1 /*
2 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
3 *
4 * This code is free software; you can redistribute it and/or modify it
5 * under the terms of the GNU General Public License version 2 only, as
6 * published by the Free Software Foundation. Oracle designates this
7 * particular file as subject to the "Classpath" exception as provided
8 * by Oracle in the LICENSE file that accompanied this code.
9 *
10 * This code is distributed in the hope that it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License
13 * version 2 for more details (a copy is included in the LICENSE file that
14 * accompanied this code).
15 *
16 * You should have received a copy of the GNU General Public License version
17 * 2 along with this work; if not, write to the Free Software Foundation,
18 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
19 *
20 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
21 * or visit www.oracle.com if you need additional information or have any
22 * questions.
23 */
24
25 /*
26 * This file is available under and governed by the GNU General Public
27 * License version 2 only, as published by the Free Software Foundation.
28 * However, the following notice accompanied the original version of this
29 * file:
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
38 import java.util.ArrayList;
39 import java.util.Collection;
40 import java.util.Date;
41 import java.util.concurrent.TimeUnit;
42 import jdk.internal.vm.annotation.ReservedStackAccess;
43
44 /**
45 * Provides a framework for implementing blocking locks and related
46 * synchronizers (semaphores, events, etc) that rely on
47 * first-in-first-out (FIFO) wait queues. This class is designed to
48 * be a useful basis for most kinds of synchronizers that rely on a
49 * single atomic {@code int} value to represent state. Subclasses
50 * must define the protected methods that change this state, and which
51 * define what that state means in terms of this object being acquired
52 * or released. Given these, the other methods in this class carry
53 * out all queuing and blocking mechanics. Subclasses can maintain
54 * other state fields, but only the atomically updated {@code int}
55 * value manipulated using methods {@link #getState}, {@link
56 * #setState} and {@link #compareAndSetState} is tracked with respect
57 * to synchronization.
58 *
59 * <p>Subclasses should be defined as non-public internal helper
60 * classes that are used to implement the synchronization properties
61 * of their enclosing class. Class
62 * {@code AbstractQueuedSynchronizer} does not implement any
63 * synchronization interface. Instead it defines methods such as
64 * {@link #acquireInterruptibly} that can be invoked as
65 * appropriate by concrete locks and related synchronizers to
66 * implement their public methods.
67 *
68 * <p>This class supports either or both a default <em>exclusive</em>
69 * mode and a <em>shared</em> mode. When acquired in exclusive mode,
70 * attempted acquires by other threads cannot succeed. Shared mode
71 * acquires by multiple threads may (but need not) succeed. This class
72 * does not "understand" these differences except in the
73 * mechanical sense that when a shared mode acquire succeeds, the next
74 * waiting thread (if one exists) must also determine whether it can
75 * acquire as well. Threads waiting in the different modes share the
76 * same FIFO queue. Usually, implementation subclasses support only
77 * one of these modes, but both can come into play for example in a
78 * {@link ReadWriteLock}. Subclasses that support only exclusive or
79 * only shared modes need not define the methods supporting the unused mode.
80 *
81 * <p>This class defines a nested {@link ConditionObject} class that
82 * can be used as a {@link Condition} implementation by subclasses
83 * supporting exclusive mode for which method {@link
84 * #isHeldExclusively} reports whether synchronization is exclusively
85 * held with respect to the current thread, method {@link #release}
86 * invoked with the current {@link #getState} value fully releases
87 * this object, and {@link #acquire}, given this saved state value,
88 * eventually restores this object to its previous acquired state. No
89 * {@code AbstractQueuedSynchronizer} method otherwise creates such a
90 * condition, so if this constraint cannot be met, do not use it. The
91 * behavior of {@link ConditionObject} depends of course on the
92 * semantics of its synchronizer implementation.
93 *
94 * <p>This class provides inspection, instrumentation, and monitoring
95 * methods for the internal queue, as well as similar methods for
96 * condition objects. These can be exported as desired into classes
97 * using an {@code AbstractQueuedSynchronizer} for their
98 * synchronization mechanics.
99 *
100 * <p>Serialization of this class stores only the underlying atomic
101 * integer maintaining state, so deserialized objects have empty
102 * thread queues. Typical subclasses requiring serializability will
103 * define a {@code readObject} method that restores this to a known
104 * initial state upon deserialization.
105 *
106 * <h3>Usage</h3>
107 *
108 * <p>To use this class as the basis of a synchronizer, redefine the
109 * following methods, as applicable, by inspecting and/or modifying
110 * the synchronization state using {@link #getState}, {@link
111 * #setState} and/or {@link #compareAndSetState}:
112 *
113 * <ul>
114 * <li>{@link #tryAcquire}
115 * <li>{@link #tryRelease}
116 * <li>{@link #tryAcquireShared}
117 * <li>{@link #tryReleaseShared}
118 * <li>{@link #isHeldExclusively}
119 * </ul>
120 *
121 * Each of these methods by default throws {@link
122 * UnsupportedOperationException}. Implementations of these methods
123 * must be internally thread-safe, and should in general be short and
124 * not block. Defining these methods is the <em>only</em> supported
125 * means of using this class. All other methods are declared
126 * {@code final} because they cannot be independently varied.
127 *
128 * <p>You may also find the inherited methods from {@link
129 * AbstractOwnableSynchronizer} useful to keep track of the thread
130 * owning an exclusive synchronizer. You are encouraged to use them
131 * -- this enables monitoring and diagnostic tools to assist users in
132 * determining which threads hold locks.
133 *
134 * <p>Even though this class is based on an internal FIFO queue, it
135 * does not automatically enforce FIFO acquisition policies. The core
136 * of exclusive synchronization takes the form:
137 *
138 * <pre>
139 * Acquire:
140 * while (!tryAcquire(arg)) {
141 * <em>enqueue thread if it is not already queued</em>;
142 * <em>possibly block current thread</em>;
143 * }
144 *
145 * Release:
146 * if (tryRelease(arg))
147 * <em>unblock the first queued thread</em>;
148 * </pre>
149 *
150 * (Shared mode is similar but may involve cascading signals.)
151 *
152 * <p id="barging">Because checks in acquire are invoked before
153 * enqueuing, a newly acquiring thread may <em>barge</em> ahead of
154 * others that are blocked and queued. However, you can, if desired,
155 * define {@code tryAcquire} and/or {@code tryAcquireShared} to
156 * disable barging by internally invoking one or more of the inspection
157 * methods, thereby providing a <em>fair</em> FIFO acquisition order.
158 * In particular, most fair synchronizers can define {@code tryAcquire}
159 * to return {@code false} if {@link #hasQueuedPredecessors} (a method
160 * specifically designed to be used by fair synchronizers) returns
161 * {@code true}. Other variations are possible.
162 *
163 * <p>Throughput and scalability are generally highest for the
164 * default barging (also known as <em>greedy</em>,
165 * <em>renouncement</em>, and <em>convoy-avoidance</em>) strategy.
166 * While this is not guaranteed to be fair or starvation-free, earlier
167 * queued threads are allowed to recontend before later queued
168 * threads, and each recontention has an unbiased chance to succeed
169 * against incoming threads. Also, while acquires do not
170 * "spin" in the usual sense, they may perform multiple
171 * invocations of {@code tryAcquire} interspersed with other
172 * computations before blocking. This gives most of the benefits of
173 * spins when exclusive synchronization is only briefly held, without
174 * most of the liabilities when it isn't. If so desired, you can
175 * augment this by preceding calls to acquire methods with
176 * "fast-path" checks, possibly prechecking {@link #hasContended}
177 * and/or {@link #hasQueuedThreads} to only do so if the synchronizer
178 * is likely not to be contended.
179 *
180 * <p>This class provides an efficient and scalable basis for
181 * synchronization in part by specializing its range of use to
182 * synchronizers that can rely on {@code int} state, acquire, and
183 * release parameters, and an internal FIFO wait queue. When this does
184 * not suffice, you can build synchronizers from a lower level using
185 * {@link java.util.concurrent.atomic atomic} classes, your own custom
186 * {@link java.util.Queue} classes, and {@link LockSupport} blocking
187 * support.
188 *
189 * <h3>Usage Examples</h3>
190 *
191 * <p>Here is a non-reentrant mutual exclusion lock class that uses
192 * the value zero to represent the unlocked state, and one to
193 * represent the locked state. While a non-reentrant lock
194 * does not strictly require recording of the current owner
195 * thread, this class does so anyway to make usage easier to monitor.
196 * It also supports conditions and exposes
197 * one of the instrumentation methods:
198 *
199 * <pre> {@code
200 * class Mutex implements Lock, java.io.Serializable {
201 *
202 * // Our internal helper class
203 * private static class Sync extends AbstractQueuedSynchronizer {
204 * // Reports whether in locked state
205 * protected boolean isHeldExclusively() {
206 * return getState() == 1;
207 * }
208 *
209 * // Acquires the lock if state is zero
210 * public boolean tryAcquire(int acquires) {
211 * assert acquires == 1; // Otherwise unused
212 * if (compareAndSetState(0, 1)) {
213 * setExclusiveOwnerThread(Thread.currentThread());
214 * return true;
215 * }
216 * return false;
217 * }
218 *
219 * // Releases the lock by setting state to zero
220 * protected boolean tryRelease(int releases) {
221 * assert releases == 1; // Otherwise unused
222 * if (getState() == 0) throw new IllegalMonitorStateException();
223 * setExclusiveOwnerThread(null);
224 * setState(0);
225 * return true;
226 * }
227 *
228 * // Provides a Condition
229 * Condition newCondition() { return new ConditionObject(); }
230 *
231 * // Deserializes properly
232 * private void readObject(ObjectInputStream s)
233 * throws IOException, ClassNotFoundException {
234 * s.defaultReadObject();
235 * setState(0); // reset to unlocked state
236 * }
237 * }
238 *
239 * // The sync object does all the hard work. We just forward to it.
240 * private final Sync sync = new Sync();
241 *
242 * public void lock() { sync.acquire(1); }
243 * public boolean tryLock() { return sync.tryAcquire(1); }
244 * public void unlock() { sync.release(1); }
245 * public Condition newCondition() { return sync.newCondition(); }
246 * public boolean isLocked() { return sync.isHeldExclusively(); }
247 * public boolean hasQueuedThreads() { return sync.hasQueuedThreads(); }
248 * public void lockInterruptibly() throws InterruptedException {
249 * sync.acquireInterruptibly(1);
250 * }
251 * public boolean tryLock(long timeout, TimeUnit unit)
252 * throws InterruptedException {
253 * return sync.tryAcquireNanos(1, unit.toNanos(timeout));
254 * }
255 * }}</pre>
256 *
257 * <p>Here is a latch class that is like a
258 * {@link java.util.concurrent.CountDownLatch CountDownLatch}
259 * except that it only requires a single {@code signal} to
260 * fire. Because a latch is non-exclusive, it uses the {@code shared}
261 * acquire and release methods.
262 *
263 * <pre> {@code
264 * class BooleanLatch {
265 *
266 * private static class Sync extends AbstractQueuedSynchronizer {
267 * boolean isSignalled() { return getState() != 0; }
268 *
269 * protected int tryAcquireShared(int ignore) {
270 * return isSignalled() ? 1 : -1;
271 * }
272 *
273 * protected boolean tryReleaseShared(int ignore) {
274 * setState(1);
275 * return true;
276 * }
277 * }
278 *
279 * private final Sync sync = new Sync();
280 * public boolean isSignalled() { return sync.isSignalled(); }
281 * public void signal() { sync.releaseShared(1); }
282 * public void await() throws InterruptedException {
283 * sync.acquireSharedInterruptibly(1);
284 * }
285 * }}</pre>
286 *
287 * @since 1.5
288 * @author Doug Lea
289 */
290 public abstract class AbstractQueuedSynchronizer
291 extends AbstractOwnableSynchronizer
292 implements java.io.Serializable {
293
294 private static final long serialVersionUID = 7373984972572414691L;
295
296 /**
297 * Creates a new {@code AbstractQueuedSynchronizer} instance
298 * with initial synchronization state of zero.
299 */
300 protected AbstractQueuedSynchronizer() { }
301
302 /**
303 * Wait queue node class.
304 *
305 * <p>The wait queue is a variant of a "CLH" (Craig, Landin, and
306 * Hagersten) lock queue. CLH locks are normally used for
307 * spinlocks. We instead use them for blocking synchronizers, but
308 * use the same basic tactic of holding some of the control
309 * information about a thread in the predecessor of its node. A
310 * "status" field in each node keeps track of whether a thread
311 * should block. A node is signalled when its predecessor
312 * releases. Each node of the queue otherwise serves as a
313 * specific-notification-style monitor holding a single waiting
314 * thread. The status field does NOT control whether threads are
315 * granted locks etc though. A thread may try to acquire if it is
316 * first in the queue. But being first does not guarantee success;
317 * it only gives the right to contend. So the currently released
318 * contender thread may need to rewait.
319 *
320 * <p>To enqueue into a CLH lock, you atomically splice it in as new
321 * tail. To dequeue, you just set the head field.
322 * <pre>
323 * +------+ prev +-----+ +-----+
324 * head | | <---- | | <---- | | tail
325 * +------+ +-----+ +-----+
326 * </pre>
327 *
328 * <p>Insertion into a CLH queue requires only a single atomic
329 * operation on "tail", so there is a simple atomic point of
330 * demarcation from unqueued to queued. Similarly, dequeuing
331 * involves only updating the "head". However, it takes a bit
332 * more work for nodes to determine who their successors are,
333 * in part to deal with possible cancellation due to timeouts
334 * and interrupts.
335 *
336 * <p>The "prev" links (not used in original CLH locks), are mainly
337 * needed to handle cancellation. If a node is cancelled, its
338 * successor is (normally) relinked to a non-cancelled
339 * predecessor. For explanation of similar mechanics in the case
340 * of spin locks, see the papers by Scott and Scherer at
341 * http://www.cs.rochester.edu/u/scott/synchronization/
342 *
343 * <p>We also use "next" links to implement blocking mechanics.
344 * The thread id for each node is kept in its own node, so a
345 * predecessor signals the next node to wake up by traversing
346 * next link to determine which thread it is. Determination of
347 * successor must avoid races with newly queued nodes to set
348 * the "next" fields of their predecessors. This is solved
349 * when necessary by checking backwards from the atomically
350 * updated "tail" when a node's successor appears to be null.
351 * (Or, said differently, the next-links are an optimization
352 * so that we don't usually need a backward scan.)
353 *
354 * <p>Cancellation introduces some conservatism to the basic
355 * algorithms. Since we must poll for cancellation of other
356 * nodes, we can miss noticing whether a cancelled node is
357 * ahead or behind us. This is dealt with by always unparking
358 * successors upon cancellation, allowing them to stabilize on
359 * a new predecessor, unless we can identify an uncancelled
360 * predecessor who will carry this responsibility.
361 *
362 * <p>CLH queues need a dummy header node to get started. But
363 * we don't create them on construction, because it would be wasted
364 * effort if there is never contention. Instead, the node
365 * is constructed and head and tail pointers are set upon first
366 * contention.
367 *
368 * <p>Threads waiting on Conditions use the same nodes, but
369 * use an additional link. Conditions only need to link nodes
370 * in simple (non-concurrent) linked queues because they are
371 * only accessed when exclusively held. Upon await, a node is
372 * inserted into a condition queue. Upon signal, the node is
373 * transferred to the main queue. A special value of status
374 * field is used to mark which queue a node is on.
375 *
376 * <p>Thanks go to Dave Dice, Mark Moir, Victor Luchangco, Bill
377 * Scherer and Michael Scott, along with members of JSR-166
378 * expert group, for helpful ideas, discussions, and critiques
379 * on the design of this class.
380 */
381 static final class Node {
382 /** Marker to indicate a node is waiting in shared mode */
383 static final Node SHARED = new Node();
384 /** Marker to indicate a node is waiting in exclusive mode */
385 static final Node EXCLUSIVE = null;
386
387 /** waitStatus value to indicate thread has cancelled. */
388 static final int CANCELLED = 1;
389 /** waitStatus value to indicate successor's thread needs unparking. */
390 static final int SIGNAL = -1;
391 /** waitStatus value to indicate thread is waiting on condition. */
392 static final int CONDITION = -2;
393 /**
394 * waitStatus value to indicate the next acquireShared should
395 * unconditionally propagate.
396 */
397 static final int PROPAGATE = -3;
398
399 /**
400 * Status field, taking on only the values:
401 * SIGNAL: The successor of this node is (or will soon be)
402 * blocked (via park), so the current node must
403 * unpark its successor when it releases or
404 * cancels. To avoid races, acquire methods must
405 * first indicate they need a signal,
406 * then retry the atomic acquire, and then,
407 * on failure, block.
408 * CANCELLED: This node is cancelled due to timeout or interrupt.
409 * Nodes never leave this state. In particular,
410 * a thread with cancelled node never again blocks.
411 * CONDITION: This node is currently on a condition queue.
412 * It will not be used as a sync queue node
413 * until transferred, at which time the status
414 * will be set to 0. (Use of this value here has
415 * nothing to do with the other uses of the
416 * field, but simplifies mechanics.)
417 * PROPAGATE: A releaseShared should be propagated to other
418 * nodes. This is set (for head node only) in
419 * doReleaseShared to ensure propagation
420 * continues, even if other operations have
421 * since intervened.
422 * 0: None of the above
423 *
424 * The values are arranged numerically to simplify use.
425 * Non-negative values mean that a node doesn't need to
426 * signal. So, most code doesn't need to check for particular
427 * values, just for sign.
428 *
429 * The field is initialized to 0 for normal sync nodes, and
430 * CONDITION for condition nodes. It is modified using CAS
431 * (or when possible, unconditional volatile writes).
432 */
433 volatile int waitStatus;
434
435 /**
436 * Link to predecessor node that current node/thread relies on
437 * for checking waitStatus. Assigned during enqueuing, and nulled
438 * out (for sake of GC) only upon dequeuing. Also, upon
439 * cancellation of a predecessor, we short-circuit while
440 * finding a non-cancelled one, which will always exist
441 * because the head node is never cancelled: A node becomes
442 * head only as a result of successful acquire. A
443 * cancelled thread never succeeds in acquiring, and a thread only
444 * cancels itself, not any other node.
445 */
446 volatile Node prev;
447
448 /**
449 * Link to the successor node that the current node/thread
450 * unparks upon release. Assigned during enqueuing, adjusted
451 * when bypassing cancelled predecessors, and nulled out (for
452 * sake of GC) when dequeued. The enq operation does not
453 * assign next field of a predecessor until after attachment,
454 * so seeing a null next field does not necessarily mean that
455 * node is at end of queue. However, if a next field appears
456 * to be null, we can scan prev's from the tail to
457 * double-check. The next field of cancelled nodes is set to
458 * point to the node itself instead of null, to make life
459 * easier for isOnSyncQueue.
460 */
461 volatile Node next;
462
463 /**
464 * The thread that enqueued this node. Initialized on
465 * construction and nulled out after use.
466 */
467 volatile Thread thread;
468
469 /**
470 * Link to next node waiting on condition, or the special
471 * value SHARED. Because condition queues are accessed only
472 * when holding in exclusive mode, we just need a simple
473 * linked queue to hold nodes while they are waiting on
474 * conditions. They are then transferred to the queue to
475 * re-acquire. And because conditions can only be exclusive,
476 * we save a field by using special value to indicate shared
477 * mode.
478 */
479 Node nextWaiter;
480
481 /**
482 * Returns true if node is waiting in shared mode.
483 */
484 final boolean isShared() {
485 return nextWaiter == SHARED;
486 }
487
488 /**
489 * Returns previous node, or throws NullPointerException if null.
490 * Use when predecessor cannot be null. The null check could
491 * be elided, but is present to help the VM.
492 *
493 * @return the predecessor of this node
494 */
495 final Node predecessor() throws NullPointerException {
496 Node p = prev;
497 if (p == null)
498 throw new NullPointerException();
499 else
500 return p;
501 }
502
503 /** Establishes initial head or SHARED marker. */
504 Node() {}
505
506 /** Constructor used by addWaiter. */
507 Node(Node nextWaiter) {
508 this.nextWaiter = nextWaiter;
509 U.putObject(this, THREAD, Thread.currentThread());
510 }
511
512 /** Constructor used by addConditionWaiter. */
513 Node(int waitStatus) {
514 U.putInt(this, WAITSTATUS, waitStatus);
515 U.putObject(this, THREAD, Thread.currentThread());
516 }
517
518 /** CASes waitStatus field. */
519 final boolean compareAndSetWaitStatus(int expect, int update) {
520 return U.compareAndSwapInt(this, WAITSTATUS, expect, update);
521 }
522
523 /** CASes next field. */
524 final boolean compareAndSetNext(Node expect, Node update) {
525 return U.compareAndSwapObject(this, NEXT, expect, update);
526 }
527
528 private static final sun.misc.Unsafe U = sun.misc.Unsafe.getUnsafe();
529 private static final long NEXT;
530 static final long PREV;
531 private static final long THREAD;
532 private static final long WAITSTATUS;
533 static {
534 try {
535 NEXT = U.objectFieldOffset
536 (Node.class.getDeclaredField("next"));
537 PREV = U.objectFieldOffset
538 (Node.class.getDeclaredField("prev"));
539 THREAD = U.objectFieldOffset
540 (Node.class.getDeclaredField("thread"));
541 WAITSTATUS = U.objectFieldOffset
542 (Node.class.getDeclaredField("waitStatus"));
543 } catch (ReflectiveOperationException e) {
544 throw new Error(e);
545 }
546 }
547 }
548
549 /**
550 * Head of the wait queue, lazily initialized. Except for
551 * initialization, it is modified only via method setHead. Note:
552 * If head exists, its waitStatus is guaranteed not to be
553 * CANCELLED.
554 */
555 private transient volatile Node head;
556
557 /**
558 * Tail of the wait queue, lazily initialized. Modified only via
559 * method enq to add new wait node.
560 */
561 private transient volatile Node tail;
562
563 /**
564 * The synchronization state.
565 */
566 private volatile int state;
567
568 /**
569 * Returns the current value of synchronization state.
570 * This operation has memory semantics of a {@code volatile} read.
571 * @return current state value
572 */
573 protected final int getState() {
574 return state;
575 }
576
577 /**
578 * Sets the value of synchronization state.
579 * This operation has memory semantics of a {@code volatile} write.
580 * @param newState the new state value
581 */
582 protected final void setState(int newState) {
583 state = newState;
584 }
585
586 /**
587 * Atomically sets synchronization state to the given updated
588 * value if the current state value equals the expected value.
589 * This operation has memory semantics of a {@code volatile} read
590 * and write.
591 *
592 * @param expect the expected value
593 * @param update the new value
594 * @return {@code true} if successful. False return indicates that the actual
595 * value was not equal to the expected value.
596 */
597 protected final boolean compareAndSetState(int expect, int update) {
598 return U.compareAndSwapInt(this, STATE, expect, update);
599 }
600
601 // Queuing utilities
602
603 /**
604 * The number of nanoseconds for which it is faster to spin
605 * rather than to use timed park. A rough estimate suffices
606 * to improve responsiveness with very short timeouts.
607 */
608 static final long SPIN_FOR_TIMEOUT_THRESHOLD = 1000L;
609
610 /**
611 * Inserts node into queue, initializing if necessary. See picture above.
612 * @param node the node to insert
613 * @return node's predecessor
614 */
615 private Node enq(Node node) {
616 for (;;) {
617 Node oldTail = tail;
618 if (oldTail != null) {
619 U.putObject(node, Node.PREV, oldTail);
620 if (compareAndSetTail(oldTail, node)) {
621 oldTail.next = node;
622 return oldTail;
623 }
624 } else {
625 initializeSyncQueue();
626 }
627 }
628 }
629
630 /**
631 * Creates and enqueues node for current thread and given mode.
632 *
633 * @param mode Node.EXCLUSIVE for exclusive, Node.SHARED for shared
634 * @return the new node
635 */
636 private Node addWaiter(Node mode) {
637 Node node = new Node(mode);
638
639 for (;;) {
640 Node oldTail = tail;
641 if (oldTail != null) {
642 U.putObject(node, Node.PREV, oldTail);
643 if (compareAndSetTail(oldTail, node)) {
644 oldTail.next = node;
645 return node;
646 }
647 } else {
648 initializeSyncQueue();
649 }
650 }
651 }
652
653 /**
654 * Sets head of queue to be node, thus dequeuing. Called only by
655 * acquire methods. Also nulls out unused fields for sake of GC
656 * and to suppress unnecessary signals and traversals.
657 *
658 * @param node the node
659 */
660 private void setHead(Node node) {
661 head = node;
662 node.thread = null;
663 node.prev = null;
664 }
665
666 /**
667 * Wakes up node's successor, if one exists.
668 *
669 * @param node the node
670 */
671 private void unparkSuccessor(Node node) {
672 /*
673 * If status is negative (i.e., possibly needing signal) try
674 * to clear in anticipation of signalling. It is OK if this
675 * fails or if status is changed by waiting thread.
676 */
677 int ws = node.waitStatus;
678 if (ws < 0)
679 node.compareAndSetWaitStatus(ws, 0);
680
681 /*
682 * Thread to unpark is held in successor, which is normally
683 * just the next node. But if cancelled or apparently null,
684 * traverse backwards from tail to find the actual
685 * non-cancelled successor.
686 */
687 Node s = node.next;
688 if (s == null || s.waitStatus > 0) {
689 s = null;
690 for (Node p = tail; p != node && p != null; p = p.prev)
691 if (p.waitStatus <= 0)
692 s = p;
693 }
694 if (s != null)
695 LockSupport.unpark(s.thread);
696 }
697
698 /**
699 * Release action for shared mode -- signals successor and ensures
700 * propagation. (Note: For exclusive mode, release just amounts
701 * to calling unparkSuccessor of head if it needs signal.)
702 */
703 private void doReleaseShared() {
704 /*
705 * Ensure that a release propagates, even if there are other
706 * in-progress acquires/releases. This proceeds in the usual
707 * way of trying to unparkSuccessor of head if it needs
708 * signal. But if it does not, status is set to PROPAGATE to
709 * ensure that upon release, propagation continues.
710 * Additionally, we must loop in case a new node is added
711 * while we are doing this. Also, unlike other uses of
712 * unparkSuccessor, we need to know if CAS to reset status
713 * fails, if so rechecking.
714 */
715 for (;;) {
716 Node h = head;
717 if (h != null && h != tail) {
718 int ws = h.waitStatus;
719 if (ws == Node.SIGNAL) {
720 if (!h.compareAndSetWaitStatus(Node.SIGNAL, 0))
721 continue; // loop to recheck cases
722 unparkSuccessor(h);
723 }
724 else if (ws == 0 &&
725 !h.compareAndSetWaitStatus(0, Node.PROPAGATE))
726 continue; // loop on failed CAS
727 }
728 if (h == head) // loop if head changed
729 break;
730 }
731 }
732
733 /**
734 * Sets head of queue, and checks if successor may be waiting
735 * in shared mode, if so propagating if either propagate > 0 or
736 * PROPAGATE status was set.
737 *
738 * @param node the node
739 * @param propagate the return value from a tryAcquireShared
740 */
741 private void setHeadAndPropagate(Node node, int propagate) {
742 Node h = head; // Record old head for check below
743 setHead(node);
744 /*
745 * Try to signal next queued node if:
746 * Propagation was indicated by caller,
747 * or was recorded (as h.waitStatus either before
748 * or after setHead) by a previous operation
749 * (note: this uses sign-check of waitStatus because
750 * PROPAGATE status may transition to SIGNAL.)
751 * and
752 * The next node is waiting in shared mode,
753 * or we don't know, because it appears null
754 *
755 * The conservatism in both of these checks may cause
756 * unnecessary wake-ups, but only when there are multiple
757 * racing acquires/releases, so most need signals now or soon
758 * anyway.
759 */
760 if (propagate > 0 || h == null || h.waitStatus < 0 ||
761 (h = head) == null || h.waitStatus < 0) {
762 Node s = node.next;
763 if (s == null || s.isShared())
764 doReleaseShared();
765 }
766 }
767
768 // Utilities for various versions of acquire
769
770 /**
771 * Cancels an ongoing attempt to acquire.
772 *
773 * @param node the node
774 */
775 private void cancelAcquire(Node node) {
776 // Ignore if node doesn't exist
777 if (node == null)
778 return;
779
780 node.thread = null;
781
782 // Skip cancelled predecessors
783 Node pred = node.prev;
784 while (pred.waitStatus > 0)
785 node.prev = pred = pred.prev;
786
787 // predNext is the apparent node to unsplice. CASes below will
788 // fail if not, in which case, we lost race vs another cancel
789 // or signal, so no further action is necessary.
790 Node predNext = pred.next;
791
792 // Can use unconditional write instead of CAS here.
793 // After this atomic step, other Nodes can skip past us.
794 // Before, we are free of interference from other threads.
795 node.waitStatus = Node.CANCELLED;
796
797 // If we are the tail, remove ourselves.
798 if (node == tail && compareAndSetTail(node, pred)) {
799 pred.compareAndSetNext(predNext, null);
800 } else {
801 // If successor needs signal, try to set pred's next-link
802 // so it will get one. Otherwise wake it up to propagate.
803 int ws;
804 if (pred != head &&
805 ((ws = pred.waitStatus) == Node.SIGNAL ||
806 (ws <= 0 && pred.compareAndSetWaitStatus(ws, Node.SIGNAL))) &&
807 pred.thread != null) {
808 Node next = node.next;
809 if (next != null && next.waitStatus <= 0)
810 pred.compareAndSetNext(predNext, next);
811 } else {
812 unparkSuccessor(node);
813 }
814
815 node.next = node; // help GC
816 }
817 }
818
819 /**
820 * Checks and updates status for a node that failed to acquire.
821 * Returns true if thread should block. This is the main signal
822 * control in all acquire loops. Requires that pred == node.prev.
823 *
824 * @param pred node's predecessor holding status
825 * @param node the node
826 * @return {@code true} if thread should block
827 */
828 private static boolean shouldParkAfterFailedAcquire(Node pred, Node node) {
829 int ws = pred.waitStatus;
830 if (ws == Node.SIGNAL)
831 /*
832 * This node has already set status asking a release
833 * to signal it, so it can safely park.
834 */
835 return true;
836 if (ws > 0) {
837 /*
838 * Predecessor was cancelled. Skip over predecessors and
839 * indicate retry.
840 */
841 do {
842 node.prev = pred = pred.prev;
843 } while (pred.waitStatus > 0);
844 pred.next = node;
845 } else {
846 /*
847 * waitStatus must be 0 or PROPAGATE. Indicate that we
848 * need a signal, but don't park yet. Caller will need to
849 * retry to make sure it cannot acquire before parking.
850 */
851 pred.compareAndSetWaitStatus(ws, Node.SIGNAL);
852 }
853 return false;
854 }
855
856 /**
857 * Convenience method to interrupt current thread.
858 */
859 static void selfInterrupt() {
860 Thread.currentThread().interrupt();
861 }
862
863 /**
864 * Convenience method to park and then check if interrupted.
865 *
866 * @return {@code true} if interrupted
867 */
868 private final boolean parkAndCheckInterrupt() {
869 LockSupport.park(this);
870 return Thread.interrupted();
871 }
872
873 /*
874 * Various flavors of acquire, varying in exclusive/shared and
875 * control modes. Each is mostly the same, but annoyingly
876 * different. Only a little bit of factoring is possible due to
877 * interactions of exception mechanics (including ensuring that we
878 * cancel if tryAcquire throws exception) and other control, at
879 * least not without hurting performance too much.
880 */
881
882 /**
883 * Acquires in exclusive uninterruptible mode for thread already in
884 * queue. Used by condition wait methods as well as acquire.
885 *
886 * @param node the node
887 * @param arg the acquire argument
888 * @return {@code true} if interrupted while waiting
889 */
890 @ReservedStackAccess
891 final boolean acquireQueued(final Node node, int arg) {
892 try {
893 boolean interrupted = false;
894 for (;;) {
895 final Node p = node.predecessor();
896 if (p == head && tryAcquire(arg)) {
897 setHead(node);
898 p.next = null; // help GC
899 return interrupted;
900 }
901 if (shouldParkAfterFailedAcquire(p, node) &&
902 parkAndCheckInterrupt())
903 interrupted = true;
904 }
905 } catch (Throwable t) {
906 cancelAcquire(node);
907 throw t;
908 }
909 }
910
911 /**
912 * Acquires in exclusive interruptible mode.
913 * @param arg the acquire argument
914 */
915 private void doAcquireInterruptibly(int arg)
916 throws InterruptedException {
917 final Node node = addWaiter(Node.EXCLUSIVE);
918 try {
919 for (;;) {
920 final Node p = node.predecessor();
921 if (p == head && tryAcquire(arg)) {
922 setHead(node);
923 p.next = null; // help GC
924 return;
925 }
926 if (shouldParkAfterFailedAcquire(p, node) &&
927 parkAndCheckInterrupt())
928 throw new InterruptedException();
929 }
930 } catch (Throwable t) {
931 cancelAcquire(node);
932 throw t;
933 }
934 }
935
936 /**
937 * Acquires in exclusive timed mode.
938 *
939 * @param arg the acquire argument
940 * @param nanosTimeout max wait time
941 * @return {@code true} if acquired
942 */
943 private boolean doAcquireNanos(int arg, long nanosTimeout)
944 throws InterruptedException {
945 if (nanosTimeout <= 0L)
946 return false;
947 final long deadline = System.nanoTime() + nanosTimeout;
948 final Node node = addWaiter(Node.EXCLUSIVE);
949 try {
950 for (;;) {
951 final Node p = node.predecessor();
952 if (p == head && tryAcquire(arg)) {
953 setHead(node);
954 p.next = null; // help GC
955 return true;
956 }
957 nanosTimeout = deadline - System.nanoTime();
958 if (nanosTimeout <= 0L) {
959 cancelAcquire(node);
960 return false;
961 }
962 if (shouldParkAfterFailedAcquire(p, node) &&
963 nanosTimeout > SPIN_FOR_TIMEOUT_THRESHOLD)
964 LockSupport.parkNanos(this, nanosTimeout);
965 if (Thread.interrupted())
966 throw new InterruptedException();
967 }
968 } catch (Throwable t) {
969 cancelAcquire(node);
970 throw t;
971 }
972 }
973
974 /**
975 * Acquires in shared uninterruptible mode.
976 * @param arg the acquire argument
977 */
978 private void doAcquireShared(int arg) {
979 final Node node = addWaiter(Node.SHARED);
980 try {
981 boolean interrupted = false;
982 for (;;) {
983 final Node p = node.predecessor();
984 if (p == head) {
985 int r = tryAcquireShared(arg);
986 if (r >= 0) {
987 setHeadAndPropagate(node, r);
988 p.next = null; // help GC
989 if (interrupted)
990 selfInterrupt();
991 return;
992 }
993 }
994 if (shouldParkAfterFailedAcquire(p, node) &&
995 parkAndCheckInterrupt())
996 interrupted = true;
997 }
998 } catch (Throwable t) {
999 cancelAcquire(node);
1000 throw t;
1001 }
1002 }
1003
1004 /**
1005 * Acquires in shared interruptible mode.
1006 * @param arg the acquire argument
1007 */
1008 private void doAcquireSharedInterruptibly(int arg)
1009 throws InterruptedException {
1010 final Node node = addWaiter(Node.SHARED);
1011 try {
1012 for (;;) {
1013 final Node p = node.predecessor();
1014 if (p == head) {
1015 int r = tryAcquireShared(arg);
1016 if (r >= 0) {
1017 setHeadAndPropagate(node, r);
1018 p.next = null; // help GC
1019 return;
1020 }
1021 }
1022 if (shouldParkAfterFailedAcquire(p, node) &&
1023 parkAndCheckInterrupt())
1024 throw new InterruptedException();
1025 }
1026 } catch (Throwable t) {
1027 cancelAcquire(node);
1028 throw t;
1029 }
1030 }
1031
1032 /**
1033 * Acquires in shared timed mode.
1034 *
1035 * @param arg the acquire argument
1036 * @param nanosTimeout max wait time
1037 * @return {@code true} if acquired
1038 */
1039 private boolean doAcquireSharedNanos(int arg, long nanosTimeout)
1040 throws InterruptedException {
1041 if (nanosTimeout <= 0L)
1042 return false;
1043 final long deadline = System.nanoTime() + nanosTimeout;
1044 final Node node = addWaiter(Node.SHARED);
1045 try {
1046 for (;;) {
1047 final Node p = node.predecessor();
1048 if (p == head) {
1049 int r = tryAcquireShared(arg);
1050 if (r >= 0) {
1051 setHeadAndPropagate(node, r);
1052 p.next = null; // help GC
1053 return true;
1054 }
1055 }
1056 nanosTimeout = deadline - System.nanoTime();
1057 if (nanosTimeout <= 0L) {
1058 cancelAcquire(node);
1059 return false;
1060 }
1061 if (shouldParkAfterFailedAcquire(p, node) &&
1062 nanosTimeout > SPIN_FOR_TIMEOUT_THRESHOLD)
1063 LockSupport.parkNanos(this, nanosTimeout);
1064 if (Thread.interrupted())
1065 throw new InterruptedException();
1066 }
1067 } catch (Throwable t) {
1068 cancelAcquire(node);
1069 throw t;
1070 }
1071 }
1072
1073 // Main exported methods
1074
1075 /**
1076 * Attempts to acquire in exclusive mode. This method should query
1077 * if the state of the object permits it to be acquired in the
1078 * exclusive mode, and if so to acquire it.
1079 *
1080 * <p>This method is always invoked by the thread performing
1081 * acquire. If this method reports failure, the acquire method
1082 * may queue the thread, if it is not already queued, until it is
1083 * signalled by a release from some other thread. This can be used
1084 * to implement method {@link Lock#tryLock()}.
1085 *
1086 * <p>The default
1087 * implementation throws {@link UnsupportedOperationException}.
1088 *
1089 * @param arg the acquire argument. This value is always the one
1090 * passed to an acquire method, or is the value saved on entry
1091 * to a condition wait. The value is otherwise uninterpreted
1092 * and can represent anything you like.
1093 * @return {@code true} if successful. Upon success, this object has
1094 * been acquired.
1095 * @throws IllegalMonitorStateException if acquiring would place this
1096 * synchronizer in an illegal state. This exception must be
1097 * thrown in a consistent fashion for synchronization to work
1098 * correctly.
1099 * @throws UnsupportedOperationException if exclusive mode is not supported
1100 */
1101 protected boolean tryAcquire(int arg) {
1102 throw new UnsupportedOperationException();
1103 }
1104
1105 /**
1106 * Attempts to set the state to reflect a release in exclusive
1107 * mode.
1108 *
1109 * <p>This method is always invoked by the thread performing release.
1110 *
1111 * <p>The default implementation throws
1112 * {@link UnsupportedOperationException}.
1113 *
1114 * @param arg the release argument. This value is always the one
1115 * passed to a release method, or the current state value upon
1116 * entry to a condition wait. The value is otherwise
1117 * uninterpreted and can represent anything you like.
1118 * @return {@code true} if this object is now in a fully released
1119 * state, so that any waiting threads may attempt to acquire;
1120 * and {@code false} otherwise.
1121 * @throws IllegalMonitorStateException if releasing would place this
1122 * synchronizer in an illegal state. This exception must be
1123 * thrown in a consistent fashion for synchronization to work
1124 * correctly.
1125 * @throws UnsupportedOperationException if exclusive mode is not supported
1126 */
1127 protected boolean tryRelease(int arg) {
1128 throw new UnsupportedOperationException();
1129 }
1130
1131 /**
1132 * Attempts to acquire in shared mode. This method should query if
1133 * the state of the object permits it to be acquired in the shared
1134 * mode, and if so to acquire it.
1135 *
1136 * <p>This method is always invoked by the thread performing
1137 * acquire. If this method reports failure, the acquire method
1138 * may queue the thread, if it is not already queued, until it is
1139 * signalled by a release from some other thread.
1140 *
1141 * <p>The default implementation throws {@link
1142 * UnsupportedOperationException}.
1143 *
1144 * @param arg the acquire argument. This value is always the one
1145 * passed to an acquire method, or is the value saved on entry
1146 * to a condition wait. The value is otherwise uninterpreted
1147 * and can represent anything you like.
1148 * @return a negative value on failure; zero if acquisition in shared
1149 * mode succeeded but no subsequent shared-mode acquire can
1150 * succeed; and a positive value if acquisition in shared
1151 * mode succeeded and subsequent shared-mode acquires might
1152 * also succeed, in which case a subsequent waiting thread
1153 * must check availability. (Support for three different
1154 * return values enables this method to be used in contexts
1155 * where acquires only sometimes act exclusively.) Upon
1156 * success, this object has been acquired.
1157 * @throws IllegalMonitorStateException if acquiring would place this
1158 * synchronizer in an illegal state. This exception must be
1159 * thrown in a consistent fashion for synchronization to work
1160 * correctly.
1161 * @throws UnsupportedOperationException if shared mode is not supported
1162 */
1163 protected int tryAcquireShared(int arg) {
1164 throw new UnsupportedOperationException();
1165 }
1166
1167 /**
1168 * Attempts to set the state to reflect a release in shared mode.
1169 *
1170 * <p>This method is always invoked by the thread performing release.
1171 *
1172 * <p>The default implementation throws
1173 * {@link UnsupportedOperationException}.
1174 *
1175 * @param arg the release argument. This value is always the one
1176 * passed to a release method, or the current state value upon
1177 * entry to a condition wait. The value is otherwise
1178 * uninterpreted and can represent anything you like.
1179 * @return {@code true} if this release of shared mode may permit a
1180 * waiting acquire (shared or exclusive) to succeed; and
1181 * {@code false} otherwise
1182 * @throws IllegalMonitorStateException if releasing would place this
1183 * synchronizer in an illegal state. This exception must be
1184 * thrown in a consistent fashion for synchronization to work
1185 * correctly.
1186 * @throws UnsupportedOperationException if shared mode is not supported
1187 */
1188 protected boolean tryReleaseShared(int arg) {
1189 throw new UnsupportedOperationException();
1190 }
1191
1192 /**
1193 * Returns {@code true} if synchronization is held exclusively with
1194 * respect to the current (calling) thread. This method is invoked
1195 * upon each call to a non-waiting {@link ConditionObject} method.
1196 * (Waiting methods instead invoke {@link #release}.)
1197 *
1198 * <p>The default implementation throws {@link
1199 * UnsupportedOperationException}. This method is invoked
1200 * internally only within {@link ConditionObject} methods, so need
1201 * not be defined if conditions are not used.
1202 *
1203 * @return {@code true} if synchronization is held exclusively;
1204 * {@code false} otherwise
1205 * @throws UnsupportedOperationException if conditions are not supported
1206 */
1207 protected boolean isHeldExclusively() {
1208 throw new UnsupportedOperationException();
1209 }
1210
1211 /**
1212 * Acquires in exclusive mode, ignoring interrupts. Implemented
1213 * by invoking at least once {@link #tryAcquire},
1214 * returning on success. Otherwise the thread is queued, possibly
1215 * repeatedly blocking and unblocking, invoking {@link
1216 * #tryAcquire} until success. This method can be used
1217 * to implement method {@link Lock#lock}.
1218 *
1219 * @param arg the acquire argument. This value is conveyed to
1220 * {@link #tryAcquire} but is otherwise uninterpreted and
1221 * can represent anything you like.
1222 */
1223 @ReservedStackAccess
1224 public final void acquire(int arg) {
1225 if (!tryAcquire(arg) &&
1226 acquireQueued(addWaiter(Node.EXCLUSIVE), arg))
1227 selfInterrupt();
1228 }
1229
1230 /**
1231 * Acquires in exclusive mode, aborting if interrupted.
1232 * Implemented by first checking interrupt status, then invoking
1233 * at least once {@link #tryAcquire}, returning on
1234 * success. Otherwise the thread is queued, possibly repeatedly
1235 * blocking and unblocking, invoking {@link #tryAcquire}
1236 * until success or the thread is interrupted. This method can be
1237 * used to implement method {@link Lock#lockInterruptibly}.
1238 *
1239 * @param arg the acquire argument. This value is conveyed to
1240 * {@link #tryAcquire} but is otherwise uninterpreted and
1241 * can represent anything you like.
1242 * @throws InterruptedException if the current thread is interrupted
1243 */
1244 public final void acquireInterruptibly(int arg)
1245 throws InterruptedException {
1246 if (Thread.interrupted())
1247 throw new InterruptedException();
1248 if (!tryAcquire(arg))
1249 doAcquireInterruptibly(arg);
1250 }
1251
1252 /**
1253 * Attempts to acquire in exclusive mode, aborting if interrupted,
1254 * and failing if the given timeout elapses. Implemented by first
1255 * checking interrupt status, then invoking at least once {@link
1256 * #tryAcquire}, returning on success. Otherwise, the thread is
1257 * queued, possibly repeatedly blocking and unblocking, invoking
1258 * {@link #tryAcquire} until success or the thread is interrupted
1259 * or the timeout elapses. This method can be used to implement
1260 * method {@link Lock#tryLock(long, TimeUnit)}.
1261 *
1262 * @param arg the acquire argument. This value is conveyed to
1263 * {@link #tryAcquire} but is otherwise uninterpreted and
1264 * can represent anything you like.
1265 * @param nanosTimeout the maximum number of nanoseconds to wait
1266 * @return {@code true} if acquired; {@code false} if timed out
1267 * @throws InterruptedException if the current thread is interrupted
1268 */
1269 public final boolean tryAcquireNanos(int arg, long nanosTimeout)
1270 throws InterruptedException {
1271 if (Thread.interrupted())
1272 throw new InterruptedException();
1273 return tryAcquire(arg) ||
1274 doAcquireNanos(arg, nanosTimeout);
1275 }
1276
1277 /**
1278 * Releases in exclusive mode. Implemented by unblocking one or
1279 * more threads if {@link #tryRelease} returns true.
1280 * This method can be used to implement method {@link Lock#unlock}.
1281 *
1282 * @param arg the release argument. This value is conveyed to
1283 * {@link #tryRelease} but is otherwise uninterpreted and
1284 * can represent anything you like.
1285 * @return the value returned from {@link #tryRelease}
1286 */
1287 @ReservedStackAccess
1288 public final boolean release(int arg) {
1289 if (tryRelease(arg)) {
1290 Node h = head;
1291 if (h != null && h.waitStatus != 0)
1292 unparkSuccessor(h);
1293 return true;
1294 }
1295 return false;
1296 }
1297
1298 /**
1299 * Acquires in shared mode, ignoring interrupts. Implemented by
1300 * first invoking at least once {@link #tryAcquireShared},
1301 * returning on success. Otherwise the thread is queued, possibly
1302 * repeatedly blocking and unblocking, invoking {@link
1303 * #tryAcquireShared} until success.
1304 *
1305 * @param arg the acquire argument. This value is conveyed to
1306 * {@link #tryAcquireShared} but is otherwise uninterpreted
1307 * and can represent anything you like.
1308 */
1309 public final void acquireShared(int arg) {
1310 if (tryAcquireShared(arg) < 0)
1311 doAcquireShared(arg);
1312 }
1313
1314 /**
1315 * Acquires in shared mode, aborting if interrupted. Implemented
1316 * by first checking interrupt status, then invoking at least once
1317 * {@link #tryAcquireShared}, returning on success. Otherwise the
1318 * thread is queued, possibly repeatedly blocking and unblocking,
1319 * invoking {@link #tryAcquireShared} until success or the thread
1320 * is interrupted.
1321 * @param arg the acquire argument.
1322 * This value is conveyed to {@link #tryAcquireShared} but is
1323 * otherwise uninterpreted and can represent anything
1324 * you like.
1325 * @throws InterruptedException if the current thread is interrupted
1326 */
1327 public final void acquireSharedInterruptibly(int arg)
1328 throws InterruptedException {
1329 if (Thread.interrupted())
1330 throw new InterruptedException();
1331 if (tryAcquireShared(arg) < 0)
1332 doAcquireSharedInterruptibly(arg);
1333 }
1334
1335 /**
1336 * Attempts to acquire in shared mode, aborting if interrupted, and
1337 * failing if the given timeout elapses. Implemented by first
1338 * checking interrupt status, then invoking at least once {@link
1339 * #tryAcquireShared}, returning on success. Otherwise, the
1340 * thread is queued, possibly repeatedly blocking and unblocking,
1341 * invoking {@link #tryAcquireShared} until success or the thread
1342 * is interrupted or the timeout elapses.
1343 *
1344 * @param arg the acquire argument. This value is conveyed to
1345 * {@link #tryAcquireShared} but is otherwise uninterpreted
1346 * and can represent anything you like.
1347 * @param nanosTimeout the maximum number of nanoseconds to wait
1348 * @return {@code true} if acquired; {@code false} if timed out
1349 * @throws InterruptedException if the current thread is interrupted
1350 */
1351 public final boolean tryAcquireSharedNanos(int arg, long nanosTimeout)
1352 throws InterruptedException {
1353 if (Thread.interrupted())
1354 throw new InterruptedException();
1355 return tryAcquireShared(arg) >= 0 ||
1356 doAcquireSharedNanos(arg, nanosTimeout);
1357 }
1358
1359 /**
1360 * Releases in shared mode. Implemented by unblocking one or more
1361 * threads if {@link #tryReleaseShared} returns true.
1362 *
1363 * @param arg the release argument. This value is conveyed to
1364 * {@link #tryReleaseShared} but is otherwise uninterpreted
1365 * and can represent anything you like.
1366 * @return the value returned from {@link #tryReleaseShared}
1367 */
1368 @ReservedStackAccess
1369 public final boolean releaseShared(int arg) {
1370 if (tryReleaseShared(arg)) {
1371 doReleaseShared();
1372 return true;
1373 }
1374 return false;
1375 }
1376
1377 // Queue inspection methods
1378
1379 /**
1380 * Queries whether any threads are waiting to acquire. Note that
1381 * because cancellations due to interrupts and timeouts may occur
1382 * at any time, a {@code true} return does not guarantee that any
1383 * other thread will ever acquire.
1384 *
1385 * <p>In this implementation, this operation returns in
1386 * constant time.
1387 *
1388 * @return {@code true} if there may be other threads waiting to acquire
1389 */
1390 public final boolean hasQueuedThreads() {
1391 return head != tail;
1392 }
1393
1394 /**
1395 * Queries whether any threads have ever contended to acquire this
1396 * synchronizer; that is if an acquire method has ever blocked.
1397 *
1398 * <p>In this implementation, this operation returns in
1399 * constant time.
1400 *
1401 * @return {@code true} if there has ever been contention
1402 */
1403 public final boolean hasContended() {
1404 return head != null;
1405 }
1406
1407 /**
1408 * Returns the first (longest-waiting) thread in the queue, or
1409 * {@code null} if no threads are currently queued.
1410 *
1411 * <p>In this implementation, this operation normally returns in
1412 * constant time, but may iterate upon contention if other threads are
1413 * concurrently modifying the queue.
1414 *
1415 * @return the first (longest-waiting) thread in the queue, or
1416 * {@code null} if no threads are currently queued
1417 */
1418 public final Thread getFirstQueuedThread() {
1419 // handle only fast path, else relay
1420 return (head == tail) ? null : fullGetFirstQueuedThread();
1421 }
1422
1423 /**
1424 * Version of getFirstQueuedThread called when fastpath fails.
1425 */
1426 private Thread fullGetFirstQueuedThread() {
1427 /*
1428 * The first node is normally head.next. Try to get its
1429 * thread field, ensuring consistent reads: If thread
1430 * field is nulled out or s.prev is no longer head, then
1431 * some other thread(s) concurrently performed setHead in
1432 * between some of our reads. We try this twice before
1433 * resorting to traversal.
1434 */
1435 Node h, s;
1436 Thread st;
1437 if (((h = head) != null && (s = h.next) != null &&
1438 s.prev == head && (st = s.thread) != null) ||
1439 ((h = head) != null && (s = h.next) != null &&
1440 s.prev == head && (st = s.thread) != null))
1441 return st;
1442
1443 /*
1444 * Head's next field might not have been set yet, or may have
1445 * been unset after setHead. So we must check to see if tail
1446 * is actually first node. If not, we continue on, safely
1447 * traversing from tail back to head to find first,
1448 * guaranteeing termination.
1449 */
1450
1451 Node t = tail;
1452 Thread firstThread = null;
1453 while (t != null && t != head) {
1454 Thread tt = t.thread;
1455 if (tt != null)
1456 firstThread = tt;
1457 t = t.prev;
1458 }
1459 return firstThread;
1460 }
1461
1462 /**
1463 * Returns true if the given thread is currently queued.
1464 *
1465 * <p>This implementation traverses the queue to determine
1466 * presence of the given thread.
1467 *
1468 * @param thread the thread
1469 * @return {@code true} if the given thread is on the queue
1470 * @throws NullPointerException if the thread is null
1471 */
1472 public final boolean isQueued(Thread thread) {
1473 if (thread == null)
1474 throw new NullPointerException();
1475 for (Node p = tail; p != null; p = p.prev)
1476 if (p.thread == thread)
1477 return true;
1478 return false;
1479 }
1480
1481 /**
1482 * Returns {@code true} if the apparent first queued thread, if one
1483 * exists, is waiting in exclusive mode. If this method returns
1484 * {@code true}, and the current thread is attempting to acquire in
1485 * shared mode (that is, this method is invoked from {@link
1486 * #tryAcquireShared}) then it is guaranteed that the current thread
1487 * is not the first queued thread. Used only as a heuristic in
1488 * ReentrantReadWriteLock.
1489 */
1490 final boolean apparentlyFirstQueuedIsExclusive() {
1491 Node h, s;
1492 return (h = head) != null &&
1493 (s = h.next) != null &&
1494 !s.isShared() &&
1495 s.thread != null;
1496 }
1497
1498 /**
1499 * Queries whether any threads have been waiting to acquire longer
1500 * than the current thread.
1501 *
1502 * <p>An invocation of this method is equivalent to (but may be
1503 * more efficient than):
1504 * <pre> {@code
1505 * getFirstQueuedThread() != Thread.currentThread() &&
1506 * hasQueuedThreads()}</pre>
1507 *
1508 * <p>Note that because cancellations due to interrupts and
1509 * timeouts may occur at any time, a {@code true} return does not
1510 * guarantee that some other thread will acquire before the current
1511 * thread. Likewise, it is possible for another thread to win a
1512 * race to enqueue after this method has returned {@code false},
1513 * due to the queue being empty.
1514 *
1515 * <p>This method is designed to be used by a fair synchronizer to
1516 * avoid <a href="AbstractQueuedSynchronizer.html#barging">barging</a>.
1517 * Such a synchronizer's {@link #tryAcquire} method should return
1518 * {@code false}, and its {@link #tryAcquireShared} method should
1519 * return a negative value, if this method returns {@code true}
1520 * (unless this is a reentrant acquire). For example, the {@code
1521 * tryAcquire} method for a fair, reentrant, exclusive mode
1522 * synchronizer might look like this:
1523 *
1524 * <pre> {@code
1525 * protected boolean tryAcquire(int arg) {
1526 * if (isHeldExclusively()) {
1527 * // A reentrant acquire; increment hold count
1528 * return true;
1529 * } else if (hasQueuedPredecessors()) {
1530 * return false;
1531 * } else {
1532 * // try to acquire normally
1533 * }
1534 * }}</pre>
1535 *
1536 * @return {@code true} if there is a queued thread preceding the
1537 * current thread, and {@code false} if the current thread
1538 * is at the head of the queue or the queue is empty
1539 * @since 1.7
1540 */
1541 public final boolean hasQueuedPredecessors() {
1542 // The correctness of this depends on head being initialized
1543 // before tail and on head.next being accurate if the current
1544 // thread is first in queue.
1545 Node t = tail; // Read fields in reverse initialization order
1546 Node h = head;
1547 Node s;
1548 return h != t &&
1549 ((s = h.next) == null || s.thread != Thread.currentThread());
1550 }
1551
1552
1553 // Instrumentation and monitoring methods
1554
1555 /**
1556 * Returns an estimate of the number of threads waiting to
1557 * acquire. The value is only an estimate because the number of
1558 * threads may change dynamically while this method traverses
1559 * internal data structures. This method is designed for use in
1560 * monitoring system state, not for synchronization control.
1561 *
1562 * @return the estimated number of threads waiting to acquire
1563 */
1564 public final int getQueueLength() {
1565 int n = 0;
1566 for (Node p = tail; p != null; p = p.prev) {
1567 if (p.thread != null)
1568 ++n;
1569 }
1570 return n;
1571 }
1572
1573 /**
1574 * Returns a collection containing threads that may be waiting to
1575 * acquire. Because the actual set of threads may change
1576 * dynamically while constructing this result, the returned
1577 * collection is only a best-effort estimate. The elements of the
1578 * returned collection are in no particular order. This method is
1579 * designed to facilitate construction of subclasses that provide
1580 * more extensive monitoring facilities.
1581 *
1582 * @return the collection of threads
1583 */
1584 public final Collection<Thread> getQueuedThreads() {
1585 ArrayList<Thread> list = new ArrayList<>();
1586 for (Node p = tail; p != null; p = p.prev) {
1587 Thread t = p.thread;
1588 if (t != null)
1589 list.add(t);
1590 }
1591 return list;
1592 }
1593
1594 /**
1595 * Returns a collection containing threads that may be waiting to
1596 * acquire in exclusive mode. This has the same properties
1597 * as {@link #getQueuedThreads} except that it only returns
1598 * those threads waiting due to an exclusive acquire.
1599 *
1600 * @return the collection of threads
1601 */
1602 public final Collection<Thread> getExclusiveQueuedThreads() {
1603 ArrayList<Thread> list = new ArrayList<>();
1604 for (Node p = tail; p != null; p = p.prev) {
1605 if (!p.isShared()) {
1606 Thread t = p.thread;
1607 if (t != null)
1608 list.add(t);
1609 }
1610 }
1611 return list;
1612 }
1613
1614 /**
1615 * Returns a collection containing threads that may be waiting to
1616 * acquire in shared mode. This has the same properties
1617 * as {@link #getQueuedThreads} except that it only returns
1618 * those threads waiting due to a shared acquire.
1619 *
1620 * @return the collection of threads
1621 */
1622 public final Collection<Thread> getSharedQueuedThreads() {
1623 ArrayList<Thread> list = new ArrayList<>();
1624 for (Node p = tail; p != null; p = p.prev) {
1625 if (p.isShared()) {
1626 Thread t = p.thread;
1627 if (t != null)
1628 list.add(t);
1629 }
1630 }
1631 return list;
1632 }
1633
1634 /**
1635 * Returns a string identifying this synchronizer, as well as its state.
1636 * The state, in brackets, includes the String {@code "State ="}
1637 * followed by the current value of {@link #getState}, and either
1638 * {@code "nonempty"} or {@code "empty"} depending on whether the
1639 * queue is empty.
1640 *
1641 * @return a string identifying this synchronizer, as well as its state
1642 */
1643 public String toString() {
1644 return super.toString()
1645 + "[State = " + getState() + ", "
1646 + (hasQueuedThreads() ? "non" : "") + "empty queue]";
1647 }
1648
1649
1650 // Internal support methods for Conditions
1651
1652 /**
1653 * Returns true if a node, always one that was initially placed on
1654 * a condition queue, is now waiting to reacquire on sync queue.
1655 * @param node the node
1656 * @return true if is reacquiring
1657 */
1658 final boolean isOnSyncQueue(Node node) {
1659 if (node.waitStatus == Node.CONDITION || node.prev == null)
1660 return false;
1661 if (node.next != null) // If has successor, it must be on queue
1662 return true;
1663 /*
1664 * node.prev can be non-null, but not yet on queue because
1665 * the CAS to place it on queue can fail. So we have to
1666 * traverse from tail to make sure it actually made it. It
1667 * will always be near the tail in calls to this method, and
1668 * unless the CAS failed (which is unlikely), it will be
1669 * there, so we hardly ever traverse much.
1670 */
1671 return findNodeFromTail(node);
1672 }
1673
1674 /**
1675 * Returns true if node is on sync queue by searching backwards from tail.
1676 * Called only when needed by isOnSyncQueue.
1677 * @return true if present
1678 */
1679 private boolean findNodeFromTail(Node node) {
1680 // We check for node first, since it's likely to be at or near tail.
1681 // tail is known to be non-null, so we could re-order to "save"
1682 // one null check, but we leave it this way to help the VM.
1683 for (Node p = tail;;) {
1684 if (p == node)
1685 return true;
1686 if (p == null)
1687 return false;
1688 p = p.prev;
1689 }
1690 }
1691
1692 /**
1693 * Transfers a node from a condition queue onto sync queue.
1694 * Returns true if successful.
1695 * @param node the node
1696 * @return true if successfully transferred (else the node was
1697 * cancelled before signal)
1698 */
1699 final boolean transferForSignal(Node node) {
1700 /*
1701 * If cannot change waitStatus, the node has been cancelled.
1702 */
1703 if (!node.compareAndSetWaitStatus(Node.CONDITION, 0))
1704 return false;
1705
1706 /*
1707 * Splice onto queue and try to set waitStatus of predecessor to
1708 * indicate that thread is (probably) waiting. If cancelled or
1709 * attempt to set waitStatus fails, wake up to resync (in which
1710 * case the waitStatus can be transiently and harmlessly wrong).
1711 */
1712 Node p = enq(node);
1713 int ws = p.waitStatus;
1714 if (ws > 0 || !p.compareAndSetWaitStatus(ws, Node.SIGNAL))
1715 LockSupport.unpark(node.thread);
1716 return true;
1717 }
1718
1719 /**
1720 * Transfers node, if necessary, to sync queue after a cancelled wait.
1721 * Returns true if thread was cancelled before being signalled.
1722 *
1723 * @param node the node
1724 * @return true if cancelled before the node was signalled
1725 */
1726 final boolean transferAfterCancelledWait(Node node) {
1727 if (node.compareAndSetWaitStatus(Node.CONDITION, 0)) {
1728 enq(node);
1729 return true;
1730 }
1731 /*
1732 * If we lost out to a signal(), then we can't proceed
1733 * until it finishes its enq(). Cancelling during an
1734 * incomplete transfer is both rare and transient, so just
1735 * spin.
1736 */
1737 while (!isOnSyncQueue(node))
1738 Thread.yield();
1739 return false;
1740 }
1741
1742 /**
1743 * Invokes release with current state value; returns saved state.
1744 * Cancels node and throws exception on failure.
1745 * @param node the condition node for this wait
1746 * @return previous sync state
1747 */
1748 final int fullyRelease(Node node) {
1749 try {
1750 int savedState = getState();
1751 if (release(savedState))
1752 return savedState;
1753 throw new IllegalMonitorStateException();
1754 } catch (Throwable t) {
1755 node.waitStatus = Node.CANCELLED;
1756 throw t;
1757 }
1758 }
1759
1760 // Instrumentation methods for conditions
1761
1762 /**
1763 * Queries whether the given ConditionObject
1764 * uses this synchronizer as its lock.
1765 *
1766 * @param condition the condition
1767 * @return {@code true} if owned
1768 * @throws NullPointerException if the condition is null
1769 */
1770 public final boolean owns(ConditionObject condition) {
1771 return condition.isOwnedBy(this);
1772 }
1773
1774 /**
1775 * Queries whether any threads are waiting on the given condition
1776 * associated with this synchronizer. Note that because timeouts
1777 * and interrupts may occur at any time, a {@code true} return
1778 * does not guarantee that a future {@code signal} will awaken
1779 * any threads. This method is designed primarily for use in
1780 * monitoring of the system state.
1781 *
1782 * @param condition the condition
1783 * @return {@code true} if there are any waiting threads
1784 * @throws IllegalMonitorStateException if exclusive synchronization
1785 * is not held
1786 * @throws IllegalArgumentException if the given condition is
1787 * not associated with this synchronizer
1788 * @throws NullPointerException if the condition is null
1789 */
1790 public final boolean hasWaiters(ConditionObject condition) {
1791 if (!owns(condition))
1792 throw new IllegalArgumentException("Not owner");
1793 return condition.hasWaiters();
1794 }
1795
1796 /**
1797 * Returns an estimate of the number of threads waiting on the
1798 * given condition associated with this synchronizer. Note that
1799 * because timeouts and interrupts may occur at any time, the
1800 * estimate serves only as an upper bound on the actual number of
1801 * waiters. This method is designed for use in monitoring system
1802 * state, not for synchronization control.
1803 *
1804 * @param condition the condition
1805 * @return the estimated number of waiting threads
1806 * @throws IllegalMonitorStateException if exclusive synchronization
1807 * is not held
1808 * @throws IllegalArgumentException if the given condition is
1809 * not associated with this synchronizer
1810 * @throws NullPointerException if the condition is null
1811 */
1812 public final int getWaitQueueLength(ConditionObject condition) {
1813 if (!owns(condition))
1814 throw new IllegalArgumentException("Not owner");
1815 return condition.getWaitQueueLength();
1816 }
1817
1818 /**
1819 * Returns a collection containing those threads that may be
1820 * waiting on the given condition associated with this
1821 * synchronizer. Because the actual set of threads may change
1822 * dynamically while constructing this result, the returned
1823 * collection is only a best-effort estimate. The elements of the
1824 * returned collection are in no particular order.
1825 *
1826 * @param condition the condition
1827 * @return the collection of threads
1828 * @throws IllegalMonitorStateException if exclusive synchronization
1829 * is not held
1830 * @throws IllegalArgumentException if the given condition is
1831 * not associated with this synchronizer
1832 * @throws NullPointerException if the condition is null
1833 */
1834 public final Collection<Thread> getWaitingThreads(ConditionObject condition) {
1835 if (!owns(condition))
1836 throw new IllegalArgumentException("Not owner");
1837 return condition.getWaitingThreads();
1838 }
1839
1840 /**
1841 * Condition implementation for a {@link
1842 * AbstractQueuedSynchronizer} serving as the basis of a {@link
1843 * Lock} implementation.
1844 *
1845 * <p>Method documentation for this class describes mechanics,
1846 * not behavioral specifications from the point of view of Lock
1847 * and Condition users. Exported versions of this class will in
1848 * general need to be accompanied by documentation describing
1849 * condition semantics that rely on those of the associated
1850 * {@code AbstractQueuedSynchronizer}.
1851 *
1852 * <p>This class is Serializable, but all fields are transient,
1853 * so deserialized conditions have no waiters.
1854 */
1855 public class ConditionObject implements Condition, java.io.Serializable {
1856 private static final long serialVersionUID = 1173984872572414699L;
1857 /** First node of condition queue. */
1858 private transient Node firstWaiter;
1859 /** Last node of condition queue. */
1860 private transient Node lastWaiter;
1861
1862 /**
1863 * Creates a new {@code ConditionObject} instance.
1864 */
1865 public ConditionObject() { }
1866
1867 // Internal methods
1868
1869 /**
1870 * Adds a new waiter to wait queue.
1871 * @return its new wait node
1872 */
1873 private Node addConditionWaiter() {
1874 Node t = lastWaiter;
1875 // If lastWaiter is cancelled, clean out.
1876 if (t != null && t.waitStatus != Node.CONDITION) {
1877 unlinkCancelledWaiters();
1878 t = lastWaiter;
1879 }
1880
1881 Node node = new Node(Node.CONDITION);
1882
1883 if (t == null)
1884 firstWaiter = node;
1885 else
1886 t.nextWaiter = node;
1887 lastWaiter = node;
1888 return node;
1889 }
1890
1891 /**
1892 * Removes and transfers nodes until hit non-cancelled one or
1893 * null. Split out from signal in part to encourage compilers
1894 * to inline the case of no waiters.
1895 * @param first (non-null) the first node on condition queue
1896 */
1897 private void doSignal(Node first) {
1898 do {
1899 if ( (firstWaiter = first.nextWaiter) == null)
1900 lastWaiter = null;
1901 first.nextWaiter = null;
1902 } while (!transferForSignal(first) &&
1903 (first = firstWaiter) != null);
1904 }
1905
1906 /**
1907 * Removes and transfers all nodes.
1908 * @param first (non-null) the first node on condition queue
1909 */
1910 private void doSignalAll(Node first) {
1911 lastWaiter = firstWaiter = null;
1912 do {
1913 Node next = first.nextWaiter;
1914 first.nextWaiter = null;
1915 transferForSignal(first);
1916 first = next;
1917 } while (first != null);
1918 }
1919
1920 /**
1921 * Unlinks cancelled waiter nodes from condition queue.
1922 * Called only while holding lock. This is called when
1923 * cancellation occurred during condition wait, and upon
1924 * insertion of a new waiter when lastWaiter is seen to have
1925 * been cancelled. This method is needed to avoid garbage
1926 * retention in the absence of signals. So even though it may
1927 * require a full traversal, it comes into play only when
1928 * timeouts or cancellations occur in the absence of
1929 * signals. It traverses all nodes rather than stopping at a
1930 * particular target to unlink all pointers to garbage nodes
1931 * without requiring many re-traversals during cancellation
1932 * storms.
1933 */
1934 private void unlinkCancelledWaiters() {
1935 Node t = firstWaiter;
1936 Node trail = null;
1937 while (t != null) {
1938 Node next = t.nextWaiter;
1939 if (t.waitStatus != Node.CONDITION) {
1940 t.nextWaiter = null;
1941 if (trail == null)
1942 firstWaiter = next;
1943 else
1944 trail.nextWaiter = next;
1945 if (next == null)
1946 lastWaiter = trail;
1947 }
1948 else
1949 trail = t;
1950 t = next;
1951 }
1952 }
1953
1954 // public methods
1955
1956 /**
1957 * Moves the longest-waiting thread, if one exists, from the
1958 * wait queue for this condition to the wait queue for the
1959 * owning lock.
1960 *
1961 * @throws IllegalMonitorStateException if {@link #isHeldExclusively}
1962 * returns {@code false}
1963 */
1964 public final void signal() {
1965 if (!isHeldExclusively())
1966 throw new IllegalMonitorStateException();
1967 Node first = firstWaiter;
1968 if (first != null)
1969 doSignal(first);
1970 }
1971
1972 /**
1973 * Moves all threads from the wait queue for this condition to
1974 * the wait queue for the owning lock.
1975 *
1976 * @throws IllegalMonitorStateException if {@link #isHeldExclusively}
1977 * returns {@code false}
1978 */
1979 public final void signalAll() {
1980 if (!isHeldExclusively())
1981 throw new IllegalMonitorStateException();
1982 Node first = firstWaiter;
1983 if (first != null)
1984 doSignalAll(first);
1985 }
1986
1987 /**
1988 * Implements uninterruptible condition wait.
1989 * <ol>
1990 * <li>Save lock state returned by {@link #getState}.
1991 * <li>Invoke {@link #release} with saved state as argument,
1992 * throwing IllegalMonitorStateException if it fails.
1993 * <li>Block until signalled.
1994 * <li>Reacquire by invoking specialized version of
1995 * {@link #acquire} with saved state as argument.
1996 * </ol>
1997 */
1998 public final void awaitUninterruptibly() {
1999 Node node = addConditionWaiter();
2000 int savedState = fullyRelease(node);
2001 boolean interrupted = false;
2002 while (!isOnSyncQueue(node)) {
2003 LockSupport.park(this);
2004 if (Thread.interrupted())
2005 interrupted = true;
2006 }
2007 if (acquireQueued(node, savedState) || interrupted)
2008 selfInterrupt();
2009 }
2010
2011 /*
2012 * For interruptible waits, we need to track whether to throw
2013 * InterruptedException, if interrupted while blocked on
2014 * condition, versus reinterrupt current thread, if
2015 * interrupted while blocked waiting to re-acquire.
2016 */
2017
2018 /** Mode meaning to reinterrupt on exit from wait */
2019 private static final int REINTERRUPT = 1;
2020 /** Mode meaning to throw InterruptedException on exit from wait */
2021 private static final int THROW_IE = -1;
2022
2023 /**
2024 * Checks for interrupt, returning THROW_IE if interrupted
2025 * before signalled, REINTERRUPT if after signalled, or
2026 * 0 if not interrupted.
2027 */
2028 private int checkInterruptWhileWaiting(Node node) {
2029 return Thread.interrupted() ?
2030 (transferAfterCancelledWait(node) ? THROW_IE : REINTERRUPT) :
2031 0;
2032 }
2033
2034 /**
2035 * Throws InterruptedException, reinterrupts current thread, or
2036 * does nothing, depending on mode.
2037 */
2038 private void reportInterruptAfterWait(int interruptMode)
2039 throws InterruptedException {
2040 if (interruptMode == THROW_IE)
2041 throw new InterruptedException();
2042 else if (interruptMode == REINTERRUPT)
2043 selfInterrupt();
2044 }
2045
2046 /**
2047 * Implements interruptible condition wait.
2048 * <ol>
2049 * <li>If current thread is interrupted, throw InterruptedException.
2050 * <li>Save lock state returned by {@link #getState}.
2051 * <li>Invoke {@link #release} with saved state as argument,
2052 * throwing IllegalMonitorStateException if it fails.
2053 * <li>Block until signalled or interrupted.
2054 * <li>Reacquire by invoking specialized version of
2055 * {@link #acquire} with saved state as argument.
2056 * <li>If interrupted while blocked in step 4, throw InterruptedException.
2057 * </ol>
2058 */
2059 public final void await() throws InterruptedException {
2060 if (Thread.interrupted())
2061 throw new InterruptedException();
2062 Node node = addConditionWaiter();
2063 int savedState = fullyRelease(node);
2064 int interruptMode = 0;
2065 while (!isOnSyncQueue(node)) {
2066 LockSupport.park(this);
2067 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0)
2068 break;
2069 }
2070 if (acquireQueued(node, savedState) && interruptMode != THROW_IE)
2071 interruptMode = REINTERRUPT;
2072 if (node.nextWaiter != null) // clean up if cancelled
2073 unlinkCancelledWaiters();
2074 if (interruptMode != 0)
2075 reportInterruptAfterWait(interruptMode);
2076 }
2077
2078 /**
2079 * Implements timed condition wait.
2080 * <ol>
2081 * <li>If current thread is interrupted, throw InterruptedException.
2082 * <li>Save lock state returned by {@link #getState}.
2083 * <li>Invoke {@link #release} with saved state as argument,
2084 * throwing IllegalMonitorStateException if it fails.
2085 * <li>Block until signalled, interrupted, or timed out.
2086 * <li>Reacquire by invoking specialized version of
2087 * {@link #acquire} with saved state as argument.
2088 * <li>If interrupted while blocked in step 4, throw InterruptedException.
2089 * </ol>
2090 */
2091 public final long awaitNanos(long nanosTimeout)
2092 throws InterruptedException {
2093 if (Thread.interrupted())
2094 throw new InterruptedException();
2095 // We don't check for nanosTimeout <= 0L here, to allow
2096 // awaitNanos(0) as a way to "yield the lock".
2097 final long deadline = System.nanoTime() + nanosTimeout;
2098 long initialNanos = nanosTimeout;
2099 Node node = addConditionWaiter();
2100 int savedState = fullyRelease(node);
2101 int interruptMode = 0;
2102 while (!isOnSyncQueue(node)) {
2103 if (nanosTimeout <= 0L) {
2104 transferAfterCancelledWait(node);
2105 break;
2106 }
2107 if (nanosTimeout >= SPIN_FOR_TIMEOUT_THRESHOLD)
2108 LockSupport.parkNanos(this, nanosTimeout);
2109 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0)
2110 break;
2111 nanosTimeout = deadline - System.nanoTime();
2112 }
2113 if (acquireQueued(node, savedState) && interruptMode != THROW_IE)
2114 interruptMode = REINTERRUPT;
2115 if (node.nextWaiter != null)
2116 unlinkCancelledWaiters();
2117 if (interruptMode != 0)
2118 reportInterruptAfterWait(interruptMode);
2119 long remaining = deadline - System.nanoTime(); // avoid overflow
2120 return (remaining <= initialNanos) ? remaining : Long.MIN_VALUE;
2121 }
2122
2123 /**
2124 * Implements absolute timed condition wait.
2125 * <ol>
2126 * <li>If current thread is interrupted, throw InterruptedException.
2127 * <li>Save lock state returned by {@link #getState}.
2128 * <li>Invoke {@link #release} with saved state as argument,
2129 * throwing IllegalMonitorStateException if it fails.
2130 * <li>Block until signalled, interrupted, or timed out.
2131 * <li>Reacquire by invoking specialized version of
2132 * {@link #acquire} with saved state as argument.
2133 * <li>If interrupted while blocked in step 4, throw InterruptedException.
2134 * <li>If timed out while blocked in step 4, return false, else true.
2135 * </ol>
2136 */
2137 public final boolean awaitUntil(Date deadline)
2138 throws InterruptedException {
2139 long abstime = deadline.getTime();
2140 if (Thread.interrupted())
2141 throw new InterruptedException();
2142 Node node = addConditionWaiter();
2143 int savedState = fullyRelease(node);
2144 boolean timedout = false;
2145 int interruptMode = 0;
2146 while (!isOnSyncQueue(node)) {
2147 if (System.currentTimeMillis() >= abstime) {
2148 timedout = transferAfterCancelledWait(node);
2149 break;
2150 }
2151 LockSupport.parkUntil(this, abstime);
2152 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0)
2153 break;
2154 }
2155 if (acquireQueued(node, savedState) && interruptMode != THROW_IE)
2156 interruptMode = REINTERRUPT;
2157 if (node.nextWaiter != null)
2158 unlinkCancelledWaiters();
2159 if (interruptMode != 0)
2160 reportInterruptAfterWait(interruptMode);
2161 return !timedout;
2162 }
2163
2164 /**
2165 * Implements timed condition wait.
2166 * <ol>
2167 * <li>If current thread is interrupted, throw InterruptedException.
2168 * <li>Save lock state returned by {@link #getState}.
2169 * <li>Invoke {@link #release} with saved state as argument,
2170 * throwing IllegalMonitorStateException if it fails.
2171 * <li>Block until signalled, interrupted, or timed out.
2172 * <li>Reacquire by invoking specialized version of
2173 * {@link #acquire} with saved state as argument.
2174 * <li>If interrupted while blocked in step 4, throw InterruptedException.
2175 * <li>If timed out while blocked in step 4, return false, else true.
2176 * </ol>
2177 */
2178 public final boolean await(long time, TimeUnit unit)
2179 throws InterruptedException {
2180 long nanosTimeout = unit.toNanos(time);
2181 if (Thread.interrupted())
2182 throw new InterruptedException();
2183 // We don't check for nanosTimeout <= 0L here, to allow
2184 // await(0, unit) as a way to "yield the lock".
2185 final long deadline = System.nanoTime() + nanosTimeout;
2186 Node node = addConditionWaiter();
2187 int savedState = fullyRelease(node);
2188 boolean timedout = false;
2189 int interruptMode = 0;
2190 while (!isOnSyncQueue(node)) {
2191 if (nanosTimeout <= 0L) {
2192 timedout = transferAfterCancelledWait(node);
2193 break;
2194 }
2195 if (nanosTimeout >= SPIN_FOR_TIMEOUT_THRESHOLD)
2196 LockSupport.parkNanos(this, nanosTimeout);
2197 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0)
2198 break;
2199 nanosTimeout = deadline - System.nanoTime();
2200 }
2201 if (acquireQueued(node, savedState) && interruptMode != THROW_IE)
2202 interruptMode = REINTERRUPT;
2203 if (node.nextWaiter != null)
2204 unlinkCancelledWaiters();
2205 if (interruptMode != 0)
2206 reportInterruptAfterWait(interruptMode);
2207 return !timedout;
2208 }
2209
2210 // support for instrumentation
2211
2212 /**
2213 * Returns true if this condition was created by the given
2214 * synchronization object.
2215 *
2216 * @return {@code true} if owned
2217 */
2218 final boolean isOwnedBy(AbstractQueuedSynchronizer sync) {
2219 return sync == AbstractQueuedSynchronizer.this;
2220 }
2221
2222 /**
2223 * Queries whether any threads are waiting on this condition.
2224 * Implements {@link AbstractQueuedSynchronizer#hasWaiters(ConditionObject)}.
2225 *
2226 * @return {@code true} if there are any waiting threads
2227 * @throws IllegalMonitorStateException if {@link #isHeldExclusively}
2228 * returns {@code false}
2229 */
2230 protected final boolean hasWaiters() {
2231 if (!isHeldExclusively())
2232 throw new IllegalMonitorStateException();
2233 for (Node w = firstWaiter; w != null; w = w.nextWaiter) {
2234 if (w.waitStatus == Node.CONDITION)
2235 return true;
2236 }
2237 return false;
2238 }
2239
2240 /**
2241 * Returns an estimate of the number of threads waiting on
2242 * this condition.
2243 * Implements {@link AbstractQueuedSynchronizer#getWaitQueueLength(ConditionObject)}.
2244 *
2245 * @return the estimated number of waiting threads
2246 * @throws IllegalMonitorStateException if {@link #isHeldExclusively}
2247 * returns {@code false}
2248 */
2249 protected final int getWaitQueueLength() {
2250 if (!isHeldExclusively())
2251 throw new IllegalMonitorStateException();
2252 int n = 0;
2253 for (Node w = firstWaiter; w != null; w = w.nextWaiter) {
2254 if (w.waitStatus == Node.CONDITION)
2255 ++n;
2256 }
2257 return n;
2258 }
2259
2260 /**
2261 * Returns a collection containing those threads that may be
2262 * waiting on this Condition.
2263 * Implements {@link AbstractQueuedSynchronizer#getWaitingThreads(ConditionObject)}.
2264 *
2265 * @return the collection of threads
2266 * @throws IllegalMonitorStateException if {@link #isHeldExclusively}
2267 * returns {@code false}
2268 */
2269 protected final Collection<Thread> getWaitingThreads() {
2270 if (!isHeldExclusively())
2271 throw new IllegalMonitorStateException();
2272 ArrayList<Thread> list = new ArrayList<>();
2273 for (Node w = firstWaiter; w != null; w = w.nextWaiter) {
2274 if (w.waitStatus == Node.CONDITION) {
2275 Thread t = w.thread;
2276 if (t != null)
2277 list.add(t);
2278 }
2279 }
2280 return list;
2281 }
2282 }
2283
2284 /**
2285 * Setup to support compareAndSet. We need to natively implement
2286 * this here: For the sake of permitting future enhancements, we
2287 * cannot explicitly subclass AtomicInteger, which would be
2288 * efficient and useful otherwise. So, as the lesser of evils, we
2289 * natively implement using hotspot intrinsics API. And while we
2290 * are at it, we do the same for other CASable fields (which could
2291 * otherwise be done with atomic field updaters).
2292 */
2293 private static final sun.misc.Unsafe U = sun.misc.Unsafe.getUnsafe();
2294 private static final long STATE;
2295 private static final long HEAD;
2296 private static final long TAIL;
2297
2298 static {
2299 try {
2300 STATE = U.objectFieldOffset
2301 (AbstractQueuedSynchronizer.class.getDeclaredField("state"));
2302 HEAD = U.objectFieldOffset
2303 (AbstractQueuedSynchronizer.class.getDeclaredField("head"));
2304 TAIL = U.objectFieldOffset
2305 (AbstractQueuedSynchronizer.class.getDeclaredField("tail"));
2306 } catch (ReflectiveOperationException e) {
2307 throw new Error(e);
2308 }
2309
2310 // Reduce the risk of rare disastrous classloading in first call to
2311 // LockSupport.park: https://bugs.openjdk.java.net/browse/JDK-8074773
2312 Class<?> ensureLoaded = LockSupport.class;
2313 }
2314
2315 /**
2316 * Initializes head and tail fields on first contention.
2317 */
2318 private final void initializeSyncQueue() {
2319 if (U.compareAndSwapObject(this, HEAD, null, new Node()))
2320 tail = head;
2321 }
2322
2323 /**
2324 * CASes tail field.
2325 */
2326 private final boolean compareAndSetTail(Node expect, Node update) {
2327 return U.compareAndSwapObject(this, TAIL, expect, update);
2328 }
2329 }