net/mlx5e: E-Switch, Fix shared fdb error flow

On error flow resources being freed in esw_master_egress_destroy_resources()
but pointers not being set to null if error flow is from creating a
bounce rule. Then in esw_acl_egress_ofld_cleanup() we try to access already
freed pointers. Fix it by resetting the pointers to null.
Also if error is from creating a second or later bounce rule then the
flow group and table being used and cannot and should not be freed.
Add a check to destroy the flow group and table if there are no bounce
rules.

mlx5_core.sf mlx5_core.sf.2: mlx5_destroy_flow_group:2306:(pid 2235): Flow group 4 wasn't destroyed, refcount > 1
mlx5_core.sf mlx5_core.sf.2: mlx5_destroy_flow_table:2295:(pid 2235): Flow table 3 wasn't destroyed, refcount > 1

Fixes: 5e0202e ("net/mlx5: E-switch, Handle multiple master egress rules")
Signed-off-by: Roi Dayan <roid@nvidia.com>
Reviewed-by: Shay Drory <shayd@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

drivers/net/ethernet/mellanox/mlx5/core/eswitch_offloads.c

@@ -2618,8 +2618,12 @@ static int esw_master_egress_create_resources(struct mlx5_eswitch *esw,
 
 static void esw_master_egress_destroy_resources(struct mlx5_vport *vport)
 {
+	if (!xa_empty(&vport->egress.offloads.bounce_rules))
+		return;
 	mlx5_destroy_flow_group(vport->egress.offloads.bounce_grp);
+	vport->egress.offloads.bounce_grp = NULL;
 	mlx5_destroy_flow_table(vport->egress.acl);
+	vport->egress.acl = NULL;
 }
 
 static int esw_set_master_egress_rule(struct mlx5_core_dev *master,