[dpdk-dev] [PATCH v4 20/58] net/txgbe: add Rx and Tx init

[Jiawen Wu <jiawenwu@trustnetic.com>]

Add receive and transmit initialize unit.

Signed-off-by: Jiawen Wu <jiawenwu@trustnetic.com>
---
 doc/guides/nics/features/txgbe.ini  |   4 +
 doc/guides/nics/txgbe.rst           |   1 +
 drivers/net/txgbe/base/txgbe_type.h |   2 +
 drivers/net/txgbe/txgbe_ethdev.h    |   8 +
 drivers/net/txgbe/txgbe_rxtx.c      | 338 ++++++++++++++++++++++++++++
 drivers/net/txgbe/txgbe_rxtx.h      |  32 +++
 6 files changed, 385 insertions(+)

diff --git a/doc/guides/nics/features/txgbe.ini b/doc/guides/nics/features/txgbe.ini
index 115a8699b..707f64131 100644
--- a/doc/guides/nics/features/txgbe.ini
+++ b/doc/guides/nics/features/txgbe.ini
@@ -7,8 +7,12 @@
 Speed capabilities   = Y
 Link status          = Y
 Link status event    = Y
+Jumbo frame          = Y
+Scattered Rx         = Y
 Unicast MAC filter   = Y
 Multicast MAC filter = Y
+CRC offload          = P
+VLAN offload         = P
 Linux UIO            = Y
 Linux VFIO           = Y
 ARMv8                = Y
diff --git a/doc/guides/nics/txgbe.rst b/doc/guides/nics/txgbe.rst
index 0ec4148e2..9ae359c9b 100644
--- a/doc/guides/nics/txgbe.rst
+++ b/doc/guides/nics/txgbe.rst
@@ -12,6 +12,7 @@ Features
 
 - Multiple queues for TX and RX
 - MAC filtering
+- Jumbo frames
 - Link state information
 
 Prerequisites
diff --git a/drivers/net/txgbe/base/txgbe_type.h b/drivers/net/txgbe/base/txgbe_type.h
index 8a8ca963f..747ada0f9 100644
--- a/drivers/net/txgbe/base/txgbe_type.h
+++ b/drivers/net/txgbe/base/txgbe_type.h
@@ -8,6 +8,8 @@
 #define TXGBE_LINK_UP_TIME	90 /* 9.0 Seconds */
 #define TXGBE_AUTO_NEG_TIME	45 /* 4.5 Seconds */
 
+#define TXGBE_FRAME_SIZE_MAX	(9728) /* Maximum frame size, +FCS */
+#define TXGBE_FRAME_SIZE_DFT	(1518) /* Default frame size, +FCS */
 #define TXGBE_MAX_UTA		128
 
 #define TXGBE_ALIGN		128 /* as intel did */
diff --git a/drivers/net/txgbe/txgbe_ethdev.h b/drivers/net/txgbe/txgbe_ethdev.h
index 8fd7a068e..096b17673 100644
--- a/drivers/net/txgbe/txgbe_ethdev.h
+++ b/drivers/net/txgbe/txgbe_ethdev.h
@@ -18,6 +18,7 @@
  * Defines that were not part of txgbe_type.h as they are not used by the
  * FreeBSD driver.
  */
+#define TXGBE_VLAN_TAG_SIZE 4
 #define TXGBE_HKEY_MAX_INDEX 10
 /*Default value of Max Rx Queue*/
 #define TXGBE_MAX_RX_QUEUE_NUM	128
@@ -76,6 +77,13 @@ struct txgbe_adapter {
 #define TXGBE_DEV_UTA_INFO(dev) \
 	(&((struct txgbe_adapter *)(dev)->data->dev_private)->uta_info)
 
+/*
+ * RX/TX function prototypes
+ */
+int txgbe_dev_rx_init(struct rte_eth_dev *dev);
+
+void txgbe_dev_tx_init(struct rte_eth_dev *dev);
+
 void txgbe_set_ivar_map(struct txgbe_hw *hw, int8_t direction,
 			       uint8_t queue, uint8_t msix_vector);
 
diff --git a/drivers/net/txgbe/txgbe_rxtx.c b/drivers/net/txgbe/txgbe_rxtx.c
index 8a7282328..eadc06bcf 100644
--- a/drivers/net/txgbe/txgbe_rxtx.c
+++ b/drivers/net/txgbe/txgbe_rxtx.c
@@ -10,6 +10,8 @@
 
 #include <rte_common.h>
 #include <rte_ethdev.h>
+#include <rte_ethdev_driver.h>
+#include <rte_mbuf.h>
 
 #include "txgbe_logs.h"
 #include "base/txgbe.h"
@@ -110,3 +112,339 @@ txgbe_get_tx_port_offloads(struct rte_eth_dev *dev)
 	return tx_offload_capa;
 }
 
+void __rte_cold
+txgbe_set_rx_function(struct rte_eth_dev *dev)
+{
+	RTE_SET_USED(dev);
+}
+
+/**
+ * txgbe_get_rscctl_maxdesc
+ *
+ * @pool Memory pool of the Rx queue
+ */
+static inline uint32_t
+txgbe_get_rscctl_maxdesc(struct rte_mempool *pool)
+{
+	struct rte_pktmbuf_pool_private *mp_priv = rte_mempool_get_priv(pool);
+
+	uint16_t maxdesc =
+		RTE_IPV4_MAX_PKT_LEN /
+			(mp_priv->mbuf_data_room_size - RTE_PKTMBUF_HEADROOM);
+
+	if (maxdesc >= 16)
+		return TXGBE_RXCFG_RSCMAX_16;
+	else if (maxdesc >= 8)
+		return TXGBE_RXCFG_RSCMAX_8;
+	else if (maxdesc >= 4)
+		return TXGBE_RXCFG_RSCMAX_4;
+	else
+		return TXGBE_RXCFG_RSCMAX_1;
+}
+
+/**
+ * txgbe_set_rsc - configure RSC related port HW registers
+ *
+ * Configures the port's RSC related registers.
+ *
+ * @dev port handle
+ *
+ * Returns 0 in case of success or a non-zero error code
+ */
+static int
+txgbe_set_rsc(struct rte_eth_dev *dev)
+{
+	struct rte_eth_rxmode *rx_conf = &dev->data->dev_conf.rxmode;
+	struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+	struct rte_eth_dev_info dev_info = { 0 };
+	bool rsc_capable = false;
+	uint16_t i;
+	uint32_t rdrxctl;
+	uint32_t rfctl;
+
+	/* Sanity check */
+	dev->dev_ops->dev_infos_get(dev, &dev_info);
+	if (dev_info.rx_offload_capa & DEV_RX_OFFLOAD_TCP_LRO)
+		rsc_capable = true;
+
+	if (!rsc_capable && (rx_conf->offloads & DEV_RX_OFFLOAD_TCP_LRO)) {
+		PMD_INIT_LOG(CRIT, "LRO is requested on HW that doesn't "
+				   "support it");
+		return -EINVAL;
+	}
+
+	/* RSC global configuration */
+
+	if ((rx_conf->offloads & DEV_RX_OFFLOAD_KEEP_CRC) &&
+	     (rx_conf->offloads & DEV_RX_OFFLOAD_TCP_LRO)) {
+		PMD_INIT_LOG(CRIT, "LRO can't be enabled when HW CRC "
+				    "is disabled");
+		return -EINVAL;
+	}
+
+	rfctl = rd32(hw, TXGBE_PSRCTL);
+	if (rsc_capable && (rx_conf->offloads & DEV_RX_OFFLOAD_TCP_LRO))
+		rfctl &= ~TXGBE_PSRCTL_RSCDIA;
+	else
+		rfctl |= TXGBE_PSRCTL_RSCDIA;
+	wr32(hw, TXGBE_PSRCTL, rfctl);
+
+	/* If LRO hasn't been requested - we are done here. */
+	if (!(rx_conf->offloads & DEV_RX_OFFLOAD_TCP_LRO))
+		return 0;
+
+	/* Set PSRCTL.RSCACK bit */
+	rdrxctl = rd32(hw, TXGBE_PSRCTL);
+	rdrxctl |= TXGBE_PSRCTL_RSCACK;
+	wr32(hw, TXGBE_PSRCTL, rdrxctl);
+
+	/* Per-queue RSC configuration */
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		struct txgbe_rx_queue *rxq = dev->data->rx_queues[i];
+		uint32_t srrctl =
+			rd32(hw, TXGBE_RXCFG(rxq->reg_idx));
+		uint32_t psrtype =
+			rd32(hw, TXGBE_POOLRSS(rxq->reg_idx));
+		uint32_t eitr =
+			rd32(hw, TXGBE_ITR(rxq->reg_idx));
+
+		/*
+		 * txgbe PMD doesn't support header-split at the moment.
+		 */
+		srrctl &= ~TXGBE_RXCFG_HDRLEN_MASK;
+		srrctl |= TXGBE_RXCFG_HDRLEN(128);
+
+		/*
+		 * TODO: Consider setting the Receive Descriptor Minimum
+		 * Threshold Size for an RSC case. This is not an obviously
+		 * beneficiary option but the one worth considering...
+		 */
+
+		srrctl |= TXGBE_RXCFG_RSCENA;
+		srrctl &= ~TXGBE_RXCFG_RSCMAX_MASK;
+		srrctl |= txgbe_get_rscctl_maxdesc(rxq->mb_pool);
+		psrtype |= TXGBE_POOLRSS_L4HDR;
+
+		/*
+		 * RSC: Set ITR interval corresponding to 2K ints/s.
+		 *
+		 * Full-sized RSC aggregations for a 10Gb/s link will
+		 * arrive at about 20K aggregation/s rate.
+		 *
+		 * 2K inst/s rate will make only 10% of the
+		 * aggregations to be closed due to the interrupt timer
+		 * expiration for a streaming at wire-speed case.
+		 *
+		 * For a sparse streaming case this setting will yield
+		 * at most 500us latency for a single RSC aggregation.
+		 */
+		eitr &= ~TXGBE_ITR_IVAL_MASK;
+		eitr |= TXGBE_ITR_IVAL_10G(TXGBE_QUEUE_ITR_INTERVAL_DEFAULT);
+		eitr |= TXGBE_ITR_WRDSA;
+
+		wr32(hw, TXGBE_RXCFG(rxq->reg_idx), srrctl);
+		wr32(hw, TXGBE_POOLRSS(rxq->reg_idx), psrtype);
+		wr32(hw, TXGBE_ITR(rxq->reg_idx), eitr);
+
+		/*
+		 * RSC requires the mapping of the queue to the
+		 * interrupt vector.
+		 */
+		txgbe_set_ivar_map(hw, 0, rxq->reg_idx, i);
+	}
+
+	dev->data->lro = 1;
+
+	PMD_INIT_LOG(DEBUG, "enabling LRO mode");
+
+	return 0;
+}
+
+/*
+ * Initializes Receive Unit.
+ */
+int __rte_cold
+txgbe_dev_rx_init(struct rte_eth_dev *dev)
+{
+	struct txgbe_hw *hw;
+	struct txgbe_rx_queue *rxq;
+	uint64_t bus_addr;
+	uint32_t fctrl;
+	uint32_t hlreg0;
+	uint32_t srrctl;
+	uint32_t rdrxctl;
+	uint32_t rxcsum;
+	uint16_t buf_size;
+	uint16_t i;
+	struct rte_eth_rxmode *rx_conf = &dev->data->dev_conf.rxmode;
+	int rc;
+
+	PMD_INIT_FUNC_TRACE();
+	hw = TXGBE_DEV_HW(dev);
+
+	/*
+	 * Make sure receives are disabled while setting
+	 * up the RX context (registers, descriptor rings, etc.).
+	 */
+	wr32m(hw, TXGBE_MACRXCFG, TXGBE_MACRXCFG_ENA, 0);
+	wr32m(hw, TXGBE_PBRXCTL, TXGBE_PBRXCTL_ENA, 0);
+
+	/* Enable receipt of broadcasted frames */
+	fctrl = rd32(hw, TXGBE_PSRCTL);
+	fctrl |= TXGBE_PSRCTL_BCA;
+	wr32(hw, TXGBE_PSRCTL, fctrl);
+
+	/*
+	 * Configure CRC stripping, if any.
+	 */
+	hlreg0 = rd32(hw, TXGBE_SECRXCTL);
+	if (rx_conf->offloads & DEV_RX_OFFLOAD_KEEP_CRC)
+		hlreg0 &= ~TXGBE_SECRXCTL_CRCSTRIP;
+	else
+		hlreg0 |= TXGBE_SECRXCTL_CRCSTRIP;
+	wr32(hw, TXGBE_SECRXCTL, hlreg0);
+
+	/*
+	 * Configure jumbo frame support, if any.
+	 */
+	if (rx_conf->offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
+		wr32m(hw, TXGBE_FRMSZ, TXGBE_FRMSZ_MAX_MASK,
+			TXGBE_FRMSZ_MAX(rx_conf->max_rx_pkt_len));
+	} else {
+		wr32m(hw, TXGBE_FRMSZ, TXGBE_FRMSZ_MAX_MASK,
+			TXGBE_FRMSZ_MAX(TXGBE_FRAME_SIZE_DFT));
+	}
+
+	/*
+	 * If loopback mode is configured, set LPBK bit.
+	 */
+	hlreg0 = rd32(hw, TXGBE_PSRCTL);
+	if (hw->mac.type == txgbe_mac_raptor &&
+	    dev->data->dev_conf.lpbk_mode)
+		hlreg0 |= TXGBE_PSRCTL_LBENA;
+	else
+		hlreg0 &= ~TXGBE_PSRCTL_LBENA;
+
+	wr32(hw, TXGBE_PSRCTL, hlreg0);
+
+	/*
+	 * Assume no header split and no VLAN strip support
+	 * on any Rx queue first .
+	 */
+	rx_conf->offloads &= ~DEV_RX_OFFLOAD_VLAN_STRIP;
+
+	/* Setup RX queues */
+	for (i = 0; i < dev->data->nb_rx_queues; i++) {
+		rxq = dev->data->rx_queues[i];
+
+		/*
+		 * Reset crc_len in case it was changed after queue setup by a
+		 * call to configure.
+		 */
+		if (rx_conf->offloads & DEV_RX_OFFLOAD_KEEP_CRC)
+			rxq->crc_len = RTE_ETHER_CRC_LEN;
+		else
+			rxq->crc_len = 0;
+
+		/* Setup the Base and Length of the Rx Descriptor Rings */
+		bus_addr = rxq->rx_ring_phys_addr;
+		wr32(hw, TXGBE_RXBAL(rxq->reg_idx),
+				(uint32_t)(bus_addr & BIT_MASK32));
+		wr32(hw, TXGBE_RXBAH(rxq->reg_idx),
+				(uint32_t)(bus_addr >> 32));
+		wr32(hw, TXGBE_RXRP(rxq->reg_idx), 0);
+		wr32(hw, TXGBE_RXWP(rxq->reg_idx), 0);
+
+		srrctl = TXGBE_RXCFG_RNGLEN(rxq->nb_rx_desc);
+
+		/* Set if packets are dropped when no descriptors available */
+		if (rxq->drop_en)
+			srrctl |= TXGBE_RXCFG_DROP;
+
+		/*
+		 * Configure the RX buffer size in the PKTLEN field of
+		 * the RXCFG register of the queue.
+		 * The value is in 1 KB resolution. Valid values can be from
+		 * 1 KB to 16 KB.
+		 */
+		buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
+			RTE_PKTMBUF_HEADROOM);
+		buf_size = ROUND_UP(buf_size, 0x1 << 10);
+		srrctl |= TXGBE_RXCFG_PKTLEN(buf_size);
+
+		wr32(hw, TXGBE_RXCFG(rxq->reg_idx), srrctl);
+
+		/* It adds dual VLAN length for supporting dual VLAN */
+		if (dev->data->dev_conf.rxmode.max_rx_pkt_len +
+					    2 * TXGBE_VLAN_TAG_SIZE > buf_size)
+			dev->data->scattered_rx = 1;
+		if (rxq->offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
+			rx_conf->offloads |= DEV_RX_OFFLOAD_VLAN_STRIP;
+	}
+
+	if (rx_conf->offloads & DEV_RX_OFFLOAD_SCATTER)
+		dev->data->scattered_rx = 1;
+
+	/*
+	 * Setup the Checksum Register.
+	 * Disable Full-Packet Checksum which is mutually exclusive with RSS.
+	 * Enable IP/L4 checksum computation by hardware if requested to do so.
+	 */
+	rxcsum = rd32(hw, TXGBE_PSRCTL);
+	rxcsum |= TXGBE_PSRCTL_PCSD;
+	if (rx_conf->offloads & DEV_RX_OFFLOAD_CHECKSUM)
+		rxcsum |= TXGBE_PSRCTL_L4CSUM;
+	else
+		rxcsum &= ~TXGBE_PSRCTL_L4CSUM;
+
+	wr32(hw, TXGBE_PSRCTL, rxcsum);
+
+	if (hw->mac.type == txgbe_mac_raptor) {
+		rdrxctl = rd32(hw, TXGBE_SECRXCTL);
+		if (rx_conf->offloads & DEV_RX_OFFLOAD_KEEP_CRC)
+			rdrxctl &= ~TXGBE_SECRXCTL_CRCSTRIP;
+		else
+			rdrxctl |= TXGBE_SECRXCTL_CRCSTRIP;
+		wr32(hw, TXGBE_SECRXCTL, rdrxctl);
+	}
+
+	rc = txgbe_set_rsc(dev);
+	if (rc)
+		return rc;
+
+	txgbe_set_rx_function(dev);
+
+	return 0;
+}
+
+/*
+ * Initializes Transmit Unit.
+ */
+void __rte_cold
+txgbe_dev_tx_init(struct rte_eth_dev *dev)
+{
+	struct txgbe_hw     *hw;
+	struct txgbe_tx_queue *txq;
+	uint64_t bus_addr;
+	uint16_t i;
+
+	PMD_INIT_FUNC_TRACE();
+	hw = TXGBE_DEV_HW(dev);
+
+	/* Setup the Base and Length of the Tx Descriptor Rings */
+	for (i = 0; i < dev->data->nb_tx_queues; i++) {
+		txq = dev->data->tx_queues[i];
+
+		bus_addr = txq->tx_ring_phys_addr;
+		wr32(hw, TXGBE_TXBAL(txq->reg_idx),
+				(uint32_t)(bus_addr & BIT_MASK32));
+		wr32(hw, TXGBE_TXBAH(txq->reg_idx),
+				(uint32_t)(bus_addr >> 32));
+		wr32m(hw, TXGBE_TXCFG(txq->reg_idx), TXGBE_TXCFG_BUFLEN_MASK,
+			TXGBE_TXCFG_BUFLEN(txq->nb_tx_desc));
+		/* Setup the HW Tx Head and TX Tail descriptor pointers */
+		wr32(hw, TXGBE_TXRP(txq->reg_idx), 0);
+		wr32(hw, TXGBE_TXWP(txq->reg_idx), 0);
+	}
+}
+
diff --git a/drivers/net/txgbe/txgbe_rxtx.h b/drivers/net/txgbe/txgbe_rxtx.h
index 9488c2b75..7d3d9c275 100644
--- a/drivers/net/txgbe/txgbe_rxtx.h
+++ b/drivers/net/txgbe/txgbe_rxtx.h
@@ -5,8 +5,40 @@
 #ifndef _TXGBE_RXTX_H_
 #define _TXGBE_RXTX_H_
 
+#define RTE_PMD_TXGBE_TX_MAX_BURST 32
+#define RTE_PMD_TXGBE_RX_MAX_BURST 32
+
 #define TXGBE_TX_MAX_SEG                    40
 
+/**
+ * Structure associated with each RX queue.
+ */
+struct txgbe_rx_queue {
+	struct rte_mempool  *mb_pool; /**< mbuf pool to populate RX ring. */
+	uint64_t            rx_ring_phys_addr; /**< RX ring DMA address. */
+	uint16_t            nb_rx_desc; /**< number of RX descriptors. */
+	uint16_t            reg_idx;  /**< RX queue register index. */
+	uint8_t             crc_len;  /**< 0 if CRC stripped, 4 otherwise. */
+	uint8_t             drop_en;  /**< If not 0, set SRRCTL.Drop_En. */
+	uint64_t	    offloads; /**< Rx offloads with DEV_RX_OFFLOAD_* */
+};
+
+/**
+ * Structure associated with each TX queue.
+ */
+struct txgbe_tx_queue {
+	uint64_t            tx_ring_phys_addr; /**< TX ring DMA address. */
+	uint16_t            nb_tx_desc;    /**< number of TX descriptors. */
+	/**< Start freeing TX buffers if there are less free descriptors than
+	 *   this value.
+	 */
+	uint16_t            tx_free_thresh;
+	uint16_t            reg_idx;       /**< TX queue register index. */
+	uint64_t offloads; /**< Tx offload flags of DEV_TX_OFFLOAD_* */
+};
+
+void txgbe_set_rx_function(struct rte_eth_dev *dev);
+
 uint64_t txgbe_get_tx_port_offloads(struct rte_eth_dev *dev);
 uint64_t txgbe_get_rx_queue_offloads(struct rte_eth_dev *dev);
 uint64_t txgbe_get_rx_port_offloads(struct rte_eth_dev *dev);
-- 
2.18.4