On Fri, Jan 11, 2002 at 05:03:53AM +1030, Paul wrote:
> A question for the developers. Once the eth0,eth1 interfaces have been > bound to br0, is any queuing discipline attached to them still used? Yes. > I have an application of QoS where I would like to bridge between two > 100BaseTX interfaces, one of which ends up going to an ADSL link > (eth0). > > eth0 eth1 > --[ ADSL Router] ------[ Bridge ]------[ Local Subnet ] > br0 > > The options are: > > - Adding a single 'cbq' to br0 which distinquishes between traffic > going out on eth0 and eth1 (via through destination > address). Traffic going directly to the router would have a local > subnet address, and so it would be classified as 'eth1' traffic > and receive appropriate treatment. (This may not be such a bad thing.) br0 is not a 'real' interface (with appropriate packet queue and such), so a bunch of shaping things like idle time estimation will not work. Simple tbf might work, but cbq will probably mess up. In fact, I'm not so sure whether tbf will work, as queueing disciplines don't seem to be used on devices without queue (such as bridge devices), by staring at the code. > - (If it works?) Adding QoS queueing disciplines to 'eth0' and > 'eth1' directly. This would only work if these get called when the > packet gets sent out on the appropriate interface. This sounds better. This'll most definitely work. And it just makes more sense to do QoS on the physical devices. cheers, Lennert _______________________________________________ Bridge mailing list [EMAIL PROTECTED] http://www.math.leidenuniv.nl/mailman/listinfo/bridge
