nftables

Nftables

Overview
Role Variables
- OS Specific Variables
- Rules Dictionaries
Examples
- With playbooks
- With group_vars and host_vars
Configuration
- Fail2ban integration
Development
License
Author Information

Overview

A role to manage Nftables rules and packages.

Highly inspired by Mike Gleason firewall role (3 levels of rules definition and template), thanks ! I hope i haven't complexify his philosophy… (I'm pretty sure, i now did complexify it :D) ^^

Role Variables

nft_enabled : Enable or disable support for Nftables [default : true].
nft_pkg_state : State of new nftables package(s) [default : present].
nft_old_pkg_list : The list of useless packages to remove (such as Iptables,…) [default : iptables].
nft_old_pkg_state : State of old package(s) [default : absent].
nft_old_pkg_manage : If old package(s) should be managed with this role [default : true].
nft_conf_dir_path : Directory to store the differents Nftables configuration files [default : /etc/nftables.d].
nft_main_conf_path : Main configuration file loaded by systemd unit [default : /etc/nftables.conf].
nft_main_conf_content : Template used to generate the previous main configuration file [default : etc/nftables.conf.j2].
nft_input_conf_path : Input configuration file include in main configuration file [default : {{ nft_conf_dir_path }}/filter-input.nft].
nft_input_conf_content : Template used to generate the previous input configuration file [default : etc/nftables.d/filter-input.nft.j2].
nft_output_conf_path : Output configuration file include in main configuration file [default : {{ nft_conf_dir_path }}/filter-output.nft].
nft_output_conf_content : Template used to generate the previous output configuration file [default : etc/nftables.d/filter-output.nft.j2].
nft_forward_conf_path : forward configuration file include in main configuration file [default : {{ nft_conf_dir_path }}/filter-forward.nft].
nft_forward_conf_content : Template used to generate the previous forward configuration file [default : etc/nftables.d/filter-forward.nft.j2].
nft_define_conf_path : Vars definition file include in main configuration file [default : {{ nft_conf_dir_path }}/defines.nft].
nft_define_conf_content : Template used to generate the previous vars definition file [default : etc/nftables.d/defines.nft.j2].
nft_sets_conf_path : Sets and maps definition file include in main configuration file [default : {{ nft_conf_dir_path }}/sets.nft].
nft_sets_conf_content : Template used to generate the previous sets and maps definition file [default : etc/nftables.d/sets.nft.j2].
nft_global_default_rules : Set default rules for global chain. Other chains will jump to global before apply their specific rules.
nft_global_rules : You can add global rules or override those defined by nft_global_default_rules for all hosts.
nft_global_group_rules : You can add global rules or override those defined by nft_global_default_rules and nft_global_rules for a group.
nft_global_host_rules : Hosts can also add or override all previours rules.
nft__custom_content : Custom content (tables, include,…) to add in Nftables configuration [default : ''].
nft_input_default_rules : Set default rules for input chain.
nft_input_rules : You can add input rules or override those defined by nft_input_default_rules for all hosts.
nft_input_group_rules : You can add input rules or override those defined by nft_input_default_rules and nft_input_rules for a group.
nft_input_host_rules: : Hosts can also add or override all previous input rules.
nft_output_default_rules : Set default rules for output chain.
nft_output_rules : You can add output rules or override those defined by nft_output_default_rules for all hosts.
nft_output_group_rules : You can add output rules or override those defined by nft_output_default_rules and nft_output_rules for a group.
nft_output_host_rules : Hosts can also add or override all previous output rules.
nft_forward_default_rules : Set default rules for forward chain.
nft_forward_rules : You can add forward rules or override those defined by nft_forward_default_rules for all hosts.
nft_forward_group_rules : You can add forward rules or override those defined by nft_forward_default_rules and nft_forward_rules for a group.
nft_forward_host_rules : Hosts can also add or override all previous forward rules.
nft__forward_table_manage : If the forward table should be managed [default : False].
nft__nat_table_manage : If the nat table should be managed [default : False].
nft__nat_default_prerouting_rules : Set default rules for prerouting chain of nat table.
nft__nat_prerouting_rules : Set rules for prerouting chain of nat table for all hosts in the Ansible inventory.
nft__nat_group_prerouting_rules : Set rules for prerouting chain of nat table for hosts in specific Ansible inventory group.
nft__nat_host_prerouting_rules : Set rules for prerouting chain of nat table for specific hosts the Ansible inventory.
nft__nat_prerouting_conf_path : Prerouting configuration file include in the main configuration [default : {{ nft_conf_dir_path }}/nat-prerouting.nft].
nft__nat_prerouting_conf_content : Template used to generate the previous prerouting configuration file [default : etc/nftables.d/nat-prerouting.nft.j2].
nft__nat_default_postrouting_rules : Set default rules for postrouting chain of nat table.
nft__nat_postrouting_rules : Set rules for postrouting chain of nat table for all hosts in the Ansible inventory.
nft__nat_group_postrouting_rules : Set rules for postrouting chain of nat table for hosts in specific Ansible inventory group.
nft__nat_host_postrouting_rules : Set rules for postrouting chain of nat table for specific hosts the Ansible inventory.
nft__nat_postrouting_conf_path : postrouting configuration file include in the main configuration [default : {{ nft_conf_dir_path }}/nat-postrouting.nft].
nft__nat_postrouting_conf_content : Template used to generate the previous postrouting configuration file [default : etc/nftables.d/nat-postrouting.nft.j2].
nft_define_default : Set default vars available in all rules.
nft_define : You can add vars or override those defined by nft_define_default for all hosts.
nft_define_group : You can add vars or override those defined by nft_define_default and nft_define for a group.
nft_define_host : You can add or override all previous vars.
nft_service_manage : If nftables service should be managed with this role [default : true].
nft_service_name : nftables service name [default : nftables].
nft_service_enabled : Set nftables service available at startup [default : true].
nft__service_protect : If systemd unit should protect system and home [default : true].
nft_merged_groups : If variables from the hosts Ansible groups should be merged [default : false].
nft_merged_groups_dir : The dictionary where the nftables group rules, named like the Ansible groups, are located in [default : vars/].
nft_debug : Toggle more verbose output on/off. [default: 'false'].

OS Specific Variables

Please see default value by Operating System file in [vars][vars directory] directory.

nft_pkg_list : The list of package(s) to provide nftables.
nft__bin_location : Path to nftables executable. [default : /usr/sbin/nft]

Rule Templates

The nft_templates dictionary contains a library of useful rules, intended to be standardized and ready to use in your firewall. For example {{ nft_templates.allow_mdns }} will expand to the following rules, covering mDNS for both IPv4 and IPv6:

  - meta l4proto udp ip6 daddr ff02::fb    udp dport mdns counter accept comment "mDNS IPv6 service discovery"
  - meta l4proto udp ip  daddr 224.0.0.251 udp dport mdns counter accept comment "mDNS IPv4 service discovery"

Intended usage in custom rule sets:

nft_host_input_rules:
  ...
  010 allow mdns: "{{ nft_templates.allow_mdns }}"

Among others, the nft_templates dictionary contains rules implementing full recommendations for forwarding and input traffic firewalls as defined in RFC 4890, RFC 7126 and RFC 9288. For details and examples see defaults/main.yml.

Rules Dictionaries

Each type of rules dictionaries will be merged and rules will be applied in the alphabetical order of the keys (the reason to use 000 to 999 as prefix). So :

nft_*_default_rules : Define default rules for all nodes. You can define it in group_vars/all.
nft_*_rules : Can add rules and override those defined by nft_*_default_rules. You can define it in group_vars/all.
nft_*_group_rules : Can add rules and override those defined by nft_*_default_rules and nft_*_rules. You can define it in group_vars/webservers.
- If 'nft_merged_groups' is set to true, multiple group rules from the ansible groups will also be merged together.
nft_*_host_rules : Can add rules and override those define by nft_*_default_rules, nft_*_group_rules and nft_*_rules. You can define it in host_vars/www.local.domain.

defaults/main.yml:

# rules
nft_global_default_rules:
  005 state management:
    - ct state established,related accept
    - ct state invalid drop
nft_global_rules: {}
nft_merged_groups: false
nft_merged_groups_dir: vars/
nft_global_group_rules: {}
nft_global_host_rules: {}

nft_input_default_rules:
  000 policy:
    - type filter hook input priority 0; policy drop;
  005 global:
    - jump global
  010 drop unwanted:
    - ip daddr @blackhole counter drop
  015 localhost:
    - iif lo accept
  210 input tcp accepted:
    - tcp dport @in_tcp_accept ct state new accept
nft_input_rules: {}
nft_input_group_rules: {}
nft_input_host_rules: {}

nft_output_default_rules:
  000 policy:
    - type filter hook output priority 0; policy drop;
  005 global:
    - jump global
  015 localhost:
    - oif lo accept
  050 icmp:
    - ip protocol icmp accept
    - ip6 nexthdr icmpv6 counter accept
  200 output udp accepted:
    - udp dport @out_udp_accept ct state new accept
  210 output tcp accepted:
    - tcp dport @out_tcp_accept ct state new accept
nft_output_rules: {}
nft_output_group_rules: {}
nft_output_host_rules: {}

# define nft vars
nft_define_default:
  broadcast and multicast:
    desc: 'broadcast and multicast'
    name: badcast_addr
    value: '{ 255.255.255.255, 224.0.0.1, 224.0.0.251 }'
  input tcp accepted:
    name: in_tcp_accept
    value: '{ ssh }'
  output tcp accepted:
    name: out_tcp_accept
    value: '{ http, https, hkp }'
  output udp accepted:
    name: out_udp_accept
    value: '{ bootps, domain, ntp }'
nft_define: {}
nft_define_group: {}
nft_define_host: {}

# sets and maps
nft_set_default:
  blackhole:
    - type ipv4_addr;
    - elements = $badcast_addr
  in_tcp_accept:
    - type inet_service; flags interval;
    - elements = $in_tcp_accept
  out_tcp_accept:
    - type inet_service; flags interval;
    - elements = $out_tcp_accept
  out_udp_accept:
    - type inet_service; flags interval;
    - elements = $out_udp_accept
nft_set: {}
nft_set_group: {}
nft_set_host: {}

Those default will generate the following configuration :

#!/usr/sbin/nft -f
# Ansible managed

# clean
flush ruleset

include "/etc/nftables.d/defines.nft"

table inet filter {
    chain global {
        # 000 state management
        ct state established,related accept
        ct state invalid drop
    }
    include "/etc/nftables.d/sets.nft"
    include "/etc/nftables.d/filter-input.nft"
    include "/etc/nftables.d/filter-output.nft"
}

And you can get all rules and definitons by displaying the ruleset on the host : $ nft list ruleset :

table inet filter {
    set blackhole {
        type ipv4_addr
        elements = { 255.255.255.255, 224.0.0.1, 224.0.0.251}
    }

    set out_tcp_accept {
        type inet_service
        flags interval
        elements = { http, https, hkp}
    }

    set out_udp_accept {
        type inet_service
        flags interval
        elements = { domain, bootps, ntp}
    }

    chain global {
        ct state established,related accept
        ct state invalid drop
    }

    chain input {
        type filter hook input priority 0; policy drop;
        jump global
        ip daddr @blackhole counter packets 0 bytes 0 drop
        iif "lo" accept
        tcp dport @in_tcp_accept ct state new accept
    }

    chain output {
        type filter hook output priority 0; policy drop;
        jump global
        oif "lo" accept
        ip protocol icmp accept
        udp dport @out_udp_accept ct state new accept
        tcp dport @out_tcp_accept ct state new accept
    }
}

Examples

With playbooks

Manage Nftables with defaults vars (click to expand)

- hosts: serverXYZ
  roles:
    - role: ipr-cnrs.nftables

Add a new simple filter rule for incoming traffic (eg. 1 port for UDP/torrent) (click to expand)

- hosts: serverXYZ
  vars:
      nft_input_rules:
        400 input torrent accepted:
          - udp dport 6881 ct state new accept
  roles:
    - role: ipr-cnrs.nftables

nft_input_group_rules or nft_input_host_rules variables can also be used.
The weight (400) allow to order all merged rules (from nft_input_*rules dictionaries).
The text following the weight (input torrent accepted) is a small description that will be added as a comment in nft_input_conf_path file on the remote host.

Add a new multi-ports filter rule for incoming traffic (eg. TCP/http, https, http-alt,…) (click to expand)

- hosts: serverXYZ
  vars:
      nft_input_rules:
        400 input http accepted:
          - tcp dport { 80, 443, 8080-8082 } ct state new accept
  roles:
    - role: ipr-cnrs.nftables

nft_input_group_rules or nft_input_host_rules variables can also be used.
The weight (400) allow to order all merged rules (from nft_input_*rules dictionaries).
The text following the weight (input http accepted) is a small description that will be added as a comment in nft_input_conf_path file on the remote host.
In this case, brackets are useful and define a anonymous set. For a single element (port, IP address,…), brackets are overkill and the singleton definition is enought.

Add a new rule with a variable (click to expand)

Nftables variables can be useful if you define somes generic rules for all hosts with such variables (called with $) and override variable's value for some groups or hosts.

- hosts: serverXYZ
  vars:
    - nft_define_group:
        input http accepted:
          desc: HTTP and HTTPS
          name: in_http_accept
          value: '{ 80, 443 }'
      nft_input_group_rules:
        400 input http accepted:
          - tcp dport $in_http_accept ct state new accept
  roles:
    - role: ipr-cnrs.nftables

Add a new variable with define for HTTP ports.
Add a new rule for incoming traffic and use the previous defined variable.

Result of nft list ruleset on the remote host will be :

table inet filter {
        …

        chain input {
                …
                tcp dport { http, https } ct state new accept
                …
        }
        …
}

No mention of $in_http_accept variable.

nft_define or nft_define_host variables can also be used.
nft_input_rules or nft_input_host_rules variables can also be used.
The weight (400) allow to order all merged rules (from nft_input_*rules dictionaries).
The text following the weight (input http accepted) is a small description that will be added as a comment in nft_input_conf_path file on the remote host.

Add a new rule with a named set (click to expand)

Quite similar to Nftables variables, named set can be useful if you define somes generic rules and sets (eg. for all hosts) and override only the set in some case (eg. for a group or some hosts).

In addition to variables, it is possible to add content to named sets on the fly from the host without completely rewrite the rule.

- hosts: serverXYZ
  vars:
      nft_set_group:
        in_udp_accept:
          - type inet_service; flags interval;
          - elements = { 6881-6887, 6889 }
      nft_input_group_rules:
        200 input udp accepted:
          - udp dport @in_udp_accept ct state new accept
  roles:
    - role: ipr-cnrs.nftables

Add a new named set with nft_set_group dictionary (eg. for torrent ports).
Add a new rule for incoming traffic and use the previous defined set.
On the remote host, if you try to add a port to this set : nft add element inet filter in_udp_accept \{ 6999 \}

Result of nft list ruleset on the remote host will now be :

table inet filter {
        …
        set in_udp_accept {
                type inet_service
                flags interval
                elements = { 6881-6887, 6889, 6999 }
        }
        chain input {
                …
                udp dport @in_udp_accept ct state new accept
                …
        }
        …
}

nft_set or nft_set_host variables can also be used.
nft_input_rules or nft_input_host_rules variables can also be used.
The weight (200) allow to order all merged rules (from nft_input_*rules dictionaries).
The text following the weight (input upd accepted) is a small description that will be added as a comment in nft_input_conf_path file on the remote host.

Override a default rule with 2 new rules (click to expand)

- hosts: serverXYZ
  vars:
      nft_input_host_rules:
        050 icmp:
          - ip protocol icmp  ip saddr != 192.168.0.0/24  counter  drop
          - ip protocol icmp  icmp type echo-request  ip length <= 84  counter  limit rate 10/minute  accept
  roles:
    - role: ipr-cnrs.nftables

Get rule description from defaults/main.yml file (eg. 050 icmp).
Drop any ICMP request that doesn't come from 192.168.0.0 network.
Ensure the request is less or equal to 84 bytes and set up a limit to 10 requests per minute.

nft_input_rules or nft_input_group_rules variables can also be used.
The weight (050) allow to order all merged rules (from nft_input_*rules dictionaries).
The text following the weight (icmp) is a small description that will be added as a comment in nft_input_conf_path file on the remote host.

Override some of the default defined sets (click to expand)

- hosts: serverXYZ
  vars:
    - nft_define:
      input tcp accepted:
        desc: Custom SSH port and torrent
        name: in_tcp_accept
        value: '{ 2201, 6881 }'
  roles:
    - role: ipr-cnrs.nftables

Get item name (eg. input tcp accepted) and variable name (eg. in_tcp_accept) from defaults/main.yml file.
Set a new value (eg. '{ 2201, 6881 }').
You can add a desc attribute that will be set as a comment in nft_input_conf_path file on the remote host.

nft_define_group or nft_define_host variables can also be used.

Override all default rules (eg. for outgoing traffic) (click to expand)

If the default rules are too permissive, if you already override most of them,… In some case, i guess, it can be interesting to redefine the default variable :

- hosts: serverXYZ
  vars:
      nft_output_default_rules:
        000 policy:
          - type filter hook output priority 0; policy drop;
        005 state management:
          - ct state established,related accept
          - ct state invalid drop
        015 localhost:
          - oif lo accept
        050 my rule for XXX hosts and services:
          - tcp dport 2000  ip saddr { xxx.xxx.xxx.xxx, yyy.yyy.yyy.yyy }  ct state new  accept
        250 reset-ssh:  # allow the host to reset SSH connections to avoid 10 min delay from Ansible controller
          - tcp sport ssh tcp flags { rst, psh | ack } counter accept
  roles:
    - role: ipr-cnrs.nftables

At least, don't forget to :
1. set a default policy.
2. manage already established state.
3. accept rst, psh | ack flags for ssh to avoid a 10 minutes delay at the first run of this Nftables role (see #1).
Then add your own rules with the wanted weight to order all merged rules (from nft_output_*rules dictionaries) and descriptions.

Remove a default rule (click to expand)

- hosts: serverXYZ
  vars:
      nft_output_host_rules:
        210 output tcp accepted:
          -
  roles:
    - role: ipr-cnrs.nftables

Get rule description from defaults/main.yml file (210 output tcp accepted).
The default policy for outgoing traffic (drop) will now be applied to ports defined in out_tcp_accept variable. Be sure of what you are doing.
The rule will no longer be present in nft list ruleset result, just a comment will remain (210 output tcp accepted) in nft_output_conf_path file on the remote host.

nft_output_rules or nft_output_group_rules variables can also be used.
The weight (210) allow to order all merged rules (from nft_output_*rules dictionaries).

With group_vars and host_vars

Use default rules and allow, for first_group, incoming ICMP and count both ICMP and default policy (drop) packets (click to expand)

group_vars/first_group :

nft_input_group_rules:
  020 icmp:
    - ip protocol icmp icmp type echo-request ip length <= 84 counter limit rate 1/minute accept
  999 count policy packet:
    - counter

Use merged group rules from multiple ansible groups (click to expand)

Enable to merge group's variables :

- hosts: serverXYZ
  vars:
    nft_merged_groups: true
    nft_merged_groups_dir: vars/
  roles:
    - role: ipr-cnrs.nftables

Put extra rules inside the "vars" folder named after your ansible groups for serverXYZ :

vars/first_group :

nft_input_group_rules:
  020 icmp:
    - ip protocol icmp icmp type echo-request ip length <= 84 counter limit rate 1/minute accept
  999 count policy packet:
    - counter

vars/second_group :

nft_input_group_rules:
  021 LAN:
    - iif eth0 accept

These rulesets, from the two groups, will be merged if the host is a member of these groups.

Configuration

This role will :

Install nftables on the system.
Enable nftables service by default at startup.
Generate a default configuration file which include all following files and loaded by systemd unit.
Generate input and output rules files include called by the main configuration file.
Generate vars in a file and sets and maps in another file.
Ensure nftables is started and enabled on boot
(re)Start nftables service at first run or when systemd units are modified
Reload nftables service at next runs to avoid to let the host without firewall rules due to invalid syntax.

Fail2ban integration

Before Debian Bullseye, systemd unit for Fail2ban doesn't come with a decent integration with Nftables. This role will create override file for fail2ban unit, unless nft_fail2ban_service_override is set to false. Default is to add it even if it's not (yet) available on the host. This ensures :