关于 python 中死锁模拟的问题

Question

当我尝试实现一个python代码来模拟死锁时，我遇到了一些有趣的问题：

1）我用下面的代码来模拟死锁。

  1 from threading import *
  2 import time
  3
  4
  5 def thread_one(lock1, lock2):
  6     print("thread 1 is trying to acquire lock 1")
  7     lock1.acquire()
  8     print("lock1 acquired by thread 1")
  9     time.sleep(1)
 10     print("thread 1 is trying to acquire lock 2")
 11     lock2.acquire()
 12
 13
 14 def thread_two(lock1, lock2):
 15     print("thread 2 is trying to acquire lock 2")
 16     lock2.acquire()
 17     print("lock2 acquired by thread 2")
 18     time.sleep(1)
 19     print("thread 2 is trying to acquire lock 1")
 20     lock1.acquire()
 21
 22
 23 if __name__ == "__main__":
 24     lock1 = Lock()
 25     lock2 = Lock()
 26
 27     t1 = Thread(target=thread_one, args=(lock1, lock2))
 28     t2 = Thread(target=thread_two, args=(lock1, lock2))
 29
 30     t1.start()
 31     t2.start()
 32
 33     t1.join()
 34     t2.join()

这些是我的输出：

thread 1 is trying to acquire lock 1
lock1 acquired by thread 1
thread 2 is trying to acquire lock 2
lock2 acquired by thread 2
thread 1 is trying to acquire lock 2
thread 2 is trying to acquire lock 1
(program stuck here)

如有错误请指正。我认为我的模拟是正确的，两个线程卡在获取第二个锁的步骤。

2）然后我做了如下修改：

 10     print("thread 1 is trying to release lock 2")
 11     lock2.release()

 28     t2 = Thread(target=thread_one, args=(lock1, lock2))

基本上，我希望两个线程实例都运行同一个函数thread_one并且在函数中它试图释放尚未获取的锁2。然后我得到了这些输出：

thread 1 is trying to acquire lock 1
lock1 acquired by thread 1
thread 1 is trying to acquire lock 1
thread 1 is trying to release lock 2
Exception in thread Thread-1:
Traceback (most recent call last):
  File "/Users/bawang/.pyenv/versions/3.6.5/lib/python3.6/threading.py", line 916, in _bootstrap_inner
    self.run()
  File "/Users/bawang/.pyenv/versions/3.6.5/lib/python3.6/threading.py", line 864, in run
    self._target(*self._args, **self._kwargs)
  File "test.py", line 11, in thread_one
    lock2.release()
RuntimeError: release unlocked lock
(program stuck here)

我的问题是：为什么两个线程都挂在那里（我需要按两次 ctrl+c 来取消它们）？我知道第二个线程正在等待第一个线程释放 lock1。但是，为什么第一个线程抛出异常后就卡住了呢？

3) 然后我通过进行以下更改进一步移动：

  5 def thread_one(lock1, lock2):
  6     print("thread 1 is trying to acquire lock 1")
  7     lock1.acquire()
  8     print("lock1 acquired by thread 1")
  9     time.sleep(1)
 10     print("thread 1 is trying to release lock 2")
 11     lock2.release()
 12
 13
 14 def thread_two(lock1, lock2):
 15     print("thread 2 is trying to acquire lock 2")
 16     lock2.acquire()
 17     print("lock2 acquired by thread 2")
 18     time.sleep(1)
 19     print("thread 2 is trying to release lock 1")
 20     lock1.release()

 27     t1 = Thread(target=thread_one, args=(lock1, lock2))
 28     t2 = Thread(target=thread_two, args=(lock1, lock2))

这次我想看看如果lock1&lock2被一个线程获取而另一个线程释放会发生什么。这些是我的输出：

thread 1 is trying to acquire lock 1
lock1 acquired by thread 1
thread 2 is trying to acquire lock 2
lock2 acquired by thread 2
thread 1 is trying to release lock 2
thread 2 is trying to release lock 1
(Program completes)

我的问题是为什么没有例外？在这种情况下，我确实期望有两个 RuntimeError: release unlocked lock。

Answer 1

1) 是的。你是对的。

2) 如您所料，第一个线程已停止，但您还有两个线程：第二个线程和主线程。第一个 cntl + c 杀死主线程。您可以查看 KeybboardInterrupt 的消息。第一个发生在 t2.join().

3) 两个线程都正确获取和释放。不同的线程可以获得同一个锁。因此，线程1只是释放了线程2获得的锁2，反之亦然。

Answer 2

您的代码的执行顺序没有保证，因此线程 1 可能会在线程 2 获取锁 2 之前释放锁 2。但是您可以再次运行代码，线程 2 可以在线程 1 然后释放它之前获取它。我不确定您在示例中途尝试实现的目标。

最后一个例子没有产生任何异常的原因应该从输出中很清楚：

thread 1 is trying to acquire lock 1
lock1 acquired by thread 1
thread 2 is trying to acquire lock 2
lock2 acquired by thread 2
thread 1 is trying to release lock 2
thread 2 is trying to release lock 1

线程尝试做的所有事情，它们都会成功。由于您将相同的锁传递给两个线程，因此没有理由线程 1 无法获取锁 1 然后线程 2 释放它，而线程 2 获取锁 2 然后线程 1 释放它 -程序可以成功完成。

由于 sleep(1)，您几乎可以保证获得该输出，尽管不完全是 - 如果某些东西阻止线程 2 在一秒钟内启动，线程 1 可能仍会尝试释放锁 2 之前收购 - 依靠超时不是一种安全的编码方式。

关于 python 中死锁模拟的问题

Question about deadlock simulation in python

python

multithreading

deadlock