[ovs-dev] [nic-401] ofproto-dpif: Do not mirror L2 multicast switch protocols to VLANs.

[Ben Pfaff]

On Fri, Jul 15, 2011 at 09:59:35AM -0700, Justin Pettit wrote:
> On Jul 11, 2011, at 3:44 PM, Ben Pfaff wrote:
> 
> > +        if (dst[0] == 0x01 && dst[1] == 0x00 && dst[2] == 0x0c) {
> > +            /* Cisco OUI. */
> > +            if ((dst[3] & 0xfe) == 0xcc &&
> > +                (dst[4] & 0xfe) == 0xcc &&
> > +                (dst[5] & 0xfe) == 0xcc) {
> > +                /* Drop the following protocols plus others following the same
> > +                   pattern:
> > +
> > +                   CDP, VTP, DTP, PAgP  (01-00-0c-cc-cc-cc)
> > +                   Spanning Tree PVSTP+ (01-00-0c-cc-cc-cd)
> > +                   STP Uplink Fast      (01-00-0c-cd-cd-cd) */
> > +                return false;
> > +            }
> 
> Do we know that the other cc/cd combinations are safe to not
> forward?  I'm guessing yes, but we may want to confirm with a Cisco
> expat.

I was using the pattern as a clue, reasoning that it's safe to *not*
mirror anything.  You can't screw up a switch that way, as far as I
know.  The worst you get is a bug report that "Open vSwitch isn't
mirroring my Cisco Magic FUBAR packets, why not?"  Right?

> > +          <dt><code>01:00:c2:00:00:00</code></dt>
> > +          <dd>IEEE 802.1D Spanning Tree Protocol (STP).</dd>
> > +
> > +          <dt><code>01:00:c2:00:00:01</code></dt>
> > +          <dd>IEEE Pause frame.</dd>
> > +
> > +          <dt><code>01:00:c2:00:00:0<var>x</var></code></dt>
> > +          <dd>Other reserved protocols.</dd>
> 
> I believe the second octet on these previous three addresses is
> supposed to be 80.

You're right, oops.  Fixed.  I did get it right in the code.