Handler,Message,Looper,ActivityThread
关键词 nativePollOnce nativeWake ActivityThread并不是Thread ActivityThread使用attach方法将主进程与AMS服务绑定起来 消息队列并不会阻塞 因为所有的生命周期和UI刷新全部用Handler消息机制执行
Handler
Handler 构造函数
public Handler(Callback callback, boolean async) {
if (FIND_POTENTIAL_LEAKS) {
final Class<? extends Handler> klass = getClass();
if ((klass.isAnonymousClass() || klass.isMemberClass() || klass.isLocalClass()) &&
(klass.getModifiers() & Modifier.STATIC) == 0) {
Log.w(TAG, "The following Handler class should be static or leaks might occur: " +
klass.getCanonicalName());
}
}
mLooper = Looper.myLooper();
if (mLooper == null) {
throw new RuntimeException(
"Can't create handler inside thread that has not called Looper.prepare()");
}
mQueue = mLooper.mQueue;
mCallback = callback;
mAsynchronous = async;
}
Handler 主要的几个方法
/*
*消息进队
*/
private boolean enqueueMessage(MessageQueue queue, Message msg, long uptimeMillis) {
msg.target = this;
if (mAsynchronous) {
msg.setAsynchronous(true);
}
return queue.enqueueMessage(msg, uptimeMillis);
}
MessageQueue
MessageQueue构造函数
MessageQueue(boolean quitAllowed) {
mQuitAllowed = quitAllowed;
mPtr = nativeInit();
}
MessageQueue的主要方法
Message next() {
// Return here if the message loop has already quit and been disposed.
// This can happen if the application tries to restart a looper after quit
// which is not supported.
final long ptr = mPtr;
if (ptr == 0) {
return null;
}
int pendingIdleHandlerCount = -1; // -1 only during first iteration
int nextPollTimeoutMillis = 0;
for (;;) {
if (nextPollTimeoutMillis != 0) {
Binder.flushPendingCommands();
}
nativePollOnce(ptr, nextPollTimeoutMillis);
synchronized (this) {
// Try to retrieve the next message. Return if found.
final long now = SystemClock.uptimeMillis();
Message prevMsg = null;
Message msg = mMessages;
if (msg != null && msg.target == null) {
// Stalled by a barrier. Find the next asynchronous message in the queue.
do {
prevMsg = msg;
msg = msg.next;
} while (msg != null && !msg.isAsynchronous());
}
if (msg != null) {
if (now < msg.when) {
// Next message is not ready. Set a timeout to wake up when it is ready.
nextPollTimeoutMillis = (int) Math.min(msg.when - now, Integer.MAX_VALUE);
} else {
// Got a message.
mBlocked = false;
if (prevMsg != null) {
prevMsg.next = msg.next;
} else {
mMessages = msg.next;
}
msg.next = null;
if (DEBUG) Log.v(TAG, "Returning message: " + msg);
msg.markInUse();
return msg;
}
} else {
// No more messages.
nextPollTimeoutMillis = -1;
}
// Process the quit message now that all pending messages have been handled.
if (mQuitting) {
dispose();
return null;
}
// If first time idle, then get the number of idlers to run.
// Idle handles only run if the queue is empty or if the first message
// in the queue (possibly a barrier) is due to be handled in the future.
if (pendingIdleHandlerCount < 0
&& (mMessages == null || now < mMessages.when)) {
pendingIdleHandlerCount = mIdleHandlers.size();
}
if (pendingIdleHandlerCount <= 0) {
// No idle handlers to run. Loop and wait some more.
mBlocked = true;
continue;
}
if (mPendingIdleHandlers == null) {
mPendingIdleHandlers = new IdleHandler[Math.max(pendingIdleHandlerCount, 4)];
}
mPendingIdleHandlers = mIdleHandlers.toArray(mPendingIdleHandlers);
}
// Run the idle handlers.
// We only ever reach this code block during the first iteration.
for (int i = 0; i < pendingIdleHandlerCount; i++) {
final IdleHandler idler = mPendingIdleHandlers[i];
mPendingIdleHandlers[i] = null; // release the reference to the handler
boolean keep = false;
try {
keep = idler.queueIdle();
} catch (Throwable t) {
Log.wtf(TAG, "IdleHandler threw exception", t);
}
if (!keep) {
synchronized (this) {
mIdleHandlers.remove(idler);
}
}
}
// Reset the idle handler count to 0 so we do not run them again.
pendingIdleHandlerCount = 0;
// While calling an idle handler, a new message could have been delivered
// so go back and look again for a pending message without waiting.
nextPollTimeoutMillis = 0;
}
}
Looper
Looper的生成函数
static final ThreadLocal<Looper> sThreadLocal = new ThreadLocal<Looper>();
private static Looper sMainLooper; // guarded by Looper.class
final MessageQueue mQueue;
final Thread mThread;
/*
* 生成Looper
*/
private static void prepare(boolean quitAllowed) {
if (sThreadLocal.get() != null) {
throw new RuntimeException("Only one Looper may be created per thread");
}
sThreadLocal.set(new Looper(quitAllowed));
}
/**
* 生成ActivityThread中的Looper
*/
public static void prepareMainLooper() {
prepare(false);
synchronized (Looper.class) {
if (sMainLooper != null) {
throw new IllegalStateException("The main Looper has already been prepared.");
}
sMainLooper = myLooper();
}
}
Looper主要方法
public static void loop() {
final Looper me = myLooper();
if (me == null) {
throw new RuntimeException("No Looper; Looper.prepare() wasn't called on this thread.");
}
final MessageQueue queue = me.mQueue;
// Make sure the identity of this thread is that of the local process,
// and keep track of what that identity token actually is.
Binder.clearCallingIdentity();
final long ident = Binder.clearCallingIdentity();
for (;;) {
Message msg = queue.next(); // might block
if (msg == null) {
// No message indicates that the message queue is quitting.
return;
}
// This must be in a local variable, in case a UI event sets the logger
final Printer logging = me.mLogging;
if (logging != null) {
logging.println(">>>>> Dispatching to " + msg.target + " " +
msg.callback + ": " + msg.what);
}
final long slowDispatchThresholdMs = me.mSlowDispatchThresholdMs;
final long traceTag = me.mTraceTag;
if (traceTag != 0 && Trace.isTagEnabled(traceTag)) {
Trace.traceBegin(traceTag, msg.target.getTraceName(msg));
}
final long start = (slowDispatchThresholdMs == 0) ? 0 : SystemClock.uptimeMillis();
final long end;
try {
msg.target.dispatchMessage(msg);
end = (slowDispatchThresholdMs == 0) ? 0 : SystemClock.uptimeMillis();
} finally {
if (traceTag != 0) {
Trace.traceEnd(traceTag);
}
}
if (slowDispatchThresholdMs > 0) {
final long time = end - start;
if (time > slowDispatchThresholdMs) {
Slog.w(TAG, "Dispatch took " + time + "ms on "
+ Thread.currentThread().getName() + ", h=" +
msg.target + " cb=" + msg.callback + " msg=" + msg.what);
}
}
if (logging != null) {
logging.println("<<<<< Finished to " + msg.target + " " + msg.callback);
}
// Make sure that during the course of dispatching the
// identity of the thread wasn't corrupted.
final long newIdent = Binder.clearCallingIdentity();
if (ident != newIdent) {
Log.wtf(TAG, "Thread identity changed from 0x"
+ Long.toHexString(ident) + " to 0x"
+ Long.toHexString(newIdent) + " while dispatching to "
+ msg.target.getClass().getName() + " "
+ msg.callback + " what=" + msg.what);
}
msg.recycleUnchecked();
}
}
遇到主要问题
- 消息机制是怎么运转起来的?
一般认为消息机制的Handler,Looper,MessageQueue在消息机制中承担的角色分别是Handler是消息的发射器和分发器,Looper是消息机制的发动机,MessageQueue是消息的存储队列。
Looper.loop()会运行一个死循环去MessageQueue取得Message,拿到Message中的Callback然后分发执行。
在loop方法中通过for (;;)死循环调用queue.next()来得到下一个处理的Message示例来执行,当queue.next()返回为null时,loop方法跳出执行,死循环结束。
而MessageQueue的next方法中同样有一个for (;;)死循环去调用nativePollOnce(ptr, nextPollTimeoutMillis);得到MessageQueue中的下一个Message实例。nativePollOnce()方法和另外一个方法nativeWake(mPtr);实现了Linux的epoll的I/O阻塞模型。(Linux中的epoll阻塞模型是经典的阻塞模型之一)
所以之前的想法是错误的 假设消息队列中有1000个消息,那么Looper,loop()方法的死循环执行1000次便跳出的想法是错误的。不管在loop方法中还是queue的next方法中的死循环在应用的生命周期中都不会被跳出,假设有1000个消息,消息全部处理完后,循环阻塞在nativePollOnce()方法中等待唤醒,当有Handler发射一个消息并且执行enqueueMessage方法调用queue.enqueueMessage(msg, uptimeMillis);方法在enqueueMessage中的nativeWake(mPtr)方法使得阻塞释放重新取出消息进行执行。
- Looper.loop方法会不会阻塞主线程ActivityThread?
简单讲有两个原因
- epoll模型 当没有消息的时候会epoll.wait,等待句柄写的时候再唤醒,这个时候其实是阻塞的。
- 所有的ui操作都通过handler来发消息操作。比如屏幕刷新16ms一个消息,你的各种点击事件,所以就会有句柄写操作,唤醒上文的wait操作,所以不会被卡死了。
详细讲
要完全彻底理解这个问题,需要准备以下4方面的知识:Process/Thread,Android Binder IPC,Handler/Looper/MessageQueue消息机制,Linux pipe/epoll机制。总结一下楼主主要有3个疑惑:1.Android中为什么主线程不会因为Looper.loop()里的死循环卡死? 2.没看见哪里有相关代码为这个死循环准备了一个新线程去运转? 3.Activity的生命周期这些方法这些都是在主线程里执行的吧,那这些生命周期方法是怎么实现在死循环体外能够执行起来的?---------
(1) Android中为什么主线程不会因为Looper.loop()里的死循环卡死? 这里涉及线程,先说说进程/线程,进程:每个app运行时前首先创建一个进程,该进程是由Zygote fork出来的,用于承载App上运行的各种Activity/Service等组件。进程对于上层应用来说是完全透明的,这也是google有意为之,让App程序都是运行在Android Runtime。大多数情况一个App就运行在一个进程中,除非在AndroidManifest.xml中配置Android:process属性,或通过native代码fork进程。线程:线程对应用来说非常常见,比如每次new Thread().start都会创建一个新的线程。该线程与App所在进程之间资源共享,从Linux角度来说进程与线程除了是否共享资源外,并没有本质的区别,都是一个task_struct结构体,在CPU看来进程或线程无非就是一段可执行的代码,CPU采用CFS调度算法,保证每个task都尽可能公平的享有CPU时间片。有了这么准备,再说说死循环问题:对于线程既然是一段可执行的代码,当可执行代码执行完成后,线程生命周期便该终止了,线程退出。而对于主线程,我们是绝不希望会被运行一段时间,自己就退出,那么如何保证能一直存活呢?简单做法就是可执行代码是能一直执行下去的,死循环便能保证不会被退出,例如,binder线程也是采用死循环的方法,通过循环方式不同与Binder驱动进行读写操作,当然并非简单地死循环,无消息时会休眠。但这里可能又引发了另一个问题,既然是死循环又如何去处理其他事务呢?通过创建新线程的方式。真正会卡死主线程的操作是在回调方法onCreate/onStart/onResume等操作时间过长,会导致掉帧,甚至发生ANR,looper.loop本身不会导致应用卡死。
(2) 没看见哪里有相关代码为这个死循环准备了一个新线程去运转? 事实上,会在进入死循环之前便创建了新binder线程,在代码ActivityThread.main()中:
public static void main(String[] args) {
//创建Looper和MessageQueue对象,用于处理主线程的消息
Looper.prepareMainLooper();
//创建ActivityThread对象
ActivityThread thread = new ActivityThread();
//建立Binder通道 (创建新线程)
thread.attach(false);
Looper.loop(); //消息循环运行
throw new RuntimeException("Main thread loop unexpectedly exited");
}
thread.attach(false);便会创建一个Binder线程(具体是指ApplicationThread,Binder的服务端,用于接收系统服务AMS发送来的事件),该Binder线程通过Handler将Message发送给主线程,具体过程可查看 startService流程分析,这里不展开说,简单说Binder用于进程间通信,采用C/S架构。另外,ActivityThread实际上并非线程,不像HandlerThread类,ActivityThread并没有真正继承Thread类,只是往往运行在主线程,该人以线程的感觉,其实承载ActivityThread的主线程就是由Zygote fork而创建的进程。主线程的死循环一直运行是不是特别消耗CPU资源呢? 其实不然,这里就涉及到Linux pipe/epoll机制,简单说就是在主线程的MessageQueue没有消息时,便阻塞在loop的queue.next()中的nativePollOnce()方法里,此时主线程会释放CPU资源进入休眠状态,直到下个消息到达或者有事务发生,通过往pipe管道写端写入数据来唤醒主线程工作。这里采用的epoll机制,是一种IO多路复用机制,可以同时监控多个描述符,当某个描述符就绪(读或写就绪),则立刻通知相应程序进行读或写操作,本质同步I/O,即读写是阻塞的。 所以说,主线程大多数时候都是处于休眠状态,并不会消耗大量CPU资源。
(3) Activity的生命周期是怎么实现在死循环体外能够执行起来的?ActivityThread的内部类H继承于Handler,通过handler消息机制,简单说Handler机制用于同一个进程的线程间通信。Activity的生命周期都是依靠主线程的Looper.loop,当收到不同Message时则采用相应措施:在H.handleMessage(msg)方法中,根据接收到不同的msg,执行相应的生命周期。 比如收到msg=H.LAUNCH_ACTIVITY,则调用ActivityThread.handleLaunchActivity()方法,最终会通过反射机制,创建Activity实例,然后再执行Activity.onCreate()等方法; 再比如收到msg=H.PAUSE_ACTIVITY,则调用ActivityThread.handlePauseActivity()方法,最终会执行Activity.onPause()等方法。 上述过程,我只挑核心逻辑讲,真正该过程远比这复杂。主线程的消息又是哪来的呢?当然是App进程中的其他线程通过Handler发送给主线程,请看接下来的内容:--------------------------------------------------------------------------------------------------------------------------------------最后,从进程与线程间通信的角度,通过一张图加深大家对App运行过程的理解:
system_server进程是系统进程,java framework框架的核心载体,里面运行了大量的系统服务,比如这里提供ApplicationThreadProxy(简称ATP),ActivityManagerService(简称AMS),这个两个服务都运行在system_server进程的不同线程中,由于ATP和AMS都是基于IBinder接口,都是binder线程,binder线程的创建与销毁都是由binder驱动来决定的。App进程则是我们常说的应用程序,主线程主要负责Activity/Service等组件的生命周期以及UI相关操作都运行在这个线程; 另外,每个App进程中至少会有两个binder线程 ApplicationThread(简称AT)和ActivityManagerProxy(简称AMP),除了图中画的线程,其中还有很多线程,比如signal catcher线程等,这里就不一一列举。Binder用于不同进程之间通信,由一个进程的Binder客户端向另一个进程的服务端发送事务,比如图中线程2向线程4发送事务;而handler用于同一个进程中不同线程的通信,比如图中线程4向主线程发送消息。结合图说说Activity生命周期,比如暂停Activity,流程如下:线程1的AMS中调用线程2的ATP;(由于同一个进程的线程间资源共享,可以相互直接调用,但需要注意多线程并发问题)线程2通过binder传输到App进程的线程4;线程4通过handler消息机制,将暂停Activity的消息发送给主线程;主线程在looper.loop()中循环遍历消息,当收到暂停Activity的消息时,便将消息分发给ActivityThread.H.handleMessage()方法,再经过方法的调用,最后便会调用到Activity.onPause(),当onPause()处理完后,继续循环loop下去。
private class H extends Handler {},其实例final H mH = new H();通过发送Message的方式完成Activity的生命周期的调用。例如
public final void scheduleStopActivity(IBinder token, boolean showWindow,
int configChanges) {
int seq = getLifecycleSeq();
if (DEBUG_ORDER) Slog.d(TAG, "stopActivity " + ActivityThread.this
+ " operation received seq: " + seq);
sendMessage(
showWindow ? H.STOP_ACTIVITY_SHOW : H.STOP_ACTIVITY_HIDE,
token, 0, configChanges, seq);
}
会在ActivityThread中的handleMessage方法中
case STOP_ACTIVITY_SHOW: {
Trace.traceBegin(Trace.TRACE_TAG_ACTIVITY_MANAGER, "activityStop");
SomeArgs args = (SomeArgs) msg.obj;
handleStopActivity((IBinder) args.arg1, true, args.argi2, args.argi3);
Trace.traceEnd(Trace.TRACE_TAG_ACTIVITY_MANAGER);
} break;
Looper 怎么保证的在一个线程中只有一个Looper 的实例的?
在Looper中有这样一个变量
static final ThreadLocal<Looper> sThreadLocal = new ThreadLocal<Looper>();
这里采用ThreadLocal来保证各个线程的Looper的唯一性
ThreadLocal的主要方法有
public T get() {
Thread t = Thread.currentThread();
ThreadLocalMap map = getMap(t);
if (map != null) {
ThreadLocalMap.Entry e = map.getEntry(this);
if (e != null) {
@SuppressWarnings("unchecked")
T result = (T)e.value;
return result;
}
}
return setInitialValue();
}
public void set(T value) {
Thread t = Thread.currentThread();
ThreadLocalMap map = getMap(t);
if (map != null)
map.set(this, value);
else
createMap(t, value);
}
static class ThreadLocalMap {}
而在Thread中存在
/* ThreadLocal values pertaining to this thread. This map is maintained
* by the ThreadLocal class. */
ThreadLocal.ThreadLocalMap threadLocals = null;
通过这样的方式保证了Looper的唯一性