examples/vhost: copy old vhost example
authorHuawei Xie <huawei.xie@intel.com>
Mon, 20 Oct 2014 04:38:13 +0000 (12:38 +0800)
committerThomas Monjalon <thomas.monjalon@6wind.com>
Wed, 22 Oct 2014 17:00:45 +0000 (19:00 +0200)
This patch copies two files main.c/main.h from most recent vhost example
(before transforming into a library) as the base for new vhost example.

Signed-off-by: Huawei Xie <huawei.xie@intel.com>
Acked-by: Changchun Ouyang <changchun.ouyang@intel.com>
examples/vhost/main.c [new file with mode: 0644]
examples/vhost/main.h [new file with mode: 0644]

diff --git a/examples/vhost/main.c b/examples/vhost/main.c
new file mode 100644 (file)
index 0000000..c23d453
--- /dev/null
@@ -0,0 +1,3727 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2014 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.
+ */
+
+#include <arpa/inet.h>
+#include <getopt.h>
+#include <linux/if_ether.h>
+#include <linux/if_vlan.h>
+#include <linux/virtio_net.h>
+#include <linux/virtio_ring.h>
+#include <signal.h>
+#include <stdint.h>
+#include <sys/eventfd.h>
+#include <sys/param.h>
+#include <unistd.h>
+
+#include <rte_atomic.h>
+#include <rte_cycles.h>
+#include <rte_ethdev.h>
+#include <rte_log.h>
+#include <rte_string_fns.h>
+#include <rte_malloc.h>
+
+#include "main.h"
+#include "virtio-net.h"
+#include "vhost-net-cdev.h"
+
+#define MAX_QUEUES 128
+
+/* the maximum number of external ports supported */
+#define MAX_SUP_PORTS 1
+
+/*
+ * Calculate the number of buffers needed per port
+ */
+#define NUM_MBUFS_PER_PORT ((MAX_QUEUES*RTE_TEST_RX_DESC_DEFAULT) +            \
+                                                       (num_switching_cores*MAX_PKT_BURST) +                   \
+                                                       (num_switching_cores*RTE_TEST_TX_DESC_DEFAULT) +\
+                                                       (num_switching_cores*MBUF_CACHE_SIZE))
+
+#define MBUF_CACHE_SIZE 128
+#define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
+
+/*
+ * No frame data buffer allocated from host are required for zero copy
+ * implementation, guest will allocate the frame data buffer, and vhost
+ * directly use it.
+ */
+#define VIRTIO_DESCRIPTOR_LEN_ZCP 1518
+#define MBUF_SIZE_ZCP (VIRTIO_DESCRIPTOR_LEN_ZCP + sizeof(struct rte_mbuf) \
+       + RTE_PKTMBUF_HEADROOM)
+#define MBUF_CACHE_SIZE_ZCP 0
+
+/*
+ * RX and TX Prefetch, Host, and Write-back threshold values should be
+ * carefully set for optimal performance. Consult the network
+ * controller's datasheet and supporting DPDK documentation for guidance
+ * on how these parameters should be set.
+ */
+#define RX_PTHRESH 8 /* Default values of RX prefetch threshold reg. */
+#define RX_HTHRESH 8 /* Default values of RX host threshold reg. */
+#define RX_WTHRESH 4 /* Default values of RX write-back threshold reg. */
+
+/*
+ * These default values are optimized for use with the Intel(R) 82599 10 GbE
+ * Controller and the DPDK ixgbe PMD. Consider using other values for other
+ * network controllers and/or network drivers.
+ */
+#define TX_PTHRESH 36 /* Default values of TX prefetch threshold reg. */
+#define TX_HTHRESH 0  /* Default values of TX host threshold reg. */
+#define TX_WTHRESH 0  /* Default values of TX write-back threshold reg. */
+
+#define MAX_PKT_BURST 32               /* Max burst size for RX/TX */
+#define MAX_MRG_PKT_BURST 16   /* Max burst for merge buffers. Set to 1 due to performance issue. */
+#define BURST_TX_DRAIN_US 100  /* TX drain every ~100us */
+
+#define BURST_RX_WAIT_US 15    /* Defines how long we wait between retries on RX */
+#define BURST_RX_RETRIES 4             /* Number of retries on RX. */
+
+#define JUMBO_FRAME_MAX_SIZE    0x2600
+
+/* State of virtio device. */
+#define DEVICE_MAC_LEARNING 0
+#define DEVICE_RX                      1
+#define DEVICE_SAFE_REMOVE     2
+
+/* Config_core_flag status definitions. */
+#define REQUEST_DEV_REMOVAL 1
+#define ACK_DEV_REMOVAL 0
+
+/* Configurable number of RX/TX ring descriptors */
+#define RTE_TEST_RX_DESC_DEFAULT 1024
+#define RTE_TEST_TX_DESC_DEFAULT 512
+
+/*
+ * Need refine these 2 macros for legacy and DPDK based front end:
+ * Max vring avail descriptor/entries from guest - MAX_PKT_BURST
+ * And then adjust power 2.
+ */
+/*
+ * For legacy front end, 128 descriptors,
+ * half for virtio header, another half for mbuf.
+ */
+#define RTE_TEST_RX_DESC_DEFAULT_ZCP 32   /* legacy: 32, DPDK virt FE: 128. */
+#define RTE_TEST_TX_DESC_DEFAULT_ZCP 64   /* legacy: 64, DPDK virt FE: 64.  */
+
+/* Get first 4 bytes in mbuf headroom. */
+#define MBUF_HEADROOM_UINT32(mbuf) (*(uint32_t *)((uint8_t *)(mbuf) \
+               + sizeof(struct rte_mbuf)))
+
+/* true if x is a power of 2 */
+#define POWEROF2(x) ((((x)-1) & (x)) == 0)
+
+#define INVALID_PORT_ID 0xFF
+
+/* Max number of devices. Limited by vmdq. */
+#define MAX_DEVICES 64
+
+/* Size of buffers used for snprintfs. */
+#define MAX_PRINT_BUFF 6072
+
+/* Maximum character device basename size. */
+#define MAX_BASENAME_SZ 10
+
+/* Maximum long option length for option parsing. */
+#define MAX_LONG_OPT_SZ 64
+
+/* Used to compare MAC addresses. */
+#define MAC_ADDR_CMP 0xFFFFFFFFFFFFULL
+
+/* Number of descriptors per cacheline. */
+#define DESC_PER_CACHELINE (CACHE_LINE_SIZE / sizeof(struct vring_desc))
+
+/* mask of enabled ports */
+static uint32_t enabled_port_mask = 0;
+
+/*Number of switching cores enabled*/
+static uint32_t num_switching_cores = 0;
+
+/* number of devices/queues to support*/
+static uint32_t num_queues = 0;
+uint32_t num_devices = 0;
+
+/*
+ * Enable zero copy, pkts buffer will directly dma to hw descriptor,
+ * disabled on default.
+ */
+static uint32_t zero_copy;
+
+/* number of descriptors to apply*/
+static uint32_t num_rx_descriptor = RTE_TEST_RX_DESC_DEFAULT_ZCP;
+static uint32_t num_tx_descriptor = RTE_TEST_TX_DESC_DEFAULT_ZCP;
+
+/* max ring descriptor, ixgbe, i40e, e1000 all are 4096. */
+#define MAX_RING_DESC 4096
+
+struct vpool {
+       struct rte_mempool *pool;
+       struct rte_ring *ring;
+       uint32_t buf_size;
+} vpool_array[MAX_QUEUES+MAX_QUEUES];
+
+/* Enable VM2VM communications. If this is disabled then the MAC address compare is skipped. */
+typedef enum {
+       VM2VM_DISABLED = 0,
+       VM2VM_SOFTWARE = 1,
+       VM2VM_HARDWARE = 2,
+       VM2VM_LAST
+} vm2vm_type;
+static vm2vm_type vm2vm_mode = VM2VM_SOFTWARE;
+
+/* The type of host physical address translated from guest physical address. */
+typedef enum {
+       PHYS_ADDR_CONTINUOUS = 0,
+       PHYS_ADDR_CROSS_SUBREG = 1,
+       PHYS_ADDR_INVALID = 2,
+       PHYS_ADDR_LAST
+} hpa_type;
+
+/* Enable stats. */
+static uint32_t enable_stats = 0;
+/* Enable retries on RX. */
+static uint32_t enable_retry = 1;
+/* Specify timeout (in useconds) between retries on RX. */
+static uint32_t burst_rx_delay_time = BURST_RX_WAIT_US;
+/* Specify the number of retries on RX. */
+static uint32_t burst_rx_retry_num = BURST_RX_RETRIES;
+
+/* Character device basename. Can be set by user. */
+static char dev_basename[MAX_BASENAME_SZ] = "vhost-net";
+
+/* Charater device index. Can be set by user. */
+static uint32_t dev_index = 0;
+
+/* This can be set by the user so it is made available here. */
+extern uint64_t VHOST_FEATURES;
+
+/* Default configuration for rx and tx thresholds etc. */
+static struct rte_eth_rxconf rx_conf_default = {
+       .rx_thresh = {
+               .pthresh = RX_PTHRESH,
+               .hthresh = RX_HTHRESH,
+               .wthresh = RX_WTHRESH,
+       },
+       .rx_drop_en = 1,
+};
+
+/*
+ * These default values are optimized for use with the Intel(R) 82599 10 GbE
+ * Controller and the DPDK ixgbe/igb PMD. Consider using other values for other
+ * network controllers and/or network drivers.
+ */
+static struct rte_eth_txconf tx_conf_default = {
+       .tx_thresh = {
+               .pthresh = TX_PTHRESH,
+               .hthresh = TX_HTHRESH,
+               .wthresh = TX_WTHRESH,
+       },
+       .tx_free_thresh = 0, /* Use PMD default values */
+       .tx_rs_thresh = 0, /* Use PMD default values */
+};
+
+/* empty vmdq configuration structure. Filled in programatically */
+static struct rte_eth_conf vmdq_conf_default = {
+       .rxmode = {
+               .mq_mode        = ETH_MQ_RX_VMDQ_ONLY,
+               .split_hdr_size = 0,
+               .header_split   = 0, /**< Header Split disabled */
+               .hw_ip_checksum = 0, /**< IP checksum offload disabled */
+               .hw_vlan_filter = 0, /**< VLAN filtering disabled */
+               /*
+                * It is necessary for 1G NIC such as I350,
+                * this fixes bug of ipv4 forwarding in guest can't
+                * forward pakets from one virtio dev to another virtio dev.
+                */
+               .hw_vlan_strip  = 1, /**< VLAN strip enabled. */
+               .jumbo_frame    = 0, /**< Jumbo Frame Support disabled */
+               .hw_strip_crc   = 0, /**< CRC stripped by hardware */
+       },
+
+       .txmode = {
+               .mq_mode = ETH_MQ_TX_NONE,
+       },
+       .rx_adv_conf = {
+               /*
+                * should be overridden separately in code with
+                * appropriate values
+                */
+               .vmdq_rx_conf = {
+                       .nb_queue_pools = ETH_8_POOLS,
+                       .enable_default_pool = 0,
+                       .default_pool = 0,
+                       .nb_pool_maps = 0,
+                       .pool_map = {{0, 0},},
+               },
+       },
+};
+
+static unsigned lcore_ids[RTE_MAX_LCORE];
+static uint8_t ports[RTE_MAX_ETHPORTS];
+static unsigned num_ports = 0; /**< The number of ports specified in command line */
+
+static const uint16_t external_pkt_default_vlan_tag = 2000;
+const uint16_t vlan_tags[] = {
+       1000, 1001, 1002, 1003, 1004, 1005, 1006, 1007,
+       1008, 1009, 1010, 1011, 1012, 1013, 1014, 1015,
+       1016, 1017, 1018, 1019, 1020, 1021, 1022, 1023,
+       1024, 1025, 1026, 1027, 1028, 1029, 1030, 1031,
+       1032, 1033, 1034, 1035, 1036, 1037, 1038, 1039,
+       1040, 1041, 1042, 1043, 1044, 1045, 1046, 1047,
+       1048, 1049, 1050, 1051, 1052, 1053, 1054, 1055,
+       1056, 1057, 1058, 1059, 1060, 1061, 1062, 1063,
+};
+
+/* ethernet addresses of ports */
+static struct ether_addr vmdq_ports_eth_addr[RTE_MAX_ETHPORTS];
+
+/* heads for the main used and free linked lists for the data path. */
+static struct virtio_net_data_ll *ll_root_used = NULL;
+static struct virtio_net_data_ll *ll_root_free = NULL;
+
+/* Array of data core structures containing information on individual core linked lists. */
+static struct lcore_info lcore_info[RTE_MAX_LCORE];
+
+/* Used for queueing bursts of TX packets. */
+struct mbuf_table {
+       unsigned len;
+       unsigned txq_id;
+       struct rte_mbuf *m_table[MAX_PKT_BURST];
+};
+
+/* TX queue for each data core. */
+struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
+
+/* TX queue fori each virtio device for zero copy. */
+struct mbuf_table tx_queue_zcp[MAX_QUEUES];
+
+/* Vlan header struct used to insert vlan tags on TX. */
+struct vlan_ethhdr {
+       unsigned char   h_dest[ETH_ALEN];
+       unsigned char   h_source[ETH_ALEN];
+       __be16          h_vlan_proto;
+       __be16          h_vlan_TCI;
+       __be16          h_vlan_encapsulated_proto;
+};
+
+/* IPv4 Header */
+struct ipv4_hdr {
+       uint8_t  version_ihl;           /**< version and header length */
+       uint8_t  type_of_service;       /**< type of service */
+       uint16_t total_length;          /**< length of packet */
+       uint16_t packet_id;             /**< packet ID */
+       uint16_t fragment_offset;       /**< fragmentation offset */
+       uint8_t  time_to_live;          /**< time to live */
+       uint8_t  next_proto_id;         /**< protocol ID */
+       uint16_t hdr_checksum;          /**< header checksum */
+       uint32_t src_addr;              /**< source address */
+       uint32_t dst_addr;              /**< destination address */
+} __attribute__((__packed__));
+
+/* Header lengths. */
+#define VLAN_HLEN       4
+#define VLAN_ETH_HLEN   18
+
+/* Per-device statistics struct */
+struct device_statistics {
+       uint64_t tx_total;
+       rte_atomic64_t rx_total_atomic;
+       uint64_t rx_total;
+       uint64_t tx;
+       rte_atomic64_t rx_atomic;
+       uint64_t rx;
+} __rte_cache_aligned;
+struct device_statistics dev_statistics[MAX_DEVICES];
+
+/*
+ * Builds up the correct configuration for VMDQ VLAN pool map
+ * according to the pool & queue limits.
+ */
+static inline int
+get_eth_conf(struct rte_eth_conf *eth_conf, uint32_t num_devices)
+{
+       struct rte_eth_vmdq_rx_conf conf;
+       unsigned i;
+
+       memset(&conf, 0, sizeof(conf));
+       conf.nb_queue_pools = (enum rte_eth_nb_pools)num_devices;
+       conf.nb_pool_maps = num_devices;
+       conf.enable_loop_back =
+               vmdq_conf_default.rx_adv_conf.vmdq_rx_conf.enable_loop_back;
+
+       for (i = 0; i < conf.nb_pool_maps; i++) {
+               conf.pool_map[i].vlan_id = vlan_tags[ i ];
+               conf.pool_map[i].pools = (1UL << i);
+       }
+
+       (void)(rte_memcpy(eth_conf, &vmdq_conf_default, sizeof(*eth_conf)));
+       (void)(rte_memcpy(&eth_conf->rx_adv_conf.vmdq_rx_conf, &conf,
+                  sizeof(eth_conf->rx_adv_conf.vmdq_rx_conf)));
+       return 0;
+}
+
+/*
+ * Validate the device number according to the max pool number gotten form
+ * dev_info. If the device number is invalid, give the error message and
+ * return -1. Each device must have its own pool.
+ */
+static inline int
+validate_num_devices(uint32_t max_nb_devices)
+{
+       if (num_devices > max_nb_devices) {
+               RTE_LOG(ERR, VHOST_PORT, "invalid number of devices\n");
+               return -1;
+       }
+       return 0;
+}
+
+/*
+ * Initialises a given port using global settings and with the rx buffers
+ * coming from the mbuf_pool passed as parameter
+ */
+static inline int
+port_init(uint8_t port)
+{
+       struct rte_eth_dev_info dev_info;
+       struct rte_eth_conf port_conf;
+       uint16_t rx_rings, tx_rings;
+       uint16_t rx_ring_size, tx_ring_size;
+       int retval;
+       uint16_t q;
+
+       /* The max pool number from dev_info will be used to validate the pool number specified in cmd line */
+       rte_eth_dev_info_get (port, &dev_info);
+
+       /*configure the number of supported virtio devices based on VMDQ limits */
+       num_devices = dev_info.max_vmdq_pools;
+       num_queues = dev_info.max_rx_queues;
+
+       if (zero_copy) {
+               rx_ring_size = num_rx_descriptor;
+               tx_ring_size = num_tx_descriptor;
+               tx_rings = dev_info.max_tx_queues;
+       } else {
+               rx_ring_size = RTE_TEST_RX_DESC_DEFAULT;
+               tx_ring_size = RTE_TEST_TX_DESC_DEFAULT;
+               tx_rings = (uint16_t)rte_lcore_count();
+       }
+
+       retval = validate_num_devices(MAX_DEVICES);
+       if (retval < 0)
+               return retval;
+
+       /* Get port configuration. */
+       retval = get_eth_conf(&port_conf, num_devices);
+       if (retval < 0)
+               return retval;
+
+       if (port >= rte_eth_dev_count()) return -1;
+
+       rx_rings = (uint16_t)num_queues,
+       /* Configure ethernet device. */
+       retval = rte_eth_dev_configure(port, rx_rings, tx_rings, &port_conf);
+       if (retval != 0)
+               return retval;
+
+       /* Setup the queues. */
+       for (q = 0; q < rx_rings; q ++) {
+               retval = rte_eth_rx_queue_setup(port, q, rx_ring_size,
+                                               rte_eth_dev_socket_id(port), &rx_conf_default,
+                                               vpool_array[q].pool);
+               if (retval < 0)
+                       return retval;
+       }
+       for (q = 0; q < tx_rings; q ++) {
+               retval = rte_eth_tx_queue_setup(port, q, tx_ring_size,
+                                               rte_eth_dev_socket_id(port), &tx_conf_default);
+               if (retval < 0)
+                       return retval;
+       }
+
+       /* Start the device. */
+       retval  = rte_eth_dev_start(port);
+       if (retval < 0) {
+               RTE_LOG(ERR, VHOST_DATA, "Failed to start the device.\n");
+               return retval;
+       }
+
+       rte_eth_macaddr_get(port, &vmdq_ports_eth_addr[port]);
+       RTE_LOG(INFO, VHOST_PORT, "Max virtio devices supported: %u\n", num_devices);
+       RTE_LOG(INFO, VHOST_PORT, "Port %u MAC: %02"PRIx8" %02"PRIx8" %02"PRIx8
+                       " %02"PRIx8" %02"PRIx8" %02"PRIx8"\n",
+                       (unsigned)port,
+                       vmdq_ports_eth_addr[port].addr_bytes[0],
+                       vmdq_ports_eth_addr[port].addr_bytes[1],
+                       vmdq_ports_eth_addr[port].addr_bytes[2],
+                       vmdq_ports_eth_addr[port].addr_bytes[3],
+                       vmdq_ports_eth_addr[port].addr_bytes[4],
+                       vmdq_ports_eth_addr[port].addr_bytes[5]);
+
+       return 0;
+}
+
+/*
+ * Set character device basename.
+ */
+static int
+us_vhost_parse_basename(const char *q_arg)
+{
+       /* parse number string */
+
+       if (strnlen(q_arg, MAX_BASENAME_SZ) > MAX_BASENAME_SZ)
+               return -1;
+       else
+               snprintf((char*)&dev_basename, MAX_BASENAME_SZ, "%s", q_arg);
+
+       return 0;
+}
+
+/*
+ * Parse the portmask provided at run time.
+ */
+static int
+parse_portmask(const char *portmask)
+{
+       char *end = NULL;
+       unsigned long pm;
+
+       errno = 0;
+
+       /* parse hexadecimal string */
+       pm = strtoul(portmask, &end, 16);
+       if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0') || (errno != 0))
+               return -1;
+
+       if (pm == 0)
+               return -1;
+
+       return pm;
+
+}
+
+/*
+ * Parse num options at run time.
+ */
+static int
+parse_num_opt(const char *q_arg, uint32_t max_valid_value)
+{
+       char *end = NULL;
+       unsigned long num;
+
+       errno = 0;
+
+       /* parse unsigned int string */
+       num = strtoul(q_arg, &end, 10);
+       if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0') || (errno != 0))
+               return -1;
+
+       if (num > max_valid_value)
+               return -1;
+
+       return num;
+
+}
+
+/*
+ * Display usage
+ */
+static void
+us_vhost_usage(const char *prgname)
+{
+       RTE_LOG(INFO, VHOST_CONFIG, "%s [EAL options] -- -p PORTMASK\n"
+       "               --vm2vm [0|1|2]\n"
+       "               --rx_retry [0|1] --mergeable [0|1] --stats [0-N]\n"
+       "               --dev-basename <name> --dev-index [0-N]\n"
+       "               --nb-devices ND\n"
+       "               -p PORTMASK: Set mask for ports to be used by application\n"
+       "               --vm2vm [0|1|2]: disable/software(default)/hardware vm2vm comms\n"
+       "               --rx-retry [0|1]: disable/enable(default) retries on rx. Enable retry if destintation queue is full\n"
+       "               --rx-retry-delay [0-N]: timeout(in usecond) between retries on RX. This makes effect only if retries on rx enabled\n"
+       "               --rx-retry-num [0-N]: the number of retries on rx. This makes effect only if retries on rx enabled\n"
+       "               --mergeable [0|1]: disable(default)/enable RX mergeable buffers\n"
+       "               --stats [0-N]: 0: Disable stats, N: Time in seconds to print stats\n"
+       "               --dev-basename: The basename to be used for the character device.\n"
+       "               --dev-index [0-N]: Defaults to zero if not used. Index is appended to basename.\n"
+       "               --zero-copy [0|1]: disable(default)/enable rx/tx "
+                       "zero copy\n"
+       "               --rx-desc-num [0-N]: the number of descriptors on rx, "
+                       "used only when zero copy is enabled.\n"
+       "               --tx-desc-num [0-N]: the number of descriptors on tx, "
+                       "used only when zero copy is enabled.\n",
+              prgname);
+}
+
+/*
+ * Parse the arguments given in the command line of the application.
+ */
+static int
+us_vhost_parse_args(int argc, char **argv)
+{
+       int opt, ret;
+       int option_index;
+       unsigned i;
+       const char *prgname = argv[0];
+       static struct option long_option[] = {
+               {"vm2vm", required_argument, NULL, 0},
+               {"rx-retry", required_argument, NULL, 0},
+               {"rx-retry-delay", required_argument, NULL, 0},
+               {"rx-retry-num", required_argument, NULL, 0},
+               {"mergeable", required_argument, NULL, 0},
+               {"stats", required_argument, NULL, 0},
+               {"dev-basename", required_argument, NULL, 0},
+               {"dev-index", required_argument, NULL, 0},
+               {"zero-copy", required_argument, NULL, 0},
+               {"rx-desc-num", required_argument, NULL, 0},
+               {"tx-desc-num", required_argument, NULL, 0},
+               {NULL, 0, 0, 0},
+       };
+
+       /* Parse command line */
+       while ((opt = getopt_long(argc, argv, "p:",long_option, &option_index)) != EOF) {
+               switch (opt) {
+               /* Portmask */
+               case 'p':
+                       enabled_port_mask = parse_portmask(optarg);
+                       if (enabled_port_mask == 0) {
+                               RTE_LOG(INFO, VHOST_CONFIG, "Invalid portmask\n");
+                               us_vhost_usage(prgname);
+                               return -1;
+                       }
+                       break;
+
+               case 0:
+                       /* Enable/disable vm2vm comms. */
+                       if (!strncmp(long_option[option_index].name, "vm2vm",
+                               MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, (VM2VM_LAST - 1));
+                               if (ret == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG,
+                                               "Invalid argument for "
+                                               "vm2vm [0|1|2]\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else {
+                                       vm2vm_mode = (vm2vm_type)ret;
+                               }
+                       }
+
+                       /* Enable/disable retries on RX. */
+                       if (!strncmp(long_option[option_index].name, "rx-retry", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, 1);
+                               if (ret == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for rx-retry [0|1]\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else {
+                                       enable_retry = ret;
+                               }
+                       }
+
+                       /* Specify the retries delay time (in useconds) on RX. */
+                       if (!strncmp(long_option[option_index].name, "rx-retry-delay", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, INT32_MAX);
+                               if (ret == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for rx-retry-delay [0-N]\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else {
+                                       burst_rx_delay_time = ret;
+                               }
+                       }
+
+                       /* Specify the retries number on RX. */
+                       if (!strncmp(long_option[option_index].name, "rx-retry-num", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, INT32_MAX);
+                               if (ret == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for rx-retry-num [0-N]\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else {
+                                       burst_rx_retry_num = ret;
+                               }
+                       }
+
+                       /* Enable/disable RX mergeable buffers. */
+                       if (!strncmp(long_option[option_index].name, "mergeable", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, 1);
+                               if (ret == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for mergeable [0|1]\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else {
+                                       if (ret) {
+                                               vmdq_conf_default.rxmode.jumbo_frame = 1;
+                                               vmdq_conf_default.rxmode.max_rx_pkt_len
+                                                       = JUMBO_FRAME_MAX_SIZE;
+                                               VHOST_FEATURES = (1ULL << VIRTIO_NET_F_MRG_RXBUF);
+                                       }
+                               }
+                       }
+
+                       /* Enable/disable stats. */
+                       if (!strncmp(long_option[option_index].name, "stats", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, INT32_MAX);
+                               if (ret == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for stats [0..N]\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else {
+                                       enable_stats = ret;
+                               }
+                       }
+
+                       /* Set character device basename. */
+                       if (!strncmp(long_option[option_index].name, "dev-basename", MAX_LONG_OPT_SZ)) {
+                               if (us_vhost_parse_basename(optarg) == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for character device basename (Max %d characters)\n", MAX_BASENAME_SZ);
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               }
+                       }
+
+                       /* Set character device index. */
+                       if (!strncmp(long_option[option_index].name, "dev-index", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, INT32_MAX);
+                               if (ret == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG, "Invalid argument for character device index [0..N]\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else
+                                       dev_index = ret;
+                       }
+
+                       /* Enable/disable rx/tx zero copy. */
+                       if (!strncmp(long_option[option_index].name,
+                               "zero-copy", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, 1);
+                               if (ret == -1) {
+                                       RTE_LOG(INFO, VHOST_CONFIG,
+                                               "Invalid argument"
+                                               " for zero-copy [0|1]\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else
+                                       zero_copy = ret;
+
+                               if (zero_copy) {
+#ifdef RTE_MBUF_REFCNT
+                                       RTE_LOG(ERR, VHOST_CONFIG, "Before running "
+                                       "zero copy vhost APP, please "
+                                       "disable RTE_MBUF_REFCNT\n"
+                                       "in config file and then rebuild DPDK "
+                                       "core lib!\n"
+                                       "Otherwise please disable zero copy "
+                                       "flag in command line!\n");
+                                       return -1;
+#endif
+                               }
+                       }
+
+                       /* Specify the descriptor number on RX. */
+                       if (!strncmp(long_option[option_index].name,
+                               "rx-desc-num", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, MAX_RING_DESC);
+                               if ((ret == -1) || (!POWEROF2(ret))) {
+                                       RTE_LOG(INFO, VHOST_CONFIG,
+                                       "Invalid argument for rx-desc-num[0-N],"
+                                       "power of 2 required.\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else {
+                                       num_rx_descriptor = ret;
+                               }
+                       }
+
+                       /* Specify the descriptor number on TX. */
+                       if (!strncmp(long_option[option_index].name,
+                               "tx-desc-num", MAX_LONG_OPT_SZ)) {
+                               ret = parse_num_opt(optarg, MAX_RING_DESC);
+                               if ((ret == -1) || (!POWEROF2(ret))) {
+                                       RTE_LOG(INFO, VHOST_CONFIG,
+                                       "Invalid argument for tx-desc-num [0-N],"
+                                       "power of 2 required.\n");
+                                       us_vhost_usage(prgname);
+                                       return -1;
+                               } else {
+                                       num_tx_descriptor = ret;
+                               }
+                       }
+
+                       break;
+
+                       /* Invalid option - print options. */
+               default:
+                       us_vhost_usage(prgname);
+                       return -1;
+               }
+       }
+
+       for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
+               if (enabled_port_mask & (1 << i))
+                       ports[num_ports++] = (uint8_t)i;
+       }
+
+       if ((num_ports ==  0) || (num_ports > MAX_SUP_PORTS)) {
+               RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
+                       "but only %u port can be enabled\n",num_ports, MAX_SUP_PORTS);
+               return -1;
+       }
+
+       if ((zero_copy == 1) && (vm2vm_mode == VM2VM_SOFTWARE)) {
+               RTE_LOG(INFO, VHOST_PORT,
+                       "Vhost zero copy doesn't support software vm2vm,"
+                       "please specify 'vm2vm 2' to use hardware vm2vm.\n");
+               return -1;
+       }
+
+       if ((zero_copy == 1) && (vmdq_conf_default.rxmode.jumbo_frame == 1)) {
+               RTE_LOG(INFO, VHOST_PORT,
+                       "Vhost zero copy doesn't support jumbo frame,"
+                       "please specify '--mergeable 0' to disable the "
+                       "mergeable feature.\n");
+               return -1;
+       }
+
+       return 0;
+}
+
+/*
+ * Update the global var NUM_PORTS and array PORTS according to system ports number
+ * and return valid ports number
+ */
+static unsigned check_ports_num(unsigned nb_ports)
+{
+       unsigned valid_num_ports = num_ports;
+       unsigned portid;
+
+       if (num_ports > nb_ports) {
+               RTE_LOG(INFO, VHOST_PORT, "\nSpecified port number(%u) exceeds total system port number(%u)\n",
+                       num_ports, nb_ports);
+               num_ports = nb_ports;
+       }
+
+       for (portid = 0; portid < num_ports; portid ++) {
+               if (ports[portid] >= nb_ports) {
+                       RTE_LOG(INFO, VHOST_PORT, "\nSpecified port ID(%u) exceeds max system port ID(%u)\n",
+                               ports[portid], (nb_ports - 1));
+                       ports[portid] = INVALID_PORT_ID;
+                       valid_num_ports--;
+               }
+       }
+       return valid_num_ports;
+}
+
+/*
+ * Macro to print out packet contents. Wrapped in debug define so that the
+ * data path is not effected when debug is disabled.
+ */
+#ifdef DEBUG
+#define PRINT_PACKET(device, addr, size, header) do {                                                                                                                          \
+       char *pkt_addr = (char*)(addr);                                                                                                                                                                 \
+       unsigned int index;                                                                                                                                                                                             \
+       char packet[MAX_PRINT_BUFF];                                                                                                                                                                    \
+                                                                                                                                                                                                                                       \
+       if ((header))                                                                                                                                                                                                   \
+               snprintf(packet, MAX_PRINT_BUFF, "(%"PRIu64") Header size %d: ", (device->device_fh), (size));                          \
+       else                                                                                                                                                                                                                    \
+               snprintf(packet, MAX_PRINT_BUFF, "(%"PRIu64") Packet size %d: ", (device->device_fh), (size));                          \
+       for (index = 0; index < (size); index++) {                                                                                                                                              \
+               snprintf(packet + strnlen(packet, MAX_PRINT_BUFF), MAX_PRINT_BUFF - strnlen(packet, MAX_PRINT_BUFF),    \
+                       "%02hhx ", pkt_addr[index]);                                                                                                                                                    \
+       }                                                                                                                                                                                                                               \
+       snprintf(packet + strnlen(packet, MAX_PRINT_BUFF), MAX_PRINT_BUFF - strnlen(packet, MAX_PRINT_BUFF), "\n");     \
+                                                                                                                                                                                                                                       \
+       LOG_DEBUG(VHOST_DATA, "%s", packet);                                                                                                                                                                    \
+} while(0)
+#else
+#define PRINT_PACKET(device, addr, size, header) do{} while(0)
+#endif
+
+/*
+ * Function to convert guest physical addresses to vhost virtual addresses. This
+ * is used to convert virtio buffer addresses.
+ */
+static inline uint64_t __attribute__((always_inline))
+gpa_to_vva(struct virtio_net *dev, uint64_t guest_pa)
+{
+       struct virtio_memory_regions *region;
+       uint32_t regionidx;
+       uint64_t vhost_va = 0;
+
+       for (regionidx = 0; regionidx < dev->mem->nregions; regionidx++) {
+               region = &dev->mem->regions[regionidx];
+               if ((guest_pa >= region->guest_phys_address) &&
+                       (guest_pa <= region->guest_phys_address_end)) {
+                       vhost_va = region->address_offset + guest_pa;
+                       break;
+               }
+       }
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") GPA %p| VVA %p\n",
+               dev->device_fh, (void*)(uintptr_t)guest_pa, (void*)(uintptr_t)vhost_va);
+
+       return vhost_va;
+}
+
+/*
+ * Function to convert guest physical addresses to vhost physical addresses.
+ * This is used to convert virtio buffer addresses.
+ */
+static inline uint64_t __attribute__((always_inline))
+gpa_to_hpa(struct virtio_net *dev, uint64_t guest_pa,
+       uint32_t buf_len, hpa_type *addr_type)
+{
+       struct virtio_memory_regions_hpa *region;
+       uint32_t regionidx;
+       uint64_t vhost_pa = 0;
+
+       *addr_type = PHYS_ADDR_INVALID;
+
+       for (regionidx = 0; regionidx < dev->mem->nregions_hpa; regionidx++) {
+               region = &dev->mem->regions_hpa[regionidx];
+               if ((guest_pa >= region->guest_phys_address) &&
+                       (guest_pa <= region->guest_phys_address_end)) {
+                       vhost_pa = region->host_phys_addr_offset + guest_pa;
+                       if (likely((guest_pa + buf_len - 1)
+                               <= region->guest_phys_address_end))
+                               *addr_type = PHYS_ADDR_CONTINUOUS;
+                       else
+                               *addr_type = PHYS_ADDR_CROSS_SUBREG;
+                       break;
+               }
+       }
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") GPA %p| HPA %p\n",
+               dev->device_fh, (void *)(uintptr_t)guest_pa,
+               (void *)(uintptr_t)vhost_pa);
+
+       return vhost_pa;
+}
+
+/*
+ * This function adds buffers to the virtio devices RX virtqueue. Buffers can
+ * be received from the physical port or from another virtio device. A packet
+ * count is returned to indicate the number of packets that were succesfully
+ * added to the RX queue. This function works when mergeable is disabled.
+ */
+static inline uint32_t __attribute__((always_inline))
+virtio_dev_rx(struct virtio_net *dev, struct rte_mbuf **pkts, uint32_t count)
+{
+       struct vhost_virtqueue *vq;
+       struct vring_desc *desc;
+       struct rte_mbuf *buff;
+       /* The virtio_hdr is initialised to 0. */
+       struct virtio_net_hdr_mrg_rxbuf virtio_hdr = {{0,0,0,0,0,0},0};
+       uint64_t buff_addr = 0;
+       uint64_t buff_hdr_addr = 0;
+       uint32_t head[MAX_PKT_BURST], packet_len = 0;
+       uint32_t head_idx, packet_success = 0;
+       uint32_t retry = 0;
+       uint16_t avail_idx, res_cur_idx;
+       uint16_t res_base_idx, res_end_idx;
+       uint16_t free_entries;
+       uint8_t success = 0;
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") virtio_dev_rx()\n", dev->device_fh);
+       vq = dev->virtqueue[VIRTIO_RXQ];
+       count = (count > MAX_PKT_BURST) ? MAX_PKT_BURST : count;
+
+       /* As many data cores may want access to available buffers, they need to be reserved. */
+       do {
+               res_base_idx = vq->last_used_idx_res;
+               avail_idx = *((volatile uint16_t *)&vq->avail->idx);
+
+               free_entries = (avail_idx - res_base_idx);
+               /* If retry is enabled and the queue is full then we wait and retry to avoid packet loss. */
+               if (enable_retry && unlikely(count > free_entries)) {
+                       for (retry = 0; retry < burst_rx_retry_num; retry++) {
+                               rte_delay_us(burst_rx_delay_time);
+                               avail_idx =
+                                       *((volatile uint16_t *)&vq->avail->idx);
+                               free_entries = (avail_idx - res_base_idx);
+                               if (count <= free_entries)
+                                       break;
+                       }
+               }
+
+               /*check that we have enough buffers*/
+               if (unlikely(count > free_entries))
+                       count = free_entries;
+
+               if (count == 0)
+                       return 0;
+
+               res_end_idx = res_base_idx + count;
+               /* vq->last_used_idx_res is atomically updated. */
+               success = rte_atomic16_cmpset(&vq->last_used_idx_res, res_base_idx,
+                                                                       res_end_idx);
+       } while (unlikely(success == 0));
+       res_cur_idx = res_base_idx;
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") Current Index %d| End Index %d\n", dev->device_fh, res_cur_idx, res_end_idx);
+
+       /* Prefetch available ring to retrieve indexes. */
+       rte_prefetch0(&vq->avail->ring[res_cur_idx & (vq->size - 1)]);
+
+       /* Retrieve all of the head indexes first to avoid caching issues. */
+       for (head_idx = 0; head_idx < count; head_idx++)
+               head[head_idx] = vq->avail->ring[(res_cur_idx + head_idx) & (vq->size - 1)];
+
+       /*Prefetch descriptor index. */
+       rte_prefetch0(&vq->desc[head[packet_success]]);
+
+       while (res_cur_idx != res_end_idx) {
+               /* Get descriptor from available ring */
+               desc = &vq->desc[head[packet_success]];
+
+               buff = pkts[packet_success];
+
+               /* Convert from gpa to vva (guest physical addr -> vhost virtual addr) */
+               buff_addr = gpa_to_vva(dev, desc->addr);
+               /* Prefetch buffer address. */
+               rte_prefetch0((void*)(uintptr_t)buff_addr);
+
+               /* Copy virtio_hdr to packet and increment buffer address */
+               buff_hdr_addr = buff_addr;
+               packet_len = rte_pktmbuf_data_len(buff) + vq->vhost_hlen;
+
+               /*
+                * If the descriptors are chained the header and data are
+                * placed in separate buffers.
+                */
+               if (desc->flags & VRING_DESC_F_NEXT) {
+                       desc->len = vq->vhost_hlen;
+                       desc = &vq->desc[desc->next];
+                       /* Buffer address translation. */
+                       buff_addr = gpa_to_vva(dev, desc->addr);
+                       desc->len = rte_pktmbuf_data_len(buff);
+               } else {
+                       buff_addr += vq->vhost_hlen;
+                       desc->len = packet_len;
+               }
+
+               /* Update used ring with desc information */
+               vq->used->ring[res_cur_idx & (vq->size - 1)].id = head[packet_success];
+               vq->used->ring[res_cur_idx & (vq->size - 1)].len = packet_len;
+
+               /* Copy mbuf data to buffer */
+               rte_memcpy((void *)(uintptr_t)buff_addr,
+                       rte_pktmbuf_mtod(buff, const void *),
+                       rte_pktmbuf_data_len(buff));
+               PRINT_PACKET(dev, (uintptr_t)buff_addr,
+                       rte_pktmbuf_data_len(buff), 0);
+
+               res_cur_idx++;
+               packet_success++;
+
+               rte_memcpy((void *)(uintptr_t)buff_hdr_addr,
+                       (const void *)&virtio_hdr, vq->vhost_hlen);
+
+               PRINT_PACKET(dev, (uintptr_t)buff_hdr_addr, vq->vhost_hlen, 1);
+
+               if (res_cur_idx < res_end_idx) {
+                       /* Prefetch descriptor index. */
+                       rte_prefetch0(&vq->desc[head[packet_success]]);
+               }
+       }
+
+       rte_compiler_barrier();
+
+       /* Wait until it's our turn to add our buffer to the used ring. */
+       while (unlikely(vq->last_used_idx != res_base_idx))
+               rte_pause();
+
+       *(volatile uint16_t *)&vq->used->idx += count;
+       vq->last_used_idx = res_end_idx;
+
+       /* Kick the guest if necessary. */
+       if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT))
+               eventfd_write((int)vq->kickfd, 1);
+       return count;
+}
+
+static inline uint32_t __attribute__((always_inline))
+copy_from_mbuf_to_vring(struct virtio_net *dev,
+       uint16_t res_base_idx, uint16_t res_end_idx,
+       struct rte_mbuf *pkt)
+{
+       uint32_t vec_idx = 0;
+       uint32_t entry_success = 0;
+       struct vhost_virtqueue *vq;
+       /* The virtio_hdr is initialised to 0. */
+       struct virtio_net_hdr_mrg_rxbuf virtio_hdr = {
+               {0, 0, 0, 0, 0, 0}, 0};
+       uint16_t cur_idx = res_base_idx;
+       uint64_t vb_addr = 0;
+       uint64_t vb_hdr_addr = 0;
+       uint32_t seg_offset = 0;
+       uint32_t vb_offset = 0;
+       uint32_t seg_avail;
+       uint32_t vb_avail;
+       uint32_t cpy_len, entry_len;
+
+       if (pkt == NULL)
+               return 0;
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") Current Index %d| "
+               "End Index %d\n",
+               dev->device_fh, cur_idx, res_end_idx);
+
+       /*
+        * Convert from gpa to vva
+        * (guest physical addr -> vhost virtual addr)
+        */
+       vq = dev->virtqueue[VIRTIO_RXQ];
+       vb_addr =
+               gpa_to_vva(dev, vq->buf_vec[vec_idx].buf_addr);
+       vb_hdr_addr = vb_addr;
+
+       /* Prefetch buffer address. */
+       rte_prefetch0((void *)(uintptr_t)vb_addr);
+
+       virtio_hdr.num_buffers = res_end_idx - res_base_idx;
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") RX: Num merge buffers %d\n",
+               dev->device_fh, virtio_hdr.num_buffers);
+
+       rte_memcpy((void *)(uintptr_t)vb_hdr_addr,
+               (const void *)&virtio_hdr, vq->vhost_hlen);
+
+       PRINT_PACKET(dev, (uintptr_t)vb_hdr_addr, vq->vhost_hlen, 1);
+
+       seg_avail = rte_pktmbuf_data_len(pkt);
+       vb_offset = vq->vhost_hlen;
+       vb_avail =
+               vq->buf_vec[vec_idx].buf_len - vq->vhost_hlen;
+
+       entry_len = vq->vhost_hlen;
+
+       if (vb_avail == 0) {
+               uint32_t desc_idx =
+                       vq->buf_vec[vec_idx].desc_idx;
+               vq->desc[desc_idx].len = vq->vhost_hlen;
+
+               if ((vq->desc[desc_idx].flags
+                       & VRING_DESC_F_NEXT) == 0) {
+                       /* Update used ring with desc information */
+                       vq->used->ring[cur_idx & (vq->size - 1)].id
+                               = vq->buf_vec[vec_idx].desc_idx;
+                       vq->used->ring[cur_idx & (vq->size - 1)].len
+                               = entry_len;
+
+                       entry_len = 0;
+                       cur_idx++;
+                       entry_success++;
+               }
+
+               vec_idx++;
+               vb_addr =
+                       gpa_to_vva(dev, vq->buf_vec[vec_idx].buf_addr);
+
+               /* Prefetch buffer address. */
+               rte_prefetch0((void *)(uintptr_t)vb_addr);
+               vb_offset = 0;
+               vb_avail = vq->buf_vec[vec_idx].buf_len;
+       }
+
+       cpy_len = RTE_MIN(vb_avail, seg_avail);
+
+       while (cpy_len > 0) {
+               /* Copy mbuf data to vring buffer */
+               rte_memcpy((void *)(uintptr_t)(vb_addr + vb_offset),
+                       (const void *)(rte_pktmbuf_mtod(pkt, char*) + seg_offset),
+                       cpy_len);
+
+               PRINT_PACKET(dev,
+                       (uintptr_t)(vb_addr + vb_offset),
+                       cpy_len, 0);
+
+               seg_offset += cpy_len;
+               vb_offset += cpy_len;
+               seg_avail -= cpy_len;
+               vb_avail -= cpy_len;
+               entry_len += cpy_len;
+
+               if (seg_avail != 0) {
+                       /*
+                        * The virtio buffer in this vring
+                        * entry reach to its end.
+                        * But the segment doesn't complete.
+                        */
+                       if ((vq->desc[vq->buf_vec[vec_idx].desc_idx].flags &
+                               VRING_DESC_F_NEXT) == 0) {
+                               /* Update used ring with desc information */
+                               vq->used->ring[cur_idx & (vq->size - 1)].id
+                                       = vq->buf_vec[vec_idx].desc_idx;
+                               vq->used->ring[cur_idx & (vq->size - 1)].len
+                                       = entry_len;
+                               entry_len = 0;
+                               cur_idx++;
+                               entry_success++;
+                       }
+
+                       vec_idx++;
+                       vb_addr = gpa_to_vva(dev,
+                               vq->buf_vec[vec_idx].buf_addr);
+                       vb_offset = 0;
+                       vb_avail = vq->buf_vec[vec_idx].buf_len;
+                       cpy_len = RTE_MIN(vb_avail, seg_avail);
+               } else {
+                       /*
+                        * This current segment complete, need continue to
+                        * check if the whole packet complete or not.
+                        */
+                       pkt = pkt->next;
+                       if (pkt != NULL) {
+                               /*
+                                * There are more segments.
+                                */
+                               if (vb_avail == 0) {
+                                       /*
+                                        * This current buffer from vring is
+                                        * used up, need fetch next buffer
+                                        * from buf_vec.
+                                        */
+                                       uint32_t desc_idx =
+                                               vq->buf_vec[vec_idx].desc_idx;
+                                       vq->desc[desc_idx].len = vb_offset;
+
+                                       if ((vq->desc[desc_idx].flags &
+                                               VRING_DESC_F_NEXT) == 0) {
+                                               uint16_t wrapped_idx =
+                                                       cur_idx & (vq->size - 1);
+                                               /*
+                                                * Update used ring with the
+                                                * descriptor information
+                                                */
+                                               vq->used->ring[wrapped_idx].id
+                                                       = desc_idx;
+                                               vq->used->ring[wrapped_idx].len
+                                                       = entry_len;
+                                               entry_success++;
+                                               entry_len = 0;
+                                               cur_idx++;
+                                       }
+
+                                       /* Get next buffer from buf_vec. */
+                                       vec_idx++;
+                                       vb_addr = gpa_to_vva(dev,
+                                               vq->buf_vec[vec_idx].buf_addr);
+                                       vb_avail =
+                                               vq->buf_vec[vec_idx].buf_len;
+                                       vb_offset = 0;
+                               }
+
+                               seg_offset = 0;
+                               seg_avail = rte_pktmbuf_data_len(pkt);
+                               cpy_len = RTE_MIN(vb_avail, seg_avail);
+                       } else {
+                               /*
+                                * This whole packet completes.
+                                */
+                               uint32_t desc_idx =
+                                       vq->buf_vec[vec_idx].desc_idx;
+                               vq->desc[desc_idx].len = vb_offset;
+
+                               while (vq->desc[desc_idx].flags &
+                                       VRING_DESC_F_NEXT) {
+                                       desc_idx = vq->desc[desc_idx].next;
+                                        vq->desc[desc_idx].len = 0;
+                               }
+
+                               /* Update used ring with desc information */
+                               vq->used->ring[cur_idx & (vq->size - 1)].id
+                                       = vq->buf_vec[vec_idx].desc_idx;
+                               vq->used->ring[cur_idx & (vq->size - 1)].len
+                                       = entry_len;
+                               entry_len = 0;
+                               cur_idx++;
+                               entry_success++;
+                               seg_avail = 0;
+                               cpy_len = RTE_MIN(vb_avail, seg_avail);
+                       }
+               }
+       }
+
+       return entry_success;
+}
+
+/*
+ * This function adds buffers to the virtio devices RX virtqueue. Buffers can
+ * be received from the physical port or from another virtio device. A packet
+ * count is returned to indicate the number of packets that were succesfully
+ * added to the RX queue. This function works for mergeable RX.
+ */
+static inline uint32_t __attribute__((always_inline))
+virtio_dev_merge_rx(struct virtio_net *dev, struct rte_mbuf **pkts,
+       uint32_t count)
+{
+       struct vhost_virtqueue *vq;
+       uint32_t pkt_idx = 0, entry_success = 0;
+       uint32_t retry = 0;
+       uint16_t avail_idx, res_cur_idx;
+       uint16_t res_base_idx, res_end_idx;
+       uint8_t success = 0;
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") virtio_dev_merge_rx()\n",
+               dev->device_fh);
+       vq = dev->virtqueue[VIRTIO_RXQ];
+       count = RTE_MIN((uint32_t)MAX_PKT_BURST, count);
+
+       if (count == 0)
+               return 0;
+
+       for (pkt_idx = 0; pkt_idx < count; pkt_idx++) {
+               uint32_t secure_len = 0;
+               uint16_t need_cnt;
+               uint32_t vec_idx = 0;
+               uint32_t pkt_len = pkts[pkt_idx]->pkt_len + vq->vhost_hlen;
+               uint16_t i, id;
+
+               do {
+                       /*
+                        * As many data cores may want access to available
+                        * buffers, they need to be reserved.
+                        */
+                       res_base_idx = vq->last_used_idx_res;
+                       res_cur_idx = res_base_idx;
+
+                       do {
+                               avail_idx = *((volatile uint16_t *)&vq->avail->idx);
+                               if (unlikely(res_cur_idx == avail_idx)) {
+                                       /*
+                                        * If retry is enabled and the queue is
+                                        * full then we wait and retry to avoid
+                                        * packet loss.
+                                        */
+                                       if (enable_retry) {
+                                               uint8_t cont = 0;
+                                               for (retry = 0; retry < burst_rx_retry_num; retry++) {
+                                                       rte_delay_us(burst_rx_delay_time);
+                                                       avail_idx =
+                                                               *((volatile uint16_t *)&vq->avail->idx);
+                                                       if (likely(res_cur_idx != avail_idx)) {
+                                                               cont = 1;
+                                                               break;
+                                                       }
+                                               }
+                                               if (cont == 1)
+                                                       continue;
+                                       }
+
+                                       LOG_DEBUG(VHOST_DATA,
+                                               "(%"PRIu64") Failed "
+                                               "to get enough desc from "
+                                               "vring\n",
+                                               dev->device_fh);
+                                       return pkt_idx;
+                               } else {
+                                       uint16_t wrapped_idx =
+                                               (res_cur_idx) & (vq->size - 1);
+                                       uint32_t idx =
+                                               vq->avail->ring[wrapped_idx];
+                                       uint8_t next_desc;
+
+                                       do {
+                                               next_desc = 0;
+                                               secure_len += vq->desc[idx].len;
+                                               if (vq->desc[idx].flags &
+                                                       VRING_DESC_F_NEXT) {
+                                                       idx = vq->desc[idx].next;
+                                                       next_desc = 1;
+                                               }
+                                       } while (next_desc);
+
+                                       res_cur_idx++;
+                               }
+                       } while (pkt_len > secure_len);
+
+                       /* vq->last_used_idx_res is atomically updated. */
+                       success = rte_atomic16_cmpset(&vq->last_used_idx_res,
+                                                       res_base_idx,
+                                                       res_cur_idx);
+               } while (success == 0);
+
+               id = res_base_idx;
+               need_cnt = res_cur_idx - res_base_idx;
+
+               for (i = 0; i < need_cnt; i++, id++) {
+                       uint16_t wrapped_idx = id & (vq->size - 1);
+                       uint32_t idx = vq->avail->ring[wrapped_idx];
+                       uint8_t next_desc;
+                       do {
+                               next_desc = 0;
+                               vq->buf_vec[vec_idx].buf_addr =
+                                       vq->desc[idx].addr;
+                               vq->buf_vec[vec_idx].buf_len =
+                                       vq->desc[idx].len;
+                               vq->buf_vec[vec_idx].desc_idx = idx;
+                               vec_idx++;
+
+                               if (vq->desc[idx].flags & VRING_DESC_F_NEXT) {
+                                       idx = vq->desc[idx].next;
+                                       next_desc = 1;
+                               }
+                       } while (next_desc);
+               }
+
+               res_end_idx = res_cur_idx;
+
+               entry_success = copy_from_mbuf_to_vring(dev, res_base_idx,
+                       res_end_idx, pkts[pkt_idx]);
+
+               rte_compiler_barrier();
+
+               /*
+                * Wait until it's our turn to add our buffer
+                * to the used ring.
+                */
+               while (unlikely(vq->last_used_idx != res_base_idx))
+                       rte_pause();
+
+               *(volatile uint16_t *)&vq->used->idx += entry_success;
+               vq->last_used_idx = res_end_idx;
+
+               /* Kick the guest if necessary. */
+               if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT))
+                       eventfd_write((int)vq->kickfd, 1);
+       }
+
+       return count;
+}
+
+/*
+ * Compares a packet destination MAC address to a device MAC address.
+ */
+static inline int __attribute__((always_inline))
+ether_addr_cmp(struct ether_addr *ea, struct ether_addr *eb)
+{
+       return (((*(uint64_t *)ea ^ *(uint64_t *)eb) & MAC_ADDR_CMP) == 0);
+}
+
+/*
+ * This function learns the MAC address of the device and registers this along with a
+ * vlan tag to a VMDQ.
+ */
+static int
+link_vmdq(struct virtio_net *dev, struct rte_mbuf *m)
+{
+       struct ether_hdr *pkt_hdr;
+       struct virtio_net_data_ll *dev_ll;
+       int i, ret;
+
+       /* Learn MAC address of guest device from packet */
+       pkt_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
+
+       dev_ll = ll_root_used;
+
+       while (dev_ll != NULL) {
+               if (ether_addr_cmp(&(pkt_hdr->s_addr), &dev_ll->dev->mac_address)) {
+                       RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") WARNING: This device is using an existing MAC address and has not been registered.\n", dev->device_fh);
+                       return -1;
+               }
+               dev_ll = dev_ll->next;
+       }
+
+       for (i = 0; i < ETHER_ADDR_LEN; i++)
+               dev->mac_address.addr_bytes[i] = pkt_hdr->s_addr.addr_bytes[i];
+
+       /* vlan_tag currently uses the device_id. */
+       dev->vlan_tag = vlan_tags[dev->device_fh];
+
+       /* Print out VMDQ registration info. */
+       RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") MAC_ADDRESS %02x:%02x:%02x:%02x:%02x:%02x and VLAN_TAG %d registered\n",
+               dev->device_fh,
+               dev->mac_address.addr_bytes[0], dev->mac_address.addr_bytes[1],
+               dev->mac_address.addr_bytes[2], dev->mac_address.addr_bytes[3],
+               dev->mac_address.addr_bytes[4], dev->mac_address.addr_bytes[5],
+               dev->vlan_tag);
+
+       /* Register the MAC address. */
+       ret = rte_eth_dev_mac_addr_add(ports[0], &dev->mac_address, (uint32_t)dev->device_fh);
+       if (ret)
+               RTE_LOG(ERR, VHOST_DATA, "(%"PRIu64") Failed to add device MAC address to VMDQ\n",
+                                       dev->device_fh);
+
+       /* Enable stripping of the vlan tag as we handle routing. */
+       rte_eth_dev_set_vlan_strip_on_queue(ports[0], (uint16_t)dev->vmdq_rx_q, 1);
+
+       /* Set device as ready for RX. */
+       dev->ready = DEVICE_RX;
+
+       return 0;
+}
+
+/*
+ * Removes MAC address and vlan tag from VMDQ. Ensures that nothing is adding buffers to the RX
+ * queue before disabling RX on the device.
+ */
+static inline void
+unlink_vmdq(struct virtio_net *dev)
+{
+       unsigned i = 0;
+       unsigned rx_count;
+       struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
+
+       if (dev->ready == DEVICE_RX) {
+               /*clear MAC and VLAN settings*/
+               rte_eth_dev_mac_addr_remove(ports[0], &dev->mac_address);
+               for (i = 0; i < 6; i++)
+                       dev->mac_address.addr_bytes[i] = 0;
+
+               dev->vlan_tag = 0;
+
+               /*Clear out the receive buffers*/
+               rx_count = rte_eth_rx_burst(ports[0],
+                                       (uint16_t)dev->vmdq_rx_q, pkts_burst, MAX_PKT_BURST);
+
+               while (rx_count) {
+                       for (i = 0; i < rx_count; i++)
+                               rte_pktmbuf_free(pkts_burst[i]);
+
+                       rx_count = rte_eth_rx_burst(ports[0],
+                                       (uint16_t)dev->vmdq_rx_q, pkts_burst, MAX_PKT_BURST);
+               }
+
+               dev->ready = DEVICE_MAC_LEARNING;
+       }
+}
+
+/*
+ * Check if the packet destination MAC address is for a local device. If so then put
+ * the packet on that devices RX queue. If not then return.
+ */
+static inline unsigned __attribute__((always_inline))
+virtio_tx_local(struct virtio_net *dev, struct rte_mbuf *m)
+{
+       struct virtio_net_data_ll *dev_ll;
+       struct ether_hdr *pkt_hdr;
+       uint64_t ret = 0;
+
+       pkt_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
+
+       /*get the used devices list*/
+       dev_ll = ll_root_used;
+
+       while (dev_ll != NULL) {
+               if ((dev_ll->dev->ready == DEVICE_RX) && ether_addr_cmp(&(pkt_hdr->d_addr),
+                                         &dev_ll->dev->mac_address)) {
+
+                       /* Drop the packet if the TX packet is destined for the TX device. */
+                       if (dev_ll->dev->device_fh == dev->device_fh) {
+                               LOG_DEBUG(VHOST_DATA, "(%"PRIu64") TX: Source and destination MAC addresses are the same. Dropping packet.\n",
+                                                       dev_ll->dev->device_fh);
+                               return 0;
+                       }
+
+
+                       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") TX: MAC address is local\n", dev_ll->dev->device_fh);
+
+                       if (dev_ll->dev->remove) {
+                               /*drop the packet if the device is marked for removal*/
+                               LOG_DEBUG(VHOST_DATA, "(%"PRIu64") Device is marked for removal\n", dev_ll->dev->device_fh);
+                       } else {
+                               uint32_t mergeable =
+                                       dev_ll->dev->features &
+                                       (1 << VIRTIO_NET_F_MRG_RXBUF);
+
+                               /*send the packet to the local virtio device*/
+                               if (likely(mergeable == 0))
+                                       ret = virtio_dev_rx(dev_ll->dev, &m, 1);
+                               else
+                                       ret = virtio_dev_merge_rx(dev_ll->dev,
+                                               &m, 1);
+
+                               if (enable_stats) {
+                                       rte_atomic64_add(
+                                       &dev_statistics[dev_ll->dev->device_fh].rx_total_atomic,
+                                       1);
+                                       rte_atomic64_add(
+                                       &dev_statistics[dev_ll->dev->device_fh].rx_atomic,
+                                       ret);
+                                       dev_statistics[dev->device_fh].tx_total++;
+                                       dev_statistics[dev->device_fh].tx += ret;
+                               }
+                       }
+
+                       return 0;
+               }
+               dev_ll = dev_ll->next;
+       }
+
+       return -1;
+}
+
+/*
+ * This function routes the TX packet to the correct interface. This may be a local device
+ * or the physical port.
+ */
+static inline void __attribute__((always_inline))
+virtio_tx_route(struct virtio_net* dev, struct rte_mbuf *m, struct rte_mempool *mbuf_pool, uint16_t vlan_tag)
+{
+       struct mbuf_table *tx_q;
+       struct vlan_ethhdr *vlan_hdr;
+       struct rte_mbuf **m_table;
+       struct rte_mbuf *mbuf, *prev;
+       unsigned len, ret, offset = 0;
+       const uint16_t lcore_id = rte_lcore_id();
+       struct virtio_net_data_ll *dev_ll = ll_root_used;
+       struct ether_hdr *pkt_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
+
+       /*check if destination is local VM*/
+       if ((vm2vm_mode == VM2VM_SOFTWARE) && (virtio_tx_local(dev, m) == 0))
+               return;
+
+       if (vm2vm_mode == VM2VM_HARDWARE) {
+               while (dev_ll != NULL) {
+                       if ((dev_ll->dev->ready == DEVICE_RX)
+                               && ether_addr_cmp(&(pkt_hdr->d_addr),
+                               &dev_ll->dev->mac_address)) {
+                               /*
+                                * Drop the packet if the TX packet is
+                                * destined for the TX device.
+                                */
+                               if (dev_ll->dev->device_fh == dev->device_fh) {
+                                       LOG_DEBUG(VHOST_DATA,
+                                       "(%"PRIu64") TX: Source and destination"
+                                       " MAC addresses are the same. Dropping "
+                                       "packet.\n",
+                                       dev_ll->dev->device_fh);
+                                       return;
+                               }
+                               offset = 4;
+                               vlan_tag =
+                               (uint16_t)
+                               vlan_tags[(uint16_t)dev_ll->dev->device_fh];
+
+                               LOG_DEBUG(VHOST_DATA,
+                               "(%"PRIu64") TX: pkt to local VM device id:"
+                               "(%"PRIu64") vlan tag: %d.\n",
+                               dev->device_fh, dev_ll->dev->device_fh,
+                               vlan_tag);
+
+                               break;
+                       }
+                       dev_ll = dev_ll->next;
+               }
+       }
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") TX: MAC address is external\n", dev->device_fh);
+
+       /*Add packet to the port tx queue*/
+       tx_q = &lcore_tx_queue[lcore_id];
+       len = tx_q->len;
+
+       /* Allocate an mbuf and populate the structure. */
+       mbuf = rte_pktmbuf_alloc(mbuf_pool);
+       if (unlikely(mbuf == NULL)) {
+               RTE_LOG(ERR, VHOST_DATA,
+                       "Failed to allocate memory for mbuf.\n");
+               return;
+       }
+
+       mbuf->data_len = m->data_len + VLAN_HLEN + offset;
+       mbuf->pkt_len = m->pkt_len + VLAN_HLEN + offset;
+       mbuf->nb_segs = m->nb_segs;
+
+       /* Copy ethernet header to mbuf. */
+       rte_memcpy(rte_pktmbuf_mtod(mbuf, void *),
+               rte_pktmbuf_mtod(m, const void *),
+               ETH_HLEN);
+
+
+       /* Setup vlan header. Bytes need to be re-ordered for network with htons()*/
+       vlan_hdr = rte_pktmbuf_mtod(mbuf, struct vlan_ethhdr *);
+       vlan_hdr->h_vlan_encapsulated_proto = vlan_hdr->h_vlan_proto;
+       vlan_hdr->h_vlan_proto = htons(ETH_P_8021Q);
+       vlan_hdr->h_vlan_TCI = htons(vlan_tag);
+
+       /* Copy the remaining packet contents to the mbuf. */
+       rte_memcpy((void *)(rte_pktmbuf_mtod(mbuf, uint8_t *) + VLAN_ETH_HLEN),
+               (const void *)(rte_pktmbuf_mtod(m, uint8_t *) + ETH_HLEN),
+               (m->data_len - ETH_HLEN));
+
+       /* Copy the remaining segments for the whole packet. */
+       prev = mbuf;
+       while (m->next) {
+               /* Allocate an mbuf and populate the structure. */
+               struct rte_mbuf *next_mbuf = rte_pktmbuf_alloc(mbuf_pool);
+               if (unlikely(next_mbuf == NULL)) {
+                       rte_pktmbuf_free(mbuf);
+                       RTE_LOG(ERR, VHOST_DATA,
+                               "Failed to allocate memory for mbuf.\n");
+                       return;
+               }
+
+               m = m->next;
+               prev->next = next_mbuf;
+               prev = next_mbuf;
+               next_mbuf->data_len = m->data_len;
+
+               /* Copy data to next mbuf. */
+               rte_memcpy(rte_pktmbuf_mtod(next_mbuf, void *),
+                       rte_pktmbuf_mtod(m, const void *), m->data_len);
+       }
+
+       tx_q->m_table[len] = mbuf;
+       len++;
+       if (enable_stats) {
+               dev_statistics[dev->device_fh].tx_total++;
+               dev_statistics[dev->device_fh].tx++;
+       }
+
+       if (unlikely(len == MAX_PKT_BURST)) {
+               m_table = (struct rte_mbuf **)tx_q->m_table;
+               ret = rte_eth_tx_burst(ports[0], (uint16_t)tx_q->txq_id, m_table, (uint16_t) len);
+               /* Free any buffers not handled by TX and update the port stats. */
+               if (unlikely(ret < len)) {
+                       do {
+                               rte_pktmbuf_free(m_table[ret]);
+                       } while (++ret < len);
+               }
+
+               len = 0;
+       }
+
+       tx_q->len = len;
+       return;
+}
+
+static inline void __attribute__((always_inline))
+virtio_dev_tx(struct virtio_net* dev, struct rte_mempool *mbuf_pool)
+{
+       struct rte_mbuf m;
+       struct vhost_virtqueue *vq;
+       struct vring_desc *desc;
+       uint64_t buff_addr = 0;
+       uint32_t head[MAX_PKT_BURST];
+       uint32_t used_idx;
+       uint32_t i;
+       uint16_t free_entries, packet_success = 0;
+       uint16_t avail_idx;
+
+       vq = dev->virtqueue[VIRTIO_TXQ];
+       avail_idx =  *((volatile uint16_t *)&vq->avail->idx);
+
+       /* If there are no available buffers then return. */
+       if (vq->last_used_idx == avail_idx)
+               return;
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") virtio_dev_tx()\n", dev->device_fh);
+
+       /* Prefetch available ring to retrieve head indexes. */
+       rte_prefetch0(&vq->avail->ring[vq->last_used_idx & (vq->size - 1)]);
+
+       /*get the number of free entries in the ring*/
+       free_entries = (avail_idx - vq->last_used_idx);
+
+       /* Limit to MAX_PKT_BURST. */
+       if (free_entries > MAX_PKT_BURST)
+               free_entries = MAX_PKT_BURST;
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") Buffers available %d\n", dev->device_fh, free_entries);
+       /* Retrieve all of the head indexes first to avoid caching issues. */
+       for (i = 0; i < free_entries; i++)
+               head[i] = vq->avail->ring[(vq->last_used_idx + i) & (vq->size - 1)];
+
+       /* Prefetch descriptor index. */
+       rte_prefetch0(&vq->desc[head[packet_success]]);
+       rte_prefetch0(&vq->used->ring[vq->last_used_idx & (vq->size - 1)]);
+
+       while (packet_success < free_entries) {
+               desc = &vq->desc[head[packet_success]];
+
+               /* Discard first buffer as it is the virtio header */
+               desc = &vq->desc[desc->next];
+
+               /* Buffer address translation. */
+               buff_addr = gpa_to_vva(dev, desc->addr);
+               /* Prefetch buffer address. */
+               rte_prefetch0((void*)(uintptr_t)buff_addr);
+
+               used_idx = vq->last_used_idx & (vq->size - 1);
+
+               if (packet_success < (free_entries - 1)) {
+                       /* Prefetch descriptor index. */
+                       rte_prefetch0(&vq->desc[head[packet_success+1]]);
+                       rte_prefetch0(&vq->used->ring[(used_idx + 1) & (vq->size - 1)]);
+               }
+
+               /* Update used index buffer information. */
+               vq->used->ring[used_idx].id = head[packet_success];
+               vq->used->ring[used_idx].len = 0;
+
+               /* Setup dummy mbuf. This is copied to a real mbuf if transmitted out the physical port. */
+               m.data_len = desc->len;
+               m.pkt_len = desc->len;
+               m.data_off = 0;
+
+               PRINT_PACKET(dev, (uintptr_t)buff_addr, desc->len, 0);
+
+               /* If this is the first received packet we need to learn the MAC and setup VMDQ */
+               if (dev->ready == DEVICE_MAC_LEARNING) {
+                       if (dev->remove || (link_vmdq(dev, &m) == -1)) {
+                               /*discard frame if device is scheduled for removal or a duplicate MAC address is found. */
+                               packet_success += free_entries;
+                               vq->last_used_idx += packet_success;
+                               break;
+                       }
+               }
+               virtio_tx_route(dev, &m, mbuf_pool, (uint16_t)dev->device_fh);
+
+               vq->last_used_idx++;
+               packet_success++;
+       }
+
+       rte_compiler_barrier();
+       vq->used->idx += packet_success;
+       /* Kick guest if required. */
+       if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT))
+               eventfd_write((int)vq->kickfd, 1);
+}
+
+/* This function works for TX packets with mergeable feature enabled. */
+static inline void __attribute__((always_inline))
+virtio_dev_merge_tx(struct virtio_net *dev, struct rte_mempool *mbuf_pool)
+{
+       struct rte_mbuf *m, *prev;
+       struct vhost_virtqueue *vq;
+       struct vring_desc *desc;
+       uint64_t vb_addr = 0;
+       uint32_t head[MAX_PKT_BURST];
+       uint32_t used_idx;
+       uint32_t i;
+       uint16_t free_entries, entry_success = 0;
+       uint16_t avail_idx;
+       uint32_t buf_size = MBUF_SIZE - (sizeof(struct rte_mbuf)
+                       + RTE_PKTMBUF_HEADROOM);
+
+       vq = dev->virtqueue[VIRTIO_TXQ];
+       avail_idx =  *((volatile uint16_t *)&vq->avail->idx);
+
+       /* If there are no available buffers then return. */
+       if (vq->last_used_idx == avail_idx)
+               return;
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") virtio_dev_merge_tx()\n",
+               dev->device_fh);
+
+       /* Prefetch available ring to retrieve head indexes. */
+       rte_prefetch0(&vq->avail->ring[vq->last_used_idx & (vq->size - 1)]);
+
+       /*get the number of free entries in the ring*/
+       free_entries = (avail_idx - vq->last_used_idx);
+
+       /* Limit to MAX_PKT_BURST. */
+       free_entries = RTE_MIN(free_entries, MAX_PKT_BURST);
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") Buffers available %d\n",
+               dev->device_fh, free_entries);
+       /* Retrieve all of the head indexes first to avoid caching issues. */
+       for (i = 0; i < free_entries; i++)
+               head[i] = vq->avail->ring[(vq->last_used_idx + i) & (vq->size - 1)];
+
+       /* Prefetch descriptor index. */
+       rte_prefetch0(&vq->desc[head[entry_success]]);
+       rte_prefetch0(&vq->used->ring[vq->last_used_idx & (vq->size - 1)]);
+
+       while (entry_success < free_entries) {
+               uint32_t vb_avail, vb_offset;
+               uint32_t seg_avail, seg_offset;
+               uint32_t cpy_len;
+               uint32_t seg_num = 0;
+               struct rte_mbuf *cur;
+               uint8_t alloc_err = 0;
+
+               desc = &vq->desc[head[entry_success]];
+
+               /* Discard first buffer as it is the virtio header */
+               desc = &vq->desc[desc->next];
+
+               /* Buffer address translation. */
+               vb_addr = gpa_to_vva(dev, desc->addr);
+               /* Prefetch buffer address. */
+               rte_prefetch0((void *)(uintptr_t)vb_addr);
+
+               used_idx = vq->last_used_idx & (vq->size - 1);
+
+               if (entry_success < (free_entries - 1)) {
+                       /* Prefetch descriptor index. */
+                       rte_prefetch0(&vq->desc[head[entry_success+1]]);
+                       rte_prefetch0(&vq->used->ring[(used_idx + 1) & (vq->size - 1)]);
+               }
+
+               /* Update used index buffer information. */
+               vq->used->ring[used_idx].id = head[entry_success];
+               vq->used->ring[used_idx].len = 0;
+
+               vb_offset = 0;
+               vb_avail = desc->len;
+               seg_offset = 0;
+               seg_avail = buf_size;
+               cpy_len = RTE_MIN(vb_avail, seg_avail);
+
+               PRINT_PACKET(dev, (uintptr_t)vb_addr, desc->len, 0);
+
+               /* Allocate an mbuf and populate the structure. */
+               m = rte_pktmbuf_alloc(mbuf_pool);
+               if (unlikely(m == NULL)) {
+                       RTE_LOG(ERR, VHOST_DATA,
+                               "Failed to allocate memory for mbuf.\n");
+                       return;
+               }
+
+               seg_num++;
+               cur = m;
+               prev = m;
+               while (cpy_len != 0) {
+                       rte_memcpy((void *)(rte_pktmbuf_mtod(cur, char *) + seg_offset),
+                               (void *)((uintptr_t)(vb_addr + vb_offset)),
+                               cpy_len);
+
+                       seg_offset += cpy_len;
+                       vb_offset += cpy_len;
+                       vb_avail -= cpy_len;
+                       seg_avail -= cpy_len;
+
+                       if (vb_avail != 0) {
+                               /*
+                                * The segment reachs to its end,
+                                * while the virtio buffer in TX vring has
+                                * more data to be copied.
+                                */
+                               cur->data_len = seg_offset;
+                               m->pkt_len += seg_offset;
+                               /* Allocate mbuf and populate the structure. */
+                               cur = rte_pktmbuf_alloc(mbuf_pool);
+                               if (unlikely(cur == NULL)) {
+                                       RTE_LOG(ERR, VHOST_DATA, "Failed to "
+                                               "allocate memory for mbuf.\n");
+                                       rte_pktmbuf_free(m);
+                                       alloc_err = 1;
+                                       break;
+                               }
+
+                               seg_num++;
+                               prev->next = cur;
+                               prev = cur;
+                               seg_offset = 0;
+                               seg_avail = buf_size;
+                       } else {
+                               if (desc->flags & VRING_DESC_F_NEXT) {
+                                       /*
+                                        * There are more virtio buffers in
+                                        * same vring entry need to be copied.
+                                        */
+                                       if (seg_avail == 0) {
+                                               /*
+                                                * The current segment hasn't
+                                                * room to accomodate more
+                                                * data.
+                                                */
+                                               cur->data_len = seg_offset;
+                                               m->pkt_len += seg_offset;
+                                               /*
+                                                * Allocate an mbuf and
+                                                * populate the structure.
+                                                */
+                                               cur = rte_pktmbuf_alloc(mbuf_pool);
+                                               if (unlikely(cur == NULL)) {
+                                                       RTE_LOG(ERR,
+                                                               VHOST_DATA,
+                                                               "Failed to "
+                                                               "allocate memory "
+                                                               "for mbuf\n");
+                                                       rte_pktmbuf_free(m);
+                                                       alloc_err = 1;
+                                                       break;
+                                               }
+                                               seg_num++;
+                                               prev->next = cur;
+                                               prev = cur;
+                                               seg_offset = 0;
+                                               seg_avail = buf_size;
+                                       }
+
+                                       desc = &vq->desc[desc->next];
+
+                                       /* Buffer address translation. */
+                                       vb_addr = gpa_to_vva(dev, desc->addr);
+                                       /* Prefetch buffer address. */
+                                       rte_prefetch0((void *)(uintptr_t)vb_addr);
+                                       vb_offset = 0;
+                                       vb_avail = desc->len;
+
+                                       PRINT_PACKET(dev, (uintptr_t)vb_addr,
+                                               desc->len, 0);
+                               } else {
+                                       /* The whole packet completes. */
+                                       cur->data_len = seg_offset;
+                                       m->pkt_len += seg_offset;
+                                       vb_avail = 0;
+                               }
+                       }
+
+                       cpy_len = RTE_MIN(vb_avail, seg_avail);
+               }
+
+               if (unlikely(alloc_err == 1))
+                       break;
+
+               m->nb_segs = seg_num;
+
+               /*
+                * If this is the first received packet we need to learn
+                * the MAC and setup VMDQ
+                */
+               if (dev->ready == DEVICE_MAC_LEARNING) {
+                       if (dev->remove || (link_vmdq(dev, m) == -1)) {
+                               /*
+                                * Discard frame if device is scheduled for
+                                * removal or a duplicate MAC address is found.
+                                */
+                               entry_success = free_entries;
+                               vq->last_used_idx += entry_success;
+                               rte_pktmbuf_free(m);
+                               break;
+                       }
+               }
+
+               virtio_tx_route(dev, m, mbuf_pool, (uint16_t)dev->device_fh);
+               vq->last_used_idx++;
+               entry_success++;
+               rte_pktmbuf_free(m);
+       }
+
+       rte_compiler_barrier();
+       vq->used->idx += entry_success;
+       /* Kick guest if required. */
+       if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT))
+               eventfd_write((int)vq->kickfd, 1);
+
+}
+
+/*
+ * This function is called by each data core. It handles all RX/TX registered with the
+ * core. For TX the specific lcore linked list is used. For RX, MAC addresses are compared
+ * with all devices in the main linked list.
+ */
+static int
+switch_worker(__attribute__((unused)) void *arg)
+{
+       struct rte_mempool *mbuf_pool = arg;
+       struct virtio_net *dev = NULL;
+       struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
+       struct virtio_net_data_ll *dev_ll;
+       struct mbuf_table *tx_q;
+       volatile struct lcore_ll_info *lcore_ll;
+       const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
+       uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
+       unsigned ret, i;
+       const uint16_t lcore_id = rte_lcore_id();
+       const uint16_t num_cores = (uint16_t)rte_lcore_count();
+       uint16_t rx_count = 0;
+       uint32_t mergeable = 0;
+
+       RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
+       lcore_ll = lcore_info[lcore_id].lcore_ll;
+       prev_tsc = 0;
+
+       tx_q = &lcore_tx_queue[lcore_id];
+       for (i = 0; i < num_cores; i ++) {
+               if (lcore_ids[i] == lcore_id) {
+                       tx_q->txq_id = i;
+                       break;
+               }
+       }
+
+       while(1) {
+               cur_tsc = rte_rdtsc();
+               /*
+                * TX burst queue drain
+                */
+               diff_tsc = cur_tsc - prev_tsc;
+               if (unlikely(diff_tsc > drain_tsc)) {
+
+                       if (tx_q->len) {
+                               LOG_DEBUG(VHOST_DATA, "TX queue drained after timeout with burst size %u \n", tx_q->len);
+
+                               /*Tx any packets in the queue*/
+                               ret = rte_eth_tx_burst(ports[0], (uint16_t)tx_q->txq_id,
+                                                                          (struct rte_mbuf **)tx_q->m_table,
+                                                                          (uint16_t)tx_q->len);
+                               if (unlikely(ret < tx_q->len)) {
+                                       do {
+                                               rte_pktmbuf_free(tx_q->m_table[ret]);
+                                       } while (++ret < tx_q->len);
+                               }
+
+                               tx_q->len = 0;
+                       }
+
+                       prev_tsc = cur_tsc;
+
+               }
+
+               rte_prefetch0(lcore_ll->ll_root_used);
+               /*
+                * Inform the configuration core that we have exited the linked list and that no devices are
+                * in use if requested.
+                */
+               if (lcore_ll->dev_removal_flag == REQUEST_DEV_REMOVAL)
+                       lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
+
+               /*
+                * Process devices
+                */
+               dev_ll = lcore_ll->ll_root_used;
+
+               while (dev_ll != NULL) {
+                       /*get virtio device ID*/
+                       dev = dev_ll->dev;
+                       mergeable =
+                               dev->features & (1 << VIRTIO_NET_F_MRG_RXBUF);
+
+                       if (dev->remove) {
+                               dev_ll = dev_ll->next;
+                               unlink_vmdq(dev);
+                               dev->ready = DEVICE_SAFE_REMOVE;
+                               continue;
+                       }
+                       if (likely(dev->ready == DEVICE_RX)) {
+                               /*Handle guest RX*/
+                               rx_count = rte_eth_rx_burst(ports[0],
+                                       (uint16_t)dev->vmdq_rx_q, pkts_burst, MAX_PKT_BURST);
+
+                               if (rx_count) {
+                                       if (likely(mergeable == 0))
+                                               ret_count =
+                                                       virtio_dev_rx(dev,
+                                                       pkts_burst, rx_count);
+                                       else
+                                               ret_count =
+                                                       virtio_dev_merge_rx(dev,
+                                                       pkts_burst, rx_count);
+
+                                       if (enable_stats) {
+                                               rte_atomic64_add(
+                                               &dev_statistics[dev_ll->dev->device_fh].rx_total_atomic,
+                                               rx_count);
+                                               rte_atomic64_add(
+                                               &dev_statistics[dev_ll->dev->device_fh].rx_atomic, ret_count);
+                                       }
+                                       while (likely(rx_count)) {
+                                               rx_count--;
+                                               rte_pktmbuf_free(pkts_burst[rx_count]);
+                                       }
+
+                               }
+                       }
+
+                       if (!dev->remove) {
+                               /*Handle guest TX*/
+                               if (likely(mergeable == 0))
+                                       virtio_dev_tx(dev, mbuf_pool);
+                               else
+                                       virtio_dev_merge_tx(dev, mbuf_pool);
+                       }
+
+                       /*move to the next device in the list*/
+                       dev_ll = dev_ll->next;
+               }
+       }
+
+       return 0;
+}
+
+/*
+ * This function gets available ring number for zero copy rx.
+ * Only one thread will call this funciton for a paticular virtio device,
+ * so, it is designed as non-thread-safe function.
+ */
+static inline uint32_t __attribute__((always_inline))
+get_available_ring_num_zcp(struct virtio_net *dev)
+{
+       struct vhost_virtqueue *vq = dev->virtqueue[VIRTIO_RXQ];
+       uint16_t avail_idx;
+
+       avail_idx = *((volatile uint16_t *)&vq->avail->idx);
+       return (uint32_t)(avail_idx - vq->last_used_idx_res);
+}
+
+/*
+ * This function gets available ring index for zero copy rx,
+ * it will retry 'burst_rx_retry_num' times till it get enough ring index.
+ * Only one thread will call this funciton for a paticular virtio device,
+ * so, it is designed as non-thread-safe function.
+ */
+static inline uint32_t __attribute__((always_inline))
+get_available_ring_index_zcp(struct virtio_net *dev,
+       uint16_t *res_base_idx, uint32_t count)
+{
+       struct vhost_virtqueue *vq = dev->virtqueue[VIRTIO_RXQ];
+       uint16_t avail_idx;
+       uint32_t retry = 0;
+       uint16_t free_entries;
+
+       *res_base_idx = vq->last_used_idx_res;
+       avail_idx = *((volatile uint16_t *)&vq->avail->idx);
+       free_entries = (avail_idx - *res_base_idx);
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") in get_available_ring_index_zcp: "
+                       "avail idx: %d, "
+                       "res base idx:%d, free entries:%d\n",
+                       dev->device_fh, avail_idx, *res_base_idx,
+                       free_entries);
+
+       /*
+        * If retry is enabled and the queue is full then we wait
+        * and retry to avoid packet loss.
+        */
+       if (enable_retry && unlikely(count > free_entries)) {
+               for (retry = 0; retry < burst_rx_retry_num; retry++) {
+                       rte_delay_us(burst_rx_delay_time);
+                       avail_idx = *((volatile uint16_t *)&vq->avail->idx);
+                       free_entries = (avail_idx - *res_base_idx);
+                       if (count <= free_entries)
+                               break;
+               }
+       }
+
+       /*check that we have enough buffers*/
+       if (unlikely(count > free_entries))
+               count = free_entries;
+
+       if (unlikely(count == 0)) {
+               LOG_DEBUG(VHOST_DATA,
+                       "(%"PRIu64") Fail in get_available_ring_index_zcp: "
+                       "avail idx: %d, res base idx:%d, free entries:%d\n",
+                       dev->device_fh, avail_idx,
+                       *res_base_idx, free_entries);
+               return 0;
+       }
+
+       vq->last_used_idx_res = *res_base_idx + count;
+
+       return count;
+}
+
+/*
+ * This function put descriptor back to used list.
+ */
+static inline void __attribute__((always_inline))
+put_desc_to_used_list_zcp(struct vhost_virtqueue *vq, uint16_t desc_idx)
+{
+       uint16_t res_cur_idx = vq->last_used_idx;
+       vq->used->ring[res_cur_idx & (vq->size - 1)].id = (uint32_t)desc_idx;
+       vq->used->ring[res_cur_idx & (vq->size - 1)].len = 0;
+       rte_compiler_barrier();
+       *(volatile uint16_t *)&vq->used->idx += 1;
+       vq->last_used_idx += 1;
+
+       /* Kick the guest if necessary. */
+       if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT))
+               eventfd_write((int)vq->kickfd, 1);
+}
+
+/*
+ * This function get available descriptor from vitio vring and un-attached mbuf
+ * from vpool->ring, and then attach them together. It needs adjust the offset
+ * for buff_addr and phys_addr accroding to PMD implementation, otherwise the
+ * frame data may be put to wrong location in mbuf.
+ */
+static inline void __attribute__((always_inline))
+attach_rxmbuf_zcp(struct virtio_net *dev)
+{
+       uint16_t res_base_idx, desc_idx;
+       uint64_t buff_addr, phys_addr;
+       struct vhost_virtqueue *vq;
+       struct vring_desc *desc;
+       struct rte_mbuf *mbuf = NULL;
+       struct vpool *vpool;
+       hpa_type addr_type;
+
+       vpool = &vpool_array[dev->vmdq_rx_q];
+       vq = dev->virtqueue[VIRTIO_RXQ];
+
+       do {
+               if (unlikely(get_available_ring_index_zcp(dev, &res_base_idx,
+                               1) != 1))
+                       return;
+               desc_idx = vq->avail->ring[(res_base_idx) & (vq->size - 1)];
+
+               desc = &vq->desc[desc_idx];
+               if (desc->flags & VRING_DESC_F_NEXT) {
+                       desc = &vq->desc[desc->next];
+                       buff_addr = gpa_to_vva(dev, desc->addr);
+                       phys_addr = gpa_to_hpa(dev, desc->addr, desc->len,
+                                       &addr_type);
+               } else {
+                       buff_addr = gpa_to_vva(dev,
+                                       desc->addr + vq->vhost_hlen);
+                       phys_addr = gpa_to_hpa(dev,
+                                       desc->addr + vq->vhost_hlen,
+                                       desc->len, &addr_type);
+               }
+
+               if (unlikely(addr_type == PHYS_ADDR_INVALID)) {
+                       RTE_LOG(ERR, VHOST_DATA, "(%"PRIu64") Invalid frame buffer"
+                               " address found when attaching RX frame buffer"
+                               " address!\n", dev->device_fh);
+                       put_desc_to_used_list_zcp(vq, desc_idx);
+                       continue;
+               }
+
+               /*
+                * Check if the frame buffer address from guest crosses
+                * sub-region or not.
+                */
+               if (unlikely(addr_type == PHYS_ADDR_CROSS_SUBREG)) {
+                       RTE_LOG(ERR, VHOST_DATA,
+                               "(%"PRIu64") Frame buffer address cross "
+                               "sub-regioin found when attaching RX frame "
+                               "buffer address!\n",
+                               dev->device_fh);
+                       put_desc_to_used_list_zcp(vq, desc_idx);
+                       continue;
+               }
+       } while (unlikely(phys_addr == 0));
+
+       rte_ring_sc_dequeue(vpool->ring, (void **)&mbuf);
+       if (unlikely(mbuf == NULL)) {
+               LOG_DEBUG(VHOST_DATA,
+                       "(%"PRIu64") in attach_rxmbuf_zcp: "
+                       "ring_sc_dequeue fail.\n",
+                       dev->device_fh);
+               put_desc_to_used_list_zcp(vq, desc_idx);
+               return;
+       }
+
+       if (unlikely(vpool->buf_size > desc->len)) {
+               LOG_DEBUG(VHOST_DATA,
+                       "(%"PRIu64") in attach_rxmbuf_zcp: frame buffer "
+                       "length(%d) of descriptor idx: %d less than room "
+                       "size required: %d\n",
+                       dev->device_fh, desc->len, desc_idx, vpool->buf_size);
+               put_desc_to_used_list_zcp(vq, desc_idx);
+               rte_ring_sp_enqueue(vpool->ring, (void *)mbuf);
+               return;
+       }
+
+       mbuf->buf_addr = (void *)(uintptr_t)(buff_addr - RTE_PKTMBUF_HEADROOM);
+       mbuf->data_off = RTE_PKTMBUF_HEADROOM;
+       mbuf->buf_physaddr = phys_addr - RTE_PKTMBUF_HEADROOM;
+       mbuf->data_len = desc->len;
+       MBUF_HEADROOM_UINT32(mbuf) = (uint32_t)desc_idx;
+
+       LOG_DEBUG(VHOST_DATA,
+               "(%"PRIu64") in attach_rxmbuf_zcp: res base idx:%d, "
+               "descriptor idx:%d\n",
+               dev->device_fh, res_base_idx, desc_idx);
+
+       __rte_mbuf_raw_free(mbuf);
+
+       return;
+}
+
+/*
+ * Detach an attched packet mbuf -
+ *  - restore original mbuf address and length values.
+ *  - reset pktmbuf data and data_len to their default values.
+ *  All other fields of the given packet mbuf will be left intact.
+ *
+ * @param m
+ *   The attached packet mbuf.
+ */
+static inline void pktmbuf_detach_zcp(struct rte_mbuf *m)
+{
+       const struct rte_mempool *mp = m->pool;
+       void *buf = RTE_MBUF_TO_BADDR(m);
+       uint32_t buf_ofs;
+       uint32_t buf_len = mp->elt_size - sizeof(*m);
+       m->buf_physaddr = rte_mempool_virt2phy(mp, m) + sizeof(*m);
+
+       m->buf_addr = buf;
+       m->buf_len = (uint16_t)buf_len;
+
+       buf_ofs = (RTE_PKTMBUF_HEADROOM <= m->buf_len) ?
+                       RTE_PKTMBUF_HEADROOM : m->buf_len;
+       m->data_off = buf_ofs;
+
+       m->data_len = 0;
+}
+
+/*
+ * This function is called after packets have been transimited. It fetchs mbuf
+ * from vpool->pool, detached it and put into vpool->ring. It also update the
+ * used index and kick the guest if necessary.
+ */
+static inline uint32_t __attribute__((always_inline))
+txmbuf_clean_zcp(struct virtio_net *dev, struct vpool *vpool)
+{
+       struct rte_mbuf *mbuf;
+       struct vhost_virtqueue *vq = dev->virtqueue[VIRTIO_TXQ];
+       uint32_t used_idx = vq->last_used_idx & (vq->size - 1);
+       uint32_t index = 0;
+       uint32_t mbuf_count = rte_mempool_count(vpool->pool);
+
+       LOG_DEBUG(VHOST_DATA,
+               "(%"PRIu64") in txmbuf_clean_zcp: mbuf count in mempool before "
+               "clean is: %d\n",
+               dev->device_fh, mbuf_count);
+       LOG_DEBUG(VHOST_DATA,
+               "(%"PRIu64") in txmbuf_clean_zcp: mbuf count in  ring before "
+               "clean  is : %d\n",
+               dev->device_fh, rte_ring_count(vpool->ring));
+
+       for (index = 0; index < mbuf_count; index++) {
+               mbuf = __rte_mbuf_raw_alloc(vpool->pool);
+               if (likely(RTE_MBUF_INDIRECT(mbuf)))
+                       pktmbuf_detach_zcp(mbuf);
+               rte_ring_sp_enqueue(vpool->ring, mbuf);
+
+               /* Update used index buffer information. */
+               vq->used->ring[used_idx].id = MBUF_HEADROOM_UINT32(mbuf);
+               vq->used->ring[used_idx].len = 0;
+
+               used_idx = (used_idx + 1) & (vq->size - 1);
+       }
+
+       LOG_DEBUG(VHOST_DATA,
+               "(%"PRIu64") in txmbuf_clean_zcp: mbuf count in mempool after "
+               "clean is: %d\n",
+               dev->device_fh, rte_mempool_count(vpool->pool));
+       LOG_DEBUG(VHOST_DATA,
+               "(%"PRIu64") in txmbuf_clean_zcp: mbuf count in  ring after "
+               "clean  is : %d\n",
+               dev->device_fh, rte_ring_count(vpool->ring));
+       LOG_DEBUG(VHOST_DATA,
+               "(%"PRIu64") in txmbuf_clean_zcp: before updated "
+               "vq->last_used_idx:%d\n",
+               dev->device_fh, vq->last_used_idx);
+
+       vq->last_used_idx += mbuf_count;
+
+       LOG_DEBUG(VHOST_DATA,
+               "(%"PRIu64") in txmbuf_clean_zcp: after updated "
+               "vq->last_used_idx:%d\n",
+               dev->device_fh, vq->last_used_idx);
+
+       rte_compiler_barrier();
+
+       *(volatile uint16_t *)&vq->used->idx += mbuf_count;
+
+       /* Kick guest if required. */
+       if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT))
+               eventfd_write((int)vq->kickfd, 1);
+
+       return 0;
+}
+
+/*
+ * This function is called when a virtio device is destroy.
+ * It fetchs mbuf from vpool->pool, and detached it, and put into vpool->ring.
+ */
+static void mbuf_destroy_zcp(struct vpool *vpool)
+{
+       struct rte_mbuf *mbuf = NULL;
+       uint32_t index, mbuf_count = rte_mempool_count(vpool->pool);
+
+       LOG_DEBUG(VHOST_CONFIG,
+               "in mbuf_destroy_zcp: mbuf count in mempool before "
+               "mbuf_destroy_zcp is: %d\n",
+               mbuf_count);
+       LOG_DEBUG(VHOST_CONFIG,
+               "in mbuf_destroy_zcp: mbuf count in  ring before "
+               "mbuf_destroy_zcp  is : %d\n",
+               rte_ring_count(vpool->ring));
+
+       for (index = 0; index < mbuf_count; index++) {
+               mbuf = __rte_mbuf_raw_alloc(vpool->pool);
+               if (likely(mbuf != NULL)) {
+                       if (likely(RTE_MBUF_INDIRECT(mbuf)))
+                               pktmbuf_detach_zcp(mbuf);
+                       rte_ring_sp_enqueue(vpool->ring, (void *)mbuf);
+               }
+       }
+
+       LOG_DEBUG(VHOST_CONFIG,
+               "in mbuf_destroy_zcp: mbuf count in mempool after "
+               "mbuf_destroy_zcp is: %d\n",
+               rte_mempool_count(vpool->pool));
+       LOG_DEBUG(VHOST_CONFIG,
+               "in mbuf_destroy_zcp: mbuf count in ring after "
+               "mbuf_destroy_zcp is : %d\n",
+               rte_ring_count(vpool->ring));
+}
+
+/*
+ * This function update the use flag and counter.
+ */
+static inline uint32_t __attribute__((always_inline))
+virtio_dev_rx_zcp(struct virtio_net *dev, struct rte_mbuf **pkts,
+       uint32_t count)
+{
+       struct vhost_virtqueue *vq;
+       struct vring_desc *desc;
+       struct rte_mbuf *buff;
+       /* The virtio_hdr is initialised to 0. */
+       struct virtio_net_hdr_mrg_rxbuf virtio_hdr
+               = {{0, 0, 0, 0, 0, 0}, 0};
+       uint64_t buff_hdr_addr = 0;
+       uint32_t head[MAX_PKT_BURST], packet_len = 0;
+       uint32_t head_idx, packet_success = 0;
+       uint16_t res_cur_idx;
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") virtio_dev_rx()\n", dev->device_fh);
+
+       if (count == 0)
+               return 0;
+
+       vq = dev->virtqueue[VIRTIO_RXQ];
+       count = (count > MAX_PKT_BURST) ? MAX_PKT_BURST : count;
+
+       res_cur_idx = vq->last_used_idx;
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") Current Index %d| End Index %d\n",
+               dev->device_fh, res_cur_idx, res_cur_idx + count);
+
+       /* Retrieve all of the head indexes first to avoid caching issues. */
+       for (head_idx = 0; head_idx < count; head_idx++)
+               head[head_idx] = MBUF_HEADROOM_UINT32(pkts[head_idx]);
+
+       /*Prefetch descriptor index. */
+       rte_prefetch0(&vq->desc[head[packet_success]]);
+
+       while (packet_success != count) {
+               /* Get descriptor from available ring */
+               desc = &vq->desc[head[packet_success]];
+
+               buff = pkts[packet_success];
+               LOG_DEBUG(VHOST_DATA,
+                       "(%"PRIu64") in dev_rx_zcp: update the used idx for "
+                       "pkt[%d] descriptor idx: %d\n",
+                       dev->device_fh, packet_success,
+                       MBUF_HEADROOM_UINT32(buff));
+
+               PRINT_PACKET(dev,
+                       (uintptr_t)(((uint64_t)(uintptr_t)buff->buf_addr)
+                       + RTE_PKTMBUF_HEADROOM),
+                       rte_pktmbuf_data_len(buff), 0);
+
+               /* Buffer address translation for virtio header. */
+               buff_hdr_addr = gpa_to_vva(dev, desc->addr);
+               packet_len = rte_pktmbuf_data_len(buff) + vq->vhost_hlen;
+
+               /*
+                * If the descriptors are chained the header and data are
+                * placed in separate buffers.
+                */
+               if (desc->flags & VRING_DESC_F_NEXT) {
+                       desc->len = vq->vhost_hlen;
+                       desc = &vq->desc[desc->next];
+                       desc->len = rte_pktmbuf_data_len(buff);
+               } else {
+                       desc->len = packet_len;
+               }
+
+               /* Update used ring with desc information */
+               vq->used->ring[res_cur_idx & (vq->size - 1)].id
+                       = head[packet_success];
+               vq->used->ring[res_cur_idx & (vq->size - 1)].len
+                       = packet_len;
+               res_cur_idx++;
+               packet_success++;
+
+               /* A header is required per buffer. */
+               rte_memcpy((void *)(uintptr_t)buff_hdr_addr,
+                       (const void *)&virtio_hdr, vq->vhost_hlen);
+
+               PRINT_PACKET(dev, (uintptr_t)buff_hdr_addr, vq->vhost_hlen, 1);
+
+               if (likely(packet_success < count)) {
+                       /* Prefetch descriptor index. */
+                       rte_prefetch0(&vq->desc[head[packet_success]]);
+               }
+       }
+
+       rte_compiler_barrier();
+
+       LOG_DEBUG(VHOST_DATA,
+               "(%"PRIu64") in dev_rx_zcp: before update used idx: "
+               "vq.last_used_idx: %d, vq->used->idx: %d\n",
+               dev->device_fh, vq->last_used_idx, vq->used->idx);
+
+       *(volatile uint16_t *)&vq->used->idx += count;
+       vq->last_used_idx += count;
+
+       LOG_DEBUG(VHOST_DATA,
+               "(%"PRIu64") in dev_rx_zcp: after  update used idx: "
+               "vq.last_used_idx: %d, vq->used->idx: %d\n",
+               dev->device_fh, vq->last_used_idx, vq->used->idx);
+
+       /* Kick the guest if necessary. */
+       if (!(vq->avail->flags & VRING_AVAIL_F_NO_INTERRUPT))
+               eventfd_write((int)vq->kickfd, 1);
+
+       return count;
+}
+
+/*
+ * This function routes the TX packet to the correct interface.
+ * This may be a local device or the physical port.
+ */
+static inline void __attribute__((always_inline))
+virtio_tx_route_zcp(struct virtio_net *dev, struct rte_mbuf *m,
+       uint32_t desc_idx, uint8_t need_copy)
+{
+       struct mbuf_table *tx_q;
+       struct rte_mbuf **m_table;
+       struct rte_mbuf *mbuf = NULL;
+       unsigned len, ret, offset = 0;
+       struct vpool *vpool;
+       struct virtio_net_data_ll *dev_ll = ll_root_used;
+       struct ether_hdr *pkt_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
+       uint16_t vlan_tag = (uint16_t)vlan_tags[(uint16_t)dev->device_fh];
+
+       /*Add packet to the port tx queue*/
+       tx_q = &tx_queue_zcp[(uint16_t)dev->vmdq_rx_q];
+       len = tx_q->len;
+
+       /* Allocate an mbuf and populate the structure. */
+       vpool = &vpool_array[MAX_QUEUES + (uint16_t)dev->vmdq_rx_q];
+       rte_ring_sc_dequeue(vpool->ring, (void **)&mbuf);
+       if (unlikely(mbuf == NULL)) {
+               struct vhost_virtqueue *vq = dev->virtqueue[VIRTIO_TXQ];
+               RTE_LOG(ERR, VHOST_DATA,
+                       "(%"PRIu64") Failed to allocate memory for mbuf.\n",
+                       dev->device_fh);
+               put_desc_to_used_list_zcp(vq, desc_idx);
+               return;
+       }
+
+       if (vm2vm_mode == VM2VM_HARDWARE) {
+               /* Avoid using a vlan tag from any vm for external pkt, such as
+                * vlan_tags[dev->device_fh], oterwise, it conflicts when pool
+                * selection, MAC address determines it as an external pkt
+                * which should go to network, while vlan tag determine it as
+                * a vm2vm pkt should forward to another vm. Hardware confuse
+                * such a ambiguous situation, so pkt will lost.
+                */
+               vlan_tag = external_pkt_default_vlan_tag;
+               while (dev_ll != NULL) {
+                       if (likely(dev_ll->dev->ready == DEVICE_RX) &&
+                               ether_addr_cmp(&(pkt_hdr->d_addr),
+                               &dev_ll->dev->mac_address)) {
+
+                               /*
+                                * Drop the packet if the TX packet is destined
+                                * for the TX device.
+                                */
+                               if (unlikely(dev_ll->dev->device_fh
+                                       == dev->device_fh)) {
+                                       LOG_DEBUG(VHOST_DATA,
+                                       "(%"PRIu64") TX: Source and destination"
+                                       "MAC addresses are the same. Dropping "
+                                       "packet.\n",
+                                       dev_ll->dev->device_fh);
+                                       MBUF_HEADROOM_UINT32(mbuf)
+                                               = (uint32_t)desc_idx;
+                                       __rte_mbuf_raw_free(mbuf);
+                                       return;
+                               }
+
+                               /*
+                                * Packet length offset 4 bytes for HW vlan
+                                * strip when L2 switch back.
+                                */
+                               offset = 4;
+                               vlan_tag =
+                               (uint16_t)
+                               vlan_tags[(uint16_t)dev_ll->dev->device_fh];
+
+                               LOG_DEBUG(VHOST_DATA,
+                               "(%"PRIu64") TX: pkt to local VM device id:"
+                               "(%"PRIu64") vlan tag: %d.\n",
+                               dev->device_fh, dev_ll->dev->device_fh,
+                               vlan_tag);
+
+                               break;
+                       }
+                       dev_ll = dev_ll->next;
+               }
+       }
+
+       mbuf->nb_segs = m->nb_segs;
+       mbuf->next = m->next;
+       mbuf->data_len = m->data_len + offset;
+       mbuf->pkt_len = mbuf->data_len;
+       if (unlikely(need_copy)) {
+               /* Copy the packet contents to the mbuf. */
+               rte_memcpy(rte_pktmbuf_mtod(mbuf, void *),
+                       rte_pktmbuf_mtod(m, void *),
+                       m->data_len);
+       } else {
+               mbuf->data_off = m->data_off;
+               mbuf->buf_physaddr = m->buf_physaddr;
+               mbuf->buf_addr = m->buf_addr;
+       }
+       mbuf->ol_flags = PKT_TX_VLAN_PKT;
+       mbuf->vlan_tci = vlan_tag;
+       mbuf->l2_len = sizeof(struct ether_hdr);
+       mbuf->l3_len = sizeof(struct ipv4_hdr);
+       MBUF_HEADROOM_UINT32(mbuf) = (uint32_t)desc_idx;
+
+       tx_q->m_table[len] = mbuf;
+       len++;
+
+       LOG_DEBUG(VHOST_DATA,
+               "(%"PRIu64") in tx_route_zcp: pkt: nb_seg: %d, next:%s\n",
+               dev->device_fh,
+               mbuf->nb_segs,
+               (mbuf->next == NULL) ? "null" : "non-null");
+
+       if (enable_stats) {
+               dev_statistics[dev->device_fh].tx_total++;
+               dev_statistics[dev->device_fh].tx++;
+       }
+
+       if (unlikely(len == MAX_PKT_BURST)) {
+               m_table = (struct rte_mbuf **)tx_q->m_table;
+               ret = rte_eth_tx_burst(ports[0],
+                       (uint16_t)tx_q->txq_id, m_table, (uint16_t) len);
+
+               /*
+                * Free any buffers not handled by TX and update
+                * the port stats.
+                */
+               if (unlikely(ret < len)) {
+                       do {
+                               rte_pktmbuf_free(m_table[ret]);
+                       } while (++ret < len);
+               }
+
+               len = 0;
+               txmbuf_clean_zcp(dev, vpool);
+       }
+
+       tx_q->len = len;
+
+       return;
+}
+
+/*
+ * This function TX all available packets in virtio TX queue for one
+ * virtio-net device. If it is first packet, it learns MAC address and
+ * setup VMDQ.
+ */
+static inline void __attribute__((always_inline))
+virtio_dev_tx_zcp(struct virtio_net *dev)
+{
+       struct rte_mbuf m;
+       struct vhost_virtqueue *vq;
+       struct vring_desc *desc;
+       uint64_t buff_addr = 0, phys_addr;
+       uint32_t head[MAX_PKT_BURST];
+       uint32_t i;
+       uint16_t free_entries, packet_success = 0;
+       uint16_t avail_idx;
+       uint8_t need_copy = 0;
+       hpa_type addr_type;
+
+       vq = dev->virtqueue[VIRTIO_TXQ];
+       avail_idx =  *((volatile uint16_t *)&vq->avail->idx);
+
+       /* If there are no available buffers then return. */
+       if (vq->last_used_idx_res == avail_idx)
+               return;
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") virtio_dev_tx()\n", dev->device_fh);
+
+       /* Prefetch available ring to retrieve head indexes. */
+       rte_prefetch0(&vq->avail->ring[vq->last_used_idx_res & (vq->size - 1)]);
+
+       /* Get the number of free entries in the ring */
+       free_entries = (avail_idx - vq->last_used_idx_res);
+
+       /* Limit to MAX_PKT_BURST. */
+       free_entries
+               = (free_entries > MAX_PKT_BURST) ? MAX_PKT_BURST : free_entries;
+
+       LOG_DEBUG(VHOST_DATA, "(%"PRIu64") Buffers available %d\n",
+               dev->device_fh, free_entries);
+
+       /* Retrieve all of the head indexes first to avoid caching issues. */
+       for (i = 0; i < free_entries; i++)
+               head[i]
+                       = vq->avail->ring[(vq->last_used_idx_res + i)
+                       & (vq->size - 1)];
+
+       vq->last_used_idx_res += free_entries;
+
+       /* Prefetch descriptor index. */
+       rte_prefetch0(&vq->desc[head[packet_success]]);
+       rte_prefetch0(&vq->used->ring[vq->last_used_idx & (vq->size - 1)]);
+
+       while (packet_success < free_entries) {
+               desc = &vq->desc[head[packet_success]];
+
+               /* Discard first buffer as it is the virtio header */
+               desc = &vq->desc[desc->next];
+
+               /* Buffer address translation. */
+               buff_addr = gpa_to_vva(dev, desc->addr);
+               phys_addr = gpa_to_hpa(dev, desc->addr, desc->len, &addr_type);
+
+               if (likely(packet_success < (free_entries - 1)))
+                       /* Prefetch descriptor index. */
+                       rte_prefetch0(&vq->desc[head[packet_success + 1]]);
+
+               if (unlikely(addr_type == PHYS_ADDR_INVALID)) {
+                       RTE_LOG(ERR, VHOST_DATA,
+                               "(%"PRIu64") Invalid frame buffer address found"
+                               "when TX packets!\n",
+                               dev->device_fh);
+                       packet_success++;
+                       continue;
+               }
+
+               /* Prefetch buffer address. */
+               rte_prefetch0((void *)(uintptr_t)buff_addr);
+
+               /*
+                * Setup dummy mbuf. This is copied to a real mbuf if
+                * transmitted out the physical port.
+                */
+               m.data_len = desc->len;
+               m.nb_segs = 1;
+               m.next = NULL;
+               m.data_off = 0;
+               m.buf_addr = (void *)(uintptr_t)buff_addr;
+               m.buf_physaddr = phys_addr;
+
+               /*
+                * Check if the frame buffer address from guest crosses
+                * sub-region or not.
+                */
+               if (unlikely(addr_type == PHYS_ADDR_CROSS_SUBREG)) {
+                       RTE_LOG(ERR, VHOST_DATA,
+                               "(%"PRIu64") Frame buffer address cross "
+                               "sub-regioin found when attaching TX frame "
+                               "buffer address!\n",
+                               dev->device_fh);
+                       need_copy = 1;
+               } else
+                       need_copy = 0;
+
+               PRINT_PACKET(dev, (uintptr_t)buff_addr, desc->len, 0);
+
+               /*
+                * If this is the first received packet we need to learn
+                * the MAC and setup VMDQ
+                */
+               if (unlikely(dev->ready == DEVICE_MAC_LEARNING)) {
+                       if (dev->remove || (link_vmdq(dev, &m) == -1)) {
+                               /*
+                                * Discard frame if device is scheduled for
+                                * removal or a duplicate MAC address is found.
+                                */
+                               packet_success += free_entries;
+                               vq->last_used_idx += packet_success;
+                               break;
+                       }
+               }
+
+               virtio_tx_route_zcp(dev, &m, head[packet_success], need_copy);
+               packet_success++;
+       }
+}
+
+/*
+ * This function is called by each data core. It handles all RX/TX registered
+ * with the core. For TX the specific lcore linked list is used. For RX, MAC
+ * addresses are compared with all devices in the main linked list.
+ */
+static int
+switch_worker_zcp(__attribute__((unused)) void *arg)
+{
+       struct virtio_net *dev = NULL;
+       struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
+       struct virtio_net_data_ll *dev_ll;
+       struct mbuf_table *tx_q;
+       volatile struct lcore_ll_info *lcore_ll;
+       const uint64_t drain_tsc
+               = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S
+               * BURST_TX_DRAIN_US;
+       uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
+       unsigned ret;
+       const uint16_t lcore_id = rte_lcore_id();
+       uint16_t count_in_ring, rx_count = 0;
+
+       RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
+
+       lcore_ll = lcore_info[lcore_id].lcore_ll;
+       prev_tsc = 0;
+
+       while (1) {
+               cur_tsc = rte_rdtsc();
+
+               /* TX burst queue drain */
+               diff_tsc = cur_tsc - prev_tsc;
+               if (unlikely(diff_tsc > drain_tsc)) {
+                       /*
+                        * Get mbuf from vpool.pool and detach mbuf and
+                        * put back into vpool.ring.
+                        */
+                       dev_ll = lcore_ll->ll_root_used;
+                       while ((dev_ll != NULL) && (dev_ll->dev != NULL)) {
+                               /* Get virtio device ID */
+                               dev = dev_ll->dev;
+
+                               if (likely(!dev->remove)) {
+                                       tx_q = &tx_queue_zcp[(uint16_t)dev->vmdq_rx_q];
+                                       if (tx_q->len) {
+                                               LOG_DEBUG(VHOST_DATA,
+                                               "TX queue drained after timeout"
+                                               " with burst size %u\n",
+                                               tx_q->len);
+
+                                               /*
+                                                * Tx any packets in the queue
+                                                */
+                                               ret = rte_eth_tx_burst(
+                                                       ports[0],
+                                                       (uint16_t)tx_q->txq_id,
+                                                       (struct rte_mbuf **)
+                                                       tx_q->m_table,
+                                                       (uint16_t)tx_q->len);
+                                               if (unlikely(ret < tx_q->len)) {
+                                                       do {
+                                                               rte_pktmbuf_free(
+                                                                       tx_q->m_table[ret]);
+                                                       } while (++ret < tx_q->len);
+                                               }
+                                               tx_q->len = 0;
+
+                                               txmbuf_clean_zcp(dev,
+                                                       &vpool_array[MAX_QUEUES+dev->vmdq_rx_q]);
+                                       }
+                               }
+                               dev_ll = dev_ll->next;
+                       }
+                       prev_tsc = cur_tsc;
+               }
+
+               rte_prefetch0(lcore_ll->ll_root_used);
+
+               /*
+                * Inform the configuration core that we have exited the linked
+                * list and that no devices are in use if requested.
+                */
+               if (lcore_ll->dev_removal_flag == REQUEST_DEV_REMOVAL)
+                       lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
+
+               /* Process devices */
+               dev_ll = lcore_ll->ll_root_used;
+
+               while ((dev_ll != NULL) && (dev_ll->dev != NULL)) {
+                       dev = dev_ll->dev;
+                       if (unlikely(dev->remove)) {
+                               dev_ll = dev_ll->next;
+                               unlink_vmdq(dev);
+                               dev->ready = DEVICE_SAFE_REMOVE;
+                               continue;
+                       }
+
+                       if (likely(dev->ready == DEVICE_RX)) {
+                               uint32_t index = dev->vmdq_rx_q;
+                               uint16_t i;
+                               count_in_ring
+                               = rte_ring_count(vpool_array[index].ring);
+                               uint16_t free_entries
+                               = (uint16_t)get_available_ring_num_zcp(dev);
+
+                               /*
+                                * Attach all mbufs in vpool.ring and put back
+                                * into vpool.pool.
+                                */
+                               for (i = 0;
+                               i < RTE_MIN(free_entries,
+                               RTE_MIN(count_in_ring, MAX_PKT_BURST));
+                               i++)
+                                       attach_rxmbuf_zcp(dev);
+
+                               /* Handle guest RX */
+                               rx_count = rte_eth_rx_burst(ports[0],
+                                       (uint16_t)dev->vmdq_rx_q, pkts_burst,
+                                       MAX_PKT_BURST);
+
+                               if (rx_count) {
+                                       ret_count = virtio_dev_rx_zcp(dev,
+                                                       pkts_burst, rx_count);
+                                       if (enable_stats) {
+                                               dev_statistics[dev->device_fh].rx_total
+                                                       += rx_count;
+                                               dev_statistics[dev->device_fh].rx
+                                                       += ret_count;
+                                       }
+                                       while (likely(rx_count)) {
+                                               rx_count--;
+                                               pktmbuf_detach_zcp(
+                                                       pkts_burst[rx_count]);
+                                               rte_ring_sp_enqueue(
+                                                       vpool_array[index].ring,
+                                                       (void *)pkts_burst[rx_count]);
+                                       }
+                               }
+                       }
+
+                       if (likely(!dev->remove))
+                               /* Handle guest TX */
+                               virtio_dev_tx_zcp(dev);
+
+                       /* Move to the next device in the list */
+                       dev_ll = dev_ll->next;
+               }
+       }
+
+       return 0;
+}
+
+
+/*
+ * Add an entry to a used linked list. A free entry must first be found
+ * in the free linked list using get_data_ll_free_entry();
+ */
+static void
+add_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
+       struct virtio_net_data_ll *ll_dev)
+{
+       struct virtio_net_data_ll *ll = *ll_root_addr;
+
+       /* Set next as NULL and use a compiler barrier to avoid reordering. */
+       ll_dev->next = NULL;
+       rte_compiler_barrier();
+
+       /* If ll == NULL then this is the first device. */
+       if (ll) {
+               /* Increment to the tail of the linked list. */
+               while ((ll->next != NULL) )
+                       ll = ll->next;
+
+               ll->next = ll_dev;
+       } else {
+               *ll_root_addr = ll_dev;
+       }
+}
+
+/*
+ * Remove an entry from a used linked list. The entry must then be added to
+ * the free linked list using put_data_ll_free_entry().
+ */
+static void
+rm_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
+       struct virtio_net_data_ll *ll_dev,
+       struct virtio_net_data_ll *ll_dev_last)
+{
+       struct virtio_net_data_ll *ll = *ll_root_addr;
+
+       if (unlikely((ll == NULL) || (ll_dev == NULL)))
+               return;
+
+       if (ll_dev == ll)
+               *ll_root_addr = ll_dev->next;
+       else
+               if (likely(ll_dev_last != NULL))
+                       ll_dev_last->next = ll_dev->next;
+               else
+                       RTE_LOG(ERR, VHOST_CONFIG, "Remove entry form ll failed.\n");
+}
+
+/*
+ * Find and return an entry from the free linked list.
+ */
+static struct virtio_net_data_ll *
+get_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr)
+{
+       struct virtio_net_data_ll *ll_free = *ll_root_addr;
+       struct virtio_net_data_ll *ll_dev;
+
+       if (ll_free == NULL)
+               return NULL;
+
+       ll_dev = ll_free;
+       *ll_root_addr = ll_free->next;
+
+       return ll_dev;
+}
+
+/*
+ * Place an entry back on to the free linked list.
+ */
+static void
+put_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr,
+       struct virtio_net_data_ll *ll_dev)
+{
+       struct virtio_net_data_ll *ll_free = *ll_root_addr;
+
+       if (ll_dev == NULL)
+               return;
+
+       ll_dev->next = ll_free;
+       *ll_root_addr = ll_dev;
+}
+
+/*
+ * Creates a linked list of a given size.
+ */
+static struct virtio_net_data_ll *
+alloc_data_ll(uint32_t size)
+{
+       struct virtio_net_data_ll *ll_new;
+       uint32_t i;
+
+       /* Malloc and then chain the linked list. */
+       ll_new = malloc(size * sizeof(struct virtio_net_data_ll));
+       if (ll_new == NULL) {
+               RTE_LOG(ERR, VHOST_CONFIG, "Failed to allocate memory for ll_new.\n");
+               return NULL;
+       }
+
+       for (i = 0; i < size - 1; i++) {
+               ll_new[i].dev = NULL;
+               ll_new[i].next = &ll_new[i+1];
+       }
+       ll_new[i].next = NULL;
+
+       return (ll_new);
+}
+
+/*
+ * Create the main linked list along with each individual cores linked list. A used and a free list
+ * are created to manage entries.
+ */
+static int
+init_data_ll (void)
+{
+       int lcore;
+
+       RTE_LCORE_FOREACH_SLAVE(lcore) {
+               lcore_info[lcore].lcore_ll = malloc(sizeof(struct lcore_ll_info));
+               if (lcore_info[lcore].lcore_ll == NULL) {
+                       RTE_LOG(ERR, VHOST_CONFIG, "Failed to allocate memory for lcore_ll.\n");
+                       return -1;
+               }
+
+               lcore_info[lcore].lcore_ll->device_num = 0;
+               lcore_info[lcore].lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
+               lcore_info[lcore].lcore_ll->ll_root_used = NULL;
+               if (num_devices % num_switching_cores)
+                       lcore_info[lcore].lcore_ll->ll_root_free = alloc_data_ll((num_devices / num_switching_cores) + 1);
+               else
+                       lcore_info[lcore].lcore_ll->ll_root_free = alloc_data_ll(num_devices / num_switching_cores);
+       }
+
+       /* Allocate devices up to a maximum of MAX_DEVICES. */
+       ll_root_free = alloc_data_ll(MIN((num_devices), MAX_DEVICES));
+
+       return 0;
+}
+
+/*
+ * Set virtqueue flags so that we do not receive interrupts.
+ */
+static void
+set_irq_status (struct virtio_net *dev)
+{
+       dev->virtqueue[VIRTIO_RXQ]->used->flags = VRING_USED_F_NO_NOTIFY;
+       dev->virtqueue[VIRTIO_TXQ]->used->flags = VRING_USED_F_NO_NOTIFY;
+}
+
+/*
+ * Remove a device from the specific data core linked list and from the main linked list. Synchonization
+ * occurs through the use of the lcore dev_removal_flag. Device is made volatile here to avoid re-ordering
+ * of dev->remove=1 which can cause an infinite loop in the rte_pause loop.
+ */
+static void
+destroy_device (volatile struct virtio_net *dev)
+{
+       struct virtio_net_data_ll *ll_lcore_dev_cur;
+       struct virtio_net_data_ll *ll_main_dev_cur;
+       struct virtio_net_data_ll *ll_lcore_dev_last = NULL;
+       struct virtio_net_data_ll *ll_main_dev_last = NULL;
+       int lcore;
+
+       dev->flags &= ~VIRTIO_DEV_RUNNING;
+
+       /*set the remove flag. */
+       dev->remove = 1;
+
+       while(dev->ready != DEVICE_SAFE_REMOVE) {
+               rte_pause();
+       }
+
+       /* Search for entry to be removed from lcore ll */
+       ll_lcore_dev_cur = lcore_info[dev->coreid].lcore_ll->ll_root_used;
+       while (ll_lcore_dev_cur != NULL) {
+               if (ll_lcore_dev_cur->dev == dev) {
+                       break;
+               } else {
+                       ll_lcore_dev_last = ll_lcore_dev_cur;
+                       ll_lcore_dev_cur = ll_lcore_dev_cur->next;
+               }
+       }
+
+       if (ll_lcore_dev_cur == NULL) {
+               RTE_LOG(ERR, VHOST_CONFIG,
+                       "(%"PRIu64") Failed to find the dev to be destroy.\n",
+                       dev->device_fh);
+               return;
+       }
+
+       /* Search for entry to be removed from main ll */
+       ll_main_dev_cur = ll_root_used;
+       ll_main_dev_last = NULL;
+       while (ll_main_dev_cur != NULL) {
+               if (ll_main_dev_cur->dev == dev) {
+                       break;
+               } else {
+                       ll_main_dev_last = ll_main_dev_cur;
+                       ll_main_dev_cur = ll_main_dev_cur->next;
+               }
+       }
+
+       /* Remove entries from the lcore and main ll. */
+       rm_data_ll_entry(&lcore_info[ll_lcore_dev_cur->dev->coreid].lcore_ll->ll_root_used, ll_lcore_dev_cur, ll_lcore_dev_last);
+       rm_data_ll_entry(&ll_root_used, ll_main_dev_cur, ll_main_dev_last);
+
+       /* Set the dev_removal_flag on each lcore. */
+       RTE_LCORE_FOREACH_SLAVE(lcore) {
+               lcore_info[lcore].lcore_ll->dev_removal_flag = REQUEST_DEV_REMOVAL;
+       }
+
+       /*
+        * Once each core has set the dev_removal_flag to ACK_DEV_REMOVAL we can be sure that
+        * they can no longer access the device removed from the linked lists and that the devices
+        * are no longer in use.
+        */
+       RTE_LCORE_FOREACH_SLAVE(lcore) {
+               while (lcore_info[lcore].lcore_ll->dev_removal_flag != ACK_DEV_REMOVAL) {
+                       rte_pause();
+               }
+       }
+
+       /* Add the entries back to the lcore and main free ll.*/
+       put_data_ll_free_entry(&lcore_info[ll_lcore_dev_cur->dev->coreid].lcore_ll->ll_root_free, ll_lcore_dev_cur);
+       put_data_ll_free_entry(&ll_root_free, ll_main_dev_cur);
+
+       /* Decrement number of device on the lcore. */
+       lcore_info[ll_lcore_dev_cur->dev->coreid].lcore_ll->device_num--;
+
+       RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been removed from data core\n", dev->device_fh);
+
+       if (zero_copy) {
+               struct vpool *vpool = &vpool_array[dev->vmdq_rx_q];
+
+               /* Stop the RX queue. */
+               if (rte_eth_dev_rx_queue_stop(ports[0], dev->vmdq_rx_q) != 0) {
+                       LOG_DEBUG(VHOST_CONFIG,
+                               "(%"PRIu64") In destroy_device: Failed to stop "
+                               "rx queue:%d\n",
+                               dev->device_fh,
+                               dev->vmdq_rx_q);
+               }
+
+               LOG_DEBUG(VHOST_CONFIG,
+                       "(%"PRIu64") in destroy_device: Start put mbuf in "
+                       "mempool back to ring for RX queue: %d\n",
+                       dev->device_fh, dev->vmdq_rx_q);
+
+               mbuf_destroy_zcp(vpool);
+
+               /* Stop the TX queue. */
+               if (rte_eth_dev_tx_queue_stop(ports[0], dev->vmdq_rx_q) != 0) {
+                       LOG_DEBUG(VHOST_CONFIG,
+                               "(%"PRIu64") In destroy_device: Failed to "
+                               "stop tx queue:%d\n",
+                               dev->device_fh, dev->vmdq_rx_q);
+               }
+
+               vpool = &vpool_array[dev->vmdq_rx_q + MAX_QUEUES];
+
+               LOG_DEBUG(VHOST_CONFIG,
+                       "(%"PRIu64") destroy_device: Start put mbuf in mempool "
+                       "back to ring for TX queue: %d, dev:(%"PRIu64")\n",
+                       dev->device_fh, (dev->vmdq_rx_q + MAX_QUEUES),
+                       dev->device_fh);
+
+               mbuf_destroy_zcp(vpool);
+       }
+
+}
+
+/*
+ * A new device is added to a data core. First the device is added to the main linked list
+ * and the allocated to a specific data core.
+ */
+static int
+new_device (struct virtio_net *dev)
+{
+       struct virtio_net_data_ll *ll_dev;
+       int lcore, core_add = 0;
+       uint32_t device_num_min = num_devices;
+
+       /* Add device to main ll */
+       ll_dev = get_data_ll_free_entry(&ll_root_free);
+       if (ll_dev == NULL) {
+               RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") No free entry found in linked list. Device limit "
+                       "of %d devices per core has been reached\n",
+                       dev->device_fh, num_devices);
+               return -1;
+       }
+       ll_dev->dev = dev;
+       add_data_ll_entry(&ll_root_used, ll_dev);
+       ll_dev->dev->vmdq_rx_q
+               = ll_dev->dev->device_fh * (num_queues / num_devices);
+
+       if (zero_copy) {
+               uint32_t index = ll_dev->dev->vmdq_rx_q;
+               uint32_t count_in_ring, i;
+               struct mbuf_table *tx_q;
+
+               count_in_ring = rte_ring_count(vpool_array[index].ring);
+
+               LOG_DEBUG(VHOST_CONFIG,
+                       "(%"PRIu64") in new_device: mbuf count in mempool "
+                       "before attach is: %d\n",
+                       dev->device_fh,
+                       rte_mempool_count(vpool_array[index].pool));
+               LOG_DEBUG(VHOST_CONFIG,
+                       "(%"PRIu64") in new_device: mbuf count in  ring "
+                       "before attach  is : %d\n",
+                       dev->device_fh, count_in_ring);
+
+               /*
+                * Attach all mbufs in vpool.ring and put back intovpool.pool.
+                */
+               for (i = 0; i < count_in_ring; i++)
+                       attach_rxmbuf_zcp(dev);
+
+               LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") in new_device: mbuf count in "
+                       "mempool after attach is: %d\n",
+                       dev->device_fh,
+                       rte_mempool_count(vpool_array[index].pool));
+               LOG_DEBUG(VHOST_CONFIG, "(%"PRIu64") in new_device: mbuf count in "
+                       "ring after attach  is : %d\n",
+                       dev->device_fh,
+                       rte_ring_count(vpool_array[index].ring));
+
+               tx_q = &tx_queue_zcp[(uint16_t)dev->vmdq_rx_q];
+               tx_q->txq_id = dev->vmdq_rx_q;
+
+               if (rte_eth_dev_tx_queue_start(ports[0], dev->vmdq_rx_q) != 0) {
+                       struct vpool *vpool = &vpool_array[dev->vmdq_rx_q];
+
+                       LOG_DEBUG(VHOST_CONFIG,
+                               "(%"PRIu64") In new_device: Failed to start "
+                               "tx queue:%d\n",
+                               dev->device_fh, dev->vmdq_rx_q);
+
+                       mbuf_destroy_zcp(vpool);
+                       return -1;
+               }
+
+               if (rte_eth_dev_rx_queue_start(ports[0], dev->vmdq_rx_q) != 0) {
+                       struct vpool *vpool = &vpool_array[dev->vmdq_rx_q];
+
+                       LOG_DEBUG(VHOST_CONFIG,
+                               "(%"PRIu64") In new_device: Failed to start "
+                               "rx queue:%d\n",
+                               dev->device_fh, dev->vmdq_rx_q);
+
+                       /* Stop the TX queue. */
+                       if (rte_eth_dev_tx_queue_stop(ports[0],
+                               dev->vmdq_rx_q) != 0) {
+                               LOG_DEBUG(VHOST_CONFIG,
+                                       "(%"PRIu64") In new_device: Failed to "
+                                       "stop tx queue:%d\n",
+                                       dev->device_fh, dev->vmdq_rx_q);
+                       }
+
+                       mbuf_destroy_zcp(vpool);
+                       return -1;
+               }
+
+       }
+
+       /*reset ready flag*/
+       dev->ready = DEVICE_MAC_LEARNING;
+       dev->remove = 0;
+
+       /* Find a suitable lcore to add the device. */
+       RTE_LCORE_FOREACH_SLAVE(lcore) {
+               if (lcore_info[lcore].lcore_ll->device_num < device_num_min) {
+                       device_num_min = lcore_info[lcore].lcore_ll->device_num;
+                       core_add = lcore;
+               }
+       }
+       /* Add device to lcore ll */
+       ll_dev->dev->coreid = core_add;
+       ll_dev = get_data_ll_free_entry(&lcore_info[ll_dev->dev->coreid].lcore_ll->ll_root_free);
+       if (ll_dev == NULL) {
+               RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Failed to add device to data core\n", dev->device_fh);
+               dev->ready = DEVICE_SAFE_REMOVE;
+               destroy_device(dev);
+               return -1;
+       }
+       ll_dev->dev = dev;
+       add_data_ll_entry(&lcore_info[ll_dev->dev->coreid].lcore_ll->ll_root_used, ll_dev);
+
+       /* Initialize device stats */
+       memset(&dev_statistics[dev->device_fh], 0, sizeof(struct device_statistics));
+
+       /* Disable notifications. */
+       set_irq_status(dev);
+       lcore_info[ll_dev->dev->coreid].lcore_ll->device_num++;
+       dev->flags |= VIRTIO_DEV_RUNNING;
+
+       RTE_LOG(INFO, VHOST_DATA, "(%"PRIu64") Device has been added to data core %d\n", dev->device_fh, dev->coreid);
+
+       return 0;
+}
+
+/*
+ * These callback allow devices to be added to the data core when configuration
+ * has been fully complete.
+ */
+static const struct virtio_net_device_ops virtio_net_device_ops =
+{
+       .new_device =  new_device,
+       .destroy_device = destroy_device,
+};
+
+/*
+ * This is a thread will wake up after a period to print stats if the user has
+ * enabled them.
+ */
+static void
+print_stats(void)
+{
+       struct virtio_net_data_ll *dev_ll;
+       uint64_t tx_dropped, rx_dropped;
+       uint64_t tx, tx_total, rx, rx_total;
+       uint32_t device_fh;
+       const char clr[] = { 27, '[', '2', 'J', '\0' };
+       const char top_left[] = { 27, '[', '1', ';', '1', 'H','\0' };
+
+       while(1) {
+               sleep(enable_stats);
+
+               /* Clear screen and move to top left */
+               printf("%s%s", clr, top_left);
+
+               printf("\nDevice statistics ====================================");
+
+               dev_ll = ll_root_used;
+               while (dev_ll != NULL) {
+                       device_fh = (uint32_t)dev_ll->dev->device_fh;
+                       tx_total = dev_statistics[device_fh].tx_total;
+                       tx = dev_statistics[device_fh].tx;
+                       tx_dropped = tx_total - tx;
+                       if (zero_copy == 0) {
+                               rx_total = rte_atomic64_read(
+                                       &dev_statistics[device_fh].rx_total_atomic);
+                               rx = rte_atomic64_read(
+                                       &dev_statistics[device_fh].rx_atomic);
+                       } else {
+                               rx_total = dev_statistics[device_fh].rx_total;
+                               rx = dev_statistics[device_fh].rx;
+                       }
+                       rx_dropped = rx_total - rx;
+
+                       printf("\nStatistics for device %"PRIu32" ------------------------------"
+                                       "\nTX total:            %"PRIu64""
+                                       "\nTX dropped:          %"PRIu64""
+                                       "\nTX successful:               %"PRIu64""
+                                       "\nRX total:            %"PRIu64""
+                                       "\nRX dropped:          %"PRIu64""
+                                       "\nRX successful:               %"PRIu64"",
+                                       device_fh,
+                                       tx_total,
+                                       tx_dropped,
+                                       tx,
+                                       rx_total,
+                                       rx_dropped,
+                                       rx);
+
+                       dev_ll = dev_ll->next;
+               }
+               printf("\n======================================================\n");
+       }
+}
+
+static void
+setup_mempool_tbl(int socket, uint32_t index, char *pool_name,
+       char *ring_name, uint32_t nb_mbuf)
+{
+       uint16_t roomsize = VIRTIO_DESCRIPTOR_LEN_ZCP + RTE_PKTMBUF_HEADROOM;
+       vpool_array[index].pool
+               = rte_mempool_create(pool_name, nb_mbuf, MBUF_SIZE_ZCP,
+               MBUF_CACHE_SIZE_ZCP, sizeof(struct rte_pktmbuf_pool_private),
+               rte_pktmbuf_pool_init, (void *)(uintptr_t)roomsize,
+               rte_pktmbuf_init, NULL, socket, 0);
+       if (vpool_array[index].pool != NULL) {
+               vpool_array[index].ring
+                       = rte_ring_create(ring_name,
+                               rte_align32pow2(nb_mbuf + 1),
+                               socket, RING_F_SP_ENQ | RING_F_SC_DEQ);
+               if (likely(vpool_array[index].ring != NULL)) {
+                       LOG_DEBUG(VHOST_CONFIG,
+                               "in setup_mempool_tbl: mbuf count in "
+                               "mempool is: %d\n",
+                               rte_mempool_count(vpool_array[index].pool));
+                       LOG_DEBUG(VHOST_CONFIG,
+                               "in setup_mempool_tbl: mbuf count in "
+                               "ring   is: %d\n",
+                               rte_ring_count(vpool_array[index].ring));
+               } else {
+                       rte_exit(EXIT_FAILURE, "ring_create(%s) failed",
+                               ring_name);
+               }
+
+               /* Need consider head room. */
+               vpool_array[index].buf_size = roomsize - RTE_PKTMBUF_HEADROOM;
+       } else {
+               rte_exit(EXIT_FAILURE, "mempool_create(%s) failed", pool_name);
+       }
+}
+
+
+/*
+ * Main function, does initialisation and calls the per-lcore functions. The CUSE
+ * device is also registered here to handle the IOCTLs.
+ */
+int
+MAIN(int argc, char *argv[])
+{
+       struct rte_mempool *mbuf_pool = NULL;
+       unsigned lcore_id, core_id = 0;
+       unsigned nb_ports, valid_num_ports;
+       int ret;
+       uint8_t portid, queue_id = 0;
+       static pthread_t tid;
+
+       /* init EAL */
+       ret = rte_eal_init(argc, argv);
+       if (ret < 0)
+               rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
+       argc -= ret;
+       argv += ret;
+
+       /* parse app arguments */
+       ret = us_vhost_parse_args(argc, argv);
+       if (ret < 0)
+               rte_exit(EXIT_FAILURE, "Invalid argument\n");
+
+       for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id ++)
+               if (rte_lcore_is_enabled(lcore_id))
+                       lcore_ids[core_id ++] = lcore_id;
+
+       if (rte_lcore_count() > RTE_MAX_LCORE)
+               rte_exit(EXIT_FAILURE,"Not enough cores\n");
+
+       /*set the number of swithcing cores available*/
+       num_switching_cores = rte_lcore_count()-1;
+
+       /* Get the number of physical ports. */
+       nb_ports = rte_eth_dev_count();
+       if (nb_ports > RTE_MAX_ETHPORTS)
+               nb_ports = RTE_MAX_ETHPORTS;
+
+       /*
+        * Update the global var NUM_PORTS and global array PORTS
+        * and get value of var VALID_NUM_PORTS according to system ports number
+        */
+       valid_num_ports = check_ports_num(nb_ports);
+
+       if ((valid_num_ports ==  0) || (valid_num_ports > MAX_SUP_PORTS)) {
+               RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
+                       "but only %u port can be enabled\n",num_ports, MAX_SUP_PORTS);
+               return -1;
+       }
+
+       if (zero_copy == 0) {
+               /* Create the mbuf pool. */
+               mbuf_pool = rte_mempool_create(
+                               "MBUF_POOL",
+                               NUM_MBUFS_PER_PORT
+                               * valid_num_ports,
+                               MBUF_SIZE, MBUF_CACHE_SIZE,
+                               sizeof(struct rte_pktmbuf_pool_private),
+                               rte_pktmbuf_pool_init, NULL,
+                               rte_pktmbuf_init, NULL,
+                               rte_socket_id(), 0);
+               if (mbuf_pool == NULL)
+                       rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
+
+               for (queue_id = 0; queue_id < MAX_QUEUES + 1; queue_id++)
+                       vpool_array[queue_id].pool = mbuf_pool;
+
+               if (vm2vm_mode == VM2VM_HARDWARE) {
+                       /* Enable VT loop back to let L2 switch to do it. */
+                       vmdq_conf_default.rx_adv_conf.vmdq_rx_conf.enable_loop_back = 1;
+                       LOG_DEBUG(VHOST_CONFIG,
+                               "Enable loop back for L2 switch in vmdq.\n");
+               }
+       } else {
+               uint32_t nb_mbuf;
+               char pool_name[RTE_MEMPOOL_NAMESIZE];
+               char ring_name[RTE_MEMPOOL_NAMESIZE];
+
+               /*
+                * Zero copy defers queue RX/TX start to the time when guest
+                * finishes its startup and packet buffers from that guest are
+                * available.
+                */
+               rx_conf_default.rx_deferred_start = (uint8_t)zero_copy;
+               rx_conf_default.rx_drop_en = 0;
+               tx_conf_default.tx_deferred_start = (uint8_t)zero_copy;
+               nb_mbuf = num_rx_descriptor
+                       + num_switching_cores * MBUF_CACHE_SIZE_ZCP
+                       + num_switching_cores * MAX_PKT_BURST;
+
+               for (queue_id = 0; queue_id < MAX_QUEUES; queue_id++) {
+                       snprintf(pool_name, sizeof(pool_name),
+                               "rxmbuf_pool_%u", queue_id);
+                       snprintf(ring_name, sizeof(ring_name),
+                               "rxmbuf_ring_%u", queue_id);
+                       setup_mempool_tbl(rte_socket_id(), queue_id,
+                               pool_name, ring_name, nb_mbuf);
+               }
+
+               nb_mbuf = num_tx_descriptor
+                               + num_switching_cores * MBUF_CACHE_SIZE_ZCP
+                               + num_switching_cores * MAX_PKT_BURST;
+
+               for (queue_id = 0; queue_id < MAX_QUEUES; queue_id++) {
+                       snprintf(pool_name, sizeof(pool_name),
+                               "txmbuf_pool_%u", queue_id);
+                       snprintf(ring_name, sizeof(ring_name),
+                               "txmbuf_ring_%u", queue_id);
+                       setup_mempool_tbl(rte_socket_id(),
+                               (queue_id + MAX_QUEUES),
+                               pool_name, ring_name, nb_mbuf);
+               }
+
+               if (vm2vm_mode == VM2VM_HARDWARE) {
+                       /* Enable VT loop back to let L2 switch to do it. */
+                       vmdq_conf_default.rx_adv_conf.vmdq_rx_conf.enable_loop_back = 1;
+                       LOG_DEBUG(VHOST_CONFIG,
+                               "Enable loop back for L2 switch in vmdq.\n");
+               }
+       }
+       /* Set log level. */
+       rte_set_log_level(LOG_LEVEL);
+
+       /* initialize all ports */
+       for (portid = 0; portid < nb_ports; portid++) {
+               /* skip ports that are not enabled */
+               if ((enabled_port_mask & (1 << portid)) == 0) {
+                       RTE_LOG(INFO, VHOST_PORT,
+                               "Skipping disabled port %d\n", portid);
+                       continue;
+               }
+               if (port_init(portid) != 0)
+                       rte_exit(EXIT_FAILURE,
+                               "Cannot initialize network ports\n");
+       }
+
+       /* Initialise all linked lists. */
+       if (init_data_ll() == -1)
+               rte_exit(EXIT_FAILURE, "Failed to initialize linked list\n");
+
+       /* Initialize device stats */
+       memset(&dev_statistics, 0, sizeof(dev_statistics));
+
+       /* Enable stats if the user option is set. */
+       if (enable_stats)
+               pthread_create(&tid, NULL, (void*)print_stats, NULL );
+
+       /* Launch all data cores. */
+       if (zero_copy == 0) {
+               RTE_LCORE_FOREACH_SLAVE(lcore_id) {
+                       rte_eal_remote_launch(switch_worker,
+                               mbuf_pool, lcore_id);
+               }
+       } else {
+               uint32_t count_in_mempool, index, i;
+               for (index = 0; index < 2*MAX_QUEUES; index++) {
+                       /* For all RX and TX queues. */
+                       count_in_mempool
+                               = rte_mempool_count(vpool_array[index].pool);
+
+                       /*
+                        * Transfer all un-attached mbufs from vpool.pool
+                        * to vpoo.ring.
+                        */
+                       for (i = 0; i < count_in_mempool; i++) {
+                               struct rte_mbuf *mbuf
+                                       = __rte_mbuf_raw_alloc(
+                                               vpool_array[index].pool);
+                               rte_ring_sp_enqueue(vpool_array[index].ring,
+                                               (void *)mbuf);
+                       }
+
+                       LOG_DEBUG(VHOST_CONFIG,
+                               "in MAIN: mbuf count in mempool at initial "
+                               "is: %d\n", count_in_mempool);
+                       LOG_DEBUG(VHOST_CONFIG,
+                               "in MAIN: mbuf count in  ring at initial  is :"
+                               " %d\n",
+                               rte_ring_count(vpool_array[index].ring));
+               }
+
+               RTE_LCORE_FOREACH_SLAVE(lcore_id)
+                       rte_eal_remote_launch(switch_worker_zcp, NULL,
+                               lcore_id);
+       }
+
+       /* Register CUSE device to handle IOCTLs. */
+       ret = register_cuse_device((char*)&dev_basename, dev_index, get_virtio_net_callbacks());
+       if (ret != 0)
+               rte_exit(EXIT_FAILURE,"CUSE device setup failure.\n");
+
+       init_virtio_net(&virtio_net_device_ops);
+
+       /* Start CUSE session. */
+       start_cuse_session_loop();
+       return 0;
+
+}
+
diff --git a/examples/vhost/main.h b/examples/vhost/main.h
new file mode 100644 (file)
index 0000000..c15d938
--- /dev/null
@@ -0,0 +1,86 @@
+/*-
+ *   BSD LICENSE
+ *
+ *   Copyright(c) 2010-2014 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.
+ */
+
+#ifndef _MAIN_H_
+#define _MAIN_H_
+
+#ifdef RTE_EXEC_ENV_BAREMETAL
+#define MAIN _main
+#else
+#define MAIN main
+#endif
+
+//#define DEBUG
+
+#ifdef DEBUG
+#define LOG_LEVEL RTE_LOG_DEBUG
+#define LOG_DEBUG(log_type, fmt, args...) do { \
+       RTE_LOG(DEBUG, log_type, fmt, ##args);          \
+} while (0)
+#else
+#define LOG_LEVEL RTE_LOG_INFO
+#define LOG_DEBUG(log_type, fmt, args...) do{} while(0)
+#endif
+
+/* Macros for printing using RTE_LOG */
+#define RTE_LOGTYPE_VHOST_CONFIG RTE_LOGTYPE_USER1
+#define RTE_LOGTYPE_VHOST_DATA   RTE_LOGTYPE_USER2
+#define RTE_LOGTYPE_VHOST_PORT   RTE_LOGTYPE_USER3
+
+/*
+ * Device linked list structure for data path.
+ */
+struct virtio_net_data_ll
+{
+       struct virtio_net                       *dev;   /* Pointer to device created by configuration core. */
+       struct virtio_net_data_ll       *next;  /* Pointer to next device in linked list. */
+};
+
+/*
+ * Structure containing data core specific information.
+ */
+struct lcore_ll_info
+{
+       struct virtio_net_data_ll       *ll_root_free;          /* Pointer to head in free linked list. */
+       struct virtio_net_data_ll       *ll_root_used;          /* Pointer to head of used linked list. */
+       uint32_t                                        device_num;                     /* Number of devices on lcore. */
+       volatile uint8_t                        dev_removal_flag;       /* Flag to synchronize device removal. */
+};
+
+struct lcore_info
+{
+       struct lcore_ll_info    *lcore_ll;      /* Pointer to data core specific lcore_ll_info struct */
+};
+
+int MAIN(int argc, char **argv);
+#endif /* _MAIN_H_ */