回顧
在上一篇 Java併發核心淺談 我們大概瞭解到了Lock
和synchronized
的共同點,再簡單總結下:
Lock
主要是自定義一個 counter,從而利用CAS
對其實現原子操作,而synchronized
是c++ hotspot
實現的 monitor(具體的咱也沒看,咱就不說)- 二者都可重入(遞歸,互調,循環),其本質都是維護一個可計數的 counter,在其它線程訪問加鎖對象時會判斷 counter 是否爲 0
- 理論上講二者都是阻塞式的,因爲線程在拿鎖時,如果拿不到,最終的結果只能等待(前提是線程的最終目的就是要獲取鎖)讀寫鎖分離成兩把鎖了,所以不一樣
舉個例子:線程 A 持有了某個對象的 monitor,其它線程在訪問該對象時,發現 monitor 不爲 0,所以只能阻塞掛起或者加入等待隊列,等着線程 A 處理完退出後將 monitor 置爲 0。在線程 A 處理任務期間,其它線程要麼循環訪問 monitor,要麼一直阻塞等着線程 A 喚醒,再不濟就真的如我所說,放棄鎖的競爭,去處理別的任務。但是應該做不到去處理別的任務後,任務處理到一半,被線程 A 通知後再回去搶鎖
公平鎖與非公平鎖
不共享 counter
// 非公平鎖在第一次拿鎖失敗也會調用該方法
public final void acquire(int arg) {
// 沒拿到鎖就加入隊列
if (!tryAcquire(arg) &&
acquireQueued(addWaiter(Node.EXCLUSIVE), arg))
selfInterrupt();
}
// 非公平鎖方法
final void lock() {
// 走來就嘗試獲取鎖
if (compareAndSetState(0, 1))
setExclusiveOwnerThread(Thread.currentThread());
else
acquire(1); // 上面那個方法
}
// 公平鎖 Acquire 計數
protected final boolean tryAcquire(int acquires) {
final Thread current = Thread.currentThread();
// 拿到計數
int c = getState();
if (c == 0) {
// 公平鎖會先嚐試排隊 非公平鎖少個 !hasQueuedPredecessors() 其它代碼一樣
if (!hasQueuedPredecessors() &&
compareAndSetState(0, acquires)) {
setExclusiveOwnerThread(current);
return true;
}
}
else if (current == getExclusiveOwnerThread()) {
int nextc = c + acquires;
if (nextc < 0) // overflow
throw new Error("Maximum lock count exceeded");
setState(nextc);
return true;
}
return false;
}
/**
* @return {@code true} if there is a queued thread preceding the // 當前線程前有線程等待,則排隊
* current thread, and {@code false} if the current thread
* is at the head of the queue or the queue is empty // 隊列爲空不用排隊
* @since 1.7
*/
public final boolean hasQueuedPredecessors() {
// The correctness of this depends on head being initialized
// before tail and on head.next being accurate if the current
// thread is first in queue.
Node t = tail; // Read fields in reverse initialization order
Node h = head;
Node s;
// 當前線程處於頭節點的下一個且不爲空則不用排隊
// 或該線程就是當前持有鎖的線程,即重入鎖,也不用排隊
return h != t &&
((s = h.next) == null || s.thread != Thread.currentThread());
}
// 加入等待隊列
final boolean acquireQueued(final Node node, int arg) {
boolean failed = true;
try {
boolean interrupted = false;
for (;;) {
final Node p = node.predecessor();
if (p == head && tryAcquire(arg)) {
setHead(node);
p.next = null; // help GC
failed = false;
return interrupted;
}
// 獲取失敗會檢查節點狀態
// 然後 park 線程
if (shouldParkAfterFailedAcquire(p, node) &&
parkAndCheckInterrupt())
interrupted = true;
}
} finally {
if (failed)
cancelAcquire(node);
}
}
/** waitStatus value to indicate thread has cancelled */
static final int CANCELLED = 1; // 線程取消加鎖
/** waitStatus value to indicate successor's thread needs unparking */
static final int SIGNAL = -1; // 解除線程 park
/** waitStatus value to indicate thread is waiting on condition */ //
static final int CONDITION = -2; // 線程被阻塞
/**
* waitStatus value to indicate the next acquireShared should
* unconditionally propagate
*/
static final int PROPAGATE = -3; // 廣播
// 官方註釋
/**
* Status field, taking on only the values:
* SIGNAL: The successor of this node is (or will soon be)
* blocked (via park), so the current node must
* unpark its successor when it releases or
* cancels. To avoid races, acquire methods must
* first indicate they need a signal,
* then retry the atomic acquire, and then,
* on failure, block.
* CANCELLED: This node is cancelled due to timeout or interrupt.
* Nodes never leave this state. In particular,
* a thread with cancelled node never again blocks.
* CONDITION: This node is currently on a condition queue.
* It will not be used as a sync queue node
* until transferred, at which time the status
* will be set to 0. (Use of this value here has
* nothing to do with the other uses of the
* field, but simplifies mechanics.)
* PROPAGATE: A releaseShared should be propagated to other
* nodes. This is set (for head node only) in
* doReleaseShared to ensure propagation
* continues, even if other operations have
* since intervened.
* 0: None of the above
*
* The values are arranged numerically to simplify use.
* Non-negative values mean that a node doesn't need to
* signal. So, most code doesn't need to check for particular
* values, just for sign.
*
* The field is initialized to 0 for normal sync nodes, and
* CONDITION for condition nodes. It is modified using CAS
* (or when possible, unconditional volatile writes).
*/
volatile int waitStatus;
讀鎖與寫鎖(共享鎖與排他鎖)
讀鎖:共享 counter
寫鎖:不共享 counter
// 讀寫鎖和線程池的類似之處
// 高 16 位爲讀計數,低 16 位爲寫計數
static final int SHARED_SHIFT = 16;
static final int SHARED_UNIT = (1 << SHARED_SHIFT);
static final int MAX_COUNT = (1 << SHARED_SHIFT) - 1;
static final int EXCLUSIVE_MASK = (1 << SHARED_SHIFT) - 1;
/** Returns the number of shared holds represented in count. */ // 獲取讀計數
static int sharedCount(int c) { return c >>> SHARED_SHIFT; }
/** Returns the number of exclusive holds represented in count. */ // 獲取寫計數
static int exclusiveCount(int c) { return c & EXCLUSIVE_MASK; }
/**
* A counter for per-thread read hold counts. 每個線程自己的讀計數
* Maintained as a ThreadLocal; cached in cachedHoldCounter.
*/
static final class HoldCounter {
int count; // initially 0
// Use id, not reference, to avoid garbage retention
final long tid = LockSupport.getThreadId(Thread.currentThread()); // 線程 id
}
// 嘗試獲取讀鎖
protected final int tryAcquireShared(int unused) {
// ReentrantReadWriteLock ReadLock 讀鎖
/*
* Walkthrough:
* 1. If write lock held by another thread, fail.
* 2. Otherwise, this thread is eligible for
* lock wrt state, so ask if it should block
* because of queue policy. If not, try
* to grant by CASing state and updating count.
* Note that step does not check for reentrant
* acquires, which is postponed to full version
* to avoid having to check hold count in
* the more typical non-reentrant case.
* 3. If step 2 fails either because thread
* apparently not eligible or CAS fails or count
* saturated, chain to version with full retry loop.
*/
Thread current = Thread.currentThread();
int c = getState();
// 如果寫鎖計數不爲零,且當前線程不是寫鎖持有線程,則可以獲得讀鎖
// 言外之意,獲得寫鎖的線程不可以再獲得讀鎖
// 個人認爲不用判斷寫計數也行
if (exclusiveCount(c) != 0 &&
getExclusiveOwnerThread() != current)
return -1;
// 獲得讀計數
int r = sharedCount(c);
// 檢查等待隊列 讀計數上限
if (!readerShouldBlock() &&
r < MAX_COUNT &&
// 在高 16 位更新
compareAndSetState(c, c + SHARED_UNIT)) {
if (r == 0) {
firstReader = current;
firstReaderHoldCount = 1;
} else if (firstReader == current) {
firstReaderHoldCount++;
} else {
HoldCounter rh = cachedHoldCounter;
if (rh == null ||
rh.tid != LockSupport.getThreadId(current))
// cachedHoldCounter 每個線程自己的讀計數,非共享。但是鎖計數與其它讀操作共享,不與寫操作共享
// readHolds 爲ThreadLocalHoldCounter,繼承於 ThreadLocal,存 cachedHoldCounter
cachedHoldCounter = rh = readHolds.get();
else if (rh.count == 0)
readHolds.set(rh);
rh.count++;
}
return 1;
}
// 說明在排隊中,就一直遍歷,直到隊首,實際起作用的代碼和上面代碼差不多
// 大師本人也說了代碼有冗餘
/*
* This code is in part redundant with that in
* tryAcquireShared but is simpler overall by not
* complicating tryAcquireShared with interactions between
* retries and lazily reading hold counts.
*/
return fullTryAcquireShared(current);
}
// 獲得寫鎖
protected final boolean tryAcquire(int acquires) {
/*
* Walkthrough:
* 1. If read count nonzero or write count nonzero
* and owner is a different thread, fail.
* 讀鎖不爲零(讀鎖排除寫鎖,但是與讀鎖共享)
* 寫鎖不爲零且鎖持有者不爲當前線程,則獲得鎖失敗
* 2. If count would saturate, fail. (This can only
* happen if count is already nonzero.) // 計數器已達最大值,獲得鎖失敗
* 3. Otherwise, this thread is eligible for lock if
* it is either a reentrant acquire or
* queue policy allows it. If so, update state
* and set owner. // 重入是可以的;隊列策略也是可以的,會在下面解釋
*/
Thread current = Thread.currentThread();
int c = getState();
// 獲得寫計數
int w = exclusiveCount(c);
if (c != 0) {
// (Note: if c != 0 and w == 0 then shared count != 0)
// 檢查所持有線程
if (w == 0 || current != getExclusiveOwnerThread())
return false;
// 檢查最大計數
if (w + exclusiveCount(acquires) > MAX_COUNT)
throw new Error("Maximum lock count exceeded");
// Reentrant acquire 線程重入獲得鎖,直接更新計數
setState(c + acquires);
return true;
}
// 隊列策略
// state 爲 0,檢查是否需要排隊
// 針對公平鎖:去排隊,如果當前線程在隊首或等待隊列爲空,則返回 false,自然會走後面的 CAS
// 否則就返回 true,則進入 return false;
// 針對非公平鎖:寫死爲 false,直接 CAS
if (writerShouldBlock() ||
!compareAndSetState(c, c + acquires))
return false;
// 設置當前寫鎖持有線程
setExclusiveOwnerThread(current);
return true;
}
// 因爲讀鎖是多個線程共享讀計數,各自維護了自己的讀計數,所以釋放的時候比寫鎖釋放要多些操作
protected final boolean tryReleaseShared(int unused) {
Thread current = Thread.currentThread();
// 當前線程是第一讀線程的操作
// firstReader 作爲字段緩存,是考慮到第一次讀的線程使用率高?
if (firstReader == current) {
// assert firstReaderHoldCount > 0;
if (firstReaderHoldCount == 1)
firstReader = null;
else
firstReaderHoldCount--;
} else {
HoldCounter rh = cachedHoldCounter;
if (rh == null ||
rh.tid != LockSupport.getThreadId(current))
rh = readHolds.get();
int count = rh.count;
if (count <= 1) {
readHolds.remove();
if (count <= 0)
throw unmatchedUnlockException();
}
--rh.count;
}
for (;;) {
int c = getState();
int nextc = c - SHARED_UNIT;
if (compareAndSetState(c, nextc))
// Releasing the read lock has no effect on readers,
// but it may allow waiting writers to proceed if
// both read and write locks are now free.
return nextc == 0;
}
}
總結一下
公平鎖和非公平鎖的“鎖”實現是基於CAS
,公平性基於內部維護的Node
鏈表
讀寫鎖,可以粗略的理解爲讀和寫兩種狀態,所以這兒的設計類似線程池的狀態。只不過,讀計數是可以多個讀線程是共享的(排除寫鎖),每個讀的線程都會維護自己的讀計數。寫鎖的話,獨佔寫計數,排除一切其他線程。