net/mlx5: Node-aware allocation for the EQs

Prefer the aware allocation, use the device NUMA node.
Signed-off-by: Tariq Toukan <tariqt@nvidia.com>
Reviewed-by: Moshe Shemesh <moshe@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>

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

@@ -728,7 +728,8 @@ struct mlx5_eq *
 mlx5_eq_create_generic(struct mlx5_core_dev *dev,
 		       struct mlx5_eq_param *param)
 {
-	struct mlx5_eq *eq = kvzalloc(sizeof(*eq), GFP_KERNEL);
+	struct mlx5_eq *eq = kvzalloc_node(sizeof(*eq), GFP_KERNEL,
+					   dev->priv.numa_node);
 	int err;
 
 	if (!eq)
@@ -888,10 +889,11 @@ static int create_comp_eqs(struct mlx5_core_dev *dev)
 		return ncomp_eqs;
 	INIT_LIST_HEAD(&table->comp_eqs_list);
 	nent = comp_eq_depth_devlink_param_get(dev);
+
 	for (i = 0; i < ncomp_eqs; i++) {
 		struct mlx5_eq_param param = {};
 
-		eq = kzalloc(sizeof(*eq), GFP_KERNEL);
+		eq = kzalloc_node(sizeof(*eq), GFP_KERNEL, dev->priv.numa_node);
 		if (!eq) {
 			err = -ENOMEM;
 			goto clean;