net/mlx5e: Term table handling of internal port rules
authorAriel Levkovich <lariel@nvidia.com>
Tue, 7 Sep 2021 20:30:59 +0000 (23:30 +0300)
committerSaeed Mahameed <saeedm@nvidia.com>
Fri, 29 Oct 2021 20:53:31 +0000 (13:53 -0700)
Adjust termination table logic to handle rules which
involve internal port as filter or forwarding device.

For cases where the rule forwards from internal port
to uplink, always choose to go via termination table.
This is because it is not known from where the packet
originally arrived to the internal port and it is possible
that it came from the uplink itself, in which case
a term table is required to perform hairpin.
If the packet arrived from a vport, going via term
table has no effect.

For cases where the rule forwards to an internal port
from uplink the rep pointer will point to the uplink rep,
avoid going via termination table as it is not required.

Signed-off-by: Ariel Levkovich <lariel@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads_termtbl.c

index d0407b3..182306b 100644 (file)
@@ -220,7 +220,7 @@ mlx5_eswitch_termtbl_required(struct mlx5_eswitch *esw,
        if (!MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, termination_table) ||
            !MLX5_CAP_ESW_FLOWTABLE_FDB(esw->dev, ignore_flow_level) ||
            mlx5_esw_attr_flags_skip(attr->flags) ||
-           !mlx5_eswitch_offload_is_uplink_port(esw, spec))
+           (!mlx5_eswitch_offload_is_uplink_port(esw, spec) && !esw_attr->int_port))
                return false;
 
        /* push vlan on RX */
@@ -229,7 +229,7 @@ mlx5_eswitch_termtbl_required(struct mlx5_eswitch *esw,
 
        /* hairpin */
        for (i = esw_attr->split_count; i < esw_attr->out_count; i++)
-               if (esw_attr->dests[i].rep &&
+               if (!esw_attr->dest_int_port && esw_attr->dests[i].rep &&
                    esw_attr->dests[i].rep->vport == MLX5_VPORT_UPLINK)
                        return true;