Date: Tue, 11 Jul 2006 23:51:49 +0100 (BST) From: Robert Watson <rwatson@FreeBSD.org> To: Brooks Davis <brooks@one-eyed-alien.net> Cc: Andre Oppermann <andre@freebsd.org>, Andrew Thompson <thompsa@freebsd.org>, freebsd-arch@freebsd.org Subject: Re: vlans and cloning Message-ID: <20060711234931.P14749@fledge.watson.org> In-Reply-To: <20060711162953.GC20418@odin.ac.hmc.edu> References: <44B15511.206@errno.com> <44B21551.5050002@freebsd.org> <20060710160441.GB31026@odin.ac.hmc.edu> <20060710202714.GC16054@heff.fud.org.nz> <20060711162953.GC20418@odin.ac.hmc.edu>
next in thread | previous in thread | raw e-mail | index | archive | help
On Tue, 11 Jul 2006, Brooks Davis wrote: >>> Unless cause ifconfig to autoload all if_ modules when cloning fails, it's >>> impossiable to support this without having if_vlan loaded. That said the >>> current plan it to eliminate if_vlan and integrate vlan support directly >>> into if_ethersubr.c to allow use to correctly handle the default vlan case >>> among otherthings. >> >> Is anyone working on this? The bridge code needs access to the vlan tag to >> properly hash the address as each vlan is a seperate broadcast domain, this >> would be much easier with vlan merged to if_ethersubr.c. > > I think it ended up with Robert's name on it at the last devsummit, but he's > got a lot of higher priority stuff on his plate. It doesn't look like this > change would be all that much work. The one thing that might be worth > investigating is seeing if there's a sane way to make vlan tag parsing part > of ether_input, but keep if_vlan.c around as a module for actual support of > trunks so we get most of the architectural benefits of correctly treating > vlan tags as part of the spec, but let embedded users who don't need trunks > avoid the overhead. Yes -- the specific proposal I have made is that we combine if_vlan.c into if_ethersubr.c, as well as LLC encapsulation decapsulation. Vlans and LLC bits are all considered standard ethernet features today, and integrating the basic support (regardless of virtual interfaces) into if_ethersubr.c makes sense. It would also make the book keeping and event handling go a bit more naturally, I think. I've not had a chance to prototype this to explore what the above words actually mean, so I think some experimental prototyping is called for. It's on my todo list but something I'm likely not to get to for a few months, so if someone else wants to look into this, I think that would be great. Robert N M Watson Computer Laboratory University of Cambridge
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20060711234931.P14749>