[ovs-dev] [PATCH] mirroring: Fix check to avoid outputing to input port.

Ben Pfaff blp at nicira.com
Thu Oct 8 16:35:58 UTC 2009


Jesse Gross <jesse at nicira.com> writes:

> When RSPANing with a bond, ensure that we don't receive a packet on
> one interface and then output it on the other interface of the bond.
> Also avoid sending the packet out the same interface it was received
> on when an interfaced is configured with the VLAN that we are using
> for RSPAN.

I had to stare at this for a very long time before I saw how it
would affect bond behavior.  Would you mind breaking it into two
patches, one for the bond fix and one for the vlan fix, then, to
make it clear which is which?

The bond fix makes perfect sense to me.  Good catch.

I don't understand the vlan fix.  It appears to be making sure
that 0 and OFP_VLAN_NONE are treated the same way.  But I think
that we have this property already, because of these invariants
on the values in the existing test "dst->vlan == vlan":

        * dst->vlan is in 0...4095 (although perhaps we should
          only allow mirroring to nonzero VLANs).

        * vlan is in 0...4095 (see the assignment and test under
          the comment "Figure out what VLAN this packet belongs
          to" in process_flow()).




More information about the dev mailing list