~jdstrand/ufw/trunk

UFW
---
https://wiki.ubuntu.com/UbuntuFirewall


What's in a name?
-----------------
What does it mean?  It has come to mean 'Uncomplicated Firewall', but you
can change it to something more suitable if you want. If you like it, you might
pick 'Universal Firewall', or 'Ultimate Firewall'. If you are not a fan,
perhaps 'Unbearable Firewall'. Have fun!


Requirements
------------
python 2.6-2.7, 3.4-3.5 (known to work with 2.6.2+, 2.7.0+, 3.4+, 3.5*)*
iptables 1.4**
gettext
/proc filesystem support
Linux kernel configured with the following modules (not exhaustive):
  addrtype
  comment
  hl (IPv6)
  limit
  multiport
  recent
  conntrack***
  iptable_nat,ip6table_nat****

*    python2.5 and python3 < 3.4 are no longer supported
**   Systems with iptables below 1.4 will not have IPv6 application rule
     support. ufw will give a warning when users try to use this functionality,
     but ufw will otherwise work fine. ufw is known to work with iptables 1.3.8
     in this degraded mode.
***  As of 0.34, the 'conntrack' modules are used instead of 'state'
**** As of 0.36, ufw supports 'nat' for both IPv4 and IPv6

ufw has been widely tested on Linux 2.6.24 and higher kernels. You may also
use the check-requirements script in the tests/ directory to see if your
system has all the required iptables/netfilter functionality.


Install
-------
Users can install with:
# python ./setup.py install
$ python ./setup.py install --home=PREFIX

The interpreter used for setup.py is the one used for ufw. So if your system
python is 2.6, but you have python2.7 available, use something like:
$ python2.7 ./setup.py install --home=PREFIX

Distributions which install to a build directory for packaging can install
with:
$ python ./setup.py install --root=PREFIX

Eg:
$ python ./setup.py install --root=/tmp/ufw

When installing ufw from source, you will also need to integrate it into your
boot process for the firewall to start when you restart your system. Depending
on your needs, this can be as simple as adding the following to a startup
script (eg rc.local for systems that use it):

# /lib/ufw/ufw-init start

For systems that use SysV initscripts, an example script is provided in
doc/initscript.example. See doc/upstart.example for an Upstart example and
doc/systemd.example as a systemd example. Consult your distribution's
documentation for the proper way to modify your boot process.


Basic Layout
------------
/usr/sbin/ufw 			is the UI for people (have different backends)
/etc/defaults/ufw 		high level configuration
/etc/ufw/before[6].rules 	rules evaluated before UI added rules
/etc/ufw/after[6].rules 	rules evaluated after UI added rules
/lib/ufw/user[6].rules		UI added rules (not to be modified)
/etc/ufw/sysctl.conf 		kernel network tunables
/lib/ufw/ufw-init		start script


Usage
-----
ufw enable|disable 		turn firewall on and off (including at boot)
ufw default allow|deny		updates default policy
ufw logging on|off		updates backend logging (*.rules)
ufw status			displays firewall status (user.rules only)
ufw allow|deny|limit RULE	add RULE to firewall
ufw route allow|deny|limit RULE	add routing (FORWARD) RULE to firewall

See 'man ufw' and also Ubuntu's tutorial at:
http://doc.ubuntu.com/ubuntu/serverguide/C/firewall.html


Chains
------
ufw uses several chains to allow ease of use and flexibility. Control flow
through the various chains is (essentially) as follows:

INPUT ->
  ufw-before-logging-input ->
  ufw-before-input ->
    ufw-user-input ->
      ufw-user-logging-input (rule specific) ->
  ufw-after-input ->
  ufw-after-logging-input ->
  ufw-reject-input -> return to INPUT

OUTPUT ->
  ufw-before-logging-output ->
  ufw-before-output ->
    ufw-user-output ->
      ufw-user-logging-output (rule specific) ->
  ufw-after-output ->
  ufw-after-logging-output ->
  ufw-reject-output -> return to OUTPUT

FORWARD ->
  ufw-before-logging-forward ->
  ufw-before-forward ->
    ufw-user-forward ->
      ufw-user-logging-forward (not used) ->
  ufw-after-forward ->
  ufw-after-logging-forward ->
  ufw-reject-forward -> return to FORWARD

The 'before' chains are setup in 'before.rules', the 'after' chains in
'after.rules' and the 'user' chains are maintained by ufw. If an administrator
wants to add rules manually, the rules should be added to 'before.rules' and
'after.rules'. The 'reject' chains are used for when the default policy is
set to REJECT (because iptables does not support REJECT as a target at this
time). Keep in mind, when using REJECT as default policy, ufw may end up
rejecting rules that are added outside of ufw and after ufw is started.

There is some default configuration in both 'before.rules' and 'after.rules',
and this configuration is not displayed with 'ufw status' (but can always
be viewed with 'iptables -L -n' or 'iptables -L [chain] -n'. See the iptables
man page for details. There are also 3 chains (for both IPv4 and IPv6) that
can be used to immediately go to POLICY, which are mostly useful to avoid
logging (these chains are used in the default ufw after*.rules configuration to
avoid logging noisy services by default):
  ufw-skip-to-policy-input
  ufw-skip-to-policy-output
  ufw-skip-to-policy-forward

The primary chains are ufw-before-*, ufw-after-* and ufw-reject-*. The
treatment of iptables' built-in chains can be controlled with the
MANAGE_BUILTINS configuration option (in /etc/default/ufw). By default this is
set to 'no', which means that other than adding the primary chains, the
built-in chains will remain untouched. This also means that these primary
chains will stay in the table, even after disabling ufw. This is to make sure
that the primary chains don't move around other non-ufw rules and chains. To
completely flush the built-in chains with this configuration, you can use:

# /lib/ufw/ufw-init flush-all

Alternately, ufw may also take full control of the firewall by setting
MANAGE_BUILTINS=yes in /etc/defaults/ufw. This will flush all the built-in
rules and delete the non-built-in rules on start, stop and reload.

In addition to the above, the following tracking chains are setup last to
support stateful tracking of NEW connections when the default policy is set to
ACCEPT:
  ufw-track-input
  ufw-track-output
  ufw-track-forward

When default policy is set to ACCEPT, an '-m conntrack --ctstate NEW -j ACCEPT'
rule is added to the appropriate chain for tcp and udp connections, otherwise
the chain is empty and the default policy for the builtin chain.


Advanced Configuration
----------------------
ufw can be thought of two parts, the ufw command-line program and the ufw
framework. The ufw command is intentionally kept as simple as possible, so
users can do common tasks more easily. The framework (ie the bootscripts, setup
of the chains (see above), sysctl configuration, etc) is very flexible, and
since ufw is simply a frontend for iptables, anything that can be done with
iptables can be done within the ufw framework.

As an example, to perform port redirection, users can add to the top of
/etc/ufw/before.rules, before the '*filter' section:
*nat
:PREROUTING ACCEPT [0:0]
# redirect all incoming requests to tcp port 80 to tcp port 22
-A PREROUTING -p tcp --dport 80 -j REDIRECT --to-port 22
COMMIT

then run:
# ufw disable
# ufw enable
# ufw allow 80/tcp	(required only if ufw blocks requests to this port)


To add NAT masquerading to the above, change the nat table that was just added
to something like:
*nat
:PREROUTING ACCEPT [0:0]
:POSTROUTING ACCEPT [0:0]
-A PREROUTING -p tcp --dport 80 -j REDIRECT --to-port 22
# Forward traffic from eth1 through eth0.
-A POSTROUTING -s 192.168.0.0/24 -o eth0 -j MASQUERADE
COMMIT

then adjust /etc/default/ufw to have:
DEFAULT_FORWARD_POLICY="ACCEPT"

and /etc/ufw/sysctl.conf to have:
net.ipv4.ip_forward=1

then run:
# ufw disable
# ufw enable


It's important to remember that ufw will only flush the chains and tables it
manages, so if if you need to flush the nat table to restart anew, please do:
# iptables -F -t nat

Similarly, to see what rules are in the nat table's chains, use:
# iptables -L -n -t nat

See 'man iptables' for details.


Default ruleset
---------------
Enabling ufw creates a ruleset that is intended to protect the host while
allowing some common traffic such as DHCP, ping and mDNS. These defaults are
setup in the before*.rules and after*.rules files (see 'man iptables' for
terminology):
- Default DROP on INPUT
- Default DROP on FORWARD
- Default ACCEPT on OUTPUT
- ACCEPT all on lo
- DROP packets with RH0 headers
- ACCEPT all RELATED and ESTABLISHED on INPUT and OUTPUT
- ACCEPT all RELATED and ESTABLISHED on FORWARD (ip forwarding must be enabled
  via sysctl for this to be in effect)
- DROP INVALID packets (packets not associated with a known connection)
- ACCEPT certain icmp packets (INPUT and FORWARD):
  - destination-unreachable, source-quench, time-exceeded, parameter-problem,
    and echo-request for IPv4
  - destination-unreachable, packet-too-big, time-exceeded, parameter-problem,
    and echo-request
- ACCEPT certain icmpv6 packets for stateless autoconfiguration (INPUT):
  neighbor-solicitation, neighbor-advertisement, router-solicitation
- ACCEPT mDNS (zeroconf/bonjour/avahi 224.0.0.251 for IPv4 and ff02::fb for
  IPv6) for service discovery (INPUT)
- ACCEPT UPnP (239.255.255.250 for IPv4 and ff02::f for IPv6) for service
  discovery (INPUT)
- ACCEPT ping replies from IPv6 link-local (ffe8::/10) addresses (INPUT)
- ACCEPT DHCP client traffic (INPUT)
- Log all blocked packets not matching the default policy with rate limiting

If you are using a packaged version of ufw supplied by your distribution, the
default ruleset may be different.


Remote Management
-----------------
On /lib/ufw/ufw-init start and 'ufw enable' the chains are flushed, so
ssh may drop. This is needed so ufw is in a consistent state. Once the ufw is
'enabled' it will insert rules into the existing chains, and therefore not
flush the chains (but will when modifying a rule or changing the default
policy).

You can insert rules before enabling the firewall however, so it is often
a good idea to to:
# ufw allow proto tcp from any to any port 22
# ufw enable

In this case, the chains are still flushed, but the ssh port will be open
after enabling the firewall.


IPV6
----
ufw has full support for IPv6, and it is enabled by default. To disable, modify
/etc/default/ufw (or wherever this is installed) to have:

IPV6=no

Then do:
# ufw disable
# ufw enable


Application Integration
-----------------------
ufw has support for application integration. This allows for administrators
and developers to put profiles in /etc/ufw/applications.d and have users use
these profiles in their rules. Profiles use the .INI syntax, and examples
can be found in the examples/ directory. See 'man ufw' for details.


Upgrading
---------
If upgrading from 0.17 or below to 0.18, new chains to support the 'limit'
command will be added automatically.


Distributions
-------------
While it certainly ok to use /lib/ufw/ufw-init as the initscript for
ufw, this script is meant to be used by ufw itself, and therefore not
particularly user friendly. See doc/initscript.example for a simple
implementation that can be adapted to your distribution.

Simple rules for rsyslog support can be found in doc/rsyslog.example.


Testing
-------
$ mkdir -p /tmp/ufw/usr /tmp/ufw/etc
$ UFW_SKIP_CHECKS=1 python ./setup.py install --home=/tmp/ufw

(ufw does a number of checks since it is normally run as root. When testing,
this is usually not the case, so you can use UFW_SKIP_CHECKS=1 with setup.py
to disable these checks.)

Now do:
$ /tmp/ufw/usr/sbin/ufw help

Here is a command to do it all at once:
$ rm -rf /tmp/ufw && mkdir -p /tmp/ufw/usr /tmp/ufw/etc && UFW_SKIP_CHECKS=1 python ./setup.py install --home=/tmp/ufw

Then test with:
$ PYTHONPATH=$PYTHONPATH:/tmp/ufw/lib/python /tmp/ufw/usr/sbin/ufw ...

$ sudo sh -c "PYTHONPATH=$PYTHONPATH:/tmp/ufw/lib/python /tmp/ufw/usr/sbin/ufw ..."

Can also just run from the source directory:
$ ./run_tests.sh -s

You may also specify an interpreter for the tests. Eg:
$ ./run_tests.sh -s -i /usr/local/bin/python2.7

Or for the root tests (these are iptables version dependent, will modify your
existing firewall and insert kernel modules, so they require root privileges
and aren't run by default):
# ./run_tests.sh -s root

Finally, ufw's behavior may differ based on available kernel features. The
root_kern tests assume all kernel features supported by check-requirements
are enabled. They behave just like the root tests.


Unit Tests
----------
ufw unit tests are in tests/unit and can be called like so:
$ make unittest

This is equivalent to something like:
$ python ./tests/unit/runner.py

Alternatively, can install python-coverage and do:
$ make coverage         # like 'unittest', but with -m coverage
$ make coverage-report

To run an individual test, use:
$ python ./tests/unit/runner.py test_<individual test script>.py

Copyright 2008-2014 Canonical Ltd.