On Sat, Aug 25, 2018 at 12:57:07PM +1200, Brad Cowie wrote: > On 25 August 2018 at 12:07, Ben Pfaff <b...@ovn.org> wrote: > > > On Sat, Aug 25, 2018 at 11:09:41AM +1200, Brad Cowie wrote: > > > By the way we can dynamically generate our openflow pipeline now such > > > that tables are automatically sized (more important for real hardware) > > > and disable tables that aren't used (e.g. FIB tables are now turned > > > off when routing disabled, ACL tables disabled when no ACLs have been > > > applied). We are mostly doing this by sending a OFPMP_TABLE_FEATURES > > > packet with the tables and matches we wish to use in each (we also > > > include a name for each table which I see OVS 2.10.0 now > > > supports). However we haven't been able to get OVS to successfully > > > parse our OFPMP_TABLE_FEATURES packet yet. I guess I should open a new > > > thread on ovs-dev to discuss whether OVS actually supports this or > > > not? > > > > So far, OVS provides a pretty bare OFPMP_TABLE_FEATURES implementation, > > which allows a controller to query the features that it supports. It > > provides very little (maybe nothing?) in terms of allowing a controller > > to modify the features supported. That's primarily because OVS is > > mainly a software implementation where we can support all the features > > all the time. The motivation for OFPMP_TABLE_FEATURES when it was added > > to the OpenFlow specs was to allow a controller to configure hardware > > features that offer trade-offs. OVS does of course run on some hardware > > switches; I don't know whether the vendors who customize it for their > > hardware add support for OFPMP_TABLE_FEATURES so that it can configure > > trade-offs, though I suspect they do not. > > > > That pretty much answers all my questions right there. We would never > really need to modify the OVS pipeline as you say it does everything > we need in software. It would be interesting for us to be able to use > it to be able to use openflow as an API to rename flow tables to make > the output of "ovs-ofctl dump-flows" easy to understand.
Well, that wouldn't be too hard, I guess. It's too bad that there's no way with a Table Features Request to just modify *some* of the properties of a table (such as the name). I guess that Faucet would have to spit back at OVS all the capabilities that OVS says it supports, and then OVS would have to verify that it was actually the same. That seems a little prone to error, but probably we could get it right. _______________________________________________ discuss mailing list disc...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-discuss
