Hello and many thanks.
Below my answer to your last question.
Regards
Corinne
-----Message d'origine-----
De : Ben Pfaff [mailto:[email protected]] 
Envoyé : jeudi 16 juin 2011 19:08
À : SAINT JALME Corinne RD-CORE-LAN
Cc : [email protected]
Objet : Re: [ovs-discuss] Why does OpenFlow OpenVSwitch not learn L2 MAC 
adresses?

On Thu, Jun 16, 2011 at 04:24:19PM +0200, [email protected] 
wrote:
> - activation of the ovs-controller (for L2 MAC learning) :  ovs-controller 
> ptcp:6633:127.0.0.1

Don't use ovs-controller.  It's just for experimenting with OpenFlow
controllers.  It's not actually a useful controller for production.

> - activation of the ovs-openflow :  ovs-openflowd dp0 tcp:127.0.0.1:6633 
> -listen=ptcp:6634:127.0.0.1 -detach

Use ovs-vswitchd instead.  It has a MAC learning switch built-in, with
no need for a controller.

> - configuration of the flows/instruction (use of ovs-ofctl) :      
> ovs-ofctl add-flow tcp:127.0.0.1.6634 in_port=1,actions=output:2
>  
> ovs-ofctl add-flow tcp:127.0.0.1.6634 in_port=2,actions=output:1

If you want MAC learning, why would you set up flows that do not match
on MAC address?

CSJ : in fact we assume that we need MAC learning, because once the generic 
flows above are configured our ovs-switch (all frames which arrive on eth1 are 
sent to eth2 and reversely), we observe that only frames with destination MAC 
address = FF cross the ovs-switch and all the frames with a destination MAC 
address != FF are dropped somewhere before the virtual interface of the VM 
which implements the switch.

Assuming it should work without MAC learning then we can suspect a wrong 
behaviour due to Virtual Box internal bridge.

_______________________________________________
discuss mailing list
[email protected]
http://openvswitch.org/mailman/listinfo/discuss

Reply via email to