Bug #958
closedreply-to for 1:1 from other directly connected subnets not functioning correctly
Added by Chris Buechler almost 14 years ago. Updated almost 14 years ago.
0%
Description
Where you have a system with two WANs, such as WAN1 and WAN2, when sourcing traffic from a host on the WAN1's IP subnet to a 1:1 NAT on WAN2, connectivity does not work. The SYN is passed in correctly, the internal host responds with its SYN ACK, but the firewall sends that SYN ACK out of WAN1, with the private IP as the source IP.
Ermal committed a potential fix for this today, just opening this for tracking.
Updated by Chris Buechler almost 14 years ago
- Status changed from Feedback to New
this reportedly not fixed
Updated by Adam Thompson almost 14 years ago
Here's some hard data describing the problem. The host "lisa.muug.mb.ca" is connected nearby via MRNet, but still needs to send email via the Terago (public) interface. The most obvious symptom is that email stops working for all MRNet-connected hosts (which includes all of CA*Net, Internet2, etc. - all research & education sites, basically), but this does affect all NAT'd connections as far as I can tell. I ran "nc remote.c3a.ca 25" from lisa.muug.mb.ca to generate these traces - "telnet remote.c3a.ca 25" or just sending an email produces the same behaviour.
-Adam
MRNET (wan) -> em2 -> 192.139.69.168
C3A_LAN (lan) -> em0 -> 192.168.232.1
MRPS_LAN (opt1) -> em1_vlan233 -> 192.168.233.1
TERAGO (opt2) -> em3 -> 67.226.137.177
First, with BGP FIB coupling turned OFF, the path is out em3:
[2.0-BETA4][root@pfsense.c3a.ca]/root(8): traceroute lisa.muug.mb.ca
traceroute to lisa.muug.mb.ca (130.179.31.46), 64 hops max, 40 byte packets
1 172.16.30.30 (172.16.30.30) 3.206 ms 3.634 ms 4.214 ms
2 10.64.24.9 (10.64.24.9) 7.587 ms 12.950 ms 7.557 ms
3 10.64.24.58 (10.64.24.58) 4.972 ms 5.165 ms 5.103 ms
4 h64-141-108-9.bigpipeinc.com (64.141.108.9) 7.094 ms 8.673 ms 6.847 ms
5 * * h64-141-108-18.bigpipeinc.com (64.141.108.18) 8.440 ms
6 cwrouter-ci.cc.umanitoba.ca (192.139.114.22) 7.467 ms 7.274 ms 7.842 ms
7 enrouter-cw.cc.umanitoba.ca (130.179.42.202) 8.598 ms 7.804 ms 8.729 ms
8 e2b1-enrouter.cc.umanitoba.ca (130.179.41.42) 9.470 ms 10.094 ms 10.064 ms
9 * * *
10 * * *
11 *^C
[2.0-BETA4][root@pfsense.c3a.ca]/root(1): tcpdump -i em3 host lisa.muug.mb.ca
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on em3, link-type EN10MB (Ethernet), capture size 96 bytes
20:10:51.463801 IP lisa.muug.mb.ca.54542 > static-67-226-137-178.ptr.terago.net.smtp: Flags [S], seq 1291705018, win 5840, options [mss 1460,sackOK,TS val 379025534 ecr 0,nop,wscale 7], length 0
20:10:51.464846 IP static-67-226-137-178.ptr.terago.net.smtp > lisa.muug.mb.ca.54542: Flags [S.], seq 1974267749, ack 1291705019, win 8192, options [mss 1460,nop,wscale 8,sackOK,TS val 30140665 ecr 379025534], length 0
20:10:51.471046 IP lisa.muug.mb.ca.54542 > static-67-226-137-178.ptr.terago.net.smtp: Flags [.], ack 1, win 46, options [nop,nop,TS val 379025542 ecr 30140665], length 0
20:10:51.476244 IP static-67-226-137-178.ptr.terago.net.smtp > lisa.muug.mb.ca.54542: Flags [P.], ack 1, win 514, options [nop,nop,TS val 30140666 ecr 379025542], length 89
20:10:51.484662 IP lisa.muug.mb.ca.54542 > static-67-226-137-178.ptr.terago.net.smtp: Flags [.], ack 90, win 46, options [nop,nop,TS val 379025555 ecr 30140666], length 0
20:10:54.721215 IP lisa.muug.mb.ca.54542 > static-67-226-137-178.ptr.terago.net.smtp: Flags [F.], seq 1, ack 90, win 46, options [nop,nop,TS val 379028792 ecr 30140666], length 0
20:10:54.721670 IP static-67-226-137-178.ptr.terago.net.smtp > lisa.muug.mb.ca.54542: Flags [.], ack 2, win 514, options [nop,nop,TS val 30140990 ecr 379028792], length 0
20:10:54.722275 IP static-67-226-137-178.ptr.terago.net.smtp > lisa.muug.mb.ca.54542: Flags [F.], seq 90, ack 2, win 514, options [nop,nop,TS val 30140991 ecr 379028792], length 0
20:10:54.727965 IP lisa.muug.mb.ca.54542 > static-67-226-137-178.ptr.terago.net.smtp: Flags [.], ack 91, win 46, options [nop,nop,TS val 379028800 ecr 30140991], length 0
[2.0-BETA4][root@pfsense.c3a.ca]/root(1): tcpdump -i em0 host lisa.muug.mb.ca
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on em0, link-type EN10MB (Ethernet), capture size 96 bytes
20:10:51.464003 IP lisa.muug.mb.ca.54542 > remote.c3a.ca.smtp: Flags [S], seq 1291705018, win 5840, options [mss 1460,sackOK,TS val 379025534 ecr 0,nop,wscale 7], length 0
20:10:51.464746 IP remote.c3a.ca.smtp > lisa.muug.mb.ca.54542: Flags [S.], seq 1974267749, ack 1291705019, win 8192, options [mss 1460,nop,wscale 8,sackOK,TS val 30140665 ecr 379025534], length 0
20:10:51.471082 IP lisa.muug.mb.ca.54542 > remote.c3a.ca.smtp: Flags [.], ack 1, win 46, options [nop,nop,TS val 379025542 ecr 30140665], length 0
20:10:51.476209 IP remote.c3a.ca.smtp > lisa.muug.mb.ca.54542: Flags [P.], ack 1, win 514, options [nop,nop,TS val 30140666 ecr 379025542], length 89
20:10:51.484692 IP lisa.muug.mb.ca.54542 > remote.c3a.ca.smtp: Flags [.], ack 90, win 46, options [nop,nop,TS val 379025555 ecr 30140666], length 0
20:10:54.721293 IP lisa.muug.mb.ca.54542 > remote.c3a.ca.smtp: Flags [F.], seq 1, ack 90, win 46, options [nop,nop,TS val 379028792 ecr 30140666], length 0
20:10:54.721566 IP remote.c3a.ca.smtp > lisa.muug.mb.ca.54542: Flags [.], ack 2, win 514, options [nop,nop,TS val 30140990 ecr 379028792], length 0
20:10:54.722249 IP remote.c3a.ca.smtp > lisa.muug.mb.ca.54542: Flags [F.], seq 90, ack 2, win 514, options [nop,nop,TS val 30140991 ecr 379028792], length 0
20:10:54.728006 IP lisa.muug.mb.ca.54542 > remote.c3a.ca.smtp: Flags [.], ack 91, win 46, options [nop,nop,TS val 379028800 ecr 30140991], length 0
(no matching traffic recorded on either em1 or em2)
Then, with BGP FIB coupling turned ON, the path is out em2 instead:
[2.0-BETA4][root@pfsense.c3a.ca]/root(20): traceroute lisa.muug.mb.ca
traceroute to lisa.muug.mb.ca (130.179.31.46), 64 hops max, 40 byte packets
1 192.139.69.161 (192.139.69.161) 0.765 ms 0.395 ms 0.780 ms
2 * cirouter-mr.cc.umanitoba.ca (192.139.69.18) 0.865 ms 0.804 ms
3 cwrouter-ci.cc.umanitoba.ca (192.139.114.22) 1.334 ms 1.594 ms 1.416 ms
4 alrouter-cw.cc.umanitoba.ca (130.179.42.206) 1.370 ms 1.416 ms 1.305 ms
5 e2b1-alrouter.cc.umanitoba.ca (130.179.41.46) 2.351 ms 2.978 ms 2.202 ms
6 lisa.muug.mb.ca (130.179.31.46) 1.496 ms 1.385 ms 1.478 ms
[2.0-BETA4][root@pfsense.c3a.ca]/root(3): tcpdump -i em0 host lisa.muug.mb.ca
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on em0, link-type EN10MB (Ethernet), capture size 96 bytes
20:16:27.241040 IP lisa.muug.mb.ca.43897 > remote.c3a.ca.smtp: Flags [S], seq 1645264572, win 5840, options [mss 1460,sackOK,TS val 379361355 ecr 0,nop,wscale 7], length 0
20:16:27.241308 IP remote.c3a.ca.smtp > lisa.muug.mb.ca.43897: Flags [S.], seq 2209064005, ack 1645264573, win 8192, options [mss 1460,nop,wscale 8,sackOK,TS val 30174242 ecr 379361355], length 0
20:16:30.231888 IP remote.c3a.ca.smtp > lisa.muug.mb.ca.43897: Flags [S.], seq 2209064005, ack 1645264573, win 8192, options [mss 1460,nop,wscale 8,sackOK,TS val 30174541 ecr 379361355], length 0
20:16:36.227075 IP remote.c3a.ca.smtp > lisa.muug.mb.ca.43897: Flags [S.], seq 2209064005, ack 1645264573, win 65535, options [mss 1460,sackOK,TS val 30175141 ecr 379361355], length 0
20:16:48.226350 IP remote.c3a.ca.smtp > lisa.muug.mb.ca.43897: Flags [R], seq 2209064006, win 0, length 0
(no matching traffic on em1)
[2.0-BETA4][root@pfsense.c3a.ca]/root(2): tcpdump -i em2 host lisa.muug.mb.ca
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on em2, link-type EN10MB (Ethernet), capture size 96 bytes
20:16:17.058992 IP static-67-226-137-178.ptr.terago.net.smtp > lisa.muug.mb.ca.43896: Flags [R], seq 4175060663, win 0, length 0
20:16:27.241380 IP static-67-226-137-178.ptr.terago.net.smtp > lisa.muug.mb.ca.43897: Flags [S.], seq 2209064005, ack 1645264573, win 8192, options [mss 1460,nop,wscale 8,sackOK,TS val 30174242 ecr 379361355], length 0
20:16:30.231960 IP static-67-226-137-178.ptr.terago.net.smtp > lisa.muug.mb.ca.43897: Flags [S.], seq 2209064005, ack 1645264573, win 8192, options [mss 1460,nop,wscale 8,sackOK,TS val 30174541 ecr 379361355], length 0
20:16:36.227147 IP static-67-226-137-178.ptr.terago.net.smtp > lisa.muug.mb.ca.43897: Flags [S.], seq 2209064005, ack 1645264573, win 65535, options [mss 1460,sackOK,TS val 30175141 ecr 379361355], length 0
20:16:48.226430 IP static-67-226-137-178.ptr.terago.net.smtp > lisa.muug.mb.ca.43897: Flags [R], seq 2209064006, win 0, length 0
[2.0-BETA4][root@pfsense.c3a.ca]/root(5): tcpdump -i em3 host lisa.muug.mb.ca
tcpdump: verbose output suppressed, use -v or -vv for full protocol decode
listening on em3, link-type EN10MB (Ethernet), capture size 96 bytes
20:16:27.240856 IP lisa.muug.mb.ca.43897 > static-67-226-137-178.ptr.terago.net.smtp: Flags [S], seq 1645264572, win 5840, options [mss 1460,sackOK,TS val 379361355 ecr 0,nop,wscale 7], length 0
Note that the reply's source address is correct, it's just leaving out the wrong interface. So the NAT is half-working... this looks like something that I ought to be able to work around using reply-to rules, no?
Updated by Adam Thompson almost 14 years ago
I'm open to workarounds as well as "fixing" the problem - this is suddenly getting a lot more important for us.
Updated by Chris Buechler almost 14 years ago
- Priority changed from Normal to High
This attempted fix has caused issues with reply-to across the board.
Updated by Ermal Luçi almost 14 years ago
- Status changed from New to Feedback
Updated by Jim Pingle almost 14 years ago
- Status changed from Feedback to Resolved
Updated by Adam Thompson almost 14 years ago
I finally had a chance to upgrade, and I'm sorry, but this still doesn't work (for me).
Now testing with 2.0-BETA4 (i386) built on Fri Nov 5 01:04:55 EDT 2010 FreeBSD 8.1-RELEASE-p1.
Using the exact same scenario as above, I still see the reply traffic trying to leave on em2 instead of em3.
In fact, it looks like I see exactly the same behaviour as previously documented.
1) do I need to be running a different binary snapshot to test this?
2) are you sure I don't need to specify a gateway or anything special in the firewall rules? Right now, the inbound NAT rule is a floating rule...
Updated by Chris Buechler almost 14 years ago
- Status changed from Resolved to Feedback
Updated by Ermal Luçi almost 14 years ago
Can you show some debugging info's?!
Tracing of traffic etc to actually see what the issue is for you?
Updated by Adam Thompson almost 14 years ago
As I mentioned in #7, the tcpdump output didn't change at all after updating to the snapshot mentioned there.
So the debug output is still identical what what I posted in #2.
I'll update after business hours today and re-test.
Please let me know what other tracing/debug information might help fix this problem, I'll be happy to provide that as soon as possible.
Updated by Adam Thompson almost 14 years ago
I know you said it wasn't relevant, but I tried setting the Gateway parameter on the firewall rule: no effect whatsoever, reply packets still aren't going the right way.
Updated by Ermal Luçi almost 14 years ago
Can you post your config here?
Please sanitize you ip's but leave the subnets real per se.
I still do not understand why you say it does not work! Please post even /tmp/rules.debug to check.
Updated by Adam Thompson almost 14 years ago
I've just upgraded and confirmed that it does work perfectly for port-forward NAT, but it does still break for 1:1 NAT.
I'll email you an unencrypted config.xml (as-is, not anonymized) separately for now, to your @pfsense.org address - I'll be late for an appointment if I stop to sanitize the data right now!
Updated by Ermal Luçi almost 14 years ago
Can you please try adding rules to your opt2 interface to allow the traffic for all your 1:1 ips.
This will make sure that reply-to is added to rules for those addresses and that will make sure it will work.
Since you are doing this from floating rules there is no reply-to added to the resulting rules and that is your problem.
Updated by Adam Thompson almost 14 years ago
Well, I did mention it was a floating rule, back in comment #7...
Adding the rules to the specific interface works perfectly.
Thank you for making this work!!!
Updated by Chris Buechler almost 14 years ago
- Status changed from Feedback to Resolved