The Advanced FireWall (afw) for Chef

__A__dvanced __F__ire__W__all (AFW) for Chef and Linux that uses Iptables to dynamically configure inbound and outbound rules on nodes.

AFW uses Chef searches extensively to open access to systems. Instead of specifying the IP addresses of a set of sources or destinations, AFW allows you to specify searches in Chef syntax to list those systems. It is designed to be a lot more dynamic and maintainable than regular firewalls, and allows for filtering inbound and outbound traffic.

AFW support raw rules, that are just straight iptables syntax passed to the template. For these, Chef searches are not supported, but it gives you access to the full set of features iptables provides.

AFW runs on any Linux 2.6 or 3.x. It does not rely on distribution specific wrapper, such as ufw for ubuntu, but calls iptables-restore directly.

Rules definitions

Rules must be added in roles or nodes attributes. A typical rule looks like:

:afw =>
  :rules =>
    'MongoDB App Entry Point' => {
      :protocol => 'tcp',
      :direction => 'in',
      :interface => 'default',
      :user => 'mongodb',
      :source => ['(roles:*nodejs-app-node OR roles:*python-worker-node OR roles:*python-api-node) AND SAMETAG',
                  '10.4.76.2',
                  'backup-server.example.net']
      :dport => '27017'
    },
}

Rules must be added into node[:afw][:rules], and follow this syntax:

:afw =>
  :rules =>
    '<rule name>' =>
      :direction => '<in|out>',
      :protocol => '<udp|tcp|icmp>',
      :user => '<local user from /etc/passwd>',
      :interface => '<default|all|eth0|eth1|br0|...>',
      :source => '<ip|fqdn|chef search>|['<ip|fqdn|chef search>',...]'>',
      :sport => '<integer(:integer))>',
      :destination => '<ip|fqdn|chef search>|['<ip|fqdn|chef search>',...]'>',
      :dport => '<integer(:integer)>',
      :env => '<production|staging|...>',
      :options => ['disable_env_limit', 'disable_syntax_check', ...]
}

• Rule Name : A string that identifies the rule as uniquely as possible.

• direction (mandatory): in is for inbound firewall rules. out for outbound firewall rules. Select whether the rule will apply to packets entering the system (in) or leaving it (out).

• protocol (mandatory): Select the L4 protocol this rule applies to: udp, tcp or icmp.

• user (mandatory): Set the local user allowed to transmit packets. The user is checked only for outbound firewall rules (iptables limitation) but must be set for inbound rules as well, to ease auditing rules & systems. Note: if the local user doesn't exists, the provisioning will fail at the end, when the rules are loaded. If the user is installed by a package, the next chef run will succeed and fix the issue.

• dport (mandatory): Set the destination port of the connections being filtered. This is mandatory. Except when it's not (eg. icmp).

• interface (optional): Select the network interface. If absent of set to all, the interface parameter won't be set at all, setting the rule on all interfaces.

• source (mandatory for in rules): Set the source IP of the packets. This parameter can either be a single IP or network (eg. 10.1.2.0/32), a Fully Qualified Domain Name (eg. bob.colo.lair) or a Chef Search (eg. roles:mongodb). By default, searches are limited to the same chef_environment (eg. staging), to allow for firewall rules that open connections between environments, you will need an heresy parameter. In a chef-search, you can also use they keyword SAMETAG that will limit the search to the nodes that share the same tags. This is useful if, for example, you want to open connections to a database from all nodes within the same service tag, but not beyond. The syntax for a search would look like: 'roles:whatever-api AND SAMETAG'. If you have multiple sources and destinations, you can set them in a array:

'AMQP Producers' => {
  :direction => 'in',
  :user => 'rabbitmq',
  :protocol => 'tcp',
  :interface => 'default',
  :source => ['producer1.internal.domain.com',
              'producer2.internal.domain.com',
              '192.168.1.1',
              'roles:megaserver'],
  :dport => '5672'
}

• destination (Same as source)

• sport (optional): Set the source port in the firewall rule. For in rules, that means the source port of the remote machine. For out rules, that means the source port of this node when establishing a connection to a remote node.

• env (optional): The env parameters can be used to limit the application of a rule to a specific environment. If :env => 'staging' is set, the rule will be applied to nodes in the staging environment only.

• options (optional):

• disable_env_limit: disable_env_limit can also used to cross environment. If :options => ['disable_env_limit'] is set, the source and destination searches will return results from all environments, instead of limiting the result to the environment the node lives in. The following rule will allow production workers to connect to staging BackendDB. Don't do that. Keep environments isolated as much as possible.

      'Production Worker to Staging BackendDB' => {
        :protocol => 'tcp',
        :direction => 'out',
        :user => 'application-user',
        :destination => 'roles:*backenddb* AND SAMETAG AND chef_environment:staging',
        :dport => '15017',
        :env => 'production',
        :options => ['disable_env_limit']
      }

• disable_syntax_check: disable_syntax_check will turn off the rule validation step for a specific rule. This is useful if you want to define a rule without source/destination, and let another recipe populate the source/destination arrays later on.

      'API calls from servers' => {
        :protocol => 'tcp',
        :direction => 'in',
        :user => 'www-data',
        :dport => '80',
        :options => ['disable_syntax_check']
      },

You would then have a separate recipe that inserts IPs in this rule:

node[:afw][:rules]['API calls from servers'][:source] = []
ip_list = ['1.2.3.4', '4.5.6.7', '12.32.43.54']
ip_list.each do |ip|
  node[:afw][:rules]['API calls from servers'][:source].push(ip)
end

Creating rules from external cookbooks

If you want a cookbook to create firewal rules directly, as opposed to storing these rules in a roles, then you need to use the create_rule() function from the AFW module. Example: create outbound firewall rule for haproxy in the haproxy cookbook

depend in AFW in the metadata

cookbooks/haproxy/metadata.rb

[...]
depends 'AFW'

create the rule from the recipe using ruby

 # Call the AFW module to create the rule
 AFW.create_rule(node,
                 "Haproxy outbound to #{destination}:#{port}",
                 {'protocol' => 'tcp',
                  'direction' => 'out',
                  'user' => 'haproxy',
                  'destination' => "#{destination}",
                  'dport' => "#{port}"
                 })

Note that AFW.create_rule() must be called from a normal section of ruby code directly (not from a ruby_block) to ensure that the rules are compiled at chef compile time. The AFW template will later (at runtime) populate these rules into the iptables-restore file.

Predefined rules

Predefined rules are iptables rules that are used directly by AFW. Those rules are used for specific purposes only, such as using a very particular module for which AFW wouldn't have any support. Predefined rules only support 2 arguments: table and rule.

• table: the netfilter table on which this rule must be applied. One of nat, raw, mangle or `filter.

• rule: the firewall rule itself, in iptables-save format (do not specify a table in this format, or it will fail).

example:

  :afw => {
    :rules => {

      'Accept all packets router through the bridge' => {
        :table => 'filter',
        :rule => '-I FORWARD -o br0 -m physdev --physdev-is-bridged -j ACCEPT'
      },

      'Drop connection to the admin panel on the eth0 interface' => {
        :table => 'mangle',
        :rule => '-A INPUT -i eth0 -p tcp --dport 80 -m string --string "get /admin http/1.1" --icase --algo bm -m conntrack --ctstate ESTABLISHED -j DROP'
      },

      'DNAT a source IP to change the destination port' => {
        :table => 'nat',
        :rule => '-A PREROUTING -i eth3 -s 201.23.72.3 -p tcp --dport 8008 -j DNAT --to-destination 127.0.0.1:1234'
      },

      'Dont do conntrack on this specific user's UDP packets' => {
        :table => 'raw',
        :rule => '-A OUTPUT -o eth0 -p udp -m owner --uid-owner 105 -j NOTRACK'
      }
    }
  }

Rules Generation

The recipe will generate a rule file in /etc/firewall/rules.iptables that conforms to the iptables-save/restore syntax. At the end of the chef-run, and if the rules file has been modified during the run, the iptables-restore command will reload the entire ruleset.

Here's an example of templated ruleset. Notice how the rules are regrouped by system user, to make them easier to read.

# Generated by AFW on Tue Sep 25 02:55:32 +0000 2012

*raw
:PREROUTING ACCEPT [0:0]
:OUTPUT ACCEPT [0:0]
-A OUTPUT -o lo -j NOTRACK
-A PREROUTING -i lo -j NOTRACK
COMMIT

*mangle
:PREROUTING ACCEPT [0:0]
:INPUT ACCEPT [0:0]
:FORWARD ACCEPT [0:0]
:OUTPUT ACCEPT [0:0]
:POSTROUTING ACCEPT [0:0]
COMMIT

*nat
:PREROUTING ACCEPT [0:0]
:POSTROUTING ACCEPT [0:0]
:OUTPUT ACCEPT [0:0]
COMMIT

*filter
:INPUT ACCEPT [0:0]
:FORWARD ACCEPT [0:0]
:OUTPUT ACCEPT [0:0]
-A INPUT -i lo -j ACCEPT
-A OUTPUT -o lo -j ACCEPT
-A INPUT -m conntrack --ctstate ESTABLISHED,RELATED -j ACCEPT
-A OUTPUT -m conntrack --ctstate ESTABLISHED,RELATED -j ACCEPT
# some default rules we want to open everywhere
-A INPUT -p tcp --dport 22 -s 1.0.0.0/8 -j ACCEPT
-A INPUT -p tcp --dport 22 -s 12.30.0.0/16 -j ACCEPT
-A OUTPUT -p udp --dport 53 -d 1.0.0.0/8 -j ACCEPT

:app-user - [0:0]
-A OUTPUT -m owner --uid-owner 999 -m state --state NEW -j app-user
-A app-user -o eth0 -p tcp --dport 2003 -d 1.2.3.4 -m conntrack --ctstate NEW -j ACCEPT
-A app-user -o eth0 -p tcp --dport 27017 -d 1.1.4.3 -m conntrack --ctstate NEW -j ACCEPT
-A app-user -o eth0 -p tcp --dport 5672 -d 1.1.4.9 -m conntrack --ctstate NEW -j ACCEPT
-A app-user -o eth0 -p tcp --dport 80 -d 1.1.4.10 -m conntrack --ctstate NEW -j ACCEPT
-A app-user -o eth0 -p udp --dport 8125 -d 1.1.4.74 -m conntrack --ctstate NEW -j ACCEPT
-A app-user -j LOG --log-prefix "DROP_AFW_OUTPUT_app-user " --log-uid --log-tcp-sequence

:haproxy - [0:0]
-A OUTPUT -m owner --uid-owner 110 -m state --state NEW -j haproxy
-A INPUT -i eth0 -p tcp --dport 10097 -s 0.0.0.0/0 -m conntrack --ctstate NEW -j ACCEPT
-A haproxy -o eth0 -p tcp --dport 2003 -d 1.1.0.25 -m conntrack --ctstate NEW -j ACCEPT
-A haproxy -o eth0 -p tcp --dport 2003 -d 1.1.0.37 -m conntrack --ctstate NEW -j ACCEPT
-A haproxy -o eth0 -p tcp --dport 80 -d 1.1.4.58 -m conntrack --ctstate NEW -j ACCEPT
-A haproxy -j LOG --log-prefix "DROP_AFW_OUTPUT_haproxy " --log-uid --log-tcp-sequence

:nagios - [0:0]
-A OUTPUT -m owner --uid-owner 105 -m state --state NEW -j nagios
-A INPUT -i eth0 -p tcp --dport 5666 -s 1.1.0.33 -m conntrack --ctstate NEW -j ACCEPT
-A nagios -j LOG --log-prefix "DROP_AFW_OUTPUT_nagios " --log-uid --log-tcp-sequence

:ntp - [0:0]
-A OUTPUT -m owner --uid-owner 104 -m state --state NEW -j ntp
-A ntp -o eth0 -p udp --dport 123 -d ntp.example.net -m conntrack --ctstate NEW -j ACCEPT
-A ntp -j LOG --log-prefix "DROP_AFW_OUTPUT_ntp " --log-uid --log-tcp-sequence

:root - [0:0]
-A OUTPUT -m owner --uid-owner 0 -m state --state NEW -j root
-A INPUT  -p icmp -s 0.0.0.0/0 -m conntrack --ctstate NEW -j ACCEPT
-A root -o eth0 -p tcp --dport 514 -d 1.1.3.18 -m conntrack --ctstate NEW -j ACCEPT
-A root -o eth0 -p udp --dport 514 -d 1.1.3.18 -m conntrack --ctstate NEW -j ACCEPT
-A root -p icmp -d 0.0.0.0/0 -m conntrack --ctstate NEW -j ACCEPT
-A root -j LOG --log-prefix "DROP_AFW_OUTPUT_root " --log-uid --log-tcp-sequence

:snmp - [0:0]
-A OUTPUT -m owner --uid-owner 108 -m state --state NEW -j snmp
-A snmp -o eth0 -p udp --dport 162 -d 1.1.0.23 -m conntrack --ctstate NEW -j ACCEPT
-A snmp -j LOG --log-prefix "DROP_AFW_OUTPUT_snmp " --log-uid --log-tcp-sequence

:www-data - [0:0]
-A OUTPUT -m owner --uid-owner 33 -m state --state NEW -j www-data
-A INPUT -i eth0 -p tcp --dport 80 -s 1.1.4.4 -m conntrack --ctstate NEW -j ACCEPT
-A INPUT -i eth0 -p tcp --dport 80 -s 1.1.4.1 -m conntrack --ctstate NEW -j ACCEPT
-A www-data -j LOG --log-prefix "DROP_AFW_OUTPUT_www-data " --log-uid --log-tcp-sequence

-A INPUT -j LOG --log-prefix "DROP_AFW_INPUT " --log-uid --log-tcp-sequence
-A INPUT -j DROP

-A OUTPUT -j LOG --log-prefix "DROP_AFW_OUTPUT " --log-uid --log-tcp-sequence
COMMIT

The INPUT chain contains all of the rules for incoming connections. It does not redirect packets to other chains, but accept or drop them directly.

The OUTPUT chain is a little different. Depending on the owner of the socket emitting packets, it will direct the packets to a different chain, named after the socket owner. In the example above, the packet from the snmp user will be directed to the chain named snmp. You can see in this chain that the first two rules accept packet, while the 3rd one will LOG to syslog when it is reached (it shouldn't be). Eventually, a DROP will follow that log rule to drop packets that aren't suppose to be sent.

ATTRIBUTES

• default[:afw][:enable] = true : enable or disable the firewall restore command. If set the false, the rules will still be populated in /etc/firewall/rules.iptables but the restore command will not be issued.

• default[:afw][:enable_input_drop] = true : DROP all input packets by defaut

• default[:afw][:enable_output_drop] = true : DROP all output packets by defaut

• default[:afw][:enable_input_drop_log] = true : LOG when DROP input packets

• default[:afw][:enable_output_drop_log] = true : LOG when DROP output packets

Dependencies

Ohai network_addr

This cookbooks relies on the custom AWeber Ohai plugin which sets the two following attributes: node[:network][:lanip] is set to the IP of the node on the LAN network node[:network][:laniface] is set to the network interface on the LAN While AFW can probably run with this plugin, it makes you life easier when you want to open rules on your LAN interface. We typically enfore eth0 to be the LAN interface, but feel free to use whatever you want.

Plugin source:

provides "network"

require_plugin "hostname"
require_plugin "#{os}::network"

network['interfaces'].each do |iface, addrs|

  addrs['addresses'].each do |ip, params|
    network["ipaddress_#{iface}"] = ip if params['family'].eql?('inet')
    network["ipaddress6_#{iface}"] = ip if params['family'].eql?('inet6')
    network["macaddress_#{iface}"] = ip if params['family'].eql?('lladdr')
  end

end

laniface = from("[ -e /vagrant ] && echo eth1 || echo eth0")

network['lanip'] = network["ipaddress_#{laniface}"]
network['laniface'] = laniface

network

LICENSE

 +-------------------------------------------------------------------------+
 | Advanced FireWall (AFW) Cookbook for Opscode Chef                       |
 |                                                                         |
 | Copyright (C) 2012, AWeber, Julien Vehent                               |
 |                                                                         |
 | This program is free software; you can redistribute it and/or modify    |
 | it under the terms of the GNU General Public License version 2          |
 | as published by the Free Software Foundation.                           |
 |                                                                         |
 | This program is distributed in the hope that it will be useful,         |
 | but WITHOUT ANY WARRANTY; without even the implied warranty of          |
 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the           |
 | GNU General Public License for more details.                            |
 |                                                                         |
 | You should have received a copy of the GNU General Public License along |
 | with this program; if not, write to the Free Software Foundation, Inc., |
 | 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.             |
 |                                                                         |
 +-------------------------------------------------------------------------+