> On Mar 29, 2016, at 2:07 PM, Russell Bryant <russ...@ovn.org> wrote: > > On Tue, Mar 29, 2016 at 5:00 PM, Justin Pettit <jpet...@ovn.org> wrote: > > > On Mar 28, 2016, at 2:33 PM, Russell Bryant <russ...@ovn.org> wrote: > > > > >>> > Yeah, no tests. I honestly wasn't sure how to test it since we can't use >>> > ct() in the test suite. I was hoping that we could start adding some >>> > tests for this stuff once the userspace conntrack patches go in. >>> >> There actually are some tests written that are executed when you run "make >> check-kernel", but it has to load the kernel module locally and run them. >> Those tests don't do bitwise manipulation, though. Obviously, userspace >> conntrack would make all of this easier. Daniele plans to start working on >> those patches again soon. > > Yeah, we don't have any OVN tests in the check-kernel suite. There's nothing > kernel datapath specific in OVN, so ideally we could keep all of our tests in > the other suite, which is easier to run in more places.
If we can avoid OVN tests with check-kernel, that would be ideal. I was just referring to testing the OVS functionality, which only recently had the ability to run kernel unit tests. >>> > Bit-level twiddling would indeed be nice. I didn't have a need for it in >>> > this series, though. Are you OK with it coming as a future enhancement, >>> > or would you like to see it now? >>> >> In general, I think its nicer to provide a completed interface if it doesn't >> add too much complexity. (I don't think it requires much (if any) more code >> since we have functions that parse bit-level setting.) Also, the intent of >> the next patch is really to do flag manipulation and not setting a 128-bit >> field. However, there should have been some bit-level tests for the >> functionality that's already checked in, so I'm happy to just add the tests >> and this flexibility to my to-do list. >> > Yeah, you're right, it does matching on the bit. It's setting the whole > value, which only works because this was the first use of the field. It'll > be needed before we add a second bit. > > I can add it now. I was mostly just being lazy. It may be next week before > I get to it, though. I'm traveling all week and hacking time is hit or miss. I definitely understand. :-) If you can do that, it would be great, but I'm okay if we want to add the support once we clearly document and test the capability in OVS itself. I've sent Joe a message confirming whether this is indeed supported. Thanks! --Justin _______________________________________________ dev mailing list dev@openvswitch.org http://openvswitch.org/mailman/listinfo/dev
