使用分析
四种读写情况:读读 读写 写读 写写, 接下来分四种情况看最终结果
情况一:读读使用
public static void main(String[] args) throws InterruptedException { ReentrantReadWriteLock readWriteLock = new ReentrantReadWriteLock(); ReentrantReadWriteLock.ReadLock readLock = readWriteLock.readLock(); ReentrantReadWriteLock.WriteLock writeLock = readWriteLock.writeLock(); new Thread( ()->{ readLock.lock(); try { System.out.println("t1 try get readLock"); Thread.sleep(2000L); System.out.println("t1 got readLock"); }catch (Exception e){ }finally { readLock.unlock(); System.out.println("t1 relase readLock"); } }, "t1").start(); new Thread( ()->{ readLock.lock(); try { System.out.println("t2 try get readLock"); Thread.sleep(1000L); System.out.println("t2 got readLock"); }catch (Exception e){ }finally { readLock.unlock(); System.out.println("t2 relase readLock"); } }, "t2").start(); } 运行结果: t1 try get readLock t2 try get readLock t2 got readLock t2 relase readLock t1 got readLock t1 relase readLock
从运行结果可知: 读读非互斥
情况二:读写使用
public static void main(String[] args) throws InterruptedException { ReentrantReadWriteLock readWriteLock = new ReentrantReadWriteLock(); ReentrantReadWriteLock.ReadLock readLock = readWriteLock.readLock(); ReentrantReadWriteLock.WriteLock writeLock = readWriteLock.writeLock(); new Thread( ()->{ readLock.lock(); try { System.out.println("t1 try get readLock"); Thread.sleep(2000L); System.out.println("t1 got readLock"); }catch (Exception e){ }finally { readLock.unlock(); System.out.println("t1 relase readLock"); } }, "t1").start(); new Thread( ()->{ writeLock.lock(); try { System.out.println("t2 try get writeLock"); Thread.sleep(1000L); System.out.println("t2 got writeLock"); }catch (Exception e){ }finally { writeLock.unlock(); System.out.println("t2 relase writeLock"); } }, "t2").start(); } 打印结果: t1 try get readLock t1 got readLock t1 relase readLock t2 try get writeLock t2 got writeLock t2 relase writeLock
从运行结果可知: 读写互斥
情况三:写读使用
public static void main(String[] args) throws InterruptedException { ReentrantReadWriteLock readWriteLock = new ReentrantReadWriteLock(); ReentrantReadWriteLock.ReadLock readLock = readWriteLock.readLock(); ReentrantReadWriteLock.WriteLock writeLock = readWriteLock.writeLock(); new Thread( ()->{ writeLock.lock(); try { System.out.println("t1 try get writeLock"); Thread.sleep(2000L); System.out.println("t1 got writeLock"); }catch (Exception e){ }finally { writeLock.unlock(); System.out.println("t1 relase writeLock"); } }, "t1").start(); new Thread( ()->{ readLock.lock(); try { System.out.println("t2 try get readLock"); Thread.sleep(1000L); System.out.println("t2 got readLock"); }catch (Exception e){ }finally { readLock.unlock(); System.out.println("t2 relase readLock"); } }, "t2").start(); } 打印结果: t1 try get writeLock t1 got writeLock t1 relase writeLock t2 try get readLock t2 got readLock t2 relase readLock
从运行结果可知: 写读互斥
情况四:写写使用
public static void main(String[] args) throws InterruptedException { ReentrantReadWriteLock readWriteLock = new ReentrantReadWriteLock(); ReentrantReadWriteLock.ReadLock readLock = readWriteLock.readLock(); ReentrantReadWriteLock.WriteLock writeLock = readWriteLock.writeLock(); new Thread( ()->{ writeLock.lock(); try { System.out.println("t1 try get writeLock"); Thread.sleep(2000L); System.out.println("t1 got writeLock"); }catch (Exception e){ }finally { writeLock.unlock(); System.out.println("t1 relase writeLock"); } }, "t1").start(); new Thread( ()->{ writeLock.lock(); try { System.out.println("t2 try get writeLock"); Thread.sleep(1000L); System.out.println("t2 got writeLock"); }catch (Exception e){ }finally { writeLock.unlock(); System.out.println("t2 relase writeLock"); } }, "t2").start(); } 打印结果: t1 try get writeLock t1 got writeLock t1 relase writeLock t2 try get writeLock t2 got writeLock t2 relase writeLock
从运行结果可知: 写写互斥
使用总结: 除读读共享锁,其他情况表现为互斥
源码分析
构造方法分析
以下代码可以看出ReadLock WriteLock都是依赖ReentrantReadWriteLock类的同步器进行锁操作, 下面重点关注ReentrantReadWriteLock类中Sync的代码
ReentrantReadWriteLock类 public class ReentrantReadWriteLock implements ReadWriteLock, java.io.Serializable { private static final long serialVersionUID = -6992448646407690164L; //读锁 private final ReentrantReadWriteLock.ReadLock readerLock; //写锁 private final ReentrantReadWriteLock.WriteLock writerLock; //同步器 获取锁和释放锁具体实现类 final Sync sync; //默认构造器 public ReentrantReadWriteLock() { this(false); } //实际构造器 //初始化上面三个参数 readerLock writerLock sync public ReentrantReadWriteLock(boolean fair) { sync = fair ? new FairSync() : new NonfairSync(); readerLock = new ReadLock(this); writerLock = new WriteLock(this); } public ReentrantReadWriteLock.WriteLock writeLock() { return writerLock; } public ReentrantReadWriteLock.ReadLock readLock() { return readerLock; } } ReadLock类 public static class ReadLock implements Lock, java.io.Serializable { private final Sync sync; //使用ReentrantReadWriteLock.sync赋值给ReadLock.sync protected ReadLock(ReentrantReadWriteLock lock) { sync = lock.sync; } } WriteLock类 public static class WriteLock implements Lock, java.io.Serializable { private final Sync sync; //使用ReentrantReadWriteLock.WriteLock.sync protected WriteLock(ReentrantReadWriteLock lock) { sync = lock.sync; } }
Sync类
abstract static class Sync extends AbstractQueuedSynchronizer { static final int SHARED_SHIFT = 16; static final int SHARED_UNIT = (1 << SHARED_SHIFT); static final int MAX_COUNT = (1 << SHARED_SHIFT) - 1; static final int EXCLUSIVE_MASK = (1 << SHARED_SHIFT) - 1; //读锁重入数 static int sharedCount(int c) { return c >>> SHARED_SHIFT; } //写锁重入数 static int exclusiveCount(int c) { return c & EXCLUSIVE_MASK; } //获取当前线程重入次数 //这里比较难理解, ThreadLocalHoldCounter继承自ThreadLocal,每一个线程都有自己重入次数 //这里的作用是将重入次数放在线程缓存中, 当对应执行线程需要获取重入次数时只需要使用get()方法 private transient ThreadLocalHoldCounter readHolds; //最后一个获取读锁的线程 //举例:读读共享锁, 可能存在若干读锁, 读(fist) 读 读 读(last), cachedHoldCounter记录的是last线程读锁重入次数 private transient HoldCounter cachedHoldCounter; //第一个获取读锁的线程 private transient Thread firstReader = null; //第一个获取读锁的线程获取读锁重入数 private transient int firstReaderHoldCount; }
读写锁的核心是把AQS中的state设置为一个32位的值, 高16位表示读锁重入次数, 低16位表示写锁重入次数
读锁获取: 1 若线程需要进入CLH等待队列, 则state的值不变, 例:读写读 写读读
2若是锁持有线程为读锁, 且写状态为0,此时需要修改state的值, 例 读读
读锁释放: state减去 1<<16
写锁获取: 1若持有线程为当前线程state加1(因为写状态为低位直接加1就行)
2若非持有线程,则需要判断锁没有被其他线程持有,state加1
写锁释放: state减去1
1 获取读锁
public void lock() { sync.acquireShared(1); } //获取共享锁 public final void acquireShared(int arg) { if (tryAcquireShared(arg) < 0) doAcquireShared(arg); } //尝试获取共享锁返回值-1表示线程需要加入CLH队列等待 否则放行 protected final int tryAcquireShared(int unused) { Thread current = Thread.currentThread(); int c = getState(); //如果排他锁(写锁)被其他线程(非当前线程),返回失败 if (exclusiveCount(c) != 0 && getExclusiveOwnerThread() != current) return -1; //获取共享锁(读锁)状态 持有锁的线程共享锁重入次数 int r = sharedCount(c); if ( //核心逻辑判断是第一个判断readerShouldBlock, 判断读锁是否需要阻塞 //公平模式判断需要入列 :存在CLH队列,并且当前锁占有线程非当前线程 //非公平模式判断需要入列:增加首节点是否为写锁等待,优先写锁原则,其他类似公平模式 !readerShouldBlock() //重入个数未到最大值 && r < MAX_COUNT //CAS修改值 && compareAndSetState(c, c + SHARED_UNIT)) { //---进入此处代码段则表示锁获取成功 以下代码作用是辅助属性的赋值 //r==0读锁持有者重入次数为0 if (r == 0) { firstReader = current; firstReaderHoldCount = 1; } //如果首个读节点为当前节点 表示重入 计数+1 else if (firstReader == current) { firstReaderHoldCount++; } //等价于 r!=0 且 读锁持有线程非当前线程 //读读共享锁 else { //设置重入次数 HoldCounter rh = cachedHoldCounter; if (rh == null || rh.tid != getThreadId(current)) cachedHoldCounter = rh = readHolds.get(); else if (rh.count == 0) readHolds.set(rh); rh.count++; } return 1; } //补偿获取共享锁 return fullTryAcquireShared(current); }
2 释放读锁
/** ReentrantReadWriteLock.ReadLock 类 */ public void unlock() { sync.releaseShared(1); } /** AbstractQueuedSynchronizer类 */ public final boolean releaseShared(int arg) { if (tryReleaseShared(arg)) { doReleaseShared(); return true; } return false; } /** ReentrantReadWriteLock.ReadLock 类 */ protected final boolean tryReleaseShared(int unused) { Thread current = Thread.currentThread(); //如果第一个读锁为当前锁 if (firstReader == current) { //线程只获取一次读锁 则将firstReader置为空 if (firstReaderHoldCount == 1) firstReader = null; //线程有重入获取读锁 则将重入数减一 else firstReaderHoldCount--; } //如果当前线程非第一个读锁获取者 则从线程缓存中(ThreadLocal原理)中获取重入数, //然后执行清除缓存(未重入) 或者 重入次数减一(有重入) else { HoldCounter rh = cachedHoldCounter; if (rh == null || rh.tid != getThreadId(current)) rh = readHolds.get(); int count = rh.count; if (count <= 1) { readHolds.remove(); if (count <= 0) throw unmatchedUnlockException(); } --rh.count; } //无限循环尝试compareAndSetState (CAS操作) 直到成功 //因为读锁为共享锁 在执行CAS操作的同时可能会存在别的获取读锁线程正在获取或释放锁 //所以有可能失败的 就需要再次尝试CAS操作 for (;;) { int c = getState(); int nextc = c - SHARED_UNIT; if (compareAndSetState(c, nextc)) // Releasing the read lock has no effect on readers, // but it may allow waiting writers to proceed if // both read and write locks are now free. return nextc == 0; } }
3 获取写锁
public final void acquire(int arg) { if (!tryAcquire(arg) && acquireQueued(addWaiter(Node.EXCLUSIVE), arg)) selfInterrupt(); } //获取锁 public final void acquire(int arg) { if (!tryAcquire(arg) && acquireQueued(addWaiter(Node.EXCLUSIVE), arg)) selfInterrupt(); } //尝试获取锁 protected final boolean tryAcquire(int acquires) { Thread current = Thread.currentThread(); int c = getState(); int w = exclusiveCount(c); //如果c!=0则存在读锁或写锁 if (c != 0) { //如果写锁等于0,说明读锁不等于0, 若持有线程非当前线程 失败 if (w == 0 || current != getExclusiveOwnerThread()) return false; //超过最大写锁设定值 抛出异常 if (w + exclusiveCount(acquires) > MAX_COUNT) throw new Error("Maximum lock count exceeded"); //代码能走到这里 说明写锁不等于0, 且持有当前线程, 锁重入 setState(c + acquires); return true; } //以上代码必然直接返回结果, 代码走到这里说明c==0,不存在读写锁持有情况 if ( //判断当前线程获取写锁是否需要入CLH等待队列 //writerShouldBlock()方法代码很简单,只是CLH队列是否有等待的线程 //公平模式下如果有等待返回false(不等待),非公平模式直接返回false(不等待) writerShouldBlock() || //不需要需要写锁进入CLH, 然后修改AQS state值失败, 则返回false !compareAndSetState(c, c + acquires)) return false; //代码走到这前提 1不需要入CLH队列等待 2修改AQS state成功 直接设置锁占用线程为当前线程即可 setExclusiveOwnerThread(current); return true; }
4 释放写锁
//释放锁 public final boolean release(int arg) { //尝试释放锁 并判断重入次数是否全部释放完成 //若全部重入释放完成 则unpark CLH等待队列的第一个节点 if (tryRelease(arg)) { Node h = head; if (h != null && h.waitStatus != 0) unparkSuccessor(h); return true; } return false; } //尝试释放 protected final boolean tryRelease(int releases) { //持锁线程非当前线程 抛出异常 if (!isHeldExclusively()) throw new IllegalMonitorStateException(); //状态计算 int nextc = getState() - releases; //判断重入情况 boolean free = exclusiveCount(nextc) == 0; //若重入为0 表示此线程可以让出占用 if (free) setExclusiveOwnerThread(null); //状态设置 setState(nextc); //返回是否释放线程 return free; }
附:
1 高低位增加一次重入时AQS状态值的变更:
假设有十进制数字 11115555 如果说希望高4位用来做读锁值 低四位用来做写锁值,
那么增加一个读锁之后期望的到的结果是11125555,通过减法11125555-11115555=10000可知, 高位如果需要加1则实际需要加10000
同理增加一个写锁之后期望的到的结果是11115556,通过减法11115556-11115555=1可知,低位直接按实际值增加即可
回到AQS增加读锁重入时 state = state+(1<<16) , AQS增加写锁重入时 state = state+1
2 函数获取读锁数 int sharedCount(int c) { return c >>> SHARED_SHIFT; }
这个应该比较好理解, 就是将原有的32位向右移16位, 那么低(右侧)消失, 举个例子 1111 1111 0000 000 向右移8位变成 1111 1111
3 函数获取写锁数 int exclusiveCount(int c) { return c & EXCLUSIVE_MASK; }
需要掌握一些按位与的知识, EXCLUSIVE_MASK = 0000 0000 0000 0000 1111 1111 1111 1111
那么 c & EXCLUSIVE_MASK 表示同时为1 结果为1 否则为0,
假定 c = 0000 0000 0000 0001 0000 0000 0000 0010 表示读状态为1 写状态为2
0000 0000 0000 0000 1111 1111 1111 1111
&
0000 0000 0000 0001 0000 0000 0000 0010
--------------------------------------------------------------------------
= 0000 0000 0000 0000 0000 0000 0000 0010
c & EXCLUSIVE_MASK 得到的最终结果 0000 0000 0000 0000 0000 0000 0000 0010, 从而筛选出写状态值为2