net/bnxt: fail init when mbuf allocation fails
[dpdk.git] / drivers / net / igc / igc_ethdev.c
index 1c8507c..a1f1a97 100644 (file)
@@ -8,23 +8,18 @@
 #include <rte_string_fns.h>
 #include <rte_pci.h>
 #include <rte_bus_pci.h>
-#include <rte_ethdev_driver.h>
-#include <rte_ethdev_pci.h>
+#include <ethdev_driver.h>
+#include <ethdev_pci.h>
 #include <rte_malloc.h>
 #include <rte_alarm.h>
 
 #include "igc_logs.h"
 #include "igc_txrx.h"
+#include "igc_filter.h"
+#include "igc_flow.h"
 
 #define IGC_INTEL_VENDOR_ID            0x8086
 
-/*
- * The overhead from MTU to max frame size.
- * Considering VLAN so tag needs to be counted.
- */
-#define IGC_ETH_OVERHEAD               (RTE_ETHER_HDR_LEN + \
-                                       RTE_ETHER_CRC_LEN + VLAN_TAG_SIZE)
-
 #define IGC_FC_PAUSE_TIME              0x0680
 #define IGC_LINK_UPDATE_CHECK_TIMEOUT  90  /* 9s */
 #define IGC_LINK_UPDATE_CHECK_INTERVAL 100 /* ms */
 /* External VLAN Enable bit mask */
 #define IGC_CTRL_EXT_EXT_VLAN          (1u << 26)
 
+/* Speed select */
+#define IGC_CTRL_SPEED_MASK            (7u << 8)
+#define IGC_CTRL_SPEED_2500            (6u << 8)
+
+/* External VLAN Ether Type bit mask and shift */
+#define IGC_VET_EXT                    0xFFFF0000
+#define IGC_VET_EXT_SHIFT              16
+
+/* Force EEE Auto-negotiation */
+#define IGC_EEER_EEE_FRC_AN            (1u << 28)
+
 /* Per Queue Good Packets Received Count */
 #define IGC_PQGPRC(idx)                (0x10010 + 0x100 * (idx))
 /* Per Queue Good Octets Received Count */
@@ -166,11 +172,11 @@ static const struct rte_igc_xstats_name_off rte_igc_stats_strings[] = {
 
 static int eth_igc_configure(struct rte_eth_dev *dev);
 static int eth_igc_link_update(struct rte_eth_dev *dev, int wait_to_complete);
-static void eth_igc_stop(struct rte_eth_dev *dev);
+static int eth_igc_stop(struct rte_eth_dev *dev);
 static int eth_igc_start(struct rte_eth_dev *dev);
 static int eth_igc_set_link_up(struct rte_eth_dev *dev);
 static int eth_igc_set_link_down(struct rte_eth_dev *dev);
-static void eth_igc_close(struct rte_eth_dev *dev);
+static int eth_igc_close(struct rte_eth_dev *dev);
 static int eth_igc_reset(struct rte_eth_dev *dev);
 static int eth_igc_promiscuous_enable(struct rte_eth_dev *dev);
 static int eth_igc_promiscuous_disable(struct rte_eth_dev *dev);
@@ -203,12 +209,35 @@ static int eth_igc_xstats_get_names(struct rte_eth_dev *dev,
                                struct rte_eth_xstat_name *xstats_names,
                                unsigned int size);
 static int eth_igc_xstats_get_names_by_id(struct rte_eth_dev *dev,
-               struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
+               const uint64_t *ids, struct rte_eth_xstat_name *xstats_names,
                unsigned int limit);
 static int eth_igc_xstats_reset(struct rte_eth_dev *dev);
 static int
 eth_igc_queue_stats_mapping_set(struct rte_eth_dev *dev,
        uint16_t queue_id, uint8_t stat_idx, uint8_t is_rx);
+static int
+eth_igc_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id);
+static int
+eth_igc_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id);
+static int
+eth_igc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf);
+static int
+eth_igc_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf);
+static int eth_igc_rss_reta_update(struct rte_eth_dev *dev,
+                       struct rte_eth_rss_reta_entry64 *reta_conf,
+                       uint16_t reta_size);
+static int eth_igc_rss_reta_query(struct rte_eth_dev *dev,
+                      struct rte_eth_rss_reta_entry64 *reta_conf,
+                      uint16_t reta_size);
+static int eth_igc_rss_hash_update(struct rte_eth_dev *dev,
+                       struct rte_eth_rss_conf *rss_conf);
+static int eth_igc_rss_hash_conf_get(struct rte_eth_dev *dev,
+                       struct rte_eth_rss_conf *rss_conf);
+static int
+eth_igc_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on);
+static int eth_igc_vlan_offload_set(struct rte_eth_dev *dev, int mask);
+static int eth_igc_vlan_tpid_set(struct rte_eth_dev *dev,
+                     enum rte_vlan_type vlan_type, uint16_t tpid);
 
 static const struct eth_dev_ops eth_igc_ops = {
        .dev_configure          = eth_igc_configure,
@@ -236,10 +265,6 @@ static const struct eth_dev_ops eth_igc_ops = {
 
        .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,
@@ -253,6 +278,19 @@ static const struct eth_dev_ops eth_igc_ops = {
        .stats_reset            = eth_igc_xstats_reset,
        .xstats_reset           = eth_igc_xstats_reset,
        .queue_stats_mapping_set = eth_igc_queue_stats_mapping_set,
+       .rx_queue_intr_enable   = eth_igc_rx_queue_intr_enable,
+       .rx_queue_intr_disable  = eth_igc_rx_queue_intr_disable,
+       .flow_ctrl_get          = eth_igc_flow_ctrl_get,
+       .flow_ctrl_set          = eth_igc_flow_ctrl_set,
+       .reta_update            = eth_igc_rss_reta_update,
+       .reta_query             = eth_igc_rss_reta_query,
+       .rss_hash_update        = eth_igc_rss_hash_update,
+       .rss_hash_conf_get      = eth_igc_rss_hash_conf_get,
+       .vlan_filter_set        = eth_igc_vlan_filter_set,
+       .vlan_offload_set       = eth_igc_vlan_offload_set,
+       .vlan_tpid_set          = eth_igc_vlan_tpid_set,
+       .vlan_strip_queue_set   = eth_igc_vlan_strip_queue_set,
+       .flow_ops_get           = eth_igc_flow_ops_get,
 };
 
 /*
@@ -269,8 +307,8 @@ igc_check_mq_mode(struct rte_eth_dev *dev)
                return -EINVAL;
        }
 
-       if (rx_mq_mode != ETH_MQ_RX_NONE &&
-               rx_mq_mode != ETH_MQ_RX_RSS) {
+       if (rx_mq_mode != RTE_ETH_MQ_RX_NONE &&
+               rx_mq_mode != RTE_ETH_MQ_RX_RSS) {
                /* RSS together with VMDq not supported*/
                PMD_INIT_LOG(ERR, "RX mode %d is not supported.",
                                rx_mq_mode);
@@ -280,7 +318,7 @@ igc_check_mq_mode(struct rte_eth_dev *dev)
        /* To no break software that set invalid mode, only display
         * warning if invalid mode is used.
         */
-       if (tx_mq_mode != ETH_MQ_TX_NONE)
+       if (tx_mq_mode != RTE_ETH_MQ_TX_NONE)
                PMD_INIT_LOG(WARNING,
                        "TX mode %d is not supported. Due to meaningless in this driver, just ignore",
                        tx_mq_mode);
@@ -296,6 +334,9 @@ eth_igc_configure(struct rte_eth_dev *dev)
 
        PMD_INIT_FUNC_TRACE();
 
+       if (dev->data->dev_conf.rxmode.mq_mode & RTE_ETH_MQ_RX_RSS_FLAG)
+               dev->data->dev_conf.rxmode.offloads |= RTE_ETH_RX_OFFLOAD_RSS_HASH;
+
        ret  = igc_check_mq_mode(dev);
        if (ret != 0)
                return ret;
@@ -336,7 +377,7 @@ igc_intr_other_disable(struct rte_eth_dev *dev)
 {
        struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
        struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
 
        if (rte_intr_allow_others(intr_handle) &&
                dev->data->dev_conf.intr_conf.lsc) {
@@ -356,7 +397,7 @@ igc_intr_other_enable(struct rte_eth_dev *dev)
        struct igc_interrupt *intr = IGC_DEV_PRIVATE_INTR(dev);
        struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
        struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
 
        if (rte_intr_allow_others(intr_handle) &&
                dev->data->dev_conf.intr_conf.lsc) {
@@ -432,12 +473,12 @@ eth_igc_link_update(struct rte_eth_dev *dev, int wait_to_complete)
                uint16_t duplex, speed;
                hw->mac.ops.get_link_up_info(hw, &speed, &duplex);
                link.link_duplex = (duplex == FULL_DUPLEX) ?
-                               ETH_LINK_FULL_DUPLEX :
-                               ETH_LINK_HALF_DUPLEX;
+                               RTE_ETH_LINK_FULL_DUPLEX :
+                               RTE_ETH_LINK_HALF_DUPLEX;
                link.link_speed = speed;
-               link.link_status = ETH_LINK_UP;
+               link.link_status = RTE_ETH_LINK_UP;
                link.link_autoneg = !(dev->data->dev_conf.link_speeds &
-                               ETH_LINK_SPEED_FIXED);
+                               RTE_ETH_LINK_SPEED_FIXED);
 
                if (speed == SPEED_2500) {
                        uint32_t tipg = IGC_READ_REG(hw, IGC_TIPG);
@@ -449,9 +490,9 @@ eth_igc_link_update(struct rte_eth_dev *dev, int wait_to_complete)
                }
        } else {
                link.link_speed = 0;
-               link.link_duplex = ETH_LINK_HALF_DUPLEX;
-               link.link_status = ETH_LINK_DOWN;
-               link.link_autoneg = ETH_LINK_FIXED;
+               link.link_duplex = RTE_ETH_LINK_HALF_DUPLEX;
+               link.link_status = RTE_ETH_LINK_DOWN;
+               link.link_autoneg = RTE_ETH_LINK_FIXED;
        }
 
        return rte_eth_linkstatus_set(dev, &link);
@@ -484,7 +525,7 @@ eth_igc_interrupt_action(struct rte_eth_dev *dev)
                                " Port %d: Link Up - speed %u Mbps - %s",
                                dev->data->port_id,
                                (unsigned int)link.link_speed,
-                               link.link_duplex == ETH_LINK_FULL_DUPLEX ?
+                               link.link_duplex == RTE_ETH_LINK_FULL_DUPLEX ?
                                "full-duplex" : "half-duplex");
                else
                        PMD_DRV_LOG(INFO, " Port %d: Link Down",
@@ -495,8 +536,7 @@ eth_igc_interrupt_action(struct rte_eth_dev *dev)
                                pci_dev->addr.bus,
                                pci_dev->addr.devid,
                                pci_dev->addr.function);
-               _rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC,
-                               NULL);
+               rte_eth_dev_callback_process(dev, RTE_ETH_EVENT_INTR_LSC, NULL);
        }
 }
 
@@ -563,15 +603,16 @@ eth_igc_rxtx_control(struct rte_eth_dev *dev, bool enable)
  *  This routine disables all traffic on the adapter by issuing a
  *  global reset on the MAC.
  */
-static void
+static int
 eth_igc_stop(struct rte_eth_dev *dev)
 {
        struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
        struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
        struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
        struct rte_eth_link link;
 
+       dev->data->dev_started = 0;
        adapter->stopped = 1;
 
        /* disable receive and transmit */
@@ -596,6 +637,9 @@ eth_igc_stop(struct rte_eth_dev *dev)
        /* disable all wake up */
        IGC_WRITE_REG(hw, IGC_WUC, 0);
 
+       /* disable checking EEE operation in MAC loopback mode */
+       igc_read_reg_check_clear_bits(hw, IGC_EEER, IGC_EEER_EEE_FRC_AN);
+
        /* Set bit for Go Link disconnect */
        igc_read_reg_check_set_bits(hw, IGC_82580_PHY_POWER_MGMT,
                        IGC_82580_PM_GO_LINKD);
@@ -617,6 +661,55 @@ eth_igc_stop(struct rte_eth_dev *dev)
 
        /* Clean datapath event and queue/vec mapping */
        rte_intr_efd_disable(intr_handle);
+       rte_intr_vec_list_free(intr_handle);
+
+       return 0;
+}
+
+/*
+ * write interrupt vector allocation register
+ * @hw
+ *  board private structure
+ * @queue_index
+ *  queue index, valid 0,1,2,3
+ * @tx
+ *  tx:1, rx:0
+ * @msix_vector
+ *  msix-vector, valid 0,1,2,3,4
+ */
+static void
+igc_write_ivar(struct igc_hw *hw, uint8_t queue_index,
+               bool tx, uint8_t msix_vector)
+{
+       uint8_t offset = 0;
+       uint8_t reg_index = queue_index >> 1;
+       uint32_t val;
+
+       /*
+        * IVAR(0)
+        * bit31...24   bit23...16      bit15...8       bit7...0
+        * TX1          RX1             TX0             RX0
+        *
+        * IVAR(1)
+        * bit31...24   bit23...16      bit15...8       bit7...0
+        * TX3          RX3             TX2             RX2
+        */
+
+       if (tx)
+               offset = 8;
+
+       if (queue_index & 1)
+               offset += 16;
+
+       val = IGC_READ_REG_ARRAY(hw, IGC_IVAR0, reg_index);
+
+       /* clear bits */
+       val &= ~((uint32_t)0xFF << offset);
+
+       /* write vector and valid bit */
+       val |= (uint32_t)(msix_vector | IGC_IVAR_VALID) << offset;
+
+       IGC_WRITE_REG_ARRAY(hw, IGC_IVAR0, reg_index, val);
 }
 
 /* Sets up the hardware to generate MSI-X interrupts properly
@@ -628,23 +721,39 @@ igc_configure_msix_intr(struct rte_eth_dev *dev)
 {
        struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
        struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
 
        uint32_t intr_mask;
+       uint32_t vec = IGC_MISC_VEC_ID;
+       uint32_t base = IGC_MISC_VEC_ID;
+       uint32_t misc_shift = 0;
+       int i, nb_efd;
 
        /* won't configure msix register if no mapping is done
         * between intr vector and event fd
         */
-       if (!rte_intr_dp_is_en(intr_handle) ||
-               !dev->data->dev_conf.intr_conf.lsc)
+       if (!rte_intr_dp_is_en(intr_handle))
                return;
 
+       if (rte_intr_allow_others(intr_handle)) {
+               base = IGC_RX_VEC_START;
+               vec = base;
+               misc_shift = 1;
+       }
+
        /* turn on MSI-X capability first */
        IGC_WRITE_REG(hw, IGC_GPIE, IGC_GPIE_MSIX_MODE |
                                IGC_GPIE_PBA | IGC_GPIE_EIAME |
                                IGC_GPIE_NSICR);
 
-       intr_mask = (1u << IGC_MSIX_OTHER_INTR_VEC);
+       nb_efd = rte_intr_nb_efd_get(intr_handle);
+       if (nb_efd < 0)
+               return;
+
+       intr_mask = RTE_LEN2MASK(nb_efd, uint32_t) << misc_shift;
+
+       if (dev->data->dev_conf.intr_conf.lsc)
+               intr_mask |= (1u << IGC_MSIX_OTHER_INTR_VEC);
 
        /* enable msix auto-clear */
        igc_read_reg_check_set_bits(hw, IGC_EIAC, intr_mask);
@@ -656,6 +765,13 @@ igc_configure_msix_intr(struct rte_eth_dev *dev)
        /* enable auto-mask */
        igc_read_reg_check_set_bits(hw, IGC_EIAM, intr_mask);
 
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               igc_write_ivar(hw, i, 0, vec);
+               rte_intr_vec_list_index_set(intr_handle, i, vec);
+               if (vec < base + rte_intr_nb_efd_get(intr_handle) - 1)
+                       vec++;
+       }
+
        IGC_WRITE_FLUSH(hw);
 }
 
@@ -678,6 +794,34 @@ igc_lsc_interrupt_setup(struct rte_eth_dev *dev, uint8_t on)
                intr->mask &= ~IGC_ICR_LSC;
 }
 
+/*
+ * It enables the interrupt.
+ * It will be called once only during nic initialized.
+ */
+static void
+igc_rxq_interrupt_setup(struct rte_eth_dev *dev)
+{
+       uint32_t mask;
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
+       int misc_shift = rte_intr_allow_others(intr_handle) ? 1 : 0;
+       int nb_efd;
+
+       /* won't configure msix register if no mapping is done
+        * between intr vector and event fd
+        */
+       if (!rte_intr_dp_is_en(intr_handle))
+               return;
+
+       nb_efd = rte_intr_nb_efd_get(intr_handle);
+       if (nb_efd < 0)
+               return;
+
+       mask = RTE_LEN2MASK(nb_efd, uint32_t) << misc_shift;
+       IGC_WRITE_REG(hw, IGC_EIMS, mask);
+}
+
 /*
  *  Get hardware rx-buffer size.
  */
@@ -768,7 +912,7 @@ eth_igc_start(struct rte_eth_dev *dev)
        struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
        struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
        struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
-       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
        uint32_t *speeds;
        int ret;
 
@@ -798,7 +942,25 @@ eth_igc_start(struct rte_eth_dev *dev)
        }
        adapter->stopped = 0;
 
-       /* confiugre msix for rx interrupt */
+       /* check and configure queue intr-vector mapping */
+       if (rte_intr_cap_multiple(intr_handle) &&
+               dev->data->dev_conf.intr_conf.rxq) {
+               uint32_t intr_vector = dev->data->nb_rx_queues;
+               if (rte_intr_efd_enable(intr_handle, intr_vector))
+                       return -1;
+       }
+
+       if (rte_intr_dp_is_en(intr_handle)) {
+               if (rte_intr_vec_list_alloc(intr_handle, "intr_vec",
+                                                 dev->data->nb_rx_queues)) {
+                       PMD_DRV_LOG(ERR,
+                               "Failed to allocate %d rx_queues intr_vec",
+                               dev->data->nb_rx_queues);
+                       return -ENOMEM;
+               }
+       }
+
+       /* configure msix for rx interrupt */
        igc_configure_msix_intr(dev);
 
        igc_tx_init(dev);
@@ -813,62 +975,61 @@ eth_igc_start(struct rte_eth_dev *dev)
 
        igc_clear_hw_cntrs_base_generic(hw);
 
+       /* VLAN Offload Settings */
+       eth_igc_vlan_offload_set(dev,
+               RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
+               RTE_ETH_VLAN_EXTEND_MASK);
+
        /* Setup link speed and duplex */
        speeds = &dev->data->dev_conf.link_speeds;
-       if (*speeds == ETH_LINK_SPEED_AUTONEG) {
+       if (*speeds == RTE_ETH_LINK_SPEED_AUTONEG) {
                hw->phy.autoneg_advertised = IGC_ALL_SPEED_DUPLEX_2500;
                hw->mac.autoneg = 1;
        } else {
                int num_speeds = 0;
-               bool autoneg = (*speeds & ETH_LINK_SPEED_FIXED) == 0;
 
-               /* Reset */
+               if (*speeds & RTE_ETH_LINK_SPEED_FIXED) {
+                       PMD_DRV_LOG(ERR,
+                                   "Force speed mode currently not supported");
+                       igc_dev_clear_queues(dev);
+                       return -EINVAL;
+               }
+
                hw->phy.autoneg_advertised = 0;
+               hw->mac.autoneg = 1;
 
-               if (*speeds & ~(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 |
-                               ETH_LINK_SPEED_FIXED)) {
+               if (*speeds & ~(RTE_ETH_LINK_SPEED_10M_HD | RTE_ETH_LINK_SPEED_10M |
+                               RTE_ETH_LINK_SPEED_100M_HD | RTE_ETH_LINK_SPEED_100M |
+                               RTE_ETH_LINK_SPEED_1G | RTE_ETH_LINK_SPEED_2_5G)) {
                        num_speeds = -1;
                        goto error_invalid_config;
                }
-               if (*speeds & ETH_LINK_SPEED_10M_HD) {
+               if (*speeds & RTE_ETH_LINK_SPEED_10M_HD) {
                        hw->phy.autoneg_advertised |= ADVERTISE_10_HALF;
                        num_speeds++;
                }
-               if (*speeds & ETH_LINK_SPEED_10M) {
+               if (*speeds & RTE_ETH_LINK_SPEED_10M) {
                        hw->phy.autoneg_advertised |= ADVERTISE_10_FULL;
                        num_speeds++;
                }
-               if (*speeds & ETH_LINK_SPEED_100M_HD) {
+               if (*speeds & RTE_ETH_LINK_SPEED_100M_HD) {
                        hw->phy.autoneg_advertised |= ADVERTISE_100_HALF;
                        num_speeds++;
                }
-               if (*speeds & ETH_LINK_SPEED_100M) {
+               if (*speeds & RTE_ETH_LINK_SPEED_100M) {
                        hw->phy.autoneg_advertised |= ADVERTISE_100_FULL;
                        num_speeds++;
                }
-               if (*speeds & ETH_LINK_SPEED_1G) {
+               if (*speeds & RTE_ETH_LINK_SPEED_1G) {
                        hw->phy.autoneg_advertised |= ADVERTISE_1000_FULL;
                        num_speeds++;
                }
-               if (*speeds & ETH_LINK_SPEED_2_5G) {
+               if (*speeds & RTE_ETH_LINK_SPEED_2_5G) {
                        hw->phy.autoneg_advertised |= ADVERTISE_2500_FULL;
                        num_speeds++;
                }
-               if (num_speeds == 0 || (!autoneg && num_speeds > 1))
+               if (num_speeds == 0)
                        goto error_invalid_config;
-
-               /* Set/reset the mac.autoneg based on the link speed,
-                * fixed or not
-                */
-               if (!autoneg) {
-                       hw->mac.autoneg = 0;
-                       hw->mac.forced_speed_duplex =
-                                       hw->phy.autoneg_advertised;
-               } else {
-                       hw->mac.autoneg = 1;
-               }
        }
 
        igc_setup_link(hw);
@@ -894,12 +1055,30 @@ eth_igc_start(struct rte_eth_dev *dev)
        rte_eal_alarm_set(IGC_ALARM_INTERVAL,
                        igc_update_queue_stats_handler, dev);
 
+       /* check if rxq interrupt is enabled */
+       if (dev->data->dev_conf.intr_conf.rxq &&
+                       rte_intr_dp_is_en(intr_handle))
+               igc_rxq_interrupt_setup(dev);
+
        /* resume enabled intr since hw reset */
        igc_intr_other_enable(dev);
 
        eth_igc_rxtx_control(dev, true);
        eth_igc_link_update(dev, 0);
 
+       /* configure MAC-loopback mode */
+       if (dev->data->dev_conf.lpbk_mode == 1) {
+               uint32_t reg_val;
+
+               reg_val = IGC_READ_REG(hw, IGC_CTRL);
+               reg_val &= ~IGC_CTRL_SPEED_MASK;
+               reg_val |= IGC_CTRL_SLU | IGC_CTRL_FRCSPD |
+                       IGC_CTRL_FRCDPX | IGC_CTRL_FD | IGC_CTRL_SPEED_2500;
+               IGC_WRITE_REG(hw, IGC_CTRL, reg_val);
+
+               igc_read_reg_check_set_bits(hw, IGC_EEER, IGC_EEER_EEE_FRC_AN);
+       }
+
        return 0;
 
 error_invalid_config:
@@ -972,31 +1151,37 @@ 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]);
+               eth_igc_rx_queue_release(dev, 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]);
+               eth_igc_tx_queue_release(dev, i);
                dev->data->tx_queues[i] = NULL;
        }
        dev->data->nb_tx_queues = 0;
 }
 
-static void
+static int
 eth_igc_close(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;
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
        struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
        struct igc_adapter *adapter = IGC_DEV_PRIVATE(dev);
        int retry = 0;
+       int ret = 0;
 
        PMD_INIT_FUNC_TRACE();
+       if (rte_eal_process_type() != RTE_PROC_PRIMARY)
+               return 0;
 
        if (!adapter->stopped)
-               eth_igc_stop(dev);
+               ret = eth_igc_stop(dev);
+
+       igc_flow_flush(dev, NULL);
+       igc_clear_all_filter(dev);
 
        igc_intr_other_disable(dev);
        do {
@@ -1015,6 +1200,8 @@ eth_igc_close(struct rte_eth_dev *dev)
 
        /* Reset any pending lock */
        igc_reset_swfw_lock(hw);
+
+       return ret;
 }
 
 static void
@@ -1038,6 +1225,9 @@ eth_igc_dev_init(struct rte_eth_dev *dev)
 
        PMD_INIT_FUNC_TRACE();
        dev->dev_ops = &eth_igc_ops;
+       dev->rx_queue_count = eth_igc_rx_queue_count;
+       dev->rx_descriptor_status = eth_igc_rx_descriptor_status;
+       dev->tx_descriptor_status = eth_igc_tx_descriptor_status;
 
        /*
         * for secondary processes, we don't initialize any further as primary
@@ -1048,6 +1238,7 @@ eth_igc_dev_init(struct rte_eth_dev *dev)
                return 0;
 
        rte_eth_copy_pci_info(dev, pci_dev);
+       dev->data->dev_flags |= RTE_ETH_DEV_AUTOFILL_QUEUE_XSTATS;
 
        hw->back = pci_dev;
        hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
@@ -1133,11 +1324,6 @@ eth_igc_dev_init(struct rte_eth_dev *dev)
                goto err_late;
        }
 
-       /* Pass the information to the rte_eth_dev_close() that it should also
-        * release the private port resources.
-        */
-       dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
-
        hw->mac.get_link_status = 1;
        igc->stopped = 0;
 
@@ -1150,11 +1336,11 @@ eth_igc_dev_init(struct rte_eth_dev *dev)
                        dev->data->port_id, pci_dev->id.vendor_id,
                        pci_dev->id.device_id);
 
-       rte_intr_callback_register(&pci_dev->intr_handle,
+       rte_intr_callback_register(pci_dev->intr_handle,
                        eth_igc_interrupt_handler, (void *)dev);
 
        /* enable uio/vfio intr/eventfd mapping */
-       rte_intr_enable(&pci_dev->intr_handle);
+       rte_intr_enable(pci_dev->intr_handle);
 
        /* enable support intr */
        igc_intr_other_enable(dev);
@@ -1164,6 +1350,9 @@ eth_igc_dev_init(struct rte_eth_dev *dev)
                igc->txq_stats_map[i] = -1;
                igc->rxq_stats_map[i] = -1;
        }
+
+       igc_flow_init(dev);
+       igc_clear_all_filter(dev);
        return 0;
 
 err_late:
@@ -1175,10 +1364,6 @@ static int
 eth_igc_dev_uninit(__rte_unused struct rte_eth_dev *eth_dev)
 {
        PMD_INIT_FUNC_TRACE();
-
-       if (rte_eal_process_type() != RTE_PROC_PRIMARY)
-               return 0;
-
        eth_igc_close(eth_dev);
        return 0;
 }
@@ -1282,9 +1467,11 @@ eth_igc_fw_version_get(struct rte_eth_dev *dev, char *fw_version,
                                 fw.eep_build);
                }
        }
+       if (ret < 0)
+               return -EINVAL;
 
        ret += 1; /* add the size of '\0' */
-       if (fw_size < (u32)ret)
+       if (fw_size < (size_t)ret)
                return ret;
        else
                return 0;
@@ -1298,15 +1485,17 @@ 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->dev_capa &= ~RTE_ETH_DEV_CAPA_FLOW_RULE_KEEP;
        dev_info->rx_offload_capa = IGC_RX_OFFLOAD_ALL;
        dev_info->tx_offload_capa = IGC_TX_OFFLOAD_ALL;
+       dev_info->rx_queue_offload_capa = RTE_ETH_RX_OFFLOAD_VLAN_STRIP;
 
        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->reta_size = RTE_ETH_RSS_RETA_SIZE_128;
        dev_info->flow_type_rss_offloads = IGC_RSS_OFFLOAD_ALL;
 
        dev_info->default_rxconf = (struct rte_eth_rxconf) {
@@ -1332,9 +1521,9 @@ eth_igc_infos_get(struct rte_eth_dev *dev, struct rte_eth_dev_info *dev_info)
        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;
+       dev_info->speed_capa = RTE_ETH_LINK_SPEED_10M_HD | RTE_ETH_LINK_SPEED_10M |
+                       RTE_ETH_LINK_SPEED_100M_HD | RTE_ETH_LINK_SPEED_100M |
+                       RTE_ETH_LINK_SPEED_1G | RTE_ETH_LINK_SPEED_2_5G;
 
        dev_info->max_mtu = dev_info->max_rx_pktlen - IGC_ETH_OVERHEAD;
        dev_info->min_mtu = RTE_ETHER_MIN_MTU;
@@ -1392,38 +1581,24 @@ eth_igc_mtu_set(struct rte_eth_dev *dev, uint16_t mtu)
        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 device is started, 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)
+       if (dev->data->dev_started && !dev->data->scattered_rx &&
+           frame_size > dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM) {
+               PMD_INIT_LOG(ERR, "Stop port first.");
                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;
+       if (mtu > RTE_ETHER_MTU)
                rctl |= IGC_RCTL_LPE;
-       } else {
-               dev->data->dev_conf.rxmode.offloads &=
-                       ~DEV_RX_OFFLOAD_JUMBO_FRAME;
+       else
                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);
+       IGC_WRITE_REG(hw, IGC_RLPML, frame_size);
 
        return 0;
 }
@@ -1707,8 +1882,7 @@ eth_igc_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *rte_stats)
 
        /* Rx Errors */
        rte_stats->imissed = stats->mpc;
-       rte_stats->ierrors = stats->crcerrs +
-                       stats->rlec + stats->ruc + stats->roc +
+       rte_stats->ierrors = stats->crcerrs + stats->rlec +
                        stats->rxerrc + stats->algnerrc;
 
        /* Tx Errors */
@@ -1821,7 +1995,7 @@ eth_igc_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
 
 static int
 eth_igc_xstats_get_names_by_id(struct rte_eth_dev *dev,
-               struct rte_eth_xstat_name *xstats_names, const uint64_t *ids,
+               const uint64_t *ids, struct rte_eth_xstat_name *xstats_names,
                unsigned int limit)
 {
        unsigned int i;
@@ -1903,6 +2077,496 @@ eth_igc_queue_stats_mapping_set(struct rte_eth_dev *dev,
        return 0;
 }
 
+static int
+eth_igc_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(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 vec = IGC_MISC_VEC_ID;
+
+       if (rte_intr_allow_others(intr_handle))
+               vec = IGC_RX_VEC_START;
+
+       uint32_t mask = 1u << (queue_id + vec);
+
+       IGC_WRITE_REG(hw, IGC_EIMC, mask);
+       IGC_WRITE_FLUSH(hw);
+
+       return 0;
+}
+
+static int
+eth_igc_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(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 vec = IGC_MISC_VEC_ID;
+
+       if (rte_intr_allow_others(intr_handle))
+               vec = IGC_RX_VEC_START;
+
+       uint32_t mask = 1u << (queue_id + vec);
+
+       IGC_WRITE_REG(hw, IGC_EIMS, mask);
+       IGC_WRITE_FLUSH(hw);
+
+       rte_intr_enable(intr_handle);
+
+       return 0;
+}
+
+static int
+eth_igc_flow_ctrl_get(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t ctrl;
+       int tx_pause;
+       int rx_pause;
+
+       fc_conf->pause_time = hw->fc.pause_time;
+       fc_conf->high_water = hw->fc.high_water;
+       fc_conf->low_water = hw->fc.low_water;
+       fc_conf->send_xon = hw->fc.send_xon;
+       fc_conf->autoneg = hw->mac.autoneg;
+
+       /*
+        * Return rx_pause and tx_pause status according to actual setting of
+        * the TFCE and RFCE bits in the CTRL register.
+        */
+       ctrl = IGC_READ_REG(hw, IGC_CTRL);
+       if (ctrl & IGC_CTRL_TFCE)
+               tx_pause = 1;
+       else
+               tx_pause = 0;
+
+       if (ctrl & IGC_CTRL_RFCE)
+               rx_pause = 1;
+       else
+               rx_pause = 0;
+
+       if (rx_pause && tx_pause)
+               fc_conf->mode = RTE_ETH_FC_FULL;
+       else if (rx_pause)
+               fc_conf->mode = RTE_ETH_FC_RX_PAUSE;
+       else if (tx_pause)
+               fc_conf->mode = RTE_ETH_FC_TX_PAUSE;
+       else
+               fc_conf->mode = RTE_ETH_FC_NONE;
+
+       return 0;
+}
+
+static int
+eth_igc_flow_ctrl_set(struct rte_eth_dev *dev, struct rte_eth_fc_conf *fc_conf)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t rx_buf_size;
+       uint32_t max_high_water;
+       uint32_t rctl;
+       int err;
+
+       if (fc_conf->autoneg != hw->mac.autoneg)
+               return -ENOTSUP;
+
+       rx_buf_size = igc_get_rx_buffer_size(hw);
+       PMD_DRV_LOG(DEBUG, "Rx packet buffer size = 0x%x", rx_buf_size);
+
+       /* At least reserve one Ethernet frame for watermark */
+       max_high_water = rx_buf_size - RTE_ETHER_MAX_LEN;
+       if (fc_conf->high_water > max_high_water ||
+               fc_conf->high_water < fc_conf->low_water) {
+               PMD_DRV_LOG(ERR,
+                       "Incorrect high(%u)/low(%u) water value, max is %u",
+                       fc_conf->high_water, fc_conf->low_water,
+                       max_high_water);
+               return -EINVAL;
+       }
+
+       switch (fc_conf->mode) {
+       case RTE_ETH_FC_NONE:
+               hw->fc.requested_mode = igc_fc_none;
+               break;
+       case RTE_ETH_FC_RX_PAUSE:
+               hw->fc.requested_mode = igc_fc_rx_pause;
+               break;
+       case RTE_ETH_FC_TX_PAUSE:
+               hw->fc.requested_mode = igc_fc_tx_pause;
+               break;
+       case RTE_ETH_FC_FULL:
+               hw->fc.requested_mode = igc_fc_full;
+               break;
+       default:
+               PMD_DRV_LOG(ERR, "unsupported fc mode: %u", fc_conf->mode);
+               return -EINVAL;
+       }
+
+       hw->fc.pause_time     = fc_conf->pause_time;
+       hw->fc.high_water     = fc_conf->high_water;
+       hw->fc.low_water      = fc_conf->low_water;
+       hw->fc.send_xon       = fc_conf->send_xon;
+
+       err = igc_setup_link_generic(hw);
+       if (err == IGC_SUCCESS) {
+               /**
+                * check if we want to forward MAC frames - driver doesn't have
+                * native capability to do that, so we'll write the registers
+                * ourselves
+                **/
+               rctl = IGC_READ_REG(hw, IGC_RCTL);
+
+               /* set or clear MFLCN.PMCF bit depending on configuration */
+               if (fc_conf->mac_ctrl_frame_fwd != 0)
+                       rctl |= IGC_RCTL_PMCF;
+               else
+                       rctl &= ~IGC_RCTL_PMCF;
+
+               IGC_WRITE_REG(hw, IGC_RCTL, rctl);
+               IGC_WRITE_FLUSH(hw);
+
+               return 0;
+       }
+
+       PMD_DRV_LOG(ERR, "igc_setup_link_generic = 0x%x", err);
+       return -EIO;
+}
+
+static int
+eth_igc_rss_reta_update(struct rte_eth_dev *dev,
+                       struct rte_eth_rss_reta_entry64 *reta_conf,
+                       uint16_t reta_size)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint16_t i;
+
+       if (reta_size != RTE_ETH_RSS_RETA_SIZE_128) {
+               PMD_DRV_LOG(ERR,
+                       "The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
+                       reta_size, RTE_ETH_RSS_RETA_SIZE_128);
+               return -EINVAL;
+       }
+
+       RTE_BUILD_BUG_ON(RTE_ETH_RSS_RETA_SIZE_128 % IGC_RSS_RDT_REG_SIZE);
+
+       /* set redirection table */
+       for (i = 0; i < RTE_ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
+               union igc_rss_reta_reg reta, reg;
+               uint16_t idx, shift;
+               uint8_t j, mask;
+
+               idx = i / RTE_ETH_RETA_GROUP_SIZE;
+               shift = i % RTE_ETH_RETA_GROUP_SIZE;
+               mask = (uint8_t)((reta_conf[idx].mask >> shift) &
+                               IGC_RSS_RDT_REG_SIZE_MASK);
+
+               /* if no need to update the register */
+               if (!mask ||
+                   shift > (RTE_ETH_RETA_GROUP_SIZE - IGC_RSS_RDT_REG_SIZE))
+                       continue;
+
+               /* check mask whether need to read the register value first */
+               if (mask == IGC_RSS_RDT_REG_SIZE_MASK)
+                       reg.dword = 0;
+               else
+                       reg.dword = IGC_READ_REG_LE_VALUE(hw,
+                                       IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
+
+               /* update the register */
+               RTE_BUILD_BUG_ON(sizeof(reta.bytes) != IGC_RSS_RDT_REG_SIZE);
+               for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
+                       if (mask & (1u << j))
+                               reta.bytes[j] =
+                                       (uint8_t)reta_conf[idx].reta[shift + j];
+                       else
+                               reta.bytes[j] = reg.bytes[j];
+               }
+               IGC_WRITE_REG_LE_VALUE(hw,
+                       IGC_RETA(i / IGC_RSS_RDT_REG_SIZE), reta.dword);
+       }
+
+       return 0;
+}
+
+static int
+eth_igc_rss_reta_query(struct rte_eth_dev *dev,
+                      struct rte_eth_rss_reta_entry64 *reta_conf,
+                      uint16_t reta_size)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint16_t i;
+
+       if (reta_size != RTE_ETH_RSS_RETA_SIZE_128) {
+               PMD_DRV_LOG(ERR,
+                       "The size of RSS redirection table configured(%d) doesn't match the number hardware can supported(%d)",
+                       reta_size, RTE_ETH_RSS_RETA_SIZE_128);
+               return -EINVAL;
+       }
+
+       RTE_BUILD_BUG_ON(RTE_ETH_RSS_RETA_SIZE_128 % IGC_RSS_RDT_REG_SIZE);
+
+       /* read redirection table */
+       for (i = 0; i < RTE_ETH_RSS_RETA_SIZE_128; i += IGC_RSS_RDT_REG_SIZE) {
+               union igc_rss_reta_reg reta;
+               uint16_t idx, shift;
+               uint8_t j, mask;
+
+               idx = i / RTE_ETH_RETA_GROUP_SIZE;
+               shift = i % RTE_ETH_RETA_GROUP_SIZE;
+               mask = (uint8_t)((reta_conf[idx].mask >> shift) &
+                               IGC_RSS_RDT_REG_SIZE_MASK);
+
+               /* if no need to read register */
+               if (!mask ||
+                   shift > (RTE_ETH_RETA_GROUP_SIZE - IGC_RSS_RDT_REG_SIZE))
+                       continue;
+
+               /* read register and get the queue index */
+               RTE_BUILD_BUG_ON(sizeof(reta.bytes) != IGC_RSS_RDT_REG_SIZE);
+               reta.dword = IGC_READ_REG_LE_VALUE(hw,
+                               IGC_RETA(i / IGC_RSS_RDT_REG_SIZE));
+               for (j = 0; j < IGC_RSS_RDT_REG_SIZE; j++) {
+                       if (mask & (1u << j))
+                               reta_conf[idx].reta[shift + j] = reta.bytes[j];
+               }
+       }
+
+       return 0;
+}
+
+static int
+eth_igc_rss_hash_update(struct rte_eth_dev *dev,
+                       struct rte_eth_rss_conf *rss_conf)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       igc_hw_rss_hash_set(hw, rss_conf);
+       return 0;
+}
+
+static int
+eth_igc_rss_hash_conf_get(struct rte_eth_dev *dev,
+                       struct rte_eth_rss_conf *rss_conf)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t *hash_key = (uint32_t *)rss_conf->rss_key;
+       uint32_t mrqc;
+       uint64_t rss_hf;
+
+       if (hash_key != NULL) {
+               int i;
+
+               /* if not enough space for store hash key */
+               if (rss_conf->rss_key_len != IGC_HKEY_SIZE) {
+                       PMD_DRV_LOG(ERR,
+                               "RSS hash key size %u in parameter doesn't match the hardware hash key size %u",
+                               rss_conf->rss_key_len, IGC_HKEY_SIZE);
+                       return -EINVAL;
+               }
+
+               /* read RSS key from register */
+               for (i = 0; i < IGC_HKEY_MAX_INDEX; i++)
+                       hash_key[i] = IGC_READ_REG_LE_VALUE(hw, IGC_RSSRK(i));
+       }
+
+       /* get RSS functions configured in MRQC register */
+       mrqc = IGC_READ_REG(hw, IGC_MRQC);
+       if ((mrqc & IGC_MRQC_ENABLE_RSS_4Q) == 0)
+               return 0;
+
+       rss_hf = 0;
+       if (mrqc & IGC_MRQC_RSS_FIELD_IPV4)
+               rss_hf |= RTE_ETH_RSS_IPV4;
+       if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_TCP)
+               rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_TCP;
+       if (mrqc & IGC_MRQC_RSS_FIELD_IPV6)
+               rss_hf |= RTE_ETH_RSS_IPV6;
+       if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_EX)
+               rss_hf |= RTE_ETH_RSS_IPV6_EX;
+       if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP)
+               rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_TCP;
+       if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_TCP_EX)
+               rss_hf |= RTE_ETH_RSS_IPV6_TCP_EX;
+       if (mrqc & IGC_MRQC_RSS_FIELD_IPV4_UDP)
+               rss_hf |= RTE_ETH_RSS_NONFRAG_IPV4_UDP;
+       if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP)
+               rss_hf |= RTE_ETH_RSS_NONFRAG_IPV6_UDP;
+       if (mrqc & IGC_MRQC_RSS_FIELD_IPV6_UDP_EX)
+               rss_hf |= RTE_ETH_RSS_IPV6_UDP_EX;
+
+       rss_conf->rss_hf |= rss_hf;
+       return 0;
+}
+
+static int
+eth_igc_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
+       uint32_t vfta;
+       uint32_t vid_idx;
+       uint32_t vid_bit;
+
+       vid_idx = (vlan_id >> IGC_VFTA_ENTRY_SHIFT) & IGC_VFTA_ENTRY_MASK;
+       vid_bit = 1u << (vlan_id & IGC_VFTA_ENTRY_BIT_SHIFT_MASK);
+       vfta = shadow_vfta->vfta[vid_idx];
+       if (on)
+               vfta |= vid_bit;
+       else
+               vfta &= ~vid_bit;
+       IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, vid_idx, vfta);
+
+       /* update local VFTA copy */
+       shadow_vfta->vfta[vid_idx] = vfta;
+
+       return 0;
+}
+
+static void
+igc_vlan_hw_filter_disable(struct rte_eth_dev *dev)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       igc_read_reg_check_clear_bits(hw, IGC_RCTL,
+                       IGC_RCTL_CFIEN | IGC_RCTL_VFE);
+}
+
+static void
+igc_vlan_hw_filter_enable(struct rte_eth_dev *dev)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       struct igc_vfta *shadow_vfta = IGC_DEV_PRIVATE_VFTA(dev);
+       uint32_t reg_val;
+       int i;
+
+       /* Filter Table Enable, CFI not used for packet acceptance */
+       reg_val = IGC_READ_REG(hw, IGC_RCTL);
+       reg_val &= ~IGC_RCTL_CFIEN;
+       reg_val |= IGC_RCTL_VFE;
+       IGC_WRITE_REG(hw, IGC_RCTL, reg_val);
+
+       /* restore VFTA table */
+       for (i = 0; i < IGC_VFTA_SIZE; i++)
+               IGC_WRITE_REG_ARRAY(hw, IGC_VFTA, i, shadow_vfta->vfta[i]);
+}
+
+static void
+igc_vlan_hw_strip_disable(struct rte_eth_dev *dev)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+
+       igc_read_reg_check_clear_bits(hw, IGC_CTRL, IGC_CTRL_VME);
+}
+
+static void
+igc_vlan_hw_strip_enable(struct rte_eth_dev *dev)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+
+       igc_read_reg_check_set_bits(hw, IGC_CTRL, IGC_CTRL_VME);
+}
+
+static int
+igc_vlan_hw_extend_disable(struct rte_eth_dev *dev)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t frame_size = dev->data->mtu + IGC_ETH_OVERHEAD;
+       uint32_t ctrl_ext;
+
+       ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
+
+       /* if extend vlan hasn't been enabled */
+       if ((ctrl_ext & IGC_CTRL_EXT_EXT_VLAN) == 0)
+               return 0;
+
+       /* Update maximum packet length */
+       if (frame_size < RTE_ETHER_MIN_MTU + VLAN_TAG_SIZE) {
+               PMD_DRV_LOG(ERR, "Maximum packet length %u error, min is %u",
+                       frame_size, VLAN_TAG_SIZE + RTE_ETHER_MIN_MTU);
+               return -EINVAL;
+       }
+       IGC_WRITE_REG(hw, IGC_RLPML, frame_size - VLAN_TAG_SIZE);
+
+       IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext & ~IGC_CTRL_EXT_EXT_VLAN);
+       return 0;
+}
+
+static int
+igc_vlan_hw_extend_enable(struct rte_eth_dev *dev)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t frame_size = dev->data->mtu + IGC_ETH_OVERHEAD;
+       uint32_t ctrl_ext;
+
+       ctrl_ext = IGC_READ_REG(hw, IGC_CTRL_EXT);
+
+       /* if extend vlan has been enabled */
+       if (ctrl_ext & IGC_CTRL_EXT_EXT_VLAN)
+               return 0;
+
+       /* Update maximum packet length */
+       if (frame_size > MAX_RX_JUMBO_FRAME_SIZE) {
+               PMD_DRV_LOG(ERR, "Maximum packet length %u error, max is %u",
+                       frame_size, MAX_RX_JUMBO_FRAME_SIZE);
+               return -EINVAL;
+       }
+       IGC_WRITE_REG(hw, IGC_RLPML, frame_size);
+
+       IGC_WRITE_REG(hw, IGC_CTRL_EXT, ctrl_ext | IGC_CTRL_EXT_EXT_VLAN);
+       return 0;
+}
+
+static int
+eth_igc_vlan_offload_set(struct rte_eth_dev *dev, int mask)
+{
+       struct rte_eth_rxmode *rxmode;
+
+       rxmode = &dev->data->dev_conf.rxmode;
+       if (mask & RTE_ETH_VLAN_STRIP_MASK) {
+               if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP)
+                       igc_vlan_hw_strip_enable(dev);
+               else
+                       igc_vlan_hw_strip_disable(dev);
+       }
+
+       if (mask & RTE_ETH_VLAN_FILTER_MASK) {
+               if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_FILTER)
+                       igc_vlan_hw_filter_enable(dev);
+               else
+                       igc_vlan_hw_filter_disable(dev);
+       }
+
+       if (mask & RTE_ETH_VLAN_EXTEND_MASK) {
+               if (rxmode->offloads & RTE_ETH_RX_OFFLOAD_VLAN_EXTEND)
+                       return igc_vlan_hw_extend_enable(dev);
+               else
+                       return igc_vlan_hw_extend_disable(dev);
+       }
+
+       return 0;
+}
+
+static int
+eth_igc_vlan_tpid_set(struct rte_eth_dev *dev,
+                     enum rte_vlan_type vlan_type,
+                     uint16_t tpid)
+{
+       struct igc_hw *hw = IGC_DEV_PRIVATE_HW(dev);
+       uint32_t reg_val;
+
+       /* only outer TPID of double VLAN can be configured*/
+       if (vlan_type == RTE_ETH_VLAN_TYPE_OUTER) {
+               reg_val = IGC_READ_REG(hw, IGC_VET);
+               reg_val = (reg_val & (~IGC_VET_EXT)) |
+                       ((uint32_t)tpid << IGC_VET_EXT_SHIFT);
+               IGC_WRITE_REG(hw, IGC_VET, reg_val);
+
+               return 0;
+       }
+
+       /* all other TPID values are read-only*/
+       PMD_DRV_LOG(ERR, "Not supported");
+       return -ENOTSUP;
+}
+
 static int
 eth_igc_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
        struct rte_pci_device *pci_dev)