[ovs-discuss] Flow to forward LLDP/STP packets

Padmanabhan Krishnan kprad1 at yahoo.com
Wed May 29 14:29:00 UTC 2013


Thanks for the reply. Sure, i can submit a patch sometime next week.

I am learning the code, so didn't know which one has more precedence. Looking at:

http://openvswitch.org/pipermail/git/2011-August/001692.html

I was of the opinion that flows cannot be installed for special Mcast packets (like LLDP, STP), where in i can give the action as flood or forward to all ports. Are you saying that's not the case?

And, i am aware of the 5sec ageout in kernel datapath, so i was sending packets every sec and monitoring the kernel flow. 

Thanks,
KP



________________________________
 From: Jesse Gross <jesse at nicira.com>
To: Padmanabhan Krishnan <kprad1 at yahoo.com> 
Cc: "discuss at openvswitch.org" <discuss at openvswitch.org> 
Sent: Friday, May 24, 2013 3:47 PM
Subject: Re: [ovs-discuss] Flow to forward LLDP/STP packets
 

On Thu, May 23, 2013 at 11:11 PM, Padmanabhan Krishnan <kprad1 at yahoo.com> wrote:
> Hello,
>
> I am looking for an option wherein i can get the OVS Bridge to forward  LLDP
> packets.  There are sent with Nearest Bridge (0180 c200 000e) or Nearest
> Customer Bridge Multicast (0180 c200 0000) address. I am aware that native
> Linux Bridge consumes such packets. Is it the same default behavior for OVS?

If you are explicitly writing flows instead of using the NORMAL action
then there is no special behavior for dropping these packets.

> I am running this in Ubuntu 12.04 and OVS is  1.10.90.
>
> I have a OVS bridge names OVS-test and it has a physical interface and a tap
> interface as its ports.
> I tried the following, and both of these didn't work:
>
> a)
> I installed the following flow:
>
> sudo ovs-ofctl add-flow OVS-test
> dl_dst=01:80:c2:00:00:0e/ff:00:00:00:00:00,actions=all
>
> This doesn't seem to work. I mean Mcast packets sent/received on physical or
> tap port doesn't get forwarded to the other port via the OVS bridge.
>
> The dp-flow seem to be empty:
>
> $ sudo ovs-dpctl dump-flows
> $

If traffic isn't being seen by the kernel then the userspace
configuration won't make a difference. However, the flows age out of
the kernel after 5 seconds (assuming the system is otherwise idle) so
perhaps there was just too long of a delay before you ran this
command.

> b)
> I came across this function "ofproto_set_forward_bpdu" and also the below
> earlier post:
>

> http://openvswitch.org/pipermail/git/2011-August/001692.html
>
> Setting this using "ovs-vsctl set bridge OVS-test
> other-config:forward-bpdu=true" also didn't solve my problem.
>
> BTW, does the MAC address in function "eth_addr_is_reserved" look right?
> Shouldn't they be 0180…. instead of 0108?

Yes, it appears that you are correct. Care to submit a patch?
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://openvswitch.org/pipermail/ovs-discuss/attachments/20130529/84ebd2f7/attachment.html>


More information about the discuss mailing list