python多线程编程(4): 死锁和可重入锁

From: http://www.cnblogs.com/holbrook/archive/2012/03/08/2385449.html

在线程间共享多个资源的时候，如果两个线程分别占有一部分资源并且同时等待对方的资源，就会造成死锁。尽管死锁很少发生，但一旦发生就会造成应用的停止响应。下面看一个死锁的例子：

# encoding: UTF-8
import threading
import time

class MyThread(threading.Thread):
    def do1(self):
        global resA, resB
        if mutexA.acquire():
             msg = self.name+' got resA'
             print msg
             
             if mutexB.acquire(1):
                 msg = self.name+' got resB'
                 print msg
                 mutexB.release()
             mutexA.release()
    def do2(self):
        global resA, resB
        if mutexB.acquire():
             msg = self.name+' got resB'
             print msg
             
             if mutexA.acquire(1):
                 msg = self.name+' got resA'
                 print msg
                 mutexA.release()
             mutexB.release()
 
    
    def run(self):
        self.do1()
        self.do2()
resA = 0
resB = 0

mutexA = threading.Lock()
mutexB = threading.Lock()

def test():
    for i in range(5):
        t = MyThread()
        t.start()
if __name__ == '__main__':
    test()

执行结果：

Thread-1 got resA
Thread-1 got resB
Thread-1 got resB
Thread-1 got resA
Thread-2 got resA
Thread-2 got resB
Thread-2 got resB
Thread-2 got resA
Thread-3 got resA
Thread-3 got resB
Thread-3 got resB
Thread-3 got resA
Thread-5 got resA
Thread-5 got resB
Thread-5 got resB
Thread-4 got resA

此时进程已经死掉。

可重入锁

更简单的死锁情况是一个线程“迭代”请求同一个资源，直接就会造成死锁：

import threading
import time

class MyThread(threading.Thread):
    def run(self):
        global num 
        time.sleep(1)

        if mutex.acquire(1):  
            num = num+1
            msg = self.name+' set num to '+str(num)
            print msg
            mutex.acquire()
            mutex.release()
            mutex.release()
num = 0
mutex = threading.Lock()
def test():
    for i in range(5):
        t = MyThread()
        t.start()
if __name__ == '__main__':
    test()

为了支持在同一线程中多次请求同一资源，python提供了“可重入锁”：threading.RLock。RLock内部维护着一个Lock和一个counter变量，counter记录了acquire的次数，从而使得资源可以被多次require。直到一个线程所有的acquire都被release，其他的线程才能获得资源。上面的例子如果使用RLock代替Lock，则不会发生死锁：

import threading
import time

class MyThread(threading.Thread):
    def run(self):
        global num 
        time.sleep(1)

        if mutex.acquire(1):  
            num = num+1
            msg = self.name+' set num to '+str(num)
            print msg
            mutex.acquire()
            mutex.release()
            mutex.release()
num = 0
mutex = threading.RLock()
def test():
    for i in range(5):
        t = MyThread()
        t.start()
if __name__ == '__main__':
    test()