Wed, Jan 27, 2016 at 04:51:14PM CET, and...@lunn.ch wrote: >On Wed, Jan 27, 2016 at 03:46:57PM +0100, Jiri Pirko wrote: >> Hi All. >> >> Just want to send a small note here about what we use for mlxsw driver >> automated testing. It can be handy for other people using switchdev >> infra, namely DSA guys. > >Hi Jiri > >I have a few hacked together python scripts combined with the Ostinato >drone, integrated into a Jenkins CI system, for testing DSA. > >Taking a quick look at the XML files, it looks like we took different >approaches. I have a host with 8 Ethernet interfaces connected to the >switch. You appear to have many more machines, with one or two >interfaces connected to the switch per machine. This is going to make >it difficult for me to run your tests using my setup.
You can run 2 lnst-slave instances on a single host, each using 4 NICs. Then you can run our tests as well. Or you can setup 2 VMs and put NICs in them using pci passthrough. > >I only took a quick look at the tests, but it also does not seem like >the linux machine in the switch itself is involved in the >tests. e.g. l2-000-minimal.py, you put IP addresses on the two >machines connected to the switch, but what about putting an address on >the linux bridge interface and make sure you can ping that as well? >This is important for DSA, since the switch is often embedded in a >WiFi access point and a lot of traffic is coming from the wifi device >attached to the linux host. This usecase we do not test now. But it is very easy to add another recipe for this. > >> It's just a start. We'd like to make them more complete, possibly with >> your help. > >If i can make them work with my testbed setup, and i can get test >results in a form Jenkins understands, i would be interested in using >them. AFAIK, we run LNST from Jenkins as well. That should not be a problem. > >Are you going to be in the netdev 1.1 switchdev BoF? Maybe we can >discuss testing there? I will be there. Sure, lets discuss it there.
