net: add rte prefix to ether structures
[dpdk.git] / examples / vmdq_dcb / main.c
index 659f46f..75cf31a 100644 (file)
@@ -1,35 +1,5 @@
-/*-
- *   BSD LICENSE
- * 
- *   Copyright(c) 2010-2013 Intel Corporation. All rights reserved.
- *   All rights reserved.
- * 
- *   Redistribution and use in source and binary forms, with or without 
- *   modification, are permitted provided that the following conditions 
- *   are met:
- * 
- *     * Redistributions of source code must retain the above copyright 
- *       notice, this list of conditions and the following disclaimer.
- *     * Redistributions in binary form must reproduce the above copyright 
- *       notice, this list of conditions and the following disclaimer in 
- *       the documentation and/or other materials provided with the 
- *       distribution.
- *     * Neither the name of Intel Corporation nor the names of its 
- *       contributors may be used to endorse or promote products derived 
- *       from this software without specific prior written permission.
- * 
- *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 
- *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 
- *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 
- *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 
- *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 
- *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 
- *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 
- *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 
- *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 
- *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 
- *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
- * 
+/* SPDX-License-Identifier: BSD-3-Clause
+ * Copyright(c) 2010-2014 Intel Corporation
  */
 
 #include <stdint.h>
 #include <rte_log.h>
 #include <rte_memory.h>
 #include <rte_memcpy.h>
-#include <rte_memzone.h>
-#include <rte_tailq.h>
 #include <rte_eal.h>
-#include <rte_per_lcore.h>
 #include <rte_launch.h>
 #include <rte_atomic.h>
 #include <rte_cycles.h>
 #include <rte_per_lcore.h>
 #include <rte_branch_prediction.h>
 #include <rte_interrupts.h>
-#include <rte_pci.h>
 #include <rte_random.h>
 #include <rte_debug.h>
 #include <rte_ether.h>
 #include <rte_ethdev.h>
-#include <rte_ring.h>
-#include <rte_log.h>
 #include <rte_mempool.h>
 #include <rte_mbuf.h>
-#include <rte_memcpy.h>
-
-#include "main.h"
 
 /* basic constants used in application */
-#define SOCKET0 0
-#define SOCKET1 1
-
-#define NUM_QUEUES 128
-
-#define NUM_MBUFS 64*1024
+#define MAX_QUEUES 1024
+/*
+ * 1024 queues require to meet the needs of a large number of vmdq_pools.
+ * (RX/TX_queue_nb * RX/TX_ring_descriptors_nb) per port.
+ */
+#define NUM_MBUFS_PER_PORT (MAX_QUEUES * RTE_MAX(RTE_TEST_RX_DESC_DEFAULT, \
+                                               RTE_TEST_TX_DESC_DEFAULT))
 #define MBUF_CACHE_SIZE 64
-#define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
-
-#define RX_PORT 0
-#define TX_PORT 1
 
-/* number of pools (if user does not specify any, 16 by default */
-static enum rte_eth_nb_pools num_pools = ETH_16_POOLS;
+#define MAX_PKT_BURST 32
 
 /*
- * 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.
+ * Configurable number of RX/TX ring descriptors
  */
-/* Default configuration for rx and tx thresholds etc. */
-static const struct rte_eth_rxconf rx_conf_default = {
-       .rx_thresh = {
-               .pthresh = 8,
-               .hthresh = 8,
-               .wthresh = 4,
-       },
-};
+#define RTE_TEST_RX_DESC_DEFAULT 1024
+#define RTE_TEST_TX_DESC_DEFAULT 1024
 
-/*
- * 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.
- */
-static const struct rte_eth_txconf tx_conf_default = {
-       .tx_thresh = {
-               .pthresh = 36,
-               .hthresh = 0,
-               .wthresh = 0,
-       },
-       .tx_free_thresh = 0, /* Use PMD default values */
-       .tx_rs_thresh = 0, /* Use PMD default values */
-};
+#define INVALID_PORT_ID 0xFF
+
+/* mask of enabled ports */
+static uint32_t enabled_port_mask;
+static uint16_t ports[RTE_MAX_ETHPORTS];
+static unsigned num_ports;
+
+/* number of pools (if user does not specify any, 32 by default */
+static enum rte_eth_nb_pools num_pools = ETH_32_POOLS;
+static enum rte_eth_nb_tcs   num_tcs   = ETH_4_TCS;
+static uint16_t num_queues, num_vmdq_queues;
+static uint16_t vmdq_pool_base, vmdq_queue_base;
+static uint8_t rss_enable;
 
 /* empty vmdq+dcb configuration structure. Filled in programatically */
 static const struct rte_eth_conf vmdq_dcb_conf_default = {
        .rxmode = {
-               .mq_mode        = ETH_VMDQ_DCB,
+               .mq_mode        = ETH_MQ_RX_VMDQ_DCB,
                .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 */
-               .jumbo_frame    = 0, /**< Jumbo Frame Support disabled */
        },
        .txmode = {
-               .mq_mode = ETH_DCB_NONE,
+               .mq_mode = ETH_MQ_TX_VMDQ_DCB,
        },
+       /*
+        * should be overridden separately in code with
+        * appropriate values
+        */
        .rx_adv_conf = {
-               /*
-                * should be overridden separately in code with
-                * appropriate values
-                */
                .vmdq_dcb_conf = {
-                       .nb_queue_pools = ETH_16_POOLS,
+                       .nb_queue_pools = ETH_32_POOLS,
                        .enable_default_pool = 0,
                        .default_pool = 0,
                        .nb_pool_maps = 0,
                        .pool_map = {{0, 0},},
-                       .dcb_queue = {0},
+                       .dcb_tc = {0},
+               },
+               .dcb_rx_conf = {
+                               .nb_tcs = ETH_4_TCS,
+                               /** Traffic class each UP mapped to. */
+                               .dcb_tc = {0},
+               },
+               .vmdq_rx_conf = {
+                       .nb_queue_pools = ETH_32_POOLS,
+                       .enable_default_pool = 0,
+                       .default_pool = 0,
+                       .nb_pool_maps = 0,
+                       .pool_map = {{0, 0},},
+               },
+       },
+       .tx_adv_conf = {
+               .vmdq_dcb_tx_conf = {
+                       .nb_queue_pools = ETH_32_POOLS,
+                       .dcb_tc = {0},
                },
        },
 };
 
 /* array used for printing out statistics */
-volatile unsigned long rxPackets[ NUM_QUEUES ] = {0};
+volatile unsigned long rxPackets[MAX_QUEUES] = {0};
 
 const uint16_t vlan_tags[] = {
        0,  1,  2,  3,  4,  5,  6,  7,
@@ -158,30 +119,64 @@ const uint16_t vlan_tags[] = {
        24, 25, 26, 27, 28, 29, 30, 31
 };
 
+const uint16_t num_vlans = RTE_DIM(vlan_tags);
+/* pool mac addr template, pool mac addr is like: 52 54 00 12 port# pool# */
+static struct rte_ether_addr pool_addr_template = {
+       .addr_bytes = {0x52, 0x54, 0x00, 0x12, 0x00, 0x00}
+};
+
+/* ethernet addresses of ports */
+static struct rte_ether_addr vmdq_ports_eth_addr[RTE_MAX_ETHPORTS];
+
 /* Builds up the correct configuration for vmdq+dcb based on the vlan tags array
  * given above, and the number of traffic classes available for use. */
 static inline int
-get_eth_conf(struct rte_eth_conf *eth_conf, enum rte_eth_nb_pools num_pools)
+get_eth_conf(struct rte_eth_conf *eth_conf)
 {
        struct rte_eth_vmdq_dcb_conf conf;
-       unsigned i;
-
-       if (num_pools != ETH_16_POOLS && num_pools != ETH_32_POOLS ) return -1;
-
-       conf.nb_queue_pools = num_pools;
+       struct rte_eth_vmdq_rx_conf  vmdq_conf;
+       struct rte_eth_dcb_rx_conf   dcb_conf;
+       struct rte_eth_vmdq_dcb_tx_conf tx_conf;
+       uint8_t i;
+
+       conf.nb_queue_pools = (enum rte_eth_nb_pools)num_pools;
+       vmdq_conf.nb_queue_pools = (enum rte_eth_nb_pools)num_pools;
+       tx_conf.nb_queue_pools = (enum rte_eth_nb_pools)num_pools;
+       conf.nb_pool_maps = num_pools;
+       vmdq_conf.nb_pool_maps = num_pools;
        conf.enable_default_pool = 0;
+       vmdq_conf.enable_default_pool = 0;
        conf.default_pool = 0; /* set explicit value, even if not used */
-       conf.nb_pool_maps = sizeof( vlan_tags )/sizeof( vlan_tags[ 0 ]);
-       for (i = 0; i < conf.nb_pool_maps; i++){
-               conf.pool_map[i].vlan_id = vlan_tags[ i ];
-               conf.pool_map[i].pools = 1 << (i % num_pools);
+       vmdq_conf.default_pool = 0;
+
+       for (i = 0; i < conf.nb_pool_maps; i++) {
+               conf.pool_map[i].vlan_id = vlan_tags[i];
+               vmdq_conf.pool_map[i].vlan_id = vlan_tags[i];
+               conf.pool_map[i].pools = 1UL << i;
+               vmdq_conf.pool_map[i].pools = 1UL << i;
        }
        for (i = 0; i < ETH_DCB_NUM_USER_PRIORITIES; i++){
-               conf.dcb_queue[i] = (uint8_t)(i % (NUM_QUEUES/num_pools));
+               conf.dcb_tc[i] = i % num_tcs;
+               dcb_conf.dcb_tc[i] = i % num_tcs;
+               tx_conf.dcb_tc[i] = i % num_tcs;
        }
+       dcb_conf.nb_tcs = (enum rte_eth_nb_tcs)num_tcs;
        (void)(rte_memcpy(eth_conf, &vmdq_dcb_conf_default, sizeof(*eth_conf)));
        (void)(rte_memcpy(&eth_conf->rx_adv_conf.vmdq_dcb_conf, &conf,
-                  sizeof(eth_conf->rx_adv_conf.vmdq_dcb_conf)));
+                         sizeof(conf)));
+       (void)(rte_memcpy(&eth_conf->rx_adv_conf.dcb_rx_conf, &dcb_conf,
+                         sizeof(dcb_conf)));
+       (void)(rte_memcpy(&eth_conf->rx_adv_conf.vmdq_rx_conf, &vmdq_conf,
+                         sizeof(vmdq_conf)));
+       (void)(rte_memcpy(&eth_conf->tx_adv_conf.vmdq_dcb_tx_conf, &tx_conf,
+                         sizeof(tx_conf)));
+       if (rss_enable) {
+               eth_conf->rxmode.mq_mode = ETH_MQ_RX_VMDQ_DCB_RSS;
+               eth_conf->rx_adv_conf.rss_conf.rss_hf = ETH_RSS_IP |
+                                                       ETH_RSS_UDP |
+                                                       ETH_RSS_TCP |
+                                                       ETH_RSS_SCTP;
+       }
        return 0;
 }
 
@@ -190,43 +185,170 @@ get_eth_conf(struct rte_eth_conf *eth_conf, enum rte_eth_nb_pools num_pools)
  * coming from the mbuf_pool passed as parameter
  */
 static inline int
-port_init(uint8_t port, struct rte_mempool *mbuf_pool)
+port_init(uint16_t port, struct rte_mempool *mbuf_pool)
 {
-       struct rte_eth_conf port_conf;
-       const uint16_t rxRings = ETH_VMDQ_DCB_NUM_QUEUES,
-               txRings = (uint16_t)rte_lcore_count();
-       const uint16_t rxRingSize = 128, txRingSize = 512;
+       struct rte_eth_dev_info dev_info;
+       struct rte_eth_conf port_conf = {0};
+       uint16_t rxRingSize = RTE_TEST_RX_DESC_DEFAULT;
+       uint16_t txRingSize = RTE_TEST_TX_DESC_DEFAULT;
        int retval;
        uint16_t q;
+       uint16_t queues_per_pool;
+       uint32_t max_nb_pools;
+       struct rte_eth_txconf txq_conf;
+       uint64_t rss_hf_tmp;
+
+       /*
+        * 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);
+       max_nb_pools = (uint32_t)dev_info.max_vmdq_pools;
+       /*
+        * We allow to process part of VMDQ pools specified by num_pools in
+        * command line.
+        */
+       if (num_pools > max_nb_pools) {
+               printf("num_pools %d >max_nb_pools %d\n",
+                       num_pools, max_nb_pools);
+               return -1;
+       }
+
+       /*
+        * NIC queues are divided into pf queues and vmdq queues.
+        * There is assumption here all ports have the same configuration!
+       */
+       vmdq_queue_base = dev_info.vmdq_queue_base;
+       vmdq_pool_base  = dev_info.vmdq_pool_base;
+       printf("vmdq queue base: %d pool base %d\n",
+               vmdq_queue_base, vmdq_pool_base);
+       if (vmdq_pool_base == 0) {
+               num_vmdq_queues = dev_info.max_rx_queues;
+               num_queues = dev_info.max_rx_queues;
+               if (num_tcs != num_vmdq_queues / num_pools) {
+                       printf("nb_tcs %d is invalid considering with"
+                               " nb_pools %d, nb_tcs * nb_pools should = %d\n",
+                               num_tcs, num_pools, num_vmdq_queues);
+                       return -1;
+               }
+       } else {
+               queues_per_pool = dev_info.vmdq_queue_num /
+                                 dev_info.max_vmdq_pools;
+               if (num_tcs > queues_per_pool) {
+                       printf("num_tcs %d > num of queues per pool %d\n",
+                               num_tcs, queues_per_pool);
+                       return -1;
+               }
+               num_vmdq_queues = num_pools * queues_per_pool;
+               num_queues = vmdq_queue_base + num_vmdq_queues;
+               printf("Configured vmdq pool num: %u,"
+                       " each vmdq pool has %u queues\n",
+                       num_pools, queues_per_pool);
+       }
+
+       if (!rte_eth_dev_is_valid_port(port))
+               return -1;
 
-       retval = get_eth_conf(&port_conf, num_pools);
+       retval = get_eth_conf(&port_conf);
        if (retval < 0)
                return retval;
 
-       if (port >= rte_eth_dev_count()) return -1;
+       rte_eth_dev_info_get(port, &dev_info);
+       if (dev_info.tx_offload_capa & DEV_TX_OFFLOAD_MBUF_FAST_FREE)
+               port_conf.txmode.offloads |=
+                       DEV_TX_OFFLOAD_MBUF_FAST_FREE;
+
+       rss_hf_tmp = port_conf.rx_adv_conf.rss_conf.rss_hf;
+       port_conf.rx_adv_conf.rss_conf.rss_hf &=
+               dev_info.flow_type_rss_offloads;
+       if (port_conf.rx_adv_conf.rss_conf.rss_hf != rss_hf_tmp) {
+               printf("Port %u modified RSS hash function based on hardware support,"
+                       "requested:%#"PRIx64" configured:%#"PRIx64"\n",
+                       port,
+                       rss_hf_tmp,
+                       port_conf.rx_adv_conf.rss_conf.rss_hf);
+       }
+
+       /*
+        * Though in this example, all queues including pf queues are setup.
+        * This is because VMDQ queues doesn't always start from zero, and the
+        * PMD layer doesn't support selectively initialising part of rx/tx
+        * queues.
+        */
+       retval = rte_eth_dev_configure(port, num_queues, num_queues, &port_conf);
+       if (retval != 0)
+               return retval;
 
-       retval = rte_eth_dev_configure(port, rxRings, txRings, &port_conf);
+       retval = rte_eth_dev_adjust_nb_rx_tx_desc(port, &rxRingSize,
+                               &txRingSize);
        if (retval != 0)
                return retval;
+       if (RTE_MAX(rxRingSize, txRingSize) >
+           RTE_MAX(RTE_TEST_RX_DESC_DEFAULT, RTE_TEST_TX_DESC_DEFAULT)) {
+               printf("Mbuf pool has an insufficient size for port %u.\n",
+                       port);
+               return -1;
+       }
 
-       for (q = 0; q < rxRings; q ++) {
+       for (q = 0; q < num_queues; q++) {
                retval = rte_eth_rx_queue_setup(port, q, rxRingSize,
-                                               SOCKET0, &rx_conf_default,
-                                               mbuf_pool);
-               if (retval < 0)
+                                       rte_eth_dev_socket_id(port),
+                                       NULL,
+                                       mbuf_pool);
+               if (retval < 0) {
+                       printf("initialize rx queue %d failed\n", q);
                        return retval;
+               }
        }
 
-       for (q = 0; q < txRings; q ++) {
+       txq_conf = dev_info.default_txconf;
+       txq_conf.offloads = port_conf.txmode.offloads;
+       for (q = 0; q < num_queues; q++) {
                retval = rte_eth_tx_queue_setup(port, q, txRingSize,
-                                               SOCKET0, &tx_conf_default);
-               if (retval < 0)
+                                       rte_eth_dev_socket_id(port),
+                                       &txq_conf);
+               if (retval < 0) {
+                       printf("initialize tx queue %d failed\n", q);
                        return retval;
+               }
        }
 
        retval  = rte_eth_dev_start(port);
-       if (retval < 0)
+       if (retval < 0) {
+               printf("port %d start failed\n", port);
                return retval;
+       }
+
+       rte_eth_macaddr_get(port, &vmdq_ports_eth_addr[port]);
+       printf("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]);
+
+       /* Set mac for each pool.*/
+       for (q = 0; q < num_pools; q++) {
+               struct rte_ether_addr mac;
+
+               mac = pool_addr_template;
+               mac.addr_bytes[4] = port;
+               mac.addr_bytes[5] = q;
+               printf("Port %u vmdq pool %u set mac %02x:%02x:%02x:%02x:%02x:%02x\n",
+                       port, q,
+                       mac.addr_bytes[0], mac.addr_bytes[1],
+                       mac.addr_bytes[2], mac.addr_bytes[3],
+                       mac.addr_bytes[4], mac.addr_bytes[5]);
+               retval = rte_eth_dev_mac_addr_add(port, &mac,
+                               q + vmdq_pool_base);
+               if (retval) {
+                       printf("mac addr add failed at pool %d\n", q);
+                       return retval;
+               }
+       }
 
        return 0;
 }
@@ -252,12 +374,53 @@ vmdq_parse_num_pools(const char *q_arg)
        return 0;
 }
 
+/* Check num_tcs parameter and set it if OK*/
+static int
+vmdq_parse_num_tcs(const char *q_arg)
+{
+       char *end = NULL;
+       int n;
+
+       /* parse number string */
+       n = strtol(q_arg, &end, 10);
+       if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
+               return -1;
+
+       if (n != 4 && n != 8)
+               return -1;
+       if (n == 4)
+               num_tcs = ETH_4_TCS;
+       else
+               num_tcs = ETH_8_TCS;
+
+       return 0;
+}
+
+static int
+parse_portmask(const char *portmask)
+{
+       char *end = NULL;
+       unsigned long pm;
+
+       /* parse hexadecimal string */
+       pm = strtoul(portmask, &end, 16);
+       if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
+               return -1;
+
+       if (pm == 0)
+               return -1;
+
+       return pm;
+}
+
 /* Display usage */
 static void
 vmdq_usage(const char *prgname)
 {
-       printf("%s [EAL options] -- [-nb-pools NP]\n"
-              "  -nb-pools NP: number of pools (16 default, 32)\n",
+       printf("%s [EAL options] -- -p PORTMASK]\n"
+       "  --nb-pools NP: number of pools (32 default, 16)\n"
+       "  --nb-tcs NP: number of TCs (4 default, 8)\n"
+       "  --enable-rss: enable RSS (disabled by default)\n",
               prgname);
 }
 
@@ -267,99 +430,210 @@ vmdq_parse_args(int argc, char **argv)
 {
        int opt;
        int option_index;
+       unsigned i;
        const char *prgname = argv[0];
-
        static struct option long_option[] = {
                {"nb-pools", required_argument, NULL, 0},
+               {"nb-tcs", required_argument, NULL, 0},
+               {"enable-rss", 0, NULL, 0},
                {NULL, 0, 0, 0}
        };
 
        /* Parse command line */
-       while ((opt = getopt_long_only(argc, argv, "",long_option,&option_index)) != EOF) {
+       while ((opt = getopt_long(argc, argv, "p:", long_option,
+               &option_index)) != EOF) {
                switch (opt) {
-               case 0:
-                       if (vmdq_parse_num_pools(optarg) == -1){
-                               printf("invalid number of pools\n");
+               /* portmask */
+               case 'p':
+                       enabled_port_mask = parse_portmask(optarg);
+                       if (enabled_port_mask == 0) {
+                               printf("invalid portmask\n");
                                vmdq_usage(prgname);
                                return -1;
                        }
                        break;
+               case 0:
+                       if (!strcmp(long_option[option_index].name, "nb-pools")) {
+                               if (vmdq_parse_num_pools(optarg) == -1) {
+                                       printf("invalid number of pools\n");
+                                       return -1;
+                               }
+                       }
+
+                       if (!strcmp(long_option[option_index].name, "nb-tcs")) {
+                               if (vmdq_parse_num_tcs(optarg) == -1) {
+                                       printf("invalid number of tcs\n");
+                                       return -1;
+                               }
+                       }
+
+                       if (!strcmp(long_option[option_index].name, "enable-rss"))
+                               rss_enable = 1;
+                       break;
+
                default:
                        vmdq_usage(prgname);
                        return -1;
                }
        }
-       return 0;
 
+       for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
+               if (enabled_port_mask & (1 << i))
+                       ports[num_ports++] = (uint8_t)i;
+       }
+
+       if (num_ports < 2 || num_ports % 2) {
+               printf("Current enabled port number is %u,"
+                       " but it should be even and at least 2\n", num_ports);
+               return -1;
+       }
 
+       return 0;
 }
 
+static void
+update_mac_address(struct rte_mbuf *m, unsigned dst_port)
+{
+       struct rte_ether_hdr *eth;
+       void *tmp;
+
+       eth = rte_pktmbuf_mtod(m, struct rte_ether_hdr *);
+
+       /* 02:00:00:00:00:xx */
+       tmp = &eth->d_addr.addr_bytes[0];
+       *((uint64_t *)tmp) = 0x000000000002 + ((uint64_t)dst_port << 40);
+
+       /* src addr */
+       ether_addr_copy(&vmdq_ports_eth_addr[dst_port], &eth->s_addr);
+}
 
-#ifndef RTE_EXEC_ENV_BAREMETAL
 /* When we receive a HUP signal, print out our stats */
 static void
 sighup_handler(int signum)
 {
-       unsigned q;
-       for (q = 0; q < NUM_QUEUES; q ++) {
-               if (q % (NUM_QUEUES/num_pools) == 0)
-                       printf("\nPool %u: ", q/(NUM_QUEUES/num_pools));
-               printf("%lu ", rxPackets[ q ]);
+       unsigned q = vmdq_queue_base;
+
+       for (; q < num_queues; q++) {
+               if (q % (num_vmdq_queues / num_pools) == 0)
+                       printf("\nPool %u: ", (q - vmdq_queue_base) /
+                                             (num_vmdq_queues / num_pools));
+               printf("%lu ", rxPackets[q]);
        }
        printf("\nFinished handling signal %d\n", signum);
 }
-#endif
 
 /*
  * Main thread that does the work, reading from INPUT_PORT
  * and writing to OUTPUT_PORT
  */
-static  __attribute__((noreturn)) int
+static int
 lcore_main(void *arg)
 {
        const uintptr_t core_num = (uintptr_t)arg;
        const unsigned num_cores = rte_lcore_count();
-       uint16_t startQueue = (uint16_t)(core_num * (NUM_QUEUES/num_cores));
-       uint16_t endQueue = (uint16_t)(startQueue + (NUM_QUEUES/num_cores));
-       uint16_t q, i;
+       uint16_t startQueue, endQueue;
+       uint16_t q, i, p;
+       const uint16_t quot = (uint16_t)(num_vmdq_queues / num_cores);
+       const uint16_t remainder = (uint16_t)(num_vmdq_queues % num_cores);
+
+
+       if (remainder) {
+               if (core_num < remainder) {
+                       startQueue = (uint16_t)(core_num * (quot + 1));
+                       endQueue = (uint16_t)(startQueue + quot + 1);
+               } else {
+                       startQueue = (uint16_t)(core_num * quot + remainder);
+                       endQueue = (uint16_t)(startQueue + quot);
+               }
+       } else {
+               startQueue = (uint16_t)(core_num * quot);
+               endQueue = (uint16_t)(startQueue + quot);
+       }
 
+       /* vmdq queue idx doesn't always start from zero.*/
+       startQueue += vmdq_queue_base;
+       endQueue   += vmdq_queue_base;
        printf("Core %u(lcore %u) reading queues %i-%i\n", (unsigned)core_num,
               rte_lcore_id(), startQueue, endQueue - 1);
 
+       if (startQueue == endQueue) {
+               printf("lcore %u has nothing to do\n", (unsigned)core_num);
+               return 0;
+       }
+
        for (;;) {
-               struct rte_mbuf *buf[32];
+               struct rte_mbuf *buf[MAX_PKT_BURST];
                const uint16_t buf_size = sizeof(buf) / sizeof(buf[0]);
+               for (p = 0; p < num_ports; p++) {
+                       const uint8_t src = ports[p];
+                       const uint8_t dst = ports[p ^ 1]; /* 0 <-> 1, 2 <-> 3 etc */
 
-               for (q = startQueue; q < endQueue; q++) {
-                       const uint16_t rxCount = rte_eth_rx_burst(RX_PORT,
-                                       q, buf, buf_size);
-                       if (rxCount == 0)
+                       if ((src == INVALID_PORT_ID) || (dst == INVALID_PORT_ID))
                                continue;
-                       rxPackets[q] += rxCount;
 
-                       const uint16_t txCount = rte_eth_tx_burst(TX_PORT,
-                                       (uint16_t)core_num, buf, rxCount);
-                       if (txCount != rxCount) {
-                               for (i = txCount; i < rxCount; i++)
-                                       rte_pktmbuf_free(buf[i]);
+                       for (q = startQueue; q < endQueue; q++) {
+                               const uint16_t rxCount = rte_eth_rx_burst(src,
+                                       q, buf, buf_size);
+
+                               if (unlikely(rxCount == 0))
+                                       continue;
+
+                               rxPackets[q] += rxCount;
+
+                               for (i = 0; i < rxCount; i++)
+                                       update_mac_address(buf[i], dst);
+
+                               const uint16_t txCount = rte_eth_tx_burst(dst,
+                                       q, buf, rxCount);
+                               if (txCount != rxCount) {
+                                       for (i = txCount; i < rxCount; i++)
+                                               rte_pktmbuf_free(buf[i]);
+                               }
                        }
                }
        }
 }
 
+/*
+ * 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) {
+               printf("\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 (!rte_eth_dev_is_valid_port(ports[portid])) {
+                       printf("\nSpecified port ID(%u) is not valid\n",
+                               ports[portid]);
+                       ports[portid] = INVALID_PORT_ID;
+                       valid_num_ports--;
+               }
+       }
+       return valid_num_ports;
+}
+
+
 /* Main function, does initialisation and calls the per-lcore functions */
 int
-MAIN(int argc, char *argv[])
+main(int argc, char *argv[])
 {
        unsigned cores;
        struct rte_mempool *mbuf_pool;
        unsigned lcore_id;
        uintptr_t i;
        int ret;
+       unsigned nb_ports, valid_num_ports;
+       uint16_t portid;
 
-#ifndef RTE_EXEC_ENV_BAREMETAL
        signal(SIGHUP, sighup_handler);
-#endif
 
        /* init EAL */
        ret = rte_eal_init(argc, argv);
@@ -373,28 +647,41 @@ MAIN(int argc, char *argv[])
        if (ret < 0)
                rte_exit(EXIT_FAILURE, "Invalid VMDQ argument\n");
 
-       if (rte_ixgbe_pmd_init() != 0 ||
-                       rte_eal_pci_probe() != 0)
-               rte_exit(EXIT_FAILURE, "Error with NIC driver initialization\n");
-
        cores = rte_lcore_count();
-       if ((cores & (cores - 1)) != 0 || cores > 16) {
-               rte_exit(EXIT_FAILURE,
-                        "This program can only run on 2,4,8 or 16 cores\n\n");
+       if ((cores & (cores - 1)) != 0 || cores > RTE_MAX_LCORE) {
+               rte_exit(EXIT_FAILURE,"This program can only run on an even"
+                               " number of cores(1-%d)\n\n", RTE_MAX_LCORE);
+       }
+
+       nb_ports = rte_eth_dev_count_avail();
+
+       /*
+        * 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 < 2 || valid_num_ports % 2) {
+               printf("Current valid ports number is %u\n", valid_num_ports);
+               rte_exit(EXIT_FAILURE, "Error with valid ports number is not even or less than 2\n");
        }
 
-       mbuf_pool = rte_mempool_create("MBUF_POOL", NUM_MBUFS,
-                                      MBUF_SIZE, MBUF_CACHE_SIZE,
-                                      sizeof(struct rte_pktmbuf_pool_private),
-                                      rte_pktmbuf_pool_init, NULL,
-                                      rte_pktmbuf_init, NULL,
-                                      SOCKET0, 0);
+       mbuf_pool = rte_pktmbuf_pool_create("MBUF_POOL",
+               NUM_MBUFS_PER_PORT * nb_ports, MBUF_CACHE_SIZE,
+               0, RTE_MBUF_DEFAULT_BUF_SIZE, rte_socket_id());
        if (mbuf_pool == NULL)
                rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
 
-       if (port_init(RX_PORT, mbuf_pool) != 0 ||
-           port_init(TX_PORT, mbuf_pool) != 0)
-               rte_exit(EXIT_FAILURE, "Cannot initialize network ports\n");
+       /* initialize all ports */
+       RTE_ETH_FOREACH_DEV(portid) {
+               /* skip ports that are not enabled */
+               if ((enabled_port_mask & (1 << portid)) == 0) {
+                       printf("\nSkipping disabled port %d\n", portid);
+                       continue;
+               }
+               if (port_init(portid, mbuf_pool) != 0)
+                       rte_exit(EXIT_FAILURE, "Cannot initialize network ports\n");
+       }
 
        /* call lcore_main() on every slave lcore */
        i = 0;