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; 37 38 import java.util.ArrayList; 39 import java.util.ConcurrentModificationException; 40 import java.util.HashSet; 41 import java.util.Iterator; 42 import java.util.List; 43 import java.util.concurrent.atomic.AtomicInteger; 44 import java.util.concurrent.locks.AbstractQueuedSynchronizer; 45 import java.util.concurrent.locks.Condition; 46 import java.util.concurrent.locks.ReentrantLock; 47 48 // BEGIN android-note 49 // removed security manager docs 50 // END android-note 51 52 /** 53 * An {@link ExecutorService} that executes each submitted task using 54 * one of possibly several pooled threads, normally configured 55 * using {@link Executors} factory methods. 56 * 57 * <p>Thread pools address two different problems: they usually 58 * provide improved performance when executing large numbers of 59 * asynchronous tasks, due to reduced per-task invocation overhead, 60 * and they provide a means of bounding and managing the resources, 61 * including threads, consumed when executing a collection of tasks. 62 * Each {@code ThreadPoolExecutor} also maintains some basic 63 * statistics, such as the number of completed tasks. 64 * 65 * <p>To be useful across a wide range of contexts, this class 66 * provides many adjustable parameters and extensibility 67 * hooks. However, programmers are urged to use the more convenient 68 * {@link Executors} factory methods {@link 69 * Executors#newCachedThreadPool} (unbounded thread pool, with 70 * automatic thread reclamation), {@link Executors#newFixedThreadPool} 71 * (fixed size thread pool) and {@link 72 * Executors#newSingleThreadExecutor} (single background thread), that 73 * preconfigure settings for the most common usage 74 * scenarios. Otherwise, use the following guide when manually 75 * configuring and tuning this class: 76 * 77 * <dl> 78 * 79 * <dt>Core and maximum pool sizes</dt> 80 * 81 * <dd style="font-family:'DejaVu Sans', Arial, Helvetica, sans-serif"> 82 * A {@code ThreadPoolExecutor} will automatically adjust the 83 * pool size (see {@link #getPoolSize}) 84 * according to the bounds set by 85 * corePoolSize (see {@link #getCorePoolSize}) and 86 * maximumPoolSize (see {@link #getMaximumPoolSize}). 87 * 88 * When a new task is submitted in method {@link #execute(Runnable)}, 89 * and fewer than corePoolSize threads are running, a new thread is 90 * created to handle the request, even if other worker threads are 91 * idle. If there are more than corePoolSize but less than 92 * maximumPoolSize threads running, a new thread will be created only 93 * if the queue is full. By setting corePoolSize and maximumPoolSize 94 * the same, you create a fixed-size thread pool. By setting 95 * maximumPoolSize to an essentially unbounded value such as {@code 96 * Integer.MAX_VALUE}, you allow the pool to accommodate an arbitrary 97 * number of concurrent tasks. Most typically, core and maximum pool 98 * sizes are set only upon construction, but they may also be changed 99 * dynamically using {@link #setCorePoolSize} and {@link 100 * #setMaximumPoolSize}. </dd> 101 * 102 * <dt>On-demand construction</dt> 103 * 104 * <dd style="font-family:'DejaVu Sans', Arial, Helvetica, sans-serif"> 105 * By default, even core threads are initially created and 106 * started only when new tasks arrive, but this can be overridden 107 * dynamically using method {@link #prestartCoreThread} or {@link 108 * #prestartAllCoreThreads}. You probably want to prestart threads if 109 * you construct the pool with a non-empty queue. </dd> 110 * 111 * <dt>Creating new threads</dt> 112 * 113 * <dd style="font-family:'DejaVu Sans', Arial, Helvetica, sans-serif"> 114 * New threads are created using a {@link ThreadFactory}. If not 115 * otherwise specified, a {@link Executors#defaultThreadFactory} is 116 * used, that creates threads to all be in the same {@link 117 * ThreadGroup} and with the same {@code NORM_PRIORITY} priority and 118 * non-daemon status. By supplying a different ThreadFactory, you can 119 * alter the thread's name, thread group, priority, daemon status, 120 * etc. If a {@code ThreadFactory} fails to create a thread when asked 121 * by returning null from {@code newThread}, the executor will 122 * continue, but might not be able to execute any tasks. Threads 123 * should possess the "modifyThread" {@code RuntimePermission}. If 124 * worker threads or other threads using the pool do not possess this 125 * permission, service may be degraded: configuration changes may not 126 * take effect in a timely manner, and a shutdown pool may remain in a 127 * state in which termination is possible but not completed.</dd> 128 * 129 * <dt>Keep-alive times</dt> 130 * 131 * <dd style="font-family:'DejaVu Sans', Arial, Helvetica, sans-serif"> 132 * If the pool currently has more than corePoolSize threads, 133 * excess threads will be terminated if they have been idle for more 134 * than the keepAliveTime (see {@link #getKeepAliveTime(TimeUnit)}). 135 * This provides a means of reducing resource consumption when the 136 * pool is not being actively used. If the pool becomes more active 137 * later, new threads will be constructed. This parameter can also be 138 * changed dynamically using method {@link #setKeepAliveTime(long, 139 * TimeUnit)}. Using a value of {@code Long.MAX_VALUE} {@link 140 * TimeUnit#NANOSECONDS} effectively disables idle threads from ever 141 * terminating prior to shut down. By default, the keep-alive policy 142 * applies only when there are more than corePoolSize threads, but 143 * method {@link #allowCoreThreadTimeOut(boolean)} can be used to 144 * apply this time-out policy to core threads as well, so long as the 145 * keepAliveTime value is non-zero. </dd> 146 * 147 * <dt>Queuing</dt> 148 * 149 * <dd style="font-family:'DejaVu Sans', Arial, Helvetica, sans-serif"> 150 * Any {@link BlockingQueue} may be used to transfer and hold 151 * submitted tasks. The use of this queue interacts with pool sizing: 152 * 153 * <ul> 154 * 155 * <li>If fewer than corePoolSize threads are running, the Executor 156 * always prefers adding a new thread 157 * rather than queuing. 158 * 159 * <li>If corePoolSize or more threads are running, the Executor 160 * always prefers queuing a request rather than adding a new 161 * thread. 162 * 163 * <li>If a request cannot be queued, a new thread is created unless 164 * this would exceed maximumPoolSize, in which case, the task will be 165 * rejected. 166 * 167 * </ul> 168 * 169 * There are three general strategies for queuing: 170 * <ol> 171 * 172 * <li><em> Direct handoffs.</em> A good default choice for a work 173 * queue is a {@link SynchronousQueue} that hands off tasks to threads 174 * without otherwise holding them. Here, an attempt to queue a task 175 * will fail if no threads are immediately available to run it, so a 176 * new thread will be constructed. This policy avoids lockups when 177 * handling sets of requests that might have internal dependencies. 178 * Direct handoffs generally require unbounded maximumPoolSizes to 179 * avoid rejection of new submitted tasks. This in turn admits the 180 * possibility of unbounded thread growth when commands continue to 181 * arrive on average faster than they can be processed. 182 * 183 * <li><em> Unbounded queues.</em> Using an unbounded queue (for 184 * example a {@link LinkedBlockingQueue} without a predefined 185 * capacity) will cause new tasks to wait in the queue when all 186 * corePoolSize threads are busy. Thus, no more than corePoolSize 187 * threads will ever be created. (And the value of the maximumPoolSize 188 * therefore doesn't have any effect.) This may be appropriate when 189 * each task is completely independent of others, so tasks cannot 190 * affect each others execution; for example, in a web page server. 191 * While this style of queuing can be useful in smoothing out 192 * transient bursts of requests, it admits the possibility of 193 * unbounded work queue growth when commands continue to arrive on 194 * average faster than they can be processed. 195 * 196 * <li><em>Bounded queues.</em> A bounded queue (for example, an 197 * {@link ArrayBlockingQueue}) helps prevent resource exhaustion when 198 * used with finite maximumPoolSizes, but can be more difficult to 199 * tune and control. Queue sizes and maximum pool sizes may be traded 200 * off for each other: Using large queues and small pools minimizes 201 * CPU usage, OS resources, and context-switching overhead, but can 202 * lead to artificially low throughput. If tasks frequently block (for 203 * example if they are I/O bound), a system may be able to schedule 204 * time for more threads than you otherwise allow. Use of small queues 205 * generally requires larger pool sizes, which keeps CPUs busier but 206 * may encounter unacceptable scheduling overhead, which also 207 * decreases throughput. 208 * 209 * </ol> 210 * 211 * </dd> 212 * 213 * <dt>Rejected tasks</dt> 214 * 215 * <dd style="font-family:'DejaVu Sans', Arial, Helvetica, sans-serif"> 216 * New tasks submitted in method {@link #execute(Runnable)} will be 217 * <em>rejected</em> when the Executor has been shut down, and also when 218 * the Executor uses finite bounds for both maximum threads and work queue 219 * capacity, and is saturated. In either case, the {@code execute} method 220 * invokes the {@link 221 * RejectedExecutionHandler#rejectedExecution(Runnable, ThreadPoolExecutor)} 222 * method of its {@link RejectedExecutionHandler}. Four predefined handler 223 * policies are provided: 224 * 225 * <ol> 226 * 227 * <li>In the default {@link ThreadPoolExecutor.AbortPolicy}, the 228 * handler throws a runtime {@link RejectedExecutionException} upon 229 * rejection. 230 * 231 * <li>In {@link ThreadPoolExecutor.CallerRunsPolicy}, the thread 232 * that invokes {@code execute} itself runs the task. This provides a 233 * simple feedback control mechanism that will slow down the rate that 234 * new tasks are submitted. 235 * 236 * <li>In {@link ThreadPoolExecutor.DiscardPolicy}, a task that 237 * cannot be executed is simply dropped. 238 * 239 * <li>In {@link ThreadPoolExecutor.DiscardOldestPolicy}, if the 240 * executor is not shut down, the task at the head of the work queue 241 * is dropped, and then execution is retried (which can fail again, 242 * causing this to be repeated.) 243 * 244 * </ol> 245 * 246 * It is possible to define and use other kinds of {@link 247 * RejectedExecutionHandler} classes. Doing so requires some care 248 * especially when policies are designed to work only under particular 249 * capacity or queuing policies. </dd> 250 * 251 * <dt>Hook methods</dt> 252 * 253 * <dd style="font-family:'DejaVu Sans', Arial, Helvetica, sans-serif"> 254 * This class provides {@code protected} overridable 255 * {@link #beforeExecute(Thread, Runnable)} and 256 * {@link #afterExecute(Runnable, Throwable)} methods that are called 257 * before and after execution of each task. These can be used to 258 * manipulate the execution environment; for example, reinitializing 259 * ThreadLocals, gathering statistics, or adding log entries. 260 * Additionally, method {@link #terminated} can be overridden to perform 261 * any special processing that needs to be done once the Executor has 262 * fully terminated. 263 * 264 * <p>If hook, callback, or BlockingQueue methods throw exceptions, 265 * internal worker threads may in turn fail, abruptly terminate, and 266 * possibly be replaced.</dd> 267 * 268 * <dt>Queue maintenance</dt> 269 * 270 * <dd style="font-family:'DejaVu Sans', Arial, Helvetica, sans-serif"> 271 * Method {@link #getQueue()} allows access to the work queue 272 * for purposes of monitoring and debugging. Use of this method for 273 * any other purpose is strongly discouraged. Two supplied methods, 274 * {@link #remove(Runnable)} and {@link #purge} are available to 275 * assist in storage reclamation when large numbers of queued tasks 276 * become cancelled.</dd> 277 * 278 * <dt>Finalization</dt> 279 * 280 * <dd style="font-family:'DejaVu Sans', Arial, Helvetica, sans-serif"> 281 * A pool that is no longer referenced in a program <em>AND</em> 282 * has no remaining threads will be {@code shutdown} automatically. If 283 * you would like to ensure that unreferenced pools are reclaimed even 284 * if users forget to call {@link #shutdown}, then you must arrange 285 * that unused threads eventually die, by setting appropriate 286 * keep-alive times, using a lower bound of zero core threads and/or 287 * setting {@link #allowCoreThreadTimeOut(boolean)}. </dd> 288 * 289 * </dl> 290 * 291 * <p><b>Extension example</b>. Most extensions of this class 292 * override one or more of the protected hook methods. For example, 293 * here is a subclass that adds a simple pause/resume feature: 294 * 295 * <pre> {@code 296 * class PausableThreadPoolExecutor extends ThreadPoolExecutor { 297 * private boolean isPaused; 298 * private ReentrantLock pauseLock = new ReentrantLock(); 299 * private Condition unpaused = pauseLock.newCondition(); 300 * 301 * public PausableThreadPoolExecutor(...) { super(...); } 302 * 303 * protected void beforeExecute(Thread t, Runnable r) { 304 * super.beforeExecute(t, r); 305 * pauseLock.lock(); 306 * try { 307 * while (isPaused) unpaused.await(); 308 * } catch (InterruptedException ie) { 309 * t.interrupt(); 310 * } finally { 311 * pauseLock.unlock(); 312 * } 313 * } 314 * 315 * public void pause() { 316 * pauseLock.lock(); 317 * try { 318 * isPaused = true; 319 * } finally { 320 * pauseLock.unlock(); 321 * } 322 * } 323 * 324 * public void resume() { 325 * pauseLock.lock(); 326 * try { 327 * isPaused = false; 328 * unpaused.signalAll(); 329 * } finally { 330 * pauseLock.unlock(); 331 * } 332 * } 333 * }}</pre> 334 * 335 * @since 1.5 336 * @author Doug Lea 337 */ 338 public class ThreadPoolExecutor extends AbstractExecutorService { 339 /** 340 * The main pool control state, ctl, is an atomic integer packing 341 * two conceptual fields 342 * workerCount, indicating the effective number of threads 343 * runState, indicating whether running, shutting down etc 344 * 345 * In order to pack them into one int, we limit workerCount to 346 * (2^29)-1 (about 500 million) threads rather than (2^31)-1 (2 347 * billion) otherwise representable. If this is ever an issue in 348 * the future, the variable can be changed to be an AtomicLong, 349 * and the shift/mask constants below adjusted. But until the need 350 * arises, this code is a bit faster and simpler using an int. 351 * 352 * The workerCount is the number of workers that have been 353 * permitted to start and not permitted to stop. The value may be 354 * transiently different from the actual number of live threads, 355 * for example when a ThreadFactory fails to create a thread when 356 * asked, and when exiting threads are still performing 357 * bookkeeping before terminating. The user-visible pool size is 358 * reported as the current size of the workers set. 359 * 360 * The runState provides the main lifecycle control, taking on values: 361 * 362 * RUNNING: Accept new tasks and process queued tasks 363 * SHUTDOWN: Don't accept new tasks, but process queued tasks 364 * STOP: Don't accept new tasks, don't process queued tasks, 365 * and interrupt in-progress tasks 366 * TIDYING: All tasks have terminated, workerCount is zero, 367 * the thread transitioning to state TIDYING 368 * will run the terminated() hook method 369 * TERMINATED: terminated() has completed 370 * 371 * The numerical order among these values matters, to allow 372 * ordered comparisons. The runState monotonically increases over 373 * time, but need not hit each state. The transitions are: 374 * 375 * RUNNING -> SHUTDOWN 376 * On invocation of shutdown(), perhaps implicitly in finalize() 377 * (RUNNING or SHUTDOWN) -> STOP 378 * On invocation of shutdownNow() 379 * SHUTDOWN -> TIDYING 380 * When both queue and pool are empty 381 * STOP -> TIDYING 382 * When pool is empty 383 * TIDYING -> TERMINATED 384 * When the terminated() hook method has completed 385 * 386 * Threads waiting in awaitTermination() will return when the 387 * state reaches TERMINATED. 388 * 389 * Detecting the transition from SHUTDOWN to TIDYING is less 390 * straightforward than you'd like because the queue may become 391 * empty after non-empty and vice versa during SHUTDOWN state, but 392 * we can only terminate if, after seeing that it is empty, we see 393 * that workerCount is 0 (which sometimes entails a recheck -- see 394 * below). 395 */ 396 private final AtomicInteger ctl = new AtomicInteger(ctlOf(RUNNING, 0)); 397 private static final int COUNT_BITS = Integer.SIZE - 3; 398 private static final int CAPACITY = (1 << COUNT_BITS) - 1; 399 400 // runState is stored in the high-order bits 401 private static final int RUNNING = -1 << COUNT_BITS; 402 private static final int SHUTDOWN = 0 << COUNT_BITS; 403 private static final int STOP = 1 << COUNT_BITS; 404 private static final int TIDYING = 2 << COUNT_BITS; 405 private static final int TERMINATED = 3 << COUNT_BITS; 406 407 // Packing and unpacking ctl runStateOf(int c)408 private static int runStateOf(int c) { return c & ~CAPACITY; } workerCountOf(int c)409 private static int workerCountOf(int c) { return c & CAPACITY; } ctlOf(int rs, int wc)410 private static int ctlOf(int rs, int wc) { return rs | wc; } 411 412 /* 413 * Bit field accessors that don't require unpacking ctl. 414 * These depend on the bit layout and on workerCount being never negative. 415 */ 416 runStateLessThan(int c, int s)417 private static boolean runStateLessThan(int c, int s) { 418 return c < s; 419 } 420 runStateAtLeast(int c, int s)421 private static boolean runStateAtLeast(int c, int s) { 422 return c >= s; 423 } 424 isRunning(int c)425 private static boolean isRunning(int c) { 426 return c < SHUTDOWN; 427 } 428 429 /** 430 * Attempts to CAS-increment the workerCount field of ctl. 431 */ compareAndIncrementWorkerCount(int expect)432 private boolean compareAndIncrementWorkerCount(int expect) { 433 return ctl.compareAndSet(expect, expect + 1); 434 } 435 436 /** 437 * Attempts to CAS-decrement the workerCount field of ctl. 438 */ compareAndDecrementWorkerCount(int expect)439 private boolean compareAndDecrementWorkerCount(int expect) { 440 return ctl.compareAndSet(expect, expect - 1); 441 } 442 443 /** 444 * Decrements the workerCount field of ctl. This is called only on 445 * abrupt termination of a thread (see processWorkerExit). Other 446 * decrements are performed within getTask. 447 */ decrementWorkerCount()448 private void decrementWorkerCount() { 449 do {} while (! compareAndDecrementWorkerCount(ctl.get())); 450 } 451 452 /** 453 * The queue used for holding tasks and handing off to worker 454 * threads. We do not require that workQueue.poll() returning 455 * null necessarily means that workQueue.isEmpty(), so rely 456 * solely on isEmpty to see if the queue is empty (which we must 457 * do for example when deciding whether to transition from 458 * SHUTDOWN to TIDYING). This accommodates special-purpose 459 * queues such as DelayQueues for which poll() is allowed to 460 * return null even if it may later return non-null when delays 461 * expire. 462 */ 463 private final BlockingQueue<Runnable> workQueue; 464 465 /** 466 * Lock held on access to workers set and related bookkeeping. 467 * While we could use a concurrent set of some sort, it turns out 468 * to be generally preferable to use a lock. Among the reasons is 469 * that this serializes interruptIdleWorkers, which avoids 470 * unnecessary interrupt storms, especially during shutdown. 471 * Otherwise exiting threads would concurrently interrupt those 472 * that have not yet interrupted. It also simplifies some of the 473 * associated statistics bookkeeping of largestPoolSize etc. We 474 * also hold mainLock on shutdown and shutdownNow, for the sake of 475 * ensuring workers set is stable while separately checking 476 * permission to interrupt and actually interrupting. 477 */ 478 private final ReentrantLock mainLock = new ReentrantLock(); 479 480 /** 481 * Set containing all worker threads in pool. Accessed only when 482 * holding mainLock. 483 */ 484 private final HashSet<Worker> workers = new HashSet<>(); 485 486 /** 487 * Wait condition to support awaitTermination. 488 */ 489 private final Condition termination = mainLock.newCondition(); 490 491 /** 492 * Tracks largest attained pool size. Accessed only under 493 * mainLock. 494 */ 495 private int largestPoolSize; 496 497 /** 498 * Counter for completed tasks. Updated only on termination of 499 * worker threads. Accessed only under mainLock. 500 */ 501 private long completedTaskCount; 502 503 /* 504 * All user control parameters are declared as volatiles so that 505 * ongoing actions are based on freshest values, but without need 506 * for locking, since no internal invariants depend on them 507 * changing synchronously with respect to other actions. 508 */ 509 510 /** 511 * Factory for new threads. All threads are created using this 512 * factory (via method addWorker). All callers must be prepared 513 * for addWorker to fail, which may reflect a system or user's 514 * policy limiting the number of threads. Even though it is not 515 * treated as an error, failure to create threads may result in 516 * new tasks being rejected or existing ones remaining stuck in 517 * the queue. 518 * 519 * We go further and preserve pool invariants even in the face of 520 * errors such as OutOfMemoryError, that might be thrown while 521 * trying to create threads. Such errors are rather common due to 522 * the need to allocate a native stack in Thread.start, and users 523 * will want to perform clean pool shutdown to clean up. There 524 * will likely be enough memory available for the cleanup code to 525 * complete without encountering yet another OutOfMemoryError. 526 */ 527 private volatile ThreadFactory threadFactory; 528 529 /** 530 * Handler called when saturated or shutdown in execute. 531 */ 532 private volatile RejectedExecutionHandler handler; 533 534 /** 535 * Timeout in nanoseconds for idle threads waiting for work. 536 * Threads use this timeout when there are more than corePoolSize 537 * present or if allowCoreThreadTimeOut. Otherwise they wait 538 * forever for new work. 539 */ 540 private volatile long keepAliveTime; 541 542 /** 543 * If false (default), core threads stay alive even when idle. 544 * If true, core threads use keepAliveTime to time out waiting 545 * for work. 546 */ 547 private volatile boolean allowCoreThreadTimeOut; 548 549 /** 550 * Core pool size is the minimum number of workers to keep alive 551 * (and not allow to time out etc) unless allowCoreThreadTimeOut 552 * is set, in which case the minimum is zero. 553 */ 554 private volatile int corePoolSize; 555 556 /** 557 * Maximum pool size. Note that the actual maximum is internally 558 * bounded by CAPACITY. 559 */ 560 private volatile int maximumPoolSize; 561 562 /** 563 * The default rejected execution handler. 564 */ 565 private static final RejectedExecutionHandler defaultHandler = 566 new AbortPolicy(); 567 568 /** 569 * Permission required for callers of shutdown and shutdownNow. 570 * We additionally require (see checkShutdownAccess) that callers 571 * have permission to actually interrupt threads in the worker set 572 * (as governed by Thread.interrupt, which relies on 573 * ThreadGroup.checkAccess, which in turn relies on 574 * SecurityManager.checkAccess). Shutdowns are attempted only if 575 * these checks pass. 576 * 577 * All actual invocations of Thread.interrupt (see 578 * interruptIdleWorkers and interruptWorkers) ignore 579 * SecurityExceptions, meaning that the attempted interrupts 580 * silently fail. In the case of shutdown, they should not fail 581 * unless the SecurityManager has inconsistent policies, sometimes 582 * allowing access to a thread and sometimes not. In such cases, 583 * failure to actually interrupt threads may disable or delay full 584 * termination. Other uses of interruptIdleWorkers are advisory, 585 * and failure to actually interrupt will merely delay response to 586 * configuration changes so is not handled exceptionally. 587 */ 588 private static final RuntimePermission shutdownPerm = 589 new RuntimePermission("modifyThread"); 590 591 /** 592 * Class Worker mainly maintains interrupt control state for 593 * threads running tasks, along with other minor bookkeeping. 594 * This class opportunistically extends AbstractQueuedSynchronizer 595 * to simplify acquiring and releasing a lock surrounding each 596 * task execution. This protects against interrupts that are 597 * intended to wake up a worker thread waiting for a task from 598 * instead interrupting a task being run. We implement a simple 599 * non-reentrant mutual exclusion lock rather than use 600 * ReentrantLock because we do not want worker tasks to be able to 601 * reacquire the lock when they invoke pool control methods like 602 * setCorePoolSize. Additionally, to suppress interrupts until 603 * the thread actually starts running tasks, we initialize lock 604 * state to a negative value, and clear it upon start (in 605 * runWorker). 606 */ 607 private final class Worker 608 extends AbstractQueuedSynchronizer 609 implements Runnable 610 { 611 /** 612 * This class will never be serialized, but we provide a 613 * serialVersionUID to suppress a javac warning. 614 */ 615 private static final long serialVersionUID = 6138294804551838833L; 616 617 /** Thread this worker is running in. Null if factory fails. */ 618 final Thread thread; 619 /** Initial task to run. Possibly null. */ 620 Runnable firstTask; 621 /** Per-thread task counter */ 622 volatile long completedTasks; 623 624 /** 625 * Creates with given first task and thread from ThreadFactory. 626 * @param firstTask the first task (null if none) 627 */ Worker(Runnable firstTask)628 Worker(Runnable firstTask) { 629 setState(-1); // inhibit interrupts until runWorker 630 this.firstTask = firstTask; 631 this.thread = getThreadFactory().newThread(this); 632 } 633 634 /** Delegates main run loop to outer runWorker. */ run()635 public void run() { 636 runWorker(this); 637 } 638 639 // Lock methods 640 // 641 // The value 0 represents the unlocked state. 642 // The value 1 represents the locked state. 643 isHeldExclusively()644 protected boolean isHeldExclusively() { 645 return getState() != 0; 646 } 647 tryAcquire(int unused)648 protected boolean tryAcquire(int unused) { 649 if (compareAndSetState(0, 1)) { 650 setExclusiveOwnerThread(Thread.currentThread()); 651 return true; 652 } 653 return false; 654 } 655 tryRelease(int unused)656 protected boolean tryRelease(int unused) { 657 setExclusiveOwnerThread(null); 658 setState(0); 659 return true; 660 } 661 lock()662 public void lock() { acquire(1); } tryLock()663 public boolean tryLock() { return tryAcquire(1); } unlock()664 public void unlock() { release(1); } isLocked()665 public boolean isLocked() { return isHeldExclusively(); } 666 interruptIfStarted()667 void interruptIfStarted() { 668 Thread t; 669 if (getState() >= 0 && (t = thread) != null && !t.isInterrupted()) { 670 try { 671 t.interrupt(); 672 } catch (SecurityException ignore) { 673 } 674 } 675 } 676 } 677 678 /* 679 * Methods for setting control state 680 */ 681 682 /** 683 * Transitions runState to given target, or leaves it alone if 684 * already at least the given target. 685 * 686 * @param targetState the desired state, either SHUTDOWN or STOP 687 * (but not TIDYING or TERMINATED -- use tryTerminate for that) 688 */ advanceRunState(int targetState)689 private void advanceRunState(int targetState) { 690 // assert targetState == SHUTDOWN || targetState == STOP; 691 for (;;) { 692 int c = ctl.get(); 693 if (runStateAtLeast(c, targetState) || 694 ctl.compareAndSet(c, ctlOf(targetState, workerCountOf(c)))) 695 break; 696 } 697 } 698 699 /** 700 * Transitions to TERMINATED state if either (SHUTDOWN and pool 701 * and queue empty) or (STOP and pool empty). If otherwise 702 * eligible to terminate but workerCount is nonzero, interrupts an 703 * idle worker to ensure that shutdown signals propagate. This 704 * method must be called following any action that might make 705 * termination possible -- reducing worker count or removing tasks 706 * from the queue during shutdown. The method is non-private to 707 * allow access from ScheduledThreadPoolExecutor. 708 */ tryTerminate()709 final void tryTerminate() { 710 for (;;) { 711 int c = ctl.get(); 712 if (isRunning(c) || 713 runStateAtLeast(c, TIDYING) || 714 (runStateOf(c) == SHUTDOWN && ! workQueue.isEmpty())) 715 return; 716 if (workerCountOf(c) != 0) { // Eligible to terminate 717 interruptIdleWorkers(ONLY_ONE); 718 return; 719 } 720 721 final ReentrantLock mainLock = this.mainLock; 722 mainLock.lock(); 723 try { 724 if (ctl.compareAndSet(c, ctlOf(TIDYING, 0))) { 725 try { 726 terminated(); 727 } finally { 728 ctl.set(ctlOf(TERMINATED, 0)); 729 termination.signalAll(); 730 } 731 return; 732 } 733 } finally { 734 mainLock.unlock(); 735 } 736 // else retry on failed CAS 737 } 738 } 739 740 /* 741 * Methods for controlling interrupts to worker threads. 742 */ 743 744 /** 745 * If there is a security manager, makes sure caller has 746 * permission to shut down threads in general (see shutdownPerm). 747 * If this passes, additionally makes sure the caller is allowed 748 * to interrupt each worker thread. This might not be true even if 749 * first check passed, if the SecurityManager treats some threads 750 * specially. 751 */ checkShutdownAccess()752 private void checkShutdownAccess() { 753 SecurityManager security = System.getSecurityManager(); 754 if (security != null) { 755 security.checkPermission(shutdownPerm); 756 final ReentrantLock mainLock = this.mainLock; 757 mainLock.lock(); 758 try { 759 for (Worker w : workers) 760 security.checkAccess(w.thread); 761 } finally { 762 mainLock.unlock(); 763 } 764 } 765 } 766 767 /** 768 * Interrupts all threads, even if active. Ignores SecurityExceptions 769 * (in which case some threads may remain uninterrupted). 770 */ interruptWorkers()771 private void interruptWorkers() { 772 final ReentrantLock mainLock = this.mainLock; 773 mainLock.lock(); 774 try { 775 for (Worker w : workers) 776 w.interruptIfStarted(); 777 } finally { 778 mainLock.unlock(); 779 } 780 } 781 782 /** 783 * Interrupts threads that might be waiting for tasks (as 784 * indicated by not being locked) so they can check for 785 * termination or configuration changes. Ignores 786 * SecurityExceptions (in which case some threads may remain 787 * uninterrupted). 788 * 789 * @param onlyOne If true, interrupt at most one worker. This is 790 * called only from tryTerminate when termination is otherwise 791 * enabled but there are still other workers. In this case, at 792 * most one waiting worker is interrupted to propagate shutdown 793 * signals in case all threads are currently waiting. 794 * Interrupting any arbitrary thread ensures that newly arriving 795 * workers since shutdown began will also eventually exit. 796 * To guarantee eventual termination, it suffices to always 797 * interrupt only one idle worker, but shutdown() interrupts all 798 * idle workers so that redundant workers exit promptly, not 799 * waiting for a straggler task to finish. 800 */ interruptIdleWorkers(boolean onlyOne)801 private void interruptIdleWorkers(boolean onlyOne) { 802 final ReentrantLock mainLock = this.mainLock; 803 mainLock.lock(); 804 try { 805 for (Worker w : workers) { 806 Thread t = w.thread; 807 if (!t.isInterrupted() && w.tryLock()) { 808 try { 809 t.interrupt(); 810 } catch (SecurityException ignore) { 811 } finally { 812 w.unlock(); 813 } 814 } 815 if (onlyOne) 816 break; 817 } 818 } finally { 819 mainLock.unlock(); 820 } 821 } 822 823 /** 824 * Common form of interruptIdleWorkers, to avoid having to 825 * remember what the boolean argument means. 826 */ interruptIdleWorkers()827 private void interruptIdleWorkers() { 828 interruptIdleWorkers(false); 829 } 830 831 private static final boolean ONLY_ONE = true; 832 833 /* 834 * Misc utilities, most of which are also exported to 835 * ScheduledThreadPoolExecutor 836 */ 837 838 /** 839 * Invokes the rejected execution handler for the given command. 840 * Package-protected for use by ScheduledThreadPoolExecutor. 841 */ reject(Runnable command)842 final void reject(Runnable command) { 843 handler.rejectedExecution(command, this); 844 } 845 846 /** 847 * Performs any further cleanup following run state transition on 848 * invocation of shutdown. A no-op here, but used by 849 * ScheduledThreadPoolExecutor to cancel delayed tasks. 850 */ onShutdown()851 void onShutdown() { 852 } 853 854 /** 855 * State check needed by ScheduledThreadPoolExecutor to 856 * enable running tasks during shutdown. 857 * 858 * @param shutdownOK true if should return true if SHUTDOWN 859 */ isRunningOrShutdown(boolean shutdownOK)860 final boolean isRunningOrShutdown(boolean shutdownOK) { 861 int rs = runStateOf(ctl.get()); 862 return rs == RUNNING || (rs == SHUTDOWN && shutdownOK); 863 } 864 865 /** 866 * Drains the task queue into a new list, normally using 867 * drainTo. But if the queue is a DelayQueue or any other kind of 868 * queue for which poll or drainTo may fail to remove some 869 * elements, it deletes them one by one. 870 */ drainQueue()871 private List<Runnable> drainQueue() { 872 BlockingQueue<Runnable> q = workQueue; 873 ArrayList<Runnable> taskList = new ArrayList<>(); 874 q.drainTo(taskList); 875 if (!q.isEmpty()) { 876 for (Runnable r : q.toArray(new Runnable[0])) { 877 if (q.remove(r)) 878 taskList.add(r); 879 } 880 } 881 return taskList; 882 } 883 884 /* 885 * Methods for creating, running and cleaning up after workers 886 */ 887 888 /** 889 * Checks if a new worker can be added with respect to current 890 * pool state and the given bound (either core or maximum). If so, 891 * the worker count is adjusted accordingly, and, if possible, a 892 * new worker is created and started, running firstTask as its 893 * first task. This method returns false if the pool is stopped or 894 * eligible to shut down. It also returns false if the thread 895 * factory fails to create a thread when asked. If the thread 896 * creation fails, either due to the thread factory returning 897 * null, or due to an exception (typically OutOfMemoryError in 898 * Thread.start()), we roll back cleanly. 899 * 900 * @param firstTask the task the new thread should run first (or 901 * null if none). Workers are created with an initial first task 902 * (in method execute()) to bypass queuing when there are fewer 903 * than corePoolSize threads (in which case we always start one), 904 * or when the queue is full (in which case we must bypass queue). 905 * Initially idle threads are usually created via 906 * prestartCoreThread or to replace other dying workers. 907 * 908 * @param core if true use corePoolSize as bound, else 909 * maximumPoolSize. (A boolean indicator is used here rather than a 910 * value to ensure reads of fresh values after checking other pool 911 * state). 912 * @return true if successful 913 */ addWorker(Runnable firstTask, boolean core)914 private boolean addWorker(Runnable firstTask, boolean core) { 915 retry: 916 for (;;) { 917 int c = ctl.get(); 918 int rs = runStateOf(c); 919 920 // Check if queue empty only if necessary. 921 if (rs >= SHUTDOWN && 922 ! (rs == SHUTDOWN && 923 firstTask == null && 924 ! workQueue.isEmpty())) 925 return false; 926 927 for (;;) { 928 int wc = workerCountOf(c); 929 if (wc >= CAPACITY || 930 wc >= (core ? corePoolSize : maximumPoolSize)) 931 return false; 932 if (compareAndIncrementWorkerCount(c)) 933 break retry; 934 c = ctl.get(); // Re-read ctl 935 if (runStateOf(c) != rs) 936 continue retry; 937 // else CAS failed due to workerCount change; retry inner loop 938 } 939 } 940 941 boolean workerStarted = false; 942 boolean workerAdded = false; 943 Worker w = null; 944 try { 945 w = new Worker(firstTask); 946 final Thread t = w.thread; 947 if (t != null) { 948 final ReentrantLock mainLock = this.mainLock; 949 mainLock.lock(); 950 try { 951 // Recheck while holding lock. 952 // Back out on ThreadFactory failure or if 953 // shut down before lock acquired. 954 int rs = runStateOf(ctl.get()); 955 956 if (rs < SHUTDOWN || 957 (rs == SHUTDOWN && firstTask == null)) { 958 if (t.isAlive()) // precheck that t is startable 959 throw new IllegalThreadStateException(); 960 workers.add(w); 961 int s = workers.size(); 962 if (s > largestPoolSize) 963 largestPoolSize = s; 964 workerAdded = true; 965 } 966 } finally { 967 mainLock.unlock(); 968 } 969 if (workerAdded) { 970 t.start(); 971 workerStarted = true; 972 } 973 } 974 } finally { 975 if (! workerStarted) 976 addWorkerFailed(w); 977 } 978 return workerStarted; 979 } 980 981 /** 982 * Rolls back the worker thread creation. 983 * - removes worker from workers, if present 984 * - decrements worker count 985 * - rechecks for termination, in case the existence of this 986 * worker was holding up termination 987 */ addWorkerFailed(Worker w)988 private void addWorkerFailed(Worker w) { 989 final ReentrantLock mainLock = this.mainLock; 990 mainLock.lock(); 991 try { 992 if (w != null) 993 workers.remove(w); 994 decrementWorkerCount(); 995 tryTerminate(); 996 } finally { 997 mainLock.unlock(); 998 } 999 } 1000 1001 /** 1002 * Performs cleanup and bookkeeping for a dying worker. Called 1003 * only from worker threads. Unless completedAbruptly is set, 1004 * assumes that workerCount has already been adjusted to account 1005 * for exit. This method removes thread from worker set, and 1006 * possibly terminates the pool or replaces the worker if either 1007 * it exited due to user task exception or if fewer than 1008 * corePoolSize workers are running or queue is non-empty but 1009 * there are no workers. 1010 * 1011 * @param w the worker 1012 * @param completedAbruptly if the worker died due to user exception 1013 */ processWorkerExit(Worker w, boolean completedAbruptly)1014 private void processWorkerExit(Worker w, boolean completedAbruptly) { 1015 if (completedAbruptly) // If abrupt, then workerCount wasn't adjusted 1016 decrementWorkerCount(); 1017 1018 final ReentrantLock mainLock = this.mainLock; 1019 mainLock.lock(); 1020 try { 1021 completedTaskCount += w.completedTasks; 1022 workers.remove(w); 1023 } finally { 1024 mainLock.unlock(); 1025 } 1026 1027 tryTerminate(); 1028 1029 int c = ctl.get(); 1030 if (runStateLessThan(c, STOP)) { 1031 if (!completedAbruptly) { 1032 int min = allowCoreThreadTimeOut ? 0 : corePoolSize; 1033 if (min == 0 && ! workQueue.isEmpty()) 1034 min = 1; 1035 if (workerCountOf(c) >= min) 1036 return; // replacement not needed 1037 } 1038 addWorker(null, false); 1039 } 1040 } 1041 1042 /** 1043 * Performs blocking or timed wait for a task, depending on 1044 * current configuration settings, or returns null if this worker 1045 * must exit because of any of: 1046 * 1. There are more than maximumPoolSize workers (due to 1047 * a call to setMaximumPoolSize). 1048 * 2. The pool is stopped. 1049 * 3. The pool is shutdown and the queue is empty. 1050 * 4. This worker timed out waiting for a task, and timed-out 1051 * workers are subject to termination (that is, 1052 * {@code allowCoreThreadTimeOut || workerCount > corePoolSize}) 1053 * both before and after the timed wait, and if the queue is 1054 * non-empty, this worker is not the last thread in the pool. 1055 * 1056 * @return task, or null if the worker must exit, in which case 1057 * workerCount is decremented 1058 */ getTask()1059 private Runnable getTask() { 1060 boolean timedOut = false; // Did the last poll() time out? 1061 1062 for (;;) { 1063 int c = ctl.get(); 1064 int rs = runStateOf(c); 1065 1066 // Check if queue empty only if necessary. 1067 if (rs >= SHUTDOWN && (rs >= STOP || workQueue.isEmpty())) { 1068 decrementWorkerCount(); 1069 return null; 1070 } 1071 1072 int wc = workerCountOf(c); 1073 1074 // Are workers subject to culling? 1075 boolean timed = allowCoreThreadTimeOut || wc > corePoolSize; 1076 1077 if ((wc > maximumPoolSize || (timed && timedOut)) 1078 && (wc > 1 || workQueue.isEmpty())) { 1079 if (compareAndDecrementWorkerCount(c)) 1080 return null; 1081 continue; 1082 } 1083 1084 try { 1085 Runnable r = timed ? 1086 workQueue.poll(keepAliveTime, TimeUnit.NANOSECONDS) : 1087 workQueue.take(); 1088 if (r != null) 1089 return r; 1090 timedOut = true; 1091 } catch (InterruptedException retry) { 1092 timedOut = false; 1093 } 1094 } 1095 } 1096 1097 /** 1098 * Main worker run loop. Repeatedly gets tasks from queue and 1099 * executes them, while coping with a number of issues: 1100 * 1101 * 1. We may start out with an initial task, in which case we 1102 * don't need to get the first one. Otherwise, as long as pool is 1103 * running, we get tasks from getTask. If it returns null then the 1104 * worker exits due to changed pool state or configuration 1105 * parameters. Other exits result from exception throws in 1106 * external code, in which case completedAbruptly holds, which 1107 * usually leads processWorkerExit to replace this thread. 1108 * 1109 * 2. Before running any task, the lock is acquired to prevent 1110 * other pool interrupts while the task is executing, and then we 1111 * ensure that unless pool is stopping, this thread does not have 1112 * its interrupt set. 1113 * 1114 * 3. Each task run is preceded by a call to beforeExecute, which 1115 * might throw an exception, in which case we cause thread to die 1116 * (breaking loop with completedAbruptly true) without processing 1117 * the task. 1118 * 1119 * 4. Assuming beforeExecute completes normally, we run the task, 1120 * gathering any of its thrown exceptions to send to afterExecute. 1121 * We separately handle RuntimeException, Error (both of which the 1122 * specs guarantee that we trap) and arbitrary Throwables. 1123 * Because we cannot rethrow Throwables within Runnable.run, we 1124 * wrap them within Errors on the way out (to the thread's 1125 * UncaughtExceptionHandler). Any thrown exception also 1126 * conservatively causes thread to die. 1127 * 1128 * 5. After task.run completes, we call afterExecute, which may 1129 * also throw an exception, which will also cause thread to 1130 * die. According to JLS Sec 14.20, this exception is the one that 1131 * will be in effect even if task.run throws. 1132 * 1133 * The net effect of the exception mechanics is that afterExecute 1134 * and the thread's UncaughtExceptionHandler have as accurate 1135 * information as we can provide about any problems encountered by 1136 * user code. 1137 * 1138 * @param w the worker 1139 */ runWorker(Worker w)1140 final void runWorker(Worker w) { 1141 Thread wt = Thread.currentThread(); 1142 Runnable task = w.firstTask; 1143 w.firstTask = null; 1144 w.unlock(); // allow interrupts 1145 boolean completedAbruptly = true; 1146 try { 1147 while (task != null || (task = getTask()) != null) { 1148 w.lock(); 1149 // If pool is stopping, ensure thread is interrupted; 1150 // if not, ensure thread is not interrupted. This 1151 // requires a recheck in second case to deal with 1152 // shutdownNow race while clearing interrupt 1153 if ((runStateAtLeast(ctl.get(), STOP) || 1154 (Thread.interrupted() && 1155 runStateAtLeast(ctl.get(), STOP))) && 1156 !wt.isInterrupted()) 1157 wt.interrupt(); 1158 try { 1159 beforeExecute(wt, task); 1160 Throwable thrown = null; 1161 try { 1162 task.run(); 1163 } catch (RuntimeException x) { 1164 thrown = x; throw x; 1165 } catch (Error x) { 1166 thrown = x; throw x; 1167 } catch (Throwable x) { 1168 thrown = x; throw new Error(x); 1169 } finally { 1170 afterExecute(task, thrown); 1171 } 1172 } finally { 1173 task = null; 1174 w.completedTasks++; 1175 w.unlock(); 1176 } 1177 } 1178 completedAbruptly = false; 1179 } finally { 1180 processWorkerExit(w, completedAbruptly); 1181 } 1182 } 1183 1184 // Public constructors and methods 1185 1186 /** 1187 * Creates a new {@code ThreadPoolExecutor} with the given initial 1188 * parameters and default thread factory and rejected execution handler. 1189 * It may be more convenient to use one of the {@link Executors} factory 1190 * methods instead of this general purpose constructor. 1191 * 1192 * @param corePoolSize the number of threads to keep in the pool, even 1193 * if they are idle, unless {@code allowCoreThreadTimeOut} is set 1194 * @param maximumPoolSize the maximum number of threads to allow in the 1195 * pool 1196 * @param keepAliveTime when the number of threads is greater than 1197 * the core, this is the maximum time that excess idle threads 1198 * will wait for new tasks before terminating. 1199 * @param unit the time unit for the {@code keepAliveTime} argument 1200 * @param workQueue the queue to use for holding tasks before they are 1201 * executed. This queue will hold only the {@code Runnable} 1202 * tasks submitted by the {@code execute} method. 1203 * @throws IllegalArgumentException if one of the following holds:<br> 1204 * {@code corePoolSize < 0}<br> 1205 * {@code keepAliveTime < 0}<br> 1206 * {@code maximumPoolSize <= 0}<br> 1207 * {@code maximumPoolSize < corePoolSize} 1208 * @throws NullPointerException if {@code workQueue} is null 1209 */ ThreadPoolExecutor(int corePoolSize, int maximumPoolSize, long keepAliveTime, TimeUnit unit, BlockingQueue<Runnable> workQueue)1210 public ThreadPoolExecutor(int corePoolSize, 1211 int maximumPoolSize, 1212 long keepAliveTime, 1213 TimeUnit unit, 1214 BlockingQueue<Runnable> workQueue) { 1215 this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue, 1216 Executors.defaultThreadFactory(), defaultHandler); 1217 } 1218 1219 /** 1220 * Creates a new {@code ThreadPoolExecutor} with the given initial 1221 * parameters and default rejected execution handler. 1222 * 1223 * @param corePoolSize the number of threads to keep in the pool, even 1224 * if they are idle, unless {@code allowCoreThreadTimeOut} is set 1225 * @param maximumPoolSize the maximum number of threads to allow in the 1226 * pool 1227 * @param keepAliveTime when the number of threads is greater than 1228 * the core, this is the maximum time that excess idle threads 1229 * will wait for new tasks before terminating. 1230 * @param unit the time unit for the {@code keepAliveTime} argument 1231 * @param workQueue the queue to use for holding tasks before they are 1232 * executed. This queue will hold only the {@code Runnable} 1233 * tasks submitted by the {@code execute} method. 1234 * @param threadFactory the factory to use when the executor 1235 * creates a new thread 1236 * @throws IllegalArgumentException if one of the following holds:<br> 1237 * {@code corePoolSize < 0}<br> 1238 * {@code keepAliveTime < 0}<br> 1239 * {@code maximumPoolSize <= 0}<br> 1240 * {@code maximumPoolSize < corePoolSize} 1241 * @throws NullPointerException if {@code workQueue} 1242 * or {@code threadFactory} is null 1243 */ ThreadPoolExecutor(int corePoolSize, int maximumPoolSize, long keepAliveTime, TimeUnit unit, BlockingQueue<Runnable> workQueue, ThreadFactory threadFactory)1244 public ThreadPoolExecutor(int corePoolSize, 1245 int maximumPoolSize, 1246 long keepAliveTime, 1247 TimeUnit unit, 1248 BlockingQueue<Runnable> workQueue, 1249 ThreadFactory threadFactory) { 1250 this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue, 1251 threadFactory, defaultHandler); 1252 } 1253 1254 /** 1255 * Creates a new {@code ThreadPoolExecutor} with the given initial 1256 * parameters and default thread factory. 1257 * 1258 * @param corePoolSize the number of threads to keep in the pool, even 1259 * if they are idle, unless {@code allowCoreThreadTimeOut} is set 1260 * @param maximumPoolSize the maximum number of threads to allow in the 1261 * pool 1262 * @param keepAliveTime when the number of threads is greater than 1263 * the core, this is the maximum time that excess idle threads 1264 * will wait for new tasks before terminating. 1265 * @param unit the time unit for the {@code keepAliveTime} argument 1266 * @param workQueue the queue to use for holding tasks before they are 1267 * executed. This queue will hold only the {@code Runnable} 1268 * tasks submitted by the {@code execute} method. 1269 * @param handler the handler to use when execution is blocked 1270 * because the thread bounds and queue capacities are reached 1271 * @throws IllegalArgumentException if one of the following holds:<br> 1272 * {@code corePoolSize < 0}<br> 1273 * {@code keepAliveTime < 0}<br> 1274 * {@code maximumPoolSize <= 0}<br> 1275 * {@code maximumPoolSize < corePoolSize} 1276 * @throws NullPointerException if {@code workQueue} 1277 * or {@code handler} is null 1278 */ ThreadPoolExecutor(int corePoolSize, int maximumPoolSize, long keepAliveTime, TimeUnit unit, BlockingQueue<Runnable> workQueue, RejectedExecutionHandler handler)1279 public ThreadPoolExecutor(int corePoolSize, 1280 int maximumPoolSize, 1281 long keepAliveTime, 1282 TimeUnit unit, 1283 BlockingQueue<Runnable> workQueue, 1284 RejectedExecutionHandler handler) { 1285 this(corePoolSize, maximumPoolSize, keepAliveTime, unit, workQueue, 1286 Executors.defaultThreadFactory(), handler); 1287 } 1288 1289 /** 1290 * Creates a new {@code ThreadPoolExecutor} with the given initial 1291 * parameters. 1292 * 1293 * @param corePoolSize the number of threads to keep in the pool, even 1294 * if they are idle, unless {@code allowCoreThreadTimeOut} is set 1295 * @param maximumPoolSize the maximum number of threads to allow in the 1296 * pool 1297 * @param keepAliveTime when the number of threads is greater than 1298 * the core, this is the maximum time that excess idle threads 1299 * will wait for new tasks before terminating. 1300 * @param unit the time unit for the {@code keepAliveTime} argument 1301 * @param workQueue the queue to use for holding tasks before they are 1302 * executed. This queue will hold only the {@code Runnable} 1303 * tasks submitted by the {@code execute} method. 1304 * @param threadFactory the factory to use when the executor 1305 * creates a new thread 1306 * @param handler the handler to use when execution is blocked 1307 * because the thread bounds and queue capacities are reached 1308 * @throws IllegalArgumentException if one of the following holds:<br> 1309 * {@code corePoolSize < 0}<br> 1310 * {@code keepAliveTime < 0}<br> 1311 * {@code maximumPoolSize <= 0}<br> 1312 * {@code maximumPoolSize < corePoolSize} 1313 * @throws NullPointerException if {@code workQueue} 1314 * or {@code threadFactory} or {@code handler} is null 1315 */ ThreadPoolExecutor(int corePoolSize, int maximumPoolSize, long keepAliveTime, TimeUnit unit, BlockingQueue<Runnable> workQueue, ThreadFactory threadFactory, RejectedExecutionHandler handler)1316 public ThreadPoolExecutor(int corePoolSize, 1317 int maximumPoolSize, 1318 long keepAliveTime, 1319 TimeUnit unit, 1320 BlockingQueue<Runnable> workQueue, 1321 ThreadFactory threadFactory, 1322 RejectedExecutionHandler handler) { 1323 if (corePoolSize < 0 || 1324 maximumPoolSize <= 0 || 1325 maximumPoolSize < corePoolSize || 1326 keepAliveTime < 0) 1327 throw new IllegalArgumentException(); 1328 if (workQueue == null || threadFactory == null || handler == null) 1329 throw new NullPointerException(); 1330 this.corePoolSize = corePoolSize; 1331 this.maximumPoolSize = maximumPoolSize; 1332 this.workQueue = workQueue; 1333 this.keepAliveTime = unit.toNanos(keepAliveTime); 1334 this.threadFactory = threadFactory; 1335 this.handler = handler; 1336 } 1337 1338 /** 1339 * Executes the given task sometime in the future. The task 1340 * may execute in a new thread or in an existing pooled thread. 1341 * 1342 * If the task cannot be submitted for execution, either because this 1343 * executor has been shutdown or because its capacity has been reached, 1344 * the task is handled by the current {@code RejectedExecutionHandler}. 1345 * 1346 * @param command the task to execute 1347 * @throws RejectedExecutionException at discretion of 1348 * {@code RejectedExecutionHandler}, if the task 1349 * cannot be accepted for execution 1350 * @throws NullPointerException if {@code command} is null 1351 */ execute(Runnable command)1352 public void execute(Runnable command) { 1353 if (command == null) 1354 throw new NullPointerException(); 1355 /* 1356 * Proceed in 3 steps: 1357 * 1358 * 1. If fewer than corePoolSize threads are running, try to 1359 * start a new thread with the given command as its first 1360 * task. The call to addWorker atomically checks runState and 1361 * workerCount, and so prevents false alarms that would add 1362 * threads when it shouldn't, by returning false. 1363 * 1364 * 2. If a task can be successfully queued, then we still need 1365 * to double-check whether we should have added a thread 1366 * (because existing ones died since last checking) or that 1367 * the pool shut down since entry into this method. So we 1368 * recheck state and if necessary roll back the enqueuing if 1369 * stopped, or start a new thread if there are none. 1370 * 1371 * 3. If we cannot queue task, then we try to add a new 1372 * thread. If it fails, we know we are shut down or saturated 1373 * and so reject the task. 1374 */ 1375 int c = ctl.get(); 1376 if (workerCountOf(c) < corePoolSize) { 1377 if (addWorker(command, true)) 1378 return; 1379 c = ctl.get(); 1380 } 1381 if (isRunning(c) && workQueue.offer(command)) { 1382 int recheck = ctl.get(); 1383 if (! isRunning(recheck) && remove(command)) 1384 reject(command); 1385 else if (workerCountOf(recheck) == 0) 1386 addWorker(null, false); 1387 } 1388 else if (!addWorker(command, false)) 1389 reject(command); 1390 } 1391 1392 /** 1393 * Initiates an orderly shutdown in which previously submitted 1394 * tasks are executed, but no new tasks will be accepted. 1395 * Invocation has no additional effect if already shut down. 1396 * 1397 * <p>This method does not wait for previously submitted tasks to 1398 * complete execution. Use {@link #awaitTermination awaitTermination} 1399 * to do that. 1400 */ 1401 // android-note: Removed @throws SecurityException shutdown()1402 public void shutdown() { 1403 final ReentrantLock mainLock = this.mainLock; 1404 mainLock.lock(); 1405 try { 1406 checkShutdownAccess(); 1407 advanceRunState(SHUTDOWN); 1408 interruptIdleWorkers(); 1409 onShutdown(); // hook for ScheduledThreadPoolExecutor 1410 } finally { 1411 mainLock.unlock(); 1412 } 1413 tryTerminate(); 1414 } 1415 1416 /** 1417 * Attempts to stop all actively executing tasks, halts the 1418 * processing of waiting tasks, and returns a list of the tasks 1419 * that were awaiting execution. These tasks are drained (removed) 1420 * from the task queue upon return from this method. 1421 * 1422 * <p>This method does not wait for actively executing tasks to 1423 * terminate. Use {@link #awaitTermination awaitTermination} to 1424 * do that. 1425 * 1426 * <p>There are no guarantees beyond best-effort attempts to stop 1427 * processing actively executing tasks. This implementation 1428 * interrupts tasks via {@link Thread#interrupt}; any task that 1429 * fails to respond to interrupts may never terminate. 1430 */ 1431 // android-note: Removed @throws SecurityException shutdownNow()1432 public List<Runnable> shutdownNow() { 1433 List<Runnable> tasks; 1434 final ReentrantLock mainLock = this.mainLock; 1435 mainLock.lock(); 1436 try { 1437 checkShutdownAccess(); 1438 advanceRunState(STOP); 1439 interruptWorkers(); 1440 tasks = drainQueue(); 1441 } finally { 1442 mainLock.unlock(); 1443 } 1444 tryTerminate(); 1445 return tasks; 1446 } 1447 isShutdown()1448 public boolean isShutdown() { 1449 return ! isRunning(ctl.get()); 1450 } 1451 1452 /** 1453 * Returns true if this executor is in the process of terminating 1454 * after {@link #shutdown} or {@link #shutdownNow} but has not 1455 * completely terminated. This method may be useful for 1456 * debugging. A return of {@code true} reported a sufficient 1457 * period after shutdown may indicate that submitted tasks have 1458 * ignored or suppressed interruption, causing this executor not 1459 * to properly terminate. 1460 * 1461 * @return {@code true} if terminating but not yet terminated 1462 */ isTerminating()1463 public boolean isTerminating() { 1464 int c = ctl.get(); 1465 return ! isRunning(c) && runStateLessThan(c, TERMINATED); 1466 } 1467 isTerminated()1468 public boolean isTerminated() { 1469 return runStateAtLeast(ctl.get(), TERMINATED); 1470 } 1471 awaitTermination(long timeout, TimeUnit unit)1472 public boolean awaitTermination(long timeout, TimeUnit unit) 1473 throws InterruptedException { 1474 long nanos = unit.toNanos(timeout); 1475 final ReentrantLock mainLock = this.mainLock; 1476 mainLock.lock(); 1477 try { 1478 while (!runStateAtLeast(ctl.get(), TERMINATED)) { 1479 if (nanos <= 0L) 1480 return false; 1481 nanos = termination.awaitNanos(nanos); 1482 } 1483 return true; 1484 } finally { 1485 mainLock.unlock(); 1486 } 1487 } 1488 1489 /** 1490 * Invokes {@code shutdown} when this executor is no longer 1491 * referenced and it has no threads. 1492 */ finalize()1493 protected void finalize() { 1494 shutdown(); 1495 } 1496 1497 /** 1498 * Sets the thread factory used to create new threads. 1499 * 1500 * @param threadFactory the new thread factory 1501 * @throws NullPointerException if threadFactory is null 1502 * @see #getThreadFactory 1503 */ setThreadFactory(ThreadFactory threadFactory)1504 public void setThreadFactory(ThreadFactory threadFactory) { 1505 if (threadFactory == null) 1506 throw new NullPointerException(); 1507 this.threadFactory = threadFactory; 1508 } 1509 1510 /** 1511 * Returns the thread factory used to create new threads. 1512 * 1513 * @return the current thread factory 1514 * @see #setThreadFactory(ThreadFactory) 1515 */ getThreadFactory()1516 public ThreadFactory getThreadFactory() { 1517 return threadFactory; 1518 } 1519 1520 /** 1521 * Sets a new handler for unexecutable tasks. 1522 * 1523 * @param handler the new handler 1524 * @throws NullPointerException if handler is null 1525 * @see #getRejectedExecutionHandler 1526 */ setRejectedExecutionHandler(RejectedExecutionHandler handler)1527 public void setRejectedExecutionHandler(RejectedExecutionHandler handler) { 1528 if (handler == null) 1529 throw new NullPointerException(); 1530 this.handler = handler; 1531 } 1532 1533 /** 1534 * Returns the current handler for unexecutable tasks. 1535 * 1536 * @return the current handler 1537 * @see #setRejectedExecutionHandler(RejectedExecutionHandler) 1538 */ getRejectedExecutionHandler()1539 public RejectedExecutionHandler getRejectedExecutionHandler() { 1540 return handler; 1541 } 1542 1543 /** 1544 * Sets the core number of threads. This overrides any value set 1545 * in the constructor. If the new value is smaller than the 1546 * current value, excess existing threads will be terminated when 1547 * they next become idle. If larger, new threads will, if needed, 1548 * be started to execute any queued tasks. 1549 * 1550 * @param corePoolSize the new core size 1551 * @throws IllegalArgumentException if {@code corePoolSize < 0} 1552 * @see #getCorePoolSize 1553 */ 1554 // Android-changed: Reverted code that threw an IAE when 1555 // {@code corePoolSize} is greater than the {@linkplain #getMaximumPoolSize() 1556 // maximum pool size}. This is due to defective code in a commonly used third 1557 // party library that does something like : 1558 // 1559 // exec.setCorePoolSize(N); 1560 // exec.setMaxPoolSize(N); setCorePoolSize(int corePoolSize)1561 public void setCorePoolSize(int corePoolSize) { 1562 if (corePoolSize < 0) 1563 throw new IllegalArgumentException(); 1564 int delta = corePoolSize - this.corePoolSize; 1565 this.corePoolSize = corePoolSize; 1566 if (workerCountOf(ctl.get()) > corePoolSize) 1567 interruptIdleWorkers(); 1568 else if (delta > 0) { 1569 // We don't really know how many new threads are "needed". 1570 // As a heuristic, prestart enough new workers (up to new 1571 // core size) to handle the current number of tasks in 1572 // queue, but stop if queue becomes empty while doing so. 1573 int k = Math.min(delta, workQueue.size()); 1574 while (k-- > 0 && addWorker(null, true)) { 1575 if (workQueue.isEmpty()) 1576 break; 1577 } 1578 } 1579 } 1580 1581 /** 1582 * Returns the core number of threads. 1583 * 1584 * @return the core number of threads 1585 * @see #setCorePoolSize 1586 */ getCorePoolSize()1587 public int getCorePoolSize() { 1588 return corePoolSize; 1589 } 1590 1591 /** 1592 * Starts a core thread, causing it to idly wait for work. This 1593 * overrides the default policy of starting core threads only when 1594 * new tasks are executed. This method will return {@code false} 1595 * if all core threads have already been started. 1596 * 1597 * @return {@code true} if a thread was started 1598 */ prestartCoreThread()1599 public boolean prestartCoreThread() { 1600 return workerCountOf(ctl.get()) < corePoolSize && 1601 addWorker(null, true); 1602 } 1603 1604 /** 1605 * Same as prestartCoreThread except arranges that at least one 1606 * thread is started even if corePoolSize is 0. 1607 */ ensurePrestart()1608 void ensurePrestart() { 1609 int wc = workerCountOf(ctl.get()); 1610 if (wc < corePoolSize) 1611 addWorker(null, true); 1612 else if (wc == 0) 1613 addWorker(null, false); 1614 } 1615 1616 /** 1617 * Starts all core threads, causing them to idly wait for work. This 1618 * overrides the default policy of starting core threads only when 1619 * new tasks are executed. 1620 * 1621 * @return the number of threads started 1622 */ prestartAllCoreThreads()1623 public int prestartAllCoreThreads() { 1624 int n = 0; 1625 while (addWorker(null, true)) 1626 ++n; 1627 return n; 1628 } 1629 1630 /** 1631 * Returns true if this pool allows core threads to time out and 1632 * terminate if no tasks arrive within the keepAlive time, being 1633 * replaced if needed when new tasks arrive. When true, the same 1634 * keep-alive policy applying to non-core threads applies also to 1635 * core threads. When false (the default), core threads are never 1636 * terminated due to lack of incoming tasks. 1637 * 1638 * @return {@code true} if core threads are allowed to time out, 1639 * else {@code false} 1640 * 1641 * @since 1.6 1642 */ allowsCoreThreadTimeOut()1643 public boolean allowsCoreThreadTimeOut() { 1644 return allowCoreThreadTimeOut; 1645 } 1646 1647 /** 1648 * Sets the policy governing whether core threads may time out and 1649 * terminate if no tasks arrive within the keep-alive time, being 1650 * replaced if needed when new tasks arrive. When false, core 1651 * threads are never terminated due to lack of incoming 1652 * tasks. When true, the same keep-alive policy applying to 1653 * non-core threads applies also to core threads. To avoid 1654 * continual thread replacement, the keep-alive time must be 1655 * greater than zero when setting {@code true}. This method 1656 * should in general be called before the pool is actively used. 1657 * 1658 * @param value {@code true} if should time out, else {@code false} 1659 * @throws IllegalArgumentException if value is {@code true} 1660 * and the current keep-alive time is not greater than zero 1661 * 1662 * @since 1.6 1663 */ allowCoreThreadTimeOut(boolean value)1664 public void allowCoreThreadTimeOut(boolean value) { 1665 if (value && keepAliveTime <= 0) 1666 throw new IllegalArgumentException("Core threads must have nonzero keep alive times"); 1667 if (value != allowCoreThreadTimeOut) { 1668 allowCoreThreadTimeOut = value; 1669 if (value) 1670 interruptIdleWorkers(); 1671 } 1672 } 1673 1674 /** 1675 * Sets the maximum allowed number of threads. This overrides any 1676 * value set in the constructor. If the new value is smaller than 1677 * the current value, excess existing threads will be 1678 * terminated when they next become idle. 1679 * 1680 * @param maximumPoolSize the new maximum 1681 * @throws IllegalArgumentException if the new maximum is 1682 * less than or equal to zero, or 1683 * less than the {@linkplain #getCorePoolSize core pool size} 1684 * @see #getMaximumPoolSize 1685 */ setMaximumPoolSize(int maximumPoolSize)1686 public void setMaximumPoolSize(int maximumPoolSize) { 1687 if (maximumPoolSize <= 0 || maximumPoolSize < corePoolSize) 1688 throw new IllegalArgumentException(); 1689 this.maximumPoolSize = maximumPoolSize; 1690 if (workerCountOf(ctl.get()) > maximumPoolSize) 1691 interruptIdleWorkers(); 1692 } 1693 1694 /** 1695 * Returns the maximum allowed number of threads. 1696 * 1697 * @return the maximum allowed number of threads 1698 * @see #setMaximumPoolSize 1699 */ getMaximumPoolSize()1700 public int getMaximumPoolSize() { 1701 return maximumPoolSize; 1702 } 1703 1704 /** 1705 * Sets the thread keep-alive time, which is the amount of time 1706 * that threads may remain idle before being terminated. 1707 * Threads that wait this amount of time without processing a 1708 * task will be terminated if there are more than the core 1709 * number of threads currently in the pool, or if this pool 1710 * {@linkplain #allowsCoreThreadTimeOut() allows core thread timeout}. 1711 * This overrides any value set in the constructor. 1712 * 1713 * @param time the time to wait. A time value of zero will cause 1714 * excess threads to terminate immediately after executing tasks. 1715 * @param unit the time unit of the {@code time} argument 1716 * @throws IllegalArgumentException if {@code time} less than zero or 1717 * if {@code time} is zero and {@code allowsCoreThreadTimeOut} 1718 * @see #getKeepAliveTime(TimeUnit) 1719 */ setKeepAliveTime(long time, TimeUnit unit)1720 public void setKeepAliveTime(long time, TimeUnit unit) { 1721 if (time < 0) 1722 throw new IllegalArgumentException(); 1723 if (time == 0 && allowsCoreThreadTimeOut()) 1724 throw new IllegalArgumentException("Core threads must have nonzero keep alive times"); 1725 long keepAliveTime = unit.toNanos(time); 1726 long delta = keepAliveTime - this.keepAliveTime; 1727 this.keepAliveTime = keepAliveTime; 1728 if (delta < 0) 1729 interruptIdleWorkers(); 1730 } 1731 1732 /** 1733 * Returns the thread keep-alive time, which is the amount of time 1734 * that threads may remain idle before being terminated. 1735 * Threads that wait this amount of time without processing a 1736 * task will be terminated if there are more than the core 1737 * number of threads currently in the pool, or if this pool 1738 * {@linkplain #allowsCoreThreadTimeOut() allows core thread timeout}. 1739 * 1740 * @param unit the desired time unit of the result 1741 * @return the time limit 1742 * @see #setKeepAliveTime(long, TimeUnit) 1743 */ getKeepAliveTime(TimeUnit unit)1744 public long getKeepAliveTime(TimeUnit unit) { 1745 return unit.convert(keepAliveTime, TimeUnit.NANOSECONDS); 1746 } 1747 1748 /* User-level queue utilities */ 1749 1750 /** 1751 * Returns the task queue used by this executor. Access to the 1752 * task queue is intended primarily for debugging and monitoring. 1753 * This queue may be in active use. Retrieving the task queue 1754 * does not prevent queued tasks from executing. 1755 * 1756 * @return the task queue 1757 */ getQueue()1758 public BlockingQueue<Runnable> getQueue() { 1759 return workQueue; 1760 } 1761 1762 /** 1763 * Removes this task from the executor's internal queue if it is 1764 * present, thus causing it not to be run if it has not already 1765 * started. 1766 * 1767 * <p>This method may be useful as one part of a cancellation 1768 * scheme. It may fail to remove tasks that have been converted 1769 * into other forms before being placed on the internal queue. 1770 * For example, a task entered using {@code submit} might be 1771 * converted into a form that maintains {@code Future} status. 1772 * However, in such cases, method {@link #purge} may be used to 1773 * remove those Futures that have been cancelled. 1774 * 1775 * @param task the task to remove 1776 * @return {@code true} if the task was removed 1777 */ remove(Runnable task)1778 public boolean remove(Runnable task) { 1779 boolean removed = workQueue.remove(task); 1780 tryTerminate(); // In case SHUTDOWN and now empty 1781 return removed; 1782 } 1783 1784 /** 1785 * Tries to remove from the work queue all {@link Future} 1786 * tasks that have been cancelled. This method can be useful as a 1787 * storage reclamation operation, that has no other impact on 1788 * functionality. Cancelled tasks are never executed, but may 1789 * accumulate in work queues until worker threads can actively 1790 * remove them. Invoking this method instead tries to remove them now. 1791 * However, this method may fail to remove tasks in 1792 * the presence of interference by other threads. 1793 */ purge()1794 public void purge() { 1795 final BlockingQueue<Runnable> q = workQueue; 1796 try { 1797 Iterator<Runnable> it = q.iterator(); 1798 while (it.hasNext()) { 1799 Runnable r = it.next(); 1800 if (r instanceof Future<?> && ((Future<?>)r).isCancelled()) 1801 it.remove(); 1802 } 1803 } catch (ConcurrentModificationException fallThrough) { 1804 // Take slow path if we encounter interference during traversal. 1805 // Make copy for traversal and call remove for cancelled entries. 1806 // The slow path is more likely to be O(N*N). 1807 for (Object r : q.toArray()) 1808 if (r instanceof Future<?> && ((Future<?>)r).isCancelled()) 1809 q.remove(r); 1810 } 1811 1812 tryTerminate(); // In case SHUTDOWN and now empty 1813 } 1814 1815 /* Statistics */ 1816 1817 /** 1818 * Returns the current number of threads in the pool. 1819 * 1820 * @return the number of threads 1821 */ getPoolSize()1822 public int getPoolSize() { 1823 final ReentrantLock mainLock = this.mainLock; 1824 mainLock.lock(); 1825 try { 1826 // Remove rare and surprising possibility of 1827 // isTerminated() && getPoolSize() > 0 1828 return runStateAtLeast(ctl.get(), TIDYING) ? 0 1829 : workers.size(); 1830 } finally { 1831 mainLock.unlock(); 1832 } 1833 } 1834 1835 /** 1836 * Returns the approximate number of threads that are actively 1837 * executing tasks. 1838 * 1839 * @return the number of threads 1840 */ getActiveCount()1841 public int getActiveCount() { 1842 final ReentrantLock mainLock = this.mainLock; 1843 mainLock.lock(); 1844 try { 1845 int n = 0; 1846 for (Worker w : workers) 1847 if (w.isLocked()) 1848 ++n; 1849 return n; 1850 } finally { 1851 mainLock.unlock(); 1852 } 1853 } 1854 1855 /** 1856 * Returns the largest number of threads that have ever 1857 * simultaneously been in the pool. 1858 * 1859 * @return the number of threads 1860 */ getLargestPoolSize()1861 public int getLargestPoolSize() { 1862 final ReentrantLock mainLock = this.mainLock; 1863 mainLock.lock(); 1864 try { 1865 return largestPoolSize; 1866 } finally { 1867 mainLock.unlock(); 1868 } 1869 } 1870 1871 /** 1872 * Returns the approximate total number of tasks that have ever been 1873 * scheduled for execution. Because the states of tasks and 1874 * threads may change dynamically during computation, the returned 1875 * value is only an approximation. 1876 * 1877 * @return the number of tasks 1878 */ getTaskCount()1879 public long getTaskCount() { 1880 final ReentrantLock mainLock = this.mainLock; 1881 mainLock.lock(); 1882 try { 1883 long n = completedTaskCount; 1884 for (Worker w : workers) { 1885 n += w.completedTasks; 1886 if (w.isLocked()) 1887 ++n; 1888 } 1889 return n + workQueue.size(); 1890 } finally { 1891 mainLock.unlock(); 1892 } 1893 } 1894 1895 /** 1896 * Returns the approximate total number of tasks that have 1897 * completed execution. Because the states of tasks and threads 1898 * may change dynamically during computation, the returned value 1899 * is only an approximation, but one that does not ever decrease 1900 * across successive calls. 1901 * 1902 * @return the number of tasks 1903 */ getCompletedTaskCount()1904 public long getCompletedTaskCount() { 1905 final ReentrantLock mainLock = this.mainLock; 1906 mainLock.lock(); 1907 try { 1908 long n = completedTaskCount; 1909 for (Worker w : workers) 1910 n += w.completedTasks; 1911 return n; 1912 } finally { 1913 mainLock.unlock(); 1914 } 1915 } 1916 1917 /** 1918 * Returns a string identifying this pool, as well as its state, 1919 * including indications of run state and estimated worker and 1920 * task counts. 1921 * 1922 * @return a string identifying this pool, as well as its state 1923 */ toString()1924 public String toString() { 1925 long ncompleted; 1926 int nworkers, nactive; 1927 final ReentrantLock mainLock = this.mainLock; 1928 mainLock.lock(); 1929 try { 1930 ncompleted = completedTaskCount; 1931 nactive = 0; 1932 nworkers = workers.size(); 1933 for (Worker w : workers) { 1934 ncompleted += w.completedTasks; 1935 if (w.isLocked()) 1936 ++nactive; 1937 } 1938 } finally { 1939 mainLock.unlock(); 1940 } 1941 int c = ctl.get(); 1942 String runState = 1943 runStateLessThan(c, SHUTDOWN) ? "Running" : 1944 runStateAtLeast(c, TERMINATED) ? "Terminated" : 1945 "Shutting down"; 1946 return super.toString() + 1947 "[" + runState + 1948 ", pool size = " + nworkers + 1949 ", active threads = " + nactive + 1950 ", queued tasks = " + workQueue.size() + 1951 ", completed tasks = " + ncompleted + 1952 "]"; 1953 } 1954 1955 /* Extension hooks */ 1956 1957 /** 1958 * Method invoked prior to executing the given Runnable in the 1959 * given thread. This method is invoked by thread {@code t} that 1960 * will execute task {@code r}, and may be used to re-initialize 1961 * ThreadLocals, or to perform logging. 1962 * 1963 * <p>This implementation does nothing, but may be customized in 1964 * subclasses. Note: To properly nest multiple overridings, subclasses 1965 * should generally invoke {@code super.beforeExecute} at the end of 1966 * this method. 1967 * 1968 * @param t the thread that will run task {@code r} 1969 * @param r the task that will be executed 1970 */ beforeExecute(Thread t, Runnable r)1971 protected void beforeExecute(Thread t, Runnable r) { } 1972 1973 /** 1974 * Method invoked upon completion of execution of the given Runnable. 1975 * This method is invoked by the thread that executed the task. If 1976 * non-null, the Throwable is the uncaught {@code RuntimeException} 1977 * or {@code Error} that caused execution to terminate abruptly. 1978 * 1979 * <p>This implementation does nothing, but may be customized in 1980 * subclasses. Note: To properly nest multiple overridings, subclasses 1981 * should generally invoke {@code super.afterExecute} at the 1982 * beginning of this method. 1983 * 1984 * <p><b>Note:</b> When actions are enclosed in tasks (such as 1985 * {@link FutureTask}) either explicitly or via methods such as 1986 * {@code submit}, these task objects catch and maintain 1987 * computational exceptions, and so they do not cause abrupt 1988 * termination, and the internal exceptions are <em>not</em> 1989 * passed to this method. If you would like to trap both kinds of 1990 * failures in this method, you can further probe for such cases, 1991 * as in this sample subclass that prints either the direct cause 1992 * or the underlying exception if a task has been aborted: 1993 * 1994 * <pre> {@code 1995 * class ExtendedExecutor extends ThreadPoolExecutor { 1996 * // ... 1997 * protected void afterExecute(Runnable r, Throwable t) { 1998 * super.afterExecute(r, t); 1999 * if (t == null 2000 * && r instanceof Future<?> 2001 * && ((Future<?>)r).isDone()) { 2002 * try { 2003 * Object result = ((Future<?>) r).get(); 2004 * } catch (CancellationException ce) { 2005 * t = ce; 2006 * } catch (ExecutionException ee) { 2007 * t = ee.getCause(); 2008 * } catch (InterruptedException ie) { 2009 * // ignore/reset 2010 * Thread.currentThread().interrupt(); 2011 * } 2012 * } 2013 * if (t != null) 2014 * System.out.println(t); 2015 * } 2016 * }}</pre> 2017 * 2018 * @param r the runnable that has completed 2019 * @param t the exception that caused termination, or null if 2020 * execution completed normally 2021 */ afterExecute(Runnable r, Throwable t)2022 protected void afterExecute(Runnable r, Throwable t) { } 2023 2024 /** 2025 * Method invoked when the Executor has terminated. Default 2026 * implementation does nothing. Note: To properly nest multiple 2027 * overridings, subclasses should generally invoke 2028 * {@code super.terminated} within this method. 2029 */ terminated()2030 protected void terminated() { } 2031 2032 /* Predefined RejectedExecutionHandlers */ 2033 2034 /** 2035 * A handler for rejected tasks that runs the rejected task 2036 * directly in the calling thread of the {@code execute} method, 2037 * unless the executor has been shut down, in which case the task 2038 * is discarded. 2039 */ 2040 public static class CallerRunsPolicy implements RejectedExecutionHandler { 2041 /** 2042 * Creates a {@code CallerRunsPolicy}. 2043 */ CallerRunsPolicy()2044 public CallerRunsPolicy() { } 2045 2046 /** 2047 * Executes task r in the caller's thread, unless the executor 2048 * has been shut down, in which case the task is discarded. 2049 * 2050 * @param r the runnable task requested to be executed 2051 * @param e the executor attempting to execute this task 2052 */ rejectedExecution(Runnable r, ThreadPoolExecutor e)2053 public void rejectedExecution(Runnable r, ThreadPoolExecutor e) { 2054 if (!e.isShutdown()) { 2055 r.run(); 2056 } 2057 } 2058 } 2059 2060 /** 2061 * A handler for rejected tasks that throws a 2062 * {@code RejectedExecutionException}. 2063 */ 2064 public static class AbortPolicy implements RejectedExecutionHandler { 2065 /** 2066 * Creates an {@code AbortPolicy}. 2067 */ AbortPolicy()2068 public AbortPolicy() { } 2069 2070 /** 2071 * Always throws RejectedExecutionException. 2072 * 2073 * @param r the runnable task requested to be executed 2074 * @param e the executor attempting to execute this task 2075 * @throws RejectedExecutionException always 2076 */ rejectedExecution(Runnable r, ThreadPoolExecutor e)2077 public void rejectedExecution(Runnable r, ThreadPoolExecutor e) { 2078 throw new RejectedExecutionException("Task " + r.toString() + 2079 " rejected from " + 2080 e.toString()); 2081 } 2082 } 2083 2084 /** 2085 * A handler for rejected tasks that silently discards the 2086 * rejected task. 2087 */ 2088 public static class DiscardPolicy implements RejectedExecutionHandler { 2089 /** 2090 * Creates a {@code DiscardPolicy}. 2091 */ DiscardPolicy()2092 public DiscardPolicy() { } 2093 2094 /** 2095 * Does nothing, which has the effect of discarding task r. 2096 * 2097 * @param r the runnable task requested to be executed 2098 * @param e the executor attempting to execute this task 2099 */ rejectedExecution(Runnable r, ThreadPoolExecutor e)2100 public void rejectedExecution(Runnable r, ThreadPoolExecutor e) { 2101 } 2102 } 2103 2104 /** 2105 * A handler for rejected tasks that discards the oldest unhandled 2106 * request and then retries {@code execute}, unless the executor 2107 * is shut down, in which case the task is discarded. 2108 */ 2109 public static class DiscardOldestPolicy implements RejectedExecutionHandler { 2110 /** 2111 * Creates a {@code DiscardOldestPolicy} for the given executor. 2112 */ DiscardOldestPolicy()2113 public DiscardOldestPolicy() { } 2114 2115 /** 2116 * Obtains and ignores the next task that the executor 2117 * would otherwise execute, if one is immediately available, 2118 * and then retries execution of task r, unless the executor 2119 * is shut down, in which case task r is instead discarded. 2120 * 2121 * @param r the runnable task requested to be executed 2122 * @param e the executor attempting to execute this task 2123 */ rejectedExecution(Runnable r, ThreadPoolExecutor e)2124 public void rejectedExecution(Runnable r, ThreadPoolExecutor e) { 2125 if (!e.isShutdown()) { 2126 e.getQueue().poll(); 2127 e.execute(r); 2128 } 2129 } 2130 } 2131 } 2132