thread.join 是干什么的？原理是什么？

Thread.join

---

加了join，表示join的线程的修改对于join之外的代码是可见的。
代码示例：

public class JoinDemo {
    private static int i = 1000;

    public static void main(String[] args) {
        new Thread(()->{
            i = 3000;
        }).start();
        
        System.out.println("i="+i);
    }
}

我们在main线程中定义一个i，初始值为1000。
在main方法中创建一个线程，将i值设置为3000，启动线程。
然后main方法打印 i 的值。
请问，i 值为多少？

我们执行后的输出结果为：

i=1000

这是由于main线程先于thread线程，所以输出的i 值为1000。
如果我们希望 Thread线程中的执行结果对main线程可见，怎么办？
使用 thread.join()，如下所示：

public static void main(String[] args) throws InterruptedException {
    Thread thread = new Thread(()->{
        i = 3000;
    });
    thread.start();
    thread.join();
    System.out.println("i="+i);
}

执行结果：

i=3000

所以说，如果我们希望结果是可见的话，可以通过join来做。
那么join的实现原理是什么？请看下图：

main线程中，创建了一个线程t1；
t1.start() 启动线程；
t1线程修改了i的值为3000；而且让修改可见。说明t1线程阻塞的main线程，使它无法打印。
t1线程继续执行直到终止。
t1终止后，去唤醒被它阻塞的线程。main线程继续执行，打印出 i=3000。

有阻塞，就一定要唤醒，否则线程无法释放。我们是怎么触发唤醒的？
我们进入join的源码：

public final synchronized void join(long millis)
    throws InterruptedException {
        long base = System.currentTimeMillis();
        long now = 0;

        if (millis < 0) {
            throw new IllegalArgumentException("timeout value is negative");
        }

        if (millis == 0) {
            while (isAlive()) {
                wait(0);
            }
        } else {
            while (isAlive()) {
                long delay = millis - now;
                if (delay <= 0) {
                    break;
                }
                wait(delay);
                now = System.currentTimeMillis() - base;
            }
        }
    }

join是个synchronized的方法，里面有个 wait(0) 方法来阻塞。
那它是如何唤醒的呢？
我们去看hotspot的源码，其中的thread.cpp：

// For any new cleanup additions, please check to see if they need to be applied to
// cleanup_failed_attach_current_thread as well.
void JavaThread::exit(bool destroy_vm, ExitType exit_type) {
  assert(this == JavaThread::current(),  "thread consistency check");

  HandleMark hm(this);
  Handle uncaught_exception(this, this->pending_exception());
  this->clear_pending_exception();
  Handle threadObj(this, this->threadObj());
  assert(threadObj.not_null(), "Java thread object should be created");

  if (get_thread_profiler() != NULL) {
    get_thread_profiler()->disengage();
    ResourceMark rm;
    get_thread_profiler()->print(get_thread_name());
  }


  // FIXIT: This code should be moved into else part, when reliable 1.2/1.3 check is in place
  {
    EXCEPTION_MARK;

    CLEAR_PENDING_EXCEPTION;
  }
  // FIXIT: The is_null check is only so it works better on JDK1.2 VM's. This
  // has to be fixed by a runtime query method
  if (!destroy_vm || JDK_Version::is_jdk12x_version()) {
    // JSR-166: change call from from ThreadGroup.uncaughtException to
    // java.lang.Thread.dispatchUncaughtException
    if (uncaught_exception.not_null()) {
      Handle group(this, java_lang_Thread::threadGroup(threadObj()));
      {
        EXCEPTION_MARK;
        // Check if the method Thread.dispatchUncaughtException() exists. If so
        // call it.  Otherwise we have an older library without the JSR-166 changes,
        // so call ThreadGroup.uncaughtException()
        KlassHandle recvrKlass(THREAD, threadObj->klass());
        CallInfo callinfo;
        KlassHandle thread_klass(THREAD, SystemDictionary::Thread_klass());
        LinkResolver::resolve_virtual_call(callinfo, threadObj, recvrKlass, thread_klass,
                                           vmSymbols::dispatchUncaughtException_name(),
                                           vmSymbols::throwable_void_signature(),
                                           KlassHandle(), false, false, THREAD);
        CLEAR_PENDING_EXCEPTION;
        methodHandle method = callinfo.selected_method();
        if (method.not_null()) {
          JavaValue result(T_VOID);
          JavaCalls::call_virtual(&result,
                                  threadObj, thread_klass,
                                  vmSymbols::dispatchUncaughtException_name(),
                                  vmSymbols::throwable_void_signature(),
                                  uncaught_exception,
                                  THREAD);
        } else {
          KlassHandle thread_group(THREAD, SystemDictionary::ThreadGroup_klass());
          JavaValue result(T_VOID);
          JavaCalls::call_virtual(&result,
                                  group, thread_group,
                                  vmSymbols::uncaughtException_name(),
                                  vmSymbols::thread_throwable_void_signature(),
                                  threadObj,           // Arg 1
                                  uncaught_exception,  // Arg 2
                                  THREAD);
        }
        if (HAS_PENDING_EXCEPTION) {
          ResourceMark rm(this);
          jio_fprintf(defaultStream::error_stream(),
                "\nException: %s thrown from the UncaughtExceptionHandler"
                " in thread \"%s\"\n",
                pending_exception()->klass()->external_name(),
                get_thread_name());
          CLEAR_PENDING_EXCEPTION;
        }
      }
    }

    // Called before the java thread exit since we want to read info
    // from java_lang_Thread object
    EventThreadEnd event;
    if (event.should_commit()) {
        event.set_javalangthread(java_lang_Thread::thread_id(this->threadObj()));
        event.commit();
    }

    // Call after last event on thread
    EVENT_THREAD_EXIT(this);

    // Call Thread.exit(). We try 3 times in case we got another Thread.stop during
    // the execution of the method. If that is not enough, then we don't really care. Thread.stop
    // is deprecated anyhow.
    if (!is_Compiler_thread()) {
      int count = 3;
      while (java_lang_Thread::threadGroup(threadObj()) != NULL && (count-- > 0)) {
        EXCEPTION_MARK;
        JavaValue result(T_VOID);
        KlassHandle thread_klass(THREAD, SystemDictionary::Thread_klass());
        JavaCalls::call_virtual(&result,
                              threadObj, thread_klass,
                              vmSymbols::exit_method_name(),
                              vmSymbols::void_method_signature(),
                              THREAD);
        CLEAR_PENDING_EXCEPTION;
      }
    }
    // notify JVMTI
    if (JvmtiExport::should_post_thread_life()) {
      JvmtiExport::post_thread_end(this);
    }

    // We have notified the agents that we are exiting, before we go on,
    // we must check for a pending external suspend request and honor it
    // in order to not surprise the thread that made the suspend request.
    while (true) {
      {
        MutexLockerEx ml(SR_lock(), Mutex::_no_safepoint_check_flag);
        if (!is_external_suspend()) {
          set_terminated(_thread_exiting);
          ThreadService::current_thread_exiting(this);
          break;
        }
        // Implied else:
        // Things get a little tricky here. We have a pending external
        // suspend request, but we are holding the SR_lock so we
        // can't just self-suspend. So we temporarily drop the lock
        // and then self-suspend.
      }

      ThreadBlockInVM tbivm(this);
      java_suspend_self();

      // We're done with this suspend request, but we have to loop around
      // and check again. Eventually we will get SR_lock without a pending
      // external suspend request and will be able to mark ourselves as
      // exiting.
    }
    // no more external suspends are allowed at this point
  } else {
    // before_exit() has already posted JVMTI THREAD_END events
  }

  // Notify waiters on thread object. This has to be done after exit() is called
  // on the thread (if the thread is the last thread in a daemon ThreadGroup the
  // group should have the destroyed bit set before waiters are notified).
  ensure_join(this);
  assert(!this->has_pending_exception(), "ensure_join should have cleared");

  // 6282335 JNI DetachCurrentThread spec states that all Java monitors
  // held by this thread must be released.  A detach operation must only
  // get here if there are no Java frames on the stack.  Therefore, any
  // owned monitors at this point MUST be JNI-acquired monitors which are
  // pre-inflated and in the monitor cache.
  //
  // ensure_join() ignores IllegalThreadStateExceptions, and so does this.
  if (exit_type == jni_detach && JNIDetachReleasesMonitors) {
    assert(!this->has_last_Java_frame(), "detaching with Java frames?");
    ObjectSynchronizer::release_monitors_owned_by_thread(this);
    assert(!this->has_pending_exception(), "release_monitors should have cleared");
  }

  // These things needs to be done while we are still a Java Thread. Make sure that thread
  // is in a consistent state, in case GC happens
  assert(_privileged_stack_top == NULL, "must be NULL when we get here");

  if (active_handles() != NULL) {
    JNIHandleBlock* block = active_handles();
    set_active_handles(NULL);
    JNIHandleBlock::release_block(block);
  }

  if (free_handle_block() != NULL) {
    JNIHandleBlock* block = free_handle_block();
    set_free_handle_block(NULL);
    JNIHandleBlock::release_block(block);
  }

  // These have to be removed while this is still a valid thread.
  remove_stack_guard_pages();

  if (UseTLAB) {
    tlab().make_parsable(true);  // retire TLAB
  }

  if (JvmtiEnv::environments_might_exist()) {
    JvmtiExport::cleanup_thread(this);
  }

  // We must flush any deferred card marks before removing a thread from
  // the list of active threads.
  Universe::heap()->flush_deferred_store_barrier(this);
  assert(deferred_card_mark().is_empty(), "Should have been flushed");

#if INCLUDE_ALL_GCS
  // We must flush the G1-related buffers before removing a thread
  // from the list of active threads. We must do this after any deferred
  // card marks have been flushed (above) so that any entries that are
  // added to the thread's dirty card queue as a result are not lost.
  if (UseG1GC) {
    flush_barrier_queues();
  }
#endif // INCLUDE_ALL_GCS

  // Remove from list of active threads list, and notify VM thread if we are the last non-daemon thread
  Threads::remove(this);
}

线程终止的时候，会调用JavaThread::exit 方法，这个退出方法中，有个清理的工作：

  // Notify waiters on thread object. This has to be done after exit() is called
  // on the thread (if the thread is the last thread in a daemon ThreadGroup the
  // group should have the destroyed bit set before waiters are notified).
  ensure_join(this);

唤醒当前线程对象上的阻塞线程，这是在调用exit之后被完成的。
我们来看 ensure_join 方法具体做了些什么？

static void ensure_join(JavaThread* thread) {
  // We do not need to grap the Threads_lock, since we are operating on ourself.
  Handle threadObj(thread, thread->threadObj());
  assert(threadObj.not_null(), "java thread object must exist");
  ObjectLocker lock(threadObj, thread);
  // Ignore pending exception (ThreadDeath), since we are exiting anyway
  thread->clear_pending_exception();
  // Thread is exiting. So set thread_status field in  java.lang.Thread class to TERMINATED.
  java_lang_Thread::set_thread_status(threadObj(), java_lang_Thread::TERMINATED);
  // Clear the native thread instance - this makes isAlive return false and allows the join()
  // to complete once we've done the notify_all below
  java_lang_Thread::set_thread(threadObj(), NULL);
  lock.notify_all(thread);
  // Ignore pending exception (ThreadDeath), since we are exiting anyway
  thread->clear_pending_exception();
}

其中关键的是 lock.notify_all(thread); wait方法锁的是当前t1实例，t1退出的时候，拿到t1实例，然后拿到t1的锁，然后notify_all。

以上就是我们对thread.join()的全部解读。