Questions about examples/digest_messages about p4-learning

[edgar-costa]

Hi @butuizd

1. There is no official documentation about digest. The only way to know how to use it, is by the little information you can find in mailing lists, issues, or here. Now I can not find more info, but basically:

device_id is the switch id, ctx_id, list_id and buffer_id somehow identify the digested packet. And it needs to be used to acknowledge it to the switch https://github.com/nsg-ethz/p4-learning/blob/master/examples/digest_messages/get_digest.py#L29

num is the number of aggregated digests.

2. Digests are somehow buffered, if you call digest very fast it will in theory batch several packets together, I guess the time it buffers can be found in the simple_switch code. Also, if you digest the same thing multiple times it will only appear one time in the "buffer".

If you want a different behaviour, use copy to cpu.

3. Again, there is no more info than the code itself (unless there is new documentation). The, function is used to acknowledge one digest message, something needed to continue digesting new packets.

[butuizd]

Thanks a lot @edgar-costa . If I use the p4runtime API, whether I can process the digest message in controller more easily ,which means I don't have to care the fields such as topic, device_id, ctx_id, list_id, buffer_id, num .

[edgar-costa]

Well, then, all good!

Our new version of p4-utils, will have a P4runtime based controller.