[ovs-dev] [PATCH] Add VLAN splinters unit test.

Alex Wang alexw at nicira.com
Tue Jun 25 20:33:33 UTC 2013


Thanks Ben,

I noticed that my third comment was wrong by experiment. The following is
just a resend of my previous email with the code review:

"

> 3. The packets with vlan header are missed in kernel and examined in
> "ofproto/ofproto-dpif.c". The "vsp_vlandev_to_realdev()" and
> "vsp_realdev_to_vlandev()" are used to convert between "vlandev port" and
> "realdev port". And the performance cost is in that there is no datapath
> flow installed.
>
>
I think I was wrong here. I experimented using scapy by sending
"Ether(src="00:00:00:00:00:02",
dst="00:00:00:00:00:01")/Dot1Q(vlan=11)/IP(src="10.24.0.2",
dst="10.24.0.1")/"payload"" and I saw the datapath flow installed with
correct datapath port number.

But I'm still not very sure what is the performance cost? Is that in that
the using device driver can be faster than matching datapath flow?

Besides sorry for this delayed review, the code makes sense and looks good
to me, there is only one redundant line below

On Tue, Jun 11, 2013 at 1:25 PM, Ben Pfaff <blp at nicira.com> wrote:

+ovs-appctl dpif/show
> +ovs-appctl dpif/show | sed -n '
> +s/\./_/g
> +s/^[[  ]]*\([[^        ]][[^   ]]*\) [[0-9]]*\/\([[0-9]]*\).*/\1=\2/p
> +' > port-numbers
> +cat port-numbers
> +. ./port-numbers
>

The "+ovs-appctl dpif/show" is redundant.
"
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.openvswitch.org/pipermail/ovs-dev/attachments/20130625/3b5a41c6/attachment-0003.html>


More information about the dev mailing list