examples/tep_termination/main.c

   1 /*-
   2  *   BSD LICENSE
   3  *
   4  *   Copyright(c) 2010-2015 Intel Corporation. All rights reserved.
   5  *   All rights reserved.
   6  *
   7  *   Redistribution and use in source and binary forms, with or without
   8  *   modification, are permitted provided that the following conditions
   9  *   are met:
  10  *
  11  *     * Redistributions of source code must retain the above copyright
  12  *       notice, this list of conditions and the following disclaimer.
  13  *     * Redistributions in binary form must reproduce the above copyright
  14  *       notice, this list of conditions and the following disclaimer in
  15  *       the documentation and/or other materials provided with the
  16  *       distribution.
  17  *     * Neither the name of Intel Corporation nor the names of its
  18  *       contributors may be used to endorse or promote products derived
  19  *       from this software without specific prior written permission.
  20  *
  21  *   THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
  22  *   "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
  23  *   LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
  24  *   A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT
  25  *   OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  26  *   SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT
  27  *   LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
  28  *   DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
  29  *   THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT
  30  *   (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
  31  *   OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
  32  */
  33
  34 #include <arpa/inet.h>
  35 #include <getopt.h>
  36 #include <linux/if_ether.h>
  37 #include <linux/if_vlan.h>
  38 #include <linux/virtio_net.h>
  39 #include <linux/virtio_ring.h>
  40 #include <signal.h>
  41 #include <stdint.h>
  42 #include <sys/eventfd.h>
  43 #include <sys/param.h>
  44 #include <unistd.h>
  45
  46 #include <rte_atomic.h>
  47 #include <rte_cycles.h>
  48 #include <rte_ethdev.h>
  49 #include <rte_log.h>
  50 #include <rte_string_fns.h>
  51 #include <rte_malloc.h>
  52 #include <rte_virtio_net.h>
  53
  54 #include "main.h"
  55 #include "vxlan.h"
  56 #include "vxlan_setup.h"
  57
  58 /* the maximum number of external ports supported */
  59 #define MAX_SUP_PORTS 1
  60
  61 /**
  62  * Calculate the number of buffers needed per port
  63  */
  64 #define NUM_MBUFS_PER_PORT ((MAX_QUEUES * RTE_TEST_RX_DESC_DEFAULT) +\
  65                                 (nb_switching_cores * MAX_PKT_BURST) +\
  66                                 (nb_switching_cores * \
  67                                 RTE_TEST_TX_DESC_DEFAULT) +\
  68                                 (nb_switching_cores * MBUF_CACHE_SIZE))
  69
  70 #define MBUF_CACHE_SIZE 128
  71 #define MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
  72
  73 #define MAX_PKT_BURST 32        /* Max burst size for RX/TX */
  74 #define BURST_TX_DRAIN_US 100   /* TX drain every ~100us */
  75
  76 /* Defines how long we wait between retries on RX */
  77 #define BURST_RX_WAIT_US 15
  78
  79 #define BURST_RX_RETRIES 4      /* Number of retries on RX. */
  80
  81 #define JUMBO_FRAME_MAX_SIZE    0x2600
  82
  83 /* State of virtio device. */
  84 #define DEVICE_MAC_LEARNING 0
  85 #define DEVICE_RX           1
  86 #define DEVICE_SAFE_REMOVE  2
  87
  88 /* Config_core_flag status definitions. */
  89 #define REQUEST_DEV_REMOVAL 1
  90 #define ACK_DEV_REMOVAL     0
  91
  92 /* Configurable number of RX/TX ring descriptors */
  93 #define RTE_TEST_RX_DESC_DEFAULT 1024
  94 #define RTE_TEST_TX_DESC_DEFAULT 512
  95
  96 /* Get first 4 bytes in mbuf headroom. */
  97 #define MBUF_HEADROOM_UINT32(mbuf) (*(uint32_t *)((uint8_t *)(mbuf) \
  98                 + sizeof(struct rte_mbuf)))
  99
 100 #define INVALID_PORT_ID 0xFF
 101
 102 /* Size of buffers used for snprintfs. */
 103 #define MAX_PRINT_BUFF 6072
 104
 105 /* Maximum character device basename size. */
 106 #define MAX_BASENAME_SZ 20
 107
 108 /* Maximum long option length for option parsing. */
 109 #define MAX_LONG_OPT_SZ 64
 110
 111 /* Used to compare MAC addresses. */
 112 #define MAC_ADDR_CMP 0xFFFFFFFFFFFFULL
 113
 114 #define CMD_LINE_OPT_NB_DEVICES "nb-devices"
 115 #define CMD_LINE_OPT_UDP_PORT "udp-port"
 116 #define CMD_LINE_OPT_TX_CHECKSUM "tx-checksum"
 117 #define CMD_LINE_OPT_TSO_SEGSZ "tso-segsz"
 118 #define CMD_LINE_OPT_FILTER_TYPE "filter-type"
 119 #define CMD_LINE_OPT_ENCAP "encap"
 120 #define CMD_LINE_OPT_DECAP "decap"
 121 #define CMD_LINE_OPT_RX_RETRY "rx-retry"
 122 #define CMD_LINE_OPT_RX_RETRY_DELAY "rx-retry-delay"
 123 #define CMD_LINE_OPT_RX_RETRY_NUM "rx-retry-num"
 124 #define CMD_LINE_OPT_STATS "stats"
 125 #define CMD_LINE_OPT_DEV_BASENAME "dev-basename"
 126
 127 /* mask of enabled ports */
 128 static uint32_t enabled_port_mask;
 129
 130 /*Number of switching cores enabled*/
 131 static uint32_t nb_switching_cores;
 132
 133 /* number of devices/queues to support*/
 134 uint16_t nb_devices = 2;
 135
 136 /* max ring descriptor, ixgbe, i40e, e1000 all are 4096. */
 137 #define MAX_RING_DESC 4096
 138
 139 struct vpool {
 140         struct rte_mempool *pool;
 141         struct rte_ring *ring;
 142         uint32_t buf_size;
 143 } vpool_array[MAX_QUEUES+MAX_QUEUES];
 144
 145 /* UDP tunneling port */
 146 uint16_t udp_port = 4789;
 147
 148 /* enable/disable inner TX checksum */
 149 uint8_t tx_checksum = 0;
 150
 151 /* TCP segment size */
 152 uint16_t tso_segsz = 0;
 153
 154 /* enable/disable decapsulation */
 155 uint8_t rx_decap = 1;
 156
 157 /* enable/disable encapsulation */
 158 uint8_t tx_encap = 1;
 159
 160 /* RX filter type for tunneling packet */
 161 uint8_t filter_idx = 1;
 162
 163 /* overlay packet operation */
 164 struct ol_switch_ops overlay_options = {
 165         .port_configure = vxlan_port_init,
 166         .tunnel_setup = vxlan_link,
 167         .tunnel_destroy = vxlan_unlink,
 168         .tx_handle = vxlan_tx_pkts,
 169         .rx_handle = vxlan_rx_pkts,
 170         .param_handle = NULL,
 171 };
 172
 173 /* Enable stats. */
 174 uint32_t enable_stats = 0;
 175 /* Enable retries on RX. */
 176 static uint32_t enable_retry = 1;
 177 /* Specify timeout (in useconds) between retries on RX. */
 178 static uint32_t burst_rx_delay_time = BURST_RX_WAIT_US;
 179 /* Specify the number of retries on RX. */
 180 static uint32_t burst_rx_retry_num = BURST_RX_RETRIES;
 181
 182 /* Character device basename. Can be set by user. */
 183 static char dev_basename[MAX_BASENAME_SZ] = "vhost-net";
 184
 185 static unsigned lcore_ids[RTE_MAX_LCORE];
 186 uint8_t ports[RTE_MAX_ETHPORTS];
 187
 188 static unsigned nb_ports; /**< The number of ports specified in command line */
 189
 190 /* ethernet addresses of ports */
 191 struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
 192
 193 /* heads for the main used and free linked lists for the data path. */
 194 static struct virtio_net_data_ll *ll_root_used;
 195 static struct virtio_net_data_ll *ll_root_free;
 196
 197 /**
 198  * Array of data core structures containing information on
 199  * individual core linked lists.
 200  */
 201 static struct lcore_info lcore_info[RTE_MAX_LCORE];
 202
 203 /* Used for queueing bursts of TX packets. */
 204 struct mbuf_table {
 205         unsigned len;
 206         unsigned txq_id;
 207         struct rte_mbuf *m_table[MAX_PKT_BURST];
 208 };
 209
 210 /* TX queue for each data core. */
 211 struct mbuf_table lcore_tx_queue[RTE_MAX_LCORE];
 212
 213 struct device_statistics dev_statistics[MAX_DEVICES];
 214
 215 /**
 216  * Set character device basename.
 217  */
 218 static int
 219 us_vhost_parse_basename(const char *q_arg)
 220 {
 221         /* parse number string */
 222         if (strlen(q_arg) >= MAX_BASENAME_SZ)
 223                 return -1;
 224         else
 225                 snprintf((char *)&dev_basename, MAX_BASENAME_SZ, "%s", q_arg);
 226
 227         return 0;
 228 }
 229
 230 /**
 231  * Parse the portmask provided at run time.
 232  */
 233 static int
 234 parse_portmask(const char *portmask)
 235 {
 236         char *end = NULL;
 237         unsigned long pm;
 238
 239         /* parse hexadecimal string */
 240         pm = strtoul(portmask, &end, 16);
 241         if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
 242                 return -1;
 243
 244         if (pm == 0)
 245                 return -1;
 246
 247         return pm;
 248 }
 249
 250 /**
 251  * Parse num options at run time.
 252  */
 253 static int
 254 parse_num_opt(const char *q_arg, uint32_t max_valid_value)
 255 {
 256         char *end = NULL;
 257         unsigned long num;
 258
 259         /* parse unsigned int string */
 260         num = strtoul(q_arg, &end, 10);
 261         if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
 262                 return -1;
 263
 264         if (num > max_valid_value)
 265                 return -1;
 266
 267         return num;
 268 }
 269
 270 /**
 271  * Display usage
 272  */
 273 static void
 274 tep_termination_usage(const char *prgname)
 275 {
 276         RTE_LOG(INFO, VHOST_CONFIG, "%s [EAL options] -- -p PORTMASK\n"
 277         "               --udp-port: UDP destination port for VXLAN packet\n"
 278         "               --nb-devices[1-64]: The number of virtIO device\n"
 279         "               --tx-checksum [0|1]: inner Tx checksum offload\n"
 280         "               --tso-segsz [0-N]: TCP segment size\n"
 281         "               --decap [0|1]: tunneling packet decapsulation\n"
 282         "               --encap [0|1]: tunneling packet encapsulation\n"
 283         "               --filter-type[1-3]: filter type for tunneling packet\n"
 284         "                   1: Inner MAC and tenent ID\n"
 285         "                   2: Inner MAC and VLAN, and tenent ID\n"
 286         "                   3: Outer MAC, Inner MAC and tenent ID\n"
 287         "               -p PORTMASK: Set mask for ports to be used by application\n"
 288         "               --rx-retry [0|1]: disable/enable(default) retries on rx."
 289         "                Enable retry if destintation queue is full\n"
 290         "               --rx-retry-delay [0-N]: timeout(in usecond) between retries on RX."
 291         "                This makes effect only if retries on rx enabled\n"
 292         "               --rx-retry-num [0-N]: the number of retries on rx."
 293         "                This makes effect only if retries on rx enabled\n"
 294         "               --stats [0-N]: 0: Disable stats, N: Time in seconds to print stats\n"
 295         "               --dev-basename: The basename to be used for the character device.\n",
 296                prgname);
 297 }
 298
 299 /**
 300  * Parse the arguments given in the command line of the application.
 301  */
 302 static int
 303 tep_termination_parse_args(int argc, char **argv)
 304 {
 305         int opt, ret;
 306         int option_index;
 307         unsigned i;
 308         const char *prgname = argv[0];
 309         static struct option long_option[] = {
 310                 {CMD_LINE_OPT_NB_DEVICES, required_argument, NULL, 0},
 311                 {CMD_LINE_OPT_UDP_PORT, required_argument, NULL, 0},
 312                 {CMD_LINE_OPT_TX_CHECKSUM, required_argument, NULL, 0},
 313                 {CMD_LINE_OPT_TSO_SEGSZ, required_argument, NULL, 0},
 314                 {CMD_LINE_OPT_DECAP, required_argument, NULL, 0},
 315                 {CMD_LINE_OPT_ENCAP, required_argument, NULL, 0},
 316                 {CMD_LINE_OPT_FILTER_TYPE, required_argument, NULL, 0},
 317                 {CMD_LINE_OPT_RX_RETRY, required_argument, NULL, 0},
 318                 {CMD_LINE_OPT_RX_RETRY_DELAY, required_argument, NULL, 0},
 319                 {CMD_LINE_OPT_RX_RETRY_NUM, required_argument, NULL, 0},
 320                 {CMD_LINE_OPT_STATS, required_argument, NULL, 0},
 321                 {CMD_LINE_OPT_DEV_BASENAME, required_argument, NULL, 0},
 322                 {NULL, 0, 0, 0},
 323         };
 324
 325         /* Parse command line */
 326         while ((opt = getopt_long(argc, argv, "p:",
 327                         long_option, &option_index)) != EOF) {
 328                 switch (opt) {
 329                 /* Portmask */
 330                 case 'p':
 331                         enabled_port_mask = parse_portmask(optarg);
 332                         if (enabled_port_mask == 0) {
 333                                 RTE_LOG(INFO, VHOST_CONFIG,
 334                                         "Invalid portmask\n");
 335                                 tep_termination_usage(prgname);
 336                                 return -1;
 337                         }
 338                         break;
 339                 case 0:
 340                         if (!strncmp(long_option[option_index].name,
 341                                 CMD_LINE_OPT_NB_DEVICES,
 342                                 sizeof(CMD_LINE_OPT_NB_DEVICES))) {
 343                                 ret = parse_num_opt(optarg, MAX_DEVICES);
 344                                 if (ret == -1) {
 345                                         RTE_LOG(INFO, VHOST_CONFIG,
 346                                         "Invalid argument for nb-devices [0-%d]\n",
 347                                         MAX_DEVICES);
 348                                         tep_termination_usage(prgname);
 349                                         return -1;
 350                                 } else
 351                                         nb_devices = ret;
 352                         }
 353
 354                         /* Enable/disable retries on RX. */
 355                         if (!strncmp(long_option[option_index].name,
 356                                 CMD_LINE_OPT_RX_RETRY,
 357                                 sizeof(CMD_LINE_OPT_RX_RETRY))) {
 358                                 ret = parse_num_opt(optarg, 1);
 359                                 if (ret == -1) {
 360                                         RTE_LOG(INFO, VHOST_CONFIG,
 361                                                 "Invalid argument for rx-retry [0|1]\n");
 362                                         tep_termination_usage(prgname);
 363                                         return -1;
 364                                 } else
 365                                         enable_retry = ret;
 366                         }
 367
 368                         if (!strncmp(long_option[option_index].name,
 369                                 CMD_LINE_OPT_TSO_SEGSZ,
 370                                 sizeof(CMD_LINE_OPT_TSO_SEGSZ))) {
 371                                 ret = parse_num_opt(optarg, INT16_MAX);
 372                                 if (ret == -1) {
 373                                         RTE_LOG(INFO, VHOST_CONFIG,
 374                                                 "Invalid argument for TCP segment size [0-N]\n");
 375                                         tep_termination_usage(prgname);
 376                                         return -1;
 377                                 } else
 378                                         tso_segsz = ret;
 379                         }
 380
 381                         if (!strncmp(long_option[option_index].name,
 382                                         CMD_LINE_OPT_UDP_PORT,
 383                                         sizeof(CMD_LINE_OPT_UDP_PORT))) {
 384                                 ret = parse_num_opt(optarg, INT16_MAX);
 385                                 if (ret == -1) {
 386                                         RTE_LOG(INFO, VHOST_CONFIG,
 387                                                 "Invalid argument for UDP port [0-N]\n");
 388                                         tep_termination_usage(prgname);
 389                                         return -1;
 390                                 } else
 391                                         udp_port = ret;
 392                         }
 393
 394                         /* Specify the retries delay time (in useconds) on RX.*/
 395                         if (!strncmp(long_option[option_index].name,
 396                                 CMD_LINE_OPT_RX_RETRY_DELAY,
 397                                 sizeof(CMD_LINE_OPT_RX_RETRY_DELAY))) {
 398                                 ret = parse_num_opt(optarg, INT32_MAX);
 399                                 if (ret == -1) {
 400                                         RTE_LOG(INFO, VHOST_CONFIG,
 401                                                 "Invalid argument for rx-retry-delay [0-N]\n");
 402                                         tep_termination_usage(prgname);
 403                                         return -1;
 404                                 } else
 405                                         burst_rx_delay_time = ret;
 406                         }
 407
 408                         /* Specify the retries number on RX. */
 409                         if (!strncmp(long_option[option_index].name,
 410                                 CMD_LINE_OPT_RX_RETRY_NUM,
 411                                 sizeof(CMD_LINE_OPT_RX_RETRY_NUM))) {
 412                                 ret = parse_num_opt(optarg, INT32_MAX);
 413                                 if (ret == -1) {
 414                                         RTE_LOG(INFO, VHOST_CONFIG,
 415                                                 "Invalid argument for rx-retry-num [0-N]\n");
 416                                         tep_termination_usage(prgname);
 417                                         return -1;
 418                                 } else
 419                                         burst_rx_retry_num = ret;
 420                         }
 421
 422                         if (!strncmp(long_option[option_index].name,
 423                                 CMD_LINE_OPT_TX_CHECKSUM,
 424                                 sizeof(CMD_LINE_OPT_TX_CHECKSUM))) {
 425                                 ret = parse_num_opt(optarg, 1);
 426                                 if (ret == -1) {
 427                                         RTE_LOG(INFO, VHOST_CONFIG,
 428                                                 "Invalid argument for tx-checksum [0|1]\n");
 429                                         tep_termination_usage(prgname);
 430                                         return -1;
 431                                 } else
 432                                         tx_checksum = ret;
 433                         }
 434
 435                         if (!strncmp(long_option[option_index].name,
 436                                         CMD_LINE_OPT_FILTER_TYPE,
 437                                         sizeof(CMD_LINE_OPT_FILTER_TYPE))) {
 438                                 ret = parse_num_opt(optarg, 3);
 439                                 if ((ret == -1) || (ret == 0)) {
 440                                         RTE_LOG(INFO, VHOST_CONFIG,
 441                                                 "Invalid argument for filter type [1-3]\n");
 442                                         tep_termination_usage(prgname);
 443                                         return -1;
 444                                 } else
 445                                         filter_idx = ret - 1;
 446                         }
 447
 448                         /* Enable/disable encapsulation on RX. */
 449                         if (!strncmp(long_option[option_index].name,
 450                                 CMD_LINE_OPT_DECAP,
 451                                 sizeof(CMD_LINE_OPT_DECAP))) {
 452                                 ret = parse_num_opt(optarg, 1);
 453                                 if (ret == -1) {
 454                                         RTE_LOG(INFO, VHOST_CONFIG,
 455                                                 "Invalid argument for decap [0|1]\n");
 456                                         tep_termination_usage(prgname);
 457                                         return -1;
 458                                 } else
 459                                         rx_decap = ret;
 460                         }
 461
 462                         /* Enable/disable encapsulation on TX. */
 463                         if (!strncmp(long_option[option_index].name,
 464                                 CMD_LINE_OPT_ENCAP,
 465                                 sizeof(CMD_LINE_OPT_ENCAP))) {
 466                                 ret = parse_num_opt(optarg, 1);
 467                                 if (ret == -1) {
 468                                         RTE_LOG(INFO, VHOST_CONFIG,
 469                                                 "Invalid argument for encap [0|1]\n");
 470                                         tep_termination_usage(prgname);
 471                                         return -1;
 472                                 } else
 473                                         tx_encap = ret;
 474                         }
 475
 476                         /* Enable/disable stats. */
 477                         if (!strncmp(long_option[option_index].name,
 478                                 CMD_LINE_OPT_STATS,
 479                                 sizeof(CMD_LINE_OPT_STATS))) {
 480                                 ret = parse_num_opt(optarg, INT32_MAX);
 481                                 if (ret == -1) {
 482                                         RTE_LOG(INFO, VHOST_CONFIG,
 483                                                         "Invalid argument for stats [0..N]\n");
 484                                         tep_termination_usage(prgname);
 485                                         return -1;
 486                                 } else
 487                                         enable_stats = ret;
 488                         }
 489
 490                         /* Set character device basename. */
 491                         if (!strncmp(long_option[option_index].name,
 492                                 CMD_LINE_OPT_DEV_BASENAME,
 493                                 sizeof(CMD_LINE_OPT_DEV_BASENAME))) {
 494                                 if (us_vhost_parse_basename(optarg) == -1) {
 495                                         RTE_LOG(INFO, VHOST_CONFIG,
 496                                                 "Invalid argument for character "
 497                                                 "device basename (Max %d characters)\n",
 498                                                 MAX_BASENAME_SZ);
 499                                         tep_termination_usage(prgname);
 500                                         return -1;
 501                                 }
 502                         }
 503
 504                         break;
 505
 506                         /* Invalid option - print options. */
 507                 default:
 508                         tep_termination_usage(prgname);
 509                         return -1;
 510                 }
 511         }
 512
 513         for (i = 0; i < RTE_MAX_ETHPORTS; i++) {
 514                 if (enabled_port_mask & (1 << i))
 515                         ports[nb_ports++] = (uint8_t)i;
 516         }
 517
 518         if ((nb_ports ==  0) || (nb_ports > MAX_SUP_PORTS)) {
 519                 RTE_LOG(INFO, VHOST_PORT, "Current enabled port number is %u,"
 520                         "but only %u port can be enabled\n", nb_ports,
 521                         MAX_SUP_PORTS);
 522                 return -1;
 523         }
 524
 525         return 0;
 526 }
 527
 528 /**
 529  * Update the global var NB_PORTS and array PORTS
 530  * according to system ports number and return valid ports number
 531  */
 532 static unsigned
 533 check_ports_num(unsigned max_nb_ports)
 534 {
 535         unsigned valid_nb_ports = nb_ports;
 536         unsigned portid;
 537
 538         if (nb_ports > max_nb_ports) {
 539                 RTE_LOG(INFO, VHOST_PORT, "\nSpecified port number(%u) "
 540                         " exceeds total system port number(%u)\n",
 541                         nb_ports, max_nb_ports);
 542                 nb_ports = max_nb_ports;
 543         }
 544
 545         for (portid = 0; portid < nb_ports; portid++) {
 546                 if (ports[portid] >= max_nb_ports) {
 547                         RTE_LOG(INFO, VHOST_PORT,
 548                                 "\nSpecified port ID(%u) exceeds max "
 549                                 " system port ID(%u)\n",
 550                                 ports[portid], (max_nb_ports - 1));
 551                         ports[portid] = INVALID_PORT_ID;
 552                         valid_nb_ports--;
 553                 }
 554         }
 555         return valid_nb_ports;
 556 }
 557
 558 /**
 559  * This function routes the TX packet to the correct interface. This may be a local device
 560  * or the physical port.
 561  */
 562 static inline void __attribute__((always_inline))
 563 virtio_tx_route(struct vhost_dev *vdev, struct rte_mbuf *m)
 564 {
 565         struct mbuf_table *tx_q;
 566         struct rte_mbuf **m_table;
 567         unsigned len, ret = 0;
 568         const uint16_t lcore_id = rte_lcore_id();
 569
 570         RTE_LOG(DEBUG, VHOST_DATA, "(%d) TX: MAC address is external\n",
 571                 vdev->vid);
 572
 573         /* Add packet to the port tx queue */
 574         tx_q = &lcore_tx_queue[lcore_id];
 575         len = tx_q->len;
 576
 577         tx_q->m_table[len] = m;
 578         len++;
 579         if (enable_stats) {
 580                 dev_statistics[vdev->vid].tx_total++;
 581                 dev_statistics[vdev->vid].tx++;
 582         }
 583
 584         if (unlikely(len == MAX_PKT_BURST)) {
 585                 m_table = (struct rte_mbuf **)tx_q->m_table;
 586                 ret = overlay_options.tx_handle(ports[0],
 587                         (uint16_t)tx_q->txq_id, m_table,
 588                         (uint16_t)tx_q->len);
 589
 590                 /* Free any buffers not handled by TX and update
 591                  * the port stats.
 592                  */
 593                 if (unlikely(ret < len)) {
 594                         do {
 595                                 rte_pktmbuf_free(m_table[ret]);
 596                         } while (++ret < len);
 597                 }
 598
 599                 len = 0;
 600         }
 601
 602         tx_q->len = len;
 603         return;
 604 }
 605
 606 /**
 607  * This function is called by each data core. It handles all
 608  * RX/TX registered with the core. For TX the specific lcore
 609  * linked list is used. For RX, MAC addresses are compared
 610  * with all devices in the main linked list.
 611  */
 612 static int
 613 switch_worker(__rte_unused void *arg)
 614 {
 615         struct rte_mempool *mbuf_pool = arg;
 616         struct vhost_dev *vdev = NULL;
 617         struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
 618         struct virtio_net_data_ll *dev_ll;
 619         struct mbuf_table *tx_q;
 620         volatile struct lcore_ll_info *lcore_ll;
 621         const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1)
 622                                         / US_PER_S * BURST_TX_DRAIN_US;
 623         uint64_t prev_tsc, diff_tsc, cur_tsc, ret_count = 0;
 624         unsigned i, ret = 0;
 625         const uint16_t lcore_id = rte_lcore_id();
 626         const uint16_t num_cores = (uint16_t)rte_lcore_count();
 627         uint16_t rx_count = 0;
 628         uint16_t tx_count;
 629         uint32_t retry = 0;
 630
 631         RTE_LOG(INFO, VHOST_DATA, "Procesing on Core %u started\n", lcore_id);
 632         lcore_ll = lcore_info[lcore_id].lcore_ll;
 633         prev_tsc = 0;
 634
 635         tx_q = &lcore_tx_queue[lcore_id];
 636         for (i = 0; i < num_cores; i++) {
 637                 if (lcore_ids[i] == lcore_id) {
 638                         tx_q->txq_id = i;
 639                         break;
 640                 }
 641         }
 642
 643         while (1) {
 644                 cur_tsc = rte_rdtsc();
 645                 /*
 646                  * TX burst queue drain
 647                  */
 648                 diff_tsc = cur_tsc - prev_tsc;
 649                 if (unlikely(diff_tsc > drain_tsc)) {
 650
 651                         if (tx_q->len) {
 652                                 RTE_LOG(DEBUG, VHOST_DATA, "TX queue drained after "
 653                                         "timeout with burst size %u\n",
 654                                         tx_q->len);
 655                                 ret = overlay_options.tx_handle(ports[0],
 656                                         (uint16_t)tx_q->txq_id,
 657                                         (struct rte_mbuf **)tx_q->m_table,
 658                                         (uint16_t)tx_q->len);
 659                                 if (unlikely(ret < tx_q->len)) {
 660                                         do {
 661                                                 rte_pktmbuf_free(tx_q->m_table[ret]);
 662                                         } while (++ret < tx_q->len);
 663                                 }
 664
 665                                 tx_q->len = 0;
 666                         }
 667
 668                         prev_tsc = cur_tsc;
 669
 670                 }
 671
 672                 rte_prefetch0(lcore_ll->ll_root_used);
 673
 674                 /**
 675                  * Inform the configuration core that we have exited
 676                  * the linked list and that no devices are
 677                  * in use if requested.
 678                  */
 679                 if (lcore_ll->dev_removal_flag == REQUEST_DEV_REMOVAL)
 680                         lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
 681
 682                 /*
 683                  * Process devices
 684                  */
 685                 dev_ll = lcore_ll->ll_root_used;
 686
 687                 while (dev_ll != NULL) {
 688                         vdev = dev_ll->vdev;
 689
 690                         if (unlikely(vdev->remove)) {
 691                                 dev_ll = dev_ll->next;
 692                                 overlay_options.tunnel_destroy(vdev);
 693                                 vdev->ready = DEVICE_SAFE_REMOVE;
 694                                 continue;
 695                         }
 696                         if (likely(vdev->ready == DEVICE_RX)) {
 697                                 /* Handle guest RX */
 698                                 rx_count = rte_eth_rx_burst(ports[0],
 699                                         vdev->rx_q, pkts_burst, MAX_PKT_BURST);
 700
 701                                 if (rx_count) {
 702                                         /*
 703                                         * Retry is enabled and the queue is
 704                                         * full then we wait and retry to
 705                                         * avoid packet loss. Here MAX_PKT_BURST
 706                                         * must be less than virtio queue size
 707                                         */
 708                                         if (enable_retry && unlikely(rx_count >
 709                                                 rte_vhost_avail_entries(vdev->vid, VIRTIO_RXQ))) {
 710                                                 for (retry = 0; retry < burst_rx_retry_num;
 711                                                         retry++) {
 712                                                         rte_delay_us(burst_rx_delay_time);
 713                                                         if (rx_count <= rte_vhost_avail_entries(vdev->vid, VIRTIO_RXQ))
 714                                                                 break;
 715                                                 }
 716                                         }
 717
 718                                         ret_count = overlay_options.rx_handle(vdev->vid, pkts_burst, rx_count);
 719                                         if (enable_stats) {
 720                                                 rte_atomic64_add(
 721                                                 &dev_statistics[vdev->vid].rx_total_atomic,
 722                                                 rx_count);
 723                                                 rte_atomic64_add(
 724                                                 &dev_statistics[vdev->vid].rx_atomic, ret_count);
 725                                         }
 726                                         while (likely(rx_count)) {
 727                                                 rx_count--;
 728                                                 rte_pktmbuf_free(pkts_burst[rx_count]);
 729                                         }
 730
 731                                 }
 732                         }
 733
 734                         if (likely(!vdev->remove)) {
 735                                 /* Handle guest TX*/
 736                                 tx_count = rte_vhost_dequeue_burst(vdev->vid,
 737                                                 VIRTIO_TXQ, mbuf_pool,
 738                                                 pkts_burst, MAX_PKT_BURST);
 739                                 /* If this is the first received packet we need to learn the MAC */
 740                                 if (unlikely(vdev->ready == DEVICE_MAC_LEARNING) && tx_count) {
 741                                         if (vdev->remove ||
 742                                                 (overlay_options.tunnel_setup(vdev, pkts_burst[0]) == -1)) {
 743                                                 while (tx_count)
 744                                                         rte_pktmbuf_free(pkts_burst[--tx_count]);
 745                                         }
 746                                 }
 747                                 while (tx_count)
 748                                         virtio_tx_route(vdev, pkts_burst[--tx_count]);
 749                         }
 750
 751                         /* move to the next device in the list */
 752                         dev_ll = dev_ll->next;
 753                 }
 754         }
 755
 756         return 0;
 757 }
 758
 759 /**
 760  * Add an entry to a used linked list. A free entry must first be found
 761  * in the free linked list using get_data_ll_free_entry();
 762  */
 763 static void
 764 add_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
 765         struct virtio_net_data_ll *ll_dev)
 766 {
 767         struct virtio_net_data_ll *ll = *ll_root_addr;
 768
 769         /* Set next as NULL and use a compiler barrier to avoid reordering. */
 770         ll_dev->next = NULL;
 771         rte_compiler_barrier();
 772
 773         /* If ll == NULL then this is the first device. */
 774         if (ll) {
 775                 /* Increment to the tail of the linked list. */
 776                 while (ll->next != NULL)
 777                         ll = ll->next;
 778
 779                 ll->next = ll_dev;
 780         } else {
 781                 *ll_root_addr = ll_dev;
 782         }
 783 }
 784
 785 /**
 786  * Remove an entry from a used linked list. The entry must then be added to
 787  * the free linked list using put_data_ll_free_entry().
 788  */
 789 static void
 790 rm_data_ll_entry(struct virtio_net_data_ll **ll_root_addr,
 791         struct virtio_net_data_ll *ll_dev,
 792         struct virtio_net_data_ll *ll_dev_last)
 793 {
 794         struct virtio_net_data_ll *ll = *ll_root_addr;
 795
 796         if (unlikely((ll == NULL) || (ll_dev == NULL)))
 797                 return;
 798
 799         if (ll_dev == ll)
 800                 *ll_root_addr = ll_dev->next;
 801         else
 802                 if (likely(ll_dev_last != NULL))
 803                         ll_dev_last->next = ll_dev->next;
 804                 else
 805                         RTE_LOG(ERR, VHOST_CONFIG,
 806                                 "Remove entry form ll failed.\n");
 807 }
 808
 809 /**
 810  * Find and return an entry from the free linked list.
 811  */
 812 static struct virtio_net_data_ll *
 813 get_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr)
 814 {
 815         struct virtio_net_data_ll *ll_free = *ll_root_addr;
 816         struct virtio_net_data_ll *ll_dev;
 817
 818         if (ll_free == NULL)
 819                 return NULL;
 820
 821         ll_dev = ll_free;
 822         *ll_root_addr = ll_free->next;
 823
 824         return ll_dev;
 825 }
 826
 827 /**
 828  * Place an entry back on to the free linked list.
 829  */
 830 static void
 831 put_data_ll_free_entry(struct virtio_net_data_ll **ll_root_addr,
 832         struct virtio_net_data_ll *ll_dev)
 833 {
 834         struct virtio_net_data_ll *ll_free = *ll_root_addr;
 835
 836         if (ll_dev == NULL)
 837                 return;
 838
 839         ll_dev->next = ll_free;
 840         *ll_root_addr = ll_dev;
 841 }
 842
 843 /**
 844  * Creates a linked list of a given size.
 845  */
 846 static struct virtio_net_data_ll *
 847 alloc_data_ll(uint32_t size)
 848 {
 849         struct virtio_net_data_ll *ll_new;
 850         uint32_t i;
 851
 852         /* Malloc and then chain the linked list. */
 853         ll_new = malloc(size * sizeof(struct virtio_net_data_ll));
 854         if (ll_new == NULL) {
 855                 RTE_LOG(ERR, VHOST_CONFIG,
 856                         "Failed to allocate memory for ll_new.\n");
 857                 return NULL;
 858         }
 859
 860         for (i = 0; i < size - 1; i++) {
 861                 ll_new[i].vdev = NULL;
 862                 ll_new[i].next = &ll_new[i+1];
 863         }
 864         ll_new[i].next = NULL;
 865
 866         return ll_new;
 867 }
 868
 869 /**
 870  * Create the main linked list along with each individual cores
 871  * linked list. A used and a free list are created to manage entries.
 872  */
 873 static int
 874 init_data_ll(void)
 875 {
 876         int lcore;
 877
 878         RTE_LCORE_FOREACH_SLAVE(lcore) {
 879                 lcore_info[lcore].lcore_ll =
 880                         malloc(sizeof(struct lcore_ll_info));
 881                 if (lcore_info[lcore].lcore_ll == NULL) {
 882                         RTE_LOG(ERR, VHOST_CONFIG,
 883                                 "Failed to allocate memory for lcore_ll.\n");
 884                         return -1;
 885                 }
 886
 887                 lcore_info[lcore].lcore_ll->device_num = 0;
 888                 lcore_info[lcore].lcore_ll->dev_removal_flag = ACK_DEV_REMOVAL;
 889                 lcore_info[lcore].lcore_ll->ll_root_used = NULL;
 890                 if (nb_devices % nb_switching_cores)
 891                         lcore_info[lcore].lcore_ll->ll_root_free =
 892                                 alloc_data_ll((nb_devices / nb_switching_cores)
 893                                                 + 1);
 894                 else
 895                         lcore_info[lcore].lcore_ll->ll_root_free =
 896                                 alloc_data_ll(nb_devices / nb_switching_cores);
 897         }
 898
 899         /* Allocate devices up to a maximum of MAX_DEVICES. */
 900         ll_root_free = alloc_data_ll(MIN((nb_devices), MAX_DEVICES));
 901
 902         return 0;
 903 }
 904
 905 /**
 906  * Remove a device from the specific data core linked list and
 907  * from the main linked list. Synchonization occurs through the use
 908  * of the lcore dev_removal_flag.
 909  */
 910 static void
 911 destroy_device(int vid)
 912 {
 913         struct virtio_net_data_ll *ll_lcore_dev_cur;
 914         struct virtio_net_data_ll *ll_main_dev_cur;
 915         struct virtio_net_data_ll *ll_lcore_dev_last = NULL;
 916         struct virtio_net_data_ll *ll_main_dev_last = NULL;
 917         struct vhost_dev *vdev = NULL;
 918         int lcore;
 919
 920         ll_main_dev_cur = ll_root_used;
 921         while (ll_main_dev_cur != NULL) {
 922                 if (ll_main_dev_cur->vdev->vid == vid) {
 923                         vdev = ll_main_dev_cur->vdev;
 924                         break;
 925                 }
 926         }
 927         if (!vdev)
 928                 return;
 929
 930         /* set the remove flag. */
 931         vdev->remove = 1;
 932         while (vdev->ready != DEVICE_SAFE_REMOVE)
 933                 rte_pause();
 934
 935         /* Search for entry to be removed from lcore ll */
 936         ll_lcore_dev_cur = lcore_info[vdev->coreid].lcore_ll->ll_root_used;
 937         while (ll_lcore_dev_cur != NULL) {
 938                 if (ll_lcore_dev_cur->vdev == vdev) {
 939                         break;
 940                 } else {
 941                         ll_lcore_dev_last = ll_lcore_dev_cur;
 942                         ll_lcore_dev_cur = ll_lcore_dev_cur->next;
 943                 }
 944         }
 945
 946         if (ll_lcore_dev_cur == NULL) {
 947                 RTE_LOG(ERR, VHOST_CONFIG,
 948                         "(%d) Failed to find the dev to be destroy.\n", vid);
 949                 return;
 950         }
 951
 952         /* Search for entry to be removed from main ll */
 953         ll_main_dev_cur = ll_root_used;
 954         ll_main_dev_last = NULL;
 955         while (ll_main_dev_cur != NULL) {
 956                 if (ll_main_dev_cur->vdev == vdev) {
 957                         break;
 958                 } else {
 959                         ll_main_dev_last = ll_main_dev_cur;
 960                         ll_main_dev_cur = ll_main_dev_cur->next;
 961                 }
 962         }
 963
 964         /* Remove entries from the lcore and main ll. */
 965         rm_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
 966                         ll_lcore_dev_cur, ll_lcore_dev_last);
 967         rm_data_ll_entry(&ll_root_used, ll_main_dev_cur, ll_main_dev_last);
 968
 969         /* Set the dev_removal_flag on each lcore. */
 970         RTE_LCORE_FOREACH_SLAVE(lcore) {
 971                 lcore_info[lcore].lcore_ll->dev_removal_flag =
 972                         REQUEST_DEV_REMOVAL;
 973         }
 974
 975         /*
 976          * Once each core has set the dev_removal_flag to
 977          * ACK_DEV_REMOVAL we can be sure that they can no longer access
 978          * the device removed from the linked lists and that the devices
 979          * are no longer in use.
 980          */
 981         RTE_LCORE_FOREACH_SLAVE(lcore) {
 982                 while (lcore_info[lcore].lcore_ll->dev_removal_flag
 983                         != ACK_DEV_REMOVAL)
 984                         rte_pause();
 985         }
 986
 987         /* Add the entries back to the lcore and main free ll.*/
 988         put_data_ll_free_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_free,
 989                                 ll_lcore_dev_cur);
 990         put_data_ll_free_entry(&ll_root_free, ll_main_dev_cur);
 991
 992         /* Decrement number of device on the lcore. */
 993         lcore_info[vdev->coreid].lcore_ll->device_num--;
 994
 995         RTE_LOG(INFO, VHOST_DATA, "(%d) Device has been removed "
 996                 "from data core\n", vid);
 997
 998         rte_free(vdev);
 999
1000 }
1001
1002 /**
1003  * A new device is added to a data core. First the device is added
1004  * to the main linked list and the allocated to a specific data core.
1005  */
1006 static int
1007 new_device(int vid)
1008 {
1009         struct virtio_net_data_ll *ll_dev;
1010         int lcore, core_add = 0;
1011         uint32_t device_num_min = nb_devices;
1012         struct vhost_dev *vdev;
1013
1014         vdev = rte_zmalloc("vhost device", sizeof(*vdev), RTE_CACHE_LINE_SIZE);
1015         if (vdev == NULL) {
1016                 RTE_LOG(INFO, VHOST_DATA,
1017                         "(%d) Couldn't allocate memory for vhost dev\n", vid);
1018                 return -1;
1019         }
1020         vdev->vid = vid;
1021         /* Add device to main ll */
1022         ll_dev = get_data_ll_free_entry(&ll_root_free);
1023         if (ll_dev == NULL) {
1024                 RTE_LOG(INFO, VHOST_DATA, "(%d) No free entry found in"
1025                         " linked list Device limit of %d devices per core"
1026                         " has been reached\n", vid, nb_devices);
1027                 if (vdev->regions_hpa)
1028                         rte_free(vdev->regions_hpa);
1029                 rte_free(vdev);
1030                 return -1;
1031         }
1032         ll_dev->vdev = vdev;
1033         add_data_ll_entry(&ll_root_used, ll_dev);
1034         vdev->rx_q = vid;
1035
1036         /* reset ready flag */
1037         vdev->ready = DEVICE_MAC_LEARNING;
1038         vdev->remove = 0;
1039
1040         /* Find a suitable lcore to add the device. */
1041         RTE_LCORE_FOREACH_SLAVE(lcore) {
1042                 if (lcore_info[lcore].lcore_ll->device_num < device_num_min) {
1043                         device_num_min = lcore_info[lcore].lcore_ll->device_num;
1044                         core_add = lcore;
1045                 }
1046         }
1047         /* Add device to lcore ll */
1048         ll_dev = get_data_ll_free_entry(&lcore_info[core_add].lcore_ll->ll_root_free);
1049         if (ll_dev == NULL) {
1050                 RTE_LOG(INFO, VHOST_DATA,
1051                         "(%d) Failed to add device to data core\n",
1052                         vid);
1053                 vdev->ready = DEVICE_SAFE_REMOVE;
1054                 destroy_device(vid);
1055                 rte_free(vdev->regions_hpa);
1056                 rte_free(vdev);
1057                 return -1;
1058         }
1059         ll_dev->vdev = vdev;
1060         vdev->coreid = core_add;
1061
1062         add_data_ll_entry(&lcore_info[vdev->coreid].lcore_ll->ll_root_used,
1063                         ll_dev);
1064
1065         /* Initialize device stats */
1066         memset(&dev_statistics[vid], 0,
1067                 sizeof(struct device_statistics));
1068
1069         /* Disable notifications. */
1070         rte_vhost_enable_guest_notification(vid, VIRTIO_RXQ, 0);
1071         rte_vhost_enable_guest_notification(vid, VIRTIO_TXQ, 0);
1072         lcore_info[vdev->coreid].lcore_ll->device_num++;
1073
1074         RTE_LOG(INFO, VHOST_DATA, "(%d) Device has been added to data core %d\n",
1075                 vid, vdev->coreid);
1076
1077         return 0;
1078 }
1079
1080 /**
1081  * These callback allow devices to be added to the data core when configuration
1082  * has been fully complete.
1083  */
1084 static const struct virtio_net_device_ops virtio_net_device_ops = {
1085         .new_device =  new_device,
1086         .destroy_device = destroy_device,
1087 };
1088
1089 /**
1090  * This is a thread will wake up after a period to print stats if the user has
1091  * enabled them.
1092  */
1093 static void
1094 print_stats(void)
1095 {
1096         struct virtio_net_data_ll *dev_ll;
1097         uint64_t tx_dropped, rx_dropped;
1098         uint64_t tx, tx_total, rx, rx_total, rx_ip_csum, rx_l4_csum;
1099         int vid;
1100         const char clr[] = { 27, '[', '2', 'J', '\0' };
1101         const char top_left[] = { 27, '[', '1', ';', '1', 'H', '\0' };
1102
1103         while (1) {
1104                 sleep(enable_stats);
1105
1106                 /* Clear screen and move to top left */
1107                 printf("%s%s", clr, top_left);
1108
1109                 printf("\nDevice statistics ================================");
1110
1111                 dev_ll = ll_root_used;
1112                 while (dev_ll != NULL) {
1113                         vid = dev_ll->vdev->vid;
1114                         tx_total = dev_statistics[vid].tx_total;
1115                         tx = dev_statistics[vid].tx;
1116                         tx_dropped = tx_total - tx;
1117
1118                         rx_total = rte_atomic64_read(
1119                                 &dev_statistics[vid].rx_total_atomic);
1120                         rx = rte_atomic64_read(
1121                                 &dev_statistics[vid].rx_atomic);
1122                         rx_dropped = rx_total - rx;
1123                         rx_ip_csum = rte_atomic64_read(
1124                                 &dev_statistics[vid].rx_bad_ip_csum);
1125                         rx_l4_csum = rte_atomic64_read(
1126                                 &dev_statistics[vid].rx_bad_l4_csum);
1127
1128                         printf("\nStatistics for device %d ----------"
1129                                         "\nTX total:            %"PRIu64""
1130                                         "\nTX dropped:          %"PRIu64""
1131                                         "\nTX successful:               %"PRIu64""
1132                                         "\nRX total:            %"PRIu64""
1133                                         "\nRX bad IP csum:      %"PRIu64""
1134                                         "\nRX bad L4 csum:      %"PRIu64""
1135                                         "\nRX dropped:          %"PRIu64""
1136                                         "\nRX successful:               %"PRIu64"",
1137                                         vid,
1138                                         tx_total,
1139                                         tx_dropped,
1140                                         tx,
1141                                         rx_total,
1142                                         rx_ip_csum,
1143                                         rx_l4_csum,
1144                                         rx_dropped,
1145                                         rx);
1146
1147                         dev_ll = dev_ll->next;
1148                 }
1149                 printf("\n================================================\n");
1150         }
1151 }
1152
1153 /**
1154  * Main function, does initialisation and calls the per-lcore functions.
1155  */
1156 int
1157 main(int argc, char *argv[])
1158 {
1159         struct rte_mempool *mbuf_pool = NULL;
1160         unsigned lcore_id, core_id = 0;
1161         unsigned nb_ports, valid_nb_ports;
1162         int ret;
1163         uint8_t portid;
1164         uint16_t queue_id;
1165         static pthread_t tid;
1166         char thread_name[RTE_MAX_THREAD_NAME_LEN];
1167
1168         /* init EAL */
1169         ret = rte_eal_init(argc, argv);
1170         if (ret < 0)
1171                 rte_exit(EXIT_FAILURE, "Error with EAL initialization\n");
1172         argc -= ret;
1173         argv += ret;
1174
1175         /* parse app arguments */
1176         ret = tep_termination_parse_args(argc, argv);
1177         if (ret < 0)
1178                 rte_exit(EXIT_FAILURE, "Invalid argument\n");
1179
1180         for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++)
1181                 if (rte_lcore_is_enabled(lcore_id))
1182                         lcore_ids[core_id++] = lcore_id;
1183
1184         /* set the number of swithcing cores available */
1185         nb_switching_cores = rte_lcore_count()-1;
1186
1187         /* Get the number of physical ports. */
1188         nb_ports = rte_eth_dev_count();
1189
1190         /*
1191          * Update the global var NB_PORTS and global array PORTS
1192          * and get value of var VALID_NB_PORTS according to system ports number
1193          */
1194         valid_nb_ports = check_ports_num(nb_ports);
1195
1196         if ((valid_nb_ports == 0) || (valid_nb_ports > MAX_SUP_PORTS)) {
1197                 rte_exit(EXIT_FAILURE, "Current enabled port number is %u,"
1198                         "but only %u port can be enabled\n", nb_ports,
1199                         MAX_SUP_PORTS);
1200         }
1201         /* Create the mbuf pool. */
1202         mbuf_pool = rte_mempool_create(
1203                         "MBUF_POOL",
1204                         NUM_MBUFS_PER_PORT
1205                         * valid_nb_ports,
1206                         MBUF_SIZE, MBUF_CACHE_SIZE,
1207                         sizeof(struct rte_pktmbuf_pool_private),
1208                         rte_pktmbuf_pool_init, NULL,
1209                         rte_pktmbuf_init, NULL,
1210                         rte_socket_id(), 0);
1211         if (mbuf_pool == NULL)
1212                 rte_exit(EXIT_FAILURE, "Cannot create mbuf pool\n");
1213
1214         for (queue_id = 0; queue_id < MAX_QUEUES + 1; queue_id++)
1215                 vpool_array[queue_id].pool = mbuf_pool;
1216
1217         /* initialize all ports */
1218         for (portid = 0; portid < nb_ports; portid++) {
1219                 /* skip ports that are not enabled */
1220                 if ((enabled_port_mask & (1 << portid)) == 0) {
1221                         RTE_LOG(INFO, VHOST_PORT,
1222                                 "Skipping disabled port %d\n", portid);
1223                         continue;
1224                 }
1225                 if (overlay_options.port_configure(portid, mbuf_pool) != 0)
1226                         rte_exit(EXIT_FAILURE,
1227                                 "Cannot initialize network ports\n");
1228         }
1229
1230         /* Initialise all linked lists. */
1231         if (init_data_ll() == -1)
1232                 rte_exit(EXIT_FAILURE, "Failed to initialize linked list\n");
1233
1234         /* Initialize device stats */
1235         memset(&dev_statistics, 0, sizeof(dev_statistics));
1236
1237         /* Enable stats if the user option is set. */
1238         if (enable_stats) {
1239                 ret = pthread_create(&tid, NULL, (void *)print_stats, NULL);
1240                 if (ret != 0)
1241                         rte_exit(EXIT_FAILURE, "Cannot create print-stats thread\n");
1242                 snprintf(thread_name, RTE_MAX_THREAD_NAME_LEN, "print-stats");
1243                 ret = rte_thread_setname(tid, thread_name);
1244                 if (ret != 0)
1245                         RTE_LOG(DEBUG, VHOST_CONFIG, "Cannot set print-stats name\n");
1246         }
1247
1248         /* Launch all data cores. */
1249         RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1250                 rte_eal_remote_launch(switch_worker,
1251                         mbuf_pool, lcore_id);
1252         }
1253         rte_vhost_feature_disable(1ULL << VIRTIO_NET_F_MRG_RXBUF);
1254
1255         ret = rte_vhost_driver_register((char *)&dev_basename, 0);
1256         if (ret != 0)
1257                 rte_exit(EXIT_FAILURE, "failed to register vhost driver.\n");
1258
1259         rte_vhost_driver_callback_register(&virtio_net_device_ops);
1260
1261         rte_vhost_driver_session_start();
1262
1263         return 0;
1264 }