Java细粒度锁实现的3种方式

1. 分段锁

借鉴concurrentHashMap的分段思想，先生成一定数量的锁，具体使用的时候再根据key来返回对应的lock。这是几个实现里最简单，性能最高，也是最终被采用的锁策略，代码如下：

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35

/**   * 分段锁，系统提供一定数量的原始锁，根据传入对象的哈希值获取对应的锁并加锁   * 注意：要锁的对象的哈希值如果发生改变，有可能导致锁无法成功释放!!!   */ public class SegmentLock<T> {      private Integer segments = 16 ; //默认分段数量      private final HashMap<Integer, ReentrantLock> lockMap = new HashMap<>();        public SegmentLock() {          init( null , false );      }        public SegmentLock(Integer counts, boolean fair) {          init(counts, fair);      }        private void init(Integer counts, boolean fair) {          if (counts != null ) {              segments = counts;          }          for ( int i = 0 ; i < segments; i++) {              lockMap.put(i, new ReentrantLock(fair));          }      }        public void lock(T key) {          ReentrantLock lock = lockMap.get(key.hashCode() % segments);          lock.lock();      }        public void unlock(T key) {          ReentrantLock lock = lockMap.get(key.hashCode() % segments);          lock.unlock();      } }

2. 哈希锁

上述分段锁的基础上发展起来的第二种锁策略，目的是实现真正意义上的细粒度锁。每个哈希值不同的对象都能获得自己独立的锁。在测试中，在被锁住的代码执行速度飞快的情况下，效率比分段锁慢 30% 左右。如果有长耗时操作，感觉表现应该会更好。代码如下：

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62

public class HashLock<T> {      private boolean isFair = false ;      private final SegmentLock<T> segmentLock = new SegmentLock<>(); //分段锁      private final ConcurrentHashMap<T, LockInfo> lockMap = new ConcurrentHashMap<>();        public HashLock() {      }        public HashLock( boolean fair) {          isFair = fair;      }        public void lock(T key) {          LockInfo lockInfo;          segmentLock.lock(key);          try {              lockInfo = lockMap.get(key);              if (lockInfo == null ) {                  lockInfo = new LockInfo(isFair);                  lockMap.put(key, lockInfo);              } else {                  lockInfo.count.incrementAndGet();              }          } finally {              segmentLock.unlock(key);          }          lockInfo.lock.lock();      }        public void unlock(T key) {          LockInfo lockInfo = lockMap.get(key);          if (lockInfo.count.get() == 1 ) {              segmentLock.lock(key);              try {                  if (lockInfo.count.get() == 1 ) {                      lockMap.remove(key);                  }              } finally {                  segmentLock.unlock(key);              }          }          lockInfo.count.decrementAndGet();          lockInfo.unlock();      }        private static class LockInfo {          public ReentrantLock lock;          public AtomicInteger count = new AtomicInteger( 1 );            private LockInfo( boolean fair) {              this .lock = new ReentrantLock(fair);          }            public void lock() {              this .lock.lock();          }            public void unlock() {              this .lock.unlock();          }      } }

3. 弱引用锁

哈希锁因为引入的分段锁来保证锁创建和销毁的同步，总感觉有点瑕疵，所以写了第三个锁来寻求更好的性能和更细粒度的锁。这个锁的思想是借助java的弱引用来创建锁，把锁的销毁交给jvm的垃圾回收，来避免额外的消耗。

有点遗憾的是因为使用了ConcurrentHashMap作为锁的容器，所以没能真正意义上的摆脱分段锁。这个锁的性能比 HashLock 快10% 左右。锁代码：

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43

/**   * 弱引用锁，为每个独立的哈希值提供独立的锁功能   */ public class WeakHashLock<T> {      private ConcurrentHashMap<T, WeakLockRef<T, ReentrantLock>> lockMap = new ConcurrentHashMap<>();      private ReferenceQueue<ReentrantLock> queue = new ReferenceQueue<>();        public ReentrantLock get(T key) {          if (lockMap.size() > 1000 ) {              clearEmptyRef();          }          WeakReference<ReentrantLock> lockRef = lockMap.get(key);          ReentrantLock lock = (lockRef == null ? null : lockRef.get());          while (lock == null ) {              lockMap.putIfAbsent(key, new WeakLockRef<>( new ReentrantLock(), queue, key));              lockRef = lockMap.get(key);              lock = (lockRef == null ? null : lockRef.get());              if (lock != null ) {                  return lock;              }              clearEmptyRef();          }          return lock;      }        @SuppressWarnings ( "unchecked" )      private void clearEmptyRef() {          Reference<? extends ReentrantLock> ref;          while ((ref = queue.poll()) != null ) {              WeakLockRef<T, ? extends ReentrantLock> weakLockRef = (WeakLockRef<T, ? extends ReentrantLock>) ref;              lockMap.remove(weakLockRef.key);          }      }        private static final class WeakLockRef<T, K> extends WeakReference<K> {          final T key;            private WeakLockRef(K referent, ReferenceQueue<? super K> q, T key) {              super (referent, q);              this .key = key;          }      } }

后记

最开始想借助 locksupport 和 AQS 来实现细粒度锁，写着写着发现正在实现的东西和java 原生的锁区别不大，于是放弃改为对java自带锁的封装，浪费了不少时间。