在Doug lea的那本著名的《Java并发编程—设计原则与模式》,英文名" Concurrent Programming in Java™: Design Principles and Patterns, Second Edition",书中提到可以用信号量Semaphore实现互斥锁Mutex。虽然java中是通过synchronize关键字提供锁,并用这个基础设施实现信号量的。在有的系统中只有信号量这一原语,锁是通过信号量实现的。代码如下:
import java.util.concurrent.Semaphore;
public class Mutex ...{ private Semaphore s = new Semaphore(1);
public void acquire() throws InterruptedException ...{ s.acquire(); } public void release()...{ s.release(); } public boolean attempt(int ms) throws InterruptedException ...{ return s.tryAcquire(ms); } }
上面的代码只能在java5中编译通过,因为Semaphore是在java5中才提供的。我在读上面的代码时有疑问。因为如果错误的连续调用release两次,然后两个线程都调用acquire,岂不是这两个线程都可以同时运行,从而违背了互斥锁的定义?为了证明我的猜测,写了如下的代码:
public class TestMutex ...{ public static void main(String[] args) throws InterruptedException...{ Mutex mutex=new Mutex(); mutex.acquire(); mutex.release(); mutex.release(); new MyThread(mutex).start(); new MyThread(mutex).start(); }
}
class MyThread extends Thread...{ private Mutex mutex;
public MyThread(Mutex mutex) ...{ this.mutex=mutex; }
public void run()...{ try ...{ mutex.acquire(); } catch (InterruptedException e1) ...{ throw new RuntimeException(e1); } for(int i=0;i<10;i++)...{ System.out.print(i); if(i%3==0)...{ try ...{ Thread.sleep(100); } catch (InterruptedException e) ...{ e.printStackTrace(); } } } mutex.release(); } } 该程序的输出如下: 00123123456456789789 从而证实了我的猜测。 |