Difference between revisions of "Iptables"

From ArchWiki
Jump to: navigation, search
m (Added nftables to related articles)
(Chains: add to explain processing order of default policies)
Line 49: Line 49:
 
See {{ic|man 8 iptables}} for a description of built-in chains in other tables.
 
See {{ic|man 8 iptables}} for a description of built-in chains in other tables.
  
By default, none of the chains contain any rules.  It is up to you to append rules to the chains that you want to use.  Chains ''do'' have a default policy, which is generally set to {{ic|ACCEPT}}, but can be reset to {{ic|DROP}}, if you want to be sure that nothing slips through your ruleset.
+
By default, none of the chains contain any rules.  It is up to you to append rules to the chains that you want to use.  Chains ''do'' have a default policy, which is generally set to {{ic|ACCEPT}}, but can be reset to {{ic|DROP}}, if you want to be sure that nothing slips through your ruleset. The default policy always applies at the end of a chain only. Hence, all existing rules in the chain will be passed through by the packet before the default policy is applied.
  
 
User-defined chains can be added to make rulesets more efficient or more easily modifiable.  See [[Simple stateful firewall]] for an example of how user-defined chains are used.
 
User-defined chains can be added to make rulesets more efficient or more easily modifiable.  See [[Simple stateful firewall]] for an example of how user-defined chains are used.

Revision as of 11:35, 9 August 2014

Iptables is a powerful firewall built into the Linux kernel and is part of the netfilter project. It can be configured directly, or by using one of the many frontends and GUIs. iptables is used for IPv4 and ip6tables is used for IPv6.

nftables was released in release with Linux kernel 3.13, and will one day replace iptables as the main Linux firewall utility.

Installation

The stock Arch Linux kernel is compiled with iptables support. You will only need to install the userland utilities, which are provided by the package iptables in the official repositories. (The iproute2 package from the base group depends on iptables, so the iptables package should be installed on your system by default.)

Basic concepts

iptables is used to inspect, modify, forward, redirect, and/or drop IPv4 packets. The code for filtering IPv4 packets is already built into the kernel and is organized into a collection of tables, each with a specific purpose. The tables are made up of a set of predefined chains, and the chains contain rules which are traversed in order. Each rule consists of a predicate of potential matches and a corresponding action (called a target) which is executed if the predicate is true; i.e. the conditions are matched. iptables is the user utility which allow you to work with these chains/rules. Most new users find the complexities of linux IP routing quite daunting, but, in practice, the most common use cases (NAT and/or basic Internet firewall) are considerably less complex.

The key to understanding how iptables works is this chart. The lowercase word on top is the table and the upper case word below is the chain. Every IP packet that that comes in on any network interface passes through this flow chart from top to bottom. A common source of confusion is that packets entering from, say, an internal interface are handled differently than packets from an Internet-facing interface. All interfaces are handled the same way; it's up to you to define rules that treat them differently. Of course some packets are intended for local processes, hence come in from the top of the chart and stop at <Local Process>, while other packets are generated by local processes; hence start at <Local Process> and proceed downward through the flowchart. A detailed explanation of how this flow chart works can be found here.

Tables

iptables contains five tables:

  1. raw is used only for configuring packets so that they are exempt from connection tracking.
  2. filter is the default table, and is where all the actions typically associated with a firewall take place.
  3. nat is used for network address translation (e.g. port forwarding).
  4. mangle is used for specialized packet alterations (see Mangled packet).
  5. security is used for Mandatory Access Control networking rules (e.g. SELinux -- see this article for more details).


In most common use cases you will only use two of these: filter and nat. The other tables are aimed at complex configurations involving multiple routers and routing decisions and are in any case beyond the scope of these introductory remarks.

Chains

Tables consist of chains, which are lists of rules which are followed in order. The default table, filter, contains three built-in chains: INPUT, OUTPUT and FORWARD which are activated at different points of the packet filtering process, as illustrated in the Andreasson flow chart. The nat table includes PREROUTING, POSTROUTING, and OUTPUT chains.

See man 8 iptables for a description of built-in chains in other tables.

By default, none of the chains contain any rules. It is up to you to append rules to the chains that you want to use. Chains do have a default policy, which is generally set to ACCEPT, but can be reset to DROP, if you want to be sure that nothing slips through your ruleset. The default policy always applies at the end of a chain only. Hence, all existing rules in the chain will be passed through by the packet before the default policy is applied.

User-defined chains can be added to make rulesets more efficient or more easily modifiable. See Simple stateful firewall for an example of how user-defined chains are used.

Rules

Packet filtering is based on rules, which are specified by multiple matches (conditions the packet must satisfy so that the rule can be applied), and one target (action taken when the packet matches all conditions). The typical things a rule might match on are what interface the packet came in on (e.g eth0 or eth1), what type of packet it is (ICMP, TCP, or UDP), or the destination port of the packet.

Targets are specified using the -j or --jump option. Targets can be either user-defined chains (i.e. if these conditions are matched, jump to the following user-defined chain and continue processing there, one of the special built-in targets, or a target extension. Built-in targets are ACCEPT, DROP, QUEUE and RETURN, target extensions are, for example, REJECT and LOG. If the target is a built-in target, the fate of the packet is decided immediately and processing of the packet in current table is stopped. If the target is a user-defined chain and the packet passes successfully through this second chain, it will move to the next rule in the original chain. Target extensions can be either terminating (as built-in targets) or non-terminating (as user-defined chains), see man 8 iptables-extensions for details.

Traversing Chains

A network packet received on any interface traverses the traffic control tables.chains in the order shown in the Andreasson flow chart. The first routing decision involves deciding if the final destination of the packet is the local machine (in which case the packet traverses through the INPUT chains) or elsewhere (in which case the packet traverses through the FORWARD chains). Subseqeunt routing decisions involve deciding what interface to assign to an outgoing packet. At each chain in the path, every rule in that chain is evaluated in order and whenever a rule matches, the corresponding target/jump action is executed. The 3 most commonly used targets are ACCEPT, DROP, and jump to a user-defined chain. While built-in chains can have default policies, user-defined chains can not. If every rule in a chain that you jumped to fails to provide a complete match, the packet is dropped back into the calling chain as illustrated here. If at any time a complete match is achieved for a rule with a DROP target, the packet is dropped and no further processing is done. If a packet is ACCEPTed within a chain, it will be ACCEPT'ed in all superset chains also and it will not traverse any of the superset chains any further. However, be aware that the packet will continue to traverse all other chains in other tables in the normal fashion.

Modules

There are many modules which can be used to extend iptables such as connlimit, conntrack, limit and recent. These modules add extra functionality to allow complex filtering rules.

Configuring and Running iptables

iptables is a Systemd service and is started accordingly:

# systemctl start iptables

However, the service won't start unless it finds an /etc/iptables/iptables.rules file, and the Arch iptables package does not come with a default iptables.rule file. So to start the service for the first time:

# touch /etc/iptables/iptables.rules
# systemctl start iptables

or

# cp /etc/iptables/empty.rules /etc/iptables/iptables.rules
# systemctl start iptables

As with other services, if you want iptables to be loaded automatically on boot, you must enable it:

# systemctl enable iptables

From the command line

Showing the current rules

You can check the current ruleset and the number of hits per rule by using the command:

# iptables -nvL
Chain INPUT (policy ACCEPT 0 packets, 0 bytes)
 pkts bytes target     prot opt in     out     source               destination   
     
Chain FORWARD (policy ACCEPT 0 packets, 0 bytes)
 pkts bytes target     prot opt in     out     source               destination    
    
Chain OUTPUT (policy ACCEPT 0K packets, 0 bytes)
 pkts bytes target     prot opt in     out     source               destination

If the output looks like the above, then there are no rules. Nothing is blocked.

To show the line numbers when listing rules, append --line-numbers to that input. This is useful when deleting and adding individual rules.

Editing rules

Rules can be added either by appending a rule to a chain or inserting them at a specific position on the chain. We will explore both methods here.

First of all, our computer is not a router (unless, of course, it is a router). We want to change the default policy on the FORWARD chain from ACCEPT to DROP.

# iptables -P FORWARD DROP
Warning: The rest of this section is meant to teach the syntax and concepts behind iptables rules. It is not intended as a means for securing servers. For improving the security of your system, see Simple stateful firewall for a minimally secure iptables configuration and Security for hardening Arch Linux in general.

The Dropbox LAN sync feature broadcasts packets every 30 seconds to all computers it can see. If we happen to be on a LAN with Dropbox clients and do not use this feature, then we might wish to reject those packets.

# iptables -A INPUT -p tcp --dport 17500 -j REJECT --reject-with icmp-port-unreachable
# iptables -nvL --line-numbers
Chain INPUT (policy ACCEPT 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         
1        0     0 REJECT     tcp  --  *      *       0.0.0.0/0            0.0.0.0/0            tcp dpt:17500 reject-with icmp-port-unreachable

Chain FORWARD (policy DROP 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         

Chain OUTPUT (policy ACCEPT 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         

Note: We use REJECT rather than DROP here, because RFC 1122 3.3.8 requires hosts return ICMP errors whenever possible, instead of dropping packets. In reality, it is best to REJECT packets from hosts who should know about your server's existence, and DROP packets from hosts who should not even know your server exists, or those who appear "up to something".

Now, say we change our mind about Dropbox and decide to install it on our computer. We also want to LAN sync, but only with one particular IP on our network. So we should use -R to replace our old rule. Where 10.0.0.85 is our other IP:

# iptables -R INPUT 1 -p tcp --dport 17500 ! -s 10.0.0.85 -j REJECT --reject-with icmp-port-unreachable
# iptables -nvL --line-numbers
Chain INPUT (policy ACCEPT 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         
1        0     0 REJECT     tcp  --  *      *      !10.0.0.85            0.0.0.0/0            tcp dpt:17500 reject-with icmp-port-unreachable

Chain FORWARD (policy DROP 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         

Chain OUTPUT (policy ACCEPT 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         

We have now replaced our original rule with one that allows 10.0.0.85 to access port 17500 on our computer. But now we realize that this is not scalable. If our friendly Dropbox user is attempting to access port 17500 on our device, we should allow him immediately, not test him against any firewall rules that might come afterwards!

So we write a new rule to allow our trusted user immediately. Using -I to insert the new rule before our old one:

# iptables -I INPUT 1 -p tcp --dport 17500 -s 10.0.0.85 -j ACCEPT -m comment --comment "Friendly Dropbox"
# iptables -nvL --line-numbers
Chain INPUT (policy ACCEPT 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         
1        0     0 ACCEPT     tcp  --  *      *       10.0.0.85            0.0.0.0/0            tcp dpt:17500 /* Friendly Dropbox */
2        0     0 REJECT     tcp  --  *      *      !10.0.0.85            0.0.0.0/0            tcp dpt:17500 reject-with icmp-port-unreachable

Chain FORWARD (policy DROP 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         

Chain OUTPUT (policy ACCEPT 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination    

And replace our second rule with one that rejects everything on port 17500:

# iptables -R INPUT 2 -p tcp --dport 17500 -j REJECT --reject-with icmp-port-unreachable 

Our final rule list now looks like this:

# iptables -nvL --line-numbers
Chain INPUT (policy ACCEPT 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         
1        0     0 ACCEPT     tcp  --  *      *       10.0.0.85            0.0.0.0/0            tcp dpt:17500 /* Friendly Dropbox */
2        0     0 REJECT     tcp  --  *      *       0.0.0.0/0            0.0.0.0/0            tcp dpt:17500 reject-with icmp-port-unreachable

Chain FORWARD (policy DROP 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         

Chain OUTPUT (policy ACCEPT 0 packets, 0 bytes)
num   pkts bytes target     prot opt in     out     source               destination         

Resetting rules

You can flush and reset iptables to default using these commands:

# iptables -F
# iptables -X
# iptables -t nat -F
# iptables -t nat -X
# iptables -t mangle -F
# iptables -t mangle -X
# iptables -t raw -F
# iptables -t raw -X
# iptables -t security -F
# iptables -t security -X
# iptables -P INPUT ACCEPT
# iptables -P FORWARD ACCEPT
# iptables -P OUTPUT ACCEPT

The -F command with no arguments flushes all the chains in its current table. Similarly, -X deletes all empty non-default chains in a table.

Individual chains may be flushed or deleted by following -F and -X with a [chain] argument.

Configuration file

Iptables rules are, by default in Arch Linux, stored in /etc/iptables/iptables.rules. They are, however, not loaded automatically, instead you can enable the iptables.service which reads this file and loads the rules at boot or when started:

# systemctl enable iptables
# systemctl start iptables

Iptables rules for IPv6 are, by default, stored in /etc/iptables/ip6tables.rules, this file is read by ip6tables.service. You can start it the same way as above.

After adding rules via command-line, the configuration file is not changed automatically — you have to save it manually:

# iptables-save > /etc/iptables/iptables.rules

If you edit the configuration file manually, you have to reload it:

# systemctl reload iptables

Or you can load it directly through iptables:

# iptables-restore < /etc/iptables/iptables.rules

Guides

Logging

The LOG target can be used to log packets that hit a rule. Unlike other targets like ACCEPT or DROP, the packet will continue moving through the chain after hitting a LOG target. This means that in order to enable logging for all dropped packets, you would have to add a duplicate LOG rule before each DROP rule. Since this reduces efficiency and makes things less simple, a logdrop chain can be created instead.

Create the chain with:

# iptables -N logdrop

Then define it:

## /etc/iptables/iptables.rules

*filter
:INPUT DROP [0:0]
:FORWARD DROP [0:0]
:OUTPUT ACCEPT [0:0]

... other user defined chains ..

## logdrop chain
:logdrop - [0:0]

-A logdrop -m limit --limit 5/m --limit-burst 10 -j LOG
-A logdrop -j DROP

... rules ...

## log AND drop packets that hit this rule:
-A INPUT -m conntrack --ctstate INVALID -j logdrop

... more rules ...

Limiting log rate

The limit module should be used to prevent your iptables log from growing too large or causing needless hard drive writes. Without limiting, an attacker could fill your drive (or at least your /var partition) by causing writes to the iptables log.

-m limit is used to call on the limit module. You can then use --limit to set an average rate and --limit-burst to set an initial burst rate. Example:

-A logdrop -m limit --limit 5/m --limit-burst 10 -j LOG

This appends a rule to the logdrop chain which will log all packets that pass through it. The first 10 packets will the be logged, and from then on only 5 packets per minute will be logged. The "limit burst" is restored by one every time the "limit rate" is not broken.

syslog-ng

Tango-view-refresh-red.pngThis article or section is out of date.Tango-view-refresh-red.png

Reason: Syslog-ng is no longer the default logger. This section should likely be updated to reflect the systemd journal. (Discuss in Talk:Iptables#)

Assuming you are using syslog-ng, you can control where iptables' log output goes this way:

filter f_everything { level(debug..emerg) and not facility(auth, authpriv); };

to

filter f_everything { level(debug..emerg) and not facility(auth, authpriv) and not filter(f_iptables); };

This will stop logging iptables output to /var/log/everything.log.

If you also want iptables to log to a different file than /var/log/iptables.log, you can simply change the file value of destination d_iptables here (still in syslog-ng.conf)

destination d_iptables { file("/var/log/iptables.log"); };

ulogd

ulogd is a specialized userspace packet logging daemon for netfilter that can replace the default LOG target. The package ulogd is available in the [community] repository.

See also

Template:Wikipedia