robshakir
commented
14 hours ago It looks like this might be something that overlaps with the reworked structure in https://github.com/openconfig/gribi/pull/93 -- could you PTAL, it might allow you to remove the augment that you have specified.
Thanks! r.
we are using gribi and aft to forward 5G GTP-U packets, the data plane destination look-up sequence uses the following network header fields:
Using AFT model, each FT is logically separated and entry is added as oneof entry {...}; therefore grouping multiple FT as another FT is not feasible. The current implementation augments the gribi-aft.yang to support a gtp-u table as: | +--ro afts | | +--ro gtp-u | | +--ro gtp-u-entry* [outer-ip-source outer-ip-destination teid] | | +--ro outer-ip-source // -> ../state/outer-ip-source | | +--ro outer-ip-destination // -> ../state/outer-ip-destination | | +--ro teid // -> ../state/teid | | +--ro state | | | +--ro outer-ip-source? // -> ../../gtp-u-in-ip/state/outer-ip/ipv4/source-address | | | +--ro outer-ip-destination? // -> ../../gtp-u-in-ip/state/outer-ip/ipv4/destination-address | | | +--ro teid? // -> ../../gtp-u-in-ip/state/tunnel/teid | | | +--ro traffic-class-mapping-profile? // -> ../../../../../../../qos/traffic-class-mapping-profiles/traffic-class-mapping-profile/config/name | | | +--ro next-hop-group? // -> /oc-ni:network-instances/network-instance/afts/next-hop-groups/next-hop-group/state/id | | | +--ro next-hop-group-network-instance? oc-ni:network-instance-ref | | +--ro gtp-u-in-ip | | | +--ro state | | | +--ro l2 | | | | +--ro vlan? uint16 | | | +--ro outer-ip | | | | +--ro ipv4 | | | | +--ro source-address? oc-inet:ipv4-prefix | | | | +--ro destination-address? oc-inet:ipv4-prefix | | | +--ro udp | | | | +--ro port? oc-inet:port-number | | | +--ro tunnel | | | | +--ro teid? uint64 | | | | +--ro qfi? uint8 | | | +--ro inner-ip | | | +--ro ip-address-set? ip-address-set. // enum for v4 or v6 ip | | | +--ro ipv4 | | | | +--ro source-address? oc-inet:ipv4-prefix | | | | +--ro destination-address? oc-inet:ipv4-prefix | | | | +--ro dscp? oc-inet:dscp | | | +--ro ipv6 | | | +--ro source-address? oc-inet:ipv6-prefix | | | +--ro source-flow-label? oc-inet:ipv6-flow-label | | | +--ro destination-address? oc-inet:ipv6-prefix | | | +--ro destination-flow-label? oc-inet:ipv6-flow-label | | | +--ro dscp? oc-inet:dscp
There are other augmentations like traffic-class-mapping-profile included here to enable per flow QoS, but ignore for now,, we will create a thread in QoS.
Looking forward to getting your feedback. Thanks
David