1 /*- 2 * BSD LICENSE 3 * 4 * Copyright(c) 2010-2012 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 * version: DPDK.L.1.2.3-3 34 */ 35 36 #include <stdio.h> 37 #include <stdlib.h> 38 #include <stdint.h> 39 #include <inttypes.h> 40 #include <sys/types.h> 41 #include <string.h> 42 #include <sys/queue.h> 43 #include <stdarg.h> 44 #include <errno.h> 45 #include <getopt.h> 46 47 #include <rte_common.h> 48 #include <rte_byteorder.h> 49 #include <rte_log.h> 50 #include <rte_tailq.h> 51 #include <rte_memory.h> 52 #include <rte_memcpy.h> 53 #include <rte_memzone.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_hash_crc.h> 74 #include <rte_fbk_hash.h> 75 #include <rte_ip.h> 76 77 #include "main.h" 78 79 #define RTE_LOGTYPE_IPv4_MULTICAST RTE_LOGTYPE_USER1 80 81 #define MAX_PORTS 16 82 83 #define MCAST_CLONE_PORTS 2 84 #define MCAST_CLONE_SEGS 2 85 86 #define PKT_MBUF_SIZE (2048 + sizeof(struct rte_mbuf) + RTE_PKTMBUF_HEADROOM) 87 #define NB_PKT_MBUF 8192 88 89 #define HDR_MBUF_SIZE (sizeof(struct rte_mbuf) + 2 * RTE_PKTMBUF_HEADROOM) 90 #define NB_HDR_MBUF (NB_PKT_MBUF * MAX_PORTS) 91 92 #define CLONE_MBUF_SIZE (sizeof(struct rte_mbuf)) 93 #define NB_CLONE_MBUF (NB_PKT_MBUF * MCAST_CLONE_PORTS * MCAST_CLONE_SEGS * 2) 94 95 /* allow max jumbo frame 9.5 KB */ 96 #define JUMBO_FRAME_MAX_SIZE 0x2600 97 98 /* 99 * RX and TX Prefetch, Host, and Write-back threshold values should be 100 * carefully set for optimal performance. Consult the network 101 * controller's datasheet and supporting DPDK documentation for guidance 102 * on how these parameters should be set. 103 */ 104 #define RX_PTHRESH 8 /**< Default values of RX prefetch threshold reg. */ 105 #define RX_HTHRESH 8 /**< Default values of RX host threshold reg. */ 106 #define RX_WTHRESH 4 /**< Default values of RX write-back threshold reg. */ 107 108 /* 109 * These default values are optimized for use with the Intel(R) 82599 10 GbE 110 * Controller and the DPDK ixgbe PMD. Consider using other values for other 111 * network controllers and/or network drivers. 112 */ 113 #define TX_PTHRESH 36 /**< Default values of TX prefetch threshold reg. */ 114 #define TX_HTHRESH 0 /**< Default values of TX host threshold reg. */ 115 #define TX_WTHRESH 0 /**< Default values of TX write-back threshold reg. */ 116 117 #define MAX_PKT_BURST 32 118 #define BURST_TX_DRAIN 200000ULL /* around 100us at 2 Ghz */ 119 120 #define SOCKET0 0 121 122 /* Configure how many packets ahead to prefetch, when reading packets */ 123 #define PREFETCH_OFFSET 3 124 125 /* 126 * Construct Ethernet multicast address from IPv4 multicast address. 127 * Citing RFC 1112, section 6.4: 128 * "An IP host group address is mapped to an Ethernet multicast address 129 * by placing the low-order 23-bits of the IP address into the low-order 130 * 23 bits of the Ethernet multicast address 01-00-5E-00-00-00 (hex)." 131 */ 132 #define ETHER_ADDR_FOR_IPV4_MCAST(x) \ 133 (rte_cpu_to_be_64(0x01005e000000ULL | ((x) & 0x7fffff)) >> 16) 134 135 /* 136 * Configurable number of RX/TX ring descriptors 137 */ 138 #define RTE_TEST_RX_DESC_DEFAULT 128 139 #define RTE_TEST_TX_DESC_DEFAULT 512 140 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT; 141 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT; 142 143 /* ethernet addresses of ports */ 144 static struct ether_addr ports_eth_addr[MAX_PORTS]; 145 146 /* mask of enabled ports */ 147 static uint32_t enabled_port_mask = 0; 148 149 static uint8_t nb_ports = 0; 150 151 static int rx_queue_per_lcore = 1; 152 153 struct mbuf_table { 154 uint16_t len; 155 struct rte_mbuf *m_table[MAX_PKT_BURST]; 156 }; 157 158 #define MAX_RX_QUEUE_PER_LCORE 16 159 #define MAX_TX_QUEUE_PER_PORT 16 160 struct lcore_queue_conf { 161 uint64_t tx_tsc; 162 uint16_t n_rx_queue; 163 uint8_t rx_queue_list[MAX_RX_QUEUE_PER_LCORE]; 164 uint16_t tx_queue_id[MAX_PORTS]; 165 struct mbuf_table tx_mbufs[MAX_PORTS]; 166 } __rte_cache_aligned; 167 static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE]; 168 169 static const struct rte_eth_conf port_conf = { 170 .rxmode = { 171 .max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE, 172 .split_hdr_size = 0, 173 .header_split = 0, /**< Header Split disabled */ 174 .hw_ip_checksum = 0, /**< IP checksum offload disabled */ 175 .hw_vlan_filter = 0, /**< VLAN filtering disabled */ 176 .jumbo_frame = 1, /**< Jumbo Frame Support enabled */ 177 .hw_strip_crc = 0, /**< CRC stripped by hardware */ 178 }, 179 .txmode = { 180 }, 181 }; 182 183 static const struct rte_eth_rxconf rx_conf = { 184 .rx_thresh = { 185 .pthresh = RX_PTHRESH, 186 .hthresh = RX_HTHRESH, 187 .wthresh = RX_WTHRESH, 188 }, 189 }; 190 191 static const struct rte_eth_txconf tx_conf = { 192 .tx_thresh = { 193 .pthresh = TX_PTHRESH, 194 .hthresh = TX_HTHRESH, 195 .wthresh = TX_WTHRESH, 196 }, 197 .tx_free_thresh = 0, /* Use PMD default values */ 198 .tx_rs_thresh = 0, /* Use PMD default values */ 199 }; 200 201 static struct rte_mempool *packet_pool, *header_pool, *clone_pool; 202 203 204 /* Multicast */ 205 static struct rte_fbk_hash_params mcast_hash_params = { 206 .name = "MCAST_HASH", 207 .entries = 1024, 208 .entries_per_bucket = 4, 209 .socket_id = SOCKET0, 210 .hash_func = NULL, 211 .init_val = 0, 212 }; 213 214 struct rte_fbk_hash_table *mcast_hash = NULL; 215 216 struct mcast_group_params { 217 uint32_t ip; 218 uint16_t port_mask; 219 }; 220 221 static struct mcast_group_params mcast_group_table[] = { 222 {IPv4(224,0,0,101), 0x1}, 223 {IPv4(224,0,0,102), 0x2}, 224 {IPv4(224,0,0,103), 0x3}, 225 {IPv4(224,0,0,104), 0x4}, 226 {IPv4(224,0,0,105), 0x5}, 227 {IPv4(224,0,0,106), 0x6}, 228 {IPv4(224,0,0,107), 0x7}, 229 {IPv4(224,0,0,108), 0x8}, 230 {IPv4(224,0,0,109), 0x9}, 231 {IPv4(224,0,0,110), 0xA}, 232 {IPv4(224,0,0,111), 0xB}, 233 {IPv4(224,0,0,112), 0xC}, 234 {IPv4(224,0,0,113), 0xD}, 235 {IPv4(224,0,0,114), 0xE}, 236 {IPv4(224,0,0,115), 0xF}, 237 }; 238 239 #define N_MCAST_GROUPS \ 240 (sizeof (mcast_group_table) / sizeof (mcast_group_table[0])) 241 242 243 /* Send burst of packets on an output interface */ 244 static void 245 send_burst(struct lcore_queue_conf *qconf, uint8_t port) 246 { 247 struct rte_mbuf **m_table; 248 uint16_t n, queueid; 249 int ret; 250 251 queueid = qconf->tx_queue_id[port]; 252 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table; 253 n = qconf->tx_mbufs[port].len; 254 255 ret = rte_eth_tx_burst(port, queueid, m_table, n); 256 while (unlikely (ret < n)) { 257 rte_pktmbuf_free(m_table[ret]); 258 ret++; 259 } 260 261 qconf->tx_mbufs[port].len = 0; 262 } 263 264 /* Get number of bits set. */ 265 static inline uint32_t 266 bitcnt(uint32_t v) 267 { 268 uint32_t n; 269 270 for (n = 0; v != 0; v &= v - 1, n++) 271 ; 272 273 return (n); 274 } 275 276 /** 277 * Create the output multicast packet based on the given input packet. 278 * There are two approaches for creating outgoing packet, though both 279 * are based on data zero-copy idea, they differ in few details: 280 * First one creates a clone of the input packet, e.g - walk though all 281 * segments of the input packet, and for each of them create a new packet 282 * mbuf and attach that new mbuf to the segment (refer to rte_pktmbuf_clone() 283 * for more details). Then new mbuf is allocated for the packet header 284 * and is prepended to the 'clone' mbuf. 285 * Second approach doesn't make a clone, it just increment refcnt for all 286 * input packet segments. Then it allocates new mbuf for the packet header 287 * and prepends it to the input packet. 288 * Basically first approach reuses only input packet's data, but creates 289 * it's own copy of packet's metadata. Second approach reuses both input's 290 * packet data and metadata. 291 * The advantage of first approach - is that each outgoing packet has it's 292 * own copy of metadata, so we can safely modify data pointer of the 293 * input packet. That allows us to skip creation if the output packet for 294 * the last destination port, but instead modify input packet's header inplace, 295 * e.g: for N destination ports we need to invoke mcast_out_pkt (N-1) times. 296 * The advantage of second approach - less work for each outgoing packet, 297 * e.g: we skip "clone" operation completely. Though it comes with a price - 298 * input packet's metadata has to be intact. So for N destination ports we 299 * need to invoke mcast_out_pkt N times. 300 * So for small number of outgoing ports (and segments in the input packet) 301 * first approach will be faster. 302 * As number of outgoing ports (and/or input segments) will grow, 303 * second way will become more preferable. 304 * 305 * @param pkt 306 * Input packet mbuf. 307 * @param use_clone 308 * Control which of the two approaches described above should be used: 309 * - 0 - use second approach: 310 * Don't "clone" input packet. 311 * Prepend new header directly to the input packet 312 * - 1 - use first approach: 313 * Make a "clone" of input packet first. 314 * Prepend new header to the clone of the input packet 315 * @return 316 * - The pointer to the new outgoing packet. 317 * - NULL if operation failed. 318 */ 319 static inline struct rte_mbuf * 320 mcast_out_pkt(struct rte_mbuf *pkt, int use_clone) 321 { 322 struct rte_mbuf *hdr; 323 324 /* Create new mbuf for the header. */ 325 if (unlikely ((hdr = rte_pktmbuf_alloc(header_pool)) == NULL)) 326 return (NULL); 327 328 /* If requested, then make a new clone packet. */ 329 if (use_clone != 0 && 330 unlikely ((pkt = rte_pktmbuf_clone(pkt, clone_pool)) == NULL)) { 331 rte_pktmbuf_free(hdr); 332 return (NULL); 333 } 334 335 /* prepend new header */ 336 hdr->pkt.next = pkt; 337 338 339 /* update header's fields */ 340 hdr->pkt.pkt_len = (uint16_t)(hdr->pkt.data_len + pkt->pkt.pkt_len); 341 hdr->pkt.nb_segs = (uint8_t)(pkt->pkt.nb_segs + 1); 342 343 /* copy metadata from source packet*/ 344 hdr->pkt.in_port = pkt->pkt.in_port; 345 hdr->pkt.vlan_tci = pkt->pkt.vlan_tci; 346 hdr->pkt.l2_len = pkt->pkt.l2_len; 347 hdr->pkt.l3_len = pkt->pkt.l3_len; 348 hdr->pkt.hash = pkt->pkt.hash; 349 350 hdr->ol_flags = pkt->ol_flags; 351 352 __rte_mbuf_sanity_check(hdr, RTE_MBUF_PKT, 1); 353 return (hdr); 354 } 355 356 /* 357 * Write new Ethernet header to the outgoing packet, 358 * and put it into the outgoing queue for the given port. 359 */ 360 static inline void 361 mcast_send_pkt(struct rte_mbuf *pkt, struct ether_addr *dest_addr, 362 struct lcore_queue_conf *qconf, uint8_t port) 363 { 364 struct ether_hdr *ethdr; 365 uint16_t len; 366 367 /* Construct Ethernet header. */ 368 ethdr = (struct ether_hdr *)rte_pktmbuf_prepend(pkt, (uint16_t)sizeof(*ethdr)); 369 RTE_MBUF_ASSERT(ethdr != NULL); 370 371 ether_addr_copy(dest_addr, ðdr->d_addr); 372 ether_addr_copy(&ports_eth_addr[port], ðdr->s_addr); 373 ethdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4); 374 375 /* Put new packet into the output queue */ 376 len = qconf->tx_mbufs[port].len; 377 qconf->tx_mbufs[port].m_table[len] = pkt; 378 qconf->tx_mbufs[port].len = ++len; 379 380 /* Transmit packets */ 381 if (unlikely(MAX_PKT_BURST == len)) 382 send_burst(qconf, port); 383 } 384 385 /* Multicast forward of the input packet */ 386 static inline void 387 mcast_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf) 388 { 389 struct rte_mbuf *mc; 390 struct ipv4_hdr *iphdr; 391 uint32_t dest_addr, port_mask, port_num, use_clone; 392 int32_t hash; 393 uint8_t port; 394 union { 395 uint64_t as_int; 396 struct ether_addr as_addr; 397 } dst_eth_addr; 398 399 /* Remove the Ethernet header from the input packet */ 400 iphdr = (struct ipv4_hdr *)rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr)); 401 RTE_MBUF_ASSERT(iphdr != NULL); 402 403 dest_addr = rte_be_to_cpu_32(iphdr->dst_addr); 404 405 /* 406 * Check that it is a valid multicast address and 407 * we have some active ports assigned to it. 408 */ 409 if(!IS_IPV4_MCAST(dest_addr) || 410 (hash = rte_fbk_hash_lookup(mcast_hash, dest_addr)) <= 0 || 411 (port_mask = hash & enabled_port_mask) == 0) { 412 rte_pktmbuf_free(m); 413 return; 414 } 415 416 /* Calculate number of destination ports. */ 417 port_num = bitcnt(port_mask); 418 419 /* Should we use rte_pktmbuf_clone() or not. */ 420 use_clone = (port_num <= MCAST_CLONE_PORTS && 421 m->pkt.nb_segs <= MCAST_CLONE_SEGS); 422 423 /* Mark all packet's segments as referenced port_num times */ 424 if (use_clone == 0) 425 rte_pktmbuf_refcnt_update(m, (uint16_t)port_num); 426 427 /* construct destination ethernet address */ 428 dst_eth_addr.as_int = ETHER_ADDR_FOR_IPV4_MCAST(dest_addr); 429 430 for (port = 0; use_clone != port_mask; port_mask >>= 1, port++) { 431 432 /* Prepare output packet and send it out. */ 433 if ((port_mask & 1) != 0) { 434 if (likely ((mc = mcast_out_pkt(m, use_clone)) != NULL)) 435 mcast_send_pkt(mc, &dst_eth_addr.as_addr, 436 qconf, port); 437 else if (use_clone == 0) 438 rte_pktmbuf_free(m); 439 } 440 } 441 442 /* 443 * If we making clone packets, then, for the last destination port, 444 * we can overwrite input packet's metadata. 445 */ 446 if (use_clone != 0) 447 mcast_send_pkt(m, &dst_eth_addr.as_addr, qconf, port); 448 else 449 rte_pktmbuf_free(m); 450 } 451 452 /* Send burst of outgoing packet, if timeout expires. */ 453 static inline void 454 send_timeout_burst(struct lcore_queue_conf *qconf) 455 { 456 uint64_t cur_tsc; 457 uint8_t portid; 458 459 cur_tsc = rte_rdtsc(); 460 if (likely (cur_tsc < qconf->tx_tsc + BURST_TX_DRAIN)) 461 return; 462 463 for (portid = 0; portid < MAX_PORTS; portid++) { 464 if (qconf->tx_mbufs[portid].len != 0) 465 send_burst(qconf, portid); 466 } 467 qconf->tx_tsc = cur_tsc; 468 } 469 470 /* main processing loop */ 471 static __attribute__((noreturn)) int 472 main_loop(__rte_unused void *dummy) 473 { 474 struct rte_mbuf *pkts_burst[MAX_PKT_BURST]; 475 uint32_t lcore_id; 476 int i, j, nb_rx; 477 uint8_t portid; 478 struct lcore_queue_conf *qconf; 479 480 lcore_id = rte_lcore_id(); 481 qconf = &lcore_queue_conf[lcore_id]; 482 483 484 if (qconf->n_rx_queue == 0) { 485 RTE_LOG(INFO, IPv4_MULTICAST, "lcore %u has nothing to do\n", 486 lcore_id); 487 while(1); 488 } 489 490 RTE_LOG(INFO, IPv4_MULTICAST, "entering main loop on lcore %u\n", 491 lcore_id); 492 493 for (i = 0; i < qconf->n_rx_queue; i++) { 494 495 portid = qconf->rx_queue_list[i]; 496 RTE_LOG(INFO, IPv4_MULTICAST, " -- lcoreid=%u portid=%d\n", 497 lcore_id, (int) portid); 498 } 499 500 while (1) { 501 502 /* 503 * Read packet from RX queues 504 */ 505 for (i = 0; i < qconf->n_rx_queue; i++) { 506 507 portid = qconf->rx_queue_list[i]; 508 nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst, 509 MAX_PKT_BURST); 510 511 /* Prefetch first packets */ 512 for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) { 513 rte_prefetch0(rte_pktmbuf_mtod( 514 pkts_burst[j], void *)); 515 } 516 517 /* Prefetch and forward already prefetched packets */ 518 for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) { 519 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[ 520 j + PREFETCH_OFFSET], void *)); 521 mcast_forward(pkts_burst[j], qconf); 522 } 523 524 /* Forward remaining prefetched packets */ 525 for (; j < nb_rx; j++) { 526 mcast_forward(pkts_burst[j], qconf); 527 } 528 } 529 530 /* Send out packets from TX queues */ 531 send_timeout_burst(qconf); 532 } 533 } 534 535 /* display usage */ 536 static void 537 print_usage(const char *prgname) 538 { 539 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n" 540 " -p PORTMASK: hexadecimal bitmask of ports to configure\n" 541 " -q NQ: number of queue (=ports) per lcore (default is 1)\n", 542 prgname); 543 } 544 545 static uint32_t 546 parse_portmask(const char *portmask) 547 { 548 char *end = NULL; 549 unsigned long pm; 550 551 /* parse hexadecimal string */ 552 pm = strtoul(portmask, &end, 16); 553 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0')) 554 return 0; 555 556 return ((uint32_t)pm); 557 } 558 559 static int 560 parse_nqueue(const char *q_arg) 561 { 562 char *end = NULL; 563 unsigned long n; 564 565 /* parse numerical string */ 566 errno = 0; 567 n = strtoul(q_arg, &end, 0); 568 if (errno != 0 || end == NULL || *end != '\0' || 569 n == 0 || n >= MAX_RX_QUEUE_PER_LCORE) 570 return (-1); 571 572 return (n); 573 } 574 575 /* Parse the argument given in the command line of the application */ 576 static int 577 parse_args(int argc, char **argv) 578 { 579 int opt, ret; 580 char **argvopt; 581 int option_index; 582 char *prgname = argv[0]; 583 static struct option lgopts[] = { 584 {NULL, 0, 0, 0} 585 }; 586 587 argvopt = argv; 588 589 while ((opt = getopt_long(argc, argvopt, "p:q:", 590 lgopts, &option_index)) != EOF) { 591 592 switch (opt) { 593 /* portmask */ 594 case 'p': 595 enabled_port_mask = parse_portmask(optarg); 596 if (enabled_port_mask == 0) { 597 printf("invalid portmask\n"); 598 print_usage(prgname); 599 return -1; 600 } 601 break; 602 603 /* nqueue */ 604 case 'q': 605 rx_queue_per_lcore = parse_nqueue(optarg); 606 if (rx_queue_per_lcore < 0) { 607 printf("invalid queue number\n"); 608 print_usage(prgname); 609 return -1; 610 } 611 break; 612 613 default: 614 print_usage(prgname); 615 return -1; 616 } 617 } 618 619 if (optind >= 0) 620 argv[optind-1] = prgname; 621 622 ret = optind-1; 623 optind = 0; /* reset getopt lib */ 624 return ret; 625 } 626 627 static void 628 print_ethaddr(const char *name, struct ether_addr *eth_addr) 629 { 630 printf("%s%02X:%02X:%02X:%02X:%02X:%02X", name, 631 eth_addr->addr_bytes[0], 632 eth_addr->addr_bytes[1], 633 eth_addr->addr_bytes[2], 634 eth_addr->addr_bytes[3], 635 eth_addr->addr_bytes[4], 636 eth_addr->addr_bytes[5]); 637 } 638 639 static int 640 init_mcast_hash(void) 641 { 642 uint32_t i; 643 644 mcast_hash = rte_fbk_hash_create(&mcast_hash_params); 645 if (mcast_hash == NULL){ 646 return -1; 647 } 648 649 for (i = 0; i < N_MCAST_GROUPS; i ++){ 650 if (rte_fbk_hash_add_key(mcast_hash, 651 mcast_group_table[i].ip, 652 mcast_group_table[i].port_mask) < 0) { 653 return -1; 654 } 655 } 656 657 return 0; 658 } 659 660 int 661 MAIN(int argc, char **argv) 662 { 663 struct lcore_queue_conf *qconf; 664 struct rte_eth_link link; 665 int ret; 666 uint16_t queueid; 667 unsigned lcore_id = 0, rx_lcore_id = 0;; 668 uint32_t n_tx_queue, nb_lcores; 669 uint8_t portid; 670 671 /* init EAL */ 672 ret = rte_eal_init(argc, argv); 673 if (ret < 0) 674 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n"); 675 argc -= ret; 676 argv += ret; 677 678 /* parse application arguments (after the EAL ones) */ 679 ret = parse_args(argc, argv); 680 if (ret < 0) 681 rte_exit(EXIT_FAILURE, "Invalid IPV4_MULTICAST parameters\n"); 682 683 /* create the mbuf pools */ 684 packet_pool = rte_mempool_create("packet_pool", NB_PKT_MBUF, 685 PKT_MBUF_SIZE, 32, sizeof(struct rte_pktmbuf_pool_private), 686 rte_pktmbuf_pool_init, NULL, rte_pktmbuf_init, NULL, 687 SOCKET0, 0); 688 689 if (packet_pool == NULL) 690 rte_exit(EXIT_FAILURE, "Cannot init packet mbuf pool\n"); 691 692 header_pool = rte_mempool_create("header_pool", NB_HDR_MBUF, 693 HDR_MBUF_SIZE, 32, 0, NULL, NULL, rte_pktmbuf_init, NULL, 694 SOCKET0, 0); 695 696 if (header_pool == NULL) 697 rte_exit(EXIT_FAILURE, "Cannot init header mbuf pool\n"); 698 699 clone_pool = rte_mempool_create("clone_pool", NB_CLONE_MBUF, 700 CLONE_MBUF_SIZE, 32, 0, NULL, NULL, rte_pktmbuf_init, NULL, 701 SOCKET0, 0); 702 703 if (clone_pool == NULL) 704 rte_exit(EXIT_FAILURE, "Cannot init clone mbuf pool\n"); 705 706 /* init driver */ 707 #ifdef RTE_LIBRTE_IGB_PMD 708 if (rte_igb_pmd_init() < 0) 709 rte_exit(EXIT_FAILURE, "Cannot init igb pmd\n"); 710 #endif 711 #ifdef RTE_LIBRTE_IXGBE_PMD 712 if (rte_ixgbe_pmd_init() < 0) 713 rte_exit(EXIT_FAILURE, "Cannot init ixgbe pmd\n"); 714 #endif 715 716 if (rte_eal_pci_probe() < 0) 717 rte_exit(EXIT_FAILURE, "Cannot probe PCI\n"); 718 719 nb_ports = rte_eth_dev_count(); 720 if (nb_ports == 0) 721 rte_exit(EXIT_FAILURE, "No physical ports!\n"); 722 if (nb_ports > MAX_PORTS) 723 nb_ports = MAX_PORTS; 724 725 nb_lcores = rte_lcore_count(); 726 727 /* initialize all ports */ 728 for (portid = 0; portid < nb_ports; portid++) { 729 /* skip ports that are not enabled */ 730 if ((enabled_port_mask & (1 << portid)) == 0) { 731 printf("Skipping disabled port %d\n", portid); 732 continue; 733 } 734 735 qconf = &lcore_queue_conf[rx_lcore_id]; 736 737 /* get the lcore_id for this port */ 738 while (rte_lcore_is_enabled(rx_lcore_id) == 0 || 739 qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) { 740 741 rx_lcore_id ++; 742 qconf = &lcore_queue_conf[rx_lcore_id]; 743 744 if (rx_lcore_id >= RTE_MAX_LCORE) 745 rte_exit(EXIT_FAILURE, "Not enough cores\n"); 746 } 747 qconf->rx_queue_list[qconf->n_rx_queue] = portid; 748 qconf->n_rx_queue++; 749 750 /* init port */ 751 printf("Initializing port %d on lcore %u... ", portid, 752 rx_lcore_id); 753 fflush(stdout); 754 755 n_tx_queue = nb_lcores; 756 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT) 757 n_tx_queue = MAX_TX_QUEUE_PER_PORT; 758 ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue, 759 &port_conf); 760 if (ret < 0) 761 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n", 762 ret, portid); 763 764 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]); 765 print_ethaddr(" Address:", &ports_eth_addr[portid]); 766 printf(", "); 767 768 /* init one RX queue */ 769 queueid = 0; 770 printf("rxq=%hu ", queueid); 771 fflush(stdout); 772 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd, 773 SOCKET0, &rx_conf, 774 packet_pool); 775 if (ret < 0) 776 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%d\n", 777 ret, portid); 778 779 /* init one TX queue per couple (lcore,port) */ 780 queueid = 0; 781 782 RTE_LCORE_FOREACH(lcore_id) { 783 if (rte_lcore_is_enabled(lcore_id) == 0) 784 continue; 785 printf("txq=%u,%hu ", lcore_id, queueid); 786 fflush(stdout); 787 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd, 788 SOCKET0, &tx_conf); 789 if (ret < 0) 790 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, " 791 "port=%d\n", ret, portid); 792 793 qconf = &lcore_queue_conf[lcore_id]; 794 qconf->tx_queue_id[portid] = queueid; 795 queueid++; 796 } 797 798 /* Start device */ 799 ret = rte_eth_dev_start(portid); 800 if (ret < 0) 801 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n", 802 ret, portid); 803 804 printf("done: "); 805 806 /* get link status */ 807 rte_eth_link_get(portid, &link); 808 if (link.link_status) { 809 printf(" Link Up - speed %u Mbps - %s\n", 810 (uint32_t) link.link_speed, 811 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ? 812 ("full-duplex") : ("half-duplex\n")); 813 rte_eth_promiscuous_enable(portid); 814 rte_eth_allmulticast_enable(portid); 815 } else { 816 printf(" Link Down\n"); 817 } 818 } 819 820 821 /* initialize the multicast hash */ 822 int retval = init_mcast_hash(); 823 if (retval != 0) 824 rte_exit(EXIT_FAILURE, "Cannot build the multicast hash\n"); 825 826 /* launch per-lcore init on every lcore */ 827 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER); 828 RTE_LCORE_FOREACH_SLAVE(lcore_id) { 829 if (rte_eal_wait_lcore(lcore_id) < 0) 830 return -1; 831 } 832 833 return 0; 834 } 835