1 /* SPDX-License-Identifier: BSD-3-Clause 2 * Copyright(c) 2010-2014 Intel Corporation 3 */ 4 5 #include <stdio.h> 6 #include <stdlib.h> 7 #include <stdint.h> 8 #include <inttypes.h> 9 #include <sys/types.h> 10 #include <sys/param.h> 11 #include <string.h> 12 #include <sys/queue.h> 13 #include <stdarg.h> 14 #include <errno.h> 15 #include <getopt.h> 16 17 #include <rte_common.h> 18 #include <rte_byteorder.h> 19 #include <rte_log.h> 20 #include <rte_memory.h> 21 #include <rte_memcpy.h> 22 #include <rte_eal.h> 23 #include <rte_launch.h> 24 #include <rte_atomic.h> 25 #include <rte_cycles.h> 26 #include <rte_prefetch.h> 27 #include <rte_lcore.h> 28 #include <rte_per_lcore.h> 29 #include <rte_branch_prediction.h> 30 #include <rte_interrupts.h> 31 #include <rte_random.h> 32 #include <rte_debug.h> 33 #include <rte_ether.h> 34 #include <rte_ethdev.h> 35 #include <rte_mempool.h> 36 #include <rte_mbuf.h> 37 #include <rte_lpm.h> 38 #include <rte_lpm6.h> 39 #include <rte_ip.h> 40 #include <rte_string_fns.h> 41 42 #include <rte_ip_frag.h> 43 44 #define RTE_LOGTYPE_IP_FRAG RTE_LOGTYPE_USER1 45 46 /* allow max jumbo frame 9.5 KB */ 47 #define JUMBO_FRAME_MAX_SIZE 0x2600 48 49 #define ROUNDUP_DIV(a, b) (((a) + (b) - 1) / (b)) 50 51 /* 52 * Default byte size for the IPv6 Maximum Transfer Unit (MTU). 53 * This value includes the size of IPv6 header. 54 */ 55 #define IPV4_MTU_DEFAULT RTE_ETHER_MTU 56 #define IPV6_MTU_DEFAULT RTE_ETHER_MTU 57 58 /* 59 * The overhead from max frame size to MTU. 60 * We have to consider the max possible overhead. 61 */ 62 #define MTU_OVERHEAD \ 63 (RTE_ETHER_HDR_LEN + RTE_ETHER_CRC_LEN + \ 64 2 * sizeof(struct rte_vlan_hdr)) 65 66 /* 67 * Default payload in bytes for the IPv6 packet. 68 */ 69 #define IPV4_DEFAULT_PAYLOAD (IPV4_MTU_DEFAULT - sizeof(struct rte_ipv4_hdr)) 70 #define IPV6_DEFAULT_PAYLOAD (IPV6_MTU_DEFAULT - sizeof(struct rte_ipv6_hdr)) 71 72 /* 73 * Max number of fragments per packet expected - defined by config file. 74 */ 75 #define MAX_PACKET_FRAG RTE_LIBRTE_IP_FRAG_MAX_FRAG 76 77 #define NB_MBUF 8192 78 79 #define MAX_PKT_BURST 32 80 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */ 81 82 /* Configure how many packets ahead to prefetch, when reading packets */ 83 #define PREFETCH_OFFSET 3 84 85 /* 86 * Configurable number of RX/TX ring descriptors 87 */ 88 #define RTE_TEST_RX_DESC_DEFAULT 1024 89 #define RTE_TEST_TX_DESC_DEFAULT 1024 90 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT; 91 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT; 92 93 /* ethernet addresses of ports */ 94 static struct rte_ether_addr ports_eth_addr[RTE_MAX_ETHPORTS]; 95 96 #ifndef IPv4_BYTES 97 #define IPv4_BYTES_FMT "%" PRIu8 ".%" PRIu8 ".%" PRIu8 ".%" PRIu8 98 #define IPv4_BYTES(addr) \ 99 (uint8_t) (((addr) >> 24) & 0xFF),\ 100 (uint8_t) (((addr) >> 16) & 0xFF),\ 101 (uint8_t) (((addr) >> 8) & 0xFF),\ 102 (uint8_t) ((addr) & 0xFF) 103 #endif 104 105 #ifndef IPv6_BYTES 106 #define IPv6_BYTES_FMT "%02x%02x:%02x%02x:%02x%02x:%02x%02x:"\ 107 "%02x%02x:%02x%02x:%02x%02x:%02x%02x" 108 #define IPv6_BYTES(addr) \ 109 addr[0], addr[1], addr[2], addr[3], \ 110 addr[4], addr[5], addr[6], addr[7], \ 111 addr[8], addr[9], addr[10], addr[11],\ 112 addr[12], addr[13],addr[14], addr[15] 113 #endif 114 115 #define IPV6_ADDR_LEN 16 116 117 /* mask of enabled ports */ 118 static int enabled_port_mask = 0; 119 120 static int rx_queue_per_lcore = 1; 121 122 #define MBUF_TABLE_SIZE (2 * MAX(MAX_PKT_BURST, MAX_PACKET_FRAG)) 123 124 struct mbuf_table { 125 uint16_t len; 126 struct rte_mbuf *m_table[MBUF_TABLE_SIZE]; 127 }; 128 129 struct rx_queue { 130 struct rte_mempool *direct_pool; 131 struct rte_mempool *indirect_pool; 132 struct rte_lpm *lpm; 133 struct rte_lpm6 *lpm6; 134 uint16_t portid; 135 }; 136 137 #define MAX_RX_QUEUE_PER_LCORE 16 138 #define MAX_TX_QUEUE_PER_PORT 16 139 struct lcore_queue_conf { 140 uint16_t n_rx_queue; 141 uint16_t tx_queue_id[RTE_MAX_ETHPORTS]; 142 struct rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE]; 143 struct mbuf_table tx_mbufs[RTE_MAX_ETHPORTS]; 144 } __rte_cache_aligned; 145 struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE]; 146 147 static struct rte_eth_conf port_conf = { 148 .rxmode = { 149 .max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE, 150 .split_hdr_size = 0, 151 .offloads = (DEV_RX_OFFLOAD_CHECKSUM | 152 DEV_RX_OFFLOAD_SCATTER | 153 DEV_RX_OFFLOAD_JUMBO_FRAME), 154 }, 155 .txmode = { 156 .mq_mode = ETH_MQ_TX_NONE, 157 .offloads = (DEV_TX_OFFLOAD_IPV4_CKSUM | 158 DEV_TX_OFFLOAD_MULTI_SEGS), 159 }, 160 }; 161 162 /* 163 * IPv4 forwarding table 164 */ 165 struct l3fwd_ipv4_route { 166 uint32_t ip; 167 uint8_t depth; 168 uint8_t if_out; 169 }; 170 171 /* Default l3fwd_ipv4_route_array table. 8< */ 172 struct l3fwd_ipv4_route l3fwd_ipv4_route_array[] = { 173 {RTE_IPV4(100,10,0,0), 16, 0}, 174 {RTE_IPV4(100,20,0,0), 16, 1}, 175 {RTE_IPV4(100,30,0,0), 16, 2}, 176 {RTE_IPV4(100,40,0,0), 16, 3}, 177 {RTE_IPV4(100,50,0,0), 16, 4}, 178 {RTE_IPV4(100,60,0,0), 16, 5}, 179 {RTE_IPV4(100,70,0,0), 16, 6}, 180 {RTE_IPV4(100,80,0,0), 16, 7}, 181 }; 182 /* >8 End of default l3fwd_ipv4_route_array table */ 183 184 /* 185 * IPv6 forwarding table 186 */ 187 188 struct l3fwd_ipv6_route { 189 uint8_t ip[IPV6_ADDR_LEN]; 190 uint8_t depth; 191 uint8_t if_out; 192 }; 193 194 /* Default l3fwd_ipv6_route_array table. 8< */ 195 static struct l3fwd_ipv6_route l3fwd_ipv6_route_array[] = { 196 {{1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 0}, 197 {{2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 1}, 198 {{3,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 2}, 199 {{4,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 3}, 200 {{5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 4}, 201 {{6,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 5}, 202 {{7,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 6}, 203 {{8,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 7}, 204 }; 205 /* >8 End of default l3fwd_ipv6_route_array table. */ 206 207 #define LPM_MAX_RULES 1024 208 #define LPM6_MAX_RULES 1024 209 #define LPM6_NUMBER_TBL8S (1 << 16) 210 211 struct rte_lpm6_config lpm6_config = { 212 .max_rules = LPM6_MAX_RULES, 213 .number_tbl8s = LPM6_NUMBER_TBL8S, 214 .flags = 0 215 }; 216 217 static struct rte_mempool *socket_direct_pool[RTE_MAX_NUMA_NODES]; 218 static struct rte_mempool *socket_indirect_pool[RTE_MAX_NUMA_NODES]; 219 static struct rte_lpm *socket_lpm[RTE_MAX_NUMA_NODES]; 220 static struct rte_lpm6 *socket_lpm6[RTE_MAX_NUMA_NODES]; 221 222 /* Send burst of packets on an output interface */ 223 static inline int 224 send_burst(struct lcore_queue_conf *qconf, uint16_t n, uint16_t port) 225 { 226 struct rte_mbuf **m_table; 227 int ret; 228 uint16_t queueid; 229 230 queueid = qconf->tx_queue_id[port]; 231 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table; 232 233 ret = rte_eth_tx_burst(port, queueid, m_table, n); 234 if (unlikely(ret < n)) { 235 do { 236 rte_pktmbuf_free(m_table[ret]); 237 } while (++ret < n); 238 } 239 240 return 0; 241 } 242 243 static inline void 244 l3fwd_simple_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf, 245 uint8_t queueid, uint16_t port_in) 246 { 247 struct rx_queue *rxq; 248 uint32_t i, len, next_hop; 249 uint16_t port_out, ether_type; 250 int32_t len2; 251 uint64_t ol_flags; 252 const struct rte_ether_hdr *eth; 253 254 ol_flags = 0; 255 rxq = &qconf->rx_queue_list[queueid]; 256 257 /* by default, send everything back to the source port */ 258 port_out = port_in; 259 260 /* save ether type of the incoming packet */ 261 eth = rte_pktmbuf_mtod(m, const struct rte_ether_hdr *); 262 ether_type = eth->ether_type; 263 264 /* Remove the Ethernet header and trailer from the input packet */ 265 rte_pktmbuf_adj(m, (uint16_t)sizeof(struct rte_ether_hdr)); 266 267 /* Build transmission burst */ 268 len = qconf->tx_mbufs[port_out].len; 269 270 /* if this is an IPv4 packet */ 271 if (RTE_ETH_IS_IPV4_HDR(m->packet_type)) { 272 struct rte_ipv4_hdr *ip_hdr; 273 uint32_t ip_dst; 274 /* Read the lookup key (i.e. ip_dst) from the input packet */ 275 ip_hdr = rte_pktmbuf_mtod(m, struct rte_ipv4_hdr *); 276 ip_dst = rte_be_to_cpu_32(ip_hdr->dst_addr); 277 278 /* Find destination port */ 279 if (rte_lpm_lookup(rxq->lpm, ip_dst, &next_hop) == 0 && 280 (enabled_port_mask & 1 << next_hop) != 0) { 281 port_out = next_hop; 282 283 /* Build transmission burst for new port */ 284 len = qconf->tx_mbufs[port_out].len; 285 } 286 287 /* if we don't need to do any fragmentation */ 288 if (likely (IPV4_MTU_DEFAULT >= m->pkt_len)) { 289 qconf->tx_mbufs[port_out].m_table[len] = m; 290 len2 = 1; 291 } else { 292 len2 = rte_ipv4_fragment_packet(m, 293 &qconf->tx_mbufs[port_out].m_table[len], 294 (uint16_t)(MBUF_TABLE_SIZE - len), 295 IPV4_MTU_DEFAULT, 296 rxq->direct_pool, rxq->indirect_pool); 297 298 /* Free input packet */ 299 rte_pktmbuf_free(m); 300 301 /* request HW to regenerate IPv4 cksum */ 302 ol_flags |= (PKT_TX_IPV4 | PKT_TX_IP_CKSUM); 303 304 /* If we fail to fragment the packet */ 305 if (unlikely (len2 < 0)) 306 return; 307 } 308 } else if (RTE_ETH_IS_IPV6_HDR(m->packet_type)) { 309 /* if this is an IPv6 packet */ 310 struct rte_ipv6_hdr *ip_hdr; 311 312 /* Read the lookup key (i.e. ip_dst) from the input packet */ 313 ip_hdr = rte_pktmbuf_mtod(m, struct rte_ipv6_hdr *); 314 315 /* Find destination port */ 316 if (rte_lpm6_lookup(rxq->lpm6, ip_hdr->dst_addr, 317 &next_hop) == 0 && 318 (enabled_port_mask & 1 << next_hop) != 0) { 319 port_out = next_hop; 320 321 /* Build transmission burst for new port */ 322 len = qconf->tx_mbufs[port_out].len; 323 } 324 325 /* if we don't need to do any fragmentation */ 326 if (likely (IPV6_MTU_DEFAULT >= m->pkt_len)) { 327 qconf->tx_mbufs[port_out].m_table[len] = m; 328 len2 = 1; 329 } else { 330 len2 = rte_ipv6_fragment_packet(m, 331 &qconf->tx_mbufs[port_out].m_table[len], 332 (uint16_t)(MBUF_TABLE_SIZE - len), 333 IPV6_MTU_DEFAULT, 334 rxq->direct_pool, rxq->indirect_pool); 335 336 /* Free input packet */ 337 rte_pktmbuf_free(m); 338 339 /* If we fail to fragment the packet */ 340 if (unlikely (len2 < 0)) 341 return; 342 } 343 } 344 /* else, just forward the packet */ 345 else { 346 qconf->tx_mbufs[port_out].m_table[len] = m; 347 len2 = 1; 348 } 349 350 for (i = len; i < len + len2; i ++) { 351 void *d_addr_bytes; 352 353 m = qconf->tx_mbufs[port_out].m_table[i]; 354 struct rte_ether_hdr *eth_hdr = (struct rte_ether_hdr *) 355 rte_pktmbuf_prepend(m, 356 (uint16_t)sizeof(struct rte_ether_hdr)); 357 if (eth_hdr == NULL) { 358 rte_panic("No headroom in mbuf.\n"); 359 } 360 361 m->ol_flags |= ol_flags; 362 m->l2_len = sizeof(struct rte_ether_hdr); 363 364 /* 02:00:00:00:00:xx */ 365 d_addr_bytes = ð_hdr->dst_addr.addr_bytes[0]; 366 *((uint64_t *)d_addr_bytes) = 0x000000000002 + 367 ((uint64_t)port_out << 40); 368 369 /* src addr */ 370 rte_ether_addr_copy(&ports_eth_addr[port_out], 371 ð_hdr->src_addr); 372 eth_hdr->ether_type = ether_type; 373 } 374 375 len += len2; 376 377 if (likely(len < MAX_PKT_BURST)) { 378 qconf->tx_mbufs[port_out].len = (uint16_t)len; 379 return; 380 } 381 382 /* Transmit packets */ 383 send_burst(qconf, (uint16_t)len, port_out); 384 qconf->tx_mbufs[port_out].len = 0; 385 } 386 387 /* main processing loop */ 388 static int 389 main_loop(__rte_unused void *dummy) 390 { 391 struct rte_mbuf *pkts_burst[MAX_PKT_BURST]; 392 unsigned lcore_id; 393 uint64_t prev_tsc, diff_tsc, cur_tsc; 394 int i, j, nb_rx; 395 uint16_t portid; 396 struct lcore_queue_conf *qconf; 397 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US; 398 399 prev_tsc = 0; 400 401 lcore_id = rte_lcore_id(); 402 qconf = &lcore_queue_conf[lcore_id]; 403 404 if (qconf->n_rx_queue == 0) { 405 RTE_LOG(INFO, IP_FRAG, "lcore %u has nothing to do\n", lcore_id); 406 return 0; 407 } 408 409 RTE_LOG(INFO, IP_FRAG, "entering main loop on lcore %u\n", lcore_id); 410 411 for (i = 0; i < qconf->n_rx_queue; i++) { 412 413 portid = qconf->rx_queue_list[i].portid; 414 RTE_LOG(INFO, IP_FRAG, " -- lcoreid=%u portid=%d\n", lcore_id, 415 portid); 416 } 417 418 while (1) { 419 420 cur_tsc = rte_rdtsc(); 421 422 /* 423 * TX burst queue drain 424 */ 425 diff_tsc = cur_tsc - prev_tsc; 426 if (unlikely(diff_tsc > drain_tsc)) { 427 428 /* 429 * This could be optimized (use queueid instead of 430 * portid), but it is not called so often 431 */ 432 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) { 433 if (qconf->tx_mbufs[portid].len == 0) 434 continue; 435 send_burst(&lcore_queue_conf[lcore_id], 436 qconf->tx_mbufs[portid].len, 437 portid); 438 qconf->tx_mbufs[portid].len = 0; 439 } 440 441 prev_tsc = cur_tsc; 442 } 443 444 /* 445 * Read packet from RX queues 446 */ 447 for (i = 0; i < qconf->n_rx_queue; i++) { 448 449 portid = qconf->rx_queue_list[i].portid; 450 nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst, 451 MAX_PKT_BURST); 452 453 /* Prefetch first packets */ 454 for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) { 455 rte_prefetch0(rte_pktmbuf_mtod( 456 pkts_burst[j], void *)); 457 } 458 459 /* Prefetch and forward already prefetched packets */ 460 for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) { 461 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[ 462 j + PREFETCH_OFFSET], void *)); 463 l3fwd_simple_forward(pkts_burst[j], qconf, i, portid); 464 } 465 466 /* Forward remaining prefetched packets */ 467 for (; j < nb_rx; j++) { 468 l3fwd_simple_forward(pkts_burst[j], qconf, i, portid); 469 } 470 } 471 } 472 } 473 474 /* display usage */ 475 static void 476 print_usage(const char *prgname) 477 { 478 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n" 479 " -p PORTMASK: hexadecimal bitmask of ports to configure\n" 480 " -q NQ: number of queue (=ports) per lcore (default is 1)\n", 481 prgname); 482 } 483 484 static int 485 parse_portmask(const char *portmask) 486 { 487 char *end = NULL; 488 unsigned long pm; 489 490 /* parse hexadecimal string */ 491 pm = strtoul(portmask, &end, 16); 492 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0')) 493 return -1; 494 495 if (pm == 0) 496 return -1; 497 498 return pm; 499 } 500 501 static int 502 parse_nqueue(const char *q_arg) 503 { 504 char *end = NULL; 505 unsigned long n; 506 507 /* parse hexadecimal string */ 508 n = strtoul(q_arg, &end, 10); 509 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0')) 510 return -1; 511 if (n == 0) 512 return -1; 513 if (n >= MAX_RX_QUEUE_PER_LCORE) 514 return -1; 515 516 return n; 517 } 518 519 /* Parse the argument given in the command line of the application */ 520 static int 521 parse_args(int argc, char **argv) 522 { 523 int opt, ret; 524 char **argvopt; 525 int option_index; 526 char *prgname = argv[0]; 527 static struct option lgopts[] = { 528 {NULL, 0, 0, 0} 529 }; 530 531 argvopt = argv; 532 533 while ((opt = getopt_long(argc, argvopt, "p:q:", 534 lgopts, &option_index)) != EOF) { 535 536 switch (opt) { 537 /* portmask */ 538 case 'p': 539 enabled_port_mask = parse_portmask(optarg); 540 if (enabled_port_mask < 0) { 541 printf("invalid portmask\n"); 542 print_usage(prgname); 543 return -1; 544 } 545 break; 546 547 /* nqueue */ 548 case 'q': 549 rx_queue_per_lcore = parse_nqueue(optarg); 550 if (rx_queue_per_lcore < 0) { 551 printf("invalid queue number\n"); 552 print_usage(prgname); 553 return -1; 554 } 555 break; 556 557 /* long options */ 558 case 0: 559 print_usage(prgname); 560 return -1; 561 562 default: 563 print_usage(prgname); 564 return -1; 565 } 566 } 567 568 if (enabled_port_mask == 0) { 569 printf("portmask not specified\n"); 570 print_usage(prgname); 571 return -1; 572 } 573 574 if (optind >= 0) 575 argv[optind-1] = prgname; 576 577 ret = optind-1; 578 optind = 1; /* reset getopt lib */ 579 return ret; 580 } 581 582 static void 583 print_ethaddr(const char *name, struct rte_ether_addr *eth_addr) 584 { 585 char buf[RTE_ETHER_ADDR_FMT_SIZE]; 586 rte_ether_format_addr(buf, RTE_ETHER_ADDR_FMT_SIZE, eth_addr); 587 printf("%s%s", name, buf); 588 } 589 590 /* Check the link status of all ports in up to 9s, and print them finally */ 591 static void 592 check_all_ports_link_status(uint32_t port_mask) 593 { 594 #define CHECK_INTERVAL 100 /* 100ms */ 595 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */ 596 uint16_t portid; 597 uint8_t count, all_ports_up, print_flag = 0; 598 struct rte_eth_link link; 599 int ret; 600 char link_status_text[RTE_ETH_LINK_MAX_STR_LEN]; 601 602 printf("\nChecking link status"); 603 fflush(stdout); 604 for (count = 0; count <= MAX_CHECK_TIME; count++) { 605 all_ports_up = 1; 606 RTE_ETH_FOREACH_DEV(portid) { 607 if ((port_mask & (1 << portid)) == 0) 608 continue; 609 memset(&link, 0, sizeof(link)); 610 ret = rte_eth_link_get_nowait(portid, &link); 611 if (ret < 0) { 612 all_ports_up = 0; 613 if (print_flag == 1) 614 printf("Port %u link get failed: %s\n", 615 portid, rte_strerror(-ret)); 616 continue; 617 } 618 /* print link status if flag set */ 619 if (print_flag == 1) { 620 rte_eth_link_to_str(link_status_text, 621 sizeof(link_status_text), &link); 622 printf("Port %d %s\n", portid, 623 link_status_text); 624 continue; 625 } 626 /* clear all_ports_up flag if any link down */ 627 if (link.link_status == ETH_LINK_DOWN) { 628 all_ports_up = 0; 629 break; 630 } 631 } 632 /* after finally printing all link status, get out */ 633 if (print_flag == 1) 634 break; 635 636 if (all_ports_up == 0) { 637 printf("."); 638 fflush(stdout); 639 rte_delay_ms(CHECK_INTERVAL); 640 } 641 642 /* set the print_flag if all ports up or timeout */ 643 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) { 644 print_flag = 1; 645 printf("\ndone\n"); 646 } 647 } 648 } 649 650 /* Check L3 packet type detection capability of the NIC port */ 651 static int 652 check_ptype(int portid) 653 { 654 int i, ret; 655 int ptype_l3_ipv4 = 0, ptype_l3_ipv6 = 0; 656 uint32_t ptype_mask = RTE_PTYPE_L3_MASK; 657 658 ret = rte_eth_dev_get_supported_ptypes(portid, ptype_mask, NULL, 0); 659 if (ret <= 0) 660 return 0; 661 662 uint32_t ptypes[ret]; 663 664 ret = rte_eth_dev_get_supported_ptypes(portid, ptype_mask, ptypes, ret); 665 for (i = 0; i < ret; ++i) { 666 if (ptypes[i] & RTE_PTYPE_L3_IPV4) 667 ptype_l3_ipv4 = 1; 668 if (ptypes[i] & RTE_PTYPE_L3_IPV6) 669 ptype_l3_ipv6 = 1; 670 } 671 672 if (ptype_l3_ipv4 == 0) 673 printf("port %d cannot parse RTE_PTYPE_L3_IPV4\n", portid); 674 675 if (ptype_l3_ipv6 == 0) 676 printf("port %d cannot parse RTE_PTYPE_L3_IPV6\n", portid); 677 678 if (ptype_l3_ipv4 && ptype_l3_ipv6) 679 return 1; 680 681 return 0; 682 683 } 684 685 /* Parse packet type of a packet by SW */ 686 static inline void 687 parse_ptype(struct rte_mbuf *m) 688 { 689 struct rte_ether_hdr *eth_hdr; 690 uint32_t packet_type = RTE_PTYPE_UNKNOWN; 691 uint16_t ether_type; 692 693 eth_hdr = rte_pktmbuf_mtod(m, struct rte_ether_hdr *); 694 ether_type = eth_hdr->ether_type; 695 if (ether_type == rte_cpu_to_be_16(RTE_ETHER_TYPE_IPV4)) 696 packet_type |= RTE_PTYPE_L3_IPV4_EXT_UNKNOWN; 697 else if (ether_type == rte_cpu_to_be_16(RTE_ETHER_TYPE_IPV6)) 698 packet_type |= RTE_PTYPE_L3_IPV6_EXT_UNKNOWN; 699 700 m->packet_type = packet_type; 701 } 702 703 /* callback function to detect packet type for a queue of a port */ 704 static uint16_t 705 cb_parse_ptype(uint16_t port __rte_unused, uint16_t queue __rte_unused, 706 struct rte_mbuf *pkts[], uint16_t nb_pkts, 707 uint16_t max_pkts __rte_unused, 708 void *user_param __rte_unused) 709 { 710 uint16_t i; 711 712 for (i = 0; i < nb_pkts; ++i) 713 parse_ptype(pkts[i]); 714 715 return nb_pkts; 716 } 717 718 static int 719 init_routing_table(void) 720 { 721 struct rte_lpm *lpm; 722 struct rte_lpm6 *lpm6; 723 int socket, ret; 724 unsigned i; 725 726 for (socket = 0; socket < RTE_MAX_NUMA_NODES; socket++) { 727 if (socket_lpm[socket]) { 728 lpm = socket_lpm[socket]; 729 /* populate the LPM table */ 730 for (i = 0; i < RTE_DIM(l3fwd_ipv4_route_array); i++) { 731 ret = rte_lpm_add(lpm, 732 l3fwd_ipv4_route_array[i].ip, 733 l3fwd_ipv4_route_array[i].depth, 734 l3fwd_ipv4_route_array[i].if_out); 735 736 if (ret < 0) { 737 RTE_LOG(ERR, IP_FRAG, "Unable to add entry %i to the l3fwd " 738 "LPM table\n", i); 739 return -1; 740 } 741 742 RTE_LOG(INFO, IP_FRAG, "Socket %i: adding route " IPv4_BYTES_FMT 743 "/%d (port %d)\n", 744 socket, 745 IPv4_BYTES(l3fwd_ipv4_route_array[i].ip), 746 l3fwd_ipv4_route_array[i].depth, 747 l3fwd_ipv4_route_array[i].if_out); 748 } 749 } 750 751 if (socket_lpm6[socket]) { 752 lpm6 = socket_lpm6[socket]; 753 /* populate the LPM6 table */ 754 for (i = 0; i < RTE_DIM(l3fwd_ipv6_route_array); i++) { 755 ret = rte_lpm6_add(lpm6, 756 l3fwd_ipv6_route_array[i].ip, 757 l3fwd_ipv6_route_array[i].depth, 758 l3fwd_ipv6_route_array[i].if_out); 759 760 if (ret < 0) { 761 RTE_LOG(ERR, IP_FRAG, "Unable to add entry %i to the l3fwd " 762 "LPM6 table\n", i); 763 return -1; 764 } 765 766 RTE_LOG(INFO, IP_FRAG, "Socket %i: adding route " IPv6_BYTES_FMT 767 "/%d (port %d)\n", 768 socket, 769 IPv6_BYTES(l3fwd_ipv6_route_array[i].ip), 770 l3fwd_ipv6_route_array[i].depth, 771 l3fwd_ipv6_route_array[i].if_out); 772 } 773 } 774 } 775 return 0; 776 } 777 778 static int 779 init_mem(void) 780 { 781 char buf[PATH_MAX]; 782 struct rte_mempool *mp; 783 struct rte_lpm *lpm; 784 struct rte_lpm6 *lpm6; 785 struct rte_lpm_config lpm_config; 786 int socket; 787 unsigned lcore_id; 788 789 /* traverse through lcores and initialize structures on each socket */ 790 791 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) { 792 793 if (rte_lcore_is_enabled(lcore_id) == 0) 794 continue; 795 796 socket = rte_lcore_to_socket_id(lcore_id); 797 798 if (socket == SOCKET_ID_ANY) 799 socket = 0; 800 801 if (socket_direct_pool[socket] == NULL) { 802 RTE_LOG(INFO, IP_FRAG, "Creating direct mempool on socket %i\n", 803 socket); 804 snprintf(buf, sizeof(buf), "pool_direct_%i", socket); 805 806 mp = rte_pktmbuf_pool_create(buf, NB_MBUF, 32, 807 0, RTE_MBUF_DEFAULT_BUF_SIZE, socket); 808 if (mp == NULL) { 809 RTE_LOG(ERR, IP_FRAG, "Cannot create direct mempool\n"); 810 return -1; 811 } 812 socket_direct_pool[socket] = mp; 813 } 814 815 if (socket_indirect_pool[socket] == NULL) { 816 RTE_LOG(INFO, IP_FRAG, "Creating indirect mempool on socket %i\n", 817 socket); 818 snprintf(buf, sizeof(buf), "pool_indirect_%i", socket); 819 820 mp = rte_pktmbuf_pool_create(buf, NB_MBUF, 32, 0, 0, 821 socket); 822 if (mp == NULL) { 823 RTE_LOG(ERR, IP_FRAG, "Cannot create indirect mempool\n"); 824 return -1; 825 } 826 socket_indirect_pool[socket] = mp; 827 } 828 829 if (socket_lpm[socket] == NULL) { 830 RTE_LOG(INFO, IP_FRAG, "Creating LPM table on socket %i\n", socket); 831 snprintf(buf, sizeof(buf), "IP_FRAG_LPM_%i", socket); 832 833 lpm_config.max_rules = LPM_MAX_RULES; 834 lpm_config.number_tbl8s = 256; 835 lpm_config.flags = 0; 836 837 lpm = rte_lpm_create(buf, socket, &lpm_config); 838 if (lpm == NULL) { 839 RTE_LOG(ERR, IP_FRAG, "Cannot create LPM table\n"); 840 return -1; 841 } 842 socket_lpm[socket] = lpm; 843 } 844 845 if (socket_lpm6[socket] == NULL) { 846 RTE_LOG(INFO, IP_FRAG, "Creating LPM6 table on socket %i\n", socket); 847 snprintf(buf, sizeof(buf), "IP_FRAG_LPM_%i", socket); 848 849 lpm6 = rte_lpm6_create(buf, socket, &lpm6_config); 850 if (lpm6 == NULL) { 851 RTE_LOG(ERR, IP_FRAG, "Cannot create LPM table\n"); 852 return -1; 853 } 854 socket_lpm6[socket] = lpm6; 855 } 856 } 857 858 return 0; 859 } 860 861 int 862 main(int argc, char **argv) 863 { 864 struct lcore_queue_conf *qconf; 865 struct rte_eth_dev_info dev_info; 866 struct rte_eth_txconf *txconf; 867 struct rx_queue *rxq; 868 int socket, ret; 869 uint16_t nb_ports; 870 uint16_t queueid = 0; 871 unsigned lcore_id = 0, rx_lcore_id = 0; 872 uint32_t n_tx_queue, nb_lcores; 873 uint16_t portid; 874 875 /* init EAL */ 876 ret = rte_eal_init(argc, argv); 877 if (ret < 0) 878 rte_exit(EXIT_FAILURE, "rte_eal_init failed"); 879 argc -= ret; 880 argv += ret; 881 882 /* parse application arguments (after the EAL ones) */ 883 ret = parse_args(argc, argv); 884 if (ret < 0) 885 rte_exit(EXIT_FAILURE, "Invalid arguments"); 886 887 nb_ports = rte_eth_dev_count_avail(); 888 if (nb_ports == 0) 889 rte_exit(EXIT_FAILURE, "No ports found!\n"); 890 891 nb_lcores = rte_lcore_count(); 892 893 /* initialize structures (mempools, lpm etc.) */ 894 if (init_mem() < 0) 895 rte_panic("Cannot initialize memory structures!\n"); 896 897 /* check if portmask has non-existent ports */ 898 if (enabled_port_mask & ~(RTE_LEN2MASK(nb_ports, unsigned))) 899 rte_exit(EXIT_FAILURE, "Non-existent ports in portmask!\n"); 900 901 /* initialize all ports */ 902 RTE_ETH_FOREACH_DEV(portid) { 903 struct rte_eth_conf local_port_conf = port_conf; 904 struct rte_eth_rxconf rxq_conf; 905 906 /* skip ports that are not enabled */ 907 if ((enabled_port_mask & (1 << portid)) == 0) { 908 printf("Skipping disabled port %d\n", portid); 909 continue; 910 } 911 912 qconf = &lcore_queue_conf[rx_lcore_id]; 913 914 /* limit the frame size to the maximum supported by NIC */ 915 ret = rte_eth_dev_info_get(portid, &dev_info); 916 if (ret != 0) 917 rte_exit(EXIT_FAILURE, 918 "Error during getting device (port %u) info: %s\n", 919 portid, strerror(-ret)); 920 921 local_port_conf.rxmode.max_rx_pkt_len = RTE_MIN( 922 dev_info.max_rx_pktlen, 923 local_port_conf.rxmode.max_rx_pkt_len); 924 925 /* get the lcore_id for this port */ 926 while (rte_lcore_is_enabled(rx_lcore_id) == 0 || 927 qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) { 928 929 rx_lcore_id ++; 930 if (rx_lcore_id >= RTE_MAX_LCORE) 931 rte_exit(EXIT_FAILURE, "Not enough cores\n"); 932 933 qconf = &lcore_queue_conf[rx_lcore_id]; 934 } 935 936 socket = (int) rte_lcore_to_socket_id(rx_lcore_id); 937 if (socket == SOCKET_ID_ANY) 938 socket = 0; 939 940 rxq = &qconf->rx_queue_list[qconf->n_rx_queue]; 941 rxq->portid = portid; 942 rxq->direct_pool = socket_direct_pool[socket]; 943 rxq->indirect_pool = socket_indirect_pool[socket]; 944 rxq->lpm = socket_lpm[socket]; 945 rxq->lpm6 = socket_lpm6[socket]; 946 qconf->n_rx_queue++; 947 948 /* init port */ 949 printf("Initializing port %d on lcore %u...", portid, 950 rx_lcore_id); 951 fflush(stdout); 952 953 n_tx_queue = nb_lcores; 954 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT) 955 n_tx_queue = MAX_TX_QUEUE_PER_PORT; 956 ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue, 957 &local_port_conf); 958 if (ret < 0) { 959 printf("\n"); 960 rte_exit(EXIT_FAILURE, "Cannot configure device: " 961 "err=%d, port=%d\n", 962 ret, portid); 963 } 964 965 /* set the mtu to the maximum received packet size */ 966 ret = rte_eth_dev_set_mtu(portid, 967 local_port_conf.rxmode.max_rx_pkt_len - MTU_OVERHEAD); 968 if (ret < 0) { 969 printf("\n"); 970 rte_exit(EXIT_FAILURE, "Set MTU failed: " 971 "err=%d, port=%d\n", 972 ret, portid); 973 } 974 975 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd, 976 &nb_txd); 977 if (ret < 0) { 978 printf("\n"); 979 rte_exit(EXIT_FAILURE, "Cannot adjust number of " 980 "descriptors: err=%d, port=%d\n", ret, portid); 981 } 982 983 /* init one RX queue */ 984 rxq_conf = dev_info.default_rxconf; 985 rxq_conf.offloads = local_port_conf.rxmode.offloads; 986 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd, 987 socket, &rxq_conf, 988 socket_direct_pool[socket]); 989 if (ret < 0) { 990 printf("\n"); 991 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: " 992 "err=%d, port=%d\n", 993 ret, portid); 994 } 995 996 ret = rte_eth_macaddr_get(portid, &ports_eth_addr[portid]); 997 if (ret < 0) { 998 printf("\n"); 999 rte_exit(EXIT_FAILURE, 1000 "rte_eth_macaddr_get: err=%d, port=%d\n", 1001 ret, portid); 1002 } 1003 1004 print_ethaddr(" Address:", &ports_eth_addr[portid]); 1005 printf("\n"); 1006 1007 /* init one TX queue per couple (lcore,port) */ 1008 ret = rte_eth_dev_info_get(portid, &dev_info); 1009 if (ret != 0) 1010 rte_exit(EXIT_FAILURE, 1011 "Error during getting device (port %u) info: %s\n", 1012 portid, strerror(-ret)); 1013 1014 queueid = 0; 1015 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) { 1016 if (rte_lcore_is_enabled(lcore_id) == 0) 1017 continue; 1018 1019 if (queueid >= dev_info.nb_tx_queues) 1020 break; 1021 1022 socket = (int) rte_lcore_to_socket_id(lcore_id); 1023 printf("txq=%u,%d ", lcore_id, queueid); 1024 fflush(stdout); 1025 1026 txconf = &dev_info.default_txconf; 1027 txconf->offloads = local_port_conf.txmode.offloads; 1028 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd, 1029 socket, txconf); 1030 if (ret < 0) { 1031 printf("\n"); 1032 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: " 1033 "err=%d, port=%d\n", ret, portid); 1034 } 1035 1036 qconf = &lcore_queue_conf[lcore_id]; 1037 qconf->tx_queue_id[portid] = queueid; 1038 queueid++; 1039 } 1040 1041 printf("\n"); 1042 } 1043 1044 printf("\n"); 1045 1046 /* start ports */ 1047 RTE_ETH_FOREACH_DEV(portid) { 1048 if ((enabled_port_mask & (1 << portid)) == 0) { 1049 continue; 1050 } 1051 /* Start device */ 1052 ret = rte_eth_dev_start(portid); 1053 if (ret < 0) 1054 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n", 1055 ret, portid); 1056 1057 ret = rte_eth_promiscuous_enable(portid); 1058 if (ret != 0) 1059 rte_exit(EXIT_FAILURE, 1060 "rte_eth_promiscuous_enable: err=%s, port=%d\n", 1061 rte_strerror(-ret), portid); 1062 1063 if (check_ptype(portid) == 0) { 1064 rte_eth_add_rx_callback(portid, 0, cb_parse_ptype, NULL); 1065 printf("Add Rx callback function to detect L3 packet type by SW :" 1066 " port = %d\n", portid); 1067 } 1068 } 1069 1070 if (init_routing_table() < 0) 1071 rte_exit(EXIT_FAILURE, "Cannot init routing table\n"); 1072 1073 check_all_ports_link_status(enabled_port_mask); 1074 1075 /* launch per-lcore init on every lcore */ 1076 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MAIN); 1077 RTE_LCORE_FOREACH_WORKER(lcore_id) { 1078 if (rte_eal_wait_lcore(lcore_id) < 0) 1079 return -1; 1080 } 1081 1082 /* clean up the EAL */ 1083 rte_eal_cleanup(); 1084 1085 return 0; 1086 } 1087