1/* 2 * Written by Doug Lea, Bill Scherer, and Michael Scott with 3 * assistance from members of JCP JSR-166 Expert Group and released to 4 * the public domain, as explained at 5 * http://creativecommons.org/publicdomain/zero/1.0/ 6 */ 7 8package java.util.concurrent; 9import java.util.concurrent.atomic.*; 10import java.util.concurrent.locks.LockSupport; 11 12/** 13 * A synchronization point at which threads can pair and swap elements 14 * within pairs. Each thread presents some object on entry to the 15 * {@link #exchange exchange} method, matches with a partner thread, 16 * and receives its partner's object on return. An Exchanger may be 17 * viewed as a bidirectional form of a {@link SynchronousQueue}. 18 * Exchangers may be useful in applications such as genetic algorithms 19 * and pipeline designs. 20 * 21 * <p><b>Sample Usage:</b> 22 * Here are the highlights of a class that uses an {@code Exchanger} 23 * to swap buffers between threads so that the thread filling the 24 * buffer gets a freshly emptied one when it needs it, handing off the 25 * filled one to the thread emptying the buffer. 26 * <pre> {@code 27 * class FillAndEmpty { 28 * Exchanger<DataBuffer> exchanger = new Exchanger<DataBuffer>(); 29 * DataBuffer initialEmptyBuffer = ... a made-up type 30 * DataBuffer initialFullBuffer = ... 31 * 32 * class FillingLoop implements Runnable { 33 * public void run() { 34 * DataBuffer currentBuffer = initialEmptyBuffer; 35 * try { 36 * while (currentBuffer != null) { 37 * addToBuffer(currentBuffer); 38 * if (currentBuffer.isFull()) 39 * currentBuffer = exchanger.exchange(currentBuffer); 40 * } 41 * } catch (InterruptedException ex) { ... handle ... } 42 * } 43 * } 44 * 45 * class EmptyingLoop implements Runnable { 46 * public void run() { 47 * DataBuffer currentBuffer = initialFullBuffer; 48 * try { 49 * while (currentBuffer != null) { 50 * takeFromBuffer(currentBuffer); 51 * if (currentBuffer.isEmpty()) 52 * currentBuffer = exchanger.exchange(currentBuffer); 53 * } 54 * } catch (InterruptedException ex) { ... handle ...} 55 * } 56 * } 57 * 58 * void start() { 59 * new Thread(new FillingLoop()).start(); 60 * new Thread(new EmptyingLoop()).start(); 61 * } 62 * }}</pre> 63 * 64 * <p>Memory consistency effects: For each pair of threads that 65 * successfully exchange objects via an {@code Exchanger}, actions 66 * prior to the {@code exchange()} in each thread 67 * <a href="package-summary.html#MemoryVisibility"><i>happen-before</i></a> 68 * those subsequent to a return from the corresponding {@code exchange()} 69 * in the other thread. 70 * 71 * @since 1.5 72 * @author Doug Lea and Bill Scherer and Michael Scott 73 * @param <V> The type of objects that may be exchanged 74 */ 75public class Exchanger<V> { 76 /* 77 * Algorithm Description: 78 * 79 * The basic idea is to maintain a "slot", which is a reference to 80 * a Node containing both an Item to offer and a "hole" waiting to 81 * get filled in. If an incoming "occupying" thread sees that the 82 * slot is null, it CAS'es (compareAndSets) a Node there and waits 83 * for another to invoke exchange. That second "fulfilling" thread 84 * sees that the slot is non-null, and so CASes it back to null, 85 * also exchanging items by CASing the hole, plus waking up the 86 * occupying thread if it is blocked. In each case CAS'es may 87 * fail because a slot at first appears non-null but is null upon 88 * CAS, or vice-versa. So threads may need to retry these 89 * actions. 90 * 91 * This simple approach works great when there are only a few 92 * threads using an Exchanger, but performance rapidly 93 * deteriorates due to CAS contention on the single slot when 94 * there are lots of threads using an exchanger. So instead we use 95 * an "arena"; basically a kind of hash table with a dynamically 96 * varying number of slots, any one of which can be used by 97 * threads performing an exchange. Incoming threads pick slots 98 * based on a hash of their Thread ids. If an incoming thread 99 * fails to CAS in its chosen slot, it picks an alternative slot 100 * instead. And similarly from there. If a thread successfully 101 * CASes into a slot but no other thread arrives, it tries 102 * another, heading toward the zero slot, which always exists even 103 * if the table shrinks. The particular mechanics controlling this 104 * are as follows: 105 * 106 * Waiting: Slot zero is special in that it is the only slot that 107 * exists when there is no contention. A thread occupying slot 108 * zero will block if no thread fulfills it after a short spin. 109 * In other cases, occupying threads eventually give up and try 110 * another slot. Waiting threads spin for a while (a period that 111 * should be a little less than a typical context-switch time) 112 * before either blocking (if slot zero) or giving up (if other 113 * slots) and restarting. There is no reason for threads to block 114 * unless there are unlikely to be any other threads present. 115 * Occupants are mainly avoiding memory contention so sit there 116 * quietly polling for a shorter period than it would take to 117 * block and then unblock them. Non-slot-zero waits that elapse 118 * because of lack of other threads waste around one extra 119 * context-switch time per try, which is still on average much 120 * faster than alternative approaches. 121 * 122 * Sizing: Usually, using only a few slots suffices to reduce 123 * contention. Especially with small numbers of threads, using 124 * too many slots can lead to just as poor performance as using 125 * too few of them, and there's not much room for error. The 126 * variable "max" maintains the number of slots actually in 127 * use. It is increased when a thread sees too many CAS 128 * failures. (This is analogous to resizing a regular hash table 129 * based on a target load factor, except here, growth steps are 130 * just one-by-one rather than proportional.) Growth requires 131 * contention failures in each of three tried slots. Requiring 132 * multiple failures for expansion copes with the fact that some 133 * failed CASes are not due to contention but instead to simple 134 * races between two threads or thread pre-emptions occurring 135 * between reading and CASing. Also, very transient peak 136 * contention can be much higher than the average sustainable 137 * levels. An attempt to decrease the max limit is usually made 138 * when a non-slot-zero wait elapses without being fulfilled. 139 * Threads experiencing elapsed waits move closer to zero, so 140 * eventually find existing (or future) threads even if the table 141 * has been shrunk due to inactivity. The chosen mechanics and 142 * thresholds for growing and shrinking are intrinsically 143 * entangled with indexing and hashing inside the exchange code, 144 * and can't be nicely abstracted out. 145 * 146 * Hashing: Each thread picks its initial slot to use in accord 147 * with a simple hashcode. The sequence is the same on each 148 * encounter by any given thread, but effectively random across 149 * threads. Using arenas encounters the classic cost vs quality 150 * tradeoffs of all hash tables. Here, we use a one-step FNV-1a 151 * hash code based on the current thread's Thread.getId(), along 152 * with a cheap approximation to a mod operation to select an 153 * index. The downside of optimizing index selection in this way 154 * is that the code is hardwired to use a maximum table size of 155 * 32. But this value more than suffices for known platforms and 156 * applications. 157 * 158 * Probing: On sensed contention of a selected slot, we probe 159 * sequentially through the table, analogously to linear probing 160 * after collision in a hash table. (We move circularly, in 161 * reverse order, to mesh best with table growth and shrinkage 162 * rules.) Except that to minimize the effects of false-alarms 163 * and cache thrashing, we try the first selected slot twice 164 * before moving. 165 * 166 * Padding: Even with contention management, slots are heavily 167 * contended, so use cache-padding to avoid poor memory 168 * performance. Because of this, slots are lazily constructed 169 * only when used, to avoid wasting this space unnecessarily. 170 * While isolation of locations is not much of an issue at first 171 * in an application, as time goes on and garbage-collectors 172 * perform compaction, slots are very likely to be moved adjacent 173 * to each other, which can cause much thrashing of cache lines on 174 * MPs unless padding is employed. 175 * 176 * This is an improvement of the algorithm described in the paper 177 * "A Scalable Elimination-based Exchange Channel" by William 178 * Scherer, Doug Lea, and Michael Scott in Proceedings of SCOOL05 179 * workshop. Available at: http://hdl.handle.net/1802/2104 180 */ 181 182 /** The number of CPUs, for sizing and spin control */ 183 private static final int NCPU = Runtime.getRuntime().availableProcessors(); 184 185 /** 186 * The capacity of the arena. Set to a value that provides more 187 * than enough space to handle contention. On small machines 188 * most slots won't be used, but it is still not wasted because 189 * the extra space provides some machine-level address padding 190 * to minimize interference with heavily CAS'ed Slot locations. 191 * And on very large machines, performance eventually becomes 192 * bounded by memory bandwidth, not numbers of threads/CPUs. 193 * This constant cannot be changed without also modifying 194 * indexing and hashing algorithms. 195 */ 196 private static final int CAPACITY = 32; 197 198 /** 199 * The value of "max" that will hold all threads without 200 * contention. When this value is less than CAPACITY, some 201 * otherwise wasted expansion can be avoided. 202 */ 203 private static final int FULL = 204 Math.max(0, Math.min(CAPACITY, NCPU / 2) - 1); 205 206 /** 207 * The number of times to spin (doing nothing except polling a 208 * memory location) before blocking or giving up while waiting to 209 * be fulfilled. Should be zero on uniprocessors. On 210 * multiprocessors, this value should be large enough so that two 211 * threads exchanging items as fast as possible block only when 212 * one of them is stalled (due to GC or preemption), but not much 213 * longer, to avoid wasting CPU resources. Seen differently, this 214 * value is a little over half the number of cycles of an average 215 * context switch time on most systems. The value here is 216 * approximately the average of those across a range of tested 217 * systems. 218 */ 219 private static final int SPINS = (NCPU == 1) ? 0 : 2000; 220 221 /** 222 * The number of times to spin before blocking in timed waits. 223 * Timed waits spin more slowly because checking the time takes 224 * time. The best value relies mainly on the relative rate of 225 * System.nanoTime vs memory accesses. The value is empirically 226 * derived to work well across a variety of systems. 227 */ 228 private static final int TIMED_SPINS = SPINS / 20; 229 230 /** 231 * Sentinel item representing cancellation of a wait due to 232 * interruption, timeout, or elapsed spin-waits. This value is 233 * placed in holes on cancellation, and used as a return value 234 * from waiting methods to indicate failure to set or get hole. 235 */ 236 private static final Object CANCEL = new Object(); 237 238 /** 239 * Value representing null arguments/returns from public 240 * methods. This disambiguates from internal requirement that 241 * holes start out as null to mean they are not yet set. 242 */ 243 private static final Object NULL_ITEM = new Object(); 244 245 /** 246 * Nodes hold partially exchanged data. This class 247 * opportunistically subclasses AtomicReference to represent the 248 * hole. So get() returns hole, and compareAndSet CAS'es value 249 * into hole. This class cannot be parameterized as "V" because 250 * of the use of non-V CANCEL sentinels. 251 */ 252 private static final class Node extends AtomicReference<Object> { 253 /** The element offered by the Thread creating this node. */ 254 public final Object item; 255 256 /** The Thread waiting to be signalled; null until waiting. */ 257 public volatile Thread waiter; 258 259 /** 260 * Creates node with given item and empty hole. 261 * @param item the item 262 */ 263 public Node(Object item) { 264 this.item = item; 265 } 266 } 267 268 /** 269 * A Slot is an AtomicReference with heuristic padding to lessen 270 * cache effects of this heavily CAS'ed location. While the 271 * padding adds noticeable space, all slots are created only on 272 * demand, and there will be more than one of them only when it 273 * would improve throughput more than enough to outweigh using 274 * extra space. 275 */ 276 private static final class Slot extends AtomicReference<Object> { 277 // Improve likelihood of isolation on <= 128 byte cache lines. 278 // We used to target 64 byte cache lines, but some x86s (including 279 // i7 under some BIOSes) actually use 128 byte cache lines. 280 long q0, q1, q2, q3, q4, q5, q6, q7, q8, q9, qa, qb, qc, qd, qe; 281 } 282 283 /** 284 * Slot array. Elements are lazily initialized when needed. 285 * Declared volatile to enable double-checked lazy construction. 286 */ 287 private volatile Slot[] arena = new Slot[CAPACITY]; 288 289 /** 290 * The maximum slot index being used. The value sometimes 291 * increases when a thread experiences too many CAS contentions, 292 * and sometimes decreases when a spin-wait elapses. Changes 293 * are performed only via compareAndSet, to avoid stale values 294 * when a thread happens to stall right before setting. 295 */ 296 private final AtomicInteger max = new AtomicInteger(); 297 298 /** 299 * Main exchange function, handling the different policy variants. 300 * Uses Object, not "V" as argument and return value to simplify 301 * handling of sentinel values. Callers from public methods decode 302 * and cast accordingly. 303 * 304 * @param item the (non-null) item to exchange 305 * @param timed true if the wait is timed 306 * @param nanos if timed, the maximum wait time 307 * @return the other thread's item, or CANCEL if interrupted or timed out 308 */ 309 private Object doExchange(Object item, boolean timed, long nanos) { 310 Node me = new Node(item); // Create in case occupying 311 int index = hashIndex(); // Index of current slot 312 int fails = 0; // Number of CAS failures 313 314 for (;;) { 315 Object y; // Contents of current slot 316 Slot slot = arena[index]; 317 if (slot == null) // Lazily initialize slots 318 createSlot(index); // Continue loop to reread 319 else if ((y = slot.get()) != null && // Try to fulfill 320 slot.compareAndSet(y, null)) { 321 Node you = (Node)y; // Transfer item 322 if (you.compareAndSet(null, item)) { 323 LockSupport.unpark(you.waiter); 324 return you.item; 325 } // Else cancelled; continue 326 } 327 else if (y == null && // Try to occupy 328 slot.compareAndSet(null, me)) { 329 if (index == 0) // Blocking wait for slot 0 330 return timed ? 331 awaitNanos(me, slot, nanos) : 332 await(me, slot); 333 Object v = spinWait(me, slot); // Spin wait for non-0 334 if (v != CANCEL) 335 return v; 336 me = new Node(item); // Throw away cancelled node 337 int m = max.get(); 338 if (m > (index >>>= 1)) // Decrease index 339 max.compareAndSet(m, m - 1); // Maybe shrink table 340 } 341 else if (++fails > 1) { // Allow 2 fails on 1st slot 342 int m = max.get(); 343 if (fails > 3 && m < FULL && max.compareAndSet(m, m + 1)) 344 index = m + 1; // Grow on 3rd failed slot 345 else if (--index < 0) 346 index = m; // Circularly traverse 347 } 348 } 349 } 350 351 /** 352 * Returns a hash index for the current thread. Uses a one-step 353 * FNV-1a hash code (http://www.isthe.com/chongo/tech/comp/fnv/) 354 * based on the current thread's Thread.getId(). These hash codes 355 * have more uniform distribution properties with respect to small 356 * moduli (here 1-31) than do other simple hashing functions. 357 * 358 * <p>To return an index between 0 and max, we use a cheap 359 * approximation to a mod operation, that also corrects for bias 360 * due to non-power-of-2 remaindering (see {@link 361 * java.util.Random#nextInt}). Bits of the hashcode are masked 362 * with "nbits", the ceiling power of two of table size (looked up 363 * in a table packed into three ints). If too large, this is 364 * retried after rotating the hash by nbits bits, while forcing new 365 * top bit to 0, which guarantees eventual termination (although 366 * with a non-random-bias). This requires an average of less than 367 * 2 tries for all table sizes, and has a maximum 2% difference 368 * from perfectly uniform slot probabilities when applied to all 369 * possible hash codes for sizes less than 32. 370 * 371 * @return a per-thread-random index, 0 <= index < max 372 */ 373 private final int hashIndex() { 374 long id = Thread.currentThread().getId(); 375 int hash = (((int)(id ^ (id >>> 32))) ^ 0x811c9dc5) * 0x01000193; 376 377 int m = max.get(); 378 int nbits = (((0xfffffc00 >> m) & 4) | // Compute ceil(log2(m+1)) 379 ((0x000001f8 >>> m) & 2) | // The constants hold 380 ((0xffff00f2 >>> m) & 1)); // a lookup table 381 int index; 382 while ((index = hash & ((1 << nbits) - 1)) > m) // May retry on 383 hash = (hash >>> nbits) | (hash << (33 - nbits)); // non-power-2 m 384 return index; 385 } 386 387 /** 388 * Creates a new slot at given index. Called only when the slot 389 * appears to be null. Relies on double-check using builtin 390 * locks, since they rarely contend. This in turn relies on the 391 * arena array being declared volatile. 392 * 393 * @param index the index to add slot at 394 */ 395 private void createSlot(int index) { 396 // Create slot outside of lock to narrow sync region 397 Slot newSlot = new Slot(); 398 Slot[] a = arena; 399 synchronized (a) { 400 if (a[index] == null) 401 a[index] = newSlot; 402 } 403 } 404 405 /** 406 * Tries to cancel a wait for the given node waiting in the given 407 * slot, if so, helping clear the node from its slot to avoid 408 * garbage retention. 409 * 410 * @param node the waiting node 411 * @param slot the slot it is waiting in 412 * @return true if successfully cancelled 413 */ 414 private static boolean tryCancel(Node node, Slot slot) { 415 if (!node.compareAndSet(null, CANCEL)) 416 return false; 417 if (slot.get() == node) // pre-check to minimize contention 418 slot.compareAndSet(node, null); 419 return true; 420 } 421 422 // Three forms of waiting. Each just different enough not to merge 423 // code with others. 424 425 /** 426 * Spin-waits for hole for a non-0 slot. Fails if spin elapses 427 * before hole filled. Does not check interrupt, relying on check 428 * in public exchange method to abort if interrupted on entry. 429 * 430 * @param node the waiting node 431 * @return on success, the hole; on failure, CANCEL 432 */ 433 private static Object spinWait(Node node, Slot slot) { 434 int spins = SPINS; 435 for (;;) { 436 Object v = node.get(); 437 if (v != null) 438 return v; 439 else if (spins > 0) 440 --spins; 441 else 442 tryCancel(node, slot); 443 } 444 } 445 446 /** 447 * Waits for (by spinning and/or blocking) and gets the hole 448 * filled in by another thread. Fails if interrupted before 449 * hole filled. 450 * 451 * When a node/thread is about to block, it sets its waiter field 452 * and then rechecks state at least one more time before actually 453 * parking, thus covering race vs fulfiller noticing that waiter 454 * is non-null so should be woken. 455 * 456 * Thread interruption status is checked only surrounding calls to 457 * park. The caller is assumed to have checked interrupt status 458 * on entry. 459 * 460 * @param node the waiting node 461 * @return on success, the hole; on failure, CANCEL 462 */ 463 private static Object await(Node node, Slot slot) { 464 Thread w = Thread.currentThread(); 465 int spins = SPINS; 466 for (;;) { 467 Object v = node.get(); 468 if (v != null) 469 return v; 470 else if (spins > 0) // Spin-wait phase 471 --spins; 472 else if (node.waiter == null) // Set up to block next 473 node.waiter = w; 474 else if (w.isInterrupted()) // Abort on interrupt 475 tryCancel(node, slot); 476 else // Block 477 LockSupport.park(node); 478 } 479 } 480 481 /** 482 * Waits for (at index 0) and gets the hole filled in by another 483 * thread. Fails if timed out or interrupted before hole filled. 484 * Same basic logic as untimed version, but a bit messier. 485 * 486 * @param node the waiting node 487 * @param nanos the wait time 488 * @return on success, the hole; on failure, CANCEL 489 */ 490 private Object awaitNanos(Node node, Slot slot, long nanos) { 491 int spins = TIMED_SPINS; 492 long lastTime = 0; 493 Thread w = null; 494 for (;;) { 495 Object v = node.get(); 496 if (v != null) 497 return v; 498 long now = System.nanoTime(); 499 if (w == null) 500 w = Thread.currentThread(); 501 else 502 nanos -= now - lastTime; 503 lastTime = now; 504 if (nanos > 0) { 505 if (spins > 0) 506 --spins; 507 else if (node.waiter == null) 508 node.waiter = w; 509 else if (w.isInterrupted()) 510 tryCancel(node, slot); 511 else 512 LockSupport.parkNanos(node, nanos); 513 } 514 else if (tryCancel(node, slot) && !w.isInterrupted()) 515 return scanOnTimeout(node); 516 } 517 } 518 519 /** 520 * Sweeps through arena checking for any waiting threads. Called 521 * only upon return from timeout while waiting in slot 0. When a 522 * thread gives up on a timed wait, it is possible that a 523 * previously-entered thread is still waiting in some other 524 * slot. So we scan to check for any. This is almost always 525 * overkill, but decreases the likelihood of timeouts when there 526 * are other threads present to far less than that in lock-based 527 * exchangers in which earlier-arriving threads may still be 528 * waiting on entry locks. 529 * 530 * @param node the waiting node 531 * @return another thread's item, or CANCEL 532 */ 533 private Object scanOnTimeout(Node node) { 534 Object y; 535 for (int j = arena.length - 1; j >= 0; --j) { 536 Slot slot = arena[j]; 537 if (slot != null) { 538 while ((y = slot.get()) != null) { 539 if (slot.compareAndSet(y, null)) { 540 Node you = (Node)y; 541 if (you.compareAndSet(null, node.item)) { 542 LockSupport.unpark(you.waiter); 543 return you.item; 544 } 545 } 546 } 547 } 548 } 549 return CANCEL; 550 } 551 552 /** 553 * Creates a new Exchanger. 554 */ 555 public Exchanger() { 556 } 557 558 /** 559 * Waits for another thread to arrive at this exchange point (unless 560 * the current thread is {@linkplain Thread#interrupt interrupted}), 561 * and then transfers the given object to it, receiving its object 562 * in return. 563 * 564 * <p>If another thread is already waiting at the exchange point then 565 * it is resumed for thread scheduling purposes and receives the object 566 * passed in by the current thread. The current thread returns immediately, 567 * receiving the object passed to the exchange by that other thread. 568 * 569 * <p>If no other thread is already waiting at the exchange then the 570 * current thread is disabled for thread scheduling purposes and lies 571 * dormant until one of two things happens: 572 * <ul> 573 * <li>Some other thread enters the exchange; or 574 * <li>Some other thread {@linkplain Thread#interrupt interrupts} 575 * the current thread. 576 * </ul> 577 * <p>If the current thread: 578 * <ul> 579 * <li>has its interrupted status set on entry to this method; or 580 * <li>is {@linkplain Thread#interrupt interrupted} while waiting 581 * for the exchange, 582 * </ul> 583 * then {@link InterruptedException} is thrown and the current thread's 584 * interrupted status is cleared. 585 * 586 * @param x the object to exchange 587 * @return the object provided by the other thread 588 * @throws InterruptedException if the current thread was 589 * interrupted while waiting 590 */ 591 public V exchange(V x) throws InterruptedException { 592 if (!Thread.interrupted()) { 593 Object v = doExchange((x == null) ? NULL_ITEM : x, false, 0); 594 if (v == NULL_ITEM) 595 return null; 596 if (v != CANCEL) 597 return (V)v; 598 Thread.interrupted(); // Clear interrupt status on IE throw 599 } 600 throw new InterruptedException(); 601 } 602 603 /** 604 * Waits for another thread to arrive at this exchange point (unless 605 * the current thread is {@linkplain Thread#interrupt interrupted} or 606 * the specified waiting time elapses), and then transfers the given 607 * object to it, receiving its object in return. 608 * 609 * <p>If another thread is already waiting at the exchange point then 610 * it is resumed for thread scheduling purposes and receives the object 611 * passed in by the current thread. The current thread returns immediately, 612 * receiving the object passed to the exchange by that other thread. 613 * 614 * <p>If no other thread is already waiting at the exchange then the 615 * current thread is disabled for thread scheduling purposes and lies 616 * dormant until one of three things happens: 617 * <ul> 618 * <li>Some other thread enters the exchange; or 619 * <li>Some other thread {@linkplain Thread#interrupt interrupts} 620 * the current thread; or 621 * <li>The specified waiting time elapses. 622 * </ul> 623 * <p>If the current thread: 624 * <ul> 625 * <li>has its interrupted status set on entry to this method; or 626 * <li>is {@linkplain Thread#interrupt interrupted} while waiting 627 * for the exchange, 628 * </ul> 629 * then {@link InterruptedException} is thrown and the current thread's 630 * interrupted status is cleared. 631 * 632 * <p>If the specified waiting time elapses then {@link 633 * TimeoutException} is thrown. If the time is less than or equal 634 * to zero, the method will not wait at all. 635 * 636 * @param x the object to exchange 637 * @param timeout the maximum time to wait 638 * @param unit the time unit of the <tt>timeout</tt> argument 639 * @return the object provided by the other thread 640 * @throws InterruptedException if the current thread was 641 * interrupted while waiting 642 * @throws TimeoutException if the specified waiting time elapses 643 * before another thread enters the exchange 644 */ 645 public V exchange(V x, long timeout, TimeUnit unit) 646 throws InterruptedException, TimeoutException { 647 if (!Thread.interrupted()) { 648 Object v = doExchange((x == null) ? NULL_ITEM : x, 649 true, unit.toNanos(timeout)); 650 if (v == NULL_ITEM) 651 return null; 652 if (v != CANCEL) 653 return (V)v; 654 if (!Thread.interrupted()) 655 throw new TimeoutException(); 656 } 657 throw new InterruptedException(); 658 } 659} 660