Design considerations with Spanning Tree when Wireless bridges are involved? [closed]

Question

The lab is configured with PVST+ on all switches. Currently I am using wireless bridges to connection different sections of a lab environment for the sake of mobility. Basically just bridging the the wireless interface and ethernet interface. All ethernet ports on the bridge are in the same VLAN.

SW1 ------ Wireless_Bridge_1 ~~~~~~~~~~~~~Wireless_Bridge_2 ------SW2

When frames go through the wireless bridge, the source MAC address is changed to MAC address of the bridge's wireless interface. Does this cause problems with layer 2 features (spanning tree, VTP etc.) At the moment STP seems to be working fine. What if I connect more STP speaking switches to the bridge? I think BPDUs should have all the information needed for STP to work, but the source address overwriting can be kinda confusing. Which layer 2 features can I expect to work over this type of wireless link (aka features that work just like it's an Ethernet link)?

Answer 1

If your DD-WRT devices are setup as AP - WDS and STATION - WDS, then they should act as an ethernet cable... forwarding all macs they see, and not forcing traffic that passes through them to inherit the interface MAC of the wireless bridges.

WDS is not cross-vendor, keep that in mind. There was never any official standard for it.

I can imagine that STP would "break" because it's only seeing a repeating MAC (wireless bridge), but I can't confirm this.

Answer 2

I have dealt with Ethernet-to-WiFi bridges that rewrite MAC addresses (MOXAs, in my case), and I can confirm absolutely that it will wreck your STP, and ARP/ND will get broken if you end up in a scenario where you have multiple paths (e.g., if you use the wireless as a backup to a hard-wired Ethernet connection). You can probably get away with it if you'll only ever be using the wireless connection, but it's essentially the same scenario as NAT-PT.

Check to make sure that the AP really is doing this; the ARP tables on hosts connected to the switches should be able to tell you. DD-WRT is based on Linux, and I haven't seen Linux do "MAC NATting" like this.

How this fails:

Try connecting a host to a switch that has such a wireless bridge connected to one port, and create a mirrored replica on the other side, but don't wire the switches together:

HostA---SwA         SwB---HostB
         |           |
        WiFi--------WiFi

You should be able to make connections and ping back and forth between HostA and HostB. Check the ARP tables, and observe that the MAC address is actually that of the other AP. Now connect a cable between the switches:

HostA---SwA---------SwB---HostB
         |           |
        WiFi--------WiFi

Your connections will stay up, since the network still knows how to send the frames the "long way around" through the WiFi bridge.

If that bridge goes down, however, and you're left with just the hardwired network (that's not rewriting MACs), your connections fall over until the ARP tables age out and the hosts re-request MACs.