[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-atm
Subject:    Re: Question on bridging 802.1Q (VLAN) Ethernet <--> ATM
From:       Ben Greear <greearb () agcs ! com>
Date:       1999-01-21 21:43:39
[Download RAW message or body]

Heikki Vatiainen wrote:
> 
> Ben Greear wrote:
> 
> > I am working on briding 802.1Q VLAN ethernet packets to PVCs (or SVCs).
> >
> > I've been reading some books and the latest ATM release (0.52) code,
> > and it looks like there is already a bunch of work done for MPOA,
> > which should be useful for what I want to do...
> 
> If you want to do it at layer 2 (bridge packets), then MPOA may
> not be what you want. MPOA works on layer 3 and does shortcut
> routing.

Ok, don't wanna go to layer 3.

> 
> > I want to do all this at layer-2.  I have the VLAN Ethernet part
> > coded up, but I have yet to figure out how to connect it to the
> > PVC's.  In other words, if someone (ATM system) can hand me an
> > Ethernet frame, I can tag it with a VLAN ID (VID) and toss it onto
> > the ethernet device.  When I get a VLAN frame FROM the device, I know
> > how to turn
> > it into a regular Ethernet packet:  I just need to find the correct
> > part of the ATM system to hand it to.
> 
> When the next release, ATM-0.53, comes out, it is possible to
> use LANE to bridge packets between Ethernet and ATM. There
> should be no difference between the interface provided by an
> Ethernet driver and the interface provided by the LANE kernel
> part.
> 
> So if you 802.1Q code works with normal Ethernet drivers, it
> will also work with lane.

This sounds like what I want.  If possible, I'd like to do this entirely
in the kernel.  Could you point me at the file or files where I might
find the methods to interface into the LANE?  Is it in kernel
or user space?  Also, when is .53 coming out?

> 
> Above you talk about bridging, but have thought about providing
> multiple logical interfaces using only one Ethernet card.
> For example, if the Ethernet card receives 802.1Q tagged packets
> for 5 different VLANs, will the host have interface in each VLAN?
> Interfaces vlan[0-4] maybe?

Yes, that is how I implement it.  Each VLAN 'device' can have different layer
three's too, so each one can have an IP/MASK.  I'm installing my
second linux box now to actually test end-to-end VLANs.

> 
> Also, since an ATM card can have multiple Classical-IP and LANE
> interfaces configured on it, isn't there a possibility to have
> e.g. 5 ELANS which all contain/carry traffic for multiple 802.1Q
> VLANs. Add a couple of Classical-IP interfaces and you will have
> lets say -- 30 different interfaces which all use the same physical
> interface.

I don't know much about LANE or ELANs.  The first thing I want to
get working is a simple bridging.  I could actually care less about
layer 3 at this point.  However, for the general community and uses,
layer 3 will have to be addressed.

I basically just want to be able to map a VLAN device to a 
PVC (Maybe what I really mean is LANE instead of PVC??) and
have traffic pass back and forth fast and efficiently.  The other
thing that is slightly wierd:  I'm talking about 1000+ VLAN devices,
and a corresponding number of PVCs.  I'm going to eat kernel memory
liberally to get maximum speed, but will try to code with a less
drastic number of VLAN's in mind for the common user!



> 
> Too much coffee, I know :)
> 
> > Can someone give me a hint as to the best way to do this?
> 
> I hope this helps.
> 
> > Thanks,
> > Ben
> 
> // Heikki
> --
> Heikki Vatiainen                  * hessu@cs.tut.fi
> Tampere University of Technology  * Tampere, Finland

-- 
Ben Greear	  greearb@agcs.com   Pager: 202-2717
(602) 581 4980    "More weight!" -- _The Crucible._
http://hydrogen:8080/home/greearb/public_html/index.html

[prev in list] [next in list] [prev in thread] [next in thread]