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

[Ben Pfaff]

On Mon, Jun 24, 2013 at 10:16:15PM -0700, Alex Wang wrote:
> On Mon, Jun 24, 2013 at 6:13 PM, Alex Wang <alexw at nicira.com> wrote:
> 
> > Thanks for the answers,
> >
> > On Mon, Jun 24, 2013 at 5:03 PM, Ben Pfaff <blp at nicira.com> wrote:
> >
> >> On Mon, Jun 24, 2013 at 03:56:31PM -0700, Alex Wang wrote:
> >> > 1. What is usually the bug in the buggy driver? How can using
> >> vlan-splinter
> >> > avoid that? I'm a bit confused here, since the "vlandev.c" code still
> >> talks
> >> > to linux device driver (e.g. when creating ADD_VLAN_CMD) via the "ioctl"
> >> > call.
> >>
> >> ovs-vlan-bug-workaround(8) has a lot of information:
> >>
> >> ......
> >
> >
> >
> > This is very helpful. Also, Do you which file I should check to see how
> > packet (with vlan header) arrived on a trunk port is mapped to
> > corresponding vlandev port?
> >
> 
> 
> Hey Ben,
> 
> I checked more code and seem to understand more about my question.
> 
> Here is my understanding about how vlan-splinter works,
> 
> 1. The "set interface p1 other-config:enable-vlan-splinters=true" enables
> the vlan splinter of the port in ovsdb.

Yes.

> 2. In "vswitch/bridge.c", the vlandev is added and ports are created for
> each vlandev. Then, it calls functions in "ofproto/ofproto.c" and
> "ofproto/ofproto-dpif.c" to set vlandev port as slave to the realdev port.

Yes.

> 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.

It should be possible to set up datapath flows for VLAN splinters.