wireless-to-wired bridging

[Reid Linnemann]

Not so much a Microsoft thing as a general networking thing. I would like
for netbios traffic to work correctly for windows file sharing/samba, as
well as broadcast LAN traffic for gaming and the like. I _could_ alter
bridge.c to always return a copy of the packet to the caller, but that
would just be a quick hack and I don't even know if it would work.

Dummynet works on the IP level, so it wouldn't solve my problem. Else
I'd jump all over it. =(

On 2/13/2005, "Lowell Gilbert"
<freebsd-questions-local at be-well.ilk.org> wrote:

>"Reid Linnemann" <lreid at cs.okstate.edu> writes:
>
>> I'm bridging the devices so that the wired and wireless nets will appear
>> to be on the same physical network to eachother.
>
>Well, yes, that's what bridging means.  Why do you want that?  [Is it
>a Microsoft thing?]
>
>> I think I was really tired when I wrote my original email.. so let me
>> rewrite my hypothesis:
>>
>> I am suspicious that, since the wireless interface on the BSD machine
>> operates in AP mode, if a wireless client wants to send a packet to
>> another wireless client, it must be first sent to the wireless interface
>> of the BSD machine, which should theoretically redirect the packet to
>> the appropriate host on the wireless net. In the wired network, a switch
>> handles this case automagically on the datalink layer before any
>> messages can hit the rl1 interface of the BSD router. I've looked at
>> the bridge code, and it seems that unless a packet is multicast or
>> broadcast it will be copied to the other bridged interfaces but not
>> returned to the original caller. Since the packets being sent from one
>> wireless client to another are not broadcast, I think that the bridge
>> module may be dumping them into the black hole of the wired LAN, and
>> they are not being processed and pumped back out through the ath
>> interface. Is this a correct assumption? Are there ways I can overcome
>> this problem?
>
>On a quick look, I think you might be on the right track.  The
>bridging code seems in a number of spots to be built specifically for
>Ethernet.  I have always maintained that bridging unlike media was a
>hack bound for problems...
>
>You might have more success using dummynet for bridging rather than
>trying to fix things in the protocol stack.
>
>Good luck.
>--
>Lowell Gilbert, embedded/networking software engineer, Boston area
>		http://be-well.ilk.org/~lowell/