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_eal.h> 54 #include <rte_per_lcore.h> 55 #include <rte_launch.h> 56 #include <rte_atomic.h> 57 #include <rte_cycles.h> 58 #include <rte_prefetch.h> 59 #include <rte_lcore.h> 60 #include <rte_per_lcore.h> 61 #include <rte_branch_prediction.h> 62 #include <rte_interrupts.h> 63 #include <rte_pci.h> 64 #include <rte_random.h> 65 #include <rte_debug.h> 66 #include <rte_ether.h> 67 #include <rte_ethdev.h> 68 #include <rte_ring.h> 69 #include <rte_mempool.h> 70 #include <rte_mbuf.h> 71 #include <rte_malloc.h> 72 #include <rte_ip.h> 73 #include <rte_tcp.h> 74 #include <rte_udp.h> 75 #include <rte_string_fns.h> 76 #include <rte_lpm.h> 77 #include <rte_lpm6.h> 78 79 #include <rte_ip_frag.h> 80 81 #define MAX_PKT_BURST 32 82 83 84 #define RTE_LOGTYPE_IP_RSMBL RTE_LOGTYPE_USER1 85 86 #define MAX_JUMBO_PKT_LEN 9600 87 88 #define BUF_SIZE RTE_MBUF_DEFAULT_DATAROOM 89 #define MBUF_SIZE \ 90 (BUF_SIZE + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM) 91 92 #define NB_MBUF 8192 93 94 /* allow max jumbo frame 9.5 KB */ 95 #define JUMBO_FRAME_MAX_SIZE 0x2600 96 97 #define MAX_FLOW_NUM UINT16_MAX 98 #define MIN_FLOW_NUM 1 99 #define DEF_FLOW_NUM 0x1000 100 101 /* TTL numbers are in ms. */ 102 #define MAX_FLOW_TTL (3600 * MS_PER_S) 103 #define MIN_FLOW_TTL 1 104 #define DEF_FLOW_TTL MS_PER_S 105 106 #define MAX_FRAG_NUM RTE_LIBRTE_IP_FRAG_MAX_FRAG 107 108 /* Should be power of two. */ 109 #define IP_FRAG_TBL_BUCKET_ENTRIES 16 110 111 static uint32_t max_flow_num = DEF_FLOW_NUM; 112 static uint32_t max_flow_ttl = DEF_FLOW_TTL; 113 114 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */ 115 116 #define NB_SOCKETS 8 117 118 /* Configure how many packets ahead to prefetch, when reading packets */ 119 #define PREFETCH_OFFSET 3 120 121 /* 122 * Configurable number of RX/TX ring descriptors 123 */ 124 #define RTE_TEST_RX_DESC_DEFAULT 128 125 #define RTE_TEST_TX_DESC_DEFAULT 512 126 127 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT; 128 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT; 129 130 /* ethernet addresses of ports */ 131 static struct ether_addr ports_eth_addr[RTE_MAX_ETHPORTS]; 132 133 #ifndef IPv4_BYTES 134 #define IPv4_BYTES_FMT "%" PRIu8 ".%" PRIu8 ".%" PRIu8 ".%" PRIu8 135 #define IPv4_BYTES(addr) \ 136 (uint8_t) (((addr) >> 24) & 0xFF),\ 137 (uint8_t) (((addr) >> 16) & 0xFF),\ 138 (uint8_t) (((addr) >> 8) & 0xFF),\ 139 (uint8_t) ((addr) & 0xFF) 140 #endif 141 142 #ifndef IPv6_BYTES 143 #define IPv6_BYTES_FMT "%02x%02x:%02x%02x:%02x%02x:%02x%02x:"\ 144 "%02x%02x:%02x%02x:%02x%02x:%02x%02x" 145 #define IPv6_BYTES(addr) \ 146 addr[0], addr[1], addr[2], addr[3], \ 147 addr[4], addr[5], addr[6], addr[7], \ 148 addr[8], addr[9], addr[10], addr[11],\ 149 addr[12], addr[13],addr[14], addr[15] 150 #endif 151 152 #define IPV6_ADDR_LEN 16 153 154 /* mask of enabled ports */ 155 static uint32_t enabled_port_mask = 0; 156 157 static int rx_queue_per_lcore = 1; 158 159 struct mbuf_table { 160 uint32_t len; 161 uint32_t head; 162 uint32_t tail; 163 struct rte_mbuf *m_table[0]; 164 }; 165 166 struct rx_queue { 167 struct rte_ip_frag_tbl *frag_tbl; 168 struct rte_mempool *pool; 169 struct rte_lpm *lpm; 170 struct rte_lpm6 *lpm6; 171 uint8_t portid; 172 }; 173 174 struct tx_lcore_stat { 175 uint64_t call; 176 uint64_t drop; 177 uint64_t queue; 178 uint64_t send; 179 }; 180 181 #define MAX_RX_QUEUE_PER_LCORE 16 182 #define MAX_TX_QUEUE_PER_PORT 16 183 #define MAX_RX_QUEUE_PER_PORT 128 184 185 struct lcore_queue_conf { 186 uint16_t n_rx_queue; 187 struct rx_queue rx_queue_list[MAX_RX_QUEUE_PER_LCORE]; 188 uint16_t tx_queue_id[RTE_MAX_ETHPORTS]; 189 struct rte_ip_frag_death_row death_row; 190 struct mbuf_table *tx_mbufs[RTE_MAX_ETHPORTS]; 191 struct tx_lcore_stat tx_stat; 192 } __rte_cache_aligned; 193 static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE]; 194 195 static struct rte_eth_conf port_conf = { 196 .rxmode = { 197 .mq_mode = ETH_MQ_RX_RSS, 198 .max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE, 199 .split_hdr_size = 0, 200 .header_split = 0, /**< Header Split disabled */ 201 .hw_ip_checksum = 1, /**< IP checksum offload enabled */ 202 .hw_vlan_filter = 0, /**< VLAN filtering disabled */ 203 .jumbo_frame = 1, /**< Jumbo Frame Support disabled */ 204 .hw_strip_crc = 0, /**< CRC stripped by hardware */ 205 }, 206 .rx_adv_conf = { 207 .rss_conf = { 208 .rss_key = NULL, 209 .rss_hf = ETH_RSS_IP, 210 }, 211 }, 212 .txmode = { 213 .mq_mode = ETH_MQ_TX_NONE, 214 }, 215 }; 216 217 /* 218 * IPv4 forwarding table 219 */ 220 struct l3fwd_ipv4_route { 221 uint32_t ip; 222 uint8_t depth; 223 uint8_t if_out; 224 }; 225 226 struct l3fwd_ipv4_route l3fwd_ipv4_route_array[] = { 227 {IPv4(100,10,0,0), 16, 0}, 228 {IPv4(100,20,0,0), 16, 1}, 229 {IPv4(100,30,0,0), 16, 2}, 230 {IPv4(100,40,0,0), 16, 3}, 231 {IPv4(100,50,0,0), 16, 4}, 232 {IPv4(100,60,0,0), 16, 5}, 233 {IPv4(100,70,0,0), 16, 6}, 234 {IPv4(100,80,0,0), 16, 7}, 235 }; 236 237 /* 238 * IPv6 forwarding table 239 */ 240 241 struct l3fwd_ipv6_route { 242 uint8_t ip[IPV6_ADDR_LEN]; 243 uint8_t depth; 244 uint8_t if_out; 245 }; 246 247 static struct l3fwd_ipv6_route l3fwd_ipv6_route_array[] = { 248 {{1,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 0}, 249 {{2,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 1}, 250 {{3,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 2}, 251 {{4,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 3}, 252 {{5,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 4}, 253 {{6,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 5}, 254 {{7,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 6}, 255 {{8,1,1,1,1,1,1,1,1,1,1,1,1,1,1,1}, 48, 7}, 256 }; 257 258 #define LPM_MAX_RULES 1024 259 #define LPM6_MAX_RULES 1024 260 #define LPM6_NUMBER_TBL8S (1 << 16) 261 262 struct rte_lpm6_config lpm6_config = { 263 .max_rules = LPM6_MAX_RULES, 264 .number_tbl8s = LPM6_NUMBER_TBL8S, 265 .flags = 0 266 }; 267 268 static struct rte_lpm *socket_lpm[RTE_MAX_NUMA_NODES]; 269 static struct rte_lpm6 *socket_lpm6[RTE_MAX_NUMA_NODES]; 270 271 #ifdef RTE_LIBRTE_IP_FRAG_TBL_STAT 272 #define TX_LCORE_STAT_UPDATE(s, f, v) ((s)->f += (v)) 273 #else 274 #define TX_LCORE_STAT_UPDATE(s, f, v) do {} while (0) 275 #endif /* RTE_LIBRTE_IP_FRAG_TBL_STAT */ 276 277 /* 278 * If number of queued packets reached given threahold, then 279 * send burst of packets on an output interface. 280 */ 281 static inline uint32_t 282 send_burst(struct lcore_queue_conf *qconf, uint32_t thresh, uint8_t port) 283 { 284 uint32_t fill, len, k, n; 285 struct mbuf_table *txmb; 286 287 txmb = qconf->tx_mbufs[port]; 288 len = txmb->len; 289 290 if ((int32_t)(fill = txmb->head - txmb->tail) < 0) 291 fill += len; 292 293 if (fill >= thresh) { 294 n = RTE_MIN(len - txmb->tail, fill); 295 296 k = rte_eth_tx_burst(port, qconf->tx_queue_id[port], 297 txmb->m_table + txmb->tail, (uint16_t)n); 298 299 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, call, 1); 300 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, send, k); 301 302 fill -= k; 303 if ((txmb->tail += k) == len) 304 txmb->tail = 0; 305 } 306 307 return fill; 308 } 309 310 /* Enqueue a single packet, and send burst if queue is filled */ 311 static inline int 312 send_single_packet(struct rte_mbuf *m, uint8_t port) 313 { 314 uint32_t fill, lcore_id, len; 315 struct lcore_queue_conf *qconf; 316 struct mbuf_table *txmb; 317 318 lcore_id = rte_lcore_id(); 319 qconf = &lcore_queue_conf[lcore_id]; 320 321 txmb = qconf->tx_mbufs[port]; 322 len = txmb->len; 323 324 fill = send_burst(qconf, MAX_PKT_BURST, port); 325 326 if (fill == len - 1) { 327 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, drop, 1); 328 rte_pktmbuf_free(txmb->m_table[txmb->tail]); 329 if (++txmb->tail == len) 330 txmb->tail = 0; 331 } 332 333 TX_LCORE_STAT_UPDATE(&qconf->tx_stat, queue, 1); 334 txmb->m_table[txmb->head] = m; 335 if(++txmb->head == len) 336 txmb->head = 0; 337 338 return 0; 339 } 340 341 static inline void 342 reassemble(struct rte_mbuf *m, uint8_t portid, uint32_t queue, 343 struct lcore_queue_conf *qconf, uint64_t tms) 344 { 345 struct ether_hdr *eth_hdr; 346 struct rte_ip_frag_tbl *tbl; 347 struct rte_ip_frag_death_row *dr; 348 struct rx_queue *rxq; 349 void *d_addr_bytes; 350 uint32_t next_hop_ipv4; 351 uint8_t next_hop_ipv6, dst_port; 352 353 rxq = &qconf->rx_queue_list[queue]; 354 355 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *); 356 357 dst_port = portid; 358 359 /* if packet is IPv4 */ 360 if (RTE_ETH_IS_IPV4_HDR(m->packet_type)) { 361 struct ipv4_hdr *ip_hdr; 362 uint32_t ip_dst; 363 364 ip_hdr = (struct ipv4_hdr *)(eth_hdr + 1); 365 366 /* if it is a fragmented packet, then try to reassemble. */ 367 if (rte_ipv4_frag_pkt_is_fragmented(ip_hdr)) { 368 struct rte_mbuf *mo; 369 370 tbl = rxq->frag_tbl; 371 dr = &qconf->death_row; 372 373 /* prepare mbuf: setup l2_len/l3_len. */ 374 m->l2_len = sizeof(*eth_hdr); 375 m->l3_len = sizeof(*ip_hdr); 376 377 /* process this fragment. */ 378 mo = rte_ipv4_frag_reassemble_packet(tbl, dr, m, tms, ip_hdr); 379 if (mo == NULL) 380 /* no packet to send out. */ 381 return; 382 383 /* we have our packet reassembled. */ 384 if (mo != m) { 385 m = mo; 386 eth_hdr = rte_pktmbuf_mtod(m, 387 struct ether_hdr *); 388 ip_hdr = (struct ipv4_hdr *)(eth_hdr + 1); 389 } 390 } 391 ip_dst = rte_be_to_cpu_32(ip_hdr->dst_addr); 392 393 /* Find destination port */ 394 if (rte_lpm_lookup(rxq->lpm, ip_dst, &next_hop_ipv4) == 0 && 395 (enabled_port_mask & 1 << next_hop_ipv4) != 0) { 396 dst_port = next_hop_ipv4; 397 } 398 399 eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4); 400 } else if (RTE_ETH_IS_IPV6_HDR(m->packet_type)) { 401 /* if packet is IPv6 */ 402 struct ipv6_extension_fragment *frag_hdr; 403 struct ipv6_hdr *ip_hdr; 404 405 ip_hdr = (struct ipv6_hdr *)(eth_hdr + 1); 406 407 frag_hdr = rte_ipv6_frag_get_ipv6_fragment_header(ip_hdr); 408 409 if (frag_hdr != NULL) { 410 struct rte_mbuf *mo; 411 412 tbl = rxq->frag_tbl; 413 dr = &qconf->death_row; 414 415 /* prepare mbuf: setup l2_len/l3_len. */ 416 m->l2_len = sizeof(*eth_hdr); 417 m->l3_len = sizeof(*ip_hdr) + sizeof(*frag_hdr); 418 419 mo = rte_ipv6_frag_reassemble_packet(tbl, dr, m, tms, ip_hdr, frag_hdr); 420 if (mo == NULL) 421 return; 422 423 if (mo != m) { 424 m = mo; 425 eth_hdr = rte_pktmbuf_mtod(m, struct ether_hdr *); 426 ip_hdr = (struct ipv6_hdr *)(eth_hdr + 1); 427 } 428 } 429 430 /* Find destination port */ 431 if (rte_lpm6_lookup(rxq->lpm6, ip_hdr->dst_addr, &next_hop_ipv6) == 0 && 432 (enabled_port_mask & 1 << next_hop_ipv6) != 0) { 433 dst_port = next_hop_ipv6; 434 } 435 436 eth_hdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv6); 437 } 438 /* if packet wasn't IPv4 or IPv6, it's forwarded to the port it came from */ 439 440 /* 02:00:00:00:00:xx */ 441 d_addr_bytes = ð_hdr->d_addr.addr_bytes[0]; 442 *((uint64_t *)d_addr_bytes) = 0x000000000002 + ((uint64_t)dst_port << 40); 443 444 /* src addr */ 445 ether_addr_copy(&ports_eth_addr[dst_port], ð_hdr->s_addr); 446 447 send_single_packet(m, dst_port); 448 } 449 450 /* main processing loop */ 451 static int 452 main_loop(__attribute__((unused)) void *dummy) 453 { 454 struct rte_mbuf *pkts_burst[MAX_PKT_BURST]; 455 unsigned lcore_id; 456 uint64_t diff_tsc, cur_tsc, prev_tsc; 457 int i, j, nb_rx; 458 uint8_t portid; 459 struct lcore_queue_conf *qconf; 460 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US; 461 462 prev_tsc = 0; 463 464 lcore_id = rte_lcore_id(); 465 qconf = &lcore_queue_conf[lcore_id]; 466 467 if (qconf->n_rx_queue == 0) { 468 RTE_LOG(INFO, IP_RSMBL, "lcore %u has nothing to do\n", lcore_id); 469 return 0; 470 } 471 472 RTE_LOG(INFO, IP_RSMBL, "entering main loop on lcore %u\n", lcore_id); 473 474 for (i = 0; i < qconf->n_rx_queue; i++) { 475 476 portid = qconf->rx_queue_list[i].portid; 477 RTE_LOG(INFO, IP_RSMBL, " -- lcoreid=%u portid=%hhu\n", lcore_id, 478 portid); 479 } 480 481 while (1) { 482 483 cur_tsc = rte_rdtsc(); 484 485 /* 486 * TX burst queue drain 487 */ 488 diff_tsc = cur_tsc - prev_tsc; 489 if (unlikely(diff_tsc > drain_tsc)) { 490 491 /* 492 * This could be optimized (use queueid instead of 493 * portid), but it is not called so often 494 */ 495 for (portid = 0; portid < RTE_MAX_ETHPORTS; portid++) { 496 if ((enabled_port_mask & (1 << portid)) != 0) 497 send_burst(qconf, 1, portid); 498 } 499 500 prev_tsc = cur_tsc; 501 } 502 503 /* 504 * Read packet from RX queues 505 */ 506 for (i = 0; i < qconf->n_rx_queue; ++i) { 507 508 portid = qconf->rx_queue_list[i].portid; 509 510 nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst, 511 MAX_PKT_BURST); 512 513 /* Prefetch first packets */ 514 for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) { 515 rte_prefetch0(rte_pktmbuf_mtod( 516 pkts_burst[j], void *)); 517 } 518 519 /* Prefetch and forward already prefetched packets */ 520 for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) { 521 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[ 522 j + PREFETCH_OFFSET], void *)); 523 reassemble(pkts_burst[j], portid, 524 i, qconf, cur_tsc); 525 } 526 527 /* Forward remaining prefetched packets */ 528 for (; j < nb_rx; j++) { 529 reassemble(pkts_burst[j], portid, 530 i, qconf, cur_tsc); 531 } 532 533 rte_ip_frag_free_death_row(&qconf->death_row, 534 PREFETCH_OFFSET); 535 } 536 } 537 } 538 539 /* display usage */ 540 static void 541 print_usage(const char *prgname) 542 { 543 printf("%s [EAL options] -- -p PORTMASK [-q NQ]" 544 " [--max-pkt-len PKTLEN]" 545 " [--maxflows=<flows>] [--flowttl=<ttl>[(s|ms)]]\n" 546 " -p PORTMASK: hexadecimal bitmask of ports to configure\n" 547 " -q NQ: number of RX queues per lcore\n" 548 " --maxflows=<flows>: optional, maximum number of flows " 549 "supported\n" 550 " --flowttl=<ttl>[(s|ms)]: optional, maximum TTL for each " 551 "flow\n", 552 prgname); 553 } 554 555 static uint32_t 556 parse_flow_num(const char *str, uint32_t min, uint32_t max, uint32_t *val) 557 { 558 char *end; 559 uint64_t v; 560 561 /* parse decimal string */ 562 errno = 0; 563 v = strtoul(str, &end, 10); 564 if (errno != 0 || *end != '\0') 565 return -EINVAL; 566 567 if (v < min || v > max) 568 return -EINVAL; 569 570 *val = (uint32_t)v; 571 return 0; 572 } 573 574 static int 575 parse_flow_ttl(const char *str, uint32_t min, uint32_t max, uint32_t *val) 576 { 577 char *end; 578 uint64_t v; 579 580 static const char frmt_sec[] = "s"; 581 static const char frmt_msec[] = "ms"; 582 583 /* parse decimal string */ 584 errno = 0; 585 v = strtoul(str, &end, 10); 586 if (errno != 0) 587 return -EINVAL; 588 589 if (*end != '\0') { 590 if (strncmp(frmt_sec, end, sizeof(frmt_sec)) == 0) 591 v *= MS_PER_S; 592 else if (strncmp(frmt_msec, end, sizeof (frmt_msec)) != 0) 593 return -EINVAL; 594 } 595 596 if (v < min || v > max) 597 return -EINVAL; 598 599 *val = (uint32_t)v; 600 return 0; 601 } 602 603 static int 604 parse_portmask(const char *portmask) 605 { 606 char *end = NULL; 607 unsigned long pm; 608 609 /* parse hexadecimal string */ 610 pm = strtoul(portmask, &end, 16); 611 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0')) 612 return -1; 613 614 if (pm == 0) 615 return -1; 616 617 return pm; 618 } 619 620 static int 621 parse_nqueue(const char *q_arg) 622 { 623 char *end = NULL; 624 unsigned long n; 625 626 printf("%p\n", q_arg); 627 628 /* parse hexadecimal string */ 629 n = strtoul(q_arg, &end, 10); 630 if ((q_arg[0] == '\0') || (end == NULL) || (*end != '\0')) 631 return -1; 632 if (n == 0) 633 return -1; 634 if (n >= MAX_RX_QUEUE_PER_LCORE) 635 return -1; 636 637 return n; 638 } 639 640 /* Parse the argument given in the command line of the application */ 641 static int 642 parse_args(int argc, char **argv) 643 { 644 int opt, ret; 645 char **argvopt; 646 int option_index; 647 char *prgname = argv[0]; 648 static struct option lgopts[] = { 649 {"max-pkt-len", 1, 0, 0}, 650 {"maxflows", 1, 0, 0}, 651 {"flowttl", 1, 0, 0}, 652 {NULL, 0, 0, 0} 653 }; 654 655 argvopt = argv; 656 657 while ((opt = getopt_long(argc, argvopt, "p:q:", 658 lgopts, &option_index)) != EOF) { 659 660 switch (opt) { 661 /* portmask */ 662 case 'p': 663 enabled_port_mask = parse_portmask(optarg); 664 if (enabled_port_mask == 0) { 665 printf("invalid portmask\n"); 666 print_usage(prgname); 667 return -1; 668 } 669 break; 670 671 /* nqueue */ 672 case 'q': 673 rx_queue_per_lcore = parse_nqueue(optarg); 674 if (rx_queue_per_lcore < 0) { 675 printf("invalid queue number\n"); 676 print_usage(prgname); 677 return -1; 678 } 679 break; 680 681 /* long options */ 682 case 0: 683 if (!strncmp(lgopts[option_index].name, 684 "maxflows", 8)) { 685 if ((ret = parse_flow_num(optarg, MIN_FLOW_NUM, 686 MAX_FLOW_NUM, 687 &max_flow_num)) != 0) { 688 printf("invalid value: \"%s\" for " 689 "parameter %s\n", 690 optarg, 691 lgopts[option_index].name); 692 print_usage(prgname); 693 return ret; 694 } 695 } 696 697 if (!strncmp(lgopts[option_index].name, "flowttl", 7)) { 698 if ((ret = parse_flow_ttl(optarg, MIN_FLOW_TTL, 699 MAX_FLOW_TTL, 700 &max_flow_ttl)) != 0) { 701 printf("invalid value: \"%s\" for " 702 "parameter %s\n", 703 optarg, 704 lgopts[option_index].name); 705 print_usage(prgname); 706 return ret; 707 } 708 } 709 710 break; 711 712 default: 713 print_usage(prgname); 714 return -1; 715 } 716 } 717 718 if (optind >= 0) 719 argv[optind-1] = prgname; 720 721 ret = optind-1; 722 optind = 0; /* reset getopt lib */ 723 return ret; 724 } 725 726 static void 727 print_ethaddr(const char *name, const struct ether_addr *eth_addr) 728 { 729 char buf[ETHER_ADDR_FMT_SIZE]; 730 ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr); 731 printf("%s%s", name, buf); 732 } 733 734 /* Check the link status of all ports in up to 9s, and print them finally */ 735 static void 736 check_all_ports_link_status(uint8_t port_num, uint32_t port_mask) 737 { 738 #define CHECK_INTERVAL 100 /* 100ms */ 739 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */ 740 uint8_t portid, count, all_ports_up, print_flag = 0; 741 struct rte_eth_link link; 742 743 printf("\nChecking link status"); 744 fflush(stdout); 745 for (count = 0; count <= MAX_CHECK_TIME; count++) { 746 all_ports_up = 1; 747 for (portid = 0; portid < port_num; portid++) { 748 if ((port_mask & (1 << portid)) == 0) 749 continue; 750 memset(&link, 0, sizeof(link)); 751 rte_eth_link_get_nowait(portid, &link); 752 /* print link status if flag set */ 753 if (print_flag == 1) { 754 if (link.link_status) 755 printf("Port %d Link Up - speed %u " 756 "Mbps - %s\n", (uint8_t)portid, 757 (unsigned)link.link_speed, 758 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ? 759 ("full-duplex") : ("half-duplex\n")); 760 else 761 printf("Port %d Link Down\n", 762 (uint8_t)portid); 763 continue; 764 } 765 /* clear all_ports_up flag if any link down */ 766 if (link.link_status == 0) { 767 all_ports_up = 0; 768 break; 769 } 770 } 771 /* after finally printing all link status, get out */ 772 if (print_flag == 1) 773 break; 774 775 if (all_ports_up == 0) { 776 printf("."); 777 fflush(stdout); 778 rte_delay_ms(CHECK_INTERVAL); 779 } 780 781 /* set the print_flag if all ports up or timeout */ 782 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) { 783 print_flag = 1; 784 printf("\ndone\n"); 785 } 786 } 787 } 788 789 static int 790 init_routing_table(void) 791 { 792 struct rte_lpm *lpm; 793 struct rte_lpm6 *lpm6; 794 int socket, ret; 795 unsigned i; 796 797 for (socket = 0; socket < RTE_MAX_NUMA_NODES; socket++) { 798 if (socket_lpm[socket]) { 799 lpm = socket_lpm[socket]; 800 /* populate the LPM table */ 801 for (i = 0; i < RTE_DIM(l3fwd_ipv4_route_array); i++) { 802 ret = rte_lpm_add(lpm, 803 l3fwd_ipv4_route_array[i].ip, 804 l3fwd_ipv4_route_array[i].depth, 805 l3fwd_ipv4_route_array[i].if_out); 806 807 if (ret < 0) { 808 RTE_LOG(ERR, IP_RSMBL, "Unable to add entry %i to the l3fwd " 809 "LPM table\n", i); 810 return -1; 811 } 812 813 RTE_LOG(INFO, IP_RSMBL, "Socket %i: adding route " IPv4_BYTES_FMT 814 "/%d (port %d)\n", 815 socket, 816 IPv4_BYTES(l3fwd_ipv4_route_array[i].ip), 817 l3fwd_ipv4_route_array[i].depth, 818 l3fwd_ipv4_route_array[i].if_out); 819 } 820 } 821 822 if (socket_lpm6[socket]) { 823 lpm6 = socket_lpm6[socket]; 824 /* populate the LPM6 table */ 825 for (i = 0; i < RTE_DIM(l3fwd_ipv6_route_array); i++) { 826 ret = rte_lpm6_add(lpm6, 827 l3fwd_ipv6_route_array[i].ip, 828 l3fwd_ipv6_route_array[i].depth, 829 l3fwd_ipv6_route_array[i].if_out); 830 831 if (ret < 0) { 832 RTE_LOG(ERR, IP_RSMBL, "Unable to add entry %i to the l3fwd " 833 "LPM6 table\n", i); 834 return -1; 835 } 836 837 RTE_LOG(INFO, IP_RSMBL, "Socket %i: adding route " IPv6_BYTES_FMT 838 "/%d (port %d)\n", 839 socket, 840 IPv6_BYTES(l3fwd_ipv6_route_array[i].ip), 841 l3fwd_ipv6_route_array[i].depth, 842 l3fwd_ipv6_route_array[i].if_out); 843 } 844 } 845 } 846 return 0; 847 } 848 849 static int 850 setup_port_tbl(struct lcore_queue_conf *qconf, uint32_t lcore, int socket, 851 uint32_t port) 852 { 853 struct mbuf_table *mtb; 854 uint32_t n; 855 size_t sz; 856 857 n = RTE_MAX(max_flow_num, 2UL * MAX_PKT_BURST); 858 sz = sizeof (*mtb) + sizeof (mtb->m_table[0]) * n; 859 860 if ((mtb = rte_zmalloc_socket(__func__, sz, RTE_CACHE_LINE_SIZE, 861 socket)) == NULL) { 862 RTE_LOG(ERR, IP_RSMBL, "%s() for lcore: %u, port: %u " 863 "failed to allocate %zu bytes\n", 864 __func__, lcore, port, sz); 865 return -1; 866 } 867 868 mtb->len = n; 869 qconf->tx_mbufs[port] = mtb; 870 871 return 0; 872 } 873 874 static int 875 setup_queue_tbl(struct rx_queue *rxq, uint32_t lcore, uint32_t queue) 876 { 877 int socket; 878 uint32_t nb_mbuf; 879 uint64_t frag_cycles; 880 char buf[RTE_MEMPOOL_NAMESIZE]; 881 882 socket = rte_lcore_to_socket_id(lcore); 883 if (socket == SOCKET_ID_ANY) 884 socket = 0; 885 886 frag_cycles = (rte_get_tsc_hz() + MS_PER_S - 1) / MS_PER_S * 887 max_flow_ttl; 888 889 if ((rxq->frag_tbl = rte_ip_frag_table_create(max_flow_num, 890 IP_FRAG_TBL_BUCKET_ENTRIES, max_flow_num, frag_cycles, 891 socket)) == NULL) { 892 RTE_LOG(ERR, IP_RSMBL, "ip_frag_tbl_create(%u) on " 893 "lcore: %u for queue: %u failed\n", 894 max_flow_num, lcore, queue); 895 return -1; 896 } 897 898 /* 899 * At any given moment up to <max_flow_num * (MAX_FRAG_NUM)> 900 * mbufs could be stored int the fragment table. 901 * Plus, each TX queue can hold up to <max_flow_num> packets. 902 */ 903 904 nb_mbuf = RTE_MAX(max_flow_num, 2UL * MAX_PKT_BURST) * MAX_FRAG_NUM; 905 nb_mbuf *= (port_conf.rxmode.max_rx_pkt_len + BUF_SIZE - 1) / BUF_SIZE; 906 nb_mbuf *= 2; /* ipv4 and ipv6 */ 907 nb_mbuf += RTE_TEST_RX_DESC_DEFAULT + RTE_TEST_TX_DESC_DEFAULT; 908 909 nb_mbuf = RTE_MAX(nb_mbuf, (uint32_t)NB_MBUF); 910 911 snprintf(buf, sizeof(buf), "mbuf_pool_%u_%u", lcore, queue); 912 913 if ((rxq->pool = rte_mempool_create(buf, nb_mbuf, MBUF_SIZE, 0, 914 sizeof(struct rte_pktmbuf_pool_private), 915 rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL, 916 socket, MEMPOOL_F_SP_PUT | MEMPOOL_F_SC_GET)) == NULL) { 917 RTE_LOG(ERR, IP_RSMBL, "mempool_create(%s) failed", buf); 918 return -1; 919 } 920 921 return 0; 922 } 923 924 static int 925 init_mem(void) 926 { 927 char buf[PATH_MAX]; 928 struct rte_lpm *lpm; 929 struct rte_lpm6 *lpm6; 930 struct rte_lpm_config lpm_config; 931 int socket; 932 unsigned lcore_id; 933 934 /* traverse through lcores and initialize structures on each socket */ 935 936 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) { 937 938 if (rte_lcore_is_enabled(lcore_id) == 0) 939 continue; 940 941 socket = rte_lcore_to_socket_id(lcore_id); 942 943 if (socket == SOCKET_ID_ANY) 944 socket = 0; 945 946 if (socket_lpm[socket] == NULL) { 947 RTE_LOG(INFO, IP_RSMBL, "Creating LPM table on socket %i\n", socket); 948 snprintf(buf, sizeof(buf), "IP_RSMBL_LPM_%i", socket); 949 950 lpm_config.max_rules = LPM_MAX_RULES; 951 lpm_config.number_tbl8s = 256; 952 lpm_config.flags = 0; 953 954 lpm = rte_lpm_create(buf, socket, &lpm_config); 955 if (lpm == NULL) { 956 RTE_LOG(ERR, IP_RSMBL, "Cannot create LPM table\n"); 957 return -1; 958 } 959 socket_lpm[socket] = lpm; 960 } 961 962 if (socket_lpm6[socket] == NULL) { 963 RTE_LOG(INFO, IP_RSMBL, "Creating LPM6 table on socket %i\n", socket); 964 snprintf(buf, sizeof(buf), "IP_RSMBL_LPM_%i", socket); 965 966 lpm6 = rte_lpm6_create("IP_RSMBL_LPM6", socket, &lpm6_config); 967 if (lpm6 == NULL) { 968 RTE_LOG(ERR, IP_RSMBL, "Cannot create LPM table\n"); 969 return -1; 970 } 971 socket_lpm6[socket] = lpm6; 972 } 973 } 974 975 return 0; 976 } 977 978 static void 979 queue_dump_stat(void) 980 { 981 uint32_t i, lcore; 982 const struct lcore_queue_conf *qconf; 983 984 for (lcore = 0; lcore < RTE_MAX_LCORE; lcore++) { 985 if (rte_lcore_is_enabled(lcore) == 0) 986 continue; 987 988 qconf = &lcore_queue_conf[lcore]; 989 for (i = 0; i < qconf->n_rx_queue; i++) { 990 991 fprintf(stdout, " -- lcoreid=%u portid=%hhu " 992 "frag tbl stat:\n", 993 lcore, qconf->rx_queue_list[i].portid); 994 rte_ip_frag_table_statistics_dump(stdout, 995 qconf->rx_queue_list[i].frag_tbl); 996 fprintf(stdout, "TX bursts:\t%" PRIu64 "\n" 997 "TX packets _queued:\t%" PRIu64 "\n" 998 "TX packets dropped:\t%" PRIu64 "\n" 999 "TX packets send:\t%" PRIu64 "\n", 1000 qconf->tx_stat.call, 1001 qconf->tx_stat.queue, 1002 qconf->tx_stat.drop, 1003 qconf->tx_stat.send); 1004 } 1005 } 1006 } 1007 1008 static void 1009 signal_handler(int signum) 1010 { 1011 queue_dump_stat(); 1012 if (signum != SIGUSR1) 1013 rte_exit(0, "received signal: %d, exiting\n", signum); 1014 } 1015 1016 int 1017 main(int argc, char **argv) 1018 { 1019 struct lcore_queue_conf *qconf; 1020 struct rte_eth_dev_info dev_info; 1021 struct rte_eth_txconf *txconf; 1022 struct rx_queue *rxq; 1023 int ret, socket; 1024 unsigned nb_ports; 1025 uint16_t queueid; 1026 unsigned lcore_id = 0, rx_lcore_id = 0; 1027 uint32_t n_tx_queue, nb_lcores; 1028 uint8_t portid; 1029 1030 /* init EAL */ 1031 ret = rte_eal_init(argc, argv); 1032 if (ret < 0) 1033 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n"); 1034 argc -= ret; 1035 argv += ret; 1036 1037 /* parse application arguments (after the EAL ones) */ 1038 ret = parse_args(argc, argv); 1039 if (ret < 0) 1040 rte_exit(EXIT_FAILURE, "Invalid IP reassembly parameters\n"); 1041 1042 nb_ports = rte_eth_dev_count(); 1043 if (nb_ports > RTE_MAX_ETHPORTS) 1044 nb_ports = RTE_MAX_ETHPORTS; 1045 else if (nb_ports == 0) 1046 rte_exit(EXIT_FAILURE, "No ports found!\n"); 1047 1048 nb_lcores = rte_lcore_count(); 1049 1050 /* initialize structures (mempools, lpm etc.) */ 1051 if (init_mem() < 0) 1052 rte_panic("Cannot initialize memory structures!\n"); 1053 1054 /* check if portmask has non-existent ports */ 1055 if (enabled_port_mask & ~(RTE_LEN2MASK(nb_ports, unsigned))) 1056 rte_exit(EXIT_FAILURE, "Non-existent ports in portmask!\n"); 1057 1058 /* initialize all ports */ 1059 for (portid = 0; portid < nb_ports; portid++) { 1060 /* skip ports that are not enabled */ 1061 if ((enabled_port_mask & (1 << portid)) == 0) { 1062 printf("\nSkipping disabled port %d\n", portid); 1063 continue; 1064 } 1065 1066 qconf = &lcore_queue_conf[rx_lcore_id]; 1067 1068 /* get the lcore_id for this port */ 1069 while (rte_lcore_is_enabled(rx_lcore_id) == 0 || 1070 qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) { 1071 1072 rx_lcore_id++; 1073 if (rx_lcore_id >= RTE_MAX_LCORE) 1074 rte_exit(EXIT_FAILURE, "Not enough cores\n"); 1075 1076 qconf = &lcore_queue_conf[rx_lcore_id]; 1077 } 1078 1079 socket = rte_lcore_to_socket_id(portid); 1080 if (socket == SOCKET_ID_ANY) 1081 socket = 0; 1082 1083 queueid = qconf->n_rx_queue; 1084 rxq = &qconf->rx_queue_list[queueid]; 1085 rxq->portid = portid; 1086 rxq->lpm = socket_lpm[socket]; 1087 rxq->lpm6 = socket_lpm6[socket]; 1088 if (setup_queue_tbl(rxq, rx_lcore_id, queueid) < 0) 1089 rte_exit(EXIT_FAILURE, "Failed to set up queue table\n"); 1090 qconf->n_rx_queue++; 1091 1092 /* init port */ 1093 printf("Initializing port %d ... ", portid ); 1094 fflush(stdout); 1095 1096 n_tx_queue = nb_lcores; 1097 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT) 1098 n_tx_queue = MAX_TX_QUEUE_PER_PORT; 1099 ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue, 1100 &port_conf); 1101 if (ret < 0) { 1102 printf("\n"); 1103 rte_exit(EXIT_FAILURE, "Cannot configure device: " 1104 "err=%d, port=%d\n", 1105 ret, portid); 1106 } 1107 1108 /* init one RX queue */ 1109 ret = rte_eth_rx_queue_setup(portid, 0, nb_rxd, 1110 socket, NULL, 1111 rxq->pool); 1112 if (ret < 0) { 1113 printf("\n"); 1114 rte_exit(EXIT_FAILURE, "rte_eth_rx_queue_setup: " 1115 "err=%d, port=%d\n", 1116 ret, portid); 1117 } 1118 1119 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]); 1120 print_ethaddr(" Address:", &ports_eth_addr[portid]); 1121 printf("\n"); 1122 1123 /* init one TX queue per couple (lcore,port) */ 1124 queueid = 0; 1125 for (lcore_id = 0; lcore_id < RTE_MAX_LCORE; lcore_id++) { 1126 if (rte_lcore_is_enabled(lcore_id) == 0) 1127 continue; 1128 1129 socket = (int) rte_lcore_to_socket_id(lcore_id); 1130 1131 printf("txq=%u,%d,%d ", lcore_id, queueid, socket); 1132 fflush(stdout); 1133 1134 rte_eth_dev_info_get(portid, &dev_info); 1135 txconf = &dev_info.default_txconf; 1136 txconf->txq_flags = 0; 1137 1138 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd, 1139 socket, txconf); 1140 if (ret < 0) 1141 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, " 1142 "port=%d\n", ret, portid); 1143 1144 qconf = &lcore_queue_conf[lcore_id]; 1145 qconf->tx_queue_id[portid] = queueid; 1146 setup_port_tbl(qconf, lcore_id, socket, portid); 1147 queueid++; 1148 } 1149 printf("\n"); 1150 } 1151 1152 printf("\n"); 1153 1154 /* start ports */ 1155 for (portid = 0; portid < nb_ports; portid++) { 1156 if ((enabled_port_mask & (1 << portid)) == 0) { 1157 continue; 1158 } 1159 /* Start device */ 1160 ret = rte_eth_dev_start(portid); 1161 if (ret < 0) 1162 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n", 1163 ret, portid); 1164 1165 rte_eth_promiscuous_enable(portid); 1166 } 1167 1168 if (init_routing_table() < 0) 1169 rte_exit(EXIT_FAILURE, "Cannot init routing table\n"); 1170 1171 check_all_ports_link_status((uint8_t)nb_ports, enabled_port_mask); 1172 1173 signal(SIGUSR1, signal_handler); 1174 signal(SIGTERM, signal_handler); 1175 signal(SIGINT, signal_handler); 1176 1177 /* launch per-lcore init on every lcore */ 1178 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER); 1179 RTE_LCORE_FOREACH_SLAVE(lcore_id) { 1180 if (rte_eal_wait_lcore(lcore_id) < 0) 1181 return -1; 1182 } 1183 1184 return 0; 1185 } 1186