net: add rte prefix to ether structures
[dpdk.git] / drivers / net / vmxnet3 / vmxnet3_ethdev.c
index 51e2d4c..3ad1e9e 100644 (file)
@@ -1,34 +1,5 @@
-/*-
- *   BSD LICENSE
- *
- *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
- *   All rights reserved.
- *
- *   Redistribution and use in source and binary forms, with or without
- *   modification, are permitted provided that the following conditions
- *   are met:
- *
- *     * Redistributions of source code must retain the above copyright
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright
- *       notice, this list of conditions and the following disclaimer in
- *       the documentation and/or other materials provided with the
- *       distribution.
- *     * Neither the name of Intel Corporation nor the names of its
- *       contributors may be used to endorse or promote products derived
- *       from this software without specific prior written permission.
- *
- *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
- *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
- *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
- *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
- *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
- *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
- *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
- *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
- *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
- *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
- *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2015 Intel Corporation
  */
 
 #include <sys/queue.h>
 #include <rte_log.h>
 #include <rte_debug.h>
 #include <rte_pci.h>
-#include <rte_atomic.h>
+#include <rte_bus_pci.h>
 #include <rte_branch_prediction.h>
 #include <rte_memory.h>
 #include <rte_memzone.h>
 #include <rte_eal.h>
 #include <rte_alarm.h>
 #include <rte_ether.h>
-#include <rte_ethdev.h>
-#include <rte_atomic.h>
+#include <rte_ethdev_driver.h>
+#include <rte_ethdev_pci.h>
 #include <rte_string_fns.h>
 #include <rte_malloc.h>
 #include <rte_dev.h>
 
 #define PROCESS_SYS_EVENTS 0
 
+#define        VMXNET3_TX_MAX_SEG      UINT8_MAX
+
+#define VMXNET3_TX_OFFLOAD_CAP         \
+       (DEV_TX_OFFLOAD_VLAN_INSERT |   \
+        DEV_TX_OFFLOAD_IPV4_CKSUM |    \
+        DEV_TX_OFFLOAD_TCP_CKSUM |     \
+        DEV_TX_OFFLOAD_UDP_CKSUM |     \
+        DEV_TX_OFFLOAD_TCP_TSO |       \
+        DEV_TX_OFFLOAD_MULTI_SEGS)
+
+#define VMXNET3_RX_OFFLOAD_CAP         \
+       (DEV_RX_OFFLOAD_VLAN_STRIP |    \
+        DEV_RX_OFFLOAD_VLAN_FILTER |   \
+        DEV_RX_OFFLOAD_SCATTER |       \
+        DEV_RX_OFFLOAD_IPV4_CKSUM |    \
+        DEV_RX_OFFLOAD_UDP_CKSUM |     \
+        DEV_RX_OFFLOAD_TCP_CKSUM |     \
+        DEV_RX_OFFLOAD_TCP_LRO |       \
+        DEV_RX_OFFLOAD_JUMBO_FRAME)
+
 static int eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev);
 static int eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev);
 static int vmxnet3_dev_configure(struct rte_eth_dev *dev);
@@ -80,23 +71,33 @@ static void vmxnet3_dev_promiscuous_enable(struct rte_eth_dev *dev);
 static void vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev);
 static void vmxnet3_dev_allmulticast_enable(struct rte_eth_dev *dev);
 static void vmxnet3_dev_allmulticast_disable(struct rte_eth_dev *dev);
+static int __vmxnet3_dev_link_update(struct rte_eth_dev *dev,
+                                    int wait_to_complete);
 static int vmxnet3_dev_link_update(struct rte_eth_dev *dev,
                                   int wait_to_complete);
-static void vmxnet3_dev_stats_get(struct rte_eth_dev *dev,
+static void vmxnet3_hw_stats_save(struct vmxnet3_hw *hw);
+static int vmxnet3_dev_stats_get(struct rte_eth_dev *dev,
                                  struct rte_eth_stats *stats);
+static void vmxnet3_dev_stats_reset(struct rte_eth_dev *dev);
+static int vmxnet3_dev_xstats_get_names(struct rte_eth_dev *dev,
+                                       struct rte_eth_xstat_name *xstats,
+                                       unsigned int n);
+static int vmxnet3_dev_xstats_get(struct rte_eth_dev *dev,
+                                 struct rte_eth_xstat *xstats, unsigned int n);
 static void vmxnet3_dev_info_get(struct rte_eth_dev *dev,
                                 struct rte_eth_dev_info *dev_info);
 static const uint32_t *
 vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev);
 static int vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev,
                                       uint16_t vid, int on);
-static void vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
-static void vmxnet3_mac_addr_set(struct rte_eth_dev *dev,
-                                struct ether_addr *mac_addr);
+static int vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask);
+static int vmxnet3_mac_addr_set(struct rte_eth_dev *dev,
+                                struct rte_ether_addr *mac_addr);
+static void vmxnet3_interrupt_handler(void *param);
+
+int vmxnet3_logtype_init;
+int vmxnet3_logtype_driver;
 
-#if PROCESS_SYS_EVENTS == 1
-static void vmxnet3_process_events(struct vmxnet3_hw *);
-#endif
 /*
  * The set of PCI devices this driver supports
  */
@@ -118,6 +119,9 @@ static const struct eth_dev_ops vmxnet3_eth_dev_ops = {
        .allmulticast_disable = vmxnet3_dev_allmulticast_disable,
        .link_update          = vmxnet3_dev_link_update,
        .stats_get            = vmxnet3_dev_stats_get,
+       .xstats_get_names     = vmxnet3_dev_xstats_get_names,
+       .xstats_get           = vmxnet3_dev_xstats_get,
+       .stats_reset          = vmxnet3_dev_stats_reset,
        .mac_addr_set         = vmxnet3_mac_addr_set,
        .dev_infos_get        = vmxnet3_dev_info_get,
        .dev_supported_ptypes_get = vmxnet3_dev_supported_ptypes_get,
@@ -129,6 +133,27 @@ static const struct eth_dev_ops vmxnet3_eth_dev_ops = {
        .tx_queue_release     = vmxnet3_dev_tx_queue_release,
 };
 
+struct vmxnet3_xstats_name_off {
+       char name[RTE_ETH_XSTATS_NAME_SIZE];
+       unsigned int offset;
+};
+
+/* tx_qX_ is prepended to the name string here */
+static const struct vmxnet3_xstats_name_off vmxnet3_txq_stat_strings[] = {
+       {"drop_total",         offsetof(struct vmxnet3_txq_stats, drop_total)},
+       {"drop_too_many_segs", offsetof(struct vmxnet3_txq_stats, drop_too_many_segs)},
+       {"drop_tso",           offsetof(struct vmxnet3_txq_stats, drop_tso)},
+       {"tx_ring_full",       offsetof(struct vmxnet3_txq_stats, tx_ring_full)},
+};
+
+/* rx_qX_ is prepended to the name string here */
+static const struct vmxnet3_xstats_name_off vmxnet3_rxq_stat_strings[] = {
+       {"drop_total",           offsetof(struct vmxnet3_rxq_stats, drop_total)},
+       {"drop_err",             offsetof(struct vmxnet3_rxq_stats, drop_err)},
+       {"drop_fcs",             offsetof(struct vmxnet3_rxq_stats, drop_fcs)},
+       {"rx_buf_alloc_failure", offsetof(struct vmxnet3_rxq_stats, rx_buf_alloc_failure)},
+};
+
 static const struct rte_memzone *
 gpa_zone_reserve(struct rte_eth_dev *dev, uint32_t size,
                 const char *post_string, int socket_id,
@@ -137,89 +162,67 @@ gpa_zone_reserve(struct rte_eth_dev *dev, uint32_t size,
        char z_name[RTE_MEMZONE_NAMESIZE];
        const struct rte_memzone *mz;
 
-       snprintf(z_name, sizeof(z_name), "%s_%d_%s",
-                dev->driver->pci_drv.driver.name, dev->data->port_id, post_string);
+       snprintf(z_name, sizeof(z_name), "eth_p%d_%s",
+                       dev->data->port_id, post_string);
 
        mz = rte_memzone_lookup(z_name);
        if (!reuse) {
                if (mz)
                        rte_memzone_free(mz);
                return rte_memzone_reserve_aligned(z_name, size, socket_id,
-                                                  0, align);
+                               RTE_MEMZONE_IOVA_CONTIG, align);
        }
 
        if (mz)
                return mz;
 
-       return rte_memzone_reserve_aligned(z_name, size, socket_id, 0, align);
+       return rte_memzone_reserve_aligned(z_name, size, socket_id,
+                       RTE_MEMZONE_IOVA_CONTIG, align);
 }
 
-/**
- * Atomically reads the link status information from global
- * structure rte_eth_dev.
- *
- * @param dev
- *   - Pointer to the structure rte_eth_dev to read from.
- *   - Pointer to the buffer to be saved with the link status.
- *
- * @return
- *   - On success, zero.
- *   - On failure, negative value.
+/*
+ * This function is based on vmxnet3_disable_intr()
  */
-
-static int
-vmxnet3_dev_atomic_read_link_status(struct rte_eth_dev *dev,
-                                   struct rte_eth_link *link)
+static void
+vmxnet3_disable_intr(struct vmxnet3_hw *hw)
 {
-       struct rte_eth_link *dst = link;
-       struct rte_eth_link *src = &(dev->data->dev_link);
+       int i;
 
-       if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
-                               *(uint64_t *)src) == 0)
-               return -1;
+       PMD_INIT_FUNC_TRACE();
 
-       return 0;
+       hw->shared->devRead.intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
+       for (i = 0; i < hw->num_intrs; i++)
+               VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 1);
 }
 
-/**
- * Atomically writes the link status information into global
- * structure rte_eth_dev.
- *
- * @param dev
- *   - Pointer to the structure rte_eth_dev to write to.
- *   - Pointer to the buffer to be saved with the link status.
- *
- * @return
- *   - On success, zero.
- *   - On failure, negative value.
- */
-static int
-vmxnet3_dev_atomic_write_link_status(struct rte_eth_dev *dev,
-                                    struct rte_eth_link *link)
+static void
+vmxnet3_enable_intr(struct vmxnet3_hw *hw)
 {
-       struct rte_eth_link *dst = &(dev->data->dev_link);
-       struct rte_eth_link *src = link;
+       int i;
 
-       if (rte_atomic64_cmpset((uint64_t *)dst, *(uint64_t *)dst,
-                               *(uint64_t *)src) == 0)
-               return -1;
+       PMD_INIT_FUNC_TRACE();
 
-       return 0;
+       hw->shared->devRead.intrConf.intrCtrl &= ~VMXNET3_IC_DISABLE_ALL;
+       for (i = 0; i < hw->num_intrs; i++)
+               VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 0);
 }
 
 /*
- * This function is based on vmxnet3_disable_intr()
+ * Gets tx data ring descriptor size.
  */
-static void
-vmxnet3_disable_intr(struct vmxnet3_hw *hw)
+static uint16_t
+eth_vmxnet3_txdata_get(struct vmxnet3_hw *hw)
 {
-       int i;
+       uint16 txdata_desc_size;
 
-       PMD_INIT_FUNC_TRACE();
+       VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
+                              VMXNET3_CMD_GET_TXDATA_DESC_SIZE);
+       txdata_desc_size = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
 
-       hw->shared->devRead.intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
-       for (i = 0; i < VMXNET3_MAX_INTRS; i++)
-               VMXNET3_WRITE_BAR0_REG(hw, VMXNET3_REG_IMR + i * 8, 1);
+       return (txdata_desc_size < VMXNET3_TXDATA_DESC_MIN_SIZE ||
+               txdata_desc_size > VMXNET3_TXDATA_DESC_MAX_SIZE ||
+               txdata_desc_size & VMXNET3_TXDATA_DESC_SIZE_MASK) ?
+               sizeof(struct Vmxnet3_TxDataDesc) : txdata_desc_size;
 }
 
 /*
@@ -231,13 +234,15 @@ eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev)
        struct rte_pci_device *pci_dev;
        struct vmxnet3_hw *hw = eth_dev->data->dev_private;
        uint32_t mac_hi, mac_lo, ver;
+       struct rte_eth_link link;
 
        PMD_INIT_FUNC_TRACE();
 
        eth_dev->dev_ops = &vmxnet3_eth_dev_ops;
        eth_dev->rx_pkt_burst = &vmxnet3_recv_pkts;
        eth_dev->tx_pkt_burst = &vmxnet3_xmit_pkts;
-       pci_dev = eth_dev->pci_dev;
+       eth_dev->tx_pkt_prepare = vmxnet3_prep_pkts;
+       pci_dev = RTE_ETH_DEV_TO_PCI(eth_dev);
 
        /*
         * for secondary processes, we don't initialize any further as primary
@@ -261,13 +266,30 @@ eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev)
        /* Check h/w version compatibility with driver. */
        ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_VRRS);
        PMD_INIT_LOG(DEBUG, "Hardware version : %d", ver);
-       if (ver & 0x1)
-               VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS, 1);
-       else {
-               PMD_INIT_LOG(ERR, "Incompatible h/w version, should be 0x1");
+
+       if (ver & (1 << VMXNET3_REV_4)) {
+               VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
+                                      1 << VMXNET3_REV_4);
+               hw->version = VMXNET3_REV_4 + 1;
+       } else if (ver & (1 << VMXNET3_REV_3)) {
+               VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
+                                      1 << VMXNET3_REV_3);
+               hw->version = VMXNET3_REV_3 + 1;
+       } else if (ver & (1 << VMXNET3_REV_2)) {
+               VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
+                                      1 << VMXNET3_REV_2);
+               hw->version = VMXNET3_REV_2 + 1;
+       } else if (ver & (1 << VMXNET3_REV_1)) {
+               VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_VRRS,
+                                      1 << VMXNET3_REV_1);
+               hw->version = VMXNET3_REV_1 + 1;
+       } else {
+               PMD_INIT_LOG(ERR, "Incompatible hardware version: %d", ver);
                return -EIO;
        }
 
+       PMD_INIT_LOG(DEBUG, "Using device version %d\n", hw->version);
+
        /* Check UPT version compatibility with driver. */
        ver = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_UVRS);
        PMD_INIT_LOG(DEBUG, "UPT hardware version : %d", ver);
@@ -294,19 +316,45 @@ eth_vmxnet3_dev_init(struct rte_eth_dev *eth_dev)
                return -ENOMEM;
        }
        /* Copy the permanent MAC address */
-       ether_addr_copy((struct ether_addr *) hw->perm_addr,
+       ether_addr_copy((struct rte_ether_addr *)hw->perm_addr,
                        &eth_dev->data->mac_addrs[0]);
 
        PMD_INIT_LOG(DEBUG, "MAC Address : %02x:%02x:%02x:%02x:%02x:%02x",
                     hw->perm_addr[0], hw->perm_addr[1], hw->perm_addr[2],
                     hw->perm_addr[3], hw->perm_addr[4], hw->perm_addr[5]);
 
+       /* Flag to call rte_eth_dev_release_port() in rte_eth_dev_close(). */
+       eth_dev->data->dev_flags |= RTE_ETH_DEV_CLOSE_REMOVE;
+
        /* Put device in Quiesce Mode */
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
 
        /* allow untagged pkts */
        VMXNET3_SET_VFTABLE_ENTRY(hw->shadow_vfta, 0);
 
+       hw->txdata_desc_size = VMXNET3_VERSION_GE_3(hw) ?
+               eth_vmxnet3_txdata_get(hw) : sizeof(struct Vmxnet3_TxDataDesc);
+
+       hw->rxdata_desc_size = VMXNET3_VERSION_GE_3(hw) ?
+               VMXNET3_DEF_RXDATA_DESC_SIZE : 0;
+       RTE_ASSERT((hw->rxdata_desc_size & ~VMXNET3_RXDATA_DESC_SIZE_MASK) ==
+                  hw->rxdata_desc_size);
+
+       /* clear shadow stats */
+       memset(hw->saved_tx_stats, 0, sizeof(hw->saved_tx_stats));
+       memset(hw->saved_rx_stats, 0, sizeof(hw->saved_rx_stats));
+
+       /* clear snapshot stats */
+       memset(hw->snapshot_tx_stats, 0, sizeof(hw->snapshot_tx_stats));
+       memset(hw->snapshot_rx_stats, 0, sizeof(hw->snapshot_rx_stats));
+
+       /* set the initial link status */
+       memset(&link, 0, sizeof(link));
+       link.link_duplex = ETH_LINK_FULL_DUPLEX;
+       link.link_speed = ETH_SPEED_NUM_10G;
+       link.link_autoneg = ETH_LINK_FIXED;
+       rte_eth_linkstatus_set(eth_dev, &link);
+
        return 0;
 }
 
@@ -320,29 +368,36 @@ eth_vmxnet3_dev_uninit(struct rte_eth_dev *eth_dev)
        if (rte_eal_process_type() != RTE_PROC_PRIMARY)
                return 0;
 
-       if (hw->adapter_stopped == 0)
-               vmxnet3_dev_close(eth_dev);
+       if (hw->adapter_stopped == 0) {
+               PMD_INIT_LOG(DEBUG, "Device has not been closed.");
+               return -EBUSY;
+       }
 
        eth_dev->dev_ops = NULL;
        eth_dev->rx_pkt_burst = NULL;
        eth_dev->tx_pkt_burst = NULL;
-
-       rte_free(eth_dev->data->mac_addrs);
-       eth_dev->data->mac_addrs = NULL;
+       eth_dev->tx_pkt_prepare = NULL;
 
        return 0;
 }
 
-static struct eth_driver rte_vmxnet3_pmd = {
-       .pci_drv = {
-               .id_table = pci_id_vmxnet3_map,
-               .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_DETACHABLE,
-               .probe = rte_eth_dev_pci_probe,
-               .remove = rte_eth_dev_pci_remove,
-       },
-       .eth_dev_init = eth_vmxnet3_dev_init,
-       .eth_dev_uninit = eth_vmxnet3_dev_uninit,
-       .dev_private_size = sizeof(struct vmxnet3_hw),
+static int eth_vmxnet3_pci_probe(struct rte_pci_driver *pci_drv __rte_unused,
+       struct rte_pci_device *pci_dev)
+{
+       return rte_eth_dev_pci_generic_probe(pci_dev,
+               sizeof(struct vmxnet3_hw), eth_vmxnet3_dev_init);
+}
+
+static int eth_vmxnet3_pci_remove(struct rte_pci_device *pci_dev)
+{
+       return rte_eth_dev_pci_generic_remove(pci_dev, eth_vmxnet3_dev_uninit);
+}
+
+static struct rte_pci_driver rte_vmxnet3_pmd = {
+       .id_table = pci_id_vmxnet3_map,
+       .drv_flags = RTE_PCI_DRV_NEED_MAPPING | RTE_PCI_DRV_INTR_LSC,
+       .probe = eth_vmxnet3_pci_probe,
+       .remove = eth_vmxnet3_pci_remove,
 };
 
 static int
@@ -388,7 +443,7 @@ vmxnet3_dev_configure(struct rte_eth_dev *dev)
        memset(mz->addr, 0, mz->len);
 
        hw->shared = mz->addr;
-       hw->sharedPA = mz->phys_addr;
+       hw->sharedPA = mz->iova;
 
        /*
         * Allocate a memzone for Vmxnet3_RxQueueDesc - Vmxnet3_TxQueueDesc
@@ -409,7 +464,7 @@ vmxnet3_dev_configure(struct rte_eth_dev *dev)
        hw->tqd_start = (Vmxnet3_TxQueueDesc *)mz->addr;
        hw->rqd_start = (Vmxnet3_RxQueueDesc *)(hw->tqd_start + hw->num_tx_queues);
 
-       hw->queueDescPA = mz->phys_addr;
+       hw->queueDescPA = mz->iova;
        hw->queue_desc_len = (uint16_t)size;
 
        if (dev->data->dev_conf.rxmode.mq_mode == ETH_MQ_RX_RSS) {
@@ -425,7 +480,7 @@ vmxnet3_dev_configure(struct rte_eth_dev *dev)
                memset(mz->addr, 0, mz->len);
 
                hw->rss_conf = mz->addr;
-               hw->rss_confPA = mz->phys_addr;
+               hw->rss_confPA = mz->iova;
        }
 
        return 0;
@@ -441,13 +496,99 @@ vmxnet3_write_mac(struct vmxnet3_hw *hw, const uint8_t *addr)
                     addr[0], addr[1], addr[2],
                     addr[3], addr[4], addr[5]);
 
-       val = *(const uint32_t *)addr;
+       memcpy(&val, addr, 4);
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACL, val);
 
-       val = (addr[5] << 8) | addr[4];
+       memcpy(&val, addr + 4, 2);
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_MACH, val);
 }
 
+static int
+vmxnet3_dev_setup_memreg(struct rte_eth_dev *dev)
+{
+       struct vmxnet3_hw *hw = dev->data->dev_private;
+       Vmxnet3_DriverShared *shared = hw->shared;
+       Vmxnet3_CmdInfo *cmdInfo;
+       struct rte_mempool *mp[VMXNET3_MAX_RX_QUEUES];
+       uint8_t index[VMXNET3_MAX_RX_QUEUES + VMXNET3_MAX_TX_QUEUES];
+       uint32_t num, i, j, size;
+
+       if (hw->memRegsPA == 0) {
+               const struct rte_memzone *mz;
+
+               size = sizeof(Vmxnet3_MemRegs) +
+                       (VMXNET3_MAX_RX_QUEUES + VMXNET3_MAX_TX_QUEUES) *
+                       sizeof(Vmxnet3_MemoryRegion);
+
+               mz = gpa_zone_reserve(dev, size, "memRegs", rte_socket_id(), 8,
+                                     1);
+               if (mz == NULL) {
+                       PMD_INIT_LOG(ERR, "ERROR: Creating memRegs zone");
+                       return -ENOMEM;
+               }
+               memset(mz->addr, 0, mz->len);
+               hw->memRegs = mz->addr;
+               hw->memRegsPA = mz->iova;
+       }
+
+       num = hw->num_rx_queues;
+
+       for (i = 0; i < num; i++) {
+               vmxnet3_rx_queue_t *rxq = dev->data->rx_queues[i];
+
+               mp[i] = rxq->mp;
+               index[i] = 1 << i;
+       }
+
+       /*
+        * The same mempool could be used by multiple queues. In such a case,
+        * remove duplicate mempool entries. Only one entry is kept with
+        * bitmask indicating queues that are using this mempool.
+        */
+       for (i = 1; i < num; i++) {
+               for (j = 0; j < i; j++) {
+                       if (mp[i] == mp[j]) {
+                               mp[i] = NULL;
+                               index[j] |= 1 << i;
+                               break;
+                       }
+               }
+       }
+
+       j = 0;
+       for (i = 0; i < num; i++) {
+               if (mp[i] == NULL)
+                       continue;
+
+               Vmxnet3_MemoryRegion *mr = &hw->memRegs->memRegs[j];
+
+               mr->startPA =
+                       (uintptr_t)STAILQ_FIRST(&mp[i]->mem_list)->iova;
+               mr->length = STAILQ_FIRST(&mp[i]->mem_list)->len <= INT32_MAX ?
+                       STAILQ_FIRST(&mp[i]->mem_list)->len : INT32_MAX;
+               mr->txQueueBits = index[i];
+               mr->rxQueueBits = index[i];
+
+               PMD_INIT_LOG(INFO,
+                            "index: %u startPA: %" PRIu64 " length: %u, "
+                            "rxBits: %x",
+                            j, mr->startPA, mr->length, mr->rxQueueBits);
+               j++;
+       }
+       hw->memRegs->numRegs = j;
+       PMD_INIT_LOG(INFO, "numRegs: %u", j);
+
+       size = sizeof(Vmxnet3_MemRegs) +
+               (j - 1) * sizeof(Vmxnet3_MemoryRegion);
+
+       cmdInfo = &shared->cu.cmdInfo;
+       cmdInfo->varConf.confVer = 1;
+       cmdInfo->varConf.confLen = size;
+       cmdInfo->varConf.confPA = hw->memRegsPA;
+
+       return 0;
+}
+
 static int
 vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
 {
@@ -456,9 +597,12 @@ vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
        uint32_t mtu = dev->data->mtu;
        Vmxnet3_DriverShared *shared = hw->shared;
        Vmxnet3_DSDevRead *devRead = &shared->devRead;
+       uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
        uint32_t i;
        int ret;
 
+       hw->mtu = mtu;
+
        shared->magic = VMXNET3_REV1_MAGIC;
        devRead->misc.driverInfo.version = VMXNET3_DRIVER_VERSION_NUM;
 
@@ -477,17 +621,19 @@ vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
 
        /*
         * Set number of interrupts to 1
-        * PMD disables all the interrupts but this is MUST to activate device
-        * It needs at least one interrupt for link events to handle
-        * So we'll disable it later after device activation if needed
+        * PMD by default disables all the interrupts but this is MUST
+        * to activate device. It needs at least one interrupt for
+        * link events to handle
         */
-       devRead->intrConf.numIntrs = 1;
+       hw->num_intrs = devRead->intrConf.numIntrs = 1;
        devRead->intrConf.intrCtrl |= VMXNET3_IC_DISABLE_ALL;
 
        for (i = 0; i < hw->num_tx_queues; i++) {
                Vmxnet3_TxQueueDesc *tqd = &hw->tqd_start[i];
                vmxnet3_tx_queue_t *txq  = dev->data->tx_queues[i];
 
+               txq->shared = &hw->tqd_start[i];
+
                tqd->ctrl.txNumDeferred  = 0;
                tqd->ctrl.txThreshold    = 1;
                tqd->conf.txRingBasePA   = txq->cmd_ring.basePA;
@@ -497,6 +643,7 @@ vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
                tqd->conf.txRingSize   = txq->cmd_ring.size;
                tqd->conf.compRingSize = txq->comp_ring.size;
                tqd->conf.dataRingSize = txq->data_ring.size;
+               tqd->conf.txDataRingDescSize = txq->txdata_desc_size;
                tqd->conf.intrIdx      = txq->comp_ring.intr_idx;
                tqd->status.stopped    = TRUE;
                tqd->status.error      = 0;
@@ -507,6 +654,8 @@ vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
                Vmxnet3_RxQueueDesc *rqd  = &hw->rqd_start[i];
                vmxnet3_rx_queue_t *rxq   = dev->data->rx_queues[i];
 
+               rxq->shared = &hw->rqd_start[i];
+
                rqd->conf.rxRingBasePA[0] = rxq->cmd_ring[0].basePA;
                rqd->conf.rxRingBasePA[1] = rxq->cmd_ring[1].basePA;
                rqd->conf.compRingBasePA  = rxq->comp_ring.basePA;
@@ -515,6 +664,10 @@ vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
                rqd->conf.rxRingSize[1]   = rxq->cmd_ring[1].size;
                rqd->conf.compRingSize    = rxq->comp_ring.size;
                rqd->conf.intrIdx         = rxq->comp_ring.intr_idx;
+               if (VMXNET3_VERSION_GE_3(hw)) {
+                       rqd->conf.rxDataRingBasePA = rxq->data_ring.basePA;
+                       rqd->conf.rxDataRingDescSize = rxq->data_desc_size;
+               }
                rqd->status.stopped       = TRUE;
                rqd->status.error         = 0;
                memset(&rqd->stats, 0, sizeof(rqd->stats));
@@ -524,10 +677,10 @@ vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
        devRead->rxFilterConf.rxMode = 0;
 
        /* Setting up feature flags */
-       if (dev->data->dev_conf.rxmode.hw_ip_checksum)
+       if (rx_offloads & DEV_RX_OFFLOAD_CHECKSUM)
                devRead->misc.uptFeatures |= VMXNET3_F_RXCSUM;
 
-       if (dev->data->dev_conf.rxmode.enable_lro) {
+       if (rx_offloads & DEV_RX_OFFLOAD_TCP_LRO) {
                devRead->misc.uptFeatures |= VMXNET3_F_LRO;
                devRead->misc.maxNumRxSG = 0;
        }
@@ -543,10 +696,12 @@ vmxnet3_setup_driver_shared(struct rte_eth_dev *dev)
                devRead->rssConfDesc.confPA  = hw->rss_confPA;
        }
 
-       vmxnet3_dev_vlan_offload_set(dev,
-                                    ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK);
+       ret = vmxnet3_dev_vlan_offload_set(dev,
+                       ETH_VLAN_STRIP_MASK | ETH_VLAN_FILTER_MASK);
+       if (ret)
+               return ret;
 
-       vmxnet3_write_mac(hw, hw->perm_addr);
+       vmxnet3_write_mac(hw, dev->data->mac_addrs->addr_bytes);
 
        return VMXNET3_SUCCESS;
 }
@@ -564,10 +719,27 @@ vmxnet3_dev_start(struct rte_eth_dev *dev)
 
        PMD_INIT_FUNC_TRACE();
 
+       /* Save stats before it is reset by CMD_ACTIVATE */
+       vmxnet3_hw_stats_save(hw);
+
        ret = vmxnet3_setup_driver_shared(dev);
        if (ret != VMXNET3_SUCCESS)
                return ret;
 
+       /* check if lsc interrupt feature is enabled */
+       if (dev->data->dev_conf.intr_conf.lsc) {
+               struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
+
+               /* Setup interrupt callback  */
+               rte_intr_callback_register(&pci_dev->intr_handle,
+                                          vmxnet3_interrupt_handler, dev);
+
+               if (rte_intr_enable(&pci_dev->intr_handle) < 0) {
+                       PMD_INIT_LOG(ERR, "interrupt enable failed");
+                       return -EIO;
+               }
+       }
+
        /* Exchange shared data with device */
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL,
                               VMXNET3_GET_ADDR_LO(hw->sharedPA));
@@ -583,6 +755,29 @@ vmxnet3_dev_start(struct rte_eth_dev *dev)
                return -EINVAL;
        }
 
+       /* Setup memory region for rx buffers */
+       ret = vmxnet3_dev_setup_memreg(dev);
+       if (ret == 0) {
+               VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
+                                      VMXNET3_CMD_REGISTER_MEMREGS);
+               ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
+               if (ret != 0)
+                       PMD_INIT_LOG(DEBUG,
+                                    "Failed in setup memory region cmd\n");
+               ret = 0;
+       } else {
+               PMD_INIT_LOG(DEBUG, "Failed to setup memory region\n");
+       }
+
+       if (VMXNET3_VERSION_GE_4(hw)) {
+               /* Check for additional RSS  */
+               ret = vmxnet3_v4_rss_configure(dev);
+               if (ret != VMXNET3_SUCCESS) {
+                       PMD_INIT_LOG(ERR, "Failed to configure v4 RSS");
+                       return ret;
+               }
+       }
+
        /* Disable interrupts */
        vmxnet3_disable_intr(hw);
 
@@ -596,17 +791,24 @@ vmxnet3_dev_start(struct rte_eth_dev *dev)
                return ret;
        }
 
+       hw->adapter_stopped = FALSE;
+
        /* Setting proper Rx Mode and issue Rx Mode Update command */
        vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_UCAST | VMXNET3_RXM_BCAST, 1);
 
-       /*
-        * Don't need to handle events for now
-        */
-#if PROCESS_SYS_EVENTS == 1
-       events = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_ECR);
-       PMD_INIT_LOG(DEBUG, "Reading events: 0x%X", events);
-       vmxnet3_process_events(hw);
-#endif
+       if (dev->data->dev_conf.intr_conf.lsc) {
+               vmxnet3_enable_intr(hw);
+
+               /*
+                * Update link state from device since this won't be
+                * done upon starting with lsc in use. This is done
+                * only after enabling interrupts to avoid any race
+                * where the link state could change without an
+                * interrupt being fired.
+                */
+               __vmxnet3_dev_link_update(dev, 0);
+       }
+
        return VMXNET3_SUCCESS;
 }
 
@@ -622,13 +824,22 @@ vmxnet3_dev_stop(struct rte_eth_dev *dev)
        PMD_INIT_FUNC_TRACE();
 
        if (hw->adapter_stopped == 1) {
-               PMD_INIT_LOG(DEBUG, "Device already closed.");
+               PMD_INIT_LOG(DEBUG, "Device already stopped.");
                return;
        }
 
        /* disable interrupts */
        vmxnet3_disable_intr(hw);
 
+       if (dev->data->dev_conf.intr_conf.lsc) {
+               struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
+
+               rte_intr_disable(&pci_dev->intr_handle);
+
+               rte_intr_callback_unregister(&pci_dev->intr_handle,
+                                            vmxnet3_interrupt_handler, dev);
+       }
+
        /* quiesce the device first */
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_QUIESCE_DEV);
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_DSAL, 0);
@@ -637,13 +848,39 @@ vmxnet3_dev_stop(struct rte_eth_dev *dev)
        /* reset the device */
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_RESET_DEV);
        PMD_INIT_LOG(DEBUG, "Device reset.");
-       hw->adapter_stopped = 0;
 
        vmxnet3_dev_clear_queues(dev);
 
        /* Clear recorded link status */
        memset(&link, 0, sizeof(link));
-       vmxnet3_dev_atomic_write_link_status(dev, &link);
+       link.link_duplex = ETH_LINK_FULL_DUPLEX;
+       link.link_speed = ETH_SPEED_NUM_10G;
+       link.link_autoneg = ETH_LINK_FIXED;
+       rte_eth_linkstatus_set(dev, &link);
+
+       hw->adapter_stopped = 1;
+}
+
+static void
+vmxnet3_free_queues(struct rte_eth_dev *dev)
+{
+       int i;
+
+       PMD_INIT_FUNC_TRACE();
+
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               void *rxq = dev->data->rx_queues[i];
+
+               vmxnet3_dev_rx_queue_release(rxq);
+       }
+       dev->data->nb_rx_queues = 0;
+
+       for (i = 0; i < dev->data->nb_tx_queues; i++) {
+               void *txq = dev->data->tx_queues[i];
+
+               vmxnet3_dev_tx_queue_release(txq);
+       }
+       dev->data->nb_tx_queues = 0;
 }
 
 /*
@@ -652,72 +889,287 @@ vmxnet3_dev_stop(struct rte_eth_dev *dev)
 static void
 vmxnet3_dev_close(struct rte_eth_dev *dev)
 {
-       struct vmxnet3_hw *hw = dev->data->dev_private;
-
        PMD_INIT_FUNC_TRACE();
 
        vmxnet3_dev_stop(dev);
-       hw->adapter_stopped = 1;
+       vmxnet3_free_queues(dev);
+}
+
+static void
+vmxnet3_hw_tx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
+                       struct UPT1_TxStats *res)
+{
+#define VMXNET3_UPDATE_TX_STAT(h, i, f, r)             \
+               ((r)->f = (h)->tqd_start[(i)].stats.f + \
+                       (h)->saved_tx_stats[(i)].f)
+
+       VMXNET3_UPDATE_TX_STAT(hw, q, ucastPktsTxOK, res);
+       VMXNET3_UPDATE_TX_STAT(hw, q, mcastPktsTxOK, res);
+       VMXNET3_UPDATE_TX_STAT(hw, q, bcastPktsTxOK, res);
+       VMXNET3_UPDATE_TX_STAT(hw, q, ucastBytesTxOK, res);
+       VMXNET3_UPDATE_TX_STAT(hw, q, mcastBytesTxOK, res);
+       VMXNET3_UPDATE_TX_STAT(hw, q, bcastBytesTxOK, res);
+       VMXNET3_UPDATE_TX_STAT(hw, q, pktsTxError, res);
+       VMXNET3_UPDATE_TX_STAT(hw, q, pktsTxDiscard, res);
+
+#undef VMXNET3_UPDATE_TX_STAT
+}
+
+static void
+vmxnet3_hw_rx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
+                       struct UPT1_RxStats *res)
+{
+#define VMXNET3_UPDATE_RX_STAT(h, i, f, r)             \
+               ((r)->f = (h)->rqd_start[(i)].stats.f + \
+                       (h)->saved_rx_stats[(i)].f)
+
+       VMXNET3_UPDATE_RX_STAT(hw, q, ucastPktsRxOK, res);
+       VMXNET3_UPDATE_RX_STAT(hw, q, mcastPktsRxOK, res);
+       VMXNET3_UPDATE_RX_STAT(hw, q, bcastPktsRxOK, res);
+       VMXNET3_UPDATE_RX_STAT(hw, q, ucastBytesRxOK, res);
+       VMXNET3_UPDATE_RX_STAT(hw, q, mcastBytesRxOK, res);
+       VMXNET3_UPDATE_RX_STAT(hw, q, bcastBytesRxOK, res);
+       VMXNET3_UPDATE_RX_STAT(hw, q, pktsRxError, res);
+       VMXNET3_UPDATE_RX_STAT(hw, q, pktsRxOutOfBuf, res);
+
+#undef VMXNET3_UPDATE_RX_STAT
 }
 
 static void
+vmxnet3_tx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
+                                       struct UPT1_TxStats *res)
+{
+               vmxnet3_hw_tx_stats_get(hw, q, res);
+
+#define VMXNET3_REDUCE_SNAPSHOT_TX_STAT(h, i, f, r)    \
+               ((r)->f -= (h)->snapshot_tx_stats[(i)].f)
+
+       VMXNET3_REDUCE_SNAPSHOT_TX_STAT(hw, q, ucastPktsTxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_TX_STAT(hw, q, mcastPktsTxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_TX_STAT(hw, q, bcastPktsTxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_TX_STAT(hw, q, ucastBytesTxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_TX_STAT(hw, q, mcastBytesTxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_TX_STAT(hw, q, bcastBytesTxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_TX_STAT(hw, q, pktsTxError, res);
+       VMXNET3_REDUCE_SNAPSHOT_TX_STAT(hw, q, pktsTxDiscard, res);
+
+#undef VMXNET3_REDUCE_SNAPSHOT_TX_STAT
+}
+
+static void
+vmxnet3_rx_stats_get(struct vmxnet3_hw *hw, unsigned int q,
+                                       struct UPT1_RxStats *res)
+{
+               vmxnet3_hw_rx_stats_get(hw, q, res);
+
+#define VMXNET3_REDUCE_SNAPSHOT_RX_STAT(h, i, f, r)    \
+               ((r)->f -= (h)->snapshot_rx_stats[(i)].f)
+
+       VMXNET3_REDUCE_SNAPSHOT_RX_STAT(hw, q, ucastPktsRxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_RX_STAT(hw, q, mcastPktsRxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_RX_STAT(hw, q, bcastPktsRxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_RX_STAT(hw, q, ucastBytesRxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_RX_STAT(hw, q, mcastBytesRxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_RX_STAT(hw, q, bcastBytesRxOK, res);
+       VMXNET3_REDUCE_SNAPSHOT_RX_STAT(hw, q, pktsRxError, res);
+       VMXNET3_REDUCE_SNAPSHOT_RX_STAT(hw, q, pktsRxOutOfBuf, res);
+
+#undef VMXNET3_REDUCE_SNAPSHOT_RX_STAT
+}
+
+static void
+vmxnet3_hw_stats_save(struct vmxnet3_hw *hw)
+{
+       unsigned int i;
+
+       VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
+
+       RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
+
+       for (i = 0; i < hw->num_tx_queues; i++)
+               vmxnet3_hw_tx_stats_get(hw, i, &hw->saved_tx_stats[i]);
+       for (i = 0; i < hw->num_rx_queues; i++)
+               vmxnet3_hw_rx_stats_get(hw, i, &hw->saved_rx_stats[i]);
+}
+
+static int
+vmxnet3_dev_xstats_get_names(struct rte_eth_dev *dev,
+                            struct rte_eth_xstat_name *xstats_names,
+                            unsigned int n)
+{
+       unsigned int i, t, count = 0;
+       unsigned int nstats =
+               dev->data->nb_tx_queues * RTE_DIM(vmxnet3_txq_stat_strings) +
+               dev->data->nb_rx_queues * RTE_DIM(vmxnet3_rxq_stat_strings);
+
+       if (!xstats_names || n < nstats)
+               return nstats;
+
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               if (!dev->data->rx_queues[i])
+                       continue;
+
+               for (t = 0; t < RTE_DIM(vmxnet3_rxq_stat_strings); t++) {
+                       snprintf(xstats_names[count].name,
+                                sizeof(xstats_names[count].name),
+                                "rx_q%u_%s", i,
+                                vmxnet3_rxq_stat_strings[t].name);
+                       count++;
+               }
+       }
+
+       for (i = 0; i < dev->data->nb_tx_queues; i++) {
+               if (!dev->data->tx_queues[i])
+                       continue;
+
+               for (t = 0; t < RTE_DIM(vmxnet3_txq_stat_strings); t++) {
+                       snprintf(xstats_names[count].name,
+                                sizeof(xstats_names[count].name),
+                                "tx_q%u_%s", i,
+                                vmxnet3_txq_stat_strings[t].name);
+                       count++;
+               }
+       }
+
+       return count;
+}
+
+static int
+vmxnet3_dev_xstats_get(struct rte_eth_dev *dev, struct rte_eth_xstat *xstats,
+                      unsigned int n)
+{
+       unsigned int i, t, count = 0;
+       unsigned int nstats =
+               dev->data->nb_tx_queues * RTE_DIM(vmxnet3_txq_stat_strings) +
+               dev->data->nb_rx_queues * RTE_DIM(vmxnet3_rxq_stat_strings);
+
+       if (n < nstats)
+               return nstats;
+
+       for (i = 0; i < dev->data->nb_rx_queues; i++) {
+               struct vmxnet3_rx_queue *rxq = dev->data->rx_queues[i];
+
+               if (rxq == NULL)
+                       continue;
+
+               for (t = 0; t < RTE_DIM(vmxnet3_rxq_stat_strings); t++) {
+                       xstats[count].value = *(uint64_t *)(((char *)&rxq->stats) +
+                               vmxnet3_rxq_stat_strings[t].offset);
+                       xstats[count].id = count;
+                       count++;
+               }
+       }
+
+       for (i = 0; i < dev->data->nb_tx_queues; i++) {
+               struct vmxnet3_tx_queue *txq = dev->data->tx_queues[i];
+
+               if (txq == NULL)
+                       continue;
+
+               for (t = 0; t < RTE_DIM(vmxnet3_txq_stat_strings); t++) {
+                       xstats[count].value = *(uint64_t *)(((char *)&txq->stats) +
+                               vmxnet3_txq_stat_strings[t].offset);
+                       xstats[count].id = count;
+                       count++;
+               }
+       }
+
+       return count;
+}
+
+static int
 vmxnet3_dev_stats_get(struct rte_eth_dev *dev, struct rte_eth_stats *stats)
 {
        unsigned int i;
        struct vmxnet3_hw *hw = dev->data->dev_private;
+       struct UPT1_TxStats txStats;
+       struct UPT1_RxStats rxStats;
 
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
 
        RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
        for (i = 0; i < hw->num_tx_queues; i++) {
-               struct UPT1_TxStats *txStats = &hw->tqd_start[i].stats;
+               vmxnet3_tx_stats_get(hw, i, &txStats);
 
-               stats->q_opackets[i] = txStats->ucastPktsTxOK +
-                                       txStats->mcastPktsTxOK +
-                                       txStats->bcastPktsTxOK;
-               stats->q_obytes[i] = txStats->ucastBytesTxOK +
-                                       txStats->mcastBytesTxOK +
-                                       txStats->bcastBytesTxOK;
+               stats->q_opackets[i] = txStats.ucastPktsTxOK +
+                       txStats.mcastPktsTxOK +
+                       txStats.bcastPktsTxOK;
+
+               stats->q_obytes[i] = txStats.ucastBytesTxOK +
+                       txStats.mcastBytesTxOK +
+                       txStats.bcastBytesTxOK;
 
                stats->opackets += stats->q_opackets[i];
                stats->obytes += stats->q_obytes[i];
-               stats->oerrors += txStats->pktsTxError + txStats->pktsTxDiscard;
+               stats->oerrors += txStats.pktsTxError + txStats.pktsTxDiscard;
        }
 
        RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_RX_QUEUES);
        for (i = 0; i < hw->num_rx_queues; i++) {
-               struct UPT1_RxStats *rxStats = &hw->rqd_start[i].stats;
+               vmxnet3_rx_stats_get(hw, i, &rxStats);
 
-               stats->q_ipackets[i] = rxStats->ucastPktsRxOK +
-                                       rxStats->mcastPktsRxOK +
-                                       rxStats->bcastPktsRxOK;
+               stats->q_ipackets[i] = rxStats.ucastPktsRxOK +
+                       rxStats.mcastPktsRxOK +
+                       rxStats.bcastPktsRxOK;
 
-               stats->q_ibytes[i] = rxStats->ucastBytesRxOK +
-                                       rxStats->mcastBytesRxOK +
-                                       rxStats->bcastBytesRxOK;
+               stats->q_ibytes[i] = rxStats.ucastBytesRxOK +
+                       rxStats.mcastBytesRxOK +
+                       rxStats.bcastBytesRxOK;
 
                stats->ipackets += stats->q_ipackets[i];
                stats->ibytes += stats->q_ibytes[i];
 
-               stats->q_errors[i] = rxStats->pktsRxError;
-               stats->ierrors += rxStats->pktsRxError;
-               stats->rx_nombuf += rxStats->pktsRxOutOfBuf;
+               stats->q_errors[i] = rxStats.pktsRxError;
+               stats->ierrors += rxStats.pktsRxError;
+               stats->imissed += rxStats.pktsRxOutOfBuf;
        }
+
+       return 0;
 }
 
 static void
-vmxnet3_dev_info_get(__rte_unused struct rte_eth_dev *dev,
+vmxnet3_dev_stats_reset(struct rte_eth_dev *dev)
+{
+       unsigned int i;
+       struct vmxnet3_hw *hw = dev->data->dev_private;
+       struct UPT1_TxStats txStats;
+       struct UPT1_RxStats rxStats;
+
+       VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_STATS);
+
+       RTE_BUILD_BUG_ON(RTE_ETHDEV_QUEUE_STAT_CNTRS < VMXNET3_MAX_TX_QUEUES);
+
+       for (i = 0; i < hw->num_tx_queues; i++) {
+               vmxnet3_hw_tx_stats_get(hw, i, &txStats);
+               memcpy(&hw->snapshot_tx_stats[i], &txStats,
+                       sizeof(hw->snapshot_tx_stats[0]));
+       }
+       for (i = 0; i < hw->num_rx_queues; i++) {
+               vmxnet3_hw_rx_stats_get(hw, i, &rxStats);
+               memcpy(&hw->snapshot_rx_stats[i], &rxStats,
+                       sizeof(hw->snapshot_rx_stats[0]));
+       }
+}
+
+static void
+vmxnet3_dev_info_get(struct rte_eth_dev *dev __rte_unused,
                     struct rte_eth_dev_info *dev_info)
 {
+       struct vmxnet3_hw *hw = dev->data->dev_private;
+
        dev_info->max_rx_queues = VMXNET3_MAX_RX_QUEUES;
        dev_info->max_tx_queues = VMXNET3_MAX_TX_QUEUES;
        dev_info->min_rx_bufsize = 1518 + RTE_PKTMBUF_HEADROOM;
        dev_info->max_rx_pktlen = 16384; /* includes CRC, cf MAXFRS register */
+       dev_info->speed_capa = ETH_LINK_SPEED_10G;
        dev_info->max_mac_addrs = VMXNET3_MAX_MAC_ADDRS;
 
-       dev_info->default_txconf.txq_flags = ETH_TXQ_FLAGS_NOXSUMSCTP;
        dev_info->flow_type_rss_offloads = VMXNET3_RSS_OFFLOAD_ALL;
 
+       if (VMXNET3_VERSION_GE_4(hw)) {
+               dev_info->flow_type_rss_offloads |= VMXNET3_V4_RSS_MASK;
+       }
+
        dev_info->rx_desc_lim = (struct rte_eth_desc_lim) {
                .nb_max = VMXNET3_RX_RING_MAX_SIZE,
                .nb_min = VMXNET3_DEF_RX_RING_SIZE,
@@ -728,19 +1180,14 @@ vmxnet3_dev_info_get(__rte_unused struct rte_eth_dev *dev,
                .nb_max = VMXNET3_TX_RING_MAX_SIZE,
                .nb_min = VMXNET3_DEF_TX_RING_SIZE,
                .nb_align = 1,
+               .nb_seg_max = VMXNET3_TX_MAX_SEG,
+               .nb_mtu_seg_max = VMXNET3_MAX_TXD_PER_PKT,
        };
 
-       dev_info->rx_offload_capa =
-               DEV_RX_OFFLOAD_VLAN_STRIP |
-               DEV_RX_OFFLOAD_UDP_CKSUM |
-               DEV_RX_OFFLOAD_TCP_CKSUM |
-               DEV_RX_OFFLOAD_TCP_LRO;
-
-       dev_info->tx_offload_capa =
-               DEV_TX_OFFLOAD_VLAN_INSERT |
-               DEV_TX_OFFLOAD_TCP_CKSUM |
-               DEV_TX_OFFLOAD_UDP_CKSUM |
-               DEV_TX_OFFLOAD_TCP_TSO;
+       dev_info->rx_offload_capa = VMXNET3_RX_OFFLOAD_CAP;
+       dev_info->rx_queue_offload_capa = 0;
+       dev_info->tx_offload_capa = VMXNET3_TX_OFFLOAD_CAP;
+       dev_info->tx_queue_offload_capa = 0;
 }
 
 static const uint32_t *
@@ -757,43 +1204,47 @@ vmxnet3_dev_supported_ptypes_get(struct rte_eth_dev *dev)
        return NULL;
 }
 
-static void
-vmxnet3_mac_addr_set(struct rte_eth_dev *dev, struct ether_addr *mac_addr)
+static int
+vmxnet3_mac_addr_set(struct rte_eth_dev *dev, struct rte_ether_addr *mac_addr)
 {
        struct vmxnet3_hw *hw = dev->data->dev_private;
 
+       ether_addr_copy(mac_addr, (struct rte_ether_addr *)(hw->perm_addr));
        vmxnet3_write_mac(hw, mac_addr->addr_bytes);
+       return 0;
 }
 
 /* return 0 means link status changed, -1 means not changed */
 static int
-vmxnet3_dev_link_update(struct rte_eth_dev *dev,
-                       __rte_unused int wait_to_complete)
+__vmxnet3_dev_link_update(struct rte_eth_dev *dev,
+                         __rte_unused int wait_to_complete)
 {
        struct vmxnet3_hw *hw = dev->data->dev_private;
-       struct rte_eth_link old, link;
+       struct rte_eth_link link;
        uint32_t ret;
 
-       /* Link status doesn't change for stopped dev */
-       if (dev->data->dev_started == 0)
-               return -1;
-
        memset(&link, 0, sizeof(link));
-       vmxnet3_dev_atomic_read_link_status(dev, &old);
 
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD, VMXNET3_CMD_GET_LINK);
        ret = VMXNET3_READ_BAR1_REG(hw, VMXNET3_REG_CMD);
 
-       if (ret & 0x1) {
+       if (ret & 0x1)
                link.link_status = ETH_LINK_UP;
-               link.link_duplex = ETH_LINK_FULL_DUPLEX;
-               link.link_speed = ETH_SPEED_NUM_10G;
-               link.link_autoneg = ETH_LINK_SPEED_FIXED;
-       }
+       link.link_duplex = ETH_LINK_FULL_DUPLEX;
+       link.link_speed = ETH_SPEED_NUM_10G;
+       link.link_autoneg = ETH_LINK_FIXED;
 
-       vmxnet3_dev_atomic_write_link_status(dev, &link);
+       return rte_eth_linkstatus_set(dev, &link);
+}
 
-       return (old.link_status == link.link_status) ? -1 : 0;
+static int
+vmxnet3_dev_link_update(struct rte_eth_dev *dev, int wait_to_complete)
+{
+       /* Link status doesn't change for stopped dev */
+       if (dev->data->dev_started == 0)
+               return -1;
+
+       return __vmxnet3_dev_link_update(dev, wait_to_complete);
 }
 
 /* Updating rxmode through Vmxnet3_DriverShared structure in adapter */
@@ -830,8 +1281,12 @@ vmxnet3_dev_promiscuous_disable(struct rte_eth_dev *dev)
 {
        struct vmxnet3_hw *hw = dev->data->dev_private;
        uint32_t *vf_table = hw->shared->devRead.rxFilterConf.vfTable;
+       uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
 
-       memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
+       if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
+               memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
+       else
+               memset(vf_table, 0xff, VMXNET3_VFT_TABLE_SIZE);
        vmxnet3_dev_set_rxmode(hw, VMXNET3_RXM_PROMISC, 0);
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
                               VMXNET3_CMD_UPDATE_VLAN_FILTERS);
@@ -884,15 +1339,16 @@ vmxnet3_dev_vlan_filter_set(struct rte_eth_dev *dev, uint16_t vid, int on)
        return 0;
 }
 
-static void
+static int
 vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
 {
        struct vmxnet3_hw *hw = dev->data->dev_private;
        Vmxnet3_DSDevRead *devRead = &hw->shared->devRead;
        uint32_t *vf_table = devRead->rxFilterConf.vfTable;
+       uint64_t rx_offloads = dev->data->dev_conf.rxmode.offloads;
 
        if (mask & ETH_VLAN_STRIP_MASK) {
-               if (dev->data->dev_conf.rxmode.hw_vlan_strip)
+               if (rx_offloads & DEV_RX_OFFLOAD_VLAN_STRIP)
                        devRead->misc.uptFeatures |= UPT1_F_RXVLAN;
                else
                        devRead->misc.uptFeatures &= ~UPT1_F_RXVLAN;
@@ -902,7 +1358,7 @@ vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
        }
 
        if (mask & ETH_VLAN_FILTER_MASK) {
-               if (dev->data->dev_conf.rxmode.hw_vlan_filter)
+               if (rx_offloads & DEV_RX_OFFLOAD_VLAN_FILTER)
                        memcpy(vf_table, hw->shadow_vfta, VMXNET3_VFT_TABLE_SIZE);
                else
                        memset(vf_table, 0xff, VMXNET3_VFT_TABLE_SIZE);
@@ -910,18 +1366,18 @@ vmxnet3_dev_vlan_offload_set(struct rte_eth_dev *dev, int mask)
                VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_CMD,
                                       VMXNET3_CMD_UPDATE_VLAN_FILTERS);
        }
+
+       return 0;
 }
 
-#if PROCESS_SYS_EVENTS == 1
 static void
-vmxnet3_process_events(struct vmxnet3_hw *hw)
+vmxnet3_process_events(struct rte_eth_dev *dev)
 {
+       struct vmxnet3_hw *hw = dev->data->dev_private;
        uint32_t events = hw->shared->ecr;
 
-       if (!events) {
-               PMD_INIT_LOG(ERR, "No events to process");
+       if (!events)
                return;
-       }
 
        /*
         * ECR bits when written with 1b are cleared. Hence write
@@ -930,10 +1386,13 @@ vmxnet3_process_events(struct vmxnet3_hw *hw)
        VMXNET3_WRITE_BAR1_REG(hw, VMXNET3_REG_ECR, events);
 
        /* Check if link state has changed */
-       if (events & VMXNET3_ECR_LINK)
-               PMD_INIT_LOG(ERR,
-                            "Process events in %s(): VMXNET3_ECR_LINK event",
-                            __func__);
+       if (events & VMXNET3_ECR_LINK) {
+               PMD_DRV_LOG(DEBUG, "Process events: VMXNET3_ECR_LINK event");
+               if (vmxnet3_dev_link_update(dev, 0) == 0)
+                       _rte_eth_dev_callback_process(dev,
+                                                     RTE_ETH_EVENT_INTR_LSC,
+                                                     NULL);
+       }
 
        /* Check if there is an error on xmit/recv queues */
        if (events & (VMXNET3_ECR_TQERR | VMXNET3_ECR_RQERR)) {
@@ -941,11 +1400,11 @@ vmxnet3_process_events(struct vmxnet3_hw *hw)
                                       VMXNET3_CMD_GET_QUEUE_STATUS);
 
                if (hw->tqd_start->status.stopped)
-                       PMD_INIT_LOG(ERR, "tq error 0x%x",
-                                    hw->tqd_start->status.error);
+                       PMD_DRV_LOG(ERR, "tq error 0x%x",
+                                   hw->tqd_start->status.error);
 
                if (hw->rqd_start->status.stopped)
-                       PMD_INIT_LOG(ERR, "rq error 0x%x",
+                       PMD_DRV_LOG(ERR, "rq error 0x%x",
                                     hw->rqd_start->status.error);
 
                /* Reset the device */
@@ -953,12 +1412,34 @@ vmxnet3_process_events(struct vmxnet3_hw *hw)
        }
 
        if (events & VMXNET3_ECR_DIC)
-               PMD_INIT_LOG(ERR, "Device implementation change event.");
+               PMD_DRV_LOG(DEBUG, "Device implementation change event.");
 
        if (events & VMXNET3_ECR_DEBUG)
-               PMD_INIT_LOG(ERR, "Debug event generated by device.");
+               PMD_DRV_LOG(DEBUG, "Debug event generated by device.");
 }
-#endif
 
-DRIVER_REGISTER_PCI(net_vmxnet3, rte_vmxnet3_pmd.pci_drv);
-DRIVER_REGISTER_PCI_TABLE(net_vmxnet3, pci_id_vmxnet3_map);
+static void
+vmxnet3_interrupt_handler(void *param)
+{
+       struct rte_eth_dev *dev = param;
+       struct rte_pci_device *pci_dev = RTE_DEV_TO_PCI(dev->device);
+
+       vmxnet3_process_events(dev);
+
+       if (rte_intr_enable(&pci_dev->intr_handle) < 0)
+               PMD_DRV_LOG(ERR, "interrupt enable failed");
+}
+
+RTE_PMD_REGISTER_PCI(net_vmxnet3, rte_vmxnet3_pmd);
+RTE_PMD_REGISTER_PCI_TABLE(net_vmxnet3, pci_id_vmxnet3_map);
+RTE_PMD_REGISTER_KMOD_DEP(net_vmxnet3, "* igb_uio | uio_pci_generic | vfio-pci");
+
+RTE_INIT(vmxnet3_init_log)
+{
+       vmxnet3_logtype_init = rte_log_register("pmd.net.vmxnet3.init");
+       if (vmxnet3_logtype_init >= 0)
+               rte_log_set_level(vmxnet3_logtype_init, RTE_LOG_NOTICE);
+       vmxnet3_logtype_driver = rte_log_register("pmd.net.vmxnet3.driver");
+       if (vmxnet3_logtype_driver >= 0)
+               rte_log_set_level(vmxnet3_logtype_driver, RTE_LOG_NOTICE);
+}