drivers: remove direct access to interrupt handle
[dpdk.git] / drivers / net / txgbe / txgbe_ethdev_vf.c
index 559af5b..4dda55b 100644 (file)
@@ -1,5 +1,6 @@
 /* SPDX-License-Identifier: BSD-3-Clause
- * Copyright(c) 2015-2020
+ * Copyright(c) 2015-2020 Beijing WangXun Technology Co., Ltd.
+ * Copyright(c) 2010-2017 Intel Corporation
  */
 
 #include <sys/queue.h>
 #include <string.h>
 #include <rte_log.h>
 #include <ethdev_pci.h>
+#include <rte_alarm.h>
 
 #include "txgbe_logs.h"
 #include "base/txgbe.h"
 #include "txgbe_ethdev.h"
 #include "txgbe_rxtx.h"
+#include "txgbe_regs_group.h"
+
+static const struct reg_info txgbevf_regs_general[] = {
+       {TXGBE_VFRST, 1, 1, "TXGBE_VFRST"},
+       {TXGBE_VFSTATUS, 1, 1, "TXGBE_VFSTATUS"},
+       {TXGBE_VFMBCTL, 1, 1, "TXGBE_VFMAILBOX"},
+       {TXGBE_VFMBX, 16, 4, "TXGBE_VFMBX"},
+       {TXGBE_VFPBWRAP, 1, 1, "TXGBE_VFPBWRAP"},
+       {0, 0, 0, ""}
+};
+
+static const struct reg_info txgbevf_regs_interrupt[] = {
+       {0, 0, 0, ""}
+};
+
+static const struct reg_info txgbevf_regs_rxdma[] = {
+       {0, 0, 0, ""}
+};
+
+static const struct reg_info txgbevf_regs_tx[] = {
+       {0, 0, 0, ""}
+};
+
+/* VF registers */
+static const struct reg_info *txgbevf_regs[] = {
+                               txgbevf_regs_general,
+                               txgbevf_regs_interrupt,
+                               txgbevf_regs_rxdma,
+                               txgbevf_regs_tx,
+                               NULL};
 
+static int txgbevf_dev_xstats_get(struct rte_eth_dev *dev,
+                                 struct rte_eth_xstat *xstats, unsigned int n);
 static int txgbevf_dev_info_get(struct rte_eth_dev *dev,
                                 struct rte_eth_dev_info *dev_info);
 static int  txgbevf_dev_configure(struct rte_eth_dev *dev);
+static int  txgbevf_dev_start(struct rte_eth_dev *dev);
 static int txgbevf_dev_link_update(struct rte_eth_dev *dev,
                                   int wait_to_complete);
+static int txgbevf_dev_stop(struct rte_eth_dev *dev);
 static int txgbevf_dev_close(struct rte_eth_dev *dev);
 static void txgbevf_intr_disable(struct rte_eth_dev *dev);
 static void txgbevf_intr_enable(struct rte_eth_dev *dev);
+static int txgbevf_dev_stats_reset(struct rte_eth_dev *dev);
+static int txgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask);
+static void txgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on);
 static void txgbevf_configure_msix(struct rte_eth_dev *dev);
+static int txgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev);
+static int txgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev);
 static void txgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index);
 static void txgbevf_dev_interrupt_handler(void *param);
 
@@ -31,8 +72,8 @@ static void txgbevf_dev_interrupt_handler(void *param);
  * The set of PCI devices this driver supports (for VF)
  */
 static const struct rte_pci_id pci_id_txgbevf_map[] = {
-       { RTE_PCI_DEVICE(PCI_VENDOR_ID_WANGXUN, TXGBE_DEV_ID_RAPTOR_VF) },
-       { RTE_PCI_DEVICE(PCI_VENDOR_ID_WANGXUN, TXGBE_DEV_ID_RAPTOR_VF_HV) },
+       { RTE_PCI_DEVICE(PCI_VENDOR_ID_WANGXUN, TXGBE_DEV_ID_SP1000_VF) },
+       { RTE_PCI_DEVICE(PCI_VENDOR_ID_WANGXUN, TXGBE_DEV_ID_WX1820_VF) },
        { .vendor_id = 0, /* sentinel */ },
 };
 
@@ -52,6 +93,28 @@ static const struct rte_eth_desc_lim tx_desc_lim = {
 
 static const struct eth_dev_ops txgbevf_eth_dev_ops;
 
+static const struct rte_txgbe_xstats_name_off rte_txgbevf_stats_strings[] = {
+       {"rx_multicast_packets_0",
+                       offsetof(struct txgbevf_hw_stats, qp[0].vfmprc)},
+       {"rx_multicast_packets_1",
+                       offsetof(struct txgbevf_hw_stats, qp[1].vfmprc)},
+       {"rx_multicast_packets_2",
+                       offsetof(struct txgbevf_hw_stats, qp[2].vfmprc)},
+       {"rx_multicast_packets_3",
+                       offsetof(struct txgbevf_hw_stats, qp[3].vfmprc)},
+       {"rx_multicast_packets_4",
+                       offsetof(struct txgbevf_hw_stats, qp[4].vfmprc)},
+       {"rx_multicast_packets_5",
+                       offsetof(struct txgbevf_hw_stats, qp[5].vfmprc)},
+       {"rx_multicast_packets_6",
+                       offsetof(struct txgbevf_hw_stats, qp[6].vfmprc)},
+       {"rx_multicast_packets_7",
+                       offsetof(struct txgbevf_hw_stats, qp[7].vfmprc)}
+};
+
+#define TXGBEVF_NB_XSTATS (sizeof(rte_txgbevf_stats_strings) / \
+               sizeof(rte_txgbevf_stats_strings[0]))
+
 /*
  * Negotiate mailbox API version with the PF.
  * After reset API version is always set to the basic one (txgbe_mbox_api_10).
@@ -103,8 +166,10 @@ eth_txgbevf_dev_init(struct rte_eth_dev *eth_dev)
        int err;
        uint32_t tc, tcs;
        struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
-       struct rte_intr_handle *intr_handle = &pci_dev->intr_handle;
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
        struct txgbe_hw *hw = TXGBE_DEV_HW(eth_dev);
+       struct txgbe_vfta *shadow_vfta = TXGBE_DEV_VFTA(eth_dev);
+       struct txgbe_hwstrip *hwstrip = TXGBE_DEV_HWSTRIP(eth_dev);
        struct rte_ether_addr *perm_addr =
                        (struct rte_ether_addr *)hw->mac.perm_addr;
 
@@ -148,6 +213,12 @@ eth_txgbevf_dev_init(struct rte_eth_dev *eth_dev)
        hw->subsystem_vendor_id = pci_dev->id.subsystem_vendor_id;
        hw->hw_addr = (void *)pci_dev->mem_resource[0].addr;
 
+       /* initialize the vfta */
+       memset(shadow_vfta, 0, sizeof(*shadow_vfta));
+
+       /* initialize the hw strip bitmap*/
+       memset(hwstrip, 0, sizeof(*hwstrip));
+
        /* Initialize the shared code (base driver) */
        err = txgbe_init_shared_code(hw);
        if (err != 0) {
@@ -159,6 +230,9 @@ eth_txgbevf_dev_init(struct rte_eth_dev *eth_dev)
        /* init_mailbox_params */
        hw->mbx.init_params(hw);
 
+       /* Reset the hw statistics */
+       txgbevf_dev_stats_reset(eth_dev);
+
        /* Disable the interrupts for VF */
        txgbevf_intr_disable(eth_dev);
 
@@ -208,13 +282,8 @@ eth_txgbevf_dev_init(struct rte_eth_dev *eth_dev)
                }
                PMD_INIT_LOG(INFO, "\tVF MAC address not assigned by Host PF");
                PMD_INIT_LOG(INFO, "\tAssign randomly generated MAC address "
-                            "%02x:%02x:%02x:%02x:%02x:%02x",
-                            perm_addr->addr_bytes[0],
-                            perm_addr->addr_bytes[1],
-                            perm_addr->addr_bytes[2],
-                            perm_addr->addr_bytes[3],
-                            perm_addr->addr_bytes[4],
-                            perm_addr->addr_bytes[5]);
+                            RTE_ETHER_ADDR_PRT_FMT,
+                                RTE_ETHER_ADDR_BYTES(perm_addr));
        }
 
        /* Copy the permanent MAC address */
@@ -227,6 +296,9 @@ eth_txgbevf_dev_init(struct rte_eth_dev *eth_dev)
                return -EIO;
        }
 
+       /* enter promiscuous mode */
+       txgbevf_dev_promiscuous_enable(eth_dev);
+
        rte_intr_callback_register(intr_handle,
                                   txgbevf_dev_interrupt_handler, eth_dev);
        rte_intr_enable(intr_handle);
@@ -275,6 +347,131 @@ static struct rte_pci_driver rte_txgbevf_pmd = {
        .remove = eth_txgbevf_pci_remove,
 };
 
+static int txgbevf_dev_xstats_get_names(__rte_unused struct rte_eth_dev *dev,
+       struct rte_eth_xstat_name *xstats_names, unsigned int limit)
+{
+       unsigned int i;
+
+       if (limit < TXGBEVF_NB_XSTATS && xstats_names != NULL)
+               return -ENOMEM;
+
+       if (xstats_names != NULL)
+               for (i = 0; i < TXGBEVF_NB_XSTATS; i++)
+                       snprintf(xstats_names[i].name,
+                               sizeof(xstats_names[i].name),
+                               "%s", rte_txgbevf_stats_strings[i].name);
+       return TXGBEVF_NB_XSTATS;
+}
+
+static void
+txgbevf_update_stats(struct rte_eth_dev *dev)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       struct txgbevf_hw_stats *hw_stats = (struct txgbevf_hw_stats *)
+                         TXGBE_DEV_STATS(dev);
+       unsigned int i;
+
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               /* Good Rx packet, include VF loopback */
+               TXGBE_UPDCNT32(TXGBE_QPRXPKT(i),
+               hw_stats->qp[i].last_vfgprc, hw_stats->qp[i].vfgprc);
+
+               /* Good Rx octets, include VF loopback */
+               TXGBE_UPDCNT36(TXGBE_QPRXOCTL(i),
+               hw_stats->qp[i].last_vfgorc, hw_stats->qp[i].vfgorc);
+
+               /* Rx Multicst Packet */
+               TXGBE_UPDCNT32(TXGBE_QPRXMPKT(i),
+               hw_stats->qp[i].last_vfmprc, hw_stats->qp[i].vfmprc);
+       }
+       hw->rx_loaded = 0;
+
+       for (i = 0; i < dev->data->nb_tx_queues; i++) {
+               /* Good Tx packet, include VF loopback */
+               TXGBE_UPDCNT32(TXGBE_QPTXPKT(i),
+               hw_stats->qp[i].last_vfgptc, hw_stats->qp[i].vfgptc);
+
+               /* Good Tx octets, include VF loopback */
+               TXGBE_UPDCNT36(TXGBE_QPTXOCTL(i),
+               hw_stats->qp[i].last_vfgotc, hw_stats->qp[i].vfgotc);
+       }
+       hw->offset_loaded = 0;
+}
+
+static int
+txgbevf_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
+                      unsigned int n)
+{
+       struct txgbevf_hw_stats *hw_stats = (struct txgbevf_hw_stats *)
+                       TXGBE_DEV_STATS(dev);
+       unsigned int i;
+
+       if (n < TXGBEVF_NB_XSTATS)
+               return TXGBEVF_NB_XSTATS;
+
+       txgbevf_update_stats(dev);
+
+       if (!xstats)
+               return 0;
+
+       /* Extended stats */
+       for (i = 0; i < TXGBEVF_NB_XSTATS; i++) {
+               xstats[i].id = i;
+               xstats[i].value = *(uint64_t *)(((char *)hw_stats) +
+                       rte_txgbevf_stats_strings[i].offset);
+       }
+
+       return TXGBEVF_NB_XSTATS;
+}
+
+static int
+txgbevf_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
+{
+       struct txgbevf_hw_stats *hw_stats = (struct txgbevf_hw_stats *)
+                         TXGBE_DEV_STATS(dev);
+       uint32_t i;
+
+       txgbevf_update_stats(dev);
+
+       if (stats == NULL)
+               return -EINVAL;
+
+       stats->ipackets = 0;
+       stats->ibytes = 0;
+       stats->opackets = 0;
+       stats->obytes = 0;
+
+       for (i = 0; i < 8; i++) {
+               stats->ipackets += hw_stats->qp[i].vfgprc;
+               stats->ibytes += hw_stats->qp[i].vfgorc;
+               stats->opackets += hw_stats->qp[i].vfgptc;
+               stats->obytes += hw_stats->qp[i].vfgotc;
+       }
+
+       return 0;
+}
+
+static int
+txgbevf_dev_stats_reset(struct rte_eth_dev *dev)
+{
+       struct txgbevf_hw_stats *hw_stats = (struct txgbevf_hw_stats *)
+                       TXGBE_DEV_STATS(dev);
+       uint32_t i;
+
+       /* Sync HW register to the last stats */
+       txgbevf_dev_stats_get(dev, NULL);
+
+       /* reset HW current stats*/
+       for (i = 0; i < 8; i++) {
+               hw_stats->qp[i].vfgprc = 0;
+               hw_stats->qp[i].vfgorc = 0;
+               hw_stats->qp[i].vfgptc = 0;
+               hw_stats->qp[i].vfgotc = 0;
+       }
+
+       return 0;
+}
+
 static int
 txgbevf_dev_info_get(struct rte_eth_dev *dev,
                     struct rte_eth_dev_info *dev_info)
@@ -289,14 +486,14 @@ txgbevf_dev_info_get(struct rte_eth_dev *dev,
        dev_info->max_mac_addrs = hw->mac.num_rar_entries;
        dev_info->max_hash_mac_addrs = TXGBE_VMDQ_NUM_UC_MAC;
        dev_info->max_vfs = pci_dev->max_vfs;
-       dev_info->max_vmdq_pools = ETH_64_POOLS;
+       dev_info->max_vmdq_pools = RTE_ETH_64_POOLS;
        dev_info->rx_queue_offload_capa = txgbe_get_rx_queue_offloads(dev);
        dev_info->rx_offload_capa = (txgbe_get_rx_port_offloads(dev) |
                                     dev_info->rx_queue_offload_capa);
        dev_info->tx_queue_offload_capa = txgbe_get_tx_queue_offloads(dev);
        dev_info->tx_offload_capa = txgbe_get_tx_port_offloads(dev);
        dev_info->hash_key_size = TXGBE_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 = TXGBE_RSS_OFFLOAD_ALL;
 
        dev_info->default_rxconf = (struct rte_eth_rxconf) {
@@ -377,22 +574,22 @@ txgbevf_dev_configure(struct rte_eth_dev *dev)
        PMD_INIT_LOG(DEBUG, "Configured Virtual Function port id: %d",
                     dev->data->port_id);
 
-       if (dev->data->dev_conf.rxmode.mq_mode & ETH_MQ_RX_RSS_FLAG)
-               dev->data->dev_conf.rxmode.offloads |= DEV_RX_OFFLOAD_RSS_HASH;
+       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;
 
        /*
         * VF has no ability to enable/disable HW CRC
         * Keep the persistent behavior the same as Host PF
         */
 #ifndef RTE_LIBRTE_TXGBE_PF_DISABLE_STRIP_CRC
-       if (conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC) {
+       if (conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC) {
                PMD_INIT_LOG(NOTICE, "VF can't disable HW CRC Strip");
-               conf->rxmode.offloads &= ~DEV_RX_OFFLOAD_KEEP_CRC;
+               conf->rxmode.offloads &= ~RTE_ETH_RX_OFFLOAD_KEEP_CRC;
        }
 #else
-       if (!(conf->rxmode.offloads & DEV_RX_OFFLOAD_KEEP_CRC)) {
+       if (!(conf->rxmode.offloads & RTE_ETH_RX_OFFLOAD_KEEP_CRC)) {
                PMD_INIT_LOG(NOTICE, "VF can't enable HW CRC Strip");
-               conf->rxmode.offloads |= DEV_RX_OFFLOAD_KEEP_CRC;
+               conf->rxmode.offloads |= RTE_ETH_RX_OFFLOAD_KEEP_CRC;
        }
 #endif
 
@@ -405,18 +602,165 @@ txgbevf_dev_configure(struct rte_eth_dev *dev)
        return 0;
 }
 
+static int
+txgbevf_dev_start(struct rte_eth_dev *dev)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       uint32_t intr_vector = 0;
+       struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
+
+       int err, mask = 0;
+
+       PMD_INIT_FUNC_TRACE();
+
+       /* Stop the link setup handler before resetting the HW. */
+       rte_eal_alarm_cancel(txgbe_dev_setup_link_alarm_handler, dev);
+
+       err = hw->mac.reset_hw(hw);
+       if (err) {
+               PMD_INIT_LOG(ERR, "Unable to reset vf hardware (%d)", err);
+               return err;
+       }
+       hw->mac.get_link_status = true;
+       hw->dev_start = true;
+
+       /* negotiate mailbox API version to use with the PF. */
+       txgbevf_negotiate_api(hw);
+
+       txgbevf_dev_tx_init(dev);
+
+       /* This can fail when allocating mbufs for descriptor rings */
+       err = txgbevf_dev_rx_init(dev);
+
+       /**
+        * In this case, reuses the MAC address assigned by VF
+        * initialization.
+        */
+       if (err != 0 && err != TXGBE_ERR_INVALID_MAC_ADDR) {
+               PMD_INIT_LOG(ERR, "Unable to initialize RX hardware (%d)", err);
+               txgbe_dev_clear_queues(dev);
+               return err;
+       }
+
+       /* Set vfta */
+       txgbevf_set_vfta_all(dev, 1);
+
+       /* Set HW strip */
+       mask = RTE_ETH_VLAN_STRIP_MASK | RTE_ETH_VLAN_FILTER_MASK |
+               RTE_ETH_VLAN_EXTEND_MASK;
+       err = txgbevf_vlan_offload_config(dev, mask);
+       if (err) {
+               PMD_INIT_LOG(ERR, "Unable to set VLAN offload (%d)", err);
+               txgbe_dev_clear_queues(dev);
+               return err;
+       }
+
+       txgbevf_dev_rxtx_start(dev);
+
+       /* check and configure queue intr-vector mapping */
+       if (rte_intr_cap_multiple(intr_handle) &&
+           dev->data->dev_conf.intr_conf.rxq) {
+               /* According to datasheet, only vector 0/1/2 can be used,
+                * now only one vector is used for Rx queue
+                */
+               intr_vector = 1;
+               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_INIT_LOG(ERR, "Failed to allocate %d rx_queues"
+                                    " intr_vec", dev->data->nb_rx_queues);
+                       return -ENOMEM;
+               }
+       }
+       txgbevf_configure_msix(dev);
+
+       /* When a VF port is bound to VFIO-PCI, only miscellaneous interrupt
+        * is mapped to VFIO vector 0 in eth_txgbevf_dev_init( ).
+        * If previous VFIO interrupt mapping setting in eth_txgbevf_dev_init( )
+        * is not cleared, it will fail when following rte_intr_enable( ) tries
+        * to map Rx queue interrupt to other VFIO vectors.
+        * So clear uio/vfio intr/evevnfd first to avoid failure.
+        */
+       rte_intr_disable(intr_handle);
+
+       rte_intr_enable(intr_handle);
+
+       /* Re-enable interrupt for VF */
+       txgbevf_intr_enable(dev);
+
+       /*
+        * Update link status right before return, because it may
+        * start link configuration process in a separate thread.
+        */
+       txgbevf_dev_link_update(dev, 0);
+
+       hw->adapter_stopped = false;
+
+       return 0;
+}
+
+static int
+txgbevf_dev_stop(struct rte_eth_dev *dev)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       struct txgbe_adapter *adapter = TXGBE_DEV_ADAPTER(dev);
+       struct rte_pci_device *pci_dev = RTE_ETH_DEV_TO_PCI(dev);
+       struct rte_intr_handle *intr_handle = pci_dev->intr_handle;
+
+       if (hw->adapter_stopped)
+               return 0;
+
+       PMD_INIT_FUNC_TRACE();
+
+       rte_eal_alarm_cancel(txgbe_dev_setup_link_alarm_handler, dev);
+
+       txgbevf_intr_disable(dev);
+
+       hw->adapter_stopped = 1;
+       hw->mac.stop_hw(hw);
+
+       /*
+        * Clear what we set, but we still keep shadow_vfta to
+        * restore after device starts
+        */
+       txgbevf_set_vfta_all(dev, 0);
+
+       /* Clear stored conf */
+       dev->data->scattered_rx = 0;
+
+       txgbe_dev_clear_queues(dev);
+
+       /* Clean datapath event and queue/vec mapping */
+       rte_intr_efd_disable(intr_handle);
+       rte_intr_vec_list_free(intr_handle);
+
+       adapter->rss_reta_updated = 0;
+       hw->dev_start = false;
+
+       return 0;
+}
+
 static int
 txgbevf_dev_close(struct rte_eth_dev *dev)
 {
        struct txgbe_hw *hw = TXGBE_DEV_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;
+       int ret;
+
        PMD_INIT_FUNC_TRACE();
        if (rte_eal_process_type() != RTE_PROC_PRIMARY)
                return 0;
 
        hw->mac.reset_hw(hw);
 
+       ret = txgbevf_dev_stop(dev);
+
        txgbe_dev_free_queues(dev);
 
        /**
@@ -439,6 +783,127 @@ txgbevf_dev_close(struct rte_eth_dev *dev)
        rte_intr_callback_unregister(intr_handle,
                                     txgbevf_dev_interrupt_handler, dev);
 
+       return ret;
+}
+
+/*
+ * Reset VF device
+ */
+static int
+txgbevf_dev_reset(struct rte_eth_dev *dev)
+{
+       int ret;
+
+       ret = eth_txgbevf_dev_uninit(dev);
+       if (ret)
+               return ret;
+
+       ret = eth_txgbevf_dev_init(dev);
+
+       return ret;
+}
+
+static void txgbevf_set_vfta_all(struct rte_eth_dev *dev, bool on)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       struct txgbe_vfta *shadow_vfta = TXGBE_DEV_VFTA(dev);
+       int i = 0, j = 0, vfta = 0, mask = 1;
+
+       for (i = 0; i < TXGBE_VFTA_SIZE; i++) {
+               vfta = shadow_vfta->vfta[i];
+               if (vfta) {
+                       mask = 1;
+                       for (j = 0; j < 32; j++) {
+                               if (vfta & mask)
+                                       hw->mac.set_vfta(hw, (i << 5) + j, 0,
+                                                      on, false);
+                               mask <<= 1;
+                       }
+               }
+       }
+}
+
+static int
+txgbevf_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vlan_id, int on)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       struct txgbe_vfta *shadow_vfta = TXGBE_DEV_VFTA(dev);
+       uint32_t vid_idx = 0;
+       uint32_t vid_bit = 0;
+       int ret = 0;
+
+       PMD_INIT_FUNC_TRACE();
+
+       /* vind is not used in VF driver, set to 0, check txgbe_set_vfta_vf */
+       ret = hw->mac.set_vfta(hw, vlan_id, 0, !!on, false);
+       if (ret) {
+               PMD_INIT_LOG(ERR, "Unable to set VF vlan");
+               return ret;
+       }
+       vid_idx = (uint32_t)((vlan_id >> 5) & 0x7F);
+       vid_bit = (uint32_t)(1 << (vlan_id & 0x1F));
+
+       /* Save what we set and restore it after device reset */
+       if (on)
+               shadow_vfta->vfta[vid_idx] |= vid_bit;
+       else
+               shadow_vfta->vfta[vid_idx] &= ~vid_bit;
+
+       return 0;
+}
+
+static void
+txgbevf_vlan_strip_queue_set(struct rte_eth_dev *dev, uint16_t queue, int on)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       uint32_t ctrl;
+
+       PMD_INIT_FUNC_TRACE();
+
+       if (queue >= hw->mac.max_rx_queues)
+               return;
+
+       ctrl = rd32(hw, TXGBE_RXCFG(queue));
+       txgbe_dev_save_rx_queue(hw, queue);
+       if (on)
+               ctrl |= TXGBE_RXCFG_VLAN;
+       else
+               ctrl &= ~TXGBE_RXCFG_VLAN;
+       wr32(hw, TXGBE_RXCFG(queue), 0);
+       msec_delay(100);
+       txgbe_dev_store_rx_queue(hw, queue);
+       wr32m(hw, TXGBE_RXCFG(queue),
+               TXGBE_RXCFG_VLAN | TXGBE_RXCFG_ENA, ctrl);
+
+       txgbe_vlan_hw_strip_bitmap_set(dev, queue, on);
+}
+
+static int
+txgbevf_vlan_offload_config(struct rte_eth_dev *dev, int mask)
+{
+       struct txgbe_rx_queue *rxq;
+       uint16_t i;
+       int on = 0;
+
+       /* VF function only support hw strip feature, others are not support */
+       if (mask & RTE_ETH_VLAN_STRIP_MASK) {
+               for (i = 0; i < dev->data->nb_rx_queues; i++) {
+                       rxq = dev->data->rx_queues[i];
+                       on = !!(rxq->offloads & RTE_ETH_RX_OFFLOAD_VLAN_STRIP);
+                       txgbevf_vlan_strip_queue_set(dev, i, on);
+               }
+       }
+
+       return 0;
+}
+
+static int
+txgbevf_vlan_offload_set(struct rte_eth_dev *dev, int mask)
+{
+       txgbe_config_vlan_strip_on_all_queues(dev, mask);
+
+       txgbevf_vlan_offload_config(dev, mask);
+
        return 0;
 }
 
@@ -446,7 +911,7 @@ static int
 txgbevf_dev_rx_queue_intr_enable(struct rte_eth_dev *dev, uint16_t queue_id)
 {
        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 txgbe_interrupt *intr = TXGBE_DEV_INTR(dev);
        struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
        uint32_t vec = TXGBE_MISC_VEC_ID;
@@ -468,7 +933,7 @@ txgbevf_dev_rx_queue_intr_disable(struct rte_eth_dev *dev, uint16_t queue_id)
        struct txgbe_interrupt *intr = TXGBE_DEV_INTR(dev);
        struct txgbe_hw *hw = TXGBE_DEV_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 vec = TXGBE_MISC_VEC_ID;
 
        if (rte_intr_allow_others(intr_handle))
@@ -508,7 +973,7 @@ static void
 txgbevf_configure_msix(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 txgbe_hw *hw = TXGBE_DEV_HW(dev);
        uint32_t q_idx;
        uint32_t vector_idx = TXGBE_MISC_VEC_ID;
@@ -534,8 +999,10 @@ txgbevf_configure_msix(struct rte_eth_dev *dev)
                 * as TXGBE_VF_MAXMSIVECOTR = 1
                 */
                txgbevf_set_ivar_map(hw, 0, q_idx, vector_idx);
-               intr_handle->intr_vec[q_idx] = vector_idx;
-               if (vector_idx < base + intr_handle->nb_efd - 1)
+               rte_intr_vec_list_index_set(intr_handle, q_idx,
+                                                  vector_idx);
+               if (vector_idx < base + rte_intr_nb_efd_get(intr_handle)
+                   - 1)
                        vector_idx++;
        }
 
@@ -566,14 +1033,8 @@ txgbevf_add_mac_addr(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr,
        err = txgbevf_set_uc_addr_vf(hw, 2, mac_addr->addr_bytes);
        if (err != 0)
                PMD_DRV_LOG(ERR, "Unable to add MAC address "
-                           "%02x:%02x:%02x:%02x:%02x:%02x - err=%d",
-                           mac_addr->addr_bytes[0],
-                           mac_addr->addr_bytes[1],
-                           mac_addr->addr_bytes[2],
-                           mac_addr->addr_bytes[3],
-                           mac_addr->addr_bytes[4],
-                           mac_addr->addr_bytes[5],
-                           err);
+                           RTE_ETHER_ADDR_PRT_FMT " - err=%d",
+                           RTE_ETHER_ADDR_BYTES(mac_addr), err);
        return err;
 }
 
@@ -615,15 +1076,9 @@ txgbevf_remove_mac_addr(struct rte_eth_dev *dev, uint32_t index)
                if (err != 0)
                        PMD_DRV_LOG(ERR,
                                    "Adding again MAC address "
-                                   "%02x:%02x:%02x:%02x:%02x:%02x failed "
+                                   RTE_ETHER_ADDR_PRT_FMT " failed "
                                    "err=%d",
-                                   mac_addr->addr_bytes[0],
-                                   mac_addr->addr_bytes[1],
-                                   mac_addr->addr_bytes[2],
-                                   mac_addr->addr_bytes[3],
-                                   mac_addr->addr_bytes[4],
-                                   mac_addr->addr_bytes[5],
-                                   err);
+                                   RTE_ETHER_ADDR_BYTES(mac_addr), err);
        }
 }
 
@@ -638,6 +1093,168 @@ txgbevf_set_default_mac_addr(struct rte_eth_dev *dev,
        return 0;
 }
 
+static int
+txgbevf_dev_set_mtu(struct rte_eth_dev *dev, uint16_t mtu)
+{
+       struct txgbe_hw *hw;
+       uint32_t max_frame = mtu + RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN;
+       struct rte_eth_dev_data *dev_data = dev->data;
+
+       hw = TXGBE_DEV_HW(dev);
+
+       if (mtu < RTE_ETHER_MIN_MTU ||
+                       max_frame > RTE_ETHER_MAX_JUMBO_FRAME_LEN)
+               return -EINVAL;
+
+       /* If device is started, refuse mtu that requires the support of
+        * scattered packets when this feature has not been enabled before.
+        */
+       if (dev_data->dev_started && !dev_data->scattered_rx &&
+           (max_frame + 2 * TXGBE_VLAN_TAG_SIZE >
+            dev->data->min_rx_buf_size - RTE_PKTMBUF_HEADROOM)) {
+               PMD_INIT_LOG(ERR, "Stop port first.");
+               return -EINVAL;
+       }
+
+       /*
+        * When supported by the underlying PF driver, use the TXGBE_VF_SET_MTU
+        * request of the version 2.0 of the mailbox API.
+        * For now, use the TXGBE_VF_SET_LPE request of the version 1.0
+        * of the mailbox API.
+        */
+       if (txgbevf_rlpml_set_vf(hw, max_frame))
+               return -EINVAL;
+
+       return 0;
+}
+
+static int
+txgbevf_get_reg_length(struct rte_eth_dev *dev __rte_unused)
+{
+       int count = 0;
+       int g_ind = 0;
+       const struct reg_info *reg_group;
+
+       while ((reg_group = txgbevf_regs[g_ind++]))
+               count += txgbe_regs_group_count(reg_group);
+
+       return count;
+}
+
+static int
+txgbevf_get_regs(struct rte_eth_dev *dev,
+               struct rte_dev_reg_info *regs)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       uint32_t *data = regs->data;
+       int g_ind = 0;
+       int count = 0;
+       const struct reg_info *reg_group;
+
+       if (data == NULL) {
+               regs->length = txgbevf_get_reg_length(dev);
+               regs->width = sizeof(uint32_t);
+               return 0;
+       }
+
+       /* Support only full register dump */
+       if (regs->length == 0 ||
+           regs->length == (uint32_t)txgbevf_get_reg_length(dev)) {
+               regs->version = hw->mac.type << 24 | hw->revision_id << 16 |
+                       hw->device_id;
+               while ((reg_group = txgbevf_regs[g_ind++]))
+                       count += txgbe_read_regs_group(dev, &data[count],
+                                                     reg_group);
+               return 0;
+       }
+
+       return -ENOTSUP;
+}
+
+static int
+txgbevf_dev_promiscuous_enable(struct rte_eth_dev *dev)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       int ret;
+
+       switch (hw->mac.update_xcast_mode(hw, TXGBEVF_XCAST_MODE_PROMISC)) {
+       case 0:
+               ret = 0;
+               break;
+       case TXGBE_ERR_FEATURE_NOT_SUPPORTED:
+               ret = -ENOTSUP;
+               break;
+       default:
+               ret = -EAGAIN;
+               break;
+       }
+
+       return ret;
+}
+
+static int
+txgbevf_dev_promiscuous_disable(struct rte_eth_dev *dev)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       int ret;
+
+       switch (hw->mac.update_xcast_mode(hw, TXGBEVF_XCAST_MODE_NONE)) {
+       case 0:
+               ret = 0;
+               break;
+       case TXGBE_ERR_FEATURE_NOT_SUPPORTED:
+               ret = -ENOTSUP;
+               break;
+       default:
+               ret = -EAGAIN;
+               break;
+       }
+
+       return ret;
+}
+
+static int
+txgbevf_dev_allmulticast_enable(struct rte_eth_dev *dev)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       int ret;
+
+       switch (hw->mac.update_xcast_mode(hw, TXGBEVF_XCAST_MODE_ALLMULTI)) {
+       case 0:
+               ret = 0;
+               break;
+       case TXGBE_ERR_FEATURE_NOT_SUPPORTED:
+               ret = -ENOTSUP;
+               break;
+       default:
+               ret = -EAGAIN;
+               break;
+       }
+
+       return ret;
+}
+
+static int
+txgbevf_dev_allmulticast_disable(struct rte_eth_dev *dev)
+{
+       struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
+       int ret;
+
+       switch (hw->mac.update_xcast_mode(hw, TXGBEVF_XCAST_MODE_MULTI)) {
+       case 0:
+               ret = 0;
+               break;
+       case TXGBE_ERR_FEATURE_NOT_SUPPORTED:
+               ret = -ENOTSUP;
+               break;
+       default:
+               ret = -EAGAIN;
+               break;
+       }
+
+       return ret;
+}
+
 static void txgbevf_mbx_process(struct rte_eth_dev *dev)
 {
        struct txgbe_hw *hw = TXGBE_DEV_HW(dev);
@@ -706,15 +1323,44 @@ txgbevf_dev_interrupt_handler(void *param)
  */
 static const struct eth_dev_ops txgbevf_eth_dev_ops = {
        .dev_configure        = txgbevf_dev_configure,
+       .dev_start            = txgbevf_dev_start,
+       .dev_stop             = txgbevf_dev_stop,
        .link_update          = txgbevf_dev_link_update,
+       .stats_get            = txgbevf_dev_stats_get,
+       .xstats_get           = txgbevf_dev_xstats_get,
+       .stats_reset          = txgbevf_dev_stats_reset,
+       .xstats_reset         = txgbevf_dev_stats_reset,
+       .xstats_get_names     = txgbevf_dev_xstats_get_names,
+       .dev_close            = txgbevf_dev_close,
+       .dev_reset            = txgbevf_dev_reset,
+       .promiscuous_enable   = txgbevf_dev_promiscuous_enable,
+       .promiscuous_disable  = txgbevf_dev_promiscuous_disable,
+       .allmulticast_enable  = txgbevf_dev_allmulticast_enable,
+       .allmulticast_disable = txgbevf_dev_allmulticast_disable,
        .dev_infos_get        = txgbevf_dev_info_get,
+       .dev_supported_ptypes_get = txgbe_dev_supported_ptypes_get,
+       .mtu_set              = txgbevf_dev_set_mtu,
+       .vlan_filter_set      = txgbevf_vlan_filter_set,
+       .vlan_strip_queue_set = txgbevf_vlan_strip_queue_set,
+       .vlan_offload_set     = txgbevf_vlan_offload_set,
+       .rx_queue_setup       = txgbe_dev_rx_queue_setup,
+       .rx_queue_release     = txgbe_dev_rx_queue_release,
+       .tx_queue_setup       = txgbe_dev_tx_queue_setup,
+       .tx_queue_release     = txgbe_dev_tx_queue_release,
        .rx_queue_intr_enable = txgbevf_dev_rx_queue_intr_enable,
        .rx_queue_intr_disable = txgbevf_dev_rx_queue_intr_disable,
        .mac_addr_add         = txgbevf_add_mac_addr,
        .mac_addr_remove      = txgbevf_remove_mac_addr,
+       .set_mc_addr_list     = txgbe_dev_set_mc_addr_list,
        .rxq_info_get         = txgbe_rxq_info_get,
        .txq_info_get         = txgbe_txq_info_get,
        .mac_addr_set         = txgbevf_set_default_mac_addr,
+       .get_reg              = txgbevf_get_regs,
+       .reta_update          = txgbe_dev_rss_reta_update,
+       .reta_query           = txgbe_dev_rss_reta_query,
+       .rss_hash_update      = txgbe_dev_rss_hash_update,
+       .rss_hash_conf_get    = txgbe_dev_rss_hash_conf_get,
+       .tx_done_cleanup      = txgbe_dev_tx_done_cleanup,
 };
 
 RTE_PMD_REGISTER_PCI(net_txgbe_vf, rte_txgbevf_pmd);