AQS框架源码分析-AbstractQueuedSynchronizer
前言:AQS框架在J.U.C中的地位不言而喻,可以说没有AQS就没有J.U.C包,可见其重要性,因此有必要对其原理进行详细深入的理解。
1.AQS是什么
在深入AQS之前,首先我们要搞清楚什么是AQS。AQS全称是AbstractQueuedSynchronizer,我们直接查看AQS源码的注释。
大致意思就是说:AQS提供了实现阻塞锁和相关同步器并依赖先进先出(FIFO)等待队列的框架。
AQS依赖一个原子数值作为锁的状态,子类可以有多个状态值,只能通过原子方法区操作该值,从而保证同步。
通过第一段的注释大致总结下AQS是什么:
①AQS是一个同步的基础框架,基于一个先进先出的队列。
②锁机制依赖一个原子值的状态。
③AQS的子类负责定义与操作这个状态值,但必须通过AQS提供的原子操作。
④AQS剩余的方法就是围绕队列,与线程阻塞唤醒等功能。
2.重要成员变量
AQS中有两个重要的成员变量:Node和ConditionObject。
①Node的作用是存储获取锁失败的线程,并且维护一个CLH FIFO队列,该队列是会被多线程操作的,所以Node中大部分变量都是被volatile修饰,并且通过自旋和CAS进行原子性的操作。CLH的数据结构如下:
Node有一个模式的属性:独占模式和共享模式,独占模式下资源是线程独占的,共享模式下,资源是可以被多个线程占用的。
Node源码如下:
1 static final class Node { 2 /** Marker to indicate a node is waiting in shared mode */ 3 static final Node SHARED = new Node(); // 共享模式 4 /** Marker to indicate a node is waiting in exclusive mode */ 5 static final Node EXCLUSIVE = null; // 独占模式 6 7 /** waitStatus value to indicate thread has cancelled */ 8 static final int CANCELLED = 1; // 表明线程已处于结束状态(被取消) 9 /** waitStatus value to indicate successor's thread needs unparking */ 10 static final int SIGNAL = -1; // 表明线程需要被唤醒 11 /** waitStatus value to indicate thread is waiting on condition */ 12 static final int CONDITION = -2; // 表明线程正处于条件队列上,等待某一条件 13 /** 14 * waitStatus value to indicate the next acquireShared should 15 * unconditionally propagate 16 */ 17 static final int PROPAGATE = -3; // 共享模式下同步状态会被传播 18 19 /** 20 * Status field, taking on only the values: 21 * SIGNAL: The successor of this node is (or will soon be) 22 * blocked (via park), so the current node must 23 * unpark its successor when it releases or 24 * cancels. To avoid races, acquire methods must 25 * first indicate they need a signal, 26 * then retry the atomic acquire, and then, 27 * on failure, block. 28 * CANCELLED: This node is cancelled due to timeout or interrupt. 29 * Nodes never leave this state. In particular, 30 * a thread with cancelled node never again blocks. 31 * CONDITION: This node is currently on a condition queue. 32 * It will not be used as a sync queue node 33 * until transferred, at which time the status 34 * will be set to 0. (Use of this value here has 35 * nothing to do with the other uses of the 36 * field, but simplifies mechanics.) 37 * PROPAGATE: A releaseShared should be propagated to other 38 * nodes. This is set (for head node only) in 39 * doReleaseShared to ensure propagation 40 * continues, even if other operations have 41 * since intervened. 42 * 0: None of the above 43 * 44 * The values are arranged numerically to simplify use. 45 * Non-negative values mean that a node doesn't need to 46 * signal. So, most code doesn't need to check for particular 47 * values, just for sign. 48 * 49 * The field is initialized to 0 for normal sync nodes, and 50 * CONDITION for condition nodes. It is modified using CAS 51 * (or when possible, unconditional volatile writes). 52 */ 53 volatile int waitStatus; 54 55 /** 56 * Link to predecessor node that current node/thread relies on 57 * for checking waitStatus. Assigned during enqueuing, and nulled 58 * out (for sake of GC) only upon dequeuing. Also, upon 59 * cancellation of a predecessor, we short-circuit while 60 * finding a non-cancelled one, which will always exist 61 * because the head node is never cancelled: A node becomes 62 * head only as a result of successful acquire. A 63 * cancelled thread never succeeds in acquiring, and a thread only 64 * cancels itself, not any other node. 65 */ 66 volatile Node prev; 67 68 /** 69 * Link to the successor node that the current node/thread 70 * unparks upon release. Assigned during enqueuing, adjusted 71 * when bypassing cancelled predecessors, and nulled out (for 72 * sake of GC) when dequeued. The enq operation does not 73 * assign next field of a predecessor until after attachment, 74 * so seeing a null next field does not necessarily mean that 75 * node is at end of queue. However, if a next field appears 76 * to be null, we can scan prev's from the tail to 77 * double-check. The next field of cancelled nodes is set to 78 * point to the node itself instead of null, to make life 79 * easier for isOnSyncQueue. 80 */ 81 volatile Node next; 82 83 /** 84 * The thread that enqueued this node. Initialized on 85 * construction and nulled out after use. 86 */ 87 volatile Thread thread; 88 89 /** 90 * Link to next node waiting on condition, or the special 91 * value SHARED. Because condition queues are accessed only 92 * when holding in exclusive mode, we just need a simple 93 * linked queue to hold nodes while they are waiting on 94 * conditions. They are then transferred to the queue to 95 * re-acquire. And because conditions can only be exclusive, 96 * we save a field by using special value to indicate shared 97 * mode. 98 */ 99 Node nextWaiter; 100 101 /** 102 * Returns true if node is waiting in shared mode. 103 */ 104 final boolean isShared() { 105 return nextWaiter == SHARED; 106 } 107 108 /** 109 * Returns previous node, or throws NullPointerException if null. 110 * Use when predecessor cannot be null. The null check could 111 * be elided, but is present to help the VM. 112 * 113 * @return the predecessor of this node 114 */ 115 final Node predecessor() throws NullPointerException { 116 Node p = prev; 117 if (p == null) 118 throw new NullPointerException(); 119 else 120 return p; 121 } 122 123 Node() { // Used to establish initial head or SHARED marker 124 } 125 // 线程加入等待结点 126 Node(Thread thread, Node mode) { // Used by addWaiter 127 this.nextWaiter = mode; 128 this.thread = thread; 129 } 130 // 线程加入条件对列,会带上线程的状态值waitStatus 131 Node(Thread thread, int waitStatus) { // Used by Condition 132 this.waitStatus = waitStatus; 133 this.thread = thread; 134 } 135 }
②ConditionObject:条件队列,这个类的作用从AQS的注释上可知。
该类主要是为了让子类实现独占模式。AQS框架下独占模式的获取资源、释放等操作到最后都是基于这个类实现的。只有在独占模式下才会去使用该类。
ConditionObject源码如下(对主要代码进行了注释):
1 public class ConditionObject implements Condition, java.io.Serializable { 2 private static final long serialVersionUID = 1173984872572414699L; 3 /** First node of condition queue. */ 4 private transient Node firstWaiter; // 存储条件对列中第一个节点 5 /** Last node of condition queue. */ 6 private transient Node lastWaiter; // 存储条件对列中最后一个节点 7 8 /** 9 * Creates a new {@code ConditionObject} instance. 10 */ 11 public ConditionObject() { } 12 13 // Internal methods 14 15 /** 16 * Adds a new waiter to wait queue. // 增加一个新的节点到等待队列中 17 * @return its new wait node 18 */ 19 private Node addConditionWaiter() { 20 Node t = lastWaiter; 21 // 如果最后一个节点的状态已经结束,则直接清理掉 22 // If lastWaiter is cancelled, clean out. 23 if (t != null && t.waitStatus != Node.CONDITION) { 24 // 拆分已经处于结束状态的节点 也就是清除掉这类节点 25 unlinkCancelledWaiters(); 26 t = lastWaiter; 27 } 28 // 创建一个新的节点,带上结点状态,表明结点处于条件对列上 29 Node node = new Node(Thread.currentThread(), Node.CONDITION); 30 /** 31 条件队列中加入节点都是从队尾加入,并且从下面代码可知,每次都会存储最后一个节点的值。 32 当最后一个节点为空时,说明队列中不存在节点,所以将node赋值给第一个节点,否则将节点加入对列尾 33 */ 34 if (t == null) 35 firstWaiter = node; 36 else 37 t.nextWaiter = node; 38 lastWaiter = node; // 存储最后一个节点的值 39 return node; 40 } 41 42 /** 43 * 唤醒节点 44 * 移除和转换节点直到节点状态处于未结束或者为空 (节点移除相当于唤醒) 45 * Removes and transfers nodes until hit non-cancelled one or 46 * null. Split out from signal in part to encourage compilers 47 * to inline the case of no waiters. 48 * @param first (non-null) the first node on condition queue 49 */ 50 private void doSignal(Node first) { 51 do { 52 // 当next节点为null时,则将lastWaiter赋值为null 53 if ( (firstWaiter = first.nextWaiter) == null) 54 lastWaiter = null; 55 first.nextWaiter = null; // 切断当前节点 56 } while (!transferForSignal(first) && 57 (first = firstWaiter) != null); 58 } 59 60 /** 61 * 唤醒所有节点 62 * Removes and transfers all nodes. 63 * @param first (non-null) the first node on condition queue 64 */ 65 private void doSignalAll(Node first) { 66 lastWaiter = firstWaiter = null; 67 do { 68 // 循环唤醒所有节点,代码还是比较容易理解 69 // 将每个节点直接截断即可 70 Node next = first.nextWaiter; 71 first.nextWaiter = null; 72 transferForSignal(first); 73 first = next; 74 } while (first != null); 75 } 76 77 /** 78 * Unlinks cancelled waiter nodes from condition queue. 79 * Called only while holding lock. This is called when 80 * cancellation occurred during condition wait, and upon 81 * insertion of a new waiter when lastWaiter is seen to have 82 * been cancelled. This method is needed to avoid garbage 83 * retention in the absence of signals. So even though it may 84 * require a full traversal, it comes into play only when 85 * timeouts or cancellations occur in the absence of 86 * signals. It traverses all nodes rather than stopping at a 87 * particular target to unlink all pointers to garbage nodes 88 * without requiring many re-traversals during cancellation 89 * storms. 90 */ 91 private void unlinkCancelledWaiters() { // 删除处于结束状态的节点 92 Node t = firstWaiter; 93 Node trail = null; 94 // 第一个节点为空,直接返回 95 // 这里会遍历所有节点 96 while (t != null) { 97 Node next = t.nextWaiter; // 记录下一个节点的值 98 // 当节点状态不为CONDITION 99 if (t.waitStatus != Node.CONDITION) { 100 // 首先将当前节点的下一个节点赋值为空,切断当前节点链路 101 t.nextWaiter = null; 102 // 如果追踪节点为空的时候,则存储第一个节点的值为next,因为当前节点状态不为CONDITION需要清理 103 if (trail == null) 104 firstWaiter = next; 105 else // 在追踪节点串联下一个节点,主要是为了存储最后一个节点的值 106 trail.nextWaiter = next; 107 if (next == null) // 当next为空时,则存储trail为最后一个节点,将最后一个节点值存储下来 108 lastWaiter = trail; 109 } 110 else // 当节点状态为CONDITION时,将该节点赋值给trail 111 trail = t; 112 t = next; // 将next赋值给t,继续遍历 113 } 114 } 115 116 // public methods 117 118 /** 119 * 唤醒等待时间最长的节点,使其拥有锁 120 * Moves the longest-waiting thread, if one exists, from the 121 * wait queue for this condition to the wait queue for the 122 * owning lock. 123 * 124 * @throws IllegalMonitorStateException if {@link #isHeldExclusively} 125 * returns {@code false} 126 */ 127 public final void signal() { 128 // 如果线程不是独占资源,则抛出异常,从这里也说明ConditionObject只能用在独占模式中 129 if (!isHeldExclusively()) 130 throw new IllegalMonitorStateException(); 131 Node first = firstWaiter; 132 if (first != null) 133 doSignal(first); 134 } 135 136 /** 137 * 唤醒所有等待节点 138 * Moves all threads from the wait queue for this condition to 139 * the wait queue for the owning lock. 140 * 141 * @throws IllegalMonitorStateException if {@link #isHeldExclusively} 142 * returns {@code false} 143 */ 144 public final void signalAll() { 145 if (!isHeldExclusively()) 146 throw new IllegalMonitorStateException(); 147 Node first = firstWaiter; 148 if (first != null) 149 doSignalAll(first); 150 } 151 152 /** 153 * 节点不间断等待 154 * Implements uninterruptible condition wait. 155 * <ol> 156 * <li> Save lock state returned by {@link #getState}. 157 * <li> Invoke {@link #release} with saved state as argument, 158 * throwing IllegalMonitorStateException if it fails. 159 * <li> Block until signalled. 160 * <li> Reacquire by invoking specialized version of 161 * {@link #acquire} with saved state as argument. 162 * </ol> 163 */ 164 public final void awaitUninterruptibly() { 165 Node node = addConditionWaiter(); 166 int savedState = fullyRelease(node); 167 boolean interrupted = false; 168 while (!isOnSyncQueue(node)) { 169 LockSupport.park(this); 170 if (Thread.interrupted()) 171 interrupted = true; 172 } 173 if (acquireQueued(node, savedState) || interrupted) 174 selfInterrupt(); 175 } 176 177 /* 178 * For interruptible waits, we need to track whether to throw 179 * InterruptedException, if interrupted while blocked on 180 * condition, versus reinterrupt current thread, if 181 * interrupted while blocked waiting to re-acquire. 182 */ 183 184 /** Mode meaning to reinterrupt on exit from wait */ 185 private static final int REINTERRUPT = 1; 186 /** Mode meaning to throw InterruptedException on exit from wait */ 187 private static final int THROW_IE = -1; 188 189 /** 190 * Checks for interrupt, returning THROW_IE if interrupted 191 * before signalled, REINTERRUPT if after signalled, or 192 * 0 if not interrupted. 193 */ 194 private int checkInterruptWhileWaiting(Node node) { 195 return Thread.interrupted() ? 196 (transferAfterCancelledWait(node) ? THROW_IE : REINTERRUPT) : 197 0; 198 } 199 200 /** 201 * Throws InterruptedException, reinterrupts current thread, or 202 * does nothing, depending on mode. 203 */ 204 private void reportInterruptAfterWait(int interruptMode) 205 throws InterruptedException { 206 if (interruptMode == THROW_IE) 207 throw new InterruptedException(); 208 else if (interruptMode == REINTERRUPT) 209 selfInterrupt(); 210 } 211 212 /** 213 * Implements interruptible condition wait. 214 * <ol> 215 * <li> If current thread is interrupted, throw InterruptedException. 216 * <li> Save lock state returned by {@link #getState}. 217 * <li> Invoke {@link #release} with saved state as argument, 218 * throwing IllegalMonitorStateException if it fails. 219 * <li> Block until signalled or interrupted. 220 * <li> Reacquire by invoking specialized version of 221 * {@link #acquire} with saved state as argument. 222 * <li> If interrupted while blocked in step 4, throw InterruptedException. 223 * </ol> 224 */ 225 public final void await() throws InterruptedException { 226 if (Thread.interrupted()) 227 throw new InterruptedException(); 228 Node node = addConditionWaiter(); 229 int savedState = fullyRelease(node); 230 int interruptMode = 0; 231 while (!isOnSyncQueue(node)) { 232 LockSupport.park(this); 233 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0) 234 break; 235 } 236 if (acquireQueued(node, savedState) && interruptMode != THROW_IE) 237 interruptMode = REINTERRUPT; 238 if (node.nextWaiter != null) // clean up if cancelled 239 unlinkCancelledWaiters(); 240 if (interruptMode != 0) 241 reportInterruptAfterWait(interruptMode); 242 } 243 244 /** 245 * Implements timed condition wait. 246 * <ol> 247 * <li> If current thread is interrupted, throw InterruptedException. 248 * <li> Save lock state returned by {@link #getState}. 249 * <li> Invoke {@link #release} with saved state as argument, 250 * throwing IllegalMonitorStateException if it fails. 251 * <li> Block until signalled, interrupted, or timed out. 252 * <li> Reacquire by invoking specialized version of 253 * {@link #acquire} with saved state as argument. 254 * <li> If interrupted while blocked in step 4, throw InterruptedException. 255 * </ol> 256 */ 257 public final long awaitNanos(long nanosTimeout) 258 throws InterruptedException { 259 if (Thread.interrupted()) 260 throw new InterruptedException(); 261 Node node = addConditionWaiter(); 262 int savedState = fullyRelease(node); 263 final long deadline = System.nanoTime() + nanosTimeout; 264 int interruptMode = 0; 265 while (!isOnSyncQueue(node)) { 266 if (nanosTimeout <= 0L) { 267 transferAfterCancelledWait(node); 268 break; 269 } 270 if (nanosTimeout >= spinForTimeoutThreshold) 271 LockSupport.parkNanos(this, nanosTimeout); 272 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0) 273 break; 274 nanosTimeout = deadline - System.nanoTime(); 275 } 276 if (acquireQueued(node, savedState) && interruptMode != THROW_IE) 277 interruptMode = REINTERRUPT; 278 if (node.nextWaiter != null) 279 unlinkCancelledWaiters(); 280 if (interruptMode != 0) 281 reportInterruptAfterWait(interruptMode); 282 return deadline - System.nanoTime(); 283 } 284 285 /** 286 * Implements absolute timed condition wait. 287 * <ol> 288 * <li> If current thread is interrupted, throw InterruptedException. 289 * <li> Save lock state returned by {@link #getState}. 290 * <li> Invoke {@link #release} with saved state as argument, 291 * throwing IllegalMonitorStateException if it fails. 292 * <li> Block until signalled, interrupted, or timed out. 293 * <li> Reacquire by invoking specialized version of 294 * {@link #acquire} with saved state as argument. 295 * <li> If interrupted while blocked in step 4, throw InterruptedException. 296 * <li> If timed out while blocked in step 4, return false, else true. 297 * </ol> 298 */ 299 public final boolean awaitUntil(Date deadline) 300 throws InterruptedException { 301 long abstime = deadline.getTime(); 302 if (Thread.interrupted()) 303 throw new InterruptedException(); 304 Node node = addConditionWaiter(); 305 int savedState = fullyRelease(node); 306 boolean timedout = false; 307 int interruptMode = 0; 308 while (!isOnSyncQueue(node)) { 309 if (System.currentTimeMillis() > abstime) { 310 timedout = transferAfterCancelledWait(node); 311 break; 312 } 313 LockSupport.parkUntil(this, abstime); 314 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0) 315 break; 316 } 317 if (acquireQueued(node, savedState) && interruptMode != THROW_IE) 318 interruptMode = REINTERRUPT; 319 if (node.nextWaiter != null) 320 unlinkCancelledWaiters(); 321 if (interruptMode != 0) 322 reportInterruptAfterWait(interruptMode); 323 return !timedout; 324 } 325 326 /** 327 * Implements timed condition wait. 328 * <ol> 329 * <li> If current thread is interrupted, throw InterruptedException. 330 * <li> Save lock state returned by {@link #getState}. 331 * <li> Invoke {@link #release} with saved state as argument, 332 * throwing IllegalMonitorStateException if it fails. 333 * <li> Block until signalled, interrupted, or timed out. 334 * <li> Reacquire by invoking specialized version of 335 * {@link #acquire} with saved state as argument. 336 * <li> If interrupted while blocked in step 4, throw InterruptedException. 337 * <li> If timed out while blocked in step 4, return false, else true. 338 * </ol> 339 */ 340 public final boolean await(long time, TimeUnit unit) 341 throws InterruptedException { 342 long nanosTimeout = unit.toNanos(time); 343 if (Thread.interrupted()) 344 throw new InterruptedException(); 345 Node node = addConditionWaiter(); 346 int savedState = fullyRelease(node); 347 final long deadline = System.nanoTime() + nanosTimeout; 348 boolean timedout = false; 349 int interruptMode = 0; 350 while (!isOnSyncQueue(node)) { 351 if (nanosTimeout <= 0L) { 352 timedout = transferAfterCancelledWait(node); 353 break; 354 } 355 if (nanosTimeout >= spinForTimeoutThreshold) 356 LockSupport.parkNanos(this, nanosTimeout); 357 if ((interruptMode = checkInterruptWhileWaiting(node)) != 0) 358 break; 359 nanosTimeout = deadline - System.nanoTime(); 360 } 361 if (acquireQueued(node, savedState) && interruptMode != THROW_IE) 362 interruptMode = REINTERRUPT; 363 if (node.nextWaiter != null) 364 unlinkCancelledWaiters(); 365 if (interruptMode != 0) 366 reportInterruptAfterWait(interruptMode); 367 return !timedout; 368 } 369 370 // support for instrumentation 371 372 /** 373 * Returns true if this condition was created by the given 374 * synchronization object. 375 * 376 * @return {@code true} if owned 377 */ 378 final boolean isOwnedBy(AbstractQueuedSynchronizer sync) { 379 return sync == AbstractQueuedSynchronizer.this; 380 } 381 382 /** 383 * Queries whether any threads are waiting on this condition. 384 * Implements {@link AbstractQueuedSynchronizer#hasWaiters(ConditionObject)}. 385 * 386 * @return {@code true} if there are any waiting threads 387 * @throws IllegalMonitorStateException if {@link #isHeldExclusively} 388 * returns {@code false} 389 */ 390 protected final boolean hasWaiters() { 391 if (!isHeldExclusively()) 392 throw new IllegalMonitorStateException(); 393 for (Node w = firstWaiter; w != null; w = w.nextWaiter) { 394 if (w.waitStatus == Node.CONDITION) 395 return true; 396 } 397 return false; 398 } 399 400 /** 401 * Returns an estimate of the number of threads waiting on 402 * this condition. 403 * Implements {@link AbstractQueuedSynchronizer#getWaitQueueLength(ConditionObject)}. 404 * 405 * @return the estimated number of waiting threads 406 * @throws IllegalMonitorStateException if {@link #isHeldExclusively} 407 * returns {@code false} 408 */ 409 protected final int getWaitQueueLength() { 410 if (!isHeldExclusively()) 411 throw new IllegalMonitorStateException(); 412 int n = 0; 413 for (Node w = firstWaiter; w != null; w = w.nextWaiter) { 414 if (w.waitStatus == Node.CONDITION) 415 ++n; 416 } 417 return n; 418 } 419 420 /** 421 * Returns a collection containing those threads that may be 422 * waiting on this Condition. 423 * Implements {@link AbstractQueuedSynchronizer#getWaitingThreads(ConditionObject)}. 424 * 425 * @return the collection of threads 426 * @throws IllegalMonitorStateException if {@link #isHeldExclusively} 427 * returns {@code false} 428 */ 429 protected final Collection<Thread> getWaitingThreads() { 430 if (!isHeldExclusively()) 431 throw new IllegalMonitorStateException(); 432 ArrayList<Thread> list = new ArrayList<Thread>(); 433 for (Node w = firstWaiter; w != null; w = w.nextWaiter) { 434 if (w.waitStatus == Node.CONDITION) { 435 Thread t = w.thread; 436 if (t != null) 437 list.add(t); 438 } 439 } 440 return list; 441 } 442 }
3.AQS成员函数
由于AQS分独占模式和共享模式,因此这里按独占、共享模式的顺序对AQS的成员函数进行分析。
①acquire(int arg)
独占模式下获取资源,如果获取到资源,线程直接返回,否则进入等待队列,直到获取到资源为止,整个过程忽略中断。源码如下:
1 /** 2 * Acquires in exclusive mode, ignoring interrupts. Implemented 3 * by invoking at least once {@link #tryAcquire}, 4 * returning on success. Otherwise the thread is queued, possibly 5 * repeatedly blocking and unblocking, invoking {@link 6 * #tryAcquire} until success. This method can be used 7 * to implement method {@link Lock#lock}. 8 * 9 * @param arg the acquire argument. This value is conveyed to 10 * {@link #tryAcquire} but is otherwise uninterpreted and 11 * can represent anything you like. 12 */ 13 public final void acquire(int arg) { 14 if (!tryAcquire(arg) && 15 acquireQueued(addWaiter(Node.EXCLUSIVE), arg)) 16 selfInterrupt(); 17 }
该函数执行流程:
A.如果tryAcquire()成功获取资源,则直接返回。
B.直接获取资源失败,则通过addWaiter()将线程加入队列尾,并标记为独占模式。
C.通过acquireQueued()让线程在等待队列中获取资源,通过自旋方式,一直获取到后才返回。如果在等待过程中被中断过,则返回true,否则返回false。
D.如果线程在等待获取资源的过程中被中断,只有在获取到资源后才会去响应,执行selfInterrupt进行自我中断。
#1.tryAcquire(int)
该方法是在独占模式下获取资源,成功-ture,失败-false。
1 protected boolean tryAcquire(int arg) { 2 throw new UnsupportedOperationException(); 3 }
直接调用该方法会抛出异常,因为AQS只是一个框架,只是定义该接口,具体实现需在子类中实现。
#2.addWaiter(Node mode)
将当前线程加入等待队列的队尾,并返回当前线程所在的节点。
1 private Node addWaiter(Node mode) { 2 // 创建节点,以独占模式 3 Node node = new Node(Thread.currentThread(), mode); 4 // Try the fast path of enq; backup to full enq on failure 5 // 尝试将节点快速放入队尾 6 Node pred = tail; 7 if (pred != null) { 8 node.prev = pred; 9 // 主要通过CAS入队尾 10 if (compareAndSetTail(pred, node)) { 11 pred.next = node; 12 return node; 13 } 14 } 15 // 如果快速入队尾失败,则通过enq方式入对尾 16 enq(node); 17 return node; 18 }
CAS操作后面讨论,这里先看enq(final Node node)入队尾操作。
1 private Node enq(final Node node) { 2 // 这里是CAS的“自旋”操作,直到将节点成功加入队尾 3 for (;;) { 4 Node t = tail; 5 // 因为每次入队都是从队尾加入,当队尾为null,则表明队列为null,则需初始化头结点 6 // 并将尾节点也指向头节点 7 if (t == null) { // Must initialize 8 if (compareAndSetHead(new Node())) 9 tail = head; 10 } else { // 通过CAS入队尾,自旋操作 11 node.prev = t; 12 if (compareAndSetTail(t, node)) { 13 t.next = node; 14 return t; 15 } 16 } 17 } 18 }
在线程入队尾后,就需要acquireQueued函数了,该函数的作用是让线程拿到资源,当然还是通过自旋的方式来拿资源,也是就是一个排队的过程。
1 final boolean acquireQueued(final Node node, int arg) { 2 boolean failed = true; // 标记是否成功拿到资源 3 try { 4 boolean interrupted = false; // 标记在等待过程中是否被中断过 5 // 自旋操作 6 for (;;) { 7 final Node p = node.predecessor(); // 拿到当前节点的前向节点 8 // 如果前向节点为head,则表明当前节点排在第二位了,已经得到获取资源的资格 9 if (p == head && tryAcquire(arg)) { 10 // 成功拿到资源后,将head节点指向当前节点 11 // 从这里可以看出,head节点就是当前获取到锁的节点 12 setHead(node); 13 // 将原来head节点的next设置为null,方便GC回收以前的head节点,也就意味着之前拿到锁的节点出队列了 14 p.next = null; // help GC 15 failed = false; 16 return interrupted; // 返回在排队过程中线程是否被中断过 17 } 18 // 到这里,表明线程处于等待状态,自旋直到被unpark 19 if (shouldParkAfterFailedAcquire(p, node) && 20 parkAndCheckInterrupt()) 21 interrupted = true; 22 } 23 } finally { 24 if (failed) // 获取资源失败,则将节点标记为结束状态 25 cancelAcquire(node); 26 } 27 }
在线程排队等待的过程中,有两个关键函数shouldParkAfterFailedAcquire(Node pred, Node node)和parkAndCheckInterrupt()。
1 private static boolean shouldParkAfterFailedAcquire(Node pred, Node node) { 2 int ws = pred.waitStatus; // 前驱节点的状态 3 if (ws == Node.SIGNAL) 4 // 如果前驱节点正处于被唤醒的状态,则正常排队等待即可 5 /* 6 * This node has already set status asking a release 7 * to signal it, so it can safely park. 8 */ 9 return true; 10 if (ws > 0) { // 前驱节点处于结束状态 11 /* 12 * Predecessor was cancelled. Skip over predecessors and 13 * indicate retry. 14 */ 15 /* 16 *继续向下找,一直找到处于正常等待状态的节点,将当前节点插入其后,其他 17 *无用节点形成一个链,会被GC 18 */ 19 do { 20 node.prev = pred = pred.prev; 21 } while (pred.waitStatus > 0); 22 pred.next = node; 23 } else { 24 /* 25 * waitStatus must be 0 or PROPAGATE. Indicate that we 26 * need a signal, but don't park yet. Caller will need to 27 * retry to make sure it cannot acquire before parking. 28 */ 29 // 前驱节点状态正常,则把前驱节点的状态设置为SIGNAL,这样前驱节点拿到资源后,可通知下当前节点 30 compareAndSetWaitStatus(pred, ws, Node.SIGNAL); 31 } 32 return false; 33 }
分析以上源码可知:只有当前驱节点的状态为SIGNAL时,当前节点才能正常排队等待,否则需找到一个合适的节点next位置来进行排队等待。
1 private final boolean parkAndCheckInterrupt() { 2 // 使线程进入waitting状态 3 LockSupport.park(this); 4 return Thread.interrupted(); // 返回线程是否被中断过 5 }
该函数作用:当节点正常进入排队后,让线程进入等待状态。
至此acquireQueued()函数总结完成,该函数的具体执行流程:
#1.首先检查节点是否可以立即获取资源。
#2.如果不能立即获取资源,则进行排队,这里需要找到正确的排队点,直到unpark或interrupt唤醒自己。
#3.唤醒后,判断自己是否有资格获取资源,如果拿到资源,则将head指向当前节点,并返回在等待过程是否被中断过,如果没拿到资源,则继续流程2。
acquire小结
到这里acquire(int)函数分析结束,这个函数非常重要,这里再贴上源码:
1 public final void acquire(int arg) { 2 if (!tryAcquire(arg) && 3 acquireQueued(addWaiter(Node.EXCLUSIVE), arg)) 4 selfInterrupt(); 5 }
#1.调用子类的tryAcquire直接获取资源,如果成功则返回。
#2.如果流程1失败,则将线程加入等待队列的队尾(独占模式)。
#3.在acquireQueued中排队,通过自旋获取资源,直到获取资源才返回。如果在排队过程中线程被中断过返回true,否则返回false。
#4.在排队过程中被中断是不响应的,只有获取到资源后,才进行自我中断,补上中断标记。
整个过程的流程图如下:
②release(int)独占模式释放资源。
1 public final boolean release(int arg) { 2 // 尝试释放资源 3 if (tryRelease(arg)) { 4 Node h = head; 5 if (h != null && h.waitStatus != 0) 6 unparkSuccessor(h); // 唤醒队列中下一个线程 7 return true; 8 } 9 return false; 10 }
释放锁的函数很简单,通过tryRelease尝试释放资源,然后唤醒队列中的其他线程。
tryRelease(int):
1 protected boolean tryRelease(int arg) { 2 throw new UnsupportedOperationException(); 3 }
与tryAcquire函数一样,该方法需要子类去实现,如果直接调用会抛异常。
unparkSuccessor(Node node):
唤醒等待队列中的下一个线程,这里唤醒的是等待队列中最前边那个未放弃的线程,注意看代码注释。
1 private void unparkSuccessor(Node node) { 2 /* 3 * If status is negative (i.e., possibly needing signal) try 4 * to clear in anticipation of signalling. It is OK if this 5 * fails or if status is changed by waiting thread. 6 */ 7 int ws = node.waitStatus; // 获取当前线程的状态 8 if (ws < 0) // 如果当前线程状态处于可用状态,则直接将状态值置0 9 compareAndSetWaitStatus(node, ws, 0); 10 11 /* 12 * Thread to unpark is held in successor, which is normally 13 * just the next node. But if cancelled or apparently null, 14 * traverse backwards from tail to find the actual 15 * non-cancelled successor. 16 */ 17 Node s = node.next; // 下一个节点 18 if (s == null || s.waitStatus > 0) { // 如果节点为null或节点已处于结束状态 19 s = null; 20 // 从队列尾向前遍历,找到next可用的节点,状态小于0就可用,这里的节点是队列中最前边的可用节点 21 for (Node t = tail; t != null && t != node; t = t.prev) 22 if (t.waitStatus <= 0) 23 s = t; 24 } 25 if (s != null) 26 LockSupport.unpark(s.thread);// 唤醒next线程 27 }
独占模式的主要函数分析完毕,接下来看共享模式。
②acquireShared(int)
共享模式下获取资源,如果成功则直接返回,否则进入等待队列,通过自旋直到获取资源为止。
1 public final void acquireShared(int arg) { 2 // 共享模式下获取资源,如果获取失败,则进入等待队列 3 // 同样该函数需要子类去实现 4 if (tryAcquireShared(arg) < 0) 5 doAcquireShared(arg); // 进入等待队列直到锁获取到为止 6 }
tryAcquireShared(int)函数返回值,需要注意下:
负数:表示获取失败;
0:获取成功,但没有剩余资源;
正数:获取成功,且有剩余资源;
#1.doAcquireShared(int)
将线程加入队列尾,然后通过自旋获取资源,直到得到资源才返回。
1 private void doAcquireShared(int arg) { 2 final Node node = addWaiter(Node.SHARED); // 将线程加入队尾,通过共享模式 3 boolean failed = true;// 是否成功 4 try { 5 boolean interrupted = false; // 在自旋过程中是否被中断过 6 for (;;) { 7 final Node p = node.predecessor(); // 前驱节点 8 if (p == head) { // 这里表明当前节点处于head的next位,此时node被唤醒,很可能是head用完来唤醒 9 int r = tryAcquireShared(arg); // 获取资源 10 if (r >= 0) { // 成功 11 setHeadAndPropagate(node, r);// 将head指向自己,还有剩余资源可用的话再唤醒之后的线程 12 p.next = null; // help GC 无用链,帮助GC 13 if (interrupted) // 如果等待过程中被中断过,将中断补上 14 selfInterrupt(); 15 failed = false; 16 return; 17 } 18 } 19 // 线程未排在head之后,继续排队,进入waiting状态,等着unpark 20 if (shouldParkAfterFailedAcquire(p, node) && 21 parkAndCheckInterrupt()) 22 interrupted = true; // 中断标记 23 } 24 } finally { 25 if (failed) 26 cancelAcquire(node); 27 } 28 }
整个流程与独占模式的acquireQueued很相似,只是共享模式下,在唤醒自己后,如果还有剩余资源,需要唤醒后续节点。
setHeadAndPropagate(node, int)
将head节点设置为当前节点,如果还有剩余资源,则唤醒下一个线程。
1 private void setHeadAndPropagate(Node node, int propagate) { 2 Node h = head; // Record old head for check below 3 setHead(node); // 将队列中的head执行当前节点 4 /* 5 * Try to signal next queued node if: 6 * Propagation was indicated by caller, 7 * or was recorded (as h.waitStatus either before 8 * or after setHead) by a previous operation 9 * (note: this uses sign-check of waitStatus because 10 * PROPAGATE status may transition to SIGNAL.) 11 * and 12 * The next node is waiting in shared mode, 13 * or we don't know, because it appears null 14 * 15 * The conservatism in both of these checks may cause 16 * unnecessary wake-ups, but only when there are multiple 17 * racing acquires/releases, so most need signals now or soon 18 * anyway. 19 */ 20 // 如果还有剩余资源,则唤醒后续线程 21 if (propagate > 0 || h == null || h.waitStatus < 0 || 22 (h = head) == null || h.waitStatus < 0) { 23 Node s = node.next; 24 if (s == null || s.isShared()) 25 doReleaseShared(); 26 } 27 }
这里除了将head设置成当前线程,如果有剩余资源,需要唤醒后续节点。
doReleaseShared()
1 private void doReleaseShared() { 2 /* 3 * Ensure that a release propagates, even if there are other 4 * in-progress acquires/releases. This proceeds in the usual 5 * way of trying to unparkSuccessor of head if it needs 6 * signal. But if it does not, status is set to PROPAGATE to 7 * ensure that upon release, propagation continues. 8 * Additionally, we must loop in case a new node is added 9 * while we are doing this. Also, unlike other uses of 10 * unparkSuccessor, we need to know if CAS to reset status 11 * fails, if so rechecking. 12 */ 13 // 自旋操作 14 for (;;) { 15 Node h = head; 16 if (h != null && h != tail) { 17 int ws = h.waitStatus; 18 if (ws == Node.SIGNAL) { // 如果head状态为SIGNAL,则需唤醒后续节点 19 // CAS一下当前节点的状态,判断是否为SIGNAL,如果是则置为0,否则继续循环 20 if (!compareAndSetWaitStatus(h, Node.SIGNAL, 0)) 21 continue; // loop to recheck cases 22 unparkSuccessor(h); // 唤醒后继节点 23 } 24 // 如果head节点状态为0,且CAS置为传播状态失败,则继续循环,因为if操作中会改变节点的状态 25 else if (ws == 0 && 26 !compareAndSetWaitStatus(h, 0, Node.PROPAGATE)) 27 continue; // loop on failed CAS 28 } 29 if (h == head) // 如果head节点发生了改变,则继续自旋操作,防止上述操作过程中添加了节点的情况 // loop if head changed 30 break; 31 } 32 }
该方法的作用主要是用于唤醒后续节点。
共享模式获取锁操作与独占模式基本相同:先直接获取资源,如果成功,直接返回;如果失败,则将线程加入等待队列尾,直到获取到资源才返回,整个过程忽略中断。不同点在于共享模式下自己拿到资源后,还需要唤醒后续节点。
#2.releaseShared(int)
同享模式下释放资源
1 public final boolean releaseShared(int arg) { 2 if (tryReleaseShared(arg)) { // 尝试释放资源 3 doReleaseShared(); // 唤醒后续节点,前面已经分析 4 return true; 5 } 6 return false; 7 }
共享模式释放资源与独占模式类似,但是独占模式下需要完全释放资源后,才会返回true,而共享模式没有这种要求。
总结
这里只是对AQS的顶层框架进行了简要的分析,具体需要深入其子类中去,AQS的子类按模式分类可聚合成以下几类:
#1.独占模式:
ReentrantLock:可重入锁。state=0独占锁,或者同一线程可多次获取锁(获取+1,释放-1)。
Worker(java.util.concurrent.ThreadPoolExecutor类中的内部类)线程池类。shutdown关闭空闲工作线程,中断worker工作线程是独占的,互斥的。
#2.共享模式:
Semaphore:信号量。 控制同时有多少个线程可以进入代码段。(互斥锁的拓展)
CountDownLatch:倒计时器。 初始化一个值,多线程减少这个值,直到为0,倒计时完毕,执行后续代码。
#3.独占+共享模式:
ReentrantReadWriteLock:可重入读写锁。独占写+共享读,即并发读,互斥写。
后续对这些类进行详细分析。
by Shawn Chen,2019.1.29日,下午。