On Sat, Oct 18, 2014 at 09:55:57AM -0400, Thomas F Herbert wrote: > On 10/17/14, 4:54 PM, Ben Pfaff wrote: > >On Fri, Oct 17, 2014 at 01:19:33PM -0400, Dave Benson wrote: > >>On Fri, Oct 10, 2014 at 6:43 PM, Ben Pfaff<b...@nicira.com> wrote: > >>... > > >My first thought is to define a new NXM_NX_VLAN_TPID that matches the > >TPID, and for backward compatibility to imply a match on 0x8100 if it > >is not given. > I think the assumption in OF is that the vlan-vid match is always on the > outermost tag but the description in the spec is not explicit.
Yes, that's the assumption. If it's not well described in the spec, then maybe we could work to improve that. > The 802.1ad spec requires that the outer TPID must be 0x88a8 so in > spite of some legacy use of 0x9100 or 0x8100 for the outer tpid, I > think it is OK to assume that 0x88a8 is the TPID for the outer tag and > 0x8100 is the TPID for the inner tag. > > However, an ambiguity in the OF spec is that the behavior for 802.1q and > 802.1ad is implied and not explicitly stated.For example, the entry in the > OF in table 11 (page 57, version 1.4,) the line about vlans should say that > the prerequisite of a VLAN_ID should be a TPID of 0x8100 OR 0x88a8 and the > description should be changed to include both 802.1q and 802.1ad and that > the match is on on the outer tag. I don't understand your proposal at all. It must be possible to distinguish a match on an 0x88a8 tag from one on an 0x8100 tag. How do you propose to do that? _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
