SDFAB-155 Allow forwarding of packet-outs like regular packets

[ccascone]

By default, all packet-outs are sent straight to the egress port passed as a controller packet-out metadata, bypassing the forwarding tables in the ingress pipe. With this change, the control plane can set a new packet-out metadata named do_forwarding to instruct the parser to forward packet-outs as regular packets.

This is required to support Xn-based handover for the DT trial, where PFCP agent will send "incomplete" packet-outs for the base stations carrying "end-marker" packets. It will be up to the switch to "complete" packet-outs by selecting the right egress port, VLAN tag, setting MAC addresses, etc.

When handling OutboundPacket in ONOS, the pipeconf (interpreter) uses the OUTPUT instruction with logical port TABLE to enable forwarding. This is consistent with the OpenFlow behavior, from the spec:

Required: TABLE: Represents the start of the OpenFlow pipeline (see 5.1). This port is only valid in an output action in the action list of a packet-out message (see 7.3.7), and submits the packet to the first flow table so that the packet can be processed through the regular OpenFlow pipeline.

TODO:

• Add pipeconf Java unit test for when do_forwarding=1
• Initialize filtering tables for cpu port in pipeliner

[charlesmcchan]

This could also help the following scenario if I understand correctly.

Imaging in a 2x2, host on leaf1 trying to ping switch interface IP of leaf2.
The request will get routed via host -> leaf1 -> spine1/2 -> leaf2 and get packet-in at leaf2.
Master of leaf2 will respond using packet-out. However, ONOS needs to first calculate the MPLS label of leaf1 and push that as part of the packet-out.

With this feature, ONOS could simply swap source and destination IP, set this flag, let the original routing logic on leaf2 push the MPLS label, and forward the packet back to leaf1.

[ccascone]

@charlesmcchan your understanding is correct

[codecov]

Codecov Report

Merging #262 (f535888) into main (51b0f82) will decrease coverage by 0.01%.
The diff coverage is 77.91%.

❗ Current head f535888 differs from pull request most recent head 018ea78. Consider uploading reports for the commit 018ea78 to get more accurate results

@@             Coverage Diff              @@
##               main     #262      +/-   ##
============================================
- Coverage     74.13%   74.12%   -0.02%     
- Complexity      273      360      +87     
============================================
  Files            18       31      +13     
  Lines          1875     2415     +540     
  Branches        153      211      +58     
============================================
+ Hits           1390     1790     +400     
- Misses          401      515     +114     
- Partials         84      110      +26     

Impacted Files	Coverage Δ
...viour/pipeliner/ForwardingObjectiveTranslator.java	75.40% <0.00%> (ø)
...atumproject/fabric/tna/stats/HighlightManager.java	63.08% <63.08%> (ø)
...atumproject/fabric/tna/stats/StatisticManager.java	63.45% <63.45%> (ø)
...roject/fabric/tna/behaviour/FabricInterpreter.java	76.97% <66.66%> (+1.42%)	⬆️
...org/stratumproject/fabric/tna/stats/cli/Utils.java	80.00% <80.00%> (ø)
.../stratumproject/fabric/tna/stats/HighlightKey.java	92.59% <92.59%> (ø)
...abric/tna/behaviour/pipeliner/FabricPipeliner.java	41.77% <94.56%> (-4.15%)	⬇️
.../stratumproject/fabric/tna/stats/StatisticKey.java	95.23% <95.23%> (ø)
...atumproject/fabric/tna/stats/StatisticDataKey.java	96.55% <96.55%> (ø)
...umproject/fabric/tna/stats/StatisticDataValue.java	97.50% <97.50%> (ø)
... and 21 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 51b0f82...018ea78. Read the comment docs.

[ccascone]

@abhilashendurthi will this require changes to the TV runner for fabric-tna? The packet-out metadata ID for ether_type has changed from 4 to 5

[abhilashendurthi]

Yes, this will require me to change the metadata ID for padding from 3 to 4 and metadata ID for ether_type from 4 to 5. Also for, do_forwarding metadata with ID 3, should the value be 0 in TV runner?

[ccascone]

Yes, in TV runner do_forwarding should be 0

[ccascone]

@pierventre @Yi-Tseng I think I have addressed all your comments, PTAL.

[pierventre]

Looks good to me

[pierventre]

is the TODO list updated ?

[ccascone]

@pierventre

is the TODO list updated ?

I think so. Why do you ask?