net/igc: support Rx and Tx
authorAlvin Zhang <alvinx.zhang@intel.com>
Wed, 15 Apr 2020 08:48:03 +0000 (16:48 +0800)
committerFerruh Yigit <ferruh.yigit@intel.com>
Tue, 21 Apr 2020 11:57:07 +0000 (13:57 +0200)
Below ops are added too:
mac_addr_add
mac_addr_remove
mac_addr_set
set_mc_addr_list
mtu_set
promiscuous_enable
promiscuous_disable
allmulticast_enable
allmulticast_disable
rx_queue_setup
rx_queue_release
rx_queue_count
rx_descriptor_done
rx_descriptor_status
tx_descriptor_status
tx_queue_setup
tx_queue_release
tx_done_cleanup
rxq_info_get
txq_info_get
dev_supported_ptypes_get

Signed-off-by: Alvin Zhang <alvinx.zhang@intel.com>
Reviewed-by: Ferruh Yigit <ferruh.yigit@intel.com>
doc/guides/nics/features/igc.ini
drivers/net/igc/Makefile
drivers/net/igc/igc_ethdev.c
drivers/net/igc/igc_ethdev.h
drivers/net/igc/igc_logs.h
drivers/net/igc/igc_txrx.c [new file with mode: 0644]
drivers/net/igc/igc_txrx.h [new file with mode: 0644]
drivers/net/igc/meson.build

index 0fbdf7c..f910483 100644 (file)
@@ -8,6 +8,21 @@ Link status          = Y
 Link status event    = Y
 FW version           = Y
 LED                  = Y
+Packet type parsing  = Y
+Rx descriptor status = Y
+Tx descriptor status = Y
+MTU update           = Y
+Jumbo frame          = Y
+Scattered Rx         = Y
+TSO                  = Y
+Promiscuous mode     = Y
+Allmulticast mode    = Y
+Unicast MAC filter   = Y
+Multicast MAC filter = Y
+RSS hash             = Y
+CRC offload          = Y
+L3 checksum offload  = Y
+L4 checksum offload  = Y
 Linux UIO            = Y
 Linux VFIO           = Y
 x86-64               = Y
index 0902811..c162c51 100644 (file)
@@ -33,5 +33,6 @@ SRCS-$(CONFIG_RTE_LIBRTE_IGC_PMD) += igc_osdep.c
 SRCS-$(CONFIG_RTE_LIBRTE_IGC_PMD) += igc_phy.c
 SRCS-$(CONFIG_RTE_LIBRTE_IGC_PMD) += igc_logs.c
 SRCS-$(CONFIG_RTE_LIBRTE_IGC_PMD) += igc_ethdev.c
+SRCS-$(CONFIG_RTE_LIBRTE_IGC_PMD) += igc_txrx.c
 
 include $(RTE_SDK)/mk/rte.lib.mk
index 423c817..25998db 100644 (file)
@@ -12,7 +12,7 @@
 #include <rte_malloc.h>
 
 #include "igc_logs.h"
-#include "igc_ethdev.h"
+#include "igc_txrx.h"
 
 #define IGC_INTEL_VENDOR_ID            0x8086
 
 /* MSI-X other interrupt vector */
 #define IGC_MSIX_OTHER_INTR_VEC                0
 
+/* External VLAN Enable bit mask */
+#define IGC_CTRL_EXT_EXT_VLAN          (1u << 26)
+
+static const struct rte_eth_desc_lim rx_desc_lim = {
+       .nb_max = IGC_MAX_RXD,
+       .nb_min = IGC_MIN_RXD,
+       .nb_align = IGC_RXD_ALIGN,
+};
+
+static const struct rte_eth_desc_lim tx_desc_lim = {
+       .nb_max = IGC_MAX_TXD,
+       .nb_min = IGC_MIN_TXD,
+       .nb_align = IGC_TXD_ALIGN,
+       .nb_seg_max = IGC_TX_MAX_SEG,
+       .nb_mtu_seg_max = IGC_TX_MAX_MTU_SEG,
+};
+
 static const struct rte_pci_id pci_id_igc_map[] = {
        { RTE_PCI_DEVICE(IGC_INTEL_VENDOR_ID, IGC_DEV_ID_I225_LM) },
        { RTE_PCI_DEVICE(IGC_INTEL_VENDOR_ID, IGC_DEV_ID_I225_V)  },
@@ -69,17 +86,18 @@ static int eth_igc_infos_get(struct rte_eth_dev *dev,
                        struct rte_eth_dev_info *dev_info);
 static int eth_igc_led_on(struct rte_eth_dev *dev);
 static int eth_igc_led_off(struct rte_eth_dev *dev);
-static void eth_igc_tx_queue_release(void *txq);
-static void eth_igc_rx_queue_release(void *rxq);
-static int
-eth_igc_rx_queue_setup(struct rte_eth_dev *dev, uint16_t rx_queue_id,
-               uint16_t nb_rx_desc, unsigned int socket_id,
-               const struct rte_eth_rxconf *rx_conf,
-               struct rte_mempool *mb_pool);
-static int
-eth_igc_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-               uint16_t nb_desc, unsigned int socket_id,
-               const struct rte_eth_txconf *tx_conf);
+static const uint32_t *eth_igc_supported_ptypes_get(struct rte_eth_dev *dev);
+static int eth_igc_rar_set(struct rte_eth_dev *dev,
+               struct rte_ether_addr *mac_addr, uint32_t index, uint32_t pool);
+static void eth_igc_rar_clear(struct rte_eth_dev *dev, uint32_t index);
+static int eth_igc_default_mac_addr_set(struct rte_eth_dev *dev,
+                       struct rte_ether_addr *addr);
+static int eth_igc_set_mc_addr_list(struct rte_eth_dev *dev,
+                        struct rte_ether_addr *mc_addr_set,
+                        uint32_t nb_mc_addr);
+static int eth_igc_allmulticast_enable(struct rte_eth_dev *dev);
+static int eth_igc_allmulticast_disable(struct rte_eth_dev *dev);
+static int eth_igc_mtu_set(struct rte_eth_dev *dev, uint16_t mtu);
 
 static const struct eth_dev_ops eth_igc_ops = {
        .dev_configure          = eth_igc_configure,
@@ -92,16 +110,30 @@ static const struct eth_dev_ops eth_igc_ops = {
        .dev_set_link_down      = eth_igc_set_link_down,
        .promiscuous_enable     = eth_igc_promiscuous_enable,
        .promiscuous_disable    = eth_igc_promiscuous_disable,
-
+       .allmulticast_enable    = eth_igc_allmulticast_enable,
+       .allmulticast_disable   = eth_igc_allmulticast_disable,
        .fw_version_get         = eth_igc_fw_version_get,
        .dev_infos_get          = eth_igc_infos_get,
        .dev_led_on             = eth_igc_led_on,
        .dev_led_off            = eth_igc_led_off,
+       .dev_supported_ptypes_get = eth_igc_supported_ptypes_get,
+       .mtu_set                = eth_igc_mtu_set,
+       .mac_addr_add           = eth_igc_rar_set,
+       .mac_addr_remove        = eth_igc_rar_clear,
+       .mac_addr_set           = eth_igc_default_mac_addr_set,
+       .set_mc_addr_list       = eth_igc_set_mc_addr_list,
 
        .rx_queue_setup         = eth_igc_rx_queue_setup,
        .rx_queue_release       = eth_igc_rx_queue_release,
+       .rx_queue_count         = eth_igc_rx_queue_count,
+       .rx_descriptor_done     = eth_igc_rx_descriptor_done,
+       .rx_descriptor_status   = eth_igc_rx_descriptor_status,
+       .tx_descriptor_status   = eth_igc_tx_descriptor_status,
        .tx_queue_setup         = eth_igc_tx_queue_setup,
        .tx_queue_release       = eth_igc_tx_queue_release,
+       .tx_done_cleanup        = eth_igc_tx_done_cleanup,
+       .rxq_info_get           = eth_igc_rxq_info_get,
+       .txq_info_get           = eth_igc_txq_info_get,
 };
 
 /*
@@ -366,6 +398,32 @@ eth_igc_interrupt_handler(void *param)
        eth_igc_interrupt_action(dev);
 }
 
+/*
+ * rx,tx enable/disable
+ */
+static void
+eth_igc_rxtx_control(struct rte_eth_dev *dev, bool enable)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t tctl, rctl;
+
+       tctl = IGC_READ_REG(hw, IGC_TCTL);
+       rctl = IGC_READ_REG(hw, IGC_RCTL);
+
+       if (enable) {
+               /* enable Tx/Rx */
+               tctl |= IGC_TCTL_EN;
+               rctl |= IGC_RCTL_EN;
+       } else {
+               /* disable Tx/Rx */
+               tctl &= ~IGC_TCTL_EN;
+               rctl &= ~IGC_RCTL_EN;
+       }
+       IGC_WRITE_REG(hw, IGC_TCTL, tctl);
+       IGC_WRITE_REG(hw, IGC_RCTL, rctl);
+       IGC_WRITE_FLUSH(hw);
+}
+
 /*
  *  This routine disables all traffic on the adapter by issuing a
  *  global reset on the MAC.
@@ -381,6 +439,9 @@ eth_igc_stop(struct rte_eth_dev *dev)
 
        adapter->stopped = 1;
 
+       /* disable receive and transmit */
+       eth_igc_rxtx_control(dev, false);
+
        /* disable all MSI-X interrupts */
        IGC_WRITE_REG(hw, IGC_EIMC, 0x1f);
        IGC_WRITE_FLUSH(hw);
@@ -405,6 +466,8 @@ eth_igc_stop(struct rte_eth_dev *dev)
        /* Power down the phy. Needed to make the link go Down */
        eth_igc_set_link_down(dev);
 
+       igc_dev_clear_queues(dev);
+
        /* clear the recorded link status */
        memset(&link, 0, sizeof(link));
        rte_eth_linkstatus_set(dev, &link);
@@ -570,8 +633,7 @@ eth_igc_start(struct rte_eth_dev *dev)
        struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
        struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
        uint32_t *speeds;
-       int num_speeds;
-       bool autoneg;
+       int ret;
 
        PMD_INIT_FUNC_TRACE();
 
@@ -602,6 +664,16 @@ eth_igc_start(struct rte_eth_dev *dev)
        /* confiugre msix for rx interrupt */
        igc_configure_msix_intr(dev);
 
+       igc_tx_init(dev);
+
+       /* This can fail when allocating mbufs for descriptor rings */
+       ret = igc_rx_init(dev);
+       if (ret) {
+               PMD_DRV_LOG(ERR, "Unable to initialize RX hardware");
+               igc_dev_clear_queues(dev);
+               return ret;
+       }
+
        igc_clear_hw_cntrs_base_generic(hw);
 
        /* Setup link speed and duplex */
@@ -610,8 +682,8 @@ eth_igc_start(struct rte_eth_dev *dev)
                hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
                hw->mac.autoneg = 1;
        } else {
-               num_speeds = 0;
-               autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
+               int num_speeds = 0;
+               bool autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
 
                /* Reset */
                hw->phy.autoneg_advertised = 0;
@@ -685,6 +757,7 @@ eth_igc_start(struct rte_eth_dev *dev)
        /* resume enabled intr since hw reset */
        igc_intr_other_enable(dev);
 
+       eth_igc_rxtx_control(dev, true);
        eth_igc_link_update(dev, 0);
 
        return 0;
@@ -692,6 +765,7 @@ eth_igc_start(struct rte_eth_dev *dev)
 error_invalid_config:
        PMD_DRV_LOG(ERR, "Invalid advertised speeds (%u) for port %u",
                     dev->data->dev_conf.link_speeds, dev->data->port_id);
+       igc_dev_clear_queues(dev);
        return -EINVAL;
 }
 
@@ -749,6 +823,27 @@ igc_reset_swfw_lock(struct igc_hw *hw)
        return IGC_SUCCESS;
 }
 
+/*
+ * free all rx/tx queues.
+ */
+static void
+igc_dev_free_queues(struct rte_eth_dev *dev)
+{
+       uint16_t i;
+
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               eth_igc_rx_queue_release(dev->data->rx_queues[i]);
+               dev->data->rx_queues[i] = NULL;
+       }
+       dev->data->nb_rx_queues = 0;
+
+       for (i = 0; i < dev->data->nb_tx_queues; i++) {
+               eth_igc_tx_queue_release(dev->data->tx_queues[i]);
+               dev->data->tx_queues[i] = NULL;
+       }
+       dev->data->nb_tx_queues = 0;
+}
+
 static void
 eth_igc_close(struct rte_eth_dev *dev)
 {
@@ -776,6 +871,7 @@ eth_igc_close(struct rte_eth_dev *dev)
 
        igc_phy_hw_reset(hw);
        igc_hw_control_release(hw);
+       igc_dev_free_queues(dev);
 
        /* Reset any pending lock */
        igc_reset_swfw_lock(hw);
@@ -959,16 +1055,55 @@ eth_igc_reset(struct rte_eth_dev *dev)
 static int
 eth_igc_promiscuous_enable(struct rte_eth_dev *dev)
 {
-       PMD_INIT_FUNC_TRACE();
-       RTE_SET_USED(dev);
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t rctl;
+
+       rctl = IGC_READ_REG(hw, IGC_RCTL);
+       rctl |= (IGC_RCTL_UPE | IGC_RCTL_MPE);
+       IGC_WRITE_REG(hw, IGC_RCTL, rctl);
        return 0;
 }
 
 static int
 eth_igc_promiscuous_disable(struct rte_eth_dev *dev)
 {
-       PMD_INIT_FUNC_TRACE();
-       RTE_SET_USED(dev);
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t rctl;
+
+       rctl = IGC_READ_REG(hw, IGC_RCTL);
+       rctl &= (~IGC_RCTL_UPE);
+       if (dev->data->all_multicast == 1)
+               rctl |= IGC_RCTL_MPE;
+       else
+               rctl &= (~IGC_RCTL_MPE);
+       IGC_WRITE_REG(hw, IGC_RCTL, rctl);
+       return 0;
+}
+
+static int
+eth_igc_allmulticast_enable(struct rte_eth_dev *dev)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t rctl;
+
+       rctl = IGC_READ_REG(hw, IGC_RCTL);
+       rctl |= IGC_RCTL_MPE;
+       IGC_WRITE_REG(hw, IGC_RCTL, rctl);
+       return 0;
+}
+
+static int
+eth_igc_allmulticast_disable(struct rte_eth_dev *dev)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t rctl;
+
+       if (dev->data->promiscuous == 1)
+               return 0;       /* must remain in all_multicast mode */
+
+       rctl = IGC_READ_REG(hw, IGC_RCTL);
+       rctl &= (~IGC_RCTL_MPE);
+       IGC_WRITE_REG(hw, IGC_RCTL, rctl);
        return 0;
 }
 
@@ -1018,10 +1153,40 @@ eth_igc_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
        dev_info->min_rx_bufsize = 256; /* See BSIZE field of RCTL register. */
        dev_info->max_rx_pktlen = MAX_RX_JUMBO_FRAME_SIZE;
        dev_info->max_mac_addrs = hw->mac.rar_entry_count;
+       dev_info->rx_offload_capa = IGC_RX_OFFLOAD_ALL;
+       dev_info->tx_offload_capa = IGC_TX_OFFLOAD_ALL;
+
        dev_info->max_rx_queues = IGC_QUEUE_PAIRS_NUM;
        dev_info->max_tx_queues = IGC_QUEUE_PAIRS_NUM;
        dev_info->max_vmdq_pools = 0;
 
+       dev_info->hash_key_size = IGC_HKEY_MAX_INDEX * sizeof(uint32_t);
+       dev_info->reta_size = ETH_RSS_RETA_SIZE_128;
+       dev_info->flow_type_rss_offloads = IGC_RSS_OFFLOAD_ALL;
+
+       dev_info->default_rxconf = (struct rte_eth_rxconf) {
+               .rx_thresh = {
+                       .pthresh = IGC_DEFAULT_RX_PTHRESH,
+                       .hthresh = IGC_DEFAULT_RX_HTHRESH,
+                       .wthresh = IGC_DEFAULT_RX_WTHRESH,
+               },
+               .rx_free_thresh = IGC_DEFAULT_RX_FREE_THRESH,
+               .rx_drop_en = 0,
+               .offloads = 0,
+       };
+
+       dev_info->default_txconf = (struct rte_eth_txconf) {
+               .tx_thresh = {
+                       .pthresh = IGC_DEFAULT_TX_PTHRESH,
+                       .hthresh = IGC_DEFAULT_TX_HTHRESH,
+                       .wthresh = IGC_DEFAULT_TX_WTHRESH,
+               },
+               .offloads = 0,
+       };
+
+       dev_info->rx_desc_lim = rx_desc_lim;
+       dev_info->tx_desc_lim = tx_desc_lim;
+
        dev_info->speed_capa = ETH_LINK_SPEED_10M_HD | ETH_LINK_SPEED_10M |
                        ETH_LINK_SPEED_100M_HD | ETH_LINK_SPEED_100M |
                        ETH_LINK_SPEED_1G | ETH_LINK_SPEED_2_5G;
@@ -1047,44 +1212,115 @@ eth_igc_led_off(struct rte_eth_dev *dev)
        return igc_led_off(hw) == IGC_SUCCESS ? 0 : -ENOTSUP;
 }
 
+static const uint32_t *
+eth_igc_supported_ptypes_get(__rte_unused struct rte_eth_dev *dev)
+{
+       static const uint32_t ptypes[] = {
+               /* refers to rx_desc_pkt_info_to_pkt_type() */
+               RTE_PTYPE_L2_ETHER,
+               RTE_PTYPE_L3_IPV4,
+               RTE_PTYPE_L3_IPV4_EXT,
+               RTE_PTYPE_L3_IPV6,
+               RTE_PTYPE_L3_IPV6_EXT,
+               RTE_PTYPE_L4_TCP,
+               RTE_PTYPE_L4_UDP,
+               RTE_PTYPE_L4_SCTP,
+               RTE_PTYPE_TUNNEL_IP,
+               RTE_PTYPE_INNER_L3_IPV6,
+               RTE_PTYPE_INNER_L3_IPV6_EXT,
+               RTE_PTYPE_INNER_L4_TCP,
+               RTE_PTYPE_INNER_L4_UDP,
+               RTE_PTYPE_UNKNOWN
+       };
+
+       return ptypes;
+}
+
 static int
-eth_igc_rx_queue_setup(struct rte_eth_dev *dev, uint16_t rx_queue_id,
-               uint16_t nb_rx_desc, unsigned int socket_id,
-               const struct rte_eth_rxconf *rx_conf,
-               struct rte_mempool *mb_pool)
+eth_igc_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
 {
-       PMD_INIT_FUNC_TRACE();
-       RTE_SET_USED(dev);
-       RTE_SET_USED(rx_queue_id);
-       RTE_SET_USED(nb_rx_desc);
-       RTE_SET_USED(socket_id);
-       RTE_SET_USED(rx_conf);
-       RTE_SET_USED(mb_pool);
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t frame_size = mtu + IGC_ETH_OVERHEAD;
+       uint32_t rctl;
+
+       /* if extend vlan has been enabled */
+       if (IGC_READ_REG(hw, IGC_CTRL_EXT) & IGC_CTRL_EXT_EXT_VLAN)
+               frame_size += VLAN_TAG_SIZE;
+
+       /* check that mtu is within the allowed range */
+       if (mtu < RTE_ETHER_MIN_MTU ||
+               frame_size > MAX_RX_JUMBO_FRAME_SIZE)
+               return -EINVAL;
+
+       /*
+        * refuse mtu that requires the support of scattered packets when
+        * this feature has not been enabled before.
+        */
+       if (!dev->data->scattered_rx &&
+           frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)
+               return -EINVAL;
+
+       rctl = IGC_READ_REG(hw, IGC_RCTL);
+
+       /* switch to jumbo mode if needed */
+       if (mtu > RTE_ETHER_MTU) {
+               dev->data->dev_conf.rxmode.offloads |=
+                       DEV_RX_OFFLOAD_JUMBO_FRAME;
+               rctl |= IGC_RCTL_LPE;
+       } else {
+               dev->data->dev_conf.rxmode.offloads &=
+                       ~DEV_RX_OFFLOAD_JUMBO_FRAME;
+               rctl &= ~IGC_RCTL_LPE;
+       }
+       IGC_WRITE_REG(hw, IGC_RCTL, rctl);
+
+       /* update max frame size */
+       dev->data->dev_conf.rxmode.max_rx_pkt_len = frame_size;
+
+       IGC_WRITE_REG(hw, IGC_RLPML,
+                       dev->data->dev_conf.rxmode.max_rx_pkt_len);
+
        return 0;
 }
 
 static int
-eth_igc_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
-               uint16_t nb_desc, unsigned int socket_id,
-               const struct rte_eth_txconf *tx_conf)
+eth_igc_rar_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
+               uint32_t index, uint32_t pool)
 {
-       PMD_INIT_FUNC_TRACE();
-       RTE_SET_USED(dev);
-       RTE_SET_USED(queue_idx);
-       RTE_SET_USED(nb_desc);
-       RTE_SET_USED(socket_id);
-       RTE_SET_USED(tx_conf);
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+
+       igc_rar_set(hw, mac_addr->addr_bytes, index);
+       RTE_SET_USED(pool);
        return 0;
 }
 
-static void eth_igc_tx_queue_release(void *txq)
+static void
+eth_igc_rar_clear(struct rte_eth_dev *dev, uint32_t index)
 {
-       RTE_SET_USED(txq);
+       uint8_t addr[RTE_ETHER_ADDR_LEN];
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+
+       memset(addr, 0, sizeof(addr));
+       igc_rar_set(hw, addr, index);
 }
 
-static void eth_igc_rx_queue_release(void *rxq)
+static int
+eth_igc_default_mac_addr_set(struct rte_eth_dev *dev,
+                       struct rte_ether_addr *addr)
 {
-       RTE_SET_USED(rxq);
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       igc_rar_set(hw, addr->addr_bytes, 0);
+       return 0;
+}
+
+static int
+eth_igc_set_mc_addr_list(struct rte_eth_dev *dev,
+                        struct rte_ether_addr *mc_addr_set,
+                        uint32_t nb_mc_addr)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       igc_update_mc_addr_list(hw, (u8 *)mc_addr_set, nb_mc_addr);
+       return 0;
 }
 
 static int
index c1e3217..3910371 100644 (file)
@@ -18,12 +18,74 @@ extern "C" {
 
 #define IGC_QUEUE_PAIRS_NUM            4
 
+#define IGC_HKEY_MAX_INDEX             10
+#define IGC_RSS_RDT_SIZD               128
+
+/*
+ * TDBA/RDBA should be aligned on 16 byte boundary. But TDLEN/RDLEN should be
+ * multiple of 128 bytes. So we align TDBA/RDBA on 128 byte boundary.
+ * This will also optimize cache line size effect.
+ * H/W supports up to cache line size 128.
+ */
+#define IGC_ALIGN                      128
+
+#define IGC_TX_DESCRIPTOR_MULTIPLE     8
+#define IGC_RX_DESCRIPTOR_MULTIPLE     8
+
+#define IGC_RXD_ALIGN  ((uint16_t)(IGC_ALIGN / \
+               sizeof(union igc_adv_rx_desc)))
+#define IGC_TXD_ALIGN  ((uint16_t)(IGC_ALIGN / \
+               sizeof(union igc_adv_tx_desc)))
+#define IGC_MIN_TXD    IGC_TX_DESCRIPTOR_MULTIPLE
+#define IGC_MAX_TXD    ((uint16_t)(0x80000 / sizeof(union igc_adv_tx_desc)))
+#define IGC_MIN_RXD    IGC_RX_DESCRIPTOR_MULTIPLE
+#define IGC_MAX_RXD    ((uint16_t)(0x80000 / sizeof(union igc_adv_rx_desc)))
+
+#define IGC_TX_MAX_SEG         UINT8_MAX
+#define IGC_TX_MAX_MTU_SEG     UINT8_MAX
+
+#define IGC_RX_OFFLOAD_ALL     (    \
+       DEV_RX_OFFLOAD_IPV4_CKSUM  | \
+       DEV_RX_OFFLOAD_UDP_CKSUM   | \
+       DEV_RX_OFFLOAD_TCP_CKSUM   | \
+       DEV_RX_OFFLOAD_SCTP_CKSUM  | \
+       DEV_RX_OFFLOAD_JUMBO_FRAME | \
+       DEV_RX_OFFLOAD_KEEP_CRC    | \
+       DEV_RX_OFFLOAD_SCATTER)
+
+#define IGC_TX_OFFLOAD_ALL     (    \
+       DEV_TX_OFFLOAD_VLAN_INSERT | \
+       DEV_TX_OFFLOAD_IPV4_CKSUM  | \
+       DEV_TX_OFFLOAD_UDP_CKSUM   | \
+       DEV_TX_OFFLOAD_TCP_CKSUM   | \
+       DEV_TX_OFFLOAD_SCTP_CKSUM  | \
+       DEV_TX_OFFLOAD_TCP_TSO     | \
+       DEV_TX_OFFLOAD_UDP_TSO     | \
+       DEV_TX_OFFLOAD_MULTI_SEGS)
+
+#define IGC_RSS_OFFLOAD_ALL    (    \
+       ETH_RSS_IPV4               | \
+       ETH_RSS_NONFRAG_IPV4_TCP   | \
+       ETH_RSS_NONFRAG_IPV4_UDP   | \
+       ETH_RSS_IPV6               | \
+       ETH_RSS_NONFRAG_IPV6_TCP   | \
+       ETH_RSS_NONFRAG_IPV6_UDP   | \
+       ETH_RSS_IPV6_EX            | \
+       ETH_RSS_IPV6_TCP_EX        | \
+       ETH_RSS_IPV6_UDP_EX)
+
 /* structure for interrupt relative data */
 struct igc_interrupt {
        uint32_t flags;
        uint32_t mask;
 };
 
+/* Union of RSS redirect table register */
+union igc_rss_reta_reg {
+       uint32_t dword;
+       uint8_t  bytes[4];
+};
+
 /*
  * Structure to store private data for each driver instance (for each port).
  */
index 67b1699..6457c4d 100644 (file)
@@ -20,6 +20,20 @@ extern int igc_logtype_driver;
 
 #define PMD_INIT_FUNC_TRACE() PMD_INIT_LOG(DEBUG, " >>")
 
+#ifdef RTE_LIBRTE_IGC_DEBUG_RX
+#define PMD_RX_LOG(level, fmt, args...) \
+       RTE_LOG(level, PMD, "%s(): " fmt "\n", __func__, ## args)
+#else
+#define PMD_RX_LOG(level, fmt, args...) do { } while (0)
+#endif
+
+#ifdef RTE_LIBRTE_IGC_DEBUG_TX
+#define PMD_TX_LOG(level, fmt, args...) \
+       RTE_LOG(level, PMD, "%s(): " fmt "\n", __func__, ## args)
+#else
+#define PMD_TX_LOG(level, fmt, args...) do { } while (0)
+#endif
+
 #define PMD_DRV_LOG_RAW(level, fmt, args...) \
        rte_log(RTE_LOG_ ## level, igc_logtype_driver, "%s(): " fmt, \
                __func__, ## args)
diff --git a/drivers/net/igc/igc_txrx.c b/drivers/net/igc/igc_txrx.c
new file mode 100644 (file)
index 0000000..3437c4d
--- /dev/null
@@ -0,0 +1,2109 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2020 Intel Corporation
+ */
+
+#include <rte_config.h>
+#include <rte_malloc.h>
+#include <rte_ethdev_driver.h>
+#include <rte_net.h>
+
+#include "igc_logs.h"
+#include "igc_txrx.h"
+
+#ifdef RTE_PMD_USE_PREFETCH
+#define rte_igc_prefetch(p)            rte_prefetch0(p)
+#else
+#define rte_igc_prefetch(p)            do {} while (0)
+#endif
+
+#ifdef RTE_PMD_PACKET_PREFETCH
+#define rte_packet_prefetch(p)         rte_prefetch1(p)
+#else
+#define rte_packet_prefetch(p)         do {} while (0)
+#endif
+
+/* Multicast / Unicast table offset mask. */
+#define IGC_RCTL_MO_MSK                        (3u << IGC_RCTL_MO_SHIFT)
+
+/* Loopback mode. */
+#define IGC_RCTL_LBM_SHIFT             6
+#define IGC_RCTL_LBM_MSK               (3u << IGC_RCTL_LBM_SHIFT)
+
+/* Hash select for MTA */
+#define IGC_RCTL_HSEL_SHIFT            8
+#define IGC_RCTL_HSEL_MSK              (3u << IGC_RCTL_HSEL_SHIFT)
+#define IGC_RCTL_PSP                   (1u << 21)
+
+/* Receive buffer size for header buffer */
+#define IGC_SRRCTL_BSIZEHEADER_SHIFT   8
+
+/* RX descriptor status and error flags */
+#define IGC_RXD_STAT_L4CS              (1u << 5)
+#define IGC_RXD_STAT_VEXT              (1u << 9)
+#define IGC_RXD_STAT_LLINT             (1u << 11)
+#define IGC_RXD_STAT_SCRC              (1u << 12)
+#define IGC_RXD_STAT_SMDT_MASK         (3u << 13)
+#define IGC_RXD_STAT_MC                        (1u << 19)
+#define IGC_RXD_EXT_ERR_L4E            (1u << 29)
+#define IGC_RXD_EXT_ERR_IPE            (1u << 30)
+#define IGC_RXD_EXT_ERR_RXE            (1u << 31)
+#define IGC_RXD_RSS_TYPE_MASK          0xfu
+#define IGC_RXD_PCTYPE_MASK            (0x7fu << 4)
+#define IGC_RXD_ETQF_SHIFT             12
+#define IGC_RXD_ETQF_MSK               (0xfu << IGC_RXD_ETQF_SHIFT)
+#define IGC_RXD_VPKT                   (1u << 16)
+
+/* TXD control bits */
+#define IGC_TXDCTL_PTHRESH_SHIFT       0
+#define IGC_TXDCTL_HTHRESH_SHIFT       8
+#define IGC_TXDCTL_WTHRESH_SHIFT       16
+#define IGC_TXDCTL_PTHRESH_MSK         (0x1fu << IGC_TXDCTL_PTHRESH_SHIFT)
+#define IGC_TXDCTL_HTHRESH_MSK         (0x1fu << IGC_TXDCTL_HTHRESH_SHIFT)
+#define IGC_TXDCTL_WTHRESH_MSK         (0x1fu << IGC_TXDCTL_WTHRESH_SHIFT)
+
+/* RXD control bits */
+#define IGC_RXDCTL_PTHRESH_SHIFT       0
+#define IGC_RXDCTL_HTHRESH_SHIFT       8
+#define IGC_RXDCTL_WTHRESH_SHIFT       16
+#define IGC_RXDCTL_PTHRESH_MSK         (0x1fu << IGC_RXDCTL_PTHRESH_SHIFT)
+#define IGC_RXDCTL_HTHRESH_MSK         (0x1fu << IGC_RXDCTL_HTHRESH_SHIFT)
+#define IGC_RXDCTL_WTHRESH_MSK         (0x1fu << IGC_RXDCTL_WTHRESH_SHIFT)
+
+#define IGC_TSO_MAX_HDRLEN             512
+#define IGC_TSO_MAX_MSS                        9216
+
+/* Bit Mask to indicate what bits required for building TX context */
+#define IGC_TX_OFFLOAD_MASK (          \
+               PKT_TX_OUTER_IPV4 |     \
+               PKT_TX_IPV6 |           \
+               PKT_TX_IPV4 |           \
+               PKT_TX_VLAN_PKT |       \
+               PKT_TX_IP_CKSUM |       \
+               PKT_TX_L4_MASK |        \
+               PKT_TX_TCP_SEG |        \
+               PKT_TX_UDP_SEG)
+
+#define IGC_TX_OFFLOAD_SEG     (PKT_TX_TCP_SEG | PKT_TX_UDP_SEG)
+
+#define IGC_ADVTXD_POPTS_TXSM  0x00000200 /* L4 Checksum offload request */
+#define IGC_ADVTXD_POPTS_IXSM  0x00000100 /* IP Checksum offload request */
+
+/* L4 Packet TYPE of Reserved */
+#define IGC_ADVTXD_TUCMD_L4T_RSV       0x00001800
+
+#define IGC_TX_OFFLOAD_NOTSUP_MASK (PKT_TX_OFFLOAD_MASK ^ IGC_TX_OFFLOAD_MASK)
+
+/**
+ * Structure associated with each descriptor of the RX ring of a RX queue.
+ */
+struct igc_rx_entry {
+       struct rte_mbuf *mbuf; /**< mbuf associated with RX descriptor. */
+};
+
+/**
+ * Structure associated with each RX queue.
+ */
+struct igc_rx_queue {
+       struct rte_mempool  *mb_pool;   /**< mbuf pool to populate RX ring. */
+       volatile union igc_adv_rx_desc *rx_ring;
+       /**< RX ring virtual address. */
+       uint64_t            rx_ring_phys_addr; /**< RX ring DMA address. */
+       volatile uint32_t   *rdt_reg_addr; /**< RDT register address. */
+       volatile uint32_t   *rdh_reg_addr; /**< RDH register address. */
+       struct igc_rx_entry *sw_ring;   /**< address of RX software ring. */
+       struct rte_mbuf *pkt_first_seg; /**< First segment of current packet. */
+       struct rte_mbuf *pkt_last_seg;  /**< Last segment of current packet. */
+       uint16_t            nb_rx_desc; /**< number of RX descriptors. */
+       uint16_t            rx_tail;    /**< current value of RDT register. */
+       uint16_t            nb_rx_hold; /**< number of held free RX desc. */
+       uint16_t            rx_free_thresh; /**< max free RX desc to hold. */
+       uint16_t            queue_id;   /**< RX queue index. */
+       uint16_t            reg_idx;    /**< RX queue register index. */
+       uint16_t            port_id;    /**< Device port identifier. */
+       uint8_t             pthresh;    /**< Prefetch threshold register. */
+       uint8_t             hthresh;    /**< Host threshold register. */
+       uint8_t             wthresh;    /**< Write-back threshold register. */
+       uint8_t             crc_len;    /**< 0 if CRC stripped, 4 otherwise. */
+       uint8_t             drop_en;    /**< If not 0, set SRRCTL.Drop_En. */
+       uint32_t            flags;      /**< RX flags. */
+       uint64_t            offloads;   /**< offloads of DEV_RX_OFFLOAD_* */
+};
+
+/** Offload features */
+union igc_tx_offload {
+       uint64_t data;
+       struct {
+               uint64_t l3_len:9; /**< L3 (IP) Header Length. */
+               uint64_t l2_len:7; /**< L2 (MAC) Header Length. */
+               uint64_t vlan_tci:16;
+               /**< VLAN Tag Control Identifier(CPU order). */
+               uint64_t l4_len:8; /**< L4 (TCP/UDP) Header Length. */
+               uint64_t tso_segsz:16; /**< TCP TSO segment size. */
+               /* uint64_t unused:8; */
+       };
+};
+
+/*
+ * Compare mask for igc_tx_offload.data,
+ * should be in sync with igc_tx_offload layout.
+ */
+#define TX_MACIP_LEN_CMP_MASK  0x000000000000FFFFULL /**< L2L3 header mask. */
+#define TX_VLAN_CMP_MASK       0x00000000FFFF0000ULL /**< Vlan mask. */
+#define TX_TCP_LEN_CMP_MASK    0x000000FF00000000ULL /**< TCP header mask. */
+#define TX_TSO_MSS_CMP_MASK    0x00FFFF0000000000ULL /**< TSO segsz mask. */
+/** Mac + IP + TCP + Mss mask. */
+#define TX_TSO_CMP_MASK        \
+       (TX_MACIP_LEN_CMP_MASK | TX_TCP_LEN_CMP_MASK | TX_TSO_MSS_CMP_MASK)
+
+/**
+ * Structure to check if new context need be built
+ */
+struct igc_advctx_info {
+       uint64_t flags;           /**< ol_flags related to context build. */
+       /** tx offload: vlan, tso, l2-l3-l4 lengths. */
+       union igc_tx_offload tx_offload;
+       /** compare mask for tx offload. */
+       union igc_tx_offload tx_offload_mask;
+};
+
+/**
+ * Hardware context number
+ */
+enum {
+       IGC_CTX_0    = 0, /**< CTX0    */
+       IGC_CTX_1    = 1, /**< CTX1    */
+       IGC_CTX_NUM  = 2, /**< CTX_NUM */
+};
+
+/**
+ * Structure associated with each descriptor of the TX ring of a TX queue.
+ */
+struct igc_tx_entry {
+       struct rte_mbuf *mbuf; /**< mbuf associated with TX desc, if any. */
+       uint16_t next_id; /**< Index of next descriptor in ring. */
+       uint16_t last_id; /**< Index of last scattered descriptor. */
+};
+
+/**
+ * Structure associated with each TX queue.
+ */
+struct igc_tx_queue {
+       volatile union igc_adv_tx_desc *tx_ring; /**< TX ring address */
+       uint64_t               tx_ring_phys_addr; /**< TX ring DMA address. */
+       struct igc_tx_entry    *sw_ring; /**< virtual address of SW ring. */
+       volatile uint32_t      *tdt_reg_addr; /**< Address of TDT register. */
+       uint32_t               txd_type;      /**< Device-specific TXD type */
+       uint16_t               nb_tx_desc;    /**< number of TX descriptors. */
+       uint16_t               tx_tail;  /**< Current value of TDT register. */
+       uint16_t               tx_head;
+       /**< Index of first used TX descriptor. */
+       uint16_t               queue_id; /**< TX queue index. */
+       uint16_t               reg_idx;  /**< TX queue register index. */
+       uint16_t               port_id;  /**< Device port identifier. */
+       uint8_t                pthresh;  /**< Prefetch threshold register. */
+       uint8_t                hthresh;  /**< Host threshold register. */
+       uint8_t                wthresh;  /**< Write-back threshold register. */
+       uint8_t                ctx_curr;
+
+       /**< Start context position for transmit queue. */
+       struct igc_advctx_info ctx_cache[IGC_CTX_NUM];
+       /**< Hardware context history.*/
+       uint64_t               offloads; /**< offloads of DEV_TX_OFFLOAD_* */
+};
+
+static inline uint64_t
+rx_desc_statuserr_to_pkt_flags(uint32_t statuserr)
+{
+       static uint64_t l4_chksum_flags[] = {0, 0, PKT_RX_L4_CKSUM_GOOD,
+                       PKT_RX_L4_CKSUM_BAD};
+
+       static uint64_t l3_chksum_flags[] = {0, 0, PKT_RX_IP_CKSUM_GOOD,
+                       PKT_RX_IP_CKSUM_BAD};
+       uint64_t pkt_flags = 0;
+       uint32_t tmp;
+
+       if (statuserr & IGC_RXD_STAT_VP)
+               pkt_flags |= PKT_RX_VLAN_STRIPPED;
+
+       tmp = !!(statuserr & (IGC_RXD_STAT_L4CS | IGC_RXD_STAT_UDPCS));
+       tmp = (tmp << 1) | (uint32_t)!!(statuserr & IGC_RXD_EXT_ERR_L4E);
+       pkt_flags |= l4_chksum_flags[tmp];
+
+       tmp = !!(statuserr & IGC_RXD_STAT_IPCS);
+       tmp = (tmp << 1) | (uint32_t)!!(statuserr & IGC_RXD_EXT_ERR_IPE);
+       pkt_flags |= l3_chksum_flags[tmp];
+
+       return pkt_flags;
+}
+
+#define IGC_PACKET_TYPE_IPV4              0X01
+#define IGC_PACKET_TYPE_IPV4_TCP          0X11
+#define IGC_PACKET_TYPE_IPV4_UDP          0X21
+#define IGC_PACKET_TYPE_IPV4_SCTP         0X41
+#define IGC_PACKET_TYPE_IPV4_EXT          0X03
+#define IGC_PACKET_TYPE_IPV4_EXT_SCTP     0X43
+#define IGC_PACKET_TYPE_IPV6              0X04
+#define IGC_PACKET_TYPE_IPV6_TCP          0X14
+#define IGC_PACKET_TYPE_IPV6_UDP          0X24
+#define IGC_PACKET_TYPE_IPV6_EXT          0X0C
+#define IGC_PACKET_TYPE_IPV6_EXT_TCP      0X1C
+#define IGC_PACKET_TYPE_IPV6_EXT_UDP      0X2C
+#define IGC_PACKET_TYPE_IPV4_IPV6         0X05
+#define IGC_PACKET_TYPE_IPV4_IPV6_TCP     0X15
+#define IGC_PACKET_TYPE_IPV4_IPV6_UDP     0X25
+#define IGC_PACKET_TYPE_IPV4_IPV6_EXT     0X0D
+#define IGC_PACKET_TYPE_IPV4_IPV6_EXT_TCP 0X1D
+#define IGC_PACKET_TYPE_IPV4_IPV6_EXT_UDP 0X2D
+#define IGC_PACKET_TYPE_MAX               0X80
+#define IGC_PACKET_TYPE_MASK              0X7F
+#define IGC_PACKET_TYPE_SHIFT             0X04
+
+static inline uint32_t
+rx_desc_pkt_info_to_pkt_type(uint32_t pkt_info)
+{
+       static const uint32_t
+               ptype_table[IGC_PACKET_TYPE_MAX] __rte_cache_aligned = {
+               [IGC_PACKET_TYPE_IPV4] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4,
+               [IGC_PACKET_TYPE_IPV4_EXT] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4_EXT,
+               [IGC_PACKET_TYPE_IPV6] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6,
+               [IGC_PACKET_TYPE_IPV4_IPV6] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_TUNNEL_IP |
+                       RTE_PTYPE_INNER_L3_IPV6,
+               [IGC_PACKET_TYPE_IPV6_EXT] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6_EXT,
+               [IGC_PACKET_TYPE_IPV4_IPV6_EXT] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_TUNNEL_IP |
+                       RTE_PTYPE_INNER_L3_IPV6_EXT,
+               [IGC_PACKET_TYPE_IPV4_TCP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_TCP,
+               [IGC_PACKET_TYPE_IPV6_TCP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6 | RTE_PTYPE_L4_TCP,
+               [IGC_PACKET_TYPE_IPV4_IPV6_TCP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_TUNNEL_IP |
+                       RTE_PTYPE_INNER_L3_IPV6 | RTE_PTYPE_INNER_L4_TCP,
+               [IGC_PACKET_TYPE_IPV6_EXT_TCP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6_EXT | RTE_PTYPE_L4_TCP,
+               [IGC_PACKET_TYPE_IPV4_IPV6_EXT_TCP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_TUNNEL_IP |
+                       RTE_PTYPE_INNER_L3_IPV6_EXT | RTE_PTYPE_INNER_L4_TCP,
+               [IGC_PACKET_TYPE_IPV4_UDP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_UDP,
+               [IGC_PACKET_TYPE_IPV6_UDP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6 | RTE_PTYPE_L4_UDP,
+               [IGC_PACKET_TYPE_IPV4_IPV6_UDP] =  RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_TUNNEL_IP |
+                       RTE_PTYPE_INNER_L3_IPV6 | RTE_PTYPE_INNER_L4_UDP,
+               [IGC_PACKET_TYPE_IPV6_EXT_UDP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV6_EXT | RTE_PTYPE_L4_UDP,
+               [IGC_PACKET_TYPE_IPV4_IPV6_EXT_UDP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_TUNNEL_IP |
+                       RTE_PTYPE_INNER_L3_IPV6_EXT | RTE_PTYPE_INNER_L4_UDP,
+               [IGC_PACKET_TYPE_IPV4_SCTP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4 | RTE_PTYPE_L4_SCTP,
+               [IGC_PACKET_TYPE_IPV4_EXT_SCTP] = RTE_PTYPE_L2_ETHER |
+                       RTE_PTYPE_L3_IPV4_EXT | RTE_PTYPE_L4_SCTP,
+       };
+       if (unlikely(pkt_info & IGC_RXDADV_PKTTYPE_ETQF))
+               return RTE_PTYPE_UNKNOWN;
+
+       pkt_info = (pkt_info >> IGC_PACKET_TYPE_SHIFT) & IGC_PACKET_TYPE_MASK;
+
+       return ptype_table[pkt_info];
+}
+
+static inline void
+rx_desc_get_pkt_info(struct igc_rx_queue *rxq, struct rte_mbuf *rxm,
+               union igc_adv_rx_desc *rxd, uint32_t staterr)
+{
+       uint64_t pkt_flags;
+       uint32_t hlen_type_rss;
+       uint16_t pkt_info;
+
+       /* Prefetch data of first segment, if configured to do so. */
+       rte_packet_prefetch((char *)rxm->buf_addr + rxm->data_off);
+
+       rxm->port = rxq->port_id;
+       hlen_type_rss = rte_le_to_cpu_32(rxd->wb.lower.lo_dword.data);
+       rxm->hash.rss = rte_le_to_cpu_32(rxd->wb.lower.hi_dword.rss);
+       rxm->vlan_tci = rte_le_to_cpu_16(rxd->wb.upper.vlan);
+
+       pkt_flags = (hlen_type_rss & IGC_RXD_RSS_TYPE_MASK) ?
+                       PKT_RX_RSS_HASH : 0;
+
+       if (hlen_type_rss & IGC_RXD_VPKT)
+               pkt_flags |= PKT_RX_VLAN;
+
+       pkt_flags |= rx_desc_statuserr_to_pkt_flags(staterr);
+
+       rxm->ol_flags = pkt_flags;
+       pkt_info = rte_le_to_cpu_16(rxd->wb.lower.lo_dword.hs_rss.pkt_info);
+       rxm->packet_type = rx_desc_pkt_info_to_pkt_type(pkt_info);
+}
+
+static uint16_t
+igc_recv_pkts(void *rx_queue, struct rte_mbuf **rx_pkts, uint16_t nb_pkts)
+{
+       struct igc_rx_queue * const rxq = rx_queue;
+       volatile union igc_adv_rx_desc * const rx_ring = rxq->rx_ring;
+       struct igc_rx_entry * const sw_ring = rxq->sw_ring;
+       uint16_t rx_id = rxq->rx_tail;
+       uint16_t nb_rx = 0;
+       uint16_t nb_hold = 0;
+
+       while (nb_rx < nb_pkts) {
+               volatile union igc_adv_rx_desc *rxdp;
+               struct igc_rx_entry *rxe;
+               struct rte_mbuf *rxm;
+               struct rte_mbuf *nmb;
+               union igc_adv_rx_desc rxd;
+               uint32_t staterr;
+               uint16_t data_len;
+
+               /*
+                * The order of operations here is important as the DD status
+                * bit must not be read after any other descriptor fields.
+                * rx_ring and rxdp are pointing to volatile data so the order
+                * of accesses cannot be reordered by the compiler. If they were
+                * not volatile, they could be reordered which could lead to
+                * using invalid descriptor fields when read from rxd.
+                */
+               rxdp = &rx_ring[rx_id];
+               staterr = rte_cpu_to_le_32(rxdp->wb.upper.status_error);
+               if (!(staterr & IGC_RXD_STAT_DD))
+                       break;
+               rxd = *rxdp;
+
+               /*
+                * End of packet.
+                *
+                * If the IGC_RXD_STAT_EOP flag is not set, the RX packet is
+                * likely to be invalid and to be dropped by the various
+                * validation checks performed by the network stack.
+                *
+                * Allocate a new mbuf to replenish the RX ring descriptor.
+                * If the allocation fails:
+                *    - arrange for that RX descriptor to be the first one
+                *      being parsed the next time the receive function is
+                *      invoked [on the same queue].
+                *
+                *    - Stop parsing the RX ring and return immediately.
+                *
+                * This policy does not drop the packet received in the RX
+                * descriptor for which the allocation of a new mbuf failed.
+                * Thus, it allows that packet to be later retrieved if
+                * mbuf have been freed in the mean time.
+                * As a side effect, holding RX descriptors instead of
+                * systematically giving them back to the NIC may lead to
+                * RX ring exhaustion situations.
+                * However, the NIC can gracefully prevent such situations
+                * to happen by sending specific "back-pressure" flow control
+                * frames to its peer(s).
+                */
+               PMD_RX_LOG(DEBUG,
+                       "port_id=%u queue_id=%u rx_id=%u staterr=0x%x data_len=%u",
+                       rxq->port_id, rxq->queue_id, rx_id, staterr,
+                       rte_le_to_cpu_16(rxd.wb.upper.length));
+
+               nmb = rte_mbuf_raw_alloc(rxq->mb_pool);
+               if (nmb == NULL) {
+                       unsigned int id;
+                       PMD_RX_LOG(DEBUG,
+                               "RX mbuf alloc failed, port_id=%u queue_id=%u",
+                               rxq->port_id, rxq->queue_id);
+                       id = rxq->port_id;
+                       rte_eth_devices[id].data->rx_mbuf_alloc_failed++;
+                       break;
+               }
+
+               nb_hold++;
+               rxe = &sw_ring[rx_id];
+               rx_id++;
+               if (rx_id >= rxq->nb_rx_desc)
+                       rx_id = 0;
+
+               /* Prefetch next mbuf while processing current one. */
+               rte_igc_prefetch(sw_ring[rx_id].mbuf);
+
+               /*
+                * When next RX descriptor is on a cache-line boundary,
+                * prefetch the next 4 RX descriptors and the next 8 pointers
+                * to mbufs.
+                */
+               if ((rx_id & 0x3) == 0) {
+                       rte_igc_prefetch(&rx_ring[rx_id]);
+                       rte_igc_prefetch(&sw_ring[rx_id]);
+               }
+
+               /*
+                * Update RX descriptor with the physical address of the new
+                * data buffer of the new allocated mbuf.
+                */
+               rxm = rxe->mbuf;
+               rxe->mbuf = nmb;
+               rxdp->read.hdr_addr = 0;
+               rxdp->read.pkt_addr =
+                       rte_cpu_to_le_64(rte_mbuf_data_iova_default(nmb));
+               rxm->next = NULL;
+
+               rxm->data_off = RTE_PKTMBUF_HEADROOM;
+               data_len = rte_le_to_cpu_16(rxd.wb.upper.length) - rxq->crc_len;
+               rxm->data_len = data_len;
+               rxm->pkt_len = data_len;
+               rxm->nb_segs = 1;
+
+               rx_desc_get_pkt_info(rxq, rxm, &rxd, staterr);
+
+               /*
+                * Store the mbuf address into the next entry of the array
+                * of returned packets.
+                */
+               rx_pkts[nb_rx++] = rxm;
+       }
+       rxq->rx_tail = rx_id;
+
+       /*
+        * If the number of free RX descriptors is greater than the RX free
+        * threshold of the queue, advance the Receive Descriptor Tail (RDT)
+        * register.
+        * Update the RDT with the value of the last processed RX descriptor
+        * minus 1, to guarantee that the RDT register is never equal to the
+        * RDH register, which creates a "full" ring situation from the
+        * hardware point of view...
+        */
+       nb_hold = nb_hold + rxq->nb_rx_hold;
+       if (nb_hold > rxq->rx_free_thresh) {
+               PMD_RX_LOG(DEBUG,
+                       "port_id=%u queue_id=%u rx_tail=%u nb_hold=%u nb_rx=%u",
+                       rxq->port_id, rxq->queue_id, rx_id, nb_hold, nb_rx);
+               rx_id = (rx_id == 0) ? (rxq->nb_rx_desc - 1) : (rx_id - 1);
+               IGC_PCI_REG_WRITE(rxq->rdt_reg_addr, rx_id);
+               nb_hold = 0;
+       }
+       rxq->nb_rx_hold = nb_hold;
+       return nb_rx;
+}
+
+static uint16_t
+igc_recv_scattered_pkts(void *rx_queue, struct rte_mbuf **rx_pkts,
+                       uint16_t nb_pkts)
+{
+       struct igc_rx_queue * const rxq = rx_queue;
+       volatile union igc_adv_rx_desc * const rx_ring = rxq->rx_ring;
+       struct igc_rx_entry * const sw_ring = rxq->sw_ring;
+       struct rte_mbuf *first_seg = rxq->pkt_first_seg;
+       struct rte_mbuf *last_seg = rxq->pkt_last_seg;
+
+       uint16_t rx_id = rxq->rx_tail;
+       uint16_t nb_rx = 0;
+       uint16_t nb_hold = 0;
+
+       while (nb_rx < nb_pkts) {
+               volatile union igc_adv_rx_desc *rxdp;
+               struct igc_rx_entry *rxe;
+               struct rte_mbuf *rxm;
+               struct rte_mbuf *nmb;
+               union igc_adv_rx_desc rxd;
+               uint32_t staterr;
+               uint16_t data_len;
+
+next_desc:
+               /*
+                * The order of operations here is important as the DD status
+                * bit must not be read after any other descriptor fields.
+                * rx_ring and rxdp are pointing to volatile data so the order
+                * of accesses cannot be reordered by the compiler. If they were
+                * not volatile, they could be reordered which could lead to
+                * using invalid descriptor fields when read from rxd.
+                */
+               rxdp = &rx_ring[rx_id];
+               staterr = rte_cpu_to_le_32(rxdp->wb.upper.status_error);
+               if (!(staterr & IGC_RXD_STAT_DD))
+                       break;
+               rxd = *rxdp;
+
+               /*
+                * Descriptor done.
+                *
+                * Allocate a new mbuf to replenish the RX ring descriptor.
+                * If the allocation fails:
+                *    - arrange for that RX descriptor to be the first one
+                *      being parsed the next time the receive function is
+                *      invoked [on the same queue].
+                *
+                *    - Stop parsing the RX ring and return immediately.
+                *
+                * This policy does not drop the packet received in the RX
+                * descriptor for which the allocation of a new mbuf failed.
+                * Thus, it allows that packet to be later retrieved if
+                * mbuf have been freed in the mean time.
+                * As a side effect, holding RX descriptors instead of
+                * systematically giving them back to the NIC may lead to
+                * RX ring exhaustion situations.
+                * However, the NIC can gracefully prevent such situations
+                * to happen by sending specific "back-pressure" flow control
+                * frames to its peer(s).
+                */
+               PMD_RX_LOG(DEBUG,
+                       "port_id=%u queue_id=%u rx_id=%u staterr=0x%x data_len=%u",
+                       rxq->port_id, rxq->queue_id, rx_id, staterr,
+                       rte_le_to_cpu_16(rxd.wb.upper.length));
+
+               nmb = rte_mbuf_raw_alloc(rxq->mb_pool);
+               if (nmb == NULL) {
+                       unsigned int id;
+                       PMD_RX_LOG(DEBUG,
+                               "RX mbuf alloc failed, port_id=%u queue_id=%u",
+                               rxq->port_id, rxq->queue_id);
+                       id = rxq->port_id;
+                       rte_eth_devices[id].data->rx_mbuf_alloc_failed++;
+                       break;
+               }
+
+               nb_hold++;
+               rxe = &sw_ring[rx_id];
+               rx_id++;
+               if (rx_id >= rxq->nb_rx_desc)
+                       rx_id = 0;
+
+               /* Prefetch next mbuf while processing current one. */
+               rte_igc_prefetch(sw_ring[rx_id].mbuf);
+
+               /*
+                * When next RX descriptor is on a cache-line boundary,
+                * prefetch the next 4 RX descriptors and the next 8 pointers
+                * to mbufs.
+                */
+               if ((rx_id & 0x3) == 0) {
+                       rte_igc_prefetch(&rx_ring[rx_id]);
+                       rte_igc_prefetch(&sw_ring[rx_id]);
+               }
+
+               /*
+                * Update RX descriptor with the physical address of the new
+                * data buffer of the new allocated mbuf.
+                */
+               rxm = rxe->mbuf;
+               rxe->mbuf = nmb;
+               rxdp->read.hdr_addr = 0;
+               rxdp->read.pkt_addr =
+                       rte_cpu_to_le_64(rte_mbuf_data_iova_default(nmb));
+               rxm->next = NULL;
+
+               /*
+                * Set data length & data buffer address of mbuf.
+                */
+               rxm->data_off = RTE_PKTMBUF_HEADROOM;
+               data_len = rte_le_to_cpu_16(rxd.wb.upper.length);
+               rxm->data_len = data_len;
+
+               /*
+                * If this is the first buffer of the received packet,
+                * set the pointer to the first mbuf of the packet and
+                * initialize its context.
+                * Otherwise, update the total length and the number of segments
+                * of the current scattered packet, and update the pointer to
+                * the last mbuf of the current packet.
+                */
+               if (first_seg == NULL) {
+                       first_seg = rxm;
+                       first_seg->pkt_len = data_len;
+                       first_seg->nb_segs = 1;
+               } else {
+                       first_seg->pkt_len += data_len;
+                       first_seg->nb_segs++;
+                       last_seg->next = rxm;
+               }
+
+               /*
+                * If this is not the last buffer of the received packet,
+                * update the pointer to the last mbuf of the current scattered
+                * packet and continue to parse the RX ring.
+                */
+               if (!(staterr & IGC_RXD_STAT_EOP)) {
+                       last_seg = rxm;
+                       goto next_desc;
+               }
+
+               /*
+                * This is the last buffer of the received packet.
+                * If the CRC is not stripped by the hardware:
+                *   - Subtract the CRC length from the total packet length.
+                *   - If the last buffer only contains the whole CRC or a part
+                *     of it, free the mbuf associated to the last buffer.
+                *     If part of the CRC is also contained in the previous
+                *     mbuf, subtract the length of that CRC part from the
+                *     data length of the previous mbuf.
+                */
+               if (unlikely(rxq->crc_len > 0)) {
+                       first_seg->pkt_len -= RTE_ETHER_CRC_LEN;
+                       if (data_len <= RTE_ETHER_CRC_LEN) {
+                               rte_pktmbuf_free_seg(rxm);
+                               first_seg->nb_segs--;
+                               last_seg->data_len = last_seg->data_len -
+                                        (RTE_ETHER_CRC_LEN - data_len);
+                               last_seg->next = NULL;
+                       } else {
+                               rxm->data_len = (uint16_t)
+                                       (data_len - RTE_ETHER_CRC_LEN);
+                       }
+               }
+
+               rx_desc_get_pkt_info(rxq, first_seg, &rxd, staterr);
+
+               /*
+                * Store the mbuf address into the next entry of the array
+                * of returned packets.
+                */
+               rx_pkts[nb_rx++] = first_seg;
+
+               /* Setup receipt context for a new packet. */
+               first_seg = NULL;
+       }
+       rxq->rx_tail = rx_id;
+
+       /*
+        * Save receive context.
+        */
+       rxq->pkt_first_seg = first_seg;
+       rxq->pkt_last_seg = last_seg;
+
+       /*
+        * If the number of free RX descriptors is greater than the RX free
+        * threshold of the queue, advance the Receive Descriptor Tail (RDT)
+        * register.
+        * Update the RDT with the value of the last processed RX descriptor
+        * minus 1, to guarantee that the RDT register is never equal to the
+        * RDH register, which creates a "full" ring situation from the
+        * hardware point of view...
+        */
+       nb_hold = nb_hold + rxq->nb_rx_hold;
+       if (nb_hold > rxq->rx_free_thresh) {
+               PMD_RX_LOG(DEBUG,
+                       "port_id=%u queue_id=%u rx_tail=%u nb_hold=%u nb_rx=%u",
+                       rxq->port_id, rxq->queue_id, rx_id, nb_hold, nb_rx);
+               rx_id = (rx_id == 0) ? (rxq->nb_rx_desc - 1) : (rx_id - 1);
+               IGC_PCI_REG_WRITE(rxq->rdt_reg_addr, rx_id);
+               nb_hold = 0;
+       }
+       rxq->nb_rx_hold = nb_hold;
+       return nb_rx;
+}
+
+static void
+igc_rx_queue_release_mbufs(struct igc_rx_queue *rxq)
+{
+       unsigned int i;
+
+       if (rxq->sw_ring != NULL) {
+               for (i = 0; i < rxq->nb_rx_desc; i++) {
+                       if (rxq->sw_ring[i].mbuf != NULL) {
+                               rte_pktmbuf_free_seg(rxq->sw_ring[i].mbuf);
+                               rxq->sw_ring[i].mbuf = NULL;
+                       }
+               }
+       }
+}
+
+static void
+igc_rx_queue_release(struct igc_rx_queue *rxq)
+{
+       igc_rx_queue_release_mbufs(rxq);
+       rte_free(rxq->sw_ring);
+       rte_free(rxq);
+}
+
+void eth_igc_rx_queue_release(void *rxq)
+{
+       if (rxq)
+               igc_rx_queue_release(rxq);
+}
+
+uint32_t eth_igc_rx_queue_count(struct rte_eth_dev *dev,
+               uint16_t rx_queue_id)
+{
+       /**
+        * Check the DD bit of a rx descriptor of each 4 in a group,
+        * to avoid checking too frequently and downgrading performance
+        * too much.
+        */
+#define IGC_RXQ_SCAN_INTERVAL 4
+
+       volatile union igc_adv_rx_desc *rxdp;
+       struct igc_rx_queue *rxq;
+       uint16_t desc = 0;
+
+       rxq = dev->data->rx_queues[rx_queue_id];
+       rxdp = &rxq->rx_ring[rxq->rx_tail];
+
+       while (desc < rxq->nb_rx_desc - rxq->rx_tail) {
+               if (unlikely(!(rxdp->wb.upper.status_error &
+                               IGC_RXD_STAT_DD)))
+                       return desc;
+               desc += IGC_RXQ_SCAN_INTERVAL;
+               rxdp += IGC_RXQ_SCAN_INTERVAL;
+       }
+       rxdp = &rxq->rx_ring[rxq->rx_tail + desc - rxq->nb_rx_desc];
+
+       while (desc < rxq->nb_rx_desc &&
+               (rxdp->wb.upper.status_error & IGC_RXD_STAT_DD)) {
+               desc += IGC_RXQ_SCAN_INTERVAL;
+               rxdp += IGC_RXQ_SCAN_INTERVAL;
+       }
+
+       return desc;
+}
+
+int eth_igc_rx_descriptor_done(void *rx_queue, uint16_t offset)
+{
+       volatile union igc_adv_rx_desc *rxdp;
+       struct igc_rx_queue *rxq = rx_queue;
+       uint32_t desc;
+
+       if (unlikely(!rxq || offset >= rxq->nb_rx_desc))
+               return 0;
+
+       desc = rxq->rx_tail + offset;
+       if (desc >= rxq->nb_rx_desc)
+               desc -= rxq->nb_rx_desc;
+
+       rxdp = &rxq->rx_ring[desc];
+       return !!(rxdp->wb.upper.status_error &
+                       rte_cpu_to_le_32(IGC_RXD_STAT_DD));
+}
+
+int eth_igc_rx_descriptor_status(void *rx_queue, uint16_t offset)
+{
+       struct igc_rx_queue *rxq = rx_queue;
+       volatile uint32_t *status;
+       uint32_t desc;
+
+       if (unlikely(!rxq || offset >= rxq->nb_rx_desc))
+               return -EINVAL;
+
+       if (offset >= rxq->nb_rx_desc - rxq->nb_rx_hold)
+               return RTE_ETH_RX_DESC_UNAVAIL;
+
+       desc = rxq->rx_tail + offset;
+       if (desc >= rxq->nb_rx_desc)
+               desc -= rxq->nb_rx_desc;
+
+       status = &rxq->rx_ring[desc].wb.upper.status_error;
+       if (*status & rte_cpu_to_le_32(IGC_RXD_STAT_DD))
+               return RTE_ETH_RX_DESC_DONE;
+
+       return RTE_ETH_RX_DESC_AVAIL;
+}
+
+static int
+igc_alloc_rx_queue_mbufs(struct igc_rx_queue *rxq)
+{
+       struct igc_rx_entry *rxe = rxq->sw_ring;
+       uint64_t dma_addr;
+       unsigned int i;
+
+       /* Initialize software ring entries. */
+       for (i = 0; i < rxq->nb_rx_desc; i++) {
+               volatile union igc_adv_rx_desc *rxd;
+               struct rte_mbuf *mbuf = rte_mbuf_raw_alloc(rxq->mb_pool);
+
+               if (mbuf == NULL) {
+                       PMD_DRV_LOG(ERR, "RX mbuf alloc failed, queue_id=%hu",
+                               rxq->queue_id);
+                       return -ENOMEM;
+               }
+               dma_addr = rte_cpu_to_le_64(rte_mbuf_data_iova_default(mbuf));
+               rxd = &rxq->rx_ring[i];
+               rxd->read.hdr_addr = 0;
+               rxd->read.pkt_addr = dma_addr;
+               rxe[i].mbuf = mbuf;
+       }
+
+       return 0;
+}
+
+/*
+ * RSS random key supplied in section 7.1.2.9.3 of the Intel I225 datasheet.
+ * Used as the default key.
+ */
+static uint8_t default_rss_key[40] = {
+       0x6D, 0x5A, 0x56, 0xDA, 0x25, 0x5B, 0x0E, 0xC2,
+       0x41, 0x67, 0x25, 0x3D, 0x43, 0xA3, 0x8F, 0xB0,
+       0xD0, 0xCA, 0x2B, 0xCB, 0xAE, 0x7B, 0x30, 0xB4,
+       0x77, 0xCB, 0x2D, 0xA3, 0x80, 0x30, 0xF2, 0x0C,
+       0x6A, 0x42, 0xB7, 0x3B, 0xBE, 0xAC, 0x01, 0xFA,
+};
+
+static void
+igc_rss_disable(struct rte_eth_dev *dev)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t mrqc;
+
+       mrqc = IGC_READ_REG(hw, IGC_MRQC);
+       mrqc &= ~IGC_MRQC_ENABLE_MASK;
+       IGC_WRITE_REG(hw, IGC_MRQC, mrqc);
+}
+
+static void
+igc_hw_rss_hash_set(struct igc_hw *hw, struct rte_eth_rss_conf *rss_conf)
+{
+       uint32_t *hash_key = (uint32_t *)rss_conf->rss_key;
+       uint32_t mrqc;
+       uint64_t rss_hf;
+
+       if (hash_key != NULL) {
+               uint8_t i;
+
+               /* Fill in RSS hash key */
+               for (i = 0; i < IGC_HKEY_MAX_INDEX; i++)
+                       IGC_WRITE_REG_LE_VALUE(hw, IGC_RSSRK(i), hash_key[i]);
+       }
+
+       /* Set configured hashing protocols in MRQC register */
+       rss_hf = rss_conf->rss_hf;
+       mrqc = IGC_MRQC_ENABLE_RSS_4Q; /* RSS enabled. */
+       if (rss_hf & ETH_RSS_IPV4)
+               mrqc |= IGC_MRQC_RSS_FIELD_IPV4;
+       if (rss_hf & ETH_RSS_NONFRAG_IPV4_TCP)
+               mrqc |= IGC_MRQC_RSS_FIELD_IPV4_TCP;
+       if (rss_hf & ETH_RSS_IPV6)
+               mrqc |= IGC_MRQC_RSS_FIELD_IPV6;
+       if (rss_hf & ETH_RSS_IPV6_EX)
+               mrqc |= IGC_MRQC_RSS_FIELD_IPV6_EX;
+       if (rss_hf & ETH_RSS_NONFRAG_IPV6_TCP)
+               mrqc |= IGC_MRQC_RSS_FIELD_IPV6_TCP;
+       if (rss_hf & ETH_RSS_IPV6_TCP_EX)
+               mrqc |= IGC_MRQC_RSS_FIELD_IPV6_TCP_EX;
+       if (rss_hf & ETH_RSS_NONFRAG_IPV4_UDP)
+               mrqc |= IGC_MRQC_RSS_FIELD_IPV4_UDP;
+       if (rss_hf & ETH_RSS_NONFRAG_IPV6_UDP)
+               mrqc |= IGC_MRQC_RSS_FIELD_IPV6_UDP;
+       if (rss_hf & ETH_RSS_IPV6_UDP_EX)
+               mrqc |= IGC_MRQC_RSS_FIELD_IPV6_UDP_EX;
+       IGC_WRITE_REG(hw, IGC_MRQC, mrqc);
+}
+
+static void
+igc_rss_configure(struct rte_eth_dev *dev)
+{
+       struct rte_eth_rss_conf rss_conf;
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint16_t i;
+
+       /* Fill in redirection table. */
+       for (i = 0; i < IGC_RSS_RDT_SIZD; i++) {
+               union igc_rss_reta_reg reta;
+               uint16_t q_idx, reta_idx;
+
+               q_idx = (uint8_t)((dev->data->nb_rx_queues > 1) ?
+                                  i % dev->data->nb_rx_queues : 0);
+               reta_idx = i % sizeof(reta);
+               reta.bytes[reta_idx] = q_idx;
+               if (reta_idx == sizeof(reta) - 1)
+                       IGC_WRITE_REG_LE_VALUE(hw,
+                               IGC_RETA(i / sizeof(reta)), reta.dword);
+       }
+
+       /*
+        * Configure the RSS key and the RSS protocols used to compute
+        * the RSS hash of input packets.
+        */
+       rss_conf = dev->data->dev_conf.rx_adv_conf.rss_conf;
+       if (rss_conf.rss_key == NULL)
+               rss_conf.rss_key = default_rss_key;
+       igc_hw_rss_hash_set(hw, &rss_conf);
+}
+
+static int
+igc_dev_mq_rx_configure(struct rte_eth_dev *dev)
+{
+       if (RTE_ETH_DEV_SRIOV(dev).active) {
+               PMD_DRV_LOG(ERR, "SRIOV unsupported!");
+               return -EINVAL;
+       }
+
+       switch (dev->data->dev_conf.rxmode.mq_mode) {
+       case ETH_MQ_RX_RSS:
+               igc_rss_configure(dev);
+               break;
+       case ETH_MQ_RX_NONE:
+               /*
+                * configure RSS register for following,
+                * then disable the RSS logic
+                */
+               igc_rss_configure(dev);
+               igc_rss_disable(dev);
+               break;
+       default:
+               PMD_DRV_LOG(ERR, "rx mode(%d) not supported!",
+                       dev->data->dev_conf.rxmode.mq_mode);
+               return -EINVAL;
+       }
+       return 0;
+}
+
+int
+igc_rx_init(struct rte_eth_dev *dev)
+{
+       struct igc_rx_queue *rxq;
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint64_t offloads = dev->data->dev_conf.rxmode.offloads;
+       uint32_t max_rx_pkt_len = dev->data->dev_conf.rxmode.max_rx_pkt_len;
+       uint32_t rctl;
+       uint32_t rxcsum;
+       uint16_t buf_size;
+       uint16_t rctl_bsize;
+       uint16_t i;
+       int ret;
+
+       dev->rx_pkt_burst = igc_recv_pkts;
+
+       /*
+        * Make sure receives are disabled while setting
+        * up the descriptor ring.
+        */
+       rctl = IGC_READ_REG(hw, IGC_RCTL);
+       IGC_WRITE_REG(hw, IGC_RCTL, rctl & ~IGC_RCTL_EN);
+
+       /* Configure support of jumbo frames, if any. */
+       if (offloads & DEV_RX_OFFLOAD_JUMBO_FRAME) {
+               rctl |= IGC_RCTL_LPE;
+
+               /*
+                * Set maximum packet length by default, and might be updated
+                * together with enabling/disabling dual VLAN.
+                */
+               IGC_WRITE_REG(hw, IGC_RLPML, max_rx_pkt_len);
+       } else {
+               rctl &= ~IGC_RCTL_LPE;
+       }
+
+       /* Configure and enable each RX queue. */
+       rctl_bsize = 0;
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               uint64_t bus_addr;
+               uint32_t rxdctl;
+               uint32_t srrctl;
+
+               rxq = dev->data->rx_queues[i];
+               rxq->flags = 0;
+
+               /* Allocate buffers for descriptor rings and set up queue */
+               ret = igc_alloc_rx_queue_mbufs(rxq);
+               if (ret)
+                       return ret;
+
+               /*
+                * Reset crc_len in case it was changed after queue setup by a
+                * call to configure
+                */
+               rxq->crc_len = (offloads & DEV_RX_OFFLOAD_KEEP_CRC) ?
+                               RTE_ETHER_CRC_LEN : 0;
+
+               bus_addr = rxq->rx_ring_phys_addr;
+               IGC_WRITE_REG(hw, IGC_RDLEN(rxq->reg_idx),
+                               rxq->nb_rx_desc *
+                               sizeof(union igc_adv_rx_desc));
+               IGC_WRITE_REG(hw, IGC_RDBAH(rxq->reg_idx),
+                               (uint32_t)(bus_addr >> 32));
+               IGC_WRITE_REG(hw, IGC_RDBAL(rxq->reg_idx),
+                               (uint32_t)bus_addr);
+
+               /* set descriptor configuration */
+               srrctl = IGC_SRRCTL_DESCTYPE_ADV_ONEBUF;
+
+               srrctl |= (uint32_t)(RTE_PKTMBUF_HEADROOM / 64) <<
+                               IGC_SRRCTL_BSIZEHEADER_SHIFT;
+               /*
+                * Configure RX buffer size.
+                */
+               buf_size = (uint16_t)(rte_pktmbuf_data_room_size(rxq->mb_pool) -
+                       RTE_PKTMBUF_HEADROOM);
+               if (buf_size >= 1024) {
+                       /*
+                        * Configure the BSIZEPACKET field of the SRRCTL
+                        * register of the queue.
+                        * Value is in 1 KB resolution, from 1 KB to 16 KB.
+                        * If this field is equal to 0b, then RCTL.BSIZE
+                        * determines the RX packet buffer size.
+                        */
+
+                       srrctl |= ((buf_size >> IGC_SRRCTL_BSIZEPKT_SHIFT) &
+                                  IGC_SRRCTL_BSIZEPKT_MASK);
+                       buf_size = (uint16_t)((srrctl &
+                                       IGC_SRRCTL_BSIZEPKT_MASK) <<
+                                       IGC_SRRCTL_BSIZEPKT_SHIFT);
+
+                       /* It adds dual VLAN length for supporting dual VLAN */
+                       if (max_rx_pkt_len + 2 * VLAN_TAG_SIZE > buf_size)
+                               dev->data->scattered_rx = 1;
+               } else {
+                       /*
+                        * Use BSIZE field of the device RCTL register.
+                        */
+                       if (rctl_bsize == 0 || rctl_bsize > buf_size)
+                               rctl_bsize = buf_size;
+                       dev->data->scattered_rx = 1;
+               }
+
+               /* Set if packets are dropped when no descriptors available */
+               if (rxq->drop_en)
+                       srrctl |= IGC_SRRCTL_DROP_EN;
+
+               IGC_WRITE_REG(hw, IGC_SRRCTL(rxq->reg_idx), srrctl);
+
+               /* Enable this RX queue. */
+               rxdctl = IGC_RXDCTL_QUEUE_ENABLE;
+               rxdctl |= ((uint32_t)rxq->pthresh << IGC_RXDCTL_PTHRESH_SHIFT) &
+                               IGC_RXDCTL_PTHRESH_MSK;
+               rxdctl |= ((uint32_t)rxq->hthresh << IGC_RXDCTL_HTHRESH_SHIFT) &
+                               IGC_RXDCTL_HTHRESH_MSK;
+               rxdctl |= ((uint32_t)rxq->wthresh << IGC_RXDCTL_WTHRESH_SHIFT) &
+                               IGC_RXDCTL_WTHRESH_MSK;
+               IGC_WRITE_REG(hw, IGC_RXDCTL(rxq->reg_idx), rxdctl);
+       }
+
+       if (offloads & DEV_RX_OFFLOAD_SCATTER)
+               dev->data->scattered_rx = 1;
+
+       if (dev->data->scattered_rx) {
+               PMD_DRV_LOG(DEBUG, "forcing scatter mode");
+               dev->rx_pkt_burst = igc_recv_scattered_pkts;
+       }
+       /*
+        * Setup BSIZE field of RCTL register, if needed.
+        * Buffer sizes >= 1024 are not [supposed to be] setup in the RCTL
+        * register, since the code above configures the SRRCTL register of
+        * the RX queue in such a case.
+        * All configurable sizes are:
+        * 16384: rctl |= (IGC_RCTL_SZ_16384 | IGC_RCTL_BSEX);
+        *  8192: rctl |= (IGC_RCTL_SZ_8192  | IGC_RCTL_BSEX);
+        *  4096: rctl |= (IGC_RCTL_SZ_4096  | IGC_RCTL_BSEX);
+        *  2048: rctl |= IGC_RCTL_SZ_2048;
+        *  1024: rctl |= IGC_RCTL_SZ_1024;
+        *   512: rctl |= IGC_RCTL_SZ_512;
+        *   256: rctl |= IGC_RCTL_SZ_256;
+        */
+       if (rctl_bsize > 0) {
+               if (rctl_bsize >= 512) /* 512 <= buf_size < 1024 - use 512 */
+                       rctl |= IGC_RCTL_SZ_512;
+               else /* 256 <= buf_size < 512 - use 256 */
+                       rctl |= IGC_RCTL_SZ_256;
+       }
+
+       /*
+        * Configure RSS if device configured with multiple RX queues.
+        */
+       igc_dev_mq_rx_configure(dev);
+
+       /* Update the rctl since igc_dev_mq_rx_configure may change its value */
+       rctl |= IGC_READ_REG(hw, IGC_RCTL);
+
+       /*
+        * Setup the Checksum Register.
+        * Receive Full-Packet Checksum Offload is mutually exclusive with RSS.
+        */
+       rxcsum = IGC_READ_REG(hw, IGC_RXCSUM);
+       rxcsum |= IGC_RXCSUM_PCSD;
+
+       /* Enable both L3/L4 rx checksum offload */
+       if (offloads & DEV_RX_OFFLOAD_IPV4_CKSUM)
+               rxcsum |= IGC_RXCSUM_IPOFL;
+       else
+               rxcsum &= ~IGC_RXCSUM_IPOFL;
+
+       if (offloads &
+               (DEV_RX_OFFLOAD_TCP_CKSUM | DEV_RX_OFFLOAD_UDP_CKSUM)) {
+               rxcsum |= IGC_RXCSUM_TUOFL;
+               offloads |= DEV_RX_OFFLOAD_SCTP_CKSUM;
+       } else {
+               rxcsum &= ~IGC_RXCSUM_TUOFL;
+       }
+
+       if (offloads & DEV_RX_OFFLOAD_SCTP_CKSUM)
+               rxcsum |= IGC_RXCSUM_CRCOFL;
+       else
+               rxcsum &= ~IGC_RXCSUM_CRCOFL;
+
+       IGC_WRITE_REG(hw, IGC_RXCSUM, rxcsum);
+
+       /* Setup the Receive Control Register. */
+       if (offloads & DEV_RX_OFFLOAD_KEEP_CRC)
+               rctl &= ~IGC_RCTL_SECRC; /* Do not Strip Ethernet CRC. */
+       else
+               rctl |= IGC_RCTL_SECRC; /* Strip Ethernet CRC. */
+
+       rctl &= ~IGC_RCTL_MO_MSK;
+       rctl &= ~IGC_RCTL_LBM_MSK;
+       rctl |= IGC_RCTL_EN | IGC_RCTL_BAM | IGC_RCTL_LBM_NO |
+                       IGC_RCTL_DPF |
+                       (hw->mac.mc_filter_type << IGC_RCTL_MO_SHIFT);
+
+       rctl &= ~(IGC_RCTL_HSEL_MSK | IGC_RCTL_CFIEN | IGC_RCTL_CFI |
+                       IGC_RCTL_PSP | IGC_RCTL_PMCF);
+
+       /* Make sure VLAN Filters are off. */
+       rctl &= ~IGC_RCTL_VFE;
+       /* Don't store bad packets. */
+       rctl &= ~IGC_RCTL_SBP;
+
+       /* Enable Receives. */
+       IGC_WRITE_REG(hw, IGC_RCTL, rctl);
+
+       /*
+        * Setup the HW Rx Head and Tail Descriptor Pointers.
+        * This needs to be done after enable.
+        */
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               rxq = dev->data->rx_queues[i];
+               IGC_WRITE_REG(hw, IGC_RDH(rxq->reg_idx), 0);
+               IGC_WRITE_REG(hw, IGC_RDT(rxq->reg_idx),
+                               rxq->nb_rx_desc - 1);
+       }
+
+       return 0;
+}
+
+static void
+igc_reset_rx_queue(struct igc_rx_queue *rxq)
+{
+       static const union igc_adv_rx_desc zeroed_desc = { {0} };
+       unsigned int i;
+
+       /* Zero out HW ring memory */
+       for (i = 0; i < rxq->nb_rx_desc; i++)
+               rxq->rx_ring[i] = zeroed_desc;
+
+       rxq->rx_tail = 0;
+       rxq->pkt_first_seg = NULL;
+       rxq->pkt_last_seg = NULL;
+}
+
+int
+eth_igc_rx_queue_setup(struct rte_eth_dev *dev,
+                        uint16_t queue_idx,
+                        uint16_t nb_desc,
+                        unsigned int socket_id,
+                        const struct rte_eth_rxconf *rx_conf,
+                        struct rte_mempool *mp)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       const struct rte_memzone *rz;
+       struct igc_rx_queue *rxq;
+       unsigned int size;
+
+       /*
+        * Validate number of receive descriptors.
+        * It must not exceed hardware maximum, and must be multiple
+        * of IGC_RX_DESCRIPTOR_MULTIPLE.
+        */
+       if (nb_desc % IGC_RX_DESCRIPTOR_MULTIPLE != 0 ||
+               nb_desc > IGC_MAX_RXD || nb_desc < IGC_MIN_RXD) {
+               PMD_DRV_LOG(ERR,
+                       "RX descriptor must be multiple of %u(cur: %u) and between %u and %u",
+                       IGC_RX_DESCRIPTOR_MULTIPLE, nb_desc,
+                       IGC_MIN_RXD, IGC_MAX_RXD);
+               return -EINVAL;
+       }
+
+       /* Free memory prior to re-allocation if needed */
+       if (dev->data->rx_queues[queue_idx] != NULL) {
+               igc_rx_queue_release(dev->data->rx_queues[queue_idx]);
+               dev->data->rx_queues[queue_idx] = NULL;
+       }
+
+       /* First allocate the RX queue data structure. */
+       rxq = rte_zmalloc("ethdev RX queue", sizeof(struct igc_rx_queue),
+                         RTE_CACHE_LINE_SIZE);
+       if (rxq == NULL)
+               return -ENOMEM;
+       rxq->offloads = rx_conf->offloads;
+       rxq->mb_pool = mp;
+       rxq->nb_rx_desc = nb_desc;
+       rxq->pthresh = rx_conf->rx_thresh.pthresh;
+       rxq->hthresh = rx_conf->rx_thresh.hthresh;
+       rxq->wthresh = rx_conf->rx_thresh.wthresh;
+       rxq->drop_en = rx_conf->rx_drop_en;
+       rxq->rx_free_thresh = rx_conf->rx_free_thresh;
+       rxq->queue_id = queue_idx;
+       rxq->reg_idx = queue_idx;
+       rxq->port_id = dev->data->port_id;
+
+       /*
+        *  Allocate RX ring hardware descriptors. A memzone large enough to
+        *  handle the maximum ring size is allocated in order to allow for
+        *  resizing in later calls to the queue setup function.
+        */
+       size = sizeof(union igc_adv_rx_desc) * IGC_MAX_RXD;
+       rz = rte_eth_dma_zone_reserve(dev, "rx_ring", queue_idx, size,
+                                     IGC_ALIGN, socket_id);
+       if (rz == NULL) {
+               igc_rx_queue_release(rxq);
+               return -ENOMEM;
+       }
+       rxq->rdt_reg_addr = IGC_PCI_REG_ADDR(hw, IGC_RDT(rxq->reg_idx));
+       rxq->rdh_reg_addr = IGC_PCI_REG_ADDR(hw, IGC_RDH(rxq->reg_idx));
+       rxq->rx_ring_phys_addr = rz->iova;
+       rxq->rx_ring = (union igc_adv_rx_desc *)rz->addr;
+
+       /* Allocate software ring. */
+       rxq->sw_ring = rte_zmalloc("rxq->sw_ring",
+                                  sizeof(struct igc_rx_entry) * nb_desc,
+                                  RTE_CACHE_LINE_SIZE);
+       if (rxq->sw_ring == NULL) {
+               igc_rx_queue_release(rxq);
+               return -ENOMEM;
+       }
+
+       PMD_DRV_LOG(DEBUG, "sw_ring=%p hw_ring=%p dma_addr=0x%" PRIx64,
+               rxq->sw_ring, rxq->rx_ring, rxq->rx_ring_phys_addr);
+
+       dev->data->rx_queues[queue_idx] = rxq;
+       igc_reset_rx_queue(rxq);
+
+       return 0;
+}
+
+/* prepare packets for transmit */
+static uint16_t
+eth_igc_prep_pkts(__rte_unused void *tx_queue, struct rte_mbuf **tx_pkts,
+               uint16_t nb_pkts)
+{
+       int i, ret;
+       struct rte_mbuf *m;
+
+       for (i = 0; i < nb_pkts; i++) {
+               m = tx_pkts[i];
+
+               /* Check some limitations for TSO in hardware */
+               if (m->ol_flags & IGC_TX_OFFLOAD_SEG)
+                       if (m->tso_segsz > IGC_TSO_MAX_MSS ||
+                               m->l2_len + m->l3_len + m->l4_len >
+                               IGC_TSO_MAX_HDRLEN) {
+                               rte_errno = EINVAL;
+                               return i;
+                       }
+
+               if (m->ol_flags & IGC_TX_OFFLOAD_NOTSUP_MASK) {
+                       rte_errno = ENOTSUP;
+                       return i;
+               }
+
+#ifdef RTE_LIBRTE_ETHDEV_DEBUG
+               ret = rte_validate_tx_offload(m);
+               if (ret != 0) {
+                       rte_errno = -ret;
+                       return i;
+               }
+#endif
+               ret = rte_net_intel_cksum_prepare(m);
+               if (ret != 0) {
+                       rte_errno = -ret;
+                       return i;
+               }
+       }
+
+       return i;
+}
+
+/*
+ *There're some limitations in hardware for TCP segmentation offload. We
+ *should check whether the parameters are valid.
+ */
+static inline uint64_t
+check_tso_para(uint64_t ol_req, union igc_tx_offload ol_para)
+{
+       if (!(ol_req & IGC_TX_OFFLOAD_SEG))
+               return ol_req;
+       if (ol_para.tso_segsz > IGC_TSO_MAX_MSS || ol_para.l2_len +
+               ol_para.l3_len + ol_para.l4_len > IGC_TSO_MAX_HDRLEN) {
+               ol_req &= ~IGC_TX_OFFLOAD_SEG;
+               ol_req |= PKT_TX_TCP_CKSUM;
+       }
+       return ol_req;
+}
+
+/*
+ * Check which hardware context can be used. Use the existing match
+ * or create a new context descriptor.
+ */
+static inline uint32_t
+what_advctx_update(struct igc_tx_queue *txq, uint64_t flags,
+               union igc_tx_offload tx_offload)
+{
+       uint32_t curr = txq->ctx_curr;
+
+       /* If match with the current context */
+       if (likely(txq->ctx_cache[curr].flags == flags &&
+               txq->ctx_cache[curr].tx_offload.data ==
+               (txq->ctx_cache[curr].tx_offload_mask.data &
+               tx_offload.data))) {
+               return curr;
+       }
+
+       /* Total two context, if match with the second context */
+       curr ^= 1;
+       if (likely(txq->ctx_cache[curr].flags == flags &&
+               txq->ctx_cache[curr].tx_offload.data ==
+               (txq->ctx_cache[curr].tx_offload_mask.data &
+               tx_offload.data))) {
+               txq->ctx_curr = curr;
+               return curr;
+       }
+
+       /* Mismatch, create new one */
+       return IGC_CTX_NUM;
+}
+
+/*
+ * This is a separate function, looking for optimization opportunity here
+ * Rework required to go with the pre-defined values.
+ */
+static inline void
+igc_set_xmit_ctx(struct igc_tx_queue *txq,
+               volatile struct igc_adv_tx_context_desc *ctx_txd,
+               uint64_t ol_flags, union igc_tx_offload tx_offload)
+{
+       uint32_t type_tucmd_mlhl;
+       uint32_t mss_l4len_idx;
+       uint32_t ctx_curr;
+       uint32_t vlan_macip_lens;
+       union igc_tx_offload tx_offload_mask;
+
+       /* Use the previous context */
+       txq->ctx_curr ^= 1;
+       ctx_curr = txq->ctx_curr;
+
+       tx_offload_mask.data = 0;
+       type_tucmd_mlhl = 0;
+
+       /* Specify which HW CTX to upload. */
+       mss_l4len_idx = (ctx_curr << IGC_ADVTXD_IDX_SHIFT);
+
+       if (ol_flags & PKT_TX_VLAN_PKT)
+               tx_offload_mask.vlan_tci = 0xffff;
+
+       /* check if TCP segmentation required for this packet */
+       if (ol_flags & IGC_TX_OFFLOAD_SEG) {
+               /* implies IP cksum in IPv4 */
+               if (ol_flags & PKT_TX_IP_CKSUM)
+                       type_tucmd_mlhl = IGC_ADVTXD_TUCMD_IPV4 |
+                               IGC_ADVTXD_DTYP_CTXT | IGC_ADVTXD_DCMD_DEXT;
+               else
+                       type_tucmd_mlhl = IGC_ADVTXD_TUCMD_IPV6 |
+                               IGC_ADVTXD_DTYP_CTXT | IGC_ADVTXD_DCMD_DEXT;
+
+               if (ol_flags & PKT_TX_TCP_SEG)
+                       type_tucmd_mlhl |= IGC_ADVTXD_TUCMD_L4T_TCP;
+               else
+                       type_tucmd_mlhl |= IGC_ADVTXD_TUCMD_L4T_UDP;
+
+               tx_offload_mask.data |= TX_TSO_CMP_MASK;
+               mss_l4len_idx |= (uint32_t)tx_offload.tso_segsz <<
+                               IGC_ADVTXD_MSS_SHIFT;
+               mss_l4len_idx |= (uint32_t)tx_offload.l4_len <<
+                               IGC_ADVTXD_L4LEN_SHIFT;
+       } else { /* no TSO, check if hardware checksum is needed */
+               if (ol_flags & (PKT_TX_IP_CKSUM | PKT_TX_L4_MASK))
+                       tx_offload_mask.data |= TX_MACIP_LEN_CMP_MASK;
+
+               if (ol_flags & PKT_TX_IP_CKSUM)
+                       type_tucmd_mlhl = IGC_ADVTXD_TUCMD_IPV4;
+
+               switch (ol_flags & PKT_TX_L4_MASK) {
+               case PKT_TX_TCP_CKSUM:
+                       type_tucmd_mlhl |= IGC_ADVTXD_TUCMD_L4T_TCP |
+                               IGC_ADVTXD_DTYP_CTXT | IGC_ADVTXD_DCMD_DEXT;
+                       mss_l4len_idx |= (uint32_t)sizeof(struct rte_tcp_hdr)
+                               << IGC_ADVTXD_L4LEN_SHIFT;
+                       break;
+               case PKT_TX_UDP_CKSUM:
+                       type_tucmd_mlhl |= IGC_ADVTXD_TUCMD_L4T_UDP |
+                               IGC_ADVTXD_DTYP_CTXT | IGC_ADVTXD_DCMD_DEXT;
+                       mss_l4len_idx |= (uint32_t)sizeof(struct rte_udp_hdr)
+                               << IGC_ADVTXD_L4LEN_SHIFT;
+                       break;
+               case PKT_TX_SCTP_CKSUM:
+                       type_tucmd_mlhl |= IGC_ADVTXD_TUCMD_L4T_SCTP |
+                               IGC_ADVTXD_DTYP_CTXT | IGC_ADVTXD_DCMD_DEXT;
+                       mss_l4len_idx |= (uint32_t)sizeof(struct rte_sctp_hdr)
+                               << IGC_ADVTXD_L4LEN_SHIFT;
+                       break;
+               default:
+                       type_tucmd_mlhl |= IGC_ADVTXD_TUCMD_L4T_RSV |
+                               IGC_ADVTXD_DTYP_CTXT | IGC_ADVTXD_DCMD_DEXT;
+                       break;
+               }
+       }
+
+       txq->ctx_cache[ctx_curr].flags = ol_flags;
+       txq->ctx_cache[ctx_curr].tx_offload.data =
+               tx_offload_mask.data & tx_offload.data;
+       txq->ctx_cache[ctx_curr].tx_offload_mask = tx_offload_mask;
+
+       ctx_txd->type_tucmd_mlhl = rte_cpu_to_le_32(type_tucmd_mlhl);
+       vlan_macip_lens = (uint32_t)tx_offload.data;
+       ctx_txd->vlan_macip_lens = rte_cpu_to_le_32(vlan_macip_lens);
+       ctx_txd->mss_l4len_idx = rte_cpu_to_le_32(mss_l4len_idx);
+       ctx_txd->u.launch_time = 0;
+}
+
+static inline uint32_t
+tx_desc_vlan_flags_to_cmdtype(uint64_t ol_flags)
+{
+       uint32_t cmdtype;
+       static uint32_t vlan_cmd[2] = {0, IGC_ADVTXD_DCMD_VLE};
+       static uint32_t tso_cmd[2] = {0, IGC_ADVTXD_DCMD_TSE};
+       cmdtype = vlan_cmd[(ol_flags & PKT_TX_VLAN_PKT) != 0];
+       cmdtype |= tso_cmd[(ol_flags & IGC_TX_OFFLOAD_SEG) != 0];
+       return cmdtype;
+}
+
+static inline uint32_t
+tx_desc_cksum_flags_to_olinfo(uint64_t ol_flags)
+{
+       static const uint32_t l4_olinfo[2] = {0, IGC_ADVTXD_POPTS_TXSM};
+       static const uint32_t l3_olinfo[2] = {0, IGC_ADVTXD_POPTS_IXSM};
+       uint32_t tmp;
+
+       tmp  = l4_olinfo[(ol_flags & PKT_TX_L4_MASK)  != PKT_TX_L4_NO_CKSUM];
+       tmp |= l3_olinfo[(ol_flags & PKT_TX_IP_CKSUM) != 0];
+       tmp |= l4_olinfo[(ol_flags & IGC_TX_OFFLOAD_SEG) != 0];
+       return tmp;
+}
+
+static uint16_t
+igc_xmit_pkts(void *tx_queue, struct rte_mbuf **tx_pkts, uint16_t nb_pkts)
+{
+       struct igc_tx_queue * const txq = tx_queue;
+       struct igc_tx_entry * const sw_ring = txq->sw_ring;
+       struct igc_tx_entry *txe, *txn;
+       volatile union igc_adv_tx_desc * const txr = txq->tx_ring;
+       volatile union igc_adv_tx_desc *txd;
+       struct rte_mbuf *tx_pkt;
+       struct rte_mbuf *m_seg;
+       uint64_t buf_dma_addr;
+       uint32_t olinfo_status;
+       uint32_t cmd_type_len;
+       uint32_t pkt_len;
+       uint16_t slen;
+       uint64_t ol_flags;
+       uint16_t tx_end;
+       uint16_t tx_id;
+       uint16_t tx_last;
+       uint16_t nb_tx;
+       uint64_t tx_ol_req;
+       uint32_t new_ctx = 0;
+       union igc_tx_offload tx_offload = {0};
+
+       tx_id = txq->tx_tail;
+       txe = &sw_ring[tx_id];
+
+       for (nb_tx = 0; nb_tx < nb_pkts; nb_tx++) {
+               tx_pkt = *tx_pkts++;
+               pkt_len = tx_pkt->pkt_len;
+
+               RTE_MBUF_PREFETCH_TO_FREE(txe->mbuf);
+
+               /*
+                * The number of descriptors that must be allocated for a
+                * packet is the number of segments of that packet, plus 1
+                * Context Descriptor for the VLAN Tag Identifier, if any.
+                * Determine the last TX descriptor to allocate in the TX ring
+                * for the packet, starting from the current position (tx_id)
+                * in the ring.
+                */
+               tx_last = (uint16_t)(tx_id + tx_pkt->nb_segs - 1);
+
+               ol_flags = tx_pkt->ol_flags;
+               tx_ol_req = ol_flags & IGC_TX_OFFLOAD_MASK;
+
+               /* If a Context Descriptor need be built . */
+               if (tx_ol_req) {
+                       tx_offload.l2_len = tx_pkt->l2_len;
+                       tx_offload.l3_len = tx_pkt->l3_len;
+                       tx_offload.l4_len = tx_pkt->l4_len;
+                       tx_offload.vlan_tci = tx_pkt->vlan_tci;
+                       tx_offload.tso_segsz = tx_pkt->tso_segsz;
+                       tx_ol_req = check_tso_para(tx_ol_req, tx_offload);
+
+                       new_ctx = what_advctx_update(txq, tx_ol_req,
+                                       tx_offload);
+                       /* Only allocate context descriptor if required*/
+                       new_ctx = (new_ctx >= IGC_CTX_NUM);
+                       tx_last = (uint16_t)(tx_last + new_ctx);
+               }
+               if (tx_last >= txq->nb_tx_desc)
+                       tx_last = (uint16_t)(tx_last - txq->nb_tx_desc);
+
+               PMD_TX_LOG(DEBUG,
+                       "port_id=%u queue_id=%u pktlen=%u tx_first=%u tx_last=%u",
+                       txq->port_id, txq->queue_id, pkt_len, tx_id, tx_last);
+
+               /*
+                * Check if there are enough free descriptors in the TX ring
+                * to transmit the next packet.
+                * This operation is based on the two following rules:
+                *
+                *   1- Only check that the last needed TX descriptor can be
+                *      allocated (by construction, if that descriptor is free,
+                *      all intermediate ones are also free).
+                *
+                *      For this purpose, the index of the last TX descriptor
+                *      used for a packet (the "last descriptor" of a packet)
+                *      is recorded in the TX entries (the last one included)
+                *      that are associated with all TX descriptors allocated
+                *      for that packet.
+                *
+                *   2- Avoid to allocate the last free TX descriptor of the
+                *      ring, in order to never set the TDT register with the
+                *      same value stored in parallel by the NIC in the TDH
+                *      register, which makes the TX engine of the NIC enter
+                *      in a deadlock situation.
+                *
+                *      By extension, avoid to allocate a free descriptor that
+                *      belongs to the last set of free descriptors allocated
+                *      to the same packet previously transmitted.
+                */
+
+               /*
+                * The "last descriptor" of the previously sent packet, if any,
+                * which used the last descriptor to allocate.
+                */
+               tx_end = sw_ring[tx_last].last_id;
+
+               /*
+                * The next descriptor following that "last descriptor" in the
+                * ring.
+                */
+               tx_end = sw_ring[tx_end].next_id;
+
+               /*
+                * The "last descriptor" associated with that next descriptor.
+                */
+               tx_end = sw_ring[tx_end].last_id;
+
+               /*
+                * Check that this descriptor is free.
+                */
+               if (!(txr[tx_end].wb.status & IGC_TXD_STAT_DD)) {
+                       if (nb_tx == 0)
+                               return 0;
+                       goto end_of_tx;
+               }
+
+               /*
+                * Set common flags of all TX Data Descriptors.
+                *
+                * The following bits must be set in all Data Descriptors:
+                *   - IGC_ADVTXD_DTYP_DATA
+                *   - IGC_ADVTXD_DCMD_DEXT
+                *
+                * The following bits must be set in the first Data Descriptor
+                * and are ignored in the other ones:
+                *   - IGC_ADVTXD_DCMD_IFCS
+                *   - IGC_ADVTXD_MAC_1588
+                *   - IGC_ADVTXD_DCMD_VLE
+                *
+                * The following bits must only be set in the last Data
+                * Descriptor:
+                *   - IGC_TXD_CMD_EOP
+                *
+                * The following bits can be set in any Data Descriptor, but
+                * are only set in the last Data Descriptor:
+                *   - IGC_TXD_CMD_RS
+                */
+               cmd_type_len = txq->txd_type |
+                       IGC_ADVTXD_DCMD_IFCS | IGC_ADVTXD_DCMD_DEXT;
+               if (tx_ol_req & IGC_TX_OFFLOAD_SEG)
+                       pkt_len -= (tx_pkt->l2_len + tx_pkt->l3_len +
+                                       tx_pkt->l4_len);
+               olinfo_status = (pkt_len << IGC_ADVTXD_PAYLEN_SHIFT);
+
+               /*
+                * Timer 0 should be used to for packet timestamping,
+                * sample the packet timestamp to reg 0
+                */
+               if (ol_flags & PKT_TX_IEEE1588_TMST)
+                       cmd_type_len |= IGC_ADVTXD_MAC_TSTAMP;
+
+               if (tx_ol_req) {
+                       /* Setup TX Advanced context descriptor if required */
+                       if (new_ctx) {
+                               volatile struct igc_adv_tx_context_desc *
+                                       ctx_txd = (volatile struct
+                                       igc_adv_tx_context_desc *)&txr[tx_id];
+
+                               txn = &sw_ring[txe->next_id];
+                               RTE_MBUF_PREFETCH_TO_FREE(txn->mbuf);
+
+                               if (txe->mbuf != NULL) {
+                                       rte_pktmbuf_free_seg(txe->mbuf);
+                                       txe->mbuf = NULL;
+                               }
+
+                               igc_set_xmit_ctx(txq, ctx_txd, tx_ol_req,
+                                               tx_offload);
+
+                               txe->last_id = tx_last;
+                               tx_id = txe->next_id;
+                               txe = txn;
+                       }
+
+                       /* Setup the TX Advanced Data Descriptor */
+                       cmd_type_len |=
+                               tx_desc_vlan_flags_to_cmdtype(tx_ol_req);
+                       olinfo_status |=
+                               tx_desc_cksum_flags_to_olinfo(tx_ol_req);
+                       olinfo_status |= (uint32_t)txq->ctx_curr <<
+                                       IGC_ADVTXD_IDX_SHIFT;
+               }
+
+               m_seg = tx_pkt;
+               do {
+                       txn = &sw_ring[txe->next_id];
+                       RTE_MBUF_PREFETCH_TO_FREE(txn->mbuf);
+
+                       txd = &txr[tx_id];
+
+                       if (txe->mbuf != NULL)
+                               rte_pktmbuf_free_seg(txe->mbuf);
+                       txe->mbuf = m_seg;
+
+                       /* Set up transmit descriptor */
+                       slen = (uint16_t)m_seg->data_len;
+                       buf_dma_addr = rte_mbuf_data_iova(m_seg);
+                       txd->read.buffer_addr =
+                               rte_cpu_to_le_64(buf_dma_addr);
+                       txd->read.cmd_type_len =
+                               rte_cpu_to_le_32(cmd_type_len | slen);
+                       txd->read.olinfo_status =
+                               rte_cpu_to_le_32(olinfo_status);
+                       txe->last_id = tx_last;
+                       tx_id = txe->next_id;
+                       txe = txn;
+                       m_seg = m_seg->next;
+               } while (m_seg != NULL);
+
+               /*
+                * The last packet data descriptor needs End Of Packet (EOP)
+                * and Report Status (RS).
+                */
+               txd->read.cmd_type_len |=
+                       rte_cpu_to_le_32(IGC_TXD_CMD_EOP | IGC_TXD_CMD_RS);
+       }
+end_of_tx:
+       rte_wmb();
+
+       /*
+        * Set the Transmit Descriptor Tail (TDT).
+        */
+       IGC_PCI_REG_WRITE_RELAXED(txq->tdt_reg_addr, tx_id);
+       PMD_TX_LOG(DEBUG, "port_id=%u queue_id=%u tx_tail=%u nb_tx=%u",
+               txq->port_id, txq->queue_id, tx_id, nb_tx);
+       txq->tx_tail = tx_id;
+
+       return nb_tx;
+}
+
+int eth_igc_tx_descriptor_status(void *tx_queue, uint16_t offset)
+{
+       struct igc_tx_queue *txq = tx_queue;
+       volatile uint32_t *status;
+       uint32_t desc;
+
+       if (unlikely(!txq || offset >= txq->nb_tx_desc))
+               return -EINVAL;
+
+       desc = txq->tx_tail + offset;
+       if (desc >= txq->nb_tx_desc)
+               desc -= txq->nb_tx_desc;
+
+       status = &txq->tx_ring[desc].wb.status;
+       if (*status & rte_cpu_to_le_32(IGC_TXD_STAT_DD))
+               return RTE_ETH_TX_DESC_DONE;
+
+       return RTE_ETH_TX_DESC_FULL;
+}
+
+static void
+igc_tx_queue_release_mbufs(struct igc_tx_queue *txq)
+{
+       unsigned int i;
+
+       if (txq->sw_ring != NULL) {
+               for (i = 0; i < txq->nb_tx_desc; i++) {
+                       if (txq->sw_ring[i].mbuf != NULL) {
+                               rte_pktmbuf_free_seg(txq->sw_ring[i].mbuf);
+                               txq->sw_ring[i].mbuf = NULL;
+                       }
+               }
+       }
+}
+
+static void
+igc_tx_queue_release(struct igc_tx_queue *txq)
+{
+       igc_tx_queue_release_mbufs(txq);
+       rte_free(txq->sw_ring);
+       rte_free(txq);
+}
+
+void eth_igc_tx_queue_release(void *txq)
+{
+       if (txq)
+               igc_tx_queue_release(txq);
+}
+
+static void
+igc_reset_tx_queue_stat(struct igc_tx_queue *txq)
+{
+       txq->tx_head = 0;
+       txq->tx_tail = 0;
+       txq->ctx_curr = 0;
+       memset((void *)&txq->ctx_cache, 0,
+               IGC_CTX_NUM * sizeof(struct igc_advctx_info));
+}
+
+static void
+igc_reset_tx_queue(struct igc_tx_queue *txq)
+{
+       struct igc_tx_entry *txe = txq->sw_ring;
+       uint16_t i, prev;
+
+       /* Initialize ring entries */
+       prev = (uint16_t)(txq->nb_tx_desc - 1);
+       for (i = 0; i < txq->nb_tx_desc; i++) {
+               volatile union igc_adv_tx_desc *txd = &txq->tx_ring[i];
+
+               txd->wb.status = IGC_TXD_STAT_DD;
+               txe[i].mbuf = NULL;
+               txe[i].last_id = i;
+               txe[prev].next_id = i;
+               prev = i;
+       }
+
+       txq->txd_type = IGC_ADVTXD_DTYP_DATA;
+       igc_reset_tx_queue_stat(txq);
+}
+
+/*
+ * clear all rx/tx queue
+ */
+void
+igc_dev_clear_queues(struct rte_eth_dev *dev)
+{
+       uint16_t i;
+       struct igc_tx_queue *txq;
+       struct igc_rx_queue *rxq;
+
+       for (i = 0; i < dev->data->nb_tx_queues; i++) {
+               txq = dev->data->tx_queues[i];
+               if (txq != NULL) {
+                       igc_tx_queue_release_mbufs(txq);
+                       igc_reset_tx_queue(txq);
+               }
+       }
+
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               rxq = dev->data->rx_queues[i];
+               if (rxq != NULL) {
+                       igc_rx_queue_release_mbufs(rxq);
+                       igc_reset_rx_queue(rxq);
+               }
+       }
+}
+
+int eth_igc_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+               uint16_t nb_desc, unsigned int socket_id,
+               const struct rte_eth_txconf *tx_conf)
+{
+       const struct rte_memzone *tz;
+       struct igc_tx_queue *txq;
+       struct igc_hw *hw;
+       uint32_t size;
+
+       if (nb_desc % IGC_TX_DESCRIPTOR_MULTIPLE != 0 ||
+               nb_desc > IGC_MAX_TXD || nb_desc < IGC_MIN_TXD) {
+               PMD_DRV_LOG(ERR,
+                       "TX-descriptor must be a multiple of %u and between %u and %u, cur: %u",
+                       IGC_TX_DESCRIPTOR_MULTIPLE,
+                       IGC_MAX_TXD, IGC_MIN_TXD, nb_desc);
+               return -EINVAL;
+       }
+
+       hw = IGC_DEV_PRIVATE_HW(dev);
+
+       /*
+        * The tx_free_thresh and tx_rs_thresh values are not used in the 2.5G
+        * driver.
+        */
+       if (tx_conf->tx_free_thresh != 0)
+               PMD_DRV_LOG(INFO,
+                       "The tx_free_thresh parameter is not used for the 2.5G driver");
+       if (tx_conf->tx_rs_thresh != 0)
+               PMD_DRV_LOG(INFO,
+                       "The tx_rs_thresh parameter is not used for the 2.5G driver");
+       if (tx_conf->tx_thresh.wthresh == 0)
+               PMD_DRV_LOG(INFO,
+                       "To improve 2.5G driver performance, consider setting the TX WTHRESH value to 4, 8, or 16.");
+
+       /* Free memory prior to re-allocation if needed */
+       if (dev->data->tx_queues[queue_idx] != NULL) {
+               igc_tx_queue_release(dev->data->tx_queues[queue_idx]);
+               dev->data->tx_queues[queue_idx] = NULL;
+       }
+
+       /* First allocate the tx queue data structure */
+       txq = rte_zmalloc("ethdev TX queue", sizeof(struct igc_tx_queue),
+                                               RTE_CACHE_LINE_SIZE);
+       if (txq == NULL)
+               return -ENOMEM;
+
+       /*
+        * Allocate TX ring hardware descriptors. A memzone large enough to
+        * handle the maximum ring size is allocated in order to allow for
+        * resizing in later calls to the queue setup function.
+        */
+       size = sizeof(union igc_adv_tx_desc) * IGC_MAX_TXD;
+       tz = rte_eth_dma_zone_reserve(dev, "tx_ring", queue_idx, size,
+                                     IGC_ALIGN, socket_id);
+       if (tz == NULL) {
+               igc_tx_queue_release(txq);
+               return -ENOMEM;
+       }
+
+       txq->nb_tx_desc = nb_desc;
+       txq->pthresh = tx_conf->tx_thresh.pthresh;
+       txq->hthresh = tx_conf->tx_thresh.hthresh;
+       txq->wthresh = tx_conf->tx_thresh.wthresh;
+
+       txq->queue_id = queue_idx;
+       txq->reg_idx = queue_idx;
+       txq->port_id = dev->data->port_id;
+
+       txq->tdt_reg_addr = IGC_PCI_REG_ADDR(hw, IGC_TDT(txq->reg_idx));
+       txq->tx_ring_phys_addr = tz->iova;
+
+       txq->tx_ring = (union igc_adv_tx_desc *)tz->addr;
+       /* Allocate software ring */
+       txq->sw_ring = rte_zmalloc("txq->sw_ring",
+                                  sizeof(struct igc_tx_entry) * nb_desc,
+                                  RTE_CACHE_LINE_SIZE);
+       if (txq->sw_ring == NULL) {
+               igc_tx_queue_release(txq);
+               return -ENOMEM;
+       }
+       PMD_DRV_LOG(DEBUG, "sw_ring=%p hw_ring=%p dma_addr=0x%" PRIx64,
+               txq->sw_ring, txq->tx_ring, txq->tx_ring_phys_addr);
+
+       igc_reset_tx_queue(txq);
+       dev->tx_pkt_burst = igc_xmit_pkts;
+       dev->tx_pkt_prepare = &eth_igc_prep_pkts;
+       dev->data->tx_queues[queue_idx] = txq;
+       txq->offloads = tx_conf->offloads;
+
+       return 0;
+}
+
+int
+eth_igc_tx_done_cleanup(void *txqueue, uint32_t free_cnt)
+{
+       struct igc_tx_queue *txq = txqueue;
+       struct igc_tx_entry *sw_ring;
+       volatile union igc_adv_tx_desc *txr;
+       uint16_t tx_first; /* First segment analyzed. */
+       uint16_t tx_id;    /* Current segment being processed. */
+       uint16_t tx_last;  /* Last segment in the current packet. */
+       uint16_t tx_next;  /* First segment of the next packet. */
+       uint32_t count;
+
+       if (txq == NULL)
+               return -ENODEV;
+
+       count = 0;
+       sw_ring = txq->sw_ring;
+       txr = txq->tx_ring;
+
+       /*
+        * tx_tail is the last sent packet on the sw_ring. Goto the end
+        * of that packet (the last segment in the packet chain) and
+        * then the next segment will be the start of the oldest segment
+        * in the sw_ring. This is the first packet that will be
+        * attempted to be freed.
+        */
+
+       /* Get last segment in most recently added packet. */
+       tx_first = sw_ring[txq->tx_tail].last_id;
+
+       /* Get the next segment, which is the oldest segment in ring. */
+       tx_first = sw_ring[tx_first].next_id;
+
+       /* Set the current index to the first. */
+       tx_id = tx_first;
+
+       /*
+        * Loop through each packet. For each packet, verify that an
+        * mbuf exists and that the last segment is free. If so, free
+        * it and move on.
+        */
+       while (1) {
+               tx_last = sw_ring[tx_id].last_id;
+
+               if (sw_ring[tx_last].mbuf) {
+                       if (!(txr[tx_last].wb.status &
+                                       rte_cpu_to_le_32(IGC_TXD_STAT_DD)))
+                               break;
+
+                       /* Get the start of the next packet. */
+                       tx_next = sw_ring[tx_last].next_id;
+
+                       /*
+                        * Loop through all segments in a
+                        * packet.
+                        */
+                       do {
+                               rte_pktmbuf_free_seg(sw_ring[tx_id].mbuf);
+                               sw_ring[tx_id].mbuf = NULL;
+                               sw_ring[tx_id].last_id = tx_id;
+
+                               /* Move to next segemnt. */
+                               tx_id = sw_ring[tx_id].next_id;
+                       } while (tx_id != tx_next);
+
+                       /*
+                        * Increment the number of packets
+                        * freed.
+                        */
+                       count++;
+                       if (unlikely(count == free_cnt))
+                               break;
+               } else {
+                       /*
+                        * There are multiple reasons to be here:
+                        * 1) All the packets on the ring have been
+                        *    freed - tx_id is equal to tx_first
+                        *    and some packets have been freed.
+                        *    - Done, exit
+                        * 2) Interfaces has not sent a rings worth of
+                        *    packets yet, so the segment after tail is
+                        *    still empty. Or a previous call to this
+                        *    function freed some of the segments but
+                        *    not all so there is a hole in the list.
+                        *    Hopefully this is a rare case.
+                        *    - Walk the list and find the next mbuf. If
+                        *      there isn't one, then done.
+                        */
+                       if (likely(tx_id == tx_first && count != 0))
+                               break;
+
+                       /*
+                        * Walk the list and find the next mbuf, if any.
+                        */
+                       do {
+                               /* Move to next segemnt. */
+                               tx_id = sw_ring[tx_id].next_id;
+
+                               if (sw_ring[tx_id].mbuf)
+                                       break;
+
+                       } while (tx_id != tx_first);
+
+                       /*
+                        * Determine why previous loop bailed. If there
+                        * is not an mbuf, done.
+                        */
+                       if (sw_ring[tx_id].mbuf == NULL)
+                               break;
+               }
+       }
+
+       return count;
+}
+
+void
+igc_tx_init(struct rte_eth_dev *dev)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t tctl;
+       uint32_t txdctl;
+       uint16_t i;
+
+       /* Setup the Base and Length of the Tx Descriptor Rings. */
+       for (i = 0; i < dev->data->nb_tx_queues; i++) {
+               struct igc_tx_queue *txq = dev->data->tx_queues[i];
+               uint64_t bus_addr = txq->tx_ring_phys_addr;
+
+               IGC_WRITE_REG(hw, IGC_TDLEN(txq->reg_idx),
+                               txq->nb_tx_desc *
+                               sizeof(union igc_adv_tx_desc));
+               IGC_WRITE_REG(hw, IGC_TDBAH(txq->reg_idx),
+                               (uint32_t)(bus_addr >> 32));
+               IGC_WRITE_REG(hw, IGC_TDBAL(txq->reg_idx),
+                               (uint32_t)bus_addr);
+
+               /* Setup the HW Tx Head and Tail descriptor pointers. */
+               IGC_WRITE_REG(hw, IGC_TDT(txq->reg_idx), 0);
+               IGC_WRITE_REG(hw, IGC_TDH(txq->reg_idx), 0);
+
+               /* Setup Transmit threshold registers. */
+               txdctl = ((uint32_t)txq->pthresh << IGC_TXDCTL_PTHRESH_SHIFT) &
+                               IGC_TXDCTL_PTHRESH_MSK;
+               txdctl |= ((uint32_t)txq->hthresh << IGC_TXDCTL_HTHRESH_SHIFT) &
+                               IGC_TXDCTL_HTHRESH_MSK;
+               txdctl |= ((uint32_t)txq->wthresh << IGC_TXDCTL_WTHRESH_SHIFT) &
+                               IGC_TXDCTL_WTHRESH_MSK;
+               txdctl |= IGC_TXDCTL_QUEUE_ENABLE;
+               IGC_WRITE_REG(hw, IGC_TXDCTL(txq->reg_idx), txdctl);
+       }
+
+       igc_config_collision_dist(hw);
+
+       /* Program the Transmit Control Register. */
+       tctl = IGC_READ_REG(hw, IGC_TCTL);
+       tctl &= ~IGC_TCTL_CT;
+       tctl |= (IGC_TCTL_PSP | IGC_TCTL_RTLC | IGC_TCTL_EN |
+                ((uint32_t)IGC_COLLISION_THRESHOLD << IGC_CT_SHIFT));
+
+       /* This write will effectively turn on the transmit unit. */
+       IGC_WRITE_REG(hw, IGC_TCTL, tctl);
+}
+
+void
+eth_igc_rxq_info_get(struct rte_eth_dev *dev, uint16_t queue_id,
+       struct rte_eth_rxq_info *qinfo)
+{
+       struct igc_rx_queue *rxq;
+
+       rxq = dev->data->rx_queues[queue_id];
+
+       qinfo->mp = rxq->mb_pool;
+       qinfo->scattered_rx = dev->data->scattered_rx;
+       qinfo->nb_desc = rxq->nb_rx_desc;
+
+       qinfo->conf.rx_free_thresh = rxq->rx_free_thresh;
+       qinfo->conf.rx_drop_en = rxq->drop_en;
+       qinfo->conf.offloads = rxq->offloads;
+       qinfo->conf.rx_thresh.hthresh = rxq->hthresh;
+       qinfo->conf.rx_thresh.pthresh = rxq->pthresh;
+       qinfo->conf.rx_thresh.wthresh = rxq->wthresh;
+}
+
+void
+eth_igc_txq_info_get(struct rte_eth_dev *dev, uint16_t queue_id,
+       struct rte_eth_txq_info *qinfo)
+{
+       struct igc_tx_queue *txq;
+
+       txq = dev->data->tx_queues[queue_id];
+
+       qinfo->nb_desc = txq->nb_tx_desc;
+
+       qinfo->conf.tx_thresh.pthresh = txq->pthresh;
+       qinfo->conf.tx_thresh.hthresh = txq->hthresh;
+       qinfo->conf.tx_thresh.wthresh = txq->wthresh;
+       qinfo->conf.offloads = txq->offloads;
+}
diff --git a/drivers/net/igc/igc_txrx.h b/drivers/net/igc/igc_txrx.h
new file mode 100644 (file)
index 0000000..1b66b3f
--- /dev/null
@@ -0,0 +1,50 @@
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2020 Intel Corporation
+ */
+
+#ifndef _IGC_TXRX_H_
+#define _IGC_TXRX_H_
+
+#include "igc_ethdev.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/*
+ * RX/TX function prototypes
+ */
+void eth_igc_tx_queue_release(void *txq);
+void eth_igc_rx_queue_release(void *rxq);
+void igc_dev_clear_queues(struct rte_eth_dev *dev);
+int eth_igc_rx_queue_setup(struct rte_eth_dev *dev, uint16_t rx_queue_id,
+               uint16_t nb_rx_desc, unsigned int socket_id,
+               const struct rte_eth_rxconf *rx_conf,
+               struct rte_mempool *mb_pool);
+
+uint32_t eth_igc_rx_queue_count(struct rte_eth_dev *dev,
+               uint16_t rx_queue_id);
+
+int eth_igc_rx_descriptor_done(void *rx_queue, uint16_t offset);
+
+int eth_igc_rx_descriptor_status(void *rx_queue, uint16_t offset);
+
+int eth_igc_tx_descriptor_status(void *tx_queue, uint16_t offset);
+
+int eth_igc_tx_queue_setup(struct rte_eth_dev *dev, uint16_t queue_idx,
+               uint16_t nb_desc, unsigned int socket_id,
+               const struct rte_eth_txconf *tx_conf);
+int eth_igc_tx_done_cleanup(void *txqueue, uint32_t free_cnt);
+
+int igc_rx_init(struct rte_eth_dev *dev);
+void igc_tx_init(struct rte_eth_dev *dev);
+void eth_igc_rxq_info_get(struct rte_eth_dev *dev, uint16_t queue_id,
+       struct rte_eth_rxq_info *qinfo);
+void eth_igc_txq_info_get(struct rte_eth_dev *dev, uint16_t queue_id,
+       struct rte_eth_txq_info *qinfo);
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif /* _IGC_TXRX_H_ */
index aa211d6..e402f26 100644 (file)
@@ -6,7 +6,8 @@ objs = [base_objs]
 
 sources = files(
        'igc_logs.c',
-       'igc_ethdev.c'
+       'igc_ethdev.c',
+       'igc_txrx.c'
 )
 
 includes += include_directories('base')