[ovs-discuss] how code calculates per-flow statistics

[dsainz at CS.technion.ac.il]

Hi,

As I understood from previous questions I asked here, flows in datapath
are a collapsed version of the OpenFlow flows, after traversing all
pertinent tables.
My question is if, given a pipeline T1->T2->T3, how the code can keep
per-flow statistics of, say, T2 specifically. My understanding is that
after collapsing flows you only get per-flow stats of the packets that
arrive to the final table, but there could be packets arrving to T2 but
never to T3. How is that stats separation made?

Thanks,

David