Python - 带套接字的多线程提供随机结果

Question

I am really confused about my problem right now. I want to discover an open port over a list of hosts (or subnet).

So first let's show what I've done so far..

from multiprocessing.dummy import Pool as ThreadPool 
from netaddr import IPNetwork as getAddrList
import socket, sys

this = sys.modules[__name__]

def threading(ip):
    sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
    sock.setsockopt(socket.SOL_SOCKET, socket.SO_REUSEADDR, 1)
    sock.settimeout(this.timeout)

    failCode = 0

    try:
        if sock.connect_ex((str(ip), this.port)) == 0:
            #port open
            this.count += 1
        else:
            #port closed/filtered
            failCode = 1
            pass
    except Exception: 
        #host unreachable
        failCode = 2
        pass
    finally:
        sock.close()

#set thread num
threads = 64
#set socket timeout
this.timeout = 1
#set ip list
ipList = getAddrList('8.8.8.0/24')
#set port
this.port = 53
#set count
this.count = 0

#threading
Pool = ThreadPool(threads) 
Pool.map_async(threading, ipList).get(9999999)
Pool.close()
Pool.join()

#result
print str(this.count)

The Script works fine without any error. But I'm struggling about what it prints out..

So if I want to scan for example the subnet 8.8.8.0/24 and discover port 53 . I know the only server that has an open dns port is 8.8.8.8 (google-dns).

But when I run my script serveral times the print str(this.count) will randomly (as it seems to me..) return 0 or 1 .

What I also know:

• Scan only 8.8.8.8 prints always 1
• Use only 1 thread for /24 prints always 1
• Use 256 threads for /24 prints randomly 0 and 1
• changing the timeout doesn't help

So it seems like it has to do with the threading option that causes lags on my computer. But note that my CPU usage is <10% and the network usage is <50%!

But there is also another thing I don't understand.. If print str(this.count) returns 0 I would normally think it is because the threads are in conflict with each other and the socket doesn't get a connection.. but that isn't true because if this.count equals 0 , the failCode is set to 1 (on 8.8.8.8 )! Which means the port is closed.. but that must be also a bug of my script. I cannot think that this is caused by a lag of the server.. it's google there are no lags..

So additionally we know:

• output of 0 is because the server respond that the port is closed
• and we are sure the port is definitely open

So I also think about that if I have many threads that run if sock.connect_ex((str(ip), this.port)) == 0: at the same time maybe the host 8.8.8.8 looks in the wrong answer value. Maybe it struggles and look in the response of 8.8.8.7 . ..hope you know what I mean.

**max_socket_connections is set to 512 on my linux distribution

Anybody experienced with socket threading and can explain me the situation or give me an answer how to prevent this?

Solved:

.. look at the answer ..

Answer 1

The first mistake was:

But note that my CPU usage is <10% and the network usage is <50%!

Actually it was a 400 % network usage. I expected bits instead of bytes.
Very embrassing..

And the second mistake was:

and we are sure the port is definitely open

Google does actually block the port after ~5 attempts in a short period of time. They release the restriction after a few seconds or minutes.