examples/ip_reassembly/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 <string.h>
  40 #include <sys/queue.h>
  41 #include <stdarg.h>
  42 #include <errno.h>
  43 #include <getopt.h>
  44 #include <signal.h>
  45 #include <sys/param.h>
  46
  47 #include <rte_common.h>
  48 #include <rte_byteorder.h>
  49 #include <rte_log.h>
  50 #include <rte_memory.h>
  51 #include <rte_memcpy.h>
  52 #include <rte_memzone.h>
  53 #include <rte_tailq.h>
  54 #include <rte_eal.h>
  55 #include <rte_per_lcore.h>
  56 #include <rte_launch.h>
  57 #include <rte_atomic.h>
  58 #include <rte_cycles.h>
  59 #include <rte_prefetch.h>
  60 #include <rte_lcore.h>
  61 #include <rte_per_lcore.h>
  62 #include <rte_branch_prediction.h>
  63 #include <rte_interrupts.h>
  64 #include <rte_pci.h>
  65 #include <rte_random.h>
  66 #include <rte_debug.h>
  67 #include <rte_ether.h>
  68 #include <rte_ethdev.h>
  69 #include <rte_ring.h>
  70 #include <rte_mempool.h>
  71 #include <rte_mbuf.h>
  72 #include <rte_malloc.h>
  73 #include <rte_ip.h>
  74 #include <rte_tcp.h>
  75 #include <rte_udp.h>
  76 #include <rte_string_fns.h>
  77 #include <rte_lpm.h>
  78 #include <rte_lpm6.h>
  79
  80 #include <rte_ip_frag.h>
  81
  82 #include "main.h"
  83
  84 #define MAX_PKT_BURST 32
  85
  86
  87 #define RTE_LOGTYPE_IP_RSMBL RTE_LOGTYPE_USER1
  88
  89 #define MAX_JUMBO_PKT_LEN  9600
  90
  91 #define BUF_SIZE        2048
  92 #define MBUF_SIZE       \
  93         (BUF_SIZE + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM)
  94
  95 #define NB_MBUF 8192
  96
  97 /* allow max jumbo frame 9.5 KB */
  98 #define JUMBO_FRAME_MAX_SIZE    0x2600
  99
 100 #define MAX_FLOW_NUM    UINT16_MAX
 101 #define MIN_FLOW_NUM    1
 102 #define DEF_FLOW_NUM    0x1000
 103
 104 /* TTL numbers are in ms. */
 105 #define MAX_FLOW_TTL    (3600 * MS_PER_S)
 106 #define MIN_FLOW_TTL    1
 107 #define DEF_FLOW_TTL    MS_PER_S
 108
 109 #define MAX_FRAG_NUM RTE_LIBRTE_IP_FRAG_MAX_FRAG
 110
 111 /* Should be power of two. */
 112 #define IP_FRAG_TBL_BUCKET_ENTRIES      16
 113
 114 static uint32_t max_flow_num = DEF_FLOW_NUM;
 115 static uint32_t max_flow_ttl = DEF_FLOW_TTL;
 116
 117 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */
 118
 119 #define NB_SOCKETS 8
 120
 121 /* Configure how many packets ahead to prefetch, when reading packets */
 122 #define PREFETCH_OFFSET 3
 123
 124 /*
 125  * Configurable number of RX/TX ring descriptors
 126  */
 127 #define RTE_TEST_RX_DESC_DEFAULT 128
 128 #define RTE_TEST_TX_DESC_DEFAULT 512
 129
 130 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT;
 131 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT;
 132
 133 /* ethernet addresses of ports */
 134 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS];
 135
 136 #ifndef IPv4_BYTES
 137 #define IPv4_BYTES_FMT "%" PRIu8 ".%" PRIu8 ".%" PRIu8 ".%" PRIu8
 138 #define IPv4_BYTES(addr) \
 139                 (uint8_t) (((addr) >> 24) & 0xFF),\
 140                 (uint8_t) (((addr) >> 16) & 0xFF),\
 141                 (uint8_t) (((addr) >> 8) & 0xFF),\
 142                 (uint8_t) ((addr) & 0xFF)
 143 #endif
 144
 145 #ifndef IPv6_BYTES
 146 #define IPv6_BYTES_FMT "%02x%02x:%02x%02x:%02x%02x:%02x%02x:"\
 147                        "%02x%02x:%02x%02x:%02x%02x:%02x%02x"
 148 #define IPv6_BYTES(addr) \
 149         addr[0],  addr[1], addr[2],  addr[3], \
 150         addr[4],  addr[5], addr[6],  addr[7], \
 151         addr[8],  addr[9], addr[10], addr[11],\
 152         addr[12], addr[13],addr[14], addr[15]
 153 #endif
 154
 155 #define IPV6_ADDR_LEN 16
 156
 157 /* mask of enabled ports */
 158 static uint32_t enabled_port_mask = 0;
 159
 160 static int rx_queue_per_lcore = 1;
 161
 162 struct mbuf_table {
 163         uint32_t len;
 164         uint32_t head;
 165         uint32_t tail;
 166         struct rte_mbuf *m_table[0];
 167 };
 168
 169 struct rx_queue {
 170         struct rte_ip_frag_tbl *frag_tbl;
 171         struct rte_mempool *pool;
 172         struct rte_lpm *lpm;
 173         struct rte_lpm6 *lpm6;
 174         uint8_t portid;
 175 };
 176
 177 struct tx_lcore_stat {
 178         uint64_t call;
 179         uint64_t drop;
 180         uint64_t queue;
 181         uint64_t send;
 182 };
 183
 184 #define MAX_RX_QUEUE_PER_LCORE 16
 185 #define MAX_TX_QUEUE_PER_PORT 16
 186 #define MAX_RX_QUEUE_PER_PORT 128
 187
 188 struct lcore_queue_conf {
 189         uint16_t n_rx_queue;
 190         struct rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE];
 191         uint16_t tx_queue_id[RTE_MAX_ETHPORTS];
 192         struct rte_ip_frag_death_row death_row;
 193         struct mbuf_table *tx_mbufs[RTE_MAX_ETHPORTS];
 194         struct tx_lcore_stat tx_stat;
 195 } __rte_cache_aligned;
 196 static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE];
 197
 198 static struct rte_eth_conf port_conf = {
 199         .rxmode = {
 200                 .mq_mode        = ETH_MQ_RX_RSS,
 201                 .max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE,
 202                 .split_hdr_size = 0,
 203                 .header_split   = 0, /**< Header Split disabled */
 204                 .hw_ip_checksum = 1, /**< IP checksum offload enabled */
 205                 .hw_vlan_filter = 0, /**< VLAN filtering disabled */
 206                 .jumbo_frame    = 1, /**< Jumbo Frame Support disabled */
 207                 .hw_strip_crc   = 0, /**< CRC stripped by hardware */
 208         },
 209         .rx_adv_conf = {
 210                         .rss_conf = {
 211                                 .rss_key = NULL,
 212                                 .rss_hf = ETH_RSS_IP,
 213                 },
 214         },
 215         .txmode = {
 216                 .mq_mode = ETH_MQ_TX_NONE,
 217         },
 218 };
 219
 220 /*
 221  * IPv4 forwarding table
 222  */
 223 struct l3fwd_ipv4_route {
 224         uint32_t ip;
 225         uint8_t  depth;
 226         uint8_t  if_out;
 227 };
 228
 229 struct l3fwd_ipv4_route l3fwd_ipv4_route_array[] = {
 230                 {IPv4(100,10,0,0), 16, 0},
 231                 {IPv4(100,20,0,0), 16, 1},
 232                 {IPv4(100,30,0,0), 16, 2},
 233                 {IPv4(100,40,0,0), 16, 3},
 234                 {IPv4(100,50,0,0), 16, 4},
 235                 {IPv4(100,60,0,0), 16, 5},
 236                 {IPv4(100,70,0,0), 16, 6},
 237                 {IPv4(100,80,0,0), 16, 7},
 238 };
 239
 240 /*
 241  * IPv6 forwarding table
 242  */
 243
 244 struct l3fwd_ipv6_route {
 245         uint8_t ip[IPV6_ADDR_LEN];
 246         uint8_t depth;
 247         uint8_t if_out;
 248 };
 249
 250 static struct l3fwd_ipv6_route l3fwd_ipv6_route_array[] = {
 251         {{1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 0},
 252         {{2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 1},
 253         {{3,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 2},
 254         {{4,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 3},
 255         {{5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 4},
 256         {{6,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 5},
 257         {{7,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 6},
 258         {{8,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 7},
 259 };
 260
 261 #define LPM_MAX_RULES         1024
 262 #define LPM6_MAX_RULES         1024
 263 #define LPM6_NUMBER_TBL8S (1 << 16)
 264
 265 struct rte_lpm6_config lpm6_config = {
 266                 .max_rules = LPM6_MAX_RULES,
 267                 .number_tbl8s = LPM6_NUMBER_TBL8S,
 268                 .flags = 0
 269 };
 270
 271 static struct rte_lpm *socket_lpm[RTE_MAX_NUMA_NODES];
 272 static struct rte_lpm6 *socket_lpm6[RTE_MAX_NUMA_NODES];
 273
 274 #ifdef RTE_LIBRTE_IP_FRAG_TBL_STAT
 275 #define TX_LCORE_STAT_UPDATE(s, f, v)   ((s)->f += (v))
 276 #else
 277 #define TX_LCORE_STAT_UPDATE(s, f, v)   do {} while (0)
 278 #endif /* RTE_LIBRTE_IP_FRAG_TBL_STAT */
 279
 280 /*
 281  * If number of queued packets reached given threahold, then
 282  * send burst of packets on an output interface.
 283  */
 284 static inline uint32_t
 285 send_burst(struct lcore_queue_conf *qconf, uint32_t thresh, uint8_t port)
 286 {
 287         uint32_t fill, len, k, n;
 288         struct mbuf_table *txmb;
 289
 290         txmb = qconf->tx_mbufs[port];
 291         len = txmb->len;
 292
 293         if ((int32_t)(fill = txmb->head - txmb->tail) < 0)
 294                 fill += len;
 295
 296         if (fill >= thresh) {
 297                 n = RTE_MIN(len - txmb->tail, fill);
 298
 299                 k = rte_eth_tx_burst(port, qconf->tx_queue_id[port],
 300                         txmb->m_table + txmb->tail, (uint16_t)n);
 301
 302                 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, call, 1);
 303                 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, send, k);
 304
 305                 fill -= k;
 306                 if ((txmb->tail += k) == len)
 307                         txmb->tail = 0;
 308         }
 309
 310         return (fill);
 311 }
 312
 313 /* Enqueue a single packet, and send burst if queue is filled */
 314 static inline int
 315 send_single_packet(struct rte_mbuf *m, uint8_t port)
 316 {
 317         uint32_t fill, lcore_id, len;
 318         struct lcore_queue_conf *qconf;
 319         struct mbuf_table *txmb;
 320
 321         lcore_id = rte_lcore_id();
 322         qconf = &lcore_queue_conf[lcore_id];
 323
 324         txmb = qconf->tx_mbufs[port];
 325         len = txmb->len;
 326
 327         fill = send_burst(qconf, MAX_PKT_BURST, port);
 328
 329         if (fill == len - 1) {
 330                 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, drop, 1);
 331                 rte_pktmbuf_free(txmb->m_table[txmb->tail]);
 332                 if (++txmb->tail == len)
 333                         txmb->tail = 0;
 334         }
 335
 336         TX_LCORE_STAT_UPDATE(&qconf->tx_stat, queue, 1);
 337         txmb->m_table[txmb->head] = m;
 338         if(++txmb->head == len)
 339                 txmb->head = 0;
 340
 341         return (0);
 342 }
 343
 344 static inline void
 345 reassemble(struct rte_mbuf *m, uint8_t portid, uint32_t queue,
 346         struct lcore_queue_conf *qconf, uint64_t tms)
 347 {
 348         struct ether_hdr *eth_hdr;
 349         struct rte_ip_frag_tbl *tbl;
 350         struct rte_ip_frag_death_row *dr;
 351         struct rx_queue *rxq;
 352         void *d_addr_bytes;
 353         uint8_t next_hop, dst_port;
 354
 355         rxq = &qconf->rx_queue_list[queue];
 356
 357         eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
 358
 359         dst_port = portid;
 360
 361         /* if packet is IPv4 */
 362         if (m->ol_flags & (PKT_RX_IPV4_HDR)) {
 363                 struct ipv4_hdr *ip_hdr;
 364                 uint32_t ip_dst;
 365
 366                 ip_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
 367
 368                  /* if it is a fragmented packet, then try to reassemble. */
 369                 if (rte_ipv4_frag_pkt_is_fragmented(ip_hdr)) {
 370                         struct rte_mbuf *mo;
 371
 372                         tbl = rxq->frag_tbl;
 373                         dr = &qconf->death_row;
 374
 375                         /* prepare mbuf: setup l2_len/l3_len. */
 376                         m->l2_len = sizeof(*eth_hdr);
 377                         m->l3_len = sizeof(*ip_hdr);
 378
 379                         /* process this fragment. */
 380                         mo = rte_ipv4_frag_reassemble_packet(tbl, dr, m, tms, ip_hdr);
 381                         if (mo == NULL)
 382                                 /* no packet to send out. */
 383                                 return;
 384
 385                         /* we have our packet reassembled. */
 386                         if (mo != m) {
 387                                 m = mo;
 388                                 eth_hdr = rte_pktmbuf_mtod(m,
 389                                         struct ether_hdr *);
 390                                 ip_hdr = (struct ipv4_hdr *)(eth_hdr + 1);
 391                         }
 392                 }
 393                 ip_dst = rte_be_to_cpu_32(ip_hdr->dst_addr);
 394
 395                 /* Find destination port */
 396                 if (rte_lpm_lookup(rxq->lpm, ip_dst, &next_hop) == 0 &&
 397                                 (enabled_port_mask & 1 << next_hop) != 0) {
 398                         dst_port = next_hop;
 399                 }
 400
 401                 eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4);
 402         }
 403         /* if packet is IPv6 */
 404         else if (m->ol_flags & (PKT_RX_IPV6_HDR | PKT_RX_IPV6_HDR_EXT)) {
 405                 struct ipv6_extension_fragment *frag_hdr;
 406                 struct ipv6_hdr *ip_hdr;
 407
 408                 ip_hdr = (struct ipv6_hdr *)(eth_hdr + 1);
 409
 410                 frag_hdr = rte_ipv6_frag_get_ipv6_fragment_header(ip_hdr);
 411
 412                 if (frag_hdr != NULL) {
 413                         struct rte_mbuf *mo;
 414
 415                         tbl = rxq->frag_tbl;
 416                         dr  = &qconf->death_row;
 417
 418                         /* prepare mbuf: setup l2_len/l3_len. */
 419                         m->l2_len = sizeof(*eth_hdr);
 420                         m->l3_len = sizeof(*ip_hdr) + sizeof(*frag_hdr);
 421
 422                         mo = rte_ipv6_frag_reassemble_packet(tbl, dr, m, tms, ip_hdr, frag_hdr);
 423                         if (mo == NULL)
 424                                 return;
 425
 426                         if (mo != m) {
 427                                 m = mo;
 428                                 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *);
 429                                 ip_hdr = (struct ipv6_hdr *)(eth_hdr + 1);
 430                         }
 431                 }
 432
 433                 /* Find destination port */
 434                 if (rte_lpm6_lookup(rxq->lpm6, ip_hdr->dst_addr, &next_hop) == 0 &&
 435                                 (enabled_port_mask & 1 << next_hop) != 0) {
 436                         dst_port = next_hop;
 437                 }
 438
 439                 eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv6);
 440         }
 441         /* if packet wasn't IPv4 or IPv6, it's forwarded to the port it came from */
 442
 443         /* 02:00:00:00:00:xx */
 444         d_addr_bytes = &eth_hdr->d_addr.addr_bytes[0];
 445         *((uint64_t *)d_addr_bytes) = 0x000000000002 + ((uint64_t)dst_port << 40);
 446
 447         /* src addr */
 448         ether_addr_copy(&ports_eth_addr[dst_port], &eth_hdr->s_addr);
 449
 450         send_single_packet(m, dst_port);
 451 }
 452
 453 /* main processing loop */
 454 static int
 455 main_loop(__attribute__((unused)) void *dummy)
 456 {
 457         struct rte_mbuf *pkts_burst[MAX_PKT_BURST];
 458         unsigned lcore_id;
 459         uint64_t diff_tsc, cur_tsc, prev_tsc;
 460         int i, j, nb_rx;
 461         uint8_t portid;
 462         struct lcore_queue_conf *qconf;
 463         const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US;
 464
 465         prev_tsc = 0;
 466
 467         lcore_id = rte_lcore_id();
 468         qconf = &lcore_queue_conf[lcore_id];
 469
 470         if (qconf->n_rx_queue == 0) {
 471                 RTE_LOG(INFO, IP_RSMBL, "lcore %u has nothing to do\n", lcore_id);
 472                 return 0;
 473         }
 474
 475         RTE_LOG(INFO, IP_RSMBL, "entering main loop on lcore %u\n", lcore_id);
 476
 477         for (i = 0; i < qconf->n_rx_queue; i++) {
 478
 479                 portid = qconf->rx_queue_list[i].portid;
 480                 RTE_LOG(INFO, IP_RSMBL, " -- lcoreid=%u portid=%hhu\n", lcore_id,
 481                         portid);
 482         }
 483
 484         while (1) {
 485
 486                 cur_tsc = rte_rdtsc();
 487
 488                 /*
 489                  * TX burst queue drain
 490                  */
 491                 diff_tsc = cur_tsc - prev_tsc;
 492                 if (unlikely(diff_tsc > drain_tsc)) {
 493
 494                         /*
 495                          * This could be optimized (use queueid instead of
 496                          * portid), but it is not called so often
 497                          */
 498                         for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) {
 499                                 if ((enabled_port_mask & (1 << portid)) != 0)
 500                                         send_burst(qconf, 1, portid);
 501                         }
 502
 503                         prev_tsc = cur_tsc;
 504                 }
 505
 506                 /*
 507                  * Read packet from RX queues
 508                  */
 509                 for (i = 0; i < qconf->n_rx_queue; ++i) {
 510
 511                         portid = qconf->rx_queue_list[i].portid;
 512
 513                         nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst,
 514                                 MAX_PKT_BURST);
 515
 516                         /* Prefetch first packets */
 517                         for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) {
 518                                 rte_prefetch0(rte_pktmbuf_mtod(
 519                                                 pkts_burst[j], void *));
 520                         }
 521
 522                         /* Prefetch and forward already prefetched packets */
 523                         for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) {
 524                                 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[
 525                                         j + PREFETCH_OFFSET], void *));
 526                                 reassemble(pkts_burst[j], portid,
 527                                         i, qconf, cur_tsc);
 528                         }
 529
 530                         /* Forward remaining prefetched packets */
 531                         for (; j < nb_rx; j++) {
 532                                 reassemble(pkts_burst[j], portid,
 533                                         i, qconf, cur_tsc);
 534                         }
 535
 536                         rte_ip_frag_free_death_row(&qconf->death_row,
 537                                 PREFETCH_OFFSET);
 538                 }
 539         }
 540 }
 541
 542 /* display usage */
 543 static void
 544 print_usage(const char *prgname)
 545 {
 546         printf("%s [EAL options] -- -p PORTMASK [-q NQ]"
 547                 "  [--max-pkt-len PKTLEN]"
 548                 "  [--maxflows=<flows>]  [--flowttl=<ttl>[(s|ms)]]\n"
 549                 "  -p PORTMASK: hexadecimal bitmask of ports to configure\n"
 550                 "  -q NQ: number of RX queues per lcore\n"
 551                 "  --maxflows=<flows>: optional, maximum number of flows "
 552                 "supported\n"
 553                 "  --flowttl=<ttl>[(s|ms)]: optional, maximum TTL for each "
 554                 "flow\n",
 555                 prgname);
 556 }
 557
 558 static uint32_t
 559 parse_flow_num(const char *str, uint32_t min, uint32_t max, uint32_t *val)
 560 {
 561         char *end;
 562         uint64_t v;
 563
 564         /* parse decimal string */
 565         errno = 0;
 566         v = strtoul(str, &end, 10);
 567         if (errno != 0 || *end != '\0')
 568                 return (-EINVAL);
 569
 570         if (v < min || v > max)
 571                 return (-EINVAL);
 572
 573         *val = (uint32_t)v;
 574         return (0);
 575 }
 576
 577 static int
 578 parse_flow_ttl(const char *str, uint32_t min, uint32_t max, uint32_t *val)
 579 {
 580         char *end;
 581         uint64_t v;
 582
 583         static const char frmt_sec[] = "s";
 584         static const char frmt_msec[] = "ms";
 585
 586         /* parse decimal string */
 587         errno = 0;
 588         v = strtoul(str, &end, 10);
 589         if (errno != 0)
 590                 return (-EINVAL);
 591
 592         if (*end != '\0') {
 593                 if (strncmp(frmt_sec, end, sizeof(frmt_sec)) == 0)
 594                         v *= MS_PER_S;
 595                 else if (strncmp(frmt_msec, end, sizeof (frmt_msec)) != 0)
 596                         return (-EINVAL);
 597         }
 598
 599         if (v < min || v > max)
 600                 return (-EINVAL);
 601
 602         *val = (uint32_t)v;
 603         return (0);
 604 }
 605
 606 static int
 607 parse_portmask(const char *portmask)
 608 {
 609         char *end = NULL;
 610         unsigned long pm;
 611
 612         /* parse hexadecimal string */
 613         pm = strtoul(portmask, &end, 16);
 614         if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0'))
 615                 return -1;
 616
 617         if (pm == 0)
 618                 return -1;
 619
 620         return pm;
 621 }
 622
 623 static int
 624 parse_nqueue(const char *q_arg)
 625 {
 626         char *end = NULL;
 627         unsigned long n;
 628
 629         printf("%p\n", q_arg);
 630
 631         /* parse hexadecimal string */
 632         n = strtoul(q_arg, &end, 10);
 633         if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0'))
 634                 return -1;
 635         if (n == 0)
 636                 return -1;
 637         if (n >= MAX_RX_QUEUE_PER_LCORE)
 638                 return -1;
 639
 640         return n;
 641 }
 642
 643 /* Parse the argument given in the command line of the application */
 644 static int
 645 parse_args(int argc, char **argv)
 646 {
 647         int opt, ret;
 648         char **argvopt;
 649         int option_index;
 650         char *prgname = argv[0];
 651         static struct option lgopts[] = {
 652                 {"max-pkt-len", 1, 0, 0},
 653                 {"maxflows", 1, 0, 0},
 654                 {"flowttl", 1, 0, 0},
 655                 {NULL, 0, 0, 0}
 656         };
 657
 658         argvopt = argv;
 659
 660         while ((opt = getopt_long(argc, argvopt, "p:q:",
 661                                 lgopts, &option_index)) != EOF) {
 662
 663                 switch (opt) {
 664                 /* portmask */
 665                 case 'p':
 666                         enabled_port_mask = parse_portmask(optarg);
 667                         if (enabled_port_mask == 0) {
 668                                 printf("invalid portmask\n");
 669                                 print_usage(prgname);
 670                                 return -1;
 671                         }
 672                         break;
 673
 674                 /* nqueue */
 675                 case 'q':
 676                         rx_queue_per_lcore = parse_nqueue(optarg);
 677                         if (rx_queue_per_lcore < 0) {
 678                                 printf("invalid queue number\n");
 679                                 print_usage(prgname);
 680                                 return -1;
 681                         }
 682                         break;
 683
 684                 /* long options */
 685                 case 0:
 686                         if (!strncmp(lgopts[option_index].name,
 687                                         "maxflows", 8)) {
 688                                 if ((ret = parse_flow_num(optarg, MIN_FLOW_NUM,
 689                                                 MAX_FLOW_NUM,
 690                                                 &max_flow_num)) != 0) {
 691                                         printf("invalid value: \"%s\" for "
 692                                                 "parameter %s\n",
 693                                                 optarg,
 694                                                 lgopts[option_index].name);
 695                                         print_usage(prgname);
 696                                         return (ret);
 697                                 }
 698                         }
 699
 700                         if (!strncmp(lgopts[option_index].name, "flowttl", 7)) {
 701                                 if ((ret = parse_flow_ttl(optarg, MIN_FLOW_TTL,
 702                                                 MAX_FLOW_TTL,
 703                                                 &max_flow_ttl)) != 0) {
 704                                         printf("invalid value: \"%s\" for "
 705                                                 "parameter %s\n",
 706                                                 optarg,
 707                                                 lgopts[option_index].name);
 708                                         print_usage(prgname);
 709                                         return (ret);
 710                                 }
 711                         }
 712
 713                         break;
 714
 715                 default:
 716                         print_usage(prgname);
 717                         return -1;
 718                 }
 719         }
 720
 721         if (optind >= 0)
 722                 argv[optind-1] = prgname;
 723
 724         ret = optind-1;
 725         optind = 0; /* reset getopt lib */
 726         return ret;
 727 }
 728
 729 static void
 730 print_ethaddr(const char *name, const struct ether_addr *eth_addr)
 731 {
 732         printf ("%s%02X:%02X:%02X:%02X:%02X:%02X", name,
 733                 eth_addr->addr_bytes[0],
 734                 eth_addr->addr_bytes[1],
 735                 eth_addr->addr_bytes[2],
 736                 eth_addr->addr_bytes[3],
 737                 eth_addr->addr_bytes[4],
 738                 eth_addr->addr_bytes[5]);
 739 }
 740
 741 /* Check the link status of all ports in up to 9s, and print them finally */
 742 static void
 743 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask)
 744 {
 745 #define CHECK_INTERVAL 100 /* 100ms */
 746 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */
 747         uint8_t portid, count, all_ports_up, print_flag = 0;
 748         struct rte_eth_link link;
 749
 750         printf("\nChecking link status");
 751         fflush(stdout);
 752         for (count = 0; count <= MAX_CHECK_TIME; count++) {
 753                 all_ports_up = 1;
 754                 for (portid = 0; portid < port_num; portid++) {
 755                         if ((port_mask & (1 << portid)) == 0)
 756                                 continue;
 757                         memset(&link, 0, sizeof(link));
 758                         rte_eth_link_get_nowait(portid, &link);
 759                         /* print link status if flag set */
 760                         if (print_flag == 1) {
 761                                 if (link.link_status)
 762                                         printf("Port %d Link Up - speed %u "
 763                                                 "Mbps - %s\n", (uint8_t)portid,
 764                                                 (unsigned)link.link_speed,
 765                                 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ?
 766                                         ("full-duplex") : ("half-duplex\n"));
 767                                 else
 768                                         printf("Port %d Link Down\n",
 769                                                 (uint8_t)portid);
 770                                 continue;
 771                         }
 772                         /* clear all_ports_up flag if any link down */
 773                         if (link.link_status == 0) {
 774                                 all_ports_up = 0;
 775                                 break;
 776                         }
 777                 }
 778                 /* after finally printing all link status, get out */
 779                 if (print_flag == 1)
 780                         break;
 781
 782                 if (all_ports_up == 0) {
 783                         printf(".");
 784                         fflush(stdout);
 785                         rte_delay_ms(CHECK_INTERVAL);
 786                 }
 787
 788                 /* set the print_flag if all ports up or timeout */
 789                 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) {
 790                         print_flag = 1;
 791                         printf("\ndone\n");
 792                 }
 793         }
 794 }
 795
 796 static int
 797 init_routing_table(void)
 798 {
 799         struct rte_lpm *lpm;
 800         struct rte_lpm6 *lpm6;
 801         int socket, ret;
 802         unsigned i;
 803
 804         for (socket = 0; socket < RTE_MAX_NUMA_NODES; socket++) {
 805                 if (socket_lpm[socket]) {
 806                         lpm = socket_lpm[socket];
 807                         /* populate the LPM table */
 808                         for (i = 0; i < RTE_DIM(l3fwd_ipv4_route_array); i++) {
 809                                 ret = rte_lpm_add(lpm,
 810                                         l3fwd_ipv4_route_array[i].ip,
 811                                         l3fwd_ipv4_route_array[i].depth,
 812                                         l3fwd_ipv4_route_array[i].if_out);
 813
 814                                 if (ret < 0) {
 815                                         RTE_LOG(ERR, IP_RSMBL, "Unable to add entry %i to the l3fwd "
 816                                                 "LPM table\n", i);
 817                                         return -1;
 818                                 }
 819
 820                                 RTE_LOG(INFO, IP_RSMBL, "Socket %i: adding route " IPv4_BYTES_FMT
 821                                                 "/%d (port %d)\n",
 822                                         socket,
 823                                         IPv4_BYTES(l3fwd_ipv4_route_array[i].ip),
 824                                         l3fwd_ipv4_route_array[i].depth,
 825                                         l3fwd_ipv4_route_array[i].if_out);
 826                         }
 827                 }
 828
 829                 if (socket_lpm6[socket]) {
 830                         lpm6 = socket_lpm6[socket];
 831                         /* populate the LPM6 table */
 832                         for (i = 0; i < RTE_DIM(l3fwd_ipv6_route_array); i++) {
 833                                 ret = rte_lpm6_add(lpm6,
 834                                         l3fwd_ipv6_route_array[i].ip,
 835                                         l3fwd_ipv6_route_array[i].depth,
 836                                         l3fwd_ipv6_route_array[i].if_out);
 837
 838                                 if (ret < 0) {
 839                                         RTE_LOG(ERR, IP_RSMBL, "Unable to add entry %i to the l3fwd "
 840                                                 "LPM6 table\n", i);
 841                                         return -1;
 842                                 }
 843
 844                                 RTE_LOG(INFO, IP_RSMBL, "Socket %i: adding route " IPv6_BYTES_FMT
 845                                                 "/%d (port %d)\n",
 846                                         socket,
 847                                         IPv6_BYTES(l3fwd_ipv6_route_array[i].ip),
 848                                         l3fwd_ipv6_route_array[i].depth,
 849                                         l3fwd_ipv6_route_array[i].if_out);
 850                         }
 851                 }
 852         }
 853         return 0;
 854 }
 855
 856 static int
 857 setup_port_tbl(struct lcore_queue_conf *qconf, uint32_t lcore, int socket,
 858         uint32_t port)
 859 {
 860         struct mbuf_table *mtb;
 861         uint32_t n;
 862         size_t sz;
 863
 864         n = RTE_MAX(max_flow_num, 2UL * MAX_PKT_BURST);
 865         sz = sizeof (*mtb) + sizeof (mtb->m_table[0]) *  n;
 866
 867         if ((mtb = rte_zmalloc_socket(__func__, sz, CACHE_LINE_SIZE,
 868                         socket)) == NULL) {
 869                 RTE_LOG(ERR, IP_RSMBL, "%s() for lcore: %u, port: %u "
 870                         "failed to allocate %zu bytes\n",
 871                         __func__, lcore, port, sz);
 872                 return -1;
 873         }
 874
 875         mtb->len = n;
 876         qconf->tx_mbufs[port] = mtb;
 877
 878         return 0;
 879 }
 880
 881 static int
 882 setup_queue_tbl(struct rx_queue *rxq, uint32_t lcore, uint32_t queue)
 883 {
 884         int socket;
 885         uint32_t nb_mbuf;
 886         uint64_t frag_cycles;
 887         char buf[RTE_MEMPOOL_NAMESIZE];
 888
 889         socket = rte_lcore_to_socket_id(lcore);
 890         if (socket == SOCKET_ID_ANY)
 891                 socket = 0;
 892
 893         frag_cycles = (rte_get_tsc_hz() + MS_PER_S - 1) / MS_PER_S *
 894                 max_flow_ttl;
 895
 896         if ((rxq->frag_tbl = rte_ip_frag_table_create(max_flow_num,
 897                         IP_FRAG_TBL_BUCKET_ENTRIES, max_flow_num, frag_cycles,
 898                         socket)) == NULL) {
 899                 RTE_LOG(ERR, IP_RSMBL, "ip_frag_tbl_create(%u) on "
 900                         "lcore: %u for queue: %u failed\n",
 901                         max_flow_num, lcore, queue);
 902                 return -1;
 903         }
 904
 905         /*
 906          * At any given moment up to <max_flow_num * (MAX_FRAG_NUM)>
 907          * mbufs could be stored int the fragment table.
 908          * Plus, each TX queue can hold up to <max_flow_num> packets.
 909          */
 910
 911         nb_mbuf = RTE_MAX(max_flow_num, 2UL * MAX_PKT_BURST) * MAX_FRAG_NUM;
 912         nb_mbuf *= (port_conf.rxmode.max_rx_pkt_len + BUF_SIZE - 1) / BUF_SIZE;
 913         nb_mbuf *= 2; /* ipv4 and ipv6 */
 914         nb_mbuf += RTE_TEST_RX_DESC_DEFAULT + RTE_TEST_TX_DESC_DEFAULT;
 915
 916         nb_mbuf = RTE_MAX(nb_mbuf, (uint32_t)NB_MBUF);
 917
 918         snprintf(buf, sizeof(buf), "mbuf_pool_%u_%u", lcore, queue);
 919
 920         if ((rxq->pool = rte_mempool_create(buf, nb_mbuf, MBUF_SIZE, 0,
 921                         sizeof(struct rte_pktmbuf_pool_private),
 922                         rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL,
 923                         socket, MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET)) == NULL) {
 924                 RTE_LOG(ERR, IP_RSMBL, "mempool_create(%s) failed", buf);
 925                 return -1;
 926         }
 927
 928         return 0;
 929 }
 930
 931 static int
 932 init_mem(void)
 933 {
 934         char buf[PATH_MAX];
 935         struct rte_lpm *lpm;
 936         struct rte_lpm6 *lpm6;
 937         int socket;
 938         unsigned lcore_id;
 939
 940         /* traverse through lcores and initialize structures on each socket */
 941
 942         for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
 943
 944                 if (rte_lcore_is_enabled(lcore_id) == 0)
 945                         continue;
 946
 947                 socket = rte_lcore_to_socket_id(lcore_id);
 948
 949                 if (socket == SOCKET_ID_ANY)
 950                         socket = 0;
 951
 952                 if (socket_lpm[socket] == NULL) {
 953                         RTE_LOG(INFO, IP_RSMBL, "Creating LPM table on socket %i\n", socket);
 954                         snprintf(buf, sizeof(buf), "IP_RSMBL_LPM_%i", socket);
 955
 956                         lpm = rte_lpm_create(buf, socket, LPM_MAX_RULES, 0);
 957                         if (lpm == NULL) {
 958                                 RTE_LOG(ERR, IP_RSMBL, "Cannot create LPM table\n");
 959                                 return -1;
 960                         }
 961                         socket_lpm[socket] = lpm;
 962                 }
 963
 964                 if (socket_lpm6[socket] == NULL) {
 965                         RTE_LOG(INFO, IP_RSMBL, "Creating LPM6 table on socket %i\n", socket);
 966                         snprintf(buf, sizeof(buf), "IP_RSMBL_LPM_%i", socket);
 967
 968                         lpm6 = rte_lpm6_create("IP_RSMBL_LPM6", socket, &lpm6_config);
 969                         if (lpm6 == NULL) {
 970                                 RTE_LOG(ERR, IP_RSMBL, "Cannot create LPM table\n");
 971                                 return -1;
 972                         }
 973                         socket_lpm6[socket] = lpm6;
 974                 }
 975         }
 976
 977         return 0;
 978 }
 979
 980 static void
 981 queue_dump_stat(void)
 982 {
 983         uint32_t i, lcore;
 984         const struct lcore_queue_conf *qconf;
 985
 986         for (lcore = 0; lcore < RTE_MAX_LCORE; lcore++) {
 987                 if (rte_lcore_is_enabled(lcore) == 0)
 988                         continue;
 989
 990                 qconf = &lcore_queue_conf[lcore];
 991                 for (i = 0; i < qconf->n_rx_queue; i++) {
 992
 993                         fprintf(stdout, " -- lcoreid=%u portid=%hhu "
 994                                 "frag tbl stat:\n",
 995                                 lcore,  qconf->rx_queue_list[i].portid);
 996                         rte_ip_frag_table_statistics_dump(stdout,
 997                                         qconf->rx_queue_list[i].frag_tbl);
 998                         fprintf(stdout, "TX bursts:\t%" PRIu64 "\n"
 999                                 "TX packets _queued:\t%" PRIu64 "\n"
1000                                 "TX packets dropped:\t%" PRIu64 "\n"
1001                                 "TX packets send:\t%" PRIu64 "\n",
1002                                 qconf->tx_stat.call,
1003                                 qconf->tx_stat.queue,
1004                                 qconf->tx_stat.drop,
1005                                 qconf->tx_stat.send);
1006                 }
1007         }
1008 }
1009
1010 static void
1011 signal_handler(int signum)
1012 {
1013         queue_dump_stat();
1014         if (signum != SIGUSR1)
1015                 rte_exit(0, "received signal: %d, exiting\n", signum);
1016 }
1017
1018 int
1019 MAIN(int argc, char **argv)
1020 {
1021         struct lcore_queue_conf *qconf;
1022         struct rte_eth_dev_info dev_info;
1023         struct rte_eth_txconf *txconf;
1024         struct rx_queue *rxq;
1025         int ret, socket;
1026         unsigned nb_ports;
1027         uint16_t queueid;
1028         unsigned lcore_id = 0, rx_lcore_id = 0;
1029         uint32_t n_tx_queue, nb_lcores;
1030         uint8_t portid;
1031
1032         /* init EAL */
1033         ret = rte_eal_init(argc, argv);
1034         if (ret < 0)
1035                 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n");
1036         argc -= ret;
1037         argv += ret;
1038
1039         /* parse application arguments (after the EAL ones) */
1040         ret = parse_args(argc, argv);
1041         if (ret < 0)
1042                 rte_exit(EXIT_FAILURE, "Invalid IP reassembly parameters\n");
1043
1044         nb_ports = rte_eth_dev_count();
1045         if (nb_ports > RTE_MAX_ETHPORTS)
1046                 nb_ports = RTE_MAX_ETHPORTS;
1047         else if (nb_ports == 0)
1048                 rte_exit(EXIT_FAILURE, "No ports found!\n");
1049
1050         nb_lcores = rte_lcore_count();
1051
1052         /* initialize structures (mempools, lpm etc.) */
1053         if (init_mem() < 0)
1054                 rte_panic("Cannot initialize memory structures!\n");
1055
1056         /* check if portmask has non-existent ports */
1057         if (enabled_port_mask & ~(RTE_LEN2MASK(nb_ports, unsigned)))
1058                 rte_exit(EXIT_FAILURE, "Non-existent ports in portmask!\n");
1059
1060         /* initialize all ports */
1061         for (portid = 0; portid < nb_ports; portid++) {
1062                 /* skip ports that are not enabled */
1063                 if ((enabled_port_mask & (1 << portid)) == 0) {
1064                         printf("\nSkipping disabled port %d\n", portid);
1065                         continue;
1066                 }
1067
1068                 qconf = &lcore_queue_conf[rx_lcore_id];
1069
1070                 /* get the lcore_id for this port */
1071                 while (rte_lcore_is_enabled(rx_lcore_id) == 0 ||
1072                            qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) {
1073
1074                         rx_lcore_id++;
1075                         if (rx_lcore_id >= RTE_MAX_LCORE)
1076                                 rte_exit(EXIT_FAILURE, "Not enough cores\n");
1077
1078                         qconf = &lcore_queue_conf[rx_lcore_id];
1079                 }
1080
1081                 socket = rte_lcore_to_socket_id(portid);
1082                 if (socket == SOCKET_ID_ANY)
1083                         socket = 0;
1084
1085                 queueid = qconf->n_rx_queue;
1086                 rxq = &qconf->rx_queue_list[queueid];
1087                 rxq->portid = portid;
1088                 rxq->lpm = socket_lpm[socket];
1089                 rxq->lpm6 = socket_lpm6[socket];
1090                 if (setup_queue_tbl(rxq, rx_lcore_id, queueid) < 0)
1091                         rte_exit(EXIT_FAILURE, "Failed to set up queue table\n");
1092                 qconf->n_rx_queue++;
1093
1094                 /* init port */
1095                 printf("Initializing port %d ... ", portid );
1096                 fflush(stdout);
1097
1098                 n_tx_queue = nb_lcores;
1099                 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT)
1100                         n_tx_queue = MAX_TX_QUEUE_PER_PORT;
1101                 ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue,
1102                                             &port_conf);
1103                 if (ret < 0) {
1104                         printf("\n");
1105                         rte_exit(EXIT_FAILURE, "Cannot configure device: "
1106                                 "err=%d, port=%d\n",
1107                                 ret, portid);
1108                 }
1109
1110                 /* init one RX queue */
1111                 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd,
1112                                              socket, NULL,
1113                                              rxq->pool);
1114                 if (ret < 0) {
1115                         printf("\n");
1116                         rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: "
1117                                 "err=%d, port=%d\n",
1118                                 ret, portid);
1119                 }
1120
1121                 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]);
1122                 print_ethaddr(" Address:", &ports_eth_addr[portid]);
1123                 printf("\n");
1124
1125                 /* init one TX queue per couple (lcore,port) */
1126                 queueid = 0;
1127                 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) {
1128                         if (rte_lcore_is_enabled(lcore_id) == 0)
1129                                 continue;
1130
1131                         socket = (int) rte_lcore_to_socket_id(lcore_id);
1132
1133                         printf("txq=%u,%d,%d ", lcore_id, queueid, socket);
1134                         fflush(stdout);
1135
1136                         rte_eth_dev_info_get(portid, &dev_info);
1137                         txconf = &dev_info.default_txconf;
1138                         txconf->txq_flags = 0;
1139
1140                         ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd,
1141                                         socket, txconf);
1142                         if (ret < 0)
1143                                 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, "
1144                                         "port=%d\n", ret, portid);
1145
1146                         qconf = &lcore_queue_conf[lcore_id];
1147                         qconf->tx_queue_id[portid] = queueid;
1148                         setup_port_tbl(qconf, lcore_id, socket, portid);
1149                         queueid++;
1150                 }
1151                 printf("\n");
1152         }
1153
1154         printf("\n");
1155
1156         /* start ports */
1157         for (portid = 0; portid < nb_ports; portid++) {
1158                 if ((enabled_port_mask & (1 << portid)) == 0) {
1159                         continue;
1160                 }
1161                 /* Start device */
1162                 ret = rte_eth_dev_start(portid);
1163                 if (ret < 0)
1164                         rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n",
1165                                 ret, portid);
1166
1167                 rte_eth_promiscuous_enable(portid);
1168         }
1169
1170         if (init_routing_table() < 0)
1171                 rte_exit(EXIT_FAILURE, "Cannot init routing table\n");
1172
1173         check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask);
1174
1175         signal(SIGUSR1, signal_handler);
1176         signal(SIGTERM, signal_handler);
1177         signal(SIGINT, signal_handler);
1178
1179         /* launch per-lcore init on every lcore */
1180         rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER);
1181         RTE_LCORE_FOREACH_SLAVE(lcore_id) {
1182                 if (rte_eal_wait_lcore(lcore_id) < 0)
1183                         return -1;
1184         }
1185
1186         return 0;
1187 }