Matti Aarnio wrote: > > > The code is technically correct in that it seems to cause no > > > harm to existing users, and it actively provides the 802.1Q VLAN > > > feature that some people like. > > It is *mostly* ok. Something odd is going in in the multicast, > but that I haven't been able to figure out. It might be that > the API for de-registering multicast MAC sets is not good enough > for this type of use in the network devices - or that Ben's code > for handling them is shoddy. > > The odd things appear only with multicast, and there only in form > of not deregistering subscribed MC MACs correctly per interface. > The physical card stays subscribed for MACs which none of the > consumers (VLANs) are interested in. I have looked at this, and can find no obvious problem with my code (which I basically coppied from Gleb & Lennert's VLAN implementation). I also don't use multi-cast on my network, so I have a hard time figuring out how to debug this. The API for multi-cast is very difficult to work with w/regard to VLANs, but it should be possible to get it right... > Your original code did touch deeply into the packet receiver main > processing, and I haven't checked if you merged my changes which > move that processing into VLAN-protocol receiver, instead of special > treating the incoming packets with VLAN tags. Yes, the 2.4 code is much less invasive now. (I did merge most of your changes.) The 2.2 patch remains invasive, but it's probably not worth changing it because the 2.2 kernel is serious feature-freeze anyway (and the code does work, even if it's ugly.) Ben -- Ben Greear <greearb@candelatech.com> <Ben_Greear@excite.com> President of Candela Technologies Inc http://www.candelatech.com ScryMUD: http://scry.wanfear.com http://scry.wanfear.com/~greear - : send the line "unsubscribe linux-net" in the body of a message to majordomo@vger.kernel.org
