Implement acceptable interface identifier range for P4RT

[earies]

• (M) release/models/p4rt/openconfig-p4rt.yang
  • Add range statement to interface identifier in conformance to P4RT
    specification v1.3.0
  • Description cleanup to map OpenConfig terminology to P4RT
    specification terminology

Per https://p4.org/p4-spec/p4runtime/main/P4Runtime-Spec.html#sec-translation-of-port-numbers, the 'SDN Port' is the equivalent attribute to the OpenConfig P4RT interface 'id'. This change limits the acceptable uint32 range for configurable values per specification.

[OpenConfigBot]

Compatibility Report for commit 138b239:
⛔ yanglint@SO 1.10.17

[robshakir]

This LGTM. One question we could have is whether to change the type here to have enumerated values for the special cases, but given that these are actually invalid values - then I think I prefer this change as-is.

@dplore @rsgcp @xw-g -- any objections?

[earies]

This LGTM. One question we could have is whether to change the type here to have enumerated values for the special cases, but given that these are actually invalid values - then I think I prefer this change as-is.

@dplore @rsgcp @xw-g -- any objections?

Thx @robshakir - yeah, if all options were available for config or state of all possible values in a different construct/location then I would say yes but since this restriction is on the interface id itself it's locked into the range for 'SDN Ports'

[dplore]

Does an SDN controller ever need to address the SDN_PORT_RECIRCULATE or SDN_PORT_CPU? Reading from the spec it seems like this is explicitly defined and permitted? Seems like the change should permit those values.

[earies]

Does an SDN controller ever need to address the SDN_PORT_RECIRCULATE or SDN_PORT_CPU? Reading from the spec it seems like this is explicitly defined and permitted? Seems like the change should permit those values.

This restriction is solely for the interface id node - e.g. to define an identifier that falls within the valid SDN_PORT range

RECIRCULATE and CPU would not be values ever assigned to a interface in this context

[dplore]

Ok, I think you are saying /interfaces/interface is not expected to include CPU or RECIRCULATE entities and therefore we don't need to allow those IDs?

[earies]

Ok, I think you are saying /interfaces/interface is not expected to include CPU or RECIRCULATE entities and therefore we don't need to allow those IDs?

Correct

[dplore]

/gcbrun

[dplore]

Hi @wenovus we are seeing ygot CI check failing here?

[wenovus]

Hi @wenovus we are seeing ygot CI check failing here?

Is the PR in sync with master?

[dplore]

@earies please rebase/merge with master to pick up changes that should fix the CI issue. :)

[earies]

@earies please rebase/merge with master to pick up changes that should fix the CI issue. :)

thx @dplore @wenovus - branch sync'd and passing CI now