【Python实战】实时获取tcpdump输出
Python实时获取tcpdump输出。
一、背景
今天有个小需求,要确认客户端有没有往服务端发送udp包,但为了减轻工作量,不想每次到机器上手动执行tcpdump抓包命令。
于是就写了个脚本来释放人力。
二、代码实现
整个脚本我还加了一些其他功能:时间戳、发送端IP提取,数据包分析,数据持久化等。这里都先去掉,仅记录下简单的实时获取tcpdump输出功能。
代码如下:
# -*- coding: utf-8 -*-
# !/usr/bin/env python
# sudo tcpdump -tt -l -nn -c 5 -i enp4s0 udp port 514 or 51414
import subprocess
cmd = ['sudo', 'tcpdump', '-tt', '-l', '-nn', '-c', '5', '-i', 'enp4s0', 'udp', 'port', '514', 'or', '51414']
proc = subprocess.Popen(cmd, stdout=subprocess.PIPE)
while True:
line = proc.stdout.readline()
line = line.strip()
if not line:
print('tcpdump finished...')
break
print(line)
输出如下(实时):
wenyuanblog@localhost:/home/test/script# python tcpdump_udp.py
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on enp4s0, link-type EN10MB (Ethernet), capture size 262144 bytes
1499774951.124178 IP 192.168.10.210.41974 > 192.168.10.251.514: UDP, length 139
1499774953.125664 IP 192.168.10.210.54995 > 192.168.10.251.51414: UDP, length 139
1499774956.128498 IP 192.168.10.210.56748 > 192.168.10.251.514: UDP, length 139
1499774958.129918 IP 192.168.10.210.53883 > 192.168.10.251.51414: UDP, length 139
1499774961.132921 IP 192.168.10.210.58803 > 192.168.10.251.514: UDP, length 139
5 packets captured
6 packets received by filter
0 packets dropped by kernel
tcpdump finished...
以上代码相当于手动执行了 sudo tcpdump -tt -l -nn -c 5 -i enp4s0 udp port 514 or 51414
这条命令。
注意参数-l很重要(行显)。
三、代码实现(更新)
上面的代码能实现tcpdump的功能,但是有一个问题:没有做超时保护。即当程序执行时间过长时kill该进程(这里使用ctrl+c的方式)。
要实现这个功能有很多种方案,例如定时器+多线程等,这里仅演示一种方案,代码如下:
# -*- coding: utf-8 -*-
# !/usr/bin/env python
# sudo tcpdump -tt -l -nn -c 50 -i enp4s0 udp port 514 or 51414
import subprocess
import signal
import time
import os
import re
import json
class CmdServer:
def __init__(self, cmd, timeout=120):
'''
:param cmd: 执行命令(列表形式)
:param timeout: 任务超时时间(seconds,进程运行超过该时间,kill该进程)
:param taskname: 任务名称(根据该任务名称记录命令输出信息)
'''
self.cmd = cmd
self.timeout = timeout
self.base_path = reduce(lambda x, y: os.path.dirname(x), range(1), os.path.abspath(__file__))
self.output_path = os.path.join(self.base_path, 'data.json')
self.udp_flow_list = []
self.begin_time = int(time.time())
# 执行tcpdump任务
def run(self):
if os.path.exists(self.output_path):
with open(self.output_path, 'r') as f:
self.udp_flow_list = json.load(f)
proc = subprocess.Popen(self.cmd, stdout=subprocess.PIPE)
stdout = ''
while proc.poll() == None:
current_time = int(time.time())
if current_time - self.begin_time >= self.timeout:
print('tcpdump timeout...')
proc.send_signal(signal.SIGINT)
stdout = proc.stdout.read()
if proc.poll() is not None and not stdout:
print('tcpdump finished...')
stdout = proc.stdout.read()
stdout_list = stdout.split('\n')
if stdout_list:
self._merge_data(stdout_list)
self._save_data()
# 数据合并(新增/更新)
def _merge_data(self, stdout_list):
for line in stdout_list:
line = line.strip()
if not line:
continue
timestamp = int(float(line.split('IP')[0].strip())) * 1000
# 源
src_ip_port_list = re.findall(r'IP(.+?)>', line)
if not src_ip_port_list:
continue
src_ip_port_str = src_ip_port_list[0].strip()
src_ip = '.'.join(src_ip_port_str.split('.')[0:4])
# 目的
dst_ip_port_list = re.findall(r'>(.+?):', line)
if not dst_ip_port_list:
continue
dst_ip_port_str = dst_ip_port_list[0].strip()
dst_port = dst_ip_port_str.split('.')[-1]
# 新增/更新latest_timestamp
src_item = filter(lambda x: src_ip == x['src_ip'], self.udp_flow_list)
if src_item:
src_item[0]['dst_port'] = dst_port
src_item[0]['latest_timestamp'] = timestamp
else:
self.udp_flow_list.append(dict(
src_ip=src_ip,
dst_port=dst_port,
latest_timestamp=timestamp
))
# 保存数据
def _save_data(self):
# 写入文件
with open(self.output_path, 'w') as f:
json.dump(self.udp_flow_list, f, encoding="utf-8", ensure_ascii=False)
if __name__ == '__main__':
cmd = ['sudo', 'tcpdump', '-tt', '-l', '-nn', '-c', '5', '-i', 'enp4s0', 'udp', 'port', '514', 'or', '51414']
cmd_server = CmdServer(cmd, 10)
cmd_server.run()
四、总结
比较简单,仅仅是记录下。如果想基于Python的tcpdump做一些业务上的逻辑,可以参考下面的“参考链接”。
参考
https://blog.csdn.net/wskzgz/article/details/83822780
https://blog.csdn.net/wangqiuyun/article/details/46966839
http://www.cnblogs.com/idvcn/p/8716066.html
https://blog.csdn.net/kobeyan/article/details/4344192
https://blog.csdn.net/xhw88398569/article/details/48022967