[ovs-discuss] Rejecting flow-mod from a slave controller

[Ben Pfaff]

The updated definition of OFPERR_OFPBRC_IS_SLAVE will use the type and
code for OFPBRC_EPERM in OpenFlow 1.0 and 1.1, and the type and code
for OFPBRC_IS_SLAVE in OpenFlow 1.2+ without any special logic
elsewhere in the code.  In other words, it does the same thing as what
you suggested elsewhere, without having to write all that special-case
code.

On Wed, Jan 21, 2015 at 01:08:40PM -0500, Anup Khadka wrote:
> I am actually confused by the fix. OFPBRC_IS_SLAVE is an OpenFlow 1.2+
> error message, correct? The comment in the header file is changed and it
> appears that this is an OpenFlow 1.0+ error message and I am not sure if
> that's right. Likewise OFPBRC_IS_SLAVE is returned while rejecting slave
> controllers regardless of the version of the switch. If the switch is not
> running in 1.2+ mode, the code that sends the error is not going to
> understand this error message, right? Which in that case, it will probably
> send a Nicira vendor error back to the controller, instead of the
> EPERM_ERROR as the switch used to send previously. Is this what you are
> aiming to do? Am I missing something?
> 
> Thank you,
> Anup
> 
> On Wed, Jan 21, 2015 at 12:46 PM, Ben Pfaff <blp at nicira.com> wrote:
> 
> > I sent out a fix:
> >         http://openvswitch.org/pipermail/dev/2015-January/050612.html
> >
> > On Tue, Jan 20, 2015 at 03:01:27PM -0800, Ben Pfaff wrote:
> > > The OF1.2 feature was based on an Open vSwitch extension to OpenFlow
> > > first implemented for OF1.0.  We need to continue supporting that
> > > extension.
> > >
> > > On Tue, Jan 20, 2015 at 05:12:30PM -0500, Anup Khadka wrote:
> > > > Do OpenFlow versions prior to OF 1.2 even support the concept of slave
> > > > controllers? I couldn't find anything in the specs.
> > > >
> > > > On Tue, Jan 20, 2015 at 4:24 PM, Ben Pfaff <blp at nicira.com> wrote:
> > > >
> > > > > On Tue, Jan 20, 2015 at 03:55:57PM -0500, Anup Khadka wrote:
> > > > > > I have noticed that OVS rejects flow-mod request from a
> > slave-controller
> > > > > > with an OFPBRC_EPERM request code. Isn't the more appropriate code
> > here
> > > > > is:
> > > > > > OFPBRC_IS_SLAVE?
> > > > > >
> > > > > > Same thing when a slave controller sends a packet out.
> > > > >
> > > > > That does sound better, for OF1.2+.
> > > > >
> >