CH24-CountDownLatch

概览

其底层是由AQS提供支持,所以其数据结构可以参考AQS的数据结构,而AQS的数据结构核心就是两个虚拟队列: 同步队列sync queue 和条件队列condition queue,不同的条件会有不同的条件队列。

CountDownLatch典型的用法是将一个程序分为n个互相独立的可解决任务,并创建值为n的CountDownLatch。当每一个任务完成时,都会在这个锁存器上调用countDown,等待问题被解决的任务调用这个锁存器的await,将他们自己拦住,直至锁存器计数结束。

源码分析

层级结构

CountDownLatch没有显示继承哪个父类或者实现哪个父接口, 它底层是AQS是通过内部类Sync来实现的。

内部类

CountDownLatch类存在一个内部类Sync,继承自AbstractQueuedSynchronizer,其源代码如下。

private static final class Sync extends AbstractQueuedSynchronizer {
    // 版本号
    private static final long serialVersionUID = 4982264981922014374L;
    
    // 构造器
    Sync(int count) {
        setState(count);
    }
    
    // 返回当前计数
    int getCount() {
        return getState();
    }

    // 试图在共享模式下获取对象状态
    protected int tryAcquireShared(int acquires) {
        return (getState() == 0) ? 1 : -1;
    }

    // 试图设置状态来反映共享模式下的一个释放
    protected boolean tryReleaseShared(int releases) {
        // Decrement count; signal when transition to zero
        // 无限循环
        for (;;) {
            // 获取状态
            int c = getState();
            if (c == 0) // 没有被线程占有
                return false;
            // 下一个状态
            int nextc = c-1;
            if (compareAndSetState(c, nextc)) // 比较并且设置成功
                return nextc == 0;
        }
    }
}

对CountDownLatch方法的调用会转发到对Sync或AQS的方法的调用,所以,AQS对CountDownLatch提供支持。

类属性

可以看到CountDownLatch类的内部只有一个Sync类型的属性:

public class CountDownLatch {
    // 同步队列
    private final Sync sync;
}

构造函数

public CountDownLatch(int count) {
    if (count < 0) throw new IllegalArgumentException("count < 0");
    // 初始化状态数
    this.sync = new Sync(count);
}

该构造函数可以构造一个用给定计数初始化的CountDownLatch,并且构造函数内完成了sync的初始化,并设置了状态数。

核心函数:await

此函数将会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。其源码如下

public void await() throws InterruptedException {
    // 转发到sync对象上
    sync.acquireSharedInterruptibly(1);
}

对CountDownLatch对象的await的调用会转发为对Sync的acquireSharedInterruptibly(从AQS继承的方法)方法的调用。

  • acquireSharedInterruptibly源码如下:
public final void acquireSharedInterruptibly(int arg)
        throws InterruptedException {
    if (Thread.interrupted())
        throw new InterruptedException();
    if (tryAcquireShared(arg) < 0)
        doAcquireSharedInterruptibly(arg);
}

acquireSharedInterruptibly又调用了CountDownLatch的内部类Sync的tryAcquireShared和AQS的doAcquireSharedInterruptibly函数。

  • tryAcquireShared函数的源码如下:
protected int tryAcquireShared(int acquires) {
    return (getState() == 0) ? 1 : -1;
}

该函数只是简单的判断AQS的state是否为0,为0则返回1,不为0则返回-1。

  • doAcquireSharedInterruptibly函数的源码如下:
private void doAcquireSharedInterruptibly(int arg) throws InterruptedException {
    // 添加节点至等待队列
    final Node node = addWaiter(Node.SHARED);
    boolean failed = true;
    try {
        for (;;) { // 无限循环
            // 获取node的前驱节点
            final Node p = node.predecessor();
            if (p == head) { // 前驱节点为头结点
                // 试图在共享模式下获取对象状态
                int r = tryAcquireShared(arg);
                if (r >= 0) { // 获取成功
                    // 设置头结点并进行繁殖
                    setHeadAndPropagate(node, r);
                    // 设置节点next域
                    p.next = null; // help GC
                    failed = false;
                    return;
                }
            }
            if (shouldParkAfterFailedAcquire(p, node) &&
                parkAndCheckInterrupt()) // 在获取失败后是否需要禁止线程并且进行中断检查
                // 抛出异常
                throw new InterruptedException();
        }
    } finally {
        if (failed)
            cancelAcquire(node);
    }
}

在AQS的doAcquireSharedInterruptibly中可能会再次调用CountDownLatch的内部类Sync的tryAcquireShared方法和AQS的setHeadAndPropagate方法。

  • setHeadAndPropagate方法源码如下
private void setHeadAndPropagate(Node node, int propagate) {
    // 获取头结点
    Node h = head; // Record old head for check below
    // 设置头结点
    setHead(node);
    /*
        * Try to signal next queued node if:
        *   Propagation was indicated by caller,
        *     or was recorded (as h.waitStatus either before
        *     or after setHead) by a previous operation
        *     (note: this uses sign-check of waitStatus because
        *      PROPAGATE status may transition to SIGNAL.)
        * and
        *   The next node is waiting in shared mode,
        *     or we don't know, because it appears null
        *
        * The conservatism in both of these checks may cause
        * unnecessary wake-ups, but only when there are multiple
        * racing acquires/releases, so most need signals now or soon
        * anyway.
        */
    // 进行判断
    if (propagate > 0 || h == null || h.waitStatus < 0 ||
        (h = head) == null || h.waitStatus < 0) {
        // 获取节点的后继
        Node s = node.next;
        if (s == null || s.isShared()) // 后继为空或者为共享模式
            // 以共享模式进行释放
            doReleaseShared();
    }
}

该方法设置头结点并且释放头结点后面的满足条件的结点,该方法中可能会调用到AQS的doReleaseShared方法,其源码如下。

private void doReleaseShared() {
    /*
        * Ensure that a release propagates, even if there are other
        * in-progress acquires/releases.  This proceeds in the usual
        * way of trying to unparkSuccessor of head if it needs
        * signal. But if it does not, status is set to PROPAGATE to
        * ensure that upon release, propagation continues.
        * Additionally, we must loop in case a new node is added
        * while we are doing this. Also, unlike other uses of
        * unparkSuccessor, we need to know if CAS to reset status
        * fails, if so rechecking.
        */
    // 无限循环
    for (;;) {
        // 保存头结点
        Node h = head;
        if (h != null && h != tail) { // 头结点不为空并且头结点不为尾结点
            // 获取头结点的等待状态
            int ws = h.waitStatus; 
            if (ws == Node.SIGNAL) { // 状态为SIGNAL
                if (!compareAndSetWaitStatus(h, Node.SIGNAL, 0)) // 不成功就继续
                    continue;            // loop to recheck cases
                // 释放后继结点
                unparkSuccessor(h);
            }
            else if (ws == 0 &&
                        !compareAndSetWaitStatus(h, 0, Node.PROPAGATE)) // 状态为0并且不成功,继续
                continue;                // loop on failed CAS
        }
        if (h == head) // 若头结点改变,继续循环  
            break;
    }
}

该方法在共享模式下释放,具体的流程再之后会通过一个示例给出。

所以,对CountDownLatch的await调用大致会有如下的调用链。

NAME

上图给出了可能会调用到的主要方法,并非一定会调用到,之后,会通过一个示例给出详细的分析。

核心函数:countDown

此函数将递减锁存器的计数,如果计数到达零,则释放所有等待的线程

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

对countDown的调用转换为对Sync对象的releaseShared(从AQS继承而来)方法的调用。

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

此函数会以共享模式释放对象,并且在函数中会调用到CountDownLatch的tryReleaseShared函数,并且可能会调用AQS的doReleaseShared函数。

  • tryReleaseShared源码如下
protected boolean tryReleaseShared(int releases) {
    // Decrement count; signal when transition to zero
    // 无限循环
    for (;;) {
        // 获取状态
        int c = getState();
        if (c == 0) // 没有被线程占有
            return false;
        // 下一个状态
        int nextc = c-1;
        if (compareAndSetState(c, nextc)) // 比较并且设置成功
            return nextc == 0;
    }
}

此函数会试图设置状态来反映共享模式下的一个释放。具体的流程在下面的示例中会进行分析。

  • AQS的doReleaseShared的源码如下
private void doReleaseShared() {
    /*
        * Ensure that a release propagates, even if there are other
        * in-progress acquires/releases.  This proceeds in the usual
        * way of trying to unparkSuccessor of head if it needs
        * signal. But if it does not, status is set to PROPAGATE to
        * ensure that upon release, propagation continues.
        * Additionally, we must loop in case a new node is added
        * while we are doing this. Also, unlike other uses of
        * unparkSuccessor, we need to know if CAS to reset status
        * fails, if so rechecking.
        */
    // 无限循环
    for (;;) {
        // 保存头结点
        Node h = head;
        if (h != null && h != tail) { // 头结点不为空并且头结点不为尾结点
            // 获取头结点的等待状态
            int ws = h.waitStatus; 
            if (ws == Node.SIGNAL) { // 状态为SIGNAL
                if (!compareAndSetWaitStatus(h, Node.SIGNAL, 0)) // 不成功就继续
                    continue;            // loop to recheck cases
                // 释放后继结点
                unparkSuccessor(h);
            }
            else if (ws == 0 &&
                        !compareAndSetWaitStatus(h, 0, Node.PROPAGATE)) // 状态为0并且不成功,继续
                continue;                // loop on failed CAS
        }
        if (h == head) // 若头结点改变,继续循环  
            break;
    }
}

此函数在共享模式下释放资源。

所以,对CountDownLatch的countDown调用大致会有如下的调用链。

NAME

上图给出了可能会调用到的主要方法,并非一定会调用到,之后,会通过一个示例给出详细的分析。

应用示例

import java.util.concurrent.CountDownLatch;

class MyThread extends Thread {
    private CountDownLatch countDownLatch;
    
    public MyThread(String name, CountDownLatch countDownLatch) {
        super(name);
        this.countDownLatch = countDownLatch;
    }
    
    public void run() {
        System.out.println(Thread.currentThread().getName() + " doing something");
        try {
            Thread.sleep(1000);
        } catch (InterruptedException e) {
            e.printStackTrace();
        }
        System.out.println(Thread.currentThread().getName() + " finish");
        countDownLatch.countDown();
    }
}

public class CountDownLatchDemo {
    public static void main(String[] args) {
        CountDownLatch countDownLatch = new CountDownLatch(2);
        MyThread t1 = new MyThread("t1", countDownLatch);
        MyThread t2 = new MyThread("t2", countDownLatch);
        t1.start();
        t2.start();
        System.out.println("Waiting for t1 thread and t2 thread to finish");
        try {
            countDownLatch.await();
        } catch (InterruptedException e) {
            e.printStackTrace();
        }            
        System.out.println(Thread.currentThread().getName() + " continue");        
    }
}
Waiting for t1 thread and t2 thread to finish
t1 doing something
t2 doing something
t1 finish
t2 finish
main continue

本程序首先计数器初始化为2。根据结果,可能会存在如下的一种时序图。

NAME

首先main线程会调用await操作,此时main线程会被阻塞,等待被唤醒,之后t1线程执行了countDown操作,最后,t2线程执行了countDown操作,此时main线程就被唤醒了,可以继续运行。下面,进行详细分析。

  • main线程执行countDownLatch.await操作,主要调用的函数如下。
NAME

在最后,main线程就被park了,即禁止运行了。此时Sync queue(同步队列)中有两个节点,AQS的state为2,包含main线程的结点的nextWaiter指向SHARED结点。

  • t1线程执行countDownLatch.countDown操作,主要调用的函数如下。
NAME

此时,Sync queue队列里的结点个数未发生变化,但是此时,AQS的state已经变为1了。

  • t2线程执行countDownLatch.countDown操作,主要调用的函数如下。
NAME

经过调用后,AQS的state为0,并且此时,main线程会被unpark,可以继续运行。当main线程获取cpu资源后,继续运行。

  • main线程获取cpu资源,继续运行,由于main线程是在parkAndCheckInterrupt函数中被禁止的,所以此时,继续在parkAndCheckInterrupt函数运行.
NAME

main线程恢复,继续在parkAndCheckInterrupt函数中运行,之后又会回到最终达到的状态为AQS的state为0,并且head与tail指向同一个结点,该节点的额nextWaiter域还是指向SHARED结点。