Convert firewall to nftables.

Debian Buster uses the nftables framework by default.

4 files changed, 199 insertions, 55 deletions

diff --git a/roles/common/templates/etc/ipsec.conf.j2 b/roles/common/templates/etc/ipsec.conf.j2
index 0ff9fbb..6b3840f 100644
--- a/roles/common/templates/etc/ipsec.conf.j2
+++ b/roles/common/templates/etc/ipsec.conf.j2
@@ -20,7 +20,7 @@ conn %default
     leftsubnet     = {{ ipsec[inventory_hostname_short] | ipv4 }}/32
     leftid         = {{ inventory_hostname }}
     leftsigkey     = {{ inventory_hostname_short }}.pem
-    leftfirewall   = yes
+    leftfirewall   = no
     lefthostaccess = yes
     rightauth      = pubkey
     auto           = route
diff --git a/roles/common/templates/etc/iptables/services.j2 b/roles/common/templates/etc/iptables/services.j2
deleted file mode 100644
index 6dd5aae..0000000
--- a/roles/common/templates/etc/iptables/services.j2
+++ /dev/null
@@ -1,48 +0,0 @@
-# {{ ansible_managed }}
-# Do NOT edit this file directly!
-#
-# direction              protocol     destination port            source port
-# (in|out|inout)[46]?    (tcp|udp|..) (port|port:port|port,port)  (port|port:port|port,port)
-
-{% if groups.all | length > 1 %}
-inout   udp     500      500                            # ISAKMP
-{% if groups.NATed | length > 0 %}
-inout4  udp     4500     4500                           # IPsec NAT Traversal
-{% endif %}
-{% endif %}
-
-out     tcp     80,443                                  # HTTP/HTTPS
-out     udp     53                                      # DNS
-out     tcp     53                                      # DNS
-out     udp     67                                      # DHCP
-out     tcp     22                                      # SSH
-out     udp     123      123                            # NTP
-
-in      tcp     {{ ansible_port|default('22') }}        # SSH
-{% if 'LDAP-provider' in group_names %}
-in      tcp     636                                     # LDAPS
-{% elif 'MX' in group_names or 'lists' in group_names or 'nextcloud' in group_names %}
-out     tcp     636                                     # LDAPS
-{% endif %}
-{% if 'MX' in group_names %}
-in      tcp     25                                      # SMTP
-{% endif %}
-{% if 'out' in group_names or 'MSA' in group_names %}
-out     tcp     25                                      # SMTP
-{% endif %}
-{% if 'IMAP' in group_names %}
-in      tcp     993                                     # IMAPS
-in      tcp     4190                                    # MANAGESIEVE
-out     tcp     2703                                    # Razor2
-{% endif %}
-{% if 'MSA' in group_names %}
-in      tcp     465                                     # SMTP-AUTH
-in      tcp     587                                     # SMTP-AUTH
-{% endif %}
-{% if 'webmail' in group_names or 'lists' in group_names or 'wiki' in group_names or 'nextcloud' in group_names %}
-in      tcp     80,443                                  # HTTP/HTTPS
-{% endif %}
-{% if 'LDAP-provider' in group_names %}
-out     tcp     11371                                   # HKP
-out     tcp     43                                      # WHOIS
-{% endif %}
diff --git a/roles/common/templates/etc/network/if-up.d/ipsec.j2 b/roles/common/templates/etc/network/if-up.d/ipsec.j2
index caa5129..9f183d3 100755
--- a/roles/common/templates/etc/network/if-up.d/ipsec.j2
+++ b/roles/common/templates/etc/network/if-up.d/ipsec.j2
@@ -25,10 +25,9 @@ PATH=/usr/sbin:/usr/bin:/sbin:/bin
 
 # Only the device with the default, globally-scoped route, is of
 # interest here.
-ip="$( ip -4 -o route show to default scope global \
-     | sed -nr '/^default via (\S+) dev (\S+).*/ {s//\2 \1/p;q}' )"
-[ "${ip% *}" = "$IFACE" ] || exit 0
-ip="${ip##* }"
+iface="$( ip -o route show to default scope global \
+        | sed -nr '/^default via \S+ dev (\S+).*/ {s//\1/p;q}' )"
+[ "$iface" = "$IFACE" ] || exit 0
 
 vip="{{ ipsec[inventory_hostname_short] }}"
 vsubnet="{{ ipsec_subnet }}"
@@ -39,9 +38,9 @@ case "$MODE" in
            # in the absence of xfrm lookup (i.e., when there is no
            # matching IPsec Security Association).
            ip route replace prohibit "$vsubnet" proto static || true
-           ip route replace table 220 to "$vsubnet" via "$ip" dev "$IFACE" proto static src "$vip" || true
+           ip route replace table 220 to "$vsubnet" dev "$IFACE" proto static src "$vip" || true
     ;;
-    stop)  ip route del table 220 to "$vsubnet" via "$ip" dev "$IFACE" proto static src "$vip" || true
+    stop)  ip route del table 220 to "$vsubnet" dev "$IFACE" proto static src "$vip" || true
            ip route del prohibit "$vsubnet" proto static || true
            ip address del "$vip/32" dev "$IFACE" scope global || true
 esac
diff --git a/roles/common/templates/etc/nftables.conf.j2 b/roles/common/templates/etc/nftables.conf.j2
new file mode 100755
index 0000000..1e1fde2
--- /dev/null
+++ b/roles/common/templates/etc/nftables.conf.j2
@@ -0,0 +1,193 @@
+#!/usr/sbin/nft -f
+
+define in-tcp-ports = {
+    {{ ansible_port|default(22) }}
+{% if 'MX' in group_names %}
+  , 25      # SMTP
+{% endif %}
+{% if 'LDAP-provider' in group_names %}
+  , 636     # ldaps
+{% endif %}
+{% if 'IMAP' in group_names %}
+  , 993     # imaps
+  , 4190    # ManageSieve
+{% endif %}
+{% if 'MSA' in group_names %}
+  , 587     # submission [RFC4409]
+  , 465     # submission over TLS [RFC8314]
+{% endif %}
+{% if 'webmail' in group_names or 'lists' in group_names or 'wiki' in group_names or 'nextcloud' in group_names %}
+  , 80      # HTTP
+  , 443     # HTTP over SSL/TLS
+{% endif %}
+}
+
+define out-tcp-ports = {
+    22
+  , 80      # HTTP
+  , 443     # HTTP over SSL/TLS
+{% if 'out' in group_names or 'MSA' in group_names %}
+  , 25      # SMTP
+{% endif %}
+{% if 'LDAP-provider' in group_names %}
+  , 11371   # OpenPGP HTTP Keyserver
+  , 43      # whois
+{% elif 'MX' in group_names or 'lists' in group_names or 'nextcloud' in group_names %}
+  , 636     # ldaps
+{% endif %}
+{% if 'IMAP' in group_names %}
+  , 2703    # Razor2
+{% endif %}
+}
+
+
+###############################################################################
+
+flush ruleset
+
+table inet filter {
+    # blackholes
+    set fail2ban  { type ipv4_addr; timeout 10m; }
+    set fail2ban6 { type ipv6_addr; timeout 10m; }
+
+    chain input {
+        type filter hook input priority 0
+        policy drop
+
+        iif lo accept
+
+        # XXX Bullseye: this is a hack for the lack of reqid matches in
+        # nftables: we mark the esp packet and accept after decapsulation
+        # https://serverfault.com/questions/971735/how-to-match-reqid-in-nftables
+        # https://blog.fraggod.net/2016/09/25/nftables-re-injected-ipsec-matching-without-xt_policy.html
+        define IPsec.mark = 0x220
+        meta l4proto esp mark set mark | $IPsec.mark accept
+        ip saddr 172.16.0.0/24 ip daddr 172.16.0.7 mark & $IPsec.mark == $IPsec.mark accept
+
+        # rate-limiting is done directly by the kernel (net.ipv4.icmp_{ratelimit,ratemask} runtime options)
+        icmp   type { echo-reply, echo-request, destination-unreachable } counter accept
+        icmpv6 type { echo-reply, echo-request, destination-unreachable,
+                      packet-too-big, time-exceeded, parameter-problem } counter accept
+
+        # accept neighbour discovery for autoconfiguration, RFC 4890 sec. 4.4.1
+        icmpv6 type { 133,134,135,136,141,142 } ip6 hoplimit 255 counter accept
+
+        jump martian
+        jump invalid
+
+        udp sport  123 udp dport  123 ct state     related,established accept
+{% if groups.all | length > 1 %}
+        udp sport  500 udp dport  500 ct state new,related,established accept
+{% if groups.NATed | length > 0 %}
+        udp sport 4500 udp dport 4500 ct state new,related,established accept
+{% endif %}
+{% endif %}
+
+        udp sport 53 ct state related,established accept
+        tcp sport 53 ct state related,established accept
+{% if 'dhclient' in group_names %}
+        udp sport 67 ct state related,established accept
+{% endif %}
+
+        meta l4proto tcp ip  saddr @fail2ban  counter drop
+        meta l4proto tcp ip6 saddr @fail2ban6 counter drop
+
+        tcp dport $in-tcp-ports  ct state related,established accept
+        tcp dport $in-tcp-ports  ct state new counter accept
+        tcp sport $out-tcp-ports ct state related,established accept
+    }
+
+    chain output {
+        type filter hook output priority 0
+        policy drop
+
+        oif lo accept
+
+        # XXX Bullseye: unlike for input we can't use marks here,
+        # because by the time we see a packet to 172.16.0.0/24 we don't
+        # know if it'll be encapsulated
+        meta l4proto esp accept
+        ip saddr 172.16.0.7 ip daddr 172.16.0.0/24 accept
+
+        meta l4proto { icmp, icmpv6 } accept
+
+        jump martian
+        jump invalid
+
+        udp sport  123 udp dport  123 ct state new,related,established accept
+        udp sport  500 udp dport  500 ct state new,related,established accept
+        udp sport 4500 udp dport 4500 ct state new,related,established accept
+
+        udp dport 53 ct state new,related,established accept
+        tcp dport 53 ct state new,related,established accept
+{% if 'dhclient' in group_names %}
+        udp dport 67 ct state new,related,established accept
+{% endif %}
+
+        tcp sport $in-tcp-ports  ct state related,established accept
+        tcp dport $out-tcp-ports ct state related,established accept
+        tcp dport $out-tcp-ports ct state new counter accept
+
+        meta l4proto tcp counter reject with tcp reset
+        meta l4proto udp counter reject
+        counter reject
+    }
+
+    chain martian {
+        # bogon filter (cf. RFC 6890 for non-global ip addresses)
+        define invalid-ip = {
+            0.0.0.0/8          # this host, on this network (RFC 1122 sec. 3.2.1.3)
+{% if not ansible_default_ipv4.address | ipaddr('10.0.0.0/8') %}
+          , 10.0.0.0/8         # private-use (RFC 1918)
+{% endif %}
+          , 100.64.0.0/10      # shared address space (RFC 6598)
+          , 127.0.0.0/8        # loopback (RFC 1122, sec. 3.2.1.3)
+          , 169.254.0.0/16     # link local (RFC 3927)
+{% if not ansible_default_ipv4.address | ipaddr('172.16.0.0/12') %}
+          , 172.16.0.0/12      # private-use (RFC 1918)
+{% endif %}
+          , 192.0.0.0/24       # IETF protocol assignments (RFC 6890 sec. 2.1)
+          , 192.0.2.0/24       # documentation (RFC 5737)
+{% if not ansible_default_ipv4.address | ipaddr('192.168.0.0/16') %}
+          , 192.168.0.0/16     # private-use (RFC 1918)
+{% endif %}
+          , 198.18.0.0/15      # benchmarking (RFC 2544)
+          , 198.51.100.0/24    # documentation (RFC 5737)
+          , 203.0.113.0/24     # documentation (RFC 5737)
+          , 240.0.0.0/4        # reserved (RFC 1112, sec. 4)
+          , 255.255.255.255/32 # limited broadcast (RFC 0919, section 7)
+        }
+
+        define invalid-ip6 = {
+            ::1/128         # loopback address (RFC 4291)
+          , ::/128          # unspecified (RFC 4291)
+          , ::ffff:0:0/96   # IPv4-mapped address (RFC 4291)
+          , 100::/64        # discard-only address block (RFC 6666)
+          , 2001::/23       # IETF protocol assignments (RFC 2928)
+          , 2001::/32       # TEREDO (RFC 4380)
+          , 2001:2::/48     # benchmarking (RFC 5180)
+          , 2001:db8::/32   # documentation (RFC 3849)
+          , 2001:10::/28    # ORCHID (RFC 4843)
+          , 2002::/16       # 6to4 (RFC 3056)
+          , fc00::/7        # unique-local (RFC 4193)
+          , fe80::/10       # linked-scoped unicast (RFC 4291)
+        }
+
+        ip saddr $invalid-ip counter drop
+        ip daddr $invalid-ip counter drop
+
+        ip6 saddr $invalid-ip6 counter drop
+        ip6 daddr $invalid-ip6 counter drop
+    }
+
+    chain invalid {
+        ct state invalid counter reject
+
+        # drop bogus TCP packets
+        tcp flags & (fin|syn|rst|psh|ack|urg) == 0x0         counter drop # null packets
+        tcp flags != syn ct state new                        counter drop # SYN-flood attacks
+        tcp flags & (fin|syn|rst|psh|ack|urg) == fin|psh|urg counter drop # XMAS packets
+        tcp flags & (fin|syn) == fin|syn                     counter drop # bogus
+        tcp flags & (syn|rst) == syn|rst                     counter drop # bogus
+    }
+}