Firewall: Move martian and bogus TCP filters early in the packet flow.

This is more efficient: the earlier we filter the crap out the less
resources they consume.

1 files changed, 84 insertions, 64 deletions

diff --git a/roles/common/templates/etc/nftables.conf.j2 b/roles/common/templates/etc/nftables.conf.j2
index fc7691a..d788457 100755
--- a/roles/common/templates/etc/nftables.conf.j2
+++ b/roles/common/templates/etc/nftables.conf.j2
@@ -45,6 +45,90 @@ define out-tcp-ports = {
 
 flush ruleset
 
+table netdev filter {
+{% for if in ansible_interfaces %}
+{% if if != "lo" and ansible_facts[if].active %}
+{% set addr = (ansible_facts[if].ipv4 | default({'address': '0.0.0.0'})).address %}
+    chain INGRESS-{{ if }} {
+        type filter hook ingress device {{ if }} priority -499
+        policy accept
+
+        # bogon filter (cf. RFC 6890 for non-global ip addresses)
+        define bogon = {
+            0.0.0.0/8          # this host, on this network (RFC 1122 sec. 3.2.1.3)
+{% if not addr | 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 addr | 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 addr | 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)
+          , 224.0.0.0/3        # multicast - class D 224.0.0.0/4 + class E 240.0.0.0/4 (RFC 1112 sec. 4)
+          , 255.255.255.255/32 # limited broadcast (RFC 0919 sec. 7)
+        }
+
+        ip saddr $bogon counter drop
+        ip daddr $bogon counter drop
+
+        # See also https://www.team-cymru.org/Services/Bogons/fullbogons-ipv6.txt
+        define bogon6 = {
+            ::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)
+        }
+
+        ip6 saddr $bogon6 counter drop
+        ip6 saddr $bogon6 counter drop
+    }
+{% endif %}
+{% endfor %}
+}
+
+table inet raw {
+    chain PREROUTING-stateless {
+        # XXX can't add that to the ingress hook as that happens before IP defragmentation
+        # so we don't have the TCP header in later fragments (we don't want to drop IP
+        # fragments, see https://blog.cloudflare.com/ip-fragmentation-is-broken/ )
+        type filter hook prerouting priority -399 # > NF_IP_PRI_CONNTRACK_DEFRAG (-400)
+        policy accept
+
+        # stateless filter for bogus TCP packets
+        tcp flags & (fin|syn|rst|psh|ack|urg) == 0x0 counter drop # null packet
+        tcp flags & (fin|psh|urg) == fin|psh|urg     counter drop # XMAS packet
+        tcp flags & (syn|rst) == syn|rst             counter drop
+        tcp flags & (fin|rst) == fin|rst             counter drop
+        tcp flags & (fin|syn) == fin|syn             counter drop
+        tcp flags & (fin|psh|ack) == fin|psh         counter drop
+    }
+
+    chain PREROUTING {
+        type filter hook prerouting priority -199 # > NF_IP_PRI_CONNTRACK (-200)
+        policy accept
+
+        # stateful filter
+        ct state invalid counter drop
+    }
+}
+
 table inet filter {
     # blackholes
     set fail2ban  { type ipv4_addr; timeout 10m; }
@@ -75,9 +159,6 @@ table inet filter {
         # 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
@@ -116,9 +197,6 @@ table inet filter {
 
         meta l4proto { icmp, icmpv6 } accept
 
-        jump martian
-        jump invalid
-
         udp sport  123 udp dport  123 ct state new,related,established accept
 {% if groups.all | length > 1 %}
         udp sport  500 udp dport  500 ct state new,related,established accept
@@ -142,62 +220,4 @@ table inet filter {
         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
-    }
 }