ceph/src/dpdk/examples/ip_fragmentation/main.c

   1 /*-
   2  *   BSD LICENSE
   3  *
   4  *   Copyright(c) 2010-2014 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 <stdio.h>
  35 #include <stdlib.h>
  36 #include <stdint.h>
  37 #include <inttypes.h>
  38 #include <sys/types.h>
  39 #include <sys/param.h>
  40 #include <string.h>
  41 #include <sys/queue.h>
  42 #include <stdarg.h>
  43 #include <errno.h>
  44 #include <getopt.h>
  45
  46 #include <rte_common.h>
  47 #include <rte_byteorder.h>
  48 #include <rte_log.h>
  49 #include <rte_memory.h>
  50 #include <rte_memcpy.h>
  51 #include <rte_memzone.h>
  52 #include <rte_eal.h>
  53 #include <rte_per_lcore.h>
  54 #include <rte_launch.h>
  55 #include <rte_atomic.h>
  56 #include <rte_cycles.h>
  57 #include <rte_prefetch.h>
  58 #include <rte_lcore.h>
  59 #include <rte_per_lcore.h>
  60 #include <rte_branch_prediction.h>
  61 #include <rte_interrupts.h>
  62 #include <rte_pci.h>
  63 #include <rte_random.h>
  64 #include <rte_debug.h>
  65 #include <rte_ether.h>
  66 #include <rte_ethdev.h>
  67 #include <rte_mempool.h>
  68 #include <rte_mbuf.h>
  69 #include <rte_lpm.h>
  70 #include <rte_lpm6.h>
  71 #include <rte_ip.h>
  72 #include <rte_string_fns.h>
  73
  74 #include <rte_ip_frag.h>
  75
  76 #define RTE_LOGTYPE_IP_FRAG RTE_LOGTYPE_USER1
  77
  78 /* allow max jumbo frame 9.5 KB */
  79 #define JUMBO_FRAME_MAX_SIZE    0x2600
  80
  81 #define ROUNDUP_DIV(a, b)       (((a) + (b) - 1) / (b))
  82
  83 /*
  84  * Default byte size for the IPv6 Maximum Transfer Unit (MTU).
  85  * This value includes the size of IPv6 header.
  86  */
  87 #define IPV4_MTU_DEFAULT        ETHER_MTU
  88 #define IPV6_MTU_DEFAULT        ETHER_MTU
  89
  90 /*
  91  * Default payload in bytes for the IPv6 packet.
  92  */
  93 #define IPV4_DEFAULT_PAYLOAD    (IPV4_MTU_DEFAULT - sizeof(struct ipv4_hdr))
  94 #define IPV6_DEFAULT_PAYLOAD    (IPV6_MTU_DEFAULT - sizeof(struct ipv6_hdr))
  95
  96 /*
  97  * Max number of fragments per packet expected - defined by config file.
  98  */
  99 #define MAX_PACKET_FRAG RTE_LIBRTE_IP_FRAG_MAX_FRAG
 100
 101 #define NB_MBUF   8192
 102
 103 #define MAX_PKT_BURST   32
 104 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
 105
 106 /* Configure how many packets ahead to prefetch, when reading packets */
 107 #define PREFETCH_OFFSET 3
 108
 109 /*
 110  * Configurable number of RX/TX ring descriptors
 111  */
 112 #define RTE_TEST_RX_DESC_DEFAULT 128
 113 #define RTE_TEST_TX_DESC_DEFAULT 512
 114 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
 115 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
 116
 117 /* ethernet addresses of ports */
 118 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
 119
 120 #ifndef IPv4_BYTES
 121 #define IPv4_BYTES_FMT "%" PRIu8 ".%" PRIu8 ".%" PRIu8 ".%" PRIu8
 122 #define IPv4_BYTES(addr) \
 123                 (uint8_t) (((addr) >> 24) & 0xFF),\
 124                 (uint8_t) (((addr) >> 16) & 0xFF),\
 125                 (uint8_t) (((addr) >> 8) & 0xFF),\
 126                 (uint8_t) ((addr) & 0xFF)
 127 #endif
 128
 129 #ifndef IPv6_BYTES
 130 #define IPv6_BYTES_FMT "%02x%02x:%02x%02x:%02x%02x:%02x%02x:"\
 131                        "%02x%02x:%02x%02x:%02x%02x:%02x%02x"
 132 #define IPv6_BYTES(addr) \
 133         addr[0],  addr[1], addr[2],  addr[3], \
 134         addr[4],  addr[5], addr[6],  addr[7], \
 135         addr[8],  addr[9], addr[10], addr[11],\
 136         addr[12], addr[13],addr[14], addr[15]
 137 #endif
 138
 139 #define IPV6_ADDR_LEN 16
 140
 141 /* mask of enabled ports */
 142 static int enabled_port_mask = 0;
 143
 144 static int rx_queue_per_lcore = 1;
 145
 146 #define MBUF_TABLE_SIZE  (2 * MAX(MAX_PKT_BURST, MAX_PACKET_FRAG))
 147
 148 struct mbuf_table {
 149         uint16_t len;
 150         struct rte_mbuf *m_table[MBUF_TABLE_SIZE];
 151 };
 152
 153 struct rx_queue {
 154         struct rte_mempool *direct_pool;
 155         struct rte_mempool *indirect_pool;
 156         struct rte_lpm *lpm;
 157         struct rte_lpm6 *lpm6;
 158         uint8_t portid;
 159 };
 160
 161 #define MAX_RX_QUEUE_PER_LCORE 16
 162 #define MAX_TX_QUEUE_PER_PORT 16
 163 struct lcore_queue_conf {
 164         uint16_t n_rx_queue;
 165         uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
 166         struct rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
 167         struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS];
 168 } __rte_cache_aligned;
 169 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
 170
 171 static const struct rte_eth_conf port_conf = {
 172         .rxmode = {
 173                 .max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE,
 174                 .split_hdr_size = 0,
 175                 .header_split   = 0, /**< Header Split disabled */
 176                 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
 177                 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
 178                 .jumbo_frame    = 1, /**< Jumbo Frame Support enabled */
 179                 .hw_strip_crc   = 0, /**< CRC stripped by hardware */
 180         },
 181         .txmode = {
 182                 .mq_mode = ETH_MQ_TX_NONE,
 183         },
 184 };
 185
 186 /*
 187  * IPv4 forwarding table
 188  */
 189 struct l3fwd_ipv4_route {
 190         uint32_t ip;
 191         uint8_t  depth;
 192         uint8_t  if_out;
 193 };
 194
 195 struct l3fwd_ipv4_route l3fwd_ipv4_route_array[] = {
 196                 {IPv4(100,10,0,0), 16, 0},
 197                 {IPv4(100,20,0,0), 16, 1},
 198                 {IPv4(100,30,0,0), 16, 2},
 199                 {IPv4(100,40,0,0), 16, 3},
 200                 {IPv4(100,50,0,0), 16, 4},
 201                 {IPv4(100,60,0,0), 16, 5},
 202                 {IPv4(100,70,0,0), 16, 6},
 203                 {IPv4(100,80,0,0), 16, 7},
 204 };
 205
 206 /*
 207  * IPv6 forwarding table
 208  */
 209
 210 struct l3fwd_ipv6_route {
 211         uint8_t ip[IPV6_ADDR_LEN];
 212         uint8_t depth;
 213         uint8_t if_out;
 214 };
 215
 216 static struct l3fwd_ipv6_route l3fwd_ipv6_route_array[] = {
 217         {{1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 0},
 218         {{2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 1},
 219         {{3,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 2},
 220         {{4,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 3},
 221         {{5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 4},
 222         {{6,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 5},
 223         {{7,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 6},
 224         {{8,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 7},
 225 };
 226
 227 #define LPM_MAX_RULES         1024
 228 #define LPM6_MAX_RULES         1024
 229 #define LPM6_NUMBER_TBL8S (1 << 16)
 230
 231 struct rte_lpm6_config lpm6_config = {
 232                 .max_rules = LPM6_MAX_RULES,
 233                 .number_tbl8s = LPM6_NUMBER_TBL8S,
 234                 .flags = 0
 235 };
 236
 237 static struct rte_mempool *socket_direct_pool[RTE_MAX_NUMA_NODES];
 238 static struct rte_mempool *socket_indirect_pool[RTE_MAX_NUMA_NODES];
 239 static struct rte_lpm *socket_lpm[RTE_MAX_NUMA_NODES];
 240 static struct rte_lpm6 *socket_lpm6[RTE_MAX_NUMA_NODES];
 241
 242 /* Send burst of packets on an output interface */
 243 static inline int
 244 send_burst(struct lcore_queue_conf *qconf, uint16_t n, uint8_t port)
 245 {
 246         struct rte_mbuf **m_table;
 247         int ret;
 248         uint16_t queueid;
 249
 250         queueid = qconf->tx_queue_id[port];
 251         m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table;
 252
 253         ret = rte_eth_tx_burst(port, queueid, m_table, n);
 254         if (unlikely(ret < n)) {
 255                 do {
 256                         rte_pktmbuf_free(m_table[ret]);
 257                 } while (++ret < n);
 258         }
 259
 260         return 0;
 261 }
 262
 263 static inline void
 264 l3fwd_simple_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf,
 265                 uint8_t queueid, uint8_t port_in)
 266 {
 267         struct rx_queue *rxq;
 268         uint32_t i, len, next_hop_ipv4;
 269         uint8_t next_hop_ipv6, port_out, ipv6;
 270         int32_t len2;
 271
 272         ipv6 = 0;
 273         rxq = &qconf->rx_queue_list[queueid];
 274
 275         /* by default, send everything back to the source port */
 276         port_out = port_in;
 277
 278         /* Remove the Ethernet header and trailer from the input packet */
 279         rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr));
 280
 281         /* Build transmission burst */
 282         len = qconf->tx_mbufs[port_out].len;
 283
 284         /* if this is an IPv4 packet */
 285         if (RTE_ETH_IS_IPV4_HDR(m->packet_type)) {
 286                 struct ipv4_hdr *ip_hdr;
 287                 uint32_t ip_dst;
 288                 /* Read the lookup key (i.e. ip_dst) from the input packet */
 289                 ip_hdr = rte_pktmbuf_mtod(m, struct ipv4_hdr *);
 290                 ip_dst = rte_be_to_cpu_32(ip_hdr->dst_addr);
 291
 292                 /* Find destination port */
 293                 if (rte_lpm_lookup(rxq->lpm, ip_dst, &next_hop_ipv4) == 0 &&
 294                                 (enabled_port_mask & 1 << next_hop_ipv4) != 0) {
 295                         port_out = next_hop_ipv4;
 296
 297                         /* Build transmission burst for new port */
 298                         len = qconf->tx_mbufs[port_out].len;
 299                 }
 300
 301                 /* if we don't need to do any fragmentation */
 302                 if (likely (IPV4_MTU_DEFAULT >= m->pkt_len)) {
 303                         qconf->tx_mbufs[port_out].m_table[len] = m;
 304                         len2 = 1;
 305                 } else {
 306                         len2 = rte_ipv4_fragment_packet(m,
 307                                 &qconf->tx_mbufs[port_out].m_table[len],
 308                                 (uint16_t)(MBUF_TABLE_SIZE - len),
 309                                 IPV4_MTU_DEFAULT,
 310                                 rxq->direct_pool, rxq->indirect_pool);
 311
 312                         /* Free input packet */
 313                         rte_pktmbuf_free(m);
 314
 315                         /* If we fail to fragment the packet */
 316                         if (unlikely (len2 < 0))
 317                                 return;
 318                 }
 319         } else if (RTE_ETH_IS_IPV6_HDR(m->packet_type)) {
 320                 /* if this is an IPv6 packet */
 321                 struct ipv6_hdr *ip_hdr;
 322
 323                 ipv6 = 1;
 324
 325                 /* Read the lookup key (i.e. ip_dst) from the input packet */
 326                 ip_hdr = rte_pktmbuf_mtod(m, struct ipv6_hdr *);
 327
 328                 /* Find destination port */
 329                 if (rte_lpm6_lookup(rxq->lpm6, ip_hdr->dst_addr, &next_hop_ipv6) == 0 &&
 330                                 (enabled_port_mask & 1 << next_hop_ipv6) != 0) {
 331                         port_out = next_hop_ipv6;
 332
 333                         /* Build transmission burst for new port */
 334                         len = qconf->tx_mbufs[port_out].len;
 335                 }
 336
 337                 /* if we don't need to do any fragmentation */
 338                 if (likely (IPV6_MTU_DEFAULT >= m->pkt_len)) {
 339                         qconf->tx_mbufs[port_out].m_table[len] = m;
 340                         len2 = 1;
 341                 } else {
 342                         len2 = rte_ipv6_fragment_packet(m,
 343                                 &qconf->tx_mbufs[port_out].m_table[len],
 344                                 (uint16_t)(MBUF_TABLE_SIZE - len),
 345                                 IPV6_MTU_DEFAULT,
 346                                 rxq->direct_pool, rxq->indirect_pool);
 347
 348                         /* Free input packet */
 349                         rte_pktmbuf_free(m);
 350
 351                         /* If we fail to fragment the packet */
 352                         if (unlikely (len2 < 0))
 353                                 return;
 354                 }
 355         }
 356         /* else, just forward the packet */
 357         else {
 358                 qconf->tx_mbufs[port_out].m_table[len] = m;
 359                 len2 = 1;
 360         }
 361
 362         for (i = len; i < len + len2; i ++) {
 363                 void *d_addr_bytes;
 364
 365                 m = qconf->tx_mbufs[port_out].m_table[i];
 366                 struct ether_hdr *eth_hdr = (struct ether_hdr *)
 367                         rte_pktmbuf_prepend(m, (uint16_t)sizeof(struct ether_hdr));
 368                 if (eth_hdr == NULL) {
 369                         rte_panic("No headroom in mbuf.\n");
 370                 }
 371
 372                 m->l2_len = sizeof(struct ether_hdr);
 373
 374                 /* 02:00:00:00:00:xx */
 375                 d_addr_bytes = &eth_hdr->d_addr.addr_bytes[0];
 376                 *((uint64_t *)d_addr_bytes) = 0x000000000002 + ((uint64_t)port_out << 40);
 377
 378                 /* src addr */
 379                 ether_addr_copy(&ports_eth_addr[port_out], &eth_hdr->s_addr);
 380                 if (ipv6)
 381                         eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv6);
 382                 else
 383                         eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
 384         }
 385
 386         len += len2;
 387
 388         if (likely(len < MAX_PKT_BURST)) {
 389                 qconf->tx_mbufs[port_out].len = (uint16_t)len;
 390                 return;
 391         }
 392
 393         /* Transmit packets */
 394         send_burst(qconf, (uint16_t)len, port_out);
 395         qconf->tx_mbufs[port_out].len = 0;
 396 }
 397
 398 /* main processing loop */
 399 static int
 400 main_loop(__attribute__((unused)) void *dummy)
 401 {
 402         struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
 403         unsigned lcore_id;
 404         uint64_t prev_tsc, diff_tsc, cur_tsc;
 405         int i, j, nb_rx;
 406         uint8_t portid;
 407         struct lcore_queue_conf *qconf;
 408         const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
 409
 410         prev_tsc = 0;
 411
 412         lcore_id = rte_lcore_id();
 413         qconf = &lcore_queue_conf[lcore_id];
 414
 415         if (qconf->n_rx_queue == 0) {
 416                 RTE_LOG(INFO, IP_FRAG, "lcore %u has nothing to do\n", lcore_id);
 417                 return 0;
 418         }
 419
 420         RTE_LOG(INFO, IP_FRAG, "entering main loop on lcore %u\n", lcore_id);
 421
 422         for (i = 0; i < qconf->n_rx_queue; i++) {
 423
 424                 portid = qconf->rx_queue_list[i].portid;
 425                 RTE_LOG(INFO, IP_FRAG, " -- lcoreid=%u portid=%d\n", lcore_id,
 426                                 (int) portid);
 427         }
 428
 429         while (1) {
 430
 431                 cur_tsc = rte_rdtsc();
 432
 433                 /*
 434                  * TX burst queue drain
 435                  */
 436                 diff_tsc = cur_tsc - prev_tsc;
 437                 if (unlikely(diff_tsc > drain_tsc)) {
 438
 439                         /*
 440                          * This could be optimized (use queueid instead of
 441                          * portid), but it is not called so often
 442                          */
 443                         for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
 444                                 if (qconf->tx_mbufs[portid].len == 0)
 445                                         continue;
 446                                 send_burst(&lcore_queue_conf[lcore_id],
 447                                            qconf->tx_mbufs[portid].len,
 448                                            portid);
 449                                 qconf->tx_mbufs[portid].len = 0;
 450                         }
 451
 452                         prev_tsc = cur_tsc;
 453                 }
 454
 455                 /*
 456                  * Read packet from RX queues
 457                  */
 458                 for (i = 0; i < qconf->n_rx_queue; i++) {
 459
 460                         portid = qconf->rx_queue_list[i].portid;
 461                         nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst,
 462                                                  MAX_PKT_BURST);
 463
 464                         /* Prefetch first packets */
 465                         for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
 466                                 rte_prefetch0(rte_pktmbuf_mtod(
 467                                                 pkts_burst[j], void *));
 468                         }
 469
 470                         /* Prefetch and forward already prefetched packets */
 471                         for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
 472                                 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
 473                                                 j + PREFETCH_OFFSET], void *));
 474                                 l3fwd_simple_forward(pkts_burst[j], qconf, i, portid);
 475                         }
 476
 477                         /* Forward remaining prefetched packets */
 478                         for (; j < nb_rx; j++) {
 479                                 l3fwd_simple_forward(pkts_burst[j], qconf, i, portid);
 480                         }
 481                 }
 482         }
 483 }
 484
 485 /* display usage */
 486 static void
 487 print_usage(const char *prgname)
 488 {
 489         printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n"
 490                "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
 491                "  -q NQ: number of queue (=ports) per lcore (default is 1)\n",
 492                prgname);
 493 }
 494
 495 static int
 496 parse_portmask(const char *portmask)
 497 {
 498         char *end = NULL;
 499         unsigned long pm;
 500
 501         /* parse hexadecimal string */
 502         pm = strtoul(portmask, &end, 16);
 503         if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
 504                 return -1;
 505
 506         if (pm == 0)
 507                 return -1;
 508
 509         return pm;
 510 }
 511
 512 static int
 513 parse_nqueue(const char *q_arg)
 514 {
 515         char *end = NULL;
 516         unsigned long n;
 517
 518         /* parse hexadecimal string */
 519         n = strtoul(q_arg, &end, 10);
 520         if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
 521                 return -1;
 522         if (n == 0)
 523                 return -1;
 524         if (n >= MAX_RX_QUEUE_PER_LCORE)
 525                 return -1;
 526
 527         return n;
 528 }
 529
 530 /* Parse the argument given in the command line of the application */
 531 static int
 532 parse_args(int argc, char **argv)
 533 {
 534         int opt, ret;
 535         char **argvopt;
 536         int option_index;
 537         char *prgname = argv[0];
 538         static struct option lgopts[] = {
 539                 {NULL, 0, 0, 0}
 540         };
 541
 542         argvopt = argv;
 543
 544         while ((opt = getopt_long(argc, argvopt, "p:q:",
 545                                   lgopts, &option_index)) != EOF) {
 546
 547                 switch (opt) {
 548                 /* portmask */
 549                 case 'p':
 550                         enabled_port_mask = parse_portmask(optarg);
 551                         if (enabled_port_mask < 0) {
 552                                 printf("invalid portmask\n");
 553                                 print_usage(prgname);
 554                                 return -1;
 555                         }
 556                         break;
 557
 558                 /* nqueue */
 559                 case 'q':
 560                         rx_queue_per_lcore = parse_nqueue(optarg);
 561                         if (rx_queue_per_lcore < 0) {
 562                                 printf("invalid queue number\n");
 563                                 print_usage(prgname);
 564                                 return -1;
 565                         }
 566                         break;
 567
 568                 /* long options */
 569                 case 0:
 570                         print_usage(prgname);
 571                         return -1;
 572
 573                 default:
 574                         print_usage(prgname);
 575                         return -1;
 576                 }
 577         }
 578
 579         if (enabled_port_mask == 0) {
 580                 printf("portmask not specified\n");
 581                 print_usage(prgname);
 582                 return -1;
 583         }
 584
 585         if (optind >= 0)
 586                 argv[optind-1] = prgname;
 587
 588         ret = optind-1;
 589         optind = 0; /* reset getopt lib */
 590         return ret;
 591 }
 592
 593 static void
 594 print_ethaddr(const char *name, struct ether_addr *eth_addr)
 595 {
 596         char buf[ETHER_ADDR_FMT_SIZE];
 597         ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr);
 598         printf("%s%s", name, buf);
 599 }
 600
 601 /* Check the link status of all ports in up to 9s, and print them finally */
 602 static void
 603 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
 604 {
 605 #define CHECK_INTERVAL 100 /* 100ms */
 606 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
 607         uint8_t portid, count, all_ports_up, print_flag = 0;
 608         struct rte_eth_link link;
 609
 610         printf("\nChecking link status");
 611         fflush(stdout);
 612         for (count = 0; count <= MAX_CHECK_TIME; count++) {
 613                 all_ports_up = 1;
 614                 for (portid = 0; portid < port_num; portid++) {
 615                         if ((port_mask & (1 << portid)) == 0)
 616                                 continue;
 617                         memset(&link, 0, sizeof(link));
 618                         rte_eth_link_get_nowait(portid, &link);
 619                         /* print link status if flag set */
 620                         if (print_flag == 1) {
 621                                 if (link.link_status)
 622                                         printf("Port %d Link Up - speed %u "
 623                                                 "Mbps - %s\n", (uint8_t)portid,
 624                                                 (unsigned)link.link_speed,
 625                                 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
 626                                         ("full-duplex") : ("half-duplex\n"));
 627                                 else
 628                                         printf("Port %d Link Down\n",
 629                                                         (uint8_t)portid);
 630                                 continue;
 631                         }
 632                         /* clear all_ports_up flag if any link down */
 633                         if (link.link_status == ETH_LINK_DOWN) {
 634                                 all_ports_up = 0;
 635                                 break;
 636                         }
 637                 }
 638                 /* after finally printing all link status, get out */
 639                 if (print_flag == 1)
 640                         break;
 641
 642                 if (all_ports_up == 0) {
 643                         printf(".");
 644                         fflush(stdout);
 645                         rte_delay_ms(CHECK_INTERVAL);
 646                 }
 647
 648                 /* set the print_flag if all ports up or timeout */
 649                 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
 650                         print_flag = 1;
 651                         printf("\ndone\n");
 652                 }
 653         }
 654 }
 655
 656 static int
 657 init_routing_table(void)
 658 {
 659         struct rte_lpm *lpm;
 660         struct rte_lpm6 *lpm6;
 661         int socket, ret;
 662         unsigned i;
 663
 664         for (socket = 0; socket < RTE_MAX_NUMA_NODES; socket++) {
 665                 if (socket_lpm[socket]) {
 666                         lpm = socket_lpm[socket];
 667                         /* populate the LPM table */
 668                         for (i = 0; i < RTE_DIM(l3fwd_ipv4_route_array); i++) {
 669                                 ret = rte_lpm_add(lpm,
 670                                         l3fwd_ipv4_route_array[i].ip,
 671                                         l3fwd_ipv4_route_array[i].depth,
 672                                         l3fwd_ipv4_route_array[i].if_out);
 673
 674                                 if (ret < 0) {
 675                                         RTE_LOG(ERR, IP_FRAG, "Unable to add entry %i to the l3fwd "
 676                                                 "LPM table\n", i);
 677                                         return -1;
 678                                 }
 679
 680                                 RTE_LOG(INFO, IP_FRAG, "Socket %i: adding route " IPv4_BYTES_FMT
 681                                                 "/%d (port %d)\n",
 682                                         socket,
 683                                         IPv4_BYTES(l3fwd_ipv4_route_array[i].ip),
 684                                         l3fwd_ipv4_route_array[i].depth,
 685                                         l3fwd_ipv4_route_array[i].if_out);
 686                         }
 687                 }
 688
 689                 if (socket_lpm6[socket]) {
 690                         lpm6 = socket_lpm6[socket];
 691                         /* populate the LPM6 table */
 692                         for (i = 0; i < RTE_DIM(l3fwd_ipv6_route_array); i++) {
 693                                 ret = rte_lpm6_add(lpm6,
 694                                         l3fwd_ipv6_route_array[i].ip,
 695                                         l3fwd_ipv6_route_array[i].depth,
 696                                         l3fwd_ipv6_route_array[i].if_out);
 697
 698                                 if (ret < 0) {
 699                                         RTE_LOG(ERR, IP_FRAG, "Unable to add entry %i to the l3fwd "
 700                                                 "LPM6 table\n", i);
 701                                         return -1;
 702                                 }
 703
 704                                 RTE_LOG(INFO, IP_FRAG, "Socket %i: adding route " IPv6_BYTES_FMT
 705                                                 "/%d (port %d)\n",
 706                                         socket,
 707                                         IPv6_BYTES(l3fwd_ipv6_route_array[i].ip),
 708                                         l3fwd_ipv6_route_array[i].depth,
 709                                         l3fwd_ipv6_route_array[i].if_out);
 710                         }
 711                 }
 712         }
 713         return 0;
 714 }
 715
 716 static int
 717 init_mem(void)
 718 {
 719         char buf[PATH_MAX];
 720         struct rte_mempool *mp;
 721         struct rte_lpm *lpm;
 722         struct rte_lpm6 *lpm6;
 723         struct rte_lpm_config lpm_config;
 724         int socket;
 725         unsigned lcore_id;
 726
 727         /* traverse through lcores and initialize structures on each socket */
 728
 729         for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
 730
 731                 if (rte_lcore_is_enabled(lcore_id) == 0)
 732                         continue;
 733
 734                 socket = rte_lcore_to_socket_id(lcore_id);
 735
 736                 if (socket == SOCKET_ID_ANY)
 737                         socket = 0;
 738
 739                 if (socket_direct_pool[socket] == NULL) {
 740                         RTE_LOG(INFO, IP_FRAG, "Creating direct mempool on socket %i\n",
 741                                         socket);
 742                         snprintf(buf, sizeof(buf), "pool_direct_%i", socket);
 743
 744                         mp = rte_pktmbuf_pool_create(buf, NB_MBUF, 32,
 745                                 0, RTE_MBUF_DEFAULT_BUF_SIZE, socket);
 746                         if (mp == NULL) {
 747                                 RTE_LOG(ERR, IP_FRAG, "Cannot create direct mempool\n");
 748                                 return -1;
 749                         }
 750                         socket_direct_pool[socket] = mp;
 751                 }
 752
 753                 if (socket_indirect_pool[socket] == NULL) {
 754                         RTE_LOG(INFO, IP_FRAG, "Creating indirect mempool on socket %i\n",
 755                                         socket);
 756                         snprintf(buf, sizeof(buf), "pool_indirect_%i", socket);
 757
 758                         mp = rte_pktmbuf_pool_create(buf, NB_MBUF, 32, 0, 0,
 759                                 socket);
 760                         if (mp == NULL) {
 761                                 RTE_LOG(ERR, IP_FRAG, "Cannot create indirect mempool\n");
 762                                 return -1;
 763                         }
 764                         socket_indirect_pool[socket] = mp;
 765                 }
 766
 767                 if (socket_lpm[socket] == NULL) {
 768                         RTE_LOG(INFO, IP_FRAG, "Creating LPM table on socket %i\n", socket);
 769                         snprintf(buf, sizeof(buf), "IP_FRAG_LPM_%i", socket);
 770
 771                         lpm_config.max_rules = LPM_MAX_RULES;
 772                         lpm_config.number_tbl8s = 256;
 773                         lpm_config.flags = 0;
 774
 775                         lpm = rte_lpm_create(buf, socket, &lpm_config);
 776                         if (lpm == NULL) {
 777                                 RTE_LOG(ERR, IP_FRAG, "Cannot create LPM table\n");
 778                                 return -1;
 779                         }
 780                         socket_lpm[socket] = lpm;
 781                 }
 782
 783                 if (socket_lpm6[socket] == NULL) {
 784                         RTE_LOG(INFO, IP_FRAG, "Creating LPM6 table on socket %i\n", socket);
 785                         snprintf(buf, sizeof(buf), "IP_FRAG_LPM_%i", socket);
 786
 787                         lpm6 = rte_lpm6_create(buf, socket, &lpm6_config);
 788                         if (lpm6 == NULL) {
 789                                 RTE_LOG(ERR, IP_FRAG, "Cannot create LPM table\n");
 790                                 return -1;
 791                         }
 792                         socket_lpm6[socket] = lpm6;
 793                 }
 794         }
 795
 796         return 0;
 797 }
 798
 799 int
 800 main(int argc, char **argv)
 801 {
 802         struct lcore_queue_conf *qconf;
 803         struct rte_eth_dev_info dev_info;
 804         struct rte_eth_txconf *txconf;
 805         struct rx_queue *rxq;
 806         int socket, ret;
 807         unsigned nb_ports;
 808         uint16_t queueid = 0;
 809         unsigned lcore_id = 0, rx_lcore_id = 0;
 810         uint32_t n_tx_queue, nb_lcores;
 811         uint8_t portid;
 812
 813         /* init EAL */
 814         ret = rte_eal_init(argc, argv);
 815         if (ret < 0)
 816                 rte_exit(EXIT_FAILURE, "rte_eal_init failed");
 817         argc -= ret;
 818         argv += ret;
 819
 820         /* parse application arguments (after the EAL ones) */
 821         ret = parse_args(argc, argv);
 822         if (ret < 0)
 823                 rte_exit(EXIT_FAILURE, "Invalid arguments");
 824
 825         nb_ports = rte_eth_dev_count();
 826         if (nb_ports == 0)
 827                 rte_exit(EXIT_FAILURE, "No ports found!\n");
 828
 829         nb_lcores = rte_lcore_count();
 830
 831         /* initialize structures (mempools, lpm etc.) */
 832         if (init_mem() < 0)
 833                 rte_panic("Cannot initialize memory structures!\n");
 834
 835         /* check if portmask has non-existent ports */
 836         if (enabled_port_mask & ~(RTE_LEN2MASK(nb_ports, unsigned)))
 837                 rte_exit(EXIT_FAILURE, "Non-existent ports in portmask!\n");
 838
 839         /* initialize all ports */
 840         for (portid = 0; portid < nb_ports; portid++) {
 841                 /* skip ports that are not enabled */
 842                 if ((enabled_port_mask & (1 << portid)) == 0) {
 843                         printf("Skipping disabled port %d\n", portid);
 844                         continue;
 845                 }
 846
 847                 qconf = &lcore_queue_conf[rx_lcore_id];
 848
 849                 /* get the lcore_id for this port */
 850                 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
 851                        qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) {
 852
 853                         rx_lcore_id ++;
 854                         if (rx_lcore_id >= RTE_MAX_LCORE)
 855                                 rte_exit(EXIT_FAILURE, "Not enough cores\n");
 856
 857                         qconf = &lcore_queue_conf[rx_lcore_id];
 858                 }
 859
 860                 socket = (int) rte_lcore_to_socket_id(rx_lcore_id);
 861                 if (socket == SOCKET_ID_ANY)
 862                         socket = 0;
 863
 864                 rxq = &qconf->rx_queue_list[qconf->n_rx_queue];
 865                 rxq->portid = portid;
 866                 rxq->direct_pool = socket_direct_pool[socket];
 867                 rxq->indirect_pool = socket_indirect_pool[socket];
 868                 rxq->lpm = socket_lpm[socket];
 869                 rxq->lpm6 = socket_lpm6[socket];
 870                 qconf->n_rx_queue++;
 871
 872                 /* init port */
 873                 printf("Initializing port %d on lcore %u...", portid,
 874                        rx_lcore_id);
 875                 fflush(stdout);
 876
 877                 n_tx_queue = nb_lcores;
 878                 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
 879                         n_tx_queue = MAX_TX_QUEUE_PER_PORT;
 880                 ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue,
 881                                             &port_conf);
 882                 if (ret < 0) {
 883                         printf("\n");
 884                         rte_exit(EXIT_FAILURE, "Cannot configure device: "
 885                                 "err=%d, port=%d\n",
 886                                 ret, portid);
 887                 }
 888
 889                 /* init one RX queue */
 890                 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
 891                                              socket, NULL,
 892                                              socket_direct_pool[socket]);
 893                 if (ret < 0) {
 894                         printf("\n");
 895                         rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: "
 896                                 "err=%d, port=%d\n",
 897                                 ret, portid);
 898                 }
 899
 900                 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
 901                 print_ethaddr(" Address:", &ports_eth_addr[portid]);
 902                 printf("\n");
 903
 904                 /* init one TX queue per couple (lcore,port) */
 905                 queueid = 0;
 906                 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
 907                         if (rte_lcore_is_enabled(lcore_id) == 0)
 908                                 continue;
 909
 910                         socket = (int) rte_lcore_to_socket_id(lcore_id);
 911                         printf("txq=%u,%d ", lcore_id, queueid);
 912                         fflush(stdout);
 913
 914                         rte_eth_dev_info_get(portid, &dev_info);
 915                         txconf = &dev_info.default_txconf;
 916                         txconf->txq_flags = 0;
 917                         ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
 918                                                      socket, txconf);
 919                         if (ret < 0) {
 920                                 printf("\n");
 921                                 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: "
 922                                         "err=%d, port=%d\n", ret, portid);
 923                         }
 924
 925                         qconf = &lcore_queue_conf[lcore_id];
 926                         qconf->tx_queue_id[portid] = queueid;
 927                         queueid++;
 928                 }
 929
 930                 printf("\n");
 931         }
 932
 933         printf("\n");
 934
 935         /* start ports */
 936         for (portid = 0; portid < nb_ports; portid++) {
 937                 if ((enabled_port_mask & (1 << portid)) == 0) {
 938                         continue;
 939                 }
 940                 /* Start device */
 941                 ret = rte_eth_dev_start(portid);
 942                 if (ret < 0)
 943                         rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
 944                                 ret, portid);
 945
 946                 rte_eth_promiscuous_enable(portid);
 947         }
 948
 949         if (init_routing_table() < 0)
 950                 rte_exit(EXIT_FAILURE, "Cannot init routing table\n");
 951
 952         check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
 953
 954         /* launch per-lcore init on every lcore */
 955         rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
 956         RTE_LCORE_FOREACH_SLAVE(lcore_id) {
 957                 if (rte_eal_wait_lcore(lcore_id) < 0)
 958                         return -1;
 959         }
 960
 961         return 0;
 962 }