On 6/19/24 19:29, Kevin Traynor wrote: > On 19/06/2024 17:00, David Marchand wrote: >> Querying link status may get delayed for an undeterministic (long) time >> with mlx5 ports. This is a consequence of the mlx5 driver calling ethtool >> kernel API and getting stuck on the kernel RTNL lock while some other >> operation is in progress under this lock. >> >> One impact for long link status query is that it is called under the bond >> lock taken in write mode periodically in bond_run(). >> In parallel, datapath threads may block requesting to read bonding related >> info (like for example in bond_check_admissibility()). >> >> The LSC interrupt mode is available with many DPDK drivers and is used by >> default with testpmd. >> >> It seems safe enough to switch on this feature by default in OVS. >> We keep the per interface option to disable this feature in case of an >> unforeseen bug. >> >> Signed-off-by: David Marchand <[email protected]> >> Reviewed-by: Robin Jarry <[email protected]> >> Acked-by: Mike Pattrick <[email protected]> >> --- > > LGTM > > Acked-by: Kevin Traynor <[email protected]> > > Re backporting. I'm not so keen on changing a default when someone > upgrades from 3.3.1 -> 3.3.2 etc. The feature is already available so > anyone who needs it can enable it.
+1 Changing defaults in stable branches is not a good practice. Since the behavior was there for a very long time and there is a workaround for those who hit the issue, it should be enough. I'd suggest considering the change inside the DPDK itself, i.e. enabling LSC by default for all or problematic (mlx) drivers. For the patch itself, I'm not sure why WARN and DBG messages are worded differently. I'd prefer them to be the same (as in WARN). But that's a minor nit that can be fixed on commit, if you think it is needed. Best regards, Ilya Maximets. _______________________________________________ dev mailing list [email protected] https://mail.openvswitch.org/mailman/listinfo/ovs-dev
