中有此焦点的clientid，又不是delay状态，则移除栈中的这个焦点信息，
// 这部分可以简单理解为用一个listener申请了两次不同的焦点，则会移除第一次申请的焦点信息
if (!focusGrantDelayed) {
mFocusStack.pop();
// the entry that was "popped" is the same that was "peeked" above
fr.release();
}
}
// focus requester might already be somewhere below in the stack, remove it
removeFocusStackEntry(clientId, false /* signal */, false /*notifyFocusFollowers*/);
final FocusRequester nfr = new FocusRequester(aa, focusChangeHint, flags, fd, cb,
clientId, afdh, callingPackageName, Binder.getCallingUid(), this, sdk);
if (focusGrantDelayed) {
// focusGrantDelayed being true implies we can't reassign focus right now
// which implies the focus stack is not empty.
//对于申请delay的焦点直接放入栈中被delay的焦点下面，比如当前栈中的焦点是电话，那么我们申请的这个允许delay的焦点就放到电话的下面
final int requestResult = pushBelowLockedFocusOwners(nfr);
if (requestResult != AudioManager.AUDIOFOCUS_REQUEST_FAILED) {
notifyExtPolicyFocusGrant_syncAf(nfr.toAudioFocusInfo(), requestResult);
}
return requestResult;
} else {
// propagate the focus change through the stack
//处理不是delay的焦点的情况 遍历焦点栈中所有焦点与之仲裁（主要是给处理栈中已有焦点对应的listener），并将此次申请的音频焦点放入栈中
if (!mFocusStack.empty()) {
propagateFocusLossFromGain_syncAf(focusChangeHint, nfr, forceDuck);
}
// push focus requester at the top of the audio focus stack
mFocusStack.push(nfr);
nfr.handleFocusGainFromRequest(AudioManager.AUDIOFOCUS_REQUEST_GRANTED);
}
notifyExtPolicyFocusGrant_syncAf(nfr.toAudioFocusInfo(),
AudioManager.AUDIOFOCUS_REQUEST_GRANTED);
if (ENFORCE_MUTING_FOR_RING_OR_CALL & enteringRingOrCall) {
runAudioCheckerForRingOrCallAsync(true/*enteringRingOrCall*/);
}
}//synchronized(mAudioFocusLock)
return AudioManager.AUDIOFOCUS_REQUEST_GRANTED;
}
以上，其实申请焦点的过程基本到此就结束了，abandon就不多说了，再说下propagateFocusLossFromGain_syncAf(focusChangeHint, nfr, forceDuck)因为之前说过我们申请什么样音频焦点，能收到什么样的callback就是在这里处理的
private void propagateFocusLossFromGain_syncAf(int focusGain, final FocusRequester fr,
boolean forceDuck) {
final List<String> clientsToRemove = new LinkedList<String>();
// going through the audio focus stack to signal new focus, traversing order doesn't
// matter as all entries respond to the same external focus gain
for (FocusRequester focusLoser : mFocusStack) {
final boolean isDefinitiveLoss =
focusLoser.handleFocusLossFromGain(focusGain, fr, forceDuck);
if (isDefinitiveLoss) {
clientsToRemove.add(focusLoser.getClientId());
}
}
for (String clientToRemove : clientsToRemove) {
removeFocusStackEntry(clientToRemove, false /*signal*/,
true /*notifyFocusFollowers*/);
}
}
我们看到根据申请的音频焦点来跟栈中所有的音频焦点做一个类似仲裁的处理，并把状态为loss的从焦点栈中移除，也就是说如果当前焦点中只有一个QQ音乐，那么我们此时申请一个优酷视频的音频焦点（他俩申请的类型相同），那么此时就会把QQ音乐从焦点栈中移除，具体逻辑继续看
boolean handleFocusLossFromGain(int focusGain, final FocusRequester frWinner, boolean forceDuck)
{
final int focusLoss = focusLossForGainRequest(focusGain);
handleFocusLoss(focusLoss, frWinner, forceDuck);
return (focusLoss == AudioManager.AUDIOFOCUS_LOSS);
}
在看下focusLossForGainRequest
private int focusLossForGainRequest(int gainRequest) {
// gainRequest表示申请的焦点类型
switch(gainRequest) {
case AudioManager.AUDIOFOCUS_GAIN:
// mFocusLossReceived 表示当前栈中焦点状态默认none
switch(mFocusLossReceived) {
case AudioManager.AUDIOFOCUS_LOSS_TRANSIENT_CAN_DUCK:
case AudioManager.AUDIOFOCUS_LOSS_TRANSIENT:
case AudioManager.AUDIOFOCUS_LOSS:
case AudioManager.AUDIOFOCUS_NONE:
return AudioManager.AUDIOFOCUS_LOSS;
}
case AudioManager.AUDIOFOCUS_GAIN_TRANSIENT_EXCLUSIVE:
case AudioManager.AUDIOFOCUS_GAIN_TRANSIENT:
switch(mFocusLossReceived) {
case AudioManager.AUDIOFOCUS_LOSS_TRANSIENT_CAN_DUCK:
case AudioManager.AUDIOFOCUS_LOSS_TRANSIENT:
case AudioManager.AUDIOFOCUS_NONE:
return AudioManager.AUDIOFOCUS_LOSS_TRANSIENT;
case AudioManager.AUDIOFOCUS_LOSS:
return AudioManager.AUDIOFOCUS_LOSS;
}
case AudioManager.AUDIOFOCUS_GAIN_TRANSIENT_MAY_DUCK:
switch(mFocusLossReceived) {
case AudioManager.AUDIOFOCUS_NONE:
case AudioManager.AUDIOFOCUS_LOSS_TRANSIENT_CAN_DUCK:
return AudioManager.AUDIOFOCUS_LOSS_TRANSIENT_CAN_DUCK;
case AudioManager.AUDIOFOCUS_LOSS_TRANSIENT:
return AudioManager.AUDIOFOCUS_LOSS_TRANSIENT;
case AudioManager.AUDIOFOCUS_LOSS:
return AudioManager.AUDIOFOCUS_LOSS;
}
default:
Log.e(TAG, "focusLossForGainRequest() for invalid focus request "+ gainRequest);
return AudioManager.AUDIOFOCUS_NONE;
}
}
原来用了两个switch/case处理所用音频焦点的可能情况。举个例子说明一下我们先申请了一个AUDIOFOCUS_GAIN的焦点，第一次肯定成功，那么第二次我们在申请一个AUDIOFOCUS_GAIN_TRANSIENT_MAY_DUCK的焦点，那么第一次申请的焦点会对应收到AudioManager.AUDIOFOCUS_LOSS_TRANSIENT_CAN_DUCK这么一个callback并把这个值更新给了第一次申请的焦点的mFocusLossReceived，在比如如我们先申请一个AUDIOFOCUS_GAIN成功了，在申请一个AUDIOFOCUS_GAIN_TRANSIENT，那么第一次申请的焦点会收到AudioManager.AUDIOFOCUS_LOSS_TRANSIENT这么一个callback，
总结
申请音频焦点的逻辑基本整个流程就这样的，从AudioManager–>AudioServicer–>MediaFocusControl.最终将申请结果根据我们申请的类型做两个switch/case的判断后返回。如果有问题欢迎大家一起沟通解决~
1834
被折叠的 条评论
为什么被折叠?
到【灌水乐园】发言
