%PDF- %PDF-
Direktori : /backups/router/usr/local/opnsense/scripts/interfaces/ |
Current File : //backups/router/usr/local/opnsense/scripts/interfaces/capture.py |
#!/usr/local/bin/python3 """ Copyright (c) 2022 Ad Schellevis <ad@opnsense.org> All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. """ TEMP_DIR = '/tmp/captures/' import argparse import glob import subprocess import os import sys import ujson import zipfile from datetime import datetime def capture_pids(jobid): """ search for capture pids using output filename :param jobid: job uuid number :return: list of pids """ pids = [] args = ['/bin/pgrep', '-af', "%s%s" % (TEMP_DIR, jobid)] for line in subprocess.run(args, capture_output=True, text=True).stdout.split(): if line.isdigit(): pids.append(line) return pids def load_settings(filename): try: return ujson.load(open(filename, 'r')) except ValueError: return {} def pcap_reader(filename, eargs=None): args = ['/usr/sbin/tcpdump', '-n', '-e', '-tt', '-r', filename] if eargs: args = args + eargs sp = subprocess.Popen(args, stdout=subprocess.PIPE, stderr=subprocess.STDOUT) for line in sp.stdout: if not line.startswith(b'reading'): yield line.decode() sp.communicate() yield '\n' if __name__ == '__main__': result = dict() parser = argparse.ArgumentParser() parser.add_argument('--job', help='job id', default=None) parser.add_argument('--detail', help='detail level', choices=['normal', 'medium','high'], default='normal') parser.add_argument( 'action', help='action to perfom', choices=['list', 'start', 'stop', 'remove', 'view', 'archive'] ) cmd_args = parser.parse_args() all_jobs = {} if os.path.exists(TEMP_DIR): for filename in glob.glob("%s*.json" % TEMP_DIR): all_jobs[os.path.basename(filename).split('.')[0]] = filename if cmd_args.action == 'list': result['jobs'] = [] result['status'] = 'ok' for jobid in all_jobs: this_pids = capture_pids(jobid) settings = load_settings(all_jobs[jobid]) settings['id'] = jobid settings['status'] = "running" if len(this_pids) > 0 else "stopped" result['jobs'].append(settings) elif cmd_args.action == 'start' and cmd_args.job in all_jobs: this_pids = capture_pids(cmd_args.job) if len(this_pids) > 0: result['status'] = 'failed' result['status_msg'] = 'already active (pids: %s)' % ','.join(this_pids) else: result['status'] = 'ok' result['started_processes'] = 0 settings = load_settings(all_jobs[cmd_args.job]) for intf in settings.get('interface', '').split(','): result['started_processes'] += 1 args = [ '/usr/sbin/daemon', '-f', '/usr/sbin/tcpdump', '-i', intf, '-n', '-U', '-w', "%s%s_%s.pcap" % (TEMP_DIR, cmd_args.job, intf) ] if settings.get('promiscuous', '0') == '1': args.append('-p') if settings.get('snaplen', '').isdigit(): args.append('-s') args.append(settings.get('snaplen')) if settings.get('count', '').isdigit() and int(settings.get('count')) > 0: args.append('-c') args.append(settings.get('count')) filters = [] if settings.get('fam', 'any') != 'any': filters.append(settings.get('fam')) if settings.get('protocol', 'any') != 'any': if settings.get('protocol_not', '0') == '1': filters.append('proto not %s' % settings.get('protocol')) else: filters.append('proto %s' % settings.get('protocol')) if settings.get('host', '') != '': tokens = [] for token in settings.get('host').split(): if token.lower() in ['and', 'or', 'not']: tokens.append(token) elif token.find('/') > -1: tokens.append('net %s' % token) elif token.count(':') == 5 and sum([len(x) == 2 for x in token.split(':')]) == 6: tokens.append('ether host %s' % token) else: tokens.append('host %s' % token) filters.append('( %s )' % ' '.join(tokens)) if settings.get('port', '') != '': if settings.get('port_not', '0') == '1': filters.append('port not %s' % settings.get('port')) else: filters.append('port %s' % settings.get('port')) if len(filters) > 0: args.append(' and '.join(filters)) subprocess.run(args) elif cmd_args.action == 'stop' and cmd_args.job in all_jobs: result['status'] = 'ok' result['stopped_processes'] = 0 for pid in capture_pids(cmd_args.job): subprocess.run(['kill', pid]) result['stopped_processes'] += 1 elif cmd_args.action == 'remove' and cmd_args.job in all_jobs: result['status'] = 'ok' result['stopped_processes'] = 0 for pid in capture_pids(cmd_args.job): subprocess.run(['kill', pid]) result['stopped_processes'] += 1 for filename in glob.glob("%s%s*" % (TEMP_DIR, cmd_args.job)): os.remove(filename) elif cmd_args.action == 'view' and cmd_args.job in all_jobs: result['settings'] = load_settings(all_jobs[cmd_args.job]) result['status'] = 'ok' result['interfaces'] = {} for filename in glob.glob("%s%s*.pcap" % (TEMP_DIR, cmd_args.job)): intf = filename.split('_', 1)[-1].rsplit('.', 1)[0] result['interfaces'][intf] = {'rows': []} args = [] if cmd_args.detail == 'normal': args.append('-q') elif cmd_args.detail == 'medium': args.append('-v') elif cmd_args.detail == 'high': args.append('-vv') this_record = [] for line in pcap_reader(filename, args): if len(line) > 0 and not line[0] in [' ', '\t']: if this_record and this_record[0].count(' ') > 3: first_row = this_record[0].split() if first_row[0].replace('.', '').isdigit(): this_record[0] = this_record[0][this_record[0].find(',')+1:].lstrip() payload = { 'timestamp': datetime.fromtimestamp(float(first_row[0])).isoformat(), 'esrc': first_row[1] if first_row[1].count(':') == 5 else None, 'edst': first_row[3].rstrip(',') if first_row[3].count(':') == 5 > 0 else None, 'raw': ''.join(this_record) } if 'IPv4,' in first_row: payload['fam'] = 'ip' elif 'IPv6,' in first_row: payload['fam'] = 'ip6' result['interfaces'][intf]['rows'].append(payload) if len(result['interfaces'][intf]['rows']) > 500: # max frames reached break this_record = [line] else: this_record.append(line) elif cmd_args.action == 'archive' and cmd_args.job in all_jobs: result['filename'] = "%s%s.zip" % (TEMP_DIR, cmd_args.job) result['status'] = 'ok' zfh = zipfile.ZipFile(result['filename'], mode='w') for filename in glob.glob("%s%s*" % (TEMP_DIR, cmd_args.job)): if not filename.endswith('.zip'): zfh.write(filename, os.path.basename(filename)) zfh.close() print (ujson.dumps(result))