Merge branch 'dpaa_eth-rss'

Madalin Bucur says:

====================
Add RSS to DPAA 1.x Ethernet driver

This patch set introduces Receive Side Scaling for the DPAA Ethernet
driver. Documentation is updated with details related to the new
feature and limitations that apply.
Added also a small fix.

v2: removed a C++ style comment
v3: move struct fman to header file to avoid exporting a function
v4: addressed compilation issues introduced in v3
====================

Signed-off-by: David S. Miller <davem@davemloft.net>

1 files changed, 66 insertions, 10 deletions

diff --git a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
index c7fa285378b5..42258060f142 100644
--- a/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
+++ b/drivers/net/ethernet/freescale/dpaa/dpaa_eth.c
@@ -158,7 +158,7 @@ MODULE_PARM_DESC(tx_timeout, "The Tx timeout in ms");
 #define DPAA_RX_PRIV_DATA_SIZE	(u16)(DPAA_TX_PRIV_DATA_SIZE + \
 					dpaa_rx_extra_headroom)
 
-#define DPAA_ETH_RX_QUEUES	128
+#define DPAA_ETH_PCD_RXQ_NUM	128
 
 #define DPAA_ENQUEUE_RETRIES	100000
 
@@ -169,6 +169,7 @@ struct fm_port_fqs {
 	struct dpaa_fq *tx_errq;
 	struct dpaa_fq *rx_defq;
 	struct dpaa_fq *rx_errq;
+	struct dpaa_fq *rx_pcdq;
 };
 
 /* All the dpa bps in use at any moment */
@@ -235,7 +236,7 @@ static int dpaa_netdev_init(struct net_device *net_dev,
 	net_dev->max_mtu = dpaa_get_max_mtu();
 
 	net_dev->hw_features |= (NETIF_F_IP_CSUM | NETIF_F_IPV6_CSUM |
-				 NETIF_F_LLTX);
+				 NETIF_F_LLTX | NETIF_F_RXHASH);
 
 	net_dev->hw_features |= NETIF_F_SG | NETIF_F_HIGHDMA;
 	/* The kernels enables GSO automatically, if we declare NETIF_F_SG.
@@ -628,6 +629,7 @@ static inline void dpaa_assign_wq(struct dpaa_fq *fq, int idx)
 		fq->wq = 5;
 		break;
 	case FQ_TYPE_RX_DEFAULT:
+	case FQ_TYPE_RX_PCD:
 		fq->wq = 6;
 		break;
 	case FQ_TYPE_TX:
@@ -688,6 +690,7 @@ static int dpaa_alloc_all_fqs(struct device *dev, struct list_head *list,
 			      struct fm_port_fqs *port_fqs)
 {
 	struct dpaa_fq *dpaa_fq;
+	u32 fq_base, fq_base_aligned, i;
 
 	dpaa_fq = dpaa_fq_alloc(dev, 0, 1, list, FQ_TYPE_RX_ERROR);
 	if (!dpaa_fq)
@@ -701,6 +704,26 @@ static int dpaa_alloc_all_fqs(struct device *dev, struct list_head *list,
 
 	port_fqs->rx_defq = &dpaa_fq[0];
 
+	/* the PCD FQIDs range needs to be aligned for correct operation */
+	if (qman_alloc_fqid_range(&fq_base, 2 * DPAA_ETH_PCD_RXQ_NUM))
+		goto fq_alloc_failed;
+
+	fq_base_aligned = ALIGN(fq_base, DPAA_ETH_PCD_RXQ_NUM);
+
+	for (i = fq_base; i < fq_base_aligned; i++)
+		qman_release_fqid(i);
+
+	for (i = fq_base_aligned + DPAA_ETH_PCD_RXQ_NUM;
+	     i < (fq_base + 2 * DPAA_ETH_PCD_RXQ_NUM); i++)
+		qman_release_fqid(i);
+
+	dpaa_fq = dpaa_fq_alloc(dev, fq_base_aligned, DPAA_ETH_PCD_RXQ_NUM,
+				list, FQ_TYPE_RX_PCD);
+	if (!dpaa_fq)
+		goto fq_alloc_failed;
+
+	port_fqs->rx_pcdq = &dpaa_fq[0];
+
 	if (!dpaa_fq_alloc(dev, 0, DPAA_ETH_TXQ_NUM, list, FQ_TYPE_TX_CONF_MQ))
 		goto fq_alloc_failed;
 
@@ -870,13 +893,14 @@ static void dpaa_fq_setup(struct dpaa_priv *priv,
 			  const struct dpaa_fq_cbs *fq_cbs,
 			  struct fman_port *tx_port)
 {
-	int egress_cnt = 0, conf_cnt = 0, num_portals = 0, cpu;
+	int egress_cnt = 0, conf_cnt = 0, num_portals = 0, portal_cnt = 0, cpu;
 	const cpumask_t *affine_cpus = qman_affine_cpus();
-	u16 portals[NR_CPUS];
+	u16 channels[NR_CPUS];
 	struct dpaa_fq *fq;
 
 	for_each_cpu(cpu, affine_cpus)
-		portals[num_portals++] = qman_affine_channel(cpu);
+		channels[num_portals++] = qman_affine_channel(cpu);
+
 	if (num_portals == 0)
 		dev_err(priv->net_dev->dev.parent,
 			"No Qman software (affine) channels found");
@@ -890,6 +914,12 @@ static void dpaa_fq_setup(struct dpaa_priv *priv,
 		case FQ_TYPE_RX_ERROR:
 			dpaa_setup_ingress(priv, fq, &fq_cbs->rx_errq);
 			break;
+		case FQ_TYPE_RX_PCD:
+			if (!num_portals)
+				continue;
+			dpaa_setup_ingress(priv, fq, &fq_cbs->rx_defq);
+			fq->channel = channels[portal_cnt++ % num_portals];
+			break;
 		case FQ_TYPE_TX:
 			dpaa_setup_egress(priv, fq, tx_port,
 					  &fq_cbs->egress_ern);
@@ -1039,7 +1069,8 @@ static int dpaa_fq_init(struct dpaa_fq *dpaa_fq, bool td_enable)
 		/* Put all the ingress queues in our "ingress CGR". */
 		if (priv->use_ingress_cgr &&
 		    (dpaa_fq->fq_type == FQ_TYPE_RX_DEFAULT ||
-		     dpaa_fq->fq_type == FQ_TYPE_RX_ERROR)) {
+		     dpaa_fq->fq_type == FQ_TYPE_RX_ERROR ||
+		     dpaa_fq->fq_type == FQ_TYPE_RX_PCD)) {
 			initfq.we_mask |= cpu_to_be16(QM_INITFQ_WE_CGID);
 			initfq.fqd.fq_ctrl |= cpu_to_be16(QM_FQCTRL_CGE);
 			initfq.fqd.cgid = (u8)priv->ingress_cgr.cgrid;
@@ -1170,7 +1201,7 @@ static int dpaa_eth_init_tx_port(struct fman_port *port, struct dpaa_fq *errq,
 
 static int dpaa_eth_init_rx_port(struct fman_port *port, struct dpaa_bp **bps,
 				 size_t count, struct dpaa_fq *errq,
-				 struct dpaa_fq *defq,
+				 struct dpaa_fq *defq, struct dpaa_fq *pcdq,
 				 struct dpaa_buffer_layout *buf_layout)
 {
 	struct fman_buffer_prefix_content buf_prefix_content;
@@ -1190,6 +1221,10 @@ static int dpaa_eth_init_rx_port(struct fman_port *port, struct dpaa_bp **bps,
 	rx_p = &params.specific_params.rx_params;
 	rx_p->err_fqid = errq->fqid;
 	rx_p->dflt_fqid = defq->fqid;
+	if (pcdq) {
+		rx_p->pcd_base_fqid = pcdq->fqid;
+		rx_p->pcd_fqs_count = DPAA_ETH_PCD_RXQ_NUM;
+	}
 
 	count = min(ARRAY_SIZE(rx_p->ext_buf_pools.ext_buf_pool), count);
 	rx_p->ext_buf_pools.num_of_pools_used = (u8)count;
@@ -1234,7 +1269,8 @@ static int dpaa_eth_init_ports(struct mac_device *mac_dev,
 		return err;
 
 	err = dpaa_eth_init_rx_port(rxport, bps, count, port_fqs->rx_errq,
-				    port_fqs->rx_defq, &buf_layout[RX]);
+				    port_fqs->rx_defq, port_fqs->rx_pcdq,
+				    &buf_layout[RX]);
 
 	return err;
 }
@@ -2201,12 +2237,13 @@ static enum qman_cb_dqrr_result rx_default_dqrr(struct qman_portal *portal,
 	dma_addr_t addr = qm_fd_addr(fd);
 	enum qm_fd_format fd_format;
 	struct net_device *net_dev;
-	u32 fd_status;
+	u32 fd_status, hash_offset;
 	struct dpaa_bp *dpaa_bp;
 	struct dpaa_priv *priv;
 	unsigned int skb_len;
 	struct sk_buff *skb;
 	int *count_ptr;
+	void *vaddr;
 
 	fd_status = be32_to_cpu(fd->status);
 	fd_format = qm_fd_get_format(fd);
@@ -2252,7 +2289,8 @@ static enum qman_cb_dqrr_result rx_default_dqrr(struct qman_portal *portal,
 	dma_unmap_single(dpaa_bp->dev, addr, dpaa_bp->size, DMA_FROM_DEVICE);
 
 	/* prefetch the first 64 bytes of the frame or the SGT start */
-	prefetch(phys_to_virt(addr) + qm_fd_get_offset(fd));
+	vaddr = phys_to_virt(addr);
+	prefetch(vaddr + qm_fd_get_offset(fd));
 
 	fd_format = qm_fd_get_format(fd);
 	/* The only FD types that we may receive are contig and S/G */
@@ -2273,6 +2311,18 @@ static enum qman_cb_dqrr_result rx_default_dqrr(struct qman_portal *portal,
 
 	skb->protocol = eth_type_trans(skb, net_dev);
 
+	if (net_dev->features & NETIF_F_RXHASH && priv->keygen_in_use &&
+	    !fman_port_get_hash_result_offset(priv->mac_dev->port[RX],
+					      &hash_offset)) {
+		enum pkt_hash_types type;
+
+		/* if L4 exists, it was used in the hash generation */
+		type = be32_to_cpu(fd->status) & FM_FD_STAT_L4CV ?
+			PKT_HASH_TYPE_L4 : PKT_HASH_TYPE_L3;
+		skb_set_hash(skb, be32_to_cpu(*(u32 *)(vaddr + hash_offset)),
+			     type);
+	}
+
 	skb_len = skb->len;
 
 	if (unlikely(netif_receive_skb(skb) == NET_RX_DROP))
@@ -2511,6 +2561,9 @@ static struct dpaa_bp *dpaa_bp_alloc(struct device *dev)
 
 	dpaa_bp->bpid = FSL_DPAA_BPID_INV;
 	dpaa_bp->percpu_count = devm_alloc_percpu(dev, *dpaa_bp->percpu_count);
+	if (!dpaa_bp->percpu_count)
+		return ERR_PTR(-ENOMEM);
+
 	dpaa_bp->config_count = FSL_DPAA_ETH_MAX_BUF_COUNT;
 
 	dpaa_bp->seed_cb = dpaa_bp_seed;
@@ -2738,6 +2791,9 @@ static int dpaa_eth_probe(struct platform_device *pdev)
 	if (err)
 		goto init_ports_failed;
 
+	/* Rx traffic distribution based on keygen hashing defaults to on */
+	priv->keygen_in_use = true;
+
 	priv->percpu_priv = devm_alloc_percpu(dev, *priv->percpu_priv);
 	if (!priv->percpu_priv) {
 		dev_err(dev, "devm_alloc_percpu() failed\n");