[dpdk-dev] [PATCH v2] net/axgbe: add support for Scattered Rx

[ssebasti@amd.com]

From: Selwin Sebastian <selwin.sebastian@amd.com>

Enable scattered rx support and add jumbo packet receive capability

Signed-off-by: Selwin Sebastian <selwin.sebastian@amd.com>
---
 doc/guides/nics/features/axgbe.ini |   1 +
 drivers/net/axgbe/axgbe_common.h   |   2 +
 drivers/net/axgbe/axgbe_ethdev.c   |  17 +++-
 drivers/net/axgbe/axgbe_rxtx.c     | 146 +++++++++++++++++++++++++++++
 drivers/net/axgbe/axgbe_rxtx.h     |   2 +
 5 files changed, 167 insertions(+), 1 deletion(-)

diff --git a/doc/guides/nics/features/axgbe.ini b/doc/guides/nics/features/axgbe.ini
index ab4da559f..0becaa097 100644
--- a/doc/guides/nics/features/axgbe.ini
+++ b/doc/guides/nics/features/axgbe.ini
@@ -7,6 +7,7 @@
 Speed capabilities   = Y
 Link status          = Y
 Jumbo frame          = Y
+Scattered Rx         = Y
 Promiscuous mode     = Y
 Allmulticast mode    = Y
 RSS hash             = Y
diff --git a/drivers/net/axgbe/axgbe_common.h b/drivers/net/axgbe/axgbe_common.h
index fdb037dd5..fbd46150c 100644
--- a/drivers/net/axgbe/axgbe_common.h
+++ b/drivers/net/axgbe/axgbe_common.h
@@ -1135,6 +1135,8 @@
 #define RX_NORMAL_DESC3_PL_WIDTH		14
 #define RX_NORMAL_DESC3_RSV_INDEX		26
 #define RX_NORMAL_DESC3_RSV_WIDTH		1
+#define RX_NORMAL_DESC3_LD_INDEX		28
+#define RX_NORMAL_DESC3_LD_WIDTH		1
 
 #define RX_DESC3_L34T_IPV4_TCP			1
 #define RX_DESC3_L34T_IPV4_UDP			2
diff --git a/drivers/net/axgbe/axgbe_ethdev.c b/drivers/net/axgbe/axgbe_ethdev.c
index d0b6f091f..542003bda 100644
--- a/drivers/net/axgbe/axgbe_ethdev.c
+++ b/drivers/net/axgbe/axgbe_ethdev.c
@@ -249,6 +249,10 @@ axgbe_dev_start(struct rte_eth_dev *dev)
 {
 	struct axgbe_port *pdata = dev->data->dev_private;
 	int ret;
+	struct rte_eth_dev_data *dev_data = dev->data;
+	uint16_t max_pkt_len = dev_data->dev_conf.rxmode.max_rx_pkt_len;
+
+	dev->dev_ops = &axgbe_eth_dev_ops;
 
 	PMD_INIT_FUNC_TRACE();
 
@@ -279,6 +283,16 @@ axgbe_dev_start(struct rte_eth_dev *dev)
 
 	axgbe_clear_bit(AXGBE_STOPPED, &pdata->dev_state);
 	axgbe_clear_bit(AXGBE_DOWN, &pdata->dev_state);
+	if ((dev_data->dev_conf.rxmode.offloads & DEV_RX_OFFLOAD_SCATTER) ||
+				max_pkt_len > pdata->rx_buf_size)
+		dev_data->scattered_rx = 1;
+
+	/*  Scatter Rx handling */
+	if (dev_data->scattered_rx)
+		dev->rx_pkt_burst = &eth_axgbe_recv_scattered_pkts;
+	else
+		dev->rx_pkt_burst = &axgbe_recv_pkts;
+
 	return 0;
 }
 
@@ -789,6 +803,8 @@ axgbe_dev_info_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
 		DEV_RX_OFFLOAD_IPV4_CKSUM |
 		DEV_RX_OFFLOAD_UDP_CKSUM  |
 		DEV_RX_OFFLOAD_TCP_CKSUM  |
+		DEV_RX_OFFLOAD_JUMBO_FRAME	|
+		DEV_RX_OFFLOAD_SCATTER	  |
 		DEV_RX_OFFLOAD_KEEP_CRC;
 
 	dev_info->tx_offload_capa =
@@ -1020,7 +1036,6 @@ eth_axgbe_dev_init(struct rte_eth_dev *eth_dev)
 	int ret;
 
 	eth_dev->dev_ops = &axgbe_eth_dev_ops;
-	eth_dev->rx_pkt_burst = &axgbe_recv_pkts;
 
 	/*
 	 * For secondary processes, we don't initialise any further as primary
diff --git a/drivers/net/axgbe/axgbe_rxtx.c b/drivers/net/axgbe/axgbe_rxtx.c
index 96055c25b..57e2bbb34 100644
--- a/drivers/net/axgbe/axgbe_rxtx.c
+++ b/drivers/net/axgbe/axgbe_rxtx.c
@@ -307,6 +307,152 @@ axgbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
 	return nb_rx;
 }
 
+
+uint16_t eth_axgbe_recv_scattered_pkts(void *rx_queue,
+		struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+	PMD_INIT_FUNC_TRACE();
+	uint16_t nb_rx = 0;
+	struct axgbe_rx_queue *rxq = rx_queue;
+	volatile union axgbe_rx_desc *desc;
+
+	uint64_t old_dirty = rxq->dirty;
+	struct rte_mbuf *first_seg = NULL;
+	struct rte_mbuf *mbuf, *tmbuf;
+	unsigned int err;
+	uint32_t error_status;
+	uint16_t idx, pidx, data_len = 0, pkt_len = 0;
+
+	idx = AXGBE_GET_DESC_IDX(rxq, rxq->cur);
+	while (nb_rx < nb_pkts) {
+		bool eop = 0;
+next_desc:
+		if (unlikely(idx == rxq->nb_desc))
+			idx = 0;
+
+		desc = &rxq->desc[idx];
+
+		if (AXGMAC_GET_BITS_LE(desc->write.desc3, RX_NORMAL_DESC3, OWN))
+			break;
+
+		tmbuf = rte_mbuf_raw_alloc(rxq->mb_pool);
+		if (unlikely(!tmbuf)) {
+			PMD_DRV_LOG(ERR, "RX mbuf alloc failed port_id = %u"
+				    " queue_id = %u\n",
+				    (unsigned int)rxq->port_id,
+				    (unsigned int)rxq->queue_id);
+			rte_eth_devices[rxq->port_id].data->rx_mbuf_alloc_failed++;
+			break;
+		}
+
+		pidx = idx + 1;
+		if (unlikely(pidx == rxq->nb_desc))
+			pidx = 0;
+
+		rte_prefetch0(rxq->sw_ring[pidx]);
+		if ((pidx & 0x3) == 0) {
+			rte_prefetch0(&rxq->desc[pidx]);
+			rte_prefetch0(&rxq->sw_ring[pidx]);
+		}
+
+		mbuf = rxq->sw_ring[idx];
+		/* Check for any errors and free mbuf*/
+		err = AXGMAC_GET_BITS_LE(desc->write.desc3,
+					 RX_NORMAL_DESC3, ES);
+		error_status = 0;
+		if (unlikely(err)) {
+			error_status = desc->write.desc3 & AXGBE_ERR_STATUS;
+			if ((error_status != AXGBE_L3_CSUM_ERR)
+					&& (error_status != AXGBE_L4_CSUM_ERR)) {
+				rxq->errors++;
+				rte_pktmbuf_free(mbuf);
+				goto err_set;
+			}
+		}
+		rte_prefetch1(rte_pktmbuf_mtod(mbuf, void *));
+
+		if (!AXGMAC_GET_BITS_LE(desc->write.desc3,
+					RX_NORMAL_DESC3, LD)) {
+			eop = 0;
+			pkt_len = rxq->buf_size;
+			data_len = pkt_len;
+		} else {
+			eop = 1;
+			pkt_len = AXGMAC_GET_BITS_LE(desc->write.desc3,
+					RX_NORMAL_DESC3, PL);
+			data_len = pkt_len - rxq->crc_len;
+		}
+
+		if (first_seg != NULL) {
+			if (rte_pktmbuf_chain(first_seg, mbuf) != 0)
+				rte_mempool_put(rxq->mb_pool,
+						first_seg);
+		} else {
+			first_seg = mbuf;
+		}
+
+		/* Get the RSS hash */
+		if (AXGMAC_GET_BITS_LE(desc->write.desc3, RX_NORMAL_DESC3, RSV))
+			mbuf->hash.rss = rte_le_to_cpu_32(desc->write.desc1);
+
+		/* Mbuf populate */
+		mbuf->data_off = RTE_PKTMBUF_HEADROOM;
+		mbuf->data_len = data_len;
+
+err_set:
+		rxq->cur++;
+		rxq->sw_ring[idx++] = tmbuf;
+		desc->read.baddr =
+			rte_cpu_to_le_64(rte_mbuf_data_iova_default(tmbuf));
+		memset((void *)(&desc->read.desc2), 0, 8);
+		AXGMAC_SET_BITS_LE(desc->read.desc3, RX_NORMAL_DESC3, OWN, 1);
+		rxq->dirty++;
+
+		if (!eop) {
+			rte_pktmbuf_free(mbuf);
+			goto next_desc;
+		}
+
+		first_seg->pkt_len = pkt_len;
+		rxq->bytes += pkt_len;
+		mbuf->next = NULL;
+
+		first_seg->port = rxq->port_id;
+		if (rxq->pdata->rx_csum_enable) {
+			mbuf->ol_flags = 0;
+			mbuf->ol_flags |= PKT_RX_IP_CKSUM_GOOD;
+			mbuf->ol_flags |= PKT_RX_L4_CKSUM_GOOD;
+			if (unlikely(error_status == AXGBE_L3_CSUM_ERR)) {
+				mbuf->ol_flags &= ~PKT_RX_IP_CKSUM_GOOD;
+				mbuf->ol_flags |= PKT_RX_IP_CKSUM_BAD;
+				mbuf->ol_flags &= ~PKT_RX_L4_CKSUM_GOOD;
+				mbuf->ol_flags |= PKT_RX_L4_CKSUM_UNKNOWN;
+			} else if (unlikely(error_status
+						== AXGBE_L4_CSUM_ERR)) {
+				mbuf->ol_flags &= ~PKT_RX_L4_CKSUM_GOOD;
+				mbuf->ol_flags |= PKT_RX_L4_CKSUM_BAD;
+			}
+		}
+
+		rx_pkts[nb_rx++] = first_seg;
+
+		 /* Setup receipt context for a new packet.*/
+		first_seg = NULL;
+	}
+
+	/* Save receive context.*/
+	rxq->pkts += nb_rx;
+
+	if (rxq->dirty != old_dirty) {
+		rte_wmb();
+		idx = AXGBE_GET_DESC_IDX(rxq, rxq->dirty - 1);
+		AXGMAC_DMA_IOWRITE(rxq, DMA_CH_RDTR_LO,
+				   low32_value(rxq->ring_phys_addr +
+				   (idx * sizeof(union axgbe_rx_desc))));
+	}
+	return nb_rx;
+}
+
 /* Tx Apis */
 static void axgbe_tx_queue_release(struct axgbe_tx_queue *tx_queue)
 {
diff --git a/drivers/net/axgbe/axgbe_rxtx.h b/drivers/net/axgbe/axgbe_rxtx.h
index a21537df9..f6796b09b 100644
--- a/drivers/net/axgbe/axgbe_rxtx.h
+++ b/drivers/net/axgbe/axgbe_rxtx.h
@@ -179,6 +179,8 @@ int axgbe_dev_rx_queue_start(struct rte_eth_dev *dev, uint16_t rx_queue_id);
 int axgbe_dev_rx_queue_stop(struct rte_eth_dev *dev, uint16_t rx_queue_id);
 uint16_t axgbe_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
 			 uint16_t nb_pkts);
+uint16_t eth_axgbe_recv_scattered_pkts(void *rx_queue,
+		struct rte_mbuf **rx_pkts, uint16_t nb_pkts);
 uint16_t axgbe_recv_pkts_threshold_refresh(void *rx_queue,
 					   struct rte_mbuf **rx_pkts,
 					   uint16_t nb_pkts);
-- 
2.17.1