1.概述

ReentrantLock、CountDownLatch奕扣、Semaphore底層都是基于AQS實(shí)現(xiàn)的薪鹦，其中tryAcquire、tryRelease惯豆、tryAcquireShared池磁、tryReleaseShared即加鎖以及釋放鎖的邏輯則由這三個(gè)子類自己實(shí)現(xiàn)。AQS的實(shí)現(xiàn)細(xì)節(jié)詳見AbstractQueuedSynchronizer（AQS）中獨(dú)占模式與共享模式的設(shè)計(jì)與實(shí)現(xiàn)循帐。

2.ReentrantLock

2.1 ReentrantLock中加鎖邏輯的實(shí)現(xiàn)

ReentrantLock中擁有Sync類型的內(nèi)部類框仔，其中Sync繼承AQS，采用獨(dú)占模式下的AQS拄养，Sync的子類包括NonfairSync（非公平）與FairSync（公平）兩種模式，所有對(duì)AQS的操作便通過Sync進(jìn)行。在初始化的時(shí)候會(huì)實(shí)例化一個(gè)Sync對(duì)象瘪匿，默認(rèn)實(shí)現(xiàn)為非公平模式跛梗。

    /**
     * Creates an instance of {@code ReentrantLock}.
     * This is equivalent to using {@code ReentrantLock(false)}.
     */
    public ReentrantLock() {
        sync = new NonfairSync();
    }

同時(shí)也可以指定Sync的實(shí)現(xiàn)為公平模式or非公平模式

    /**
     * Creates an instance of {@code ReentrantLock} with the
     * given fairness policy.
     *
     * @param fair {@code true} if this lock should use a fair ordering policy
     */
    public ReentrantLock(boolean fair) {
        sync = fair ? new FairSync() : new NonfairSync();
    }

2.1.1 ReentrantLock非公平模式下的加鎖實(shí)現(xiàn)

非公平模式下獲取鎖的邏輯，線程第一次進(jìn)入的時(shí)候會(huì)利用cas操作更新state狀態(tài)為1棋弥，如果更新成功核偿，則設(shè)置當(dāng)前線程為owner；當(dāng)cas操作失斖缛尽（包括兩種情況1.重入鎖邏輯：已經(jīng)獲取鎖的線程再次調(diào)用lock方法漾岳；2.當(dāng)前線程持有鎖，其他線程調(diào)用lock方法粉寞，即持有鎖的線程和調(diào)用lock方法的線程不是同一個(gè)）尼荆，則進(jìn)入到acquire方法中，acquire會(huì)優(yōu)先調(diào)用子類的tryAcquire方法唧垦，具體獲取鎖的實(shí)現(xiàn)就是在該方法中捅儒。

/**
 * Sync object for non-fair locks
 */
static final class NonfairSync extends Sync {
    private static final long serialVersionUID = 7316153563782823691L;

    /**
     * Performs lock.  Try immediate barge, backing up to normal
     * acquire on failure.
     */
    final void lock() {
        /*第一次獲取鎖時(shí)，設(shè)置state為1*/
        if (compareAndSetState(0, 1))
            /*設(shè)置當(dāng)前線程為owner*/
            setExclusiveOwnerThread(Thread.currentThread());
        else
            /*搶奪鎖以及重入鎖邏輯*/
            acquire(1);
    }

    /*獲取鎖的具體實(shí)現(xiàn)*/
    protected final boolean tryAcquire(int acquires) {
        return nonfairTryAcquire(acquires);
    }
}

tryAcquire中調(diào)用nonfairTryAcquire方法振亮，采用cas操作更新state的值巧还，成功后就設(shè)置當(dāng)前線程為owner，代表獲取鎖成功坊秸；如果當(dāng)前線程等于owner麸祷，就累加state值（重入鎖的實(shí)現(xiàn)邏輯），多次調(diào)用lock方法加鎖褒搔，解鎖時(shí)也要調(diào)用同樣次數(shù)的unlock方法阶牍，直到state值減為0。

/**
 * Performs non-fair tryLock.  tryAcquire is implemented in
 * subclasses, but both need nonfair try for trylock method.
 */
final boolean nonfairTryAcquire(int acquires) {
    /*獲取當(dāng)前線程*/
    final Thread current = Thread.currentThread();
    /*獲取state值*/
    int c = getState();
    if (c == 0) {
        /*cas操作更新state狀態(tài)*/
        if (compareAndSetState(0, acquires)) {
            /*更新成功后設(shè)置當(dāng)前線程為owner*/
            setExclusiveOwnerThread(current);
            return true;
        }
    }
    /*如果當(dāng)前線程是onwer（重入鎖邏輯）*/
    else if (current == getExclusiveOwnerThread()) {
        /*獲取state的值站超，并且加上本次的acquires再更新state*/
        int nextc = c + acquires;
        if (nextc < 0) // overflow
            throw new Error("Maximum lock count exceeded");
        setState(nextc);
        return true;
    }
    /*獲取鎖失敗*/
    return false;
}

2.1.2 ReentrantLock公平模式下的加鎖實(shí)現(xiàn)

公平模式下的加鎖實(shí)現(xiàn)與非公平的比較在于多了hasQueuedPredecessors方法的判斷荸恕，該方法主要是判斷阻塞隊(duì)列的隊(duì)首是否含有元素以及如果隊(duì)首有元素是否與當(dāng)前線程相等，目的其實(shí)就是檢測(cè)有沒有排在當(dāng)前線程之前的線程需要出隊(duì)死相，嚴(yán)格保證先進(jìn)先出融求。tryAcquire除了在阻塞線程被喚醒的時(shí)候會(huì)被調(diào)用外，外部代碼調(diào)用lock方法的時(shí)候也會(huì)間接調(diào)用到算撮，當(dāng)外部代碼調(diào)用lock時(shí)生宛，就會(huì)通過hasQueuedPredecessors方法校驗(yàn)阻塞隊(duì)列中是否還有等待中的線程，如果有肮柜，就不會(huì)讓該線程獲取鎖陷舅。

/**
 * Sync object for fair locks
 */
static final class FairSync extends Sync {
    private static final long serialVersionUID = -3000897897090466540L;

    final void lock() {
        acquire(1);
    }

    /**
     * Fair version of tryAcquire.  Don't grant access unless
     * recursive call or no waiters or is first.
     */
    protected final boolean tryAcquire(int acquires) {
        final Thread current = Thread.currentThread();
        int c = getState();
        if (c == 0) {
            /*hasQueuedPredecessors用于判斷AQS的阻塞隊(duì)列里是否有等待中的線程*/
            if (!hasQueuedPredecessors() &&
                compareAndSetState(0, acquires)) {
                setExclusiveOwnerThread(current);
                return true;
            }
        }
        else if (current == getExclusiveOwnerThread()) {
            int nextc = c + acquires;
            if (nextc < 0)
                throw new Error("Maximum lock count exceeded");
            setState(nextc);
            return true;
        }
        return false;
    }
}

hasQueuedPredecessors的實(shí)現(xiàn)

    public final boolean hasQueuedPredecessors() {
        // The correctness of this depends on head being initialized
        // before tail and on head.next being accurate if the current
        // thread is first in queue.
        Node t = tail; // Read fields in reverse initialization order
        Node h = head;
        Node s;
        return h != t &&
            ((s = h.next) == null || s.thread != Thread.currentThread());
    }

2.2 ReentrantLock釋放鎖邏輯

公平模式與非公平模式下釋放鎖的實(shí)現(xiàn)都是相同的，tryRelease方法在Sync類中實(shí)現(xiàn)审洞。釋放鎖的過程就是減少state的值直到0（因此在重入鎖的情況下莱睁，調(diào)用unlock的次數(shù)要與lock的次數(shù)相同），同時(shí)將owner置空。

protected final boolean tryRelease(int releases) {
    /*獲取state值并減少對(duì)應(yīng)次數(shù)*/
    int c = getState() - releases;
    /*判斷當(dāng)前線程是否為owner*/
    if (Thread.currentThread() != getExclusiveOwnerThread())
        throw new IllegalMonitorStateException();
    boolean free = false;
    /*state減為0后仰剿，設(shè)置owner為null*/
    if (c == 0) {
        free = true;
        setExclusiveOwnerThread(null);
    }
    setState(c);
    return free;
}

2.3 ReentrantLock中lockInterruptibly響應(yīng)中斷原理

當(dāng)線程調(diào)用lock方法后并處于阻塞狀態(tài)下创淡，該線程是無法響應(yīng)中斷的，只有獲取到鎖之后才能響應(yīng)中斷南吮，但是使用lockInterruptibly就能在阻塞狀態(tài)下響應(yīng)中斷琳彩。lockInterruptibly方法調(diào)用acquireInterruptibly（該方法在AQS中實(shí)現(xiàn)）

public final void acquireInterruptibly(int arg)
        throws InterruptedException {
    /*判斷是否被中斷過，如果被中斷過則直接拋出異常*/
    if (Thread.interrupted())
        throw new InterruptedException();
    /*獲取鎖失敗則調(diào)用doAcquireInterruptibly方法*/
    if (!tryAcquire(arg))
        doAcquireInterruptibly(arg);
}

重點(diǎn)關(guān)注doAcquireInterruptibly方法部凑，該方法的實(shí)現(xiàn)與acquireQueued非常相似（acquireQueued實(shí)現(xiàn)可以參考AQS）露乏，區(qū)別在于阻塞狀態(tài)下的線程因?yàn)橹袛啾粏拘褧r(shí)，parkAndCheckInterrupt返回true涂邀，就會(huì)拋出InterruptedException異常瘟仿。

/**
 * Acquires in exclusive interruptible mode.
 * @param arg the acquire argument
 */
private void doAcquireInterruptibly(int arg)
    throws InterruptedException {
    final Node node = addWaiter(Node.EXCLUSIVE);
    boolean failed = true;
    try {
        for (;;) {
            final Node p = node.predecessor();
            if (p == head && tryAcquire(arg)) {
                setHead(node);
                p.next = null; // help GC
                failed = false;
                return;
            }
            if (shouldParkAfterFailedAcquire(p, node) &&
                parkAndCheckInterrupt())
                /*線程因?yàn)橹袛啾粏拘押笾苯訏伋霎惓?/
                throw new InterruptedException();
        }
    } finally {
        if (failed)
            cancelAcquire(node);
    }
}

3.CountDownLatch

CountDownLatch同樣也擁有一個(gè)Sync類型的內(nèi)部類，其中Sync繼承AQS必孤，使用的是共享模式下的AQS猾骡，Sync的子類只有一個(gè)，在初始化的時(shí)候?qū)嵗?code>Sync的實(shí)現(xiàn)如下敷搪，包括構(gòu)造函數(shù)（給state賦初始化值）兴想，獲取鎖邏輯（判斷state值是否為0，為0則返回1否則返回-1）赡勘，釋放鎖邏輯（state值減少1）嫂便。

/**
 * Synchronization control For CountDownLatch.
 * Uses AQS state to represent count.
 */
private static final class Sync extends AbstractQueuedSynchronizer {
    private static final long serialVersionUID = 4982264981922014374L;

    /*初始化的時(shí)候state的值*/
    Sync(int count) {
        setState(count);
    }

    int getCount() {
        return getState();
    }

    /*獲取鎖邏輯*/
    protected int tryAcquireShared(int acquires) {
        return (getState() == 0) ? 1 : -1;
    }

    /*釋放鎖邏輯*/
    protected boolean tryReleaseShared(int releases) {
        // Decrement count; signal when transition to zero
        for (;;) {
            /*獲取state狀態(tài)*/
            int c = getState();
            if (c == 0)
                return false;
            /*state值減1并使用cas更新*/
            int nextc = c-1;
            if (compareAndSetState(c, nextc))
                return nextc == 0;
        }
    }
}

2.1.1 CountDownLatch初始化構(gòu)造

CountDownLatch初始化的時(shí)候會(huì)根據(jù)傳入的值設(shè)置state的值，當(dāng)調(diào)用countDown方法的時(shí)候會(huì)把該值減少1闸与，直到減為0的時(shí)候毙替，阻塞中的線程就能繼續(xù)執(zhí)行。

public CountDownLatch(int count) {
    /*校驗(yàn)count合法性*/
    if (count < 0) throw new IllegalArgumentException("count < 0");
    /*設(shè)置state初始化值*/
    this.sync = new Sync(count);
}

2.1.2 await方法實(shí)現(xiàn)原理

調(diào)用await方法時(shí)會(huì)調(diào)用acquireSharedInterruptibly践樱，該方法會(huì)首先判斷當(dāng)前線程是否有中斷標(biāo)記厂画，如果有則直接拋出異常。接著會(huì)調(diào)用tryAcquireShared方法獲取鎖拷邢，該方法通過state是否等于0決定返回1還是-1袱院，因?yàn)樵跇?gòu)造函數(shù)中，會(huì)初始化state的值為傳遞進(jìn)去的數(shù)字瞭稼，所以在沒有調(diào)用countDown方法的前提下（同時(shí)初始化傳遞的值為非0）忽洛，tryAcquireShared獲取鎖是一定會(huì)失敗的，因此返回-1环肘，接著進(jìn)入AQS提供的doAcquireSharedInterruptibly方法中欲虚。

public final void acquireSharedInterruptibly(int arg)
        throws InterruptedException {
    /*如果有中斷標(biāo)記則直接拋出異常*/
    if (Thread.interrupted())
        throw new InterruptedException();
    /*獲取鎖失敗則進(jìn)入阻塞隊(duì)列*/
    if (tryAcquireShared(arg) < 0)
        /*AQS方法，讓當(dāng)前線程進(jìn)入阻塞隊(duì)列并阻塞*/
        doAcquireSharedInterruptibly(arg);
}

doAcquireSharedInterruptibly在該場(chǎng)景下主要完成兩件事情悔雹，1.將當(dāng)前線程加入到阻塞隊(duì)列中复哆；2.阻塞當(dāng)前線程欣喧。doAcquireSharedInterruptibly與doAcquireShared(doAcquireShared實(shí)現(xiàn)參考)的實(shí)現(xiàn)很相似，區(qū)別在于doAcquireSharedInterruptibly如果因?yàn)橹袛啾粏拘鸭盘瘢蜁?huì)直接拋出異常续誉，該方法是能響應(yīng)中斷的莱没。

/**
 * Acquires in shared interruptible mode.
 * @param arg the acquire argument
 */
private void doAcquireSharedInterruptibly(int arg)
    throws InterruptedException {
    /*進(jìn)入阻塞隊(duì)列*/    
    final Node node = addWaiter(Node.SHARED);
    boolean failed = true;
    try {
        for (;;) {
            final Node p = node.predecessor();
            if (p == head) {
                int r = tryAcquireShared(arg);
                if (r >= 0) {
                    setHeadAndPropagate(node, r);
                    p.next = null; // help GC
                    failed = false;
                    return;
                }
            }
            if (shouldParkAfterFailedAcquire(p, node) &&
                parkAndCheckInterrupt())
                /*如果是因?yàn)橹袛啾粏拘丫€程則直接拋出異常*/
                throw new InterruptedException();
        }
    } finally {
        if (failed)
            cancelAcquire(node);
    }
}

因此當(dāng)前線程調(diào)用await方法的時(shí)候初肉，如果state不為0時(shí)就會(huì)被阻塞，等待其他線程調(diào)用countDown直到state減少到0才會(huì)繼續(xù)執(zhí)行下去饰躲，同時(shí)牙咏，如果阻塞中的線程被調(diào)用中斷方法，則會(huì)直接拋出InterruptedException異常嘹裂。

2.1.3 countDown方法實(shí)現(xiàn)原理

countDown方法最終會(huì)調(diào)用CountDownLatch實(shí)現(xiàn)的tryReleaseShared釋放鎖方法妄壶，該方法會(huì)將state中的值減少1，并返回state是否等于0的信息寄狼。當(dāng)調(diào)用countDown方法直到state減少到0的時(shí)候丁寄，tryReleaseShared就會(huì)返回true，通過調(diào)用AQS中的doReleaseShared方法泊愧，喚醒阻塞隊(duì)列中因?yàn)檎{(diào)用await所阻塞的線程伊磺。

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

public final boolean releaseShared(int arg) {
    /*獲取鎖成功則喚醒等待隊(duì)列中的線程*/
    if (tryReleaseShared(arg)) {
        doReleaseShared();
        return true;
    }
    return false;
}

3.Semaphore

3.1 Semaphore的構(gòu)造方法

Semaphore同樣擁有Sync類型的內(nèi)部類，其中Sync繼承AQS删咱，采用共享模式下的AQS屑埋，Sync的子類與ReentrantLock類似，包括NonfairSync（非公平）與FairSync（公平）兩種模式痰滋，默認(rèn)實(shí)現(xiàn)為非公平模式摘能。

public Semaphore(int permits) {
    sync = new NonfairSync(permits);
}

同時(shí)也能指定使用哪一種模式

public Semaphore(int permits, boolean fair) {
    sync = fair ? new FairSync(permits) : new NonfairSync(permits);
}

Semaphore初始化的時(shí)候，會(huì)傳遞permits敲街，該值會(huì)被設(shè)置為state的值团搞。

3.2 acquire方法實(shí)現(xiàn)原理

acquire會(huì)調(diào)用acquireSharedInterruptibly方法，tryAcquireShared方法（獲取鎖邏輯）由Semaphore自己實(shí)現(xiàn)多艇。如果獲取鎖失敗逻恐，則進(jìn)入阻塞隊(duì)列。

public final void acquireSharedInterruptibly(int arg)
        throws InterruptedException {
    /*發(fā)現(xiàn)中斷標(biāo)記則直接拋出異常*/        
    if (Thread.interrupted())
        throw new InterruptedException();
    /*獲取鎖失敗則進(jìn)入阻塞隊(duì)列*/
    if (tryAcquireShared(arg) < 0)
        doAcquireSharedInterruptibly(arg);
}

tryAcquireShared有兩種不同的實(shí)現(xiàn)墩蔓，先說在非公平模式下的實(shí)現(xiàn)梢莽，直接減去acquires的大小并更新state

final int nonfairTryAcquireShared(int acquires) {
    for (;;) {
        /*獲取state的值*/
        int available = getState();
        /*較少acquires的值并更新state*/
        int remaining = available - acquires;
        if (remaining < 0 ||
            compareAndSetState(available, remaining))
            return remaining;
    }
}

公平模式下獲取鎖的邏輯相對(duì)于非公平模式下的多了hasQueuedPredecessors函數(shù)判斷出隊(duì)的元素是否位于隊(duì)首，如果是外部代碼調(diào)用的奸披，則判斷阻塞隊(duì)列中是否還有未出隊(duì)的元素昏名，此處邏輯與ReentrantLock的公平鎖實(shí)現(xiàn)邏輯相似。

protected int tryAcquireShared(int acquires) {
    for (;;) {
        if (hasQueuedPredecessors())
            return -1;
        int available = getState();
        int remaining = available - acquires;
        if (remaining < 0 ||
            compareAndSetState(available, remaining))
            return remaining;
    }
}

3.3 release方法實(shí)現(xiàn)原理

release方法會(huì)調(diào)用releaseShared阵面，其中tryReleaseShared（釋放鎖）成功后就會(huì)通過AQS提供的doReleaseShared方法喚醒阻塞隊(duì)列中的線程轻局。

public final boolean releaseShared(int arg) {
    /*成功釋放則喚醒阻塞隊(duì)列中的線程*/
    if (tryReleaseShared(arg)) {
        doReleaseShared();
        return true;
    }
    return false;
}

tryReleaseShared的邏輯非常簡(jiǎn)單洪鸭，就是增加對(duì)應(yīng)的releases值并更新state

protected final boolean tryReleaseShared(int releases) {
    for (;;) {
        int current = getState();
        /*增加state的值并更新*/
        int next = current + releases;
        if (next < current) // overflow
            throw new Error("Maximum permit count exceeded");
        if (compareAndSetState(current, next))
            return true;
    }
}

4.總結(jié)

對(duì)于ReentrantLock、CountDownLatch仑扑、Semaphore實(shí)現(xiàn)加鎖解鎖邏輯最重要的變量是AQS中的state變量览爵。

• ReentrantLock加鎖流程就是將state增加1，解鎖便是減1镇饮，所以調(diào)用多次lock方法就要調(diào)用對(duì)應(yīng)次數(shù)的unlock方法才能解鎖蜓竹。
• CountDownLatch初始化的時(shí)候設(shè)置state的值，當(dāng)其他線程調(diào)用countDown的時(shí)候就減少state的值储藐，直到state減少為0的時(shí)候俱济，因?yàn)檎{(diào)用await方法而在阻塞隊(duì)列中等待的線程，就會(huì)在最后一次調(diào)用countDown的時(shí)候被喚醒钙勃。
• Semaphore同樣也是初始化的時(shí)候設(shè)置state的值蛛碌，調(diào)用acquire成功獲取鎖的時(shí)候就會(huì)減少state的值，如果state減少到0則無法再成功獲取鎖辖源，線程就會(huì)進(jìn)入到阻塞隊(duì)列中蔚携；調(diào)用release釋放鎖的時(shí)候會(huì)增加state的值。