staging: fsl-dpaa2/eth: Flow affinity for non-forwarded traffic

The previous patch ensures Tx flow affinity for forwarded frames,
but for termination traffic the initial flow affinity is determined
based on the skb hash, which is expected to hit only a few Tx queues
when there is a small number of flows.

Instead, use XPS (transmit packet steering) to set netdevice queue
affinity to the sending core.

Signed-off-by: Bogdan Purcareata <bogdan.purcareata@nxp.com>
Signed-off-by: Ioana Radulescu <ruxandra.radulescu@nxp.com>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

drivers/staging/fsl-dpaa2/ethernet/dpaa2-eth.c

@@ -1620,9 +1620,10 @@ static struct dpaa2_eth_channel *get_affine_channel(struct dpaa2_eth_priv *priv,
 static void set_fq_affinity(struct dpaa2_eth_priv *priv)
 {
 	struct device *dev = priv->net_dev->dev.parent;
+	struct cpumask xps_mask;
 	struct dpaa2_eth_fq *fq;
 	int rx_cpu, txc_cpu;
-	int i;
+	int i, err;
 
 	/* For each FQ, pick one channel/CPU to deliver frames to.
 	 * This may well change at runtime, either through irqbalance or
@@ -1641,6 +1642,17 @@ static void set_fq_affinity(struct dpaa2_eth_priv *priv)
 			break;
 		case DPAA2_TX_CONF_FQ:
 			fq->target_cpu = txc_cpu;
+
+			/* Tell the stack to affine to txc_cpu the Tx queue
+			 * associated with the confirmation one
+			 */
+			cpumask_clear(&xps_mask);
+			cpumask_set_cpu(txc_cpu, &xps_mask);
+			err = netif_set_xps_queue(priv->net_dev, &xps_mask,
+						  fq->flowid);
+			if (err)
+				dev_err(dev, "Error setting XPS queue\n");
+
 			txc_cpu = cpumask_next(txc_cpu, &priv->dpio_cpumask);
 			if (txc_cpu >= nr_cpu_ids)
 				txc_cpu = cpumask_first(&priv->dpio_cpumask);