We had an AP that had a BVI address of 10.1.1.5. There was also an address on FA0 and D0 of 10.1.1.5. We needed to add a second address on the AP, but couldn't. (Option was there but wouldn't take.) So, we changed the address on the BVI to what we needed while connected to the 10.1.1.5 address.
Now at this point I have:
BVI - 10.100.100.5
FA0 - 10.1.1.5
I can still move around on the FA0 telnet session, and I can telnet to the new address. Can someone explain how this was possible?
Anytime a client sends an ARP request, it goes with the first response it receives. Since all interfaces are configured with the same address, they all responded to the ARP request sent and the FA0 interface responded first. So when you Telnetted into the AP, you did so via the FA0 interface. You changed the address on BVI1, which did not affect the existing Telnet session.
Is that what you're looking for? Just so you know, best practice is to only have an address on the BVI, there's no need for one on the radio or fastethernet interfaces. Of course, if you need a second address and this is working, then feel free to leave it.
Well, I've never put an address on all three, but the company that I just started working for have them like this. I'm creating VLANs for all of the APs, and I'm not putting addresses on the subinterfaces, but we had a need to change an existing address the other day.
It was just very odd because I thought that the BVI was the "all interface" address, and if that was changed then you'd lose your connection because it was binding all of the interfaces to the BVI.
I actually used to think that as well, and for APs it might as well be. But when you start configuring 800-series routers with access point modules, you get into the nitty gritty of what BVIs are bridge-groups are. It's interesting, but extremely complicated.