ImportNewXXT0101

Java多线程系列--【JUC锁09】-CountDownLatch原理和示例

参考：http://www.cnblogs.com/skywang12345/p/3533887.html

概要

前面对"独占锁"和"共享锁"有了个大致的了解；本章，我们对CountDownLatch进行学习。和ReadWriteLock.ReadLock一样，CountDownLatch的本质也是一个"共享锁"。本章的内容包括：
CountDownLatch简介
CountDownLatch数据结构
CountDownLatch源码分析(基于JDK1.7.0_40)
CountDownLatch示例

转载请注明出处：http://www.cnblogs.com/skywang12345/p/3533887.html

CountDownLatch简介

CountDownLatch是一个同步辅助类，在完成一组正在其他线程中执行的操作之前，它允许一个或多个线程一直等待。

CountDownLatch和CyclicBarrier的区别
(01) CountDownLatch的作用是允许1或N个线程等待其他线程完成执行；而CyclicBarrier则是允许N个线程相互等待。
(02) CountDownLatch的计数器无法被重置；CyclicBarrier的计数器可以被重置后使用，因此它被称为是循环的barrier。
关于CyclicBarrier的原理，后面一章再来学习。

CountDownLatch函数列表

CountDownLatch(int count)
构造一个用给定计数初始化的 CountDownLatch。

// 使当前线程在锁存器倒计数至零之前一直等待，除非线程被中断。
void await()
// 使当前线程在锁存器倒计数至零之前一直等待，除非线程被中断或超出了指定的等待时间。
boolean await(long timeout, TimeUnit unit)
// 递减锁存器的计数，如果计数到达零，则释放所有等待的线程。
void countDown()
// 返回当前计数。
long getCount()
// 返回标识此锁存器及其状态的字符串。
String toString()

CountDownLatch数据结构

CountDownLatch的UML类图如下：

CountDownLatch的数据结构很简单，它是通过"共享锁"实现的。它包含了sync对象，sync是Sync类型。Sync是实例类，它继承于AQS。

CountDownLatch源码分析(基于JDK1.7.0_40)

CountDownLatch完整源码(基于JDK1.7.0_40)

  1 /*
  2  * ORACLE PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
  3  *
  4  *
  5  *
  6  *
  7  *
  8  *
  9  *
 10  *
 11  *
 12  *
 13  *
 14  *
 15  *
 16  *
 17  *
 18  *
 19  *
 20  *
 21  *
 22  *
 23  */
 24 
 25 /*
 26  *
 27  *
 28  *
 29  *
 30  *
 31  * Written by Doug Lea with assistance from members of JCP JSR-166
 32  * Expert Group and released to the public domain, as explained at
 33  * http://creativecommons.org/publicdomain/zero/1.0/
 34  */
 35 
 36 package java.util.concurrent;
 37 import java.util.concurrent.locks.*;
 38 import java.util.concurrent.atomic.*;
 39 
 40 /**
 41  * A synchronization aid that allows one or more threads to wait until
 42  * a set of operations being performed in other threads completes.
 43  *
 44  * A {@code CountDownLatch} is initialized with a given count.
 45  * The {@link #await await} methods block until the current count reaches
 46  * zero due to invocations of the {@link #countDown} method, after which
 47  * all waiting threads are released and any subsequent invocations of
 48  * {@link #await await} return immediately.  This is a one-shot phenomenon
 49  * -- the count cannot be reset.  If you need a version that resets the
 50  * count, consider using a {@link CyclicBarrier}.
 51  *
 52  * 
A {@code CountDownLatch} is a versatile synchronization tool
 53  * and can be used for a number of purposes.  A
 54  * {@code CountDownLatch} initialized with a count of one serves as a
 55  * simple on/off latch, or gate: all threads invoking {@link #await await}
 56  * wait at the gate until it is opened by a thread invoking {@link
 57  * #countDown}.  A {@code CountDownLatch} initialized to N
 58  * can be used to make one thread wait until N threads have
 59  * completed some action, or some action has been completed N times.
 60  *
 61  * 
A useful property of a {@code CountDownLatch} is that it
 62  * doesn't require that threads calling {@code countDown} wait for
 63  * the count to reach zero before proceeding, it simply prevents any
 64  * thread from proceeding past an {@link #await await} until all
 65  * threads could pass.
 66  *
 67  * 
Sample usage: Here is a pair of classes in which a group
 68  * of worker threads use two countdown latches:
 69  * 

 70  * The first is a start signal that prevents any worker from proceeding
 71  * until the driver is ready for them to proceed;
 72  * 
The second is a completion signal that allows the driver to wait
 73  * until all workers have completed.
 74  * 
 75  *
 76  *  77  * class Driver { // ...
 78  *   void main() throws InterruptedException {
 79  *     CountDownLatch startSignal = new CountDownLatch(1);
 80  *     CountDownLatch doneSignal = new CountDownLatch(N);
 81  *
 82  *     for (int i = 0; i < N; ++i) // create and start threads
 83  *       new Thread(new Worker(startSignal, doneSignal)).start();
 84  *
 85  *     doSomethingElse();            // don't let run yet
 86  *     startSignal.countDown();      // let all threads proceed
 87  *     doSomethingElse();
 88  *     doneSignal.await();           // wait for all to finish
 89  *   }
 90  * }
 91  *
 92  * class Worker implements Runnable {
 93  *   private final CountDownLatch startSignal;
 94  *   private final CountDownLatch doneSignal;
 95  *   Worker(CountDownLatch startSignal, CountDownLatch doneSignal) {
 96  *      this.startSignal = startSignal;
 97  *      this.doneSignal = doneSignal;
 98  *   }
 99  *   public void run() {
100  *      try {
101  *        startSignal.await();
102  *        doWork();
103  *        doneSignal.countDown();
104  *      } catch (InterruptedException ex) {} // return;
105  *   }
106  *
107  *   void doWork() { ... }
108  * }
109  *
110  * 
111  *
112  * Another typical usage would be to divide a problem into N parts,
113  * describe each part with a Runnable that executes that portion and
114  * counts down on the latch, and queue all the Runnables to an
115  * Executor.  When all sub-parts are complete, the coordinating thread
116  * will be able to pass through await. (When threads must repeatedly
117  * count down in this way, instead use a {@link CyclicBarrier}.)
118  *
119  * 
120  * class Driver2 { // ...
121  *   void main() throws InterruptedException {
122  *     CountDownLatch doneSignal = new CountDownLatch(N);
123  *     Executor e = ...
124  *
125  *     for (int i = 0; i < N; ++i) // create and start threads
126  *       e.execute(new WorkerRunnable(doneSignal, i));
127  *
128  *     doneSignal.await();           // wait for all to finish
129  *   }
130  * }
131  *
132  * class WorkerRunnable implements Runnable {
133  *   private final CountDownLatch doneSignal;
134  *   private final int i;
135  *   WorkerRunnable(CountDownLatch doneSignal, int i) {
136  *      this.doneSignal = doneSignal;
137  *      this.i = i;
138  *   }
139  *   public void run() {
140  *      try {
141  *        doWork(i);
142  *        doneSignal.countDown();
143  *      } catch (InterruptedException ex) {} // return;
144  *   }
145  *
146  *   void doWork() { ... }
147  * }
148  *
149  * 
150  *
151  * Memory consistency effects: Until the count reaches
152  * zero, actions in a thread prior to calling
153  * {@code countDown()}
154  * happen-before
155  * actions following a successful return from a corresponding
156  * {@code await()} in another thread.
157  *
158  * @since 1.5
159  * @author Doug Lea
160  */
161 public class CountDownLatch {
162     /**
163      * Synchronization control For CountDownLatch.
164      * Uses AQS state to represent count.
165      */
166     private static final class Sync extends AbstractQueuedSynchronizer {
167         private static final long serialVersionUID = 4982264981922014374L;
168 
169         Sync(int count) {
170             setState(count);
171         }
172 
173         int getCount() {
174             return getState();
175         }
176 
177         protected int tryAcquireShared(int acquires) {
178             return (getState() == 0) ? 1 : -1;
179         }
180 
181         protected boolean tryReleaseShared(int releases) {
182             // Decrement count; signal when transition to zero
183             for (;;) {
184                 int c = getState();
185                 if (c == 0)
186                     return false;
187                 int nextc = c-1;
188                 if (compareAndSetState(c, nextc))
189                     return nextc == 0;
190             }
191         }
192     }
193 
194     private final Sync sync;
195 
196     /**
197      * Constructs a {@code CountDownLatch} initialized with the given count.
198      *
199      * @param count the number of times {@link #countDown} must be invoked
200      *        before threads can pass through {@link #await}
201      * @throws IllegalArgumentException if {@code count} is negative
202      */
203     public CountDownLatch(int count) {
204         if (count < 0) throw new IllegalArgumentException("count < 0");
205         this.sync = new Sync(count);
206     }
207 
208     /**
209      * Causes the current thread to wait until the latch has counted down to
210      * zero, unless the thread is {@linkplain Thread#interrupt interrupted}.
211      *
212      * 
If the current count is zero then this method returns immediately.
213      *
214      * 
If the current count is greater than zero then the current
215      * thread becomes disabled for thread scheduling purposes and lies
216      * dormant until one of two things happen:
217      * 

218      * The count reaches zero due to invocations of the
219      * {@link #countDown} method; or
220      * 
Some other thread {@linkplain Thread#interrupt interrupts}
221      * the current thread.
222      * 
223      *
224      * If the current thread:
225      * 

226      * has its interrupted status set on entry to this method; or
227      * 
is {@linkplain Thread#interrupt interrupted} while waiting,
228      * 
229      * then {@link InterruptedException} is thrown and the current thread's
230      * interrupted status is cleared.
231      *
232      * @throws InterruptedException if the current thread is interrupted
233      *         while waiting
234      */
235     public void await() throws InterruptedException {
236         sync.acquireSharedInterruptibly(1);
237     }
238 
239     /**
240      * Causes the current thread to wait until the latch has counted down to
241      * zero, unless the thread is {@linkplain Thread#interrupt interrupted},
242      * or the specified waiting time elapses.
243      *
244      * If the current count is zero then this method returns immediately
245      * with the value {@code true}.
246      *
247      * 
If the current count is greater than zero then the current
248      * thread becomes disabled for thread scheduling purposes and lies
249      * dormant until one of three things happen:
250      * 

251      * The count reaches zero due to invocations of the
252      * {@link #countDown} method; or
253      * 
Some other thread {@linkplain Thread#interrupt interrupts}
254      * the current thread; or
255      * 
The specified waiting time elapses.
256      * 
257      *
258      * If the count reaches zero then the method returns with the
259      * value {@code true}.
260      *
261      * 
If the current thread:
262      * 

263      * has its interrupted status set on entry to this method; or
264      * 
is {@linkplain Thread#interrupt interrupted} while waiting,
265      * 
266      * then {@link InterruptedException} is thrown and the current thread's
267      * interrupted status is cleared.
268      *
269      * If the specified waiting time elapses then the value {@code false}
270      * is returned.  If the time is less than or equal to zero, the method
271      * will not wait at all.
272      *
273      * @param timeout the maximum time to wait
274      * @param unit the time unit of the {@code timeout} argument
275      * @return {@code true} if the count reached zero and {@code false}
276      *         if the waiting time elapsed before the count reached zero
277      * @throws InterruptedException if the current thread is interrupted
278      *         while waiting
279      */
280     public boolean await(long timeout, TimeUnit unit)
281         throws InterruptedException {
282         return sync.tryAcquireSharedNanos(1, unit.toNanos(timeout));
283     }
284 
285     /**
286      * Decrements the count of the latch, releasing all waiting threads if
287      * the count reaches zero.
288      *
289      * 
If the current count is greater than zero then it is decremented.
290      * If the new count is zero then all waiting threads are re-enabled for
291      * thread scheduling purposes.
292      *
293      * 
If the current count equals zero then nothing happens.
294      */
295     public void countDown() {
296         sync.releaseShared(1);
297     }
298 
299     /**
300      * Returns the current count.
301      *
302      * 
This method is typically used for debugging and testing purposes.
303      *
304      * @return the current count
305      */
306     public long getCount() {
307         return sync.getCount();
308     }
309 
310     /**
311      * Returns a string identifying this latch, as well as its state.
312      * The state, in brackets, includes the String {@code "Count ="}
313      * followed by the current count.
314      *
315      * @return a string identifying this latch, as well as its state
316      */
317     public String toString() {
318         return super.toString() + "[Count = " + sync.getCount() + "]";
319     }
320 }

CountDownLatch是通过“共享锁”实现的。下面，我们分析CountDownLatch中3个核心函数: CountDownLatch(int count), await(), countDown()。

1. CountDownLatch(int count)

public CountDownLatch(int count) {
    if (count < 0) throw new IllegalArgumentException("count < 0");
    this.sync = new Sync(count);
}

说明：该函数是创建一个Sync对象，而Sync是继承于AQS类。Sync构造函数如下：

Sync(int count) {
    setState(count);
}

setState()在AQS中实现，源码如下：

protected final void setState(long newState) {
    state = newState;
}

说明：在AQS中，state是一个private volatile long类型的对象。对于CountDownLatch而言，state表示的”锁计数器“。CountDownLatch中的getCount()最终是调用AQS中的getState()，返回的state对象，即”锁计数器“。

2. await()

public void await() throws InterruptedException {
    sync.acquireSharedInterruptibly(1);
}

说明：该函数实际上是调用的AQS的acquireSharedInterruptibly(1);

AQS中的acquireSharedInterruptibly()的源码如下：

public final void acquireSharedInterruptibly(long arg)
        throws InterruptedException {
    if (Thread.interrupted())
        throw new InterruptedException();
    if (tryAcquireShared(arg) < 0)
        doAcquireSharedInterruptibly(arg);
}

说明：acquireSharedInterruptibly()的作用是获取共享锁。
如果当前线程是中断状态，则抛出异常InterruptedException。否则，调用tryAcquireShared(arg)尝试获取共享锁；尝试成功则返回，否则就调用doAcquireSharedInterruptibly()。doAcquireSharedInterruptibly()会使当前线程一直等待，直到当前线程获取到共享锁(或被中断)才返回。

tryAcquireShared()在CountDownLatch.java中被重写，它的源码如下：

protected int tryAcquireShared(int acquires) {
    return (getState() == 0) ? 1 : -1;
}

说明：tryAcquireShared()的作用是尝试获取共享锁。
如果"锁计数器=0"，即锁是可获取状态，则返回1；否则，锁是不可获取状态，则返回-1。

private void doAcquireSharedInterruptibly(long arg)
    throws InterruptedException {
    // 创建"当前线程"的Node节点，且Node中记录的锁是"共享锁"类型；并将该节点添加到CLH队列末尾。
    final Node node = addWaiter(Node.SHARED);
    boolean failed = true;
    try {
        for (;;) {
            // 获取上一个节点。
            // 如果上一节点是CLH队列的表头，则"尝试获取共享锁"。
            final Node p = node.predecessor();
            if (p == head) {
                long r = tryAcquireShared(arg);
                if (r >= 0) {
                    setHeadAndPropagate(node, r);
                    p.next = null; // help GC
                    failed = false;
                    return;
                }
            }
            // (上一节点不是CLH队列的表头) 当前线程一直等待，直到获取到共享锁。
            // 如果线程在等待过程中被中断过，则再次中断该线程(还原之前的中断状态)。
            if (shouldParkAfterFailedAcquire(p, node) &&
                parkAndCheckInterrupt())
                throw new InterruptedException();
        }
    } finally {
        if (failed)
            cancelAcquire(node);
    }
}

说明：
(01) addWaiter(Node.SHARED)的作用是，创建”当前线程“的Node节点，且Node中记录的锁的类型是”共享锁“(Node.SHARED)；并将该节点添加到CLH队列末尾。关于Node和CLH在"Java多线程系列--“JUC锁”03之公平锁(一)"已经详细介绍过，这里就不再重复说明了。
(02) node.predecessor()的作用是，获取上一个节点。如果上一节点是CLH队列的表头，则”尝试获取共享锁“。
(03) shouldParkAfterFailedAcquire()的作用和它的名称一样，如果在尝试获取锁失败之后，线程应该等待，则返回true；否则，返回false。
(04) 当shouldParkAfterFailedAcquire()返回ture时，则调用parkAndCheckInterrupt()，当前线程会进入等待状态，直到获取到共享锁才继续运行。
doAcquireSharedInterruptibly()中的shouldParkAfterFailedAcquire(), parkAndCheckInterrupt等函数在"Java多线程系列--“JUC锁”03之公平锁(一)"中介绍过，这里也就不再详细说明了。

3. countDown()

public void countDown() {
    sync.releaseShared(1);
}

说明：该函数实际上调用releaseShared(1)释放共享锁。

releaseShared()在AQS中实现，源码如下：

public final boolean releaseShared(int arg) {
    if (tryReleaseShared(arg)) {
        doReleaseShared();
        return true;
    }
    return false;
}

说明：releaseShared()的目的是让当前线程释放它所持有的共享锁。
它首先会通过tryReleaseShared()去尝试释放共享锁。尝试成功，则直接返回；尝试失败，则通过doReleaseShared()去释放共享锁。

tryReleaseShared()在CountDownLatch.java中被重写，源码如下：

protected boolean tryReleaseShared(int releases) {
    // Decrement count; signal when transition to zero
    for (;;) {
        // 获取“锁计数器”的状态
        int c = getState();
        if (c == 0)
            return false;
        // “锁计数器”-1
        int nextc = c-1;
        // 通过CAS函数进行赋值。
        if (compareAndSetState(c, nextc))
            return nextc == 0;
    }
}

说明：tryReleaseShared()的作用是释放共享锁，将“锁计数器”的值-1。

总结：CountDownLatch是通过“共享锁”实现的。在创建CountDownLatch中时，会传递一个int类型参数count，该参数是“锁计数器”的初始状态，表示该“共享锁”最多能被count给线程同时获取。当某线程调用该CountDownLatch对象的await()方法时，该线程会等待“共享锁”可用时，才能获取“共享锁”进而继续运行。而“共享锁”可用的条件，就是“锁计数器”的值为0！而“锁计数器”的初始值为count，每当一个线程调用该CountDownLatch对象的countDown()方法时，才将“锁计数器”-1；通过这种方式，必须有count个线程调用countDown()之后，“锁计数器”才为0，而前面提到的等待线程才能继续运行！

以上，就是CountDownLatch的实现原理。

CountDownLatch的使用示例

下面通过CountDownLatch实现："主线程"等待"5个子线程"全部都完成"指定的工作(休眠1000ms)"之后，再继续运行。

 1 import java.util.concurrent.CountDownLatch;
 2 import java.util.concurrent.CyclicBarrier;
 3 
 4 public class CountDownLatchTest1 {
 5 
 6     private static int LATCH_SIZE = 5;
 7     private static CountDownLatch doneSignal;
 8     public static void main(String[] args) {
 9 
10         try {
11             doneSignal = new CountDownLatch(LATCH_SIZE);
12 
13             // 新建5个任务
14             for(int i=0; i)
15                 new InnerThread().start();
16 
17             System.out.println("main await begin.");
18             // "主线程"等待线程池中5个任务的完成
19             doneSignal.await();
20 
21             System.out.println("main await finished.");
22         } catch (InterruptedException e) {
23             e.printStackTrace();
24         }
25     }
26 
27     static class InnerThread extends Thread{
28         public void run() {
29             try {
30                 Thread.sleep(1000);
31                 System.out.println(Thread.currentThread().getName() + " sleep 1000ms.");
32                 // 将CountDownLatch的数值减1
33                 doneSignal.countDown();
34             } catch (InterruptedException e) {
35                 e.printStackTrace();
36             }
37         }
38     }
39 }

运行结果：

main await begin.
Thread-0 sleep 1000ms.
Thread-2 sleep 1000ms.
Thread-1 sleep 1000ms.
Thread-4 sleep 1000ms.
Thread-3 sleep 1000ms.
main await finished.

结果说明：主线程通过doneSignal.await()等待其它线程将doneSignal递减至0。其它的5个InnerThread线程，每一个都通过doneSignal.countDown()将doneSignal的值减1；当doneSignal为0时，main被唤醒后继续执行。

概要