Cannot NAT trough OPT1 interface on multiwan

[mikeisfly]

I can confirm, I am having the same issue using the Oct 4th 64bit build. I can provide configurations if needed but I know I have configured everything correctly and the firewall blocks show no blocks. Downgrading back to 2.1.5 fixes the issues with the exact same configuration file. Just to give an over view of what I did.

created a rule under lan to have my plex machine use the second WAN.
Created a NAT (port Forward for TCP 32400) to plex machine
Rule was automatically generated.

Some fixes that I have found:

Changing default gateways to second WAN (Puts everything on second WAN), Not the desired behavior
Downgrading to 2.1.5 (everything works as expected)

I have this running on my home router so I can flash firmware again if Needed.

[eri--]

Actually seems that the one that have posted here have some issues in their configuration.

They have set a generic policy routing rule on their lan which overrides the reply-to(NAT of their OPTx) tha would make things work.

[eSpezi]

Hi Ermal,

on my Dual WAN (WAN1 = Cable Router, WAN2 = PPPoE Modem) installation I successfully can ping each WAN IF.
But I only can reach pfSenses WebGui or anything else on the current Default Gateway.
By current I mean, that if the default Default Gateway (WAN1) is marked as down (unfortunately still way too often false alarm by apinger) then I am able to reach the WebGui via WAN2.

I followed your suggestions, switched to manual outbound and added the propsed rule on top of everything, but that doesn't change anything.

Has anybody had success with this workaround?
Any other suggestions to make it work?

Thanks,
Harry

Edit:
Ermal in my case, you're right. I've policy routing on lan to a failover gateway group. That same scenario used to work fine with 2.15

[mattbunce]

Excuse my ignorance, but I can't quite figure out if there is a resolution to this yet? I see from Gloom this should be working if you set manual NAT rules on the OPT1 interface (not sure exactly how to, or what this means?).

Also, ermal suggests that the problem is that we're using a "generic policy routing rule" - and this is overriding the reply-to. Again, I'm not exactly sure what is meant by "generic policy routing rule" and if there's anything we can do to work around this?

I have run Wireshark on the NAT target machine, and whilst I don't understand what I'm looking at, you can clearly see that the http request I am making to the NAT forwarded port (from an external network) is hitting the target machine, so it does seem to be the reply that is getting lost somewhere.

My setup

WAN = DHCP client
OPT1 = PIA VPN

I have set-up forwarding via Firewall > NAT and manually created a floating rule for TCP/UDP traffic on OPT1 coming in the my target IP.

I have two LAN rules, the first is for target machine and sets the default gateway as OPT1. The second is for all devices and has the default gateway as the router default (which would be WAN).

The floating rule seems to make no difference, I see the traffic within Wireshark whether or not the floating rule is enabled.

[cmb]

It's not working, none of the workarounds are worth trying (short of for experimentation purposes if you want). This is the bug:
https://redmine.pfsense.org/issues/3760

[mattbunce]

Thanks cmb - That saves me spending any more time on this for now!

[eri--]

Please try a next coming snapshot it should be fixed.

[eSpezi]

Hi Ermal,

thanks a lot for addressing the issue!
Is the fix already in the Wed Oct 15 14:37:51 CDT 2014 Build?

Cheers,
Harry

[eri--]

Yeah it should be on that build or newer.

[mattbunce]

Unfortunately, this still doesn't seem to be working for me. I've tried used the automatically generated linked rule and manually creating a floating rule, but nomatter what I try it is still not working for me on the October 16th 09:56:37 version.

Is there anything else I should be looking at?

[phil.davis]

@mattbunce:

Unfortunately, this still doesn't seem to be working for me. I've tried used the automatically generated linked rule and manually creating a floating rule, but nomatter what I try it is still not working for me on the October 16th 09:56:37 version.

Is there anything else I should be looking at?

Follow the bug report also: https://redmine.pfsense.org/issues/3760
On that, cmb indicates he has done some testing and it is still not fixed.

[mattbunce]

Cheers Phil

I have already commented on the bug report too, I just didn't want to get too 'chatty' on there, because it seems the folks on there have done some pretty advanced diagnostics and I didn't want to distract from the good work they were doing :)

[eri--]

Another round of fixes.
Next snapshot should have the missing piece for fixing this.

[mattbunce]

I still can't seem to get this working on the 17th October, 11:27:45 version.

Have you had a chance to try it out ermal? Perhaps I am not configuring the Nat and Rules pages correctly?

M

[mikeisfly]

This is not working for me and I'm using 2.2-BETA (amd64) built on Fri Oct 17 20:02:23 CDT 2014 FreeBSD 10.1-RC2

I have attached my LAN, NAT (Port Forwrd), Manual Outbound NAT Rules.

I have also captured packets on both my WAN, WAN2 and LAN interfaces. On my WAN2 and LAN I can see the Syn, and the Syn Ack packets however I don't see the ack packet. Also there are some retransmissions. When I capture packets on my WAN and filter for port 32400 the capture is blank which tells me that the packet is not being sent out the default gateway. I have attached the .pcap files here in the form of a .txt file if you would like to look at them in wireshark the extension just needs to be changed back to .pcap . Hope this helps

Thanks,

P.S. Not sure why I was trying to embed these images from my onedrive folder but it didn't seem to work so I had to add attachments.

[capture on LAN.txt](/public/imported_attachments/1/capture on LAN.txt)
[capture on wan2.txt](/public/imported_attachments/1/capture on wan2.txt)

[Wolf666]

I think my problem is the same.
If someone has some spare time to read: https://forum.pfsense.org/index.php?topic=82944.0

I am running:

Version 2.2-BETA (amd64)
built on Fri Oct 17 20:02:23 CDT 2014
FreeBSD 10.1-RC2

I am available to test any solution and report back.

[mikeisfly]

It's a known bug and the developers are working on it. Until then if this is a deal breaker you can try going back to 2.1.5 until the issue is resolved. That is what I'm doing.

[stewgoin]

I'm seeing the same issues on the current snapshot – I've worked around the issue with some sourcerouting and having an upstream system on one of the WAN links handle NATting for now. That works (firewall rules forcing a gateway, then having manual NAT rules disabling NAT).

It looked like using firewall rules to force the alternate gateway would make the multiWAN NAT rules start working correctly, but I didn't test that yet. It definitely adds another layer of config management to do this, so I eagerly await having this one squished :)

[cmb]

Don't change anything on your system trying to fix this, it's part of the packet filter that's currently broken in snapshots. It's slightly more correct now, as reply-to does at least return route correctly. But it's sending traffic with broken checksums, which still leaves it broken. Ermal's heading home from our hackathon in a few hours. He'll fix that when he gets back home later in the week.

[stewgoin]

@cmb:

Don't change anything on your system trying to fix this, it's part of the packet filter that's currently broken in snapshots. It's slightly more correct now, as reply-to does at least return route correctly. But it's sending traffic with broken checksums, which still leaves it broken. Ermal's heading home from our hackathon in a few hours. He'll fix that when he gets back home later in the week.

It's just a little portion of the old home network, and thankfully the changes are easy backed out to test any fixes :)

Thanks a ton for the feedback that it's being worked on, aside from this particular snag, 2.2 is looking stellar for my usage.