[ovs-discuss] FF group liveness

Ben Pfaff blp at ovn.org
Tue Apr 12 16:10:45 UTC 2016


The specification isn't very precise, in most cases, about particular
error codes.  It would probably be better if it were more precise.  I'd
support a proposal to the Open Datapath working group to tighten this
up.

On Tue, Apr 12, 2016 at 10:13:44AM +0200, Miklós Pelyva wrote:
> Hi Ben,
> 
> thanks for the reply! Indeed, this error code seems reasonable.
> Then, it is more of an administrative question.
> Can this be written into the protocol, as well?
> 
> Best regards,
> Miklos
> 
> On 04/10/2016 08:44 PM, Ben Pfaff wrote:
> >On Fri, Apr 08, 2016 at 10:24:31AM +0200, Miklós Pelyva wrote:
> >>Hi!
> >>
> >>I have a question regarding some difference between OpenFlow protocol
> >>version 1.3.5 and vanilla Open vSwitch implementation version 2.5.0.
> >>
> >>If someone would like to use fast failover type group without bucket
> >>liveness set, then it would return an OFPERR_OFPGMFC_INVALID_GROUP error
> >>code.
> >>However, based on the protocol (1.3.5) invalid group error should be
> >>returned only in the following two cases:
> >>1) "If a specified group type is invalid (ie: includes fields such as weight
> >>that are unde ned for the specified group type) then the switch must refuse
> >>to add the group entry and must send an ofp_error_msg with
> >>OFPET_GROUP_MOD_FAILED type and OFPGMFC_INVALID_GROUP code."
> >>2) "If a delete request includes some buckets in the Group-Mod message, then
> >>the switch must return an ofp_error_msg with OFPET_GROUP_MOD_FAILED type and
> >>OFPGMFC_INVALID_GROUP code."
> >>
> >>The particular piece of code from ofp-util.c in ofputil_decode_group_mod,
> >>l:8878 is the following:
> >>         case OFPGT11_FF:
> >>             if (!ofputil_bucket_has_liveness(bucket)) {
> >>                 return OFPERR_OFPGMFC_INVALID_GROUP;
> >>             }
> >>             break;
> >>
> >>Is the above error code handling not written into the protocol by purpose or
> >>by accident, or is it just not explicitly determined by the protocol how to
> >>handle fast failover type groups without bucket liveness?
> >"Invalid group" seems like a reasonable error code to me for this
> >problem.  What error code do you prefer?
> 
> 



More information about the discuss mailing list