On Wed, Jul 11, 2012 at 02:10:42PM -0700, Ben Pfaff wrote: > On Mon, Jul 09, 2012 at 06:03:55PM +0900, Simon Horman wrote: > > Signed-off-by: Simon Horman <ho...@verge.net.au> > > I'm not sure about MFM_IPV6_LABEL, because it means that an exact-match > mask would be rejected, even though an exact-match mask is supposed to > be equivalent to a match without a mask. I think that we need to deal > with that somehow. The simplest way would be to just accept all masks > for MFM_IPV6_LABEL; I'd be OK with that.
Ok, I don't think that I have a problem with that. I'll see about update my patches accordingly. One thing that I do have a problem with is the correct treatment of the CFI bit in the value and mask when decoded from an OXMi VLAN_VID match. So far I have come up with four possible solutions. 1. Require it to always be set except in the case of OFPVID12_NONE 2. Require it to always be clear except in the case of OFPVID12_PRESENT - I believe that this is how things need to be handled internally 3. Ignore the CFI bit except in the case of OFPVID12_NONE and OFPVID12_PRESENT. 4. Honour the CFI bit. > I don't see other problems at first glance but please increment > FLOW_WC_SEQ to draw attention to all the places that sometimes need to > change when wildcard behavior changes. Sure, will do. FWIW, I did subsequently notice FLOW_WC_SEQ and check all the locations - I think that I have covered all the bases. _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
