CLH锁原理

CLH锁其实是为了优化自旋锁的缺点：

所有自旋锁都在一个内存地址上自旋，持有锁的线程释放锁后，会引发惊群效应，而且会造成个别线程一只拿不到锁，处在饥饿状态,CLH锁通过队列将所有线程排队，避免惊群效应，也保证所有线程都能执行。

简单的自旋锁实现：

public class SpinLock {

    private AtomicReference owner = new AtomicReference();


    public boolean lock() throws InterruptedException {
        Thread currThread = Thread.currentThread();

        while (!owner.compareAndSet(null,currThread)){
            System.out.println("本次未获得锁，继续自旋获取锁,thread name:"+currThread.getName());
            Thread.sleep(1000);
        }
        System.out.println("获得锁成功,thread name:"+currThread.getName());
        return true;
    }

    public boolean unlock() {
        Thread currThread = Thread.currentThread();
        if (owner.compareAndSet(currThread,null)){
            System.out.println("释放锁成功,thread name:"+currThread.getName());
            return true;
        }else {
            return false;
        }
    }

    public static void main(String[] args) throws InterruptedException {
        SpinLock lock = new SpinLock();
        Thread t1 = new Thread(()->{
            try {
                if(lock.lock()){
                    System.out.println("获得锁，任务开始, t:"+ Thread.currentThread().getName());
                    Thread.sleep(1000);
                    lock.unlock();
                    System.out.println("任务结束,释放锁 t:"+ Thread.currentThread().getName());
                }
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        });


        Thread t2 = new Thread(()->{
            try {
                if(lock.lock()){
                    System.out.println("Get the SpinLock, t:"+ Thread.currentThread().getName());
                    Thread.sleep(1000);
                    lock.unlock();
                }
            } catch (InterruptedException e) {
                e.printStackTrace();
            }
        });

        t1.start();
        t2.start();

    }

}

CLH原理：

将线程包装为CLH节点，locked代表线程状态，false表示释放锁或未持有锁，true表示持有锁或正在排队等待锁。

    @Data
    private static class Node{
        volatile boolean locked = false;
        private Thread curThread;
        private Node preNode;
    }

获得锁：

1.设置一个尾节点tail，初始为一个空节点，节点状态为false；

2.新加入的线程直接到队尾排队，这一步通过CAS保证成功；

3.将前一个节点set进当前节点的属性;

4.自旋监听前一个节点的状态是否为释放锁；

5.获得锁；

释放锁：

1.将当前节点的状态设置为false，表示锁已释放

2.清空前一个节点，直接设置为null（help GC）

public class CLHLock {

    @Data
    private static class Node{
        volatile boolean locked = false;
        private Thread curThread;
        private Node preNode;
    }

    AtomicReference tail =new  AtomicReference(new Node());

    ThreadLocal currNode = new ThreadLocal();

    public boolean lock(Thread t){
        Node currentNode = new Node();
        currentNode.setCurThread(t);
        Node preNode = tail.getAndSet(currentNode);
        currentNode.setPreNode(preNode);
        currentNode.setLocked(true);
        while (currentNode.getPreNode().locked);
        System.out.println("线程"+t.getName()+"获得锁");
        currNode.set(currentNode);
        return true;
    }

    public boolean unlock(Thread t){
        Node currentNode = currNode.get();
        if(null==currentNode){
            return false;
        }
        currentNode.setLocked(false);
        System.out.println("线程"+t.getName()+"释放锁成功");
        currentNode.setPreNode(null);
        //用完要释放，避免内存泄露
        currNode.remove();
        return true;
    }

    private static int counter = 0;
    public static void main(String[] args) {

        CLHLock clhLock = new CLHLock();

        Runnable task = new Runnable() {
            @Override
            public void run() {
                Thread t = Thread.currentThread();
                if (clhLock.lock(t)) {
                    counter++;
                    System.out.println("线程" + t.getName() + "执行加法成功，counter:" + counter);
                    clhLock.unlock(t);
                }
            }
        };

        Thread t1 = new Thread(task);
        Thread t2 = new Thread(task);
        Thread t3 = new Thread(task);
        Thread t4 = new Thread(task);
        Thread t5 = new Thread(task);
        t1.start();
        t2.start();
        t3.start();
        t4.start();
        t5.start();
    }
}

AQS论文地址 https://gee.cs.oswego.edu/dl/papers/aqs.pdf

AQS源码分析https://javadoop.com/post/AbstractQueuedSynchronizer