skywang12345

导航

统计
 

 

概要

Java的JUC(java.util.concurrent)包中的锁包括"独占锁"和"共享锁"。在“Java多线程系列--“JUC锁”02之 互斥锁ReentrantLock ”中,对Java的独占锁进行了说明。本章对Java的“共享锁”进行介绍,JUC中的共享锁有CountDownLatch, CyclicBarrier, Semaphore, ReentrantReadWriteLock等;本章会以ReentrantReadWriteLock为蓝本对共享锁进行说明。内容包括:
ReadWriteLock 和 ReentrantReadWriteLock介绍
ReadWriteLock 和 ReentrantReadWriteLock函数列表
ReentrantReadWriteLock数据结构

参考代码(基于JDK1.7.0_40)
  获取共享锁
  释放共享锁
  公平共享锁和非公平共享锁
ReentrantReadWriteLock示例

转载请注明出处:http://www.cnblogs.com/skywang12345/p/3505809.html

 

ReadWriteLock 和 ReentrantReadWriteLock介绍

ReadWriteLock,顾名思义,是读写锁。它维护了一对相关的锁 — — “读取锁”和“写入锁”,一个用于读取操作,另一个用于写入操作。
读取锁”用于只读操作,它是“共享锁”,能同时被多个线程获取。
写入锁”用于写入操作,它是“独占锁”,写入锁只能被一个线程锁获取。
注意:不能同时存在读取锁和写入锁!
ReadWriteLock是一个接口。ReentrantReadWriteLock是它的实现类,ReentrantReadWriteLock包括子类ReadLock和WriteLock。

 

ReadWriteLock 和 ReentrantReadWriteLock函数列表

ReadWriteLock函数列表

// 返回用于读取操作的锁。
Lock readLock()
// 返回用于写入操作的锁。
Lock writeLock()

 

ReentrantReadWriteLock函数列表

// 创建一个新的 ReentrantReadWriteLock,默认是采用“非公平策略”。
ReentrantReadWriteLock()
// 创建一个新的 ReentrantReadWriteLock,fair是“公平策略”。fair为true,意味着公平策略;否则,意味着非公平策略。
ReentrantReadWriteLock(boolean fair)

// 返回当前拥有写入锁的线程,如果没有这样的线程,则返回 null。
protected Thread getOwner()
// 返回一个 collection,它包含可能正在等待获取读取锁的线程。
protected Collection<Thread> getQueuedReaderThreads()
// 返回一个 collection,它包含可能正在等待获取读取或写入锁的线程。
protected Collection<Thread> getQueuedThreads()
// 返回一个 collection,它包含可能正在等待获取写入锁的线程。
protected Collection<Thread> getQueuedWriterThreads()
// 返回等待获取读取或写入锁的线程估计数目。
int getQueueLength()
// 查询当前线程在此锁上保持的重入读取锁数量。
int getReadHoldCount()
// 查询为此锁保持的读取锁数量。
int getReadLockCount()
// 返回一个 collection,它包含可能正在等待与写入锁相关的给定条件的那些线程。
protected Collection<Thread> getWaitingThreads(Condition condition)
// 返回正等待与写入锁相关的给定条件的线程估计数目。
int getWaitQueueLength(Condition condition)
// 查询当前线程在此锁上保持的重入写入锁数量。
int getWriteHoldCount()
// 查询是否给定线程正在等待获取读取或写入锁。
boolean hasQueuedThread(Thread thread)
// 查询是否所有的线程正在等待获取读取或写入锁。
boolean hasQueuedThreads()
// 查询是否有些线程正在等待与写入锁有关的给定条件。
boolean hasWaiters(Condition condition)
// 如果此锁将公平性设置为 ture,则返回 true。
boolean isFair()
// 查询是否某个线程保持了写入锁。
boolean isWriteLocked()
// 查询当前线程是否保持了写入锁。
boolean isWriteLockedByCurrentThread()
// 返回用于读取操作的锁。
ReentrantReadWriteLock.ReadLock readLock()
// 返回用于写入操作的锁。
ReentrantReadWriteLock.WriteLock writeLock()

 

ReentrantReadWriteLock数据结构

ReentrantReadWriteLock的UML类图如下:

从中可以看出:

(01) ReentrantReadWriteLock实现了ReadWriteLock接口。ReadWriteLock是一个读写锁的接口,提供了"获取读锁的readLock()函数" 和 "获取写锁的writeLock()函数"。
(02) ReentrantReadWriteLock中包含:sync对象,读锁readerLock和写锁writerLock。读锁ReadLock和写锁WriteLock都实现了Lock接口。读锁ReadLock和写锁WriteLock中也都分别包含了"Sync对象",它们的Sync对象和ReentrantReadWriteLock的Sync对象 是一样的,就是通过sync,读锁和写锁实现了对同一个对象的访问。
(03) 和"ReentrantLock"一样,sync是Sync类型;而且,Sync也是一个继承于AQS的抽象类。Sync也包括"公平锁"FairSync和"非公平锁"NonfairSync。sync对象是"FairSync"和"NonfairSync"中的一个,默认是"NonfairSync"。

 

参考代码(基于JDK1.7.0_40)

ReentrantReadWriteLock的完整源码

   1 /*
   2  * ORACLE PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
   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  * Written by Doug Lea with assistance from members of JCP JSR-166
  32  * Expert Group and released to the public domain, as explained at
  33  * http://creativecommons.org/publicdomain/zero/1.0/
  34  */
  35 
  36 package java.util.concurrent.locks;
  37 import java.util.concurrent.*;
  38 import java.util.concurrent.atomic.*;
  39 import java.util.*;
  40 
  41 /**
  42  * An implementation of {@link ReadWriteLock} supporting similar
  43  * semantics to {@link ReentrantLock}.
  44  * <p>This class has the following properties:
  45  *
  46  * <ul>
  47  * <li><b>Acquisition order</b>
  48  *
  49  * <p> This class does not impose a reader or writer preference
  50  * ordering for lock access.  However, it does support an optional
  51  * <em>fairness</em> policy.
  52  *
  53  * <dl>
  54  * <dt><b><i>Non-fair mode (default)</i></b>
  55  * <dd>When constructed as non-fair (the default), the order of entry
  56  * to the read and write lock is unspecified, subject to reentrancy
  57  * constraints.  A nonfair lock that is continuously contended may
  58  * indefinitely postpone one or more reader or writer threads, but
  59  * will normally have higher throughput than a fair lock.
  60  * <p>
  61  *
  62  * <dt><b><i>Fair mode</i></b>
  63  * <dd> When constructed as fair, threads contend for entry using an
  64  * approximately arrival-order policy. When the currently held lock
  65  * is released either the longest-waiting single writer thread will
  66  * be assigned the write lock, or if there is a group of reader threads
  67  * waiting longer than all waiting writer threads, that group will be
  68  * assigned the read lock.
  69  *
  70  * <p>A thread that tries to acquire a fair read lock (non-reentrantly)
  71  * will block if either the write lock is held, or there is a waiting
  72  * writer thread. The thread will not acquire the read lock until
  73  * after the oldest currently waiting writer thread has acquired and
  74  * released the write lock. Of course, if a waiting writer abandons
  75  * its wait, leaving one or more reader threads as the longest waiters
  76  * in the queue with the write lock free, then those readers will be
  77  * assigned the read lock.
  78  *
  79  * <p>A thread that tries to acquire a fair write lock (non-reentrantly)
  80  * will block unless both the read lock and write lock are free (which
  81  * implies there are no waiting threads).  (Note that the non-blocking
  82  * {@link ReadLock#tryLock()} and {@link WriteLock#tryLock()} methods
  83  * do not honor this fair setting and will acquire the lock if it is
  84  * possible, regardless of waiting threads.)
  85  * <p>
  86  * </dl>
  87  *
  88  * <li><b>Reentrancy</b>
  89  *
  90  * <p>This lock allows both readers and writers to reacquire read or
  91  * write locks in the style of a {@link ReentrantLock}. Non-reentrant
  92  * readers are not allowed until all write locks held by the writing
  93  * thread have been released.
  94  *
  95  * <p>Additionally, a writer can acquire the read lock, but not
  96  * vice-versa.  Among other applications, reentrancy can be useful
  97  * when write locks are held during calls or callbacks to methods that
  98  * perform reads under read locks.  If a reader tries to acquire the
  99  * write lock it will never succeed.
 100  *
 101  * <li><b>Lock downgrading</b>
 102  * <p>Reentrancy also allows downgrading from the write lock to a read lock,
 103  * by acquiring the write lock, then the read lock and then releasing the
 104  * write lock. However, upgrading from a read lock to the write lock is
 105  * <b>not</b> possible.
 106  *
 107  * <li><b>Interruption of lock acquisition</b>
 108  * <p>The read lock and write lock both support interruption during lock
 109  * acquisition.
 110  *
 111  * <li><b>{@link Condition} support</b>
 112  * <p>The write lock provides a {@link Condition} implementation that
 113  * behaves in the same way, with respect to the write lock, as the
 114  * {@link Condition} implementation provided by
 115  * {@link ReentrantLock#newCondition} does for {@link ReentrantLock}.
 116  * This {@link Condition} can, of course, only be used with the write lock.
 117  *
 118  * <p>The read lock does not support a {@link Condition} and
 119  * {@code readLock().newCondition()} throws
 120  * {@code UnsupportedOperationException}.
 121  *
 122  * <li><b>Instrumentation</b>
 123  * <p>This class supports methods to determine whether locks
 124  * are held or contended. These methods are designed for monitoring
 125  * system state, not for synchronization control.
 126  * </ul>
 127  *
 128  * <p>Serialization of this class behaves in the same way as built-in
 129  * locks: a deserialized lock is in the unlocked state, regardless of
 130  * its state when serialized.
 131  *
 132  * <p><b>Sample usages</b>. Here is a code sketch showing how to perform
 133  * lock downgrading after updating a cache (exception handling is
 134  * particularly tricky when handling multiple locks in a non-nested
 135  * fashion):
 136  *
 137  * <pre> {@code
 138  * class CachedData {
 139  *   Object data;
 140  *   volatile boolean cacheValid;
 141  *   final ReentrantReadWriteLock rwl = new ReentrantReadWriteLock();
 142  *
 143  *   void processCachedData() {
 144  *     rwl.readLock().lock();
 145  *     if (!cacheValid) {
 146  *        // Must release read lock before acquiring write lock
 147  *        rwl.readLock().unlock();
 148  *        rwl.writeLock().lock();
 149  *        try {
 150  *          // Recheck state because another thread might have
 151  *          // acquired write lock and changed state before we did.
 152  *          if (!cacheValid) {
 153  *            data = ...
 154  *            cacheValid = true;
 155  *          }
 156  *          // Downgrade by acquiring read lock before releasing write lock
 157  *          rwl.readLock().lock();
 158  *        } finally {
 159  *          rwl.writeLock().unlock(); // Unlock write, still hold read
 160  *        }
 161  *     }
 162  *
 163  *     try {
 164  *       use(data);
 165  *     } finally {
 166  *       rwl.readLock().unlock();
 167  *     }
 168  *   }
 169  * }}</pre>
 170  *
 171  * ReentrantReadWriteLocks can be used to improve concurrency in some
 172  * uses of some kinds of Collections. This is typically worthwhile
 173  * only when the collections are expected to be large, accessed by
 174  * more reader threads than writer threads, and entail operations with
 175  * overhead that outweighs synchronization overhead. For example, here
 176  * is a class using a TreeMap that is expected to be large and
 177  * concurrently accessed.
 178  *
 179  * <pre>{@code
 180  * class RWDictionary {
 181  *    private final Map<String, Data> m = new TreeMap<String, Data>();
 182  *    private final ReentrantReadWriteLock rwl = new ReentrantReadWriteLock();
 183  *    private final Lock r = rwl.readLock();
 184  *    private final Lock w = rwl.writeLock();
 185  *
 186  *    public Data get(String key) {
 187  *        r.lock();
 188  *        try { return m.get(key); }
 189  *        finally { r.unlock(); }
 190  *    }
 191  *    public String[] allKeys() {
 192  *        r.lock();
 193  *        try { return m.keySet().toArray(); }
 194  *        finally { r.unlock(); }
 195  *    }
 196  *    public Data put(String key, Data value) {
 197  *        w.lock();
 198  *        try { return m.put(key, value); }
 199  *        finally { w.unlock(); }
 200  *    }
 201  *    public void clear() {
 202  *        w.lock();
 203  *        try { m.clear(); }
 204  *        finally { w.unlock(); }
 205  *    }
 206  * }}</pre>
 207  *
 208  * <h3>Implementation Notes</h3>
 209  *
 210  * <p>This lock supports a maximum of 65535 recursive write locks
 211  * and 65535 read locks. Attempts to exceed these limits result in
 212  * {@link Error} throws from locking methods.
 213  *
 214  * @since 1.5
 215  * @author Doug Lea
 216  *
 217  */
 218 public class ReentrantReadWriteLock
 219         implements ReadWriteLock, java.io.Serializable {
 220     private static final long serialVersionUID = -6992448646407690164L;
 221     /** Inner class providing readlock */
 222     private final ReentrantReadWriteLock.ReadLock readerLock;
 223     /** Inner class providing writelock */
 224     private final ReentrantReadWriteLock.WriteLock writerLock;
 225     /** Performs all synchronization mechanics */
 226     final Sync sync;
 227 
 228     /**
 229      * Creates a new {@code ReentrantReadWriteLock} with
 230      * default (nonfair) ordering properties.
 231      */
 232     public ReentrantReadWriteLock() {
 233         this(false);
 234     }
 235 
 236     /**
 237      * Creates a new {@code ReentrantReadWriteLock} with
 238      * the given fairness policy.
 239      *
 240      * @param fair {@code true} if this lock should use a fair ordering policy
 241      */
 242     public ReentrantReadWriteLock(boolean fair) {
 243         sync = fair ? new FairSync() : new NonfairSync();
 244         readerLock = new ReadLock(this);
 245         writerLock = new WriteLock(this);
 246     }
 247 
 248     public ReentrantReadWriteLock.WriteLock writeLock() { return writerLock; }
 249     public ReentrantReadWriteLock.ReadLock  readLock()  { return readerLock; }
 250 
 251     /**
 252      * Synchronization implementation for ReentrantReadWriteLock.
 253      * Subclassed into fair and nonfair versions.
 254      */
 255     abstract static class Sync extends AbstractQueuedSynchronizer {
 256         private static final long serialVersionUID = 6317671515068378041L;
 257 
 258         /*
 259          * Read vs write count extraction constants and functions.
 260          * Lock state is logically divided into two unsigned shorts:
 261          * The lower one representing the exclusive (writer) lock hold count,
 262          * and the upper the shared (reader) hold count.
 263          */
 264 
 265         static final int SHARED_SHIFT   = 16;
 266         static final int SHARED_UNIT    = (1 << SHARED_SHIFT);
 267         static final int MAX_COUNT      = (1 << SHARED_SHIFT) - 1;
 268         static final int EXCLUSIVE_MASK = (1 << SHARED_SHIFT) - 1;
 269 
 270         /** Returns the number of shared holds represented in count  */
 271         static int sharedCount(int c)    { return c >>> SHARED_SHIFT; }
 272         /** Returns the number of exclusive holds represented in count  */
 273         static int exclusiveCount(int c) { return c & EXCLUSIVE_MASK; }
 274 
 275         /**
 276          * A counter for per-thread read hold counts.
 277          * Maintained as a ThreadLocal; cached in cachedHoldCounter
 278          */
 279         static final class HoldCounter {
 280             int count = 0;
 281             // Use id, not reference, to avoid garbage retention
 282             final long tid = Thread.currentThread().getId();
 283         }
 284 
 285         /**
 286          * ThreadLocal subclass. Easiest to explicitly define for sake
 287          * of deserialization mechanics.
 288          */
 289         static final class ThreadLocalHoldCounter
 290             extends ThreadLocal<HoldCounter> {
 291             public HoldCounter initialValue() {
 292                 return new HoldCounter();
 293             }
 294         }
 295 
 296         /**
 297          * The number of reentrant read locks held by current thread.
 298          * Initialized only in constructor and readObject.
 299          * Removed whenever a thread's read hold count drops to 0.
 300          */
 301         private transient ThreadLocalHoldCounter readHolds;
 302 
 303         /**
 304          * The hold count of the last thread to successfully acquire
 305          * readLock. This saves ThreadLocal lookup in the common case
 306          * where the next thread to release is the last one to
 307          * acquire. This is non-volatile since it is just used
 308          * as a heuristic, and would be great for threads to cache.
 309          *
 310          * <p>Can outlive the Thread for which it is caching the read
 311          * hold count, but avoids garbage retention by not retaining a
 312          * reference to the Thread.
 313          *
 314          * <p>Accessed via a benign data race; relies on the memory
 315          * model's final field and out-of-thin-air guarantees.
 316          */
 317         private transient HoldCounter cachedHoldCounter;
 318 
 319         /**
 320          * firstReader is the first thread to have acquired the read lock.
 321          * firstReaderHoldCount is firstReader's hold count.
 322          *
 323          * <p>More precisely, firstReader is the unique thread that last
 324          * changed the shared count from 0 to 1, and has not released the
 325          * read lock since then; null if there is no such thread.
 326          *
 327          * <p>Cannot cause garbage retention unless the thread terminated
 328          * without relinquishing its read locks, since tryReleaseShared
 329          * sets it to null.
 330          *
 331          * <p>Accessed via a benign data race; relies on the memory
 332          * model's out-of-thin-air guarantees for references.
 333          *
 334          * <p>This allows tracking of read holds for uncontended read
 335          * locks to be very cheap.
 336          */
 337         private transient Thread firstReader = null;
 338         private transient int firstReaderHoldCount;
 339 
 340         Sync() {
 341             readHolds = new ThreadLocalHoldCounter();
 342             setState(getState()); // ensures visibility of readHolds
 343         }
 344 
 345         /*
 346          * Acquires and releases use the same code for fair and
 347          * nonfair locks, but differ in whether/how they allow barging
 348          * when queues are non-empty.
 349          */
 350 
 351         /**
 352          * Returns true if the current thread, when trying to acquire
 353          * the read lock, and otherwise eligible to do so, should block
 354          * because of policy for overtaking other waiting threads.
 355          */
 356         abstract boolean readerShouldBlock();
 357 
 358         /**
 359          * Returns true if the current thread, when trying to acquire
 360          * the write lock, and otherwise eligible to do so, should block
 361          * because of policy for overtaking other waiting threads.
 362          */
 363         abstract boolean writerShouldBlock();
 364 
 365         /*
 366          * Note that tryRelease and tryAcquire can be called by
 367          * Conditions. So it is possible that their arguments contain
 368          * both read and write holds that are all released during a
 369          * condition wait and re-established in tryAcquire.
 370          */
 371 
 372         protected final boolean tryRelease(int releases) {
 373             if (!isHeldExclusively())
 374                 throw new IllegalMonitorStateException();
 375             int nextc = getState() - releases;
 376             boolean free = exclusiveCount(nextc) == 0;
 377             if (free)
 378                 setExclusiveOwnerThread(null);
 379             setState(nextc);
 380             return free;
 381         }
 382 
 383         protected final boolean tryAcquire(int acquires) {
 384             /*
 385              * Walkthrough:
 386              * 1. If read count nonzero or write count nonzero
 387              *    and owner is a different thread, fail.
 388              * 2. If count would saturate, fail. (This can only
 389              *    happen if count is already nonzero.)
 390              * 3. Otherwise, this thread is eligible for lock if
 391              *    it is either a reentrant acquire or
 392              *    queue policy allows it. If so, update state
 393              *    and set owner.
 394              */
 395             Thread current = Thread.currentThread();
 396             int c = getState();
 397             int w = exclusiveCount(c);
 398             if (c != 0) {
 399                 // (Note: if c != 0 and w == 0 then shared count != 0)
 400                 if (w == 0 || current != getExclusiveOwnerThread())
 401                     return false;
 402                 if (w + exclusiveCount(acquires) > MAX_COUNT)
 403                     throw new Error("Maximum lock count exceeded");
 404                 // Reentrant acquire
 405                 setState(c + acquires);
 406                 return true;
 407             }
 408             if (writerShouldBlock() ||
 409                 !compareAndSetState(c, c + acquires))
 410                 return false;
 411             setExclusiveOwnerThread(current);
 412             return true;
 413         }
 414 
 415         protected final boolean tryReleaseShared(int unused) {
 416             Thread current = Thread.currentThread();
 417             if (firstReader == current) {
 418                 // assert firstReaderHoldCount > 0;
 419                 if (firstReaderHoldCount == 1)
 420                     firstReader = null;
 421                 else
 422                     firstReaderHoldCount--;
 423             } else {
 424                 HoldCounter rh = cachedHoldCounter;
 425                 if (rh == null || rh.tid != current.getId())
 426                     rh = readHolds.get();
 427                 int count = rh.count;
 428                 if (count <= 1) {
 429                     readHolds.remove();
 430                     if (count <= 0)
 431                         throw unmatchedUnlockException();
 432                 }
 433                 --rh.count;
 434             }
 435             for (;;) {
 436                 int c = getState();
 437                 int nextc = c - SHARED_UNIT;
 438                 if (compareAndSetState(c, nextc))
 439                     // Releasing the read lock has no effect on readers,
 440                     // but it may allow waiting writers to proceed if
 441                     // both read and write locks are now free.
 442                     return nextc == 0;
 443             }
 444         }
 445 
 446         private IllegalMonitorStateException unmatchedUnlockException() {
 447             return new IllegalMonitorStateException(
 448                 "attempt to unlock read lock, not locked by current thread");
 449         }
 450 
 451         protected final int tryAcquireShared(int unused) {
 452             /*
 453              * Walkthrough:
 454              * 1. If write lock held by another thread, fail.
 455              * 2. Otherwise, this thread is eligible for
 456              *    lock wrt state, so ask if it should block
 457              *    because of queue policy. If not, try
 458              *    to grant by CASing state and updating count.
 459              *    Note that step does not check for reentrant
 460              *    acquires, which is postponed to full version
 461              *    to avoid having to check hold count in
 462              *    the more typical non-reentrant case.
 463              * 3. If step 2 fails either because thread
 464              *    apparently not eligible or CAS fails or count
 465              *    saturated, chain to version with full retry loop.
 466              */
 467             Thread current = Thread.currentThread();
 468             int c = getState();
 469             if (exclusiveCount(c) != 0 &&
 470                 getExclusiveOwnerThread() != current)
 471                 return -1;
 472             int r = sharedCount(c);
 473             if (!readerShouldBlock() &&
 474                 r < MAX_COUNT &&
 475                 compareAndSetState(c, c + SHARED_UNIT)) {
 476                 if (r == 0) {
 477                     firstReader = current;
 478                     firstReaderHoldCount = 1;
 479                 } else if (firstReader == current) {
 480                     firstReaderHoldCount++;
 481                 } else {
 482                     HoldCounter rh = cachedHoldCounter;
 483                     if (rh == null || rh.tid != current.getId())
 484                         cachedHoldCounter = rh = readHolds.get();
 485                     else if (rh.count == 0)
 486                         readHolds.set(rh);
 487                     rh.count++;
 488                 }
 489                 return 1;
 490             }
 491             return fullTryAcquireShared(current);
 492         }
 493 
 494         /**
 495          * Full version of acquire for reads, that handles CAS misses
 496          * and reentrant reads not dealt with in tryAcquireShared.
 497          */
 498         final int fullTryAcquireShared(Thread current) {
 499             /*
 500              * This code is in part redundant with that in
 501              * tryAcquireShared but is simpler overall by not
 502              * complicating tryAcquireShared with interactions between
 503              * retries and lazily reading hold counts.
 504              */
 505             HoldCounter rh = null;
 506             for (;;) {
 507                 int c = getState();
 508                 if (exclusiveCount(c) != 0) {
 509                     if (getExclusiveOwnerThread() != current)
 510                         return -1;
 511                     // else we hold the exclusive lock; blocking here
 512                     // would cause deadlock.
 513                 } else if (readerShouldBlock()) {
 514                     // Make sure we're not acquiring read lock reentrantly
 515                     if (firstReader == current) {
 516                         // assert firstReaderHoldCount > 0;
 517                     } else {
 518                         if (rh == null) {
 519                             rh = cachedHoldCounter;
 520                             if (rh == null || rh.tid != current.getId()) {
 521                                 rh = readHolds.get();
 522                                 if (rh.count == 0)
 523                                     readHolds.remove();
 524                             }
 525                         }
 526                         if (rh.count == 0)
 527                             return -1;
 528                     }
 529                 }
 530                 if (sharedCount(c) == MAX_COUNT)
 531                     throw new Error("Maximum lock count exceeded");
 532                 if (compareAndSetState(c, c + SHARED_UNIT)) {
 533                     if (sharedCount(c) == 0) {
 534                         firstReader = current;
 535                         firstReaderHoldCount = 1;
 536                     } else if (firstReader == current) {
 537                         firstReaderHoldCount++;
 538                     } else {
 539                         if (rh == null)
 540                             rh = cachedHoldCounter;
 541                         if (rh == null || rh.tid != current.getId())
 542                             rh = readHolds.get();
 543                         else if (rh.count == 0)
 544                             readHolds.set(rh);
 545                         rh.count++;
 546                         cachedHoldCounter = rh; // cache for release
 547                     }
 548                     return 1;
 549                 }
 550             }
 551         }
 552 
 553         /**
 554          * Performs tryLock for write, enabling barging in both modes.
 555          * This is identical in effect to tryAcquire except for lack
 556          * of calls to writerShouldBlock.
 557          */
 558         final boolean tryWriteLock() {
 559             Thread current = Thread.currentThread();
 560             int c = getState();
 561             if (c != 0) {
 562                 int w = exclusiveCount(c);
 563                 if (w == 0 || current != getExclusiveOwnerThread())
 564                     return false;
 565                 if (w == MAX_COUNT)
 566                     throw new Error("Maximum lock count exceeded");
 567             }
 568             if (!compareAndSetState(c, c + 1))
 569                 return false;
 570             setExclusiveOwnerThread(current);
 571             return true;
 572         }
 573 
 574         /**
 575          * Performs tryLock for read, enabling barging in both modes.
 576          * This is identical in effect to tryAcquireShared except for
 577          * lack of calls to readerShouldBlock.
 578          */
 579         final boolean tryReadLock() {
 580             Thread current = Thread.currentThread();
 581             for (;;) {
 582                 int c = getState();
 583                 if (exclusiveCount(c) != 0 &&
 584                     getExclusiveOwnerThread() != current)
 585                     return false;
 586                 int r = sharedCount(c);
 587                 if (r == MAX_COUNT)
 588                     throw new Error("Maximum lock count exceeded");
 589                 if (compareAndSetState(c, c + SHARED_UNIT)) {
 590                     if (r == 0) {
 591                         firstReader = current;
 592                         firstReaderHoldCount = 1;
 593                     } else if (firstReader == current) {
 594                         firstReaderHoldCount++;
 595                     } else {
 596                         HoldCounter rh = cachedHoldCounter;
 597                         if (rh == null || rh.tid != current.getId())
 598                             cachedHoldCounter = rh = readHolds.get();
 599                         else if (rh.count == 0)
 600                             readHolds.set(rh);
 601                         rh.count++;
 602                     }
 603                     return true;
 604                 }
 605             }
 606         }
 607 
 608         protected final boolean isHeldExclusively() {
 609             // While we must in general read state before owner,
 610             // we don't need to do so to check if current thread is owner
 611             return getExclusiveOwnerThread() == Thread.currentThread();
 612         }
 613 
 614         // Methods relayed to outer class
 615 
 616         final ConditionObject newCondition() {
 617             return new ConditionObject();
 618         }
 619 
 620         final Thread getOwner() {
 621             // Must read state before owner to ensure memory consistency
 622             return ((exclusiveCount(getState()) == 0) ?
 623                     null :
 624                     getExclusiveOwnerThread());
 625         }
 626 
 627         final int getReadLockCount() {
 628             return sharedCount(getState());
 629         }
 630 
 631         final boolean isWriteLocked() {
 632             return exclusiveCount(getState()) != 0;
 633         }
 634 
 635         final int getWriteHoldCount() {
 636             return isHeldExclusively() ? exclusiveCount(getState()) : 0;
 637         }
 638 
 639         final int getReadHoldCount() {
 640             if (getReadLockCount() == 0)
 641                 return 0;
 642 
 643             Thread current = Thread.currentThread();
 644             if (firstReader == current)
 645                 return firstReaderHoldCount;
 646 
 647             HoldCounter rh = cachedHoldCounter;
 648             if (rh != null && rh.tid == current.getId())
 649                 return rh.count;
 650 
 651             int count = readHolds.get().count;
 652             if (count == 0) readHolds.remove();
 653             return count;
 654         }
 655 
 656         /**
 657          * Reconstitute this lock instance from a stream
 658          * @param s the stream
 659          */
 660         private void readObject(java.io.ObjectInputStream s)
 661             throws java.io.IOException, ClassNotFoundException {
 662             s.defaultReadObject();
 663             readHolds = new ThreadLocalHoldCounter();
 664             setState(0); // reset to unlocked state
 665         }
 666 
 667         final int getCount() { return getState(); }
 668     }
 669 
 670     /**
 671      * Nonfair version of Sync
 672      */
 673     static final class NonfairSync extends Sync {
 674         private static final long serialVersionUID = -8159625535654395037L;
 675         final boolean writerShouldBlock() {
 676             return false; // writers can always barge
 677         }
 678         final boolean readerShouldBlock() {
 679             /* As a heuristic to avoid indefinite writer starvation,
 680              * block if the thread that momentarily appears to be head
 681              * of queue, if one exists, is a waiting writer.  This is
 682              * only a probabilistic effect since a new reader will not
 683              * block if there is a waiting writer behind other enabled
 684              * readers that have not yet drained from the queue.
 685              */
 686             return apparentlyFirstQueuedIsExclusive();
 687         }
 688     }
 689 
 690     /**
 691      * Fair version of Sync
 692      */
 693     static final class FairSync extends Sync {
 694         private static final long serialVersionUID = -2274990926593161451L;
 695         final boolean writerShouldBlock() {
 696             return hasQueuedPredecessors();
 697         }
 698         final boolean readerShouldBlock() {
 699             return hasQueuedPredecessors();
 700         }
 701     }
 702 
 703     /**
 704      * The lock returned by method {@link ReentrantReadWriteLock#readLock}.
 705      */
 706     public static class ReadLock implements Lock, java.io.Serializable {
 707         private static final long serialVersionUID = -5992448646407690164L;
 708         private final Sync sync;
 709 
 710         /**
 711          * Constructor for use by subclasses
 712          *
 713          * @param lock the outer lock object
 714          * @throws NullPointerException if the lock is null
 715          */
 716         protected ReadLock(ReentrantReadWriteLock lock) {
 717             sync = lock.sync;
 718         }
 719 
 720         /**
 721          * Acquires the read lock.
 722          *
 723          * <p>Acquires the read lock if the write lock is not held by
 724          * another thread and returns immediately.
 725          *
 726          * <p>If the write lock is held by another thread then
 727          * the current thread becomes disabled for thread scheduling
 728          * purposes and lies dormant until the read lock has been acquired.
 729          */
 730         public void lock() {
 731             sync.acquireShared(1);
 732         }
 733 
 734         /**
 735          * Acquires the read lock unless the current thread is
 736          * {@linkplain Thread#interrupt interrupted}.
 737          *
 738          * <p>Acquires the read lock if the write lock is not held
 739          * by another thread and returns immediately.
 740          *
 741          * <p>If the write lock is held by another thread then the
 742          * current thread becomes disabled for thread scheduling
 743          * purposes and lies dormant until one of two things happens:
 744          *
 745          * <ul>
 746          *
 747          * <li>The read lock is acquired by the current thread; or
 748          *
 749          * <li>Some other thread {@linkplain Thread#interrupt interrupts}
 750          * the current thread.
 751          *
 752          * </ul>
 753          *
 754          * <p>If the current thread:
 755          *
 756          * <ul>
 757          *
 758          * <li>has its interrupted status set on entry to this method; or
 759          *
 760          * <li>is {@linkplain Thread#interrupt interrupted} while
 761          * acquiring the read lock,
 762          *
 763          * </ul>
 764          *
 765          * then {@link InterruptedException} is thrown and the current
 766          * thread's interrupted status is cleared.
 767          *
 768          * <p>In this implementation, as this method is an explicit
 769          * interruption point, preference is given to responding to
 770          * the interrupt over normal or reentrant acquisition of the
 771          * lock.
 772          *
 773          * @throws InterruptedException if the current thread is interrupted
 774          */
 775         public void lockInterruptibly() throws InterruptedException {
 776             sync.acquireSharedInterruptibly(1);
 777         }
 778 
 779         /**
 780          * Acquires the read lock only if the write lock is not held by
 781          * another thread at the time of invocation.
 782          *
 783          * <p>Acquires the read lock if the write lock is not held by
 784          * another thread and returns immediately with the value
 785          * {@code true}. Even when this lock has been set to use a
 786          * fair ordering policy, a call to {@code tryLock()}
 787          * <em>will</em> immediately acquire the read lock if it is
 788          * available, whether or not other threads are currently
 789          * waiting for the read lock.  This &quot;barging&quot; behavior
 790          * can be useful in certain circumstances, even though it
 791          * breaks fairness. If you want to honor the fairness setting
 792          * for this lock, then use {@link #tryLock(long, TimeUnit)
 793          * tryLock(0, TimeUnit.SECONDS) } which is almost equivalent
 794          * (it also detects interruption).
 795          *
 796          * <p>If the write lock is held by another thread then
 797          * this method will return immediately with the value
 798          * {@code false}.
 799          *
 800          * @return {@code true} if the read lock was acquired
 801          */
 802         public  boolean tryLock() {
 803             return sync.tryReadLock();
 804         }
 805 
 806         /**
 807          * Acquires the read lock if the write lock is not held by
 808          * another thread within the given waiting time and the
 809          * current thread has not been {@linkplain Thread#interrupt
 810          * interrupted}.
 811          *
 812          * <p>Acquires the read lock if the write lock is not held by
 813          * another thread and returns immediately with the value
 814          * {@code true}. If this lock has been set to use a fair
 815          * ordering policy then an available lock <em>will not</em> be
 816          * acquired if any other threads are waiting for the
 817          * lock. This is in contrast to the {@link #tryLock()}
 818          * method. If you want a timed {@code tryLock} that does
 819          * permit barging on a fair lock then combine the timed and
 820          * un-timed forms together:
 821          *
 822          * <pre>if (lock.tryLock() || lock.tryLock(timeout, unit) ) { ... }
 823          * </pre>
 824          *
 825          * <p>If the write lock is held by another thread then the
 826          * current thread becomes disabled for thread scheduling
 827          * purposes and lies dormant until one of three things happens:
 828          *
 829          * <ul>
 830          *
 831          * <li>The read lock is acquired by the current thread; or
 832          *
 833          * <li>Some other thread {@linkplain Thread#interrupt interrupts}
 834          * the current thread; or
 835          *
 836          * <li>The specified waiting time elapses.
 837          *
 838          * </ul>
 839          *
 840          * <p>If the read lock is acquired then the value {@code true} is
 841          * returned.
 842          *
 843          * <p>If the current thread:
 844          *
 845          * <ul>
 846          *
 847          * <li>has its interrupted status set on entry to this method; or
 848          *
 849          * <li>is {@linkplain Thread#interrupt interrupted} while
 850          * acquiring the read lock,
 851          *
 852          * </ul> then {@link InterruptedException} is thrown and the
 853          * current thread's interrupted status is cleared.
 854          *
 855          * <p>If the specified waiting time elapses then the value
 856          * {@code false} is returned.  If the time is less than or
 857          * equal to zero, the method will not wait at all.
 858          *
 859          * <p>In this implementation, as this method is an explicit
 860          * interruption point, preference is given to responding to
 861          * the interrupt over normal or reentrant acquisition of the
 862          * lock, and over reporting the elapse of the waiting time.
 863          *
 864          * @param timeout the time to wait for the read lock
 865          * @param unit the time unit of the timeout argument
 866          * @return {@code true} if the read lock was acquired
 867          * @throws InterruptedException if the current thread is interrupted
 868          * @throws NullPointerException if the time unit is null
 869          *
 870          */
 871         public boolean tryLock(long timeout, TimeUnit unit)
 872                 throws InterruptedException {
 873             return sync.tryAcquireSharedNanos(1, unit.toNanos(timeout));
 874         }
 875 
 876         /**
 877          * Attempts to release this lock.
 878          *
 879          * <p> If the number of readers is now zero then the lock
 880          * is made available for write lock attempts.
 881          */
 882         public  void unlock() {
 883             sync.releaseShared(1);
 884         }
 885 
 886         /**
 887          * Throws {@code UnsupportedOperationException} because
 888          * {@code ReadLocks} do not support conditions.
 889          *
 890          * @throws UnsupportedOperationException always
 891          */
 892         public Condition newCondition() {
 893             throw new UnsupportedOperationException();
 894         }
 895 
 896         /**
 897          * Returns a string identifying this lock, as well as its lock state.
 898          * The state, in brackets, includes the String {@code "Read locks ="}
 899          * followed by the number of held read locks.
 900          *
 901          * @return a string identifying this lock, as well as its lock state
 902          */
 903         public String toString() {
 904             int r = sync.getReadLockCount();
 905             return super.toString() +
 906                 "[Read locks = " + r + "]";
 907         }
 908     }
 909 
 910     /**
 911      * The lock returned by method {@link ReentrantReadWriteLock#writeLock}.
 912      */
 913     public static class WriteLock implements Lock, java.io.Serializable {
 914         private static final long serialVersionUID = -4992448646407690164L;
 915         private final Sync sync;
 916 
 917         /**
 918          * Constructor for use by subclasses
 919          *
 920          * @param lock the outer lock object
 921          * @throws NullPointerException if the lock is null
 922          */
 923         protected WriteLock(ReentrantReadWriteLock lock) {
 924             sync = lock.sync;
 925         }
 926 
 927         /**
 928          * Acquires the write lock.
 929          *
 930          * <p>Acquires the write lock if neither the read nor write lock
 931          * are held by another thread
 932          * and returns immediately, setting the write lock hold count to
 933          * one.
 934          *
 935          * <p>If the current thread already holds the write lock then the
 936          * hold count is incremented by one and the method returns
 937          * immediately.
 938          *
 939          * <p>If the lock is held by another thread then the current
 940          * thread becomes disabled for thread scheduling purposes and
 941          * lies dormant until the write lock has been acquired, at which
 942          * time the write lock hold count is set to one.
 943          */
 944         public void lock() {
 945             sync.acquire(1);
 946         }
 947 
 948         /**
 949          * Acquires the write lock unless the current thread is
 950          * {@linkplain Thread#interrupt interrupted}.
 951          *
 952          * <p>Acquires the write lock if neither the read nor write lock
 953          * are held by another thread
 954          * and returns immediately, setting the write lock hold count to
 955          * one.
 956          *
 957          * <p>If the current thread already holds this lock then the
 958          * hold count is incremented by one and the method returns
 959          * immediately.
 960          *
 961          * <p>If the lock is held by another thread then the current
 962          * thread becomes disabled for thread scheduling purposes and
 963          * lies dormant until one of two things happens:
 964          *
 965          * <ul>
 966          *
 967          * <li>The write lock is acquired by the current thread; or
 968          *
 969          * <li>Some other thread {@linkplain Thread#interrupt interrupts}
 970          * the current thread.
 971          *
 972          * </ul>
 973          *
 974          * <p>If the write lock is acquired by the current thread then the
 975          * lock hold count is set to one.
 976          *
 977          * <p>If the current thread:
 978          *
 979          * <ul>
 980          *
 981          * <li>has its interrupted status set on entry to this method;
 982          * or
 983          *
 984          * <li>is {@linkplain Thread#interrupt interrupted} while
 985          * acquiring the write lock,
 986          *
 987          * </ul>
 988          *
 989          * then {@link InterruptedException} is thrown and the current
 990          * thread's interrupted status is cleared.
 991          *
 992          * <p>In this implementation, as this method is an explicit
 993          * interruption point, preference is given to responding to
 994          * the interrupt over normal or reentrant acquisition of the
 995          * lock.
 996          *
 997          * @throws InterruptedException if the current thread is interrupted
 998          */
 999         public void lockInterruptibly() throws InterruptedException {
1000             sync.acquireInterruptibly(1);
1001         }
1002 
1003         /**
1004          * Acquires the write lock only if it is not held by another thread
1005          * at the time of invocation.
1006          *
1007          * <p>Acquires the write lock if neither the read nor write lock
1008          * are held by another thread
1009          * and returns immediately with the value {@code true},
1010          * setting the write lock hold count to one. Even when this lock has
1011          * been set to use a fair ordering policy, a call to
1012          * {@code tryLock()} <em>will</em> immediately acquire the
1013          * lock if it is available, whether or not other threads are
1014          * currently waiting for the write lock.  This &quot;barging&quot;
1015          * behavior can be useful in certain circumstances, even
1016          * though it breaks fairness. If you want to honor the
1017          * fairness setting for this lock, then use {@link
1018          * #tryLock(long, TimeUnit) tryLock(0, TimeUnit.SECONDS) }
1019          * which is almost equivalent (it also detects interruption).
1020          *
1021          * <p> If the current thread already holds this lock then the
1022          * hold count is incremented by one and the method returns
1023          * {@code true}.
1024          *
1025          * <p>If the lock is held by another thread then this method
1026          * will return immediately with the value {@code false}.
1027          *
1028          * @return {@code true} if the lock was free and was acquired
1029          * by the current thread, or the write lock was already held
1030          * by the current thread; and {@code false} otherwise.
1031          */
1032         public boolean tryLock( ) {
1033             return sync.tryWriteLock();
1034         }
1035 
1036         /**
1037          * Acquires the write lock if it is not held by another thread
1038          * within the given waiting time and the current thread has
1039          * not been {@linkplain Thread#interrupt interrupted}.
1040          *
1041          * <p>Acquires the write lock if neither the read nor write lock
1042          * are held by another thread
1043          * and returns immediately with the value {@code true},
1044          * setting the write lock hold count to one. If this lock has been
1045          * set to use a fair ordering policy then an available lock
1046          * <em>will not</em> be acquired if any other threads are
1047          * waiting for the write lock. This is in contrast to the {@link
1048          * #tryLock()} method. If you want a timed {@code tryLock}
1049          * that does permit barging on a fair lock then combine the
1050          * timed and un-timed forms together:
1051          *
1052          * <pre>if (lock.tryLock() || lock.tryLock(timeout, unit) ) { ... }
1053          * </pre>
1054          *
1055          * <p>If the current thread already holds this lock then the
1056          * hold count is incremented by one and the method returns
1057          * {@code true}.
1058          *
1059          * <p>If the lock is held by another thread then the current
1060          * thread becomes disabled for thread scheduling purposes and
1061          * lies dormant until one of three things happens:
1062          *
1063          * <ul>
1064          *
1065          * <li>The write lock is acquired by the current thread; or
1066          *
1067          * <li>Some other thread {@linkplain Thread#interrupt interrupts}
1068          * the current thread; or
1069          *
1070          * <li>The specified waiting time elapses
1071          *
1072          * </ul>
1073          *
1074          * <p>If the write lock is acquired then the value {@code true} is
1075          * returned and the write lock hold count is set to one.
1076          *
1077          * <p>If the current thread:
1078          *
1079          * <ul>
1080          *
1081          * <li>has its interrupted status set on entry to this method;
1082          * or
1083          *
1084          * <li>is {@linkplain Thread#interrupt interrupted} while
1085          * acquiring the write lock,
1086          *
1087          * </ul>
1088          *
1089          * then {@link InterruptedException} is thrown and the current
1090          * thread's interrupted status is cleared.
1091          *
1092          * <p>If the specified waiting time elapses then the value
1093          * {@code false} is returned.  If the time is less than or
1094          * equal to zero, the method will not wait at all.
1095          *
1096          * <p>In this implementation, as this method is an explicit
1097          * interruption point, preference is given to responding to
1098          * the interrupt over normal or reentrant acquisition of the
1099          * lock, and over reporting the elapse of the waiting time.
1100          *
1101          * @param timeout the time to wait for the write lock
1102          * @param unit the time unit of the timeout argument
1103          *
1104          * @return {@code true} if the lock was free and was acquired
1105          * by the current thread, or the write lock was already held by the
1106          * current thread; and {@code false} if the waiting time
1107          * elapsed before the lock could be acquired.
1108          *
1109          * @throws InterruptedException if the current thread is interrupted
1110          * @throws NullPointerException if the time unit is null
1111          *
1112          */
1113         public boolean tryLock(long timeout, TimeUnit unit)
1114                 throws InterruptedException {
1115             return sync.tryAcquireNanos(1, unit.toNanos(timeout));
1116         }
1117 
1118         /**
1119          * Attempts to release this lock.
1120          *
1121          * <p>If the current thread is the holder of this lock then
1122          * the hold count is decremented. If the hold count is now
1123          * zero then the lock is released.  If the current thread is
1124          * not the holder of this lock then {@link
1125          * IllegalMonitorStateException} is thrown.
1126          *
1127          * @throws IllegalMonitorStateException if the current thread does not
1128          * hold this lock.
1129          */
1130         public void unlock() {
1131             sync.release(1);
1132         }
1133 
1134         /**
1135          * Returns a {@link Condition} instance for use with this
1136          * {@link Lock} instance.
1137          * <p>The returned {@link Condition} instance supports the same
1138          * usages as do the {@link Object} monitor methods ({@link
1139          * Object#wait() wait}, {@link Object#notify notify}, and {@link
1140          * Object#notifyAll notifyAll}) when used with the built-in
1141          * monitor lock.
1142          *
1143          * <ul>
1144          *
1145          * <li>If this write lock is not held when any {@link
1146          * Condition} method is called then an {@link
1147          * IllegalMonitorStateException} is thrown.  (Read locks are
1148          * held independently of write locks, so are not checked or
1149          * affected. However it is essentially always an error to
1150          * invoke a condition waiting method when the current thread
1151          * has also acquired read locks, since other threads that
1152          * could unblock it will not be able to acquire the write
1153          * lock.)
1154          *
1155          * <li>When the condition {@linkplain Condition#await() waiting}
1156          * methods are called the write lock is released and, before
1157          * they return, the write lock is reacquired and the lock hold
1158          * count restored to what it was when the method was called.
1159          *
1160          * <li>If a thread is {@linkplain Thread#interrupt interrupted} while
1161          * waiting then the wait will terminate, an {@link
1162          * InterruptedException} will be thrown, and the thread's
1163          * interrupted status will be cleared.
1164          *
1165          * <li> Waiting threads are signalled in FIFO order.
1166          *
1167          * <li>The ordering of lock reacquisition for threads returning
1168          * from waiting methods is the same as for threads initially
1169          * acquiring the lock, which is in the default case not specified,
1170          * but for <em>fair</em> locks favors those threads that have been
1171          * waiting the longest.
1172          *
1173          * </ul>
1174          *
1175          * @return the Condition object
1176          */
1177         public Condition newCondition() {
1178             return sync.newCondition();
1179         }
1180 
1181         /**
1182          * Returns a string identifying this lock, as well as its lock
1183          * state.  The state, in brackets includes either the String
1184          * {@code "Unlocked"} or the String {@code "Locked by"}
1185          * followed by the {@linkplain Thread#getName name} of the owning thread.
1186          *
1187          * @return a string identifying this lock, as well as its lock state
1188          */
1189         public String toString() {
1190             Thread o = sync.getOwner();
1191             return super.toString() + ((o == null) ?
1192                                        "[Unlocked]" :
1193                                        "[Locked by thread " + o.getName() + "]");
1194         }
1195 
1196         /**
1197          * Queries if this write lock is held by the current thread.
1198          * Identical in effect to {@link
1199          * ReentrantReadWriteLock#isWriteLockedByCurrentThread}.
1200          *
1201          * @return {@code true} if the current thread holds this lock and
1202          *         {@code false} otherwise
1203          * @since 1.6
1204          */
1205         public boolean isHeldByCurrentThread() {
1206             return sync.isHeldExclusively();
1207         }
1208 
1209         /**
1210          * Queries the number of holds on this write lock by the current
1211          * thread.  A thread has a hold on a lock for each lock action
1212          * that is not matched by an unlock action.  Identical in effect
1213          * to {@link ReentrantReadWriteLock#getWriteHoldCount}.
1214          *
1215          * @return the number of holds on this lock by the current thread,
1216          *         or zero if this lock is not held by the current thread
1217          * @since 1.6
1218          */
1219         public int getHoldCount() {
1220             return sync.getWriteHoldCount();
1221         }
1222     }
1223 
1224     // Instrumentation and status
1225 
1226     /**
1227      * Returns {@code true} if this lock has fairness set true.
1228      *
1229      * @return {@code true} if this lock has fairness set true
1230      */
1231     public final boolean isFair() {
1232         return sync instanceof FairSync;
1233     }
1234 
1235     /**
1236      * Returns the thread that currently owns the write lock, or
1237      * {@code null} if not owned. When this method is called by a
1238      * thread that is not the owner, the return value reflects a
1239      * best-effort approximation of current lock status. For example,
1240      * the owner may be momentarily {@code null} even if there are
1241      * threads trying to acquire the lock but have not yet done so.
1242      * This method is designed to facilitate construction of
1243      * subclasses that provide more extensive lock monitoring
1244      * facilities.
1245      *
1246      * @return the owner, or {@code null} if not owned
1247      */
1248     protected Thread getOwner() {
1249         return sync.getOwner();
1250     }
1251 
1252     /**
1253      * Queries the number of read locks held for this lock. This
1254      * method is designed for use in monitoring system state, not for
1255      * synchronization control.
1256      * @return the number of read locks held.
1257      */
1258     public int getReadLockCount() {
1259         return sync.getReadLockCount();
1260     }
1261 
1262     /**
1263      * Queries if the write lock is held by any thread. This method is
1264      * designed for use in monitoring system state, not for
1265      * synchronization control.
1266      *
1267      * @return {@code true} if any thread holds the write lock and
1268      *         {@code false} otherwise
1269      */
1270     public boolean isWriteLocked() {
1271         return sync.isWriteLocked();
1272     }
1273 
1274     /**
1275      * Queries if the write lock is held by the current thread.
1276      *
1277      * @return {@code true} if the current thread holds the write lock and
1278      *         {@code false} otherwise
1279      */
1280     public boolean isWriteLockedByCurrentThread() {
1281         return sync.isHeldExclusively();
1282     }
1283 
1284     /**
1285      * Queries the number of reentrant write holds on this lock by the
1286      * current thread.  A writer thread has a hold on a lock for
1287      * each lock action that is not matched by an unlock action.
1288      *
1289      * @return the number of holds on the write lock by the current thread,
1290      *         or zero if the write lock is not held by the current thread
1291      */
1292     public int getWriteHoldCount() {
1293         return sync.getWriteHoldCount();
1294     }
1295 
1296     /**
1297      * Queries the number of reentrant read holds on this lock by the
1298      * current thread.  A reader thread has a hold on a lock for
1299      * each lock action that is not matched by an unlock action.
1300      *
1301      * @return the number of holds on the read lock by the current thread,
1302      *         or zero if the read lock is not held by the current thread
1303      * @since 1.6
1304      */
1305     public int getReadHoldCount() {
1306         return sync.getReadHoldCount();
1307     }
1308 
1309     /**
1310      * Returns a collection containing threads that may be waiting to
1311      * acquire the write lock.  Because the actual set of threads may
1312      * change dynamically while constructing this result, the returned
1313      * collection is only a best-effort estimate.  The elements of the
1314      * returned collection are in no particular order.  This method is
1315      * designed to facilitate construction of subclasses that provide
1316      * more extensive lock monitoring facilities.
1317      *
1318      * @return the collection of threads
1319      */
1320     protected Collection<Thread> getQueuedWriterThreads() {
1321         return sync.getExclusiveQueuedThreads();
1322     }
1323 
1324     /**
1325      * Returns a collection containing threads that may be waiting to
1326      * acquire the read lock.  Because the actual set of threads may
1327      * change dynamically while constructing this result, the returned
1328      * collection is only a best-effort estimate.  The elements of the
1329      * returned collection are in no particular order.  This method is
1330      * designed to facilitate construction of subclasses that provide
1331      * more extensive lock monitoring facilities.
1332      *
1333      * @return the collection of threads
1334      */
1335     protected Collection<Thread> getQueuedReaderThreads() {
1336         return sync.getSharedQueuedThreads();
1337     }
1338 
1339     /**
1340      * Queries whether any threads are waiting to acquire the read or
1341      * write lock. Note that because cancellations may occur at any
1342      * time, a {@code true} return does not guarantee that any other
1343      * thread will ever acquire a lock.  This method is designed
1344      * primarily for use in monitoring of the system state.
1345      *
1346      * @return {@code true} if there may be other threads waiting to
1347      *         acquire the lock
1348      */
1349     public final boolean hasQueuedThreads() {
1350         return sync.hasQueuedThreads();
1351     }
1352 
1353     /**
1354      * Queries whether the given thread is waiting to acquire either
1355      * the read or write lock. Note that because cancellations may
1356      * occur at any time, a {@code true} return does not guarantee
1357      * that this thread will ever acquire a lock.  This method is
1358      * designed primarily for use in monitoring of the system state.
1359      *
1360      * @param thread the thread
1361      * @return {@code true} if the given thread is queued waiting for this lock
1362      * @throws NullPointerException if the thread is null
1363      */
1364     public final boolean hasQueuedThread(Thread thread) {
1365         return sync.isQueued(thread);
1366     }
1367 
1368     /**
1369      * Returns an estimate of the number of threads waiting to acquire
1370      * either the read or write lock.  The value is only an estimate
1371      * because the number of threads may change dynamically while this
1372      * method traverses internal data structures.  This method is
1373      * designed for use in monitoring of the system state, not for
1374      * synchronization control.
1375      *
1376      * @return the estimated number of threads waiting for this lock
1377      */
1378     public final int getQueueLength() {
1379         return sync.getQueueLength();
1380     }
1381 
1382     /**
1383      * Returns a collection containing threads that may be waiting to
1384      * acquire either the read or write lock.  Because the actual set
1385      * of threads may change dynamically while constructing this
1386      * result, the returned collection is only a best-effort estimate.
1387      * The elements of the returned collection are in no particular
1388      * order.  This method is designed to facilitate construction of
1389      * subclasses that provide more extensive monitoring facilities.
1390      *
1391      * @return the collection of threads
1392      */
1393     protected Collection<Thread> getQueuedThreads() {
1394         return sync.getQueuedThreads();
1395     }
1396 
1397     /**
1398      * Queries whether any threads are waiting on the given condition
1399      * associated with the write lock. Note that because timeouts and
1400      * interrupts may occur at any time, a {@code true} return does
1401      * not guarantee that a future {@code signal} will awaken any
1402      * threads.  This method is designed primarily for use in
1403      * monitoring of the system state.
1404      *
1405      * @param condition the condition
1406      * @return {@code true} if there are any waiting threads
1407      * @throws IllegalMonitorStateException if this lock is not held
1408      * @throws IllegalArgumentException if the given condition is
1409      *         not associated with this lock
1410      * @throws NullPointerException if the condition is null
1411      */
1412     public boolean hasWaiters(Condition condition) {
1413         if (condition == null)
1414             throw new NullPointerException();
1415         if (!(condition instanceof AbstractQueuedSynchronizer.ConditionObject))
1416             throw new IllegalArgumentException("not owner");
1417         return sync.hasWaiters((AbstractQueuedSynchronizer.ConditionObject)condition);
1418     }
1419 
1420     /**
1421      * Returns an estimate of the number of threads waiting on the
1422      * given condition associated with the write lock. Note that because
1423      * timeouts and interrupts may occur at any time, the estimate
1424      * serves only as an upper bound on the actual number of waiters.
1425      * This method is designed for use in monitoring of the system
1426      * state, not for synchronization control.
1427      *
1428      * @param condition the condition
1429      * @return the estimated number of waiting threads
1430      * @throws IllegalMonitorStateException if this lock is not held
1431      * @throws IllegalArgumentException if the given condition is
1432      *         not associated with this lock
1433      * @throws NullPointerException if the condition is null
1434      */
1435     public int getWaitQueueLength(Condition condition) {
1436         if (condition == null)
1437             throw new NullPointerException();
1438         if (!(condition instanceof AbstractQueuedSynchronizer.ConditionObject))
1439             throw new IllegalArgumentException("not owner");
1440         return sync.getWaitQueueLength((AbstractQueuedSynchronizer.ConditionObject)condition);
1441     }
1442 
1443     /**
1444      * Returns a collection containing those threads that may be
1445      * waiting on the given condition associated with the write lock.
1446      * Because the actual set of threads may change dynamically while
1447      * constructing this result, the returned collection is only a
1448      * best-effort estimate. The elements of the returned collection
1449      * are in no particular order.  This method is designed to
1450      * facilitate construction of subclasses that provide more
1451      * extensive condition monitoring facilities.
1452      *
1453      * @param condition the condition
1454      * @return the collection of threads
1455      * @throws IllegalMonitorStateException if this lock is not held
1456      * @throws IllegalArgumentException if the given condition is
1457      *         not associated with this lock
1458      * @throws NullPointerException if the condition is null
1459      */
1460     protected Collection<Thread> getWaitingThreads(Condition condition) {
1461         if (condition == null)
1462             throw new NullPointerException();
1463         if (!(condition instanceof AbstractQueuedSynchronizer.ConditionObject))
1464             throw new IllegalArgumentException("not owner");
1465         return sync.getWaitingThreads((AbstractQueuedSynchronizer.ConditionObject)condition);
1466     }
1467 
1468     /**
1469      * Returns a string identifying this lock, as well as its lock state.
1470      * The state, in brackets, includes the String {@code "Write locks ="}
1471      * followed by the number of reentrantly held write locks, and the
1472      * String {@code "Read locks ="} followed by the number of held
1473      * read locks.
1474      *
1475      * @return a string identifying this lock, as well as its lock state
1476      */
1477     public String toString() {
1478         int c = sync.getCount();
1479         int w = Sync.exclusiveCount(c);
1480         int r = Sync.sharedCount(c);
1481 
1482         return super.toString() +
1483             "[Write locks = " + w + ", Read locks = " + r + "]";
1484     }
1485 
1486 }
View Code

 

AQS的完整源码

   1 /*
   2  * ORACLE PROPRIETARY/CONFIDENTIAL. Use is subject to license terms.
   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  * Written by Doug Lea with assistance from members of JCP JSR-166
  32  * Expert Group and released to the public domain, as explained at
  33  * http://creativecommons.org/publicdomain/zero/1.0/
  34  */
  35 
  36 package java.util.concurrent.locks;
  37 import java.util.*;
  38 import java.util.concurrent.*;
  39 import java.util.concurrent.atomic.*;
  40 import sun.misc.Unsafe;
  41 
  42 /**
  43  * Provides a framework for implementing blocking locks and related
  44  * synchronizers (semaphores, events, etc) that rely on
  45  * first-in-first-out (FIFO) wait queues.  This class is designed to
  46  * be a useful basis for most kinds of synchronizers that rely on a
  47  * single atomic <tt>int</tt> value to represent state. Subclasses
  48  * must define the protected methods that change this state, and which
  49  * define what that state means in terms of this object being acquired
  50  * or released.  Given these, the other methods in this class carry
  51  * out all queuing and blocking mechanics. Subclasses can maintain
  52  * other state fields, but only the atomically updated <tt>int</tt>
  53  * value manipulated using methods {@link #getState}, {@link
  54  * #setState} and {@link #compareAndSetState} is tracked with respect
  55  * to synchronization.
  56  *
  57  * <p>Subclasses should be defined as non-public internal helper
  58  * classes that are used to implement the synchronization properties
  59  * of their enclosing class.  Class
  60  * <tt>AbstractQueuedSynchronizer</tt> does not implement any
  61  * synchronization interface.  Instead it defines methods such as
  62  * {@link #acquireInterruptibly} that can be invoked as
  63  * appropriate by concrete locks and related synchronizers to
  64  * implement their public methods.
  65  *
  66  * <p>This class supports either or both a default <em>exclusive</em>
  67  * mode and a <em>shared</em> mode. When acquired in exclusive mode,
  68  * attempted acquires by other threads cannot succeed. Shared mode
  69  * acquires by multiple threads may (but need not) succeed. This class
  70  * does not &quot;understand&quot; these differences except in the
  71  * mechanical sense that when a shared mode acquire succeeds, the next
  72  * waiting thread (if one exists) must also determine whether it can
  73  * acquire as well. Threads waiting in the different modes share the
  74  * same FIFO queue. Usually, implementation subclasses support only
  75  * one of these modes, but both can come into play for example in a
  76  * {@link ReadWriteLock}. Subclasses that support only exclusive or
  77  * only shared modes need not define the methods supporting the unused mode.
  78  *
  79  * <p>This class defines a nested {@link ConditionObject} class that
  80  * can be used as a {@link Condition} implementation by subclasses
  81  * supporting exclusive mode for which method {@link
  82  * #isHeldExclusively} reports whether synchronization is exclusively
  83  * held with respect to the current thread, method {@link #release}
  84  * invoked with the current {@link #getState} value fully releases
  85  * this object, and {@link #acquire}, given this saved state value,
  86  * eventually restores this object to its previous acquired state.  No
  87  * <tt>AbstractQueuedSynchronizer</tt> method otherwise creates such a
  88  * condition, so if this constraint cannot be met, do not use it.  The
  89  * behavior of {@link ConditionObject} depends of course on the
  90  * semantics of its synchronizer implementation.
  91  *
  92  * <p>This class provides inspection, instrumentation, and monitoring
  93  * methods for the internal queue, as well as similar methods for
  94  * condition objects. These can be exported as desired into classes
  95  * using an <tt>AbstractQueuedSynchronizer</tt> for their
  96  * synchronization mechanics.
  97  *
  98  * <p>Serialization of this class stores only the underlying atomic
  99  * integer maintaining state, so deserialized objects have empty
 100  * thread queues. Typical subclasses requiring serializability will
 101  * define a <tt>readObject</tt> method that restores this to a known
 102  * initial state upon deserialization.
 103  *
 104  * <h3>Usage</h3>
 105  *
 106  * <p>To use this class as the basis of a synchronizer, redefine the
 107  * following methods, as applicable, by inspecting and/or modifying
 108  * the synchronization state using {@link #getState}, {@link
 109  * #setState} and/or {@link #compareAndSetState}:
 110  *
 111  * <ul>
 112  * <li> {@link #tryAcquire}
 113  * <li> {@link #tryRelease}
 114  * <li> {@link #tryAcquireShared}
 115  * <li> {@link #tryReleaseShared}
 116  * <li> {@link #isHeldExclusively}
 117  *</ul>
 118  *
 119  * Each of these methods by default throws {@link
 120  * UnsupportedOperationException}.  Implementations of these methods
 121  * must be internally thread-safe, and should in general be short and
 122  * not block. Defining these methods is the <em>only</em> supported
 123  * means of using this class. All other methods are declared
 124  * <tt>final</tt> because they cannot be independently varied.
 125  *
 126  * <p>You may also find the inherited methods from {@link
 127  * AbstractOwnableSynchronizer} useful to keep track of the thread
 128  * owning an exclusive synchronizer.  You are encouraged to use them
 129  * -- this enables monitoring and diagnostic tools to assist users in
 130  * determining which threads hold locks.
 131  *
 132  * <p>Even though this class is based on an internal FIFO queue, it
 133  * does not automatically enforce FIFO acquisition policies.  The core
 134  * of exclusive synchronization takes the form:
 135  *
 136  * <pre>
 137  * Acquire:
 138  *     while (!tryAcquire(arg)) {
 139  *        <em>enqueue thread if it is not already queued</em>;
 140  *        <em>possibly block current thread</em>;
 141  *     }
 142  *
 143  * Release:
 144  *     if (tryRelease(arg))
 145  *        <em>unblock the first queued thread</em>;
 146  * </pre>
 147  *
 148  * (Shared mode is similar but may involve cascading signals.)
 149  *
 150  * <p><a name="barging">Because checks in acquire are invoked before
 151  * enqueuing, a newly acquiring thread may <em>barge</em> ahead of
 152  * others that are blocked and queued.  However, you can, if desired,
 153  * define <tt>tryAcquire</tt> and/or <tt>tryAcquireShared</tt> to
 154  * disable barging by internally invoking one or more of the inspection
 155  * methods, thereby providing a <em>fair</em> FIFO acquisition order.
 156  * In particular, most fair synchronizers can define <tt>tryAcquire</tt>
 157  * to return <tt>false</tt> if {@link #hasQueuedPredecessors} (a method
 158  * specifically designed to be used by fair synchronizers) returns
 159  * <tt>true</tt>.  Other variations are possible.
 160  *
 161  * <p>Throughput and scalability are generally highest for the
 162  * default barging (also known as <em>greedy</em>,
 163  * <em>renouncement</em>, and <em>convoy-avoidance</em>) strategy.
 164  * While this is not guaranteed to be fair or starvation-free, earlier
 165  * queued threads are allowed to recontend before later queued
 166  * threads, and each recontention has an unbiased chance to succeed
 167  * against incoming threads.  Also, while acquires do not
 168  * &quot;spin&quot; in the usual sense, they may perform multiple
 169  * invocations of <tt>tryAcquire</tt> interspersed with other
 170  * computations before blocking.  This gives most of the benefits of
 171  * spins when exclusive synchronization is only briefly held, without
 172  * most of the liabilities when it isn't. If so desired, you can
 173  * augment this by preceding calls to acquire methods with
 174  * "fast-path" checks, possibly prechecking {@link #hasContended}
 175  * and/or {@link #hasQueuedThreads} to only do so if the synchronizer
 176  * is likely not to be contended.
 177  *
 178  * <p>This class provides an efficient and scalable basis for
 179  * synchronization in part by specializing its range of use to
 180  * synchronizers that can rely on <tt>int</tt> state, acquire, and
 181  * release parameters, and an internal FIFO wait queue. When this does
 182  * not suffice, you can build synchronizers from a lower level using
 183  * {@link java.util.concurrent.atomic atomic} classes, your own custom
 184  * {@link java.util.Queue} classes, and {@link LockSupport} blocking
 185  * support.
 186  *
 187  * <h3>Usage Examples</h3>
 188  *
 189  * <p>Here is a non-reentrant mutual exclusion lock class that uses
 190  * the value zero to represent the unlocked state, and one to
 191  * represent the locked state. While a non-reentrant lock
 192  * does not strictly require recording of the current owner
 193  * thread, this class does so anyway to make usage easier to monitor.
 194  * It also supports conditions and exposes
 195  * one of the instrumentation methods:
 196  *
 197  * <pre>
 198  * class Mutex implements Lock, java.io.Serializable {
 199  *
 200  *   // Our internal helper class
 201  *   private static class Sync extends AbstractQueuedSynchronizer {
 202  *     // Report whether in locked state
 203  *     protected boolean isHeldExclusively() {
 204  *       return getState() == 1;
 205  *     }
 206  *
 207  *     // Acquire the lock if state is zero
 208  *     public boolean tryAcquire(int acquires) {
 209  *       assert acquires == 1; // Otherwise unused
 210  *       if (compareAndSetState(0, 1)) {
 211  *         setExclusiveOwnerThread(Thread.currentThread());
 212  *         return true;
 213  *       }
 214  *       return false;
 215  *     }
 216  *
 217  *     // Release the lock by setting state to zero
 218  *     protected boolean tryRelease(int releases) {
 219  *       assert releases == 1; // Otherwise unused
 220  *       if (getState() == 0) throw new IllegalMonitorStateException();
 221  *       setExclusiveOwnerThread(null);
 222  *       setState(0);
 223  *       return true;
 224  *     }
 225  *
 226  *     // Provide a Condition
 227  *     Condition newCondition() { return new ConditionObject(); }
 228  *
 229  *     // Deserialize properly
 230  *     private void readObject(ObjectInputStream s)
 231  *         throws IOException, ClassNotFoundException {
 232  *       s.defaultReadObject();
 233  *       setState(0); // reset to unlocked state
 234  *     }
 235  *   }
 236  *
 237  *   // The sync object does all the hard work. We just forward to it.
 238  *   private final Sync sync = new Sync();
 239  *
 240  *   public void lock()                { sync.acquire(1); }
 241  *   public boolean tryLock()          { return sync.tryAcquire(1); }
 242  *   public void unlock()              { sync.release(1); }
 243  *   public Condition newCondition()   { return sync.newCondition(); }
 244  *   public boolean isLocked()         { return sync.isHeldExclusively(); }
 245  *   public boolean hasQueuedThreads() { return sync.hasQueuedThreads(); }
 246  *   public void lockInterruptibly() throws InterruptedException {
 247  *     sync.acquireInterruptibly(1);
 248  *   }
 249  *   public boolean tryLock(long timeout, TimeUnit unit)
 250  *       throws InterruptedException {
 251  *     return sync.tryAcquireNanos(1, unit.toNanos(timeout));
 252  *   }
 253  * }
 254  * </pre>
 255  *
 256  * <p>Here is a latch class that is like a {@link CountDownLatch}
 257  * except that it only requires a single <tt>signal</tt> to
 258  * fire. Because a latch is non-exclusive, it uses the <tt>shared</tt>
 259  * acquire and release methods.
 260  *
 261  * <pre>
 262  * class BooleanLatch {
 263  *
 264  *   private static class Sync extends AbstractQueuedSynchronizer {
 265  *     boolean isSignalled() { return getState() != 0; }
 266  *
 267  *     protected int tryAcquireShared(int ignore) {
 268  *       return isSignalled() ? 1 : -1;
 269  *     }
 270  *
 271  *     protected boolean tryReleaseShared(int ignore) {
 272  *       setState(1);
 273  *       return true;
 274  *     }
 275  *   }
 276  *
 277  *   private final Sync sync = new Sync();
 278  *   public boolean isSignalled() { return sync.isSignalled(); }
 279  *   public void signal()         { sync.releaseShared(1); }
 280  *   public void await() throws InterruptedException {
 281  *     sync.acquireSharedInterruptibly(1);
 282  *   }
 283  * }
 284  * </pre>
 285  *
 286  * @since 1.5
 287  * @author Doug Lea
 288  */
 289 public abstract class AbstractQueuedSynchronizer
 290     extends AbstractOwnableSynchronizer
 291     implements java.io.Serializable {
 292 
 293     private static final long serialVersionUID = 7373984972572414691L;
 294 
 295     /**
 296      * Creates a new <tt>AbstractQueuedSynchronizer</tt> instance
 297      * with initial synchronization state of zero.
 298      */
 299     protected AbstractQueuedSynchronizer() { }
 300 
 301     /**
 302      * Wait queue node class.
 303      *
 304      * <p>The wait queue is a variant of a "CLH" (Craig, Landin, and
 305      * Hagersten) lock queue. CLH locks are normally used for
 306      * spinlocks.  We instead use them for blocking synchronizers, but
 307      * use the same basic tactic of holding some of the control
 308      * information about a thread in the predecessor of its node.  A
 309      * "status" field in each node keeps track of whether a thread
 310      * should block.  A node is signalled when its predecessor
 311      * releases.  Each node of the queue otherwise serves as a
 312      * specific-notification-style monitor holding a single waiting
 313      * thread. The status field does NOT control whether threads are
 314      * granted locks etc though.  A thread may try to acquire if it is
 315      * first in the queue. But being first does not guarantee success;
 316      * it only gives the right to contend.  So the currently released
 317      * contender thread may need to rewait.
 318      *
 319      * <p>To enqueue into a CLH lock, you atomically splice it in as new
 320      * tail. To dequeue, you just set the head field.
 321      * <pre>
 322      *      +------+  prev +-----+       +-----+
 323      * head |      | <---- |     | <---- |     |  tail
 324      *      +------+       +-----+       +-----+
 325      * </pre>
 326      *
 327      * <p>Insertion into a CLH queue requires only a single atomic
 328      * operation on "tail", so there is a simple atomic point of
 329      * demarcation from unqueued to queued. Similarly, dequeing
 330      * involves only updating the "head". However, it takes a bit
 331      * more work for nodes to determine who their successors are,
 332      * in part to deal with possible cancellation due to timeouts
 333      * and interrupts.
 334      *
 335      * <p>The "prev" links (not used in original CLH locks), are mainly
 336      * needed to handle cancellation. If a node is cancelled, its
 337      * successor is (normally) relinked to a non-cancelled
 338      * predecessor. For explanation of similar mechanics in the case
 339      * of spin locks, see the papers by Scott and Scherer at
 340      * http://www.cs.rochester.edu/u/scott/synchronization/
 341      *
 342      * <p>We also use "next" links to implement blocking mechanics.
 343      * The thread id for each node is kept in its own node, so a
 344      * predecessor signals the next node to wake up by traversing
 345      * next link to determine which thread it is.  Determination of
 346      * successor must avoid races with newly queued nodes to set
 347      * the "next" fields of their predecessors.  This is solved
 348      * when necessary by checking backwards from the atomically
 349      * updated "tail" when a node's successor appears to be null.
 350      * (Or, said differently, the next-links are an optimization
 351      * so that we don't usually need a backward scan.)
 352      *
 353      * <p>Cancellation introduces some conservatism to the basic
 354      * algorithms.  Since we must poll for cancellation of other
 355      * nodes, we can miss noticing whether a cancelled node is
 356      * ahead or behind us. This is dealt with by always unparking
 357      * successors upon cancellation, allowing them to stabilize on
 358      * a new predecessor, unless we can identify an uncancelled
 359      * predecessor who will carry this responsibility.
 360      *
 361      * <p>CLH queues need a dummy header node to get started. But
 362      * we don't create them on construction, because it would be wasted
 363      * effort if there is never contention. Instead, the node
 364      * is constructed and head and tail pointers are set upon first
 365      * contention.
 366      *
 367      * <p>Threads waiting on Conditions use the same nodes, but
 368      * use an additional link. Conditions only need to link nodes
 369      * in simple (non-concurrent) linked queues because they are
 370      * only accessed when exclusively held.  Upon await, a node is
 371      * inserted into a condition queue.  Upon signal, the node is
 372      * transferred to the main queue.  A special value of status
 373      * field is used to mark which queue a node is on.
 374      *
 375      * <p>Thanks go to Dave Dice, Mark Moir, Victor Luchangco, Bill
 376      * Scherer and Michael Scott, along with members of JSR-166
 377      * expert group, for helpful ideas, discussions, and critiques
 378      * on the design of this class.
 379      */
 380     static final class Node {
 381         /** Marker to indicate a node is waiting in shared mode */
 382         static final Node SHARED = new Node();
 383         /** Marker to indicate a node is waiting in exclusive mode */
 384         static final Node EXCLUSIVE = null;
 385 
 386         /** waitStatus value to indicate thread has cancelled */
 387         static final int CANCELLED =  1;
 388         /** waitStatus value to indicate successor's thread needs unparking */
 389         static final int SIGNAL    = -1;
 390         /** waitStatus value to indicate thread is waiting on condition */
 391         static final int CONDITION = -2;
 392         /**
 393          * waitStatus value to indicate the next acquireShared should
 394          * unconditionally propagate
 395          */
 396         static final int PROPAGATE = -3;
 397 
 398         /**
 399          * Status field, taking on only the values:
 400          *   SIGNAL:     The successor of this node is (or will soon be)
 401          *               blocked (via park), so the current node must
 402          *               unpark its successor when it releases or
 403          *               cancels. To avoid races, acquire methods must
 404          *               first indicate they need a signal,
 405          *               then retry the atomic acquire, and then,
 406          *               on failure, block.
 407          *   CANCELLED:  This node is cancelled due to timeout or interrupt.
 408          *               Nodes never leave this state. In particular,
 409          *               a thread with cancelled node never again blocks.
 410          *   CONDITION:  This node is currently on a condition queue.
 411          *               It will not be used as a sync queue node
 412          *               until transferred, at which time the status
 413          *               will be set to 0. (Use of this value here has
 414          *               nothing to do with the other uses of the
 415          *               field, but simplifies mechanics.)
 416          *   PROPAGATE:  A releaseShared should be propagated to other
 417          *               nodes. This is set (for head node only) in
 418          *               doReleaseShared to ensure propagation
 419          *               continues, even if other operations have
 420          *               since intervened.
 421          *   0:          None of the above
 422          *
 423          * The values are arranged numerically to simplify use.
 424          * Non-negative values mean that a node doesn't need to
 425          * signal. So, most code doesn't need to check for particular
 426          * values, just for sign.
 427          *
 428          * The field is initialized to 0 for normal sync nodes, and
 429          * CONDITION for condition nodes.  It is modified using CAS
 430          * (or when possible, unconditional volatile writes).
 431          */
 432         volatile int waitStatus;
 433 
 434         /**
 435          * Link to predecessor node that current node/thread relies on
 436          * for checking waitStatus. Assigned during enqueing, and nulled
 437          * out (for sake of GC) only upon dequeuing.  Also, upon
 438          * cancellation of a predecessor, we short-circuit while
 439          * finding a non-cancelled one, which will always exist
 440          * because the head node is never cancelled: A node becomes
 441          * head only as a result of successful acquire. A
 442          * cancelled thread never succeeds in acquiring, and a thread only
 443          * cancels itself, not any other node.
 444          */
 445         volatile Node prev;
 446 
 447         /**
 448          * Link to the successor node that the current node/thread
 449          * unparks upon release. Assigned during enqueuing, adjusted
 450          * when bypassing cancelled predecessors, and nulled out (for
 451          * sake of GC) when dequeued.  The enq operation does not
 452          * assign next field of a predecessor until after attachment,
 453          * so seeing a null next field does not necessarily mean that
 454          * node is at end of queue. However, if a next field appears
 455          * to be null, we can scan prev's from the tail to
 456          * double-check.  The next field of cancelled nodes is set to
 457          * point to the node itself instead of null, to make life
 458          * easier for isOnSyncQueue.
 459          */
 460         volatile Node next;
 461 
 462         /**
 463          * The thread that enqueued this node.  Initialized on
 464          * construction and nulled out after use.
 465          */
 466         volatile Thread thread;
 467 
 468         /**
 469          * Link to next node waiting on condition, or the special
 470          * value SHARED.  Because condition queues are accessed only
 471          * when holding in exclusive mode, we just need a simple
 472          * linked queue to hold nodes while they are waiting on
 473          * conditions. They are then transferred to the queue to
 474          * re-acquire. And because conditions can only be exclusive,
 475          * we save a field by using special value to indicate shared
 476          * mode.
 477          */
 478         Node nextWaiter;
 479 
 480         /**
 481          * Returns true if node is waiting in shared mode
 482          */
 483         final boolean isShared() {
 484             return nextWaiter == SHARED;
 485         }
 486 
 487         /**
 488          * Returns previous node, or throws NullPointerException if null.
 489          * Use when predecessor cannot be null.  The null check could
 490          * be elided, but is present to help the VM.
 491          *
 492          * @return the predecessor of this node
 493          */
 494         final Node predecessor() throws NullPointerException {
 495             Node p = prev;
 496             if (p == null)
 497                 throw new NullPointerException();
 498             else
 499                 return p;
 500         }
 501 
 502         Node() {    // Used to establish initial head or SHARED marker
 503         }
 504 
 505         Node(Thread thread, Node mode) {     // Used by addWaiter
 506             this.nextWaiter = mode;
 507             this.thread = thread;
 508         }
 509 
 510         Node(Thread thread, int waitStatus) { // Used by Condition
 511             this.waitStatus = waitStatus;
 512             this.thread = thread;
 513         }
 514     }
 515 
 516     /**
 517      * Head of the wait queue, lazily initialized.  Except for
 518      * initialization, it is modified only via method setHead.  Note:
 519      * If head exists, its waitStatus is guaranteed not to be
 520      * CANCELLED.
 521      */
 522     private transient volatile Node head;
 523 
 524     /**
 525      * Tail of the wait queue, lazily initialized.  Modified only via
 526      * method enq to add new wait node.
 527      */
 528     private transient volatile Node tail;
 529 
 530     /**
 531      * The synchronization state.
 532      */
 533     private volatile int state;
 534 
 535     /**
 536      * Returns the current value of synchronization state.
 537      * This operation has memory semantics of a <tt>volatile</tt> read.
 538      * @return current state value
 539      */
 540     protected final int getState() {
 541         return state;
 542     }
 543 
 544     /**
 545      * Sets the value of synchronization state.
 546      * This operation has memory semantics of a <tt>volatile</tt> write.
 547      * @param newState the new state value
 548      */
 549     protected final void setState(int newState) {
 550         state = newState;
 551     }
 552 
 553     /**
 554      * Atomically sets synchronization state to the given updated
 555      * value if the current state value equals the expected value.
 556      * This operation has memory semantics of a <tt>volatile</tt> read
 557      * and write.
 558      *
 559      * @param expect the expected value
 560      * @param update the new value
 561      * @return true if successful. False return indicates that the actual
 562      *         value was not equal to the expected value.
 563      */
 564     protected final boolean compareAndSetState(int expect, int update) {
 565         // See below for intrinsics setup to support this
 566         return unsafe.compareAndSwapInt(this, stateOffset, expect, update);
 567     }
 568 
 569     // Queuing utilities
 570 
 571     /**
 572      * The number of nanoseconds for which it is faster to spin
 573      * rather than to use timed park. A rough estimate suffices
 574      * to improve responsiveness with very short timeouts.
 575      */
 576     static final long spinForTimeoutThreshold = 1000L;
 577 
 578     /**
 579      * Inserts node into queue, initializing if necessary. See picture above.
 580      * @param node the node to insert
 581      * @return node's predecessor
 582      */
 583     private Node enq(final Node node) {
 584         for (;;) {
 585             Node t = tail;
 586             if (t == null) { // Must initialize
 587                 if (compareAndSetHead(new Node()))
 588                     tail = head;
 589             } else {
 590                 node.prev = t;
 591                 if (compareAndSetTail(t, node)) {
 592                     t.next = node;
 593                     return t;
 594                 }
 595             }
 596         }
 597     }
 598 
 599     /**
 600      * Creates and enqueues node for current thread and given mode.
 601      *
 602      * @param mode Node.EXCLUSIVE for exclusive, Node.SHARED for shared
 603      * @return the new node
 604      */
 605     private Node addWaiter(Node mode) {
 606         Node node = new Node(Thread.currentThread(), mode);
 607         // Try the fast path of enq; backup to full enq on failure
 608         Node pred = tail;
 609         if (pred != null) {
 610             node.prev = pred;
 611             if (compareAndSetTail(pred, node)) {
 612                 pred.next = node;
 613                 return node;
 614             }
 615         }
 616         enq(node);
 617         return node;
 618     }
 619 
 620     /**
 621      * Sets head of queue to be node, thus dequeuing. Called only by
 622      * acquire methods.  Also nulls out unused fields for sake of GC
 623      * and to suppress unnecessary signals and traversals.
 624      *
 625      * @param node the node
 626      */
 627     private void setHead(Node node) {
 628         head = node;
 629         node.thread = null;
 630         node.prev = null;
 631     }
 632 
 633     /**
 634      * Wakes up node's successor, if one exists.
 635      *
 636      * @param node the node
 637      */
 638     private void unparkSuccessor(Node node) {
 639         /*
 640          * If status is negative (i.e., possibly needing signal) try
 641          * to clear in anticipation of signalling.  It is OK if this
 642          * fails or if status is changed by waiting thread.
 643          */
 644         int ws = node.waitStatus;
 645         if (ws < 0)
 646             compareAndSetWaitStatus(node, ws, 0);
 647 
 648         /*
 649          * Thread to unpark is held in successor, which is normally
 650          * just the next node.  But if cancelled or apparently null,
 651          * traverse backwards from tail to find the actual
 652          * non-cancelled successor.
 653          */
 654         Node s = node.next;
 655         if (s == null || s.waitStatus > 0) {
 656             s = null;
 657             for (Node t = tail; t != null && t != node; t = t.prev)
 658                 if (t.waitStatus <= 0)
 659                     s = t;
 660         }
 661         if (s != null)
 662             LockSupport.unpark(s.thread);
 663     }
 664 
 665     /**
 666      * Release action for shared mode -- signal successor and ensure
 667      * propagation. (Note: For exclusive mode, release just amounts
 668      * to calling unparkSuccessor of head if it needs signal.)
 669      */
 670     private void doReleaseShared() {
 671         /*
 672          * Ensure that a release propagates, even if there are other
 673          * in-progress acquires/releases.  This proceeds in the usual
 674          * way of trying to unparkSuccessor of head if it needs
 675          * signal. But if it does not, status is set to PROPAGATE to
 676          * ensure that upon release, propagation continues.
 677          * Additionally, we must loop in case a new node is added
 678          * while we are doing this. Also, unlike other uses of
 679          * unparkSuccessor, we need to know if CAS to reset status
 680          * fails, if so rechecking.
 681          */
 682         for (;;) {
 683             Node h = head;
 684             if (h != null && h != tail) {
 685                 int ws = h.waitStatus;
 686                 if (ws == Node.SIGNAL) {
 687                     if (!compareAndSetWaitStatus(h, Node.SIGNAL, 0))
 688                         continue;            // loop to recheck cases
 689                     unparkSuccessor(h);
 690                 }
 691                 else if (ws == 0 &&
 692                          !compareAndSetWaitStatus(h, 0, Node.PROPAGATE))
 693                     continue;                // loop on failed CAS
 694             }
 695             if (h == head)                   // loop if head changed
 696                 break;
 697         }
 698     }
 699 
 700     /**
 701      * Sets head of queue, and checks if successor may be waiting
 702      * in shared mode, if so propagating if either propagate > 0 or
 703      * PROPAGATE status was set.
 704      *
 705      * @param node the node
 706      * @param propagate the return value from a tryAcquireShared
 707      */
 708     private void setHeadAndPropagate(Node node, int propagate) {
 709         Node h = head; // Record old head for check below
 710         setHead(node);
 711         /*
 712          * Try to signal next queued node if:
 713          *   Propagation was indicated by caller,
 714          *     or was recorded (as h.waitStatus) by a previous operation
 715          *     (note: this uses sign-check of waitStatus because
 716          *      PROPAGATE status may transition to SIGNAL.)
 717          * and
 718          *   The next node is waiting in shared mode,
 719          *     or we don't know, because it appears null
 720          *
 721          * The conservatism in both of these checks may cause
 722          * unnecessary wake-ups, but only when there are multiple
 723          * racing acquires/releases, so most need signals now or soon
 724          * anyway.
 725          */
 726         if (propagate > 0 || h == null || h.waitStatus < 0) {
 727             Node s = node.next;
 728             if (s == null || s.isShared())
 729                 doReleaseShared();
 730         }
 731     }
 732 
 733     // Utilities for various versions of acquire
 734 
 735     /**
 736      * Cancels an ongoing attempt to acquire.
 737      *
 738      * @param node the node
 739      */
 740     private void cancelAcquire(Node node) {
 741         // Ignore if node doesn't exist
 742         if (node == null)
 743             return;
 744 
 745         node.thread = null;
 746 
 747         // Skip cancelled predecessors
 748         Node pred = node.prev;
 749         while (pred.waitStatus > 0)
 750             node.prev = pred = pred.prev;
 751 
 752         // predNext is the apparent node to unsplice. CASes below will
 753         // fail if not, in which case, we lost race vs another cancel
 754         // or signal, so no further action is necessary.
 755         Node predNext = pred.next;
 756 
 757         // Can use unconditional write instead of CAS here.
 758         // After this atomic step, other Nodes can skip past us.
 759         // Before, we are free of interference from other threads.
 760         node.waitStatus = Node.CANCELLED;
 761 
 762         // If we are the tail, remove ourselves.
 763         if (node == tail && compareAndSetTail(node, pred)) {
 764             compareAndSetNext(pred, predNext, null);
 765         } else {
 766             // If successor needs signal, try to set pred's next-link
 767             // so it will get one. Otherwise wake it up to propagate.
 768             int ws;
 769             if (pred != head &&
 770                 ((ws = pred.waitStatus) == Node.SIGNAL ||
 771                  (ws <= 0 && compareAndSetWaitStatus(pred, ws, Node.SIGNAL))) &&
 772                 pred.thread != null) {
 773                 Node next = node.next;
 774                 if (next != null && next.waitStatus <= 0)
 775                     compareAndSetNext(pred, predNext, next);
 776             } else {
 777                 unparkSuccessor(node);
 778             }
 779 
 780             node.next = node; // help GC
 781         }
 782     }
 783 
 784     /**
 785      * Checks and updates status for a node that failed to acquire.
 786      * Returns true if thread should block. This is the main signal
 787      * control in all acquire loops.  Requires that pred == node.prev
 788      *
 789      * @param pred node's predecessor holding status
 790      * @param node the node
 791      * @return {@code true} if thread should block
 792      */
 793     private static boolean shouldParkAfterFailedAcquire(Node pred, Node node) {
 794         int ws = pred.waitStatus;
 795         if (ws == Node.SIGNAL)
 796             /*
 797              * This node has already set status asking a release
 798              * to signal it, so it can safely park.
 799              */
 800             return true;
 801         if (ws > 0) {
 802             /*
 803              * Predecessor was cancelled. Skip over predecessors and
 804              * indicate retry.
 805              */
 806             do {
 807                 node.prev = pred = pred.prev;
 808             } while (pred.waitStatus > 0);
 809             pred.next = node;
 810         } else {
 811             /*
 812              * waitStatus must be 0 or PROPAGATE.  Indicate that we
 813              * need a signal, but don't park yet.  Caller will need to
 814              * retry to make sure it cannot acquire before parking.
 815              */
 816             compareAndSetWaitStatus(pred, ws, Node.SIGNAL);
 817         }
 818         return false;
 819     }
 820 
 821     /**
 822      * Convenience method to interrupt current thread.
 823      */
 824     private static void selfInterrupt() {
 825         Thread.currentThread().interrupt();
 826     }
 827 
 828     /**
 829      * Convenience method to park and then check if interrupted
 830      *
 831      * @return {@code true} if interrupted
 832      */
 833     private final boolean parkAndCheckInterrupt() {
 834         LockSupport.park(this);
 835         return Thread.interrupted();
 836     }
 837 
 838     /*
 839      * Various flavors of acquire, varying in exclusive/shared and
 840      * control modes.  Each is mostly the same, but annoyingly
 841      * different.  Only a little bit of factoring is possible due to
 842      * interactions of exception mechanics (including ensuring that we
 843      * cancel if tryAcquire throws exception) and other control, at
 844      * least not without hurting performance too much.
 845      */
 846 
 847     /**
 848      * Acquires in exclusive uninterruptible mode for thread already in
 849      * queue. Used by condition wait methods as well as acquire.
 850      *
 851      * @param node the node
 852      * @param arg the acquire argument
 853      * @return {@code true} if interrupted while waiting
 854      */
 855     final boolean acquireQueued(final Node node, int arg) {
 856         boolean failed = true;
 857         try {
 858             boolean interrupted = false;
 859             for (;;) {
 860                 final Node p = node.predecessor();
 861                 if (p == head && tryAcquire(arg)) {
 862                     setHead(node);
 863                     p.next = null; // help GC
 864                     failed = false;
 865                     return interrupted;
 866                 }
 867                 if (shouldParkAfterFailedAcquire(p, node) &&
 868                     parkAndCheckInterrupt())
 869                     interrupted = true;
 870             }
 871         } finally {
 872             if (failed)
 873                 cancelAcquire(node);
 874         }
 875     }
 876 
 877     /**
 878      * Acquires in exclusive interruptible mode.
 879      * @param arg the acquire argument
 880      */
 881     private void doAcquireInterruptibly(int arg)
 882         throws InterruptedException {
 883         final Node node = addWaiter(Node.EXCLUSIVE);
 884         boolean failed = true;
 885         try {
 886             for (;;) {
 887                 final Node p = node.predecessor();
 888                 if (p == head && tryAcquire(arg)) {
 889                     setHead(node);
 890                     p.next = null; // help GC
 891                     failed = false;
 892                     return;
 893                 }
 894                 if (shouldParkAfterFailedAcquire(p, node) &&
 895                     parkAndCheckInterrupt())
 896                     throw new InterruptedException();
 897             }
 898         } finally {
 899             if (failed)
 900                 cancelAcquire(node);
 901         }
 902     }
 903 
 904     /**
 905      * Acquires in exclusive timed mode.
 906      *
 907      * @param arg the acquire argument
 908      * @param nanosTimeout max wait time
 909      * @return {@code true} if acquired
 910      */
 911     private boolean doAcquireNanos(int arg, long nanosTimeout)
 912         throws InterruptedException {
 913         long lastTime = System.nanoTime();
 914         final Node node = addWaiter(Node.EXCLUSIVE);
 915         boolean failed = true;
 916         try {
 917             for (;;) {
 918                 final Node p = node.predecessor();
 919                 if (p == head && tryAcquire(arg)) {
 920                     setHead(node);
 921                     p.next = null; // help GC
 922                     failed = false;
 923                     return true;
 924                 }
 925                 if (nanosTimeout <= 0)
 926                     return false;
 927                 if (shouldParkAfterFailedAcquire(p, node) &&
 928                     nanosTimeout > spinForTimeoutThreshold)
 929                     LockSupport.parkNanos(this, nanosTimeout);
 930                 long now = System.nanoTime();
 931                 nanosTimeout -= now - lastTime;
 932                 lastTime = now;
 933                 if (Thread.interrupted())
 934                     throw new InterruptedException();
 935             }
 936         } finally {
 937             if (failed)
 938                 cancelAcquire(node);
 939         }
 940     }
 941 
 942     /**
 943      * Acquires in shared uninterruptible mode.
 944      * @param arg the acquire argument
 945      */
 946     private void doAcquireShared(int arg) {
 947         final Node node = addWaiter(Node.SHARED);
 948         boolean failed = true;
 949         try {
 950             boolean interrupted = false;
 951             for (;;) {
 952                 final Node p = node.predecessor();
 953                 if (p == head) {
 954                     int r = tryAcquireShared(arg);
 955                     if (r >= 0) {
 956                         setHeadAndPropagate(node, r);
 957                         p.next = null; // help GC
 958                         if (interrupted)
 959                             selfInterrupt();
 960                         failed = false;
 961                         return;
 962                     }
 963                 }
 964                 if (shouldParkAfterFailedAcquire(p, node) &&
 965                     parkAndCheckInterrupt())
 966                     interrupted = true;
 967             }
 968         } finally {
 969             if (failed)
 970                 cancelAcquire(node);
 971         }
 972     }
 973 
 974     /**
 975      * Acquires in shared interruptible mode.
 976      * @param arg the acquire argument
 977      */
 978     private void doAcquireSharedInterruptibly(int arg)
 979         throws InterruptedException {
 980         final Node node = addWaiter(Node.SHARED);
 981         boolean failed = true;
 982         try {
 983             for (;;) {
 984                 final Node p = node.predecessor();
 985                 if (p == head) {
 986                     int r = tryAcquireShared(arg);
 987                     if (r >= 0) {
 988                         setHeadAndPropagate(node, r);
 989                         p.next = null; // help GC
 990                         failed = false;
 991                         return;
 992                     }
 993                 }
 994                 if (shouldParkAfterFailedAcquire(p, node) &&
 995                     parkAndCheckInterrupt())
 996                     throw new InterruptedException();
 997             }
 998         } finally {
 999             if (failed)
1000                 cancelAcquire(node);
1001         }
1002     }
1003 
1004     /**
1005      * Acquires in shared timed mode.
1006      *
1007      * @param arg the acquire argument
1008      * @param nanosTimeout max wait time
1009      * @return {@code true} if acquired
1010      */
1011     private boolean doAcquireSharedNanos(int arg, long nanosTimeout)
1012         throws InterruptedException {
1013 
1014         long lastTime = System.nanoTime();
1015         final Node node = addWaiter(Node.SHARED);
1016         boolean failed = true;
1017         try {
1018             for (;;) {
1019                 final Node p = node.predecessor();
1020                 if (p == head) {
1021                     int r = tryAcquireShared(arg);
1022                     if (r >= 0) {
1023                         setHeadAndPropagate(node, r);
1024                         p.next = null; // help GC
1025                         failed = false;
1026                         return true;
1027                     }
1028                 }
1029                 if (nanosTimeout <= 0)
1030                     return false;
1031                 if (shouldParkAfterFailedAcquire(p, node) &&
1032                     nanosTimeout > spinForTimeoutThreshold)
1033                     LockSupport.parkNanos(this, nanosTimeout);
1034                 long now = System.nanoTime();
1035                 nanosTimeout -= now - lastTime;
1036                 lastTime = now;
1037                 if (Thread.interrupted())
1038                     throw new InterruptedException();
1039             }
1040         } finally {
1041             if (failed)
1042                 cancelAcquire(node);
1043         }
1044     }
1045 
1046     // Main exported methods
1047 
1048     /**
1049      * Attempts to acquire in exclusive mode. This method should query
1050      * if the state of the object permits it to be acquired in the
1051      * exclusive mode, and if so to acquire it.
1052      *
1053      * <p>This method is always invoked by the thread performing
1054      * acquire.  If this method reports failure, the acquire method
1055      * may queue the thread, if it is not already queued, until it is
1056      * signalled by a release from some other thread. This can be used
1057      * to implement method {@link Lock#tryLock()}.
1058      *
1059      * <p>The default
1060      * implementation throws {@link UnsupportedOperationException}.
1061      *
1062      * @param arg the acquire argument. This value is always the one
1063      *        passed to an acquire method, or is the value saved on entry
1064      *        to a condition wait.  The value is otherwise uninterpreted
1065      *        and can represent anything you like.
1066      * @return {@code true} if successful. Upon success, this object has
1067      *         been acquired.
1068      * @throws IllegalMonitorStateException if acquiring would place this
1069      *         synchronizer in an illegal state. This exception must be
1070      *         thrown in a consistent fashion for synchronization to work
1071      *         correctly.
1072      * @throws UnsupportedOperationException if exclusive mode is not supported
1073      */
1074     protected boolean tryAcquire(int arg) {
1075         throw new UnsupportedOperationException();
1076     }
1077 
1078     /**
1079      * Attempts to set the state to reflect a release in exclusive
1080      * mode.
1081      *
1082      * <p>This method is always invoked by the thread performing release.
1083      *
1084      * <p>The default implementation throws
1085      * {@link UnsupportedOperationException}.
1086      *
1087      * @param arg the release argument. This value is always the one
1088      *        passed to a release method, or the current state value upon
1089      *        entry to a condition wait.  The value is otherwise
1090      *        uninterpreted and can represent anything you like.
1091      * @return {@code true} if this object is now in a fully released
1092      *         state, so that any waiting threads may attempt to acquire;
1093      *         and {@code false} otherwise.
1094      * @throws IllegalMonitorStateException if releasing would place this
1095      *         synchronizer in an illegal state. This exception must be
1096      *         thrown in a consistent fashion for synchronization to work
1097      *         correctly.
1098      * @throws UnsupportedOperationException if exclusive mode is not supported
1099      */
1100     protected boolean tryRelease(int arg) {
1101         throw new UnsupportedOperationException();
1102     }
1103 
1104     /**
1105      * Attempts to acquire in shared mode. This method should query if
1106      * the state of the object permits it to be acquired in the shared
1107      * mode, and if so to acquire it.
1108      *
1109      * <p>This method is always invoked by the thread performing
1110      * acquire.  If this method reports failure, the acquire method
1111      * may queue the thread, if it is not already queued, until it is
1112      * signalled by a release from some other thread.
1113      *
1114      * <p>The default implementation throws {@link
1115      * UnsupportedOperationException}.
1116      *
1117      * @param arg the acquire argument. This value is always the one
1118      *        passed to an acquire method, or is the value saved on entry
1119      *        to a condition wait.  The value is otherwise uninterpreted
1120      *        and can represent anything you like.
1121      * @return a negative value on failure; zero if acquisition in shared
1122      *         mode succeeded but no subsequent shared-mode acquire can
1123      *         succeed; and a positive value if acquisition in shared
1124      *         mode succeeded and subsequent shared-mode acquires might
1125      *         also succeed, in which case a subsequent waiting thread
1126      *         must check availability. (Support for three different
1127      *         return values enables this method to be used in contexts
1128      *         where acquires only sometimes act exclusively.)  Upon
1129      *         success, this object has been acquired.
1130      * @throws IllegalMonitorStateException if acquiring would place this
1131      *         synchronizer in an illegal state. This exception must be
1132      *         thrown in a consistent fashion for synchronization to work
1133      *         correctly.
1134      * @throws UnsupportedOperationException if shared mode is not supported
1135      */
1136     protected int tryAcquireShared(int arg) {
1137         throw new UnsupportedOperationException();
1138     }
1139 
1140     /**
1141      * Attempts to set the state to reflect a release in shared mode.
1142      *
1143      * <p>This method is always invoked by the thread performing release.
1144      *
1145      * <p>The default implementation throws
1146      * {@link UnsupportedOperationException}.
1147      *
1148      * @param arg the release argument. This value is always the one
1149      *        passed to a release method, or the current state value upon
1150      *        entry to a condition wait.  The value is otherwise
1151      *        uninterpreted and can represent anything you like.
1152      * @return {@code true} if this release of shared mode may permit a
1153      *         waiting acquire (shared or exclusive) to succeed; and
1154      *         {@code false} otherwise
1155      * @throws IllegalMonitorStateException if releasing would place this
1156      *         synchronizer in an illegal state. This exception must be
1157      *         thrown in a consistent fashion for synchronization to work
1158      *         correctly.
1159      * @throws UnsupportedOperationException if shared mode is not supported
1160      */
1161     protected boolean tryReleaseShared(int arg) {
1162         throw new UnsupportedOperationException();
1163     }
1164 
1165     /**
1166      * Returns {@code true} if synchronization is held exclusively with
1167      * respect to the current (calling) thread.  This method is invoked
1168      * upon each call to a non-waiting {@link ConditionObject} method.
1169      * (Waiting methods instead invoke {@link #release}.)
1170      *
1171      * <p>The default implementation throws {@link
1172      * UnsupportedOperationException}. This method is invoked
1173      * internally only within {@link ConditionObject} methods, so need
1174      * not be defined if conditions are not used.
1175      *
1176      * @return {@code true} if synchronization is held exclusively;
1177      *         {@code false} otherwise
1178      * @throws UnsupportedOperationException if conditions are not supported
1179      */
1180     protected boolean isHeldExclusively() {
1181         throw new UnsupportedOperationException();
1182     }
1183 
1184     /**
1185      * Acquires in exclusive mode, ignoring interrupts.  Implemented
1186      * by invoking at least once {@link #tryAcquire},
1187      * returning on success.  Otherwise the thread is queued, possibly
1188      * repeatedly blocking and unblocking, invoking {@link
1189      * #tryAcquire} until success.  This method can be used
1190      * to implement method {@link Lock#lock}.
1191      *
1192      * @param arg the acquire argument.  This value is conveyed to
1193      *        {@link #tryAcquire} but is otherwise uninterpreted and
1194      *        can represent anything you like.
1195      */
1196     public final void acquire(int arg) {
1197         if (!tryAcquire(arg) &&
1198             acquireQueued(addWaiter(Node.EXCLUSIVE), arg))
1199             selfInterrupt();
1200     }
1201 
1202     /**
1203      * Acquires in exclusive mode, aborting if interrupted.
1204      * Implemented by first checking interrupt status, then invoking
1205      * at least once {@link #tryAcquire}, returning on
1206      * success.  Otherwise the thread is queued, possibly repeatedly
1207      * blocking and unblocking, invoking {@link #tryAcquire}
1208      * until success or the thread is interrupted.  This method can be
1209      * used to implement method {@link Lock#lockInterruptibly}.
1210      *
1211      * @param arg the acquire argument.  This value is conveyed to
1212      *        {@link #tryAcquire} but is otherwise uninterpreted and
1213      *        can represent anything you like.
1214      * @throws InterruptedException if the current thread is interrupted
1215      */
1216     public final void acquireInterruptibly(int arg)
1217             throws InterruptedException {
1218         if (Thread.interrupted())
1219             throw new InterruptedException();
1220         if (!tryAcquire(arg))
1221             doAcquireInterruptibly(arg);
1222     }
1223 
1224     /**
1225      * Attempts to acquire in exclusive mode, aborting if interrupted,
1226      * and failing if the given timeout elapses.  Implemented by first
1227      * checking interrupt status, then invoking at least once {@link
1228      * #tryAcquire}, returning on success.  Otherwise, the thread is
1229      * queued, possibly repeatedly blocking and unblocking, invoking
1230      * {@link #tryAcquire} until success or the thread is interrupted
1231      * or the timeout elapses.  This method can be used to implement
1232      * method {@link Lock#tryLock(long, TimeUnit)}.
1233      *
1234      * @param arg the acquire argument.  This value is conveyed to
1235      *        {@link #tryAcquire} but is otherwise uninterpreted and
1236      *        can represent anything you like.
1237      * @param nanosTimeout the maximum number of nanoseconds to wait
1238      * @return {@code true} if acquired; {@code false} if timed out
1239      * @throws InterruptedException if the current thread is interrupted
1240      */
1241     public final boolean tryAcquireNanos(int arg, long nanosTimeout)
1242             throws InterruptedException {
1243         if (Thread.interrupted())
1244             throw new InterruptedException();
1245         return tryAcquire(arg) ||
1246             doAcquireNanos(arg, nanosTimeout);
1247     }
1248 
1249     /**
1250      * Releases in exclusive mode.  Implemented by unblocking one or
1251      * more threads if {@link #tryRelease} returns true.
1252      * This method can be used to implement method {@link Lock#unlock}.
1253      *
1254      * @param arg the release argument.  This value is conveyed to
1255      *        {@link #tryRelease} but is otherwise uninterpreted and
1256      *        can represent anything you like.
1257      * @return the value returned from {@link #tryRelease}
1258      */
1259     public final boolean release(int arg) {
1260         if (tryRelease(arg)) {
1261             Node h = head;
1262             if (h != null && h.waitStatus != 0)
1263                 unparkSuccessor(h);
1264             return true;
1265         }
1266         return false;
1267     }
1268 
1269     /**
1270      * Acquires in shared mode, ignoring interrupts.  Implemented by
1271      * first invoking at least once {@link #tryAcquireShared},
1272      * returning on success.  Otherwise the thread is queued, possibly
1273      * repeatedly blocking and unblocking, invoking {@link
1274      * #tryAcquireShared} until success.
1275      *
1276      * @param arg the acquire argument.  This value is conveyed to
1277      *        {@link #tryAcquireShared} but is otherwise uninterpreted
1278      *        and can represent anything you like.
1279      */
1280     public final void acquireShared(int arg) {
1281         if (tryAcquireShared(arg) < 0)
1282             doAcquireShared(arg);
1283     }
1284 
1285     /**
1286      * Acquires in shared mode, aborting if interrupted.  Implemented
1287      * by first checking interrupt status, then invoking at least once
1288      * {@link #tryAcquireShared}, returning on success.  Otherwise the
1289      * thread is queued, possibly repeatedly blocking and unblocking,
1290      * invoking {@link #tryAcquireShared} until success or the thread
1291      * is interrupted.
1292      * @param arg the acquire argument
1293      * This value is conveyed to {@link #tryAcquireShared} but is
1294      * otherwise uninterpreted and can represent anything
1295      * you like.
1296      * @throws InterruptedException if the current thread is interrupted
1297      */
1298     public final void acquireSharedInterruptibly(int arg)
1299             throws InterruptedException {
1300         if (Thread.interrupted())
1301             throw new InterruptedException();
1302         if (tryAcquireShared(arg) < 0)
1303             doAcquireSharedInterruptibly(arg);
1304     }
1305 
1306     /**
1307      * Attempts to acquire in shared mode, aborting if interrupted, and
1308      * failing if the given timeout elapses.  Implemented by first
1309      * checking interrupt status, then invoking at least once {@link
1310      * #tryAcquireShared}, returning on success.  Otherwise, the
1311      * thread is queued, possibly repeatedly blocking and unblocking,
1312      * invoking {@link #tryAcquireShared} until success or the thread
1313      * is interrupted or the timeout elapses.
1314      *
1315      * @param arg the acquire argument.  This value is conveyed to
1316      *        {@link #tryAcquireShared} but is otherwise uninterpreted
1317      *        and can represent anything you like.
1318      * @param nanosTimeout the maximum number of nanoseconds to wait
1319      * @return {@code true} if acquired; {@code false} if timed out
1320      * @throws InterruptedException if the current thread is interrupted
1321      */
1322     public final boolean tryAcquireSharedNanos(int arg, long nanosTimeout)
1323             throws InterruptedException {
1324         if (Thread.interrupted())
1325             throw new InterruptedException();
1326         return tryAcquireShared(arg) >= 0 ||
1327             doAcquireSharedNanos(arg, nanosTimeout);
1328     }
1329 
1330     /**
1331      * Releases in shared mode.  Implemented by unblocking one or more
1332      * threads if {@link #tryReleaseShared} returns true.
1333      *
1334      * @param arg the release argument.  This value is conveyed to
1335      *        {@link #tryReleaseShared} but is otherwise uninterpreted
1336      *        and can represent anything you like.
1337      * @return the value returned from {@link #tryReleaseShared}
1338      */
1339     public final boolean releaseShared(int arg) {
1340         if (tryReleaseShared(arg)) {
1341             doReleaseShared();
1342             return true;
1343         }
1344         return false;
1345     }
1346 
1347     // Queue inspection methods
1348 
1349     /**
1350      * Queries whether any threads are waiting to acquire. Note that
1351      * because cancellations due to interrupts and timeouts may occur
1352      * at any time, a {@code true} return does not guarantee that any
1353      * other thread will ever acquire.
1354      *
1355      * <p>In this implementation, this operation returns in
1356      * constant time.
1357      *
1358      * @return {@code true} if there may be other threads waiting to acquire
1359      */
1360     public final boolean hasQueuedThreads() {
1361         return head != tail;
1362     }
1363 
1364     /**
1365      * Queries whether any threads have ever contended to acquire this
1366      * synchronizer; that is if an acquire method has ever blocked.
1367      *
1368      * <p>In this implementation, this operation returns in
1369      * constant time.
1370      *
1371      * @return {@code true} if there has ever been contention
1372      */
1373     public final boolean hasContended() {
1374         return head != null;
1375     }
1376 
1377     /**
1378      * Returns the first (longest-waiting) thread in the queue, or
1379      * {@code null} if no threads are currently queued.
1380      *
1381      * <p>In this implementation, this operation normally returns in
1382      * constant time, but may iterate upon contention if other threads are
1383      * concurrently modifying the queue.
1384      *
1385      * @return the first (longest-waiting) thread in the queue, or
1386      *         {@code null} if no threads are currently queued
1387      */
1388     public final Thread getFirstQueuedThread() {
1389         // handle only fast path, else relay
1390         return (head == tail) ? null : fullGetFirstQueuedThread();
1391     }
1392 
1393     /**
1394      * Version of getFirstQueuedThread called when fastpath fails
1395      */
1396     private Thread fullGetFirstQueuedThread() {
1397         /*
1398          * The first node is normally head.next. Try to get its
1399          * thread field, ensuring consistent reads: If thread
1400          * field is nulled out or s.prev is no longer head, then
1401          * some other thread(s) concurrently performed setHead in
1402          * between some of our reads. We try this twice before
1403          * resorting to traversal.
1404          */
1405         Node h, s;
1406         Thread st;
1407         if (((h = head) != null && (s = h.next) != null &&
1408              s.prev == head && (st = s.thread) != null) ||
1409             ((h = head) != null && (s = h.next) != null &&
1410              s.prev == head && (st = s.thread) != null))
1411             return st;
1412 
1413         /*
1414          * Head's next field might not have been set yet, or may have
1415          * been unset after setHead. So we must check to see if tail
1416          * is actually first node. If not, we continue on, safely
1417          * traversing from tail back to head to find first,
1418          * guaranteeing termination.
1419          */
1420 
1421         Node t = tail;
1422         Thread firstThread = null;
1423         while (t != null && t != head) {
1424             Thread tt = t.thread;
1425             if (tt != null)
1426                 firstThread = tt;
1427             t = t.prev;
1428         }
1429         return firstThread;
1430     }
1431 
1432     /**
1433      * Returns true if the given thread is currently queued.
1434      *
1435      * <p>This implementation traverses the queue to determine
1436      * presence of the given thread.
1437      *
1438      * @param thread the thread
1439      * @return {@code true} if the given thread is on the queue
1440      * @throws NullPointerException if the thread is null
1441      */
1442     public final boolean isQueued(Thread thread) {
1443         if (thread == null)
1444             throw new NullPointerException();
1445         for (Node p = tail; p != null; p = p.prev)
1446             if (p.thread == thread)
1447                 return true;
1448         return false;
1449     }
1450 
1451     /**
1452      * Returns {@code true} if the apparent first queued thread, if one
1453      * exists, is waiting in exclusive mode.  If this method returns
1454      * {@code true}, and the current thread is attempting to acquire in
1455      * shared mode (that is, this method is invoked from {@link
1456      * #tryAcquireShared}) then it is guaranteed that the current thread
1457      * is not the first queued thread.  Used only as a heuristic in
1458      * ReentrantReadWriteLock.
1459      */
1460     final boolean apparentlyFirstQueuedIsExclusive() {
1461         Node h, s;
1462         return (h = head) != null &&
1463             (s = h.next)  != null &&
1464             !s.isShared()         &&
1465             s.thread != null;
1466     }
1467 
1468     /**
1469      * Queries whether any threads have been waiting to acquire longer
1470      * than the current thread.
1471      *
1472      * <p>An invocation of this method is equivalent to (but may be
1473      * more efficient than):
1474      *  <pre> {@code
1475      * getFirstQueuedThread() != Thread.currentThread() &&
1476      * hasQueuedThreads()}</pre>
1477      *
1478      * <p>Note that because cancellations due to interrupts and
1479      * timeouts may occur at any time, a {@code true} return does not
1480      * guarantee that some other thread will acquire before the current
1481      * thread.  Likewise, it is possible for another thread to win a
1482      * race to enqueue after this method has returned {@code false},
1483      * due to the queue being empty.
1484      *
1485      * <p>This method is designed to be used by a fair synchronizer to
1486      * avoid <a href="AbstractQueuedSynchronizer#barging">barging</a>.
1487      * Such a synchronizer's {@link #tryAcquire} method should return
1488      * {@code false}, and its {@link #tryAcquireShared} method should
1489      * return a negative value, if this method returns {@code true}
1490      * (unless this is a reentrant acquire).  For example, the {@code
1491      * tryAcquire} method for a fair, reentrant, exclusive mode
1492      * synchronizer might look like this:
1493      *
1494      *  <pre> {@code
1495      * protected boolean tryAcquire(int arg) {
1496      *   if (isHeldExclusively()) {
1497      *     // A reentrant acquire; increment hold count
1498      *     return true;
1499      *   } else if (hasQueuedPredecessors()) {
1500      *     return false;
1501      *   } else {
1502      *     // try to acquire normally
1503      *   }
1504      * }}</pre>
1505      *
1506      * @return {@code true} if there is a queued thread preceding the
1507      *         current thread, and {@code false} if the current thread
1508      *         is at the head of the queue or the queue is empty
1509      * @since 1.7
1510      */
1511     public final boolean hasQueuedPredecessors() {
1512         // The correctness of this depends on head being initialized
1513         // before tail and on head.next being accurate if the current
1514         // thread is first in queue.
1515         Node t = tail; // Read fields in reverse initialization order
1516         Node h = head;
1517         Node s;
1518         return h != t &&
1519             ((s = h.next) == null || s.thread != Thread.currentThread());
1520     }
1521 
1522 
1523     // Instrumentation and monitoring methods
1524 
1525     /**
1526      * Returns an estimate of the number of threads waiting to
1527      * acquire.  The value is only an estimate because the number of
1528      * threads may change dynamically while this method traverses
1529      * internal data structures.  This method is designed for use in
1530      * monitoring system state, not for synchronization
1531      * control.
1532      *
1533      * @return the estimated number of threads waiting to acquire
1534      */
1535     public final int getQueueLength() {
1536         int n = 0;
1537         for (Node p = tail; p != null; p = p.prev) {
1538             if (p.thread != null)
1539                 ++n;
1540         }
1541         return n;
1542     }
1543 
1544     /**
1545      * Returns a collection containing threads that may be waiting to
1546      * acquire.  Because the actual set of threads may change
1547      * dynamically while constructing this result, the returned
1548      * collection is only a best-effort estimate.  The elements of the
1549      * returned collection are in no particular order.  This method is
1550      * designed to facilitate construction of subclasses that provide
1551      * more extensive monitoring facilities.
1552      *
1553      * @return the collection of threads
1554      */
1555     public final Collection<Thread> getQueuedThreads() {
1556         ArrayList<Thread> list = new ArrayList<Thread>();
1557         for (Node p = tail; p != null; p = p.prev) {
1558             Thread t = p.thread;
1559             if (t != null)
1560                 list.add(t);
1561         }
1562         return list;
1563     }
1564 
1565     /**
1566      * Returns a collection containing threads that may be waiting to
1567      * acquire in exclusive mode. This has the same properties
1568      * as {@link #getQueuedThreads} except that it only returns
1569      * those threads waiting due to an exclusive acquire.
1570      *
1571      * @return the collection of threads
1572      */
1573     public final Collection<Thread> getExclusiveQueuedThreads() {
1574         ArrayList<Thread> list = new ArrayList<Thread>();
1575         for (Node p = tail; p != null; p = p.prev) {
1576             if (!p.isShared()) {
1577                 Thread t = p.thread;
1578                 if (t != null)
1579                     list.add(t);
1580             }
1581         }
1582         return list;
1583     }
1584 
1585     /**
1586      * Returns a collection containing threads that may be waiting to
1587      * acquire in shared mode. This has the same properties
1588      * as {@link #getQueuedThreads} except that it only returns
1589      * those threads waiting due to a shared acquire.
1590      *
1591      * @return the collection of threads
1592      */
1593     public final Collection<Thread> getSharedQueuedThreads() {
1594         ArrayList<Thread> list = new ArrayList<Thread>();
1595         for (Node p = tail; p != null; p = p.prev) {
1596             if (p.isShared()) {
1597                 Thread t = p.thread;
1598                 if (t != null)
1599                     list.add(t);
1600             }
1601         }
1602         return list;
1603     }
1604 
1605     /**
1606      * Returns a string identifying this synchronizer, as well as its state.
1607      * The state, in brackets, includes the String {@code "State ="}
1608      * followed by the current value of {@link #getState}, and either
1609      * {@code "nonempty"} or {@code "empty"} depending on whether the
1610      * queue is empty.
1611      *
1612      * @return a string identifying this synchronizer, as well as its state
1613      */
1614     public String toString() {
1615         int s = getState();
1616         String q  = hasQueuedThreads() ? "non" : "";
1617         return super.toString() +
1618             "[State = " + s + ", " + q + "empty queue]";
1619     }
1620 
1621 
1622     // Internal support methods for Conditions
1623 
1624     /**
1625      * Returns true if a node, always one that was initially placed on
1626      * a condition queue, is now waiting to reacquire on sync queue.
1627      * @param node the node
1628      * @return true if is reacquiring
1629      */
1630     final boolean isOnSyncQueue(Node node) {
1631         if (node.waitStatus == Node.CONDITION || node.prev == null)
1632             return false;
1633         if (node.next != null) // If has successor, it must be on queue
1634             return true;
1635         /*
1636          * node.prev can be non-null, but not yet on queue because
1637          * the CAS to place it on queue can fail. So we have to
1638          * traverse from tail to make sure it actually made it.  It
1639          * will always be near the tail in calls to this method, and
1640          * unless the CAS failed (which is unlikely), it will be
1641          * there, so we hardly ever traverse much.
1642          */
1643         return findNodeFromTail(node);
1644     }
1645 
1646     /**
1647      * Returns true if node is on sync queue by searching backwards from tail.
1648      * Called only when needed by isOnSyncQueue.
1649      * @return true if present
1650      */
1651     private boolean findNodeFromTail(Node node) {
1652         Node t = tail;
1653         for (;;) {
1654             if (t == node)
1655                 return true;
1656             if (t == null)
1657                 return false;
1658             t = t.prev;
1659         }
1660     }
1661 
1662     /**
1663      * Transfers a node from a condition queue onto sync queue.
1664      * Returns true if successful.
1665      * @param node the node
1666      * @return true if successfully transferred (else the node was
1667      * cancelled before signal).
1668      */
1669     final boolean transferForSignal(Node node) {
1670         /*
1671          * If cannot change waitStatus, the node has been cancelled.
1672          */
1673         if (!compareAndSetWaitStatus(node, Node.CONDITION, 0))
1674             return false;
1675 
1676         /*
1677          * Splice onto queue and try to set waitStatus of predecessor to
1678          * indicate that thread is (probably) waiting. If cancelled or
1679          * attempt to set waitStatus fails, wake up to resync (in which
1680          * case the waitStatus can be transiently and harmlessly wrong).
1681          */
1682         Node p = enq(node);
1683         int ws = p.waitStatus;
1684         if (ws > 0 || !compareAndSetWaitStatus(p, ws, Node.SIGNAL))
1685             LockSupport.unpark(node.thread);
1686         return true;
1687     }
1688 
1689     /**
1690      * Transfers node, if necessary, to sync queue after a cancelled
1691      * wait. Returns true if thread was cancelled before being
1692      * signalled.
1693      * @param current the waiting thread
1694      * @param node its node
1695      * @return true if cancelled before the node was signalled
1696      */
1697     final boolean transferAfterCancelledWait(Node node) {
1698         if (compareAndSetWaitStatus(node, Node.CONDITION, 0)) {
1699             enq(node);
1700             return true;
1701         }
1702         /*
1703          * If we lost out to a signal(), then we can't proceed
1704          * until it finishes its enq().  Cancelling during an
1705          * incomplete transfer is both rare and transient, so just
1706          * spin.
1707          */
1708         while (!isOnSyncQueue(node))
1709             Thread.yield();
1710         return false;
1711     }
1712 
1713     /**
1714      * Invokes release with current state value; returns saved state.
1715      * Cancels node and throws exception on failure.
1716      * @param node the condition node for this wait
1717      * @return previous sync state
1718      */
1719     final int fullyRelease(Node node) {
1720         boolean failed = true;
1721         try {
1722             int savedState = getState();
1723             if (release(savedState)) {
1724                 failed = false;
1725                 return savedState;
1726             } else {
1727                 throw new IllegalMonitorStateException();
1728             }
1729         } finally {
1730             if (failed)
1731                 node.waitStatus = Node.CANCELLED;
1732         }
1733     }
1734 
1735     // Instrumentation methods for conditions
1736 
1737     /**
1738      * Queries whether the given ConditionObject
1739      * uses this synchronizer as its lock.
1740      *
1741      * @param condition the condition
1742      * @return <tt>true</tt> if owned
1743      * @throws NullPointerException if the condition is null
1744      */
1745     public final boolean owns(ConditionObject condition) {
1746         if (condition == null)
1747             throw new NullPointerException();
1748         return condition.isOwnedBy(this);
1749     }
1750 
1751     /**
1752      * Queries whether any threads are waiting on the given condition
1753      * associated with this synchronizer. Note that because timeouts
1754      * and interrupts may occur at any time, a <tt>true</tt> return
1755      * does not guarantee that a future <tt>signal</tt> will awaken
1756      * any threads.  This method is designed primarily for use in
1757      * monitoring of the system state.
1758      *
1759      * @param condition the condition
1760      * @return <tt>true</tt> if there are any waiting threads
1761      * @throws IllegalMonitorStateException if exclusive synchronization
1762      *         is not held
1763      * @throws IllegalArgumentException if the given condition is
1764      *         not associated with this synchronizer
1765      * @throws NullPointerException if the condition is null
1766      */
1767     public final boolean hasWaiters(ConditionObject condition) {
1768         if (!owns(condition))
1769             throw new IllegalArgumentException("Not owner");
1770         return condition.hasWaiters();
1771     }
1772 
1773     /**
1774      * Returns an estimate of the number of threads waiting on the
1775      * given condition associated with this synchronizer. Note that
1776      * because timeouts and interrupts may occur at any time, the
1777      * estimate serves only as an upper bound on the actual number of
1778      * waiters.  This method is designed for use in monitoring of the
1779      * system state, not for synchronization control.
1780      *
1781      * @param condition the condition
1782      * @return the estimated number of waiting threads
1783      * @throws IllegalMonitorStateException if exclusive synchronization
1784      *         is not held
1785      * @throws IllegalArgumentException if the given condition is
1786      *         not associated with this synchronizer
1787      * @throws NullPointerException if the condition is null
1788      */
1789     public final int getWaitQueueLength(ConditionObject condition) {
1790         if (!owns(condition))
1791             throw new IllegalArgumentException("Not owner");
1792         return condition.getWaitQueueLength();
1793     }
1794 
1795     /**
1796      * Returns a collection containing those threads that may be
1797      * waiting on the given condition associated with this
1798      * synchronizer.  Because the actual set of threads may change
1799      * dynamically while constructing this result, the returned
1800      * collection is only a best-effort estimate. The elements of the
1801      * returned collection are in no particular order.
1802      *
1803      * @param condition the condition
1804      * @return the collection of threads
1805      * @throws IllegalMonitorStateException if exclusive synchronization
1806      *         is not held
1807      * @throws IllegalArgumentException if the given condition is
1808      *         not associated with this synchronizer
1809      * @throws NullPointerException if the condition is null
1810      */
1811     public final Collection<Thread> getWaitingThreads(ConditionObject condition) {
1812         if (!owns(condition))
1813             throw new IllegalArgumentException("Not owner");
1814         return condition.getWaitingThreads();
1815     }
1816 
1817     /**
1818      * Condition implementation for a {@link
1819      * AbstractQueuedSynchronizer} serving as the basis of a {@link
1820      * Lock} implementation.
1821      *
1822      * <p>Method documentation for this class describes mechanics,
1823      * not behavioral specifications from the point of view of Lock
1824      * and Condition users. Exported versions of this class will in
1825      * general need to be accompanied by documentation describing
1826      * condition semantics that rely on those of the associated
1827      * <tt>AbstractQueuedSynchronizer</tt>.
1828      *
1829      * <p>This class is Serializable, but all fields are transient,
1830      * so deserialized conditions have no waiters.
1831      */
1832     public class ConditionObject implements Condition, java.io.Serializable {
1833         private static final long serialVersionUID = 1173984872572414699L;
1834         /** First node of condition queue. */
1835         private transient Node firstWaiter;
1836         /** Last node of condition queue. */
1837         private transient Node lastWaiter;
1838 
1839         /**
1840          * Creates a new <tt>ConditionObject</tt> instance.
1841          */
1842         public ConditionObject() { }
1843 
1844         // Internal methods
1845 
1846         /**
1847          * Adds a new waiter to wait queue.
1848          * @return its new wait node
1849          */
1850         private Node addConditionWaiter() {
1851             Node t = lastWaiter;
1852             // If lastWaiter is cancelled, clean out.
1853             if (t != null && t.waitStatus != Node.CONDITION) {
1854                 unlinkCancelledWaiters();
1855                 t = lastWaiter;
1856             }
1857             Node node = new Node(Thread.currentThread(), Node.CONDITION);
1858             if (t == null)
1859                 firstWaiter = node;
1860             else
1861                 t.nextWaiter = node;
1862             lastWaiter = node;
1863             return node;
1864         }
1865 
1866         /**
1867          * Removes and transfers nodes until hit non-cancelled one or
1868          * null. Split out from signal in part to encourage compilers
1869          * to inline the case of no waiters.
1870          * @param first (non-null) the first node on condition queue
1871          */
1872         private void doSignal(Node first) {
1873             do {
1874                 if ( (firstWaiter = first.nextWaiter) == null)
1875                     lastWaiter = null;
1876                 first.nextWaiter = null;
1877             } while (!transferForSignal(first) &&
1878                      (first = firstWaiter) != null);
1879         }
1880 
1881         /**
1882          * Removes and transfers all nodes.
1883          * @param first (non-null) the first node on condition queue
1884          */
1885         private void doSignalAll(Node first) {
1886             lastWaiter = firstWaiter = null;
1887             do {
1888                 Node next = first.nextWaiter;
1889                 first.nextWaiter = null;
1890                 transferForSignal(first);
1891                 first = next;
1892             } while (first != null);
1893         }
1894 
1895         /**
1896          * Unlinks cancelled waiter nodes from condition queue.
1897          * Called only while holding lock. This is called when
1898          * cancellation occurred during condition wait, and upon
1899          * insertion of a new waiter when lastWaiter is seen to have
1900          * been cancelled. This method is needed to avoid garbage
1901          * retention in the absence of signals. So even though it may
1902          * require a full traversal, it comes into play only when
1903          * timeouts or cancellations occur in the absence of
1904          * signals. It traverses all nodes rather than stopping at a
1905          * particular target to unlink all pointers to garbage nodes
1906          * without requiring many re-traversals during cancellation
1907          * storms.
1908          */
1909         private void unlinkCancelledWaiters() {
1910             Node t = firstWaiter;
1911             Node trail = null;
1912             while (t != null) {
1913                 Node next = t.nextWaiter;
1914                 if (t.waitStatus != Node.CONDITION) {
1915                     t.nextWaiter = null;
1916                     if (trail == null)
1917                         firstWaiter = next;
1918                     else
1919                         trail.nextWaiter = next;
1920                     if (next == null)
1921                         lastWaiter = trail;
1922                 }
1923                 else
1924                     trail = t;
1925                 t = next;
1926             }
1927         }
1928 
1929         // public methods
1930 
1931         /**
1932          * Moves the longest-waiting thread, if one exists, from the
1933          * wait queue for this condition to the wait queue for the
1934          * owning lock.
1935          *
1936          * @throws IllegalMonitorStateException if {@link #isHeldExclusively}
1937          *         returns {@code false}
1938          */
1939         public final void signal() {
1940             if (!isHeldExclusively())
1941                 throw new IllegalMonitorStateException();
1942             Node first = firstWaiter;
1943             if (first != null)
1944                 doSignal(first);
1945         }
1946 
1947         /**
1948          * Moves all threads from the wait queue for this condition to
1949          * the wait queue for the owning lock.
1950          *
1951          * @throws IllegalMonitorStateException if {@link #isHeldExclusively}
1952          *         returns {@code false}
1953          */
1954         public final void signalAll() {
1955             if (!isHeldExclusively())
1956                 throw new IllegalMonitorStateException();
1957             Node first = firstWaiter;
1958             if (first != null)
1959                 doSignalAll(first);
1960         }
1961 
1962         /**
1963          * Implements uninterruptible condition wait.
1964          * <ol>
1965          * <li> Save lock state returned by {@link #getState}.
1966          * <li> Invoke {@link #release} with
1967          *      saved state as argument, throwing
1968          *      IllegalMonitorStateException if it fails.
1969          * <li> Block until signalled.
1970          * <li> Reacquire by invoking specialized version of
1971          *      {@link #acquire} with saved state as argument.
1972          * </ol>
1973          */
1974         public final void awaitUninterruptibly() {
1975             Node node = addConditionWaiter();
1976             int savedState = fullyRelease(node);
1977             boolean interrupted = false;
1978             while (!isOnSyncQueue(node)) {
1979                 LockSupport.park(this);
1980                 if (Thread.interrupted())
1981                     interrupted = true;
1982             }
1983             if (acquireQueued(node, savedState) || interrupted)
1984                 selfInterrupt();
1985         }
1986 
1987         /*
1988          * For interruptible waits, we need to track whether to throw
1989          * InterruptedException, if interrupted while blocked on
1990          * condition, versus reinterrupt current thread, if
1991          * interrupted while blocked waiting to re-acquire.
1992          */
1993 
1994         /** Mode meaning to reinterrupt on exit from wait */
1995         private static final int REINTERRUPT =  1;
1996         /** Mode meaning to throw InterruptedException on exit from wait */
1997         private static final int THROW_IE    = -1;
1998 
1999         /**
2000          * Checks for interrupt, returning THROW_IE if interrupted
2001          * before signalled, REINTERRUPT if after signalled, or
2002          * 0 if not interrupted.
2003          */
2004         private int checkInterruptWhileWaiting(Node node) {
2005             return Thread.interrupted() ?
2006                 (transferAfterCancelledWait(node) ? THROW_IE : REINTERRUPT) :
2007                 0;
2008         }
2009 
2010         /**
2011          * Throws InterruptedException, reinterrupts current thread, or
2012          * does nothing, depending on mode.
2013          */
2014         private void reportInterruptAfterWait(int interruptMode)
2015             throws InterruptedException {
2016             if (interruptMode == THROW_IE)
2017                 throw new InterruptedException();
2018             else if (interruptMode == REINTERRUPT)
2019                 selfInterrupt();
2020         }
2021 
2022         /**
2023          * Implements interruptible condition wait.
2024          * <ol>
2025          * <li> If current thread is interrupted, throw InterruptedException.
2026          * <li> Save lock state returned by {@link #getState}.
2027          * <li> Invoke {@link #release} with
2028          *      saved state as argument, throwing
2029          *      IllegalMonitorStateException if it fails.
2030          * <li> Block until signalled or interrupted.
2031          * <li> Reacquire by invoking specialized version of
2032          *      {@link #acquire} with saved state as argument.
2033          * <li> If interrupted while blocked in step 4, throw InterruptedException.
2034          * </ol>
2035          */
2036         public final void await() throws InterruptedException {
2037             if (Thread.interrupted())
2038                 throw new InterruptedException();
2039             Node node = addConditionWaiter();
2040             int savedState = fullyRelease(node);
2041             int interruptMode = 0;
2042             while (!isOnSyncQueue(node)) {
2043                 LockSupport.park(this);
2044                 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0)
2045                     break;
2046             }
2047             if (acquireQueued(node, savedState) && interruptMode != THROW_IE)
2048                 interruptMode = REINTERRUPT;
2049             if (node.nextWaiter != null) // clean up if cancelled
2050                 unlinkCancelledWaiters();
2051             if (interruptMode != 0)
2052                 reportInterruptAfterWait(interruptMode);
2053         }
2054 
2055         /**
2056          * Implements timed condition wait.
2057          * <ol>
2058          * <li> If current thread is interrupted, throw InterruptedException.
2059          * <li> Save lock state returned by {@link #getState}.
2060          * <li> Invoke {@link #release} with
2061          *      saved state as argument, throwing
2062          *      IllegalMonitorStateException if it fails.
2063          * <li> Block until signalled, interrupted, or timed out.
2064          * <li> Reacquire by invoking specialized version of
2065          *      {@link #acquire} with saved state as argument.
2066          * <li> If interrupted while blocked in step 4, throw InterruptedException.
2067          * </ol>
2068          */
2069         public final long awaitNanos(long nanosTimeout)
2070                 throws InterruptedException {
2071             if (Thread.interrupted())
2072                 throw new InterruptedException();
2073             Node node = addConditionWaiter();
2074             int savedState = fullyRelease(node);
2075             long lastTime = System.nanoTime();
2076             int interruptMode = 0;
2077             while (!isOnSyncQueue(node)) {
2078                 if (nanosTimeout <= 0L) {
2079                     transferAfterCancelledWait(node);
2080                     break;
2081                 }
2082                 LockSupport.parkNanos(this, nanosTimeout);
2083                 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0)
2084                     break;
2085 
2086                 long now = System.nanoTime();
2087                 nanosTimeout -= now - lastTime;
2088                 lastTime = now;
2089             }
2090             if (acquireQueued(node, savedState) && interruptMode != THROW_IE)
2091                 interruptMode = REINTERRUPT;
2092             if (node.nextWaiter != null)
2093                 unlinkCancelledWaiters();
2094             if (interruptMode != 0)
2095                 reportInterruptAfterWait(interruptMode);
2096             return nanosTimeout - (System.nanoTime() - lastTime);
2097         }
2098 
2099         /**
2100          * Implements absolute timed condition wait.
2101          * <ol>
2102          * <li> If current thread is interrupted, throw InterruptedException.
2103          * <li> Save lock state returned by {@link #getState}.
2104          * <li> Invoke {@link #release} with
2105          *      saved state as argument, throwing
2106          *      IllegalMonitorStateException if it fails.
2107          * <li> Block until signalled, interrupted, or timed out.
2108          * <li> Reacquire by invoking specialized version of
2109          *      {@link #acquire} with saved state as argument.
2110          * <li> If interrupted while blocked in step 4, throw InterruptedException.
2111          * <li> If timed out while blocked in step 4, return false, else true.
2112          * </ol>
2113          */
2114         public final boolean awaitUntil(Date deadline)
2115                 throws InterruptedException {
2116             if (deadline == null)
2117                 throw new NullPointerException();
2118             long abstime = deadline.getTime();
2119             if (Thread.interrupted())
2120                 throw new InterruptedException();
2121             Node node = addConditionWaiter();
2122             int savedState = fullyRelease(node);
2123             boolean timedout = false;
2124             int interruptMode = 0;
2125             while (!isOnSyncQueue(node)) {
2126                 if (System.currentTimeMillis() > abstime) {
2127                     timedout = transferAfterCancelledWait(node);
2128                     break;
2129                 }
2130                 LockSupport.parkUntil(this, abstime);
2131                 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0)
2132                     break;
2133             }
2134             if (acquireQueued(node, savedState) && interruptMode != THROW_IE)
2135                 interruptMode = REINTERRUPT;
2136             if (node.nextWaiter != null)
2137                 unlinkCancelledWaiters();
2138             if (interruptMode != 0)
2139                 reportInterruptAfterWait(interruptMode);
2140             return !timedout;
2141         }
2142 
2143         /**
2144          * Implements timed condition wait.
2145          * <ol>
2146          * <li> If current thread is interrupted, throw InterruptedException.
2147          * <li> Save lock state returned by {@link #getState}.
2148          * <li> Invoke {@link #release} with
2149          *      saved state as argument, throwing
2150          *      IllegalMonitorStateException if it fails.
2151          * <li> Block until signalled, interrupted, or timed out.
2152          * <li> Reacquire by invoking specialized version of
2153          *      {@link #acquire} with saved state as argument.
2154          * <li> If interrupted while blocked in step 4, throw InterruptedException.
2155          * <li> If timed out while blocked in step 4, return false, else true.
2156          * </ol>
2157          */
2158         public final boolean await(long time, TimeUnit unit)
2159                 throws InterruptedException {
2160             if (unit == null)
2161                 throw new NullPointerException();
2162             long nanosTimeout = unit.toNanos(time);
2163             if (Thread.interrupted())
2164                 throw new InterruptedException();
2165             Node node = addConditionWaiter();
2166             int savedState = fullyRelease(node);
2167             long lastTime = System.nanoTime();
2168             boolean timedout = false;
2169             int interruptMode = 0;
2170             while (!isOnSyncQueue(node)) {
2171                 if (nanosTimeout <= 0L) {
2172                     timedout = transferAfterCancelledWait(node);
2173                     break;
2174                 }
2175                 if (nanosTimeout >= spinForTimeoutThreshold)
2176                     LockSupport.parkNanos(this, nanosTimeout);
2177                 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0)
2178                     break;
2179                 long now = System.nanoTime();
2180                 nanosTimeout -= now - lastTime;
2181                 lastTime = now;
2182             }
2183             if (acquireQueued(node, savedState) && interruptMode != THROW_IE)
2184                 interruptMode = REINTERRUPT;
2185             if (node.nextWaiter != null)
2186                 unlinkCancelledWaiters();
2187             if (interruptMode != 0)
2188                 reportInterruptAfterWait(interruptMode);
2189             return !timedout;
2190         }
2191 
2192         //  support for instrumentation
2193 
2194         /**
2195          * Returns true if this condition was created by the given
2196          * synchronization object.
2197          *
2198          * @return {@code true} if owned
2199          */
2200         final boolean isOwnedBy(AbstractQueuedSynchronizer sync) {
2201             return sync == AbstractQueuedSynchronizer.this;
2202         }
2203 
2204         /**
2205          * Queries whether any threads are waiting on this condition.
2206          * Implements {@link AbstractQueuedSynchronizer#hasWaiters}.
2207          *
2208          * @return {@code true} if there are any waiting threads
2209          * @throws IllegalMonitorStateException if {@link #isHeldExclusively}
2210          *         returns {@code false}
2211          */
2212         protected final boolean hasWaiters() {
2213             if (!isHeldExclusively())
2214                 throw new IllegalMonitorStateException();
2215             for (Node w = firstWaiter; w != null; w = w.nextWaiter) {
2216                 if (w.waitStatus == Node.CONDITION)
2217                     return true;
2218             }
2219             return false;
2220         }
2221 
2222         /**
2223          * Returns an estimate of the number of threads waiting on
2224          * this condition.
2225          * Implements {@link AbstractQueuedSynchronizer#getWaitQueueLength}.
2226          *
2227          * @return the estimated number of waiting threads
2228          * @throws IllegalMonitorStateException if {@link #isHeldExclusively}
2229          *         returns {@code false}
2230          */
2231         protected final int getWaitQueueLength() {
2232             if (!isHeldExclusively())
2233                 throw new IllegalMonitorStateException();
2234             int n = 0;
2235             for (Node w = firstWaiter; w != null; w = w.nextWaiter) {
2236                 if (w.waitStatus == Node.CONDITION)
2237                     ++n;
2238             }
2239             return n;
2240         }
2241 
2242         /**
2243          * Returns a collection containing those threads that may be
2244          * waiting on this Condition.
2245          * Implements {@link AbstractQueuedSynchronizer#getWaitingThreads}.
2246          *
2247          * @return the collection of threads
2248          * @throws IllegalMonitorStateException if {@link #isHeldExclusively}
2249          *         returns {@code false}
2250          */
2251         protected final Collection<Thread> getWaitingThreads() {
2252             if (!isHeldExclusively())
2253                 throw new IllegalMonitorStateException();
2254             ArrayList<Thread> list = new ArrayList<Thread>();
2255             for (Node w = firstWaiter; w != null; w = w.nextWaiter) {
2256                 if (w.waitStatus == Node.CONDITION) {
2257                     Thread t = w.thread;
2258                     if (t != null)
2259                         list.add(t);
2260                 }
2261             }
2262             return list;
2263         }
2264     }
2265 
2266     /**
2267      * Setup to support compareAndSet. We need to natively implement
2268      * this here: For the sake of permitting future enhancements, we
2269      * cannot explicitly subclass AtomicInteger, which would be
2270      * efficient and useful otherwise. So, as the lesser of evils, we
2271      * natively implement using hotspot intrinsics API. And while we
2272      * are at it, we do the same for other CASable fields (which could
2273      * otherwise be done with atomic field updaters).
2274      */
2275     private static final Unsafe unsafe = Unsafe.getUnsafe();
2276     private static final long stateOffset;
2277     private static final long headOffset;
2278     private static final long tailOffset;
2279     private static final long waitStatusOffset;
2280     private static final long nextOffset;
2281 
2282     static {
2283         try {
2284             stateOffset = unsafe.objectFieldOffset
2285                 (AbstractQueuedSynchronizer.class.getDeclaredField("state"));
2286             headOffset = unsafe.objectFieldOffset
2287                 (AbstractQueuedSynchronizer.class.getDeclaredField("head"));
2288             tailOffset = unsafe.objectFieldOffset
2289                 (AbstractQueuedSynchronizer.class.getDeclaredField("tail"));
2290             waitStatusOffset = unsafe.objectFieldOffset
2291                 (Node.class.getDeclaredField("waitStatus"));
2292             nextOffset = unsafe.objectFieldOffset
2293                 (Node.class.getDeclaredField("next"));
2294 
2295         } catch (Exception ex) { throw new Error(ex); }
2296     }
2297 
2298     /**
2299      * CAS head field. Used only by enq.
2300      */
2301     private final boolean compareAndSetHead(Node update) {
2302         return unsafe.compareAndSwapObject(this, headOffset, null, update);
2303     }
2304 
2305     /**
2306      * CAS tail field. Used only by enq.
2307      */
2308     private final boolean compareAndSetTail(Node expect, Node update) {
2309         return unsafe.compareAndSwapObject(this, tailOffset, expect, update);
2310     }
2311 
2312     /**
2313      * CAS waitStatus field of a node.
2314      */
2315     private static final boolean compareAndSetWaitStatus(Node node,
2316                                                          int expect,
2317                                                          int update) {
2318         return unsafe.compareAndSwapInt(node, waitStatusOffset,
2319                                         expect, update);
2320     }
2321 
2322     /**
2323      * CAS next field of a node.
2324      */
2325     private static final boolean compareAndSetNext(Node node,
2326                                                    Node expect,
2327                                                    Node update) {
2328         return unsafe.compareAndSwapObject(node, nextOffset, expect, update);
2329     }
2330 }
View Code

 

其中,共享锁源码相关的代码如下:

public static class ReadLock implements Lock, java.io.Serializable {
    private static final long serialVersionUID = -5992448646407690164L;
    // ReentrantReadWriteLock的AQS对象
    private final Sync sync;

    protected ReadLock(ReentrantReadWriteLock lock) {
        sync = lock.sync;
    }

    // 获取“共享锁”
    public void lock() {
        sync.acquireShared(1);
    }

    // 如果线程是中断状态,则抛出一场,否则尝试获取共享锁。
    public void lockInterruptibly() throws InterruptedException {
        sync.acquireSharedInterruptibly(1);
    }

    // 尝试获取“共享锁”
    public  boolean tryLock() {
        return sync.tryReadLock();
    }

    // 在指定时间内,尝试获取“共享锁”
    public boolean tryLock(long timeout, TimeUnit unit)
            throws InterruptedException {
        return sync.tryAcquireSharedNanos(1, unit.toNanos(timeout));
    }

    // 释放“共享锁”
    public  void unlock() {
        sync.releaseShared(1);
    }

    // 新建条件
    public Condition newCondition() {
        throw new UnsupportedOperationException();
    }

    public String toString() {
        int r = sync.getReadLockCount();
        return super.toString() +
            "[Read locks = " + r + "]";
    }
}

说明
ReadLock中的sync是一个Sync对象,Sync继承于AQS类,即Sync就是一个锁。ReentrantReadWriteLock中也有一个Sync对象,而且ReadLock中的sync和ReentrantReadWriteLock中的sync是对应关系。即ReentrantReadWriteLock和ReadLock共享同一个AQS对象,共享同一把锁。
ReentrantReadWriteLock中Sync的定义如下:

final Sync sync;

下面,分别从“获取共享锁”和“释放共享锁”两个方面对共享锁进行说明。

 

获取共享锁

获取共享锁的思想(即lock函数的步骤),是先通过tryAcquireShared()尝试获取共享锁。尝试成功的话,则直接返回;尝试失败的话,则通过doAcquireShared()不断的循环并尝试获取锁,若有需要,则阻塞等待。doAcquireShared()在循环中每次尝试获取锁时,都是通过tryAcquireShared()来进行尝试的。下面看看“获取共享锁”的详细流程。

1. lock()

lock()在ReadLock中,源码如下:

public void lock() {
    sync.acquireShared(1);
}

 

2. acquireShared()

Sync继承于AQS,acquireShared()定义在AQS中。源码如下:

public final void acquireShared(int arg) {
    if (tryAcquireShared(arg) < 0)
        doAcquireShared(arg);
}

说明acquireShared()首先会通过tryAcquireShared()来尝试获取锁。
尝试成功的话,则不再做任何动作(因为已经成功获取到锁了)。
尝试失败的话,则通过doAcquireShared()来获取锁。doAcquireShared()会获取到锁了才返回。

 

3. tryAcquireShared()

tryAcquireShared()定义在ReentrantReadWriteLock.java的Sync中,源码如下:

protected final int tryAcquireShared(int unused) {
    Thread current = Thread.currentThread();
    // 获取“锁”的状态
    int c = getState();
    // 如果“锁”是“互斥锁”,并且获取锁的线程不是current线程;则返回-1。
    if (exclusiveCount(c) != 0 &&
        getExclusiveOwnerThread() != current)
        return -1;
    // 获取“读取锁”的共享计数
    int r = sharedCount(c);
    // 如果“不需要阻塞等待”,并且“读取锁”的共享计数小于MAX_COUNT;
    // 则通过CAS函数更新“锁的状态”,将“读取锁”的共享计数+1。
    if (!readerShouldBlock() &&
        r < MAX_COUNT &&
        compareAndSetState(c, c + SHARED_UNIT)) {
        // 第1次获取“读取锁”。
        if (r == 0) { 
            firstReader = current;
            firstReaderHoldCount = 1;
        // 如果想要获取锁的线程(current)是第1个获取锁(firstReader)的线程
        } else if (firstReader == current) { 
            firstReaderHoldCount++;
        } else {
            // HoldCounter是用来统计该线程获取“读取锁”的次数。
            HoldCounter rh = cachedHoldCounter;
            if (rh == null || rh.tid != current.getId())
                cachedHoldCounter = rh = readHolds.get();
            else if (rh.count == 0)
                readHolds.set(rh);
            // 将该线程获取“读取锁”的次数+1。
            rh.count++;
        }
        return 1;
    }
    return fullTryAcquireShared(current);
}

说明:tryAcquireShared()的作用是尝试获取“共享锁”。
如果在尝试获取锁时,“不需要阻塞等待”并且“读取锁的共享计数小于MAX_COUNT”,则直接通过CAS函数更新“读取锁的共享计数”,以及将“当前线程获取读取锁的次数+1”。
否则,通过fullTryAcquireShared()获取读取锁。

 

4. fullTryAcquireShared()

fullTryAcquireShared()在ReentrantReadWriteLock中定义,源码如下:

final int fullTryAcquireShared(Thread current) {
    HoldCounter rh = null;
    for (;;) {
        // 获取“锁”的状态
        int c = getState();
        // 如果“锁”是“互斥锁”,并且获取锁的线程不是current线程;则返回-1。
        if (exclusiveCount(c) != 0) {
            if (getExclusiveOwnerThread() != current)
                return -1;
        // 如果“需要阻塞等待”。
        // (01) 当“需要阻塞等待”的线程是第1个获取锁的线程的话,则继续往下执行。
        // (02) 当“需要阻塞等待”的线程获取锁的次数=0时,则返回-1。
        } else if (readerShouldBlock()) {
            // 如果想要获取锁的线程(current)是第1个获取锁(firstReader)的线程
            if (firstReader == current) {
            } else {
                if (rh == null) {
                    rh = cachedHoldCounter;
                    if (rh == null || rh.tid != current.getId()) {
                        rh = readHolds.get();
                        if (rh.count == 0)
                            readHolds.remove();
                    }
                }
                // 如果当前线程获取锁的计数=0,则返回-1。
                if (rh.count == 0)
                    return -1;
            }
        }
        // 如果“不需要阻塞等待”,则获取“读取锁”的共享统计数;
        // 如果共享统计数超过MAX_COUNT,则抛出异常。
        if (sharedCount(c) == MAX_COUNT)
            throw new Error("Maximum lock count exceeded");
        // 将线程获取“读取锁”的次数+1。
        if (compareAndSetState(c, c + SHARED_UNIT)) {
            // 如果是第1次获取“读取锁”,则更新firstReader和firstReaderHoldCount。
            if (sharedCount(c) == 0) {
                firstReader = current;
                firstReaderHoldCount = 1;
            // 如果想要获取锁的线程(current)是第1个获取锁(firstReader)的线程,
            // 则将firstReaderHoldCount+1。
            } else if (firstReader == current) {
                firstReaderHoldCount++;
            } else {
                if (rh == null)
                    rh = cachedHoldCounter;
                if (rh == null || rh.tid != current.getId())
                    rh = readHolds.get();
                else if (rh.count == 0)
                    readHolds.set(rh);
                // 更新线程的获取“读取锁”的共享计数
                rh.count++;
                cachedHoldCounter = rh; // cache for release
            }
            return 1;
        }
    }
}

说明:fullTryAcquireShared()会根据“是否需要阻塞等待”,“读取锁的共享计数是否超过限制”等等进行处理。如果不需要阻塞等待,并且锁的共享计数没有超过限制,则通过CAS尝试获取锁,并返回1。

 

5. doAcquireShared()

doAcquireShared()定义在AQS函数中,源码如下:

private void doAcquireShared(int arg) {
    // addWaiter(Node.SHARED)的作用是,创建“当前线程”对应的节点,并将该线程添加到CLH队列中。
    final Node node = addWaiter(Node.SHARED);
    boolean failed = true;
    try {
        boolean interrupted = false;
        for (;;) {
            // 获取“node”的前一节点
            final Node p = node.predecessor();
            // 如果“当前线程”是CLH队列的表头,则尝试获取共享锁。
            if (p == head) {
                int r = tryAcquireShared(arg);
                if (r >= 0) {
                    setHeadAndPropagate(node, r);
                    p.next = null; // help GC
                    if (interrupted)
                        selfInterrupt();
                    failed = false;
                    return;
                }
            }
            // 如果“当前线程”不是CLH队列的表头,则通过shouldParkAfterFailedAcquire()判断是否需要等待,
            // 需要的话,则通过parkAndCheckInterrupt()进行阻塞等待。若阻塞等待过程中,线程被中断过,则设置interrupted为true。
            if (shouldParkAfterFailedAcquire(p, node) &&
                parkAndCheckInterrupt())
                interrupted = true;
        }
    } finally {
        if (failed)
            cancelAcquire(node);
    }
}

说明:doAcquireShared()的作用是获取共享锁。
它会首先创建线程对应的CLH队列的节点,然后将该节点添加到CLH队列中。CLH队列是管理获取锁的等待线程的队列。
如果“当前线程”是CLH队列的表头,则尝试获取共享锁;否则,则需要通过shouldParkAfterFailedAcquire()判断是否阻塞等待,需要的话,则通过parkAndCheckInterrupt()进行阻塞等待。
doAcquireShared()会通过for循环,不断的进行上面的操作;目的就是获取共享锁。需要注意的是:doAcquireShared()在每一次尝试获取锁时,是通过tryAcquireShared()来执行的!

shouldParkAfterFailedAcquire(), parkAndCheckInterrupt()等函数已经在“Java多线程系列--“JUC锁”03之 公平锁(一) ”中详细介绍过,这里就不再重复说明了。

 

释放共享锁

释放共享锁的思想,是先通过tryReleaseShared()尝试释放共享锁。尝试成功的话,则通过doReleaseShared()唤醒“其他等待获取共享锁的线程”,并返回true;否则的话,返回flase。

1. unlock()

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

说明:该函数实际上调用releaseShared(1)释放共享锁。

 

2. releaseShared()

releaseShared()在AQS中实现,源码如下:

public final boolean releaseShared(int arg) {
    if (tryReleaseShared(arg)) {
        doReleaseShared();
        return true;
    }
    return false;
}

说明:releaseShared()的目的是让当前线程释放它所持有的共享锁。
它首先会通过tryReleaseShared()去尝试释放共享锁。尝试成功,则直接返回;尝试失败,则通过doReleaseShared()去释放共享锁。

 

3. tryReleaseShared()

tryReleaseShared()定义在ReentrantReadWriteLock中,源码如下:

protected final boolean tryReleaseShared(int unused) {
    // 获取当前线程,即释放共享锁的线程。
    Thread current = Thread.currentThread();
    // 如果想要释放锁的线程(current)是第1个获取锁(firstReader)的线程,
    // 并且“第1个获取锁的线程获取锁的次数”=1,则设置firstReader为null;
    // 否则,将“第1个获取锁的线程的获取次数”-1。
    if (firstReader == current) {
        // assert firstReaderHoldCount > 0;
        if (firstReaderHoldCount == 1)
            firstReader = null;
        else
            firstReaderHoldCount--;
    // 获取rh对象,并更新“当前线程获取锁的信息”。
    } else {
 
        HoldCounter rh = cachedHoldCounter;
        if (rh == null || rh.tid != current.getId())
            rh = readHolds.get();
        int count = rh.count;
        if (count <= 1) {
            readHolds.remove();
            if (count <= 0)
                throw unmatchedUnlockException();
        }
        --rh.count;
    }
    for (;;) {
        // 获取锁的状态
        int c = getState();
        // 将锁的获取次数-1。
        int nextc = c - SHARED_UNIT;
        // 通过CAS更新锁的状态。
        if (compareAndSetState(c, nextc))
            return nextc == 0;
    }
}

说明:tryReleaseShared()的作用是尝试释放共享锁。

 

4. doReleaseShared()

doReleaseShared()定义在AQS中,源码如下:

private void doReleaseShared() {
    for (;;) {
        // 获取CLH队列的头节点
        Node h = head;
        // 如果头节点不为null,并且头节点不等于tail节点。
        if (h != null && h != tail) {
            // 获取头节点对应的线程的状态
            int ws = h.waitStatus;
            // 如果头节点对应的线程是SIGNAL状态,则意味着“头节点的下一个节点所对应的线程”需要被unpark唤醒。
            if (ws == Node.SIGNAL) {
                // 设置“头节点对应的线程状态”为空状态。失败的话,则继续循环。
                if (!compareAndSetWaitStatus(h, Node.SIGNAL, 0))
                    continue;
                // 唤醒“头节点的下一个节点所对应的线程”。
                unparkSuccessor(h);
            }
            // 如果头节点对应的线程是空状态,则设置“文件点对应的线程所拥有的共享锁”为其它线程获取锁的空状态。
            else if (ws == 0 &&
                     !compareAndSetWaitStatus(h, 0, Node.PROPAGATE))
                continue;                // loop on failed CAS
        }
        // 如果头节点发生变化,则继续循环。否则,退出循环。
        if (h == head)                   // loop if head changed
            break;
    }
}

说明:doReleaseShared()会释放“共享锁”。它会从前往后的遍历CLH队列,依次“唤醒”然后“执行”队列中每个节点对应的线程;最终的目的是让这些线程释放它们所持有的锁。

 

公平共享锁和非公平共享锁

和互斥锁ReentrantLock一样,ReadLock也分为公平锁和非公平锁。

公平锁和非公平锁的区别,体现在判断是否需要阻塞的函数readerShouldBlock()是不同的。
公平锁的readerShouldBlock()的源码如下:

final boolean readerShouldBlock() {
    return hasQueuedPredecessors();
}

 

在公平共享锁中,如果在当前线程的前面有其他线程在等待获取共享锁,则返回true;否则,返回false。
非公平锁的readerShouldBlock()的源码如下:

final boolean readerShouldBlock() {
    return apparentlyFirstQueuedIsExclusive();
}

在非公平共享锁中,它会无视当前线程的前面是否有其他线程在等待获取共享锁。只要该非公平共享锁对应的线程不为null,则返回true。

 

ReentrantReadWriteLock示例

 1 import java.util.concurrent.locks.ReadWriteLock; 
 2 import java.util.concurrent.locks.ReentrantReadWriteLock; 
 3 
 4 public class ReadWriteLockTest1 { 
 5 
 6     public static void main(String[] args) { 
 7         // 创建账户
 8         MyCount myCount = new MyCount("4238920615242830", 10000); 
 9         // 创建用户,并指定账户
10         User user = new User("Tommy", myCount); 
11 
12         // 分别启动3个“读取账户金钱”的线程 和 3个“设置账户金钱”的线程
13         for (int i=0; i<3; i++) {
14             user.getCash();
15             user.setCash((i+1)*1000);
16         }
17     } 
18 } 
19 
20 class User {
21     private String name;            //用户名 
22     private MyCount myCount;        //所要操作的账户 
23     private ReadWriteLock myLock;   //执行操作所需的锁对象 
24 
25     User(String name, MyCount myCount) {
26         this.name = name; 
27         this.myCount = myCount; 
28         this.myLock = new ReentrantReadWriteLock();
29     }
30 
31     public void getCash() {
32         new Thread() {
33             public void run() {
34                 myLock.readLock().lock(); 
35                 try {
36                     System.out.println(Thread.currentThread().getName() +" getCash start"); 
37                     myCount.getCash();
38                     Thread.sleep(1);
39                     System.out.println(Thread.currentThread().getName() +" getCash end"); 
40                 } catch (InterruptedException e) {
41                 } finally {
42                     myLock.readLock().unlock(); 
43                 }
44             }
45         }.start();
46     }
47 
48     public void setCash(final int cash) {
49         new Thread() {
50             public void run() {
51                 myLock.writeLock().lock(); 
52                 try {
53                     System.out.println(Thread.currentThread().getName() +" setCash start"); 
54                     myCount.setCash(cash);
55                     Thread.sleep(1);
56                     System.out.println(Thread.currentThread().getName() +" setCash end"); 
57                 } catch (InterruptedException e) {
58                 } finally {
59                     myLock.writeLock().unlock(); 
60                 }
61             }
62         }.start();
63     }
64 }
65 
66 class MyCount {
67     private String id;         //账号 
68     private int    cash;       //账户余额 
69 
70     MyCount(String id, int cash) { 
71         this.id = id; 
72         this.cash = cash; 
73     } 
74 
75     public String getId() { 
76         return id; 
77     } 
78 
79     public void setId(String id) { 
80         this.id = id; 
81     } 
82 
83     public int getCash() { 
84         System.out.println(Thread.currentThread().getName() +" getCash cash="+ cash); 
85         return cash; 
86     } 
87 
88     public void setCash(int cash) { 
89         System.out.println(Thread.currentThread().getName() +" setCash cash="+ cash); 
90         this.cash = cash; 
91     } 
92 }

运行结果

Thread-0 getCash start
Thread-2 getCash start
Thread-0 getCash cash=10000
Thread-2 getCash cash=10000
Thread-0 getCash end
Thread-2 getCash end
Thread-1 setCash start
Thread-1 setCash cash=1000
Thread-1 setCash end
Thread-3 setCash start
Thread-3 setCash cash=2000
Thread-3 setCash end
Thread-4 getCash start
Thread-4 getCash cash=2000
Thread-4 getCash end
Thread-5 setCash start
Thread-5 setCash cash=3000
Thread-5 setCash end

结果说明
(01) 观察Thread0和Thread-2的运行结果,我们发现,Thread-0启动并获取到“读取锁”,在它还没运行完毕的时候,Thread-2也启动了并且也成功获取到“读取锁”。
因此,“读取锁”支持被多个线程同时获取。
(02) 观察Thread-1,Thread-3,Thread-5这三个“写入锁”的线程。只要“写入锁”被某线程获取,则该线程运行完毕了,才释放该锁。
因此,“写入锁”不支持被多个线程同时获取。

 


更多内容

1. Java多线程系列--“JUC锁”01之 框架 

2. Java多线程系列--“JUC锁”02之 互斥锁ReentrantLock 

3. Java多线程系列--“JUC锁”03之 公平锁(一) 

4. Java多线程系列--“JUC锁”04之 公平锁(二)

5. Java多线程系列--“JUC锁”05之 非公平锁

6. Java多线程系列--“JUC锁”06之 Condition条件

7. Java多线程系列--“JUC锁”07之 LockSupport 

8. Java多线程系列目录(共xx篇)

 

posted on 2014-01-25 21:31 如果天空不死 阅读(...) 评论(...) 编辑 收藏