Hi guys, I have a weird problem going on in one of my PFsense boxes, I
would really appreciate you insight into the problem. To give you a
heads up, when I run the box for a couple of days with VLAN trunking,
there comes a moment when the machine starts closing up access and
eventually prohibits all traffic through it. What I mean by
prohibiting is that although I have permit any/any rules on an
interface, I can only ping from the PFSense to the client machine, but
I can't ping from the client machine to the PFSense. I can even reboot
the machine and everything will stay the same from the moment that
access was blocked.

Now for the story:

I've been setting up a new PFSense box with 5 interfaces, two wans and
3 lans, in such a way that I can send traffic out the two providers,
and maintain control between the LANs. One lan is for admin staff, one
is for guests and the other is for guests over a wireless connection,
with a captive portal for access control.

The machine I was given to accomplish the task actually only has 3 pci
slots, so if I use the onboard ethernet controller, I have only 4
network interfaces. The first solution I tried was using a USB network
card but that proved too unstable, and you guys helped me by
recommending to stick to 3Com and Intel for production purposes.

What I decided to do, since I was putting in new switches that support
trunking, was to use that feature in the PFSense box and use only a
couple of interfaces.

Since the onboard interface uses de "vr" network driver that doesn't
support trunking, I installed a 3Com 509 SOHO board (xl driver) to
accomplish the trunking. The way I set up the VLANs was like this:

VLAN1: Administrative Network
VLAN2: Guests Wired Network
VLAN3: Guests Wireless Network
VLAN10: Provider 1
VLAN11: Provider 2

I used the onboard NIC for the admin network, and the 3Com NIC for all
the other networks, both connected to the same switch, but the onboard
nic was connected to an "access" or "untagged" port while the 3Com NIC
was connected to a trunk or tagged port.

I was having problems getting the box to accept any traffic and I was
seeing some of the VLAN traffic come in through the xl0 interface (the
3Com card) which was being blocked so I decided to add VLAN1 to the
3Com NIC. After I did that, everything worked fine, I could ping the
firewall from the clients and vice-versa and route the traffic as
expected. I did not configure the two providers as that was part of
the next step of the project, but I did enable advanced NAT and set up
the rules for each subnet.

I proceeded to configure all the firewall rules, captive portal,
dynamic dns update client, pptp server and dhcp server and did some
tests and everything worked fine, I rebooted the firewall just to make
sure and everything kept on working.

After about 4 or 5 days, the system blocked access just like when I
had the VLAN1 problem and there was no way of getting it to work
again, rebooting the PFSense box or changing rules led to no change, I
eventually had to remove the box from production.

Seeing the symptoms I have several hypothesis that I wanted to share with you:
+The firewall is seeing duplicate frames coming in from interface vr0
and VLAN1 and it's confusing PF or looking like a DoS or something. I
have no way of blocking VLAN1 traffic from appearing on the switch
trunk port connected to the 3Com NIC so my next test, if you think
it's worthwhile is to disconnect the vr0 nic and try routing all
traffic through the 3Com card.
+There is something going on with the NAT, seeing as I can only ping
firewall->client and not client->firewall, I thought this might have
something to do with NAT or maybe I have used advanced nat
inapropiately, currently my mappings are:
VLAN1: 192.168.1.0/24 -> outbound NAT out default interface
VLAN2: 10.1.25.0/24 -> outbound NAT out default interface
VLAN3: 10.1.26.0/24 -> outbound NAT out default interface
VLAN4: this is the default interface for nat, gets it's address via DHCP
VLAN5: this is the alternate WAN connection, hasn't been set up but
works via static IP, there aren't any rules that use it yet.
+Maybe there is something in the switch that is seeing traffic come
back from the PFSense box that it considers a loop and it shuts down
some forwarding considering it's a spanning tree loop
+There might be a bug in the vr driver which is causing the instability

I can send you the config backups if you'd like to take a look at
them. I would also like to help if I can with the VLAN documentation,
I have a lot of experience with the switch side of the connections but
not the PFSense/BSD side.

I really appreciate you taking the time to read this, any insight or
suggestions are super welcome!

Thank you all!

Esteban Zarikian

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to