什么IPC机制用于在两个Python进程之间的multiprocessing.Queue中共享数据？

Question

Here is my Python code.

#!/usr/bin/env python
import multiprocessing
import time

def worker(q):
    while True:
        data = q.get()
        print 'worker: got: %d' % data
        if data == -1:
            print 'worker: done'
            break

def master():
    q = multiprocessing.Queue()
    p = multiprocessing.Process(target=worker, args=(q,))
    p.start()

    for i in range(5):
        q.put(i)
        print 'master: put: %d' % i
        time.sleep(1)

    q.put(-1)
    p.join()

    print 'master: done'

master()
print 'exiting ...'

Here is the output when I run this code on a Debian 9 GNU/Linux system.

$ python q.py
master: put: 0
worker: got: 0
master: put: 1
worker: got: 1
master: put: 2
worker: got: 2
master: put: 3
worker: got: 3
master: put: 4
worker: got: 4
worker: got: -1
worker: done
master: done
exiting ...

I am trying to find some evidence that the the master process and the worker process are communicating the data being put in the queue via a socket, message queue or shared memory. But I cannot seem to find any evidence in favour of it.

$ ps -ef | grep python | grep -v grep; netstat -nopa | grep python; ipcs
lone      2914  9836  2 12:54 pts/1    00:00:00 python q.py
lone      2915  2914  0 12:54 pts/1    00:00:00 python q.py
(Not all processes could be identified, non-owned process info
 will not be shown, you would have to be root to see it all.)

------ Message Queues --------
key        msqid      owner      perms      used-bytes   messages

------ Shared Memory Segments --------
key        shmid      owner      perms      bytes      nattch     status
0x00000000 163840     lone       600        393216     2          dest
0x00000000 262145     lone       600        393216     2          dest
0x00000000 360450     lone       600        393216     2          dest
0x00000000 142344195  lone       600        524288     2          dest
0x00000000 101941252  lone       600        4194304    2          dest
0x00000000 950277     lone       600        524288     2          dest
0x00000000 118194183  lone       600        696320     2          dest
0x00000000 118292488  lone       600        4153344    2          dest
0x00000000 117899273  lone       600        4153344    2          dest
0x00000000 118226954  lone       600        696320     2          dest
0x00000000 123535371  lone       600        86016      2          dest
0x00000000 123338764  lone       600        1728512    2          dest
0x00000000 123240461  lone       600        1798144    2          dest
0x00000000 123568144  lone       600        86016      2          dest
0x00000000 137330705  lone       600        32768      2          dest
0x00000000 137232402  lone       600        81920      2          dest
0x00000000 29098003   lone       600        4194304    2          dest
0x00000000 137265172  lone       600        81920      2          dest
0x00000000 137297941  lone       600        151552     2          dest
0x00000000 35258391   lone       600        393216     2          dest
0x00000000 35291160   lone       600        12288      2          dest
0x00000000 35323929   lone       600        12288      2          dest
0x00000000 35356698   lone       600        393216     2          dest
0x00000000 35520539   lone       600        12288      2          dest
0x00000000 35422236   lone       600        393216     2          dest
0x00000000 35455005   lone       600        12288      2          dest

------ Semaphore Arrays --------
key        semid      owner      perms      nsems

• How is the data put into multiprocessing.Queue transferred from the master process to the worker process?
• What commands can I run to find evidence for the IPC mechanism used to transfer data in multiprocessing.Queue from the master process to the worker process?

Answer 1

Python is open source, and you can see the implementation plainly here: https://github.com/python/cpython/blob/master/Lib/multiprocessing/queues.py

It uses a Pipe which is implemented here: https://github.com/python/cpython/blob/master/Lib/multiprocessing/connection.py

Which uses os.pipe() , implemented here: https://github.com/python/cpython/blob/master/Modules/posixmodule.c

Which uses pipe2() , documented here: http://man7.org/linux/man-pages/man2/pipe.2.html