Skip to content

prepare: don't bring up simulated devices

Lubomir Rintel requested to merge lr/simdev-down into main

Bringing up our end of the Veth pair is not a good idea, with dnsmasq sending RAs and the kernel accepting them, the devices might end up being externally configured. This typically not what we'd want during testing.

Specifically, this interferes with the @ovs_cloned_mac_set_on_iface. The test creates a testX device and an ovs-testX connection for the device. It expects it to autoactivate, but that doesn't always happen because the device might end up being externally configured after getting a RA from dnsmasq.

That results in a truly sad state of affairs:

--- nmcli -f ALL d --- DEVICE TYPE STATE CONNECTION testX ethernet connected (externally) testX

--- nmcli -f ALL c --- NAME TYPE ACTIVE DEVICE STATE SLAVE testX ethernet yes testX activated -- ovs-testX ethernet no -- -- ovs-port

Edited by Lubomir Rintel

Merge request reports