qed/qede: Multi CoS support.

This patch adds support for tc mqprio offload,
using this different traffic classes on the adapter
can be utilized based on configured priority to tc map.

For example -

tc qdisc add dev eth0 root mqprio num_tc 4 map 0 1 2 3

This will cause SKBs with priority 0,1,2,3 to transmit
over tc 0,1,2,3 hardware queues respectively.
Signed-off-by: Manish Chopra <manish.chopra@cavium.com>
Signed-off-by: Ariel Elior <ariel.elior@cavium.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

drivers/net/ethernet/qlogic/qed/qed_l2.c

@@ -2188,16 +2188,17 @@ int qed_get_queue_coalesce(struct qed_hwfn *p_hwfn, u16 *p_coal, void *handle)
 static int qed_fill_eth_dev_info(struct qed_dev *cdev,
 				 struct qed_dev_eth_info *info)
 {
+	struct qed_hwfn *p_hwfn = QED_LEADING_HWFN(cdev);
 	int i;
 
 	memset(info, 0, sizeof(*info));
 
-	info->num_tc = 1;
-
 	if (IS_PF(cdev)) {
 		int max_vf_vlan_filters = 0;
 		int max_vf_mac_filters = 0;
 
+		info->num_tc = p_hwfn->hw_info.num_hw_tc;
+
 		if (cdev->int_params.out.int_mode == QED_INT_MODE_MSIX) {
 			u16 num_queues = 0;
 
@@ -2248,6 +2249,8 @@ static int qed_fill_eth_dev_info(struct qed_dev *cdev,
 	} else {
 		u16 total_cids = 0;
 
+		info->num_tc = 1;
+
 		/* Determine queues &  XDP support */
 		for_each_hwfn(cdev, i) {
 			struct qed_hwfn *p_hwfn = &cdev->hwfns[i];
@@ -2554,7 +2557,7 @@ static int qed_start_txq(struct qed_dev *cdev,
 
 	rc = qed_eth_tx_queue_start(p_hwfn,
 				    p_hwfn->hw_info.opaque_fid,
-				    p_params, 0,
+				    p_params, p_params->tc,
 				    pbl_addr, pbl_size, ret_params);
 
 	if (rc) {

drivers/net/ethernet/qlogic/qed/qed_main.c

@@ -948,13 +948,14 @@ static void qed_update_pf_params(struct qed_dev *cdev,
 		params->eth_pf_params.num_arfs_filters = 0;
 
 	/* In case we might support RDMA, don't allow qede to be greedy
-	 * with the L2 contexts. Allow for 64 queues [rx, tx, xdp] per hwfn.
+	 * with the L2 contexts. Allow for 64 queues [rx, tx cos, xdp]
+	 * per hwfn.
 	 */
 	if (QED_IS_RDMA_PERSONALITY(QED_LEADING_HWFN(cdev))) {
 		u16 *num_cons;
 
 		num_cons = &params->eth_pf_params.num_cons;
-		*num_cons = min_t(u16, *num_cons, 192);
+		*num_cons = min_t(u16, *num_cons, QED_MAX_L2_CONS);
 	}
 
 	for (i = 0; i < cdev->num_hwfns; i++) {

drivers/net/ethernet/qlogic/qede/qede.h

@@ -386,6 +386,15 @@ struct qede_tx_queue {
 #define QEDE_TXQ_XDP_TO_IDX(edev, txq)	((txq)->index - \
 					 QEDE_MAX_TSS_CNT(edev))
 #define QEDE_TXQ_IDX_TO_XDP(edev, idx)	((idx) + QEDE_MAX_TSS_CNT(edev))
+#define QEDE_NDEV_TXQ_ID_TO_FP_ID(edev, idx)	((edev)->fp_num_rx + \
+						 ((idx) % QEDE_TSS_COUNT(edev)))
+#define QEDE_NDEV_TXQ_ID_TO_TXQ_COS(edev, idx)	((idx) / QEDE_TSS_COUNT(edev))
+#define QEDE_TXQ_TO_NDEV_TXQ_ID(edev, txq)	((QEDE_TSS_COUNT(edev) * \
+						 (txq)->cos) + (txq)->index)
+#define QEDE_NDEV_TXQ_ID_TO_TXQ(edev, idx)	\
+	(&((edev)->fp_array[QEDE_NDEV_TXQ_ID_TO_FP_ID(edev, idx)].txq \
+	[QEDE_NDEV_TXQ_ID_TO_TXQ_COS(edev, idx)]))
+#define QEDE_FP_TC0_TXQ(fp)	(&((fp)->txq[0]))
 
 	/* Regular Tx requires skb + metadata for release purpose,
 	 * while XDP requires the pages and the mapped address.
@@ -399,6 +408,8 @@ struct qede_tx_queue {
 
 	/* Slowpath; Should be kept in end [unless missing padding] */
 	void *handle;
+	u16 cos;
+	u16 ndev_txq_id;
 };
 
 #define BD_UNMAP_ADDR(bd)		HILO_U64(le32_to_cpu((bd)->addr.hi), \
@@ -541,5 +552,7 @@ void qede_update_rx_prod(struct qede_dev *edev, struct qede_rx_queue *rxq);
 #define QEDE_RX_HDR_SIZE		256
 #define QEDE_MAX_JUMBO_PACKET_SIZE	9600
 #define	for_each_queue(i) for (i = 0; i < edev->num_queues; i++)
+#define for_each_cos_in_txq(edev, var) \
+	for ((var) = 0; (var) < (edev)->dev_info.num_tc; (var)++)
 
 #endif /* _QEDE_H_ */

drivers/net/ethernet/qlogic/qede/qede_ethtool.c

@@ -222,7 +222,7 @@ static void qede_get_strings_stats_txq(struct qede_dev *edev,
 				QEDE_TXQ_XDP_TO_IDX(edev, txq),
 				qede_tqstats_arr[i].string);
 		else
-			sprintf(*buf, "%d: %s", txq->index,
+			sprintf(*buf, "%d_%d: %s", txq->index, txq->cos,
 				qede_tqstats_arr[i].string);
 		*buf += ETH_GSTRING_LEN;
 	}
@@ -262,8 +262,13 @@ static void qede_get_strings_stats(struct qede_dev *edev, u8 *buf)
 		if (fp->type & QEDE_FASTPATH_XDP)
 			qede_get_strings_stats_txq(edev, fp->xdp_tx, &buf);
 
-		if (fp->type & QEDE_FASTPATH_TX)
-			qede_get_strings_stats_txq(edev, fp->txq, &buf);
+		if (fp->type & QEDE_FASTPATH_TX) {
+			int cos;
+
+			for_each_cos_in_txq(edev, cos)
+				qede_get_strings_stats_txq(edev,
+							   &fp->txq[cos], &buf);
+		}
 	}
 
 	/* Account for non-queue statistics */
@@ -338,8 +343,12 @@ static void qede_get_ethtool_stats(struct net_device *dev,
 		if (fp->type & QEDE_FASTPATH_XDP)
 			qede_get_ethtool_stats_txq(fp->xdp_tx, &buf);
 
-		if (fp->type & QEDE_FASTPATH_TX)
-			qede_get_ethtool_stats_txq(fp->txq, &buf);
+		if (fp->type & QEDE_FASTPATH_TX) {
+			int cos;
+
+			for_each_cos_in_txq(edev, cos)
+				qede_get_ethtool_stats_txq(&fp->txq[cos], &buf);
+		}
 	}
 
 	for (i = 0; i < QEDE_NUM_STATS; i++) {
@@ -366,7 +375,8 @@ static int qede_get_sset_count(struct net_device *dev, int stringset)
 				num_stats--;
 
 		/* Account for the Regular Tx statistics */
-		num_stats += QEDE_TSS_COUNT(edev) * QEDE_NUM_TQSTATS;
+		num_stats += QEDE_TSS_COUNT(edev) * QEDE_NUM_TQSTATS *
+				edev->dev_info.num_tc;
 
 		/* Account for the Regular Rx statistics */
 		num_stats += QEDE_RSS_COUNT(edev) * QEDE_NUM_RQSTATS;
@@ -741,9 +751,17 @@ static int qede_get_coalesce(struct net_device *dev,
 		}
 
 		for_each_queue(i) {
+			struct qede_tx_queue *txq;
+
 			fp = &edev->fp_array[i];
+
+			/* All TX queues of given fastpath uses same
+			 * coalescing value, so no need to iterate over
+			 * all TCs, TC0 txq should suffice.
+			 */
 			if (fp->type & QEDE_FASTPATH_TX) {
-				tx_handle = fp->txq->handle;
+				txq = QEDE_FP_TC0_TXQ(fp);
+				tx_handle = txq->handle;
 				break;
 			}
 		}
@@ -801,9 +819,17 @@ static int qede_set_coalesce(struct net_device *dev,
 		}
 
 		if (edev->fp_array[i].type & QEDE_FASTPATH_TX) {
+			struct qede_tx_queue *txq;
+
+			/* All TX queues of given fastpath uses same
+			 * coalescing value, so no need to iterate over
+			 * all TCs, TC0 txq should suffice.
+			 */
+			txq = QEDE_FP_TC0_TXQ(fp);
+
 			rc = edev->ops->common->set_coalesce(edev->cdev,
 							     0, txc,
-							     fp->txq->handle);
+							     txq->handle);
 			if (rc) {
 				DP_INFO(edev,
 					"Set TX coalesce error, rc = %d\n", rc);
@@ -1385,8 +1411,10 @@ static int qede_selftest_transmit_traffic(struct qede_dev *edev,
 	u16 val;
 
 	for_each_queue(i) {
-		if (edev->fp_array[i].type & QEDE_FASTPATH_TX) {
-			txq = edev->fp_array[i].txq;
+		struct qede_fastpath *fp = &edev->fp_array[i];
+
+		if (fp->type & QEDE_FASTPATH_TX) {
+			txq = QEDE_FP_TC0_TXQ(fp);
 			break;
 		}
 	}

drivers/net/ethernet/qlogic/qede/qede_fp.c

@@ -408,12 +408,12 @@ static void qede_xdp_tx_int(struct qede_dev *edev, struct qede_tx_queue *txq)
 
 static int qede_tx_int(struct qede_dev *edev, struct qede_tx_queue *txq)
 {
+	unsigned int pkts_compl = 0, bytes_compl = 0;
 	struct netdev_queue *netdev_txq;
 	u16 hw_bd_cons;
-	unsigned int pkts_compl = 0, bytes_compl = 0;
 	int rc;
 
-	netdev_txq = netdev_get_tx_queue(edev->ndev, txq->index);
+	netdev_txq = netdev_get_tx_queue(edev->ndev, txq->ndev_txq_id);
 
 	hw_bd_cons = le16_to_cpu(*txq->hw_cons_ptr);
 	barrier();
@@ -1365,9 +1365,14 @@ static bool qede_poll_is_more_work(struct qede_fastpath *fp)
 		if (qede_txq_has_work(fp->xdp_tx))
 			return true;
 
-	if (likely(fp->type & QEDE_FASTPATH_TX))
-		if (qede_txq_has_work(fp->txq))
-			return true;
+	if (likely(fp->type & QEDE_FASTPATH_TX)) {
+		int cos;
+
+		for_each_cos_in_txq(fp->edev, cos) {
+			if (qede_txq_has_work(&fp->txq[cos]))
+				return true;
+		}
+	}
 
 	return false;
 }
@@ -1382,8 +1387,14 @@ int qede_poll(struct napi_struct *napi, int budget)
 	struct qede_dev *edev = fp->edev;
 	int rx_work_done = 0;
 
-	if (likely(fp->type & QEDE_FASTPATH_TX) && qede_txq_has_work(fp->txq))
-		qede_tx_int(edev, fp->txq);
+	if (likely(fp->type & QEDE_FASTPATH_TX)) {
+		int cos;
+
+		for_each_cos_in_txq(fp->edev, cos) {
+			if (qede_txq_has_work(&fp->txq[cos]))
+				qede_tx_int(edev, &fp->txq[cos]);
+		}
+	}
 
 	if ((fp->type & QEDE_FASTPATH_XDP) && qede_txq_has_work(fp->xdp_tx))
 		qede_xdp_tx_int(edev, fp->xdp_tx);
@@ -1444,8 +1455,8 @@ netdev_tx_t qede_start_xmit(struct sk_buff *skb, struct net_device *ndev)
 
 	/* Get tx-queue context and netdev index */
 	txq_index = skb_get_queue_mapping(skb);
-	WARN_ON(txq_index >= QEDE_TSS_COUNT(edev));
-	txq = edev->fp_array[edev->fp_num_rx + txq_index].txq;
+	WARN_ON(txq_index >= QEDE_TSS_COUNT(edev) * edev->dev_info.num_tc);
+	txq = QEDE_NDEV_TXQ_ID_TO_TXQ(edev, txq_index);
 	netdev_txq = netdev_get_tx_queue(ndev, txq_index);
 
 	WARN_ON(qed_chain_get_elem_left(&txq->tx_pbl) < (MAX_SKB_FRAGS + 1));

drivers/net/ethernet/qlogic/qede/qede_main.c

@@ -536,6 +536,43 @@ static int qede_ioctl(struct net_device *dev, struct ifreq *ifr, int cmd)
 	return 0;
 }
 
+int qede_setup_tc(struct net_device *ndev, u8 num_tc)
+{
+	struct qede_dev *edev = netdev_priv(ndev);
+	int cos, count, offset;
+
+	if (num_tc > edev->dev_info.num_tc)
+		return -EINVAL;
+
+	netdev_reset_tc(ndev);
+	netdev_set_num_tc(ndev, num_tc);
+
+	for_each_cos_in_txq(edev, cos) {
+		count = QEDE_TSS_COUNT(edev);
+		offset = cos * QEDE_TSS_COUNT(edev);
+		netdev_set_tc_queue(ndev, cos, count, offset);
+	}
+
+	return 0;
+}
+
+static int
+qede_setup_tc_offload(struct net_device *dev, enum tc_setup_type type,
+		      void *type_data)
+{
+	struct tc_mqprio_qopt *mqprio;
+
+	switch (type) {
+	case TC_SETUP_QDISC_MQPRIO:
+		mqprio = type_data;
+
+		mqprio->hw = TC_MQPRIO_HW_OFFLOAD_TCS;
+		return qede_setup_tc(dev, mqprio->num_tc);
+	default:
+		return -EOPNOTSUPP;
+	}
+}
+
 static const struct net_device_ops qede_netdev_ops = {
 	.ndo_open = qede_open,
 	.ndo_stop = qede_close,
@@ -568,6 +605,7 @@ static const struct net_device_ops qede_netdev_ops = {
 #ifdef CONFIG_RFS_ACCEL
 	.ndo_rx_flow_steer = qede_rx_flow_steer,
 #endif
+	.ndo_setup_tc = qede_setup_tc_offload,
 };
 
 static const struct net_device_ops qede_netdev_vf_ops = {
@@ -621,7 +659,8 @@ static struct qede_dev *qede_alloc_etherdev(struct qed_dev *cdev,
 	struct qede_dev *edev;
 
 	ndev = alloc_etherdev_mqs(sizeof(*edev),
-				  info->num_queues, info->num_queues);
+				  info->num_queues * info->num_tc,
+				  info->num_queues);
 	if (!ndev) {
 		pr_err("etherdev allocation failed\n");
 		return NULL;
@@ -830,7 +869,8 @@ static int qede_alloc_fp_array(struct qede_dev *edev)
 		}
 
 		if (fp->type & QEDE_FASTPATH_TX) {
-			fp->txq = kzalloc(sizeof(*fp->txq), GFP_KERNEL);
+			fp->txq = kcalloc(edev->dev_info.num_tc,
+					  sizeof(*fp->txq), GFP_KERNEL);
 			if (!fp->txq)
 				goto err;
 		}
@@ -879,10 +919,15 @@ static void qede_sp_task(struct work_struct *work)
 static void qede_update_pf_params(struct qed_dev *cdev)
 {
 	struct qed_pf_params pf_params;
+	u16 num_cons;
 
 	/* 64 rx + 64 tx + 64 XDP */
 	memset(&pf_params, 0, sizeof(struct qed_pf_params));
-	pf_params.eth_pf_params.num_cons = (MAX_SB_PER_PF_MIMD - 1) * 3;
+
+	/* 1 rx + 1 xdp + max tx cos */
+	num_cons = QED_MIN_L2_CONS;
+
+	pf_params.eth_pf_params.num_cons = (MAX_SB_PER_PF_MIMD - 1) * num_cons;
 
 	/* Same for VFs - make sure they'll have sufficient connections
 	 * to support XDP Tx queues.
@@ -1363,8 +1408,12 @@ static void qede_free_mem_fp(struct qede_dev *edev, struct qede_fastpath *fp)
 	if (fp->type & QEDE_FASTPATH_XDP)
 		qede_free_mem_txq(edev, fp->xdp_tx);
 
-	if (fp->type & QEDE_FASTPATH_TX)
-		qede_free_mem_txq(edev, fp->txq);
+	if (fp->type & QEDE_FASTPATH_TX) {
+		int cos;
+
+		for_each_cos_in_txq(edev, cos)
+			qede_free_mem_txq(edev, &fp->txq[cos]);
+	}
 }
 
 /* This function allocates all memory needed for a single fp (i.e. an entity
@@ -1391,9 +1440,13 @@ static int qede_alloc_mem_fp(struct qede_dev *edev, struct qede_fastpath *fp)
 	}
 
 	if (fp->type & QEDE_FASTPATH_TX) {
-		rc = qede_alloc_mem_txq(edev, fp->txq);
-		if (rc)
-			goto out;
+		int cos;
+
+		for_each_cos_in_txq(edev, cos) {
+			rc = qede_alloc_mem_txq(edev, &fp->txq[cos]);
+			if (rc)
+				goto out;
+		}
 	}
 
 out:
@@ -1466,10 +1519,23 @@ static void qede_init_fp(struct qede_dev *edev)
 		}
 
 		if (fp->type & QEDE_FASTPATH_TX) {
-			fp->txq->index = txq_index++;
-			if (edev->dev_info.is_legacy)
-				fp->txq->is_legacy = 1;
-			fp->txq->dev = &edev->pdev->dev;
+			int cos;
+
+			for_each_cos_in_txq(edev, cos) {
+				struct qede_tx_queue *txq = &fp->txq[cos];
+				u16 ndev_tx_id;
+
+				txq->cos = cos;
+				txq->index = txq_index;
+				ndev_tx_id = QEDE_TXQ_TO_NDEV_TXQ_ID(edev, txq);
+				txq->ndev_txq_id = ndev_tx_id;
+
+				if (edev->dev_info.is_legacy)
+					txq->is_legacy = 1;
+				txq->dev = &edev->pdev->dev;
+			}
+
+			txq_index++;
 		}
 
 		snprintf(fp->name, sizeof(fp->name), "%s-fp-%d",
@@ -1483,7 +1549,9 @@ static int qede_set_real_num_queues(struct qede_dev *edev)
 {
 	int rc = 0;
 
-	rc = netif_set_real_num_tx_queues(edev->ndev, QEDE_TSS_COUNT(edev));
+	rc = netif_set_real_num_tx_queues(edev->ndev,
+					  QEDE_TSS_COUNT(edev) *
+					  edev->dev_info.num_tc);
 	if (rc) {
 		DP_NOTICE(edev, "Failed to set real number of Tx queues\n");
 		return rc;
@@ -1685,9 +1753,13 @@ static int qede_stop_queues(struct qede_dev *edev)
 		fp = &edev->fp_array[i];
 
 		if (fp->type & QEDE_FASTPATH_TX) {
-			rc = qede_drain_txq(edev, fp->txq, true);
-			if (rc)
-				return rc;
+			int cos;
+
+			for_each_cos_in_txq(edev, cos) {
+				rc = qede_drain_txq(edev, &fp->txq[cos], true);
+				if (rc)
+					return rc;
+			}
 		}
 
 		if (fp->type & QEDE_FASTPATH_XDP) {
@@ -1703,9 +1775,13 @@ static int qede_stop_queues(struct qede_dev *edev)
 
 		/* Stop the Tx Queue(s) */
 		if (fp->type & QEDE_FASTPATH_TX) {
-			rc = qede_stop_txq(edev, fp->txq, i);
-			if (rc)
-				return rc;
+			int cos;
+
+			for_each_cos_in_txq(edev, cos) {
+				rc = qede_stop_txq(edev, &fp->txq[cos], i);
+				if (rc)
+					return rc;
+			}
 		}
 
 		/* Stop the Rx Queue */
@@ -1758,6 +1834,7 @@ static int qede_start_txq(struct qede_dev *edev,
 
 	params.p_sb = fp->sb_info;
 	params.sb_idx = sb_idx;
+	params.tc = txq->cos;
 
 	rc = edev->ops->q_tx_start(edev->cdev, rss_id, &params, phys_table,
 				   page_cnt, &ret_params);
@@ -1877,9 +1954,14 @@ static int qede_start_queues(struct qede_dev *edev, bool clear_stats)
 		}
 
 		if (fp->type & QEDE_FASTPATH_TX) {
-			rc = qede_start_txq(edev, fp, fp->txq, i, TX_PI(0));
-			if (rc)
-				goto out;
+			int cos;
+
+			for_each_cos_in_txq(edev, cos) {
+				rc = qede_start_txq(edev, fp, &fp->txq[cos], i,
+						    TX_PI(cos));
+				if (rc)
+					goto out;
+			}
 		}
 	}
 
@@ -1973,6 +2055,7 @@ static int qede_load(struct qede_dev *edev, enum qede_load_mode mode,
 		     bool is_locked)
 {
 	struct qed_link_params link_params;
+	u8 num_tc;
 	int rc;
 
 	DP_INFO(edev, "Starting qede load\n");
@@ -2019,6 +2102,10 @@ static int qede_load(struct qede_dev *edev, enum qede_load_mode mode,
 		goto err4;
 	DP_INFO(edev, "Start VPORT, RXQ and TXQ succeeded\n");
 
+	num_tc = netdev_get_num_tc(edev->ndev);
+	num_tc = num_tc ? num_tc : edev->dev_info.num_tc;
+	qede_setup_tc(edev->ndev, num_tc);
+
 	/* Program un-configured VLANs */
 	qede_configure_vlan_filters(edev);
 
@@ -2143,7 +2230,7 @@ static bool qede_is_txq_full(struct qede_dev *edev, struct qede_tx_queue *txq)
 {
 	struct netdev_queue *netdev_txq;
 
-	netdev_txq = netdev_get_tx_queue(edev->ndev, txq->index);
+	netdev_txq = netdev_get_tx_queue(edev->ndev, txq->ndev_txq_id);
 	if (netif_xmit_stopped(netdev_txq))
 		return true;
 
@@ -2208,9 +2295,11 @@ static void qede_get_eth_tlv_data(void *dev, void *data)
 	for_each_queue(i) {
 		fp = &edev->fp_array[i];
 		if (fp->type & QEDE_FASTPATH_TX) {
-			if (fp->txq->sw_tx_cons != fp->txq->sw_tx_prod)
+			struct qede_tx_queue *txq = QEDE_FP_TC0_TXQ(fp);
+
+			if (txq->sw_tx_cons != txq->sw_tx_prod)
 				etlv->txqs_empty = false;
-			if (qede_is_txq_full(edev, fp->txq))
+			if (qede_is_txq_full(edev, txq))
 				etlv->num_txqs_full++;
 		}
 		if (fp->type & QEDE_FASTPATH_RX) {

include/linux/qed/qed_eth_if.h

@@ -39,6 +39,10 @@
 #include <linux/qed/qed_if.h>
 #include <linux/qed/qed_iov_if.h>
 
+/* 64 max queues * (1 rx + 4 tx-cos + 1 xdp) */
+#define QED_MIN_L2_CONS (2 + NUM_PHYS_TCS_4PORT_K2)
+#define QED_MAX_L2_CONS (64 * (QED_MIN_L2_CONS))
+
 struct qed_queue_start_common_params {
 	/* Should always be relative to entity sending this. */
 	u8 vport_id;
@@ -49,6 +53,8 @@ struct qed_queue_start_common_params {
 
 	struct qed_sb_info *p_sb;
 	u8 sb_idx;
+
+	u8 tc;
 };
 
 struct qed_rxq_start_ret_params {