| Message ID | c1f4cc6214c28ce9a39147db9f3b66927dbae612.1578567988.git.dcaratti@redhat.com |
|---|---|
| State | Awaiting Upstream |
| Delegated to: | David Miller |
| Headers | show |
| Series | [net-next] net/mlx5e: allow TSO on VXLAN over VLAN topologies | expand |
From: Davide Caratti <dcaratti@redhat.com> Date: Thu, 9 Jan 2020 12:07:59 +0100 > since mlx5 hardware can segment correctly TSO packets on VXLAN over VLAN > topologies, CPU usage can improve significantly if we enable tunnel > offloads in dev->vlan_features, like it was done in the past with other > NIC drivers (e.g. mlx4, be2net and ixgbe). > > Signed-off-by: Davide Caratti <dcaratti@redhat.com> Saeed, I am assuming you will review and integrate this. Thank you.
On Thu, 2020-01-09 at 10:52 -0800, David Miller wrote: > From: Davide Caratti <dcaratti@redhat.com> > Date: Thu, 9 Jan 2020 12:07:59 +0100 > > > since mlx5 hardware can segment correctly TSO packets on VXLAN over > VLAN > > topologies, CPU usage can improve significantly if we enable tunnel > > offloads in dev->vlan_features, like it was done in the past with > other > > NIC drivers (e.g. mlx4, be2net and ixgbe). > > > > Signed-off-by: Davide Caratti <dcaratti@redhat.com> > > Saeed, I am assuming you will review and integrate this. > > Thank you. LGTM, applied to net-next-mlx5, will run some tests and provide feedback if any issues will surface, while i don't expect any. Will submit to net-next in one of my upcoming net-next pull requests. Thanks, Saeed.
diff --git a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c index 78737fd42616..87267c18ff8c 100644 --- a/drivers/net/ethernet/mellanox/mlx5/core/en_main.c +++ b/drivers/net/ethernet/mellanox/mlx5/core/en_main.c @@ -4878,6 +4878,8 @@ static void mlx5e_build_nic_netdev(struct net_device *netdev) netdev->hw_enc_features |= NETIF_F_GSO_UDP_TUNNEL | NETIF_F_GSO_UDP_TUNNEL_CSUM; netdev->gso_partial_features = NETIF_F_GSO_UDP_TUNNEL_CSUM; + netdev->vlan_features |= NETIF_F_GSO_UDP_TUNNEL | + NETIF_F_GSO_UDP_TUNNEL_CSUM; } if (mlx5e_tunnel_proto_supported(mdev, IPPROTO_GRE)) {
since mlx5 hardware can segment correctly TSO packets on VXLAN over VLAN topologies, CPU usage can improve significantly if we enable tunnel offloads in dev->vlan_features, like it was done in the past with other NIC drivers (e.g. mlx4, be2net and ixgbe). Signed-off-by: Davide Caratti <dcaratti@redhat.com> --- drivers/net/ethernet/mellanox/mlx5/core/en_main.c | 2 ++ 1 file changed, 2 insertions(+)