Introduction

About python-iptables

Iptables is the tool that is used to manage netfilter, the standard packet filtering and manipulation framework under Linux. As the iptables manpage puts it:

Iptables is used to set up, maintain, and inspect the tables of IPv4 packet filter rules in the Linux kernel. Several different tables may be defined.

Each table contains a number of built-in chains and may also contain user- defined chains.

Each chain is a list of rules which can match a set of packets. Each rule specifies what to do with a packet that matches. This is called a target, which may be a jump to a user-defined chain in the same table.

Python-iptables provides a pythonesque wrapper via python bindings to iptables under Linux. Interoperability with iptables is achieved via using the iptables C libraries (libiptc, libxtables, and the iptables extensions), not calling the iptables binary and parsing its output. It is meant primarily for dynamic and/or complex routers and firewalls, where rules are often updated or changed, or Python programs wish to interface with the Linux iptables framework..

If you are looking for ebtables python bindings, check out python-ebtables.

Python-iptables supports Python 2.6, 2.7 and 3.4.

Installing via pip

The usual way:

pip install --upgrade python-iptables

Compiling from source

First make sure you have iptables installed (most Linux distributions install it by default). Python-iptables needs the shared libraries libiptc.so and libxtables.so coming with iptables, they are installed in /lib on Ubuntu.

You can compile python-iptables in the usual distutils way:

% cd python-iptables
% python setup.py build

If you like, python-iptables can also be installed into a virtualenv:

% mkvirtualenv python-iptables
% python setup.py install

If you install python-iptables as a system package, make sure the directory where distutils installs shared libraries is in the dynamic linker's search path (it's in /etc/ld.so.conf or in one of the files in the folder /etc/ld.co.conf.d). Under Ubuntu distutils by default installs into /usr/local/lib.

Now you can run the tests:

% sudo PATH=$PATH python setup.py test
WARNING: this test will manipulate iptables rules.
Don't do this on a production machine.
Would you like to continue? y/n y
[...]

The PATH=$PATH part is necessary after sudo if you have installed into a virtualenv, since sudo will reset your environment to a system setting otherwise..

Once everything is in place you can fire up python to check whether the package can be imported:

% sudo PATH=$PATH python
>>> import iptc
>>>

Of course you need to be root to be able to use iptables.

Using a custom iptables install

If you are stuck on a system with an old version of iptables, you can install a more up to date version to a custom location, and ask python-iptables to use libraries at that location.

To install iptables to /tmp/iptables:

% git clone git://git.netfilter.org/iptables && cd iptables
% ./autogen.sh
% ./configure --prefix=/tmp/iptables
% make
% make install

Make sure the dependencies iptables needs are installed.

Now you can point python-iptables to this install path via:

% sudo PATH=$PATH IPTABLES_LIBDIR=/tmp/iptables/lib XTABLES_LIBDIR=/tmp/iptables/lib/xtables python
>>> import iptc
>>>

What is supported

The basic iptables framework and all the match/target extensions are supported by python-iptables, including IPv4 and IPv6 ones. All IPv4 and IPv6 tables are supported as well.

Full documentation with API reference is available here.

Examples

High level abstractions

python-iptables implements a low-level interface that tries to closely match the underlying C libraries. The module iptc.easy improves the usability of the library by providing a rich set of high-level functions designed to simplify the interaction with the library, for example::

>>> import iptc
>>> iptc.easy.dump_table('nat', ipv6=False)
{'INPUT': [], 'OUTPUT': [], 'POSTROUTING': [], 'PREROUTING': []}
>>> iptc.easy.dump_chain('filter', 'OUTPUT', ipv6=False)
[{'comment': {'comment': 'DNS traffic to Google'},
  'dst': '8.8.8.8/32',
  'protocol': 'udp',
  'target': 'ACCEPT',
  'udp': {'dport': '53'}}]
>>> iptc.easy.add_chain('filter', 'TestChain')
True
>>> rule_d = {'protocol': 'tcp', 'target': 'ACCEPT', 'tcp': {'dport': '22'}}
>>> iptc.easy.insert_rule('filter', 'TestChain', rule_d)
>>> iptc.easy.dump_chain('filter', 'TestChain')
[{'protocol': 'tcp', 'target': 'ACCEPT', 'tcp': {'dport': '22'}}]
>>> iptc.easy.delete_chain('filter', 'TestChain', flush=True)

Rules

In python-iptables, you usually first create a rule, and set any source/destination address, in/out interface and protocol specifiers, for example:

>>> import iptc
>>> rule = iptc.Rule()
>>> rule.in_interface = "eth0"
>>> rule.src = "192.168.1.0/255.255.255.0"
>>> rule.protocol = "tcp"

This creates a rule that will match TCP packets coming in on eth0, with a source IP address of 192.168.1.0/255.255.255.0.

A rule may contain matches and a target. A match is like a filter matching certain packet attributes, while a target tells what to do with the packet (drop it, accept it, transform it somehow, etc). One can create a match or target via a Rule:

>>> rule = iptc.Rule()
>>> m = rule.create_match("tcp")
>>> t = rule.create_target("DROP")

Match and target parameters can be changed after creating them. It is also perfectly valid to create a match or target via instantiating them with their constructor, but you still need a rule and you have to add the matches and the target to their rule manually:

>>> rule = iptc.Rule()
>>> match = iptc.Match(rule, "tcp")
>>> target = iptc.Target(rule, "DROP")
>>> rule.add_match(match)
>>> rule.target = target

Any parameters a match or target might take can be set via the attributes of the object. To set the destination port for a TCP match:

>>> rule = iptc.Rule()
>>> rule.protocol = "tcp"
>>> match = rule.create_match("tcp")
>>> match.dport = "80"

To set up a rule that matches packets marked with 0xff:

>>> rule = iptc.Rule()
>>> rule.protocol = "tcp"
>>> match = rule.create_match("mark")
>>> match.mark = "0xff"

Parameters are always strings. You can supply any string as the parameter value, but note that most extensions validate their parameters. For example this:

>>> rule = iptc.Rule()
>>> rule.protocol = "tcp"
>>> rule.target = iptc.Target(rule, "ACCEPT")
>>> match = iptc.Match(rule, "state")
>>> chain = iptc.Chain(iptc.Table(iptc.Table.FILTER), "INPUT")
>>> match.state = "RELATED,ESTABLISHED"
>>> rule.add_match(match)
>>> chain.insert_rule(rule)

will work. However, if you change the state parameter:

>>> rule = iptc.Rule()
>>> rule.protocol = "tcp"
>>> rule.target = iptc.Target(rule, "ACCEPT")
>>> match = iptc.Match(rule, "state")
>>> chain = iptc.Chain(iptc.Table(iptc.Table.FILTER), "INPUT")
>>> match.state = "RELATED,ESTABLISHED,FOOBAR"
>>> rule.add_match(match)
>>> chain.insert_rule(rule)

python-iptables will throw an exception:

Traceback (most recent call last):
  File "state.py", line 7, in <module>
    match.state = "RELATED,ESTABLISHED,FOOBAR"
  File "/home/user/Projects/python-iptables/iptc/ip4tc.py", line 369, in __setattr__
    self.parse(name.replace("_", "-"), value)
  File "/home/user/Projects/python-iptables/iptc/ip4tc.py", line 286, in parse
    self._parse(argv, inv, entry)
  File "/home/user/Projects/python-iptables/iptc/ip4tc.py", line 516, in _parse
    ct.cast(self._ptrptr, ct.POINTER(ct.c_void_p)))
  File "/home/user/Projects/python-iptables/iptc/xtables.py", line 736, in new
    ret = fn(*args)
  File "/home/user/Projects/python-iptables/iptc/xtables.py", line 1031, in parse_match
    argv[1]))
iptc.xtables.XTablesError: state: parameter error -2 (RELATED,ESTABLISHED,FOOBAR)

Certain parameters take a string that optionally consists of multiple words. The comment match is a good example:

>>> rule = iptc.Rule()
>>> rule.src = "127.0.0.1"
>>> rule.protocol = "udp"
>>> rule.target = rule.create_target("ACCEPT")
>>> match = rule.create_match("comment")
>>> match.comment = "this is a test comment"
>>> chain = iptc.Chain(iptc.Table(iptc.Table.FILTER), "INPUT")
>>> chain.insert_rule(rule)

Note that this is still just one parameter value.

However, when a match or a target takes multiple parameter values, that needs to be passed in as a list. Let's assume you have created and set up an ipset called blacklist via the ipset command. To create a rule with a match for this set:

>>> rule = iptc.Rule()
>>> m = rule.create_match("set")
>>> m.match_set = ['blacklist', 'src']

Note how this time a list was used for the parameter value, since the set match match_set parameter expects two values. See the iptables manpages to find out what the extensions you use expect. See ipset for more information.

When you are ready constructing your rule, add them to the chain you want it to show up in:

>>> chain = iptc.Chain(iptc.Table(iptc.Table.FILTER), "INPUT")
>>> chain.insert_rule(rule)

This will put your rule into the INPUT chain in the filter table.

Chains and tables

You can of course also check what a rule's source/destination address, in/out inteface etc is. To print out all rules in the FILTER table:

>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> for chain in table.chains:
>>>     print "======================="
>>>     print "Chain ", chain.name
>>>     for rule in chain.rules:
>>>         print "Rule", "proto:", rule.protocol, "src:", rule.src, "dst:", \
>>>               rule.dst, "in:", rule.in_interface, "out:", rule.out_interface,
>>>         print "Matches:",
>>>         for match in rule.matches:
>>>             print match.name,
>>>         print "Target:",
>>>         print rule.target.name
>>> print "======================="

As you see in the code snippet above, rules are organized into chains, and chains are in tables. You have a fixed set of tables; for IPv4:

• FILTER,
• NAT,
• MANGLE and
• RAW.

For IPv6 the tables are:

• FILTER,
• MANGLE,
• RAW and
• SECURITY.

To access a table:

>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> print table.name
filter

To create a new chain in the FILTER table:

>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> chain = table.create_chain("testchain")

$ sudo iptables -L -n
[...]
Chain testchain (0 references)
target     prot opt source               destination

To access an existing chain:

>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> chain = iptc.Chain(table, "INPUT")
>>> chain.name
'INPUT'
>>> len(chain.rules)
10
>>>

More about matches and targets

There are basic targets, such as DROP and ACCEPT. E.g. to reject packets with source address 127.0.0.1/255.0.0.0 coming in on any of the eth interfaces:

>>> import iptc
>>> chain = iptc.Chain(iptc.Table(iptc.Table.FILTER), "INPUT")
>>> rule = iptc.Rule()
>>> rule.in_interface = "eth+"
>>> rule.src = "127.0.0.1/255.0.0.0"
>>> target = iptc.Target(rule, "DROP")
>>> rule.target = target
>>> chain.insert_rule(rule)

To instantiate a target or match, we can either create an object like above, or use the rule.create_target(target_name) and rule.create_match(match_name) methods. For example, in the code above target could have been created as:

>>> target = rule.create_target("DROP")

instead of:

>>> target = iptc.Target(rule, "DROP")
>>> rule.target = target

The former also adds the match or target to the rule, saving a call.

Another example, using a target which takes parameters. Let's mark packets going to 192.168.1.2 UDP port 1234 with 0xffff:

>>> import iptc
>>> chain = iptc.Chain(iptc.Table(iptc.Table.MANGLE), "PREROUTING")
>>> rule = iptc.Rule()
>>> rule.dst = "192.168.1.2"
>>> rule.protocol = "udp"
>>> match = iptc.Match(rule, "udp")
>>> match.dport = "1234"
>>> rule.add_match(match)
>>> target = iptc.Target(rule, "MARK")
>>> target.set_mark = "0xffff"
>>> rule.target = target
>>> chain.insert_rule(rule)

Matches are optional (specifying a target is mandatory). E.g. to insert a rule to NAT TCP packets going out via eth0:

>>> import iptc
>>> chain = iptc.Chain(iptc.Table(iptc.Table.NAT), "POSTROUTING")
>>> rule = iptc.Rule()
>>> rule.protocol = "tcp"
>>> rule.out_interface = "eth0"
>>> target = iptc.Target(rule, "MASQUERADE")
>>> target.to_ports = "1234"
>>> rule.target = target
>>> chain.insert_rule(rule)

Here only the properties of the rule decide whether the rule will be applied to a packet.

Matches are optional, but we can add multiple matches to a rule. In the following example we will do that, using the iprange and the tcp matches:

>>> import iptc
>>> rule = iptc.Rule()
>>> rule.protocol = "tcp"
>>> match = iptc.Match(rule, "tcp")
>>> match.dport = "22"
>>> rule.add_match(match)
>>> match = iptc.Match(rule, "iprange")
>>> match.src_range = "192.168.1.100-192.168.1.200"
>>> match.dst_range = "172.22.33.106"
>>> rule.add_match(match)
>>> rule.target = iptc.Target(rule, "DROP")
>>> chain = iptc.Chain(iptc.Table(iptc.Table.FILTER), "INPUT")
>>> chain.insert_rule(rule)

This is the python-iptables equivalent of the following iptables command:

# iptables -A INPUT -p tcp –destination-port 22 -m iprange –src-range 192.168.1.100-192.168.1.200 –dst-range 172.22.33.106 -j DROP

You can of course negate matches, just like when you use ! in front of a match with iptables. For example:

>>> import iptc
>>> rule = iptc.Rule()
>>> match = iptc.Match(rule, "mac")
>>> match.mac_source = "!00:11:22:33:44:55"
>>> rule.add_match(match)
>>> rule.target = iptc.Target(rule, "ACCEPT")
>>> chain = iptc.Chain(iptc.Table(iptc.Table.FILTER), "INPUT")
>>> chain.insert_rule(rule)

This results in:

$ sudo iptables -L -n
Chain INPUT (policy ACCEPT)
target     prot opt source               destination
ACCEPT     all  --  0.0.0.0/0            0.0.0.0/0            MAC ! 00:11:22:33:44:55

Chain FORWARD (policy ACCEPT)
target     prot opt source               destination

Chain OUTPUT (policy ACCEPT)
target     prot opt source               destination

Counters

You can query rule and chain counters, e.g.:

>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> chain = iptc.Chain(table, 'OUTPUT')
>>> for rule in chain.rules:
>>>         (packets, bytes) = rule.get_counters()
>>>         print packets, bytes

However, the counters are only refreshed when the underlying low-level iptables connection is refreshed in Table via table.refresh(). For example:

>>> import time, sys
>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> chain = iptc.Chain(table, 'OUTPUT')
>>> for rule in chain.rules:
>>>         (packets, bytes) = rule.get_counters()
>>>         print packets, bytes
>>> print "Please send some traffic"
>>> sys.stdout.flush()
>>> time.sleep(3)
>>> for rule in chain.rules:
>>>         # Here you will get back the same counter values as above
>>>         (packets, bytes) = rule.get_counters()
>>>         print packets, bytes

This will show you the same counter values even if there was traffic hitting your rules. You have to refresh your table to get update your counters:

>>> import time, sys
>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> chain = iptc.Chain(table, 'OUTPUT')
>>> for rule in chain.rules:
>>>         (packets, bytes) = rule.get_counters()
>>>         print packets, bytes
>>> print "Please send some traffic"
>>> sys.stdout.flush()
>>> time.sleep(3)
>>> table.refresh()  # Here: refresh table to update rule counters
>>> for rule in chain.rules:
>>>         (packets, bytes) = rule.get_counters()
>>>         print packets, bytes

What is more, if you add:

iptables -A OUTPUT -p tcp --sport 80
iptables -A OUTPUT -p tcp --sport 22

you can query rule and chain counters together with the protocol and sport(or dport), e.g.:

>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> chain = iptc.Chain(table, 'OUTPUT')
>>> for rule in chain.rules:
>>>         for match in rule.matches:
>>>             (packets, bytes) = rule.get_counters()
>>>             print packets, bytes, match.name, match.sport

Autocommit

Python-iptables by default automatically performs an iptables commit after each operation. That is, after you add a rule in python-iptables, that will take effect immediately.

It may happen that you want to batch together certain operations. A typical use case is traversing a chain and removing rules matching a specific criteria. If you do this with autocommit enabled, after the first delete operation, your chain's state will change and you have to restart the traversal. You can do something like this:

>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> removed = True
>>> chain = iptc.Chain(table, "FORWARD")
>>> while removed == True:
>>>     removed = False
>>>     for rule in chain.rules:
>>>         if rule.out_interface and "eth0" in rule.out_interface:
>>>             chain.delete_rule(rule)
>>>             removed = True
>>>             break

This is clearly not ideal and the code is not very readable. An alternative is to disable autocommits, traverse the chain, removing one or more rules, than commit it:

>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> table.autocommit = False
>>> chain = iptc.Chain(table, "FORWARD")
>>> for rule in chain.rules:
>>>     if rule.out_interface and "eth0" in rule.out_interface:
>>>         chain.delete_rule(rule)
>>> table.commit()
>>> table.autocommit = True

The drawback is that Table is a singleton, and if you disable autocommit, it will be disabled for all instances of that Table.

Easy rules with dictionaries

To simplify operations with python-iptables rules we have included support to define and convert Rules object into python dictionaries.

>>> import iptc
>>> table = iptc.Table(iptc.Table.FILTER)
>>> chain = iptc.Chain(table, "INPUT")
>>> # Create an iptc.Rule object from dictionary
>>> rule_d = {'comment': {'comment': 'Match tcp.22'}, 'protocol': 'tcp', 'target': 'ACCEPT', 'tcp': {'dport': '22'}}
>>> rule = iptc.easy.encode_iptc_rule(rule_d)
>>> # Obtain a dictionary representation from the iptc.Rule
>>> iptc.easy.decode_iptc_rule(rule)
{'tcp': {'dport': '22'}, 'protocol': 'tcp', 'comment': {'comment': 'Match tcp.22'}, 'target': 'ACCEPT'}

Known Issues

These issues are mainly caused by complex interaction with upstream's Netfilter implementation, and will require quite significant effort to fix. Workarounds are available.

• The hashlimit match requires explicitly setting hashlimit_htable_expire. See Issue #201.
• The NOTRACK target is problematic; use CT --notrack instead. See Issue #204.