I want to write simple port scanner. Do it like follows:
class Scanner(object):
def __init__(self, addr=None):
self._addr = addr if addr else '127.0.0.1'
self._lock = Lock()
self._opened_ports = []
def __enter__(self):
return self
def __exit__(self, exc_type, exc_val, exc_tb):
return True if not exc_val else False
def _scan_ports(self, thread_num):
from_port = 1000 * thread_num
ports_count = 1000
for port in range(ports_count):
try:
sock = socket.socket(socket.AF_INET, socket.SOCK_STREAM)
sock.connect((self._addr, from_port + port))
with self._lock:
self._opened_ports.append(from_port + port)
except:
pass
finally:
sock.close()
def scan(self, num_threads=None):
if not num_threads:
num_threads = len(os.sched_getaffinity(0))
pool = ThreadPool(num_threads)
thread_numbers = [thread_num for thread_num in range(num_threads)]
pool.map(self._scan_ports, thread_numbers)
pool.close()
pool.join()
return sorted(self._opened_ports)
As one can see there I'm trying to connect sockets to ports in parallel. So I run it in multithreaded mode like this:
with Scanner() as scanner:
ports = scanner.scan()
On my core i7 with 8 cores it occupies about 200 ms. If I do the same on 1 thread like this:
with Scanner() as scanner:
ports = scanner.scan(1)
it occupies 100 ms.
I don't understand why I get such result. Why scanning 1000 ports in 1 thread requires 2 times less time then scanning 1000 ports in 8 threads? Explain me please.
UPD if I replace socket stuff with simple time.sleep it will work as i expect - multithreaded mode occupies the same time as singlethreaded, so what is bad with sockets?
