HOWTO Setup iptables

Kernel Configuration

NOTE This configuration is for basic firewalling only; we don't do NAT/packet-forwarding... so, if you're reading this, and wish to use NAT/forwarding, you will be missing a few key configuration items :-O

   Typical kernel 2.6.22 and higher kernel options for our very-basic firewalling:

Networking  ---->
 Networking options  ---->
  Network packet filtering framework (Netfilter)--->
   Core Netfilter Configuration ---->
    ["enable"] Netfilter connection tracking support--->Layer 3 Independent Connection tracking
    ["enable"] Netfilter Xtables support (required for ip_tables)
    ["enable"] "state" match support 
   IP: Netfilter Configuration --->
    ["enable"] IPv4 connection tracking support (required for NAT) required by "Layer 3 Independent Connection tracking" above (caused many headaches)
    ["enable"] IP tables support (required for filtering/masq/NAT)
    ["enable"]   Packet Filtering
    ["enable"]     REJECT target support

Iptables Installation

hostname ~ # emerge iptables
hostname ~ # rc-update add iptables default

Usually, when you try to start a new installation of iptables, you get an error, sometimes like this:

 hostname ~ # /etc/init.d/iptables start
* Not starting iptables.  First create some rules then run:
* /etc/init.d/iptables save

Or, you may see this kind of error:

FATAL: Module ip_tables not found.
iptables v1.3.8: can't initialize iptables table `filter': iptables who? (do you need to insmod?)
Perhaps iptables or your kernel needs to be upgraded.

Occasionally, there are complaints about "mangle"... it may be that there is some previous ruleset which included NAT/forwarding. For our application, we do not need mangling-capability!
In any of the above startup cases, we will manually start iptables (that is, not using the init-script), and give it a very-simple command-line rule, just to get iptables going:

hostname ~ # /sbin/iptables -A INPUT -i lo -j ACCEPT

This usually (should!) result in very simple, default "all-pass" ruleset, which doesn't actually do anything except keep the init-scripts happy:

hostname ~ # iptables -L
Chain INPUT (policy ACCEPT)
target     prot opt source               destination
ACCEPT     all  --  anywhere             anywhere

Chain FORWARD (policy ACCEPT)
target     prot opt source               destination

Chain OUTPUT (policy ACCEPT)
target     prot opt source               destination

We then use the init-script to save our ruleset, which will at least then allow iptables to start properly the next time with the init-scripts:

hostname ~ # /etc/init.d/iptables save
* Saving iptables state ...

Scripting the Rules

Once iptables is up-and-running, simply execute one of the scripts below, to implement the policies, Or, use these scripts as a basis for your own similar-but-custom rulesets, which Google can help you with. Our standard is to save the ruleset-script as /etc/iptables.bak. So, once iptables is running (in very-basic-form), invoking the ruleset-script is very simple:

hostname ~ # sh /etc/iptables.bak

Creating iptables Rulesets

To develop rulesets, it's often a great starting-point to see what ports are currently in-use, and reckon that these are may need to be opened through our firewall, in order for the server to perform it's intended function. So, let's see what ports are in-use, and by what programs/services:

hostname ~ # netstat -alnp

On a complex server, there can be many, many unfamiliar ports open, and some may vary with each invocation of the associated program :-( Google around, and first see if you can identify all the services. Then, see if you can nail down these services to always use a known/specific port. Some examples are provided below, which may help.
In a worst-case scenario, you may decide it's too scary to actually begin blocking ports and breaking services to users. While this is a crummy approach, it can work OK with fail2ban! So, you will gain some protection anyway, against dictionary-attacks. This basically starts with our all-pass default filter (shown above), and along with the fail2ban script invocation adding a few extra lines for you, to perform it's blocking.

Example 1 - very-basic /etc/iptables.bak to prevent breaking extensive/weird/complex services

Example 2 - /etc/iptables.bak for a web-server with vsftpd upload, also SSH connectivity, and being monitored by nagios

#! /bin/sh
# /etc/iptables.bak

# Let's save typing & confusion with variables
IPTABLES=/sbin/iptables

# Flush active rules and custom tables
$IPTABLES --flush
$IPTABLES --delete-chain

# set the defaults so that by-default incoming packets are dropped, unless explicitly allowed;
# for a desktop workstation, we'll let lots of (unpredictable) outgoing packets go freely.
$IPTABLES -P INPUT DROP
$IPTABLES -P FORWARD DROP
$IPTABLES -P OUTPUT ACCEPT

# INBOUND POLICY
# ==============
# of course, accepting loopback is a good idea
$IPTABLES -A INPUT -i lo -j ACCEPT 

#   (Applies to packets entering our network interface from the network, 
#   and addressed to this host.)

$IPTABLES -A INPUT -m state --state INVALID -j DROP 
$IPTABLES -A INPUT -m state --state RELATED,ESTABLISHED -j ACCEPT 

# ftp incoming
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 20 -j ACCEPT
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 21 -j ACCEPT

# ssh incoming, including non-standard port (if needed)
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 22 -j ACCEPT 
#$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 222 -j ACCEPT

# web serving, let's allow it!
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 80 -j ACCEPT

# nagios (5666); monitor time (123), allow snmp (161)
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 5666 -j ACCEPT
$IPTABLES -A INPUT -p udp -m state --state NEW --dport 123 -j ACCEPT
$IPTABLES -A INPUT -p udp -m state --state NEW --dport 161 -j ACCEPT


# OUTBOUND POLICY
# ===============
# of course, accepting loopback is a good idea
$IPTABLES -A OUTPUT -o lo -j ACCEPT

#   (Applies to packets sent to the network interface from local processes)

$IPTABLES -A OUTPUT -m state --state RELATED,ESTABLISHED -j ACCEPT

Example 3 - /etc/iptables.bak for a web-server with both http and https, nfs-client, amanda-tape-backup-client, SSH connectivity, a Sassafrass keyserver, a flexlm license-server, and monitoring by nagios

#! /bin/sh
# /etc/iptables.bak

# Let's save typing & confusion with variables
IPTABLES=/sbin/iptables

# Flush active rules and custom tables
$IPTABLES --flush
$IPTABLES --delete-chain

# set the defaults so that by-default incoming packets are dropped, unless explicitly allowed;
# for a desktop workstation, we'll let lots of (unpredictable) outgoing packets go freely.
$IPTABLES -P INPUT DROP
$IPTABLES -P FORWARD DROP
$IPTABLES -P OUTPUT ACCEPT

# INBOUND POLICY
# ==============
# of course, accepting loopback is a good idea
$IPTABLES -A INPUT -i lo -j ACCEPT 

#   (Applies to packets entering our network interface from the network, 
#   and addressed to this host.)

$IPTABLES -A INPUT -m state --state INVALID -j DROP 
$IPTABLES -A INPUT -m state --state RELATED,ESTABLISHED -j ACCEPT 

# ftp incoming
#$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 20 -j ACCEPT
#$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 21 -j ACCEPT

# ssh incoming, including non-standard port (if needed)
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 22 -j ACCEPT 
#$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 222 -j ACCEPT

# this machine is a mail-server, aggregating logs + hosting mailing-lists
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 25 -j ACCEPT

# web serving, let's allow it!  Both http and https ports
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 80 -j ACCEPT
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 443 -j ACCEPT

# portmapper, in support of NFS-client
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 111 -j ACCEPT

# nagios (5666); monitor time (123), allow snmp (161)
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 5666 -j ACCEPT
$IPTABLES -A INPUT -p udp -m state --state NEW --dport 123 -j ACCEPT
$IPTABLES -A INPUT -p udp -m state --state NEW --dport 161 -j ACCEPT

# amanda tape-backups; we reach out and tape things from this machine
$IPTABLES -A INPUT -p udp -m state --state NEW --dport 10080 -j ACCEPT
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 10082 -j ACCEPT
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 10083 -j ACCEPT

# flexlm (lmgrd) license-server listens here (set in license.dat file)
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 7111 -j ACCEPT

# Sassafrass keyserver listens here on both udp and tcp
$IPTABLES -A INPUT -p tcp -m state --state NEW --dport 19283 -j ACCEPT
$IPTABLES -A INPUT -p udp -m state --state NEW --dport 19283 -j ACCEPT


# OUTBOUND POLICY
# ===============
# of course, accepting loopback is a good idea
$IPTABLES -A OUTPUT -o lo -j ACCEPT

#   (Applies to packets sent to the network interface from local processes)

$IPTABLES -A OUTPUT -m state --state RELATED,ESTABLISHED -j ACCEPT

Save the configuration:

hostname ~ # etc/init.d/iptables save

And then back up your working configuration in case you break something later you can quickly revert:

hostname ~ # cp /var/lib/iptables/rules-save /var/lib/iptables/rules.working

Viewing/checking the active ruleset:

hostname ~ # iptables -L