ReentrantLock

很常用的一个锁

一、可重入原理

//TODO acquireQueued方法还不算熟悉
//公平策略
static final class FairSync extends Sync {
	private static final long serialVersionUID = -3000897897090466540L;

	final void lock() {
		//尝试抢占
		acquire(1);
	}
	//tryAcquire返回false会进入
	public final void acquire(int arg) {
		if (!tryAcquire(arg) &&
				//加入一个新节点进入CLH
				acquireQueued(addWaiter(Node.EXCLUSIVE), arg))
			selfInterrupt();
	}

	//可重入原理:如果没人占有锁,就cas尝试占有,否则如果是当前线程，state++,
	//会检查,由此可以看到嵌套的最大层数不能超过int,两者都失败会返回false,
	protected final boolean tryAcquire(int acquires) {
		final Thread current = Thread.currentThread();
		int c = getState();
		if (c == 0) {
			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;
	}
}

static final class NonfairSync extends Sync {
	private static final long serialVersionUID = 7316153563782823691L;

	/**
	* 很鲁莽，上来就想抢占锁，把自己设置为占有锁的线程，否则尝试抢占
	*/
	final void lock() {
		if (compareAndSetState(0, 1))
			setExclusiveOwnerThread(Thread.currentThread());
		else
			acquire(1);
	}

	protected final boolean tryAcquire(int acquires) {
		return nonfairTryAcquire(acquires);
	}
}

二、Condition 的使用

//TODO