On Sat, May 11, 2002 at 08:14:39AM +0200, David GLAUDE Mailing wrote:

> > > I will be using the bridge on Bluetooth accesspoints running
> > > the BNEP protocoll. Because a user could move and change
> > > accesspoint a lot of times I would like to remove the users
> > > mac from all bridges in the LAN when they change accesspoint
> > > so the bridges don't continue to send the data on wrong ports.
> > > Is it possible to do this?
> >
> > No, there is no current possibility to remove a MAC address from
> > a port other than to wait for a timeout.
> 
> Yes it is possible, it is called Topological Change.

You still have to wait for a timeout, the timeout is just
shorter.. my statement still holds :)


> > But even if there were.. if you remove a MAC address from a port,
> > it will cause all future packets to that MAC address to be sent
> > to _all_ ports (because the bridge doesn't know where to send
> > them).  Is that what you want?
> 
> Well, that's the only way to rediscover the new location of the MAC.

Sure, but do you want all traffic to be broadcasted in this
grace period?


> If you use bi-directionnal protocol, this is not a lot of traffic.

So you think that the 802.1d 'LEARNING' state is redundant?


> > A way to 'blackhole' a MAC address would be to connect a fake
> > ethernet interface to every bridge, say an ethertap interface,
> > and send a packet from that interface with the address you
> > want to blackhole as the source address.  All future packets
> > to that MAC address will then be sent to that fake device,
> > effectively being blackholed.  You would need some kind of
> > signalling protocol for this, though.
> 
> Blackholing is more likely NOT a good solution. That MAC
> will loose connectivity wherever it is. I guess you want the
> device to recover connectivity from another access point.

Which it can, by sending a packet.  Which it will probably
do soon enough anyway.


> So I suggest the following (assuming linux bridging strictly follow the
> standard
> and you have a way to detect that a BNEP device being out of reach).
> Create a ethertap interface, that you control by your own deamon.
> When the deamon detect a device being out of range, you have to
> disable/enable
> the ethertap interface to trigger the topological change.

This is an even bigger hack than what I proposed..  really,
if one's going to do such things, one might just as well
implement MAC address addition/removal properly.


> Another option would be to generate (on demand) your own
> topological change frame. Notice that once you have that
> software, you have a pretty good DoS tool for switched network.

STP is pretty much flawed security-wise anyway.. :)


cheers,
Lennert
_______________________________________________
Bridge mailing list
[EMAIL PROTECTED]
http://www.math.leidenuniv.nl/mailman/listinfo/bridge

Reply via email to