#!/usr/sbin/nft -f
# {{ ansible_managed }}
{% macro list(map) -%}
{{ map | default({}, true) | dict2items | rejectattr('value', 'none') | map(attribute='value') | map('string') | unique | join(', ') }}
{%- endmacro %}
{% macro addr_set(name, type, map) -%}
set {{ name }} {
		type {{ type }}; flags interval; auto-merge;
{% if list(map) %}
		elements = { {{ list(map) }} }
{% endif %}
	}
{%- endmacro %}
{% macro addr_list_set(name, type, items) -%}
set {{ name }} {
		type {{ type }}; flags interval; auto-merge;
{% if items %}
		elements = { {{ items | join(', ') }} }
{% endif %}
	}
{%- endmacro %}
{% set cluster = firewall.cluster | default({}, true) %}
{# public_ips of every inventory host: IPv4 per address, IPv6 per /64 (a longer prefix is widened, nft masks the host bits) #}
{% set public_ips = groups['all'] | map('extract', hostvars, 'public_ips') | select('defined') | flatten | map('string') | unique | list %}
{% set exempt_v4 = public_ips | reject('search', ':') | map('regex_replace', '^([^/]+)$', '\\1/32') | list %}
{% set exempt_v6 = public_ips | select('search', ':') | map('regex_replace', '^([^/]+)$', '\\1/128') | map('regex_replace', '/(6[5-9]|[7-9][0-9]|1[01][0-9]|12[0-8])$', '/64') | list %}
{% set cluster_ifaces = cluster.ifaces | default({}, true) | dict2items | rejectattr('value', 'none') | map(attribute='value') | unique | list %}
{% set internal_ifaces = firewall.internal_ifaces | default({}, true) | dict2items | rejectattr('value', 'none') | map(attribute='value') | unique | list %}
{% set guest_ifaces = firewall.guest_ifaces | default({}, true) | dict2items | rejectattr('value', 'none') | map(attribute='value') | unique | list %}
{% set guest_match = 'iifname { "' ~ (guest_ifaces | join('", "')) ~ '" }' %}
{% set dnat = firewall.dnat | default({}, true) | dict2items | rejectattr('value', 'none') | map(attribute='value') | list %}
{% set masquerade = firewall.masquerade | default({}, true) | dict2items | rejectattr('value', 'none') | map(attribute='value') | unique | list %}
{% set scrapers = firewall.scrapers | default({}, true) %}
{% set forward_guard = firewall.forward_guard | default({}, true) | dict2items | rejectattr('value', 'none') | list %}
{% set blocklists = firewall.spamhaus or firewall.firehol %}
{% macro port_values(map) -%}
{{ map | default({}, true) | dict2items | rejectattr('value', 'none') | map(attribute='value') | map('string') | unique | join(' ') }}
{%- endmacro %}
{# Per-service limits apply only to ports the host serves: input for the public and geo ports, forward for DNATed traffic (port forwards of the host and Docker-published ports). #}
{% set served_input = (port_values(firewall.public.tcp) ~ ' ' ~ port_values(firewall.geo.tcp)).split() %}
{% set served_forward = (port_values(firewall.public.tcp) ~ ' ' ~ port_values(firewall.geo.tcp) ~ ' ' ~ (dnat | map(attribute='tcp') | map('default', {}, true) | map('dict2items') | sum(start=[]) | rejectattr('value', 'none') | map(attribute='value') | map('string') | join(' '))).split() %}
{% set rl = firewall.ratelimit | default({}, true) %}
{% set rl_factor = rl.range_factor | default({}, true) %}
{% set rl_services = [] %}
{% for svc in rl.services | default({}, true) | dict2items | rejectattr('value', 'none') %}
{% set ports = [svc.value.ports] | flatten | map('string') | list %}
{% set per = svc.value.per | default('minute', true) %}
{% set burst = svc.value.burst | default(svc.value.rate * 2, true) %}
{% set _ = rl_services.append({
	'name': svc.key | replace('-', '_'),
	'ports': ports | join(', '),
	'rate': svc.value.rate,
	'burst': burst,
	'per': per,
	'f4': svc.value.range_factor_v4 | default(rl_factor.v4 | default(6), true),
	'f6': svc.value.range_factor_v6 | default(rl_factor.v6 | default(3), true),
	'input': (ports | select('in', served_input) | list | length) > 0,
	'forward': (ports | select('in', served_forward) | list | length) > 0,
}) %}
{% endfor %}
{% set rl_input = rl_services | selectattr('input') | list %}
{% set rl_forward = rl_services | selectattr('forward') | list %}

# Declare-then-delete makes every load one atomic transaction that replaces
# only these tables; Cilium, Docker, NetBird and CrowdSec tables stay untouched.
table inet host {}
delete table inet host
table ip host_nat {}
delete table ip host_nat

{% if firewall.geo_countries %}
include "/etc/nftables.d/host-geo.nft"
{% endif %}
{% if firewall.spamhaus %}
include "/etc/nftables.d/host-spamhaus.nft"
{% endif %}
{% if firewall.firehol %}
include "/etc/nftables.d/host-firehol.nft"
{% endif %}
{% if firewall.scanners %}
include "/etc/nftables.d/host-scanners.nft"
{% endif %}

table inet host {
	{{ addr_set('admin_v4', 'ipv4_addr', firewall.admin.v4) }}
	{{ addr_set('admin_v6', 'ipv6_addr', firewall.admin.v6) }}
	{{ addr_set('cluster_v4', 'ipv4_addr', cluster.v4) }}
	{{ addr_set('cluster_v6', 'ipv6_addr', cluster.v6) }}
	{{ addr_set('scrapers_v4', 'ipv4_addr', scrapers.v4) }}
	{{ addr_set('scrapers_v6', 'ipv6_addr', scrapers.v6) }}
	{{ addr_list_set('rl_exempt_v4', 'ipv4_addr', exempt_v4) }}
	{{ addr_list_set('rl_exempt_v6', 'ipv6_addr', exempt_v6) }}
{% for svc in rl_services if svc.input or svc.forward %}
	set rl_{{ svc.name }}_v4 {
		type ipv4_addr; flags dynamic; timeout 10m; size 65535;
	}
	set rl_{{ svc.name }}_v6 {
		type ipv6_addr; flags dynamic; timeout 10m; size 65535;
	}
	set rl_{{ svc.name }}_v4r {
		type ipv4_addr; flags dynamic; timeout 10m; size 16384;
	}
	set rl_{{ svc.name }}_v6r {
		type ipv6_addr; flags dynamic; timeout 10m; size 16384;
	}
	set rl_hs_{{ svc.name }}_v4 {
		type ipv4_addr; flags dynamic; timeout 10m; size 65535;
	}
	set rl_hs_{{ svc.name }}_v6 {
		type ipv6_addr; flags dynamic; timeout 10m; size 65535;
	}
{% endfor %}

	chain input {
		type filter hook input priority filter; policy drop;
{% if blocklists %}

		# Before the established rule: outbound is blocked too, so listed
		# addresses have no legitimate replies.
{% if firewall.spamhaus %}
		ip saddr @spamhaus_v4 jump spamhaus_in
		ip6 saddr @spamhaus_v6 jump spamhaus_in
{% endif %}
{% if firewall.firehol %}
		ip saddr @firehol_v4 jump firehol_in
{% endif %}

{% endif %}
{% if rl_input %}
		# Before the established rule: the packet that completes a handshake is the second original-direction packet
		# (needs net.netfilter.nf_conntrack_acct).
		meta l4proto tcp ct state established ct direction original ct original packets 2 jump ratelimit_hs_in
{% endif %}
		ct state established,related accept
{% if firewall.scanners %}

		# After the established rule: only inbound connections of scanners drop.
		ip saddr @scanners_v4 jump scanners_in
		ip6 saddr @scanners_v6 jump scanners_in

{% endif %}
		ct state invalid jump drop_log
		iif lo accept

		icmp type { echo-request, destination-unreachable, time-exceeded, parameter-problem } accept
		icmpv6 type { echo-request, destination-unreachable, packet-too-big, time-exceeded, parameter-problem, nd-router-advert, nd-neighbor-solicit, nd-neighbor-advert, mld-listener-query } accept
		udp sport 67 udp dport 68 accept
		meta nfproto ipv6 udp sport 547 udp dport 546 accept
{% if cluster_ifaces or list(cluster.v4) or list(cluster.v6) %}

		# ponytail: the cluster ring trusts every port between nodes and pods;
		# per-port lists (6443, 10250, 2379-2380, 4240, 8472, ...) are the upgrade path.
{% for iface in cluster_ifaces %}
		iifname "{{ iface }}" accept
{% endfor %}
		iifname "wt0" ip saddr @cluster_v4 accept
		iifname "wt0" ip6 saddr @cluster_v6 accept
{% endif %}
{% if internal_ifaces %}

		# ponytail: internal bridges are trusted on every port; per-port lists are the upgrade path.
{% for iface in internal_ifaces %}
		iifname "{{ iface }}" accept
{% endfor %}
{% endif %}
{% if guest_ifaces %}

		# Pods on k3s guests egress through the node IP, so guest bridges get
		# only DHCP and router solicitations beyond the public zone.
		{{ guest_match }} udp dport 67 accept
		{{ guest_match }} icmpv6 type nd-router-solicit accept
{% endif %}

		# VPN peers get the vpn ports, then everything public.
		iifname "wt0" jump vpn
		jump public

		# Broadcasts and multicasts are never scans; they stay out of the log.
		meta pkttype { broadcast, multicast } drop
		jump drop_log
	}

	# A single rule makes the limit global: a flood logs at the limit and the
	# rest drops silently. The kernel log line has the iptables LOG fields that
	# the CrowdSec iptables-logs parser reads.
	chain drop_log {
		limit rate {{ firewall.log.in_rate }} log prefix "fw-in-drop: "
		drop
	}
{% if rl_services | selectattr('input') | list or rl_forward %}

	# Per-service limits on new inbound TCP connections (firewall.ratelimit): per source address (IPv4 /32, IPv6 /64) and per
	# range (IPv4 /24, IPv6 /48; the per-source rate times range_factor). Admin, cluster, VPN, guest and internal sources
	# and the public addresses of all inventory hosts (rl_exempt_v4, rl_exempt_v6) are exempt: cluster nodes reach other hosts
	# with their public address (ArgoCD git fetches, mail submission). The public set is checked only here, no other chain
	# trusts it. The SYN meters drop the excess with the log prefixes fw-rl-syn-addr: (per address) and fw-rl-syn-range: (per range),
	# which a spoofed source can trigger: evidence only. The handshake meters count the packet that completes a handshake
	# (the second original-direction packet of an established connection), which a spoofed source cannot send: the excess is
	# logged with fw-rl-hs: and reset. Only the address tier exists there; ranges are left to CrowdSec.
{% macro rl_exempt() %}
		iifname "wt0" return
{% for iface in cluster_ifaces + internal_ifaces + guest_ifaces %}
		iifname "{{ iface }}" return
{% endfor %}
		ip saddr @admin_v4 return
		ip6 saddr @admin_v6 return
		ip saddr @cluster_v4 return
		ip6 saddr @cluster_v6 return
		ip saddr @rl_exempt_v4 return
		ip6 saddr @rl_exempt_v6 return
{% endmacro %}
{% for hook, svcs in [('in', rl_input), ('fwd', rl_forward)] if svcs %}
	chain ratelimit_{{ hook }} {
{{ rl_exempt() }}
{% for svc in svcs %}
		meta l4proto tcp ct original proto-dst { {{ svc.ports }} } counter jump rl_{{ svc.name }}
{% endfor %}
	}

{% endfor %}
{% for hook, svcs in [('in', rl_input), ('fwd', rl_forward)] if svcs %}
	chain ratelimit_hs_{{ hook }} {
{{ rl_exempt() }}
{% for svc in svcs %}
		meta l4proto tcp ct original proto-dst { {{ svc.ports }} } counter jump rl_hs_{{ svc.name }}
{% endfor %}
	}

{% endfor %}
{% for svc in rl_services if svc.input or svc.forward %}
	chain rl_{{ svc.name }} {
		update @rl_{{ svc.name }}_v4 { ip saddr limit rate over {{ svc.rate }}/{{ svc.per }} burst {{ svc.burst }} packets } counter jump rl_drop_addr
		update @rl_{{ svc.name }}_v6 { ip6 saddr & ffff:ffff:ffff:ffff:: limit rate over {{ svc.rate }}/{{ svc.per }} burst {{ svc.burst }} packets } counter jump rl_drop_addr
		update @rl_{{ svc.name }}_v4r { ip saddr & 255.255.255.0 limit rate over {{ svc.rate * svc.f4 }}/{{ svc.per }} burst {{ svc.burst * svc.f4 }} packets } counter jump rl_drop_range
		update @rl_{{ svc.name }}_v6r { ip6 saddr & ffff:ffff:ffff:: limit rate over {{ svc.rate * svc.f6 }}/{{ svc.per }} burst {{ svc.burst * svc.f6 }} packets } counter jump rl_drop_range
	}

{% endfor %}
{% for svc in rl_services if svc.input or svc.forward %}
	chain rl_hs_{{ svc.name }} {
		update @rl_hs_{{ svc.name }}_v4 { ip saddr limit rate over {{ svc.rate }}/{{ svc.per }} burst {{ svc.burst }} packets } counter jump rl_hs_reject
		update @rl_hs_{{ svc.name }}_v6 { ip6 saddr & ffff:ffff:ffff:ffff:: limit rate over {{ svc.rate }}/{{ svc.per }} burst {{ svc.burst }} packets } counter jump rl_hs_reject
	}

{% endfor %}
	# One log limit per prefix, so a flood of spoofed SYNs cannot use up the budget of the handshake evidence.
	chain rl_drop_addr {
		limit rate {{ firewall.log.ratelimit_rate }} log prefix "fw-rl-syn-addr: "
		drop
	}

	chain rl_drop_range {
		limit rate {{ firewall.log.ratelimit_rate }} log prefix "fw-rl-syn-range: "
		drop
	}

	# The handshake is complete, so the connection has to be ended actively.
	chain rl_hs_reject {
		limit rate {{ firewall.log.ratelimit_rate }} log prefix "fw-rl-hs: "
		reject with tcp reset
	}
{% endif %}
{% if firewall.spamhaus %}

	# Own limits, so a listed scanner cannot use up the budget of fw-in-drop.
	# CrowdSec whitelists both prefixes: the lists already ban these networks.
	chain spamhaus_in {
		limit rate {{ firewall.log.spamhaus_rate }} log prefix "fw-spamhaus-in: "
		drop
	}

	chain spamhaus_out {
		limit rate {{ firewall.log.spamhaus_rate }} log prefix "fw-spamhaus-out: "
		reject
	}
{% endif %}
{% if firewall.firehol %}

	chain firehol_in {
		limit rate {{ firewall.log.spamhaus_rate }} log prefix "fw-firehol-in: "
		drop
	}

	chain firehol_out {
		limit rate {{ firewall.log.spamhaus_rate }} log prefix "fw-firehol-out: "
		reject
	}
{% endif %}
{% if firewall.scanners %}

	# Own limit and prefix, so scanners use up neither the fw-in-drop budget
	# nor the one of the blocklists. CrowdSec whitelists the prefix: scans of
	# these networks are known and need no decision.
	chain scanners_in {
		limit rate {{ firewall.log.spamhaus_rate }} log prefix "fw-scanners-in: "
		drop
	}
{% endif %}
{% if firewall.block_smtp_out or blocklists %}

	# Outbound mail goes through the mail host only. Loopback stays open for a
	# local MTA; the log lines carry OUT=, which CrowdSec does not count as a scan.
	chain output {
		type filter hook output priority filter; policy accept;
{% if firewall.spamhaus %}

		ip daddr @spamhaus_v4 jump spamhaus_out
		ip6 daddr @spamhaus_v6 jump spamhaus_out
{% endif %}
{% if firewall.firehol %}

		ip daddr @firehol_v4 jump firehol_out
{% endif %}
{% if firewall.block_smtp_out %}

		oifname != "lo" tcp dport 25 limit rate {{ firewall.log.out_rate }} log prefix "fw-out-reject: "
		oifname != "lo" tcp dport 25 reject with tcp reset
{% endif %}
	}
{% endif %}

	chain vpn {
{% if list(scrapers.tcp) %}
{% if list(scrapers.v4) %}
		ip saddr @scrapers_v4 tcp dport { {{ list(scrapers.tcp) }} } accept
{% endif %}
{% if list(scrapers.v6) %}
		ip6 saddr @scrapers_v6 tcp dport { {{ list(scrapers.tcp) }} } accept
{% endif %}
{% endif %}
{% if list(firewall.vpn.tcp) %}
		tcp dport { {{ list(firewall.vpn.tcp) }} } accept
{% endif %}
{% if list(firewall.vpn.udp) %}
		udp dport { {{ list(firewall.vpn.udp) }} } accept
{% endif %}
	}

	chain public {
{% if rl_input %}
		ct state new jump ratelimit_in
{% endif %}
{% if list(firewall.public.tcp) %}
		tcp dport { {{ list(firewall.public.tcp) }} } accept
{% endif %}
{% if list(firewall.public.udp) %}
		udp dport { {{ list(firewall.public.udp) }} } accept
{% endif %}
{% if list(firewall.geo.tcp) %}
		tcp dport { {{ list(firewall.geo.tcp) }} } jump geo
{% endif %}
{% if list(firewall.admin.tcp) %}
		tcp dport { {{ list(firewall.admin.tcp) }} } jump admin
{% endif %}
	}
{% if list(firewall.admin.tcp) %}

	chain admin {
		ip saddr @admin_v4 accept
		ip6 saddr @admin_v6 accept
	}
{% endif %}

	chain geo {
{% if guest_ifaces %}
		# Guest IPv6 addresses sit inside the geo sets.
		{{ guest_match }} return
{% endif %}
		ip saddr @admin_v4 accept
		ip6 saddr @admin_v6 accept
{% if firewall.geo_countries %}
		ip saddr @geo_v4 accept
		ip6 saddr @geo_v6 accept
{% endif %}
	}
{% if forward_guard or blocklists or firewall.scanners or rl_forward %}

	chain forward {
		type filter hook forward priority filter; policy accept;
{% if rl_forward %}

		# Docker-published ports and port forwards arrive here after DNAT; pod and guest egress is not DNATed.
		ct status dnat ct state new jump ratelimit_fwd
		meta l4proto tcp ct status dnat ct state established ct direction original ct original packets 2 jump ratelimit_hs_fwd
{% endif %}
{% if firewall.spamhaus %}

		ip saddr @spamhaus_v4 jump spamhaus_in
		ip6 saddr @spamhaus_v6 jump spamhaus_in
		ip daddr @spamhaus_v4 jump spamhaus_out
		ip6 daddr @spamhaus_v6 jump spamhaus_out
{% endif %}
{% if firewall.firehol %}

		ip saddr @firehol_v4 jump firehol_in
		ip daddr @firehol_v4 jump firehol_out
{% endif %}
{% if firewall.scanners %}

		ct state new ip saddr @scanners_v4 jump scanners_in
		ct state new ip6 saddr @scanners_v6 jump scanners_in
{% endif %}
{% for guard in forward_guard %}

		# {{ guard.key }}
{% set ifaces = guard.value.ifaces | default({}, true) | dict2items | rejectattr('value', 'none') | map(attribute='value') | unique | list %}
{% for fam, l3, icmp in [('v4', 'ip', 'icmp'), ('v6', 'ip6', 'icmpv6')] if list((guard.value[fam] | default({}, true)).daddr) %}
{% set zone = guard.value[fam] %}
{% set match = l3 ~ ' daddr { ' ~ list(zone.daddr) ~ ' }' %}
		{{ match }} ct state established,related accept
{% if ifaces %}
		{{ match }} iifname { "{{ ifaces | join('", "') }}" } accept
{% endif %}
{% if list(zone.tcp) %}
		{{ match }} tcp dport { {{ list(zone.tcp) }} } accept
{% endif %}
{% if list(zone.icmp) %}
		{{ match }} {{ icmp }} type { {{ list(zone.icmp) }} } accept
{% endif %}
		{{ match }} jump drop_log
{% endfor %}
{% endfor %}
	}
{% endif %}
}
{% if dnat or masquerade %}

table ip host_nat {
	chain prerouting {
		type nat hook prerouting priority dstnat; policy accept;
{% for fwd in dnat %}
{% for proto in ['tcp', 'udp'] if list(fwd[proto]) %}
		ip daddr {{ fwd.daddr }} {{ proto }} dport { {{ list(fwd[proto]) }} } dnat to {{ fwd.to }}
{% endfor %}
{% endfor %}
	}

	chain postrouting {
		type nat hook postrouting priority srcnat; policy accept;
{% for match in masquerade %}
		{{ match }} masquerade
{% endfor %}
	}
}
{% endif %}