1

I have a Linux machine which I'm using as a router. It has 5 network interfaces: three separate LANs which it routes between, and two WANs. At the moment I just have one WAN as the default route, and the other one is doing effectively nothing; I've been trying for years without success to get both WANs to work using iptables and ip rules.

The problem I've been having is this: when I try to route a ping through WAN 2 - which requires NAT - my ping gets from my client host to the Linux machine, which then forwards it out via WAN 2 correctly, and it sees the reply come back, but then it does not forward the packet back to my client machine. I've been unable to figure out why it's not forwarding back, despite many searches and reading related questions. (WAN 1 does not require NAT as this is done on an external router.)

A couple days ago I switched from iptables to nftables, since it a) makes the config much easier to read and b) actually lets me trace rule evaluation so I can see what's going on. With that I now feel I have enough to post this question.

Here's my /etc/nftables.conf:

table ip filter {
    chain INPUT {
            type filter hook input priority 0; policy accept;

            ip protocol icmp counter meta nftrace set 1

            # allow loopback
            iifname "lo" accept

            # allow established/related connections
            ct state {established, related} accept

            # allow ping
            ip protocol icmp accept

            # accept anything from local networks
            ip saddr {
                    172.23.0.0/24, # lan1
                    172.23.2.0/24, # routed through lan1
                    172.23.3.0/24, # routed through lan1
                    172.23.4.0/24, # lan2
                    172.23.5.0/24, # lan3
            } accept

            # ntp exploit protection
            udp sport ntp ct state {invalid, related, new, untracked} counter drop

            # accept SSH from anyone else
            ct state new tcp dport ssh accept

            # drop all other packets
            counter drop
    }

    chain FORWARD {
            type filter hook forward priority 0; policy accept;

            ip protocol icmp counter meta nftrace set 1

            # drop anything to old local network 172.23.1.0/24
            ip daddr 172.23.1.0/24 counter drop

            # accept all other packets
            counter accept
    }

    chain OUTPUT {
            type filter hook output priority 0; policy accept;

            # ntp exploit protection
            udp dport ntp ct state {invalid, related, untracked} counter drop
    }
}

table ip mangle {
    chain FORWARD {
            type filter hook forward priority -150; policy accept;

            ip protocol icmp counter meta nftrace set 1
    }

    chain OUTPUT {
            type filter hook output priority -150; policy accept;

            # send replies to WAN->HERE connections via the same route as where they were initiated from
            ct state related,established meta mark set ct mark
    }

    chain PREROUTING {
            type filter hook prerouting priority -150; policy accept;

            # trace ALL packets coming from enp6s0 (WAN 2)
            iifname enp6s0 counter meta nftrace set 1

            # send subsequent packets on forwarded connections via the same route as when they were initiated
            ct state related,established meta mark set ct mark

            # trace all packets with a packet mark
            meta mark != 0x0 counter meta nftrace set 1

            # all further processing is for new connections only - so everything else returns here
            ct state != new return

            # any new WAN->LAN connections from enp6s0 (WAN 2) go into route 3, for the initial and subsequent packets
            # the return on the end ensures we don't do any further processing, which checks outbound protocols
            iifname enp6s0 ct mark set 0x3 meta mark set 0x3 return

            # any new WAN->LAN connections from enp4s0 (WAN 1) shouldn't do further processing either
            iifname enp4s0 return

            # everything from this point onwards is for new outgoing LAN->WAN connections only

            # for testing - route specific protocols through WAN 2
            #tcp dport 443 ct mark set 0x3 meta mark set 0x3
            #tcp dport 80 ct mark set 0x3 meta mark set 0x3
            ip protocol icmp ct mark set 0x3 meta mark set 0x3 counter meta nftrace set 1
    }
}

table ip nat {
    chain POSTROUTING {
            type nat hook postrouting priority 100; policy accept;

            oifname enp6s0 counter meta nftrace set 1 masquerade
    }
}

ip -4 addr: (enp4s0 is WAN 1, enp6s0 is WAN 2, the others are LANs)

1: lo: <LOOPBACK,UP,LOWER_UP> mtu 65536 qdisc noqueue state UNKNOWN group default qlen 1000
    inet 127.0.0.1/8 scope host lo
       valid_lft forever preferred_lft forever
2: enp4s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    inet 192.168.0.3/24 brd 192.168.0.255 scope global enp4s0
       valid_lft forever preferred_lft forever
3: enp5s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    inet 172.23.4.3/24 brd 172.23.4.255 scope global enp5s0
       valid_lft forever preferred_lft forever
4: enp6s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc fq_codel state UP group default qlen 1000
    inet (redacted).117/22 brd 255.255.255.255 scope global enp6s0
       valid_lft forever preferred_lft forever
5: enp7s0: <NO-CARRIER,BROADCAST,MULTICAST,UP> mtu 1500 qdisc fq_codel state DOWN group default qlen 1000
    inet 172.23.5.3/24 brd 172.23.5.255 scope global enp7s0
       valid_lft forever preferred_lft forever
6: enp8s0: <BROADCAST,MULTICAST,UP,LOWER_UP> mtu 1500 qdisc mq state UP group default qlen 1000
    inet 172.23.0.3/24 brd 172.23.0.255 scope global enp8s0
       valid_lft forever preferred_lft forever

ip route:

default via 192.168.0.1 dev enp4s0
(redacted).0/22 dev enp6s0 proto kernel scope link src (redacted).117 metric 204 mtu 1500
172.23.0.0/24 dev enp8s0 proto kernel scope link src 172.23.0.3
172.23.0.0/16 via 172.23.0.2 dev enp8s0
172.23.4.0/24 dev enp5s0 proto kernel scope link src 172.23.4.3
172.23.5.0/24 dev enp7s0 proto kernel scope link src 172.23.5.3 linkdown
192.168.0.0/24 dev enp4s0 proto kernel scope link src 192.168.0.3

ip route show table 3:

default via (redacted).1 dev enp6s0
(redacted).1 dev enp6s0 scope link src (redacted).117
172.23.0.0/24 dev enp8s0 proto kernel scope link src 172.23.0.3
172.23.0.0/16 via 172.23.0.2 dev enp8s0
172.23.4.0/24 dev enp5s0 proto kernel scope link src 172.23.4.3
172.23.5.0/24 dev enp7s0 proto kernel scope link src 172.23.5.3 linkdown
192.168.0.0/24 dev enp4s0 proto kernel scope link src 192.168.0.3

ip rule:

0:      from all lookup local
32764:  from all fwmark 0x3 lookup 3
32765:  from (redacted).117 lookup 3
32766:  from all lookup main
32767:  from all lookup default

And now the fun bit, here's the output of nft monitor trace when I ping 8.8.8.8 from my client (Windows) PC:

trace id 8e85e085 ip mangle PREROUTING packet: iif "enp8s0" ether saddr dc:9f:db:16:42:b5 ether daddr 38:ea:a7:ab:f8:bc ip saddr 172.23.2.132 ip daddr 8.8.8.8 ip dscp cs0 ip ecn not-ect ip ttl 127 ip id 4170 ip length 60 icmp type echo-request icmp code 0 icmp id 1 icmp sequence 779
trace id 8e85e085 ip mangle PREROUTING rule ip protocol icmp ct mark set 0x00000003 mark set 0x00000003 counter packets 0 bytes 0 nftrace set 1 (verdict continue)
trace id 8e85e085 ip mangle PREROUTING verdict continue mark 0x00000003
trace id 8e85e085 ip mangle PREROUTING mark 0x00000003
trace id 8e85e085 ip mangle FORWARD packet: iif "enp8s0" oif "enp6s0" ether saddr dc:9f:db:16:42:b5 ether daddr 38:ea:a7:ab:f8:bc ip saddr 172.23.2.132 ip daddr 8.8.8.8 ip dscp cs0 ip ecn not-ect ip ttl 126 ip id 4170 ip length 60 icmp type echo-request icmp code 0 icmp id 1 icmp sequence 779
trace id 8e85e085 ip mangle FORWARD rule ip protocol icmp counter packets 0 bytes 0 nftrace set 1 (verdict continue)
trace id 8e85e085 ip mangle FORWARD verdict continue mark 0x00000003
trace id 8e85e085 ip mangle FORWARD mark 0x00000003
trace id 8e85e085 ip filter FORWARD packet: iif "enp8s0" oif "enp6s0" ether saddr dc:9f:db:16:42:b5 ether daddr 38:ea:a7:ab:f8:bc ip saddr 172.23.2.132 ip daddr 8.8.8.8 ip dscp cs0 ip ecn not-ect ip ttl 126 ip id 4170 ip length 60 icmp type echo-request icmp code 0 icmp id 1 icmp sequence 779
trace id 8e85e085 ip filter FORWARD rule ip protocol icmp counter packets 0 bytes 0 nftrace set 1 (verdict continue)
trace id 8e85e085 ip filter FORWARD rule counter packets 8 bytes 452 accept (verdict accept)
trace id 8e85e085 ip nat POSTROUTING packet: oif "enp6s0" ip saddr 172.23.2.132 ip daddr 8.8.8.8 ip dscp cs0 ip ecn not-ect ip ttl 126 ip id 4170 ip length 60 icmp type echo-request icmp code 0 icmp id 1 icmp sequence 779
trace id 8e85e085 ip nat POSTROUTING rule oifname "enp6s0" counter packets 0 bytes 0 nftrace set 1 masquerade (verdict accept)
trace id eae785df ip mangle PREROUTING packet: iif "enp6s0" ether saddr 00:01:5c:86:1a:47 ether daddr 00:e0:4c:68:12:d9 ip saddr 8.8.8.8 ip daddr (redacted).117 ip dscp cs0 ip ecn not-ect ip ttl 56 ip id 39719 ip length 60 icmp type echo-reply icmp code 0 icmp id 1 icmp sequence 779
trace id eae785df ip mangle PREROUTING rule iifname "enp6s0" counter packets 0 bytes 0 nftrace set 1 (verdict continue)
trace id eae785df ip mangle PREROUTING rule ct state established,related mark set ct mark (verdict continue)
trace id eae785df ip mangle PREROUTING rule mark != 0x00000000 counter packets 0 bytes 0 nftrace set 1 (verdict continue)
trace id eae785df ip mangle PREROUTING verdict return mark 0x00000003
trace id eae785df ip mangle PREROUTING mark 0x00000003
trace id eae785df ip filter INPUT packet: iif "enp6s0" ether saddr 00:01:5c:86:1a:47 ether daddr 00:e0:4c:68:12:d9 ip saddr 8.8.8.8 ip daddr (redacted).117 ip dscp cs0 ip ecn not-ect ip ttl 56 ip id 39719 ip length 60 icmp type echo-reply icmp code 0 icmp id 1 icmp sequence 779
trace id eae785df ip filter INPUT rule ip protocol icmp counter packets 0 bytes 0 nftrace set 1 (verdict continue)
trace id eae785df ip filter INPUT rule ct state { } accept (verdict accept)

And here's the relevant line from the output of conntrack -L:

icmp     1 15 src=172.23.2.132 dst=8.8.8.8 type=8 code=0 id=1 src=8.8.8.8 dst=(redacted).117 type=0 code=0 id=1 mark=3 use=1

The outbound part has a source of my client's local IP and the destination of the external server I'm pinging, but the inbound part has the external IP of the machine doing the forwarding, not my client's local IP. (I'm not sure if this is indicative of a problem or not.)

As you can see, the echo-request packet correctly has the packet mark and conntrack mark set to 3, it then picks the correct output interface thanks to the ip rules and route table 3, it then gets masqueraded correctly, and clearly gets out to the Internet since I'm getting an echo-reply. The echo-reply packet correctly copies the conntrack mark (which is still 3) to the packet mark... but then as you can see, it is not reversing the NAT which was originally performed, so it's heading into the INPUT chain, instead of being forwarded back to my client PC.

I'm sure I am missing something - I feel like there has to be a rule somewhere to tell it to reverse the NAT operation - but every page I have seen that explains how to do NAT from LAN->WAN says that the only rule you need is the masquerade one on the postrouting of the initial outbound packet (a lot of guides provide other rules for things like port forwarding for inbound connections, but these are irrelevant to simple outbound connections).

What am I missing?

2

The nftables wiki states:

"[...] you have to register the prerouting/postrouting chains even if you have no rules there since these chain will invoke the NAT engine for the packets coming in the reply direction." at https://wiki.nftables.org/wiki-nftables/index.php/Performing_Network_Address_Translation_(NAT)

You appear to have a prerouting chain of type filter but NOT of type nat. Try adding chain PREROUTING { type nat hook prerouting priority -150 ; } to the table ip nat { [...] } section in the /etc/nftable.conf file.

1

I think the problem is that your nat postrouting chain is at priority -100. According to the nftables wiki, DNAT in iptables operated at priority -100 but I think you want SNAT, which in iptables was equivalent to priority (+)100. I hope that helps.

  • Thanks, good spot. The postrouting chain should indeed have been +100 instead of -150. I've changed it, but sadly this by itself hasn't resolved the problem, nor changed the output of nft monitor trace. In case it helps though, I've added relevant output from conntrack -L to my question as well. – Keiji Nov 27 '17 at 20:24

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.