Manual: ufw-framework
UFW FRAMEWORK(8) Nov 2024 UFW FRAMEWORK(8)
NAME
ufw-framework - using the ufw framework
DESCRIPTION
ufw provides both a command line interface and a framework for managing a netfilter firewall. While the ufw command
provides an easy to use interface for managing a firewall, the ufw framework provides the administrator methods to
customize default behavior and add rules not supported by the command line tool. In this way, ufw can take full adâ
vantage of Linux netfilter's power and flexibility.
OVERVIEW
The framework provides boot time initialization, rules files for adding custom rules, a method for loading netfilter
modules, configuration of kernel parameters and configuration of IPv6. The framework consists of the following files:
/lib/ufw/ufw-init
initialization script
/etc/ufw/before.init
initialization customization script run before ufw is initialized
/etc/ufw/after.init
initialization customization script run after ufw is initialized
/etc/ufw/before[6].rules
rules file containing rules evaluated before UI added rules
/etc/ufw/user[6].rules
rules file containing UI added rules (managed with the ufw command)
/etc/ufw/after[6].rules
rules file containing rules evaluated after UI added rules
/etc/default/ufw
high level configuration
/etc/ufw/sysctl.conf
kernel network tunables
/etc/ufw/ufw.conf
additional high level configuration
BOOT INITIALIZATION
ufw is started on boot with /lib/ufw/ufw-init. This script is a standard SysV style initscript used by the ufw comâ
mand and should not be modified. The /etc/before.init and /etc/after.init scripts may be used to perform any addiâ
tional firewall configuration that is not yet supported in ufw itself and if they exist and are executable, ufw-init
will execute these scripts. ufw-init will exit with error if either of these scripts exit with error. ufw-init supâ
ports the following arguments:
start: loads the firewall
stop: unloads the firewall
restart:
reloads the firewall
force-reload:
same as restart
status:
basic status of the firewall
force-stop:
same as stop, except does not check if the firewall is already loaded
flush-all:
flushes the built-in chains, deletes all non-built-in chains and resets the policy to ACCEPT
ufw-init will call before.init and after.init with start, stop, status and flush-all, but typically, if used, these
scripts need only implement start and stop.
ufw uses many user-defined chains in addition to the built-in iptables chains. If MANAGE_BUILTINS in /etc/default/ufw
is set to 'yes', on stop and reload the built-in chains are flushed. If it is set to 'no', on stop and reload the ufw
secondary chains are removed and the ufw primary chains are flushed. In addition to flushing the ufw specific chains,
it keeps the primary chains in the same order with respect to any other user-defined chains that may have been added.
This allows for ufw to interoperate with other software that may manage their own firewall rules.
To ensure your firewall is loading on boot, you must integrate this script into the boot process. Consult your disâ
tribution's documentation for the proper way to modify your boot process if ufw is not already integrated.
RULES FILES
ufw is in part a front-end for iptables-restore, with its rules saved in /etc/ufw/before.rules, /etc/ufw/after.rules
and /etc/ufw/user.rules. Administrators can customize before.rules and after.rules as desired using the standard iptâ
ables-restore syntax. Rules are evaluated as follows: before.rules first, user.rules next, and after.rules last. IPv6
rules are evaluated in the same way, with the rules files named before6.rules, user6.rules and after6.rules. Please
note that ufw status only shows rules added with ufw and not the rules found in the /etc/ufw rules files.
Important: ufw only uses the *filter table by default. You may add any other tables such as *nat, *raw and *mangle as
desired. For each table a corresponding COMMIT statement is required.
After modifying any of these files, you must reload ufw for the rules to take effect. See the EXAMPLES section for
common uses of these rules files.
MODULES
Netfilter has many different connection tracking modules. These modules are aware of the underlying protocol and alâ
low the administrator to simplify his or her rule sets. You can adjust which netfilter modules to load by adjusting
IPT_MODULES in /etc/default/ufw. Some popular modules to load are:
nf_conntrack_ftp
nf_nat_ftp
nf_conntrack_irc
nf_nat_irc
nf_conntrack_netbios_ns
nf_conntrack_pptp
nf_conntrack_tftp
nf_nat_tftp
nf_conntrack_sane
Unconditional loading of connection tracking modules (nf_conntrack_*) in this manner is deprecated. ufw continues to
support the functionality but new configuration should only contain the specific modules required for the site. For
more information, see CONNECTION HELPERS.
KERNEL PARAMETERS
ufw will read in /etc/ufw/sysctl.conf on boot when enabled. Please note that /etc/ufw/sysctl.conf overrides values
in the system systcl.conf (usually /etc/sysctl.conf). Administrators can change the file used by modifying /etc/deâ
fault/ufw.
IPV6
IPv6 is enabled by default. When disabled, all incoming, outgoing and forwarded packets are dropped, with the excepâ
tion of traffic on the loopback interface. To adjust this behavior, set IPV6 to 'yes' in /etc/default/ufw. See the
ufw manual page for details.
EXAMPLES
As mentioned, ufw loads its rules files into the kernel by using the iptables-restore and ip6tables-restore commands.
Users wanting to add rules to the ufw rules files manually must be familiar with these as well as the iptables and
ip6tables commands. Below are some common examples of using the ufw rules files. All examples assume IPv4 only and
that DEFAULT_FORWARD_POLICY in /etc/default/ufw is set to DROP.
IP Masquerading
To allow IP masquerading for computers from the 10.0.0.0/8 network on eth1 to share the single IP address on eth0:
Edit /etc/ufw/sysctl.conf to have:
net.ipv4.ip_forward=1
Add to the end of /etc/ufw/before.rules, after the *filter section:
*nat
:POSTROUTING ACCEPT [0:0]
-A POSTROUTING -s 10.0.0.0/8 -o eth0 -j MASQUERADE
COMMIT
If your firewall is using IPv6 tunnels or 6to4 and is also doing NAT, then you should not usually masquerade protocol
'41' (ipv6) packets. For example, instead of the above, /etc/ufw/before.rules can be adjusted to have:
*nat
:POSTROUTING ACCEPT [0:0]
-A POSTROUTING -s 10.0.0.0/8 ! --protocol 41 -o eth0 -j MASQUERADE
COMMIT
Add the ufw route to allow the traffic:
ufw route allow in on eth1 out on eth0 from 10.0.0.0/8
Port Redirections
To forward tcp port 80 on eth0 to go to the webserver at 10.0.0.2:
Edit /etc/ufw/sysctl.conf to have:
net.ipv4.ip_forward=1
Add to the end of /etc/ufw/before.rules, after the *filter section:
*nat
:PREROUTING ACCEPT [0:0]
-A PREROUTING -p tcp -i eth0 --dport 80 -j DNAT \
--to-destination 10.0.0.2:80
COMMIT
Add the ufw route rule to allow the traffic:
ufw route allow in on eth0 to 10.0.0.2 port 80 proto tcp
Egress filtering
To block RFC1918 addresses going out of eth0:
Add the ufw route rules to reject the traffic:
ufw route reject out on eth0 to 10.0.0.0/8
ufw route reject out on eth0 to 172.16.0.0/12
ufw route reject out on eth0 to 192.168.0.0/16
Full example
This example combines the other examples and demonstrates a simple routing firewall. Warning: this setup is only an
example to demonstrate the functionality of the ufw framework in a concise and simple manner and should not be used
in production without understanding what each part does and does not do. Your firewall will undoubtedly want to be
less open.
This router/firewall has two interfaces: eth0 (Internet facing) and eth1 (internal LAN). Internal clients have adâ
dresses on the 10.0.0.0/8 network and should be able to connect to anywhere on the Internet. Connections to port 80
from the Internet should be forwarded to 10.0.0.2. Access to ssh port 22 from the administrative workstation
(10.0.0.100) to this machine should be allowed. Also make sure no internal traffic goes to the Internet.
Edit /etc/ufw/sysctl.conf to have:
net.ipv4.ip_forward=1
Add to the end of /etc/ufw/before.rules, after the *filter section:
*nat
:PREROUTING ACCEPT [0:0]
:POSTROUTING ACCEPT [0:0]
-A PREROUTING -p tcp -i eth0 --dport 80 -j DNAT \
--to-destination 10.0.0.2:80
-A POSTROUTING -s 10.0.0.0/8 -o eth0 -j MASQUERADE
COMMIT
Add the necessary ufw rules:
ufw route reject out on eth0 to 10.0.0.0/8
ufw route reject out on eth0 to 172.16.0.0/12
ufw route reject out on eth0 to 192.168.0.0/16
ufw route allow in on eth1 out on eth0 from 10.0.0.0/8
ufw route allow in on eth0 to 10.0.0.2 port 80 proto tcp
ufw allow in on eth1 from 10.0.0.100 to any port 22 proto tcp
CONNECTION HELPERS
Various protocols require the use of netfilter connection tracking helpers to group related packets into RELATED
flows to make rulesets clearer and more precise. For example, with a couple of kernel modules and a couple of rules,
a ruleset could simply allow a connection to FTP port 21, then the kernel would examine the traffic and mark the
other FTP data packets as RELATED to the initial connection.
When the helpers were first introduced, one could only configure the modules as part of module load (eg, if your FTP
server listened on a different port than 21, you'd have to load the nf_conntrack_ftp module specifying the correct
port). Over time it was understood that unconditionally using connection helpers could lead to abuse, in part because
some protocols allow user specified data that would allow traversing the firewall in undesired ways. As of kernel
4.7, automatic conntrack helper assignment (ie, handling packets for a given port and all IP addresses) is disabled
(the old behavior can be restored by setting net/netfilter/nf_conntrack_helper=1 in /etc/ufw/sysctl.conf). Firewalls
should now instead use the CT target to associate traffic with a particular helper and then set RELATED rules to use
the helper. This allows sites to tailor the use of helpers and help avoid abuse.
In general, to use helpers securely, the following needs to happen:
1. net/netfilter/nf_conntrack_helper should be set to 0 (default)
2. create a rule for the start of a connection (eg for FTP, port 21)
3. create a helper rule to associate the helper with this connection
4. create a helper rule to associate a RELATED flow with this connection
5. if needed, add the corresponding nf_conntrack_* module to IPT_MODULES
6. optionally add the corresponding nf_nat_* module to IPT_MODULES
In general it is desirable to make connection helper rules as specific as possible and ensure anti-spoofing is corâ
rectly setup for your site to avoid security issues in your ruleset. For more information, see ANTI-SPOOFING, above,
and .
Currently helper rules must be managed in via the RULES FILES. A future version of ufw will introduce syntax for
working with helper rules.
NOTES
When using ufw with libvirt and bridging, packets may be blocked. The libvirt team recommends that the following
sysctl's be set to disable netfilter on the bridge:
net.bridge.bridge-nf-call-ip6tables = 0
net.bridge.bridge-nf-call-iptables = 0
net.bridge.bridge-nf-call-arptables = 0
Note that the bridge module must be loaded in to the kernel before these values are set. One way to ensure this works
properly with ufw is to add 'bridge' to IPT_MODULES in /etc/default/ufw, and then add the above rules to
/etc/ufw/sysctl.conf.
Alternatively to disabling netfilter on the bridge, you can configure iptables to allow all traffic to be forwarded
across the bridge. Eg, add to /etc/ufw/before.rules within the *filter section:
-I FORWARD -m physdev --physdev-is-bridged -j ACCEPT
When using ufw with LXD or Incus, additional rules for the bridge (by default, lxdbr0) are needed for the guests for
DHCP, DNS and outbound connections. Eg:
ufw allow in on lxdbr0 to any port 67 proto udp
ufw allow in on lxdbr0 to any port 53
ufw route allow in on lxdbr0 from XXX.YYY.ZZZ.0/24
The first two rules limit access from the guests to the host to only DHCP and DNS while the 3rd rule allows access
from the guests on the XXX.YYY.ZZZ.0/24 subnet (as configured by LXD or Incus) to the world. Adjust as necessary for
your environment (site requirements, name of the bridge, bridge network, etc).
To allow traffic from the outside to your guests, then additional rules are needed. Eg, to allow HTTPS to the
10.210.254.123 guest:
ufw route allow out on lxdbr0 to 10.210.254.123 port 443 proto tcp
Note: when thinking of 'in on lxdbr0' vs 'out on lxdbr0', think of the traffic in the context of the host and the
guest: if the traffic is from the guest into or through the host, then an 'in on lxdbr0' should be used; if the trafâ
fic is from anywhere out to the guest, then an 'out on lxdbr0' should be used. As always, additional rules may be
needed depending on your default policy.
SEE ALSO
ufw(8), iptables(8), ip6tables(8), iptables-restore(8), ip6tables-restore(8), sysctl(8), sysctl.conf(5)
AUTHOR
ufw is Copyright 2008-2024, Canonical Ltd.
Nov 2024 UFW FRAMEWORK(8)