Re: CFT: new trunk(4) [results]

Ian FREISLICH wrote:
> Andrew Thompson wrote:
> > Here is a patch to add OpenBSD's trunk(4) interface, and also includes
> > LACP support which came from agr(4) on NetBSD.  Im interested in anyone
> > who wants to test this and in particular lacp mode if you have a switch
> > that supports it. 
> > 
> > http://people.freebsd.org/~thompsa/if_trunk-20070330b.diff
> 
> This looks very interesting.  I'm busy testing with a switch that
> claims 802.3ad support.

The trunk(4) device establishes LACP with the (nasty) smitch that
I have using my testbed.  It also appears to utilise both links I
put in the trunk.

The setup is as follows:

rl0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
        options=8<VLAN_MTU>
        ether 00:50:bf:09:67:8b
        media: Ethernet autoselect (100baseTX <full-duplex>)
        status: active
        trunk: trunkdev trunk0
rl1: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
        options=8<VLAN_MTU>
        ether 00:50:bf:09:67:8b
        media: Ethernet autoselect (100baseTX <full-duplex>)
        status: active
        trunk: trunkdev trunk0
trunk0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
        ether 00:50:bf:09:67:8b
        media: Ethernet autoselect
        status: active
        trunk: trunkproto lacp
                trunkport rl1 =1c<ACTIVE,COLLECTING,DISTRIBUTING>
                trunkport rl0 =1c<ACTIVE,COLLECTING,DISTRIBUTING>
vlan300: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1496
        ether 00:50:bf:09:67:8b
        inet 10.0.0.210 netmask 0xffffff00 broadcast 10.0.0.255
        media: Ethernet autoselect
        status: active
        vlan: 300 parent interface: trunk0

All is well even when one of the links goes down.  The moment I
plug the wire back in, the link may or may not continue working
depending on which link the switch decides to send the traffic down.
If I disconnect the other link, it starts working again until I
plug it back in.  This is not the way things behaved when I tested
this switch with a Cisco 6509 at our provider's premises yesterday.

When one wire has been unplugged and then plugged back again the
trunk is left in the following state:

trunk0: flags=8843<UP,BROADCAST,RUNNING,SIMPLEX,MULTICAST> metric 0 mtu 1500
        ether 00:50:bf:09:67:8b
        media: Ethernet autoselect
        status: active
        trunk: trunkproto lacp
                trunkport rl1 =18<COLLECTING,DISTRIBUTING>
                trunkport rl0 =1c<ACTIVE,COLLECTING,DISTRIBUTING>

I'm not sure if this is a feature of the NICs I'm using (I know
they're junk) or a bug in the switch or a bug in the trunk driver.
I don't think it's the switch because this test worked when it was
connected to the Cisco 6509.  I don't have access to the cisco any
more.

If I remove and then re-add the "bad" trunk port, it starts working
again.

Ian

--
Ian Freislich