Hi Kevin, Kevin Traynor, Jun 22, 2023 at 15:06: > Hi Robin, > > As discussed offline, the issue is fixed in v12. In the case where > offload has failed on device A, and device B triggers set_config the > set_config for device B will see that device B has the correct number of > queues (user_n_rxq), so there will not be a reconfigure of *that* > device. I tested this by faking some failed offloads with gdb. > > If a reconfigure for device A does occur for some other reason, then it > will attempt to apply the config again, (presumably) the offload will > fail again and the normal retry occurs without offload. This seems the > correct behaviour to me.
Somehow I did assume that netdev_dpdk_reconfigure() would be called for all ports regardless if their configuration had changed. Good to know that the problem is now fixed. > For v12, I also tested some basic increasing decreasing rxqs, enabling > hwol and it's working as expected. > > thanks, > Kevin. > > Acked-by: Kevin Traynor <ktray...@redhat.com> Thanks a lot! _______________________________________________ dev mailing list d...@openvswitch.org https://mail.openvswitch.org/mailman/listinfo/ovs-dev
