23

“类Object(及其子类)的每个实例都拥有一个锁,该锁在进入synchronized方法时获得,并在退出时自动释放”

这是否意味着我们创建的任何对象实例在内部默认都有一个“锁”(作为字段实现)?

我对这个“锁”概念感到困惑,我想知道它在内部实际上做了什么。

谁能指导我到一些可以找到更多信息的地方?

4

3 回答 3

24

与往常一样,JLS 提供了答案 (17.1)

这些方法中最基本的是同步,它是使用监视器实现的。Java 中的每个对象都与一个监视器相关联,线程可以锁定或解锁监视器。一次只有一个线程可以锁定监视器。任何其他试图锁定该监视器的线程都会被阻塞,直到它们能够获得对该监视器的锁定。一个线程 t 可能会多次锁定一个特定的监视器;每次解锁都会反转一次锁定操作的效果。

所以,不,lock不像是一个字段Object(正如您通过简单地查看Object 的源代码所看到的那样)。相反,每个Object都与一个“监视器”相关联,并且正是这个监视器被锁定或解锁。

我只是想指出一个进一步的参考资料,其中详细说明了“Java 是如何做到的”,以确保它不会被忽视。这位于@selig 在下面发现的 C++ 代码的注释中,我鼓励所有对下面内容的支持来回答他的问题。您可以在此处提供的链接中查看完整的源代码。

  126 // -----------------------------------------------------------------------------
  127 // Theory of operations -- Monitors lists, thread residency, etc:
  128 //
  129 // * A thread acquires ownership of a monitor by successfully
  130 //   CAS()ing the _owner field from null to non-null.
  131 //
  132 // * Invariant: A thread appears on at most one monitor list --
  133 //   cxq, EntryList or WaitSet -- at any one time.
  134 //
  135 // * Contending threads "push" themselves onto the cxq with CAS
  136 //   and then spin/park.
  137 //
  138 // * After a contending thread eventually acquires the lock it must
  139 //   dequeue itself from either the EntryList or the cxq.
  140 //
  141 // * The exiting thread identifies and unparks an "heir presumptive"
  142 //   tentative successor thread on the EntryList.  Critically, the
  143 //   exiting thread doesn't unlink the successor thread from the EntryList.
  144 //   After having been unparked, the wakee will recontend for ownership of
  145 //   the monitor.   The successor (wakee) will either acquire the lock or
  146 //   re-park itself.
  147 //
  148 //   Succession is provided for by a policy of competitive handoff.
  149 //   The exiting thread does _not_ grant or pass ownership to the
  150 //   successor thread.  (This is also referred to as "handoff" succession").
  151 //   Instead the exiting thread releases ownership and possibly wakes
  152 //   a successor, so the successor can (re)compete for ownership of the lock.
  153 //   If the EntryList is empty but the cxq is populated the exiting
  154 //   thread will drain the cxq into the EntryList.  It does so by
  155 //   by detaching the cxq (installing null with CAS) and folding
  156 //   the threads from the cxq into the EntryList.  The EntryList is
  157 //   doubly linked, while the cxq is singly linked because of the
  158 //   CAS-based "push" used to enqueue recently arrived threads (RATs).
  159 //
  160 // * Concurrency invariants:
  161 //
  162 //   -- only the monitor owner may access or mutate the EntryList.
  163 //      The mutex property of the monitor itself protects the EntryList
  164 //      from concurrent interference.
  165 //   -- Only the monitor owner may detach the cxq.
  166 //
  167 // * The monitor entry list operations avoid locks, but strictly speaking
  168 //   they're not lock-free.  Enter is lock-free, exit is not.
  169 //   See http://j2se.east/~dice/PERSIST/040825-LockFreeQueues.html
  170 //
  171 // * The cxq can have multiple concurrent "pushers" but only one concurrent
  172 //   detaching thread.  This mechanism is immune from the ABA corruption.
  173 //   More precisely, the CAS-based "push" onto cxq is ABA-oblivious.
  174 //
  175 // * Taken together, the cxq and the EntryList constitute or form a
  176 //   single logical queue of threads stalled trying to acquire the lock.
  177 //   We use two distinct lists to improve the odds of a constant-time
  178 //   dequeue operation after acquisition (in the ::enter() epilog) and
  179 //   to reduce heat on the list ends.  (c.f. Michael Scott's "2Q" algorithm).
  180 //   A key desideratum is to minimize queue & monitor metadata manipulation
  181 //   that occurs while holding the monitor lock -- that is, we want to
  182 //   minimize monitor lock holds times.  Note that even a small amount of
  183 //   fixed spinning will greatly reduce the # of enqueue-dequeue operations
  184 //   on EntryList|cxq.  That is, spinning relieves contention on the "inner"
  185 //   locks and monitor metadata.
  186 //
  187 //   Cxq points to the the set of Recently Arrived Threads attempting entry.
  188 //   Because we push threads onto _cxq with CAS, the RATs must take the form of
  189 //   a singly-linked LIFO.  We drain _cxq into EntryList  at unlock-time when
  190 //   the unlocking thread notices that EntryList is null but _cxq is != null.
  191 //
  192 //   The EntryList is ordered by the prevailing queue discipline and
  193 //   can be organized in any convenient fashion, such as a doubly-linked list or
  194 //   a circular doubly-linked list.  Critically, we want insert and delete operations
  195 //   to operate in constant-time.  If we need a priority queue then something akin
  196 //   to Solaris' sleepq would work nicely.  Viz.,
  197 //   http://agg.eng/ws/on10_nightly/source/usr/src/uts/common/os/sleepq.c.
  198 //   Queue discipline is enforced at ::exit() time, when the unlocking thread
  199 //   drains the cxq into the EntryList, and orders or reorders the threads on the
  200 //   EntryList accordingly.
  201 //
  202 //   Barring "lock barging", this mechanism provides fair cyclic ordering,
  203 //   somewhat similar to an elevator-scan.
  204 //
  205 // * The monitor synchronization subsystem avoids the use of native
  206 //   synchronization primitives except for the narrow platform-specific
  207 //   park-unpark abstraction.  See the comments in os_solaris.cpp regarding
  208 //   the semantics of park-unpark.  Put another way, this monitor implementation
  209 //   depends only on atomic operations and park-unpark.  The monitor subsystem
  210 //   manages all RUNNING->BLOCKED and BLOCKED->READY transitions while the
  211 //   underlying OS manages the READY<->RUN transitions.
  212 //
  213 // * Waiting threads reside on the WaitSet list -- wait() puts
  214 //   the caller onto the WaitSet.
  215 //
  216 // * notify() or notifyAll() simply transfers threads from the WaitSet to
  217 //   either the EntryList or cxq.  Subsequent exit() operations will
  218 //   unpark the notifyee.  Unparking a notifee in notify() is inefficient -
  219 //   it's likely the notifyee would simply impale itself on the lock held
  220 //   by the notifier.
  221 //
  222 // * An interesting alternative is to encode cxq as (List,LockByte) where
  223 //   the LockByte is 0 iff the monitor is owned.  _owner is simply an auxiliary
  224 //   variable, like _recursions, in the scheme.  The threads or Events that form
  225 //   the list would have to be aligned in 256-byte addresses.  A thread would
  226 //   try to acquire the lock or enqueue itself with CAS, but exiting threads
  227 //   could use a 1-0 protocol and simply STB to set the LockByte to 0.
  228 //   Note that is is *not* word-tearing, but it does presume that full-word
  229 //   CAS operations are coherent with intermix with STB operations.  That's true
  230 //   on most common processors.
  231 //
  232 // * See also http://blogs.sun.com/dave
  233 
  234 
  235 // -----------------------------------------------------------------------------
于 2013-06-28T15:47:05.517 回答
22

另一个答案描述了语言定义所说的,而不是“内部发生的”。

Java 中的每个对象都有一个两个单词的对象头。标记词和类指针。第一个字(标记字)用于存储锁定信息和缓存哈希码。第二个字是指向存储该对象的静态信息(包括方法代码)的 klass 对象的指针。

HotSpot JVM 有一些花哨的锁定东西,包括瘦锁和偏向锁定,这基本上意味着如果你从不锁定一个对象,或者如果你从来没有任何争用,那么你将永远不会创建一个监视器对象(这是存储额外锁定信息的东西)。

监视器对象有一个条目集。如果对象已被锁定,则当您锁定该对象时,您的线程将被添加到条目集中。当您解锁对象时,您会唤醒条目集中的一个线程。

并发是一个非常复杂的领域,显然还有很多细节。

更新

此处解释了对象标头,对象监视器(即等待集)发生的情况的详细信息可以在openjdk 代码中找到。

于 2013-06-28T16:42:55.330 回答
-3

Lock是Java并发的内部概念。您可以通过同步或外部锁获得它。要了解更多信息,请参阅:http ://docs.oracle.com/javase/tutorial/essential/concurrency/newlocks.html

于 2013-06-28T15:48:39.943 回答