On Wed, Dec 19, 2012 at 10:11 PM, Ethan Jackson <et...@nicira.com> wrote:
> Letting netdev-vport manage ethernet addresses itself instead of
> relying on the datapath has several advantages. It simplifies the
> code, is significantly more efficient, and will work when there is
> no longer a one to one mapping from netdev-vports to datapath
> vports.
>
> Signed-off-by: Ethan Jackson <et...@nicira.com>
The one remaining user of being able to set MAC addresses is path MTU
discovery. Currently, it will randomly generate an address and
actually send ICMP messages from that address so depending on what
filters are configured it can be important to be able to control that
address.
We've already marked PMTUD as being deprecated and are planning on
removing it on master. At the very least, before we apply this patch
we should remove the ability to turn it on from userspace (it's
already off by default). Ideally, we would also fully remove both it
and the ability to set MAC addresses through the vport layer since
both will be completely dead code. However, that part doesn't
necessarily have to happen before this patch.
> diff --git a/lib/netdev-vport.c b/lib/netdev-vport.c
> index e50103b..a527f91 100644
> --- a/lib/netdev-vport.c
> +++ b/lib/netdev-vport.c
> @@ -56,6 +56,7 @@ struct netdev_dev_vport {
> struct netdev_dev netdev_dev;
> struct ofpbuf *options;
> unsigned int change_seq;
> + uint8_t ea[ETH_ADDR_LEN];
Can you use a name that is slightly more descriptive than 'ea'? I
think there isn't necessarily enough context for the meaning to be
obvious.
_______________________________________________
dev mailing list
dev@openvswitch.org
http://openvswitch.org/mailman/listinfo/dev
