4 files changed, 91 insertions, 55 deletions

diff --git a/drivers/net/ethernet/qlogic/qede/Makefile b/drivers/net/ethernet/qlogic/qede/Makefile
index 28dc58919c85..048a230c3ce0 100644
--- a/drivers/net/ethernet/qlogic/qede/Makefile
+++ b/drivers/net/ethernet/qlogic/qede/Makefile
@@ -2,4 +2,4 @@ obj-$(CONFIG_QEDE) := qede.o
 
 qede-y := qede_main.o qede_ethtool.o
 qede-$(CONFIG_DCB) += qede_dcbnl.o
-qede-$(CONFIG_INFINIBAND_QEDR) += qede_roce.o
+qede-$(CONFIG_QED_RDMA) += qede_roce.o
diff --git a/drivers/net/ethernet/qlogic/qede/qede.h b/drivers/net/ethernet/qlogic/qede/qede.h
index 28c0e9f42c9e..974689a13337 100644
--- a/drivers/net/ethernet/qlogic/qede/qede.h
+++ b/drivers/net/ethernet/qlogic/qede/qede.h
@@ -348,12 +348,13 @@ bool qede_has_rx_work(struct qede_rx_queue *rxq);
 int qede_txq_has_work(struct qede_tx_queue *txq);
 void qede_recycle_rx_bd_ring(struct qede_rx_queue *rxq, struct qede_dev *edev,
 			     u8 count);
+void qede_update_rx_prod(struct qede_dev *edev, struct qede_rx_queue *rxq);
 
 #define RX_RING_SIZE_POW	13
 #define RX_RING_SIZE		((u16)BIT(RX_RING_SIZE_POW))
 #define NUM_RX_BDS_MAX		(RX_RING_SIZE - 1)
 #define NUM_RX_BDS_MIN		128
-#define NUM_RX_BDS_DEF		NUM_RX_BDS_MAX
+#define NUM_RX_BDS_DEF		((u16)BIT(10) - 1)
 
 #define TX_RING_SIZE_POW	13
 #define TX_RING_SIZE		((u16)BIT(TX_RING_SIZE_POW))
diff --git a/drivers/net/ethernet/qlogic/qede/qede_ethtool.c b/drivers/net/ethernet/qlogic/qede/qede_ethtool.c
index 25a9b293ee8f..7567cc464b88 100644
--- a/drivers/net/ethernet/qlogic/qede/qede_ethtool.c
+++ b/drivers/net/ethernet/qlogic/qede/qede_ethtool.c
@@ -175,16 +175,23 @@ static void qede_get_strings_stats(struct qede_dev *edev, u8 *buf)
 	for (i = 0, k = 0; i < QEDE_QUEUE_CNT(edev); i++) {
 		int tc;
 
-		for (j = 0; j < QEDE_NUM_RQSTATS; j++)
-			sprintf(buf + (k + j) * ETH_GSTRING_LEN,
-				"%d:   %s", i, qede_rqstats_arr[j].string);
-		k += QEDE_NUM_RQSTATS;
-		for (tc = 0; tc < edev->num_tc; tc++) {
-			for (j = 0; j < QEDE_NUM_TQSTATS; j++)
+		if (edev->fp_array[i].type & QEDE_FASTPATH_RX) {
+			for (j = 0; j < QEDE_NUM_RQSTATS; j++)
 				sprintf(buf + (k + j) * ETH_GSTRING_LEN,
-					"%d.%d: %s", i, tc,
-					qede_tqstats_arr[j].string);
-			k += QEDE_NUM_TQSTATS;
+					"%d:   %s", i,
+					qede_rqstats_arr[j].string);
+			k += QEDE_NUM_RQSTATS;
+		}
+
+		if (edev->fp_array[i].type & QEDE_FASTPATH_TX) {
+			for (tc = 0; tc < edev->num_tc; tc++) {
+				for (j = 0; j < QEDE_NUM_TQSTATS; j++)
+					sprintf(buf + (k + j) *
+						ETH_GSTRING_LEN,
+						"%d.%d: %s", i, tc,
+						qede_tqstats_arr[j].string);
+				k += QEDE_NUM_TQSTATS;
+			}
 		}
 	}
 
@@ -756,6 +763,8 @@ static void qede_get_channels(struct net_device *dev,
 	struct qede_dev *edev = netdev_priv(dev);
 
 	channels->max_combined = QEDE_MAX_RSS_CNT(edev);
+	channels->max_rx = QEDE_MAX_RSS_CNT(edev);
+	channels->max_tx = QEDE_MAX_RSS_CNT(edev);
 	channels->combined_count = QEDE_QUEUE_CNT(edev) - edev->fp_num_tx -
 					edev->fp_num_rx;
 	channels->tx_count = edev->fp_num_tx;
@@ -820,6 +829,13 @@ static int qede_set_channels(struct net_device *dev,
 	edev->req_queues = count;
 	edev->req_num_tx = channels->tx_count;
 	edev->req_num_rx = channels->rx_count;
+	/* Reset the indirection table if rx queue count is updated */
+	if ((edev->req_queues - edev->req_num_tx) != QEDE_RSS_COUNT(edev)) {
+		edev->rss_params_inited &= ~QEDE_RSS_INDIR_INITED;
+		memset(&edev->rss_params.rss_ind_table, 0,
+		       sizeof(edev->rss_params.rss_ind_table));
+	}
+
 	if (netif_running(dev))
 		qede_reload(edev, NULL, NULL);
 
@@ -1053,6 +1069,12 @@ static int qede_set_rxfh(struct net_device *dev, const u32 *indir,
 	struct qede_dev *edev = netdev_priv(dev);
 	int i;
 
+	if (edev->dev_info.common.num_hwfns > 1) {
+		DP_INFO(edev,
+			"RSS configuration is not supported for 100G devices\n");
+		return -EOPNOTSUPP;
+	}
+
 	if (hfunc != ETH_RSS_HASH_NO_CHANGE && hfunc != ETH_RSS_HASH_TOP)
 		return -EOPNOTSUPP;
 
@@ -1184,8 +1206,8 @@ static int qede_selftest_transmit_traffic(struct qede_dev *edev,
 	}
 
 	first_bd = (struct eth_tx_1st_bd *)qed_chain_consume(&txq->tx_pbl);
-	dma_unmap_page(&edev->pdev->dev, BD_UNMAP_ADDR(first_bd),
-		       BD_UNMAP_LEN(first_bd), DMA_TO_DEVICE);
+	dma_unmap_single(&edev->pdev->dev, BD_UNMAP_ADDR(first_bd),
+			 BD_UNMAP_LEN(first_bd), DMA_TO_DEVICE);
 	txq->sw_tx_cons++;
 	txq->sw_tx_ring[idx].skb = NULL;
 
@@ -1199,8 +1221,8 @@ static int qede_selftest_receive_traffic(struct qede_dev *edev)
 	struct qede_rx_queue *rxq = NULL;
 	struct sw_rx_data *sw_rx_data;
 	union eth_rx_cqe *cqe;
+	int i, rc = 0;
 	u8 *data_ptr;
-	int i;
 
 	for_each_queue(i) {
 		if (edev->fp_array[i].type & QEDE_FASTPATH_RX) {
@@ -1219,46 +1241,60 @@ static int qede_selftest_receive_traffic(struct qede_dev *edev)
 	 * queue and that the loopback traffic is not IP.
 	 */
 	for (i = 0; i < QEDE_SELFTEST_POLL_COUNT; i++) {
-		if (qede_has_rx_work(rxq))
+		if (!qede_has_rx_work(rxq)) {
+			usleep_range(100, 200);
+			continue;
+		}
+
+		hw_comp_cons = le16_to_cpu(*rxq->hw_cons_ptr);
+		sw_comp_cons = qed_chain_get_cons_idx(&rxq->rx_comp_ring);
+
+		/* Memory barrier to prevent the CPU from doing speculative
+		 * reads of CQE/BD before reading hw_comp_cons. If the CQE is
+		 * read before it is written by FW, then FW writes CQE and SB,
+		 * and then the CPU reads the hw_comp_cons, it will use an old
+		 * CQE.
+		 */
+		rmb();
+
+		/* Get the CQE from the completion ring */
+		cqe = (union eth_rx_cqe *)qed_chain_consume(&rxq->rx_comp_ring);
+
+		/* Get the data from the SW ring */
+		sw_rx_index = rxq->sw_rx_cons & NUM_RX_BDS_MAX;
+		sw_rx_data = &rxq->sw_rx_ring[sw_rx_index];
+		fp_cqe = &cqe->fast_path_regular;
+		len =  le16_to_cpu(fp_cqe->len_on_first_bd);
+		data_ptr = (u8 *)(page_address(sw_rx_data->data) +
+				  fp_cqe->placement_offset +
+				  sw_rx_data->page_offset);
+		if (ether_addr_equal(data_ptr,  edev->ndev->dev_addr) &&
+		    ether_addr_equal(data_ptr + ETH_ALEN,
+				     edev->ndev->dev_addr)) {
+			for (i = ETH_HLEN; i < len; i++)
+				if (data_ptr[i] != (unsigned char)(i & 0xff)) {
+					rc = -1;
+					break;
+				}
+
+			qede_recycle_rx_bd_ring(rxq, edev, 1);
+			qed_chain_recycle_consumed(&rxq->rx_comp_ring);
 			break;
-		usleep_range(100, 200);
+		}
+
+		DP_INFO(edev, "Not the transmitted packet\n");
+		qede_recycle_rx_bd_ring(rxq, edev, 1);
+		qed_chain_recycle_consumed(&rxq->rx_comp_ring);
 	}
 
-	if (!qede_has_rx_work(rxq)) {
+	if (i == QEDE_SELFTEST_POLL_COUNT) {
 		DP_NOTICE(edev, "Failed to receive the traffic\n");
 		return -1;
 	}
 
-	hw_comp_cons = le16_to_cpu(*rxq->hw_cons_ptr);
-	sw_comp_cons = qed_chain_get_cons_idx(&rxq->rx_comp_ring);
+	qede_update_rx_prod(edev, rxq);
 
-	/* Memory barrier to prevent the CPU from doing speculative reads of CQE
-	 * / BD before reading hw_comp_cons. If the CQE is read before it is
-	 * written by FW, then FW writes CQE and SB, and then the CPU reads the
-	 * hw_comp_cons, it will use an old CQE.
-	 */
-	rmb();
-
-	/* Get the CQE from the completion ring */
-	cqe = (union eth_rx_cqe *)qed_chain_consume(&rxq->rx_comp_ring);
-
-	/* Get the data from the SW ring */
-	sw_rx_index = rxq->sw_rx_cons & NUM_RX_BDS_MAX;
-	sw_rx_data = &rxq->sw_rx_ring[sw_rx_index];
-	fp_cqe = &cqe->fast_path_regular;
-	len =  le16_to_cpu(fp_cqe->len_on_first_bd);
-	data_ptr = (u8 *)(page_address(sw_rx_data->data) +
-		     fp_cqe->placement_offset + sw_rx_data->page_offset);
-	for (i = ETH_HLEN; i < len; i++)
-		if (data_ptr[i] != (unsigned char)(i & 0xff)) {
-			DP_NOTICE(edev, "Loopback test failed\n");
-			qede_recycle_rx_bd_ring(rxq, edev, 1);
-			return -1;
-		}
-
-	qede_recycle_rx_bd_ring(rxq, edev, 1);
-
-	return 0;
+	return rc;
 }
 
 static int qede_selftest_run_loopback(struct qede_dev *edev, u32 loopback_mode)
diff --git a/drivers/net/ethernet/qlogic/qede/qede_main.c b/drivers/net/ethernet/qlogic/qede/qede_main.c
index 343038ca047d..85f46dbecd5b 100644
--- a/drivers/net/ethernet/qlogic/qede/qede_main.c
+++ b/drivers/net/ethernet/qlogic/qede/qede_main.c
@@ -313,8 +313,8 @@ static int qede_free_tx_pkt(struct qede_dev *edev,
 		split_bd_len = BD_UNMAP_LEN(split);
 		bds_consumed++;
 	}
-	dma_unmap_page(&edev->pdev->dev, BD_UNMAP_ADDR(first_bd),
-		       BD_UNMAP_LEN(first_bd) + split_bd_len, DMA_TO_DEVICE);
+	dma_unmap_single(&edev->pdev->dev, BD_UNMAP_ADDR(first_bd),
+			 BD_UNMAP_LEN(first_bd) + split_bd_len, DMA_TO_DEVICE);
 
 	/* Unmap the data of the skb frags */
 	for (i = 0; i < skb_shinfo(skb)->nr_frags; i++, bds_consumed++) {
@@ -359,8 +359,8 @@ static void qede_free_failed_tx_pkt(struct qede_dev *edev,
 		nbd--;
 	}
 
-	dma_unmap_page(&edev->pdev->dev, BD_UNMAP_ADDR(first_bd),
-		       BD_UNMAP_LEN(first_bd) + split_bd_len, DMA_TO_DEVICE);
+	dma_unmap_single(&edev->pdev->dev, BD_UNMAP_ADDR(first_bd),
+			 BD_UNMAP_LEN(first_bd) + split_bd_len, DMA_TO_DEVICE);
 
 	/* Unmap the data of the skb frags */
 	for (i = 0; i < nbd; i++) {
@@ -943,8 +943,7 @@ static inline int qede_realloc_rx_buffer(struct qede_dev *edev,
 	return 0;
 }
 
-static inline void qede_update_rx_prod(struct qede_dev *edev,
-				       struct qede_rx_queue *rxq)
+void qede_update_rx_prod(struct qede_dev *edev, struct qede_rx_queue *rxq)
 {
 	u16 bd_prod = qed_chain_get_prod_idx(&rxq->rx_bd_ring);
 	u16 cqe_prod = qed_chain_get_prod_idx(&rxq->rx_comp_ring);
@@ -2840,7 +2839,7 @@ static int qede_alloc_sge_mem(struct qede_dev *edev, struct qede_rx_queue *rxq)
 		}
 
 		mapping = dma_map_page(&edev->pdev->dev, replace_buf->data, 0,
-				       rxq->rx_buf_size, DMA_FROM_DEVICE);
+				       PAGE_SIZE, DMA_FROM_DEVICE);
 		if (unlikely(dma_mapping_error(&edev->pdev->dev, mapping))) {
 			DP_NOTICE(edev,
 				  "Failed to map TPA replacement buffer\n");
@@ -2941,7 +2940,7 @@ static int qede_alloc_mem_txq(struct qede_dev *edev, struct qede_tx_queue *txq)
 	txq->num_tx_buffers = edev->q_num_tx_buffers;
 
 	/* Allocate the parallel driver ring for Tx buffers */
-	size = sizeof(*txq->sw_tx_ring) * NUM_TX_BDS_MAX;
+	size = sizeof(*txq->sw_tx_ring) * TX_RING_SIZE;
 	txq->sw_tx_ring = kzalloc(size, GFP_KERNEL);
 	if (!txq->sw_tx_ring) {
 		DP_NOTICE(edev, "Tx buffers ring allocation failed\n");
@@ -2952,7 +2951,7 @@ static int qede_alloc_mem_txq(struct qede_dev *edev, struct qede_tx_queue *txq)
 					    QED_CHAIN_USE_TO_CONSUME_PRODUCE,
 					    QED_CHAIN_MODE_PBL,
 					    QED_CHAIN_CNT_TYPE_U16,
-					    NUM_TX_BDS_MAX,
+					    TX_RING_SIZE,
 					    sizeof(*p_virt), &txq->tx_pbl);
 	if (rc)
 		goto err;